Dimensions of Possession - elchacocomoarealinguistica

Dimensions of Possession

Typological Studies in Language (TSL)A companion series to the journal Studies in LanguageGeneral EditorAssistant EditorsMichael NoonanSpike Gildea, Suzanne KemmerEditorial BoardWallace Chafe (Santa Barbara)Bernard Comrie (Leipzig)R.M.W. Dixon (Canberra)Matthew Dryer (Buffalo)John Haiman (St Paul)Kenneth Hale (Cambridge, Mass.)Bernd Heine (Köln)Paul Hopper (Pittsburgh)Andrej Kibrik (Moscow)Ronald Langacker (San Diego)Charles Li (Santa Barbara)Andrew Pawley (Canberra)Doris Payne (Oregon)Frans Plank (Konstanz)Jerrold Sadock (Chicago)Dan Slobin (Berkeley)Sandra Thompson (Santa Barbara)Volumes in this series will be functionally and typologically oriented, coveringspecific topics in language by collecting together data from a wide variety oflanguages and language typologies. The orientation of the volumes will besubstantive rather than formal, with the aim of investigating universals ofhuman language via as broadly defined a data base as possible, leaning towardcross-linguistic, diachronic, developmental and live-discourse data.Volume 47Dimensions of PossessionEdited by Irène Baron, Michael Herslund and Finn Sørensen

Dimensions of PossessionEdited byIrène BaronMichael HerslundFinn SørensenCopenhagen Business SchoolJohn Benjamins Publishing CompanyAmsterdam/Philadelphia

TM paper used in this publication meets the minimum requirements of American8TheNational Standard for Information Sciences – Permanence of Paper for PrintedLibrary Materials, ansi z39.48-1984.Library of Congress Cataloging-in-Publication DataDimensions of Possession / edited by Irène Baron, Michael Herslund and Finn Sørensen.p. cm. (Typological Studies in Language, issn 0167–7373 ; v. 47)Includes bibliographical references and indexes.1. Grammar, Comparative and general--Possessives. I. Baron, Irène. II.Herslund, Michael. III. Sørensen, Finn, 1942- IV. Series.P299.P67 D56 2001415--dc21 2001035089isbn 90 272 29511 (Eur.) / 1 58811 0621 (US) (Hb; alk. paper)© 2001 – John Benjamins B.V.No part of this book may be reproduced in any form, by print, photoprint, microfilm, or anyother means, without written permission from the publisher.John Benjamins Publishing Co. · P.O.Box 36224 · 1020 me Amsterdam · The NetherlandsJohn Benjamins North America · P.O.Box 27519 · Philadelphia pa 19118-0519 · usa

Table of ContentsIntroduction: Dimensions of possession 1Michael Herslund and Irène Baron1. The operational basis of possession:A dimensional approach revisited 27Hansjakob Seiler2. The concept of possession in Danish grammar 41Ole Togeby3. Possession spaces in Danish 57Finn Sørensen4. The verb have in Nyulnyulan languages 67William McGregor5. Semantics of the verb have 85Irène Baron and Michael Herslund6. Possessum-oriented and possessor-oriented constructionsin Russian 99Per Durst-Andersen7. Datives and comitatives as neighbouring spousesThe case of indirect objects and comitatives in Danish 115Lars Heltoft8. Towards a typology of French NP de NP structures or how muchpossession is there in complex noun phrases with de in French? 147Inge Bartning9. Spanish N de N structures from a cognitive perspective 169Henrik Høeg Müller10. The grammatical category “Possession’’and the part–whole relation in French 187Martin Riegel

viContents11. Kinship in grammar 201Östen Dahl and Maria Koptjevskaja-Tamm12. (In)alienability and (in)determination in Portuguese 227Anne-Marie Spanoghe13. Possessives with extensive use:A source of definite articles? 243Kari Fraurud14. Possessors and experiencers in Classical Latin 269A. Machtelt Bolkestein15. The difference a category makes in the expressionof possession and inalienability 285Marianne Mithun16. Ways of explaining possession 311Bernd HeineIndex of languages 329Index of authors 331Index of subjects 334

IntroductionDimensions of possessionMichael Herslund and Irène Baron1. IntroductionPossession is a very elusive notion. It is a complex and vague notion which inorder to be expressed has to draw on different linguistic structures. For thereis namely no such thing as an established possessive construction in the samesense as there are locational or equational structures which can be identifiedby their linguistic material, cf. the contribution to the present volume bySørensen. Whether it is approached from the semasiological or theonomasiological angle, the problems present themselves immediately. Thesemasiological approach seems offhand hopeless because of the mere fact thatno form or construction in any language conveys exclusively what everybodywould agree on identifying as possession. And one could argue, as doesTogeby in the present volume, that in a language like Danish possession is notat all grammaticalised. As is well known, even the genitive, which, where itexists, should be the expression of possession par excellence has in mostlanguages numerous uses unlikely to qualify as possessive. And the same istrue of predicative constructions with verbs with the general meaning ‘have’,which also have numerous uses that only with difficulty can be reconciled witha common pretheoretical understanding of possession.Since possession is inherently a concept, not a linguistic construction, theonly way to approach it seems to be from the onomasiological angle. Thisapproach is consequently the one chosen by most grammarians and linguists.Starting from an intuitive notion of possession, the domain is made more preciseby listing the different kinds of relations which seem to fall under such a commonnotion. This is true of traditional grammarians as well as of modernlinguists. An example of the first category is Nygaard’s characterisation of thepossessive genitive from his syntax of Old Norse: “In connection with a noun

2 Michael Herslund and Irène Baronhaving a concrete meaning the genitive is used with the name of the person orthing, to whom or which something belongs or is related by virtue of ownership— family relation — friendship or enmity — superior or subaltern position —origin, source, reason, cause — natural or necessary (conceptual) connection —established order or a relation created by circumstances’’ (1906: 129). As anexample of the second category, the modern linguists, one can quote Seiler’scharacterisation: “Semantically, the domain of POSSESSION can be defined asbio-cultural. It is the relationship between a human being and his kinsmen, hisbody parts, his material belongings, his cultural and intellectual products. In amore extended view, it is the relationship between parts and wholes of anorganism’’ (1983a: 4). Such conceptual delimitations of possession can finally becoined into something like the seven types of predicative possession identified byHeine (1997: 34ff.): Physical possession, Temporary possession, Permanent possession,Inalienable possession, Abstract possession, Inanimate inalienable possession,Inanimate alienable possession. Some of these are of course less prototypical thanothers: there may be less agreement among linguists on the proper inclusion intothe realm of possession of the Abstract possession (Ihavenotime, Ihaveacold)orthe Inanimate alienable possession (The tree has crows on it). The prototypical caseof possession is expounded as a cluster of properties such as ‘human possessor,concrete possessee, possessor having the right to use the possessee, spatialproximity between the two, no temporal limit on the possessive relation’ (Heine1997: 39ff.). This characterisation is also close to Seiler’s: “Linguistic POSSES-SION consists of the representation of a relationship between a substance andanother substance. Substance A, called the POSSESSOR, is prototypically[+animate], more specifically [+human], and still more specifically [+EGO] orclose to the speaker’’ (1983a: 4 and this volume). As the comparisons show, thetraditional grammarians’ insight is not conspicuously inferior to that of themodern linguists’. It is of course the same problem that they are facing.Following Seiler we can say that what is normally called possession is thelinguistic expression of the relation between two entities, a Possessor and aPossessum, such that one, the Possessor, is seen as being in some way relatedto the other, the Possessum, as having it near or controlling it. The kind ofrelation between the two can be of various sorts, as the possibilities listed byNygaard quoted above, but is often deducible from the combined lexicalcontent of the Possessor and the Possessum nouns, cf. Baron and Herslund(this volume). We can illustrate this by three simple examples:(1) a. Mary has a daughter.Mary’s daughter

Introduction 3b. Mary has a house.Mary’s housec. Mary has a pen.Mary’s penIn (1a) the relational noun daughter induces a kinship reading in both thepredicative and the attributive expression. In (1b) the noun house induces anownership reading in the predicative expression, but the correspondingattributive expression is much more polysemous. In (1c) finally, the word pencan in the predicative construction induce an ownership reading as in (4b),but the availability reading, i.e. ‘Mary has a pen at her disposal right now’, ismore probable, and the same is true of the attributive phrase, viz. ‘You can useMary’s pen. Her name is on it’ (ownership) or ‘Mary isn’t here right now, butthere is a pen on her desk. We can use Mary’s pen’ (availability).But as soon as one moves on from rather simple cases like these, numerousproblems occur. One source of such problems is the fundamentallypolysemous nature of the verbs occurring in predicative possessive expressionssuch as English have. In many languages, a have verb has at least threedistinct uses of which two are not clearly of the possessive kind. The first useis the one illustrated in (2), i.e. a stative relation between two distinctentities:(2) [Mary] has [a cat].This is what we regard as prototypical possession corresponding with Heine’spermanent possession, cf. above. The second use is the situation where theconstruction only denotes one entity in a certain state:(3) [Mary] has a cold.This is the case of ‘abstract possession’, cf. Heine (1997: 34), but it is debatablewhether it should be included in the concept of possession, or whether, rather,the Possessum in such cases merges with the verb to form a verbo-nominalpredicate, cf. Baron and Herslund (1998). It is true, however, that there existsan attributive expression in such cases just as in the case of possession proper,viz. Mary’s cold and Mary’s cat.The third case is the use of have as an auxiliary verb:(4) [Mary] has bought a cat.This construction can historically be traced back to a possessive construction,i.e. a construction where the object is only the object of have, the participle

4 Michael Herslund and Irène Baronbeing the complement of the object: Mary has [a cat bought]. But it seems bestto disregard it in a discussion of possession, because the have-part of it nowclearly involves only one entity in a state: the other entity, a cat in (4), isrelated to the first entity not by have, but by the lexical verb. It is the lexicalmeaning of the verb buy which accounts for the possessive content of (4),which accordingly disappears when other verbs are chosen:(5) Mary has {seen, fed, caressed, chased ...}acat.The following discussion of possession is, like the articles in the presentvolume, articulated along the three major dimensions which constitute theprincipal linguistic realisations of possession: predicative possession(Section 2), attributive possession (Section 3) and the constructions “inbetween’’, i.e. the phenomena variously labelled external possession orpossessor ascension (Section 4).2. Predicative possessionAs seen above, possession occurs in both predicative and attributive constructions,i.e. the possessive relation is either expressed by a verb, hence explicitlyasserted, or by a nominal (a genitive phrase), in which case the relation ispresupposed. As seen in the examples above, the attributive construction istherefore inherently much more polysemous than the predicative one, cf.Heine (1997: 25ff.). We return to this question in 3.1.2.1 Possessor, Location and ExperiencerWhereas predicative possession, i.e. the asserted relation between a Possessorand a Possessum, is often realised by the Agent–Patient role schema, there aretwo other basic role configurations which seem to be of more immediateimportance for the interpretation of possession:Location – ArgumentExperiencer – StimulusFigure 1.These two patterns have one interesting and quite fundamental property incommon with possessive structures. Possession involves, as seen above, therelation between two entities, but these two entities are only assigned the roles

Introduction 5of Possessor and Possessum in virtue of one another, i.e. there is no Possessorwithout a Possessum, no Possessum without a Possessor. Seiler (1983a: 4)speaks in fact of a “strictly binary relation’’. There is thus a basic solidarity inthe underlying semantic role schema which is also found in the case ofLocation-Argument and Experiencer-Stimulus. Nothing is a place untilsomething is placed there, thus becoming the argument, as it were, of the place(cf. Guillet and Leclère 1992; Baron and Herslund 1997a: 135): in e.g. The bookis on the table, the table denotes only a place by virtue of the book being placedupon it. Nothing is an experiencer until a stimulus is present, and nothing isa stimulus if not perceived as such: in e.g. The noise scared Sharon, the noisedenotes an objectively occurring physical phenomenon, which only obtains thestatus of stimulus by being perceived and reacted to. In this respect, the Agent-Patient role schema as referred to above, is however quite different: somethingcan be an agent without the presence of a patient, viz. Peter jumps, andsomething can be a patient without any identifiable agent, viz. The waterevaporates. The conclusion is of course that possession has more in commonwith the Location and Experiencer schemata than with the Agent schema. Andit is the Location schema which, as the simplest and most primitive andconcrete notion, seems more basic, hence the traditional and widespread ideathat possession is a kind of “sophisticated’’ Location (cf. among others Lyons1967, 1968; Clark 1978; Freeze 1992; Spanoghe 1995: 30ff.; Baron andHerslund 1997a; Sørensen 1997a), whereas other scholars admit that locationis important in connection with possession, but that possession cannot bereduced to it, cf. in particular Isačenko (1974), Seiler (1983a), Heine (1997)and Heltoft (this volume).2.1.1The first argument in favour of a locative interpretation of possession is theobservation that many languages exhibit striking structural resemblancesbetween existential, locative and possessive constructions, cf. e.g. Lyons(1967); Christie (1970); Clark (1978). Existentials are often found to patternwith one kind of possessive construction whereas locatives pattern withanother, as regards word order, definiteness of the Possessum as well as choiceof verb (cf. Clark 1978). This can be illustrated by French:(6) a. ExistentialIl y a un livre sur la table.it there has a book on the table‘There is a book on the table’

6 Michael Herslund and Irène BaronPossessive 1Jeanaunlivre.‘Jean has a book’b. LocativeLe livre est sur la table.‘The book is on the table’Possessive 2Le livre est à Jean.the book is to Jean‘The book is Jean’s’In the a-examples un livre ‘a book’ is indefinite and follows the locativemarker y, 1 which anticipates the locative phrase sur la table ‘on the table’, aswell as the Possessor subject Jean. In the b-examples le livre ‘the book’ isdefinite and precedes the locative sur la table ‘on the table’ as well as thePossessor à Jean. These patterns are no coincidence: whereas many languageshave the same verb in all four cases, some have two different verbs manifestingthe pattern shown in (6), where one verb appears in the existential andthe possessive 1 (the have-construction), another in the locative and thepossessive 2 (the be-construction). And in languages with two verbs, it isnever the case that “the existential and possessive 2 [. . .] appear as a pair’’(Clark 1978: 109).2.1.2A second argument for the basically locative nature of possession takes itspoint of departure in Benveniste’s (1966a) observation that have is aninverted be. 2 If have is an inverted be, then the passive of have should beequivalent to a be-clause, i.e. have an existential meaning. Now, passives ofhave are rare, probably due to the fact that have like be is a state verb, andsuch verbs are not readily passivisable. Yet, the Danish verb have has a passive,haves, which occurs in restricted uses with the meaning ‘be’, ‘exist’, ‘beavailable’, i.e. precisely in existential constructions: 3(7) Røget fisk haves.smoked fish have-pass‘Smoked fish for sale’Den kvalitet haves ikke.that quality have-pass neg‘That quality is not available’

Introduction 7As seen in (7), the passive of have occurs, like most passives, normally withoutan agent-phrase. But one interesting thing about the demoted subject of anactive clause is that its “true nature’’, its semantic role, in many languages is soto speak revealed in the passive where it is realised as an oblique complement.So the active subject of locative verbs like Danish rumme, indeholde ‘contain’will in the passive not have the normal agentive preposition af ‘by’, but rathera locative preposition like i ‘in’: 4(8) Activea. Det er utroligt hvad den bil rummer.it is incredible what that car contains‘It is incredible how much that car contains’b. Kassen indeholder tredive flasker.box-def contains thirty bottle-pl‘The box contains thirty bottles’Passivec. Det er utroligt hvad der rummes i den bil.it is incredible what there contain-pass in that car‘It is incredible how much is contained in that car’d. Tredive flasker indeholdes i kassen.thirty bottle-pl contain-pass in box-def‘Thirty bottles are contained in the box’This is true also of the passive of have:(9) a. Den kvalitet haves i alle butikker.that quality have-pass in all shop-pl‘That quality is found in all shops’b. Oplysninger haves hos indehaveren. 5information have-pass at owner-def‘Information available from the owner’These facts seem to indicate the profound locative nature of have. A furtherpiece of evidence is the fact that certain have-clauses with a definite objectactually convey a rather concrete locative meaning insofar as they answerquestions in where, cf. Wierzbicka (1988: 345):(10) – Where are the children?– John has them/the children.

8 Michael Herslund and Irène BaronSuch data also seem to contradict Isačenko’s contention (1974: 76) that haveno longer denotes relative positions in space.2.1.3Apart from the arguments for the locative nature of possession, even inhave-constructions, adduced so far, there is a further argument in favour of alocative interpretation in the shape of the notion of Sub-Place often found inhave-constructions. This notion covers the case of prepositional complementswhich specify the actual location of the Possessum. The Possessum is thusrepresented as located, first, in a possessive relation to the Possessor, and then ina locative relation with respect to another place. Different patterns of denotationalinclusion can be identified in such constructions. In examples like (11):(11) a. The bank has a branch in Bristol.b. Mary has a child on her lap.A branch in (11a) is denotatively included in the bank via a Part–Wholerelation and locally situated by in Bristol; in (11b) it is her lap which isdenotatively included in the subject in a Part–Whole relation whereas theobject is simply located with respect to the subject by have and with respect toher lap by the local preposition on, cf. Baron and Herslund (1997b) and thisvolume. Such structures are neatly explained under the assumption that haveis a basically locative verb which can be accompanied by a sub-place whichlocates the object noun. This notion of sub-place only makes sense if there alsois a place. And this place can only be the subject of have. The property ofcombining with a sub-place is shared by stative locative verbs such as Englishcontain, keep, etc.:(12) He has his money in an old sock.The museum contains some fine pieces in its Degas collection.He keeps his goats in a garden shed.The fact that have-constructions display an apparent Agent-Patient configuration,i.e. a transitive construction with a non-local verb originally meaning‘take, grasp’, is often adduced as an argument against the locative nature ofpossession, cf. e.g. the contribution by Heltoft. This argument is howeverweakened by the fact that uncontroversial locative relations can also beexpressed by a transitive schema in languages which have chosen this way ofexpressing possession, cf. English verbs like contain, hold, keep, occupy (aposition), etc. Since the transitive Agent-Patient schema thus covers both

Introduction 9possession and location, it cannot be used as an argument in favour of thenon-locative nature of possession. There are, on the contrary, as shown by thedata adduced so far good reasons for believing that the subject of have issemantically a Location only posing syntactically as an Agent.2.2 The two-sided nature of possession: have vs. belongA major distinction within predicative possession is the one between haveand belong constructions, i.e. on the one hand constructions with thePossessor as topic and grammatical subject, on the other constructions havingthe Possessum as topic and subject. This distinction is reminiscent of the twopossessive structures, the existential and the locative, discussed in 2.1.1 above.And historically, it seems primary insofar as it can be traced back to theearliest Indo-European, cf. Benveniste (1949, 1966a); Baron (1997: 117). Butalso cross-linguistically it seems fundamental: it is apparently found in alllanguages (Heine 1997: 33), although Isačenko (1974: 64f.) expresses someserious reservations as to its universality.There seems to be a characteristic asymmetry between the two constructions,as contended e.g. by Seiler (1983a: 61ff.): whereas have-constructionstend to be very polysemous, but to include ownership in their meaning, cf.Heine (1997: 32), belong-constructions are more restricted denoting onlyownership, i.e. a sub-category of possession. We do not think, however, thatthe importance of the asymmetry should be exaggerated. belong is indeedpolysemous and is not restricted to expressing ownership, at least in theEuropean languages we are familiar with: Eng. belong, Fr.appartenir, Span.pertenecer, Germ. gehören, Dan. tilhøre, all have both (concrete) locative andpossessive meanings, the two being often distinguished by partially differentconstructions. This is for instance the case in French, where two differentchoices of clitics, lui and y, reveal the possessive and the locative readingsrespectively (cf. Kotschi 1981: 93). And a similar distinction is found inDanish, where the systematic alternation between a verbal prefix, with apredominantly abstract meaning, and a postverbal particle, with a predominantlyconcrete meaning, is put to use in order to distinguish the possessivefrom the (concrete) locative meaning of the verb:(13) Frencha. Cette équipe, Jean y appartient.this team Jean there belongs‘This team, Jean belongs (= is part of) to it’

10 Michael Herslund and Irène Baronb. Cette équipe, Jean lui appartient.this team Jean it-dat belongs‘This team, Jean belongs (=is the property of) to it’Danisha. Bogen tilhører biblioteket.Book-def belongs library-def‘The book belongs to the library’b. Bogen hører til på nederste hylde.Book-def belongs to on bottom shelf‘The book belongs on the bottom shelf’cf. also the different translations and choice of preposition in the two cases inEnglish.Even if the two verbs share a common core of meaning, i.e. location andownership, belong seems to be more concrete than have, tobemoreofa“fully’’ lexical verb than have (cf. Isačenko 1974: 64f. and Seiler 1983a: 62ff.),the latter having, as may be seen, a larger array of increasingly abstractmeanings:HAVELocationOwnershipAvailabilityKinshipAbstractpossession...Figure 2.BELONGBasically however, the two verbs seem to be converses and to constitute whatlooks very much like a voice distinction: the transitive have-constructiontopicalises the Possessor, as the active topicalises the Agent, whereas theintransitive belong-construction topicalises the Possessum as the passivetopicalises the Patient. And it seems indeed natural that the belong-constructionand the passive should both have a more restrictedmeaning potential than the have-construction and the active, because theoccurrence of the different constructions is determined by the lexical contentof their subjects, and the possible class of subjects of belong, the Possessum,is inherently less “topic-worthy’’ than that of have, the Possessor. As the

Introduction 11passive is a marked choice vis-à-vis the active, the belong-construction ismarked vis-à-vis the have-construction, and is therefore more restricted andprecise in meaning.But one could ask what exactly it is that makes ‘ownership’ contrary to‘availability, kinship, etc.’, so important that it can be lexically expressed intwo diametrically opposite ways, i.e. why is it precisely the ownership submeaningof possession that exhibits a lexical choice? The answer could be thatthe ownership relation is central and even prototypical (cf. Bartning 1993: 78ff.and Heine 1997: 39) because ownership — of course a very culture dependentnotion — is the most salient representative of the possessive relation, i.e. abasically locative relation between two distinct entities enriched with “somethingmore’’, this “more’’ being of an institutionalised or legal sense. Ownershipconstitutes thus a central point on the semantic scale stretching frominalienable possession, or the Part–Whole relation, to mere availability:Part–Whole Ownership Availability< >Figure 3.As seen in Figure 2 above, belong selects only the ownership meaning,besides its basic locative meaning, and is, as the marked choice, necessarywhen this meaning is crucial, cf. (14):(14) This watch was your father’s, so now it belongs to you.? This watch was your father’s, so now you have it.The concept of ‘ownership’, which occupies the central point on the scale ofFigure 3, thus seems important enough to be expressible in two differentways by a lexical opposition between an unmarked (have) and a marked(belong) member. But it is, again, the two-sided nature of possession, thesolidarity between Possessor and Possessum, which makes this oppositionavailable. And the unmarked status of have in this opposition does notnecessarily amount to its reduction to a kind of copular verb, as pointed outby McGregor in his study of the verb have in languages of the AustralianNorth-West.The marked status of belong-constructions is also related to the fact thatlanguages with a have-verb seem to have developed such a verb precisely inorder to align their possessive and experiential construction with their canonicalagentive pattern, viz. John has a boat and John feels a pain on a par withJohn makes a boat, thereby identifying the three major semantic functions

12 Michael Herslund and Irène BaronAgent, Location, Experiencer, with Subject and Topic in one and the samesyntactic schema in Figure 4.{Agent, Loc, Exp} – {Patient, Arg, Stim}Subject – ObjectTopic – CommentFigure 4.The basic distinction between have and belong is crucially linked to thefundamental aspects of the semantics of possession as noted above in 2.1 inthe discussion of the semantic roles associated with it, aspects which hithertohave not fully received the attention they deserve, we believe. The two-sidednature of possession, a nature it shares with location and experience, has manywider and important ramifications among which the syntactic phenomenarelated to Russian verbs with the general meaning ‘give, present, supply’studied by Durst-Andersen and the syntax of indirect objects and co-subjectsin Danish discussed by Heltoft in his contribution. The fundamental solidaritybetween the two members of the possessive relation also accounts for themorphological distinction between subject-like and object-like possessivepronouns and determiners, as studied by Seiler (1983b), and it is furtherreflected in the extended uses of possessives presented by Fraurud in thepresent volume.3. Attributive possessionWhen we come to attributive possession, we enter a domain where parameterssuch as alienable vs. inalienable become salient, cf. the contributions by Riegel,Dahl and Koptjevskaja-Tamm, Spanoghe, and Mithun. Many languages makea distinction between these two kinds of possession, and the distinction ismainly found in attributive constructions, and in constructions with externalpossessors, to which we return in 4. below.3.1 Alienable and inalienable possessionAttributive possessive constructions are in general much more polysemousthan the predicative constructions. Whereas Mary has a house probably onlyallows one reading, the corresponding attributive construction, Mary’s house,

Introduction 13allows several, viz. ‘the house Mary has designed, lives in, is always talkingabout, etc.’, besides of course ‘the house Mary owns’. This inherently polysemousnature of genitival constructions has been pointed out several times, cf.e.g. Benveniste (1966b), Bartning (1993), Sørensen (1997b) as well as thecontributions to the present volume by Bartning and Høeg Müller. It is relatedto the fact that the predicative construction, as mentioned, asserts the possessivelink between two entities, and sometimes even lexically specifies it byusing concrete verbs like own, possess, etc., whereas this link is simply presupposedin the attributive construction. That is why the predicative constructionis primarily used in connection with what Seiler (1983a: 62ff.) calls ‘establishedpossession’, as opposed to ‘inherent possession’, because the verb itselfestablishes and asserts the possessive link. With inherently possessed entities,i.e. relational nouns, the predicative construction is rather marginal, at least inthe unmodified affirmative, due to the uninformative nature of expressionslike Mary has a mother. But when the Possessum is modified, viz. Mary hastwo mothers, or negated, viz. The Greek goddess Athena had no mother, thesentences become informative and potentially relevant.Attributive possession seems to be the preferred domain for distinctionsalong the alienable-inalienable dimension. If attributive constructions presupposea possessive relation rather than assert and establish it, it is obvious thatthey are unambiguous only when the possessive relation is so to speak alreadycontained in the Possessum noun, i.e. when it is a relational noun. Now, apartfrom kinship terms, other so-called relational nouns are inherently ambiguousinsofar as they can be used both relationally and non-relationally. Kinshipterms, cf. Dahl and Koptjevskaja-Tamm’s paper, are indeed special insofar asthey denote both persons and roles, i.e. a relational noun such as fatherdenotes both an adult male and a certain kinship relation, but they name onlythe relation. Other “relational’’ terms such as body parts simply denote thingsand do not specify whether these things should be seen as objects or as parts ofother things, i.e. nose denotes simply a nose and nothing else. So whereas (15a)is hardly ambiguous, (15b) potentially is:(15) a. Mary’s fatherb. Mary’s noseThat is probably why many languages feel the necessity of disambiguatingphrases like (15b) by specifying whether nose should be seen as a part of Mary’sface, or as an object belonging to or produced by Mary, because she is a sculptoror a plastic surgeon. And in languages making such a distinction, many nouns

14 Michael Herslund and Irène Baronwhich are not offhand conceived as relational can occur in the inalienableconstruction thereby acquiring a special meaning or status as being presentedas extensions of the Possessor rather than as autonomous objects, cf. e.g.Herslund (1980: 106ff.); Spanoghe (1995: 49); Velázquez-Castillo (1996: 69ff.). 63.2 Different genitivesAnother aspect of attributive possession is the fact that certain languages makeuse of different genitival constructions, also outside the alienable-inalienabledistinction. The best known example of this is English with its “Saxon’’’s-genitive and its “Norman’’ of-genitive. Also in Danish there is a distinctionbetween the preposed genitive and different postposed prepositional constructions:(16) a. hestens hovedhorse-def-gen head‘the horse’s head’b. hovedet på hestenhead-def on horse-def‘the head of the horse’Such alternations are however far from random but subject to systematicchoices along the theme-rheme or topic-comment dimension, cf. Baron (1996a,1996b, 1997). The a. construction is preferred when hesten ‘the horse’ is topical,i.e. when talking about or describing an already introduced horse, the b.constructionwhenitishovedet ‘the head’ which is topical, i.e. talking forinstance of a riding accident where several heads may be involved. The mostinteresting aspect of such distinctions is however that languages with twogenitives thus reproduce within the noun phrase the basic existentialpossessive1 vs. locative-possessive 2 and the have-belong distinctions, cf.Sections 2.1 and 2.2 above: they distinguish a construction with the Possessoras theme-topic, from a construction with the Possessum as theme-topic, justlike the predicative constructions do.4. Between predicative and attributive possession: External possessorsOne much studied area within the grammar of possession is that of externalpossessors or possessor ascension. Such labels cover the cases where the

Introduction 15Possessor is not expressed in the same noun phrase as the Possessum, viz. as agenitival dependent of the Possessum as in (17):(17) She slapped Tom’s face.but is instead realised as an argument of the verb:(18) She slapped Tom in the face.What we have is in a sense something in between predicative and attributivepossession: the external possessive construction shares with predicativepossession the feature that the possessive link between Possessor andPossessum is conveyed by a verb; but it shares with attributive possession thefeature that the possessive link is not asserted by a verb, but presupposed. Theeffect of the external construction is thus a promotion of the Possessor, whichinstead of being realised as a dependent of the Possessum becomes a primaryclause member, cf. the label ‘possessor ascension’.4.1 Constraints on External PossessionNot all possessive structures can be realised externally. There are two mainconstraints on external possession. The first of these is that the link betweenPossessum and Possessor is, or can be construed as, a Part–Whole relation, i.e.an inalienable possessive relation. The second constraint, which is morevariable cross-linguistically, cf. below, is that the verb must be dynamic. Thecombined outcome of these two constraints is that the Possessor, promoted tothe status of primary clause member, is presented as in some way affected bywhatever the verb does to the Possessum. This is illustrated in the followingexample from Serbo-Croat (Partridge 1964: 105):(19) Želja za duvanom dražila mu je jezik.desire for tobacco tickled him-dat aux tongue‘The desire for tobacco tickled his tongue’The presupposed Part–Whole relationship between Possessum and Possessorbased upon the lexical content of the two terms is sufficient in order to ensurethe correct interpretation. In (19) jezik ‘tongue’ can only be construed as partof mu ‘him’.4.1.1In most Indo-European languages structures with external possessors are thus

16 Michael Herslund and Irène Baronmainly limited to body or similar organic parts, but in the Balkan languages anexternal Possessor in the dative, often homonymous with the genitive, iscurrent with kinship terms as well, cf. the following examples from Rumanianand Serbo-Croat:(20) RumanianEl mi-e frate.he me-dat is brother‘He is my brother’Serbo-CroatOn mi je brat.he me-dat is brother‘He is my brother’As shown by examples (20) the Balkan languages also have external Possessorswith stative verbs contrary to most other European languages. This generalpreference for external Possessors over internal ones (König and Haspelmath1998: 567f.) seems to be a common feature of the Balkan languages. It musthowever be related to the general structural feature that the Balkan languagestend to use verbal clitics (dative or genitive pronouns) instead of nominaldeterminers with all kinds of nouns. 7 This feature can in fact be traced back toHellenistic Greek koinè:(21) F× ,É AXJDoH, 6"Â ¦BÂ J"bJ® J± BXJD‘ oÆ6o*o:ZFT :oL J¬

Introduction 17(22) DanishDe renoverede husets facade.They renovated house-def-gen facade‘They renovated the facade of the house’De facaderenoverede huset.They facade-renovated house-defSwedish (Josefsson 1993):Läkaren opererade patientens hjärta.doctor-def operated patient-def-gen heart‘The doctor operated the patient’s heart’Läkaren hjärtopererade patienten.Doctor-def heart-operated patient-def‘The doctor performed heart surgery on the patient’Similar procedures seem widespread in Australian languages, cf. Evans (1996),Walsh (1996), and is also found in a polysynthetic language such as Mohawk(Mithun 1996: 643):(23) tá:khwasket2:sg:agt/1:sg:pat-back-scratch‘Scratch my back!’ (Literally: ‘Backscratch me.’)4.2 External possessors as experiencersThe interpretation of external possession as a promotion of the Possessor,which is presented as affected by the verbal action via the Possessum, leadsnaturally to the question of the semantic interpretation of such constructions.In many languages the dative case is used with external possessors. Since thedative is, par excellence, the case of the Experiencer, the question is of coursewhether external Possessors are really still Possessors or should rather be seenas Experiencers, cf. Bolkestein (1983) and the papers by Bolkestein andMithun in the present volume. One argument in favour of such a conclusionis of course the fact that external constructions mainly or even only occur withdynamic verbs denoting situations where the Possessor is presented as physicallyaffected through the Possessum, cf. e.g. French:

18 Michael Herslund and Irène Baron(24) Il lui a baisé la main.he her-dat aux kissed the hand‘He kissed her hand’In most languages external constructions are therefore not found with sensoryverbs. But this constraint is not absolute, cf. Italian (König and Haspelmath1998: 568):(25) Le ho visto le gambe.her-dat have-1sg seen the legs‘I have seen her legs’Constructions like (25) are also current in Portuguese, see the contribution bySpanoghe. And similar constructions are found in French too, where one forinstance has things like:(26) On te voit le dos.indef you-dat see-3sg the back‘One can see your back’This is not, however, a serious argument against the Experiencer interpretationsince utterances such as (26) are typically used not as pure statements of facts,but rather as mild reproaches or warnings. The Experiencer interpretationthen becomes relevant, because the Possessor is not threatened by somethingphysical, but as an Experiencer gets exposed to something potentially disagreeable.The Experiencer aspect of the external Possessor further accounts for thedifference which is often observable between an “inadvertent’’ reading withinternal (attributive) possession as opposed to a “willful act’’ with externalpossession, as in the following examples from French, cf. examples (17) and(18) above, and Herslund (1997: 33f.):(27) a. Elle a heurté le dos de son mari.she aux hit the back of her husband‘She bumped into her husband’s back’b. Elle a heurté son mari dans le dos.she aux hit her husband in the back‘She hit her husband in the back (in order to make him . . .)’As seen in 2.1 above, possessive structures have important features in commonboth with Location and with Experience structures, most conspicuously in thesolidarity between the two arguments of such structures. There is thus an

Introduction 19important shared semantic core between the three, and external possession canthen, according to language, be represented as either locative or as experiential.In French, for instance, the two ways of realising the external constructionexist, either with the Possessor as direct object and the Possessum as a place,(28a), or the Possessor as a dative (Experiencer) and the Possessum as thedirect object, (28b):(28) a. Elle l’ a frappé sur la joue.she him-acc aux hit on the cheek‘She hit him on the cheek’b. Elle lui a frappé la joue.she him-dat aux hit the cheek‘She hit his cheek’In Danish, the only way seems to be the locative, either with the Whole(the Possessor) retaining the role as a place, the Part (the Possessum) beingrealised as a direct object, (29a), or with the Part taking over the role as aplace while the Whole (the Possessor) as in French is promoted to directobject, (29b):(29) a. Hun vred armen om på ham.she twisted arm-def part on him‘She twisted his arm’b. Hun slog ham på kinden.she hit him on cheek-def‘She hit him on the cheek’The second construction, the Part as place, (29b), is reminiscent of the subplaceof predicative possession discussed in 2.1.3 above. In both predicativepossession with the Possessor as subject and in external possession with thePossessor as object a sub-place can be specified. Notice the parallelism betweenthe following:(30) a. Han har et ar på kinden.he has a scar on cheek-def‘He has a scar on his cheek’b. Hun slog ham på kinden.she hit him on cheek-def‘She hit him on the cheek’In the locative version of external possession, (30b), only the sub-place subsists

20 Michael Herslund and Irène Baronas a remnant of the basically locative structure, the Possessor (the principalplace) being promoted to direct object of the clause.4.3 Textual functions of external possessionIn general, the external construction operates, as seen, on a Part–Wholeconfiguration and its function is to highlight the Possessor, the Whole, eitherby syntactically promoting this constituent or by demoting the Possessum, thePart, syntactically. This interpretation can be further corroborated by Rumanian.In the external construction of this language the demotion of the Parthas as a corollary that when it is realised as the object it cannot be marked bythe preposition pe or the pronominal doubling, which is otherwise used withindividualised object phrases as in (31), cf. Manoliu-Manea (1994: 29):(31) O văd pe fată.her-acc see-pres-1sg prep girl‘I see the girl’But in (32) there is no preposition and no pronominal doubling of the object:(32) Ion şi-a trimis copilul la părinţi.Ion refl-has sent child-def to parent-pl‘Ion has sent his child to his parents’This contrast shows the dependent status of the object phrase (the Possessum)in external possession as being part of a semantic unit, Whole–Part, where itis incapable of being individualised or independently highlighted: copilul ‘thechild’ of (32) does not enjoy the same syntactic autonomy as fată ‘girl’ of (31).As shown by different authors (Velázquez-Castillo 1996: 165ff., Mithun1996 and the present volume, Herslund 1997: 32ff.), the use of externalpossession is in most languages primarily a question of textual presentationrather than of lexical classification. It is sometimes useful and relevant topresent a Possessor not just as a Possessor or as the Whole of a Part–Wholestructure, but first and foremost as a participant in the related events, forinstance as an Experiencer affected by something (un)pleasant. And contraryto what is often claimed, cf. e.g. König and Haspelmath (1998: 531), there iscertainly no reason to restrict the occurrence of external possessors to animates.It is, again, a matter of presentation and textual salience. In French forinstance, if for some reason one is mad and wants to take it out on an inanimatething, the external construction is the only appropriate choice:

Introduction 21(33) il (. . .) revint vers la carriole et lui assena un coup dehe (. . .) returned towards the cart and it-dat stroke a blow ofpied dans les rouesfoot in the wheels‘he (. . .) came back to the cart and gave it a kick in the wheels’ (DurasUn barrage contre le Pacifique 18)5. ConclusionSumming up, we can say that possession is not a primitive linguistic notion. Itmay be a primitive notion, but in order to be expressed in language it seems todraw on different pre-existing and/or more basic lexico-grammatical structures.The view that possession so to speak borrows its expression from otherareas is clearly stated and studied in great detail in Heine (1997), where severalsuch “source schemata’’ are identified, the most widespread of which seem tobe the action schema (e.g. have-constructions) and the location schema (e.g.the Russian U menjá kniga ‘By me book — I have a book’). The source schemaapproach is further illustrated in the present volume by the same author’sdescription of Kxoe (Central Khoisan, northeast Namibia), where the locationschema prevails. One central claim of this source schema approach is thatlanguages tend to favour one such schema at the expense of others. But apartfrom the fundamental bifurcation into have- and belong-constructions asdiscussed in 2.1.2 above, it is in fact quite common for languages to have morethan one schema, as argued for Danish by Togeby in the present volume. Butit is not impossible to decide on such issues and stipulate that Danish is indeeda language whose core possessive construction is the have-variant of Heine’saction schema.Even if possession has many ways of manifesting itself in language, it doesnot seem impossible to maintain the view expressed in the preceding pagesthat possession is in crucial and fundamental ways linked to location. The coreof the argument in favour of this view is that possession, like location, is abinary, solidaric relation between two entities which only receive their semanticinterpretation in virtue of one another, cf. 2.1. Here we return to theparallelism pointed out above in 2.1 between possession, location and experience:in possessive constructions, one can, precisely because of the fundamentalsolidarity of the two, choose either the Possessor or the Possessum assubject just as in locative and experiential constructions one can choose either

22 Michael Herslund and Irène Baronthe Location and the Experiencer as subject (The table has some books on it,Jane is afraid of spiders) or the Argument and the Stimulus (The books are onthe table, Spiders scare Jane). This explains the fact that possession sharesimportant and obvious features with both location and experience. But it doesof course not amount to saying that possession, which remains a very complexand multifaceted notion, can be reduced to location. It is more, but we thinkthat location because of its quite fundamental status as a primitive and veryconcrete notion is indeed the basic ingredient of possession.Notes1. The construction is actually clearer in Medieval French, where the pronoun y is onlyused with “full pronominal value’’, i.e. not pleonastically as in Modern French. So it doesnot occur when a locative is expressed in the same clause:(i) pres de ci a gentclose to here has people‘There is someone nearby’ (La mort le roi Artu, 48.31)2. Cf. also Isačenko’s (1974: 76) conclusion that have is be + transitivity.3. Cf. also in this connection Russian imet’sja ‘have’ + refl as synonymous with existentialbyt’ ‘be’, Isačenko (1974: 60).4. Similar phenomena are found in French, cf. Leclère (1993: 7). Compare also the Englishtranslations of the examples in (8) and (9).5. The primary function of the preposition hos, etymologically from hus ‘house’, seems tobe to convert a person into a place, cf. Herslund (1997: 30).6. See e.g. the articles by Chappell, Manoliu-Manea and Bavin in Chappell and McGregor(1996), and the comprehensive bibliography on the subject in that volume.7. On the use of the dative of the personal clitic pronoun instead of a possessive determinerin Rumanian in general, see Cristea (1974), Popescu-Ramírez and Tasmowski-De Ryck(1988), and Timoc-Bardy (1996).ReferencesBaron, I. 1996a. “Information structure and the anatomy of noun phrases. The expressionof subject and object in Danish noun phrases’’. In Content, Expression and Structure:Studies in Danish Functional Grammar [Studies in Language Companion Series 29],E. Engberg-Pedersen et al. (eds), 235–259. Amsterdam and Philadelphia: JohnBenjamins.

Introduction 23Baron, I. 1996b. “L’expression du possesseur dans le Syntagme Nominal danois’’. In Larelation d’appartenance [Faits de Langues 7], 33–42. Paris: Ophrys.Baron, I. 1997. “Possession in Noun Phrases: A Functional Analysis’’. In Possessive Structuresin Danish [KLIMT 3], I. Baron and M. Herslund (eds), 115–130. CopenhagenBusiness School.Baron, I. and Herslund, M. 1997a. “The Danish Verb have and the Notion of Possession’’.In Possessive Structures in Danish [KLIMT 3], I. Baron and M. Herslund (eds),131–157. Copenhagen Business School.Baron, I. and Herslund, M. 1997b. “Place and Sub-Place: Locative Specifications inhave-Clauses’’. In From Words to Utterances in LSP [Copenhagen Studies in Language20], D. Faber and F. Sørensen (eds), 5–21. Copenhagen Business School.Baron, I. and Herslund, M. 1998. “Support Verb Constructions as Predicate Formation’’. InThe Structure of the Lexicon in Functional Grammar [Studies in Language CompanionSeries 43], H. Olbertz, K. Hengeveld and J. Sánchez García (eds), 99–116. Amsterdamand Philadelphia: John Benjamins.Bartning, I. 1993. “La préposition de et les interprétations possibles des syntagmesnominaux complexes. Essai d’approche cognitive’’. In Les prépositions: méthodesd’analyse [Lexique 11], A.-M. Berthonneau and P. Cadiot (eds), 163–191.Bavin, E. L. 1996. “Body Parts in Acholi: Alienable and Inalienable Distinctions andExtended Uses’’. In H. Chappell and W. McGregor (eds), 1996: 841–864.Benveniste, E. 1949. “Sur l’emploi des cas en hittite’’. Archiv Orientálni XVII: 44–45.Benveniste, E. 1966a. “’’Etre’’ et “avoir’’ dans leurs fonctions linguistiques’’. In Problèmes delinguistique générale, 176–186. Paris: Gallimard.Benveniste, E. 1966b. “Pour l’analyse des fonctions casuelles: le génitif latin’’. In Problèmesde linguistique générale, 140–148. Paris: Gallimard.Bolkestein, A. M. 1983. “Genitive and Dative Possessors in Latin’’. In Advances in FunctionalGrammar, S. C. Dik (ed.), 55–91. Dordrecht: Foris.Chappell, H. 1996. “Inalienability and the Personal Domain in Mandarin Chinese Discourse’’.In H. Chappell and W. McGregor (eds), 1996: 65–109.Chappell, H. and McGregor, W. (eds). 1996. The Grammar of Inalienability. A TypologicalPerspective on Body Part Terms and the Part-Whole Relation. Berlin-New York: Moutonde Gruyter.Christie, J. J. 1970. “Locative, Possessive and Existential in Swahili’’. Foundations ofLanguage 6: 166–177.Clark, E. V. 1978. “Locationals: Existential, locative and possessive constructions’’. InUniversals of Human Language [Vol. 4: Syntax], J. Greenberg (ed.), 85–126. Stanford:Stanford University Press.Cristea, T. 1974. “Remarques sur le datif possessif en roumain et en français’’. Bulletin de laSociété Roumaine de Linguistique Romane X: 5–14.Evans, N. 1996. “The Syntax and Semantics of Body Part Incorporation in Mayali’’. In H.Chappell and W. McGregor (eds), 1996: 65–109.Freeze, R. 1992. “Existentials and other locatives’’. Language 68: 553–95.Guillet, A. and Leclère, C. 1992. La structure des phrases simples en français. Constructionstransitives locatives. Geneva-Paris: Droz.

24 Michael Herslund and Irène BaronHeine, B. 1996. “Grammaticalization and language universals’’. In La relationd’appartenance [Faits de langues 7], 11–22. Paris: Ophrys.Heine, B. 1997. Possession. Cognitive sources, forces and grammaticalization [Cambridgestudies in linguistics 83]. Cambridge: Cambridge University Press.Herslund, M. 1980. Problèmes de syntaxe de l’ancien français. Compléments datifs et génitifs.[Etudes romanes de l’Université de Copenhague 21]. Copenhagen: Akademisk Forlag.Herslund, M. 1996. “Partitivité et possession inaliénable’’. In La relation d’appartenance[Faits de Langues 7], 33–42. Paris: Ophrys.Herslund, M. 1997. “Partitivity and Inalienable Possession’’. In Possessive Structures inDanish [KLIMT 3], I. Baron and M. Herslund (eds), 1–44. Copenhagen BusinessSchool.Isačenko, A. V. 1974. “On ‘HAVE’ and ‘BE’ Languages’’. In Slavic Forum. Essays inLinguistics and Literature, M. S. Flier (ed.), 43–77. The Hague-Paris: Mouton.Josefsson, G. 1993. “Noun Incorporating Verbs in Swedish’’. In The Nordlex Project. LexicalStudies in the Scandinavian Languages [LAMBDA 18], M. Herslund and F. Sørensen(eds), 274–304. Copenhagen Business School.Kotschi, T. 1981. “Verbvalenz im Französischen’’. In Beiträge zur Linguistik desFranzösischen, T. Kotschi (ed.), 80–122. Tübingen: Gunter Narr.König, E. and Haspelmath, M. 1998. “Les constructions à possesseur externe dans leslangues d’Europe’’. In Actance et valence dans les langues de l’Europe [EmpiricalApproaches to Language Typology, Eurotyp 20–2], J. Feuillet (ed.), 525–606. Berlin-New York: Mouton de Gruyter.Leclère, C. 1993. “Classes de constructions directes sans passif’’. In Sur le passif [Langages109], G. Gross (ed.), 7–31. Paris: Larousse.Lyons, J. 1967. “A note on possessive, existential and locative sentences’’. Foundations ofLanguage 3: 390–396.Lyons, J. 1968. “Existence, location, possession and transitivity’’. In Logic, methodology andphilosophy of science III, B. van Rootselaar and J. F. Staal (eds), 495–504. Amsterdam:North-Holland.Manoliu-Manea, M. 1994. Discourse and Pragmatic Constraints on Grammatical Choices. AGrammar of Surprises [North Holland Linguistic Series 57]. Amsterdam: Elsevier.Manoliu-Manea, M. 1996. “Inalienability and Topicality in Romanian: Pragma-Semanticsof Syntax’’. In H. Chappell and W. McGregor (eds), 1996: 711–743.Mithun, M. 1996. “Multiple Reflections of Inalienability in Mohawk’’. In H. Chappell andW. McGregor (eds), 1996: 633–649.Nygaard, M. 1906. Norrøn Syntax. Oslo: Aschehoug [1966 2 ].Partridge, M. 1964. Serbo-Croatian. Practical Grammar and Reader. New York, London,Toronto: McGraw-Hill.Popescu-Ramírez, L. and Tasmowski-De Ryck, L. 1988. “Thématicité et possessivité enroumain’’. Lingvisticæ Investigationes 12: 303–335.Seiler, H. 1983a. Possession as an operational dimension of language. Tübingen: Gunter Narr.Seiler, H. 1983b. “Possessivity, Subject and Object’’. Studies in Language 7: 89–117.Spanoghe, A.-M. 1995. La syntaxe de l’appartenance inaliénable en français, en espagnol et enportugais. Bern: Peter Lang.

Introduction 25Sørensen, F. 1997a. “Possession as Location’’. In Possessive Structures in Danish [KLIMT 3],I. Baron and M. Herslund (eds), 158–168. Copenhagen Business School.Sørensen, F. 1997b. “The Nightmare of the Genitive’’. In The Valency of Nouns [OdenseWorking Papers in Language and Communication 15], K. van Durme (ed.), 119–134.Timoc-Bardy, R. 1996. “Appartenance implicite vs appartenance explicite en roumain’’. InLa relation d’appartenance [Faits de Langues 7], 241–250. Paris: Ophrys.Velázquez-Castillo, M. 1996. The Grammar of Possession. Inalienability, Incorporation andPossessor Ascension in Guaraní [Studies in Language Companion Series 33]. Amsterdamand Philadelphia: John Benjamins.Walsh, M. 1996. “Body Parts in Murrinh-Patha: Incorporation, Grammar and Metaphor’’.In H. Chappell and W. McGregor (eds), 1996: 327–380.

Chapter 1The operational basis of possessionA dimensional approach revisitedHansjakob Seiler1. IntroductionThere has been a considerable thrive in studies on possession within the pastfew years: No less than three major books (Chappell and McGregor 1996;Taylor 1996; Heine 1997), one specialized monograph (Lehmann 1995), onecollection of working papers (Baron and Herslund 1997), and this probablydoes not exhaust all that would deserve mention. My own publications wereearlier but also not few: Seiler 1973a, 1973b, 1982, 1983a, 1983b. What is sofascinating about this topic? I personally think that possession is a domainloaded with paradoxes, and the challenge consists in resolving them — whichwe shall try in the following pages. Let us point out jut a few of them:1. When I tell my non-linguistic friends that I am working on possession, thereaction is mostly one of insightful understanding: ‘ownership’, ‘belonging’,‘appurtenance’ would come to their minds. Not so some linguists, who find itmuch harder to detect something palpable behind this notion. Some wouldflatly deny that possession should have grammatical status (Weinrich 1993:433).Others, unwilling to cope with the extraordinarily wide range of variantexpressions try an escape via reductionism: reducing all of possession to a ‘‘deepstructural’’ configuration with ‘to have’; reducing ‘‘inalienable’’ possession to a‘‘deep structural’’ dative; reducing possessive constructions to subvarieties oflocal, directional, instrumental and comitative expressions (Heine 1997:76).While it is true that local and other constructions are drawn upon for thepurpose of representing possessive relations, it is also true that after completionof such a shift the resulting expressions no longer are purely local etc., butsomething new, called possession. We want to know what this ‘‘something’’ is.2. The major distinction within the domain of possession is between “inalienable’’or inherent, and “alienable’’ or established relation. 1 For several

28 Hansjakob Seilerstatements regarding this distinction a converse or opposite statement mayalso be true: (a) “Inalienability’’ is the older construction, it is more archaic,it is unmarked. ‘‘Alienability’’ is the younger construction, it is derived, it ismarked. There is ample evidence in the data in support of such a statement.Nevertheless, a view to the contrary is also being advanced. One of thearguments would go like this: ‘‘Inalienable’’ possession is attributive,‘‘alienable’’ possession is predicative. Attributive expressions normally derivefrom predicative expressions, and not the other way round. (b) One strategyfor signalling ‘‘inalienability’’ is to have the Possessor obligatory, i.e. itmust be represented, e.g. by pronominal affixation to the Possessum, asinmany American Indian languages (Seiler 1983a: 20). The opposite strategyalso occurs in suppressing the Possessor and having the definite articleinstead (Seiler 1983a: 18). (c) Particularly confusing is the situation of casemarking, especially genitive and dative: While there is evidence — to bepresented later — that the genitive signals the more intimate, and the dativethe less intimate relation, there is other evidence pointing to the exactopposite.A resolution of these antinomies cannot come from a one-sidedly categorialand morpho-syntactic approach. A dynamic, processual one must supplementit. To characterize this approach, I should submit the following tenets:1. Any attempt at defining Possession in purely categorial terms is doomedto failure.2. Any attempt at categorially delimiting ‘‘inalienable’’ from ‘‘alienable’’Possession will run into difficulties.3. In order to understand the workings of Possession it is necessary toconsider all possible variant expressions in an integrating view instead oftreating some of them, e.g. the ‘‘inalienable’’, separately.4. Determining the notion of Possession is a matter of universality inlanguage. There is no direct connection — by way of inductive generalization— between the facts of an individual language and universality.Instead, we need a combined inductive–abductive procedure that leads usin a parcours through three levels: (a) a universal level that specifiesconceptual Possession in a constructivistic operational manner, (b) anintermediate level of general comparative grammar that specifies the‘‘menu’’ of morpho-syntactic techniques and categories representingPossession, and (c) the level of each individual language with its particularchoice from the techniques provided by the ‘‘menu’’.

The operational basis of possession 29In what follows I should propose to follow the “parcours’’ across the threelevels. At the end we shall see how this will help us to resolve some of theafore-mentioned puzzles.2. The operational basis of PossessionIt is assumed here that conceptual Possession is not a once and for ever fixedcategory, but that it comes about by way of construal in three operations calledDistinction, Genesis and Essence. 2 This can be visualized by the followingschema:DISTINCTIONEGO“inalien.”unspeci²cinversionWORLD“alien.”DISTINCTIONinherentgiven–control+contacttime-stableestablishedacquired+control–contacttemporaryGENESISFigure 1.ESSENCEThe schema can be supplemented by the following definition of Possession:Def. Possession is a relation under permanent construction. It is the EGO’sgradual proactive and retroactive appropriation of the things of theexternal WORLD.Although no grammatical categories are specified, we are not on extra-linguisticground: Both the schema and the definition specify the repraesentandum,i.e. that which is to be represented by the means of language, viz. conceptualcontent and the way it is construed. This, I would hold, is an integrated andsubstantial part of language activity.The schema and the definition say that Possession is the relationbetween EGO and the things of the external WORLD. EGO, here, is to betaken in an abstract sense, not confined to 1st person singular, but includinghuman and animate Possessors in general. Now, the relation of Possessioncan be seen and construed under different perspectives, to which different

30 Hansjakob Seileroperations correspond. The first operation consists in drawing a Distinction:between ‘‘inalienable’’ and ‘‘alienable’’. The dynamics of this operationimply a vantage point from which the process radiates in two oppositedirections, each toward its ideal maximum. The vantage point itself is neutralor unspecific with regard to the distinction. It is also the point of inversionfrom more ‘‘inalienable’’ to more ‘‘alienable’’, and vice versa. 3 The distinctionas such is substantiated by a number of parameters, each with two oppositepoles. The parameters are ordered vertically by way of implication: inherent/established→ given/acquired → –control/+control → +contact/–contact→ time-stable/temporary. It is important to note that the same parameters areoperative in specifying distinctions within other dimensions studied earlier,such as Identification, Participation, etc.Once Distinction is established, we need two further operations forconstruing the relation of Possession: The Genesis line marks the startingpoint and the dynamic of increasing appropriation. The easiest and thereforemost natural starting point is by having the relation inherent in the notionsrelating to our Self: body part terms, kin terms, ‘‘sphère personnelle’’ (Bally1926): Possessor inherent in the Possessum. In addition, Possessor ismaximally salient: EGO and EGO-like. The arrow of proactivity from left toright takes us via a center of neutrality and inversion toward a maximum of‘‘alienability’’ (and a minimum of ‘‘inalienability’’) and tells us how thingsoutside the ‘‘sphère personnelle’’ are being appropriated by Self. This is doneby stepwise proceeding toward more distanciated relations. It takes us to suchnotions as ‘acquiring’, ‘ownership’ and ‘belonging’, representing the Essenceof Possession.But even a maximally established and aquired and controlled Possessummust always be related back to the Possessor, viz. to EGO. Therefore, theretroactive Essence line from right to left symbolizes the reversed perspectivetaking us back to the starting point.Now, the claim is that all of this is not invented, but is reflected in thelinguistic data, as we shall see when getting to the respective levels. Specificallyit will help us to correctly account for these data. For example, we cantell why an expression like Mary left a fortune to Peter is not a possessiveexpression: It entails Possession, but Possessum (fortune) does not relateback to EGO (Mary). We have an ordinary transitive verb with a direct andan indirect object. We can also tell what a verb of Possession would looklike: take, grasp, seize would eventually qualify, but not give; buy mightqualify, but not sell. In sum, a verb of Possession may be transitive, but it

The operational basis of possession 31must in the first place refer to the mode of relationship between the tworelata, viz. EGO and the thing appropriated. For the same reason, part-wholerelations among inanimates, e.g. a table has four legs, are not possessive inspite of the numerous similarities on the morpho-syntactic level. There is noEGO among inanimates.The schema with the two converse arrows equally extending over theentire range of the Dimension tells us that ‘‘inalienability’’ and ‘‘alienability’’can be distinguished but not separated: They are co-active throughout, albeitat different ratios: the more ‘‘alienability’’, the less ‘‘inalienability’’, and viceversa.And with this the schema also tells us that between the extremes theremust be an area where the two opposite forces are about equally strong,thereby neutralizing each other. It is a transition area, unspecific with regardto the major distinction. This shows up on the level of General ComparativeGrammar, where it appears that within this area one and the same expression,e.g. a genitive or a dative, may either represent intimate or acquired Possessionaccording to context.3. Ordering in General Comparative GrammarFigure 2 visualizes the ‘‘menu’’ of morpho-syntactic ‘‘techniques’’ and categoriesbrought into play for the purpose of representing relations of Possession.The ordering as determined by the conceptual level is in continuousfashion of increasing vs. decreasing explicitness in the representation of thepossessive relation. It can be read off from the symbols on the top line.locdirN N— Nconn.N— Nclass.N— NcaseN— NcomitN— NexistN—NVNdeftopicUnspeci²c area of inversionRelation inherent,indicativeFigure 2.Relation established,predicativeNN symbolizes juxtaposition of two nominals, as e.g. in Classical Arabic.

32 Hansjakob SeilerNconn.N stands for a connector between the two nouns, as e.g., the ezafe inModern Persian, or the linker as a prominent feature in many Austronesianlanguages.Nclass.N symbolizes classifiers in possession, a technique found in Oceaniclanguages and in languages of the Americas. The technique is moreexplicit as compared to the preceding one: The classifier not only connectsbetween the two nouns but adds classificatory information about therelation as such or about the Possessor or the Possessum.NcaseN. Here we enter the area of non-specificity with regard to a cleardistinction between ‘‘inalienability’’ and ‘‘alienability’’. The genitive is theprototypical representative of this technique; to a lesser degree also thedative and the instrumental. But all these cases contract a twofold relationship:between noun and noun, and between these and a verb, therebyadding further information about the possessive relation.N:N represents a multiple choice situation, which is typical for non-specificityin transition. Here, the possessive relation could be represented as if itwere a relation of location: ‘the house near me’, or of directionality: ‘acousin to me’, or of accompaniment: ‘the man with the hat’, etc. Note thatthe extreme techniques in the continuum do not exhibit multiple choice,or only rarely so.NexistN stands for auxiliaries (‘be’, ‘have’) mediating between the two nouns.NVN. Full verbs offer a maximum of differentiable information about thepossessive relation.The schema is open for insertion of further techniques as they may be foundin the languages of the world, e.g. possessive pronouns, possessive adjectives(Slavic, Ancient Greek), possessive compounds (bahuvrihis), etc.In the following I shall confine myself to discussing two strategic areas ofthis continuum, especially in the light of my now more consistent three-levelapproach: Genitive/Dative as representatives of the transition area, and verbsof possession.From Roman Jakobson’s pioneering interpretation (1936/71: 37ff.) welearn that the genitive focuses upon the extent to which the entity takes part inthe message, by implying that the extent is not total. This means that thegenitive is open for interpretations that convey either intimacy or, on thecontrary, distance or separation of a relation (cf. our paradox (c) above). Inthe adnominal uses the genitive is not specialized for signalling Possession.What it does signal is mere Appurtenance. Inthe destruction of the city (by

The operational basis of possession 33the enemy) it simply transposes the verbal clause (the enemy) destroyed the cityinto a nominal syntagm (Benveniste 1962/66: 150). Regarding the two varietiesof genitive in English, viz. the inflectional vs. the prepositional, it has beenremarked (Anschutz 1997:31) that the inflectional is restricted to use withanimate Possessors where Possessor is old information before Possessumnew information: the President’s daughter. The prepositional construction ischosen where the Possessum is old information before the Possessor asnew information: . . . he was the father of Mrs. Needler. . . This can be accountedfor with reference to our conceptual level where the inflectionalconstruction is in principle ‘‘more inalienable’’ and the starting point of theoperation is the Possessor, who is thus ordered before the Possessum,whereas the prepositional construction is ‘‘more alienable’’ and the startingpoint is the Possessum, ordered accordingly before the Possessor. Thecategory genitive in English thus includes a variant ‘‘more inalienable’’ and avariant ‘‘more alienable’’.A further peculiarity of the genitive in English and in many other languagesexhibiting genitives is its openness to an almost infinite number ofpossible interpretations (Seiler 1973a: 199f.). Thus Peter’s house, beside itsmost current acceptation as ‘house which Peter owns’ may be interpreted,according to context, as: ‘house where Peter lives or used to live’, ‘house Peteris always talking about’, ‘house which Peter likes’, etc. The important observationto be made here is that verbs are always involved in these interpretations.They are not actually asserted but evoked. I suggested (Seiler loc.cit.) that, tothe extent that a relation between a genitive and such verbs becomes operativefor the interpretation of the utterance, the phenomenon is rather one oftransposition, as in the above, and marginal in the domain of Possession.Now, the dative. On the one hand, it may signal less intimate Possessionas compared to the genitive. Consider the examples of Latin:(1) a. liber est Marc-ibook cop Mark-gen‘The book belongs to Mark’b. liber est Marc-obook cop Mark-dat‘Mark has a/the book’On the other hand we have the dative acting as a marked means for expressinginherence (Seiler 1983a:43f.). Consider the examples of German and French asagainst English:

34 Hansjakob Seiler(2) a. Ich habe mir das Bein gebrochenJe me suis cassé la jambe‘I broke my leg’b. ? Ich habe mein Bein gebrochenJ’ai cassé ma jambeThe utterances in (2b) have a dubious status in both German and French. Apossible interpretation might imply that some leg which is not part of thespeaker’s own body is being talked about. The contrast is implemented by acombination of dative plus personal or reflexive pronoun plus definite articleplus a verb indicating action upon Self on the one hand, and possessivepronoun plus noun on the other. In the context given, the latter exhibits an‘‘alienating’’ force, referring to a non-body part, while the former is clearly‘‘inalienating’’. The corresponding English version is neutral in this respect. Itis important to see that the dative in these constructions contracts a twofoldrelationship viz. one of inherent Possession vis-à-vis the other noun, and onewith the verb which indicates that the agent is at the same time the experiencer.To sum up, we see the genitive and the dative in functions, which, accordingto context, can be either ‘‘more inalienable’’ or ‘‘more alienable’’ orneutral with regard to both. This can be accounted for with reference to theposition which case marking occupies within the overall dimension of Possession:An area of unspecificity and of inversion, with increasing intrusionof verbs into the relation between the Possessor and the Possessum.Next, a bird’s eye view on verbs of Possession, and this time a diachronicone. We are still on the level of General Comparative Grammar. Our conceptualschema plus definition tells us that the EGO’s proactive moves strive atappropriating more and more things of the external world, and that to eachproactive move there is a corresponding retroactive one relating these thingsback to EGO. Specifically, at a certain maximum of proactivity there must bea reversal into retroactivity. Our Comparative Grammar schema tells us that,corresponding to maximal proactivity there is a maximum of explicitness inthe representation of the possessive relation, and that this is best representedby verbs exhibiting a maximum of structural information, ideally by transitiveverbs. However, in the interest of retroactivity, transitivity may not come to itsfull bloom and must be turned back to stativity and reflexivity.These processes are mirrored in historical changes. We know of thedevelopment of a verb of existence plus dative into a pseudo-transitive verb ‘tohave’ plus accusative: Older Latin mihi est domus > Later Latin habeo domum

The operational basis of possession 35‘I have a house’: A proactive move. The following etymological connectionstestify to the inverse, the retroactive move: Gothic hafjan ‘seize’ and its Latincognate capio ‘seize’ vs. Old High German habēn ‘to have’, a stativizingderivative in -ē- indicating the state of the subject. The same development maybe inferred from a comparison of Old Slavic imēti ‘to have’ vs. jęti ‘to take’ andLatin emo ‘to buy’, or Lithuanian turēti ‘to have’ vs. tverti ‘to take’, againderivatives in -ē- with stativizing and detransitivizing value. Other means ofexpression achieving the same effect are: preterito-present forms as in Gothicaigan/aih ‘possess, own’, or, middle voice formations as in Sanskrit īse ‘to bemaster, control’, where possessing affects not the object but the subject.Summing up this survey of Possession in General Comparative Grammarit appears that the morpho-syntactic techniques of representing thepossessive relation can be ordered in continuous fashion from inherent toestablishing, and that such categories as connective, classifier, case, location etal., and finally verb have a prominent role within these techniques. In short,we are presented with the menu from which the individual languages exhibittheir proper selection.4. The level of individual languagesTurning now to the grammars of individual languages, we must assess theirextraordinary plasticity and deformation potential with regard to the menuprovided by General Comparative Grammar: Some languages lack connectives,others lack classifieres, still others lack case, etc. There may be gaps, syncretisms,or overextensions. What then about our dimensional approach? We predict thatwhile the categorial and morpho-syntactic look may vary from one language toanother, the operational program for representing Possession as delineatedon the conceptual level remains the same everywhere. The evidence, here again,can be found in the possible continuous orderings of expression means.I shall exemplify this in more detail with material from one particularlanguage: Cahuilla, an American Indian language of the Takic branch ofNothern Uto-Aztecan, spoken in Southern California, which I have describedin detail. In a special study (Seiler 1982: 185f.) I examined kinship expressionsfeaturing two pronominal elements, one for the Possessor, the other beingcoreferent with the Possessum, the kin term: Translation equivalents of suchEnglish expressions as ‘she is my niece’, ‘I am her niece’, etc. There arealtogether seven combinations possible. From the point of view of English, all

36 Hansjakob Seilerthese combinations seem to be parallel. Not so for Cahuilla, where they mustbe distributed over two gradient scales, and the relation between the scales isnot one of parallelism but one of inversion. The reason for this is presented bytwo widely differing types of expression distributed over two scales in amanner to be described presently. The two types are:(3) a. ’et-ne-nésithou-my-niece‘Thou (art) my niece’b. Ø-Ø-nésishe-her-niece‘She (is) her niece’(4) a. pe-y-nési-k(a)tobj:3sg-subj:3sg-niece-orient.dirher-she-niece-toward’‘She is her aunt’, lit.: ‘She is one who is (in a relation) toward her, theniece’b. pe-y-hénes-ka(t)obj:3sg-subj:3sg-aunt-orient.dir‘She is her niece’, lit.: ‘She is one who is (in a relation) toward her,the aunt’Note that both types are predicative, sentence-generating; that is the way kinrelations are represented throughout. Example (3a) exhibits the kin term‘niece’ obligatorily preceded by a subject and a possessive prefix, which mayboth be zero in third person, (3b). This is the unmarked, the inherent type. In(4a) we have the same kin term, but now preceded by an object plus a subjectprefix and followed by a suffix specifying directionality. The literal translationwould be ‘She is one who is (in relation) toward her, the niece’, whichamounts to saying ‘She is her aunt’. This is clearly a marked expression thatestablishes a relationship by showing that it has a point of departure (thesubject, ‘she’) and a goal (the object, ‘her’), coreferential with the kin term‘niece’, toward which the relation extends. It starts ‘‘from the other end’’, as itwere, i.e. from the pronoun referring to the reciprocal kin term ‘aunt’. Therehave been cultural reasons for using such a distanciating expression, viz. whenthe kin, the ‘aunt’, is no longer living and you are not supposed to refer to herdirectly. In order to construct an establishing expression corresponding to theinherent (3b) ‘She is her niece’, one would have to start from the reciprocalterm ‘aunt’, as in (4b).

The operational basis of possession 37What is of particular interest in connection with our dimensional–operationalapproach to Possession is the ways speakers chooseamong the two types of expression, and this is tabulated in Figure 3.Possessor Possessum Expression type Possessor Possessum Expression typemy 1 3 she Almost excl.inherenther 3 1 I Onlyestablishingmy 1 2 thou Mostlyinherentthy 2 1 I Mostlyestablishingthy 2 3 she Preferablyinherenther 3 2 thou Preferablyestablishingher 3 3 she Inherent orestablishingFigure 3.The meaning of the combination can be read off by going from right to left,e.g. first line left side ‘She is my niece’. Person is additionally symbolized bynumber in a circle to make the distance between them more salient. Theinformants either volunteered or accepted or rejected an expression type for agiven combination. We see from the chart that exclusive or near-exclusive useof one vs. the other type coincides with the maximal distance between thepersons (2 digits). We find a scale of decreasing exclusivity or increasingtolerance for the other of the two types as the distance between the personsbecomes smaller. When both are third person, both types are acceptable — asin our examples (3) and (4).The following generalizations can be derived from what has just beenoutlined: The constraints in the choice of one or the other expression type arecorrelated with a scale or hierarchy of proximity with regard to the speaker.The direct, unmarked type is chosen when the person of the Possessor isnearer to EGO, i.e. to the speaker, than the person of the Possessum. This isthe natural, the expected instance. It has to be chosen when the Possessor isidentical with the speaker, i.e. 1st person. The inverse type is chosen when theperson of the Possessum is nearer to the speaker than the person of thePossessor. It has to be chosen when the Possessum is identical with thespeaker. When both persons are third, Cahuilla has the choice of presentingeither the Possessor or the Possessum as being nearer to him and ofrespectively distanciating either the Possessum or the Possessor. Thischoice is exploited for the cultural purpose of obviating a deceased relative.What we have here is a typical area of inversion. The distributional schema of

38 Hansjakob Seilerthe two expression types reflects the functional-operational schema on theconceptual level with its two converse operations and with an area of inversion— in other words: It reflects what speakers do.The lesson to be learned from this description is this: Even though it iswidely claimed that kin expressions are ‘‘inalienable’’ and that ‘‘inalienable’’expressions are always attributive, this need not be so. With language you canalways do ‘‘as if’’: You can represent a kin relation that is expected to beinherent, ‘‘inalienable’’, “as if’’ it were non-inherent, established, more‘‘alienable’’; and both expressions are predicative.5. ConclusionLet us look back at some of our paradoxes: (1) The layman and the linguist:The layman who readily associates Possession with ‘ownership’, ‘belonging’,etc. has a point there in that these may in fact represent the essence of Possession.His view is simply incomplete as he overlooks inherent relations. —The professional linguist who claims that possessive expressions are notoriginally possessive but rather derived from locative, instrumental, directionalexpressions has a point as far as our ‘‘unspecific area’’ goes. He would have tobe contradicted when claiming that this holds for the entire span of possessiveexpressions. Possession is an original relation that spans between the polesof ‘‘inalienability’’ and ‘‘alienability’’ with an ‘‘unspecific area’’ of inversion intransition. (2) ‘‘Inalienability’’ the older, ‘‘alienability’’ the younger construction— or the opposite: The overwhelming evidence crosslinguistically is infavour of the first alternative. Yet, since our conceptual scheme exhibits amental movement in the reverse, i.e. from Essence back to EGO, someindividual languages may proceed accordingly and derive ‘‘inalienables’’ from‘‘alienables’’. A clear example has been pointed out by Heine (1997: 183) forMelanesian Pidgin English, where the English possessive verb belong is used asa marker for ‘‘inalienable’’ Possession: papa bilong papa bilong me ‘father offather of me’ = ‘my grandfather’. (3) The apparently contradictory use of casemarking, especially genitive and dative, with regard to ‘‘inalienability’’ vs.‘‘alienability’’ is accounted for by assigning these to the unspecific area ofinversion.Let us also evaluate some of our tenets. (1) Defining Possession anddelimiting ‘‘alienables’’ from ‘‘inalienables’’: Categorial morpho-syntactic andlexical evidence may lend itself to certain cross-linguistic generalizations, but

The operational basis of possession 39they would never add up to anything truly universal. This is the reason why noworkable definition of Possession has been put forward in the literature —with the exception of a first instalment by this writer (Seiler 1983a: 4ff.) Forthe same reason no clearcut delimitation of ‘‘inalienability’’ from ‘‘alienability’’has been proposed. In categorial terms this is simply not possible. Wehave suggested instead that on a conceptual level it is possible to distinguishbetween them in operational terms. (2) All possible variant expressions ofPossession: We have seen how intimately ‘‘inalienability’’, ‘‘inversion area’’,and ‘‘alienability’’ are intertertwined in their language-specific realizations.Only an integrated view of the entire spectrum of linguistic manifestations ofPossession will permit us to recognize the dynamics behind them and thusto understand the conceptual content and its processual construction. (3) Ourthree-level approach: The universal, the general, the individual. Of the three,the universal and the individual are on-line in the sense that in an individuallanguage we find the same programs operative as on the level of conceptuality.The intermediate level of General Comparative Grammar is off-line in thesense that it represents the ‘‘menu’’ making available a maximum of techniquesand grammatical categories from which each individual languageexhibits a proper choice. Only by an attempt to disentangle the three levelsthat are mixed in each individual language can we hope to arrive at a trueunderstanding of the workings of Possession.Notes1. Some remarks concerning presentation and terminology are in order here; Possessionis a relation between an entity called the Possessor and another entity called thePossessum. Relational terms and relata are marked with capital letters when a level ofconceptuality is under discussion. The relation of Possession appears in two basicvarieties: ‘‘inalienable’’, symbolized as (X) Father = ‘father of X’, and ‘‘alienable’’,symbolized as (X) R (Y) = ‘X possesses Y’. The traditional terminology is misleading:‘‘Alienability’’ in the sense of separability is at best a concomitant property of the relation.What really matters is whether the relation is inherently given in one of the entities, viz. inthe Possessum: Father is inherently and necessarily someone’s father; or whether therelation is not inherently given and therefore has to be established by special means, viz. arelator R as a ‘‘third’’ between the two relata. The correct terms would thus be inherent vs.established Possession. For ease of presentation I shall maintain the traditional terms, butI shall put them in quotation marks.2. The detailed specification of this level with its constructivistic and operational aspects isan outgrowth of my renewed inquiries into the entire problem complex of Possession.

40 Hansjakob Seiler3. While constructing this schema I have profited from discussions with Jürgen Broschartand his interactional model (Broschart 1997: 83ff.)ReferencesAnschutz, A. 1997. ‘‘How to choose a possessive noun phrase construction in four easysteps’’. Studies in Language 21: 1–35.Bally, C. 1926. “L’expression des idées de la sphère personnelle et de solidarité dans leslangues indo-européennes’’. In Festschrift Louis Gauchat, F. Frankhauser and J. Jud(eds), 68–78. Arau: Sauerländer.Baron, I. and Herslund, M. (eds). 1997. Possessive Structures in Danish [KLIMT 3].Copenhagen Business School.Benveniste, E. 1962/6. ‘‘Pour l’analyse des fonctions casuelles: le génitif latin.’’ Lingua 11:10–18 [=Problèmes de Linguistique Générale I. 140–150]. Paris: Gallimard.Broschart, J. 1997. ‘‘On turning language into vision. Towards a geometry of interaction.’’In Proceedings from the Workshop on Representation between Vision and Natural LanguageProcessing. Budapest: European Conference on Artificial Intelligence 1996.Chappell, H. and McGregor, W. (eds). 1996. The Grammar of Inalienability. A typologicalperspective on body part terms and the part-whole relation [Empirical Approaches toLanguage Typology 14]. Berlin: Mouton de Gruyter.Heine, B. 1997. Possession. Cognitive sources, forces, and grammaticalization. Cambridge:Cambridge University Press.Jakobson, R. 1936/71. ‘‘Beitrag zur allgemeinen Kasuslehre.’’ In Selected Writings, Vol. II:Word and Language, 130–147. The Hague: Mouton.Lehmann, C. 1996. Possession in Yucatec Maya. Munich: Lincom Europa.Seiler, H. 1973a. ‘‘Zum Problem der sprachlichen Possessivität.’’ Folia LinguisticaVI: 231–250.Seiler, H. 1973b. “On the semanto-syntactic configuration ‘Possessor of an Act’’’. In Issuesin Linguistics. Papers in Honor of Henry and Renée Kahane, B. Kachru et al. (eds),836–853. Urbana: University of Illinois Press.Seiler, H. 1982. ‘‘Inherent vs. established relation, proximity vs. obviation, and two types ofCahuilla kinship expression.’’ IJAL 48: 186–196.Seiler, H. 1983a. Possession as an Operational Dimension of Language [Language UniversalsSeries, vol. 2]. Tübingen: Gunter Narr.Seiler, H. 1983b. ‘‘Possessivity, subject and object.’’ Studies in Language 7: 89–118.Taylor, J. R. 1996. Possessives in English. An Exploration in Cognitive Grammar. Oxford:Clarendon Press.Weinrich, H. 1993. Textgrammatik der deutschen Sprache. Mannheim: Duden Verlag.

Chapter 2The concept of possession in Danish grammarOle Togeby1. The expression of the concept of possession in textsIt is my intention to investigate how possession is expressed in the grammaticalsystem of Danish as a whole. Single constructed examples such as(1) mandens hat‘the man’s hat’are not enough to demonstrate that possession is a grammaticized feature inDanish. In order to show the typological characteristics of Danish with respectto the expression of the concept of possession it is necessary to study both howpossession is typically and frequently expressed and what the prototypicalmeaning of the grammatical forms is.As a starting point I will show in how many ways the concept of possessionis expressed in a Danish text. The text is the beginning of Hans ChristianAndersen’s fairy tale The tinder box. I quote it in two columns, the Danishoriginal in the left column, and the English translation in the right. Thetranslation is taken from Fairy Tales. Transl. by Jean Hersholt, Heritage Press,New York 1949. In the left column I have underlined the words by whichsome sort of possession is expressed.In the following section all examples are authentic sentences or phrasesfrom a corpus consisting of 30 small texts of different types.1/ Fyrtøiet2/ Der kom en Soldat marcherende henadLandevejen: 3/Een, To! 4/Een, To! 5/ hanhavde sin tornister paa Ryggen og en Sabelved siden, 6/ for han havde været i Krigen,7/ og nu skulde han hjem. 8/ Saa mødte hanen gammel Hex paa Landeveien; 9/ hun varsaa ækel, 10/ hendes Underlæbe hang hendelige ned paa Brystet. 11/ Hun sagde: 12/‘‘god Aften, Soldat! 13/ hvor du har en pænThe tinder boxThere came a soldier marching down thehigh road — one, two! one, two! He had hisknapsack on his back and his sword at hisside as he came home from the wars. Onthe road he met a witch, an ugly old witch,a witch whose lower lip dangled right downon her chest.“Good evening, soldier,’’ she said. “What afine sword you’ve got there, and what a big

42 Ole TogebySabel og en stor Tornister, 14/ du er enrigtig Soldat! 15/ Nu skal du faae saa mangePenge, du vil eie!’’16/ ‘‘Tak skal du have, din gamle Hex!’’17/ sagde Soldaten.18/ ‘‘Kan du se det store Træ?’’ 19/sagde Hexen, og pegede paa det Træ, derstod ved Siden af dem. 20/ ‘‘Det er ganskehuult indeni! 21/ Der skal du krybe op iToppen, 22/ saa seer du et Hul, som du kanlade Dig glide igjennem og komme dybt iTræet! 23/ Jeg skal binde Dig en Strikke omLivet, for at jeg kan heise Dig op igjen, naardu raaber paa mig!’’24/ ‘‘Hvad skal jeg saa nede i træet’’ 25/spurgte Soldaten.26/ ‘‘Hente Penge!’’ 27/ sagde Hexen,28/ ‘‘du skal vide, naar du kommer ned paaBunden af Træet, saa er du i en stor Gang,29/ der er ganske lyst, 30/ for der brænderover hundrede Lamper. 31/ Saa ser du treDøre, 32/ du kan lukke dem op, 33/ Nøglensidder i. 34/ Gaar du ind i det førsteKammer, da ser du midt paa Gulvet en storKiste, 35/ oven paa den sidder en Hund; 36/han har et par Øine saa store som et ParTheekopper, 37/ men det skal du ikke brydedig om! 38/ Jeg giver dig mit blaaternedeForklæde, 39/ det kan du brede ud paagulvet; 40/ gaa saa rask hen og tag Hunden,41/ sæt ham paa mit Forklæde, 42/ lukKisten op og 43/ tag ligesaa mangeSkillinger, du vil. 44/ De ere alle sammen afKobber; 45/ men vil du hellere have Sølv,saa skal du gaa ind i det næste Værelse; 46/men der sidder en Hund, der har et parØine, saa store som et Par Møllehjul; 47/men det skal du ikke bryde dig om, 48/ sætham paa mit Forklæde og 49/ tag du afPengene! 50/ Vil du derimod have Guld, detkan du ogsaa faae, og det saa meget, du vilbære, naar du gaar ind i det tredie Kammer.51/ Men Hunden, som sidder paaPengekisten her, han har to Øine, hvert saastort som Rundetaarn. 52/ Det er en rigtigHund, kan du troe! 53/ men det skal duikke bryde Dig Noget om! 54/ sæt ham bareknapsack. Aren’t you every inch a soldier!And now you shall have money, as much asyou please.’’“That’s very kind, you old witch,’’ said thesoldier.“See that big tree.’’ The witch pointed toone near by them. ’’It’s hollow to the roots.Climb to the top of the trunk and you’llfind a hole through which you can let yourselfdown deep under the tree. I’ll tie a ropearound your middle, so that when you callme I can pull you up again.’’“What would I do deep down under thattree’’’ the soldier wanted to know.“Fetch money,’’ the witch said. “Listen.When you touch bottom you’ll find yourselfin a great hall. It is very bright there,because more than a hundred lamps areburning. By their light you will see threedoors. Each door has a key in it, so you canopen them all.“If you walk into the first room, you’ll see alarge chest in the middle of the floor. On itsits a dog, and his eyes are as big as saucers.But don’t worry about that. I’ll give you myblue checked apron to spread out on thefloor. Snatch up that dog and set him onmy apron. Then you can open the chest andtake out as many pieces of money as youplease. They are all copper.“But if silver suits you better, then go intothe next room. There sits a dog and his eyesare as big as mill wheels. But don’t you careabout that. Set the dog on my apron whileyou line your pockets with silver.“Maybe you’d rather have gold. You can,you know. You can have all the gold you cancarry if you go into the third room. Theonly hitch is that there on the money-chestsits a dog, and each of his eyes is as big asthe Round Tower of Copenhagen. That’s thesort of dog he is. But never you mind howfierce he looks. Just set him on my apronand he’ll do you no harm as you help yourselffrom the chest to all the gold you want.’’

Possession in Danish grammar 43paa mit Forklæde, 55/ saa gjør han Dig ikkeNoget, 56/ og tag du af Kisten saa megetGuld, du vil!’’57/ ‘‘Det var ikke saa galt!’’ 58/ sagdeSoldaten.59/ ‘‘Men hvad skal jeg give Dig,din gamle Hex? 60/ For noget vil du velhave med, 61/ kan jeg tænke!’’62/ ‘‘Nei,’’ 63/ sagde Hexen, 62/ ‘‘ikkeen eneste Skilling vil jeg have! 64/ du skalbare tage til mig et gammelt Fyrtøi, 65/ sommin Bedstemoder glemte, da hun sidst vardernede!’’“That suits me,’’ said the soldier. “But whatdo you get out of all this, you old witch? Isuppose that you want your share.’’“No indeed,’’ said the witch. “I don’t want apenny of it. All I ask is for you to fetch mean old tinder box that my grandmotherforgot the last time she was down there.’’The concept of possession — in a wide sense of the word — is expressed inthe following ways in the Danish text:a. Genitive (possessive pronoun): 5/ sin tornister, 10/ hendes underlæbe, 38/mit blaaternede Forklæde,b. Definite form: 5/ paa Ryggen, 19/ paa Brystet, 33/ Nøglen, 34/ Gulvet,c. Prepositional phrase: 28/ Bunden af træet, 54/ (tag Du) af Kisten saa megetGuld Du vil, 64/ tage til mig et gammelt Fyrtøi,d. Indirect object: 38/ Jeg giver dig mit blaaternede Forklæde,59/hvad skal jeggive Dig,e. So called ethical dative + preposition + definite form, primarily aboutbody parts and clothing: 10/ hang hende lige ned paa Brystet, 23/ jeg skalbinde dig en Strikke om Livet,f. Verbs: 5/ han havde sin Tornister, 13/ hvor Du har en pæn Sabel,15/faa saamange Penge, Du vil eie, 26/ Hente Penge, 38/ jeg giver dig mit blaaternedeForklæde, 40/ tag Hunden, 41/ sæt ham paa mit Forklæde.From the examples above it is obvious that possession is not expressed by onegrammatical form in Danish. On the contrary possession is expressed in manygrammatical forms, and there is no immediate connection between thedifferent forms of expression. In the following paragraphs of this paper I willdiscuss the meaning of the forms mentioned above. Do these grammaticalforms have possession as their typical meaning or their core meaning, or dothey have other general meanings of which possession is a subpart?

44 Ole Togeby2. The genitiveIn Danish the genitive does not only, and not primarily express possession; thegenitive expresses meronymy (part of a whole), location, control, subjecthood,and effect, and possession is only one subpart of one of them, viz. of control.A. Meronymy: 1. part of a period: krigens næste fase ‘the next phase of thewar’, 2. property of the thing: ædecellens overflade ‘the surface of the eatercell’, husets farve ‘the colour of the house’, ordenes betydning ‘the meaningof the words’, 4. part of the body or clothes: hendes underlæbe ‘her lowerlip’, i vores krop ‘in our body’, 5. entities in the territory: hovedstadensgader ‘the streets of the capital’, Afrikas rovdyr ‘the carnivores of Africa’.B. Location (in space, time and mental space): fremtidens lærer ‘the teacher ofthe future’, denne lovs bestemmelser ‘the provisions of this law’.C. Control: 1. the controllable things of the controller: Iraks masseødelæggelsesvåben‘Iraq’s mass destruction weapons’, 2. the property of theowner: sin bolig ‘his residence’, 3. the subordinate of the superior: Eikstedtsdragoner ‘Eikstedt’s dragoons’, ypperstepræsternes tempelvagter ‘the highpriests’ temple guards’.D. Subjecthood (subjective genitive): 1. the actor’s act or activity: deresembedsførelse ‘the discharge of their office’, komiteens møder ‘the meetingsof the committee’, Jørgen Grys forsvar for materialismen ‘Jørgen Gry’sdefence of materialism’, statens ydelse til kongen ‘the payment to the kingfrom the state’, 2. the process or state of the entity: sagens forløb ‘theprogress of the case’, verdens begyndelse ‘the beginning of the world’,verdens tilblivelse ‘the origin of the world’, stemplernes bevægelse op og ned‘the movement up and down of the stamps’, diktatorens fald ‘the fall of thedictator’, 3. the attribute of the carrier (adjectival noun): deres elendighed‘their misery’, deres ansvarlighed ‘their responsibility’, kongens umyndighedeller sygdom ‘the King’s minority or illness’, forehavendets umulighed ‘theimpossibility of the project’.E. Effect: 1. Of a resultative verbal noun: jordens frugter og dyr ‘the fruits andanimals of the earth’, min egen retorik ‘my own rhetoric’, Snorres fortælling‘Snorre’s tale’, gedens mælk ‘the milk of the goat’, FN’s beslutninger ‘theresolutions of UN’, kongens underskrift ‘the king’s signature’, 2. The relatorof the relatum (relational nouns): sine disciple ‘his disciples’, hans fader ‘hisfather’, deres undersåtter ‘their subjects’, dyrenes konge ‘the king of theanimals’, hans kone og hans to små drenge ‘his wife and his two little boys’,husets ejer ‘the owner of the house’, 3. the processing of the object (objec-

Possession in Danish grammar 45tive genitive): kongens befrielse ‘the liberation of the king’, ministerietsafsked ‘the dismissal of the ministry’, forretningernes fordeling ‘the allocationof the tasks’, verdens skabelse ‘the creation of the world’, ved købetsafslutning ‘the closing of the purchase’, regeringens førelse ‘the managementof the government’.Langacker (1991: 338–9) has described the general meaning of the genitive inthis way:… the linguistic category of possession has an abstract basis (i.e., a schematiccharacterization applicable to all class members) with respect to which ownership,part/whole, and kinship relations constitute special, prototypical cases. What allpossessives share, I believe, is that one entity (the ‘‘possessor’’) is used as areference point (R) for purposes af establishing mental contact with another, thetarget (T). The reason that ownership, part/whole, and kinship relations areprototypical for possessives is that they in particular are central to our experienceand lend themselves very well to this reference-point function. [. . .] I suggest thatall possessives involve this reference-point function. [. . .] However, certainpossessive uses show very little vestige of such notions [possession, part/whole,kinship — OT], in which case the reference-point relationship becomes salient bycomparison and may assume the status of relational profile.Following Langacker the core meaning of the genitive is not possession, butreference-point.3. Definite formFrom the translation it is seen that the definite form in Danish is equivalent tothe genitive in English:2/ Der kom en Soldat marcherende henad Landeveien (...)5/han havde sinTornister paa Ryggen og en Sabel ved Siden ‘There came a soldier marchingdown the high road (. . .) He had his knapsack on the back and his sword atthe side’ is translated into (. . .) He had his knapsack on his back and his swordat his side. When there is a metonymic relation between a known entity andthe entity to be profiled in a sentence, it is sufficient to refer to the antecedentby the definite form of the new entity. The back and the side has a meronymicrelation to a person, and consequently the definite forms Ryggen and Sidenrefer back to the antecedent Soldat, who is a person.These anaphoric references all depend on metonymic relations:A. Meronymy: han havde sin Tornister paa Ryggen og en Sabel ved Siden,

46 Ole Togebylodderne (på vægten i konsultationsværelset) ‘the weights on the scales inthe consultation room’.B. Location: der kom en soldat marcherende henad Landeveien ‘there came asoldier marching down the high road’ , (karlene oppe fra) Herregården ‘thegrooms from the manor’, (lodderne på) vægten (i konsultationsværelset)‘the lots on the scales in the consultation room’.C. Control: karlene (oppe fra Herregården) ‘the grooms from the manor’,(lodderne på vægten i) konsultationsværelset ‘the lots on the scales in theconsultation room’.D. Affectedness: mens Ymer endnu lå småslumrende, sprang sveden ud af hamover hele kroppen ‘while Ymer was stil slumbering, the sweat gushed out ofhim, all over the body’. The definite form, exactly as the genitive, has thereference-point relationship as its only general meaning.4. Prepositional phrasesThe same holds for prepositional phrases; they do not only or primarilyexpress possession, but rather meronymy, location, control and effect oraffectedness.A. Meronymy: alle andre arter i kattefamilien ‘all other species of the catfamily’, ikke så mange elever i klasserne ‘not so many pupils in the classes’.B. Location: fiskene i vandet ‘the fish in the water’, dyrene i skoven ‘theanimals in the forrest’, på bunden af træet ‘at the bottom of the tree’,bladene faldt af træerne ‘the leaves fell off the trees’.C. Control: biskop i Viborg ‘bishop of Viborg’, regent i andre lande ‘sovereignin other countries’, de nationale regnskabstjenester i medlemsstaterne ‘thenational accounts departments in the member states’, den udøvende magter hos kongen ‘the executive power is with the king’.D. Subjecthood: krav fra hans klient ‘his client’s demands’, en barsk barndomfor Hanne ‘a harsh childhood for Hanne’, fornyet håb for desertører ‘renewedhope for deserters’, det sproglige indhold af reguleringsklausulen ‘thelinguistic content of the Clause of regulation’, Årsrapport fra komiteen (tiludarbejdelse af BNI) ‘Annual report from the committee for calculation ofthe GNI’.E. Affectedness: navnet på et fængsel ‘the name of a prison’, (Årsrapport fra)komiteen til udarbejdelse af BNI ‘Annual report from the committee forcalculation of the GNI’, (Årsrapport fra komiteen til) udarbejdelse af BNI

Possession in Danish grammar 47‘Annual report from the committee for elaboration of the GNI’,Bestemmelser angående regeringens førelse ‘resolutions concerning themanagement of the government’, indehaveren af møbelforretningen ‘theproprietor of the furniture shop’.F. Accompaniment and companionship: med and uden + verbal noun: udenfolketingets samtykke ‘without the consent of the parliament’, Datter skallære at leve uden mor ‘A daughter must learn to live without a mother’,gammel idiot med skæg, hvid kåbe og en palmekvist i hånden ‘old idiot withbeard, white cloak and a palm twig in the hand’.The meaning of prepositions in general is explained in the following way(Taylor 1995:110): ‘‘Prepositions, in their spatial sense, serve to locate spatiallyone entity with reference to another. This definition can be generalized to alloccurrences of prepositions.’’ The general meaning of a preposition is, like thegeneral meaning of the genitive, to give a reference point through which someitem of information is accessible to the mind.5. Indirect objectPossession is part of the meaning of verbs such as have ‘have’, få ‘get’, mangle‘lack’, miste ‘lose’, give ‘give’, tage ‘take’. That means that possession islexicalized in Danish, not that it is grammaticized, i.e. expressed by a grammaticalmorpheme which is part of the grammatical system. But in a theory ofdeep case many different predicates take the same semantic role, viz. the roleof the possessor as their subject, object or indirect object. I shall sketch sucha theory in this section.Predicates describing states of affairs (SoA) are subdivided into four types,depending on their aspect (state or accomplishment) and their orientation(part-oriented or whole-oriented) (Widell 1996):Point of view Aspect State AccomplishmentOrientation Part-oriented være ‘be’ blive ‘become’Whole-oriented have ‘have’ få ‘get’The entities that participate in an SoA can play three different roles relative tothe predicate: the relations ‘‘be’’ or ‘‘become’’ constitute an SoA with one role:Hun var så ækel ‘She was ugly’, Der brænder 100 lamper ‘a hundred lamps areburning’. I will adopt the terminology (Togeby 1996) that this entity plays the

48 Ole Togebybe-role (B). The relations ‘‘have’’ and ‘‘get’’ demand two roles to establish anSoA, e.g. Hun havde et forklæde ‘she had an apron’, han fik alle pengene ‘he gotall the money’. One of them has the be-role, et forklæde ‘an apron’ og allepengene ‘all the money’, and the other one has what I call the have-role (H).The be and have relations are mirror-image relations, i. e. the samerelation between a whole and one of its parts seen from two different viewpoints,whole-oriented or part-oriented: træerne H havde blade B ‘the trees hadleaves’ = der var blade B på træerne H ‘there were leaves on the trees’. Both eje‘own’ and tilhøre ‘belong to’ are verbs involving the be-role and the have-role:heksen H ejer forklædet B ‘the witch owns the apron’ = forklædet B tilhører heksen H‘the apron belongs to the witch’. In many cases, but not in all, there is synonymybetween ‘‘x has y’’ and ‘‘y is at x’’, and in some languages, e.g. Russian,‘‘have’’ is normally expressed by be at.This alternation between whole-oriented and part-oriented predicates ismore frequent in Danish than normally assumed. It is directly expressed inthis example: I en del grupper H findes der kun en han B , men andre H har op tilfire B ‘in some groups there is only one male, but other groups have up to four’.In the Danish constitution the separation of the powers is stated in thefollowing way: 3. Den lovgivende magt B er hos kongen og folketinget i forening H .Den udøvende magt B er hos kongen H . Den dømmende magt B er hos domstolene H .‘The legislative power is with the king and the parliament together. Theexecutive power is with the king. The judicial power is with the law-courts.’Both the have and the be-relation can be seen as effected or ‘‘done’’ bysome other processes in relation to which a third entity plays the do-role (D),e.g. hun D rakte forklædet B til soldaten H ‘she passed the apron to the soldier’means ‘she caused the apron to be with the soldier’, and hun D gav soldaten Hforklædet B ‘she gave the soldier the apron’ means ‘she caused the soldier tohave the apron’.Predicates may be monovalent, bivalent or trivalent. Monovalent predicatestake as a subject only a be-role, bivalent predicates take a be-role and ahave-role, være hos ‘be at’, a have-role and a be-role, have ‘have’, a do-roleand a be-role skrive ‘write’, or a do-role and a have-role skrive på ‘be writing’;the be-role is the effect of the doing of the do-role, hun D skrev et brev B ‘shewrote a letter’, the have-role is affected by the doing hun D skrev på en roman H‘she was working on a novel’. Trivalent predicates take all three roles. Heksen Drakte forklædet B til soldaten H ‘the witch passed the apron to the soldier’ andHeksen D gav soldaten H sit forklæde B ‘The witch gave the soldier her apron’.The trivalent SoA is constituted by a become-relation (↓) between be-role

Possession in Danish grammar 49and have-role, or a get-relation (↑) between have-role and be-role and acausative relation (→) between do-role and the other relation, i.e. either a dobecome-relation,Heksen gav forklædet til soldaten ‘The witch gave the apron tothe soldier’, or a do-get-relation, Heksen gav soldaten forklædet ‘The witchgave the soldier the apron’.Dhas the effect B (at H)be-roledo-role → predicate ↓↑ BisatH; H has BDaffects H (with B)have-roleLinguistically the three types of roles in the predicational kernel are defined asfollows:1. the be-role (B)syntactically: the transitive object, or the intransitive subjectsemantically: the thing which is referred to as being, becoming or remaining,or the effect of the doing2. the do-role (D)syntactically: transitive subject (active)semantically: the person or thing referred to as doing or causing somethingintentionally3. the have-role (H)syntactically: the indirect object, and (in most cases) the prepositionalobject or the subject of transitive mental verbs taking a that-clause asobject, such as tænke ‘think’, drømme ‘dream’, seat... ‘seethat...’.semantically: the entity being referred to as the place of the action or theperson having something or the person affected by an action or a phenomenon.(Besides the roles of the predication there are in a clause satellites of differenttypes denoting: time, place, duration, instrument, companion, material, aim,result, condition, cause, but that is irrelevant in this connection.)The have-role expresses location, the experiencer and, it is supposed, thepossessor.The have-role in many cases expresses the location, ‘‘the place in whichthe be-role is or becomes’’: 5/ Der kom [en soldat] B marcherende henad [Landevejen]H ‘There came a soldier marching down the high road’; 19/ [det Træ] B ,

50 Ole Togebyder stod ved siden af [dem] H ‘the tree near by them’; 28 naar [Du] B kommer nedpaa [Bunden af træet] H ‘when you touch bottom of the tree’.Active transitive verbs expressing mental experiences and takingthat-clauses as their objects are analysed as having the have-role as subjectand a be-role as object: [Soldaten] H så [at der stod et stort træ] B ‘the soldier sawthat there was a big tree’. And when the have-role expresses the sensor of amental experience, there is often a mirror-image verb expressing the samerelation with the phenomenon as subject: Soldaten skulle ikke bryde sig om athunden havde store øjne ‘the soldier should not worry about the fact that thedog had big eyes = at hunden havde store øjne bekymrede ikke soldaten ‘the factthat the dog had big eyes didn’t worry the soldier’.The have-role is in connection with trivalent predicates in the activerealized either as an indirect object (a) or as a prepositional phrase (b):(2) a. Heksen D gav soldaten H forklædet B‘the witch gave the soldier the apron’b. Heksen D gav forklædet B til soldaten H‘the witch gave the apron to the soldier’This alternation is called dative-shift by some authors. The subject in thepassive is either the be-role or the have-role:(3) a. I ønskes en god Jul‘You are wished a merry X-mas’b. En god jul ønskes jer‘a merry X-mas is wished you’Langacker (1991: 13) explains the difference between (2a) and (2b) in thefollowing way: ‘‘The semantic contrast resides in the relative salience of certainfacets of this complex scene: (a) ‘‘The juxtaposition of two unmarkednominals [. . .] after the verb symbolizes a possessive relationship between thefirst nominal and the second; (b) the morpheme to specifically designates thepath [. . .], thereby rendering this aspect of the conceptualization moreprominent than it would otherwise be.’’ In his explanation the indirect objectis the grammaticization of possession in opposition to the prepositionalconstruction where the concept of a path is emphasized: Jeg gav rækværket enny farve ‘I gave the fence a new colour’; *jeg gav en ny farve til rækværket‘*I gave a new colour to the fence’; jeg bagte hende en kage ‘I baked her a cake’;jeg slog hende græsplænen ‘I mowed her the lawn’.In my opinion the difference between (a) and (b) is better explained by

Possession in Danish grammar 51the general difference in orientation: (a) is whole-oriented: the witch affectsthe soldier; (b) is part-oriented: the witch has effect on the apron, it changesplace. Durst-Andersen and Herslund (1996) explain it as a difference between(a) resultative and (b) attemptative. But this explanation is do-role-oriented.In Danish grammar (Diderichsen 1946: 188) the relation between anindirect object and a direct object is explained as a secondary nexus, as areduced subject verb relation hun gav ham bogen ‘she gave him the book’implies han fik bogen ‘he got the book’, and hun lærte ham at læse ‘she taughthim to read’. According to this theory the have-role expresses subjecthoodwith reference to this secondary nexus.The general meaning of the indirect object in Danish is, according to mytheory of semantic roles, the entity in the SoA which is either the location atwhich the be-role is (or goes), the subject of a secondary nexus between thehave-role and the be-role, or the person affected by the do-role, or by apropositional be-role. Possession is only one subpart of that general meaning.6. VerbsThe relation of possession is implied by verbs involving a have-role. But thehave-role is a notion much broader than that of possession. In the followingI will describe the semantics of verbs taking the have-role. Verbs having ahave-role express meronymy, location, control, subjecthood, affectedness andexperience.A. Meronymy, often in connection with verbs indicating inalienable properties:han har et par øjne så store som tekopper ‘he has a couple of eyes as bigas saucers’, I en del grupper findes der kun en han, men andre har op til fire ‘insome groups there is only one male, but other groups have up to four’, dehavde den guddommelige egenskab . . . ‘they had the divine characteristic . . .’B. Location: intet andet anses som indeholdt i aftalen ‘nothing else is consideredto be implied by the agreement’, tag du af kisten så meget guld du vil‘take from the chest as much gold as you want’, han huggede højre øre afham ‘he cut off his right ear’.C. Control, also called alienable properties: 1. Divalent predicates: Hvor duhar en pæn sabel og en stor tornister ‘what a fine sword you’ve got there,and what a big knapsack’, lægen der havde klinikken ‘the doctor who hadthe clinic’, geden måtte have godt foder ‘the goat should have som goodfeeding stuff’, indehaveren af møbelforretningen ejede hele ejendommen ‘the

52 Ole Togebyowner of the furniture shop owned the whole block’, ...somduvileje‘that you want to own’, 2. Trivalent predicates: Jeg giver dig mit blåternedeforklæde ‘I’ll give you my blue checked apron’, Aserne havde taget den(verden) fra dem ‘The Aesir had taken it (the world) from them’,statsejendele der skal overlades kongen til brug ‘national property that is tobe at the king’s disposal’, vinderen overtager territoriet ‘the winner takesover the territory’, genstanden overgives i køberens besiddelse ‘the article isgiven into the purchaser’s possession’, hvortil ikke kræves særlig adkomst ‘towhich no claim is demanded’.D. Subjecthood: afgive sit votum til protokollen ‘submit his vote for the record’,de skal have mere selvbestemmelse ‘they should have more self-determination’,banker kan miste fradrag ‘banks may lose deductions’, som havde tresønner, (men kun en eneste ged) ‘who had three sons, but only one goat’.E. Affectedness: Kongen skal høre til den evangelisk-Lutherske kirke ‘the kingmust belong to the evangelic Lutheranian church’, for at modtage folketshyldest ‘to receive the homage of the people’, et ministerium som har fået etmistillidsvotum ‘a ministry that has received a vote of no confidence’.F. Experience: Soldaten så at der stod et stort træ ‘the soldier saw that therewas a big tree’, Soldaten skulle ikke bryde sig om at hunden havde store øjne‘the soldier should not worry because the dog had big eyes’, at hundenhavde store øjne bekymrede ikke soldaten ‘the fact that the dog had big eyesdidn’t worry the soldier’, tak skal du have ‘you have my thanks’, Gudskænke os alle et glædeligt nytår ‘may God give all of us a happy new year’,at have så stor viden som muligt ‘to have as much knowledge as possible’,fordi de har en bagtanke ‘because they have an ulterior motive’, de havdeikke andre guder dengang ‘they had no other gods at that time’.One important use of the verb have ‘have’ has to be mentioned. The verbserves, in Danish as well as in English, as an auxiliary that forms the perfecttense of most verbs: han havde været i krigen ‘he had been at war’, har den fåetnok at æde ‘has it got enough to eat’, kommiteen har holdt møde to gange ‘thecommittee has met twice’. Is this use of the verb as an auxiliary related to themeaning described above as meronymy, location and subjecthood? Langacker(1991: 339) believes that it is the case:‘...akind of subjectification, in that the prominence characteristic of profilingpasses from an objective relationship of control, access, or potency to a referencepointrelationship based on the subjective phenomenon of the conceptualizerestablishing mental contact with the target. A precisely analogous change figures

Possession in Danish grammar 53Table 1.Genitive Definiteness Preposition have-role VerbsMeronymy hendes underlæbe‘her lower lip’på ryggen ‘on the back’ bunden af træet ‘thebottom of the tree’hunden havde øjne så storesom tekopper ‘The dog hadeyes as big as saucers’Location fremtidens lærer‘the teacher of thefuture’(karlene oppe fra)Herregården ‘thegrooms from themanor’fiskene i vandet ‘thefish in the water’de gik hjem ‘theywenthome’så er du i en stor gang ‘youare in a great hall’Controller Eikstedts dragoner‘Eikstedt’sdragoons’karlene (oppe fraHerregården) ‘thegrooms from themanor’den dømmende magter hos kongen ‘thejudicial power is withthe king’jeg giver dig mitforklæde ‘I’ll give youmy apron’lægen der havde klinikken‘the doctor who had theclinic’Subjecthood deres embedsførelse‘their discharge’krav fra hans klient‘the claims from hisclient’hun lærte ham atlæse ‘she taught himto read’afgive votum ‘submit hisvote’Effect,affectednessverdens skabelse‘the creation ofthe world’sveden sprang ud af ham‘the sweat gushed out ofhim’udarbejdelse af BNI‘calculation of GNI’de malede porten rød‘they painted the gatered’modtage hyldest ‘to receivehomage’Accompanimentuden folketingetssamtykke ‘withoutthe consent of theparliamentExperiencer hun fortalte ham det‘she told him it’han så at der sad en hund‘he saw that there sat a dog’

54 Ole Togebyin the evolution of have into a marker of perfect aspect. In that case, however, therelevant sense of have is one in which the target is not a thing but rather a processconstrued atemporally and expressed by a past-participial complement. Thehypothesized starting point is thus an expression such as He has finished (...)Specifically, the precursor of the perfect have is assumed to have profiled arelationship of relevance or potency between its trajector (specified by the subject)and the prior event described by the complement. On this interpretation, He hasfinished would indicate, roughly, that the subject stands in a relationship ofaccomplishment vis-à-vis the finishing, or that the prior occurrence of finishingremains relevant to him.’7. ConclusionTo sum up: the concept of possession is expressed in the following ways inthe Danish text: (a) genitive, (b) definite form, (c) prepositional phrase,(d) indirect object, (e) so called ethical dative + preposition + definite form,primarily about body parts and clothing, and (f) verbs.In Danish all the forms expressing possession have the general meaningpotential: meronymy, location, control, subjecthood, effect or affectedness(and some of them the meaning accompaniment and experience). This isillustrated in Table 1. These senses are generalized to the sense reference-pointfor the profiling of some other relevant information about the SoA.I conclude that possession is not a grammaticized concept in Danish, butonly part of a much more abstract concept of reference-point.ReferencesDavidsen-Nielsen, N. (ed.). 1996. Sentence Analysis, Valency, and the Concept of Adject[Copenhagen Studies in Language 19]. Copenhagen: Samfundslitteratur.Diderichsen, P. 1946. Elementær Dansk Grammatik. Copenhagen: Gyldendal.Durst-Andersen, P. and Herslund, M. 1996. “Prepositional objects in Danish’’. In Proceedingsof the Thirteenth Scandinavian Conference of Linguistics, L. Heltoft and H.Haberland (eds), 93–108. Roskilde: Department of Languages and Culture, RoskildeUniversity.Halliday, M. A. K. 1994. An Introduction to Functional Grammar. Second Edition. London:Edward Arnold.Herslund, M. and Sørensen, F. 1993. “Valence Theory. An Introduction to the DanishProject’’. In The Nordlex Project. Lexical Studies in the Scandinavian Languages[LAMBDA 18], M. Herslund and F. Sørensen (eds), 1–22. Copenhagen Business School.

Possession in Danish grammar 55Herslund, M. and Sørensen, F. 1994. “A Valence Based Theory of Grammatical relations’’.In Function and Expression in Functional Grammar [Functional Grammar Series 16],E. Engberg-Pedersen et al. (eds), 81–95. Berlin and New York: Mouton de Gruyter.Herslund, M. and Sørensen, F. 1996. “Introduction’’ and “Discussion’’. In N. Davidsen-Nielsen (ed.), 9–13, 143–157.Langacker, R. W. 1991. Concepts, Image, and Symbol. The Cognitive Basis of Grammar.Berlin and New York: Mouton de Gruyter.Taylor, J. R. 1995. Linguistic Categorization. Prototypes in Linguistic Theory. Oxford:Clarendon Press.Togeby, O. 1996. “The Locative Argument’’. In N. Davidsen-Nielsen (ed.), 127–142.Widell, P. 1996. “Aspektuelle verbalklasser og semantiske roller. Den dobbelte aspektkalkule’’.In Semantiske roller [Odense Working Papers in Language and Communication10], L. Schack Rasmussen (ed.), 135–168. Odense University.

Chapter 3Possession spaces in DanishFinn Sørensen1. IntroductionThe linguistic phenomenon called possession is a rather strange phenomenon,so strange that one might say that it does not exist. There is for example nosingle set of constructions in a particular language, which can be said to be thepossessional constructions of that language by reference to some formalsyntactic criteria. And there is no unified semantic field, which unifies exactlythe possessive notions linguists want to use in their semantic characterizationof different possessional constructions. Therefore there is no explanation inany theory I know of as to why there are eight to ten different kinds ofpossession. 1 Even the well-known and well-established distinction betweenalienable and inalienable possession is more often than not just imposed bythe data rather than predicted.The aim of this paper is to sketch a tentative theory which characterizesthe basic possessive notions needed to describe possessive phenomena inDanish and which relates possession to other semantic phenomena, and inparticular to space phenomena.2. BasicsThe central idea of the theory to be described below is that languages expresslocation via binary location relations, that a space of some sort is pointed outvia entities like individuals and that possession is a particular kind of space. Sothe theory is clearly relational and localistic. 2As all other theories my theory needs to operate with objects. Among theobjects some are abstract and others are composed of material stuff. A situation(or an event) is abstract, while a person or a portion of water is a materialobject. And objects may be either discrete like situations and persons or nondiscretelike a portion of water.

58 Finn SørensenI also need location patterns which have the form 〈L, o, s(o’)〉, where L isa location relation, o and o’ are objects, and s is a space function. A particularstructure of this form states that o is within the space s(o’) in the way specifiedby the relation L. A direct encoding of the location pattern can be illustratedby (1):(1) Peter ligger under bordet.‘Peter lies under the table’where Peter denotes the located object, the table the reference object (=o’),under the table the space, i.e. the space (or area) situated between the table andthe floor on which it is placed, and ligger gives the location (‘is_within’) andthe posture of Peter. So prepositions are functions, i.e. relations which bymeans of a reference object point out spaces situated in the neighbourhood ofthis object. Change of preposition thus amounts to change of space, cf.:(2) a. i London (‘in London’)b. over London (‘over London’)c. under London (‘under London’)d. ved siden af London (‘beside London’)and not to a change of the locating part of the relation introduced in aparticular utterance.The space creating function assigned to prepositions above can also beexpressed implicitly by the location relation in location structures, by cases orby other kinds of linguistic markers.In order to integrate possession into the location theory just sketched Ipropose that there exists a function pos which operates on objects and associatesthese objects with a possession space. As other space functions pos may beexpressed explicitly or implicitly.Before applying this theory of possession to possessive constructions inDanish I want to stress the following points. A space is an empty and extendedabstract object, which only exists as part of a location structure in which someobject is claimed to be in the space created by the space creating function. Thisis a clear relativistic position towards space phenomena, and in relation tolanguages it is based on the empirical claim that (i) space expressions likeprepositional phrases always locate something and (ii) the subject of anintransitive verb taking only one argument can never denote a space. Apossessive pattern in the theory proposed above is claimed to be a locationpattern in which the space is a possession space. So a possessor is an object

Possession spaces 59which is associated with a possession space in which some relation locates anobject, the possessee. Note that pos thus is not a relation between the possessorand the possessee as is the case in many theories, but a relation between anobject and a space which assigns the possessor role to the object. 3 My third andfinal remark is that the holding of any particular location claim made by anutterance, be it possessional or not, may be subject to circumstantial restrictionshaving to do with, for example, which type of space the speaker is talkingabout. There are, I think, rather different restrictions on what one can locatein a geographical space like ‘in Paris’, in a domain space like ‘in linguistics’, orin a possession space like ‘for Peter’ (cf. this book is for Peter). 43. Possession in Danish3.1 The constructionsIn Danish (and in many other languages) one finds three major constructions,which are used to express possession, i.e. to express location of objects inpossession spaces. The three constructions are:(3) Peters to hatte. (the genitive construction)‘Peter’s two hats’(4) Peter har to hatte. (the have-construction)‘Peter has two hats’(5) Jens gav Peter to hatte. (the give-construction) 5‘Jens gave Peter two hats’Disregarding ambiguities I would say that in (3)–(5) Peter is associated with apossessive space and that the two hats are said to be located in this space inthree different ways because the involved objects, Peter and the two hats, areintroduced by different grammatical means. And the possession function posis implicitly expressed by the relational marker -s (‘s’), the verb have (‘have’),and the verb give (‘give’). It is this analysis of (3)–(5) I want to argue for onthe following pages. But first some more facts.3.2 Two sets of factsThe three constructions illustrated in (3)–(5) can be used to express several

60 Finn Sørensenslightly different semantic relations and very often a particular instance of oneof the constructions is ambiguous, especially the genitive construction. As forthe semantic flexibility of the constructions I want to stress that they all maybe used with a locational meaning (= location within a geographical space):(6) a. Byens larm.‘The town’s noise’b. Larmen i byen.‘The noise in the town’(7) Byen har mange indbyggere.‘The town has many inhabitants’(8) a. Peter efterlod Jens på banegården.‘Peter left Jens at the station’b. Peter efterlod Jens en formue.‘Peter left Jens a fortune’If (3)–(5) and (6a, 7, 8a) are analysed along the same lines it becomes naturalto find this flexibility. This is not the case if possession and location areconceived of as completely different phenomena. As for the ambiguity of theconstructions mentioned in (3)–(5) it is most clearly manifested in thegenitive construction, cf. (9) which can have the meanings given as paraphrasesin (10):(9) Byens biler.‘The town’s cars’(10) a. Bilerne i byen. (Location)‘The cars in the town’b. Bilerne der tilhører byen. (Possession)‘The cars belonging to the town’The verb have (‘have’), which clearly accepts the possessive reading, cf. (4), isdifficult to construct with a pure locational meaning except in cases like (7).But the ambiguity is found with verbs like tabe (‘lose’) which may mean bothtermination of possession, cf. (11), and change of location, cf. (12):(11) Peter tabte sine penge i spil.‘Peter lost his money in gambling’

Possession spaces 61(12) Peter tabte sine penge på gulvet.‘Peter dropped his money on the floor’Also with verbs like give/efterlade (‘give/leave’) it is difficult to find ambiguoussentences. When they are used in the pattern NP 1 VNP 2 NP 3 they have apossessional meaning, cf. (5) and (8b). In the pattern NP 1 VNP 2 prep NP 3 give(‘give’) can only have the possessional meaning while efterlade (‘leave’) alsocan have a locational meaning, cf. (8a).In Danish then, the constructions illustrated in (3)–(5) are semanticallyflexible in the sense that they can have more meanings, and they can in somecases be ambiguous. The problem now is whether these facts can be explainedin a reasonable way by the theory proposed in Section 2.4. Applying the theory to Danish4.1 An exampleConsider the following example:(13) Finns Paris er bedre end Michaels.‘Finn’s Paris is better than Michaels’For anybody who knows Finn and Michael it is clear that (13) is about thecapital of France because they both like to go to that city. It is also clear thatboth Finns Paris (‘Finn’s Paris’) and Michaels Paris (‘Michael’s Paris’) aregenitive constructions associated with a possessional meaning in the sense thatsome part of Paris belongs to Finn and some other part belongs to Michael. Soboth Michael and Finn are possessors. But the two genitives do not just saythat Michael and Finn both are possessors of Paris in which case it would notmake sense to use the comparative predicate ‘is better than’. What happens isthat the genitive -s creates a pattern of the form 〈 _, _, F s (_) 〉, where F s is avariable over functions creating spaces. So Finns (and Michaels) defines a spaceassociated with Finn (and Michael). In this space you will find, say, Finn’sexperience with Paris and in another we have Michael’s experience with Paris.The point I want to make by (13) and the analysis just given is that anevent like an experience hardly can be said to be possessed by someone or tobelong to someone. But it can easily be said to be in a certain space, here apossessional space. That is why I take the genitives as evidence for possessionalspaces even though genitives as Finn’s Paris are a bit marginal.

62 Finn Sørensen4.2 Genitive constructionsTo be consistent I now must say that all genitives are locational in some wayand that it is the genitive marker, which imposes the locational pattern on theinterpretation of genitives. Due to lack of space I cannot comment on all thedifferent kinds of genitives. But let me illustrate why I think such an approachcan explain the flexibility of genitives mentioned in Section 3.2.Consider the genitive in (14):(14) Fabrikkens vinduer.‘The factory’s windows’(14) may have three meanings. The windows may be parts of the factory as abuilding. The windows may be placed within the space delimited by the wallsof the factory. And the windows may have been produced at the factory.With respect to the placed_within case, which can also be expressed asin (15):(15) Vinduerne inde i fabrikken.‘The windows in the factory’the interpretation seems clear. The windows are said to be placed in the spaceconstituting the interior of the factory. Notice that a factory is a material objecthaving a certain purpose, the production of something. But what counts in (14)when it has the meaning illustrated in (15) is that a factory as an object has aform, which determines an empty space, the interior (of the factory). So thislocational meaning of (14) can only be assigned to genitives having agenitivised NP which has an interior. This restriction is a direct consequence ofthe theory presented here. And the crucial notion is that of space creation.When (14) is used to describe the origin of the windows the space interpretationalso seems straightforward, cf. expressions like vinduet er fra Velux(‘the window is from Velux’) (Velux is the name of a Danish window productioncompany). So the basic meaning in this case is: some object o, here thewindow, comes from the object o¹, where from the object o¹ denotes the spacewithin which the object o was created or starts its history. Note that in thiscase the factory is regarded as a space filled with a certain kind of activity, aninterpretation, which is close to the one assigned to the genitives in (13). Thethird meaning carried by (14) is the part–whole meaning. Could this meaninghave something to do with spaces? I think so. And here is the explanation. Anyspace s has, as an extended empty entity, a set of subspaces that are all part of

Possession spaces 63the superspace s. And all material objects fill out a certain space s¹ because oftheir form and extension. The material interior of s is the set of sublocationss¹ 1 ,…,s¹ n satisfying the following conditions: (i) For each s¹ i ,s¹ i isapartofs;(ii) there exists a partition PAR of the object o so that each object in PAR (o)is in an s¹ i and each s¹ i is filled out by an object in PAR. Given these rathernatural extensions of the theory proposed in Section 2, the meaning of (14) inthe part–whole case is: the windows are located within the material interior ofthe factory. So once again we have a space created by the genitive -s, this timethe material interior of the factory. And some windows are located in thisspace. From this account it follows that the windows as objects are part of thefactory as an object. But the flexibility of the genitive is also explained becauseall the meanings of (14) and similar constructions are seen as a variation of thesame pattern, the location pattern. Note that I am not saying that the part–whole distinction can be reduced to location. I am only saying that the genitiveconstruction may involve location in the material interior of an object, andwhen it does then the part–whole relation holds between the two objects. 64.3 Other constructionsThe observations I made in connection with the constructions in (3)–(5) werethat the flexibility of have- and give-constructions is more constrained than thegenitive construction. Why is this so? Is there a beginning of an explanation inthe framework proposed here? If it exists it must have something to do withthe selection of space, as this is the only natural place to make distinctions inthe location patterns I operate with.It is natural that the genitive construction has the greatest flexibilitybecause the genitive -s just introduces a variable over space functions, whichwill be fixed differently depending on the context. This is what I argued inSection 4.2.As for the verbal constructions it is natural that each verb imposes restrictionson its arguments. In the case of space arguments they ought to besensitive to different kinds of spaces. This prediction is corroborated to acertain extent. In Baron and Herslund (this volume) it is shown that themeaning of certain uses of have involves what they call inclusion and can beillustrated by the following examples:(16) a. Huset har et sort tag.‘The house has a black roof’

64 Finn Sørensenb. Peter har et hus.‘Peter has a house’c. Fabrikken har ti arbejdere.‘The factory has ten workers’In my terms have (‘have’) locates the located object, here the referent of thesyntactic object, within the material interior of the house, (16a), within theinterior of the possession space associated with Peter (16b), and within theinterior of the factory as an organization, (16c). So, in Danish have (‘have’)means ‘is_within’ and it selects the interior of some object. This is a partialsuccess because it unifies the description of have (‘have’) up to the selection ofspaces. Why possession goes together with material and abstract interiors isclearly not explained.As for verbs like give (‘give’) the choice of a possessive denoting argumentis obligatory, while the verbs efterlade (‘leave’) and tabe (‘lose’/‘drop’) canexpress both possession and location, cf. the examples in (8), (11) and (12).This can be predicted by a more fine-grained analysis of types of space, aproject I can not report on in this paper.5. ConclusionI think the sketch of a theory presented above is promising because it allows tounify possession and location via the notion of a space and because it issufficiently fine grained to predict both the distinction between possession andlocation and the different selections made by different linguistic units. I alsothink that the theory can make, and even already predicts, the distinctionbetween alienable and inalienable possession often discussed by grammariansand linguists. Note that all spaces are introduced via some abstract or materialobjects. Given this huge class of spaces there are two natural classifications ofthese spaces. In one class we have all the possession spaces. In another we haveall the other types of space. But in both subclasses we can also make a distinctionbetween spaces, which are external to the object used to create the spaces,and spaces, which are internal to the object used to create the spaces. Thecombination of these two distinctions gives exactly what is needed. Alienablepossession is location in a possessional space exterior to the reference object,and inalienable possession is location in a possessional space interior to thereference object, cf. the well-known distinction between Peter har en bil (‘Peter

Possession spaces 65has a car’) and Peter har to øjne (‘Peter has two eyes’), respectively. For reasonsof space I cannot develop this analysis. But I think I have shown that theproposed theory of spaces is promising.Notes1. See the list given in Heine (1997a: Chap. 1.3), Heine (1997b: Chap. 5.1.3).2. The theory is localistic because a location relation introduces possession as a kind ofspace. And it is relational because spaces only exist as part of a location relation.3. In many different theoretical contexts the possessee is said to be related directly to thepossessor, cf. Heine (1997a, b); Jackendoff (1996); Herslund (1997); Barker (1995).4. Some linguists and grammarians seem to think of location only in cases like ‘in Paris’which I call a geographical space. But such a restricted view of location is not very fruitfulin connection with language. In the theory proposed here the function of expressions likein Paris and in linguistics is the same in sentences such as: In Paris I liked to walk around, inlinguistics an argument is not always accepted as such. The only difference is the kind ofspace denoted by the prepositional phrases.5. For some reason this construction is rarely mentioned in connection with possession,but it is clearly possessional in the same sense as in (3) and (4).6. The presence of both the locational parameter and the possessional parameter is alsopart of the account of possession in Herslund (1997) and Baron and Herslund (1997).ReferencesBarker, C. 1995. Possessive Descriptions. Stanford: CSLI Publications.Baron, I. and Herslund, M. 1997. “The Danish Verb have and the Notion of Possession’’. InI. Baron and M. Herslund (eds), 131–157.Baron, I. and Herslund, M. (eds). 1997. Possessive Structures in Danish [KLIMT 3]. CopenhagenBusiness School.Heine, B. 1997a. Possession. Cognitive sources, forces and grammaticalization. Cambridge:Cambridge University Press.Heine, B. 1997b. Cognitive Foundations of Grammar. New York and Oxford: OxfordUniversity Press.Herslund, M. 1997. “Partitivity and Possession’’. In I. Baron and M. Herslund (eds), 1–44.Jackendoff, R. 1996. “Semantics and Cognition’’. In Contemporary Semantic Theory, S.Lappin (ed.), 539–560. Oxford: Basil Blackwell.

Chapter 4The verb have in Nyulnyulan languagesWilliam McGregor1. IntroductionA good deal is known about the expression of possession within NPs inAustralian Aboriginal languages (virtually every grammar discusses it), and toa lesser extent about external possession (“possessor raising’’) constructions(e.g. McGregor 1985, Chappell and McGregor 1995, Part 2, and McGregor1999a). By contrast, precious little is known about predicative possession,about the means available for expressing the possessive relation as the primarypredicate of a clause. No grammar of an Australian language provides adetailed account of the topic, and the majority ignore it completely; nor is thereany literature devoted specifically to it. This paper represents an initial excursusinto the domain, investigating the verb ‘have’ in the Nyulnyulan languages, anon-Pama-Nyungan family consisting of about ten languages spoken in andnear the Dampier Land region in the far north-west of Australia. 1Six Nyulnyulan languages are definitely known to show lexical expressionof predicative possession by means of a ‘have’ verb. In each case it is a simpleinflecting verb, never a compound verb construction involving an uninflectingpreverb plus an inflecting auxiliary verb (see e.g. McGregor 1996:38, 47–48).In the Western Nyulnyulan languages Nyulnyul, Jabirrjabirr, and Nimanburruthe verb is -BAKAND, whilst in Bardi it is the probably non-cognate -LABA~LEBANBAD. 2 In the Eastern Nyulnyulan group Nyikina shows the verb-BIKA ~-BI, and Warrwa, -BA. Yawuru, also an Eastern Nyulnyulan language,almost certainly does not possess a lexical verb ‘have’. Although Nekes andWorms (1953) mention -BAKAND ‘have, possess, own’ as a Yawuru verb,they do not exemplify its use, and according to Hosokawa (1991) the onlyverbal expression of predicative possession involves the verb -NI ~-NGA ~ -JI‘be’ (as do all examples cited in Nekes and Worms 1953). The three remainingNyulnyulan languages, Jawi, Ngumbarl, and Jukun are moribund, and it is notknown whether they also had a ‘have’ verb.

68 William McGregorThe discussion is organised as follows. Section 2 outlines the main propertiesof the ‘have’ verbs in Nyulnyulan languages. Following this, Section 3discusses the nature of the relationship between the two entities, the subject(possessor) and the object (possessum). Section 4 then presents evidence thatthe ‘have’ verb is not a copula, but a “full’’ verb. Section 5concludes the paperwith a brief summary and some general remarks on ‘have’ verbs in Australianlanguages.2. Main features of ‘have’ verbs in Nyulnyulan languagesBefore discussing the syntax and semantics of the ‘have’ verbs, a few remarkson their morphological characteristics are in order. The Bardi verb -LABA isformally transitive: it belongs to the na conjugation class, the members ofwhich are predominantly transitive. This is illustrated by the past tense formi-na-laba-na (3.sg.nom-tr-have-pa) ‘he/she/it had him/her/it’. The situationfor -LEBANBAD is not known: no forms of this verb are available that permitus to determine its conjugation class. By contrast, the ‘have’ verbs in the otherlanguages are all formally intransitive, that is, they belong to the intransitiveØ conjugation class. In Nyulnyul, for instance, we have i-m-bakand-an(3.sg.nom-en-have-imp) ‘he/she/it had him/her/it’, rather than *i-na-mbakand-an(3.sg.nom-tr-en-have-imp). In Warrwa, baana (i.e. Ø-ba(a)-na3.min.nom-have-imp) ‘he/she/it had him/her/it’ occurs, not *na-wa-na or *nam-ba-na,with the na (transitive) class prefix. And Nyikina shows yi-m-biga-na(3.min.nom-en-have-pa) ‘he/she/it had him/her/it’, not *yi-na-(m)biga-na.The other Nyikina form -BI is reconstructed from inflected forms given inNekes and Worms (1953), who classify it as intransitive; it does not appear inStokes (1982).Another respect in which ‘have’ in Nyulnyulan languages appears to beintransitive, or at least of reduced transitivity, is that it lacks a reflexive/reciprocal form, even in Bardi. There would seem to be no semantic reasonwhy such a form should not exist for this verb, since at least some of its sensesadmit reflexive or reciprocal interpretations (see next section).Despite the formal intransitivity of the verb ‘have’ in all languages exceptBardi, it can host a non-zero accusative pronominal enclitic cross-referencingthe possessed object, and occur in a fully transitive clause, with subject markedby the ergative postposition, as shown by examples (1) and (2). Hence there isa mismatch between the transitivity of lexical verb itself, and the transitivity of

have in Nyulnyulan languages 69the entire inflected verb construction and/or the clause, which is transitive.(We return to this observation later.) 3(1) nga-m-bika-ny-jarr-irr manja yila Nyikinal.min.nom-en-have-pa-rel-3.aug.acc many dog‘I used to have lots of dogs’ (Stokes 1982: 322)(2) kinimirr-in i-rr-bakand-in karrj jarringk Nyulnyulshark-erg 3.nom-pl-have-pres sharp tooth‘Sharks have very sharp teeth’The verbs -BAKAND, -BIKA and -BA show, in addition to their ‘have’possessive sense, also non-possessive interpretations that might be glossed‘hold’, as in (3) and (4), and ‘keep’, as in (5). In none of these examples doesthe subject possess the object in any reasonable sense of the term. Similarexamples are attested in other languages, with the possible exception of Bardi.As far as I can tell, there are no grammatical differences associated with thesedifferent senses.(3) kard wa-na-r jan nimal, mi-bakand Nyulnyulstill 2.sg.nom.fut-tr-pierce my hand 2.sg.nom-holdyadinyfor.a.while‘So keep holding my hand, hold it for a little while’(4) nyinka-n nyinka-n, jubjub-ina Ø-baa-n, Warrwathis-loc this-loc cut-ag 3.sg.nom-have-presnyin-rnirl-ma, kujarra-layina,this-pl-erg two-alone‘Three (rivets) hold the blade onto this (the handle)’(5) imber-dj yong-en djan mai ban Nyulnyulngimbirr-ij yu-ngk-in jan may bannight-dat 3.sg.nom-fut-be my food somi-bogandanmi-bakand-an2.sg.nom-hold-pres‘Keep my meal ready for this evening, I can not eat it now’ [More accurately:‘Keep my meal so it will be (ready) tonight’] (Nekes and Worms1953: 796)The lexemes -BAKAND, -BIKA, -BI and -BA are probably cognates, not just

70 William McGregorbecause of the obvious phonological relatedness, but also because they are allunexpectedly assigned to the intransitive Ø conjugation. Not inconceivablythey are reflexes of a verb *-BIKANDA, that might be traceable back to proto-Nyulnyulan. The likelihood of such a distant source is attested to by the factthat the four modern terms come from both branches of the Nyulnyulanfamily, Eastern and Western. Assuming this to be the case, what was itsmeaning in proto-Nyulnyulan? One possibility is that it meant ‘grasp’, or‘hold’. And in fact this meaning is attested in Jukun: Bates (nd) cites a numberof inflected forms of a verb she consistently transcribes as booganda —presumably -BUKANDA — an obvious reflex of *-BIKANDA, which sheglosses ‘catch hold of’ and ‘hold’. Furthermore, such a development would beconsistent with a known pathway for development of ‘have’ verbs, viz. frommore active verbs such as ‘grasp’, ‘take’, ‘obtain’, ‘hold’, ‘get’, etc. (e.g. Heine1997:47). Against this is the fact that *-BIKANDA must have belonged to theØ conjugation, or its historical source; whichever, the verb must have beenintransitive. 4 This does not sit well with the proposed active meaning of theverb. Although at this stage it is impossible to be certain about the meaning of*-BIKANDA in proto-Nyulnyulan (assuming it does indeed go back that far),or about how it changed over time, the available language internal evidencemakes it doubtful that the ‘grasp’, ‘take’, ‘get’ sense is historically prior. Thismay well be a case in which the more active meaning displayed by the Jukunverb has derived over time from a stative ‘have’ sense.‘Have’, ‘hold’, ‘keep’, and ‘grasp’ are not, however, the only senses associatedwith the Nyulnyulan ‘have’ verbs, as we will see in the next section.Examination of the corpora (see note 1) reveals that the various senses showtwo recurrent features. First, there is always some associative relationshipbetween the subject (or Agent) and object (or Undergoer): they are invariablyrelated to one another by an associative link. And second, the clause alwaysrefers to a situation of some sort: to something going on or happening in thereferent world. We examine these two characteristics in order in the followingtwo sections.3. The associative relation between the two arguments of ‘have’ verbsImpressionistically, the associative relationship between the two argumentsvaries from very close, almost inherent ones, to quite temporary and accidentalones, with a number of intermediate possibilities. This is depicted in Figure 1,

WesternBardiNimanburruJabirrjabirrNyulnyulNyikinaWarrwahave in Nyulnyulan languages 71which also indicates the range of senses attested in each language for whichrelevant information is available. (Recall that Yawuru does not have a ‘have’lexical verb and that information is lacking on Jukun, Ngumbarl, and Jawi.)Although not all points along the scale are attested for each language, it seemsnot unreasonable to suppose that the entire range indicated by the doubleheadedarrow was covered by the ‘have’ verb in the languages indicated. Themissing points most likely represent vagaries of the recorded information.Some discussion and illustration is now given of the various points alongthis continuum; in most cases illustration could have been provided bymultiple examples from more than one of the languages.At the closest extreme are parts of the subject’s body. Rarely, these areEastern {Items belonging to the personal domainInalienable parts of the bodyInalienable parts of the body with exceptional qualitiesAttached (alienable, regenerative, or non-permanent) body partsAttached parts of inanimates (things, plants, etc.)Manifestations of the bodyItems held in/on part of the bodyItems held in one’s domain of controlItems looked after generally, typically kin, petsItems owned or over which one has rights of useItems existing within one’s socio-cultural domainSecond-order (non-physical) entities concerning which one hasknowledge, beliefs, practices, etc.Figure 1. Types of associative relation expressed by have verbs in Nyulnyulanlanguages

72 William McGregorinalienable parts, but if so, they are parts that display some unusual,exceptional, or non-essential quality that marks them as different from thenorm, as illustrated by example (6). 5 More usually, however, they are the morealienable parts: attached and usually regenerative parts, such as beards (example(7)), fruit of plants, and so forth. 6(6) amb-nem i-leban budidj lamad Bardiamb-nim i-laba-n budij lamadman-erg 3.sg.nom-have-pres big cheek‘The man has big cheeks’ (Nekes and Worms 1953: 640)(7) djō gōl-nem ilebanbad djēd Bardijoo kool-nim i-lebanbad jiid2.min.obl father-erg 3.sg.nom-have beard‘Your father has a beard’ (Nekes and Worms 1953: 649)Similar remarks hold for parts of inanimates, as illustrated by the followingWarrwa example:(8) manyja-yinu mayi Ø-baa-n larrkardii-na Warrwamany-abl nut 3.min.nom-have-pres boab.tree-ergbaalutree‘The boab tree has lots of nuts’Slightly more distant associations obtain where the object represents somethingmore temporarily or less intrinsically associated with the subject, as inthe case of sicknesses and diseases; undesirable or unnatural bodily conditionssuch as prickly heat, warts, pimples, sores, etc.; and parasites on an animal’sbody. (9) provides illustration.(9) ai a-laba-n niman almb Bardingay nga-laba-n niman almbI 1.sg.nom-have-pres many prickly.heat‘I have many blisters of prickly heat’ (Nekes and Worms 1953: 318)Next items in terms of distance are things that are typically held in the hand,and invoke the ‘hold’ sense of the verb, as in example (3) above. At a slightlygreater remove are things that are (perhaps metaphorically) ‘held’ under aperson’s care or control. For an inanimate thing, this may mean a prolongedphysical proximity, as would probably have been involved in (5), where theitem was presumably intended to be held continuously under the subject’s

have in Nyulnyulan languages 73custodianship (rather than in their hands); for animate beings, less continualvigilance and physical contiguity is required, and more socially prescribed careand control may be involved, as for pets (example (1)), and kin (as in (10)).(10) juwa-na mi-m-ba-n kujarra baawa Warrwayou-erg 2.min.nom-en-have-pres two child‘You’ve got two children’Moving further down the scale we find items one has right of use over, that is,prototypical alienable possessions (example (11)), possessed under sociallyprescribed ownership. For these entities there is no implication of caring-typeinvolvement, or that they be held in physical proximity with the subjectowner.(11) warindjer djomborere mandjen a-boganden Nyulnyul, Jabirrjabirrwarinyjirr jumbarrirri -manyjin nga-bukand-inarindjaar djomborere mindjen a-leban Bardiarinyjangarr jumbarrirri -minyjin nga-laba-none knife-only 1.sg.nom-have-pres‘I only have one knife’ (Nekes and Worms 1953: 503, 686, 715)The next point on the scale — items that exist in one’s socio-cultural domain— is attested in Bardi, as exemplified by (12). This also, incidentally, showsthat the ‘have’ verb in this language at least can be used with something like anexistential sense, as is not uncommon cross-linguistically. No otherNyulnyulan language shows either sense.(12) arra a-la-rra-laba iniini jarri Bardinot 1.pl.nom-irr-pl-have emu here‘We don’t have emus here’ (Aklif 1991)Lastly, there are just a handful of examples in which the associated item is nota concrete entity, and the associative relation is an abstract mental, behavioural,or social one. Examples are (13) and (14).(13) kinya bijin-ka jalngkangurru [nya-ni wamba] Nyikinathis Pigeon-emp doctor [this-erg man. . . malu nga-la-biga nilawal kinya wamba warany-kanot l.sg.nom-irr-have name this man other-emp‘This medicine man [the first man] (was called) Pigeon, . . . but I don’tknow [literally, have] the name of that other one’ (Stokes 1982: 147)

74 William McGregor(14) ara mile-lab ara god, ai mindjen Bardiarra mi-li-lab arang god ngay-minyjinnot 2.sg.nom-irr-have other god I-only‘Thou shalt not have strange gods before me’ (Nekes and Worms1953: 715)Summing up, the associative relationship — which is invariably present inclauses involving the verb ‘have’ — can vary from a quite close relation to avery distant one. There is no evidence that any grammatical differences areassociated with any point along the scale of Figure 1. They therefore representetic, rather than emic, distinctions. The relation of possession as such, in thestrong (perhaps prototypical) sense of ownership, applies only to a very smallportion of this scale: really only to the antepenultimate point. Nevertheless, theterm “possession’’ is no less appropriate for the Nyulnyulan semantic domainof the ‘have’ verbs than it is for the semantic domain of English have, which issimilarly non-specific. And it has the advantage that it does not suggest thatthe associative relationship is restricted to the concrete and physical.4. Clauses with ‘have’ designate situationsWe turn now to the second characteristic feature of Nyulnyulan ‘have’ verbs,that the clauses they occur in always designate some ongoing situation,normally a stative one. The ‘have’ verbs, that is to say, are “full’’ referringverbs. They are not mere copulas that serve no other function than to mark orestablish a possessive link between two NPs, the possessor and possessum, ashave arguably is in English expressions such as John has a house (Lyons1968:388ff, McGregor 1997: 152). 7 There are two aspects to this claim, oneformal the other semantic.4.1 Lack of distinctive grammatical peculiaritiesThe formal facet is that there is (as far as I can determine) no significantgrammatical difference whatever between the verb ‘have’ and any other “full’’verb in any Nyulnyulan language; nor are the clauses they occur in in any waygrammatically distinctive. The verb ‘have’ cannot be distinguished from anyother verb in terms of peculiar grammatical behaviour that might be indicativeof status as a copula. Clauses with the verb ‘have’ are structurally indistin-

have in Nyulnyulan languages 75guishable from ordinary transitive clauses, with semantically “full’’ verbs orverbal constructions expressing meanings such as ‘hit’, ‘follow’, ‘leave’, ‘hear’,‘carry’, etc. Basically this means that a non-elliptical clause involving the verb‘have’ will show an NP that can be marked by the ergative postposition, andwhich is cross-referenced by the nominative pronominal prefix to the inflectingverb, and an unmarked NP that is cross-referenced by an accusativepronominal enclitic to the verb. Illustrative examples are (2), (4), (6)–(8), and(10) above.Two qualifications are in order. First, as the observant reader may havenoticed, not all subjects of ‘have’ clauses are marked by the ergative postposition,see (9), for instance. However, there is no direct link between thepresence vs. absence of the ergative postposition and the use of ‘have’ as a fullverb vs. its use as a putative copula. Thus, the ‘keep’ and ‘have’ senses do notconveniently separate from one another on this dimension. In more than oneinstance we find ergative marking of the subject where the ‘have’ sense ismanifest, and the ‘keep’ sense improbable: see examples (2) and (7).In fact, ergative marking of the subject of transitive clauses in Nyulnyulanlanguages is never obligatory (e.g. Stokes 1982: 132–135, McGregor 1996: 29,and McGregor 1999b). Omissibility of the ergative marker in clauses involvingthe ‘have’ sense of the ‘have’ verb cannot be marshalled as evidence in supportof the proposition that these clauses employ it as a copula. The motivations forpresence or absence of the ergative marker is not well understood for anyNyulnyulan language. This topic has been investigated in only one language,namely Warrwa, and somewhat superficially due to paucity of examples(McGregor 1999b). It seems that in Warrwa the ergative postposition may beomitted from the subject of a transitive clause only if the referent is: (i) low inagentivity (that is, is not particularly potent as an agent), and (ii) it is expectedas an agent of the clause, given the discourse environment.Unfortunately, clauses with -BA ‘have’ are fairly uncommon in Warrwatexts, and it is impossible to provide compelling independent evidence thatconditions (i) and (ii) motivate the omission of the ergative marker. Nevertheless,(15) and (16) do at least illustrate the proposed contrast. Both examples,it will be noted, manifest the ‘have’ sense of the verb -BA ‘have’, although theydiffer in terms of the degree of closeness between the possessor and possessum.Furthermore, (ii) is satisfied by the subjects of both examples: (15)comes from a mythological text about two snakes, and restates informationprovided in the two previous sentences; (16), on the other hand, was offeredin the description of a drawing depicting a knife with a patch of blood on it

76 William McGregorand follows a number of sentences describing the knife, and in which it wasthe topic of discourse. However, whereas the knife quite obviously shows avery low degree of semantic agentivity, the snake does not. It has a normaldegree of agentivity, both at the level of the referent situation — subjects ofclauses expressing the possession of relatives are normally marked by theergative — and in the surrounding discourse. 8(15) wayilwayil-ni Ø-baa-na jina baawa, Warrwabrown.snake-erg 3.min.nom-have-imp 3.min.obl childbidiwarra-n jina,hole-loc 3.min.obl‘The brown snake had his children in his hole’(16) kunbulu nyinka Ø-baa-n, nyinka Ø-baa-n, Warrwablood this 3.min.nom-have-pres this 3.min.nom-have-presnyinka jubjub-ina-waalu,this cut-ag-thing‘It has blood on it, the blade’For other Nyulnyulan languages I can only remark that the handful of examplesin the available corpora in which the subject is not ergatively marked are allelicited. They show the common feature that the possessor does not activelycontrol the associative relationship, and thus that the subject is low in semanticagentivity, as in the case of Warrwa. Instead, the associative relationship seemsto be more a result of external circumstances that the possessor has little poweror influence over; it is not something that they can control, or exploit to achieveanother goal. Indeed, the possessive relationship is typically something that thepossessor suffers from, that disadvantages them, as in the case of example (9).By contrast, in example (2) we observe the more inalienable possession of teethby a shark represented by a transitive clause with subject marked by theergative. Notice, however, that teeth are at least advantageous to, and useable bythe shark, even if it has no option about having them. Whether control and/oragentivity are relevant conditioning factors for the use of the ergative postpositionin clauses with other verbs in languages other than Warrwa is not knownfor certain. However, it seems a reasonable possibility, particularly granted thatit is relevant also in nearby, though at best distantly related, Gooniyandi(McGregor 1998). In sum, there is no reason to believe that ergative marking isdifferently motivated in ‘have’ clauses than any other clauses.Second, in a small number of examples the unmarked NP representing theobject (the possessum) is not cross-referenced by a bound accusative pronom-

have in Nyulnyulan languages 77inal enclitic. The reasons for this are not understood well, and require furtherstudy. However, one relevant observation is that items most likely to bepossessed are inanimates and lower order animates, and such entities are quitefrequently either not cross-referenced by bound pronominals, or are crossreferencedby third person singular forms, irrespective of the actual number ofentities in the referent set. The third person singular accusative pronominalform being zero in all Nyulnyulan languages, it is often difficult (if notimpossible) to distinguish between the two possibilities, and it may be thatmany examples that appear not to have an accusative pronominal encliticmight actually have a zero one. However, not all of the apparent exceptionscan be accounted for in this way: see for instance examples (10) and (14)above. Accusative pronominal enclitics are occasionally omitted from the verbsof transitive clauses in Nyulnyul and Warrwa at least, so the fact that they aresometimes omitted from the verb ‘have’ can’t necessarily be taken as a peculiarityof the ‘have’ verb.4.2 Semantic characteristicsTo complete the case against Nyulnyulan ‘have’ verbs as copulas, we now turnto semantic considerations. I will attempt to demonstrate that clauses with thisverb do indeed refer to situations occurring in some referent world. Obviouslythe ‘hold’ and ‘keep’ senses are consistent with this proposal. It is in respect ofother senses that situation status requires some motivation. To begin with, ashinted above, it is not unreasonable to consider that examples such as (1) and(10), which relate to kin and pets, may invoke reference to nurturing and careinvolved in their maintenance. Notice that the type of kin who are ‘had’ in theavailable examples are always close relatives, typically children. No ‘have’clauses involve possession of kin such as cousins, mother-in-laws, and the like,relatives who one does not normally nurture. This nuance may perhaps alsobe relevant to the existential sense of (12): it is surely significant that all suchexamples refer to natural species, not inanimate objects, and natural speciesare generally considered to be propagated through the performance ofincrease rites; they are also likely to be actively engaged with as sources ofsustenance, for the benefit of people. 9 Similarly, active engagement appears tobe involved in the final category in Figure 1: either through mentation, bodily,or social activity.Next, observe that virtually all “possessions’’ represented in ‘have’ clausesare temporary, non-intrinsic, and/or non-characteristic. This is uncontentious

78 William McGregorfor all but the three highest points covered by the ‘have’ verbs on the scale ofFigure 1. Let us examine these problematic cases.Bodily manifestations are rather like held items, except that they are notactively held under the control of the person; nor (prototypically) are theyheld in the hand. However, they almost always refer to temporary conditionslasting only for a short period of time, including feelings of sickness (as in(9)); blisters, sores and the like; ticks (which of course die, drop off, etc.); andfontanelles (that soon close up). These uses are consistent with the suggestionthat the clause does indeed refer to a situation, that it does construe a certainchunk of “reality’’ not just as a pair of entities that are related together bysome abstract associative relation, but as mutually interacting in some circumstantialcontext.Parts of the body also show a pattern consistent with the proposals. As wehave seen, the body part Ns that occur in the ‘have’ construction are typicallythe less inalienable ones such as hair, beards, and teeth. They are almost alwaysattachments that are not intrinsic to the animate being’s persona, and manyare regenerative or renewable; similarly for inanimates, particularly plants, therelevant parts of which are appendages that are grown and regenerated. In noinstance is the body part referred to by a prefixing N, i.e. it is never one of themore central inalienable parts. Furthermore, in the few cases in which thebody part is higher on the scale of inalienability, the N designating it isinvariably modified in some way (as in (2)), indicating that some specialquality obtains for it. In other words, it is only when there is somethingatypical or exceptional about inalienable body parts that we find representationby a ‘have’ clause. These are precisely the conditions under which wewould expect construal as a situation.To sum up, it seems clear that the types of things that are represented aspossessions in ‘have’ clauses are consistent with the claim that the clause doesdesignate a situation in the real world; something is asserted to be happeningor going on at some point in time. Such clauses do more than just establish anassociative relationship between two entities. Mere possessive associations (tothe exclusion of on-going activities) are expressed, in Nyulnyulan languages,by verbless modes of expression. Although these cover much of the same rangeof items that are covered by ‘have’ clauses, there are two significant differences.First, verbless clauses do not extend to the final two points of the scalein Figure 1, where the distance is greatest and some sort of engagement of thesubject with the object is apparently essential. And second, verbless clauses areused in the expression of possessive relations of the closest types, not expressed

have in Nyulnyulan languages 79by the ‘have’ verb: the closest inalienable aspects of the personal domain, andthe most inalienable of the parts of the body, that is, just those circumstancesin which the associative relationship is normally a given, and not in need ofmaintenance, and where it characterises the possessor. This is illustrated by thefollowing examples:(17) djugudodo wōr-njer Jabirrjabirrjukududu wurr-nyirrreef.fish horn-comit‘The reef-fish has a horn’ (Nekes and Worms 1953: 499)(18) mōge mōg nimid ginjig bāb Jabirrjabirr, Nyulnyulmuka-muk ni-mid kinyingk baablame-lame 3.sg-leg that child‘That child has a lame leg’ (Nekes and Worms 1953: 726)But even something as characteristic and permanent as the spines of anechidna, or a scar on a person, can be represented as possessions in ‘have’clauses, as (19) and (20) show. Construal of such associations in situations ispossible where there is a focus on the acquisition of the possession: (19) comesfrom a traditional story explaining how the echidna got its spines, and (20)explains the having of the scar in terms of a prior accident.(19) kana yi-m-bika-n mangul-junu Nyikinaspike 3.sg.nom-en-have-pres spear-abl‘She’s (still) got spikes from the spears’ (Stokes 1982: 101)(20) laanbirr nga-laba-n. nail-nim Bardiscar 1.sg.nom-have-pres nail-ergi-na-m-boo-na-ngayoomoorrooloo-marr ngayoo.3.sg.nom-tr-en-hit-pa-1.min.acc little-temp 1.sg.nom‘I have got a scar. A nail poked me when I was little’ (Aklif 1991)The important point is not that there are certain possessions that cannotpossibly be represented by ‘have’ clauses, and others that cannot be representedby verbless expressions. Rather, it is a matter of representation, and asa consequence, involves considerations of probability: certain representationsare simply more likely given the way the material and social worlds arestructured, and conceptualised by speakers. And in those cases where bothverbal and verbless expression of possession are, let us say, about equally likely— for the middle ranges of Figure 1—weexpecttofindsemantic contrasts

80 William McGregorbetween expression by ‘have’ and by verbless clauses that relate in some way tothe fundamental difference between situational and non-situational representation.I submit that this is indeed the case, though limitations of spaceprevent me from arguing the case in detail. Suffice it to say that verblessmodes of expression such as (17) represent such more distantly associatedpossessions as characteristic properties of their possessors, whilst other typesserve to identify them as another individual’s possession. In both circumstancesit is clear that reference is not being made to on-going situations.Contrasting ‘have’ expression with verbless modes of expression underlinesthe central thesis of this paper, that ‘have’ is not a copula, but a fullymeaningful verb that refers to a process and occurs in a clause denoting asituation. This winds up the case for treating the verb ‘have’ in Nyulnyulanlanguages as a full transitive verb in all of its uses — despite its formalintransitivity — one that is never used as a possessive copula. The onlyqualification that needs to be added is that the Warrwa ‘have’ verb -BA shows— unlike the ‘have’ verbs of other Nyulnyulan languages — marginal use as atype of “auxiliary’’ verb. This is nothing like the perfect auxiliary have ofEnglish, French, and other Indo-European languages, and is not reminiscentof a copula usage. Rather, what have sometimes been referred to as auxiliaryverbs in Nyulnyulan languages occur (as mentioned previously) in compoundverb constructions, in collocation with uninflecting verbal “particles’’, inwhich constructions they serve as verbal classifiers, indicating the type ofprocess being referred to. In Warrwa just three such collocations are attestedfor -BA ‘have’, with miird ‘hold in hand’, nganybu ‘hold by hand’, and kuk‘be/remain quiet, silent’.5. ConclusionThe various interpretations of the ‘have’ verb in Nyulnyulan languages are,I submit, etic contextualisations of a single invariant inherent meaning. Thereis a single monosemous verb ‘have’ in each language, that admits a considerablerange of contextual senses; the verb is not polysemous. Even if it deriveshistorically from a verb meaning ‘hold’ or ‘grasp’, this is not a satisfactory glossfor the verbs in the modern languages. The verbal lexemes -BAKAND, -BIKA,-BI, -BA, -LABA, and -LEBANBAD do not mean ‘hold’ in the sense that this isnot a good characterisation of their inherent meaning. Their inherent meaningis more abstract, and better glossed as ‘have’, the ‘hold’ and ‘keep’ senses being

have in Nyulnyulan languages 81contextually engendered, superimposed from the surrounding discourse andenvironment onto the abstract relational sense. Although I have not attemptedto specify the inherent meaning of these ‘have’ verbs precisely, I have suggestedthat two components are crucial: (i) there must be some sort of “association’’between the referents of the two arguments, the subject and object (see alsoBenveniste 1960/1971:172, who argues that avoir serves to establish an extrinsicrelation between the two entities); and (ii) this association must be construableas temporally, spatially, or actively locatable within a situation. A third componentthat has not been discussed in this paper, but is clearly necessary, is (iii)the association must be asymmetric.An important corollary is that there is no need to postulate a distinct‘have’-type verbal possessive construction in Nyulnyulan languages. Themeaning of a clause with the verb ‘have’ can be accounted for compositionally,in terms of the meaning of the transitive clause construction, the meaning ofthe lexical verb ‘have’, and the meanings of the other lexical and grammaticalunits in the clause. Clauses with the verb ‘have’ serve merely as vehicles for theexpression of predicative possession; they are not predicative possessionconstructions. It follows that the associative link between the subject and objectis a semantic one specified by the lexical verb of having; it is not represented byan emic-grammatical relation between the NPs designating the two entities.Nyulnyulan ‘have’ verbs show some striking differences from their Indo-European counterparts, including English have and French avoir. They arefully transitive in terms of the clause type they slot into, and their crossreferencingmorphological potentials, as against the pseudo-transitive status ofthe corresponding clauses with ‘have’ verbs in Indo-European languages(Benveniste 1960/1971: 169). On the other hand, as verbs they show (in alllanguages other than Bardi) reduced transitivity at the lexical level, and thisirrespective of whether their ‘have’ or the ‘hold’ senses are invoked. Overall,Nyulnyulan ‘have’ verbs give the impression of being rather more dynamicthan verbs such as avoir and have, referred to by Benveniste (1960/1971: 171)as stative verbs of possession.The scattered references to predicative possession in the literature onAustralian languages tend to suggest that the typical means of expression is byverbless clauses involving an NP representing the possessor along with acomitative-marked N (or NP) representing the possessum, as in Jabirrjabirrexample (17) above. Nevertheless, examination of available materials revealsthat a fair number of Australian languages have one or more verbs that incertain environments admit a ‘have’ interpretation, to the exclusion of a

82 William McGregormaterial ‘hold’ or ‘grasp’ sense. These include, among others, the non-Pama-Nyungan languages Gooniyandi, Jaminjung, Wambaya, Ndjébbana,Mangarrayi, Murriny-Patha, and Warray, and the Pama-Nyungan languagesJaru, Karajarri, Mangala, Arrernte, and Baakandji.Various questions arise. To begin with, are the ‘have’ verbs usable ascopulas in any of these languages, or are they always “full’’ verbs designatingprocesses? Second, is the verb “basically’’ a ‘have’ verb, or one of holding oracquisition? In most languages the verb also admits the stative ‘hold’ sense,and sometimes, as in Murriny-Patha, and possibly Jukun (see above), an evenmore active ‘take’ or ‘grasp’ sense. Careful investigations will doubtless revealmarked differences among the languages in terms of the uses they put theseverbs to, as well as their inherent meanings. Thus, I suspect that in somelanguages the ‘hold’ or ‘take’ senses may well be inherent, and the ‘have’contextually engendered. This is almost certainly the case for the Gooniyandiverb goorij-, which — in stark contrast to the Nyulnyulan ‘have’ verbs — isonly rarely used in non-material senses, usually only in respect of kin.It is hoped that this paper will stimulate Australianists to investigate thesefascinating verbs of possession, along with other modes of expressing predicativepossession, and address the questions of the semantic contrasts amongthem, and their appropriate grammatical analyses.AcknowledgementsThis is a revised version of a paper presented to the International Workshop on ‘Possession’,Copenhagen Business School, 28–30 May 1998. I am grateful to the audience for anumber of useful comments, and to Kari Fraurud and Anne-Marie Spanoghe for insightfulcomments on earlier drafts. The usual disclaimers apply. My fieldwork on Nyulnyulanlanguages was supported by grants from the Australian Research Council (Grant A58930745and A59332055), the Australian Institute of Aboriginal and Torres Strait Islander Studies,the National Aboriginal Languages Program, and the Max-Planck-Institut für Psycholinguistik,Nijmegen. My greatest debts are, of course, to my language instructors, MaudieLennard (Warrwa), †Carmel Charles (Nyulnyul) and †Freddy Marker (Warrwa).Notes1. The main sources of data are: my own field notes on Nyulnyul and Warrwa; Stokes 1982for Nyikina; Hosokawa 1991 and 1995for Yawuru; Metcalfe 1975and Aklif 1991 for Bardi;

have in Nyulnyulan languages 83and Nekes and Worms 1953 for Bardi, Nyulnyul, Jabirrjabirr, Nimanburru, Yawuru, andNyikina.2. This verb form is cited in Nekes and Worms 1953, but is not attested elsewhere.3. The following abbreviations are used: abl – ablative; acc – accusative; ag – agentivesuffix (like English -er); aug – augmented number; comit – comitative; dat – dative; emp– emphatic; en – epenthetic nasal; erg – ergative; fut – future; imp – imperfective; irr –irrealis; loc – locative; min – minimal number; nom – nominative; obl – oblique; pa – pasttense; pl – plural; pres – present tense; rel – relative verbal suffix; sg – singular; temp –temporal; tr – marker of transitive conjugation (na in most languages). The first threeintegers indicate the three person categories, first, second and third. Root forms ofinflecting verbs are cited in all capitals; otherwise all Nyulnyulan words cited are given initalics. Reference is provided to the sources of all examples other than those from my owncorpora. Free translations of examples enclosed in double quotes represent the wording ofthe source. Words are cited in the practical orthography I have elsewhere adopted — e.g.McGregor (1996) and McGregor (1999a). Examples from Nekes and Worms (1953) arepresented first in the orthographic style of the source, in IPA font, and then in the practicalorthography. Morpheme divisions are my own.4. There are at least three possible sources of the conjugation marker in modernNyulnyulan languages, all of which point to the na conjugation class going back totransitive verbs, the Ø class to intransitive verbs.5. Actually, I have been able to find no such examples in which the part is a “true’’ emicinalienable, that is to say, is represented by a prefixing nominal (see McGregor 1995).However, cheeks are certainly etically-semantically high on the scale of inalienability.6. It will be observed that this also applies to sharks’ teeth, as would certainly have beenknown to the coastal Nyulnyul people, see (2) above. Moreover, this example also showsmodification of the part by a nominal indicating a non-inherent, impermanent, quality.7. Although he does not use the term “copula’’ in reference to ‘have’ verbs in Indo-European languages, Benveniste (1960/1971: 169) makes essentially this point when heobserves that “[t]here can be no transitive relation between the subject and object of avoirsuch that the notion might be assumed to pass over to the object and modify it. A “to have’’verb does not state any process.’’8. (16) might also be compared with (4) above, which forms a near minimal pair with it,showing the same conceptual distance between possessor and possessum, though exhibitingthe ‘hold’ sense of -BA ‘have’. In (4), significantly, (ii) does not hold — the rivets had notbeen mentioned in the previous discussion, and furthermore are not prominent in thedrawing (the same one that (16) describes), and are not expected as agents.9. For this reason it also seems unlikely that a ‘have’ clause designating the possession ofkin other than those that are nurtured would be possible with the ‘Items existing withinone’s domain’ interpretation. Thus, for example, ‘We have mother-in-laws in this country’should be unlikely semantically for the same reason that ‘I have a mother-in-law’ would be— in Australian cultures generally one is expected to maintain physical and social distancefrom a mother-in-law, and not engage in close interaction.

84 William McGregorReferencesAklif, G. 1991. Bardi material. Electronic Archive Files. Canberra: Australian Institute ofAboriginal and Torres Strait Islander Studies.Bates, D.M. nd. Native vocabularies — Sunday Island. Typescript, Australian NationalLibrary. Section 12, 2E: 6. ANL-MS365–53/41–81.Benveniste, E. 1960/1971. “The linguistic functions of ‘to be’ and ‘to have’ ’’. In Problems ingeneral linguistics, E. Benveniste (ed.), 163–179. [Translated by Mary Meek.] CoralGables: University of Miami Press.Chappell, H. and McGregor, W. B. (eds). 1995. The grammar of inalienability: a typologicalperspective on body part terms and the part–whole relation. Berlin: Mouton de Gruyter.Heine, B. 1997. Possession: cognitive sources, forces, and grammaticalization. Cambridge:Cambridge University Press.Hosokawa, K. 1991. The Yawuru language of West Kimberley: a meaning-based description.PhD thesis, Australian National University.Hosokawa, K. 1995. “‘My face am burning!’: quasi-passive, body parts, and related issues inYawuru grammar and cultural concepts’’. In H. Chappell and W. B. McGregor (eds),1996: 155–192.Lyons, J. 1968. Introduction to theoretical linguistics. Cambridge: Cambridge UniversityPress.McGregor, W. B. 1985. “Body parts in Kuniyanti clause grammar’’. Australian Journal ofLinguistics 5: 209–232.McGregor, W. B. 1995. “Nominal prefixing in Nyulnyul’’. In H. Chappell and W. B.McGregor (eds), 1996: 251–292.McGregor, W. B. 1996. Nyulnyul. Munich: Lincom Europa.McGregor, W. B. 1997. Semiotic grammar. Oxford: Clarendon Press.McGregor, W. B. 1998. ‘‘‘Optional’ ergative marking in Gooniyandi revisited: implicationsto the theory of marking’’. Leuvense Bijdragen 87: 491–534.McGregor, W. B. 1999a. “External possession constructions in Nyulnyulan languages’’.In External possession, D. Payne and I. Barshi (eds), 429–448. Amsterdam: JohnBenjamins.McGregor, W. B. 1999b. Focal and non-focal ergative marking in Warrwa. Seminar givento Department of Linguistics, University of Groningen, November 1999, and typescript,Max-Planck-Institut für Psycholinguistik, Nijmegen.Metcalfe, C. D. 1975. Bardi verb morphology. Canberra: Pacific Linguistics.Nekes, H. and Worms, E. A. 1953. Australian languages. Fribourg: Anthropos Institut.Stokes, B. 1982. A description of Nyigina: a language of the West Kimberley, WesternAustralia. PhD thesis, Australian National University.

Chapter 5Semantics of the verb haveIrène Baron and Michael Herslund1. IntroductionThe verb have establishes a possessive relation between two entities (Seiler1983:62ff.). 1 Following the description of Benveniste (1966:199) we considerthe verb have, asillustratedbyFrenchavoir, as an inverted version of theexpression être à. Sentences of the type X avoir Y (e.g. Charles a une maison‘Charles has a house’) correspond to YêtreàX(e.g. La maison est à Charles‘The house is (belongs) to Charles’). The difference has to do with the fact thatwith avoir the expression of the possessor has passed from the status of objectof a locative preposition to that of subject of the sentence: it denotes the entityabout which the sentence is asserted. In other words we consider the notion of‘possession’ to have its origin in the primitive notion of ‘existence’ (La maisonest/existe ‘The house is/exists’) which first developed into ‘location’ (La maisonest (située) à Biarritz ‘The house is (situated) in Biarritz’), then into ‘belonging’(La maison est à Charles ‘The house is (belongs) to Charles’) and finally into‘ownership’ (Charles a une maison ‘Charles has a house’). 2 In this sense onecan say that a verb have has emerged in order to allow the abstract conceptualisationand expression of a local relation. Moreover, have like be fromwhich it is in some sort derived, does not denote a process like most transitiveverbs but a state, that of the subject. And the prototypical state is precisely alocation (e.g. The house is in Brighton/in the forest/on the hill . . .). On the localdimension of the verb have, see further Lyons (1967, 1968); Clark (1978);Freeze (1992); Sørensen (1997a); Baron and Herslund (1997a,b).We will argue that the fundamental meaning of have is a local relation.The exact nature of this relation will then depend upon the semantic linkbetween subject and object. As we shall see, this link always has the nature ofa denotative inclusion, i.e. a relation where the denotation of the object nounis included in the denotation of the subject noun. The inclusion can be ofthree kinds:

86 Irène Baron and Michael Herslunda. The object noun is part of a part–whole relation (e.g. roof–house). This isthe case when the object is a relational noun, i.e. a noun which needs acomplementary term in order to be properly identified (Benveniste1974:151; Seiler 1983:13; Herslund 1996: 39), as in (1):(1) The house has a thatched roof.b. The object noun denotes part of the subject noun’s possessions (e.g. house— Charles), which is typically the case when the subject is animate and theobject is a non-relational noun: 3(2) Charles has a house.c. The object noun constitutes one of the semantic features of the subjectnoun so that the two form a chain of isotopic inclusion (Greimas1966:69ff.). This semantic feature can be either a seme (e.g. worker–factory, a factory being distinct from e.g. a bank or a restaurant in that itemploys workers), or a sub-function as in bed–flat: the main function ofa flat is living there and one of the sub-functions is sleeping, which is alsothe function of a bed (cf. the notion of Telic Role as a component of anoun’s Qualia-structure in Pustejovsky 1995:85ff.):(3) a. The factory has two hundred workers.b. Maureen’s flat has three beds.We thus have two levels:1. A fundamental ‘local’ level where the object is merely located with respectto the subject.2. A superimposed inclusional level where the denotation of the object nounis semantically included in the denotation of the subject noun.The lexeme have, by locating the object with respect to the subject, actualisesthe denotative inclusion, and it is thus the combination of these two levelswhich allows us to speak of possession and paraphrase e.g. example (1) abovein the following way:(1¹) ‘With respect to the house the thatched roof is located in a part–wholerelation’Now have-clauses allow adverbial expansions which also have a locativecontent, e.g. Charles has a house in Brighton. In such cases the object which islocated with respect to the subject by way of have, is further located with

Semantics of the verb HAVE 87respect to a new place by way of a preposition, the new place thereby beingalso in some way located with respect to the subject. So Brighton in the aboveexample is placed in a relevance relation with respect to Charles: it is the placewhere his house is located. In what follows we shall use the term sub-place todenote such adverbial expansions, which constitute cases of what Heine(1997: 156ff., 190ff.) calls “specification’’ of the possessive relation.In the same way as a relation of denotative inclusion specifies the localrelation between subject and object established by have, it will define therelation established between subject, object and sub-place. This inclusion canbe seen as the filling of a slot in the denotation of the subject. In sentenceswithout a sub-place, it is the denotation of the object which fills the slot. Insentences with a sub-place specification, it is either the denotation of theobject noun or that of the sub-place which fills the slot opened in the denotationof the subject noun. These different possibilities determine the differentcombinations we are going to discuss in the present article. What is importantto underline from the outset is that the denotative inclusion is organised interms of a hierarchy where it is always either the object or the sub-place whichis denotatively included in the subject, never the other way around. Thishierarchy already appeared in constructions without an adverbial specification:if we have The house has a thatched roof, where ‘roof’ is a part of the whole,‘house’, cf. (1), we cannot have *A thatched roof has a house. 4 A similarorganisation is found, as we are going to see, when have-sentences containadverbial specifications: the subject always includes either the object or thesub-place, the object never includes anything. Our aim is to show that theinclusional hierarchy is superimposed upon the basic local relations to whichit is intimately linked. Since the constituents concerned are grammaticalentities, our study will at the same time deepen our general understanding ofthe grammatical relations, subject and object.We shall illustrate our study with examples from Danish, because Danish,for structural reasons, seems to illustrate the results of our analysis mostreadily, in spite of the nevertheless universal character of the findings.We shall distinguish two kinds of situations depending on whether thesubject simply includes one of the other two constituents or on whether thereexists a more complex relationship between the three. In the first case wespeak of simple inclusion (Section 2), in the second of complex inclusion(Section 3). We shall finally study to what extent the transposition of ahave-sentence to a noun phrase with a genitive confirms the existence ofinclusion relations (Section 4). 5

88 Irène Baron and Michael Herslund2. Simple InclusionIt is important to specify that there is only one available slot in the inclusionalspace defined by the subject. This can be seen as a parallel to the fact that atthe local level the subject of have needs an “argument’’, and only one, theobject, in order to be conceived as a place: an entity does not become a placeuntil something is located with respect to it, this something constituting anargument of the place function (Guillet and Leclère 1992; Baron and Herslund1997a:135). In the same way the inclusional space needs an element it caninclude in order to establish a semantic link which will specify the exact natureof the fundamental local relation. And just as the subject cannot play its roleas ‘place’ without its argument, the unique slot in the inclusional space cannotremain vacuous: it must by necessity be filled either by the object or by thesub-place.2.1 The Subject includes the ObjectIn the following examples:(4) a. Carl har et hus i Brighton.Carl has a house in Brighton‘Carl has a house in Brighton’b. Banken har en filial i provinsen.Bank-def has a branch in province-def‘The bank has a branch in the province’hus ‘house’ in (4a) is part of the possessions of Carl, and filial ‘branch’ in (4b)is a relational noun which, in order to be properly identified, needs a complementaryterm, bank, which contitutes the whole. The object nouns hus andfilial are thus both included in the denotation of the subject as being owned bya person or being part of a whole, respectively. This case can be illustrated inthe following way:(5)Subject Object Sub-placeThe sub-place is not included in the subject and the relation between subjectand object is the same as in sentences without a sub-place.

Semantics of the verb HAVE 892.2 The Subject Includes the Sub-PlaceIf the subject includes the sub-place it cannot at the same time include theobject in its denotation because, as stated above, there is only one slot availablein the space it defines. The inclusion relation between subject and object is soto speak blocked. And as the object is not included there is ipso facto nosemantic relation between the two which allows a more precise definition of thelocation denoted by have. The only relation expressed between subject andobject remains the local one, i.e. a simple location without any inclusional link.The denotative inclusion between subject and sub-place can be a part–whole relation (e.g. hånd ‘hand’ — Marie) as in (6a), or it can be an ownershiprelation (e.g. bil ‘car’ — Anne) as in (6b):(6) a. Marie har en tepotte i hånden.Marie has a teapot in hand-def‘Marie has a teapot in her hand’b. Anne har en puddelhund i bilen.Anne has a poodle in car-def‘Anne has a poodle in her car’The object noun tepotte ‘teapot’ or puddelhund ‘poodle’, whereas still being theargument of the subject-as-place to which it is linked by way of have, isnothing more than just an element which is further located by way of apreposition with respect to the entity included in the subject, i.e. the subplace.The blocking of the inclusion relation between subject and objectbecomes evident in cases where the object noun is a term which can be eitherrelational or non-relational with respect to the subject, such as barn ‘child’. Inthis case only the non-relational reading is possible because the subject’scapacity of acting as complementary term in a part–whole relation is saturatedby the sub-place. So barn ‘child’ in example (7):(7) Julie har et barn på skødet.Julie has a child on lap-def‘Julie has a child on her lap’is in a non-specified relation with Julie: it can be anybody’s child, whereas inexamples such as Julie har et barn ‘Julie has a child’ or Julie har et barn iFredericia ‘Julie has a child in Fredericia’ it is by necessity Julie’s child, thenoun acquiring a relational reading.As it appears, the sub-place wins in cases of rivalry with the object regardingthe available slot in the subject’s denotation. This is because at the level of

90 Irène Baron and Michael Herslundlocal relations the sub-place, as its name indicates, is a place located withrespect to another place, the subject. The two have the same argument, theobject, in common. The location of the sub-place is reflected at the inclusionallevel where it is the sub-place’s denotation which occupies the empty slot inthe space defined by the subject. The object has to hold back: it is first of all anargument shared by the two places and it is only designated to fill the availableslot when there is no competing sub-place. The inclusion of the sub-place canbe depicted like this:(8)SubjectSub-placeObject3. Complex InclusionAs illustrated, the only available slot in the inclusional space defined by thesubject has obligatorily to be filled by another constituent. We can here againdistinguish two types of situations: two-step inclusion and twofold inclusion.3.1 Two-Step InclusionThis is the case where the object noun is either a relational term (e.g. wart)included in a sub-place which is also a relational term (e.g. nose), cf. (9a), orthe denotation of the object noun constitutes a semantic feature of the subplace(e.g. bed–flat), which denotes a belonging of the subject, cf. (9b). Thereis consequently an inclusion on two levels: it is no longer the object alone, butthe entity ‘object-in-a-sub-place’ which enters into a part–whole or anownership relation with the subject. It is because the object is included in thedenotation of the sub-place that the object and the sub-place together as asingle entity can fill the available slot in the inclusional space defined by thesubject. The examples of (9) then have the paraphrases of (9’):(9) a. Paul har en vorte på næsen.Paul has a wart on nose-def‘Paul has a wart on his nose’b. Dorte har tre senge i lejligheden.Dorte has three beds in flat-def‘Dorte has three beds in her flat’

Semantics of the verb HAVE 91(9¹) a. ‘With respect to Paul a wart on the nose is located in a part–wholerelation’b. ‘With respect to Dorte three beds in the flat are located in an ownershiprelation’The situation can be depicted as follows:(10)Subject Object Sub-place3.2 Twofold InclusionA new subdivision is necessary here in order to show the two variants of thistype of inclusion.There is first the case where the object is included both in the subject andin the sub-place. This is the case in (11a) and (11b):(11) a. Marie har en søn i skolen.Marie has a son in school-def‘Marie has a son in the school’b. Anne har en angorakat på katteudstillingen.Anne has a Persian cat at cat show-def‘Anne has a Persian cat at the cat show’On the one hand the relational term søn ‘son’ is in a reciprocal relation withMarie, the mother, and angorakat ‘Persian cat’, being a domestic animal,establishes an ownership relation with Anne. 6 On the other, søn and angorakatboth constitute semantic features of the respective sub-places which do notrepresent parts of the subjects: søn is a constitutive part, a seme, of the semanticdescription of skole ‘school’, which denotes an institution attended bychildren, i.e. the sons and daughters of people; and angorakat likewise constitutesa subset of the defining features of katteudstilling ‘cat show’, whichdenotes a place where different kinds and breeds of cats are exhibited, cf.paragraph c. of 1. above. But neither skole nor katteudstilling are in any wayincluded in the subject. We can describe this variant of the twofold inclusionby the following picture:

92 Irène Baron and Michael Herslund(12)SubjectObjectSub-placeAs the verb have is a state verb, as mentioned in the introduction, it cannotordinarily signify that someone performs an action and consequently it cannotcombine with an object in the sense of ‘create’. The sentence Charles has ahouse thus cannot express that Charles has built or designed the house, onlythat he possesses it, cf. Pustejovsky’s idea that a verb ‘coerces’ its object intothe semantic type it is expected to combine with (1995:106ff.). But thislimitation does not apply to cases of twofold inclusion, as can be seen in (13a)and (13b):(13) a. Peter har to billeder på maleriudstillingen.Peter has two pictures at exhibition-def‘Peter has two pictures at the exhibition’b. Janne har en artikel om possession i Language.Janne has an article on possession in Language‘Janne has an article on possession in Language’Here we have creator relations between subject and object: the pictures arepainted by Peter and the article is written by Janne. The adverbial expansionconsequently influences the relation between subject and object because of theisotopic inclusion of the object in the sub-place: an art exhibition is a placewhere painters exhibit the works they have created and Language is a journalwhich contains articles by different authors. The inherent features of billeder‘pictures’ and artikel ‘article’, a painted and a written entity, respectively andhence created by an agent, are so to speak actualised by the semantic featuresof the sub-place. In order to verify this conclusion it is sufficient to replace thesub-place by an entity which does not form an isotopic chain with the object.Then the relation between subject and object changes and the creator relationis again excluded:(14) Peter har to billeder over kaminen.Peter has two pictures over fire-place-def‘Peter has two pictures above the fire-place’The second variety of the twofold inclusion situation is the case when object

Semantics of the verb HAVE 93and sub-place are both relational terms denoting inalienable parts with respectto the subject such as body parts:(15) Lise har en hånd på knæet.Lise has a hand on knee-def‘Lise has one hand on her knee’Such terms are equally included in the subject and consequently share theavailable slot. We get the following illustration:(16)Subject Object Sub-placeWhereas in (12) we had a single entity shared by two other entities, we herehave the inverse situation: two entities share the only available slot included ina larger entity.We have now seen the different kinds of inclusion which occur betweenthe three constituents joined together by have. Irrespective of whether theinclusion relation is simple or complex, its existence is nevertheless necessaryin order to speak about possession. If no inclusion between the subject andone of the other constituents can be identified, a sentence with have becomesimpossible to interpret, viz.:(17) *The lamp has a stone in the hand.*The garden has two pictures at the art exhibition.The existence of an inclusion relation can further be verified by transposingthe have-sentence into a noun phrase. This then constitutes an inclusion test,as we shall see now.4. Noun phrases as inclusion testsWhereas the relations containing a sub-place mentioned so far can be assertedin have-sentences, they cannot all be expressed in the reduced manner of agenitival noun phrase. Thus the nominal construction always presupposes aninclusion between subject and object. It can be an ownership relation, a part–

94 Irène Baron and Michael Herslundwhole relation, a creator relation, etc. (Baron 1995:46f.; Müller 1997; Sørensen1997b:121f.), but whatever its nature, the existence of such a relation isindispensable in order to transpose a have-sentence into a noun phrase. Thenominal expressions in (18) are unproblematic because the correspondingverbal constructions all contain an inclusion, simple or complex, betweensubject and object:(18) a. Simple ownership relation:Carl har et hus i Brighton.Carl has a house in Brighton‘Carl has a house in Brighton’Carls hus i BrightonCarl-gen house in Brighton‘Carl’s house in Brighton’b. Complex part–whole relation:Paul har en vorte på næsen.Paul has a wart on nose-def‘Paul has a wart on his nose’Pauls vorte på næsenPaul-gen wart on nose-def‘Paul’s wart on his nose’c. Complex creator relation:Anne har en artikel om possession i Language.Anne has an article on possession in Language‘Anne has an article on possession in Language’Annes artikel om possession i LanguageAnne-gen article on possession in Language‘Anne’s article on possession in Language’On the other hand the noun phrases of (19) are unacceptable because they aretranspositions of sentences in which the inclusion relation between subject andobject is blocked by the inclusion existing between subject and sub-place (cf.Section 2.2):(19) a. Marie har en tepotte i hånden.Marie has a teapot in hand-def‘Marie has a teapot in her hand’*Maries tepotte i håndenMarie-gen teapot in hand-def‘Marie’s teapot in her hand’

Semantics of the verb HAVE 95b. Anne har en puddelhund i bilen.Anne has a poodle in car-def‘Anne has a poodle in her car’*Annes puddelhund i bilenAnne-gen poodle in car-def‘Anne’s poodle in her car’c. Julie har et barn på skødet.Julie has a child on lap-def‘Julie has a child on her lap’*Julies barn på skødetJulie-gen child on lap-def‘Julie’s child on her lap’The noun phrases can thus be seen as tests revealing the exact nature of therelations between the constituents of a have-sentence. While *Julies barn påskødet ‘Julie’s child on the (her) lap’ is unacceptable when the lap is Julie’s, thenoun phrase becomes acceptable when the lap belongs to someone else, viz.(Anne sad med) Julies barn på skødet ‘(Anne was sitting with) Julie’s child onthe (her) lap’. The same way of reasoning applies of course to the a. and b.examples of (19).5. ConclusionThe subject of a have-sentence is the place with respect to which the object,i.e. the place argument, and a possible sub-place are located. A similar organisationis characteristic of the inclusion relations between the three constituents:the possessor always includes the possessum or the sub-place (or both atthe same time). From a grammatical point of view the fact that the subjectalways includes something and that the object never includes anythingmatches this other fact that one has on the one hand a constituent of whichthe rest of the sentence is predicated: the subject, which is thus the elementnecessary to form a proposition (subject + predicate = proposition). On theother hand, there is a constituent, the object, so closely linked to the verb thatit forms with it a syntactic predicate (Herslund and Sørensen 1994:83ff.).Since the object is the “fundamental argument’’ of the verb, it cannot includedenotatively an “external argument’’. The predicate forming constituent is inthis sense truly an “internal argument’’ (Grimshaw 1990).

96 Irène Baron and Michael HerslundThe hierarchy we mentioned in the introduction when speaking of theinclusion relations is found, then, at every one of the following levels:(20) Local: Place > Sub-place > ArgumentInclusional: Possessor > Sub-place > PossessumGrammatical: Subject > Sub-place > ObjectThe three constituents show an identical ranking at the three levels. At thelocal level, the argument and a possible sub-place are situated with respect tothe place. At the inclusional level, the possessum or the sub-place, or both atthe same time, cf. (10) and (16), are included in the possessor. These semanticfacts are finally mirrored at the grammatical level, where the object firstconstitutes a predicate with the verb have; this predicate is then expanded bythe adverbial sub-place, and finally combined with the subject thus forming aproposition. It is the local hierarchy which is fundamental in the sense that itexplains the presence of a single available slot in the inclusional space definedby the subject (cf. 2.), and it is also this level which justifies that in case ofcompetition for this slot between the sub-place and the object it is always thesub-place which wins (cf. 2.2). The opposite is never the case: the inclusionalhierarchy can never account for the local hierarchy. And the fact that locationis a constructed notion — an entity does not become a place until somethingis located with respect to it — makes possession — i.e. location + denotativeinclusion — a constructed notion too: nothing is a possessor until somethingacts as argument with respect to it, becomes a possessum. In other words,Charles of (2) Charles has a house, does not become a possessor until a houseis located in an ownership relation to him.To recapitulate, we can say that the local hierarchy is fundamental withrespect to the inclusional hierarchy, which is reflected in the grammaticalorganisation of the sentence: the constituent which is necessary for theformation of a proposition, the subject, is always the one which includesanother element, whereas the predicate constituting element, the object, isalways included. In this way, possessive sentences with have contributetowards underlining and explaining the asymmetry between subject and objectwhich is generally found in all kinds of sentences.Notes1. These two entities have to be distinct. We consider expressions such as John has aproposal as intransitive verb constructions where has a proposal is a support verb structure

Semantics of the verb HAVE 97and which, in spite of the common ground of the verb have, should not be treated aspossession. Neither should e.g. the auxiliary and causative uses of have.2. On this evolution, see among many others Lyons (1967: 392), Herslund (1988: 282 ff.),Baron (1997). Germanic languages such as English and Danish differ from French in thatthese languages use the genitive to express the notion of ‘belonging’: The house is Charles’.Benveniste speaks in such cases of ‘appartenance’ (1966: 196) and Seiler (1983) uses thelegal term ‘appurtenance’.3. Cf. Seiler’s idea (1983: 13) that have is typically used in order to establish a possessiverelation, when the possessum is not inherently relational.4. Such a contrast is only possible in a small number of cases where the subject is acollective noun and the object is a member of the set denoted by the subject noun, e.g. Theclass has a good teacher. By interchanging the subject and the object, viz. The teacher has agood class, one makes the assertion about the semantically included entity that this entityhas as part of its universe or simply possesses the collective entity of which it is a member.One thereby goes from a part–whole relation at the lexical level to a “part of possessions’’inclusion of the ‘Charles has a house’ type at the level of assertion.5. In the examples we discuss the object is indefinite and the sub-place is definite. Thiscorresponds with the most frequent usage in non-metaphorical examples which we havechosen to concentrate on here. Contrary to English, Danish uses the definite article whenthe sub-place is seen as included in the subject, viz. Hun har et barn på skødet vs. She hasa child on her lap. The definite article is in such cases motivated by the associativeanaphoric relation between the subject phrase and the sub-place. When the sub-place isnot included in the subject phrase its definite article must be motivated by an anaphoricrelation to the wider context, viz. Hun har et barn i skolen ‘She has a child in the school’(Baron and Herslund 1997a: 139). Notice incidentally that have-clauses with definiteobjects reveal the locative nature of have since such clauses can constitute answers towhere-questions: Where are the children? — John has the children/them, cf. Wierzbicka(1988: 345).6. We assimilate in this context kinship relations to part–whole relations, cf. Baron andHerslund (1997a: 142f.).ReferencesBaron, I. 1995. “Complex Noun Phrases in Danish. A Valency Perspective’’. In Studies inValency I [RASK Supplement Vol. 1], L. Schøsler and M. Talbot (eds), 29–62. Odense:Odense University Press.Baron, I. 1997. “Possession in Noun Phrases: A Functional Analysis’’. In Possessive Structuresin Danish [KLIMT 3], I. Baron and M. Herslund (eds), 115–130. CopenhagenBusiness School.Baron, I. and Herslund, M. 1997a. “The Danish Verb have and the Notion of Possession’’.In Possessive Structures in Danish [KLIMT 3], I. Baron and M. Herslund (eds),131–157. Copenhagen Business School.

98 Irène Baron and Michael HerslundBaron, I. and Herslund, M. 1997b. “Place and Sub-Place: Locative Specifications inhave-Clauses’’. In From Words to Utterances in LSP [Copenhagen Studies in Language20], D. Faber and F. Sørensen (eds), 5–21. Copenhagen Business School.Benveniste, E. 1966. “‘Etre’ et ‘avoir’ dans leurs fonctions linguistiques’’. In Problèmes delinguistique générale, 176–186. Paris: Gallimard.Benveniste, E. 1974. “Fondements syntaxiques de la composition nominale’’. In Problèmesde linguistique générale II, 145–162. Paris: Gallimard.Clark, E. V. 1978. “Locationals: Existential, locative and possessive constructions’’. InUniversals of Human Language [Vol.4: Syntax], J. Greenberg (ed.), 85–126. Stanford:Stanford University Press.Freeze, R. 1992. “Existentials and other locatives’’. Language 68: 553–595.Greimas, A. J. 1966. Sémantique structurale. Paris: Larousse.Grimshaw, J. 1990. Argument Structure. Cambridge, Massachusetts: The MIT Press.Guillet, A. and Leclère, C. 1992. La structure des phrases simples en français. Constructionstransitives locatives. Geneva-Paris: Droz.Heine, B. 1997. Possession. Cognitive sources, forces, and grammaticalization. Cambridge:Cambridge University Press.Herslund, M. 1988. Le datif en français. Louvain-Paris: Peeters.Herslund, M. 1996. “Partitivité et possession inaliénable’’. In La relation d’appartenance.[Faits de langues 7], 33–42. Paris: Ophrys.Herslund, M. and Sørensen, F. 1994. “A Valence Based Theory of Grammatical Relations’’.In Function and Expression in Functional Grammar, E. Engberg-Pedersen et al. (eds),81–95. Berlin-New York: Mouton de Gruyter.Lyons, J. 1967. “A note on possessive, existential and locative sentences’’. Foundations ofLanguage 3: 390–396.Lyons, J. 1968. “Existence, location, possession and transitivity’’. In Logic, methodology andphilosophy of science III, B. van Rootselaar and J. F. Staal (eds), 495–504. Amsterdam:North-Holland.Müller, H. H. 1997. “The Danish s-genitive and related issues’’. In Possessive Structures inDanish [KLIMT 3], I. Baron and M. Herslund (eds), 93–114. Copenhagen BusinessSchool.Pustejovsky, J. 1995. The Generative Lexicon. Cambridge, Massachusetts: The MIT Press.Seiler, H. 1983. Possession as an operational dimension of language. Tübingen: Gunter Narr.Sørensen, F. 1997a. “Possession as Location’’. In Possessive Structures in Danish [KLIMT 3],I. Baron and M. Herslund (eds), 158–168. Copenhagen Business School.Sørensen, F. 1997b. “The Nightmare of the Genitive’’. In The Valency of Nouns [OdenseWorking Papers in Language and Communication 15], K. van Durme (ed.), 119–134.Odense University.Wierzbicka, A. 1988. The Semantics of Grammar [Studies in Language Companion Series18]. Amsterdam-Philadelphia: John Benjamins.

Chapter 6Possessum-oriented and possessor-orientedconstructions in RussianPer Durst-Andersen1. IntroductionIn the present paper I shall examine the various syntactic constructions thatcan be said to express possessive relationships in Russian. Two types areseparated, viz. possessum-oriented and possessor-oriented constructions. Thepossessum-oriented construction has the Possessum Role as (inanimate)surface and/or underlying subject (cf. Fr. qc. être à qn., i.e. ‘smth. exist withsmb.’), while the possessor-oriented construction has the Possessor Role as(animate) surface and/or underlying subject (cf. Fr. qn. avoir qc., i.e. ‘smb.have at his disposal smth.’). As indicated, the two construction types are notrestricted to possession-based state verbs like prinadležat’ (ipf) ‘belong’ andvladet’ (ipf) ‘own’. They are also found within possession-based activity verbslike zavedovat’ (ipf) ‘manage’ and rukovodit’ (ipf) ‘lead’ as well as withinpossession-based action verbs like obespečit’ (pf)/obespečivat’ (ipf) ‘provide’and snabdit’ (pf)/snabžat’ (ipf) ‘provide’. The reason for that can be tracedback to the fact that, one way or the other, all three verb classes include acertain state description (cf. Durst-Andersen 1992). This means that allpossession-based verbs, irrespective of verb class membership, involve apossession description that has the shape of a state description. Whereas thisis self-evident in the case of state verbs, it seems less clear in the case ofactivity verbs and action verbs. I argue that all possession-based activity verbslogically entail a certain possessive state description. For instance, X MAN-AGE A SHOP logically entails X CONTROL THE SHOP (i.e. the manager’sactivities are defined against the background of this state). Similarly, I arguethat all possession-based action verbs imply a posssessive state description, forinstance, X PROVIDE Y WITH SUCCESS implies Y HAVE SUCCESS (InRussian the truth or falsity of the state description is determined by aspect).

100 Per Durst-AndersenIn principle, the two construction types are identified by case marking: thepossessum-oriented construction involves the dative (cf. Y-NOM/ACC EXISTWITH X-DAT), while the possessor-oriented construction involves theinstrumental case (cf. X-NOM/ACC HAVE Y-INSTR). In other words, at thislevel of description it turns out that the genitive as such has nothing to dowith possession. Since this might seem odd to the reader I shall start with abrief examination of the Russian case system which is regarded as thenominal analogue to the verbal category of mood, or — to put it differently— the structure of a case system is determined by the structure of the moodsystem (cf. Durst-Andersen 1996).2. The Russian case systemThe semantic system of pure case in Russian consists of six cases: the nominative,accusative, vocative, genitive, dative, and the instrumental (the locativehas no place in this system, since it is used only in connection with prepositions).Just as the Russian category of mood can be divided into subcategories,viz. indicative, imperative, and subjunctive, the Russian category of case can bedivided as well. I take my starting point in the distinction between “modusrectus’’ (i.e. the indicative) and “modus obliquus’’ (i.e. all other mood forms)which has been observed within Russian linguistics for decades and thedistinction between direct and oblique cases which is known from the oldGreek and Latin tradition (but which is not observed anymore). In the Russiancase system the distinction between direct and oblique cases is not thatbetween the nominative and all other cases, but that between the nominativeand the accusative on the one hand and all other cases on the other. Thisappears, for instance, from the fact that both the nominative and the accusativecases are subject to the so-called genitive of negation rule (cf. Babby 1980and Franks 1995), they behave identically with respect to numerals, and, allthings being equal, they share the same morphological endings.The nominative and the accusative function just as the indicative moodwhich is what I prefer to call a direct mood form. As direct cases they assignthe relation of equality between the copy created by the noun and the originalin the situation referred to and thus signal that the person or thing denoted bythe noun is indeed present in the de facto situation referred to by the verb —the nominative and the accusative signal local reference. In short, the nominativeand accusative forms in Russian are not empty of semantic content: they

Possessum-oriented and possessor-oriented 101both assert local existence and both carry what can be called the ExistentiatorRole. All this has important implications for the presentation of new participantsand for topic shift in Russian and for its text-grammatical structures ingeneral (cf. Jansen 1998) — implications that make Russian text-grammaticalstructures completely different from those of English (cf. Nichols 1988). Allother cases, i.e. the vocative, genitive, dative, and instrumental cases, functionas non-indicative forms which are all oblique mood forms. As oblique casesthe vocative, genitive, dative and the instrumental assign lack of equalitybetween the copy named by the noun and the original in the situation referredto, and thus they signal non-local reference, i.e. the person or thing denoted bythe noun having these case forms are not asserted to be present in the de factosituation. The vocative case (e.g., Mam! ‘Mommy!’, Pap! ‘Pappa!’, Len! ‘Lena!’,Saš! ‘Sasha!’, etc.) is limited more or less to first names, but it is included inthe Russian case system — not only because Russian case is not understood asmarking syntactic dependencies of nouns (obviously, the vocative has nothingto do with that), but also because of its high productivity among all Russians(cf. Bilý 1990). It is acquired simultaneously with the imperative mood and,not surprisingly, it functions exactly as this mood form, viz. it treats the lack ofequality between the copy and the original as a matter to be changed (followingthe script Create an original against the background of the copy being showedto you!). In other words, the speaker requests the hearer’s local reference: hewants the hearer to come into his picture in the shape of a concrete figure (inthe sense of ‘Come here’) or he wants himself to be figure in the hearer’s nextpicture (in the sense of ‘Look at me’). Note that the vocative case and theimperative mood not only share the same prosodic features, but also zeroendings— this can and should be taken as being a symptom of their sharedcontent.The remaining three cases (the genitive, dative and the instrumental)function as a broad subjunctive mood: they treat the lack of equality betweenthe copy and the original as a matter of fact, i.e. the person or thing named byone of these three oblique cases is not present in the de facto situation referredto by the noun. The difference between them seems to correspond to thedistinction between objective mood (genitive) and subjective mood (dative andinstrumental) — a mood distinction which is found in several Amerindianlanguages (cf. Whorf 1950). In Russian the distinction corresponds to outercases (they denote outer reference or involve the objective reality) and innercases (they denote inner reference or involve the subjective reality). If “Xwishes Y’’ or if “X fears Y’’, then Y will be in the genitive case, because “Y’’ is

102 Per Durst-Andersennot present in the situation, where “X’’ is having a wish for “Y’’ or is feeling afear of “Y’’. In the former case, the genitive is used as a kind of desiderative(the static variant) — in the latter case as a kind of obviative (the dynamicvariant). We are here dealing with objective reality, i.e. with external realitywhich can be touched, seen or heard: If you wish for cold weather, there is nocold weather (static use); if you fear snakes, you in fact avoid them, try to stayout of contact with them (dynamic use). The dative and the instrumental donot deal with the objective reality, but with what could be called the subjectivereality, i.e. with internal reality. If “X is unfaithful to Y’’ or if “X abuses Y’’,there is no local reference in either case: in the former case “X’’ is doingsomething with somebody else, but not with “Y’’; in the latter case “X’’ is notdoing something directly to “Y’’ — there is no contact between “X’’ and “Y’’.Therefore the accusative case is out of the question. Although they are not adirect part of the situation named by their respective verb, they are both partof another situation which belongs to internal reality. In the case of “X isunfaithful to Y’’ Y is put in the dative case, because “Y’’ is suffering indirectlyfrom “X’s’’ behaviour or is affected mentally by it (“Y’’ carries what could becalled the Male/Benefactive Role). In the case of “X abuses Y’’ Y is put in theinstrumental case, because “Y’’ is used as a kind of Means — not directly, butindirectly (“Y’’ carries the so-called Victim Role — the animate counterpart ofMeans). In the following I attribute to each of these three oblique case formsa specific viewpoint or a particular percept that marks a relation between twoentities. Section 3 and the two subsections of four will be a brief examinationof the three viewpoints and their relationship to the notion of possession.3. Possession and the genitiveIt might be the case in other so-called case languages that the genitive hassomething substantial to do with possession, but in Russian the relation seemsto be purely superficial and belong to what is normally called communicatedcontent, i.e. the genitive as a grammatical form does not denote possession, butthe genitive construction seen as a pragmatic tool communicates such arelationship. As the oblique case per se the genitive itself assigns non-localreference, i.e. the person or thing named by the genitive noun is not presentin the situation named by the verb. The genitive involves what I shall call anextrovertive viewpoint, i.e. it denotes a relation or a direction from somethingnear, A, to something distant, B. A is its starting point and B—the object of

Possessum-oriented and possessor-oriented 103comparison—its terminal point ((X) marks the locator and x} his vantagepoint):(X)x} A >—— BFigure 1. The extrovertive viewpointIf (X) is located in A and looks at B from A, then B must be distant from Aand separated from it. In that way the genitive denotes remoteness andseparation. This appears from the following examples:(1) a. On (nom) živët v kvartire (loc) brata (gen).‘He is living in his brother’s appartment’b. On (nom) živët na/v kvartire (loc) u brata (gen).‘He is living together with his brother in his appartment’In (1a) there is no local reference, i.e. his brother is not present in the situationreferred to. (1b) is an instance of the general rule that all prepositionspresuppose local reference and therefore we get the meaning “near distance’’.The fact that not only linguists, but also Russian people feel a certainpossessive-like relation in certain genitive contexts can be explained by theextrovertive viewpoint combined with a human being as starting point and alack of dynamicity. If one says dom (nom) brata (gen) ’my brother’s house’,one in fact separates “house’’ from “all my brother’s belongings’’ in order tofocus one’s attention on the house. What triggers the genitive form brata‘brother’s’ is not that “house’’ is included in the set of all his belongings, butthat something has been removed and excluded. In other words, the genitivedoes not involve possession sensu stricto, but due to the fact that exclusionpresupposes inclusion, people attach the presupposed content to the genitive.That the genitive has nothing substantial to do with possession and that therelation of the genitive to possession is established not by language, but bypeople themselves whereby the relationship can be said to belong to communicatedcontent appears indirectly from the fact that what could be called theRussian variant of inalienable possession (cf. Žurinskaja 1977, Herslund 1997)is linked to the so-called possessive adjectives (e.g. Papin dom ‘My father’shouse, Mamin komp’juter ‘My mother’s computer’, Alešina komnata ‘Alesha’sroom’, etc.). One could not use the genitive here unless the speaker deliberatelywanted to emphasize that s/he was talking to strangers simply because theentities referred to cannot be thought of as separable — and separationbetween two entities is exactly what the genitive marks (cf. trebovat’ + gen‘demand smth.’ and želat’ + gen ‘wish smth.’ as static variants and izbežat’ +

104 Per Durst-Andersengen ‘avoid smth.’ and lišit’ kogo-n. + gen ‘deprive smb. of smth.’ as dynamicvariants). In other words, the genitive does not involve possession sensu stricto(see also Durst-Andersen 1997).4. Possession and state descriptionsIn fact, possession is connected to the dative and the instrumental, i.e. to thoseoblique cases that are inner cases and not outer cases as the genitive is (cf.Durst-Andersen 1996). Those state verbs which create a ground-propositionbased on a description of a possession relationship are named possession-basedstate verbs. They fall into two groups: (1) those verbs that take the dative(Section 4.1) and (2) those verbs that take the instrumental (Section 4.2).4.1 The dative and the possessum-oriented constructionIn opposition to the genitive case the dative marks what I call an introvertiveviewpoint, i.e. it denotes a relation or a direction from something distant, B,to something near, A. Since B is the starting point of the dative and A itsterminal point, the dative and the genitive are converse cases — in otherwords, X’s location point and vantage point do not coincide:(X) A ——< B{xFigure 2. The introvertive viewpointIf a person, X, is located at A and wants to view A, X has to place his vantagepoint outside of A, in casu in B. By looking at A from the standpoint of B, X infact embraces B and in this way includes B in A. If we view this strictly fromthe point of view of B, B is received in A or is incorporated into A’s domain.In that way the dative denotes nearness (or, alternatively, closeness) andreception — the converse of remoteness and separation which were linked tothe genitive. The dative is used in possessum-oriented constructions as in (2a)with a lexical verb and (2b) with byt’ ‘be’:(2) a. Dom (nom) prinadležit prepodavatelju (dat).‘The house belongs to a high school teacher’b. Moskve (dat) —850(nom) let (gen).‘Moscow is 850 years old’

Possessum-oriented and possessor-oriented 105As we see, the subject carries the Possessum Role in both examples. Thesubject in the nominative is typically inanimate, whereas the oblique object orthe adverbial in the dative is typically animate.4.2 The instrumental and the possessor-oriented constructionThe instrumental can be said to presuppose the existence of the two otheroblique cases, since its viewpoint is correlative, i.e. it denotes an interrelationshipbeween something near, A, and something distant, B. This means thatone (i.e. (X)) is located at something near, A, but is viewing at the relationshipbetween A and something distant, B, from an external point of view (i.e. x}),i.e. from the point of view of totality. In other words, also in this case X’slocation point and vantage point do not coincide, but in contradistinction tothe dative vantage point the instrumental one is placed at an upper level orcompletely outside A and B. The result is that we leave the level of comparisonwhere A and B are felt to be either identical or non-identical, and instead weenter another level where A is incommensurable with B, but nevertheless isclosely related to B and vice versa. This interrelationship between somethingnear, A, and something distant, B, can be depicted in the following way:(X) A >——< Bx}Figure 3. The correlative viewpointSince there is only one genuine kind of interrelationship, that of part andwhole, the correlative viewpoint can be said to entail the relation of inclusionand that of inherent property. This appears from (3a) where we are dealingwith physical or acquired properties, as well as from (3b), where we areconcerned with intellectual or transferred properties:(3) a. On (nom) vladeet domom (instr).‘He owns a house’b. On (nom) obladaet bol’šimi (instr) talantami (instr).‘He has great talents’As we see, possessor-oriented constructions involve an animate subject in thenominative and an inanimate oblique object in the instrumental. In both casesthe instrumental noun denotes an inherent property of the person in question.

106 Per Durst-AndersenMoreover, it can be argued that the two possession-based state verbs in,respectively, (3a) and (3b) always imply that somebody controls something,i.e. this something is at the possessor’s disposal and can be used by him as aMeans.The meaning of inherent property is also involved in those adjectives withshort form (SF) that govern the instrumental case:(4) Strana (nom) bogata (SF)/bedna (SF) tradicijami (instr).‘Our country is (inherently) rich/poor of traditions’If we take polnyj ‘full’ that takes the genitive or the instrumental case, themeaning of inherent property attached to the instrumental becomes evenmore obvious (cf. 5a and 5b):(5) a. Rjumka (nom) polna (SF) vodki (gen).‘The glass is full of vodka’b. Rjumka (nom) polna (SF) vodkoj (instr).‘The glass is full of vodka’In the case of (5a) it is just stated that there is a glass that has some vodka init. In the case of (5b) we are dealing with a vodka glass with vodka, i.e. vodkais an inherent property of this glass and therefore vodka is put in the instrumental.4.3 SummaryWhat has been said so far, can be summarized in the following scheme:Genitive: The extrovertive viewpointSEPARATION/REMOTENESSFrom smb. near >—— to smth. distantDom (Nom) brata (Gen).‘His brother’s house’Dative: The introvertive viewpointRECEPTION/NEARNESSTo smb. near ——< from smth. distantDom (Nom) prinadležit bratu (Dat).‘The house belongs to his brother’

Possessum-oriented and possessor-oriented 107Instrumental: The correlative viewpointINCLUSION/INTERRELATIONSHIPSmb. near >——< and smth. distant is interrelatedBrat (Nom) vladeet domom (Instr).‘His brother owns a house’As will appear from the following sections, the distinction between a so-calledpossessum-oriented construction (signalled by the dative) and a so-calledpossessor-oriented one (signalled by the instrumental) is not restricted to statedescriptions — it is a characteristic feature of the entire Russian system.5. Possession and action descriptionsAction verbs are also called complex verbs because — opposed to simplexverbs, i.e. state verbs and activity verbs — they do not involve one groundproposition,but two ground-propositions, i.e. one based on an activitydescription (SOMEBODY X PRODUCE AN ACTIVITY), and another based ona state description. When speaking of possession-based action verbs (i.e. anactivity description + a description of possession) Russian distinguishestwo constructions, viz. (a) čto-n. (prinadležit/est’) komu-n./čemu-n. (thepossessum-oriented construction) and (b) kogo-n. (vladeet/obladaet) čem-n.(the possessor-oriented construction).5.1 The dative and the possesum-oriented constructionAs appears from the above, the indirect object function of the dative casesignals the possessum-oriented construction:(6) Ministr (nom) podčinil institut (acc) ministerstvu (dat).‘The minister placed the institute under the ministry’(7) On (nom) podaril ej (dat) podarok (acc).‘He gave her a gift’Both verbs prototypically take animate subjects, but differ with respect to theindirect object. (6) takes an inanimate indirect object, i.e. it instantiates the GoalRole, while (7) takes an animate indirect object, i.e. it instantiates the RecipientRole. Nevertheless, both (6) and (7) can be argued to manifest the possessumorientedconstruction where the inanimate accusative object occupies the

108 Per Durst-Andersenposition of underlying subject in the state description of the action (cf. Theinstitute (Acc) belongs to the ministry (Dat); A gift (Acc) belongs to her (Dat)).5.2 The instrumental and the possessor-oriented constructionIn possessor-oriented constructions with the instrumental we observe thereverse relations between animate and inanimate entities. Here the animateaccusative object occupies the position of underlying subject in the statedescription of the action:(8) On (nom) podaril/odaril eë (acc) vnimaniem (instr).‘He gave her his attention’While (7) states that a certain gift now belongs to her, (8) states that she has nowhis attention at her disposal, i.e. she controls his attention. Although, admittedly,podaril, but not odaril in (8) represents a rather archaic example which is usedin more poetic speech, I use it deliberately to show that some verbs allow bothconstructions. Apart from darit’/podarit’ there are only a few of them:(9) a. Trener (nom) obespečil emu (dat) uspex (acc).‘The trainer provided for his success’b. On (nom) obespečil sekretaršu (acc) francuzskimi (instr) duxami(instr).‘He provided his secretary with French perfume’It appears, however, that the possessor-oriented construction itself is quitenormal and frequent in Modern Russian:(10) On (nom) snabdil eë (acc) odeždoj (instr).‘He provided her with clothes’(11) Roditeli (nom) odarili detej (acc) igruškami (instr).‘The parents gave the children toys’(12) Firma (nom) premirovala lučšix rabotnikov (acc) časami (instr).‘The firm gave its best workers a watch as a reward’(13) Prepodavatel’ (nom) vooružil studenta (acc) znanijami (instr).‘The teacher equipped the student with knowledge’In all examples the accusative object occupies the position of underlyingsubject in the ground-proposition involving a state description and togetherwith the instrumental object it manifests the same reading, viz. somebody has

Possessum-oriented and possessor-oriented 109something at his or her disposal. Note that this meaning is responsible for thefact that as a rule the possessor-oriented construction requires an obliqueobject with a non-singular reference (cf. *On snabdil syna ženoj (sg) ‘Heprovided his son with a wife’ which is ungrammatical vs. On snabdil synaženami (pl) ‘He provided his son with wives’ which is grammatical, althougha little bit funny).5.3 Concluding remarksIt seems to be confirmed that the dative and the instrumental cases are closelyrelated to the notion of possession and that they differ from one another bybeing converse with respect to surface subject when speaking of state verbsand with respect to underlying subject when speaking of trivalent action verbs.These observations seem to be confirmed by data from French (I would like tothank Michael Herslund for having drawn my attention to these data):(14) The possessum–oriented constructionIls ont fourni des médicaments à l’hopital (ETRE à)‘They provided medicine for the hospital’(15) The possessor–oriented constructionIls ont fourni l’hopital de/en médicaments (AVOIR)‘They supplied the hospital with medicine’In (14), which is an action verb construction involving an activity as well as astate description, we can identify the well-known state verb construction qc.être à qn., and in (15), which represents another action verb construction, wecan identify the other well-known state verb construction qn. avoir qc. In thatway, French shows in a convincing way that action descriptions indeed consistof an activity description (cf. Ils ont fourni) and a state description (e.g.l’hopital a des médicaments), which in this case is asserted. (For a detailedaccount of possessive constructions in French, see Luxt 1977).6. Possession and activity descriptionsBecause any activity is and must be performed against the background of acertain state, be that a state based on location (e.g. rabotat’ ‘work’ and igrat’‘play’), possession (e.g. zavedovat’ ‘manage’ and rukovodit’ ‘lead’), experience(e.g. gljadet’ ‘look at’ and slušat’ ‘listen’), or a certain quality (e.g. blestet’

110 Per Durst-Andersen‘shine’ and bolet’ ‘hurt’), it appears that activity verbs are state verbs plus anactivity description. The role of the state description can be extracted from therole that the very state plays in a real activity. If a certain soldier is creeping, hemust be in a certain horizontal position — he could not be creeping, if thisstate did not obtain. The state is thus a necessary condition, but not thesufficient condition, for performing the activity of creeping. For naturallanguages this cannot but mean that all activity verbs involve an activitydescription which logically entails a certain state description.6.1 The instrumental and the possessor-oriented constructionAs already indicated above, those activity verbs that take oblique objects in theinstrumental case (e.g. zavedovat’ ‘manage’, rukovodit’ ‘lead’, upravljat’‘govern’, komandovat’ ‘command’, vlastvovat’ ‘control’, vedat’ ‘govern’, pravit’‘lead’, and rasporjažat’sja ‘control’) belong to so-called possession-based verbs.This means that just as the state verbs vladet’ ‘own’ and obladat’ ‘possess’ theyare possessor-oriented and denote an inherent property. The inherent propertywhich is described in the entailment structure of the verb normallyincludes a certain institution (e.g. a certain section) that has been transferredto a certain person and is now placed in his jurisdiction (cf. 16a, b):(16) a. Molodoj, talantlivyj čelovek (nom) zavedyet novym otdelom (instr)amerikanskoj firmy.‘A talented young man leads the new section of the American firm’b. Novyj komandir, staršij lejtenant Ivanov (nom), upravljaet 50-ojbatareej (instr).‘The new commander, first lieutenant Ivanov, is in charge of the 50thbattery’In other words, the section and the battalion in question are inherentlypossessed by a talented young man and first lieutenant Ivanov, respectively,and against this background they perform their respective activities. It iscrucial to note that these verbs are only used to give a characterization (showedin the English translation by the simple present tense form) — they cannot beused to describe situations in reality (One cannot say *Smotri, on upravljaetbatareej ‘Look, he is leading the battery’). This is explained by the status of theinstrumental case as an oblique and inner case. Some of the verbs in question,however, may be used in that function, but in that case they are used with apreposition (cf. 17a, b as well as 18a, b):

Possessum-oriented and possessor-oriented 111(17) a. Lena (nom) vlastvuet soboj (instr).‘Lena controls herself’b. Lena (nom) vlastvuet nad nami (instr).‘Lena is dominating us’(18) a. On (nom) komanduet polkom (instr).‘He is a brigadier’b. On (nom) komanduet nad domašnimi (instr).‘He is pushing around his family’Although we do not find a genuine possessum-oriented construction with thedative case within activity verbs, it must be emphasized that exactly these verbsaccept a synthetic passive construction despite the fact that the Russiangrammar does not allow passives with intransitive verbs. So one will not finda single example in any of the existing grammars of Russian, but has to lookthem up in real life (I would like to thank Elena Wahl, Jan Erik Nielsen, andCarl-Henrik Lund for the examples):(19) Krome togo, otdelenija (nom) upravljalis’ (passive) komandirom (instr)vzvoda s pomošč’ju signalov ...‘Besides the sections were guided by the platoon commander by means ofsignals...’(20) A poskol’ku upravljaetsja (passive) vse éto (nom) nervnoj sistemoj(instr), to xozjain — ryba ili kal’mar — možet menjat’ intensivnost’ ...‘And since all this is governed by the nerve system, then the host, a fish oran octopus, may change the intensity ...’(21) Novyj otdel (nom) amerikanskoj firmy zaveduetsja (passive) molodym,talantlivym čelovekom (instr).‘The new section of the American firm is headed by a young talented man’(22) Rabota (nom) aspirantov rukovoditsja (passive) professorom Ivanovym(instr).‘The Ph.D.students are supervised by Prof. Ivanov’(23) Rossija (nom) pravitsja (passive) novymi russkimi (instr).‘Russia is led by the new Russians’(24) Kem (instr) rukovoditsja (passive) éta rabota (nom)? Toboj (instr) ilimnoj (instr)?‘Who is in charge? Is it you or is it me?’

112 Per Durst-AndersenMy point is that these passive constructions can be created, and indeed arecreated, because they are possessum-oriented and in that way can be said to fillin a vacuum at this specific place.7. Summarizing remarksI have tried to demonstrate that the genitive is connected to the notions ofseparation and remoteness which appear to be almost the opposite of possession.Instead it was shown that the dative, which was linked to the notions ofreception and closeness, as well as the instrumental, which was linked to thenotions of interrelationship and inclusion, are naturally correlated with thenotion of possession. Two construction types denoting possession wereseparated on the basis of case marking: a possessum-oriented construction withthe oblique object in the dative and a possessor-oriented construction with theoblique object in the instrumental. In state descriptions and in activity descriptionsthe grammatical surface subject is realized by, respectively, an inanimatenoun and an animate noun in the nominative, i.e. a direct case form. In actiondescriptions, the underlying subject of the state description component isoccupied by another direct case form, viz. the accusative. That the two constructiontypes do not belong to what is normally called “accidental properties’’,but, on the contrary, have gained a firm ground in the Russian grammaticalsystem appears from the fact that the possessum-oriented construction hasgained access into activity descriptions in the shape of a synthetic passiveconstruction. This construction should be ungrammatical according to therules of Russian grammar because no intransitive verb may take a passive form.Since these intransitive activity verbs do so, the separation of two distinctpossessive constructions may also serve as a tool to explain a passive constructionthat hithertoo has been left unexplained within Russian linguistics.ReferencesBabby, L. H. 1980. Existential sentences and negation in Russian. Ann Arbor: Karoma.Bilý, M. 1990. “The case remains open. Some notes on a new book on morphological casein Russian’’. Russian Linguistics 14: 185–203.Durst-Andersen, P. 1992. Mental grammar. Russian aspect and related issues. Columbus,Ohio: Slavica.Durst-Andersen, P. 1996. “Russian case as mood’’. Journal of Slavic Linguistics 4: 177–273.

Possessum-oriented and possessor-oriented 113Durst-Andersen, P. 1997. “Possession in Danish and Russian from a typological perspective’’.In Possessive Structures in Danish [KLIMT 3], M. Herslund and I. Baron (eds),78–92. Copenhagen Business School.Franks, S. 1995. Parameters of Slavic morphosyntax. Oxford: Oxford University Press.Herslund, M. 1997. “Partitivity and inalienable possession’’. In Possessive Structures inDanish [KLIMT 3], M. Herslund and I. Baron (eds), 1–44. Copenhagen BusinessSchool.Jansen, L. 1998. “On text structure in Russian and Danish. A comparative study of Russianand Danish Frog-stories’’. In Clause combining and text structure [Copenhagen Studiesin Language 22], M. Herslund and I. Korzen (eds), 43–62. Copenhagen: Samfundslitteratur.Luxt, L. I. 1977. “Kategorii bytija i obladanija (franzuzsko-rumynskie paralleli)’’. InKategorii bytija i obladanija v jazyke, V. N. Jarceva (ed.), 125–143. Moscow: Nauka.Nichols, J. 1988. “Nominalization and assertion in scientific Russian prose’’. In Clausecombining in grammar and discourse, J. Haiman and S. A. Thompson (eds), 399–428.Amsterdam-Philadelphia: John Benjamins.Whorf, B. L. 1950. “An American Indian model of the universe’’. International Journal ofAmerican Linguistics 16: 67–72.Žurinskaja, M. A. 1977. “Imennye posessivnye konstrukcii i problema neottoržimojprinadležnosti’’. In Kategorii bytija i obladanija v jazyke, V. N. Jarceva (ed.), 194–258.Moscow: Nauka.

Chapter 7Datives and comitatives as neighbouringspousesThe case of indirect objects and comitatives inDanishLars Heltoft1. IntroductionIn many languages, datives and comitatives merge, typically with respect toinflectional case, so that the facility of distinguishing between such functionsis left to adpositional constructions. In other languages, there are no suchmergers, but only systems that maintain a sharp distinction.In Classical Greek, the inflectional dative allows both sociative andinstrumental readings along with receptive readings. Simple examples aregiven in (1a–c):(1) a. chilíais nausì diebíbase tè:n stratián (Instrumental)1000-dat ships-dat he let go across the army‘he let the army go across by means of 1000 ships’b. o’udeìs a’utô:i dielégeto ho:s ’androphóno:i ’ónti (Sociative)nobody he-dat spoke since murderer-dat being-dat‘nobody spoke with him since he was a murderer’c. Kûros dído:sin a’utô:i hèx me:nô:n misthón (Receptive)Kyros gives he-dat six months’ payment‘Kyros gives him six months’ payment’I shall return to this merger in Classical Greek in 6. Together with a mentionof the situation in Old Scandinavian, in particular Old Danish, this will besufficient to illustrate the inflectional dative-comitative merger for the presentpurpose.By contrast, Modern Danish is an example of a language with no such merger.A clear distinction must always be made between indirect objects and comi-

116 Lars Heltofttatives, the indirect object being a positionally defined grammatical function andall comitatives being prepositional. This situation is on display in Section 1.1.1.1 The neighbouring spouses — Danish IO and CSDanish has categorical sentence structure (Sasse 1987), its word order patternsare XVSO in main clauses (the unmarked realization of X is S), SVO insubordinate clauses. There are no functional case remnants of its formertypically Indo-European inflectional case manifestations of semantic rolesystems. 1 Relevant to our present topic are the word order rules for theindirect object (IO). This is identified as the first object:(2) a. De tildelte kongen en ubetydelig rolle.they allotted the king an insignificant role‘They allotted the king an insignificant role’b. Man forklarede kongen sammenhængen.they explained the king the background‘They explained the background to the king’c. De bebrejdede kongen nederlaget.they reproached the king the defeat‘They reproached the king with the defeat’The comitative is manifested by means of an adpositional construction, theco-subject construction, named ‘consubjectum’ by the Danish grammarianHøysgaard (1752). The preposition marking the co-subject (CS) is med ‘with’:(3) a. Ole spiller fodbold med Peter.Ole is playing soccer with Peter‘Ole is playing soccer with Peter’b. Hunden logrer med halen.the dog wags with its tail‘The dog is wagging its tail’The only detailed grammatical description of the CS-construction is ErikHansen (1971). I owe many examples and views to him.1.2 Constructionally grammaticalized possession in DanishBoth indirect object constructions (IO) and co-subject constructions (CS)grammaticalize possession, yet they are markedly different. Not only do they

Datives and comitatives 117differ in their expression structure, they also impress on the semantic substanceof possession a corresponding distinction of content structure — orcontent form, in the Hjelmslevian sense of this term. 2 I shall try to determineto what extent this distinction can be formulated in terms of language specificsemantic roles.In addition to semantic roles, I shall claim that the two constructions IOand CS must be analyzed in terms of a generalized notion of subjecthood, anexplicit version, in a sense to become explicit below, of the ‘logical subject’ oftraditional grammar. Mutatis mutandis, both constructions are organized onthe basis of what will be called content subjecthood. Provisionally, this meansthat both IOs and CSs must obey a specific set of subject restrictions, knownfrom explicit expression subjects elsewhere in Danish syntax.On the agenda are the following items:1. The differences between the Danish IO and CS and their relatedness, withrespect to semantic roles.2. The notion of content subjecthood as a means of relating IOs and CSs at thelevel of content syntax, and, more tentatively:3. A brief illustration of how the Modern Danish situation may relate semanticallyto older Indo-European stages, as attested by Classical Greek and OldScandinavian.There will be a fourth perspective, too:4. The content relations between dative, possessive, comitative, and instrumentalfunctions are often described as transitory steps along cognitivecontinua. The notion of a continuum will not suffice here if taken necessarilyto include fuzzy content borderlines. The stance taken by Heine, Claudi andHünnemeyer (1991) is a sensible distinction between conceptually distinct casemeanings and contextually determined continua; I shall, however, insist thatcontent distinctions must be thought of as linguistically coded, languagespecific content. This again will be claimed to imply that detailed analyses oflanguage specific content form must be included as starting points for analysesof other languages or other stages of development, along with hypothesesabout universal networks of grammaticalization.Not all border lines are fuzzy, and the content relations suggested for theDanish situation will not be vague. They will include, among others, the viewthat IO-constructions and CS-constructions differ in their organization of thesemantic role relationship of Framing (see below), and that this difference can

118 Lars Heltoftbe described in terms of options within the system of telicity, combining withthe notion of frame through presuppositional relationships and entailmentrelationships.2. Content subjects, content objects and semantic rolesIn contrast to most treatments of this topic, I do not take semantic roles tobe semantic primitives, for instance cognitively given sets of content functions.We cannot a priori assume that Recipient, Possessor, Beneficiary,Experient, Agent, Instrument, Source and Goal etc. are relevant for any givenlanguage. 3 Specific languages single out different sets of semantic roles, andthey coin these sets differently, in accordance with the design of theirgrammatical content systems in general. 42.1 Lexical items for verbsSemantic roles may be viewed as lexically controlled phenomena, and as suchthey belong to the lexicon as part of valency templates for verbs. This is infact the predominant approach in recent linguistics. Whatever the technicalor theoretical differences between for instance Gruber’s (1967, 1976[1965]),Fillmore’s (1968, 1972, 1978 and later), Jackendoff’s (1972, 1983, 1987 andlater), and Grimshaw’s (1990) approaches, they converge in being descriptionsof ‘lexical items for verbs’.The alternative is to see semantic roles as codings of the constructionsof the language in question (Eisenberg 1992, 1996) or of the positions of itsword order system. In this approach, only those roles will count asgrammaticalized — and thereby relevant for a given language — that arepicked out by one or more members of the following systems: inflectionaland adpositional paradigms, syntactic constructions, and positions of theword order system. For instance, a Modern Danish subject is open withrespect to semantic roles, meaning that the subject position in Danish isnot in itself coded with respect to semantic roles, but open to whateverrestrictions lexical items or syntactic constructions may impose on it. Thus,both perspectives are relevant. What is almost always presupposed, butcannot be, is that all relevant semantic role structure must be lexicallyconditioned.The outcome of this strategy will be descriptions of the semantic role

Datives and comitatives 119systems grammaticalized in given languages. (For this wider view ofgrammaticalization, see Heltoft 1996b).To take an example, the open subject position of Danish is somewhatunexpectedly matched by a restricted direct object position. Gruber (1976),Halliday (1970, 1996), Fillmore (1968 etc.), Jackendoff and others speak of theobject as the ‘neutral role’, ‘objective’, ‘theme’, etc., but in Danish, directobjects cannot be agents. Crucially, English examples like:(4) a. She walks her little dog in St. James’s Park.b. The sergeant marched the recruits.c. The dog wagged its tail.have no direct Danish counterparts. Danish demands the CS in such instances,see (3b).2.2 Content subjectsAny constituent that complies with the semantic subject restrictions of a givenverb, is itself a content subject. Similarly, any constituent that complies withthe semantic object restrictions of a given verb is itself a content object. Atrivial, but crucial example is the status of the ‘logical subject’ in Danishder-constructions:(5) a. Der må snart komme penge.there must soon come money‘There must soon be money’expression DO = content subject (NON-AGENT)b. Der arbejder 50 mand på projektet.there work-pres.act 50 men on the project‘50 men work/are working on the project’expression DO = content subject (AGENT)In these examples of the der-construction, the positionally identified directobjects (penge ‘money’ and 50 mand ‘50 men’) must comply with the lexicalsubject rules of the verbs komme ‘come’ and arbejde ‘work’, respectively. Morespecifically, they must comply with the semantic subject roles of these verbs.(5a,b) imply the existence (not the truth!) of sentence patterns like (5¹). In (5¹)the relationship between expression subject and content subject is isomorphic,so that the expression subject is also the content subject, but in (5a,b) thecontent subject was manifested as the direct object:

120 Lars Heltoft(5¹) a. Penge kommer altid senere end man venter det.‘money will always come later than one would expect’b. 50 mand arbejder på projektet.‘50 men work on the project’A similarly classical example of a content object is (6), the upgraded directobject of passive constructions:(6) Sagen blev grundigt undersøgt.the case was thoroughly investigated‘The case was thoroughly investigated’expression subject = content objectWhere the semantic analysis of the IO is concerned, the verbs have/få’have/get’ (and their negative counterparts mangle/miste ‘lack/lose’) will play acentral role. Many syntactic and semantic properties of the Danish IO matchthe subject properties of få, and the IO will be said to function as the contentsubject of få. Quite informally, example (2a) entails (2a¹):(2) a¹. Kongen fik en ubetydelig rolle.‘The king had (was given) an insignificant role’Quite in keeping with the ‘logical subject’ analysis of traditional grammar(Diderichsen 1946), we shall say that the IO is a content subject in (2). Giventhat the analogy between the IO and the subject of få is valid, the IO willmanifest the same semantic role as the subject of få. What this semantic roleis, will be discussed in what follows. I shall assume that it can be adequatelydescribed on the basis of a notion of abstract meronymy, the relationship ofFRAMING.Similarly, CSs are content subjects of their clausal verbs, on the basis of thefollowing entailment relationships (3a¹ and 3b¹) and presuppositional relationship(3b¹¹):(3) a¹. Ole spiller fodbold med Peter.→ Peter spiller fodbold med Ole‘Peter is playing soccer with Ole’b¹.Hunden logrer med halen.→ halen logrer‘Its tail is wagging’

Datives and comitatives 121b¹¹. Hunden logrer med halen.→ (P) hunden har en hale‘The dog has a tail’In both types, the CS qualifies as a content subject of the verb in question, asshown by the entailment relationship. A free CS like med Peter ’with Peter’ in(3a) is a content subject in the relation være med i, ’participate in’. A valencybound CS (3b) is the object of a have-relation, the content subject of which is theexpression subject (or “grammatical subject’’, as tradition misleadingly has it). 52.3 Abstract possessionIf a notion of possession is relevant here, it must be a fairly abstract one.English offers no good everyday language options, but Danish offers tilhørsforhold,the nearby equivalent of the likewise feasible German word Zugehörigkeit,or French appartenance, cf. Benveniste (1966). This notion will besuggested as the semantic base for both the IO and the CS in Danish. It willdefine the framing relationship.3. The IOThe content of the IO is complex and abstract (for an internationally accessibleexposition, see Herslund (1986), who stresses the abstract nature of thisrelationship). The IO is an object that is at the same time qualified as thecontent subject of få ‘get’. As such, it acquires the semantic role of abstractwhole or Frame: The DO is transferred to the IO, making the DO a part of alarger whole, delimited or framed by the IO, which may then be said tofunction as the Frame for this transition. We shall speak of the IO as theFrame. The Frame, again, impresses on the nonagentive direct object acontextually determined interpretation as the entity Framed.Prototypically, the IO is bound up with transitional (telic) meaning, see(2a,b). Transitional verbs proper are movement verbs (2a), others, for instanceverbs of communication, are extension verbs (2b). It follows that like inEnglish, Benefactive readings are no longer productive in IOs. A number ofDanish IO-verbs without telic meaning like (2c) have developed parallelalternative patterns:

122 Lars Heltoft(7) a. Subject V Indirect Object Direct Object >Subject V Direct Object Oblique Objectb. de bebrejdede kongen nederlaget >de bebrejdede kongen for nederlagetthey reproached the king for the defeatFigure 1 shows the mapping of subject and object of have, få, mangle, miste(‘have, get, lack, lose’) onto the IO-construction. These four fairly abstractverbs denote abstract meronymy, the ‘Zugehörigkeit’. The denotata of subjectand object are viewed as a whole, the subject being the point of departure: theframe that the object is framed into, that is, integrated into as a part.Verbs taking IO are prototypically lexical causatives of these four basicverbs. For such verbs, illustrated by (2a), subject restrictions on the IO mustcorrespond to the restriction imposed by have/få on their subjects.subject ofhave, få, mangle, misteindirect objectdirect object of have, få,mangle, mistedirect objectFigure 1. Verbs of possession syntactically related to IOconstructions.3.1 The range of få-subjectsEvery time there is a proper transitional IO-construction, there is a similar,basic få-construction, and vice versa. I can be disproved by somebody showingthat this parallel does not hold:(8) a. Hun fik en 2CV.she got a 2CV‘She got a 2CV’b. Forældrene forærede hende en 2 CV.her parents presented her a 2 CV‘Her parents presented her with a2CV’(9) a. Universiteterne mistede mange resurser.the universities lost many resourcesb. Ministeriet fratog universiteterne mange resurser.the ministry deprived the universities many resources‘The ministry deprived the universities of many resources’

Datives and comitatives 123(10) a. Hvordan elefanten fik sin snabel. (Inalienable possession)how the elephant got its trunkb. Hvordan krokodillen gav elefanten sin snabel.how the crocodile gave the elephant its trunk(11) a. Krogen fik en orm på. (Instrumental/intentional)the hook got a worm on‘The hook was provided with a worm’b. Hun gav krogen en orm på.she gave the hook a worm on‘She provided the hook with a worm’The range of these subjects does not comprise the locative domain. (11) and(12a) are meronymic constructions in the wide sense defined above, notlocatives. Indisputably locative features as shown in (12b) do not enter thefå-construction, since they do not match its semantics. The få-construction in(12a) indicates an intended part-whole relationship, and the purely locativemeaning of the der-construction (12b) is hard to squeeze into this semanticframe. In cases of pure location there is no such problem, (12c):(12) a. Han fik en hue på.he got a cap on‘He got a cap on’b. *Der kom en hue på ham.there came a cap on himc. Der kom en plet på ham.there came a stain on himCrucially, no proper locatives qualify as IOs, either, since (13a) can onlyrefer to the church as an institution, and a proper locative interpretionrequires (13b):(13) a. Han sendte kirken blomster.he sent the church flowersb. Han sendte blomster til kirken.he sent flowers to the churchIn Modern Danish, location is not part of the semantic roles of Frame andFramed (Whole and Part). 6, 7

124 Lars Heltoft3.2 The få-passiveCentral to the constructional view of the IO and of the semantic roles involvedis the existence of other grammatical constructions sensitive to the content ofthe IO. Danish has a specific passive which upgrades only the indirect object,and does not apply to objects in general, as does the ordinary passive. Characteristically,this passive has the verb få ‘get’ as its semi-auxiliary, as illustratedin (14). Even communication verbs occcur in the få-passive:(14) a. Forældrene forærede hende pengene.the parents presented her the money‘Her parents presented her with the money’b. Hun fik foræret pengene af forældrene.she got presented the money by the parents‘She was presented with the money by her parents’c. Man forklarede kongen sammenhængen.they explained the king the background‘They explained the background to the king’b. Kongen fik forklaret sammenhængen.the king got explained the background‘The king was given an explanation of the background’The få-passive identifies grammatically the role as FRAME. Only NPs qualifyingfor this interpretation can be the expression subject of få and hence of thisconstruction.3.3 Free IOSo-called free datives or adverbial datives occur as well, as examplified in(15a,b):(15) a. Han hentede os en øl i stedet.he fetched us a beer insteadb. Han købte sin kone en ny bluse.he bought his wife a new blouseSpace will not permit much detail. In short, such free IOs are not necessaryparts of the verbal semantic template. They are free adjuncts of the transitiverelation. This external status peeps through in their total lack of upgrading inany kind of passive. Whereas the genuine IOs in (2a, b) upgrade in bothpassives, (16), the adverbial IO does not, (17):

Datives and comitatives 125(16) a. Kongen fik tildelt en ubetydelig rolle.the king got allotted an insignificant roleb. Kongen blev tildelt en ubetydelig rolle.the king was allotted an insignificant role(17) a. *Hans kone fik købt en ny bluse.his wife got bought a new blouseb. *Hans kone blev købt en ny bluse.his wife was bought a new blouseWhat the free IOs add to our understanding of the IO, is the coding of theposition for IOs. Free IOs are not valency bound and do not acquire theirsemantic role from the verb, but directly from the position they end up in.The coding of the IO-position as the frame is still extant from the constantlyentailed få-clauses:(18) a. Han hentede os en øl i stedet.→ vi fik en øl i stedet‘We got a beer instead’b. Han købte sin kone en ny bluse.→ hans kone fik en ny bluse‘His wife got a new blouse’3.4 Reflexive pronouns as a test of content subjectsA more traditional feature and test, well known from traditional grammar (seefor instance Diderichsen 1946) and also pointed out by Herslund (1986), is theantecedent of the possessive reflexive pronoun sin/sit, pl. sine, which mustnormally refer to an expression subject antecedent (in the singular). Forcenturies, however, most varieties of Danish have allowed this reflexive to pickup, alternatively, the content subject latent in the indirect object. Example (19)shows expression subject antecedence (a), and content subject antecedence inthe IO, (b):(19) a. Hun fik sine penge tilbage.she got refl money back‘She got her money back’b. Vi gav hende sine penge tilbage.we gave her refl money back‘We gave her her money back’

126 Lars HeltoftPredictable ambiguities arise with singular expression subjects as in (20a). In(20b) there is no ambiguity, since this subject is in the plural:(20) a. Hun gav ham sine bukser.she gave him refl trousers‘She gave him his/her trousers’b. De gav ham sine bukser.they gave him refl trousers‘They gave him his trousers’3.5 Summing up the analysis of Danish IOThe IO in Danish, then, is a content subject, restricted in the same way as theexpression subject of få ‘get’. It is coded as the semantic role of Frame, thisrole being attached partly to Danish word order, namely to the IO-position,and to a Danish construction, the få-passive. Verbs that pattern with IOs areprototypically telic (transition) verbs.4. Co-subjectAs an expression unit, a CS is an adverbial phrase morphologically marked bythe preposition med. As a content unit, it is an argument, a Comitative, whichis never manifested itself by an expression subject, although prototypically, itshows the same content restrictions as the expression subject of the relevantclause (Hansen 1971).4.1 Symmetrical co-subjectSymmetrical CS constructions allow the NP of the CS to appear in coordinatingconstructions, see (21b), (22b). They are either valency bound orfree. As valency bound they are exemplified in (21), as free in (22):(21) a. A diskuterer med B.A discusses with B‘A is discussing with B’b. A og B diskuterer.A and B discuss‘A and B are discussing’

Datives and comitatives 127(22) a. Rist meldte sig (sammen) med et par kammerater somR. enlisted refl together with a couple of friends asfrivillige.volunteers‘Mr. Rist enlisted (together) with a couple of friends as volunteers’b. Hr. Rist og et par kammerater meldte sig som frivillige.‘Mr. Rist and a couple of friends enlisted as volunteers’4.2 The semantics of the co-subjectThe CS denotes a set of content subjects all of which qualify as expressionsubjects for V. All such content subjects are structured meronymically in thatthey are Parts. Neither the expression subject nor the CS are assigned the roleof Frame, this being instead presupposed as the total set of NPs identified bythe expression subject and the CS.A simple pseudo set theoretic exposition may help to make this pointstand out. The genuine IO construction is transitional and could be renderedas in Figure 2, as far as the movement interpretation is concerned:S has DODO passes to IOIO has DOSDOIO S→DO→IO S IO DOFigure 2. Transitional framingThe relation behind the CS is a static relation. The basic relation in thesymmetrical CS-construction is være med i ’participate in, be part of’; inasymmetrical CS-constructions (see below) it is the have ‘have’-relation. Forsymmetrical CSs the pseudo-set theoretical exposition is that of (23), for theasymmetrical version it is (24):(A) set of relevant content subjects(B) Content subject(s) ofexpression subjects(C) Content subject(s) of CSFigure 3. Framing in CS

128 Lars Heltoft(23) a. A=B+C(BandCarepartsofA,AhasBandC)(24) b. B=B+C(CispartofB,BhasC),(B=A)4.3 Asymmetrical co-subjectBy contrast, the asymmetrical CS presupposes static meronymy, a version ofthe framing relationship where the CS is framed by what is denoted by theexpression subject. In (25) we find meronymy proper:(25) a. Din cykel rasler ad helvede til med bagskærmen!your bike rattles like hell with rear guard-def‘Your bike’s rear guard rattles like hell’b. Hotellet dominerer det hele med sin grimme facade.the hotel dominates the whole with its ugly facade‘The hotel’s ugly facade dominates it all/everything’c. Byen når helt ned til havet med sine forstæder.the city extends all the way to the sea with its suburbs‘The city’s suburbs extend all the way to the sea’Asymmetrical CSs project the have-relation onto the adnominal genitive:(26) a. Din cykels bagskærm rasler.your bike’s rear guard rattles‘Your bike’s rear guard is rattling’b. Hotellets grimme facade dominerer det hele.the hotel’s ugly facade dominates the whole‘The hotel’s ugly facade dominates everything’In the asymmetrical CS have-relations are entailed, also by definition:(27) a. Din cykel har en bagskærm.your bike has a rear guardb. Hotellet har en facade.the hotel has a facadec. Byen har nogle forstæder.the city has some suburbs4.4 AbundanceThis variant is the counterpart of Fillmore’s example the garden is swarmingwith bees (1972). There is no reason, however, to analyse the subject of this

Datives and comitatives 129construction as a locative. At the level of subjecthood, this stands out as astraightforward CS-construction:(28) a. Gaden vrimlede med fattige børn.the street swarmed with poor children‘The street was swarming with poor children’b. Byen myldrer med polske gangstere.the city teems with Polish gangsters‘The city is teeming with Polish gangsters’c. Grøftekanten flagrede med sommerfugle.the roadside fluttered with butterflies‘The roadside was swarming with butterflies’The characteristic have-relation is always entailed and adnominal genitives aswell:(29) a. Gaden har fattige børn.‘The street has poor children’b. Byen har polske gangstere.‘The city has Polish gangsters’c. Grøftekanten har sommerfugle.‘The roadside has butterflies’(30) a. Gadens fattige børn.‘The street’s poor children’b. Byens polske gangstere.‘The city’s Polish gangsters’c. Grøftekantens sommerfugle.‘The roadside’s butterflies’4.5 Control causativesControl causatives (or activity causatives, cf. Heltoft 1978) are treated in Danishas a variant of the CS. Notice again that Danish direct objects cannot beagentive, therefore parallels to Halliday’s The sergeant marched the recruits orShe walked her little dog in St. James’s park pattern with the CS-construction:(31) a. Faderen læser lektier med sin søn.the father reads homework with his son‘The father is doing homework with his son’

130 Lars Heltoftb. Sergenten eksercerede med rekrutterne.the sergeant drilled with the recruits‘The sergeant drilled the recruits’c. Kan du ikke vente med dine indvendinger?can you not wait with your objections‘Can’t you put off your objections?’d. Nu kom politiet med den anholdte.now came the police with the prisoner‘Now the police brought (in) the prisoner’We shall not document these in detail, but have-relations and adnominalgenitives are possible everywhere, and entailed subject relations too, followingthe pattern:(32) a. Faderen læser lektier med sin søn.→ faderens søn læser lektier‘The father’s son is doing homework’Genuinely meronymic CS (inalienable possession) can occur in controlcausatives, provided that the expression subject has agentive content:(33) a. Den lille hund logrede med halen.the little dog wagged with its tail‘The little dog wagged its tail’b. Hvorfor hænger du med hovedet?why do you hang with your head?‘Why are you down in your mouth?’c. Klapperslangen rasler med halen før den hugger.the rattlesnake rattles with its tail before it strikes‘The rattlesnake rattles its tail before striking’4.6 InstrumentalsInstrumentals are CS as well, although a few extra restrictions make them lessprototypical. Have-relations are added: 8(34) a. Hug den rod over med en økse!chop that root over with an axe‘Chop that root in two with an axe’b. → du har en økse.‘You have an axe’

Datives and comitatives 131(35) a. Man kan som regel tage den slags pletter af med terpentin.one can, as a rule, take such stains off with terpentine‘One can — as a rule — remove such stains with terpentine’b. → man har terpentin‘One has terpentine’Sentences with instrumental subjects matching Fillmore’s the key opened thedoor (1968) cannot be plain declaratives in Danish (and probably not inEnglish, either), but must be modal generics, next to synonyms to sentencescontaining an explicit modal kan ‘can, will’:(36) a. Sådan en økse hugger roden over på et øjeblik.such an axe chops the root over in a moment’‘Such an axe chops the root in two in a moment’b. Denne nøgle åbner kælderdøren.this key opens the basement door‘This key opens the basement door’(36¹) a. Sådan en økse kan hugge roden over på et øjeblik.‘Such an axe will chop the root in two in a moment’b. Denne nøgle kan kun åbne kælderdøren.‘This key can only open the basement door’4.7 Reducing the semantic role inventoryThe direct profit from this constructional approach is that a number oftraditional or would-be semantic roles are reduced to variants of the CS.Sociative, genuine meronymic Part, Comitative, Controlled, Instrument, Co-Agents etc., are rendered superfluous as structurally constitutive notions. Theymay be relevant for contextually sensitive analyses of e.g. style or coherence,but they are not units at the level of la langue.5. Expression subjecthood and content subjecthoodas organizing principles in DanishSubjecthood is a central part of the analysis of both IOs and CS. The commondenominator for the IO and the CS in Danish is their function as contentsubjects. Both qualify as subjects of a predication entailed by the full construc-

132 Lars Heltofttion, both participate in the structuring of the semantic field we have calledabstract meronymy or framing. What distinguishes them from the outset istheir telicity. The IO is always telic (in the sense of denoting transition); theCS is always static. The entailed telic relation in the IO is prototypically thefå-relation, the presupposed static relation in the CS is the have-relation.It follows from this characterization that the distribution of the framingrelationship over the sentence members involved must differ. The IO istransitional: In prototypical IOs the DO changes its status from being framedby the expression subject to being part of a new frame, the IO. The CS,however, presupposes that S and CS are both parts of the same set or frame,and it forwards the claim that they are involved in the same activity.Danish makes this difference an important one, while at the same time(which we have not dealt with in detail here) delimiting IOs from locatives.Although we can formulate at an abstract level some degree of similarity(namely the framing relationship) between IOs and CS, such an abstractionwould in part take us outside the realm of the linguistic content of Danish. Atthe level of Hjelmslevian language specific content structure (content form, inHjelmslev’s terminology, see Harder 1996, Engberg-Pedersen et al. 1996:Preface), the point remains a question of difference. What is specificallyDanish is the sign relation, that is, the way the borderline pinned down at thelevel of content is related to two different syntactic constructions. Thisguarantees their linguistic relevance.Again, the framing relationship has a language specific design in Danishwhere it is defined as non-local. Other languages which lend themselves moreeasily to localistic analyses may include framing as special cases of localstructure. An overview of the distinctions is given in Figure 4 below.Insisting on expression criteria for any content distinction furnishes uswith a tool for putting the question of the number and quality of semanticroles on a proper empirical footing. Instruments, sociatives, comitatives andabundance PPs reduce to one and the same structure, the CS, free and valencybound. A tempting way to reintroduce the term Comitative in the analysis ofDanish, then, would be to define it as a cover term for the entire range of CSs.A Comitative in Danish is any CS viewed from the content, namely the staticco-part relationship defined in (23)–(24). Similarly, we can in this specificsense reintroduce Receptive for any IO in Danish, meaning thereby its statusas a Frame organized through the relevant system of content subjecthood.Since our approach is synchronic, the unifying feature of receptives andcomitatives has turned out to be a fairly abstract one. The relations discussed

Datives and comitatives 133Semantic substance ...Local relations ... ...Framing relations ...Local relations Framing relationsDanishcontentform(not on the agenda)IOAsserted framingAssertedtransitionCSPresupposedframingAsserted activityFigure 4. Content structure (Hjelmslevian content form) of the IO and CS in Danish.can be summarized as in Figure 4. Single vertical lines indicate borders ofDanish content form.By distinguishing IO and CS, Danish has split the field of framing into twomajor areas of content form. Telicity is one basic parameter, presuppositionvs. assertion another.All differences aside, my view of Modern Danish converges with Grimshaw’s(1990) on English in assuming firm structural relationships betweentelicity (‘aspect’, in her terminology) and semantic roles. On the other hand,such intermarriages are not universal. In well-known case languages we findexamples of grammaticalization across the borders of telicity.6. Datives and comitatives as spouses: Some case mergersIt is worthwhile to take a brief look at some case mergers in proper Indo-European case languages to see whether the substantial relatedness betweenIOs and CS may shed light on such mergers. Mergers that are immune tocatalysis (Hjelmslev 1943; namely those that cannot be resolved), should makesense at the level of content and not just remain expression phenomena.This exercise is not an attempt to revive the search for basic meanings(Grundbedeutungen) of the respective cases, neither diachronically nor cognitively.It is an attempt to find language specific organization behind mergers,and as such, it is a logical consequence of the strategy adopted in the analysis ofthe Danish IO and CS. Lists of variants such as those found in classical grammarsand in classical works on Indo-European syntax, esp. in Delbrück’s part ofBrugmann’s and Delbrück’s Grundriss will not answer this question.Heine, Claudi and Hünnemeyer (1991) point to diachronic relations —very frequent in African languages — between allatives and datives, and Blake(1995: 145) even regards a Target role as the basic meaning of datives. Datives

134 Lars Heltoftencode “entities that are the target of an activity or emotion’’, examples aredatives with Latin three-place verbs like dāre, monstrāre, mandāre, and withtwo-place verbs like parēre, fidere.As long as one restricts the scope to languages without instrumental andcomitative functions of the dative, the Target model may well be a sensibleinterpretation. But in languages like Classical Greek and an Old Germanicdialect like Old Scandinavian, the dative comprises instrumental andcomitative readings as well. In such languages we find mergers that cannot bereduced synchronically to the Target function. We shall take a closer look atthese languages below.In the approach of Heine, Claudi and Hünnemeyer (1991), meanings ofcase functions are related by means of conceptual networks, normally with adiachronic orientation. Quantitative generalizations over such networks leadthem to set up a chain of increasing grammaticalization, see below. I shall notdiscuss their points, but use their exposition as a bridgehead for my ownexposition. Of interest in this context is the presupposed, but not implausible,organization of case functions on a localist basis. Path aside, there are threebasic local functions and the networks that are formed on them as a template.As they say (1991: 159), ‘‘spatial concepts are more basic than other conceptsand therefore serve as a template for understanding [sic] nonspatial concepts’’.The local dimensions derive their increasing degree of grammaticalizationfrom another dimension, that of anthropocentricity:Spatial Anthropocentric InanimateAblative > Agent > Purpose > Time > Condition > MannerAllative Comitative Instrument CauseLocative Benefactive DativePathPossessiveNow, these hypotheses are really only very abstract and general guidelines, andempirical languages may pattern differently, both with respect to diachroniclines of development and with respect to synchronic patterning. As presented,this diagram contains no claims of interrelations along its horizontal rows, butonly with respect to its columns.Connections between these concepts may be lost in the development of anactual language — personally, I would find it wise to add that they may neverhave existed. From Schlesinger’s analysis of the polysemy of English with(1979) they render the continuum comitative > instrument > manner, but thepreposition with has no allative or ablative background, nor has the Danish

Datives and comitatives 135preposition med, for that matter. An actual network like Schlesinger’s(1979: 320):ComitativeInstrumentPossessiveFigure 5.Mannerneed not presuppose a localist analysis of the comitative-instrument-mannercontinuum. Such synchronically organized patterns may either include theirlocal origin, or they may have lost it during their historical development, oreven never have had one. 9In the light of these considerations, my own analysis is not just a nonlocalisticapproach, but also an attempt to formulate a common core meaningfor Modern Danish across the columns of case notions in Heine, Claudi andHünnemeyer. This core meaning was the relation of framing (abstract possessionor Zugehörigkeit). I have bracketed this notion to indicate that it is just anabstraction formed on the basis of the relevant constructions, in reality anotion that finds lexical support in the semantics of have ‘have’ and få ‘get’. Itis not directly grammaticalized as such in Danish, but must always combinewith options of telicity:[Framing]Dative (IO)Comitative (CS)Figure 6.Instrument (CS)I add this notion — involved case readings are in boldface — to the figure ondegrees of grammaticalization to show its non-localistic character, stressingagain that Danish have/få-relations and the Danish IO are not anthropocentric,but more abstract (cf. Herslund 1986). The framing relationship israther non-local. The brackets around the notion of possession indicate that

136 Lars Heltoftboth lexically and constructionally, the relevant notion of possession is theframing relationship:ablative agent purpose time condition mannerallative comitative instrument causelocative benefactive dativepath[possessive]What I shall propose next is a strategy relevant for the analysis of languagesthat merge with respect to comitative/sociative, instrumental and dativefunctions. It makes sense to ask whether such languages have more directgrammaticalizations of the framing relationship. It should be stressed thatwhat follows is the unfolding of such a strategy; full validity with respect to thestructures of the languages analysed is not claimed.6.1 Classical GreekThe main semantic variants of the dative merger are rendered in (37), part ofwhich were documented in (1). The remaining functions are listed in (38). Sincethe Greek system is not bound up with telicity, Greek readily admits Benefactivereadings, a semantic role excluded from the system of Modern Danish:(37) Causative (Agent/Cause)*Sociative (Comitative)InstrumentalPossessiveDativeExperient (‘Dativus judicantis’)BenefactiveReceptive*Locative(marginal)(marginal)(38) a. Pánta ’éire:to a’utê:i. Agenteverything had been said by she-dat‘Everything had been said by her’b. O’ikídion ’ésti moi diploûn. Possessivelittle house is I-dat two-storey‘I have a little two-storey house’

Datives and comitatives 137c. ’Allà gàr ’é:de: hó:ra ’apiénai, ’emoì mèn ’apothanouméno:ibut now is the time to leave, I-dat for my part die-ptc.dathymîn dè bio:soménois. Possessiveyou-dat for your part to live-ptc.dat‘But now is the time to leave, for me to die, but for you to live on.’d. Hopóteroi dè he:mô:n ’érchontai ’epì ’ámeinon prâgma,which-pl and us-gen go-pres.pl towards the better thing,’áde:lon pantì plè:n ’è: tô:i theô:i. Dativusconcealed everybody-dat but the-dat god-dat judicantis‘And which of us will go for the better alternative, is concealed toeverybody but the god.e. toîs tê:s póleo:s nómois ’epithóme:n. Receptivedef.d def.g. city-gen laws-gen obeyed-1‘I obeyed the laws of the city’f. Marathô:ni LocativeMarathon-dat‘at Marathon’g. Prosérchetaí moí tis presbûtis ’ánthropos. Locativecomes towards I-dat somebody old woman‘some old woman is coming towards me’The readings clustering around Possession are taken to constitute the corearea. I shall disregard the agentive use of the dative. Although this reading isstill extant in the classical language, its marginality is demonstrated by its nearabsence in New Testament Greek.Greek has an inflectional case system that discards the semantic differencebetween telic and static relationships. Thus, its dative is neutral with respect tothe distinctions grammaticalized in Danish. At the same time, its semanticsextends far beyond the limits of the Target reading: There is no obvioussimilarity between Targets and Comitatives/ Instruments, compare the factthat these latter readings go with the ablative in Latin. An obvious candidatefor the meaning of the Greek dative would be the Frame role. Target readingsor Allative-like readings such as (38g) may have been reinterpreted to Frames,but Instrumentals and Comitatives cannot be viewed as Targets.The issue of the status of the locative readings can be reformulated as thequestion whether Greek treats framing relationships as a special case oflocation. I shall only hint at the preconditions for carrying through such anextension of the topic of possession by emphasizing that there are at least two

138 Lars Heltoftconceivable answers to this. One strategy would be to view Classical Greek assynchronically locative-based; referring again to the networks employed inHeine, Claudi and Hünnemeyer (1991), the locative would then be interpretedas a focal member of such a network. Following the alternative strategy wouldmean taking the extant locatives to be just remnants of a former system thathas at the stage of Classical Greek almost changed into an abstract framingsystem. This issue is for experts of Greek, but one possible outcome would bethat Homeric Greek still reflects a locative based system, whereas later stages,especially Attic prose, do not.Luraghi (1991) accounts for the mergers of Greek in terms of combinationwith noun classes, much along the same lines as the distinction betweenanthropocentric case and inanimate cases quoted from Heine, Claudi andHünnemeyer. Plain inflectional datives with human NPs must be interpretedas the semantic roles of Receptive and Comitative, whereas non-human andinanimate NPs will require Instrument readings of a plain inflectional dative.To give a human NP the function as Instrument calls for an adposition; in thesame way, to give non-human NPs the Receptive function calls for anadposition. I quote two examples from Luraghi (1991:60–61), showing nonhumanReceptive and human Instrument readings:(39) Ho poimè:n ho kalòs tòn bíon parèchei hypèr tà próbata.the shepherd the good the life gives for the sheep‘The good shepherd gives his life for his sheep’(40) ’Autoì gár sphêis hoi Héllenes légontes di’ ’aggélô:n ho:s ...selves for they the Greeks saying through messengers that ...‘For the Greeks themselves said through messengers that ...’While this is a neat description of the functionality of a multipolysemousinflectional case system, there is no need to take the further step of distinguishingbetween two different dative paradigms, provided that a commoncore reading can be provided for the two sets, human and non-human datives.If one adopts something like the abstract frame reading of the dative, thedistribution of Instrument readings and Receptive readings over adpositionsand noun classes will reduce to a case of combinatorial variation, as inFigure 7.

Datives and comitatives 139ReceptiveInstrumentHuman Dative Adpos + NPNon-human Adpos + NP DativeFigure 7. Distribution of instrumentaland receptive/benefactive datives inGreek (Luraghi 1991).6.2 Old ScandinavianThe historical forerunner of the modern situation in Danish is a dative/instrumentalmerger similar to the one found in Greek. Judging from thestandard syntactic description of Old Norse (Nygaard 1906, in many respectsobsolete), Old Scandinavian retains relatively few instances of the agentive/causative use of the inflectional dative, and possession at the clausal level istypically not a dative construction, but a genitive one, or a transitive ‘have’-construction. The relevant functions are listed in (42)–(46):(41) SociativeInstrumentalReceptiveDativeExperientBenefactiveLocative(42) a. Vm man rithær annars mans hæsti. Instrumental/if a man rides another-gen man-gen horse-dat sociative‘If a man rides another man’s horse’b. Bøta threm marcum Instrumentalcompensate by means of three-dat marks-dat‘To pay three marks as compensation’(43) The mæn ...ær hanumfylg[i]a. Sociativethe men . . . that he-d accompany‘Those men ...that accompany him’(44) Æn vm man liggær á døtha digi. oc gifuær han guthi egn sina. Receptiveif a man lies on his dying day and gives he God-dat property his‘If a man lying on his dying day gives God his property’

140 Lars Heltoft(45) a. Æn mylnu dammi annars mans ma æy at fiscum fara. Locativebut millpond-dat other man’s may not for fish go‘But at another man’s millpond may one not go fishing’b. Han ma æy na hinu. ær læt var.he cannot touch that-dat which was lent out‘He cannot recapture that which was lent out’(46) Sæms them æy ær scoga ago. at swin gange samman Exper.agree they-d not that wood own. that swine may go together‘If those who own the wood cannot agree that their swine may gotogether’There is a remarkable similarity between this situation and the one displayedby Old Greek. They cover next to the same range of senses. We may again setaside the question of an originally local interpretation of the dative merger.Many historical details and facts beyond the scope of the present article pointtowards a diachronic extension of the locative case. The Old Scanic remnants— as they indeed are — of the locative use of the dative can either be includedin a synchronic continuum extending across the comitative-receptive border,or just be set aside as remnants.It is interesting to cast a glance at the development of the inflectional dativecase in later Middle Scanic. Middle Scanic of the fourteenth century has lost alllocative and instrumental/sociative uses, retaining only two main functions ofthe dative: the receptive function and an inagentive (‘experient’) reading occuringin so-called impersonal constructions. Later, the inflectional dative disappearscompletely (late fifteenth century). Examples of these two functions are:(47) The gamblæ skra thær ærchebiscup æskil oc biscup absalon . . . gafæskanungum.the old statutes that archbishop Æ. and bishop A. . . . gave scanians-dat(Ms. AM 37, 4 o , c.1300)(48) En man i sin ytarste tima. tha thøkte honum som en stor sten ofvirhonum hængde. oc vilde vpa honum falla.‘A man in his utmost hour. then thought he-A/D as if a big stone wouldfall upon him.’ (SjT 95, early 15th C.)Of these two readings, the inagentive one merges in a limited set of pronounswith the accusative to form a new oblique pronominal form, whereas separate

Datives and comitatives 141dative forms for nouns are optionally retained, disappearing altogether by thelate fifteenth century. The pronominal merger, illustrated in Figure 8, wasalready on its way in common Nordic, but extended into the later Scandinavianlanguages:Old ScanicLate Middle ScanicNom. hun Unmarked Nom. hun UnmarkedAcc.Dat.hana -Agentivehænni Receptive/ExperiencerObl.hænni -AgentiveFigure 8. Case reduction in Middle Scanic, exemplified by hun‘she’In fact, this merger brings us close to the Modern Danish situation, laterchanges of content are irrelevant in the present context. Case can no longerreflect the semantics of the IO and CS, neither as a merger nor as distinctsystems.Late Old Danish (represented here by late Middle Scanic) shows a conflictbetween two case systems, a nominative vs. oblique system and a direct vs.oblique system. In nouns, the receptive dative is in contrast to a nom/accmerger; the pronominal system has a nom. vs. oblique opposition, showing anacc/dat merger.Sg.Pl.Nom.Acc.Noun inflectionhæstkunungDat. hæsti kunungeNom.Acc.Dat.hæstascanungahæstum scanungumPronominal inflectionhanhanum/honumthethemhunhænni(hæst ‘horse’; kunung ‘king’; scanunga ‘Scanians’; han ‘he/it’;hun ‘she/it’; the ‘they’)Figure 9. Competing mergers in Late Old Scanic.The pronominal system is the only one to survive. From this point onwards,the IO, if expressed, must be so by means of position (word order), the Comitative/Instrumentmerger by means of adposition only, the CS construction.

142 Lars Heltoft6.3 The frame mergerThe dative case ending in Greek and in Old Scandinavian is a merger acrossthe borders of telicity. In Greek and Old Scandinavian we can claim thefollowing mergers — leaving the question of the locative open. For ClassicalGreek, Figure 10a, for Old Scanic, Figure 10b, the difference being that OldScandinavian normally uses a ‘have’-construction for possession in the narrowsense, not the dative.[Locative]PossessionReceptiveComitativeExperient(‘Dativus judicantis’)InstrumentFigure 10a. The frame merger in Classical Greek[Locative][Framing]ReceptiveComitativeExperientInstrumentFigure 10b. The frame merger in Old Scanic7. ConclusionIn the analysis of Modern Danish it was necessary to make a distinctionbetween two kinds of framing, receptive and comitative. These two kinds offraming are organized not as differences of semantic role, but as differentlyorganized relations of content subjecthood and telicity. A framing relationshiplies at the bottom of both constructions, but at the level of content form the

Datives and comitatives 143comitative function and the receptive function are complex notions, eachadding its own set of content subject rules to their common abstract framingrelationship. Modern Danish is a categorical subject language, and the loss ofoblique inflectional case in Danish has been compensated for by extensiveapplication of subject relations and subject rules to the realm of the originaldative. By contrast, former Indo-European stages like Old Greek and OldScandinavian are examples of direct coding of the role of frame onto theinflectional dative morpheme.Notes1. A nominative–accusative distinction is observed in a handful of personal pronouns, butthis distinction has been reinterpreted on the model of categorical sentence structure, sothat the nominative marks plain anaphorical subjects, the accusative all other cases,cataphorical subjects and non-subjects alike (Heltoft 1997).2. For necessary modifications of the Hjelmslevian apparatus, see esp. Harder (1996). Fora short statement, see Engberg Pedersen et al. (1996) (Preface); for an application, Heltoftand Falster Jakobsen (1996).3. The term ‘experient’ replaces experiencer in an attempt to harmonize the vocabulary forsemantic roles.4. Of course, we may need tentative, but proposedly general substance notions as landmarksfor our analysis of semantic role structure in specific languages. Nor do I rule out thepossibility that certain demands or restrictions on semantic roles may turn out to have nextto universal cognitive status.5. Manner adverbials may be formed as med-PPs, but these are not synchronically part ofthe CS-construction, since they do not comply with the subject criterion. Positionally, suchmanner adverbials are distinct from CS:(i) De var kørt med høj fart ind i et træ.they had run with high speed into a tree‘They had run into a tree at high speed’Some oblique objects are found with the preposition med. These fall short of the subjecttest:(ii)Han sad og fumlede med sin rosenkrans.‘He was fiddling with his rosary’6. The universalist version of the localist position, namely that all case content is universallymodeled on local structure, will not hold, nor will probably its diachronic counterpart,the claim that all case meanings are diachronically derived from local meanings.The position adopted here makes localism but one possible synchronic organization

144 Lars Heltoftprinciple, since framing is not locally based, in the sense of related intrinsically to concretelocatives. We need not for the present purpose make an issue out of the diachronic version.7. For a valence theory with localist implications, the so-called Adject Theory, se Herslund(1988), Herslund and Sørensen (1994). This theory deserves much more attention than ithas acquired so far.8. An additional authentic example like (i):(i)Se, med en lille vatpind tørrer du hendes mund med trøskesaft.‘Look, with a small swab stick you dry her mouth with glycerine of borax’→ (P)du har trøskesaftyou have glycerine of borax→ (P)du har en lille vatpindyou have a swab stickwill disprove the argument prevalent in American distributionalism that one never findstwo representations of the same category in one and the same clause.9. For another (brief) exposition of the confusion of synchrony and diachrony in localisttrends, see Heltoft (1996).ReferencesAnderson, J. M. 1971. The Grammar of Case. Towards a Localistic Theory. Cambridge:Cambridge University Press.Anderson, J. M. 1975. On Case Grammar. London: Croom Helm.Bech, G. 1955–57. Studien über das deutsche Verbum Infinitum I-II [Det kongelige danskeVidenskabernes Selskab. Historisk-Filologiske Meddelelser 35: 2, 36: 6]. Copenhagen:Ejnar Munksgaard.Benveniste, E. 1966 [1960]. “‘Être’ et ‘avoir’ dans leurs fonctions linguistiques’’. Reprintedin Problèmes de Linguistique Générale I, Emile Benveniste, 187–207. Paris: Gallimard.Blake, B. J. 1994. Case. Cambridge: Cambridge University Press.Blomqvist, J. and Jastrup, P. O. 1991. Grekisk/Græsk Grammatik. Copenhagen: AkademiskForlag.Comrie, B. 1981. Language Universals and Linguistic Typology. Oxford: Basil Blackwell.Diderichsen, P. 1946. Elementær dansk Grammatik. Copenhagen: Gyldendal.Dik, S. C. 1989. The Theory of Functional Grammar. Part 1. The Structure of the Clause.Dordrecht: Foris.Eisenberg, P. 1989. Grundriss der deutschen Grammatik. Stuttgart: Metzler.Eisenberg. P. 1992. “Syntaktische Funktionen und Semantische Rollen’’. Lecture at the 13thScandinavian Conference of Linguistics, Roskilde University, Department of Languagesand Culture, January 1992.Eisenberg. P. 1996. “Syntaktische Funktionen und Semantische Rollen. Subjekt, Direktesund Indirektes Objekt im Deutschen’’. In Proceedings of the 13th ScandinavianConference of Linguistics, L. Heltoft and H. Haberland (eds), 83–92. Department ofLanguages and Culture, Roskilde University.

Datives and comitatives 145Engberg-Pedersen, E. et al. (eds). 1996. Content, Expression and Structure. Studies in DanishFunctional Grammar. Amsterdam: John Benjamins.Engberg-Pedersen, E. 1996. “Iconicity and arbitrariness’’. In E. Engberg-Pedersen et al.(eds), 453–468.Faarlund, J. T. 1990. Syntactic Change. Toward a Theory of Historical Syntax [Trends inLinguistics. Studies and Monographs 50]. Berlin and New York: Mouton de Gruyter.Fillmore, C. 1966. “Toward a Modern Theory of Case’’. In Modern Studies in English, D.Reibel and S. A. Schane (eds), 361–375. Englewood Cliffs, N.J.: Prentice Hall.Fillmore, C. 1968. “The Case for Case’’. In Universals in Linguistic Theory, E. Bach and R.T. Harms (eds), 1–88. Cambridge, Mass.: The MIT Press.Fillmore, C. 1971. “Types of Lexical Information’’. In Semantics. An Interdisciplinary Readerin Philosophy, Linguistics and Psychology, D. D. Steinberg and L. A. Jakobovits (eds),370–392. Cambridge: Cambridge University Press.Fillmore, C. 1977. “The Case for Case Reopened’’. In Syntax and Semantics vol. 8. GrammaticalRelations, P. Cole and J. M. Sadock (eds), 59–81. New York: Academic Press.Grimshaw, J. 1990. Argument Structure. Cambridge, Mass.: The MIT Press.Goodwin, W. W. 1965[1894]. Greek Grammar. London: Macmillan.Gruber, J. S. 1967. “Look and See’’. Language 43: 936–947.Gruber, J. S. 1976 [1965]. Lexical Structures in Syntax and Semantics. Studies in LexicalRelations. Amsterdam: North-Holland.Hansen, E. 1971. “Jensen er nede i postkassen med et brev. Konstruktioner med consubjectumi moderne dansk’’. Danske Studier 1971: 5–36.Harder, P. 1996a. Functional Semantics. A Theory of Meaning, Structure and Tense inEnglish. Berlin and New York: Mouton de Gruyter.Harder, P. 1996b. “On Content and Expression in Syntax’’. In Proceedings of the 13thScandinavian Conference of Linguistics, L. Heltoft and H. Haberland (eds), 83–92.Department of Languages and Culture, Roskilde University.Heine, B., Claudi, U. and Hünnemeyer, F. 1991. Grammaticalization. A ConceptualFramework. Chicago: The University of Chicago Press.Heltoft, L. 1978. “Smæk ikke med døren når du smækker den. Static and DynamicCausatives in Danish’’. In Papers from the 4th Scandinavian Conference of Linguistics,K.Gregersen (ed.), 287–294. Odense: Odense University Press.Heltoft, L. 1996a. “Grammatikalisering af semantiske roller i dansk’’. In Semantiske roller[Odense Working Papers in Language and Communication 10], L. Schack Rasmussen(ed.), 43–64. Odense University.Heltoft, L. 1996b. “Paradigms, Word Order and Grammaticalization’’. In E. Engberg-Pedersen et al. (eds), 469–494.Heltoft, L. and Falster Jakobsen, L. 1996. “Danish Passives and Subject Positions as a MoodSystem’’. In E. Engberg-Pedersen et al. (eds), 199–234.Herslund, M. 1986. “The Double Object Construction in Danish’’. In Topics in ScandinavianSyntax, L. Hellan and K. Koch Christensen (eds), 125–147. Dordrecht: D. Reidel.Herslund, M. 1988. Le Datif en Français. Louvain-Paris: Éditions Peeters.Herslund, M. 1995. “Valens og grammatiske relationer’’. In Ny forskning i grammatik.Fællespublikation 2, P. Durst-Andersen and J. Nørgaard-Sørensen (eds), 48–72.Odense: Odense University Press.

146 Lars HeltoftHerslund, M. and Sørensen, F. 1994. “A Valence Based Theory of Grammatical Relations’’.In Function and Expression in Functional Grammar, E. Engberg-Pedersen, L. FalsterJakobsen and L. Schack Rasmussen (eds), 81–95. Berlin: Mouton de Gruyter.Hjelmslev, L. 1966 2 [1943]. Omkring Sprogteoriens Grundlæggelse. Copenhagen: AkademiskForlag. Translated by F. J. Whitfield (1963): Prolegomena to a Theory of Language.Madison: The University of Wisconsin Press.Høysgaard, J. P. 1752. Methodisk Forsøg til en fuldstændig dansk Syntax [= DanskeGrammatikere V, H. Bertelsen (ed.)]. Copenhagen: Reitzel.Jackendoff, R. M. 1972. Semantic Interpretation in Generative Grammar. Cambridge, Mass.:The MIT Press.Jackendoff, R. M. 1983. Semantics and Cognition. Cambridge, Mass.: The MIT Press.Jackendoff, R. M. 1987. “The Status of Thematic Relations in Linguistic Theory.’’ LinguisticInquiry 18: 369–411.Luraghi, S. 1991. “Paradigm size, possible syncretism, and the use of adpositions with casesin flective languages’’. In F. Plank (ed.), 57–74.Lyons, J. 1967. “A Note on Possessives, Existentials and Locatives’’. Foundations of Language3: 390–396.Lyons, J. 1977. Semantics 1–2. Cambridge: Cambridge University Press..Madvig, J. N. 1968 [1883]. Syntax der Griechischen Sprache. Repr. Copenhagen: Gyldendal.Nygaard, M. 1966 [1906]. Norrøn syntax. Oslo: Aschehoug.Plank, F. (ed.) 1991. Paradigms. The Economy of Inflection. Berlin and New York: Moutonde Gruyter.Sasse, H. J. 1987. “The Categorical vs. Thetic Distinction Revisited.’’ Linguistics 25: 511–567.Schlesinger, I. M. 1979. “Cognitive Structures and Semantic Deep Structures. The Case ofthe Instrumental’’. Journal of Linguistics 15: 307–24.Taylor, J. R. 1989. Linguistic Categorization. Prototypes in Linguistic Categorization. Oxford:Oxford University Press.

Chapter 8Towards a typology of French NP de NPstructures or how much possession is there incomplex noun phrases with de in French?Inge Bartning1. IntroductionThe purpose of this study is to present a typology of interpretations of thestructure NP de NP in French and to identify the possessive cases in the richpolysemy of relations between the two nouns conveyed by the sequence NPde NP (Bartning 1987, 1989, 1990, 1993, 1996, 1998). The preposition de hasbeen interpreted as a marker indicating a relationship which is inferred byone or both of the nouns (Bartning 1993; Cadiot 1989). The concern of thispaper is primarily attributive ‘possessive’ constructions. The term possessionis indeed elusive (see Herslund and Baron, this volume). In the following theterm possession has, above all, two senses. The broad sense is paraphrasablewith ‘avoir’ and ‘appartenir à’ and the narrow sense with ‘posséder’ (Seiler1983:90–92).In what follows a two-level model is used to present different interpretationsof the NP de NP construction. This model is subsequently supported bysyntactic tests. A short description of the use of the possessive determiner andsome of its constraints will be given in order to identify the cases of possessivemeaning in the French NP de NP system.2. A two-level model of NP de NP in FrenchBartning (1993) outlined three levels of interpretation of the NP de NPstructure: the discourse level, the pragmatic level and the prototypical level,exemplified here by (1a–c), (2) and (3) respectively:

148 Inge Bartning(1) a. Le chevalier du baccalauréat installa son camarade sur sa chaise(dossier Donald Duck) et, sans le quitter des yeux, […] il ouvrit lesplacards. (Orsenna, Grand amour, p. 31)‘The knight of/from the A-levels put his comrade on a chair and,without stopping to look at him, he opened the cupboards’b. Debout devant la cheminée était un homme de moyenne taille, à lamine haute et fière, [. . .]. De temps en temps, l’homme de la cheminéelevait les yeux de dessus les écritures. (Dumas, Mousquetaires,183–4, cité par Eriksson 1980: 399)‘Standing in front of the fireplace there was a man of medium sizewith a noble and proud expression. Now and then, the man of thefireplace/fireplace man raised his eyes above the writings’c. Et puis à le voir, lui, l’homme de la Mandchourie endormi ou mort.Celui de la main, celui du voyage. (Duras, L’Amant, p. 59)‘And then to see him, him, the man from Mandchourie sleeping ordead. The one of/with the hand, the one of/from the journey’(2) La rue de mon grand-père‘My grandfather’s street’(3) La fierté des Gaulois‘The pride of the Gauls’Discourse casesExamples (1a–c) show a type of binominal NP which is interpretable at a levelhigher than the phrase or sentence (hence the term discourse), and signalled bythe context, rather than by the complex noun phrase itself (the micro-structure).The sequence de NP2 serves as an identifier, N2 having the function of recallinga previous mentioned NP (Bartning 1993; Fraurud 1990). The text helps toestablish links between the two NPs. The ‘‘textual’’ distance (Kleiber et al. 1994)may vary a great deal: in (1a), le chevalier is the protagonist of the novel, at thebeginning of which he helped a girl he fell in love with pass her exams, and isthen, in a later chapter, ‘named’ le chevalier du baccalauréat. In (1b) la cheminéeis found only in three utterances before the NP structure. In (1c) the manreferred to is first identified by a part of his body and then by a journey duringwhich the man has put his hand on a girl’s knee (Corblin 1987: vague interpretation;Baron and Herslund 1997: topical relation; the ‘de notoire’, Englebert1992). In short, in (1b) one must already know that the man had stood in frontof the chimney earlier, or that, in (1c), the man in Duras’ novel had put his handon the girl’s knee, to be able to understand the discursive complex NPs.

Typology of French NP de NP structures 149Pragmatic casesAs is well-known (cf. Seiler 1983:40), cases such as (2) can be interpretedbased on extra-linguistic knowledge, (hence the term pragmatic) as in (4a–b):(4) a. Le chien de Sophie (= the dog which S. takes care of every day, as aveterinary surgeon)‘Sophie’s dog’b. Les secrets de famille sont comme ça. Ils vous tiennent comme lesparadrap du capitaine Haddock. Le plus agaçant, sans doute, c’estque tout le monde en a, tout le monde en fabrique, tout le monde ensouffre, mais que personne n’a encore trouvé le moyen de s’endébarasser (NO 1582: 8)‘Family secrets are like that. They hold you like Captain Haddock’ssticking plaster. The most annoying thing, without doubt, is thateverybody has them, that everybody invents them, everyone suffersfrom them, but nobody has yet found a way to get rid of them’The interpretation of the NP of (4b), which requires a good knowledge ofBelgian culture, is given by the enumeration of the ‘‘supporting’’ verbs, the lastof which recalls the true sense of the NP: ‘‘a sticking-plaster which is impossibleto get rid of’’ (cf. Hergé, L’affaire Tournesol).Prototypical casesFinally, the NP structure in (3) above shows an attributive relationship, witha semantic link which is interpretable on the basis of the characteristics of N1,which is a property noun. The sentence may be paraphrased as: ‘‘The Gaulsare proud’’ (Riegel 1985). The interpretation is predictable on the basis of theN1 which is an adjectival nominalisation and thus uni-relational as are verbalnominalisations and relational nouns (Bartning 1993 for French, Durieux1990 for English).There are, however, problems: if we insert a noun which is typically unirelationallike voyage (a predicative noun with one argument) in a minimalcontext such as (5):(5) Le voyage de Luc‘Luc’s journey’we can still obtain several possible interpretations:

150 Inge Bartning(a) the journey undertaken by Luc(b) the journey of which Luc dreams, talks, etc. (Heine 1997, ‘accidental’possession)Consequently, it seems necessary to generalize and say that, in principle, everyNP de NP may be interpreted at two different levels:1. The level of the NP itself (example 3) with a meaning imposed by thesemantic and argumental properties of the two nouns (cf. 5a). In otherwords, the relationship exists before discourse and is possible to establishon the basis of the lexical units involved. The semantic relationship is thusavailable at the level of the micro-structure of the NP.2. The discourse level, which then includesa. the contextual cases (examples 1a–c) interpreted according to knowledgegained from the preceding context without a preexisting semanticlink between the two nouns, as in the micro-level, andb. the pragmatic cases (examples 2, 4a–b, 5b) interpreted according toextra-linguistic or encyclopedic knowledge.If N2 is [+human], one can expect a large array of interpretations since mancan act in unlimited ways and act on objects in different ways. In such casesthe possibilities of pragmatic interpretations thus increase. In the discoursecases the nominal entities in N2 position may either occur earlier in the text(cf. 1a–c) or belong to the universe of global discourse or extra-linguisticknowledge (le voyage de Luc ‘‘the journey of which he dreams’’, etc.). In bothcases the interpretation requires a larger context than the NP itself. There isno preexisting semantic link between the two nouns as in the prototypicalcases.Grammaticalisation/Textual developmentIt seems to me that the contextual discursive cases may be considered cases ofgrammaticalisation by using the structure of NP de NP with the verypolyfunctional preposition de. Consider (1b) again: the man was first introducedas Debout devant la cheminée était un homme (‘the man stood in frontof the fireplace’), then at a potential intermediate stage > l’homme devant lacheminée (‘the man in front of the fireplace’) , and then finally > l’homme dela cheminée (‘the man of the fireplace’). This example can be considered anillustration of Heine’s proposal (1997:184) of the evolution of the SourceSchema in Romance languages which appears to have entailed a developmentfrom adverbial morphosyntax involving the latin preposition de (from) toa

Typology of French NP de NP structures 151nominal morphosyntax giving rise to such markers of attributive possession asFrench de, Spanish de and Italian di.There seems to be independent evidence for the two levels since they play arole in the acceptability of associative anaphora as Kleiber (1997) has shownwhen describing the semantic links in Part/Whole and locative relations whichcontain intrinsic links. As is shown in examples (6) and (7a), the prototypicalcases accept the associative anaphora but not the discursive one in (7b) whichhas no such intrinsic links (‘the vase’ in (7b) must have been introduced before):(6) Il s’abrita sous un vieux tilleul. Le tronc était tout craquelé. (part/whole)‘He took shelter under the old lime tree. The trunk was quite crackled’(7) a. Nous entrâmes dans un village. L’église était située sur une colline.(locative)‘We came into a village. The church was on a hill’b. Nous entrâmes dans une cuisine. ?Le vase était posé sur la table.(locative discourse case) (Kleiber 1997: 56)‘We came into a kitchen. The vase was on the table’3. The level of the micro-structure and the prototypical casesIn this section we will take a closer glance at the prototypical cases. These casesare distinguished by an analysis of the properties of N1 and N2 and bysyntactic tests of some NP de NP constructions.3.1 Presentation of a noun classificationWe will present three groups of nouns. Group I shows characteristics that helpto trigger the interpretation of the whole NP de NP construction, whereasgroups II and III, in contrast, contain multirelational nouns which do notconvey a given meaning (Bartning 1996). Moreover, nouns of the latter groupsare referentially autonomous, while those of the first group are oftensyncategorematic or referentially non-autonomous (Kleiber 1981, 1997,1999a). (For a classification of the NPs according to their cognitive ontologicalstatus, see Fraurud 1996.)Noun classificationGroup I contains nouns which determine the meaning of the whole NP de NP.The semantic relation between the two nouns “pre-exists’’ or exists as a

152 Inge Bartningpotential because of the lexical properties of N1 and the semantic link itcreates with N2 (cf. Pustejovsky 1993:86: the Qualia structure of nouns;Warren forthcoming; Kleiber 1999b).1. Predicative/verbal nouns which reproduce the same patterns of argumentstructure or valency as the verbs from which they are often derived(Herslund 1980; Baron 1992; Stage 1986; Gross 1991; Gaatone 1986; forNoun-adjective or Adjective-Noun constructions in French and English, seeBartning 1980; Levi 1978; Warren 1984). These are event and action nouns,which presuppose actors (verbal nominalisations with N2 as subject orobject):(8) Le retour des investisseurs étrangers fait flamber la Bourse de Paris.(InfoMatin 940720: 8)‘The foreign investors’ return’2. Attributive nouns which maintain an inherent relationship with the personor object possessing something (adjectival nominalisations, ‘N2 is Adjective’)(Guillaume 1919; Riegel 1985):(9) Enfin, il vit la gravité du regard que maître Biard posait sur lui (Engwall1984) ‘the graveness of the glance’3. Agentive nouns with NP2 as object:(10) La conductrice du camion‘the driver of the lorry’4. Psychological nouns (expressing a feeling or an attitude) which assume a‘‘patient’’, a ‘‘beneficiary’’ (Spang-Hanssen 1963; Anscombre 1995):(11) Car Bérénice avait le goût de l’absolu (Aragon, cité par Spang-Hanssen1963) ‘the taste of the absolute’5. Part/Whole nouns which assume that the N1 expresses an inherent part ofthe whole (Winston et al. 1987; Riegel 1994; Riegel et al. 1994: 240; Kleiber1997; Langacker 1995) ‘X is part of Y’:(12) la poignée de la porte vs *la poignée de la maison‘the handle of the door’ vs *‘the handle of the house’le volant de la voiture‘the steering wheel of the car’6. ‘‘Relational’’ nouns (of kinship and socio-professional status) referring

Typology of French NP de NP structures 153implicitly to another person (Cadiot 1991; Reboul 1993) or a social ranking(provided that N1 is a person):(13) a. l’épouse de Jean‘John’s wife’le frère de Marie‘Marie’s brother’b. Mais tout le monde n’a pas fait le même chemin que la fille de Lacan(NO 1583: 7)‘Lacan’s daughter’c. le président de l’entreprise, du club‘the president of the enterprise, of the club’As we shall see these six types of head nouns create together with the N2 thecore group of the prototypical cases since the N1 determines the meaning ofthe whole NP de NP.Group II. Multirelational/Non-predicative nouns but whose lexical meaningcontributes to the interpretation of the relationship, in particular N2.7. Locative nouns in N2 position: these NP de NP are paraphrased by ‘in Y,there is X’ (Fradin 1984; Kleiber 1997):(14) les bibelots du salon‘the trinkets of the living room’(15) un magasin de la ville‘a shop of the town’8. Temporal nouns in N2 position:(16) les programmes du samedi dernier‘last Saturday’s programs’son départ de la veille‘his departure of the evening before’Group III. Multirelational nouns.9. Iconic nouns (representational nouns):(17) a. l’image d’un père génial et trop connu‘the picture of a fantastic and too well-known father’

154 Inge Bartningb. c’est le portrait de son père‘the portrait of his father’Iconic nouns may offer other interpretations with the NP2: besides therepresentational interpretation there are two other possible interpretations ofN2 in (b), viz. agent (‘the father has made the portrait’) and possession (‘thefather owns the portrait’) and an example like le portrait du château also offersa locative meaning ‘the portrait is in the castle’, see example below.10. Open class: nouns expressing concrete objects (for various tests of thesenouns, see Kleiber 1981; Riegel 1985; Flaux 1992):(18) la voiture de Jean, les livres de Nicolas, les vêtements de Catherine‘John’s car, Nicolas’ books, Catherine’s clothes’For these nouns we may imagine all sorts of relations, e.g. possession, agentive,etc. as well as discourse interpretations (’’la voiture que Jean a dessinée’’, “thecar that J. drew’’) and even metonymy cases like ‘the picture of the car thatJohn drew’ (Warren, personal communication).3.2 Syntactic support for some established relationshipsBefore presenting the prototypical combinations of NP de NP I will show thatin French there are ways of differentiating binominal complex NPs, even thosewhere neither N1 nor N2 suggest a link of the NP (cf. group III above): In 1982Milner proposed two tests to demonstrate that there are different interpretationalgroups, i.e. pronominalization in de lui (which of course only concernsan animate, human N2) and the construction être de N2. Thus, the possessive:(19) la voiture de Marie‘Mary’s car’as well as the subjective and objective genitive:(20) a. la venue de Marie‘Mary’s arrival’b. l’éducation des enfants‘the education of the children/the children’s education’neither accept pronominalization with de lui nor the construction être de lui:(21) a. la voiture (est) *d’elle‘the car (is) of her’

Typology of French NP de NP structures 155b. la venue (est) *d’elle‘the arrival (is) of her’l’éducation (est) *d’eux‘the education (is) of them’,in contrast to the agentive type:(22) a. le livre (est) de Sartre‘the book (is) by Sartre/Sartre’s book’b. le livre (est) de lui‘the book (is) by him’and NPs with an iconic head noun:(23) a. la photo (est) de Marie‘the photo (is) of Mary’b. la photo (est) d’elle‘the photo (is) of her’Milner (1982) concluded that the possessive and the subjective and objectivegenitives represent a case in modern French, whereas the agentive type andthe ‘‘iconic’’ object are prepositional groups. Elsewhere we have been able toshow (Bartning 1990) that the system of NP de NP is somewhat morecomplex and that these criteria not only isolate the groups cited by Milnerbut may also be applied to other complex NPs containing de, viz. to attributivecases (=adjectival nominalisations):(24) *la gravité est de la situation‘the gravity is of the situation’and cases expressing ‘origin’:(25) ce thé est de Chine‘the tea is from China’This result places the ‘origin’ type of interpretation in the same group as theagentive relations, and the attributive type in the ‘possessive’ group. Eventhough Milner’s tests do not isolate different types of relations so that one canestablish different grammatical categories (Bartning 1990), they may significantlysupport the differentiation of the relations within the polysemy of NPde NP.

156 Inge Bartning3.3 The prototypical relations based on the previous noun classificationBasing the analysis on combinations of criteria of noun classifications presentedin 3.1 and the syntactic behaviour of the groups according to the testsin Section 3.2, we can now distinguish likely combinations of NP de NPprototypes, presented in Table 1.The prototypical cases have thus been identified and distinguished by themeaning imposed at the level of the micro-structure (group 1–2, 4, 6), by acombination between the meaning and the syntactic behaviour of the N2(group 3, 5, 6, 10, 12) and by the lexical meaning of the two nouns (N1: group7, 8, 9, 13b; N2: 11, 12, 13a).Table 1. The prototypical cases at micro-level‘‘Object’’ relationships (N2 = ‘‘object’’)1. N1: nominalization–N2: ±human, ±animatea. N1: deverbal noun expressing an event (“objective genitive’’)En finançant la publication de cette recherche sur l’origine de nos plus grands patrons(NO 1580: 10)(‘the publication of this research’)la libération des prisonniers (‘the liberation of the prisoners’) (Riegel et al.1994: 188)b. N1: agentive noun–N2: ±human, ±animatele conducteur du camion (‘the driver of the lorry’)le sauveur de Paul (‘the saviour of Paul, Paul’s saviour’)2. N1: psychological noun–N2: ±human, ±animate, ‘‘object’’/’’patient’’le goût du risque (‘the taste of/for risks’) (Petit Robert)la haine des immigrés (‘the hatred of immigrants’)3. N1: iconic noun–N2: ±human, ±animate (the iconic group)la photo, le portrait du château, du chat (‘the photo, portrait of the castle, the cat’)Quand elle a entendu dire: ‘‘Mon Dieu, c’est le portrait de son père’’, elle faillit sursauter(‘the portrait of her father’)(NO: 1582: 7)“Subjective’’ relationships (N2 = ‘‘subject’’)4. N1: nominalization–N2: ±human, ±animate (subjective genitive)Balladur réduit les dépenses de ses ministres (‘the ministers’ expenses’) (InfoMatin94.07.20, p.9)5. N1: noun expressing product, artefact, creation–N2 (agent): ±human, ±animate (‘N2makes/has made N1’, agentive relationship)Il jouait une sonate de Mozart (‘a sonata of/by Mozart’)(Le Clézio)le livre de Le Clézio (‘the book of Le Clézio/Le Clézio’s book’)

Typology of French NP de NP structures 1576. N1: property noun/adjectival nominalization–N2: ±human, ±animate (‘N2 is Adjective’,attributive relationship)la gravité du regard (‘the gravity of the glance’)la gentillesse de Paul (‘Paul’s friendliness’)7. N1: classifying noun–N2: +/- human ±animate (‘N2 is an N1’ or ‘N1 concerns N2’ )Le problème des immigrés préoccupe la plupart des gouvernements (‘the problem of theimmigrants’)[. . .] il y a aussi le thème de la culpabilité (‘the theme of guilt’) (NO 1582: 6)8. N1: relational noun–N2: 1. +human, 2. -animate (kinship relation)1. l’épouse de Jean (‘John’s wife’)2. le patron, le chef de l’entreprise (‘the head of the enterprise’)Other relationships9. N1: Part/Whole noun–N2: ±human, ±animate (‘N1 is a part of N2’, ‘N2 includes N1’,partitive relationship, inalienable possession)Tristan se souvenait des mains de sa mère (‘his mother’s hands’) jouant sur le piano noir(Le Clézio, 26)la poignée de la porte (‘the handle of the door’)la porte du restaurant (‘the door of the restaurant’)10. N1: multi-relational noun denoting an object–N2: ±human, ±animate, ‘‘subject’’ (‘N2has N1’, ‘N1 appartient à N2’, possessive relationship)a. posséder (‘possess’): N1-human, N2 ±humanle livre de Paul (‘Paul’s book’)b. avoir (‘have’)Sous prétexte de préserver les intérêts des épargnants, on confie les banques à deshauts fonctionnaires issus du Trésor (‘the savers’ interest’) (NO 1580: 10)11. N1: open category or locative noun–N2: locative noun (‘in Y, there is X’, Fradin 1984;Kleiber 1997; locative relationship)a. permanent location:Elle voyait M. Ferne sur la place du village (‘the village market place’) (Le Clézio)b. habitual location:(. . .) les gens fortunés des villas (‘the rich people in the villas’) (Le Clézio)c. temporary location:les gens du compartiment (‘the people in the compartment’)12. N1: open category — N2locative noun (dynamic relationship: ‘N1 comes from N2’,relationship of origin)le thé de Chine, le vin d’Alsace (‘the tea of/from China, the wine of/from Alsatia’)le livre de la bibliothèque (‘the library book’)13. a. N1: open category — N2: N2 locates N1 in time (temporal relationship)les programmes du samedi dernier (‘the programs of last Saturday’)Mon père était juif, confie-t-elle, sarcastique, et ma mère avait la tuberculose.Comme ça j’héritais des deux tares majeures du XX e siècle (‘the two major flaws ofthe XXth century’) (NO 1582: 8)

158 Inge Bartningb. N1: noun locating the event in time — N2 event nounla veille de l’inhumation (‘the evening before the burial’) (Ernaux, La place 19)C’est le moment de la révolte […] (‘the moment of the rebellion’) (NO 1580: 5)4. A typology of the prototypical cases and their relationsin the polysemy of NP de NPOne might claim that among the prototypical cases presented in Table 1 thereare certain relations that are interconnected (in particular groups 5, 6, 9, 10,11 and 12 in the list above). Some of these relations are more linked topossession than others. They are all expressed by the same syntactic structurebut they do not all express possession.I propose that there are two basic types of relations, at micro-level,namely possession and origin, which subsumes many other relations. Thisdivision is supported by the syntactic behaviour of the groups and correspondbetter to the original meaning of de. (Table 2 below does not cover allNP de NP structures, e.g. destination cases like le train de Paris, quantity unkilo de beurre, une heure de lit, location la région des parfums, nor quality unfripon de valet.) The relations given under A in table 2 have conceptualaffinities and syntactic behaviour resembling possession. Under B we findrelations connected to that of origin. (For the conceptual affinities, seeBartning 1993.)Table 2. Distribution under two groups of essential prototypical relationsA. Attributive Part/whole Possession Locationla gentillesse de Jean le bras de Jean la voiture de Jean les bibelots du salonB. (Source) Origin Agent (Iconic)(l’idée de Jean) le thé de Chine le livre de Sartre (la photo de Jean)The meaning component which group A has in common may be summarizedby the stative verb ‘‘avoir’’ (see e.g. Milner 1982:138; Godard 1986; Riegel1985:63 and Baron and Herslund, this volume, for the analogy betweenlocative and possessive phrases), that of group B by the dynamic verbs ‘‘venirde’’, ‘‘faire’’, both expressing ‘‘provenience’’ (cf. the original meaning of de inOld French, Herslund 1980; Heine 1997: 145: ‘Source Schema’: Y from X).This division is reflected by the possibility/impossibility of the tests ofMilner (1982), namely être de N2:

Typology of French NP de NP structures 159(26) Group A: *Le bras est de Jean, *la voiture est de Jean (possessive), *lagentillesse est de Jean, *les bibelots sont du salonhere in the locative sense; the predicative phrase is acceptable in the‘origin’ sense(27) Group B: le thé est de Chine, le livre est de la bibliothèque, l’idée est deJean, le terme est de moi, le livre est de SartreBecause of their syntactic behaviour the first group may also include thesubjective genitives (l’arrivée de Jean ‘John’s arrival’) and the objective genitive(la connaissance du dossier ‘the knowledge of the dossier’), as well as (verylikely) the NPs with the head noun expressing feelings and the N2 ‘object’ orpatient (la haine des immigrés ‘the hatred of the immigrants’) and the NPs withclassifying nouns (le problème des immigrés ‘the problem of/with the immigrants’).These groups are of course very different semantically. (For a discussionof the possibility to use subjective and agentive complex NPs in predicativestructures, see Bartning 1990).In contrast, the discourse cases such as l’homme de la cheminée, la fille dubac (‘the man of the fireplace, the girl of/from the ferry boat’) are not part ofthis system. They are constructions interpreted at a textual level which borrowthe general meaning of the NP1 de NP2 construction paraphrased as ‘N1 isassociated with N2’. As we saw above, they are decoded more precisely withthe help of the context or extra-linguistic knowledge.5. Is there any other syntactic support for the groups A and Bamong the prototypical cases?Let’s look at the tests of de lui again and then the possessive determiner to seeto what extent they support the hypothesis of the two groups A and B in orderto isolate the possessive meaning from other prototypical cases and thediscourse cases.5.1 De luiAs we saw, the de lui test distinguished the possessive cases and the objective/subjectivecases from the agentive and iconic cases. We can now extendthis distinction by stating that pronominalization in de lui is possible only forgroup B. In group A, it cannot be used for possessive or Part/Whole relationshipsor for the attributive group:

160 Inge Bartning(28) a. le livre de lui (agent)‘the book of him’b. la photo de lui (iconique)‘the photo of him’c. une gifle de toi m’humilierait (Gross 1991: 268) (agent)‘a slap of you would humiliate me’d. un signe de toi me ferait plaisir (id.) (agent)‘a sign from you would please me’e. Nakasone, lui aussi, est peintre. Dans son bureau de Tokyo, il amême une toile de moi. (COSTO: X04, PE, p.2) (agent)‘a painting of me’versus:(29) a. la voiture *de lui (possessive)‘the car of him’b. le bras *de lui (P/W)‘the arm of him’c. la gentillesse *de lui (attributive)‘the friendliness of him’5.2 The possessive determinerIf the result of the de lui test is unambiguous between the groups A and B, i.e.among the prototypical cases, that of the possessive determiner son, sa, ses, isless so (Godard 1986): the ‘subjective’ NP de NP are those which are mosteasily pronominalized by the possessive determiner, viz. possessive, agentive,subjective and attributive relationships (Bartning 1989, 1993). Thus, thepossessive determiner is not exclusively limited to the NPs within group A, butis also used within the domain of group B as shown in (30):(30) a. sa gentillesse (attr), son bras (p/w), sa voiture (poss), ses bibelots(loc)b. son idée (source), son thé (origin), son livre (agent), sa photo(iconic))However, there are other prototypical combinations in Table 1 that do notaccept the possessive determiner: the ‘‘objective’’ groups and the locative andtemporal groups. The ‘objective’ groups are exemplified in le respect desconventions/*?leur respect (‘the respect of the conventions/their respect’), la

Typology of French NP de NP structures 161haine des immigrés/*leur haine (‘the hatred of the immigrants/their hatred’)(object) and le goût de l’absolu/*son goût (‘the taste of the absolute/its taste’)(Godard 1986; Bartning 1989). The locative and temporal groups on theother hand include permanent location: les bibelots du salon/ses bibelots ‘thetrinkets of the living room/ its trinkets’ which seems to accept the possessivedeterminer, whereas neither ‘‘habitual’’ location (les gens des villas/*leurs gens‘the people of the villas/ their people’) nor ‘‘temporary’’ location (les gens ducompartiment/*ses gens ‘the people of the compartment/its people’) do. Butagain the conceptual affinity between location and the Part/Whole relationshiphas the effect that the possessive determiner is often accepted, which isthe case of l’eau des rivières/leur eau (‘the water of the rivers’) and les taresdu XX e siècle/ses tares (‘the flaws of the twentieth century’). Finally there arestill other relationships which do not accept the possessive determiner,namely classifying NPs as in le problème des immigrés/*leur problème, laquestion des immigrés/*leur question (‘their problem’, ‘their question’, cf.Bartning 1989: 181) and identifying NPs as la ville de Paris/*sa ville (‘itstown’). In the latter case the pronominalisation does not apply to a nonreferentialNP.5.3 The possessive determiner and the discourse casesThe discourse cases such as l’homme de la cheminée/*son homme (‘the man ofthe fireplace/its man’), le chevalier du baccalauréat/*son chevalier (‘the knightof the A-level/ its man’), cannot be pronominalized by the possessive determinerdespite the fact that many prototypical locative cases can be, as in lemagasin de la ville /son magasin (‘the shop of the town/its shop’). This syntacticbehaviour seems to support our hypothesis that these NP de NP constructionshave another origin or derivation than the prototypical cases (cf. la mainde l’homme/sa main ‘the hand of the man, the man’s hand/his hand’ vs.l’homme de la main/*son homme ‘the man of the hand/its man’).This is also confirmed by Kleibers’ data of associative anaphora (1997: 58):(31) Le pique-nique était raté. ?Sa bière était trop chaude (‘The picnic was afailure. Its beer was too warm’) (discourse case)(32) Un village était situé sur une butte. Son église dominait toutes lesmaisons. (‘A village was situated on a hill. Its church dominated all thehouses’) (prototypical locative case)Another interesting feature of the possessive determiner is that besides the

162 Inge Bartningprototypical cases which allow pronominalization by the possessive determiner,there are discourse (pragmatic) cases which all accept the possessivedeterminer (Bartning 1989; Gross 1986; Schapira 1997). The feature +habitual/topicalseems to be involved in these interpretations:(33) a. ton Platini‘your Platini’ (Gross 1986)b. Voyez Mme Barzach, la sainte patronne des médecins, leurprotectrice. Ce n’est pas elle qui critiquera ses chers médecins‘her dear doctors’ (NO1168: 9)c. Il nous a montré son cinéma‘his cinema’These discourse cases (‘habitual’) do not accept de lui even if the N2 is human:(34) le café de Paul/*de lui‘the café of him’(35) la rue de grand-père/*de lui‘the street of him’One can thus conclude that son and de lui are not equivalent. Son is acceptedby the possessive and habitual/topical cases, whereas de lui functions withagent and origine. The pragmatic interpretation son N1 is thus more ‘possessive’than de lui.To sum up, the possessive determiner does not distinguish between groupsA and B although there are restrictions within the groups. As far as thediscourse cases are concerned, the data are more complicated, since there isalso a ‘‘discursive’’ possessive determiner with a meaning which varies accordingto extra-linguistic knowledge (ton Platini).However, the possessive determiner is not applicable to the real discoursecases of NP de NP constructions connected to the text by previous mention,such as l’homme de la cheminée. So in this case the possessive determiner servesas one of the criteria that support the two level model.We can conclude that the feature [+animate] is important but not necessaryin pronominalization using a possessive determiner. In other words thepossessive determiner prefers human and animate N2 in prototypical relationsbut it also accepts inanimate N2 in locative and P/W relations. Additionally, inpragmatic/discourse cases, it applies to human N2 and the feature [+habitual]but not in the contextual/discourse cases. There remains, however, much to bedone in the study of the syntax and semantics of the possessive determiner in

Typology of French NP de NP structures 163French. (For a new approach of analysis of the possessive determiner, seeKleiber To appear)6. Where is ‘possession’ to be found in the NP de NP system?As we have seen, the distinction of the two levels, the discourse level and theNP level facilitates the identification of ‘possession’ in the NP de NP system.Possession is thus to be found at the micro-level as the kernel meaning of theA group and it is paraphrasable with ‘N1 avoir N2’ or ‘N2 appartenir à N1’.The B group cases do not express possession but origin and agentive relationshipsparaphrasable with ‘être de NP2’ (cf. 26 and 27 above).This micro-level is opposed to the macro-level which includes the pragmaticand discourse cases as can be seen in Table 3: the meaning of theadnominal NP2 and even the possessive determiner at the pragmatic discourselevel is not pure possession but that of ‘habitualness’ or ‘topicality’. As furtherevidence for the two levels, it has been shown that the possessive determiner isnot possible in the purely contextual cases nor are the paraphrases with ‘avoir’ou ‘appartenir à’:Table 3. Possession in the NP de NP-system1. Prototypical 2. Discursive/Pragmatic 3. Discursive/Contextual(habitual/topical)le chapeau de Paul le café de Paul l’homme de la cheminéeson chapeau son café *son hommepossession NOT possession NOT possessionMore precisely, the possessive determiner is possible in group A, as is shownin Table 4 below, in the attributive, Part/Whole and Locative cases, and is stillpossible in the interpretations of the group B: origin and agentive, among theprototypical cases (cf. 30 above).At the pragmatic level the possessive determiner has habitual or topicalmeaning but cannot be applied to the pure contextual cases like l’homme de lacheminée/*son homme. The structures être de lui and de lui are accepted by the‘origin’ and ‘agent’ cases in group B but not by the two discursive cases, as wealso can see from Table 4.

164 Inge BartningTable 4. The distribution of the possessive determinerPrototypical (micro level) Possessive determiner (être) de luiGroup A: OK *Group B: OK OKDiscursive (macro level)Pragma: OK *Context: * *7. Concluding remarksNaturally, this study does not claim to have identified the typology of interpretingNP de NP. Our model, however, shows that it is necessary to distinguishbetween two different levels, the discourse level and the complex NPlevel, the micro-structure, in order to grasp much of the work that the littleword de can perform between two NPs. Thus, it is suggested that the NP deNP of the micro-structure level is decoded by the meaning imposed by thesemantic and argumental properties of N1 and sometimes by those of N2. Wehave called them the prototypical cases of the construction, because throughsemantic resemblance they are connected to two essential relationships, that of‘possession’ and that of ‘origin’, and the number of relationships they areinvolved in is limited. Certain syntactic tests have confirmed the distinction ofthe two levels.The interpretation of the discursive NP de NP construction, on the otherhand, is accomplished at the textual level in a wide sense. These syntagms areonly understood by taking into account interpretations which depend onprevious discourse and the textual interval and/or extra-linguistic knowledge.AcknowledgementsI would like to thank Andrée Borillo, Mats Forsgren, Georges Kleiber andBeatrice Warren for their valuable comments on earlier drafts of this chapter.This is a modified and expanded version in English of Bartning (1996). Theexamples are taken from Engwall (1984), from the COSTO corpus (Engwalland Bartning 1989), French newspapers and novels, e.g. Duras, Le Clézio,Ernaux. It would be interesting to include other cases in an enlarged integratedmodel of NP structures containing de, e.g. quantitative cases (une bande

Typology of French NP de NP structures 165de voyous), qualitative cases (fripon de valet), compounds, cases of destinationand characterization (la table du piquet, l’homme de la discipline).ReferencesAnscombre, J.-C. 1995. “Morphologie et représentation événementielle: le cas des noms desentiment et d’attitude’’. Langue française 105: 40–54.Baron, I. 1992. “Les syntagmes nominaux complexes dans les textes juridiques francais’’.Hermes, Journal of linguistics 9: 19–42.Baron, I. and Herslund, M. 1997. “The Danish verb have and the notion of possession’’. InPossessive Structures in Danish [KLIMT 3], I. Baron and M. Herslund (eds), 131–157.Copenhagen Business School.Bartning, I. 1980. Remarques sur la syntaxe et la sémantique des pseudo-adjectifs. Stockholm:AWE International.Bartning, I. 1987. “L’interprétation des syntagmes binominaux en de en françaiscontemporain’’. Cahiers de grammaire 12: 1–64.Bartning, I. 1989. “Le déterminant possessif et les compléments adnominaux en de’’. Revueromane 24: 163–203.Bartning, I. 1990. “Les syntagmes binominaux en de–les types interprétatifs subjectifs etagentifs’’. Actes du X e congrès des romanistes scandinaves, Lund, 10–14 août 1987. Etudesromanes de Lund 45: 20–34.Bartning, I. 1993. “La préposition de et les interprétations possibles des syntagmesnominaux complexes. Essai d’approche cognitive’’. Lexique 11: 163–191.Bartning, I. 1996. “Eléments pour une typologie des SN complexes en de en français’’.Langue française 109: 29–43.Bartning, I. 1998. “Modèle intégré des syntagmes binominaux en de–typologie d’interprétationset reprise anaphorique’’. Romanische Forschungen 110: 165–184.Cadiot, P. 1989. “La préposition : interprétation par codage et interprétation parinférence’’. Cahiers de grammaire 14: 25–50.Cadiot, P. 1991. La préposition POUR. De la grammaire à la cognition. Paris: ÉditionsCNRS.Corblin, F. 1987. Indéfini, défini et démonstratif. Geneva, Paris : Droz.Duras, M. 1985. L’Amant. Paris: Minuit. (= L’Amant I)Duras, M. 1991. L’Amant de la Chine du Nord. Paris: Gallimard. (= L’Amant II)Durieux, F. 1990. “The meanings of the specifying genitive in English. A cognitive analysis’’.Antwerp Papers in Linguistics 66.Englebert, A. 1992. Le « petit mot » DE. Etude de sémantique historique. Geneva-Paris: Droz.Engwall, G. 1984. Vocabulaire du roman (1962–1968). Dictionnaire des fréquences. Datalinguistica 17. Stockholm: AWE International.Engwall, G. and Bartning, I. 1989. “Le COSTO — description d’un corpus journalistique’’.Moderna språk LXXXII: 343–348.Eriksson, O. 1980. “L’étiquette (fixée) sur la valise — l’étiquette de la valise — l’étiquette devalise’’. Studia Neophilologica 52: 389–414.

166 Inge BartningFlaux, N. 1992. “Les syntagmes nominaux du type Le fils d’un paysan : référence définie ouindéfinie?’’ Le Français moderne LX: 23–45.Fradin, B. 1984. “Anaphorisation et stéréotypes nominaux’’. Lingua 64: 325–369.Fraurud, K. 1990. “Definiteness and the Procession of Noun Phrases in Natural Discourse’’.Journal of Semantics 7: 395–433.Fraurud, K. 1996. “Cognitive ontology and NP form’’. In Reference and Referent Accessibility,T. Fretheim and J. K. Gundel (eds), 65–87. Amsterdam: Benjamins.Gaatone, D. 1986. “Problèmes de possessivation en français’’. Actes du 17 e congrès internationalde linguistique et philologie romanes, vol. 4: 225–231.Godard, D. 1986. “Les déterminants possessifs et les compléments de nom’’. Languefrançaise 72: 102–122.Gross, G. 1986. “Syntaxe du déterminant possessif’’. In Détermination : syntaxe etsémantique, J. David and G. Kleiber (eds), 89–111. Paris: Klincksieck.Gross, G. 1991. “Syntaxe du complément du nom’’. Lingvisticae Investigationes XV:255–284.Guillaume, G. 1919. Le problème de l’article et sa solution dans la langue française. Paris:Hachette.Heine, B. 1997. Possession. Cognitive Sources, Forces and Grammaticalization. Cambridge:Cambridge University Press.Herslund, M. 1980. Problèmes de syntaxe de l’ancien français. Compléments datifs et génitifs[Etudes romanes de l’Université de Copenhague 21]. Copenhagen: Akademisk Forlag.Kleiber, G. 1981. Problèmes de référence : descriptions définies et noms propres. Paris:Klincksieck.Kleiber, G. 1997. “Des anaphores associatives méronymiques aux anaphores associativeslocatives’’. Verbum XIX: 26–66.Kleiber, G. 1999a. “Anaphore associative et relation partie/tout: condition d’aliénation etprincipe de congruence ontologique’’. Langue française 122: 70–100.Kleiber, G. 1999b. Problèmes de sémantique. La polysémie en questions. Lille: Pressesuniversitaires du Septentrion.Kleiber, G. To appear. ‘‘Le possessif via l’anaphore associative’’. Paper presented at theinternational congress ‘‘From NP to DP’’, Antwerp, 10–12 February 2000.Kleiber, G., Schnedecker, C. and Ujma, I. 1994. “L’anaphore associative, d’une conceptionà l’autre’’. In L’anaphore associative. (Aspects linguistiques, psycholinguistiques etautomatiques) [Recherches Linguistiques de Metz, XIX], C. Schnedecker, M. Charolles,G. Kleiber and J. David (eds), 5–64. Paris: Klincksieck.Langacker, R. W. 1995. “Possession and possessive constructions’’. In Language and theCognitive Construal of the World [Trends in Linguistics 82], J. R. Taylor and R. E.MacLaury (eds), 51–79. Berlin-New York: Mouton de Gruyter.Levi, J. 1978. The Syntax and Semantics of Complex Nominals. New York, San Fransisco,London: Academic Press.Milner, J.-C. 1982. Ordres et raisons de langue. Paris: Seuil.Pustejovsky, J. 1993. “Type coercion and lexical selection’’. In Semantics and the Lexicon, J.Pustejovsky (ed.), 73–94. Dordrecht: Kluwer.Reboul, A. 1993. “Le poids des pères, le choc des fils : prédicats de phase, modificateurs etidentification’’. Cahiers de linguistique française 14: 229–46.

Typology of French NP de NP structures 167Riegel, M. 1985. L’adjectif attribut. Paris: PUF.Riegel, M. 1994. “Article défini, anaphore intra-propositionnelles et relation Partie/tout’’.In L’Anaphore associative, C. Schnedecker, M. Charolles, G. Kleiber and J. David (eds),233–250. Paris: Klincksieck.Riegel, M., Pellat, J.-C. and Rioul, R. 1994. Grammaire méthodique du français. Paris : PUF.Schapira, C. 1997. “Quelques emplois non possessifs de l’adjectif possessif: valeursaspectuelles et valeurs modales’’. In Actos do XIX Congreso Internacional de Lingüísticae Filoloxía Románicas [Tome 1. Sección 1. Lingüística Teórica e LingüísticaSincrónica], R. Lorenzo (ed.), 185–196. A Coruña: Fundación Conde de Fenosa.Seiler, H. 1983. Possession as an Operational Dimension of Language. Tübingen: GunterNarr.Stage, L. 1986. “Franske substantivers valens’’. CEBAL 8: 204–29.Spang Hanssen, E. 1963. Les prépositions incolores du français moderne. Copenhagen: Gad.Warren, B. 1984. Classifying adjectives [Gothenburg Studies in English 56]. Gothenburg:AWE International.Warren, B. Forthcom. “The role of links and/or Qualia in modifier head constructions’’.Winston, M. E., Chaffin, R., and Herrmann, D. 1987. “A Taxonomy of Part-Wholerelations’’. Cognitive Science 11: 417–444.

Chapter 9Spanish N de N structuresfrom a cognitive perspectiveHenrik Høeg Müller1. IntroductionPossession is a category that manifests itself linguistically in a number of ways,e.g. through specific verbs, pronouns, prepositions and other morphemes. Inthis paper I will confine myself to commenting on the Spanish complex NP ofthe type las aves del bosque ‘the birds of the forest’, i.e. the N de N structure.If we search in the Spanish grammatical literature for a description of themeaning of the preposition de, it is very common to run into the word“possession’’ in one of the first lines of this description (see e.g. Gili Gaya1993:251, Escandell Vidal 1995:39, Lagunilla and Rebollo 1995: 245). It isoften claimed that de, at least to a certain extent, or in some non-specifiedmanner, should be associated with the fundamental, extralinguistic category ofpossession. However, several authors moderate this possessional approach bystating that of course many applications of de cannot be said to expresspossession in a direct or overt sense, but nevertheless maybe in some sort ofabstract or indirect manner.→It is as if there is a tendency towards attaching maximum importance or akind of first priority to the domain of possession, maybe because of its status asa universal concept in our society, in the sense that all languages seem to haveconventionalized ways of encoding possession linguistically (Heine 1997: 1).This almost automatic linkage of the preposition de with the notion of possessioncould probably also be due to the assumption that de-constructions veryfrequently involve some aspects of possession. To my knowledge, however,there is no statistical evidence for such an allegation, and another problem inthis connection is the lack of a watertight definition of possession.If we look at the examples listed in (1), it would be somewhat odd, and inmy opinion intuitively wrong, to maintain that they express even the vaguestform of possession.

170 Henrik Høeg Müller(1) a. La casa de la colina Locationthe house of the hill‘The house of(on) the hill’b. La fiesta de ayer Timethe party of yesterday‘Yesterday’s party’c. Los jamones de Extremadura Originthe hams of Extremadura‘The hams of (from) Extremadura’d. El ruido del coche Causethe noise of.the car‘The noise of the car’e. Los periodistas de El País Working relationthe journalists of El País‘The journalists of El País’f. Los televisores de Philips Producerthe TV-sets of Philips‘The TV-sets of Philips’g. La dieta del equipo de fútbol The diet that they followthe diet of.the team of football‘The diet of the football team’h. La carta de María Senderthe letter of María‘The letter from María’i. Las ramas del árbol Part–wholethe branches of.the tree‘The branches of the tree’j. El padre de Juan Kinshipthe father of Juan‘The father of Juan’k. La llegada de Pedro Subjective genitivethe arrival of Pedro‘The arrival of Pedro’l. La destrucción de la ciudad Objective genitivethe destruction of the city‘The destruction of the city’

Spanish N de N structures 1712. PrototypicalityWhen we are confronted with the obvious problem that N de N structuressometimes seem to encode a relation whose interpretation deviates substantiallyfrom that of possession, the word “prototypicality’’ pops up from time totime.As a means of explaining these deviating constructions and still stick topossession as an overall explanatory frame, the notion of prototypicality comesin handy.Prototypicality is often reproduced as a core surrounded by an unspecifiednumber of concentric circles that symbolise either more or less prototypicalproperties of the notion in question or more or less prototypical members ofa certain category. The application of such a model to possession could beillustrated as in Figure 1, the core-meaning of possession being something likepermanent and physical control of an object, and the concentric circlesindicating different degrees of possession. The more distant an instance of thecategory is from the core the fewer possessive features it contains and the lesspossessive the features are, and vice versa.Permanent and physicalcontrol of objectFigure 1.It is not that I basically disagree with this way of representing meaning orsuggesting knowledge organization, but because of its fuzziness it provokessome reflections, especially regarding the theoretical consequences of such anapproach.First of all we might ask the question: How many concentric circles arethere? Or to put it in another way: How far can we move away from the coreand still claim an affinity to possession? When coming across a certain N de Nstructure which has intuitively very little to do with possession, do we justdraw another circle stating that this hypothetical example is merely an extremelyabstract instance of possession?Another, maybe a little speculative, concern of mine comprises the way weorganize prototypical meanings in larger networks in our brains, and how westore these meanings. Assuming that every prototypical meaning forms part ofa neural network, which is of course not entirely uncontroversial as we do not

172 Henrik Høeg Müllerknow that much about how we actually organise, store and retrieve knowledge,could it then be that the concentric circles from different core-meaningsoverlap each other, as shown in Figure 2.Figure 2.This would imply that the more an instance of a category moves away fromone core the closer it gets to another or to several others. This again makes onethink how far an example can diverge from the original core before it entersinto the scope of another and actually becomes part of the new core’s circles orrange of meaning. Or maybe we could imagine a more chaotic situation wherethe circles of many prototypical meanings meet, as is illustrated in Figure 3.Figure 3.Of course it would be too ambitious and difficult to try to answer all thesequestions in the present article, but the point is that the notion ofprototypicality is not very clearly defined (see, however, Geeraerts 1988 for athorough discussion of prototypicality) and more importantly it entails quitea number of problems.3. Rejection of possession as a prototypical meaning ofN de N structuresThe idea of the preposition de as giving rise to an ideal meaning which ispresent to a certain degree in all its uses and aspects must be rejected. As wesaw before de is used to establish a wide range of different semantic relationships,and only some of them can be interpreted as expressing possession inthe ordinary sense of the word. Hence explaining the Spanish N de N structurethrough an alleged basic relationship e.g. to possession, or location for thatmatter, or any other prototypical meaning one might claim could be invoked

Spanish N de N structures 173by de, seems futile. We simply cannot attribute a well-defined set of semanticrelations to the formal syntactic features of the de-construction, because thereis no systematic correlation between the syntactic and the semantic levels. Theextralinguistic category of possession should not be directly associated withspecial linguistic elements or certain morphological patterns, as they are likelyto express other concepts than for example possession.What should be emphasized is, in summary, and from a more generalperspective, that there is no one-to-one relation between linguistic facts andreal-world phenomena.But with the rejection of prototypicality and possession as being of centralimportance to the basic interpretation of the N de N structure, it would ofcourse be natural to ask what kind of relation can be rendered by de and howsuch a structure should then be explained.4. The preposition de as a cognitive primitiveAs regards the analysis of the function of the preposition de inside an NP,I suggest that we reverse the angle or turn the viewpoint upside down, so tospeak. Instead of regarding the multiple semantic relations invoked by de asdifferent forms or levels of abstraction from a core meaning, I propose thatit is precisely the abstractness or the non-specification that should constitutethe basic explanatory point of departure. This statement calls for someexplanation.As the associations encoded by de can be so multifaceted semantically, wemust conclude that no overt specification of the relation between the twonouns is imposed by de. The preposition functions as a predicative elementthat establishes a semantically highly indeterminate relation between twolinguistic representations of entities.In line with other basic cognitive image schemas such as “container’’,“path’’, “causal schemas’’, “abstract goals’’ and so forth, de is a lexical item thatperforms a basic cognitive operation, i.e. combining or relating entities to eachother. In fact de functions as a cognitive primitive, and therefore de-constructionscan be qualified as belonging to a pre-possessional category whichrepresents a much more basic cognitive phenomenon than just the concept ofpossession. The sole task undertaken by de is to relate two objects to eachother in a given way that is determined by other factors such as contextualproperties, conceptual import of head nouns and world-knowledge.

174 Henrik Høeg MüllerRegarding N de N constructions as a linguistic encoding of a very basiccognitive ability, is further corroborated by the fact that children use possessiveconstructions early and often (Barker 1995:45). The clitic -s is for exampleamong the first fully productive morphemes acquired by English speakingchildren in early multiword utterances (Villiers and Villiers 1985:68, quoted inBarker loc. cit.). It is, however, in my opinion very unlikely that children, at anearly stage, use and control a linguistic construction that basically expresses anintangible relation, i.e. possession. In this light for example Howe (1976: 29,quoted in Barker loc. cit.) and Slobin (1985:1179, quoted in Barker ibid.)suggest that the utterance that we might be tempted to interpret as possessiveactually expresses a more general relation, and it is explicitly put forward thatthere is a more general semantic notion that encompasses both possession andlocative relationships, which is actually in good keeping with my ideas.5. Analysis of the N de N structureThe analysis of the relationship that holds between the two nouns in an N deN construction will be the subject of the rest of this article.As a first step towards a description of the relationship in question it isrelevant to establish a general dichotic division between NPs where N2 isreferential, and NPs where this is not the case. The examples in (2) illustratethis difference:(2) a. El carnicero de Madridthe butcher of Madrid‘The butcher of (from) Madrid’El traductor del librothe translator of.the book‘The translator of the book’El cuadro de Miróthe picture of Miró‘The picture of Miró’b. El paté de hígadothe pâté of liver‘The liver pâté’La sortija de matrimoniothe ring of wedding‘The wedding ring’

Spanish N de N structures 175La granada de manothe grenade of hand‘The hand grenade’In the a-examples the second nouns denote an entity whereas in the b-examplesthey are non-referential, in the sense that they do not identify an entitybut they refer to a concept. This fundamental difference between the two typesof constructions can be depicted as shown in Figure 4.a.RefN1DERefN2→ ENTITYRENTITYb.RefN1DERefN2→ ENTITYFigure 4.The difference between them is that the type illustrated in Figure 4a denotestwo entities and the other in Figure 4b only one. This means that the prepositionin the case of (4a) instructs the recipient of the communication to relateN1 to N2 in such a way that they maintain their status as individual entities,whereas the instruction made by the preposition in Figure 4b is to relate N1,not to another entity, but to a type of entities which denotes an abstractconcept. N1 is related to N2 in such a way that N2 classifies or describes N1,and in such a way that they together (the whole construction) only denote onesingle entity. On this occasion I shall only have something to say about thefirst type, i.e. Figure 4a.The next step is to find out the nature of the relation that holds betweenthe two nouns. That is, what can be predicted about such a construction basedon the meaning of its parts. How does the meaning of this relation dependupon the meanings of its constituents.It has almost become trivial to say that the N de N construction is massivelyvague, but regardless of this apparent vagueness, there are strongconstraints on what the construction can mean. All combinations of nouns arecertainly not possible.5.1 The restriction/non-restriction scaleThe descriptive contents of the N de N construction depends primarily on thedenotation of N1. The conceptual knowledge about the domain of the headnoun puts constraints on, or say determines, the final interpretation of the

176 Henrik Høeg Müllerwhole construction. In this connection it is important to emphasize that I donot take contextual features into consideration.According to the nature of the head noun we may place the whole constructionon a sort of restriction/non-restriction scale, where one end ischaracterized by constructions open to very few interpretational possibilities,because of the constraints imposed by the head noun, and characteristic of theother end is that the head imposes very few interpretational constraints on theconstruction. This scale can be illustrated as shown in Figure 5.Conceptual importof head nounNon-relational nounsRelational nounsPart/wholeNouns denoting objects (bodyparts)Nouns denoting arbitrary partsNouns denoting persons (kinship)Deverbal nounsDeadjectival nounsFigure 5.InherentDerivedThe scale in Figure 5is divided into two main parts according to therelational properties of the nouns in question. The bottom part consists ofnouns that are either inherently relational (for a thorough description of theserelational nouns see e.g. Herslund 1997) or are morphologically related topredicates. 1 The upper part contains nouns that do not refer to externalentities, and consequently they are non-relational.It also appears from Figure 5that the head of a part–whole constructioncan be both relational and non-relational. In this way the part–whole structureconstitutes a sort of intermediate form where nouns that are conventionallyrecognized as being non-relational actually behave as if they were relational.This could mean that the sharp dichotic division between relational and nonrelationalis in reality blurred or maybe even non-existent, a point that I shallreturn to later on.5.2 Deadjectival nounsStarting from the bottom with the deadjectival nouns, it can be stated that in

Spanish N de N structures 177general such derivations are very reluctant to combine with any other nounthan the one referring to the entity that is described by the correspondingoriginal adjective.The predicative force of the adjective is transferred to the derived noun,and the noun creates a frame with one or two participants, dependent on thenature of the original adjective, but only one of the participants can beintroduced by the preposition de.If the nominalized adjective is derived from a stage-level predicate(Carlson 1979:53ff.), it can occur with de-phrases expressing time and place.The examples in (3) show some different deadjectival nouns combined withtheir corresponding participants, (3a,b), and with prepositional phrasesexpressing time and place respectively, (3c, d):(3) a. La inteligencia de Juanthe intelligence of Juan‘The intelligence of Juan’b. La velocidad del trenthe velocity of.the train‘The velocity of the train’c. La oportunidad del mes pasadothe opportunity of.the month last‘The opportunity of last month’d. Las dificultades de la callethe difficulties of the street‘The difficulties of the street’The ability of nominalized adjectives to combine with prepositional phrasesintroduced by de can be modelled as shown in Figure 6.SPACETIMEDEADJECTIVAL NOUNX, (Y)Figure 6.

178 Henrik Høeg MüllerThe two ‘x, (y)’ indicate the possible participant roles, of which only one, asmentioned before, can be introduced by de. The dotted lines indicate that timeand space can be expressed in certain cases, according to the nature of thepredicate.5.3 Deverbal nounsThe next step is the deverbal nouns. Deverbal nouns can be described asnominal reflections or versions of the state of affairs described by the originalbase verb and the corresponding participant roles of the verb. Thenominalized verb inherits the argument structure of the verb, and consequentlythe nominalization implies the same type and number of participantroles as the verb (see e.g. Baron 1994, Stage 1994 and Müller 1998). As is thecase with the verb, the nominalization establishes a frame of participant roles,which are interrelated in a given way, determined by the lexical make-up ofthe nominalization. This implies on the syntactic level that every nounintroduced by the preposition de must be interpreted according to the frameestablished by the deverbal head noun, i.e. syntactically they will function aseither subject or object. The examples in (4) contain deverbal head nounsexpanded with different types of participants:(4) a. La descripción del paisajethe description of.the landscape‘The description of the landscape’b. La entrega de las mercancíasthe delivery of the goods‘The delivery of the goods’c. La llegada de Juanthe arrival of Juan‘The arrival of Juan’As opposed to the deadjectival nouns, however, all deverbal nouns can alwaysbe anchored temporally. This emplies that temporal prepositional phrasesintroduced by de can occur with deverbal nouns. In Figure 7 the predicativepotential of deverbal nouns is reproduced.A general difference between deverbal nouns and other nouns that denoteprocesses, events, states or actions on one hand, and all other nouns on theother, is that the members of the first group denote abstract entities, while themembers of the other denote concrete entities.

Spanish N de N structures 179SPACETIMEDEVERBAL NOUNX, Y, (Z)Figure 7.Typical of concrete entities is that they can be located in space by aprepositional phrase, whereas their placement in time usually requires aregular predication in the form of a relative clause for example.A concrete entity combined with a prepositional phrase expressing timeforms a construction that is open to an infinite number of context dependentpossibilities of interpretation, as in (5a). (5a) could mean “the car we sawyesterday’’, “the car that broke down yesterday’’, etc. But when a concreteentity forms an NP with a place adverbial, only one interpretation is possible,i.e. the concrete entity is located in the place denoted by the de-phrase, as in(5b) which means “the car that is in the garage’’.However, when an abstract entity as head of an NP is modified by ade-phrase expressing time, the interpretational possibilities are also narroweddown to one, which can be paraphrased to something like “to take place’’. 2The combination “abstract noun + locative prepositional phrase’’ normallyrequires that N2 is introduced by a preposition different from de. These finalobservations are illustrated in (5cd):(5) a. El coche de ayerthe car of yesterday‘Yesterday’s car’b. El coche del garajethe car of.the garage‘The car in the garage’c. La entrega de ayerthe delivery of yesterday‘Yesterday’s delivery’

180 Henrik Høeg Müllerd. La llegada en/*de la estaciónthe arrival at/*of the station‘The arrival at/*of the station’We can say that the different nature of the head nouns makes them more orless inclined to combine with prepositional phrases expressing either timeor place.5.4 Inherently relational nounsIt is a well-known fact that some nouns, such as kinship terms and nounsdenoting body parts etc. have an argument structure without being morphologicallyrelated to predicates. These relational nouns refer to external entities,in that they imply a necessary existence of something or somebody to whomor which the noun in question can be related. The concept of mother forinstance necessarily entails as part of its lexical denotation an offspring. Inreality, the definition of this type of relational nouns does not differ substantiallyfrom that of nouns related to predicates.This type of relational nouns can be divided into three groups accordingto their denotational characteristics. In the first group, (6), the head nounsdenote quantities and arbitrary parts, in the second, (7), they denote persons,and in the third, (8), physical objects:(6) a. La parte de Juanthe part of Juan‘The part of Juan/Juan’s part’b. El porcentaje del bancothe percentage of.the bank‘The percentage of the bank’c. La mayoría de PSOEthe majority of PSOE‘The majority of PSOE/PSOE’s majority’d. Los centilitros de la probetathe centilitres of the test tube‘The centilitres in the test tube’(7) a. Las madres de la Plaza de Mayothe mothers of the Square of May‘The mothers of the May Square’

Spanish N de N structures 181b. Las madres solteras del mercado laboralthe mothers single of.the market labour‘The single mothers of the labour market’c. El padre del clan Kennedythe father of.the clan Kennedy‘The father of the Kennedy clan’d. La madre de la pedagogathe mother of the pedagogue‘The mother of the pedagogue’(8) a. La nariz de la chicathe nose of the girl‘The nose of the girl’b. Las narices del departamento patológicothe noses of.the department patological‘The noses of the patological department’c. La cabeza de la estatuathe head of the statue‘The head of the statue’d. Las cabezas del museothe heads of.the museum‘The heads of the museum’Looking at the examples in (6), we can observe that apart from their relationalinterpretation the de-phrases can impose another, non-relational meaning onthe whole construction. 3(6a) and (6b) could be interpreted as that part or percentage of somethingto which Juan and el banco are entitled, i.e. possession in some way, (6c) asPSOE’s majority of the voters, and (6d) must be interpreted locationally as thecentilitres (of a liquid) that are in the test tube.Passing on to the group where the head nouns denote persons i.e. (7), wenotice the same duplicity as was the case in the first group. The relationalnouns denoting persons can occur with argumental, prepositional expansions,and, as indicated in the examples, with non-argumental ones.In (7c), el padre del clan Kennedy ’the father of the Kennedy clan’, the headnoun should not be interpreted as standing in a father relation to the entitydenoted by the second noun. The head noun padre ‘father’ does not act as arelational noun, but as an independent, non-relational noun, because theperson in question (Joseph Kennedy) is not the father of all the members of

182 Henrik Høeg Müllerthe clan. In addition to the kinship reading (7d), la madre de la pedagoga ‘themother of the pedagogue’, could be interpreted in the right context as forinstance the mother with whom the pedagogue of the kindergarten has aspecial contact, and consequently not her own mother. However, such areading will undoubtedly seem marginal to most people, although it is mostcertainly possible.It is as if we are more likely to accept relational nouns in the plural as headsof non-relational constructions. Plural common nouns like mothers denote setswhose elements are identical in the sense that the members of the set all sharea mother-relation to an offspring. The creation of a set on the basis of individualrole expressions like mother enables us to deemphasize the relationalaspect of the single members of the set, and instead focus on how the set acts asan integrated individual entity in a context. We could say that the set establishesits own identity which is not necessarily linked to the relational characteristicsof the individual elements. We are simply more inclined to perceiverelational nouns in the plural as isolated groups of things or persons who canact independently of their relational properties. Consequently relational nounsin the singular generally have a stronger tendency to bind themselves to theirexternal arguments than is the case when they occur in the plural.Part–whole structuresAs indicated in Figure 5both the relational nouns denoting physical objectsand the relational nouns denoting quantities and arbitrary parts refer to theconstitutive parts of an entity and thus they presuppose the whole of whichthey form part. But whereas the nouns of the second group only receive ameaning in function of the entity to which they belong, the nouns of the firstgroup have their own specific, individual meaning. It is most certainly possibleto identify an arm or a leg independently of the body they are or were part of,but it does not make any sense to try to point out a part or a centilitre ofsomething without knowing what that something is.The examples in (8) indicate that the so-called prototypical members ofthe group of relational nouns denoting physical objects, i.e. body parts, showthe same pattern as the constructions in (6) and (7). Either they are interpretedin accordance with their relational content as in (8ac), or they act asindependent “normal’’ non-relational nouns as in (8bd).Returning to Figure 5again we can observe that part–whole constructionscan occur with a head which is conventionally considered to be non-relational.This is of course not controversial in any way, as we are used to structures like

Spanish N de N structures 183(9a,c) where the non-relational nouns ventana ‘window’ and techo ‘roof’constitute the heads of part–whole structures:(9) a. La ventana de la casathe window of the house‘The window of the house’b. La ventana del vidrierothe window of.the glazier‘The window of the glazier’c. El techo del cochethe roof of.the car‘The roof of the car’d. Los techos del almacén 4the roofs of.the stock‘The roofs of the stock’The reason why we normally think of arms and legs as being relational is thatthey imply the existence of a whole (a body) and they function as arms andlegs in relation to that body. This situation is exactly analogous to what is seenin (9ac). These two examples are part–whole constructions in which the headnouns ventana ‘window’ and techo ‘roof’ imply the existence of entities thatcan function as wholes, and the nouns can be described as entities thatfunction as a window and a roof in relation to la casa ‘the house’ and el coche‘the car’. Moreover the same head nouns in (9bd) cannot qualify as relationalin those particular contexts, because they do not function as a window or asroofs in relation to el vidriero ‘the glazier’ or el almacén ‘the stock’. 5 In otherwords we can distinguish between a functional interpretation according towhich ventana ‘window’ and techo/techos ‘roof/roofs’ act as relational nounsand a non-functional interpretation where this is not the case. So the examplesin (9) are actually totally parallel to the ones in (8). The conclusion of theseobservations must be that if a noun, no matter whether it has been preclassifiedas relational or not, forms part of a partial de-construction as head itfunctions as a relational noun in that particular context.It can also be deduced from Figure 5that not all nouns of the upper partof the scale can form part of a part–whole construction. Nouns like e.g. coche‘car’ and árbol ‘tree’, which constitute complete, independent entities, cannotbe head nouns of partial de-constructions.Another important aspect in this discussion, which shall just be mentionedbriefly here, is the interrelation between partial and locative constructions. All

184 Henrik Høeg Müllerpart–whole structures are in fact locative because an entity cannot under anycircumstances constitute a part of another entity without being positioned inrelation to that entity. However, it must be emphasized that in contrast to thepartial constructions not all de-constructions with a locative content arerelational, as example (10) indicates:(10) Los árboles del campothe trees of.the field‘The trees of(in) the field’It cannot be claimed that the head noun árboles ‘trees’ functions as trees inrelation to el campo ‘the field’, and consequently the NP in question shouldnot be qualified as a relational construction. The trees are simply placed in thefield, and hence (10) is a purely locative construction without relationalcontents.The essence of the entire discussion is that a noun is not necessarily eitherrelational or non-relational, but the relational property of a given head noun,if it is present in its lexical make-up at all, can be stimulated or triggered bythe other noun of the constellation and the context as such. That some nounsshow a stronger tendency than others towards going into relational structureshas to do with the fact that in our mental representation of the world we aremore inclined to see them either as forming part of kinship relations or asbeing a part of a whole and not as independent entities.I believe that maintaining a sharp division between relational and nonrelationalnouns gives us a distorted picture of reality. Such a division is toorigid to describe a much more complex world.6. Final remarksTo summarize, it has been my intention to illustrate four central points withthis article:1. N de N constructions should not be defined with point of departure in theconcept of prototypicality.2. Possession does not constitute an overall meaning frame within which allN de N constructions can be classified.3. The preposition de should be regarded as a cognitive primitive, whosemere function is to combine entities in a given way.

Spanish N de N structures 1854. The interpretation of the N de N structure depends on the restrictive ornon-restrictive nature of N1.In general I have hoped to show with this article that the N de N-structure isa much more complex semantic field than just possession.Notes1. Normally only the nouns with an inherent argument structure are referred to asrelational, but I also consider the derived nouns to be relational as they too have the facultyof selecting external arguments.2. In the case of some nominalizations it is difficult to establish, independently of thecontext, whether N2 has argumental status or not. The example la descripción de ayer ‘thedescription of yesterday’ can both be interpreted as “the description that took placeyesterday’’ (non-argumental reading) and as “the description of yesterday’’ (argumentalreading). If in such cases the non-argumental reading is intended, it would probably bemore natural to express the time relation in an alternative way syntactically.3. (6d) has no relational reading but only a locative one.4. The intended meaning of this construction is “the roofs that some company has onstock’’ and not “the roofs of a certain building that functions as the stock of a company’’.5. In this connection of course I ignore the metonymical reading, where el vidriero ‘theglazier’ for example could mean the glazier’s house.ReferencesBarker, C. 1995. Possessive Descriptions. Leland Stanford Junior University: Center for theStudy of Language and Information.Baron, I. 1994. “Les syntagmes nominaux français dans une perspective valentielle’’.Mélanges Lucien Tesnière [Linguistica XXXIV, 1], 29–45.Carlson, G. N. 1979. “Generics and Atemporal When’’. Linguistics and Philosophy 3: 49–98.Escandell Vidal, M. V. 1995. Los complementos del nombre. Madrid: Arco Libros.Geeraerts, D. 1988. “Where does prototypicality come from?’’ In Topics in CognitiveLinguistics, B. Rudzka-Ostyn (ed.), 207–229. Amsterdam: John Benjamins.Gili Gaya, S. 1993. Curso Superior de Sintaxis Española. Barcelona: Bibliograf.Heine, B. 1997. Possession. Cognitive sources, forces, and grammaticalization. Cambridge:Cambridge University Press.Herslund, M. 1997. “Partitivity and inalienable possession’’. In Possessive Structures inDanish [KLIMT 3], I. Baron and M. Herslund (eds), 1–44. Copenhagen BusinessSchool.

186 Henrik Høeg MüllerHowe, C. J. 1976. “The meaning of two-word utterances in the speech of young children’’.Journal of Child Language 3: 29–47.Lagunilla, M. F. and Rebollo, A. A. 1995. Sintaxis y Cognición. Introducción al conocimiento,el procesamiento y los déficits sintácticos. Madrid: Editorial Sintesis.Müller, H. H. 1998. “Substantivsyntagmer i spansk. En valensanalyse’’. Copenhagen WorkingPapers in LSP 4: 1–179.Slobin, D. I. 1985. “Crosslinguistic Evidence for the Language-making Capacity’’. In TheCrosslinguistic Study of Language Acquisition, Volume 2: Theoretical Issues, D. Slobin(ed.), 1157–1256. Hillsdale, New Jersey: Lawrence Erlbaum Associates.Stage, L. 1994. “La valence des noms en français’’. In Noun Phrase Structures [CopenhagenStudies in Language 17], 93–133.Villiers, J. G. de and de Villiers, P. A.. 1985. “The Acquisition of English’’. In TheCrosslinguistic Study of Language Acquisition, Volume 1: The Data, D. Slobin (ed.),27–139. Hillsdale, New Jersey: Lawrence Erlbaum Associates.

Chapter 10The grammatical category “Possession’’ andthe part–whole relation in FrenchMartin Riegel1. Possession and possession(s)In French there is no overlap between the fields of application of the nominaland the adjectival derivatives possession and possessif as metalinguistic labels:possession (in-)aliénable, rapport de possession, mot/adjectif/pronom/syntagmepossessif et construction/relation/structure possessive, and their everyday use.Indeed, even if there are hesitations and disagreements about its extension, thetraditional grammatical notion of “possession’’ goes largely beyond theconjunction of the three main meanings associated with the word possession incommon speech: (1) the ownership of goods (la possession d’une voiture),(2) the owned thing, especially estates and territories (les possessions d’unprince), (3), the control over one’s behaviour and faculties (être en possessionde toutes ses facultés).The same goes for the five meanings of the verbal root posséder, the threemore specific 1 of which are not enclosed within the grammatical possessionfield, but the two main meanings — (a) to have at one’s disposal a thingpossessed as its rightful owner, (b) to have an inner quality, an advantage, etc.,concerning an animate being — only cover some of the relationships commonlyviewed by the grammarians as being of a possessive nature. Theevidence is to be seen in the almost unnatural, or even deviant character of thefollowing sentences, in which the verb posséder is supposed to indicate successivelythe part–whole, the member-group and the parent–parent relationships:(1) ?* Jean possède des yeux bleus.(2) ?* Une équipe de football possède onze joueurs.(3) ?* Jean possède deux frères.This divergence between the metalinguistic concept and the ordinary notionof “possession’’ is confirmed by the fact that, except for the adjective possessif

188 Martin Riegelin its grammatical sense, the meaning of the other derivatives of the verbposséder (possédant, possesseur, possessionné, possessionnel, possessivité and possessoire)is uniformly built upon the relationship “possession/belonging’’.The ordinary language is thus of little help in grasping the real linguisticcounterpart of the metalinguistic concept of “possession’’ and its limitations,be they interpretative or formal. Indeed the concept itself — through thevarious proposals of definitions, the alternative theoretical models thatinspire them and the data that support them — appears to be tremendouslyvast because it is very flexible. 2 Under these circumstances, a well temperedinductive approach will consist in making an inventory of the categories ofpossession commonly used by grammarians and linguists, with the hope ofreducing them to a common denominator (which could accomodate othercategories than those that helped establish it). This is what I will first try todo by identifying the extension of the possession concept in French with fourbasic forms, then by reducing those forms to a unique formula whichassociates three syntactic patterns to a very global type, although welldetermined, of mereological configuration. To conclude, I will defend thehypothesis that possession thus defined is instantiated in four categories, oneof which (the organic part–whole relationship) is central in relation to thethree others.2. The possessive relationship under all aspects2.1 Four species of the same kindIn his analysis of “the semantic structure of the possessive constructions’’ offour European languages (German, English, French and Hungarian), Fónagy(1975) distinguishes ten relationships of the possessive kind, that all fit in thesyntactic mould N1–de–N2. I will keep the three first as they are, where N1 isinterpreted respectively as the thing possessed by the owner N2, as N2’s parent(but also the affective or social partner) and as part of the whole N2, 3 whileadding the set relationship between a collective entity N2 and its elements N1.Besides the instantiation of the N1–de–N2 sequence those four kinds ofrelationship have been submitted to three other tests: the predicativeverbalization of the relationship into a sentence N2–V–N1, the substitution ofa possessive determiner for de–N2 and the possibility of a characterizingprepositional construction à–definite article–N1–(Modifier):

The part–whole relation in French 189(4) Jean possède/a trois voitures.a. les trois voitures de Jean/ses trois voituresb. l’homme aux trois voitures (blanches)(5) Jean ? *possède/a dix-huit frères.a. les dix-huit frères de Jean /ses dix-huit frèresb. l’homme aux dix-huit frères (moustachus)(6) Une voiture, ça ? *possède/a quatre roues.a. les quatre roues de la voiture/ses quatre rouesb. la voiture aux roues voilées(6¹) Jean ? *possède/a un nez bulbeux.a. le nez bulbeux de Jean/son nez bulbeuxb. l’homme aux nez bulbeux/à la moustache (grise)(7) Une équipe de football ? *possède/a onze joueurs.a. les onze joueurs d’une équipe de football /ses onze joueursb. l’équipe aux (onze) joueurs millionnaires2.2 The formal unityThe result of the tests is clear. The four “possessive’’ relationships are all castin the same syntactic moulds:1. The sentence N2–avoir–N1 is used to assert the generic relationshipwhereas the verb posséder is restricted to the specific relationship of“ownership/belonging’’. 42. Inverted prepositional construction N1–de–N2 to express the samerelationship in a presuppositional mode.3. Substitution of a possessive determiner for the prepositional phrase[de–N2] of the previous construction.4. Possibility of modifying N2 by a prepositional phrase [à–definite article–N1–(Modifier)]with a characterizing value.The conjunction of these four criteria is sufficiently discriminatory to eliminatethe undesirable couples that would only satisfy the second and thirdcriterion (for instance la construction du pont/sa construction/?*Un pont a uneconstruction/?*le pont à la construction hardie). Is there a conceptual counterpartto this formal unity? And, if such is the case, how does the first partexplain the second? And if the association of these two leads to a suitable and

190 Martin Riegeloperative definition of the linguistic notion of possession, what are theextension and the internal organization of this conceptual category?2.3 The conceptual convergencesFrom a conceptual point of view, what do couples respectively composed ofan owner and of an object he or she owns, or of two related people, or of awhole and one of its parts, or even of a collective entity and one of itselements have in common? It is not the nature — indeed basically variable —of the terms that compose the couples. However it is possible to see that, inthe four specific relationships, the second term of each of these couplesestablishes with the first an instance of the same generic relationship. Thisrelationship, which needs to be very abstract, has three characteristics: it isoriented, it links two distinct (but not necessarily disconnected) entities andit locates the second, in a way or in another, in the field determined by itsrelationship with the first. This field can be of a topological nature(whole–part or set relationships), of an (inter)connective nature (kinship),and/or of a dependency nature (ownership/belonging). In the diagramsbelow, which represent the four instances of the relationship, the term thatdetermines the field is illustrated by a striped shape, the one located in thefield is shown by a black illustration, the field itself being represented by acircle or a white oval shape.(8)[I]PèreMèrePéPrXF1F2[IV][II][III]The diagram I shows the relationship between an organic part and its whole,illustrated by the sentence (6¹) above. The whole which is the person namedJean constitutes the field as it is (hence the overlap of the striped circle and thewhite one). The nose, as a part, occupies . . . a part of this field. From a

The part–whole relation in French 191topological point of view, the relationship can be identified as a unilateraloverlap: 5 what is in the part is necessarily in the whole, but what is in thewhole is not necessarily in the part. This organic relationship between a wholeand its elements is perfectly embodied in the topological continuity and thefunctional interdependence of the parts of the human body that cannot beremoved from the whole otherwise than artificially. A competing model of thispattern is to be found in the topological and functional integration of the parts(or components) into machines and mechanisms from which they are oftenremovable (hence the possibility of standard spare parts).The diagram II symbolizes the relationship, illustrated in (7), between acollective entity and its elements. It is only different from [I] because of thediscontinuity (disjunction) of the elements inside the whole, each beingendowed with an individual autonomy. Contrary to the limbs and organs ofthe human body or to the components of an engine, the members of a footballteam or the animals of a herd are individuals that exist regardless of the otherseven if their collective coexistence implies a spatial proximity and/or a functionalinterdependence. Thus, even if the collective entity represents as suchthe reference field, its members are located in a less cohesive way than theorganic parts of a whole.The kinship can be represented with a network of hierarchic relationshipsof consanguinity and alliances in which each node identifies a “parental role’’(for instance father, sister, etc.) and in which the branch that links two nodesdesignates the kinship between two individuals both endowed with a parentalrole (for instance being the sister of/brother of, for a man and woman of thesame parents). In diagram III F1 and F2 are two men, Pierre and Paul, withrespect to whom X (a third person called Jean) plays the parental role“brother’’. The three of them are united by the kinship “being brother of’’ thatdetermines the specific parental field of “brotherhood’’. In this case therelationship is oriented since Jean is identified as a brother in relation withPierre and Paul. At this point it is enough to underline that to identifysomeone as being a brother necessarily means to place him like a node in thespecific network of the kinship where it is linked to other nodes.The relation ownership/belonging represented by diagram IV andillustrated by (4) places what is possessed in a field that gathers together allthe belongings of a same owner. It is a specific dependency relation betweentwo distinct entities, such that a person exclusively possesses an individual ora distinct object (for instance a slave, a cat, a car, an estate, a book, etc.).Thus all these four relationships can be reduced to a semantic common

192 Martin Riegeldenominator that Bally (1926) defined in a seminal article as being the“solidarité’’ relationship between an element and the “sphère personnelle’’ towhich it belongs. Being the symbolic equivalent of domaine, sphère perfectlyevokes the notion of the conceptual space within which the defining spottingof the conceptual entities (the parts, the elements, the parental roles and thepossessed objects) is carried out. But the adjective personnel is more annoyingif we consider that the second term of the whole–part and element–collectiveentity relationships represents a sphere that is not necessarily a person’s. Amore thorough study of the way the four relationships instantiate the samegeneral relationship will give a more accurate idea of the paradigmatic unityand its prototypical organization.3. A paradigm with a prototypical basis3.1 The degree of integration into the field of “possession’’3.1.1 The part–whole relationAccording to the ontology of a language like French it is undoubtedly thepart–whole relation that appears to be the more integrative. As we mentionedbefore, the organic parts are linked in a consubstantial way to the whole inwhich they are continuously integrated. In this regard the parts of the body orof other natural objects such as plants, places, etc., reveal a maximum degreeof integration and of interdependence within and with the whole (and thuswith each other). There are only few exceptions (hair loss, leaves or fruitfalling from the tree) that are naturally removable. Hence the parts areimmediately viewed as constituents and thus not removable from the whole inwhich they are said, in French, to be incluses, comprises or contenues. Thisinterdependence has revealing linguistic counterparts:1. If it is impossible to define a part “naturally’’ without relating it to its wholein a partitive way, the definition of a whole is also worthless without itsprincipal parts, the ones that are functional and perceptively prominent, unlessone wishes to distort the image of the object to be defined:(9) Un manche est la partie d’un outil ou d’un instrument par lequel on peutle tenir.(10) Un couteau est un instrument pour couper fait d’une lame et d’unmanche.

The part–whole relation in French 1932. A whole severed of one of its parts loses typicality. If a conjunction of partsis missing, it jeopardizes its categorial belonging (the whole is no longeridentified as such). Finally the deletion of all the parts leads to the deletion ofthe whole and, conversely, the existence of the part implies the existence ofthe whole:(11) a. Un chat sans oreilles, c’est un drôle de chat.b. Un chat sans oreilles, sans pattes, sans queue et sans yeux, ce n’estplus un chat.c. Un chat sans oreilles, sans pattes, sans queue, sans yeux, etc. [etc.=from which all the parts are abstracted], n’existe plus.d. Une oreille qui n’est pas une oreille de quelqu’un ou d’un animal, çan’existe pas.3. When the objects that are parts are given a name, there is generally only thisname — that names them as such (hence in French the designation nom departie) — available to designate them in common speech. Whereas undernormal circumstances I can designate (and thus categorize) a person bychoosing between one of the terms homme, commerçant, échevin, voisin, etc., Ican only designate the object that is his or her knee with the term genou. It justshows that the linguistic identification of this kind of referent is mainlydetermined by its status as “part of something else’’, since it cannot really becategorized otherwise. Thus we can explain the fact that, for such a thing, theloss of the status of part is equivalent to the loss of its essence, hence of itsexistence — which is the interpretation of (11d).3.1.2 The element–collective entity relationshipThe overall structure of a collective entity such as a football team appears lesscohesive because of the discontinuity of the members that make it up. Eachplayer is an individual endowed, as a principle, with an automatic spatiotemporalautonomy that transforms itself into a strong functional interdependenceduring the games. Thus it will not be said that a player est inclus, estcompris or est contenu dans but rather appartient à, fait partie d’une équipe.Since the elements and the collective entity interdefine each other, the deletionof some elements undermines the categorial identity of the collective entity,and the cancellation of all the elements ruins its existence:(12) a. Celui qui commande une équipe sur le terrain est son capitaine.b. Une équipe de football comprend un gardien de but, un libéro, deuxarrières latéraux, etc.

194 Martin Riegelc. Une équipe de football sans gardien de but, ce n’est plus vraimentune équipe de football.d. Enlevez tous les joueurs d’une équipe de football, elle cesse d’exister.However an individual may lose his or her status as a member of a collectivity(a player may be expelled from the team, the team may be dissolved) withoutquestioning its existence. The real linguistic counterpart of this ontologicalseparation is to be seen in the fact that the referent designated by a name thatcategorizes it as an element belonging to a collective entity can still have otherdenominative categorizations:(12) e. Si Jean est exclu de l’équipe de football, il continue néanmoinsd’exister, mais n’est plus un joueur de l’équipe.f. Si l’équipe de football cesse d’exister, ses joueurs continuent d’exister,mais ne sont plus des joueurs de l’équipe.g. Le capitaine de l’équipe est un commerçant avisé, un bon père, unépoux fidèle et un excellent saxophoniste.As a result, the consubstantial continuity and the interdependence of the partswithin the whole have been replaced by the individual discontinuity of theelements found in the collective entity which influences their topologicalregrouping and/or their functional interdependence.3.1.3 The kinship relationA kinship can be described as a predicate that assigns to each of its twoarguments a parental role. A father, for instance, is a person of masculinegender viewed according to his relationship with the person(s) he fathered.The relationship is generally oriented (the two arguments cannot be swappedsalva veritate) since it accepts a converse reading: a child (son or daughter) isa person viewed according to its relationship with the two people that conceivedhim or her. Hence a given parental name is necessarily defined accordingto its link with the partner(s) that a kinship assigns to it in the field that itdetermines. The definition is also reversible:(13) a. Le père et la mère de X sont respectivement celui qui l’a engendré etcelle qui lui a donné naissance.b. Un père qui n’a pas (eu) d’enfant(s), ça n’existe pas.c. Le fils ou la fille de X est l’homme ou la femme qui a été engendrépar X.

The part–whole relation in French 195The consanguineous parental ties biologically and legally outlive the existenceof one of the partners. The ties stemming from a union, established bymarriage can be legally broken off:(13) d. Un père qui a perdu son fils unique est toujours un père.e. Paul et Jeanne ont divorcé: ils ne sont donc plus mari et femme.f. Paul est resté en bons termes avec son ex-femme.The same individual generally plays several parental roles, and of course, otherroles depending on the status he or she has in society and on the range of hisor her activities. Hence the parental names function like relational categorizingterms among a lot of others to characterize and identify an individual:(13) g. Paul est le fils du maire, le mari de Jeanne, le président du club detennis et le meilleur client du Bar-tabac de la Poste.This denominative characteristic is the linguistic reflection of a human being’sexistential autonomy in relation to his or her parental ties.If the parts and the elements are in their own way included into a wholethey, so to speak, additionally make up, it is not the same for a parent and theone he or she is parent of. If Jean is Paul’s fils, not only will we not say thatJean is included into Paul, but such a situation is absolutely inconceivable:from a referential point of view (that is to say in extension) Jean and Paul areindividuals hence disconnected entities. But, from an intensional point ofview, a fils is an individual who has a position in the network of immediateand mediate consanguineous relationships in which he is characterized by thefact that he is the first argument of the relationship BEING GENERATED BY(x, y). The referent of a parental name is thus included from a relational pointof view into the general network of kinship.If we only take into account the necessary relationship between the parentand the one he or she is a parent of, it is this relationship, and not the partner,that is included in the meaning of the kinship term: being the son of Jean doesnot mean having the propriety of including but of being linked to him by therelation of having been generated by him. From an extensional point of viewthe parent is thus a component of the referential field that is determined bythe relationship he or she has with his or her partner. 63.1.4 The owned thing–owner relationshipThe verbal forms (a) and (b) are converse forms of one and the same predicatethat expresses the fact that an object, rarely a person, is someone’s property:

196 Martin Riegel(14) a. posséder, être (le) propriétaire/possesseur deJean possède/est le propriétaire (possesseur) de ce champ.b. être à, appartenir à, être la propriété deCe champ est à/appartient à/est la propriété de Jean.To own an object means to have the right, generally approved and guaranteedby the law, to use it as one wishes and to take the responsibility of its control.All of a person’s owned things thus constitute his or her estate. This estate maybe viewed as an extension of the one made up by the parts of a whole (but itis outside the owner) or as the elements of a whole (but the whole of what isowned is physically disconnected from the owner and does not constitute himor her). It is similar to kinship since it is determined by a relationship betweenan object (the owned thing) and another (the owner), but it is also differentbecause of its mainly contingent nature. Indeed since it is not ontologicallywritten into the nature of objects and beings, this kind of relationship onlyaffects them moderately (goods are alienable, a king’s possessions are due tohis title and not to his person).For all these reasons the integration into an estate is by far the looser ofthe four types of relations mentioned (Riegel 1984). As we saw before, notonly does the existence of the owned thing and of its owner not depend on therelationship that binds them together, but this relationship is only exceptionallylexicalized. Indeed few are the couples of words such as chien–maître,esclave–maître, bien–propriétaire, that by definition keep this relationship intheir encoded meaning. Hence, in a majority of cases, the owned thing and theowner are designated by ordinary terms, i.e. that do not identify them as suchand that are thus not relational terms. As a consequence the existence of abelonging/owning relationship is necessarily and exclusively expressed thanksto lexical and grammatical markers (être à, appartenir à, avoir, posséder, de, êtrele propriétaire de) that are sometimes ambiguous (Jean is not necessarily theowner of the book in le livre de Jean).Considering what was demonstrated above, I suggest that the French termparticipation be substituted for possession in order to designate the genericrelationship defined at the beginning of Section 2.3 and which subsumes themore specific relationships (part–whole, element–whole, kinship, and ownedthing–owner). In addition to these four relationships the “relation de participation’’will also cover not only such instances as the characterized state–entityrelationships (Jean a du courage/le courage de Jean) and characteristic behaviour–characterizedentity (Voltaire a un style/le style de Voltaire) but also the

The part–whole relation in French 197kind of relationship that constitutes couples such as sujet–roi, ouvrier–patron,élève–professeur, patient–docteur, adepte–gourou, etc.inwhichthetermsaredependent in an intersubjective way.3.2 The part–whole relation, basis of the paradigmof participation relationshipsArguments other than those presented in Sections 3.1.1–4 lead us to view thepart–whole relation as the most typical instantiation among the “participation’’relationships because it is the most integrating. The part is a consubstantialcomponent of the whole so that it is not possible to assert it of a specificwhole, because it could lead to a tautology. This restriction extends to theelement-collective entity relation, insofar as the elements are simply discontinuousparts, to those kinship relations that are necessary, but not to thebelonging-property relation (except some cases in which it combines a coupleof words that integrate it in their definition):(15) a. Une voiture, ça a un moteur/des roues/un châssis, etc.b. ?? La voiture de Jean a un moteur/des roues/un châssis.(16) a. Une équipe de football, ça a des joueurs/un gardien de butb. ?? Notre équipe de football a des joueurs/un gardien de but(17) Mon voisin a ??un père/un fils/un frère, etc.(18) a. Mon voisin a une villa/une collection de tableaux.b. Ce maître a un/des esclave(s).From a syntactic point of view, the integration of the part into the whole is sosalient that the identification of it is carried out using the anaphoric definitearticle when the whole is located in the clausal environment or in the previousco-text. The phenomenon — limited to the components of the part–wholerelation — is illustrated by the sentences (19)–(24) where the coindexingindicates the partial coreference between the part and its whole:(19) Jean lève son verre/Jean i lève la main i(20) Le chat a griffé son fils/Le chat lui i a griffé la main i(21) Je lave ma voiture/Je me i lave les mains i(22) Il tient le loup i par la queue i(23) Je brosse mes dents/Je me i brosse les dents i

198 Martin Riegel(24) Sa voiture i a été gravement accidentée. La carrosserie i est très abîmée,mais le moteur i est intact.Of all the characteristics brought to the fore in the literature dedicated to thoseconstructions 7 only one will be retained here, which confirms the consubstantialinterdependence between the whole and the parts via the iconic exploitationmade of that interdependence. When the action expressed by a directtransitive verb is exerted over the part of a whole, there are two kinds ofactantial splitting in French (Riegel 1991). On the one hand, the directcomplement represents the part directly affected, but the whole indirectlyinvolved comes out as a dative complement, often pronominal (20)–(21). Onthe other hand, if the action totally affects the whole, then this whole will bethe direct object, the part being seen as an indirect locative complementshowing the impact of the action (22). The occasional use of the possessivedeterminer to identify the part implies the loss or the change of this status.Such is the case for the first version of (23), where the referent of the object isviewed as the dentures of the subject.4. In order not to conclude . . .. . . too rapidly, it would be proper to relativize the impact of the analyses andhypotheses formulated in what precedes. First, they are only valid in Frenchwhere the metalinguistic term possession is certainly not the most appropriateto evoke the grammatical category it refers to in English and in German.Replacing it by participation or partitivité is a matter of cosmetic terminologybut does not settle the problem of determining the limits of the category thatis located on the border between lexicon, syntax, and ultimately the ontologyspecific to each language. Considering this last point of view it seems that theFrench language assigns narrower limits to the category in comparison withother languages. This is at least the picture we may have if we take intoaccount the idea that the part–whole relationships represent the basis (or thecentral instance) of a more general relationship that I called participation, ofwhich three other relationships are in fact extensions with a decreasing degreeof integration.In associating the general “participation’’ category with a conjunction ofsyntactic criteria and a definition liable to be adapted to a mereologicalcalculation it has been possible to delimit the field of the relationship in a

The part–whole relation in French 199relatively homogeneous way, maybe to the detriment of a more extensiveconception. This would imply that the principle of category extension departingfrom a basic instance, cherished by Strawson, could be applied to otherinstances, more peripheral, due to prototypicality. This is no surprise if weconsider that the basic category itself has an internal organization of aprototypical type. This is seen in the study of the clause structures — such asJean a les yeux bleus (Gaatone 1991; Riegel 1996) — which connect a couplesubject-direct object — necessarily interpreted as a whole and its part — withthe property evoked by the complement of the object.It appears that this construction expresses a complex predicate whichcharacterizes the whole via a property of one of its parts. This is a configurationthat other languages lexicalize into a compound adjective (German:blauaügig, dickköpfig, langhaarig, dickbäckig; English: blue-eyed, hard-headed,long-legged, hard-hearted). Besides, couples subject-direct object do not onlyrefer to human beings and their organic parts, which only constituteprototypical instances, but also to less typical couples the second term ofwhich refers to non-human objects (Cette casserole a le fond cabossé), clothesworn by the whole (Il a la chemise ouverte), ‘Gestalt’ configuration (Elle a ladémarche souple/ondulante), perceptive faculties and psychological andintellectual abilities (Il a l’ouïe fine/la vue perçante/l’odorat développé; Alain Reyn’a pas l’érudition triste [Le Monde]), physical and mental states (Ilalajoieexplosive/le triomphe modeste/la franchise brutale/la rancune tenace), and typicalactivities or behaviours (Il a la répartie prompte/le tutoiement/le complimentfacile). Which makes one wonder about the limits of the most representativeinstantiation of the paradigm of participation relations. This is howeveranother story.Notes1. (a): to have sexual intercourse with a woman; (b): to deceive; (c): talking about ademon, to be in possession of the spirit or the body of somebody.2. To be convinced, the contributions to the present volume give a good idea of the rangeof definitions regarding the concept and the theoretical and empirical factors that are at theorigin of this diversity.3. The other relationships can be viewed as particular cases of the first three or are far fromthe notion of possession as it is commonly used in French grammar books (see for instanceBonnard, 1971–1978: 4494–4500).4. The verb posséder has converses such as appartenir à and être à, that are submitted to the

200 Martin Riegelsame restrictions: Les roues ?*appartiennent à/?*sont à/sont des parties de la voiture — Lesjoueurs ?*appartiennent à/?*sont à/font partie de l’équipe de foot — Les dix-huit frères*appartiennent à/sont à/sont ceux de Jean.5. The overlap notion is the basis of the “calculus of the individuals’’ by Leonard andGoodman (1940), a mereologic extension of the calculus of classical sets, appropriate toformalize the four relationships that are considered possessive.6. However when it is not immediate, this relationship represents a series of relationshipsbuilding up a particular sector in the general network (being John’s nephew means beingthe son of a father who is himself John’s brother).7. See for instance among many others Junker and Martineau (1987), Riegel (1991),Spanoghe (1995) and Herslund (1997).ReferencesBally, C. 1926. “L’expression des idées de sphère personnelle et de solidarité dans leslangues indo-européennes’’. In Festschrift Louis Gauchat, F. Fankhauser and J. Jud(eds), 68–78. Arau: Sauerländer.Bonnard, H. 1971–1978. “La possession et les mots possessifs’’. In Grand Larousse de lalangue française, vol. V, 4494–4500. Paris: Larousse.Fónagy, I. 1975. “La structure des constructions possessives’’. In Langue, discours et société.Pour Emile Benveniste, J. Kristeva, J.-C. Milner and N. Ruwet (eds), 44–84. Paris: Seuil.Gaatone, D. 1991. “Un calembour syntaxique en français’’. French Review Studies 1: 45–53.Herslund, M. 1997. “Partitivity and inalienable possession’’. In Possessive Structures inDanish [KLIMT 3], I. Baron and M. Herslund (eds), 1–44. Copenhagen BusinessSchool.Junker, M. O. and Martineau, F. 1987. “Les possessions inaliénables dans les constructionsobjet’’. Revue Romane 22: 194–209.Leonard, H. S. and Goodman, N. 1940. “The Calculus of Individuals’’. Journal of SymbolicLogic 5: 45–55.Riegel, M. 1984. “Pour une redéfinition linguistique des relations dites de possession etd’appartenance’’. L’information grammaticale 23: 3–7.Riegel, M. 1991. “Transitivité et conditionnements cognitifs: La relation partie-tout et lacomplémentation verbale’’. LINX 24: 133–146.Riegel, M. 1996. “L’interprétation d’une forme tactique ou quand la méronomie investitune structure syntaxique’’. In Sémantique et cognition [SCOLIA 9], M. Riegel (ed.),191–216.Spanoghe, A. M. 1995. La syntaxe de l’appartenance inaliénable en français, en espagnol et enportugais. Bern: Peter Lang.Winston, M. E., Chaffin, R. and Herman, D. 1987. “A Taxonomy of Part-Whole Relations’’.Cognitive Science 11: 417–444.

Chapter 11Kinship in grammarÖsten Dahl and Maria Koptjevskaja-Tamm1. Introductory remarksKinship terminology has always been a central subject in anthropology, withthe focus on the ways we classify our relatives and how they relate to socialstructure. In linguistic literature, kin terms are often mentioned as a group oflexical items with special properties. Most notably perhaps, they figuretogether with body part terms as the nouns that most often show up ininalienable possessive constructions. Due to their role in what is referred to as“possessor ascension’’, body part terms have been discussed quite extensivelyin recent grammatical literature (e.g. Chapell and McGregor 1996). In contrast,there seems to have be no general and systematic treatment of thegrammatical properties of kin terms. This paper is an attempt to open adiscussion on the topic.2. Some important concepts and distinctions2.1 Terminology: Anchors and referentsKin terms are by definition relational. In the typical case, a kin term is used torefer to an individual by relating him/her kinshipwise to another individual.For instance, if I say My father is sick, I am referring to my father by relatinghim to myself. In such a case, my father is the referent and I am the anchor. IfIsayDad is sick, no anchor is explicitly indicated, but the utterance is stillunderstood to refer to my father (if there are no indications to the contrary).Thus, the anchor may be explicit or implicit. In anthropological literature, theanchor is often called the ego — in a linguistic context, this term may beambiguously understood to refer to the speaker, who has to be distinguishedfrom the anchor, although (or precisely because of) being often identical to it.

202 Östen Dahl and Maria Koptjevskaja-TammThe referent and the anchor may belong to the same or to differentgenerations. In the latter case, we distinguish ascending kin terms — wherethe generation of the referent precedes that of the ego (father, mother, grandfatheretc.) — and descending kin terms — where the referent belongs to a latergeneration (son, daughter, grandchild etc.). 1 When the referent and the anchorare of the same generation, we may talk of horizontal kin terms (sister, brother,cousin etc.)Kin terms like father and mother are called parental kin terms for obviousreasons.2.2 Proper and improper kin termsAmong words used for relatives we may further distinguish proper andimproper kin terms. Proper kin terms are those where the kin use is clearly thebasic one. In the case of improper kin terms, there is also a non-kin (typicallynon-relational) use which is at least as salient as the kin use.Swedish examples of proper kin terms are pappa ‘daddy’, mamma ‘mum’,mormor ‘MoMo’ 2 , farmor ‘FaMo’, syster ‘sister’, son ‘son’. Some improper kinterms are barn ‘child’, pojke ‘boy’ (commonly used in spoken language insteadof ‘son’) and man ‘man/husband’. It is hardly a coincidence that the improperterms are either descending or denote marital relations. In particular, we knowof no language where the most common word for ‘first generation descendant’is not the same as the most common word for ‘person who has not reachedpuberty’ (i.e. ‘child’).2.3 Taxonomy of uses of kin termsKin terms are used in many different ways, and in many contexts. Oneimportant parameter in classifying uses of kin terms is the identity of theanchor. As we have already said, a common case is for the anchor to beidentical to the speaker of the utterance. Of course, it may also be the addressee,or some set of persons including the speaker and/or the addressee. Ifthe anchor includes one or more speech act participants, we say that the use isegocentric (following the definition of this term in Dahl 1997). A perhaps lessobvious distinction also pertaining to the anchor is that between in-family andout-of-family uses. In an in-family use, the speech act participants belong tothe same family as the anchor of the kin term. This is of course highly correlatedto egocentricity but does not coincide with it. If I am speaking to a

Kinship in grammar 203stranger about my father, it is an egocentric out-of-family use. Conversely, ifa man says to his wife “Is Granny coming this weekend?’’, he may be referringto his children’s grandmother rather than to his own or his wife’s, in whichcase the use is in-family but non-egocentric. In-family and egocentric uses aretogether characterized by the backgrounded and/or highly predictable characterof the anchor, making any indication of it communicatively redundant. If,in addition, the kin term has a unique referent in the context, as will normallybe the case with a word like mother, its function comes close to that of aproper name. Such proper name-like uses of kin terms will play a crucial rolein the rest of this paper.A further distinction concerns the function of the kin term in the utterance:it may have a vocative use, as in Where are you, Daddy? or a referentialuse, as in Where is Daddy?, and a predicative use, as in He is my Daddy. In thispaper, we shall concentrate on vocative and referential uses.Vocative uses, by definition, must have second-person referents.Referential uses, on the other hand, may have first, second, and third personreferents: in certain languages and certain social contexts, kin terms may beused in lieu of first and second person pronouns. Given that the anchor mayalso be any person, we would thus theoretically obtain nine combinations ofanchors and referents in referential uses of kin terms. Probably only five arerelatively common, as shown in Table 1.Table 1. Anchor-referent combinationsFirst personreferentSecond personreferentFirst person anchor + +Second person anchor + +Third person anchor +Third personreferent3. Kin terms in spoken SwedishWe shall now give some statistics about the distribution of kin terms in actualspeech as a background to the ensuing discussion of the grammatical behaviourof kin terms, basing ourselves on a corpus of spoken Swedish, “Samtal iGöteborg’’ (for details, see Löfström 1988 and Dahl 1997), consisting of abouthalf a million words of conversation between adults, some related to eachother and some not. 3

204 Östen Dahl and Maria Koptjevskaja-TammThe first observation to be made is that kin terms are indeed very frequentin spoken discourse, and make up a sizeable proportion of all lexical NPs withanimate reference. Totally, the corpus contains 950 instances of noun phrasesused referentially or vocatively as kin terms, including 733 proper and 217improper kin terms, making up about 0.2 per cent of the total number ofwords. Two other observations may make these figures more concrete: (i) Thetwo most frequent animate nouns in our material were mamma ‘mother’ andbarn ‘child’ — that is, one proper and one improper kin term. (ii) In theinvestigation reported in Dahl (1997), a syntactic analysis was made of a subcorpusof about 10,000 finite clauses from “Samtal i Göteborg’’. Of these,52 had lexical subjects containing a proper kin term as head or used as a title,which means 23 per cent of all subjects with a lexical (non-proper name) headin the sub-corpus.The kin-referring NPs in the corpus can be classified as follows withregard to their syntactic make-up:1. “Bare nouns’’ — i.e nouns used without a determiner (incl. definiteendings) or possessor — this included the following kin terms: far ‘father’,farfar ‘FaFa’, farmor ‘FaMo’, mamma, mor ‘mother’, morfar ‘MoFa’,mormor ‘MoMo’, moster ‘MoSi’, pappa ‘father’, svärmor ‘mother-in-law’.2. Possessive-marked NPs: (a) NPs containing a possessive pronoun indicatingthe anchor, such as min far ‘my father’; (b) NPs containing a genitivephrase indicating the anchor, such as Kalles far ‘Kalle’s father’.3. Definite NPs (that is, with a definite head noun), such as farbrodern ‘theuncle’.4. Others, including NPs containing a prepositional phrase indicating theanchor, such as sonen till Kalle ‘son to Kalle’, and NPs with a kin termepithet, such as farbror Kalle ‘uncle Kalle’.Table 2 shows the distribution of parental and non-parental terms in thesegroups.Table 2. Distribution of parental and non-parental termsBare nounsGenitiveNP anchorPronominalanchorDefinite Others SumParental 243 19 118 36 13 429Non-parental 76 23 213 17534 521Sum 319 42 331 211 47 950

Kinship in grammar 205What is most salient here is the large proportion of parental bare nouns, andthe dominance of non-parental terms among definite NPs.With regard to the question of egocentric reference, about 400 of the kinterms could be identified as having a first person anchor, and about 100 ashaving a second-person one.Finally, we give in Table 3 the breakdown of proper and improper kinterms and the generation of the referent relative to the anchor, showing thedominance of proper kin terms among ascending and improper kin termsamong descending relationships, and the overall dominance of the firstascending (“parental’’) generation.Table 3. Distribution of proper and improper kin terms over generationsGenerations Proper kin terms Improper kin terms Sum–3 2 2–2 9595Ascending –1 517 517 614Same generation 0 89 42 131 131Descending 1 30 175205205Sum 733 217 950 9504. Grammatical and other peculiarities of kin termsIn many languages, some or all kin terms receive special treatment, settingthem apart from other nouns. In particular, we find examples of the followingphenomena:4.1 DefinitenessLike proper names, kin terms are often treated as inherently definite. Thismeans for instance that they may be exempt from being obligatorily markedby definite articles or other determiners, leading to bare noun uses like theones already mentioned. In some languages definite articles are simplyexcluded with many kin terms. In Standard Italian, some kin terms areexceptions to the rule that possessive pronouns are usually combined with adefinite article: mia madre ‘my mother’ vs. la mia casa ‘my house’. A somewhatsimilar situation is found in many varieties of Mainland Scandinavian,where possessive pronouns usually demand a definite form of the head noun,

206 Östen Dahl and Maria Koptjevskaja-Tammwhen postposed to it, with the exception of some kin terms, which take thebare form: far min ‘my father’ vs. huset mitt ‘my house’-def.Other reflexes of the treatment as definites is e.g. that ‘naked’ kin termsobey the same constraints as NPs with an overt article and proper names insyntactic contexts such as there-constructions (English, Scandinavian) or withrespect to object marking, as in Hungarian, where they trigger definite objectendings in verbs.In languages where kin terms in proper name-like uses are unmarked fordefiniteness, the same terms with definite marking may be used with a thirdpersonanchor, giving rise to minimal pairs such as Swedish Pappa kommer‘(Our) father is coming’: Pappa-n kommer ‘The father (of a contextuallydetermined third-person referent) is coming’. This may be exploited forcreating subtle differences in empathy in written prose.4.2 “Proprial’’ articlesSome languages have so-called “proprial’’ articles whose primary use is withproper names. Such articles are often extended to kin terms in proper namelikeuses, that is, the same contexts where bare kin terms are used in otherlanguages. Examples are found in Scandinavian dialects, e.g. Northern Swedish‘nPer‘Per’, ‘n far ‘father’. Another case in point is Samoan (Austronesian), InSamoan, the locative preposition i ‘in, at’ and the directional preposition ‘i ‘to’have special forms iā and ‘iā respectively when used before proper names,personal pronouns and naked kin terms, such as ‘mother’ and ‘father’.Historically these latter variants come from the combination of the prepositionwith the proprial article a, which is still found in a few Polynesian languages.In such cases the kin terms are used egocentrically, referring to the speaker’s,or to the hearer’s parents:(1) sā ‘ou fai ‘i- -ā tinā e sauimmpast 1sg say to prop mother uns come‘I told my mother to come’ (Jonsson 1999: 37)(2) sau e fesoasoani ‘i- ā tamācome uns give.help to prop father‘Come and help your father’ (said by a father to his child) (Jonsson1999: 38)The kin terms which can be used naked and, thus, can occur in combinationswith iā and ‘iā refer to consanguineal individual relatives, the central place

Kinship in grammar 207among which is taken by parental terms. Note that kin terms combining withpossessive pronouns or articles do not take proprial articles — cf. the exampleabove with the following more or less synonymous sentence:(3) sā ‘ou fai ‘i lo’u tinā e sauimmpast 1sg say to my mother uns come‘I told my mother to come’ (ib.)4.3 Plural formationLike proper names, kin terms may sometimes lack plural forms, or they mayshare special plural endings. 4 For instance, in Dalecarlian (Germanic) asspoken in Älvdalen, Sweden, the following kin terms are said to lack plurals(Levander 1909, Steensland 1986):faðer ‘Fa’; muna ‘Mo’; fafar ‘FaFa’; muäfar ‘MoFa’; mumun ‘MoMo’;famun ‘FaMo’; hlunga ‘female cousin’; tytta ‘aunt’Other kin terms do have plurals, for instance:bruäðer ‘Br’; syster ‘Si’; faster ‘FaSi’; muäster ‘MoSi’; duäter ‘Da’In Polish, masculine animate nouns usually end in -i in the nominative plural,e.g. student: studenci. However, the ending -owie is used regularly with• proper names: Jan: Janowie• kin terms: ojciec: ojcowie ‘father’ except wnuk ‘grandchild’ and kuzyn‘cousin’• titles such as generał: generałowie4.4 Associative (group) pluralsA large number of typologically different languages has a regular way ofbuilding constructions with the general meaning ‘X + those surrounding X’ —so-called ‘‘associative plurals’’. At least in some languages, these constructionsare restricted to proper names and (certain) kin terms. For examples, this istrue for Lezgian (NE Caucasian; Haspelmath 1993: 79), where this meaning isexpressed through a combination of a reduced genitive ending and asubstantivizing plural sufix. This formation is only limited to proper namesand some kinship terms, e.g.:

208 Östen Dahl and Maria Koptjevskaja-Tamm(4) a. Dide-d-bur bazar.di-z fe-namother-gen-sbst.pl market-dat go-aor‘Mother and those with her went to the market’b. Suna xala-d-bur.u-n k’walSuna aunt-gen-sbst.pl-gen house‘the house of Suna-xala and her family’Similarly in Kpelle — and more generally in the Mande languages withinNiger-Congo — the associative plural marker is used with proper names ofpersons and personified animals, the interrogative pronoun ‘who’ and certainkin terms (Bill Anderson p.c. via Edith Moravcsik p.c.).4.5 PossessorsWhen used to express “possessors’’ in possessive constructions, kin terms alsosometimes go together with proper names. Thus, in Dalecarlian, only definites,proper names and kin terms can have a genitive in -es: wardj-em-es ‘thewolf’s’, Jerk-es ‘Eric’s’, fadher-es ‘father’s’. Similarly, in Standard German, onlyproper names and kin terms may be used as possessors before the head noun,e.g. Peters Buch ‘Peter’s book’, Vaters Auto ‘Father’s car’. In Russian, so-calledpossessive adjectives, which were once the standard way of forming possessiveNPs, are now basically restricted to proper names and kin terms in -a, e.g.Mašina stat’ja ‘Masha’s article’, papin stul ‘Dad’s chair’. In Faroese, propernames and bare kin terms have a special ‘‘possessive’’ form:(5) Jákup-sa(r)/mammu-sa(r) bókJákup-poss/mother-poss book(Lockwood 1955: 104, 106; Barnes and Weyhe 1994: 198–9).4.6 Kin terms as heads of possessive constructionsThe behaviour of kin terms in the role of heads of possessive constructions isprobably the most discussed aspect of kinship in grammar. It is a well-knownfact that kin terms and body part terms (head, foot, stomach etc.) are the twosemantic classes of nouns that are most often treated as “inalienable’’ wheneveralienability distinctions are made.Some comments on such distinctions are in place here. As we noted in anearlier paper (Dahl and Koptjevskaja-Tamm 1998), the traditional labels givethe impression that one is dealing with two kinds of possession: alienable and

Kinship in grammar 209inalienable, variously characterized in the literature in terms of e.g. thepermanence, inherentness or essentiality of the possessive relationship and/orthe relationality of the head noun. The choice between inalienable andalienable constructions is seldom predictable from such general definitions,however; rather, what the alienability distinction means in most languages isthat a set of relational nouns are singled out for special treatment, and thatthis set always includes members of one or both of the groups mentionedabove. Equally important, however, is the phenomenon of obligatory possessormarking, which tends to occur with the same nouns as those that appear ininalienable constructions.In our previous paper, we noted that although kin terms and body partsare alike in that they can be part of the core of an inalienable construction,they are also rather different in many ways. Thus, as we have already seen, kinterms are animate nouns which in their typical uses strongly resemble propernames, i.e. they are used of well-known individuals in the immediate situationalcontext — syntactically, typically subjects. In the normal case, theanchor of a kin term is indeed the “I’’ of the speech act, i.e. the speaker, andthe addition of a 1st person pronoun is thus communicatively redundant. Yet,grammaticalization processes may lead to a situation where the grammarobliges the speaker to do so.By contrast, body part terms are inanimate nouns, and tend to show up inrather different syntactic contexts than kin terms — they are usually objects oradverbials. But here also, the “possessor’’ is highly predictable, stronglytending to be identical to the referent of the subject of the sentence, oralternatively, a referent of another NP, according to the particular syntacticconstruction used (e.g. the direct object in a sentence such as I hit him in thehead). Body parts are also special in that their state and any changes in it havedirect relevance for the “whole’’ organism. Thus, in most contexts, body partterms have a low discourse status: what really matters is not so much the bodypart as such but rather the affected person or animal. This motivates syntacticconstructions such as possessor ascension/ external possession and body-partincorporation (Chappell and McGregor 1996 and König and Haspelmath1998). Generalizing, we might say that kin terms are egocentric and pragmaticallyanchored, while body parts are syntactically anchored.4.6.1 Possessive constructions restricted to kin termsThis is the type of situation that is normally described in terms of alienabilitydistinctions. We argued in our previous paper that it is the result of a

210 Östen Dahl and Maria Koptjevskaja-Tammgrammaticalization process by which a younger and expanding possessiveconstruction is encroaching on the territory of an older one, leaving it onlywith a few lexical items. Alternatively, the differentiation might arise by aphonetic reduction of possessive pronouns in certain contexts where they arehighly predictable and thus prone to lose stress.One important aspect of alienability distinctions noted in Nichols (1988)is that they are often highly lexically idiosyncratic, that is, that there is in factno general semantic rule that determines whether an inalienable constructioncan be used or not. Thus, such a construction may be possible with one kinterm and impossible with another. Such a situation commonly arises in thefinal stages of the expansion of a new construction, where only a few lexicalitems, usually high-frequency ones, are able to resist the process.We shall use kin term constructions in Catalan as an illustrative example,without making any claims as to the exact ways in which the current situationhas arisen. Catalan has two possessive constructions, as shown by the followingexamples:• ‘inalienable’: mon pare ‘my father’• ‘alienable’: a meva casa ‘my house’We can see that in addition to utilizing two different sets of possessive pronouns,the constructions differ in the presence of a definite article only in the‘alienable’ construction. This is analogous to the Italian situation referred toabove. In Catalan, however, the ‘inalienable’ construction is possible only withcertain kin terms, with considerable variation between dialects. Table 4 showsthe information given in Alcover and Moll (1956) with respect to the possiblecombinations of mon and different male kin terms in twelve different locations.(Regrettably, the corresponding female terms cannot be included, due toinsufficient information in the source.)Table 4. Distribution of mon in different locationsNoun 1 2 3 4 5 6 7 8 9 10 11 12pare ‘father’ + + + + + + + + +germà ‘brother’ + + + + + + + + +tio/oncle ‘uncle’ + + + + + + + +cosí ‘cousin’ + + + + + + +cunyat ‘brother-in-law’ + + + +nebot ‘nephew’ + + + +sogre ‘father-in-law’ +padri ‘godfather’ + +aví ‘grandfather’ +

Kinship in grammar 2114.6.2 Obligatory possessor markingThere are really two varieties of obligatory possessor marking. In the first,certain nouns obligatorily carry a possessive affix. An example is Navajo(Athabascan):(6) shi-ma ‘my mother’a-ma ‘someone’s mother, mother in general’In the second variety, the possessor is obligatory but may be either pronominalor lexical. This is found in the Tupí-Guaraní language family, e.g. SirionóJuanito ru ‘Juanito’s father’, nde-ru ‘your father’ (own data). (Cf. Velázquez-Castillo 1996:62 for the corresponding facts in Guaraní.)It may be noted that there is a parallel here to obligatory subject marking.In English or Swedish, every sentence must (in principle) have a subject, butit may be either pronominal or lexical. By contrast, many languages have anobligatory subject marker in the form of a pronoun or an affix on the verb,irrespective of the presence of a lexical subject.We may also compare with the obligatoriness of possessive pronouns withsubject-controlled body part terms in English, e.g. I hurt my foot (not ‘the foot’as in many other languages).We see obligatory possessor marking as the result of a grammaticalizationprocess by which possessive pronouns come to be used whenever possible,rather than when communicatively motivated. The frequent use of pronounsin communicatively redundant positions facilitates their reduction to affixes.4.7 Kinship verbsOur discussion has so far focussed on kin terms as a special subtype of nouns.Thus, kin terms typically share most grammatical properties with other nounsin the language, but in addition show certain peculiarities. Interestingly, insome languages kinship relations are more or less consistently expressed notby nouns, but as verbs. The following discussion is based on the detailed andinsightful treatment of this phenomenon in Evans (1999).As an example of such a language, let us take Ilgar (Northern Australia,the Iwaidjan family). Verbs in Ilgar attach person-gender prefixes for bothsubject and object. The same prefixes can be efficiently employed in referringto various kinship relations, when combined with a closed set of verbal kinshiproots, such as -wula illustrated below. Thus, in the two following expressions,the anchor is coded by the object (absolutive) prefixes, whereas the referent is

212 Östen Dahl and Maria Koptjevskaja-Tammcoded by the subject (ergative) ones, and the interpretation is something like‘an X such that X is Y’s mother’:(7) an-na-wula1sg.abs-3sg.f.erg-be.mother.to‘my mother’ (she is mother to me)(8) a-ni-wula1sg.abs-3sg.m.erg-be.mother.to‘my maternal uncle’ (he is (as a) mother to me)With appropriate subject-object prefixes, the same root can also be exploitedfor reference to the converse relations: ‘an X such that Y is X’s mother’, e.g.:(9) a-wula1sg.erg/3sg.m.abs-be.mother.to‘my son (female speaker), my sister’s son (male speaker)’(10) a-bula1sg.erg/3sg.f.abs-be.mother.toNP‘my daughter (female speaker), my sister’s daughter (male speaker)’Kinship verbs are attested in other Australian languages (e.g. the Gunwinyguanlanguages), in the Iroquoian languages (e.g. Seneca and Cayuga), inthe Yuman languages (e.g. Yuma) and in the Uto-Aztecan languages (e.g.Central Guerrero Nahuatl and Cahuilla), but probably also exist in quite a fewothers. Their semantics can vary; thus, their meanings can be ‘be K to X’, ‘haveX as K’, ‘call X a K’ and ‘consider X a K’ (where K refers to a particular kinshiprelation).The rationale behind this phenomenon is sufficiently clear: kinshipinvolves (at least) two-place relations between humans and the argumentstructure of kin terms is, thus, similar to that of transitive verbs. Relationality,on the whole, is not typical of nouns. On the other hand, kinship involvesstatic relations, as opposed to typically dynamic two-place verbs. Thus,whereas many languages take the dichotomy between ‘dynamic’ and ‘static’semantics as the basis for their lexical categorization into verbs and nouns,quite a few of them appear to use relationality as the essential underlyingfactor for such a categorization (Evans 1999). However, the stativity of kinterms makes them relatively poor members of the verb class, which is manifested,among other things, in their restricted repertoire of such typical verbalcharacteristics as tense-aspect-mood categories.

Kinship in grammar 213Also, kinship verbs typically co-exist with kinship nouns within one andthe same language, and the choice between the two types of expressions isgoverned by various factors.5. Differences among kin terms: The parental prototypeAs we have seen, kin terms as a lexical domain tend to behave in particularways, due to their special referential properties in discourse. In many respects,kin terms go together with proper names. From the point of view ofgrammaticalization processes, this works in two ways. Sometimes propernames and kin terms form the core of the domain of a certain phenomenon,such as proprial articles or special suffixes such as the nominative pluralmorpheme -owie in Polish. In other cases, proper names and kin terms arealike in resisting some grammaticalization process, such as the spread ofdefinite articles. (In the end, even proper names and kin terms take definitearticles.) With respect to possessive constructions, kin terms go together withbody part terms to form a residual domain, which is the last one to be conqueredby a new, expanding construction. But it would be wrong to think ofkin terms as a monolithic class. In actual practice, and as we saw above inseveral cases, the phenomena we have been talking about are often restrictedto a subset of all the kin terms in a language. It can be noted that lexicalidiosyncrasy seems to be a general property of residual phenomena ingrammaticalization. But it is also clear that not all kin terms, and not all usesof kin terms, are alike with respect to the properties that motivate their specialtreatment. Thus, proper name-like uses are much more frequent with certainkin terms, notably the ones that denote ascending relations, and in particularparental ones, such as father and mother. This is partly dependent on socialconventions. In the English-speaking world, one would address one’s father asfather but one’s siblings or children rather by their names. In China, on theother hand, expressions such as ‘third sister’ are regularly used for the youngermembers of a family. However, a system which is the inverse of the Englishone is not attested, to our knowledge. This suggests a universal partial orderingof kin terms (‘hierarchy’ being too strong), based on closeness to what wewould propose to call the parental prototype. In fact, parental terms weresingled out already in Joseph Greenberg’s work on markedness relations (1966,1980). Greenberg claimed that kin terms could be assigned different degrees ofmarkedness using the criteria set out in Table 5.

214 Östen Dahl and Maria Koptjevskaja-TammTable 5. Greenberg’s criteria of markedness for kin termsLess marked More marked ExampleLineal Collateral ‘father’

Kinship in grammar 215out parental terms from all others (or with the only exceptions being otherascending terms such as Swedish farfar ‘FaFa’) is that of unique reference(relative to an anchor). It is plausible that this property creates favourableconditions for proper name-like uses. It should be noted that such uses are notcharacteristic of terms like parent(s), which are in most languages only orpreferably used with plural reference.We shall now give a couple of further illustrations of differentiatedbehaviour among kin terms, giving additional support to the idea of theparental prototype.In Kera (Chadic; Ebert 1979: 171), possessive suffixes are in generalobligatory with kin terms and body part terms. In many contexts, a definitemarker -´ often realized only as a change in tone, is also obligatory. What isrelevant for our purposes is that the presence or absence of this definitemarker may distinguish otherwise homophonous pairs of kin terms, where thedefinite marker is present on the descending member and absent on theascending one:nəəndə`n ‘my MoBr’ : nəəndəń ‘my SiSo’moomədù ‘his grandfather’: moomədù´ ‘his grandson’In Somali, the same term, adeer, is employed in addressing an uncle by hisnephew or a nephew by his uncle. In referential uses, however, the term isrestricted for reference to the ascending generation, whereas the descendinggeneration is referred to by a relative clause construction, such as:(11) ínan aan adeer u ahayboy am uncle ben is‘a boy to whom I am an uncle’ (Serzisko 1983: 72).This asymmetry between the ascending and descending generations is seen inseveral other kin terms, referring to mother/father vs. child, grandmother/grandfather vs. grandchild, aunt vs. nephew/niece relations.In Georgian (Kartvelian; Thomas Widmann and Nino Amiridze, p.c.),a number of kin terms use clitic possessive pronouns, e.g., mamachemi‘my father’, dedasheni ‘your (sg.) mother’ opposed to preposed free pronounsas in chemi da ‘my sister’, sheni saxli ‘your (sg.) house’. These kin terms areas can be seen from the following list all ascending: mama ‘father’, deda‘mother’, bebia ‘grandmother’ babua or p’ap’a ‘grandfather’, deida ‘mother’ssister’, mamida ‘father’s sister’, bija ‘uncle’, bijola ‘aunt’.The examples given above from Dalecarlian, Polish, and Catalan also show

216 Östen Dahl and Maria Koptjevskaja-Tamma differentiated behaviour among kin terms, which does not fit the partialordering defined by the conditions above but still gives support to the idea ofa clustering around parental terms.In addition to the grammatical parameters we have already talked about,it seems that the parental prototype also can be used to account for otherdistinctions within the kin domain. For instance, in many languages, there arekin terms that are formed by a reduplicative CVCV, CVC or VCV pattern,such as mama, papa, ata and the like. It appears that this is most common forparental terms, relatively common for other ascending kin terms and leastcommon for descending ones. The list of Georgian kin terms just givenillustrates this. Consider also the set of reduplicative kin terms in Russian:papa ‘father’, mama ‘mother’, bab(ušk)a ‘grandmother’, ded ‘grandfather’,djadja ‘uncle’, tetja ‘aunt’.Possessive NPs in the Polynesian languages are famous for the distinctionbetween ‘‘dominant’’ possession (o-possession) and ‘‘subordinate possession’’(a-possession), which has certain similarities with, but is not exactly the sameas the traditional alienability split. The two categories correlate with thedifference in the relation between a possessee and a possessor. For our purposeit is noteworthy that parental terms and other kin terms denoting ascendingrelations cluster together with typically inherently, inalienably possessedentities, such as body parts (o-possession). Other kin terms cluster with wordswhich, for instance, denote more or less recently acquired personal possessions(a-possession) (Bauer 1993: 209–216).6. Kin term doubletsThe differentiation among kin terms that we have discussed so far has beenbased on the semantics of the terms, that is, on the properties of the kinshiprelation they express — for instance, ascending relations were distinguishedfrom descending ones. There is another type of differentiation, however, thatdistinguishes kin terms that are denotationally synonymous, that is, expressthe same kinship relations. The existence of such kin term doublets in alanguage like English is of course something that most of its speakers areaware of: fathers are called Daddy and mothers Mum or the like. It appearsthat the frequency of such doublets is another phenomenon that is correlatedwith the parental prototype. The phenomenon is not restricted to Europeanlanguages, as the following list shows:

Kinship in grammar 217• English: dad: father; mum: mother• Swedish: pappa: far ‘father’; mamma: mor ‘mother’• French: papa: père ‘father’; maman: mère ‘mother’• Russian: papa: otec ‘father’; mama: mat’ ‘mother’• Sirionó: paba: ru ‘father’; tei: si ‘mother’• Slave (Athabascan; Rice 1989): ʔabá: -ta ‘father’; ʔamá: -ne ‘mother’• Samoan (Jonsson 1999): tamā: papa ‘father’; tinā: mama ‘mother’The members of kin term doublets usually differ with respect to their use. Acommon pattern is for one member to be more frequent in vocative andegocentric uses. Sometimes, this seems to be regularized to an extent where itis motivated to talk of suppletion within a single paradigm. Statements to thiseffect are not seldom found in grammars, but it is hard to know exactly whatreality is behind them.In Axininca Campa (Arawakan) the following suppletive pairs are foundaccording to Payne (1981):nowaapati ‘my father’paapa ‘father’ (vocative)tomichi ‘my son’nochomi ‘son’ (vocative)In Slave (Rice 1989:227), the vocative of -ne ‘mother’, ʔene, is said to be used“mostly in referring to an older or deceased woman’’ whereas the usualvocative form for ‘mother’ is ʔamá.Merlan (1982: 138) reports that in certain Australian languages “referencein other than the egocentric referential mode requires the use of a set of termswhich (for some relationships) differ from vocative and egocentric referentialkin terms’’. Thus, in Jawoñ a child calls her own mother gara but whenreferring to another uses the term -ņakunñira?mi.Less drastically, the vocative function may require a special derivative formsuch as a diminutive, as in Polish: babka ‘grandmother’ would have thevocative babko! but instead the diminutive babciu! is used (F. Gladney,personal communication). In Klamath (Penutian; Scott Delancey, p.c.), kinterms used referentially take a special prefix, /b-/ (historically a 3rd personclitic pronoun). In vocative use, some kin terms lack this prefix but insteadtake a special suffix not found anywhere else in the language:p-tis-ap ‘father (referential)’tis-i:p ‘father! (vocative)’

218 Östen Dahl and Maria Koptjevskaja-Tammp-k’is-ap ‘mother’k’is-i:p ‘mother!’Corresponding to these differences in use one also frequently finds differenceswhich are directly relevant to our discussion in that the members of kin termdoublets tend to behave differently with respect to such phenomena as obligatorypossessive and definiteness marking. In French, only papa has a bare nounuse in referential function, as is illustrated by the following example:(12) ?? Père est venu/Mon père est venu/Papa est venu‘father has come’Similarly, although possessive prefixes are obligatory with kin terms in Slaveaccording to Rice (1989:227), her examples show that there are kin termdoublets where this only holds for one of the members:(13) Mary be-tá ʔeyá hiliM. 3sg-father sick is‘Mary’s father is sick’ (1989: 229)(14) ʔabá gok’eríʔéefather jacket‘father’s jacket’ (1989: 231)Likewise the obligatoriness of possessor markers in Sirionó holds only for thesecond members of the pairs listed above:(15) hue s paba u-chähere from father 3sg-flee‘(my) father fled from here’(16) aba nde-ru?who 2sg-father‘who is your father?’In the next section, we shall consider some possible diachronic explanations ofkin-term doublets.7. Renewal of kin termsThere seems to be a tendency to think of kin terms like father, mother, brotheretc. as belonging to the most stable parts of vocabulary. Indeed, the English

Kinship in grammar 219words just mentioned belong to the common Indo-European heritage,meaning that they have been around for a long time. In spite of this, renewalof kin terms does take place, sufficiently often 5 for there to be at least fourdifferent unrelated terms for ‘father’ in the Germanic languages:• as in English father etc.• as in Swedish pappa etc. (apparently from French; general spread inFrench-influenced area)• as in English daddy• as in Gothic atta (many parallels in the area, cf. Turkish ata and Russianotec)Parental and parental-like kin terms may be renewed from several differentsources:• “nursery language’’ (“Motherese’’)• loans from other languages• slang• diminutivesWhile the influence of “Motherese’’ and child language on parental terminologyhas always been acknowledged and was stressed in Jakobson’s classicalpaper (1966), the fact that parental kin terms are not seldom borrowed fromone language to another has been less often appreciated. The forms ‘Mama’and ‘Papa’, alluded to in the title of Jakobson’s paper, are in fact the mostobvious cases in point, in that they were borrowed into the Germanic andSlavic languages from French, apparently during the period when Frenchupper class culture spread in Europe. In Swedish, they have become generallyused among all groups in society only in the twentieth century. They are thusnot somehow an automatic extension of children’s speech, as is sometimesthought. However, renewals such as the introduction of ‘Mama’ and ‘Papa’seem to follow a certain pattern, which explains the typical distribution of themembers of kin term doublets. More specifically, innovated kin terms like‘Mama’ and ‘Papa’ tend to be restricted to vocative and/or egocentric andin-family uses, that is, the uses where the identity of the anchor is mostpredictable, and where the proper name-like character of the kin-referring NPsis most obvious. (The existence of vocative-referential suppletion and similarphenomena in kin terms suggests that the vocative function tends to be theoriginal renewal site.) It is therefore not astonishing that innovated kin termstend to show proper name-like behaviour with respect to possessive and

220 Östen Dahl and Maria Koptjevskaja-Tammdefinite marking. Not only are possessive and definite markers frequently nonobligatorywith such terms, they may even be impossible. As time passes,however, these properties may change. As an example, we may consider theintroduction of the words pappa and mamma into Swedish. (The accountbuilds on information given in the Swedish Academy Dictionary.)Pappa and mamma entered the Swedish language in the late seventeenthcentury, as loans from French. Until the nineteenth century, they do not seemto have been used in the plural or with definite articles. In the twentiethcentury, such uses have become quite normal. The words far and mor (shortenedforms of the lexemes fader and moder) 6 were common in in-family usesuntil fairly recently, but today, pappa and mamma are rapidly replacing themeven in out-of-family uses.The development during the last century is documented in Table 6. Itshows the frequencies of various forms of the lexemes pappa, fader/far,mamma, and moder/mor in two corpora from the Swedish Language Bank.Regrettably, some of the forms are homonymous 7 and the figures are thereforenot totally reliable. In spite of this, the table gives clear evidence of twotendencies: the general shift from fader/far and moder/mor to pappa andTable 6. The expansion of pappa and mamma: some corpus evidenceNovelsNovels19th and 1980–81 19th and 1980–81early 20th C.early 20th C.N % N % N % N %pappa 476 10.4 1,092 32.8 mamma 575 13.2 1,842 41.6pappas 49 1.1 1354.1 mammas 39 0.9 183 4.1pappan 4 0.1 44 1.3 mamman 10 0.2 89 2.0pappans 1 0.0 50.2 mammans 1 0.0 10 0.2pappor 2 0.0 10 0.3 mammor 3 0.1 19 0.4papporna 0 0.0 4 0.1 mammorna 0 0.0 0 0.0far* 3,079 67.2 1,284 38.6 mor 2,840 65.2 1,461 33.0fader 379 8.3 266 8.0 moder 203 4.7 155 3.5fadern 425 9.3 343 10.3 modern* 540 12.4 589 13.3faderns 101 2.2 87 2.6 moders 80 1.8 29 0.7fäder 40 0.9 44 1.3 mödrar 42 1.0 30 0.7fäderna 150.3 50.2 mödrarna 11 0.3 13 0.3fädernas 13 0.3 11 0.3 mödrarnas 2 0.0 2 0.0fäders 0 0.0 0 0.0 mödrars 11 0.3 2 0.0Total 4,584 3,330 4,357 4,424*Homonymous forms

Kinship in grammar 221mamma, and the increased use of the latter with definite and plural marking.The story we have just told is not unique. For instance, the Aramaic word‘abba, which is morphologically definite, was originally used only with thespeaker as anchor. It was borrowed into post-biblical Hebrew and in moderncolloquial Hebrew is apparently used without the anchor restriction (RichardSteiner, personal communication).It is also worth noting that kin terms show very different patterns ofrenewal depending on their semantics. Thus, it is obvious that words for morecomplex relationships (e.g. terms for in-laws and non-first generation kinterms) are often formed by derivation or compounding. But the synchronicskewing between proper and improper kin terms, mentioned above, is directlyrelated to the diachronic tendency to renew descending kin terms and wordsdenoting marital relations from primarily non-relational nouns such as child,man, woman etc.8. Lexical integrationThe development of kin terms like Swedish pappa and mamma is of moregeneral interest in that it can be seen as an example of what could be calledlexical integration, a process by which new lexemes are gradually pulled intothe grammatical system of the language. Lexical integration is related to themore well-known process of grammaticalization, which it both resembles andinteracts with. 8To start with the parallels, in both cases the starting-point is a renewal ofthe resources of a language, which may exploit different sources, but whichtends to repeat itself in much the same way and in the same locations. Bothlexical integration and grammaticalization are gradual and tend to involveexpansion of domains of use, from concrete to more abstract uses of items,with the final parts of the process being characterized by lexical idiosyncrasyand fossilization.But the discussion of kin terms above also illuminates the interactionbetween the two processes. As the use of a grammatical marking such as adefinite article expands, more and more lexical items are pulled into itsdomain. In this process, a lexical item which is less well integrated into thesystem has a better chance of staying outside. But over time the probabilitythat it will be treated in the same way as other lexemes increases.The gradual integration of new lexical items is most obvious with loan

222 Östen Dahl and Maria Koptjevskaja-Tammwords. But we have seen in the case of kin terms that semantic and pragmaticproperties of a word also influence the ease with which it is integrated into thelanguage. A vivid illustration of this is provided by the two words rektor ‘headmaster’and lektor ‘lecturer, senior teacher’ in Scandinavian, discussed inChristiansen (1977). (Christiansen talks about Danish but exactly the samefacts hold for Swedish.) The words in Scandinavian are borrowings from Latinrector and lector. In spite of their common origin, and their near-identicalphonological make-up and closely related semantics, the two words differ withrespect to their grammatical properties: the word rektor can be used as a barenoun, without a determiner; lektor cannot be used this way. As Christiansennotes, this clearly has to do with the fact that rektor but not lektor denotes aunique person in a school context — in addition, a figure of authority. Thus,while rektor can be compared to father, lektor is more like uncle. It can thereforebe claimed that rektor has resisted lexical integration better than lektor dueto its semantics. However, it is clear that also other factors are at play here. Aparallel case to rektor is found in the Swedish word for ‘dean’, which may beused with or without the Latin ending -us: dekanus or dekan. Only the longerform can be used without a definite article, which shows that the preservedLatin character of the word also helps it resist integration.9. ConclusionAmong nouns, kin terms have specific semantic and pragmatic properties thatexplain why they often display deviant grammatical behaviour, in particular,their role in grammaticalization processes. As we have seen above, however,kin terms do not necessarily form a well-delimited natural class. Rather, thepropensity to be singled out for special treatment depends on the closeness tothe parental prototype, with ascendance and linealness as most salient features.Moreover, a kin term may change its character over time, as it undergoeslexical integration, a process which is both similar to and interacting withgrammaticalization. The study of kin terms is therefore relevant to the understandingof central processes in synchronic and diachronic grammar.AcknowledgementsThis paper is dedicated to Ekkehard König on the occasion of his sixtieth

Kinship in grammar 223birthday. We want to thank the following persons for input and/or commentson earlier versions: Nino Amiridze, Loren A. Billings, Bart de Boer, ScottDeLancey, Alan Dench, Mark Donohue, Nicholas Evans, Zygmunt Frajzyngier,Dick Hudson, Niklas Jonsson, John E. Koontz, Bill McGregor, EdithMoravcsik, Herb Stahlke, Richard Steiner, Judit Szito, and Thomas Widmann.Notes1. Kin terms are not always unambiguously ascending or descending. In Australianlanguages, the same terms are commonly used for the second ascending and the seconddescending generations (Bill McGregor, personal communication). See section 5for similarexamples, only distinguished by the use of a definite article.2. Whenever appropriate, we use the standard notation in which e.g. ‘FaMo’ stands forfather’s mother, ‘MoBr’ for mother’s brother and so on.3. The participants in the project within which the corpus was created were asked to recordhalf an hour’s conversation with a person they knew well. In many cases, this person was amember of the same family. The corpus thus represents both in-family and out-of-familykin reference but since we did not have access to precise information about the identity ofthe referents it was not possible to distinguish them systematically.4. Another phenomenon that should be mentioned in this context is the existence of pluralsor plural-like forms with specialized meanings, for instance the “dyadic’’ forms in Australianlanguages, e.g. Panyjima kurtarra ‘two people related as brothers’ (Dench 1981:109) or formslike Icelandic feðgar ‘father and son’ and Spanish padres ‘parents’ (lit. ‘fathers’).5. There is so far no reliable global statistics on the replacement rates of individual words.Dyen et al. (1997) give values for Indo-European languages that suggest that the words‘father’ (0.67) and ‘mother’ (0.32) are relatively stable but that these words are still replacedmore often than a maximally stable word like ‘sun’ (0.19). (A word has a replacement rateof 1 if it is replaced once on the average during a time period estimated at 2,400 years.)6. The relationship between the long and short forms fader:far and moder:mor is a separatestory which is also relevant to the issue of kin term renewal. The short forms cannot takeany endings in Standard Swedish; the missing forms in the paradigm are supplied by thelong forms. Conversely, the long forms fader and moder cannot be used as bare nouns inModern Swedish (with the exception of vocative uses when addressing God or a Catholicpriest). *Fader kommer ‘Father comes’ is just ungrammatical while far kommer is OK infamilies where far is still used. The pairs of long and short forms can thus be said toconstitute a special case of a kin term doublet.7. Far ‘father’ is also the present tense of fara ‘go, travel’ and modern may be an adjectivewith the same meaning as in English.8. An alternative to ‘lexical integration’ is ‘lexicalization’ but it has been used in recentliterature with slightly different interpretations and we therefore avoid it here.

224 Östen Dahl and Maria Koptjevskaja-TammReferencesAlcover, A. and Moll, F., eds. 1956. Diccionari català-valencià-balear : inventari lexicogràfici etimològic de la llengua catalana en totes les seves formes literàries i dialectals. T.7:Lliom. Palma de Mallorca: Moll.Barnes, M. P. and Weyhe, E. 1994. ‘‘Faroese’’. In The Germanic Languages, E. König andJ. van der Auwera (eds), 190–218. London: Routledge.Bauer, W. 1993. Maori. Routledge Descriptive Grammars. London: Routledge.Chapell, H. and McGregor, W. (eds). 1996. The Grammar of Inalienability. Berlin: Moutonde Gruyter.Christiansen, H. B. 1977. “Artikelløsheden i Lild og Tømmerby sogne i Vester Hanherred’’.Dialektstudier 4: 1–28. Copenhagen: Akademisk Forlag.Dahl, Ö. 1997. ‘‘Egocentricity in Discourse and Syntax’’. http://www.ling.su.se/staff/oesten/egocentric.Dahl, Ö. and Koptjevskaja-Tamm, M. 1998. “Alienability splits and the grammaticalizationof possessive constructions’’. In Papers from the 16th Scandinavian Conference ofLinguistics, T. Haukioja (ed.). Publications of the Department of Finnish and GeneralLinguistics of the University of Turku 60. Turku: University of Turku.Dench, A. 1981. ‘‘Kin terms and pronouns of the Panyjima language of northwest WesternAustralia’’. Anthropological Forum 5: 105–120.Dyen, I., Kruskal, J. and Black, P. 1997. ‘‘IE-RATE1’’. http://www.ldc.upenn.edu/ldc/service/comp-ie/IE-RATE1.Ebert, K. 1979. Sprache und Tradition der Kera (Tschad). Teil III: Grammatik. Berlin: Verlagvon Dietrich Reimer.Evans, N. 1999. “Kinship verbs’’. In Approaches to the typology of word classes, P. M. Vogeland B. Comrie (eds). Berlin: Mouton de Gruyter.Greenberg, J. H. 1966. Language Universals [Janua Linguarum, Series Minor 59]. TheHague: Mouton.Greenberg, J. H. 1980. “Universals of kinship terminology’’. In On linguistic anthropology:Essays in honor of Harry Hoijer, J. Maquet (ed.), 9–32. Malibu: Udena Publications.Haspelmath, M. 1993. A grammar of Lezgian [Mouton grammar library 9]. Berlin: Moutonde Gruyter.Jakobson, R. 1966. ‘‘Why ‘Mama’ and ‘Papa’?’’ In Selected Writings 1. 462–74. The Hague:Mouton.Jonsson, N. 1999. Some grammatical properties of Samoan kin terms. M.A. thesis. Departmentof Linguistics, Stockholm University.König, E. and Haspelmath, M. 1998. “Les constructions à possesseur externe dans leslangues de l’Europe’’. In Actance et valence dans les langues de l’Europe, J. Feuillet (ed.),525–606. Berlin: Mouton de Gruyter.Levander, L. 1909. Älvdalsmålet i Dalarna. Ordböjning och syntax [Svenska landsmål ochsvenskt folkliv 1909, H.2]. Stockholm.Lockwood, W. B. 1955. An introduction to modern Faroese. Copenhagen: Munksgaard.Löfström, J. 1988. ‘‘Repliker utan gränser. Till studiet av syntaktisk struktur i samtal’’.Institutionen för nordiska språk, Göteborgs universitet.

Kinship in grammar 225Merlan, F. 1982. ‘‘‘Egocentric’ and ‘altercentric’ usage of kin terms in Maarayi’’. In Thelanguages of kinship in Aboriginal Australia, J. Heath, F. Merlan and A. Rumsey (eds),125–140. Sydney: University of Sydney.Murdock, G. P. 1959. “Cross-Language Parallels in Parental Kin Terms’’. AnthropologicalLinguistics 1: 1–5.Nichols, J. 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.Payne, D. L. 1981. “The phonology and morphology of Axininca Campa’’. Summer Instituteof Linguistics, no. 66.Payne, D. and Barshi, I. (eds). 1999. External Possession. Amsterdam: John Benjamins.Rice, K. 1989. A Grammar of Slave. [Mouton Grammar Library]. Berlin and New York:Mouton de Gruyter.Serzisko, F. 1983. Der Ausdruck der Possessivität in Somali. Tübingen: Gunter Narr.Steensland, L. 1986. Liten älvdalsk-svensk och svensk-älvdalsk ordbok. Älvdalen: Ulum dalska.Velázquez-Castillo, M. 1996. The grammar of possession: inalienability, incorporation andpossessor ascension in Guaraní [Studies in Language Companion Series 33]. Amsterdam:John Benjamins.

Chapter 12(In)alienability and (in)determinationin PortugueseAnne-Marie Spanoghe1. Preliminary remarksSo far there has already been done a huge amount of research on the expressionof the so-called possessive relationship in natural human language.Specifically, the so-called (in)alienable possessive relationship and the partwholerelationship are items of top priority in general linguistics and languagetypology. 1Referring to ‘inalienability’, I briefly recall that this terminology was partlyinvented by Lévy-Bruhl (1916), who used the term ‘alienable’ to characterizeone of the two existing noun classes in the Melanesian languages:The first class (nouns taking the suffix) is constituted by nouns designating bodyparts, parts of a thing, objects closely related to a person (his/her arms, his/herfishing net, etc.), kinship terms, and some prepositions expressing spatial relations:beside, on, next to, far from, etc. The second class contains all other nouns[...]〈and is characterized by〉 the fact that I possess something as a property, andas an alienable property. 〈In contrast to the alienable property, Lévy-Bruhlsupposes the existence〉 of a solidarity that cannot be broken, [. . .] an interdependence,a participation. 2In 1917 Uhlenbeck introduced the terminological opposition alienable vs.inalienable possession. 3 In a famous publication of 1926 Bally detected thepossible relevance of personal sphere and solidarity in the syntax of Indo-European languages. 4, 5Given the fact that as far as I know, there is no investigation describingexhaustively the expression of ‘inalienability’ in the Romance languages, 6 Ihave focused on noun phrases designating body parts (NPbp) in French,Spanish and Portuguese, considering the relationship between a human beingand his/her body part(s) as a good example of an inalienable relationship. Onthe other hand, assuming that the expression of inalienability is subject to

228 Anne-Marie Spanogheparameters of a different kind in French, Spanish and Portuguese my researchwill also focus on the possible relationship between inalienability and determinationin these three Romance languages through the functioning of thedefinite article, the indefinite article and the possessive determiner in EuropeanPortuguese. 7, 8From a methodological point of view, my study has a double stratification:the first level presents theoretical viewpoints and the second concentratesmore on the description of the empirical data. 9In the first place, I will underline the particular position of the Portugueselanguage among other Romance languages such as French and Spanish withregard to body part syntax (Section 2). After a short note on Brazilian Portuguese(Section 3), I describe some of the rules of the ‘discourse game’ betweenthe definite article, the indefinite article and the possessive determiner ofNPbp in European Portuguese (Section 4).2. ‘Body part syntax’ 10 in French, Spanish and Portuguese 112.1 Syntactic functions of NPbpIf we take into account the syntactic functions of NPbp , we observe thatFrench, Spanish and Portuguese NPbp prefer, in decreasing order, the functionsof direct object (‘DO’), oblique object (‘OO’), subject (‘S’) and indirectobject (IO):DO>OO>S>IO.For the Portuguese NPbp the statistics are as follows:Direct Object (52% 12 )O Escritor ri, encolhe os ombros e abre os braços [. . .]. (M. Alegre, Jornada deÁfrica, 157)‘The Author laughs, shrugs the shoulders and opens the arms’ 13Oblique Object (36%)Procuro com a mão o despertador que está a tocar há mais de meio minuto.(L. de Sttau Monteiro, Um homem não chora, 111)‘I search with the hand the alarm which has been calling for more than half aminute’

(In)alienability and (in)determination in Portuguese 229Subject (11.9%)O seu rosto de porcelana, com olhos dum azul incandescente e cabelos cor depalha, era um modelo de perfeição [. . .].(J. de Melo, Gente feliz com lágrimas,194)‘His china face, with incandescent blue eyes and straw hair, was a model ofperfection’Indirect Object (1 item)Os olhos piscavam mais intensamente por detrás dos óculos redondos e brancos,sob a intrincada cabeleira que lhe roubava espaço à testa. (A. Faria, Emenda esoneto, 195)‘The eyes were blinking more intensely behind the round and white spectacles,beneath the knotty hair which was robbing him of space on the front’2.2 Syntactic functions of the PossessorYet, if we examine the syntactic functions of the NP denoting the possessor ofthe respective body part(s) (‘NPP’), it appears that the Portuguese languagediffers clearly from French and Spanish. In Portuguese the NPP is expressed inorder of preference by: 14Subject function (56%)O Escritor ri, encolhe os ombros e abre os braços [...]. (M. Alegre, Jornada deÁfrica, 157)‘The Author laughs, shrugs the shoulders and opens the arms’Indirect Object (24.4%)Amei-lhe as grandes mãos [. . .].(J. de Melo, Gente feliz com lágrimas, 420)‘I loved him the big hands’Pragmatic/discursive factors (10%)Recusa-se a revê-la assim, figura de barro mirando-se na água enegrecida pelasnuvens. Procura os olhos azuis de outrora, os negros cabelos vivos [. . .].(J.deMelo,Gente feliz com lágrimas, 313)‘He refuses to see her again like this, a figure of mud contemplating itself inthe water darkened by the clouds. He searches for the formerly blue eyes, thedark and lively hair’

230 Anne-Marie SpanoghePossessive determiner (6.2%)Ele estava voltado para a janela cerrada e contemplei os seus ombros descaídos[. . .]. (Baptista-Baptos, Cão velho entre flores, 109)‘He was leaning against the closed window and I contemplated his fallenshoulders’Direct object (3.4%)[...] inclinou-se, beijava-a na boca [. . .]. (Baptista-Baptos, Cão velho entreflores, 30)‘He leant forward and kissed her on the mouth’A first glance at the general statistical results mentioned and illustrated by theprevious lines allows us a first observation in connection with the “expression’’of the possessor in the Portuguese language: apparently, the absence of apossessor expression is preferred to its expression on noun phrase level (as apossessive determiner) on the one hand, and to its expression on the propositionallevel (as a DO) on the other. Parameters of a pragmatic and/or discursivenature assure its identification. This tendency of not expressing thepossessor in Portuguese can be more easily observed if we take into account anNPbp in syntactic subject function (Type: His eyes were blazing). In thissyntactic configuration, 201 out of 303 Portuguese items do not explicitly referto the possessor: 15(1) Ao fundo, com sentinelas à porta, estão dois pretos em duas camas, os olhosbrilham como brasas [. . .]. (M. Alegre, Jornada de África, 169)‘At the back, with guards at the door, there are two black people in twobeds, the eyes were shining like glowing embers’Actually, it is in this respect, i.e. non-expression of inalienability, that thePortuguese language is slightly different from both French and Spanish, whichare, from a typological point of view, relatively close.In my opinion, the data showing (or the fact) 16 that the Portugueselanguage prefers not to express the possessor and to rely on the context, shedsanother light on the problem of “expressing’’ inalienability, raising questionsabout what the possible influence could be of parameters such as determinationand “context’’. 17

(In)alienability and (in)determination in Portuguese 2313. Extralinguistic ‘context’: Brazilian PortugueseIn an earlier publication I have started to explore the possible impact ofextralinguistic context on Brazilian Portuguese syntax. In Brazilian Portugueseit is the absence of the definite article before the possessive form which causesdifficulties, 18 which, in my opinion, cannot be explained as a persistence offifteenth- or sixteenth-century Portuguese nor as an innovation of BrazilianPortuguese. 19Nevertheless, the analysis of a 2,700 item database 20 reveals so many individualdifferences between the authors 21 that the only thing we can confirm isthe fact that Brazilian authors with a solid education and living in large townsdo respect the tendency of omitting the definite article before the possessiveform. 22 We thus provisorily leave behind the sociolinguistic track 23 in order toreorient ourselves cautiously to the “intralinguistic context’’, i.e. the ‘text’, 24 andmore particularly, the macrosequential units 25 being Fernando Sabino’s tales. 264. Intralinguistic ‘context’: ‘discourse games’ between definite article,indefinite article and possessive determiner4.1 Point of departure: Fernando Sabino’s talesFernando Sabino’s tales hide, indeed, an interesting fact. It occurs to us thatthe following example develops what Givón (1987) calls “foreground information’’and “background information’’:(2) Desta vez, era o homem da televisão! Não era. Refugiado no lanço da escadaentre os andares, esperou que o elevador passasse, e voltou para a porta deseu apartamento [. . .]. Percebeu, desorientado, que estava sendo levadocada vez para mais longe de seu apartamento [. . .]. Depois experimentouapertar o botão do seu andar. (F. Sabino, O homem nu, 66–67)‘This time it was the television man! No, it wasn’t. Sheltering in the stairwellbetween the floors, he waited till the elevator passed and went back tothe door of his appartment [. . .]. He understood, disorientated , that eachtime he was being carried away somewhat further from his appartment[. . .]. Then he tried to press the button of his floor’This example shows that andar (floor) and apartamento (appartment) aretreated differently regarding the nature of their determiner. Probably, the class

232 Anne-Marie Spanogheof floors (os andares, o seu andar), presented as identifiable from the very start,can be held responsible for the use of the definite article before the possessiveform, whereas the item representing the class of appartments (“seuapartamento’’) is not presented as identifiable and is thus (?) considered to beless important in the progression of narration, its function having to belimited to the setting of the main character.In fact, this last observation constitutes a long investigation program. Ialready announced it by making use of conceptual tools such as ‘text’,‘narrative discourse’ and ‘identifiability’, terms the definition of which stillremains incomplete. On the other hand, the same observation departs fromthe hypothesis that the nature of the factors determining the functioning ofa given human natural language also has an influence on the type of itsdescriptive model. In the case of Portuguese a discourse model could be anadequate one.4.2 The discourse information-flow:from possessive determiner to indefinite articleNow that we have illustrated our theoretical and methodological premises, wecan finally attend to the expression of the so-called inalienable possessionrelationship in Portuguese.With reference to the possessive determiner, theoretically three forms canbe taken into account viz. o, omeu, meu which also gives rise to three morphologicaloppositions: 271. o vs. omeu2. omeu vs. meu3. o vs. meuThe first kind of opposition is mentioned — mainly by contrastive grammars— in the light of inalienability expression: o determines the NPbp when therespective possessor is presented as ‘known’and omeuis the determiner of theNPbp if the respective possessor has to be explicitly referred to. 28 Secondly, thepresence or absence of the definite article before the bare possessive formcould be a characteristic of both Brazilian and European Portuguese. 29 Finally,up to now we have not found any description opposing the definite article o tothe possessive form meu.But even if the productivity of the third kind of opposition still remainsopen to further investigation, checking briefly its possible existence could be

(In)alienability and (in)determination in Portuguese 233interesting for the elaboration of my investigation. In fact, I did find anexample in Brazilian Portuguese where the form meu is used in order tointroduce into the text a referent presented as unknown:(3) Um dia o capitão o desafiou com a sugestão de número inédito:– Um mágico já engoliu meu anel e depois devolveu.– Isso é fácil, também sei fazer: me dá o anel. (F. Sabino, O homem nu,173)‘One day the captain challenged him with the suggestion of an unknownact:– A magician has already swallowed my ring and then he gave it back.– This is easy, I’am also able to do this: give me the ring’ 30In my opinion, the previous passage suggests that the possessive form assumesa discursive function comparable to that of an indefinite article introducinginto the text a referent presented as unknown. 314.3 Indefinite article and ‘inalienability’So far, the role played by the indefinite article in the expression of inalienabilityhas not been thoroughly examined. 32 Generally, it is stated that within theframe of the actualisation theory the NPbp does not follow the expectedprocedure: 33 instead of being introduced in discourse preceded by an indefinitearticle, it immediately appears with a definite article. 34 However, thefollowing pages show that this “associative anaphor’’ occurs with some regularityonly if the respective prototypical possessor of an equally prototypical35, 36body part has already been introduced in discourse:(4) O rapaz alto e magro continuou a olhar para as mãos como se as estivessevendo pela primeira vez [. . .]. (L. de Sttau Monteiro, Um homem nãochora, 144)‘The tall and thin boy kept looking at the hands as if he was seeing themfor the first time’Given the preceding insights, it seems that we have taken the first steps in aninteresting direction which is defined by three crucial questions, each of themmarking a stage on what we would call the “(in)alienability continuum’’:Question 1: What could be the impact of a non-prototypical body part on thetype of determiner of the NPbp?

234 Anne-Marie SpanogheQuestion 2: Is definiteness of the NPbp the result of the respective possessor’spresence?Question 3: Does a non-prototypical possessor have an influence on theappearance of an indefinite article?Throughout the last stage of our investigation I will try to give some usefulanswers to these three questions.4.3.1 Non-prototypical body partsWhen we isolate the NPbp referring to a non-prototypical body part on thevery moment of its introduction into the text, 37, 38 we observe the existence ofat least three types of determiners, 39 indeed including the indefinite article(two items):(5) A porta da prisão rangeu. Os dedos trémulos da luz tactearam o chãoesponjoso, o visco negro das paredes e do tecto. Um homem fardado entrou,cerrou a porta nas costas.– É a hora, amigo? – perguntaram do escuro.– É a hora – responderam para o escuro.O homem fardado bateu os bolsos, raspou um fósforo na escuridão.–Trouxe uma vela – disse. – Gostava de te ver.E acendeu um toco de estearina, que destacou do negrume igual uma facelarga e pálida, uns olhos calmos, cansados de amar e de odiar. Os doishomens fitaram-se até se não estranharem. Gotas de água caíam, solitárias,do alto do tecto, empoçando na lama do chão.– Não vais demorar-te? – perguntou o preso. (Vergílio Ferreira, Saturno, ll.1–16)‘The door of the prison creaked. The trembling fingers of the lightexplored the spongy floor, the black viscosity of the walls and the ceiling.A man in uniform entered, closed the door behind him.– Is it time, my friend? – they asked from the dark.– It is time – they answered to the dark.The man in uniform slapped his pockets, lit a match in the dark.– I brought you a candle – he said. –Iwould like to see you.And he lit a bit of stearin, which made appear from the same dark a largeand pale face, a pair of calm eyes, worn out by love and hate. The two menfixed their eyes upon each other till they got acquainted. Drops of water werefalling, solitarily, from the ceiling, covering the muddy floor with water.

(In)alienability and (in)determination in Portuguese 235– You won’t be late? – asked the prisoner’I also registered 11 occurrences of the possessive determiner:(6) Fumei o meu cigarro triste de condenado. E fitando Artur na sua facecansada, nos seus olhos túmidos de sono, expliquei que também eu sofriacom o que estava acontecendo [. . .]. (Vergílio Ferreira, O jogo de Deus, ll.486–489)‘I smoked my sad cigarette of a condemned man. And fixing Arthur inhis tired face, in his drowsy eyes, I explained that I also suffered fromwhat was happening ‘And, finally, I once more noted the overwhelming presence of the definitearticle determining the NPbp at the moment of its introduction into the text(20 items):(7) O homem quebrou a cinza do cigarro, enterrou os dedos da mão livre nabarba suja e longa. (Vergílio Ferreira, Saturno, ll. 38–39)‘The man broke the ashes of his cigarette, put the fingers of the free handin the long and dirty beard’Obviously, a non-prototypical body part does not require the unconditionaluse of the indefinite article. On the contrary, in the majority of the cases thedefinite article is used.4.3.2 Non-prototypical possessorsMy second attempt at discovering parameters ruling the (exceptional) appearanceof an indefinite article as a determiner of NPbp is more fruitful andseems to suggest the possible existence of an “(in)alienability continuum’’.Actually, we could speak of a process in which the body part has increasingautonomy and rhematicity and the possessor is slowly fading away: indeterminateinalienability becomes alienability. The following examples may illustratethis process. In example (8), we have a prototypical possessor, a verb ofpossession and a NPbp occupying a rhematic position: 40(8) Pequenino, trouxera do retiro forçado uma barba cerrada e negra [. . .].(M. Torga, A criação do mundo, 388)‘When he was young, he had brought a tight and dark beard from theforced retreat’We encounter the same rhematicity of the body part when the NPbp assumes

236 Anne-Marie Spanoghethe OO function. 41 Example (9) shows that even if the respective possessor ispresent, he is presented as increasingly indeterminate:(9) E um vago primo negligentemente vestido, desengonçado, e com uma cabeçade melão, horrorrizou a sensibilidade do dandy copiado de Fradique. (M.Torga, A criação do mundo, 153)‘And a vague cousin negligently dressed, loose, and with a melon head,horrified the sensibility of the dandy copied from Fradique’Example (10), already mentioned in a previous paragraph (example (5) supra),however, still has a human possessor but he is presented as vague and indeterminate:(10) A porta da prisão rangeu. Os dedos trémulos da luz tactearam o chãoesponjoso, o visco negro das paredes e do tecto. Um homem fardado entrou,cerrou a porta nas costas.– É a hora, amigo? – perguntaram do escuro.– É a hora – responderam para o escuro.O homem fardado bateu os bolsos, raspou um fósforo na escuridão.–Trouxe uma vela – disse. – Gostava de te ver.E acendeu um toco de estearina, que destacou do negrume igual uma facelarga e pálida, uns olhos calmos, cansados de amar e de odiar. Os doishomens fitaram-se até se não estranharem. Gotas de água caíam, solitárias,do alto do tecto, empoçando na lama do chão.– Não vais demorar-te? – perguntou o preso. (Vergílio Ferreira, Saturno, ll.1–16)‘The door of the prison creaked. The trembling fingers of the lightexplored the spongy floor, the black viscosity of the walls and the ceiling.A man in uniform entered, closed the door behind him.– Is it time, my friend? – they asked from the dark.– It is time – they answered to the dark.The man in uniform slapped his pockets, lit a match in the dark.– I brought you a candle – he said. –Iwould like to see you.And he lit a bit of stearin, which made appear from the same dark a largeand pale face, a pair of calm eyes, worn out by love and hate. The twomen fixed their eyes upon each other till they got acquainted. Drops ofwater were falling, solitarily, from the ceiling, covering the muddy floorwith water.– You won’t be late? – asked the prisoner’

(In)alienability and (in)determination in Portuguese 237In the last two extracts, either the body parts are only remnants of theirpossessor:(11) Uns ossos encontrados em Maktan, Poro, foram examinados por Bailen, umantropólogo da mesma universidade. (Grande Reportagem, Nov. 1995, 44)‘Some bones 42 found in Maktan, Poro, were examined by Bailen, ananthropologist of the same university’or the body part (a pele ‘the skin’) is entirely autonomous and does not implyany relationship with a particular possessor:(12) [. . .] o ar estava macio, como se passasse pelo filtro duma pele humana[. . .]. (J. Saramago, Objecto quase, 127)‘the air was soft, as if it passed through the filter of a human skin’5. ConclusionNot entirely free from circular reasoning, characterizing almost every linguisticapproach, my investigation, which is organized from the emissor’s viewpointand based on empirical data from various Portuguese novels, raises somefundamental questions. First of all, I asked to what extent the nature of aparticular natural human language should/could determine the theoreticalmodel aiming at its description. On the other hand, adopting the emissor’spoint of view, we face the difficulties of formulating tendencies/rules on theinterface between conceptual/philosophical entities — in the present case‘inalienability’- and empirical data determination: 43 the only interface tendencybetween (in)alienability and (in)determination we found is that it iseasier to find the (exceptional) indefinite article introducing a NPbp if thereferent of that NPbp is non-prototypical, assumes a rhematic syntacticposition and whose possessor is indeterminate or simply does not exist.After all, it seems that the track I have initiated deserves some furtherthoughts about the possible link between (in)alienability /(in)determinationon the one hand, and about the measuring of the subjective extralinguisticexperience in the definition of these notions on the other. Or is it simplyimpossible to eliminate this aspect of circularity?

238 Anne-Marie SpanogheNotes1. Cf. Riegel (1984, 1991, 1992), Spanoghe (1995), Chappell and Mc Gregor (1996), Baronand Herslund (1997), Heine (1997) and a forthcoming workshop on the part-wholerelationship organized by G. Kleiber and M. Riegel.2. The original version is as follows: “La première classe (les noms qui prennent le suffixe) secompose des noms qui désignent les membres du corps, les parties d’une chose, les objets enrelation étroite avec un homme (ses armes, son filet de pêche, etc.), les relations de parenté, etde quelques prépositions exprimant des rapports dans l’espace: à côté de, au dessus de, près de,loin de, etc. La seconde classe comprend tous les autres noms. [...]〈la seconde classe secaractérise par〉 le fait que je possède quelque chose, à titre de propriété, et de propriétéaliénable’’ 〈par opposition à la propriété aliénable, Lévy-Bruhl suppose l’existence〉 d’unesolidarité qui ne peut se rompre, [...] une interdépendance, une participation’’ (Lévy-Bruhl1916: 99). By “〈〉’’ I distinguish my own paraphrasing from the original text; the bold is alsomine. For more details on the topic, see Spanoghe (1995).3. The original terms are vervreemdbaar (‘alienable’) vs. onvervreemdbaar (‘inalienable’)bezit (‘possession’). For more details see Uhlenbeck’s publication (1917: 346): “Van grootewaarde voor het bedoelde onderzoek is de omstandigheid, dat in een reeks van talenverschillende soorten van possessie door morphologische verschillen worden gereflecteerd. Welis het aantal dier grammatische onderscheidingen geenszins in overeenstemming met dewerkelijke betrekkingsverschillen, die tusschen den “bezitter’’ en het “bezetene’’ kunnen bestaan,maar althans één in het oog loopend essentieel verschil in den aard der feitelijke betrekkingwordt in die talen ook door den vorm te kennen gegeven, namelijk dat der onvervreemdbare ofintieme en der vervreemdbare of uiterlijke possessie.’’ [Emphasis mine, A-MS].4. In the original version: “sphère personnelle’’: “[. . .] comprend, ou peut comprendre, leschoses et les êtres associés à une personne d’une façon habituelle, intime, organique (p.ex. lecorps et ses parties, les vêtements, la famille, etc.’’; “solidarité’’: “[. . .] tout phénomène, action,état ou qualité qui affecte une partie quelconque de la sphère personnelle 〈et qui〉 retentit, parrépercussion spontanée, sur la personne tout entière.’’ (Ch. Bally 1926: 68).5. See Chappell and McGregor (1996) for an English translation of Bally’s article.6. This is probably due to the lack of any operational definition in intensio of“(in)alienability’’. For a good attempt, however, see Seiler (1983).7. Cf. Spanoghe (1995).8. I will include some observations in connection with Brazilian Portuguese in order toexplicitate my reasoning.9. The (conceptual) gap between the two descriptive levels is due to our search for anadequate conceptual framework for ‘inalienability descriptions’. (Many thanks to BillMcGregor for sending us his Semiotic Grammar, which we were not able to assimilatebefore the deadline of this article.)10. Cf. Fox (1981).11. This paragraph was inspired by Spanoghe (1995).12. 100% = 3,275items from contemporary Portuguese novels (see Spanoghe 1995).

(In)alienability and (in)determination in Portuguese 23913. We translated almost literally the “inalienability’’ passages of the examples in order toillustrate the particular position of Portuguese.14. This statistical result of Possessor expression clearly differs from its expression inFrench: S > possessive determiner > IO > pragmatic-discourse factors > DO and SpanishS > possessive determiner > IO > DO > pragmatic-discourse factors.15. See Spanoghe (1995). In Spanish only 38 out of 471 items do not mark the possessor inthe same syntactical configuration and in French we found 56 out of 604 items withoutreference to the possessor.16. Following Milner (1989), a fact is a data associated with the whole of the propositionsthe data falsifies.17. Cf. our terminological hesitations in connection with “pragmatic-discursive factors’’.18. Cf. Spanoghe (1998, Forthcoming).19. Invoking these parameters, one does not solve the question of the ancient language’ssources nor the problem of the constitution of creoles, cf. Teyssier (1976), Lucchesi (1993).20. This database records 2,700 items from 6 contemporary Portuguese novels, three ofwhich are written in Brazilian Portuguese, cf. Spanoghe (1998, Forthcoming).21. See Spanoghe (Forthcoming). By “individual differences“ we mean parameters ofidiosyncratic or diatopic nature.22. In our sample of 2,700 items, nearly 75% of the Brazilian items present the barepossessive form meu in contrast with 25% presenting the compound form with the definitearticle o meu see Spanoghe (Forthcoming).23. Cf. Spanoghe and Formiga (Forthcoming)24. We adopt Ducrot’s terminology: “What we call ‘text’ is mostly a discourse consideredas the object of a unique choice. The discourse end, for instance, has already been prefiguredby the author the moment he formulates the beginning“. The original version is asfollows: “Ce qu’on appelle texte c’est d’habitude un discours censé faire l’objet d’un choixunique, et dont la fin, par exemple, est déjà prévue par l’auteur au moment où il rédige ledébut.’’ (Ducrot 1984: 176).25. Brès (1994).26. The option I made to place the expression of inalienability in the context of Portuguesecontemporary narrative prose provisorily takes for granted that the principles underlyingdetermination of the NPbp are the same in European and Brazilian Portuguese. Otherwise,our observations about Brazilian Portuguese would not have had any significance in ourreasoning.27. This observation has some significance only if a functional opposition is at stake, cf.Martinet (1967: 27). For some thoughts about the ‘opposition’ between the definite article,the indefinite article and the possessive determiner in French, Spanish and Portuguese, seeSpanoghe (1996b).28. Still remains the question of what is ‘known’ exactly.

240 Anne-Marie Spanoghe29. In European Portuguese the absence of the definite article is recorded in the vocative,in some idioms, predetermining the word casa and before kinship terms. For furtherinformation, see Spanoghe (1997, 1998).30. See also example (2) seu apartamento. Concerning this example we suppose that thePortuguese language formalizes as ‘inalienable’ the referent of o anel ‘the ring’.31. Suffice it to recall that following Bally (1932) and Christophersen (1939) the indefinitearticle, unlike the definite article, serves to introduce into the text a referent presented asunknown.32. For a first attempt, see Spanoghe (1997).33. See Bally (1932), Christophersen (1939).34. See Spanoghe (1996a).35. See Kleiber (1994) for useful insights about “associative anaphors’’.36. Still remains the problem of defining exactly what is meant by a “prototypical bodypart’’ and a “prototypical possessor’’, cf. infra.37. For instance, a body part which does not define a prototypical human being- everybodyhas eyes, but not everybody has blue eyes. For some useful insights on the “prototypicity’’of a body part’s possessor, see Seiler (1983).38. The interested reader can always consult the global results of our database (VergílioFerreira, Contos) in Spanoghe (1998).39. We urgently need an investigation recording and analyzing all possible types ofdeterminers of NPbp.40. Cf. Chafe (1976), Givón (1976, 1984), Haiman (1983).41. The descriptive notion of rhematicity also depends on the emissor who, at any moment(?) in discourse flow, is able to reinterpret in a rhematic manner a referent which alreadyhas been presented as thematic: Em todas as mulheres que conhecera até ali via-lhesirremediavelmente o sexo por debaixo das saias. Um sexo que ostensivamente se oferecia ou senegava, em plena floração ou já murcho, ciosamente possessivo ou perfidamente complacente.Um sexo que não dava tréguas ao homem, a insinuar-se . . . (M. Torga, A criação do mundo,331) [On all the women he had known until then, he saw them inevitably the sex under theskirts. A sex which ostentatiously offered itself or negated itself, in full flower or alreadyfainted, hotly possessive or perfidiously benevolent. A sex which didn’t give any rest toman, insinuating themselves. . .]42. For more details about the relation between numerals, indefinite pronouns and articles,see Spanoghe (1997).43. See Corblin (1987: 7).

(In)alienability and (in)determination in Portuguese 241ReferencesBally, C. 1926. “L’expression des idées de sphère personnelle et de solidarité dans leslangues indo-européennes’’. In Festschrift Louis Gauchat, F. Fankhauser and J. Jud(eds), 68–78. Arau: Sauerländer.Bally, C. 1932. Linguistique générale et linguistique française. Bern: Francke.Banfield, A. 1995. Phrases sans parole. Paris: Seuil.Baron, I. and Herslund, M. (eds). 1997. Possessive Structures in Danish [KLIMT 3]. CopenhagenBusiness School.Brès, J. 1994. La narrativité. Louvain-la-Neuve: Duculot.Chafe, W. L. 1976. “Givenness, Contrastiveness, Definiteness, Subjects, Topics, and Pointof View’’. In Subject and Topic, C. N. Li (ed.), 25–57. New York-San Francisco-London: Academic Press.Corblin, F. 1987. Indéfini, défini et démonstratif. Genève: Droz.Christophersen, P. 1939. The articles: a study of their theory and use in English. Copenhagen:Oxford University Press.Ducrot, O. 1984. Le dire et le dit. Paris: Minuit.Fauconnier, G. 1984. Espaces mentaux. Paris: Minuit.Fox, B. 1981. “The syntax of body parts: towards a universal characterization’’. Studies inLanguage 5: 323–342.Givón, T. 1976. “Topic, Pronoun and Grammatical Agreement’’. In Subject and Topic, C.N. Li (ed.), 149–188. New York-San Francisco-London: Academic Press.Givón, T. 1984. Syntax. A Functional-Typological Introduction. Vol. I. Amsterdam andPhiladelphia: John Benjamins.Givón, T. 1987. “Beyond foreground and background’’. In Coherence and grounding indiscourse, R. S. Tomlin (ed.), 175–188. Amsterdam and Philadelphia: John Benjamins.Haiman, J. 1983. “Iconic and Economic Motivation’’. Language 59: 781–819.Kleiber, G. 1994. Anaphores et pronoms. Louvain-la-Neuve: Duculot.Lévi-Bruhl, L. 1916. ‘‘L’expression de la possession dans les langues mélanésiennes’’.Mémoires de la Société de Linguistique de Paris 19: 96–104.Lucchesi, D. 1993. “The article systems of Cape Verde and São Tomé Creole Portuguese:general principles and specific factors’’. Journal of Pidgin and Creole Languages8: 81–108.McGregor, W. B. 1997. Semiotic Grammar. Oxford: Clarendon Press.Martinet, A. 1967. Eléments de linguistique générale. Paris: A. Colin.Milner, J.-Cl. 1989. Introduction à une science du langage. Paris: Seuil.Roegiest, E. and Spanoghe, A.-M. 1991. “Relation de possession inaliénable et qualificationen français et en espagnol’’. Revue de Linguistique Romane 55: 81–94.Seiler, H. 1983. Possession as an operational dimension of language. Tübingen: Gunter Narr.Spanoghe, A.-M. 1995. La syntaxe de l’appartenance inaliénable en français, en espagnol et enportugais. Bern: Peter Lang.Spanoghe, A.-M. 1996a. “Elementos para uma interpretação discursiva do possessivoportuguês’’.Verba 23: 67–82.

242 Anne-Marie SpanogheSpanoghe, A.-M. 1996b. “La construction il lève la/sa tête en français, en espagnol et enportugais. Une approche fonctionnelle’’. In Actes du XX e Colloque International deLinguistique Fonctionnelle, 223–227. Liège-Belgium, 13–19 July 1995.Spanoghe, A.-M. 1997. “L’article indéfini portugais. Introduction à une étude sur son rôledans l’expression de l’appartenance inaliénable’’. Romanistik in Geschichte und Gegenwart3: 83–107.Spanoghe, A.-M. 1998. “L’expression de l’appartenance inaliénable en portugais, unproblème discursif?’’. In Actes du XXI e Congrès International de Linguistique et dePhilologie Romanes, Palermo, 18–24.09.1995.Spanoghe, A.-M. Forthcoming. “Meu Brasil eomeuPortugal.Lepossessif brésilien est-ilbrésilien ou portugais?’’. Cadernos de Estudos Lingüísticos .Spanoghe, A.-M. and Formiga, M. A. Forthcoming. “meu livro vs. o meu livro,brasileirismo ou arcaísmo?’’Uhlenbeck, C. C. 1917. ‘‘Het identificeerend Karakter der Possessieve Flexie in Talen vanNoord-Amerika’’. Verslagen en Mededeelingen der Koninklijke Academie van WetenschappenV, 2: 345–376. Amsterdam.Teyssier, P. 1976. Manuel de langue portugaise. Paris: Klincksieck.DatabaseAlegre, M. 1989. Jornada de África. Lisboa: Dom Quixote.Baptista-Bastos. s.d. Cão velho entre flores. Mira-Sintra:Europa-América.Faria, A. 1987. Emenda e soneto. Mem Martins: Europa-América.Ferreira, V. 1993. Contos. Venda Nova: Bertrand Editora.Melo, J. de 1990 7 . Gente feliz com lágrimas. Lisboa: Dom Quixote.Monteiro, L. De Sttau 1973 5 . Um homem não chora. Lisboa: Ática.Sabino, F. 1960. O homem nu. Rio de Janeiro-São Paulo: Ed. Record.Saramago, J. 1984. Objecto Quase. Lisboa: Caminho.Torga, M. 1991. A criação do mundo. Coimbra: Coimbra ed.

Chapter 13Possessives with extensive useA source of definite articles?Kari Fraurud1. IntroductionThat definite articles develop out of demonstratives is commonly accepted fora number of languages and language families. In this paper I want to raise thequestion of whether certain possessives can be another potential source ofdefinite articles. In particular, I will examine three of the Altaic and Uraliclanguages for which it is often claimed that their extensively used possessivesuffixes are “used as definite articles’’. Thus, the kind of constructions to beconsidered here is non-lexical attributive possession constructions, as in, forexample her cat as opposed to Ann’s cat (lexical attributive possession) and toshe has a cat or the cat belongs to her (predicative possession), 1 cf.:Predicative AttributiveLexical Ann has a cat Ann’s catNon-lexical she has a cat her catWithin non-lexical attributive possession I include not only constructions withpossessive pronouns, as in, for example, English (1a), but also their counterpartsin other languages, in particular, clitics or affixes that function aspossessive determiners or modifiers, as in (1b), cf.:(1) a. English: his hand poss.3sg.masc handb. Turkish: el-i hand-poss.3sgWhen a language with the latter type of non-lexical possessive constructionsalso has possessive pronouns, these are commonly used only optionally, forinstance, for emphasis, and in addition to the clitics/affixes. I will for short referto non-lexical possessive elements used in the unmarked possessive construction(as well as to the construction itself) as ‘possessives’, abbreviated ‘poss’. 2

244 Kari FraurudAnd, since most of the possessives relevant to the present discussion are in thethird person singular form, poss will in the following, unless otherwise stated,stand for third person singular possessive pronouns, clitics or affixes.In the following two sections I will give a background to and describe twohypotheses about possessives that are used extensively. Then I will considersome possible evidence for and against the hypotheses (Sections 4 and 5,respectively). I conclude with a discussion of alternative interpretations of theavailable data. Although some observations do indeed suggest that poss insome Uralic languages has developed into something that can be likened to thedefinite article in languages such as English, I will argue that it may be betterto avoid Eurocentric analogies and try to describe this poss in its own terms—asaposs with special characteristics distinguishing it from definite articlesas well as from poss in languages like English.2. Point of departure: Three observationsThe question posed in the title of this paper was inspired by three observationson the use of certain definite determiners. These concern: (a) diachronicpatterns in the grammaticalization of demonstratives into definite articles; (b)synchronic patterns of overlap in use between the definite article and demonstrativeand possessive determiners; and (c) the extensive use of possessives insome languages.2.1 Grammaticalization of demonstratives into definite articlesThe grammaticalization of demonstratives into definite articles has beendescribed — sometimes merely assumed — for a number of languages andlanguage families including Germanic, Romance, Bantu and Niger-Congolanguages (Greenberg 1978, 1991; but cf. Frajzyngier 1996). 3 As regards thepathway of grammaticalization, Greenberg mainly provides discussion of thelater stages of the process, where the article either vanishes or grammaticalizesfurther into a general marker of nominality or gender. About the beginning ofthe process we still have limited knowledge for many of the languages inquestion, and in the better analyzed cases there is some divergence in theaccounts of the earliest article-like uses of the demonstrative (see e.g. Laury1991, 1997; Epstein 1993, 1994; Himmelmann 1996, 1997). A particular use ofa determiner can of course be defined on different levels of generality, but in the

Possessives as a source of definite articles? 245present discussion it may be useful to refer to some of the usage types inHawkins’ (1978) well-known taxonomy of uses of the definite article in English: 4(2) a. direct anaphoric use: a car ...the carb. associative anaphoric use: a car ...the clutchc. immediate situation use: (in a car) Press the clutch!d. larger situation use: (in Denmark in 1998) ...the queenThe abbreviated examples illustrate some of the possible definite article uses inlanguages such as English. Thus, after mentioning a referent, e.g. a car, we may(2a) refer to the same referent, direct anaphora, or (2b) to something that is apart of that referent or otherwise associated with it, e.g. the clutch (of the car),associative anaphora. In the situational context of sitting in a car, we may (2c)refer to something present in the immediate environment, e.g. the clutch,immediate situation use, and, in the larger situational context of being inDenmark in the year 1998, we may (2d) refer to something that can beassociated with this context, e.g. the queen (of Denmark in 1998), largersituation use. In addition to these terms for different def-uses (that is, uses ofthe definite article or, more properly, of noun phrases determined by thedefinite article), I will use the term anchor, defined in Fraurud (1990) as anyelement of the linguistic or situational context other than the referent itself inrelation to which the reference of a noun phrase is determined. The notion ofanchor is relevant in cases of indirect (as opposed to direct) reference, whereit is possible to analyze the referent as being associated with or anchored to oneor more anchors. In the examples above, the anchor of the clutch in (2b) is ajust mentioned discourse referent, the car, and — depending on how far onewants to extend the notion of indirect reference — it is also possible toconsider the clutch in (2c) to be anchored to a discourse representation of theimmediately present car, and the queen in (2d) to the time-place anchorsDenmark and 1998. In terms of actual interpretation processes (as far as ourintuitions go), the indirectness and associativity of immediate and largersituation uses is often disputable, however. In particular when it comes toreferences to ‘unique’ entities like the moon or the sun (of the Earth), theirsimilarity to proper nouns (which refer directly) is striking. To capture this, Iwill talk about extended associative uses in cases where the anchor is somethingother than another discourse referent in the focus of attention (see Section 6);non-associative uses will refer to direct anaphora.Now, as regards the grammaticalization of demonstratives into definitearticles, my current interpretation of available data and analyses is that —

246 Kari Fraurudwith regard to the usage types described above — the process often startswith an extension from a (deictic) immediate situation use. From that maydevelop direct anaphoric use, or perhaps more generally non-deictic directreference in the sense of reference based on specific, shared knowledgeestablished either in the preceding discourse or by some earlier conversationor joint experience. 5 In the present discussion, I believe the essential propertyis that of directness of reference as opposed to indirectness and associativity.Furthermore, it seems clear that, in the extension of uses of demonstratives,usage type interacts with and is sometimes overruled by discourse parameterssuch as prominence or importance of the referent (cf. in particular Laury1991, 1997, and Epstein 1994).2.2 Overlap in use between different definite determinersThe second observation underlying the title question is synchronic and has todo with the overlap in use between different forms of semantically definitenoun phrases, viz. noun phrases with possessives (poss), with demonstrativedeterminers, abbreviated demonstratives (dem), and with definite articles(def) or, in languages without articles, bare noun phrases (Ø). 6 In this paper,I refer to all adnominal demonstrative and possessive pronouns as ‘determiners’,regardless of the possibility of analyzing (some or all) instances of theseelements in certain languages as modifiers rather than proper determiners (fora discussion, see Lyons 1986 and Plank 1992). 7 Furthermore, I will, for brevity,use def, dem and poss both for referring to the determiners and for theconstructions in which the determiners are used — definite, demonstrativeand possessive noun phrases.In some of the def-uses described in (2), we also find other determiners,such as dem or poss, and the point is that certain patterns can be seen in theseoverlaps. For example, def alternates with dem in direct anaphora, and withposs in associative anaphora, but not vice versa, cf.:(3) a. Beside the barn there is a little cottage. The/This cottage was built in1875.(but: *Its cottage. . .)b. Beside the barn there is a little cottage. The/Its roof is leaking.(but: *This roof. . .)These patterns of overlaps and non-overlaps are not confined to English, butare wide-spread among the world’s languages. For example, the occurrence of

Possessives as a source of definite articles? 247dem in direct but not associative anaphora is apparently so general that it issometimes used as a criterion for distinguishing dem from def (Himmelmann1996, 1997). Counter-examples to this generalization are continually debated(e.g. Kleiber 1990; Apothéloz and Reichler-Béguelin 1999), but could perhapsbest be described as exceptions that prove the rule, since their interpretationsalways seems to be marked in one way or another. Similarly, the occurrence ofan element classified as poss in direct anaphora might be argued to cast doubtabout its being a ‘true’ poss. In languages without a definite article, we find acorresponding alternation between Ø and dem in direct anaphora, andbetween Ø and poss in associative anaphora. The patterns of overlaps in usecan be seen both language internally and cross-linguistically. Cross-linguisticallyit becomes evident, for example, in translations, where one languageshows a stronger preference for def and another language for dem (see e.g.Jonasson 1998, 2000 and for a comparison of French and Swedish determineruse, and Aziz 1993 on English — Arabic).The patterns of overlap in use could be summarized as follows:(4) def (or Ø) ∩ dem: direct anaphoradef (or Ø) ∩ poss: associative anaphoradem ∩ poss: (null overlap)Variation and overlap of different forms is often a reflection of ongoinglanguage change such as, for example, the grammaticalization of dem into def(cf. e.g. Lehmann 1985). A certain use where we find a synchronic overlapbetween two forms will often turn out to represent the starting point for theexpansion of one of the forms into the other’s territory, initializing agrammaticalization process. Thus, it is no coincidence that dem and defoverlap in the direct anaphoric use, and that this use is an early extension areafor dem to develop into def. The question here is if — analogously — theoverlap in use between poss and def in associative anaphora may have somebearing on a possible grammaticalization process involving poss.2.3 Extensive use of possessives in some languagesThe third observation relating to poss as a potential source of definite articlesis again synchronic, or at least mainly so. The extensive use of poss in manyUralic languages is well known among Uralists; in descriptions of theselanguages it is repeatedly pointed out that the third person and sometimes —for instance in Komi (‘Zyryan’) and Mari (‘Cheremis’) — the second person

248 Kari Fraurudsingular possessive suffixes are used “determinatively’’, “as definite articles’’ or“instead of definite articles’’ (e.g. Collinder 1957, 1960: 203; Schlachter 1960;Krámský 1972; Hajdú 1975: 100, Hajdú and Domokos 1987:223; Rédei 1988;Uibopuu 1988:269; Leinonen 1998). The Uralist literature is abundant withexamples of more or less clearly non-associative uses of poss, such as in thefollowing Komi sentence (Rédei 1978a:77):(5) ver-as limj-is silema n’inforest-iness:poss.3sg snow-poss.3sg melted already‘In the forest the snow melted already’Similar observations have also been made for Turkish and other Turkiclanguages (e.g. Grönbech 1936; Nilsson 1985). Grönbech goes so far as toclaim that the Turkish poss suffix has already developed into an article. Thishas been rejected by Johanson (1991), among others, on the grounds that itsuse is not systematic. 8 In his study of definiteness and article systems inEurope, Bechert (1993) talks about “an area with definite articles: emergentdefinite articles which have nothing in common with demonstratives’’ — anarea which, as I understand it, would comprise most of the Uralic languages(with the exception of Hungarian and Mordvin) and apparently also Turkiclanguages (Bechert 1993: 13). Bechert also notes, however, that the article ine.g. Mari is optional and thus only ‘emergent’, in contrast to the ‘full-blown’article in for example German.An extensive use of poss has also been observed in one language which isgenetically and areally unrelated to the Ural-Altaic families, and which —‘already’ — has an (incipient) definite article originating in a demonstrative.Lehmann (1998:87) notes that Yucatec Maya “uses the possessed nominal [anNP determined by the proclitic/prefixed poss u — KF] in many situationswhere languages like English use a definite NP’’ (but see below for his furthercomments upon these uses), e.g. (Lehmann 1998: 87, example 97): 9(6) yàan a t’an-ik u h-men-ildeb sbj.2 call-incmpl poss.3 m-shaman-rel‘you have to call the (pertinent) shaman’3. HypothesesIf we consider the quantitatively and qualitatively extensive use of possobserved in some languages, and a possible analogy with dem in other lan-

Possessives as a source of definite articles? 249guages illustrating how a synchronic pattern of overlap in use may be areflection of a diachronic pattern of grammaticalization, we may arrive at thefollowing hypotheses:(7) H1: poss may — just like dem in some languages — grammaticalize intodef.H2: If poss grammaticalize into def, the process starts by an extensionwithin associative anaphora — not, as in the case of dem, with directanaphora.In order to evaluate these hypotheses, I have tried to follow up and check allclaims about possessives “used as definite articles’’ (or the like) found in aliterature survey of works on determiners, including general overviews such asKrámský (1972) as well as over 500 articles on specific languages or languagegroups. Among the languages mentioned in this connection, there are somefor which I have come to the conclusion that they — for separate reasons —are less likely candidates for languages with poss grammaticalizing or havinggrammaticalized into def. Two such languages are Armenian (Krámský 1972;Comrie 1981:182) and Tok Pisin (Sankoff and Mazzie 1991). I will leave theseout of the discussion here, and instead consider some of the more ‘promising’candidates.Thus, I will in the present discussion focus on four languages from threelanguage families for which I also have access to some primary data: Komi andUdmurt (‘Votyak’), two Permic languages that may represent several languagesin the Uralic family with poss claimed to be used as def; Turkish,representing several other Turkic languages and possibly other languages inthe Altaic language family; and, finally, Yucatec Maya, which as far as I knowis the only Mayan language to exhibit this phenomenon. I will also brieflyconsider the semitic language Amharic, for which I have only secondaryinformation, but which is claimed to have definite articles deriving frompossessives.In the next two sections I will discuss some possible evidence for andagainst the hypotheses by examining currently available data on theselanguages in some more detail and in the light of some things that weknow about grammaticalization processes in general (see e.g. Hopper1991; Heine, Claudi and Hünnemeyer 1991; Hopper and Traugott 1993),and the grammaticalization of demonstratives into definite articles inparticular.

250 Kari Fraurud4. Possible evidence for the hypothesesThe claims in the literature about possessives being “used as definite articles’’certainly reflects some truth; in the languages in question, possessives areindeed used extensively both in a quantitative and a qualitative sense. Theword ‘extensive’ is of course relative (that is, extensive in comparison towhat?), and it may eventually turn out that it is inadequate for describing whatis going on in these languages. But let us for the moment go on talking about“extensive use of poss’’ in the admittedly Eurocentric sense of being extensivecompared to the use of poss in languages such as English.4.1 Bound, formerly free, possessive morphemesAs we know from grammaticalization processes in general, change in meaningor use is commonly accompanied by morphological change from free forms toclitics and from clitics to affixes (e.g. Heine, Claudi and Hünnemeyer1991:213). It is therefore of interest to note that, in all the languages claimedto use poss extensively and non-associatively, poss is in the form of a clitic oran affix. At least in the case of Turkish and other Altaic (Baskakov 1975) andUralic (Janhunen 1981:31; Raun 1988: 560; Décsy 1990:67) languages poss isassumed to have developed out of a personal pronoun (which, according toRaun, in Proto-Uralic already was enclitic). For no language with a free possmorpheme I have found claims about an extensive poss use.4.2 High discourse frequency of possessivesOne common manifestation of an ongoing grammaticalization process is anincreased discourse frequency of the form in question (e.g. Heine, Claudi andHünnemeyer 1991). In the absence of diachronic data for a particularlanguage, it may be possible to use discourse frequency either as a criterionfor whether a particular form is fully grammaticalized or not, or, moregenerally, as a measure of how far a particular grammaticalization process hasgone. The discourse frequency of a category in one language is then comparedto the frequency of related categories in other languages. Such amethod of cross-linguistic quantification is proposed by Cyr (1993), who, onthe basis of cross-linguistic discourse frequency data, argues that the so-calleddemonstratives in the Algonquian language Montagnais in fact should beconsidered definite articles.

Possessives as a source of definite articles? 251In the present discussion the relevant quantitative measure is of course —just like in studies of grammaticalizing dem — the discourse frequency of possrelative to that of other forms of NPs. For languages other than the wellresearchedIndo-European languages it is often difficult to find discourse data,in particular data based on larger text corpora. The following data on discoursefrequencies of poss are rather limited, but may still, I think, give anindication of the considerable quantitative differences between English andsome languages that I here characterize as exhibiting an extensive use of poss.For Komi I do not yet have access to more than fragmentary discoursefrequency data, but for its sister language Udmurt there are statistical dataavailable from a corpus study by Suihkonen (1990). In Suihkonen’s corpus,the possessive suffix is found in 331 of 1,122, i.e. 30%, of the subject NPs andin 231 of 571, i.e. 40%, of the object NPs. In a Turkish corpus of written nonfictionprose, I found that 1,014 of 3,492 nouns, i.e. about 30%, had the thirdperson singular possessive suffix. 10 (For comparison it may be mentioned thatthe same corpus contained 46 poss: 3P, one singular and eight plural firstperson possessive suffixes and no second person possessives.) Finally, in asmall Yucatec Maya corpus of oral narrative I counted 52 third person singularpossessive prefixes on 263 nouns, i.e. about 20%. 11 Even though some of thesefigures are based on restricted data, they are notable in cross-linguisticcomparison. All these discourse frequencies of poss differ drastically fromthose in languages such as Swedish and English, where the frequency of NPswith poss (a third person singular possessive pronoun) is seldom more than afew percent. In a 85,000 word corpus of written non-fiction Swedish, less thantwo percent of the 19,329 lexical NPs, were poss (N=308, i.e. 1.6%, if weinclude reflexive and relative pronouns which, however, may also have pluralreferents, and N=45, or 0.23%, if we exclude these). 12 In English, poss is usedmore frequently, but it is still not comparable to, for instance, Udmurt. A500,000 word corpus of English fiction and non-fiction texts contained 7,739instances of the forms his, her and its including uses of his as an independentpossessive and of her as an object pronoun. 13 Even if we would assume that asmany as 7,000 of these forms actually are poss, and that the corpus containsno more than 100,000 lexical NPs, poss would not constitute more than aboutseven percent of the lexical NPs.Despite the limitations of these data and the diverse and somewhat roughmeasures used in the calculations of proportions, these figures suffice toillustrate the very high discourse frequency of poss in the languages underconsideration as compared to other languages. In Section 5.5, however, I will

252 Kari Fraurudreturn to this issue with a discussion of how to further evaluate thesequantitative data.4.3 Non-associative and extended associative uses of possessivesAbove we saw two examples of non-associative uses of poss, fromKomiandYucatec Maya. Both could presumably in Hawkins’ terminology be describedas larger situation uses, that is, as one kind of extended associative use. InKomi and Udmurt we also find examples of poss in direct anaphoric use, anon-associative use, as in the following example from the Southern Permyakdialect of Komi (Rédei 1978b: 474):(8) et-piriś sećće woktis ruć. rućis čig.once then came fox fox-poss.3sg hungry‘Once the fox [Ø] came that way. The fox [poss] was hungry.’According to Hypothesis II, this would indicate a fairly advanced stage in adevelopment of poss into def, since direct anaphora would represent asignificant extension of the original possessive or associative meaning. Forboth Turkish and Yucatec Maya, however, the use of poss appears to be lessextensive in qualitative terms. In written Turkish texts I have not found anyexamples of clearly article-like non-associative uses of poss. But spokenTurkish seems to differ somewhat from written in this respect. Schroeder(1999), provides the following example from a conversation in a tailor shop(Schroeder 1999: 131 f., example 15): 14(9) A: şimdilik içer-de beş kişi çalış-ıyor-uz.at present inside.loc five person work-pres-1platölye-ler-e dağıt-ıyor-uzworkshop-pl-dat distribute-pres-1plbiz yalnız kesim bölüm-ü-yüzwe only cutting department-cp-1plB: dikim-in-i de şey-de yap-tır-ıyor-sunuzsewing-ap-acc also thing-loc make-cause-pres-2plA: idare edi-yor-uz tabii. . . dışarı-sı da bizim. . .organize-rp certainly outside-rp also ourA: ‘At present there are five of us working inside.We distribute (the work) to the workshops [Ø]. We are only thecutting department [poss].’

Possessives as a source of definite articles? 253B: ‘So you have the sewing [poss] also done in whassit.’A: ‘We organize that, certainly. What-is-outside [poss] also belongsto us.’The first instance of poss is a compound marker (see below), but the secondand third can be described as instances of extended associative anaphoric use, asthere is no ‘possessor’, or anchor, in the close context (which becomes evidentif we try to replace the article by its in the English translation, e.g. ??its sewing).In Yucatec Maya, more clearly non-associative uses of poss, as in example(6) above, do not appear to be very frequent. Lehmann points out that “inmost of the examples there is, in fact, a part-whole, a member-of-set [. . .] orat least an associative relationship [. . .] between the possessed noun andsomething in the context’’ (Lehmann 1998:87). The following example is froma narrative about the building of a house, a whole which according toLehmann has not been mentioned for a while when reference is made to itsparts (Lehmann 1998:87, example 98):(10) [‘Now you have found four pillars; you bring them. When this is ready,’]k-a kaxt-k u báaloh-ilyou searchposs cross.beam-relka’ ts’íit báaloh y-éetel u pàach-nah-iltwo cl.long cross.beam 0-with poss back-house-rel‘you search the cross-beams [poss]; two cross-beams and the girders [poss]’In this and similar examples, Lehmann claims that “the relevant whole [. . .] istextually too far away for anaphora’’ and that what the poss does is rather thatit “associates the possessed noun to the present universe of discourse’’ (Lehmann1998:87, my emphasis). This would again be an example of an extendedassociative anaphoric use where, in English, the use of a possessive pronoun itswould not be permissible.5. Possible evidence against the hypothesesAlready in some of the examples presented above, we began to see differencesbetween the Permic languages, Turkish, and Yucatec Maya that seem to callfor a more complex account of the phenomenon of extensively used possessives.Let us consider some further observations that may suggest that thepredictions in hypotheses (7) are wrong or at least too simple.

254 Kari Fraurud5.1 No obligatorinessObligatorification is often described as an essential aspect of grammaticalizationprocesses (e.g. Greenberg 1978; Lehmann 1985). That a certainconstruction has become grammatically obligatory in one or more of its newuses may or may not be taken as a criterion for grammaticalization, or as anindication of how far the grammaticalization process has developed. Therelevance of obligatoriness to considerations about grammaticalization, Ibelieve, differ from language to language, and possibly even among categorieswithin one language.In none of the four languages is poss obligatory in any of the extendedassociative or non-associative uses. Thus if there is an on-going grammaticalizationprocess of poss developing into def, one might suggest that it has notgone very far. On the other hand, in Komi, noun phrase morphology in generalis often optional (Baker 1985). More importantly, however, for two of thelanguage families under discussion, some historical data is available that castsfurther doubt upon the likelihood of an ongoing grammaticalization process.5.2 Historical data suggest time stabilityAs regards the Uralic languages, poss is claimed to have been employed in‘non-possessive’ definite uses already in Proto-Uralic (which just like mostmodern Uralic languages had no article), cf.: “The [possessive] suffixesdenoting one owner, especially sg. 3., were probably also used as generaldefining elements’’ (Janhunen 1981:32). Décsy’s (1990) claim is even stronger:“The definiteness of the noun was expressed by the possessive endings (withoutpossessive meaning)’’ (Décsy 1990: 81, cf. also 66). The statements areinteresting, although it is important to remember that they necessarily derivefrom reconstruction based on comparisons of (more or less) contemporarylanguages. In the case of Turkish, Ergin (1983: 225) notes that in Old Turkish,spoken up to the thirteenth century, the use of poss was even more extensivethan in modern Western Turkish. This is also an observation that may berelevant to our present discussion, although we would like to see more detailabout the data and analyses behind the statement. What seems fairly clear,however, is that, in general, the morphology of Uralic and Altaic languages “isextremely resistant to both internal and external changes, [while] the syntax ismore vulnerable to influences and consequently to changes’’ (Menges1968:182, cf. also Baker 1985). 15

Possessives as a source of definite articles? 2555.3 Continued possessive use of possessivesIn the four languages under discussion, there seem to be no signs of possloosing its possessive meaning; besides its other possible uses, poss continues tobe the main means of expressing non-lexical attributive possession. This is initself no argument against an ongoing development from poss to def, since itis common in grammaticalization processes that change in meaning or useprecedes change in form, resulting in a stage where the original meaning coexistswith the new, more grammatical, meaning (Heine, Claudi and Hünnemeyer1991: 212f.). But if there is something to the analogy withdemonstratives, one would perhaps expect the upsurge of a new poss candidate,which could eventually replace the weakened old poss. One possibledevelopment would be that poss further grammaticalizes into def and graduallylooses its possessive meaning altogether, and concurrently a new form orconstruction arises as the unmarked expression of non-lexical attributivepossession —agoodcandidatebeingthecurrentlyoptional free possessivepronouns. In the Finno-Ugric languages as a group, a diachronic developmentfrom synthetic towards analytic possession constructions has indeed beenobserved. However, it appears that this development is strongest in some of thelanguages not claimed to have an extensive use of poss, e.g. Finnish, rather thanin e.g. Komi or Udmurt.5.4 Simultaneous extensive use of demonstrativesInterestingly enough, an extensive and non-associative use of poss may alsoco-occur with an extensive use of dem. This is the case in some Udmurt textsthat I have examined (from Wichmann 1901). 16 One of the texts is a fairy talenarrated in the Malmyzh-Urzhum dialect (Wichmann 1901: text 7). The storyhas a cyclic structure common to certain fairy tales in many cultures, where asimilar sequence of events is repeated with a replacement of some element, inthis case the daughter replacing the step-daughter. Let us look at a shortexcerpt from the first cycle of the story (11a) and its corresponding part in thesecond cycle (11b). (The form of the NP is indicated within brackets in theEnglish translation; absence of poss implies basic, non-possessive, declensionof the noun):

256 Kari Fraurud(11) a. (from Cycle I)so pits’i korka-n vylem odig peres’ kyshnothat little house-ine aux one old womanso peres’ kyshno so nyl min’ts’o estyny kosemthat old woman that girl sauna to.heat orderednyl-yz min’ts’o-ze estemgirl-poss.3s sauna-poss:acc heatedno so peres’ kyshno [. . .] shuemand that old woman [. . .] said‘In the little house [dem] there was an old woman [‘ONE’].The old woman [dem] ordered the girl [dem] to heat the sauna [Ø].The girl [poss] heated the sauna [poss:acc],and the old woman [dem] said: . . .’b. (from Cycle II)nyl so pits’i korka pyremgirl that little house enteredno ótyn vylem peres’ kyshnoand there aux old womanso peres’ kyshno so nyl-ez min’ts‘o estyny kosemthat old woman that girl-acc sauna to.heat orderednyl min’ts’o-jez estem, no peres’ kyshno [. . .] shuemgirl sauna-acc heated and old woman [. . .] said‘The girl [Ø] went into the little house [dem], and there was an oldwoman [Ø].The old woman [dem] ordered the girl [dem] to heat the sauna [Ø].The girl [Ø] heated the sauna [acc], and the old woman [Ø] said: . . .’The extracts illustrate the direct anaphoric use of both poss and, here frequently,dem — neither of which appears to be obligatory (cf. especially CycleII). They also exemplify the use of odig ‘one’ for introducing prominentcharacters (which may or may not represent a first step towards an indefinitearticle, cf. Givón 1981). The whole story contained 112 NPs (NPs in directspeech excluded) of which as many as 25were determined by poss and 27 bydem (Table 1). Without making too much out of such a small sample, it is still

Possessives as a source of definite articles? 257Table 1. Udmurt (Malmyz-Urzhum dialect): Distribution of lexical NP forms withregard to (i) prominence of discourse referent, and (ii) first/second cycle in a cyclicfairy taleLexical NPs ‘one’ N N-gen N N-poss ‘that’-N N TotalCycle I Main characters (N=2) 2 1 7 3 518Secondary ch. (N=1) 1 0 0 8 0 9Other participants (N=3) 2 52 4 3 16Props (N=17) 0 0 9 3 9 21total 5 6 18 18 17 64Cycle II Main actors (N=2) 0 2 1 510 18Secondary ch. (N=1) 0 0 0 1 7 8Other participants (N=3) 0 0 1 1 0 2Props (N=17) 0 0 52 13 20total 0 2 7 9 30 48interesting to note (i) that the two main and second characters (girl and oldwoman) are introduced by ‘one’ and subsequently referred to by means ofeither dem or poss (the latter also sometimes used with props) 17 , and (ii) thatthe NPs in the first cycle of the story are in general more morphologicallycomplex than those in the second cycle. Thus, the choice of NP form appears tobe governed not only by the status of the discourse referent, but also by thestatus of the particular discourse segment in which the reference occurs, a factorwhich has been shown to play a role in other languages (e.g. Cain 1991). 18Also as regards the use of dem, the two Permic languages differ considerablyfrom Turkish as well as from Yucatec Maya. Turkish, as far as I have been able toestablish, does not seem to employ dem in any article-like way, and the discoursefrequency of dem in the written corpus mentioned above is comparable to thatof e.g. written Swedish. In the Turkish corpus, 71 of 3,492 nouns were precededby dem, and in the Swedish corpus also mentioned above 422 of 19,329 lexicalNPs were determined by dem — i.e. about 2 per cent in both languages.Yucatec Maya represents an entirely different case, since it has, concurrentwith the extensively used poss, a likewise frequently used (possibly incipient)definite article originating in a demonstrative (Lehman 1991, 1998), togetherwith a uncommonly rich set of demonstratives.5.5 ‘Syntactic’ non-associative uses of possessivesAs is generally the case, gross quantitative data cannot be properly evaluatedwithout careful qualitative analyses. Thus a mere high discourse frequency of

258 Kari Fraurudposs may conceal a variety of different uses of poss, some of which have littleif anything to do with definiteness. Consider, for example, the following twouses of poss in Turkish:(12) a. kız-ın okul-ugirl-gen school-poss.3sg‘the girl’s school’b. kız okul-ugirl school-poss.3sg‘girls-school’In Turkish, poss occurs obligatorily on the head noun in possessor-possesseeconstructions (12a), a use which might be characterized as agreement (cf.Schroeder (1999) who, however, among his Turkish ‘agreement possessives’includes also those with an anchor referred to outside the NP). Anotherobligatory use of poss is the clearly non-referential head marker in most kindsof compounds (12b). Similar poss uses are found also in Komi and YucatecMaya, although perhaps not to the same extent. Thus poss in Komi occurs onthe head noun in possessor-possessee constructions, while in Yucatec Mayaposs occurs (non-obligatorily) in compounds. The space allotted to this paperdoes not allow a full discussion of these and other uses of poss which shouldprobably not be included in discourse frequency counts aimed at measuringthe possible progression of a grammaticalization process. Let it suffice to pointout that it is important to distinguish uses of poss that are governed bydiscourse parameters like referentiality, definiteness, prominence and topicality,from uses determined by morphosyntactic rules. I find it likely that, forTurkish — and at least to some extent the other languages under discussion —the latter kind of uses may both account for a large part of the high frequencyand explain many of the claims in the literature. More conclusive judgementson this issue must await further careful qualitative analysis and characterizationof different poss uses, followed by more detailed discourse frequencycounts (for a valuable contribution as regards the uses of poss in Turkish, seeSchroeder 1999).6. Concluding discussion: Two diachronically stable types of possessives?Let me try to sum up the possible evidence for and against the hypothesesformulated in (7). By analogy with dem grammaticalizing into def, we might

Possessives as a source of definite articles? 259expect the following manifestations of a grammaticalization of poss into def:a. change in form from free morphemes to clitics and affixes,b. increased discourse frequency of poss,c. employment of poss in certain def-uses (cf. dem: associative anaphora),d. obligatoriness in some of these uses,e. emergence of new forms of poss, gradually replacing the old affixedpossessives, andf. loss of essential semantic/pragmatic features (cf. dem: proximity distinctions).Due to insufficiency of diachronic as well as discourse data, it can be difficultif not impossible to ascertain whether a particular language displays one ormore of these phenomena. Before turning to the evaluation of these hypothesesin relation to my present sample of languages, I would like to brieflymention one language that is not included in my sample because of lack ofprimary discourse data, but which might actually be an even stronger candidatefor a poss-def development. It is the Semitic language Amharic, forwhich there are also some diachronic data suggesting a development fromthird person possessive suffixes to what the grammars today call ‘definitearticles’ (Armbruster 1908; Leslau 1967, 1995; Kapeliuk 1994). Withoutfurther primary data, it is, however, not possible to determine whether theAmharic poss-def is more similar to English-style def than to poss in thefour languages of my sample (poss Type II, see below). All things considered,however, I suspect that Amharic poss-def may indeed be a different story.To get back to the four languages under investigation in this paper —Komi, Udmurt, Turkish, and Yucatec Maya — the following may be notedwith respect to the hypotheses in (7):a. In all four languages poss takes the form of a clitic or affix, that probablydeveloped from a free morpheme. A change in form has thus already takenplace, but there is no evidence that this change in form would have beenassociated with a change in use.b. All four languages seem to have a very high discourse frequency of poss ascompared to Swedish and English. But a high frequency need not imply anincrease in frequency over time; on the contrary, some historical data suggesttime stability both as regards form and meaning. Furthermore, only a fractionof the many poss occurrences are likely to fall within the domain of def-uses.(This is probably different for different languages. Here my data are stillinconclusive, however.)

260 Kari Fraurudc. Among the ‘article-like’ uses of poss, we saw examples of an extendedassociative anaphoric use, in accordance with hypothesis II, and of the largersituation use, which could be seen as a further extension — an anchoring intothe situational context itself. But in Komi and Udmurt we also found directanaphoric uses. And furthermore, at least in Udmurt we observed a simultaneousextensive use of dem apparently governed by very much the samediscourse parameters as the extensive poss use.d. In none of the languages is poss obligatory in any of these uses. Nonobligatorinessdoes not exclude the possibility that a new category has developedor is developing, but it is a property that differentiates poss from definitearticles in general. The observation is also interesting in the light of thehistorical remarks on Uralic languages, which suggest a scenario where the socalled poss is a morpheme that has grammaticalized into its present form anduse already in the Proto-Uralic period. This supposition is also supported bythe observation on grammaticalization in general that change in meaning anduse tends to precede change in form (Heine, Claudi and Hünnemeyer1991:213).e. None of the four languages with extensive poss use seems to show astronger tendency than their sister languages with less extensive poss use todevelop new forms of non-lexical attributive possessive constructions, forinstance, replacing synthetic by analytical constructions.f. The employment of poss in some of the extended associative and nonassociativeuses does suggest absence of certain semantic/pragmatic featuresthat are always present in English its, for instance. But, just as in the case ofhigh or increased discourse frequency, absence need not imply loss, since wedo not know what was there in the beginning. Nevertheless, I believe that, inthe case of extensively used poss, the essential semantic/pragmatic propertyhas to do not with possessivity but with referentiality and, more specifically,with focus of attention.This last point requires some elaboration. Recall examples (9)–(10) above onextensive associative anaphoric use, and Lehmann’s remark that what I wouldcall the anchor is “to far away for anaphora’’ (1998:87). What this means isthat the poss in question does not obey any constraint on (anaphoric) pronounswith regard to focus of attention. Our focus of attention is assumed tohold what we at each moment (are able to) keep in working or short-termmemory (as opposed to long-term memory). A focussed referent is oftendefined as one that can be referred to by means of a pronoun, and (rather

Possessives as a source of definite articles? 261circularly) referring to something in focus is seen as a constraint on pronominalelements. I would argue that an essential — perhaps the essential —difference between poss in a language such as English and poss in a languagesuch as Komi is that the former but not the latter obeys the focus constraint.Without going into the details of this constraint, I submit that this criterioncan be used for differentiating between two types of poss, as per (13):(13) The focus constraint as a watershed between different types of poss:(i) poss Type I has to have an anchor in the form of a referent in the focusof attention. (ii) poss Type II, by contrast, may be anchored to nonfocussedor implicit referents (cf. extended associative anaphora), tocontextual elements like time and place (cf. larger situation use), tosecond and third order entities (cf. discourse deixis, abstract objectreference), or even merely to the linguistic or situational context (directanaphora and immediate situation use).As regards discourse deixis or abstract object reference, the cases I have inmind are constructions with poss ‘referring’, as it were, to events, propositions,and the like, as in the following constructed Turkish examples:(14) Düş-tü-m. Sonra-sı-nı bil-me-m.fall-past-1sg after-poss.3sg-acc know-neg-1sg‘I fell. I don’t know what happened after (that)’ (lit.: ‘its after’, ‘the afterof it’)(15) Gel-e-me-di.come-pot-neg-pastDaha doğru-su, gel-mek iste-me-di.more right-poss.3s come-inf want-neg-past‘He couldn’t come. Or, more correctly (than that), he didn’t want tocome’ (lit.: ‘its more right’)A common characteristic of the range of uses of poss Type II, I think, is thatassociativity is more essential than referentiality; in some cases it does not evenmake sense to talk about poss Type II as referring. This property of possType II has been partially captured in other words by Sinor (1978: 264) in his“definition’’ of the third person possessive in Uralic and Altaic languages:It is a morpheme which indicates that in the speaker’s mind the concept expressedby the word to which it is added is linked to another concept. It is important toremember that this other concept may or may not be expressed in the same

262 Kari Fraurudsentence. Indeed, it may not be expressed at all but simply implied by circumstanceswhich, on occasion, may be extralingual.The picture that emerges from this summary is of a poss Type II that (i) showsa remarkable time stability in form, use, and meaning, (ii) has manifold uses,including ones that are neither possessive nor definite, (iii) does not seem toprevent demonstratives from simultaneously occurring extensively in def-usesand eventually develop into articles, and (iv) might perhaps never develop intoan article. As regards the last point, I would like to connect to the suggestionabove about the possibly Eurocentric bias behind the notion of a poss used“extensively’’ and “as a definite article’’, and conclude by quoting Bechert(1993:37f.):But in this continuum [between the possessive suffixes of the second and thirdpersons and an emerging definite article — KF], the end seems to be neverreached, at the eastern margins of Europe and in Northern Asia, the definitearticle remains a category in statu nascendi. It might even be the case that thiscategory in Eurasian languages is a product of our Eurocentric perspective. If wereverse the perspective, we can view the European category of the definite articleas a special case of a category of belonging which is denoted indirectly, in Europe,by weakened demonstratives and would be more aptly expressed by possessivesuffixes, as it is in Northern Asia and its western outskirts.With the noteworthy exception of Bechert (1993), accounts of the developmentof definite articles often impart the impression that the lack of articles ina language is something that it will strive to make amends for, either bydeveloping articles or by using some other means for signaling definiteness. Inmy view, there is no reason why this should be so. After all, a clear majority ofthe world’s languages do fine without articles (e.g. Dryer 1989), and many ofthe purported other definiteness markers could, I think, be accounted for inan alternative way. But this issue takes us far beyond the scope of this paper.Without having been able to provide a conclusive answer to the questionposed in the title of this paper, I hope to have taken a first step towards a moreaccurate and hopefully less Eurocentric account of the uses of poss and otherdeterminers in languages such as Komi.AcknowledgementsI wish to thank Östen Dahl, Orin Gensler, Päivi Juvonen, Anne-CharlotteRendahl, and Marketta Sundman for suggestions on various points in earlier

Possessives as a source of definite articles? 263versions of this paper, and Bill McGregor for illuminating comments anddiscussion and for correcting my English. I am of course sole responsible forall the remaining deficiencies. The work reported in this paper has beencarried out within a project financed by the Bank of Sweden TercentenaryFoundation.Notes1. Alternative terms for predicative and attributive possession are verbal and (ad)nominalpossession, respectively (cf. Heine 1997).2. Cf. Manzelli’s (1990) ‘possessive adnominal modifiers’.3. Contra Greenberg, Frajzyngier claims that there are many languages, e.g. among Chadiclanguages, in which definite markers neither have or can be shown to have had a deicticfunction, and many languages in which deictic and definite markers are unrelated.4. The taxonomy is, however, not without its problems. For a critical discussion, see e.g.Fraurud (1990). I have replaced Hawkins’ term strict anaphoric by the more commonlyused direct anaphoric.5. Cf. Himmelman (1996, 1997), who — in contrast to the dominant view — suggests thatdefinite articles develop, not from the anaphoric, but from what he calls the ‘recognitional’use, characterized as follows: “the intended referent is to be identified via specific, sharedknowledge rather than through situational clues or reference to preceding segments of theongoing discourse. A central feature of this use is that the speaker anticipates problems withrespect to the information used in referring to a given referent. [. . .] Such use could alwaysbe (and in fact often is) accompanied by a you know? or remember?-type of tag question’’(1996: 230).6. The symbol ‘Ø’ is merely an abbreviation for ‘bare NP’, and should not be taken to implyan analysis of bare NPs as having a zero determiner rather than having no determiner.7. Alternative terms for demonstrative determiners are adnominal demonstratives anddemonstrative adjectives.8. “[. . .] es klar scheint, daß das Possessivsuffix in keiner Türk sprache eine systematischeRolle im Sinne eines bestimmten Artikels spielt’’ (Johanson 1991: 225).9. deb=debitive, incmpl=incompletive, M=masculine, and rel=relativizer.10. The corpus is a morphologically tagged approximately 7000 word text on archeology,prepared by The Turkish Natural Language Processing Initiative, at Bilkent University,Middle East Technical University, and Halici Computing Inc.11. The corpus consists of transcriptions by W. F. Hanks used as a course material. I thankAnne-Charlotte Rendahl for making these texts available to me and for analyzing the data.12. The corpus is Skrivsyntax: professionell prosa (‘The syntax of writing: professionalprose’), Lund University (see Teleman 1974).

264 Kari Fraurud13. The corpus is The English-Swedish Parallel Corpus (ESPC), developed by the Departmentsof English at the Universities of Lund and Gothenburg. I thank Bengt Altenberg andMikael Svensson for their kind assistance.14. Schroeder classifies instances of poss into: cp=compounding possessive, ap=agreementpossessive, rp=relational possessive.15. Menges uses ‘syntax’ in the modern sense encompassing word order, conjunctions, andthe like. Menges also mentions how, in these languages: “the possessive significance of thesuffix of the third person can completely recede when it defines or determines a noun, or,more often, a pronoun’’ (Menges 1968: 113).16. I would like to thank Eva Lindström for pre-processing the texts for me.17. In this analysis, the classification of referents into main and secondary characters, otherparticipants, and props is based on the role they play in the discourse in terms of involvementin actions and number of mentions (for a discussion see Grimes 1975).18. Cain shows how a definiteness marker in Hausa is used, not with main, but only withsecondary characters, and in the peak of the story.ReferencesArmbruster, C. H. 1908. Initia Amharica. An Introduction to Spoken Amharic. Cambridge:Cambridge University Press.Apothéloz, D. and Reichler-Béguelin, M.-J. 1999. “Interpretations and functions ofdemonstrative NPs in indirect anaphora’’. Journal of Pragmatics 31: 363–397.Aziz, Y. Y. 1993. “Explicit and Implicit Reference in Arabic-English Translation’’. Babel39: 129–150.Baker, R. 1985. The Development of the Komi Case System: A Dialectological Investigation.Helsinki: Suomalais-Ugrilainen Seura.Baskakov, N. A. 1975. “On the common origin of the categories of person and personalpossession in the Altaic languages’’. In Researches in Altaic Languages: Papers Read atthe 14th Meeting of the Permanent International Altaistic Conference Held in Szeged,August 22–28, 1971, 7–13. Budapest: Akadémia Kiadó.Bechert, J. 1993. Definiteness and article systems. EUROTYP Working Papers I (4),September 1993.Cain, B. 1991. “A discourse analysis of the Hausa fable ‘The hyena and the drum’’’. Journalof West African Languages 21: 45–52.Collinder, B. 1957. Survey of the Uralic Languages. Stockholm: Almquist och Wiksell.Collinder, B. 1960. Comparative Grammar of the Uralic Languages. Stockholm: Almqvistoch Wiksell.Comrie, B. 1981. The Languages of the Soviet Union. Cambridge: Cambridge UniversityPress.Cyr, D. 1993. “Cross-linguistic Quantification: Definite articles vs demonstratives’’.Language Sciences 15: 195–229.

Possessives as a source of definite articles? 265Décsy, G. 1990. The Uralic Protolanguage: A Comprehensive Reconstruction. Bloomington,Indiana: EUROLINGUA.Dryer, M. S. 1989. “Article-noun order’’. In Papers from the 25th Annual Regional Meetingof the Chicago Linguistic Society, Part One: The General Session, C. Wiltshire, R.Graczyk and B. Music (eds), 83–97. Chicago: Chicago Linguistic Society.Epstein, R. 1993. “The definite article: Early stages of development’’. In Historical Linguistics1991, J. van Marle (ed.), 111–134. Amsterdam and Philadelphia: John Benjamins.Epstein, R. 1994. Discourse and Definiteness: Synchronic and Diachronic Perspectives. Ph.D. Thesis. University of California, San Diego.Ergin, M. 1983. Türk Dil Bilgisi. Istanbul: Boğaziçi Yayınları.Frajzyngier, Z. 1996. “On sources of demonstratives and anaphors’’. In Studies in Anaphora,B. Fox (ed.), 169–203. Amsterdam and Philadelphia: John Benjamins.Fraurud, K. 1990. “Definiteness and the Processing of NP’s in Natural Discourse’’. Journalof Semantics 7: 395–433.Givón, T. 1981. “On the development of the numeral ‘one’ as an indefinite marker’’. FoliaLinguistica Historica II: 35–53.Greenberg, J. H. 1978. “How does a language acquire gender markers?’’. In Universals ofHuman Language III, J. H. Greenberg et al. (eds), 47–82. Stanford: Stanford UniversityPress.Greenberg, J. H. 1991. “The last stages of grammatical elements; contractive and expansivedesemanticization’’. In Approaches to Grammaticalization Volume I: Focus on Theoreticaland Methodological Issues, E. C. Traugott and B. Heine (eds), 301–314. Amsterdamand Philadelphia: John Benjamins.Grimes, J. E. 1975. The Thread of Discourse. The Hague: Mouton.Grönbech, K. 1936. Der Türkische Sprachbau. Copenhagen: Levin og Munksgaard.Hajdú, P. 1975. Finno-Ugrian Languages and Peoples (translated by G. F. Cushing). London:André Deutch.Hajdú, P. and Domokos, P. 1987. Die Uralischen Sprachen und Literaturen. Budapest:Akadémia Kiadó.Hawkins, J. A. 1978. Definiteness and Indefiniteness: A Study in Reference andGrammaticality Prediction. London: Croom Helm.Heine, B. 1997. Possession: Cognitive Sources, Forces, and Grammaticalization. Cambridge:Cambridge University Press.Heine, B., Claudi, U. and Hünnemeyer, F. 1991. Grammaticalization: A Conceptual Framework.Chicago: University of Chicago Press.Himmelmann, N. P. 1996. “Demonstratives in narrative discourse: A taxonomy of universaluses’’. In Studies in Anaphora, B. Fox (ed.), 205–254. Amsterdam and Philadelphia:John Benjamins.Himmelmann, N. P. 1997. Deiktikon, Artikel, Nominalphrase: Zur Emergenz syntaktischerStruktur. [Linguistische Arbeiten 362]. Tübingen: Niemeyer.Hopper, P. J. 1991. “On some principles of grammaticalization’’. In Approaches toGrammaticalization Volume I: Focus on Theoretical and Methodological Issues, E.C.Traugott and B. Heine (eds), 17–35. Amsterdam and Philadelphia: John Benjamins.Hopper, P. J. and Traugott, E. C. 1993. Grammaticalization. Cambridge: CambridgeUniversity Press.

266 Kari FraurudJanhunen, J. 1981. “On the structure of proto-Uralic’’. Finnisch-Ugrische Forschungen44: 23–42.Johanson, L. 1991. “Bestimmtheit und Mitteilungsperspektive im türkischen Satz’’. InLinguistische Beiträge zur Gesamtturkologie, G. Hazai (ed.), 225–250. Budapest:Akadémiai Kiadó.Jonasson, K. 1998. “Le déterminant démonstratif en français: un marqueur de quoi?’’.Travaux de Linguistique 36: 59–70.Jonasson, K. 2000. “Translation and deixis. Demonstrative determiners in a Frenchnarrative text translated into Swedish’’. In Översättning och tolkning. Rapport frånASLA:s höstsymposium Stockholm 5–6 november 1998, B. Englund Dimitrova (ed.).Uppsala: ASLA.Kapeliuk, O. 1994. Syntax of the Noun in Amharic. Wiesbaden: Harrassowitz.Kleiber, G. 1990. ‘‘Sur l’anaphore associative: article défini et adjectif démonstratif’’. Rivistadi Linguistica 2: 155–175.Krámský, J. 1972. The Article and the Concept of Definiteness in Language. The Hague:Mouton.Laury, R. 1991. “On the development of the definite article se in spoken Finnish’’. In The1991 Yearbook of the Linguistic Association of Finland, M. Vilkuna and A. Anttila (eds),93–121. Helsinki.Laury, R. 1997. Demonstratives in Interaction: The Emergence of Definite Articles in Finnish.Amsterdam and Philadelphia: John Benjamins.Leinonen, M. 1998. “The postpositive particle -to of Northern Russian dialects, comparedwith Permic languages (Komi Zyryan)’’. In Studia Slavica Finlandensia XV, J.Papinniemi, J. Lindstedt and P. Pesonen (eds), 74–90. Helsinki: Institute for Russianand East European.Lehmann, C. 1985. “Grammaticalization: synchronic variation and diachronic change’’.Lingua e Stile 20: 303–318.Lehmann, C. 1991. “Yukatekisch’’. Zeitschrift für Sprachwissenschaft 9: 28–51.Lehmann, C. 1998. Possession in Yucatec Maya: structures — functions — typology [LincomStudies in Native American Linguistics 04]. München: Lincom Europa.Leslau, W. 1967. Amharic Textbook. Wiesbaden: Otto Harrassowitz.Leslau, W. 1995. Reference Grammar of Amharic. Wiesbaden: Otto Harrassowitz.Lyons, C. 1986. “The syntax of English genitive constructions’’. Journal of Linguistics22: 123–143.Manzelli, G. 1990. Possessive adnominal modifiers. In Toward a Typology of the EuropeanLanguages, J. Bechert, G. Bernini, and C. Buridant (eds), 63–111. Berlin, New York:Mouton de Gruyter.Menges, K. H. 1968. The Turkic Languages and peoples. An Introduction to Turkic Studies.Wiesbaden: Otto Harrassowitz.Nilsson, B. 1985. Case Marking Semantics in Turkish. Ph. D. Thesis. Stockholm: Departmentof Linguistics.Plank, F. 1992. “Possessives and the distinction between determiners and modifiers (withspecial reference to German)’’. Journal of Linguistics 28: 453–468.

Possessives as a source of definite articles? 267Raun, A. 1988. “Proto-Uralic comparative historical morphosyntax’’. In The UralicLanguages: Description, History and Foreign Influences, D. Sinor (ed.), 555–571. Leiden,New York, København and Köln: E. J. Brill.Rédei, K. 1978a. Syrjänische Chrestomatie. Mit Grammatik und Glossar. Aus dem Ungarischenübersetzt von Attila Dabó. Wien: Verband der wissenschaftlichen GesellschaftenÖsterreichs.Rédei, K. (ed.). 1978b. Zyrian folklore texts (translated by I. Gombos). Budapest: AkadémiaKiadó.Rédei, K. 1988. “Die syrjänische Sprache’’. In The Uralic languages, D. Sinor (ed.), 111–130.Leiden: E. J. Brill.Sankoff, G. and Mazzie, C. 1991. ‘‘Determining noun phrases in Tok Pisin’’. Journal ofPidgin and Creole Languages 6: 1–24.Schlachter, W. 1960. Studien zum Possessivsuffix des Syrjänischen. Berlin: Akademie-Verlag.Schroeder, C. 1999. The Turkish Noun Phrase in Discourse. Wiesbaden: Otto Harrassowitz.Sinor, D. 1978. “The nature of possessive suffixes in Uralic and Altaic’’. In Linguistic andLiterary Studies in Honor of Archibald A. Hill, III: Historical and Comparative Linguistics,M. A. Jazayery, E. C. Polomé and W. Winter (eds), 257–266. The Hague: Mouton.Suihkonen, P. 1990. Korpustutkimus keilitypologiassa sovellettuna udmurttiin. Suomalaisugrilaisenseruan toimituksia Vol. 207. Helsinki: Suomalais-Ugrilainen Seura.Teleman, U. 1974. Manual för grammatisk beskrivning av talad och skriven svenska. Lund:Studentlitteratur.Uibopuu, V. 1988. Finnougrierna och deras språk (translated by H. Lagman). Lund:Studentlitteratur.Wichmann, Y. 1901. Wotjakische Sprachproben II, Sprichwörter, Rätsel, Märchen, Sagen undErzählungen. Suomalais-ugrilaisen Seuran Aikakauskirja XIX: 1. Helsinki: Suomalais-Ugrilainen Seura.

Chapter 14Possessors and experiencersin Classical LatinA. Machtelt Bolkestein1. IntroductionHeine (1997) identifies six basic syntactic construction types found in theworld’s languages for expressing the notion of ‘possession’. Possession in thisaccount covers both the concept of ‘having’ and the notion of ‘belonging’, andboth inalienable and alienable, time stable and temporary possession. Four ofthe six basic ‘participant encodings’ which he identifies are (1a–d):(1) a. the ‘Action schema’: (Possessee Obj, Possessor Subj)‘X takes/grasps/has Y’b. the ‘Goal schema’: (Possessee Subj, Possessor Dative)‘Y exists for/to X’c. the ‘Location schema’: (Possessee Subj, Possessor Location)‘Y is located at/stands/sits at X’d. the ‘Genitive schema’: (Possessee Subj, Possessor Gen)‘X’s Y exists’Latin has both construction (1a) and (1b) for expressing the notion of ‘having’in what is often called (somewhat misleadingly, as we shall see below) predicativepossession, that is, as a main clause, but no equivalent of (1d), the genitivebeing used only for NP level ‘attributive’ possession (and many other semanticrelations between nominal constituents 1 ), as in (3a) below. Another schemanot separately mentioned in Heine is the one in (1e), with its variant (1e¹),which is possible in Latin as well:(1) e. Y is X’s (Possessee Subj, Possessor Gen complement in copularpattern)e¹. Y is X’s Y (Possessee Subj, Possessor adnominal modifier ofPossessee)

270 A. Machtelt BolkesteinIn earlier discussions about differences between ‘have’ expressions as in ‘Johnhas a book’ and ‘be’ expressions as in ‘the book is John’s’ it is often pointedout that the use of the two constructions correlates with a difference in(in)definiteness of the Possessee, and this is often taken as the main factor indetermining the choice. It is not impossible, however, to find instances of‘have’ expressions with definite Possessees (‘John has the/Peter’s book’), andthe same holds for the Latin equivalent of (1b):(2) Sunt tibi regna patris ...(Verg.Aen. 12.22)are you-dat realms father-gen‘You have the realms of your father’This means that the difference should be accounted for by more than just(in)definiteness of the entities involved.A ‘possessive’ relation between two (usually nominal) entities also exists ina construction such as (3b), the so-called Dativus Possessivus or Sympatheticus(Havers 1911):(3) a. Caesaris ad pedes se iecerunt.Caesar-gen to feet refl threw‘They threw themselves at Ceasar’s (gen) feet’b. Caesari ad pedes se iecerunt.Caesar-dat to feet refl threw‘They threw themselves at Caesar’s (dat) feet’In linguistic discussions of similar variants in other languages, (3b) is oftenalluded to as an instance of ‘Possessor raising’ or ‘Possessor ascension’, as acase of ‘external possession’ (see the contributions in Payne and Barshi 1999),and sometimes again as ‘predicative’ or ‘clause level’ (vs. ‘attributive’ as in(3a)) possession.In this paper I will insist on the differences between dative and genitive‘Possessors’ rather than on their similarities, and at the same time try to showthat there are similarities between various other constructions involving Dativesin Latin (thus, in a sense, going along with the traditional accounts which tendto follow a form-function approach). In particular I will argue two points.Firstly, speaking of Possessor in instances such as (1b) on the one hand and inboth (3a) and (3b) on the other, obscures certain perhaps subtle but essentialsemantic differences between the latter two constructions, which lead to adifference in distribution between them (that is to say, the ‘Raising’/‘Ascension’account is misleading). In a number of other, unrelated languages similar

Possessors and experiencers in Classical Latin 271semantic differences between the two expression formats have been noticed(see e.g. Mithun’s contribution in this volume).Secondly, these and several other constructions involving datives in Latin(and presumably in other languages) may be described in a coherent way asbeing situated on different layers of the ‘hierarchically ordered underlying structure’currently ascribed to utterances in Functional Grammar (for a full accountsee Dik 1998; further discussion in Vet 1998. For a recent descriptive account ofthe various uses of the dative case form in Latin, see van Hoecke 1996).2. Other observations concerning the genitive and dative constructionsOne observation which one encounters regularly, cf. Kuehner and Stegmann(1912), Hoffmann and Szantyr (1965), B. Loefstedt (1963), is that using (3a)rather than (3b) is characteristic of colloquial, vulgar, or late Latin: as evidencefor this claim one points to the distribution of (Poss) genitives vs. (Poss)datives in the speeches of (socially) ‘low level’ (dative) and ‘high level’ (genitive)speakers in Petronius (cf., among others, Hoffmann and Szantyr 1965:94f.;Kuehner and Stegmann 1912; Ernout and Thomas 1953; Petersmann 1977;E. Loefstedt 1956: 225f.; Hoffmann 1951). The dative is frequent in Plautus,and in technical, non-literary authors such as Plinius NH and Celsus (seeOennerfors 1956 and Englund 1935 respectively), and the use of the dative‘instead of’ the genitive seems to increase over time. Note that speaking of‘instead of’ implies interchangeability between the two case forms.Another observation found is that more involvement (physical? emotional?)or affectedness of the ‘Possessor’ entity is implied in the case of (3b)than of (3a) (see e.g. Bennett 1914: 134f.; Hoffmann and Szantyr 1965:94,following E. Loefstedt 1956: 225f., speaks of the genitive as being ‘kuehlobjektiver’),and the dative as being both more affected in the state of affairsand being ‘subjektiver, waermer und innerlicher’. This latter statement isdifficult to interpret, and seems to imply some kind of difference in attitudeon the part of the speaker.The first factor boils down to a difference in distribution according to typeof speaker or style of speaker, or according to text type and genre, and period.The second factor may or may not correlate with the first, and is difficult tocorroborate, although a high degree of involvement of the Possessor maysometimes be deduced from the state of affairs designated and the context. Inboth cases the fact that we might be dealing with a basic semantic distinction

272 A. Machtelt Bolkesteinis neglected: and if this basic semantic difference exists, it may be responsiblefor differences in distribution as well as for differences in the degree ofinvolvement of the ‘Possessor’ entity. 23. Semantic differences between genitive and dativeand their underlying structure in FGIn Bolkestein (1983) I propose a semantic analysis and representation of (1b)and argue for recognizing a basic semantic difference between (3a) and (3b),which in a Functional Grammar formalization should also be reflected in theunderlying representation assigned to the constructions involved. I will brieflysummarize my arguments here.The semantic structure assigned to (3b) shows a similarity to (1b) at thesame time. The difference lies in the fact that in (3a) but not in (3b) (nor in 1b)a property is predicated of the Possessee (namely the (time stable, permanent)property ‘being X’s’, ‘belonging to, connected with X’), whereas no suchproperty is being predicated in the case of (1b) and (3b), where we are dealingwith (temporary) ownership. In FG terms this means that the genitive Possessorconstituent is the result of predicate formation, and that this predicate in theunderlying structure of the term functions as the second restrictor, as in (4a).In copular patterns such as (4b), the genitive is a derived predicate as well: 3(4) a. liber pueri(x: liber (x): {puer Poss} (x))book child-gen‘The child’s book’b. liber pueri est{puer Poss} (x: liber (x))book child-gen is‘The book belongs to the child’There are various kinds of evidence for the predicate status of the genitivePossessor. The first is the fact that Latin has a number of derived adjectiveswhich may be used in the same positions and expressing the same propertiesas genitive Possessors (cf. Kuehner and Stegmann 1912: I 209f., 454; Hoffmannand Szantyr 1965:66, 151f.). This substitution-possibility is ratherrestricted, such noun–adjective formation not being an extremely productiveprocess in Latin (it is however not restricted to variation with Poss genitives,

Possessors and experiencers in Classical Latin 273but may occur instead of genitives representing Agents or Patients as well,cf. (5c)):(5) a. arma fraternaweapons brother-adj‘Brotherly weapons, the weapons of a brother’b. arma fraterna suntweapons brother-adj are‘The weapons are a (one’s) brother’s, belong to a (one’s) brother’c. fraterna fuga/fraterna caedesbrother-adj fleeing/murder‘The fleeing/murder of a brother’(6) a. errare hominis estto err man-gen is‘To err is characteristic of, a property of a man (gen)’b. errare humanum estto err human is‘To err is human’Somehow using the adjective instead of a genitive in (5b) seems less natural: Iassume that if the Poss predicate is focal the genitive would normally bechosen rather that the adjective. Obviously the two constructions are nottotally synonymous, the genitive being capable of reference to specific referentsand of modification by adjectives etc.Another type of evidence of the ‘predicate’ status of the genitive constituentis the fact that one may find such constituents embedded as Object complementsof verbs governing a double accusative, such as verbs of ‘considering’(ducere, habere etc.) and verbs of causation (curare, facere) as in (7)–(8):(7) a. hanc pecuniam tuam, haud fratris ducothis money yours not brother-gen I consider‘I consider this money (to be) yours, not your brother’s, to belong toyou, not to your brother’b. hanc pecuniam patris ducothis money father-gen I consider‘I consider this money to belong to my (your etc) father’(8) a. errare hominis ducoto err man-gen I consider‘I consider erring (to be) characteristic of man’

274 A. Machtelt Bolkesteinb. errare humanum ducoto err human I consider‘I consider erring (to be) human’(9) hanc pecuniam tuam/patris faciothis money yours/father-gen I make‘I make this money your’s/father’s’It is impossible to either replace the dative in (1b) by an adjective, or to find adative as Object complement in constructions such as (7)–(9). Note, however,that the so-called Dativus Finalis or Predicative dative of abstract nouns doesin fact function as a predicate, and may therefore occur in such constructions:(10) a. hoc tibi laudi estthis you-dat praise-dat is‘This is for you a reason to be praised, makes you praiseworthy’b. hoc tibi laudi ducothis you-dat praise-dat I consider‘I deem this to be a reason to be praised for you, to make you praiseworthy’In contrast, in the case of (1b), the dative Possessor does not predicate aproperty of the Possessee, but is one of the two participants involved in thestate of affairs. This state of affairs may, but need not be an existential one: itmay also be a locative, or a copular one (due to the fact that in Latin existential‘be’ and copular ‘be’ are the same verb, that definiteness is not expressedand that word order is flexible, the surface structure in the latter case isambiguous between an existential reading, with the adjective functioning as asecond restrictor within the noun phrase (c¹), and the copular reading with theadjectival predicate as Subject complement (c):(11) a. puero liber est{EXIST} (x 1 :liber) 0 (x 2 :puer) ‘Poss’‘The boy has a book’b. puero liber in villa est{villa Loc} (x 1 :liber) 0 (x 2 :puer) ‘Poss’‘The boy has a book in the villa’c. puero liber carus estcarus (x 1 :liber) 0 (x 2 :puer) ‘Poss’‘The boy has a cherished book’ [alternative interpretation: ‘ the bookis dear to the boy’]

Possessors and experiencers in Classical Latin 275c¹.{EXIST} (x 1 :liber:carus) 0 (x 2 :puer) ‘Poss’In Bolkestein (1983) I suggest the label Experiencer for the semantic functionfulfilled by the dative on the basis of a number of considerations. Firstly thedative in construction (1b) is almost never a non-animate entity, (+human)Datives being the overwhelming majority. Thus one rarely encounters clausessuch as (12):(12) Huic templo statuae pulchrae sunt.the temple-dat statues beautiful are‘This temple has beautiful statues’This holds not only for the ‘Possessive’ dative in construction (1b), but for theconstruction (3b) as well. In other languages comparable constructions exhibitsimilar restrictions:(13) a. She touched the leg of the table.b. ?She touched the table on the leg.c. She touched his leg.d. She touched him on the leg.In Latin I found only a few exceptions to this tendency, mainly in Pliny NH,such as (14b–c) as opposed to (14a): 4(14) a. ne ...crescat gurgiti altitudo (Plin. NH 8.11)‘that not the depth of the whirlpool (dat) increases’b. (herbae) quibus flos antequam caules exeant (Plin. NH 1.21.66)‘(Herbs) whose (dat) flower blooms before its sprouts’c. folia arbori decidunt (Plin. NH 1.12.23, 1.16.33)‘The leaves of the tree (dat) fall’In the case of genitive Poss no semantic restrictions hold either on NP level orwhen the genitive is a Subject complement in a copular pattern. For examplesof the last kind cf. (15b), where the relation between the Subj ‘Possessee’ andthe genitive constituent is completely similar to that in (15a) in spite of thefact that the genitive ‘Possessor’ designates an abstract notion:(15) a. sollicita iustitia non est sapientis (Cic. Rep. 3.39)‘Worried justice is not typical of a wise man’b. hoc sentire prudentiae est (Cic. Sest. 86)‘To feel this is part of being sensible’Note that the nature of the relation between the two entities involved in the

276 A. Machtelt Bolkesteingenitive constructions changes according to the semantic properties of theconstituents involved (cf. note 1).With respect to its semantic features, the dative with the verb esse does notseem essentially different from the dative found in two other constructions inLatin, namely those where the Subject is a Gerundium (a verbal noun) as in(16b) or an action noun, as in (16a):(16) a. quid tibi (hanc) tactio est?‘What need is there for you to touch (this woman)?’b. tibi eundum est‘You have to go’c. tibi haec tangenda est‘She has to be touched by you, you have to touch her’d. tibi hanc tangendum est‘You have to touch her’I consider instances such as (16a), which is archaic and almost limited toPlautus, and (16b) and the also archaic and rare construction (16d) —which is parallel to (16a) in that the verbal noun governs an accusative caseObject — to all have the underlying structure of (11a), that is, the predicationis existential, and what is predicated to exist is a relation between apotential action and a human Experiencer participant. The fact that a‘modal’ meaning results from such a predication is not unparallelled inother languages (the potential action may be a verbal noun, or an infinitive,and the modal meaning may be one of possibility rather than deonticnecessity). 5 As I point out in Bolkestein (1980), I have seen no instanceswhere the so-called dativus Auctoris exemplified in (16b) and (16c–d)designates a non-human entity apart from a few which can be explained asimplying human participants anyway (this does not mean that the verbalpredicate involved must necessarily designate a controllable action, since Ialso found predicates such as perire ‘perish’ and algere ‘be cold’ in thisconstruction. That latter verb illustrates that states are allowed as well as[+ dynamic] states of affairs).The Gerundivum in (16c) is comparable to (11c), with the verbal adjectivepredicating a relation between the two participants referred to by tibi and haec.The ‘predicate’ status of the verbal adjective appears from its occurrence notonly as an attributive modifier within noun phrases, but also as Objectcomplement of verbs of considering such as ducere, habere etc. and causativeverbs such as curare, facere, etc.:

Possessors and experiencers in Classical Latin 277(17) a. Carthago delenda est‘Carthago has to be deleted’b. Carthaginem delendam curo‘I have Carthago be deleted’c. Carthaginem delendam duco‘I am of the opinion that Carthago has to be deleted’There is a long standing debate among Latinists which of the two (16b) or(16c) is the older construction; I tend to favour (16b) exactly on the basis ofthe existence of (16a and d), and for typological reasons. It is attractive butspeculative to view (16c) as originally having the structure of (11c¹), with theNP haec tangenda as Subj with a so-called dominant modifier, the modifierlater being reanalyzed as Subj complement in a copular pattern.Anyway, to return to the main issue, assuming that (i) the semantic role ofthe dative can indeed be characterized as an Experiencer role in both (1b) and(3b), (ii) that there is indeed a basic semantic difference between the two‘Possessive’ constructions in (3a–b) in this respect, and (iii) that the status ofthe genitive resp. dative constituent in the predication as a whole differs aswell, two further differences in semantic properties of the two constructionsfollow in a natural way, namely one concerning the type of relation betweenthe ‘Possessee’ item and the ‘Possessor’ item, and one concerning the type ofstate of affairs involved.One restriction holding for construction (3b), apart from the [+ animate]requirement on Poss, sometimes explicitly observed in the handbooks (cf. e.g.Bennett 1914: 135f.), is that we are almost always dealing with a close, usuallyinalienable, relationship between the two constituents, and especially withpart–whole relationships (body part–owner of body part). Occasionally wefind entities such as a person’s clothing, or even one’s ship:(18) confracta navis in mari est illis (Pl. Rud. 152)‘Their (dat) ship has been shipwrecked’There is no comparable restriction to close relationships between Possesseeand Possessor in the case of the genitive. Interestingly, when the close relationshipis that of a family-tie, we find genitives in instances where we mightexpect datives, as in (19):(19) a. alicuius filium corrumpere (Pl. Most. 1138)‘To corrupt someone’s son’

278 A. Machtelt Bolkesteinb. alicuius filiam vitiare (Ter. Ad. 466)‘To violate someone’s daughter’The Object Patient ‘Possessee’ constituents in instances such as (19a–b) arethemselves [+human] and therefore capable of experiencing the state ofaffairs: perhaps this explains why the Possessor entities are not presented asExperiencers.Another restriction holding for the Dative (Experiencer) construction isthe fact that the type of state of affairs must be such that the Dative participantcan indeed be affected by it. Thus in spite of a part whole relationship in(20a), the alternative (20b) is not acceptable, whereas it is acceptable in thecase of (21a–b):(20) a. membra eius videbam/cognovi‘I saw/recognized his (gen) limbs’b. ? Ei membra videbam/cognovi‘I saw/recognized him (dat) the limbs’(21) a. lassitudo/horror membra eius tenebat/occipit‘Weariness/horror held/occupied his (gen) limbs’b. Ei lassitudo/horror membra tenebat/occipit‘Weariness/horror held/occupied him (dat) the limbs’‘Seeing’ someone is not likely to affect him. Note that both state and dynamicverbs occur. The majority of verbs cooccurring with dative Experiencers,however, turn out to be dynamic rather than static, and denote the occurrenceof some kind of change in the body part and its ‘Possessor’. A notable exceptionwhere I would have expected a genitive rather than a dative is (22):(22) rostra his (sc. porphyrionibus) et praelonga crura rubent (Plin. NH 10.129)‘Their (dat) beaks and long legs are red’Some other interesting examples are (23):(23) a. patrono meo ossa bene quiescant (Petr. 39.4)‘Let my patron’s (dat) bones rest in peace’b. (apes) sedere in ore infantis tum etiam Platonis dormientis,suavitatem . . . portendentes (Plin. NH 11.55)‘(bees) sat on the lips of the then still young Plato (gen) while he wassleeping, predicting the sweetness’c. alicuius ...stomachus cibum respuit (Celsus 3.6.1)‘Someone’s (gen) stomach throws up the food’

Possessors and experiencers in Classical Latin 279Sentence (23a) illustrates that the dative is not required to actually be able tobe aware of what is happening; on the other hand in (23b) the genitive mayhave been chosen rather than the dative exactly because Poss is ‘unaware’ ofthe event. Interestingly, in (23c) (and a number of other instances in Celsus)we are dealing with a person-body part situation, but the potential affectednessof the person involved seems to be treated as irrelevant.4. Other datives and underlying utterance structureThe dative constituents discussed up till now have been described as being aparticipant in the state of affairs (the nuclear predication). For (1b) and theGerund construction with esse this description seems adequate. However, in aFunctional Grammar account of (3b), the decision has to be taken whether thedativus Sympatheticus is actually to be regarded as an extra argument of theverbal predicate involved or rather as a satellite to the state of affairs as a whole(that is, as on a par with Time, Place and Circumstance satellites). An argumentagainst the latter view might be the fact that Possessors of body parts areclearly treated as arguments of the predicate in one construction which couldbe viewed as a sort of mirror image of the dative construction, that is, theconstruction with the so-called retained accusative and its Possessor assyntactic Subject, as in (24):(24) a. caput vulneratus esthead-acc wounded-nom (he) is‘He is wounded on the head’b. bracchium ictus ceciditarm-acc hit (he) fell‘He fell hit on the arm’Whereas not all constructions in which the dative Experiencer appears have apassive equivalent such as (24), all instances of (24) seem to have an activeequivalent with the Subject entity as dative. 6Without further resolving this particular issue, I briefly want to discusstwo types of datives which function as satellites on the interpersonal level (thepropositional and the utterance layers) of the underlying structure rather thanon the representational ones (the layers of the nuclear and the extendedpredication):

280 A. Machtelt Bolkestein(25) a. a foro eunti domus a dextra parte est‘For someone coming from the forum the house is on the right’b. Quintia formosa est multis‘Many find Quintia beautiful’c. ceteris deus, sibi certe homo est‘In the eyes of the others he is a god, in his own eyes surely a man’(26) a. Quoianam vox mihi prope hic sonat? (Pl. Rud. 229)‘What kind of voice does sound here nearby?’ (surprised: do I suddenlyhear)b. Hic tibi rostra Cato advolat (Cic. Att. 1.14.5)‘Here comes Cato flying to the speaker’s chair!’ (do you see it?)c. Ecce tibi iste de traverso (Rhet. Her. 4.10.14)‘Look there that guy goes from aside’In (25) we are dealing with the so-called Dativus Iudicantis, in (26) with theDativus Ethicus. The latter is (almost?) always either a first or second personpronoun, that is, coreferent with either the speaker or the addressee, andtherefore closely connected with the here-and-now of the speech situation.In Functional Grammar the underlying structure of utterances is viewed asconsisting of a number of hierarchically ordered layers, which representdifferent order entities. The lowest layer is the predicational layer: a predicationdesignates a state of affairs (a second order entity, which can be specifiedas to its Time and Place in reality), and itself consists of relations (predicates)and participants (first order entities)). Predications are viewed as beingcontained by a proposition (a higher (third) order entity which may bespecified as to its truth value; and propositions are contained within theproduct of a speech act which carries a certain basic illocutionary force in thesituation in which it is exchanged (the other way round: utterances (speechproducts) (often) contain propositions which contain predications whichconsist of predicates and arguments situated in time and space). Whileopinions differ about what kind of entity the highest level of the structureshould represent, and what situational parameters should be included in theformal structure (see the discussions in Hannay and Bolkestein 1998), in atleast one much used approach variables for Speaker and Addressee form partof the structure itself, as in (27), which would be the representation of adeclarative sentence with non modalized positive polarity, such as ‘Quintia isbeautiful’:(27) E:DECL (S) (A) (POSPOL X: (e: beautiful (Quintia)))

Possessors and experiencers in Classical Latin 281(E=variable for speech act or product, DECL= an abstract illocutionarypredicate with S(peaker), A(ddressee) and X (=variable for proposition) asarguments). In the case of modalization by means of ‘surely’ or ‘must/mightbe’ or ‘according to me’, the proposition is expanded by modal operatorsand/or satellites, which are then attached to the propositional layer of thehierarchical structure.Whether or not one accepts this structure as in all details a correct view ofthe content of an utterance, 7 it is clear that the function of the dative constituentsexemplified by (25) is to delimit the truth value of the proposition (orpart of it) as less than universally accepted, and that they may therefore be saidto function on the propositional layer of the structure.Stretching the notion of Experiencer, one could view the dative as representingan ‘Experiencer’ of the truth value of the proposition contained in theutterance.The datives in (26) do not have anything to do with truth value (notethat they may occur in combination with different basic illocutionary forces(sentence types)) and they are not part of either the nuclear predication orthe proposition. Just like vocatives, they function on the level of the utterancesas exchanged between two speech participants (and are for that reasonpresumably impossible to be embedded as part of indirect speech). 8 Againone could stretch the notion of Experiencer and state that such dativesrepresent an ‘Experiencer’ of the speech situation. In any case, in a FGrepresentation such as (27), or one of the alternatives proposed recently (e.g.Dik 1998, Vet 1998), the dative would be situated outside (on a higher layerthan) the propositional layer. Independently, Smith (2001) reaches a similarconclusion for ‘Ethic’ pronouns in Romance.5. ConclusionIn this paper I have argued that the Possessive dative and the dativusSympatheticus in Latin, though involved as possessors/owners with respect tosome possessed item in the extra-linguistic world, linguistically behave quitedifferently from Possessive genitives: they do not function as predicatespredicating a property of the possessee, but as arguments or satellites with thesemantic function Experiencer rather than Possessor on various layers of thehierarchical structure of the utterance. This characterization correlates with anumber of semantic properties of the constructions in which they occur.

282 A. Machtelt BolkesteinWhether or not it is felicitous to extend the notion of ‘Experiencer’ to thosedatives which function as satellites on the propositional (iudicantis) and onthe utterance level (ethicus), the layered structure account can nicelyaccomodate the various uses of this case form discussed above.Notes1. Obviously I do not want to suggest that the genitive case form denotes Possessorhood:the function of the case is no more than indicating that there is a syntactic link within anoun phrase between the genitive constituent and its syntactic Head.2. A final observation, much made of in B. Loefstedt (1963), is the fact that the dative ismuch more frequently a pronoun than a full noun. This may correlate with differences inthe distribution of topicality and focality over the component parts of the two constructions(this is in fact sometimes claimed with respect to the ‘have’ and ‘be’ variants in theliterature on these variants in other languages). Statistically predicates tend to be moreoften focal in predications than referents are. However, this is no more than a tendency,and there is nothing in principle which forbids the dative to carry focus, nor the genitive tobe topical.3. For the sake of simplicity I leave out irrelevant details from the formal representation.4. In the case of Effective Objects non animate datives occur more frequently, but they mayperhaps be categorized as Beneficiaries (Dativus Commodi) rather than as ‘Poss’ (‘Exp’): cf.colorem lanis ‘to create colour for the wool’ (Plin. NH 1.319), claritatem oculis ‘clearness forthe eyes’ (Plin. NH 25.143), vilitatem auro facere ‘dullness for gold’ (Plin. NH 33.50), etc.5. In Latin modal constructions with infinitival Subjects are rare, but cf. expressions suchas: praecipue mirari est ‘one should especially wonder’, scire non est ‘it is impossible toknow’ etc. Interestingly, constructions with habere ‘have’ also develop modal meanings.6. The ‘retained’ accusative is not limited to passives, but also occurs with adjectives (flavuscapillos ‘blond of hair (acc)’ etc. — in which case the term retained is of course lessapplicable — the parallel dative construction in that case would be (1b) : ei capilli flavi sunt‘his (dat) hair was blond, he had blond hair’).7. Dik (1998) explicitly rejects incorporating S and A within the structure of the clause, asdo Vet (1998) and Bolkestein (1998). Dik does retain other than basic illocutionary forcesas elements within the structure (as operators on the speech act level), whereas Vet (1998)and Liedtke (1998) only retain basic illocutionary force (sentence type). See Hannay andBolkestein (1998) for discussion.8. These datives often occur with present tense (real or lively historical), and in combinationwith elements like hic, ecce, en etc. It would be interesting to investigate what otherspeech situation bound elements they tend to cooccur with (infinitivus historicus?exclamatives?), and what elements they tend not to be combined with.

Possessors and experiencers in Classical Latin 283ReferencesBennett, C. 1914. Syntax of Early Latin II: The Cases. Boston. Reprint: Hildesheim: Olms[1966].Bolkestein, A. M. 1980. Problems in the description of modal verbs: an investigation of Latin.Assen: Van Gorcum.Bolkestein, A. M. 1983. “Genitive and Dative Possessors in Latin’’. In Advances in FunctionalGrammar, S. C. Dik (ed.), 55–91. Dordrecht: Foris.Dik, S. C. 1998 a–b. The Theory of Functional Grammar, Part I: The structure of the clause.Part II: Complex constructions. Second revised edition, ed. by Kees Hengeveld. Berlinand New York: Mouton de Gruyter.Englund, J. 1935. Kasussyntaxen hos A. Cornelius Celsus. Goeteborg: Elander.Ernout, A. and Thomas, F. 1953. Syntaxe latine. Paris: Klincksieck.Hannay, M. and Bolkestein, A. M. (eds). 1998. Functional Grammar and Verbal Interaction.Amsterdam and Philadelphia: John Benjamins.Havers, W. 1911. Untersuchungen zur Kasussyntax der indogermanischen Sprachen.Strassburg: Truebner.Heine, B. 1997. Possession. Cognitive Sources, Forces and Grammaticalization. Cambridge:Cambridge University Press.Hoffmann, J. B. 1951. Lateinische Umgangssprache [3rd ed.]. Heidelberg: Winter.Hoffmann, J. B. and Szantyr, A. 1965. Lateinische Syntax und Stilistik. Munich: Beck.Kuehner, R. and Stegmann, C. 1912–1914. Ausführliche Grammatik der lateinischen SpracheII. Satzlehre (2 vols). Hannover: Hahnsche Buchhandlung.Liedtke, F. 1998. “Illocution and grammar: a double level approach’’. In M. Hannay andA. M. Bolkestein (eds), 107–127.Loefstedt, B. 1963. “Zum lateinischen possessiven Dativ’’. Zeitschrift für vergleichendeSprachwissenschaft 78: 64–83.Loefstedt, E. 1956. Syntactica I. Studien und Beiträge zur historischen Syntax des Lateins[2 e erw. Aufl.]. Lund: Gleerup.Oennerfors, A. 1956. Pliniana. Uppsala: Almqvist och Wiksell.Payne, D. L. and Barshi, Y. (eds). 1999. External Possession. Amsterdam and Philadelphia:John Benjamins.Petersmann, H. 1963. Petrons urbane Prosa. Untersuchungen zu Sprache und Text. Vienna:Akademie.Smith, J. C. 2001. ‘‘Illocutionary conversion, bystander deixis and Romance ‘Ethic’pronouns’’. Working Papers in Functional Grammar 74. Department of Humanities,University of Amsterdam.van Hoecke, W. 1996. ‘‘The Latin dative’’. In The Dative, W. van Belle and W. vanLangendonck (eds), 3–37. Amsterdam and Philadelphia: John Benjamins.Vet, C. 1998. “The multi-layered structure of the utterance: about illocution, modality anddiscourse moves’’. In M. Hannay and A. M. Bolkestein (eds), 1–23.

Chapter 15The difference a category makes in theexpression of possession and inalienabilityMarianne MithunA basic distinction drawn in discussions of the grammar of possession is thatbetween attributive and predicative constructions. Attributive constructionstypically presuppose possession and involve modifiers in nouns or nounphrases: my dog. Predicative constructions typically assert possession andinvolve verb phrases or clauses: The dog belongs to me or I have a dog. Yetcertain languages appear to express attributive possession in verbs. Suchstructures can be seen, for example, in languages of the Siouan family centeredon the Northern Plains in North America, the Chinookan family spoken alongthe Columbia River on the Northwest Coast, and the Iroquoian familycentered in the Northeast:(1) Lakhota (Siouan): Stan Redbird, speakerp h ehí˛ wéčašlap h ehį wa-ki-ka-šlahair 1sg.ag-poss-cause.with.instrument-be.baldhair I cut his‘I cut his hair’(2) Kathlamet (Chinook): George Cultee, speaker, to Boas 1901: 10.14uałqǽ: amiuxúlx amaimí·łatx nuałqǽ a-m-i-x-u-lx am-ai-mi-łatx nthus fut-2erg-masc.abs-poss-speak-fut masc-2poss-nephewthus you will speak to yours your nephew‘How can you tell your own nephew such a thing?’

286 Marianne Mithun(3) Mohawk (Iroquoian): Leatrice Beauvais, speakerWahinonhsóhare’se’wa-hi-nonhs-ohare-’s-’factual-1sg.ag/masc.pat-house-wash-poss-prf‘I cleaned his house’A closer look at their structures shows that they are not quite what they seem.It has long been recognized that many languages exhibit a distinction intheir nominal possessive constructions, one used for such expressions as ‘myhead’ or ‘my father’, the other for ‘my hat’ or ‘my firewood’. The first type isgenerally known as ‘inalienable’, ‘inseparable’, or ‘indivisible’ possession, andthe second as ‘alienable’, ‘separable’, or ‘divisible’ possession. The entitiesclassified grammatically as inalienable vary from language to language, butthey typically include body parts and/or kinsmen, spatial relations (the top ofan object), and often other intimately associated objects, such as one’s home,certain personal tools, footprints, or thoughts. (Extensive discussion is inChappell and McGregor 1996.)As early as 1926 Charles Bally noted that a similar distinction is signaled inIndo-European languages by other grammatical patterns, such as the clausaldative of involvement construction in French (4a), in place of a possessivedeterminer in the noun phrase (4b):(4) French inalienability in the clause: Bally 1926a. On lui tranche la têteone him.dat slice the head‘They’re cutting his head off’ (clausal construction)b. On déchire [ses habits].one rips [his.gen.pl clothes]‘They’re ripping [his clothes]’ (nominal construction)Bally linked the notion of inseparability to the sphère personnelle which ‘caninclude objects and beings associated with a person in an habitual, intimate ororganic way’ (1996 [1926]:33).Taking Bally’s observations as a point of departure, Chappell andMcGregor (1996a) show that the expression of inalienability is not limitedcross-linguistically to word or phrase-level constructions (the noun or nounphrase), but may also be carried by clause-level constructions, such as Bally’sdative of involvement in (4a), body-part locatives (The dog bit Cliff on theankle), and noun incorporation as in (3). In all of these constructions thepossessor (lui, Cliff, -hi- ‘I’/‘him’) appears as a core argument of the clause.

The difference a category makes 287These constructions have sometimes been described as the products of‘possessor ascension’, ‘possessor raising’, or ‘possessor promotion’, based onan assumption that the possessor nominal has been removed from its basicposition as a modifier within the noun phrase. As Blake (1993) and othershave pointed out, however, the appropriateness of such analyses is called intoquestion by the fact that the clausal constructions differ in meaning from theirputative sources. Not only are they limited to inalienable possession; one cansay The dog bit Cliff on the ankle but not *The dog bit Cliff on the hat. They alsorepresent the victim as more intimately affected than their nominal counterparts:The dog bit Cliff’s ankle.Several other features have been associated with the clausal constructionsas well. Bally pointed to the relevance of the part–whole relation in the dativeof involvement:Each constitutive element of the [personal] domain is regarded, not as asimple property, but as an integral part of the person . . . The idea of indivisibilityor of being part of a whole follows directly from the above: each phenomenon,action, state or quality which affects any part whatsoever of thepersonal domain, automatically affects the whole person. The part of the bodydirectly affected is only the medium for a condition which spreads to thewhole system. (Bally (1996 [1926]: 33), translated by Christine Béal and HilaryChappell.)The importance of the part–whole relation in inalienability is furtherpursued by Herslund (1997). Bally also observed that possessors in suchconstructions are typically animate. One might say, for example, On lui acoupé la jambe (one her.dat cut the leg) ‘They cut her leg off’ about a personbut not about a table. Describing Romanian, Manoliu-Manea (1996) showsthat clausal constructions typically involve topicalization as well.In the Siouan, Chinookan, and Iroquoian languages, core arguments ofthe clause are represented within the verb by pronominal prefixes. Thus verbscan and often do constitute complete, grammatical sentences in themselves,specifying both predicate and arguments within a single word. The affixestranslated as markers of possession in examples (1)–(3) are part of the verbalmorphology, but they are elements of clause-level constructions. The clausalstatus of the constructions suggests that they might be of the type discussed byBally and by Chappell and McGregor as markers of inalienable possession.Here we will examine their structures and uses more closely to determinewhether their primary function is indeed to signal attributive possession, theninvestigate the interrelationships they reflect among features that have been

288 Marianne Mithunassociated with such constructions, particularly inalienability, intimacy ofinvolvement, partitivity, animacy, and topicality.1. Nominal versus clausal constructionsIn addition to the clausal constructions seen in (1)–(3), Lakhota, Kathlamet,and Mohawk also contain nominal possessive constructions, as in (5)–(7):(5) Lakhota possession on nouns: Stan Redbird, speakernit h áhayápi kį blužážanit h a-hayapi kį wa-yu-žažayour-clothing the.past 1sg.ag-by.pulling-wash‘I washed your clothes’(6) Kathlamet possession on nouns: Boas 1901: 115.9Icktúkuipckkánaui táx ii-ck-tu-kui-pckkánaui táx iimm-masc.pl.erg-pl.abs-carry-water.to.shore all thoset´stamqut-sta-mqupl-du.poss-wood‘They carried up all their wood’(7) Mohawk possession on nouns: Warisose Kaierithon, speakeriakonnià:tha’ne raonahsire’shòn:’aiak-onni-a’t-ha’ne raon-ahsire’-shòn:’afem.ag-make-instr-imprf the masc.pl.poss-blanket-dist‘She makes their blankets with it’The fact that the nominal and clausal constructions coexist so robustly in eachlanguage suggests that they are functionally distinct. The examples seen so farindicate that clausal constructions might indeed be used for inalienablepossession (hair, nephew, house), while nominal constructions are used foralienable possession (clothing, wood, blankets). This pattern is typical innatural speech. It is not exceptionless, however. In all three languages, bothinalienable and alienable possession can be conveyed by either nominal orclausal constructions. The circumstances under which speakers choose eachare revealing.

The difference a category makes 289The use of Lakhota nominal constructions for apparently inalienablepossession can be seen in (8), and the use of clausal constructions for apparentalienable possession in (9):(8) Lakhota nominal constructions with inalienables: D 1932: 12.4, StanRedbird, speakera. hé mič h ų´ kši yúzįkta č h í˛hé mi-č h ųkši yuzį-kta č h í˛that 1sg.poss-daughter catch-irr desirethat one my daughter he will marry her he wishes‘He wishes to marry my daughter’b. mičí˛kši kį k h úžac h ami-čįkši kį k h uža=č h a3poss-son the.past ill=since‘since my sonwassick...’(9) Lakhota clausal constructions with alienables: Stan Redbird, speaker, D1932: 13.3a. hayápi waglúžažaha-yá-pi wa-ki-yu-žažaskin-cover-nomr 1ag-poss-pulling-washclothing I washed mine‘I washed my clothes’b. íš wasé glubléblel hinážįíš wasé ki-yu-bleblel hi-nážį3emph red.earth poss-pulling-break.rdp arrive-standshe warpaint untying hers she came and stood‘… she advanced with her bag of face paints open …’Kathlamet shows similar exceptions, as in (10) and (11):(10) Kathlamet nominal constructions: Boas 1901: 9.7, 159.3–4a. Aqa ikłúquat łkáx anaqa i-k-ł-u-quat ł-ka-x anthen imm-fem.erg-neut.abs-x-bathe neut-fem.poss-childthen she washed ither child‘Then she washed her child’

290 Marianne Mithunb. icłúqumst łáqauwulqt ...i-c-ł-u-qumstł-qa-uwulqtimm-masc.erg-neut.abs-away-drink neut-indef.poss-bloodhe drank ittheir blood‘He drank their blood’(11) Kathlamet clausal constructions: Boas 1901: 206.3, 133.15, cited in Hymes1955: 236a. iskixílakua istáx animi-s-ki-x-l-akuai-sta-x animimm-3du.erg-masc.abs-poss-move-around masc-3du.poss-canoethey two turned theirs aroundtheir canoe‘They turned their canoe around’b. anixúxtkamaa-n-i-x-ú-xtk-am-afut-1erg-3masc.abs-poss-from-steal-purp-futI will rob mineicxial.i-c-xial.masc-1sg.poss-workmanmy workman‘I am going to rob my workman’Similar exceptions appear in Mohawk, as in (12) and (13):(12) Mohawk nominal constructions: W. Kaierithon, T. Jacobs, speakersa. Iotkà:te’ ronwatikaratón:nisiotka’te’ ronwati-karatonni-soften fem.agt/3pl.pat-storytell-imprfoften she would tell them storiesne ronwatiien’okòn:’ane ronwati-ien’-okon’athe fem.agt/3pl.pat-child-distthe her children‘She would often tell her children stories’

The difference a category makes 291b. ienontsì:ne iahaié:na’ie-nontsi=hne i-a-ha-iena-’fem.poss-head=loc transloc-factual-masc.ag-hold-prfher headhe touched it‘He touched her head’(13) Mohawk clausal constructions: T. Jacobs, K. Jacobs, speakersa. Wahi’serehtóhare’se’wa-hi-’sereht-ohare-’s-’factual-1sg.ag/masc.pat-vehicle-wash-poss-prf‘I washed his car’b. Iah ne wén:ton ónhka’iah ne wén:ton ónhka’not the ever someonenot the ever someoneteskonwawennahrón:kente-s-konwa-wenn-ahronk-enneg-rep-indef.pl./fem-word-hear-stativedid they word-hear her again‘No one ever heard her words again’The distinction underlying the choice between the nominal and clausalconstructions is not inalienability after all. It is affectedness. The clausalconstruction is used when the individual translated as a possessor is consideredthe most significantly affected participant in an event or state. A choice ofthe construction in (8a) ‘He wishes to marry my daughter’ would haveindicated that the speaker considered himself more significantly affected by themarriage than his daughter. In (10a) ‘Then she washed her child’ it wouldhave indicated that the speaker considered herself more significantly affectedby the washing than her child. In (11a) ‘She would often tell her childrenstories’ it would have indicated that the speaker considered the storytellermore affected by the storytelling than the children.By contrast, a clausal construction was chosen in the Kathlamet ‘Theyturned their canoe around’ in (11a). By turning the canoe the two boys turnedthemselves around as well, heading toward shore. The effect of the change indirection on the boys was portrayed as more significant than its effect on thecanoe. The choice of the clausal construction in the Mohawk ‘No one everheard her words again’ in (13b) reflects the deeper point of that utterance: the

292 Marianne Mithunwoman was never heard from again, that is, she disappeared. The effect on herwas portrayed as more significant than the effect on her words.Often the rationale behind the choice of construction is not obviouswithout an understanding of the context. One might think, for example, thatdrinking someone’s blood would have a more significant effect on the victimthan on his blood, and prompt the choice of the clausal construction. Yet anominal construction was used in the Kathlamet ‘He drank their blood’ in(10b). The line comes from a tale about a man who loved blood. If he couldnot find enough, he would kill his wives and drink theirs. After he had boughtone wife, her brothers worked diligently to supply him with blood, in order toprotect their sister. They gave him five sea lions, and he drank their blood (asin (10b)). The sea lions were already dead at this point and played no otherrole in the story. The effect of the action on them was thus not portrayed assignificant. They simply characterized the kind of blood consumed.One might think that if my workman is robbed, he is more seriouslyaffected than I. Yet a clausal construction was chosen in the Kathlamet (11b)‘I am going to rob my workman’. It comes from a tale about Owl and Panther,who lived together. A young woman was sent by her father to marry Panther,but she encountered Owl first. Owl, pretending to be Panther, made her hiswife and took her home. The house was full of meat and grease, but the greaseon Owl’s side was ugly and green, taken from intestines, while that on Panther’sside was lovely and white. Owl went to Panther’s end of the house tofind some nice grease for his new wife, saying he would just get some from hisworkman. He used the clausal construction to indicate that by taking fat fromthe workman’s area he was really taking it from himself.Speakers have choices in their portrayal of affectedness. In (14a) and (15a)the speaker focused on the property and the horse with nominal possesssion.In (14b) and (15b), the focus was put on me (the speaker), with clausalconstructions:(14) Lakhota choices: Boas and Deloria 1941a. mit h áwoyuha manų´’mit h a-wa-yuha manų?1sg.poss-things-possess stealmy propertyhe stole it‘He stole something belonging to me (among other stolen property)’

The difference a category makes 293b. wóyuha mamákinų’wa-yuha ma-ma-ki-nųthings-possess steal-1sg-poss-stealproperty he stole from me‘He stole property from me’(15) Lakhota choices: Boas and Deloria 1941: 88a. mit h ášųke kį nap h é’mit h a-šųka kį na-p h e1sg.poss-dog the.past by.foot-fleemy horse the it fled‘My horse has run away’b. šų´ kak h ą ’imákiyayapi’šųka-wak h ą i-ma-ki-yaya=pidog-great to-1pat-poss-have.gone=plhorse mine have gone‘My horses have run away’ (I am horseless.)Similar choices can be seen in Kathlamet. Both sentences in (16) involve amat, but nominal possession was used in (16a), while a clausal constructionwas used in (16b):(16) Kathlamet choices: Boas 1901: 11.16, 12.1a. ikłústxula łkáq apnxi-k-ł-u-stxulał-ka-qapnximm-fem.erg-neut.abs-app-carry.on.back neut-fem.poss-matshe was carrying it on her backher mat‘She was carrying her mat’b. inłaxskam łkáq apnxi-n-ł-a-x-sk-amł-q ka-apnximm-1erg-neut.abs-fem.dat-poss-take-cmpl neut-fem.poss-matI took hersher mat‘I took her mat away’People in a village were starving. One youth could see Hunger, a supernaturalbeing, coming into the settlement in the evenings carrying a mat on her back.She would peer into the window of a house, and soon people living in thehouse would die. The mat was actually a powerful medicine bundle. Theyouth plotted to destroy Hunger by seizing her bundle. In (16a) the effect ofcarrying did not go beyond the mat. When the youth took the mat in (16b),

294 Marianne Mithunhe ultimately destroyed her, an effect expressed with a clausal construction.Similar choices can be seen in (17). A young man was told that he hadbeen kidnapped as a child, and that the woman he lived with was not hismother. He was instructed in (17a) that the way to destroy her was to cut herthroat first. The nominal construction was used here to focus on the throat,specifying just where he should cut. Once he had cut it, something roundwould jump out of it which he was told to break in order to kill his kidnapper.The clausal construction in (17b) comes from another tale. Two sistersregularly went berrypicking together. One day while they were out, the wickedelder sister Robin ate a louse she had found on the good younger sisterSalmonberry. She exclaimed at its sweetness and suggested that her youngersister would probably taste sweet too. When Salmonberry returned home shewarned her sons that if she were to disappear, they should flee, so that Robinwould not eat them, too. One day Robin returned home alone. The clausalconstruction was used in (17b) to announce the demise of Salmonberry,focusing on the effect of the event not on the neck but on Salmonberry herself:(17) Kathlamet choices: Boas 1901: 11.4, 119.9–10a. [...]łq up amiúxua icátuk.łq up a-m-i-x-u-ai-ká-tuk.cut fut-2erg-masc.abs-irr-do-fut masc-fem.poss-neckcut you will (cut) it her neck‘[If you want to kill her], cut her throat’b. Qušt, łq up ikiáxuxQušt, łq up i-k-i-á-x-uxbehold cut imm-fem.erg-masc.abs-fem.dat-poss-apparently-dobehold cut she i cut hers j‘Behold, she had cuticátukwux i akámtx ixi-ka-tukwux i a-ka-mtx ixmasc-fem.poss-neck that fem-fem.poss-younger.sisterher j neckthat her younger sister jthe throat of her younger sister’2. Diachronic sourcesThe functions of the clausal constructions are easily understood once their

The difference a category makes 295structural sources are considered. The constructions in the three languagesshow fundamental similarities.2.1 LakhotaVerbs in Lakhota, as in all Siouan languages, contain pronominal prefixes forfirst, second, and inclusive persons, but none for third. The prefixes reflect asemantic agent/patient distinction, though the choice is now lexicalized witheach stem (Mithun 1991). Participants instigating events and states arecategorized as grammatical agents (below left), while those affected but not incontrol are categorized as grammatical patients (below right):(18) Lakhota pronominal prefixes: Stan Redbird, speakerwa-híʔ ‘I came’ ma-hí˛xpaya ‘I fell’ya-híʔ ‘you came’ ni-hí˛xpaya ‘you fell’ų-híʔ ‘you and I came’ ų-hí˛xpaya ‘you and I fell’hiʔ ‘(he/she/it) came’ hįxpáya ‘(he/she/it) fell’wa-ktékte ‘I’ll kill (him/her/it)’ ma-ktékte ‘(he/she/it) will kill me’ya-ktékte ‘You’ll kill (him/her/it)’ ni-ktékte ‘(he/she/it) will kill you’ma-yá-ktekte ‘you’ll kill me’If the effect is indirect, a dative prefix ki- marks the indirectness:(19) Lakhota indirectness or dative: Stan Redbird, speakeroyá-yaka omáyakiyakao-yaka o-ma-ya-ki-yakaabout-2agt-talk about-1pat-2agt-dat-talk‘You talked about it, told a story’ ‘You talked about it to me, told me’A possessive relationship is often inferrable from the specification of indirecteffect. If the death of a horse affected someone indirectly, a likely explanationis that the horse was his:(20) Lakhota dative interpreted as possessive: Boas and Deloria 1941: 128šų´ kak h ą wą kité’šųka-wak h ą wąki-tedog-great a dat-diehorse a it died on him‘A horse died on him’ → ‘His horse died’

296 Marianne Mithun2.2 KathlametIn Kathlamet, as in other Chinookan languages, pronominal prefixes on verbsdistinguish first, second, and third persons, as well as inclusive and exclusivefirst persons, and masculine, feminine, and neuter third persons. Ergative,absolutive, and dative cases are distinguished, with the pronouns appearing inthat order within the verbal morphology:(21) Kathlamet core arguments: ergative-absolutive-dative: Boas 1901: 139.16actnlútaa-c-t-n-l-u-t-afut-masc.erg-3pl.abs-1sg.dat-to-x-give-fut‘He shall give them to me’Coreference among core arguments is indicated by a reflexive prefix -x-:(22) Kathlamet reflexive: Boas 1901: 14.5Aqa ikixquataqa ik-i-x-quatthen imm-3masc.abs-refl-wash‘Then he washed himself’As in most languages, the dative argument represents an individual indirectlyaffected by the situation. If an agent carries out an action that affects himselfor herself indirectly, the dative is coreferential with the agent. The coreferenceis marked by the reflexive prefix:(23) Kathlamet reflexive effect: Boas 1901: 104.16, cited in Hymes 1955: 236anłxáyaa-n-ł-x-á-yafut-1erg-neut.abs-refl-haul.ashore-fut‘I will haul her ashore for myself’The verb in (23) was uttered by the character Mink, who set a dish in thewater near the shore in hopes of attracting a woman. He announced, ‘Ifsomebody should come to take that dish, I will haul her ashore; I will lie downwith her all day.’ Mink apparently saw himself as the most significant beneficiaryof his act, a fact indicated by the reflexive dative.This reflexive prefix has sometimes been interpreted as a marker ofpossession within the verb. One may indeed be indirectly affected by action onone’s possessions. But the specification of indirect affectedness and possession

The difference a category makes 297are distinct in Kathlamet: affectedness is specified in the verb, and possessionis specified in the noun. The verbal marking of affectedness and the nominalmarking of possession often cooccur within sentences, because action on apossession often indirectly affects the possessor, but either can occur withoutthe other. We saw affectedness without possession in (23) ‘I will haul herashore for myself’, and possession without affectedness in (10a) ‘Then shewashed her child.’ Affectedness with and without possession can be comparedin the sentences in (24) below. The arrows in (24a) belonged to the boy, butthe sinew in (24b) had been borrowed. (The noun for ‘sinew’ is feminine ingender, but the noun contains no possessive prefix):(24) Kathlamet affectedness with and without possession: Boas 1901: 12.1, 190:2a. Aqa itxátquamaqa i-t-xa-t-qu-amthen masc.erg-3pl.abs-refl-take-completely-cmplthen he finished them on himselftiáqamacxt-ia-qamacx3pl-masc.poss-arrowhis arrows‘He used up all his arrows’b. ikaxłx um wúx i aqíłatai-k-a-x-łx u-mwúx i a-qiłataimm-fem.erg-fem.abs-refl-finish-cmpl that fem-sinewshe had finished it on herselfthat sinew‘She had used up all the sinew’The Kathlamet reflexive construction has been extended in an interestingway. If the person indirectly affected is other than the agent, this affectedperson is identified by a dative pronominal prefix. The reflexive suffix -x- canstill appear to link the absolutive argument (the object directly affected) withthe dative argument (the person indirectly affected). This time the relationshipis not perfect coreference, a fact already marked by the distinctabsolutive and dative pronouns. They share the effect: the dative is affectedindirectly through the absolutive, a situation that can arise, for example,when possessors are affected by actions on their possessions. This is thestructure seen in (2) and (16b).

298 Marianne Mithun2.3 MohawkThe pronominal prefixes on verbs in Iroquoian languages, as in Siouanlanguages, show an agent/patient pattern:(25) Mohawk pronominal prefixesGrammatical agents Grammatical patientsk-tákhe’ ‘I run’ wak-í:ta’s ‘I sleep’s-tákhe’ ‘you run’ sén-ta’s ‘you sleep’ra-tákhe’ ‘he runs’ ró:-ta’s ‘he sleeps’ie-tákhe’ ‘she runs’ iakó:-ta’s ‘she sleeps’s-k-arewáhtha’ ‘you’re hurting me’Indirect effect is indicated by a dative applicative suffix on the verb. Inconstructions containing the applicative, the grammatical patient, specifiedwithin the pronominal prefix complex, is marked as indirectly affected. Itusually refers to a semantic goal or beneficiary:(26) Mohawk benefactive: Rokwaho Dan Thompson, speakerskwatá:koshekwatákwa’ss-kwatakwshe-kwatakw-a’s2sg.ag-fix2sg.ag/fem.pat-fix-dat‘Get it ready, fix it up!’ ‘Fix it for her!’Mohawk also shows extensive noun incorporation, whereby a noun stemusually invoking a semantic patient is compounded with a verb root to yield anew compound verb stem. These stems, like others, may occur with dativeapplicatives, as in (27). Such structures have sometimes been identified aspossessive constructions (Baker 1999):(27) Mohawk clausal construction: Rokwaho Dan Thompson. speakershe’serehtakwatákwa’sshe-’sereht-a-kwatak-a’s2sg.ag/fem.pat-vehicle-epenthetic-fix-dat‘Fix the car for her!’The notion of possession is actually not part of the grammatical structure. Ifan event or state involving an object indirectly affects someone, one likelyexplanation is that the person is the owner. If you are fixing a car for her, apossible inference is that the car belongs to her. Dative arguments are thusoften interpreted as possessors of objects evoked by incorporated nouns. Butpossession is not specified by the structure.

The difference a category makes 299The sentence in (28) below has the same structure as that in (27) above,but the beneficiary ‘him’ is not interpreted as the possessor of the incorporateddoor. Inferences of possession are based on context and real-world knowledge,not this grammatical structure. The possessor of the door in (28) was actuallythe woman opening it rather than the boy outside, just as we might expectfrom our general experience with visit protocol:(28) Mohawk incorporation without possession: Niioronhia’a Montour, speakerWahonwanhotónkwahse’wa-honwa-nh-oton-kw-hs-’factual-fem.ag/masc.pat-door-close-rev-dat-prf‘She opened the door for him’3. Intimacy of affectedness: directness of effectA feature that has been associated with inalienability is ‘intimacy of effect’. Aspointed out by Blake (1990:102) and by Chappell and McGregor in theintroduction to their inalienability volume (1996a:7), the clausal constructionThe dog bit Cliff on the ankle represents the bite as more intimately affectingCliff than does the nominal in The dog bit Cliff’s ankle. Inalienability andintimacy of affectedness are indeed closely related, but grammatical structuresin Lakhota, Kathlamet, and Mohawk show that they are ultimately distinct.Two of the languages distinguish inalienability overtly in their nominalpossessive constructions. Both Lakhota and Mohawk contain two paradigmsof possessive prefixes for nouns. In Lakhota, inalienable possession is indicatedon nouns by the same pronominal prefixes as those marking patients on verbs:ma- or mi- ‘my’, ni- ‘your’, ųkí- ‘our’. (There is no third person prefix.)Alienable possession is shown by the marker t h ą- preceded by the samepronominal prefixes: mit h ą´- ‘my’, nit h ą´- ‘your’, t h a- ‘his/her/its’, ųkít h ą- ‘our’.One thus says ma-sí ‘my foot’ and ma-íte ‘my face’, but mit h á-šųka ‘my horse’and mit h á-hayapi ‘my clothing’. In Mohawk, inalienable possession is indicatedon nouns by nearly the same pronominal prefixes as those marking agents onverbs, while alienable possession is indicated on nouns by nearly the sameprefixes as those marking patients on verbs. One thus says k-ahsi’tà:ke ‘myfoot’ and k-konhsà:ke ‘my face’ but ak-itshé:nen ‘my domestic animal’ andakw-atià:tawi ‘my coat/shirt/dress’.In addition to their clausal constructions marking indirect affectedness,Lakhota, Kathlamet, and Mohawk all contain a second type of clausal con-

300 Marianne Mithunstruction that is often interpreted as attributive possession. In this type, theparticipant identified as a possessor is cast as a core argument of the clause, asin the clausal constructions described above, but there is no marker of indirectness.In Lakhota the participant is expressed as a grammatical patient, butthe verb does not contain the indirect prefix ki-:(29) Lakhota: Stan Redbird, speaker, Boas and Deloria 1941: 129p h á mayázą sí makáhų’p h á ma-yazą sí ma-kahųhead 1sg.pat-ache foot 1sg.pat-cuthead I am in pain foot (he) slashed me‘I have a headache’ ‘He foot slashed me’=‘My head aches’=‘He slashed my foot’In Kathlamet the participant identified as the possessor appears in theabsolutive rather than dative case, and there is no reflexive prefix:(30) Kathlamet: Boas 1901: 234.5qałkiuquílx mxqa-łk-i-quilx -m-xnon.imm-neut.erg-masc.abs-strike-cont-repshe struck it repeatedlyiáq aqstaqpai-ia-q aqstaq-pamasc-masc.poss-head-locon its headyáx i imúlakyax i i-mulakthat masc-elk‘She struck the elk on the head’ = ‘She struck the elk’s head’In Mohawk the participant identified as the possessor (as in Baker 1997) is castas the grammatical patient of the clause, but there is no dative suffix to markindirectness:(31) Mohawk direct effect: Kaia’titahkhe’ Jacobs, speakerwahikonhsóhare’wa-hi-konhs-ohare-’factual-1sg.ag/masc.pat.sg-face-wash-prf‘I washed his face’

The difference a category makes 301Such constructions are used primarily with body parts, just the entities that arenormally considered inalienably possessed. The sentence in (31) with incorporatednoun -konhs- ‘face’ is fine, but nearly the same sentence, with incorporatednoun -nonhs- ‘house’, is not acceptable: *wahinonhsóhare’. A dativeapplicative suffix is necessary, as in (3) above.Though there is a strong correlation between the use of these constructionsand inalienability, their primary function is to specify directness of effect.If someone cuts my foot, he cuts me directly. When the woman struck theelk’s head, she struck the elk directly. Direct affectedness and inalienabilityusually go hand in hand: action on an inalienable possession, such as a foot,head, or face, usually affects the possessor more directly than action on analienable possession such as a house or car. But where the two do not coincide,the choice of clausal structure reflects the directness of effect rather thaninalienability.In Lakhota, hair is categorized as an inalienable possession by the nounmorphology: ma-p h éhį ‘my hair’ (not *mit h á-p h ehį). But if I cut someone’s hair(as in example (1) repeated below as (32)), my action is expressed as affectinghim indirectly, with a verb containing the indirect prefix ki-, despite thegrammatical inalienability of the noun:(32) Lakhota indirect affectedness with inalienable possession: Stan Redbird,speakerp h ehí˛ wéčašlap h ehį wa-ki-ka-šlahair 1sg.ag-indirect-cause.with.instrument-be.bald‘I cut his hair’In Mohawk, one’s car is classified by the nominal morphology as alienable:akê:sere ‘my vehicle’. But in (33) no dative appears. The loan was madedirectly to the son:(33) Mohawk direct affectedness with alienable possession: K. Lazore, speakeraonsaho’seréhtani’aon-sa-ho-’sereht-ani-’opt-rep-masc.ag/masc.pat-car-lend-prf‘He (the father) would lend him (the son) the car again’Further evidence of the fact that the direct clausal construction does notdepend on inalienability can be seen in the appearance of the same noun inboth kinds of constructions. The Mohawk noun ahkwénnia’ ‘harness’ is

302 Marianne Mithuncategorized by the noun morphology as alienable. In (34a) it appears in anominal with an alienable possessive pronominal prefix. In (34b) it appearsincorporated in a direct-effect clausal construction. The difference instructure does not reflect a difference in alienability, but rather a differencein affectedness. Strapping the horses’ harness on the fence has little effect onthe horses, but removing their harness, that is, unharnessing them, affectsthem directly:(34) Mohawk: Tekaronhiokon Jacobs, speakera. Aten’enhrà:ke wahrotárhoke’ neaten’enhr-a’-ke wa-hra-otarhok-e’nefence-ns-loc factual-masc.agt-strap-prf theon the fence he strapped it theaonahkwénnia’aon-ahkwennia’zoic.pl.alien.poss-harnesstheir harness‘He strapped their harness on the fence’b. Wahshakohkwenniahrá:ko’ newa-hshako-ahkwennia-hra-ko-’nefactual-masc.agt/masc.pl.pat-harness-set-rev-prf thehe harness-removed themtheakohsá:tensako-hsaten-sthe indef.pat-carry-imprfhorses‘He removed the horses’ harness’The direct-effect construction, like the indirect-effect construction, does notactually specify possession. The sentence ‘she water-gave them’ in (35) has thesame structure as ‘I face-washed him’ in (31), but there is no implication thatthe water belonged to the guests. The idea that the face belonged to the boy in(31) but that the water did not belong to the guests in (35) is a matter ofinference from general knowledge of the world:

The difference a category makes 303(35) Mohawk clausal construction: Warisose Kaierithonwahonwatihné:kanonte’ohné:kanoswa-honwati-hnek-nont-’o-hnek-anosfactual-fem.ag/3pl.pat-liquid-feed-prf neut-liquid-be.cool.stativeshe liquid-gave themwater‘She gave them a drink of water’Inalienability and direct affectedness thus often cooccur for logical reasons,since events affecting inalienable possessions usually affect the owner moredirectly than those affecting alienable possessions, but the correlation iscircumstantial. It is important to note that the selection and interpretation ofthe direct-effect and indirect-effect clausal constructions depends not only onthe way a speaker wishes to present a situation, but also on the lexical inventoryof the language. The dative markers, reflexive, and noun incorporationare all word-formation devices, used to create lexical items. Speakers tendmost often to select existing lexical items as they speak, though they may ofcourse create neologisms on occasion. Individual lexical items often developcertain associations with circumstances surrounding their usual uses, a factwhich colors their interpretation.4. PartitivityThe notion of partitivity, or the part–whole relation, appears to play a role inthe direct-affect clausal construction. If a man cuts my foot, as in the Lakhotaexample in (29), he affects me directly because he is cutting a part of me: whenhe cuts my foot he cuts me. If a woman hits an elk on the head as in theKathlamet example in (30), she affects the elk directly because the head is partof the elk: when she strikes the head she strikes the elk. Other examples inSection 3 show, however, that while the feature of partitivity can be related todirect affectedness as a contributing factor, it is not the primary featuresignalled by the construction. When the father lent the son a car in theMohawk example in (33), he did not lend the son a part of himself. When thefarmer unharnessed his horses in (34), he did not remove a part of them.When the woman gave her guests water in (35), she did not give them a partof themselves. The direct-effect clausal constructions were used nevertheless,because the son, horses, and guests were portrayed as directly affected.

304 Marianne Mithun5. AnimacyAs noted by Bally in 1926 and others since, the participants interpreted aspossessors in the clausal constructions are typically animate. It will be recalledthat French speakers easily say On lui a coupé la jambe (‘They cut off her leg’)about a person but not about a table (even though the table is grammaticallyfeminine in French). The same close association can be seen in both kinds ofclausal constructions examined here. Those interpreted as possessors, actuallythose portrayed as indirectly or directly affected, are usually animate.But the animacy is not specified by the construction. It is a consequence ofthe fact that the effect of events and states on animates, that is, sentient beings,is often portrayed as noteworthy, but their effect on inanimate objects seldomis. An example from Kathlamet shows that inanimates are not categoricallyexcluded from the construction. To say ‘The water began to boil’, the speakercast the water in the dative case, as the entity most significantly affected by theevent. The overlap in identity between the (dative) water and its own(absolutive) froth was indicated by the reflexive prefix -x(a)-:(36) Kathlamet inanimate: Boas 1901: 239.5 cited in Hymes 1955: 237íł-ałmłmnixatlúxuaxi-ła-łmłm n-i-xa-t-l-ú-xu-axmasc-pl.poss-foam non.imm-masc.abs-rfl-pl.dat-to-on-do-imprfits foamit (its own foam) was on it (the water)łáx i łcúquałáx i ł-cuquathat masc-waterthat water‘The water became foamy’ = ‘The water began to boil’At the same time, animacy is not sufficient to prompt the choice of the clausalconstruction. The possessor of the footprints in (37) was animate, but the constructionwas not used. The person was not affected by the discovery of histracks:(37) Kathlamet animacy without affectedness: Boas 1901: 162: 13łkuałílx insktúskamł-kuałilx i-n-sk-t-u-sk-amneut-person imm-1erg.pl-3pl.abs-x-find-cmplsome person we found them

The difference a category makes 305tłáx atkt-ła-x atk3pl-neut.possessor-tracksomeone’s footprints‘We found the footprints of a person’6. TopicalityThe frequent association between the use of clausal constructions and thetopicality of possessors in Romanian was noted by Manoliu-Manea (1996). Asimilar correlation can be seen in Lakhota, Kathlamet, and Mohawk. Thereason behind the correlation is easy to understand in light of the function andform of the constructions.The primary function of the clausal constructions in all of the languages isto portray significant affectedness. Speakers show more interest in the affectednessof human or personified participants who are central to a discussionthan of peripheral characters or inanimate objects. Significantly-affectedparticipants are cast as grammatical core arguments, a status generally reservedfor topical arguments. Though the features of animacy and humanness aretypical of topical participants, they are not sufficient to render the participantstopicworthy. In the sentence in (38), those affected by blood-drinking werehuman beings, but the clausal construction was not used:(38) Kathlamet lack of topicality: Boas 1901: 13.6Tílx am łáqauwulqtt-ilx am ł-qa-uwulqtpl-person neut-indef.poss-bloodpeople their bloodančkłuqumsta.a-n-c-k-ł-u-qumst-afut-1-pl-erg-neut.abs-away-drink-prf-futwe shall drink it‘We shall drink the blood of people’The people were not expressed as core arguments because they were nottopicworthy in this context. The speakers were fleas, going out to seeksustenance. The people served only to characterize the kind of blood theywere after.

306 Marianne Mithun7. Extension of grammatical patternsThe characteristics of the clausal constructions seen so far are easily understoodin terms of the functions of their source structures. They specify thesignificant affectedness of participants by casting them as core arguments: asabsolutives, patients, or datives. The feature of possession is a secondaryinference from contexts in which they are used.But the origins of grammatical constructions do not necessarily constraintheir functions forever. Developments in the Siouan languages show howfunctions may evolve.It will be recalled that Lakhota, like other Siouan languages, contains adative prefix ki- that marks indirectness of effect. At a certain point in thedevelopment of the family, a benefactive prefix kiči- was formed from reduplicationof the dative prefix with palatalization of the second k to č induced bythe preceding i (Robert Rankin p.c. 1998). The form of the resulting prefixshows extensive phonological alternation across contexts. It specifies that anaction was done on behalf of another individual, in his or her place:(39) Lakhota dative and benefactive markers: Stan Redbird, speakerlową´ ‘sing’ma-kí-lową ‘sing to me’(Dative)m-íci-lową ‘sing for me, on my behalf, in my place’ (Benefactive)(40) Lakhota benefactive clause: Stan Redbird, speakerité wéčiyužážaité wa-kiči-yu-žažaface 1ag-benefactive-by.pulling-washface I washed for (him)‘I washed his face for him’As in Mohawk, a possessive relationship is often inferred between the objectand beneficiary of an action. If I washed a face and the washing benefittedsome person, a likely inference is that the face belonged to that person. Thedative and the benefactive constructions have now developed distinct, conventionalizedmeanings, as described by Boas and Deloria:the form ki- [dative] implies action referring to an object belonging to a persondifferent from the subject but without sanction or permission of the owner, forinstance, “I take his own without his permission’’, in other words, an action thatreflects in some way upon his interest but performed on the initiative of thesubject. The form kiči- [benefactive] expresses an action done with permission of

The difference a category makes 307the owner of an object, an action done on his initiative or in his place. (Boas andDeloria 1941: 86)The development of another verbal prefix in the Siouan languages shows thatoriginal inferences may be reinterpreted as core meaning. A second prefix hasdeveloped from reduplication of the dative ki-, a reflexive possessive prefixkik- ‘one’s own’ that marks actions directed at one’s own possessions. It hasfollowed a separate course of development from the benefactive, showing nopalatalization and distinct morphophonemic behavior:(41) Reflexive possessives: Martha St. John, Stan Redbird, speakersa. napsúkaza wakpáhųnapsukaza wa-kik-pa-hųfinger 1ag-own-by.drawing-cutfinger I cut own‘I cut my finger (with a knife)’ (Santee dialect)b. hayápi waglžažaha-yá-pi wa-ki-yu-žažaskin-cover-nomr 1ag-own-by.pulling-washclothing I washed own‘I washed my clothes’As can be seen by comparing (41a) and (41b), the construction does notdistinguish alienability.This construction now specifies possession directly rather than simplyimplying it, as is confirmed by certain items that are grammaticallyunpossessible in Lakhota. They consist primarily of objects that cannot beconsidered personal property, such as rocks, trees, and food. Nouns for themnever appear with possessive prefixes. They also never appear with verbscontaining the reflexive possessive prefix. Instead, the basic reflexive constructionis used, which otherwise marks coreference between the agent and patientor dative. The reflexive appears in its primary function in (42a) and with anunpossessible object in (42b):(42) Lakhota reflexive -ič ’ i-: Stan Redbird, speaker, Boas and Deloria 1941:103,90a. mič ’ íktektem-ič ’ i-kte=kte1sg.pat-refl-kill=irr‘I’m going to kill myself’

308 Marianne Mithunb. wamíč˛ižúžuwa-m-ič˛i-žužuby.sawing-1sg.pat-refl-butcher‘I slaughtered buffalo for myself’ for ‘I slaughtered my buffalo’8. ConclusionVerbal affixes in three genetically and geographically distinct languages ofNorth America, Lakhota, Kathlamet, and Mohawk, have sometimes beenidentified as possessive markers, in part because they are often translated assuch. The location of the markers in verbs is surprising, since attributivepossession is normally marked in nouns or noun phrases.Though they appear in verbs, the affixes are actually markers of clauselevelconstructions. In all three of the languages, the core arguments of clausesare represented by pronominal prefixes in verbs, so that every verb canconstitute a full grammatical sentence in its own right. The Lakhota, Kathlamet,and Mohawk constructions are akin to clausal structures in otherlanguages that have been said to specify inalienability.A closer examination of the uses of the constructions shows that theirprimary function is actually not to specify possession or even inalienability,but the significant affectedness of a participant. Possession, inalienability, andfeatures associated with inalienability such as partitivity and animacy, may beinferred from contexts in which the constructions are used, but they are notspecified directly by the constructions themselves.The functions and distributions of the clausal constructions reflect theirstructural origins. In each, a participant is cast as a core argument, either agrammatical patient/absolutive or a dative/beneficiary. Status as a patient orabsolutive indicates that the individual is directly affected by the event or state,while status as a dative or beneficiary indicates that the effect is indirect.Affectedness can suggest the possibility of possession, since possessors areoften affected by situations involving their possessions. Direct affectedness cansuggest inalienable possession, and indirect affectedness alienable possession,since situations involving inalienable possessions usually affect their possessorsmore directly than those involving alienable possessions. Partitivity andanimacy may be suggested as well. Situations involving a part of a participantusually affect that participant directly. The effect of situations on sentient

The difference a category makes 309beings is typically portrayed by speakers as more significant than that oninanimate objects. In the end, the forms reflect their functions.AcknowledgementsI am grateful to the following speakers who have generously shared theirexpertise on their languages: Leatrice Beauvais (Mohawk), Warisose KaierithonHorne (Mohawk), Kaia’titahkhe’ Annette Jacobs (Mohawk), TekaronhiokonFrank Jacobs (Mohawk), Karihwenhawe’ Dorothy Lazore (Mohawk),Niioronhia’a Mae Montour (Mohawk), Skawennati Montour (Mohawk),Kanerahtenhawi Hilda Nicholas (Mohawk), Stanley Redbird (Lakhota),Martha St. John (Santee Dakota), Rokwaho Dan Thompson (Mohawk).Abbreviationsabs absolutive masc masculineag agent neg negativeBD Boas and Deloria 1942 neut neutercmpl completive nomr nominalizercont continuative ns noun suffixD Deloria 1939 opt optativedat dative pat patientdist distributive pl pluraldu dual poss posssessiveemph emphatic prf perfectiveerg ergative purp purposivefem feminine rdp reduplicationfut future refl reflexivegen genitive rev reversiveimm immediate tense rfl reflexiveimprf imperfective rep repetitiveindef indefinite sg singularinstr instrumental transloc translocativeirr irrealis x unidentifiedloc locativeReferencesBaker, M. 1999. “Conditions on external possession in Mohawk: incorporation, argumentstructure, and aspect’’. In External Possession, D. Payne and I. Barsh (eds), 293–324.Amsterdam: John Benjamins.

310 Marianne MithunBally, C. 1926. “L’expression des idées de sphère personnelle et de solidarité dans leslangues indo-européennes’’. In Festschrift Louis Gauchat, F. Fankhauser and J. Jud(eds), 68–78. Aarau: Sauerländer. 68–78. English translation by Christine Béal andHilary Chappell as “The expression of concepts of the personal domain and indivisibilityin Indo-European languages’’ in H. Chappell and W. McGregor (eds), 31–61.Blake, B. 1984. “Problems of possessor ascension: some Australian examples’’. Linguistics22: 437–453.Blake, B. 1990. Relational grammar. London: Routledge.Boas, F. 1901. Kathlamet texts. Bureau of American Ethnology Bulletin 26. Washington.Boas, F. and Deloria, E. 1941. Dakota grammar. National Academy of Sciences Memoir 32.Chappell, H. and McGregor, W. (eds). 1996. The grammar of inalienability: a typologicalperspective on body part terms and the part–whole relation. Berlin: Mouton de Gruyter.Chappell, H. and McGregor, W. (1996a). “Introduction’’. In H. Chappell and W.McGregor (eds), 1–30.Heine, B. 1997. Possession: cognitive sources, forces, and grammaticalization. Cambridge:Cambridge University Press.Herslund, M. 1997. “Partitivity and inalienable possession’’. In Possessive Structures inDanish [KLIMT 3], I. Baron and M. Herslund (eds), 1–44. Copenhagen BusinessSchool.Hymes, D. 1955. The language of the Kathlamet Chinook. Ph.D. dissertation, University ofIndiana, Bloomington.Manoliu-Manea, M. 1996. “Inalienability and topicality in Romanian: pragma-semantics ofsyntax’’. In H. Chappell and W. McGregor (eds), 711–743.Mithun, M. 1984. “The evolution of noun incorporation’’. Language 60: 847–894.Mithun, M. 1991. “Active/agentive case marking and its motivations’’. Language 67:510–546.Mithun, M. 1996. “Multiple reflections of inalienability in Mohawk’’. In H. Chappell andW. McGregor (eds), 633–649.Seiler, H. 1983. Possession as an operational dimension of language. Tübingen: Gunter Narr.

Chapter 16Ways of explaining possessionBernd Heine1. IntroductionThe papers presented in this volume offer a wealth of insights into what thenotion “possession’’ is about, and they show the potential that linguistics hasfor describing and understanding this notion, which is essentially nonlinguisticin nature. The analysis of possessive constructions raises a numberof more general questions, such as the following:i. How, or to what extent, can the linguistic study of possession contributeto our extra-linguistic knowlege or understanding of this concept? Forexample, what do linguistic expressions of possession tell us aboutpsychological, legal, or historical aspects of possession?ii. Most of the chapters in this book, like most other contributions onpossessive constructions, take language structure, be that meaning orform, or both, as their starting point to generalize on the structure ofpossession. Why not start from the opposite end and examine howhuman behavior influences the linguistic form possession takes? Forexample, how do such concepts as theft, inheritance, donation, oracquisition affect the way possession is encoded linguistically?iii. To what extent is it possible to study possessive constructions as a purelylinguistic phenomenon?iv. How, or to what extent, is the linguistics of possession of a more generalconcern to linguistic theory? For example, what do observations onpossessive constructions tell us about the relation between form andmeaning, considering e.g. the fact that quite commonly there is no exactone-to-one relationship between the two?v. How to explain why possession is encoded linguistically the way it is?As the contributions to this volume suggest, each of the authors would answerthese questions somehow in a different way. We are still far away from being

312 Bernd Heineable to rely on a more widely accepted framework for describing and understandingwhat “possession’’ is about. Still, it would seem that there are a fewgeneral observations that appear to hold across languages, such as the following:a. Possession is a cross-culturally fairly stable concept; in all languages so farstudied there are conventionalized means of encoding possession. FollowingSeiler (1983: 2–4) one may define possession as a bio-cultural domaininvolving a relationship between a prototypically human possessor, inmost cases presented as the topic, and the possessee (or possessum orpossessed), normally the comment.b. There is no universal linguistic structure common to all possessive constructions.c. Possessive constructions are likely to also express concepts other thanpossessive ones.d. Possessive concepts can be expressed by linguistic forms that are nottypically associated with the domain of possession.e. That there is not necessarily a one-to-one-correspondence betweenpossessive form and possessive meaning is neither unusual nor abnormal,but it is in need of explanation.Accordingly, we may expect e.g. that in any given language there are conventionalizedforms to express predications such as ‘I have a dog’. At the sametime, however, the term “possession’’ refers to a number of different structuresand contents. First, there is a distinction between predicative possession(I have a dog) and attributive possession (my dog). The latter differs from theformer primarily in that it (a) presents typically presupposed rather thanasserted information, (b) involves object-like rather than event-like contents,and (c) has phrasal rather than clausal syntax (see the Introduction). Second,there is also a distinction between belong-constructions (The dog belongs to meor The dog is mine) and have-constructions (I have a dog). The main, thoughnot the only, difference between these two is that in the former constructionthe possessee is definite while in the latter construction it is typically indefinite(cf. Seiler 1983). Third, the term possession includes a range of differentmeanings (or notions, as I will put it here). Perhaps the most salient ones arethe following (the English examples added in parentheses illustrate therespective notions):• Physical possession (I want to fill in this form; doyouhaveapen?).• Temporary possession (I have a car that I use to go to the office but it belongsto Judy).

Ways of explaining possession 313• Permanent possession (Judy has a car but I use it all the time).• Inalienable possession (I have blue eyes).• Abstract possession (He has no time/no mercy).• Inanimate possession (My study has three windows).One may wonder whether e.g. “physical possession’’ or “inanimate possession’’should be called “possessive’’. I do not wish to discuss this issue here(see Heine 1997a for details; see also the Introduction), but as long as alinguistic construction expressing either or both of these is also used for otherpossessive notions in addition, I will not hesitate to call it a “possessiveconstruction’’.These are probably the main distinctions made in the cross-linguisticencoding of possession. Thus, most languages appear to have separate constructionsfor predicative and attributive possession, and for belong-constructionsand have-constructions. We will also expect that belong-constructionsare likely to be confined to the expression of permanent possession whilehave-constructions and attributive possession will cover several, or even all, ofthe possessive notions distinguished above, perhaps in addition to variousother meanings (Heine 1997a).My concern in this chapter is with question (v) raised above, namely: Howto explain why possession is encoded in language or languages the way it is? Iwill be concerned with a couple of peculiar structures and how they can beaccounted for. Accordingly, Section 2 deals with the notion “explanation’’ asit is understood here, in Section 3 some generalizations on the rise of possessiveconstructions are presented, and the remainder of the chapter is devotedto some problems associated with the expression of possession in Kxoe, aKhoisan language of southwestern Africa. The framework used is grammaticalizationtheory.2. On explanationIn the course of this book, a number of explanatory approaches have emerged.The kind of explanations I am concerned with here have the followingproperties (see Heine 1994: 257):a. They are multi-causal rather than mono-causal. While I will deal with onlyone causal parameter, I am aware that there are various other parametersthat also have to be taken into account. Language contact constitutes one

314 Bernd Heinesuch parameter.b. They are probabilistic rather than deductive-nomological.c. They are context-dependent rather than context-free.d. They are based on induction (or abduction) rather than on deduction.When using such a framework, it goes without saying that one may expressdoubt as to whether one is really justified to talk of “explanation’’ (cf. Lass1980). For example, according to what is said to be one of the first principlesof empirical science, every fact has one and only one explanation (cf. Sanders1974:18). I would argue that the phenomena that one is commonly confrontedwith in linguistics are such that looking for mono-causal explanationswould be either a sterile or a futile exercise.It is also widely held that explanation must rely on laws. What constitutesa “law’’ in linguistics is hard to establish; many will say that there are at best“rules’’. Some linguists have therefore proposed to baptize laws rules and todeal with exceptions to rules e.g. in the following way:[. . .] to say that something is irregular — for example, that men as a plural ofman is irregular — is merely to say that there is one statement (the “regular’’ rule)universally quantified over the set of all Q except E (e.g. all plural nouns exceptthose meaning ‘men’) and another statement (the “irregular’’ rule) universallyquantified over the set E (e.g. all plural nouns meaning ‘men’). An irregular ruleis thus simply a law whose domain is complementary to and more specific thanthat of its corresponding regular rule. (Sanders 1974: 8)This is not the procedure adopted here. I will not be concerned with laws orrules but with patterns of language use. Whether, or to what extent, thesepatterns constitute “laws’’ or “rules’’ is immaterial to the way the notion“explanation’’ is applied here. As used here, this notion has to do with viewingphenomena in a wider context, with understanding the facts under scrutiny,first, with reference to a wider range of phenomena and, second, with referenceto alternative perspectives of understanding.Furthermore, my concern will be primarily with external rather than withinternal explanation. This means in particular that I argue that in order toexplain the linguistic structure of possession, one needs to look at phenomenathat are, first, extra-linguistic and, second, relate to other phenomena thathave nothing to do with possession. The approach used here, however, alsohas some bearing on language-internal issues and, accordingly, is also materialto achieving language-internal explanation. I will return to the latter issue inthe second part of my presentation (Sections 5and 6).

Ways of explaining possession 315A number of attempts have been made to explain the structure of possessivecategories. Explanatory accounts have centered mostly around thequestion: How can the fact be accounted for that categories A and B to befound in a given language share some systematic regularity? The answer givenis usually of any of the following kinds:a. A is included in (or is derived from, or is a special case of) B and, accordingly,the structure of A can be explained with reference to the moreinclusive category B (or vice versa).b. Both A and B are included in some overarching category C and theircommon property or properties can be explained with reference to C.With regard to the ontological or other status of these categories, the followingalternatives were among the ones most frequently considered:c. While A is possessive in nature, B is not, or vice versa.d. Both A and B are possessive while C is not.Combining these four options, we get seven basic alternatives for explanation.These alternatives are listed in (1).(1) Kinds of explanation for possession (where ((A)B) = “A is included in B’’,((A, B) C) = “A and B are both included in C’’, etc.)i. ((poss) poss)ii. ((poss) non-poss)iii. ((non-poss) poss)iv. ((poss, poss) non-poss)v. ((poss, non-poss) poss)vi. ((poss, non-poss) non-poss)vii. ((non-poss, non-poss) poss)Not all of these alternatives have been exploited in treatments of possession,but at least four of them commonly have, namely i, ii, iv, and vi. 1 Iwillsay that explanations in terms of i are internal while explanations in terms ofii,iv,orviareexternal since they invoke categories that are non-possessivein nature.Examples for i can be found abundantly. A particularly common oneconcerns the relationship between attributive and predicative possession: It hasbeen claimed, especially but not only by students of generative paradigms, thatpredicative possession is the more inclusive category, that is, attributive

316 Bernd Heinepossession is included in, or derived from it or, as Baron (1997: 117) puts it,attributive possession is a “reduced’’ form of predicative possession (=“haveclauses’’in her terminology).Examples of ii are also easy to come by. Perhaps the most popular one isthat used by adherents of the localism paradigm according to which possessionis included in location (see Heine 1997a).An example of vi can be found in Herslund (1997). Herslund notes thatthe notions of inalienable possession and partitivity have a number of propertiesin common and he concludes that the former is a special case of the latter— more specifically, that inalienability is an extreme case of what he calls“organic partitivity’’.Explanation as proposed here relates to iv (and vi), that is, I am concernedwith external explanations, but the kind of explanation proposed here differsfrom most accounts represented in this volume, as well as from most previousaccounts, in that it is diachronic in nature. Consequently, when I say that A isderived from B, or A and B are both derived from C, then the term “derivation’’stands for “diachronic derivation’’ rather than for some abstract synchronicmechanism.3. Sources for possessive constructionsMy concern here is exclusively with expressions for predicative possession, andhere again with have-constructions. While a number of studies have demonstratedthat possession shares various structural similarities with such domainsas location, experience and existence (see the Introduction to this volume), theevolution of have-constructions reveals a more complex situation. A crosslinguisticstudy of these constructions suggests that have-constructions arealmost invariably derived from any of the source schemas listed in (2).(2) A formulaic description of schemas used as sources for possessionFormula Source schemaX takes Y ActionY is located at X LocationY exists for/to X GoalX’s Y exists GenitiveX is with Y Companion SchemaAs for X, Y exists Topic Schema

Ways of explaining possession 317The schemas listed in (2) are also essentially the ones that are used for attributivepossession. Note, however, that predicative possession and attributivepossession should be strictly kept apart, especially for the following reasons(Heine 1997a; see also the Introduction to this volume):a. While their conceptual sources are similar, they are not identical. Forexample, there is no equivalent for the Action Schema in attributivepossession.b. Almost invariably, constructions for attibutive possession develop independentlyfrom those for predicative possession, and vice versa. Thisapplies even in those relatively rare cases where the same source schema isused for both kinds of possession in one and the same language.One may wonder why possession should be derived from other concepts at all,that is, from concepts relating to what one does (Action), where one is located(Location), who one is accompanied by (Companion), etc. There is an obviousanswer: Possession is a fairly abstract concept; a stolen apple is hard todistinguish from one that is not (Miller and Johnson-Laird 1976). It is mucheasier to describe where a given item is, what happens to it, or who it isaccompanied by, e.g. by using body movements. Thus, action, location,accompaniment, etc. provide convenient structural templates to expresspossession: An item that I have taken or got, that is habitually located where Iam, or that I am regularly accompanied by may in certain situations be interpretedas belonging to me. In the course of time, a linguistic expressiondefining an action, a location, or a companion may then become conventionalizedas a possessive expression.Each of the schemas listed in (2) is associated with a particular structure ofmorphosyntactic encoding. The kinds of associations that are normally to beexpected are presented in (3) (see Heine 1997a).(3) Typical participant encoding in ‘have’-constructions according tosource schemaSource schema Possessor PossesseeAction Subject ObjectLocation Locative complement SubjectGoal Dative adjunct SubjectGenitive Genitive modifier SubjectCompanion Subject Comitative adjunctTopic Theme, subject Subject

318 Bernd HeineThese schemas and their morphosyntactic correlates are illustrated in (4).Examples are confined to African languages; for more examples, see Heine(1997a). For example, we find that in (4a), which is an instance of the ActionSchema, the possessor is encoded as the clausal subject and the possessee as theclausal object.(4) Examples of have-constructions (see Heine 1997a)a. Waata (East Cushitic, Afro-Asiatic)ani mín qaw-a ActionI house seize‘I have a house’ (Lit.: ‘I seize a house’)b. Manding (Mande, Niger-Congo)wari bε à fὲ Locationmoney be.at his place‘He has money (i.e. he is rich)’ (Lit.: ‘There is money at his place’)c. Ik (Kuliak, Nilo-Saharan)iá hoa ńci-k e Goalexist house me-dat‘I have a house’ (Lit.: ‘There is a house to/for me’)d. Gabu (Ubangi, Niger-Congo)aduturu dii lɔ mbiGenitivedog my is there‘I have a dog’ (Lit.: ‘My dog exists’)e. Swahili (Bantu, Niger-Congo)Hadija a-ta-kuwa na paka CompanionHadija 3:sg-fut-be com cat‘Hadija will have a cat’ (Lit.: ‘Hadija will be with a cat’)f. Lango (Western Nilotic, Nilo-Saharan)òkélò gwók’kέrέ‘ péTopicOkelo dog.his 3.neg.exist‘Okelo doesn’t have a dog’ (Lit.: ‘(As for) Okelo, his dog doesn’t exist’)To conclude, the motivation for drawing on non-possessive expressions todevelop possessive ones can be said to lie in the fact that the former are moreconcrete, easier to define and to refer to in linguistic discourse. Accordingly,the Waata construction illustrated in (4a) can be said to be historically derivedfrom an action proposition (the Action Schema), the Manding construction in(4b) from a locative construction or, as I will say here, from the Location

Ways of explaining possession 319Schema, the Ik construction in (4c), where the possessor is presented as adative participant, from the Goal Schema, the Gabu construction in (4d) froman existential proposition (the Genitive Schema), while the Swahili constructionin (4e) has a comitative form as its source, with the possessee apppearingas a kind of companion (Companion Schema), and the Lango construction(4f) from an existential one (like (4d)), with the possessor being presented asa topic or theme (the Topic Schema).4. Possession in KxoeThe situation summarized in Section 3 suggests that there are systematiccorrelations between the schematic-conceptual structures listed in (2) and themorphosyntactic patterns used for encoding these structures (cf. (3)). Thingsbecome more complicated once one is concerned with the situation to befound in a given language. Complications are likely to be the result of thespecific conditions obtaining in that language and, hence, can be explainedwith reference to these conditions.What this may mean is now demonstrated by looking in more detail atthe structure of predicative possession in one particular language. Thislanguage is Kxoe, a Central Khoisan (=Khoe) language spoken by some8,000 people in the western part of the Caprivi strip in northeast Namibia aswell as in adjacent areas of Angola and Botswana (see Brenzinger 1997;Vossen 1997). Like all other Khoe languages, Kxoe has an SOV basic order,that is, the verb normally stands at the end of the clause, and postpositions(rather than prepositions) are used. Another typological feature of thelanguage is a gender system, the genders distinguished being masculine (m),feminine (f), and common (c), and a threefold number system: singular(sg), dual (du), and plural (pl). Gender and number are expressed by aparadigm of portmanteau morphemes, referred to as pgn-markers (persongender-numbermarkers). 2Of the six source schemas used to encode have-constructions, three arefound in Kxoe. Example (5) is an instance of the Action Schema, (6) of theLocation Schema, and (7) of the Companion Schema. 3(5) ti ngúú ‘à wo Action1sg house obj get‘I have a house’

320 Bernd Heine(6) a. ti ki ¦á `m ¦¦góáà n=¦ uin or tìín Location1sg loc two stone sit stand‘I have two stones’b. xàḿ ki ¦áḿ ¦=xéí n ¦=uin or tìín3:m:sg loc two eye sit stand‘He has two eyes’c. xàń ki tcéka uu ¦¦òè or tìín3c:pl loc good country lie stand‘They have a nice country’d. áàm ki xó `m tìín but not: n¦=uin, *¦¦òè3:m:sg loc strength stand‘He is strong’(7) tí ¦ámbara ¦xòà tìín or tí ¦xòà ¦ámbara tìín Companion1sg hunger with stand 1sg with hunger stand‘I am hungry’ ‘I am hungry’5. The Location schemaMy interest here is exclusively with the Location Schema, illustrated in (6),considered by many to constitute the one most closely related to possession(see the Introduction). We are dealing in fact with a canonical instance of thisschema, as is suggested by the following facts:(a) The subject is reserved for the possessee (possessum, possessed);(b) the possessor is encoded as a locative complement;(c) there is a preference to place the locative before the subject participant.There are, however, a few peculiarities associated with the use of this schemain Kxoe. One peculiarity concerns the use of the locative copula (‘be at’). Asthe examples in (6) show, there appears to be a clause-final particle acting asa locative copula (‘be at’), but there is some variation in the use of thisparticle: The form it takes is either n=¦ uin or tìín in (6a) and (6b), but either¦¦òè or tìín in (6c), while in (6d) only tìín can be used.Another characteristic of the Location Schema in Kxoe is a conceptualdistinction between what I propose to call proximal vs. distal possession. Thisdistinction is expressed by the locative (loc) markers ki or ‘o on the one hand,and kà or ‘okà on the other: whenever the former markers are used, then thepossessee is close to the deictic center while the use of the latter markers

Ways of explaining possession 321entails that the possessee is at some distance from the deictic center, typicallyout of sight. Thus, in (8a), the possessee ¦óán ‘child’ is assumed to be close towhere the utterance is made, while in (8b) the possessee is at some distancefrom that deictic center.A different situation obtains in (8c): Whenever the constituent to the leftof the locative marker is a third-person, rather than a first- or second-personparticipant, like Caátco-m in (8c), then it is not the possessee but the possessorthat is at some distance from the deictic center. Thus, distal possessioninvolves the possessee in the case of “participants’’ (first or second person) butthe possessor in the case of “non-participants’’ (third person).(8) a. tí ki ¦á `m ¦óán tìín1sg loc two child stand‘I have two children (they are nearby)’b. tí okà ¦á `m ¦óán tìín1sg loc two child stand‘I have two children (they are not nearby)’c. Caátco-m okà ¦á `m ¦óán tìínCaatco-3m:sg loc two child stand‘Caatco has two children (he is not nearby)’The observations made above raise the following questions in particular:(a) Why is there a peculiar type of variation in the use of locative copulasin (6)?(b) Why can the copulas n=¦ uin and //òè be optionally replaced by tìín in (6a),(6b), and (6c), respectively?(c) Why is tìín in certain uses, like (6d), the only particle to be used?(d) Why are there two types of possession, proximal and distal?I will now try two answer these questions by using the framework ofgrammaticalization theory as a model. Kxoe has a system of postural markers,according to which nouns are inherently associated with one or more of themarkers presented in (9).(9) Postural markerstìín standn=¦ ùín sit¦¦òé lieHuman beings, to some extent animals as well, can occur with all three

322 Bernd Heinemarkers, as exemplified in (10). Whenever the stand-marker is used, theparticipant preceding the marker is perceived as being in an upright position,while the sit- and lie-markers imply a participant in a sitting or lying position,respectively.(10) Postural markers on human nounsa. tí-ì dàmàcí ‘à tìín tí ‘à múùn-à-tè1sg-poss sibling obl stand 1sg obj see-i-pres‘My younger brother (standing) is watching me’b. tí-ì dàmàcí ‘à n=¦ uin tí ‘à múùn-à-tè1sg-poss sibling obl sit 1sg obj see-i-pres‘My younger brother (sitting) is watching me’c. tí-ì dàmàcí ‘à ¦¦òè tí ‘à múùn-à-tè1sg-poss sibling obl lie 1sg obj see-I-pres‘My younger brother (lying) is watching me’Inanimate nouns are typically associated with only one of the three markers. 4Thus,• trees, hands, fingers, thorns, houses, or all abstract nouns are typicallyassociated with the stand-marker;• place names, stones, birds, or scorpions take the sit-marker, while• rivers, lakes, kinds of soil, or countries, take the lie-marker.But for most inanimate nouns it is also possible to use any of the othermarkers to convey specific shape distinctions. The noun yìí ‘tree’, for example,takes the stand-marker, as in (11a), but it may also take either of the othermarkers, as can be seen in (11b) and (11c).(11) Postural marker on inanimate nounsa. yìì ‘á tìín tí múùn-à-tètree obj stand 1sg see-i-pres‘I see a tree’b. yìì ‘á n=¦ ùín tí múùn-à-tètree obj sit 1sg see-i-pres‘I see a tree stump’c. yìì ‘á ¦¦òé tí múùn-à-tètree obj lie 1sg see-i-pres‘I see a tree that is lying on the ground’That in the possessive sentences in (6) all three markers are found can be

Ways of explaining possession 323accounted for in the following way: both the postural markers and the locativecopulas in (6) are historically verbs meaning ‘stand’, ‘sit’, and ‘lie’, respectively,and as such they are used as locative copulas in the Location Schema. Whichof the three copulas is used depends on the semantics of the subject noun.Thus, since stones sit in Kxoe, the locative copula consists of the verb for ‘sit’in (6a), while countries lie, hence, it is the verb for ‘lie’ that serves as thelocative copula in (6c), and in (6d) the copula is stand since abstract nounstrigger the ‘stand’-verb (see above).But this does not answer the question of why all possessive sentencesin (6) can optionally take the stand-marker without expressing any specialsemantic shades, unlike those in (11) where the postural markers are used forsemantic differentiation. The reason is the following: While (6) has thestructure of a locative construction, it exclusively serves the expression ofpossession and, as such, it is being conventionalized or “fossilized’’, in accordancewith well-known principles of grammaticalization (Heine, Claudi andHünnemeyer 1991). The principle relevant here is the following: The morestrongly a given construction is grammaticalized (and, hence, conventionalized),the more it loses in paradigmatic variability, and the more it tends tobecome associated with one particular form only.What this means in the present example is that the paradigm of posturalmarkers is more conventionalized (or fossilized) in possessive constructionsthan in other use patterns, with the effect that it tends to be reduced to onemarker, viz. tìín, that is, the three markers are being reduced to one, which isthe most “unmarked’’ one, namely the stand-marker. This means that thestand-marker has been generalized to all possessive uses of the LocationSchema, even if the sit- and lie-markers have not yet been discarded.What this also means is that (6) presents an intermediate stage where thesit- and lie-markers of (9) may still be used if associated with the appropriatenoun, but where the stand-marker tìín gradually emerges as the only onesurviving. The result is optional variation in (6) in cases where the possessee isinanimate and is intrinsically associated with either sit or lie, but no variation(i.e. only stand can be used) in cases where the possessee is intrinsicallyassociated with stand.We now remain with the question of why there are two types of possessionin Kxoe: proximal and distal. There are two sets of locativepostpositions: ki, ‘o, which express a location which is nearby, while kà and‘okà refer to a location which is at some distance, typically out of sight, asexemplified in (12).

324 Bernd Heine(12) a. tí kúùn-à-tè ¦¦xó `m ki1sg go-i-pres river loc‘I am going to the river (the river is nearby)’b. tí kúùn-à-tè ¦¦xó `m ‘okà1sg go-i-pres river loc‘I am going to the river (the river is far away)’Thus, the distinction between proximal and distal possession, illustrated in (8),is due to the semantics of the postpositions employed: whenever either of thepostpositions referring to a location nearby is used, the instance of theLocation Schema denotes proximal possession, while the use of the twopostpositions referring to distant items entails distal possession.This leaves us with a final question, namely why the distinction proximalvs. distal possession relates in some cases to the possessee (cf. 8a,b) and inothers to the possessor (cf. 8c). Once again, the answer can be derived fromthe locative semantics of Kxoe postpositions: In (8c), the postpositionalmeaning has been retained, even if the construction is no longer a locativeone; hence, the proximal vs. distal distinction relates to the noun phraseheaded by the postpositional phrase, that is, to the possessor. In (8a) and (8b),there is a conflict concerning the location of participants: On the one hand,the hearer must assume that the speaker/possessor is at the deictic center, thatis, at the place where the utterance is made. On the other hand, the use of ‘okàsuggests that the possessee is at some distance from the deictic center. The waythis conflict is resolved by Kxoe speakers apparently is that they imply that, atspeech time, the possessee is at some distance from the possessor.6. ConclusionsOne might argue that the situations described here are exceptional, e.g. thatmore often than not, possessive constructions are not derived from nonpossessiveones. There are in fact examples from a number of languages whereno motivation can be reconstructed, i.e. where no information on the historicalsource of possession is available. In such cases, either of the followingstances can be taken: (a) We have to look for more information in order toachieve the appropriate reconstruction. (b) Since no motivation has beenfound, there is no motivation. On the basis of the evidence that has becomeavailable so far, (a) is the alternative to be adopted: In a number of cases, at

Ways of explaining possession 325first glance seemingly no motivation has existed, yet after a more detailedanalysis it turns out that one of the schemas listed in (2) can in fact bereconstructed. Accordingly, (b) would offer a less plausible solution; a conclusionlike “I cannot see any motivation, hence, there is no motivation’’, runsthe risk of turning ignorance into a scientific dogma.In the first part of this chapter I was reiterating what I have been trying toestablish in earlier publications, namely that, in large measure, grammaticalforms like the ones used for possessive constructions are not the result oflanguage-internal factors, but are due to language-external forces, most of allto the desire to structure experience and to communicate successfully. It is notsyntactic or phonological factors that can be held responsible for the fact thatpossessors appear as locative or dative complements in some languages, assubjects in others, and as genitival modifiers of subjects in still others; rather,it is the desire to convey the notion of possession to the hearer in the best waypossible that induces people to seek out conceptual domains like action,location, and accompaniment as a means of expressing possessive concepts.Once an originally non-possessive expression is conventionalized, it tends tolose its association with its erstwhile contents and to be used for nothing elsebut for designating possessive concepts.Having established that there is an explanation for why the overall patternof predicative possession discussed here has the kind of syntactic and morphologicalmarking it has, we were left with a couple of additional questions whereno external explanations offer themselves. There does not seem to be anymeaningful way of explaining, for example, why there should be optionalvariation in the use of locative copulas in the case of certain possessee nounsin Kxoe but not of other nouns, or why Kxoe people should have invented adistinction between proximal and distal possession.The more a semantically motivated pattern is conventionalized, the moreit becomes part of the conventions characterizing already existing grammaticalmodes. For example, a conceptual distinction between proximal and distallocation has no intrinsic relevance for the grammar of possession in Kxoe.But it is part of the semantics of Kxoe locative postpositions, and once thesepostpositions happen to be part of a locative construction that, in certainuses, loses its association with space and comes to be associated primarilywith the domain of possession, these postpositions may acquire a newgrammatical significance: They give rise to a new pattern of expressingpossessive relations for which there exists no precedent in the language.Language is the result of processes that have occurred in past decades,

326 Bernd Heinecenturies, and millennia, and diachronic explanations therefore constitute apowerful tool for explaining language structure. The question of why there isa certain grammatical category and why it has the structure it has, cannot beanswered exhaustively if one is confined to the synchronic status of thiscategory; rather, what is required in addition is some knowledge of the factorsthat were responsible for the creation of this category. A “comprehensive’’explanation of the category must therefore remain unsatisfactory as long asone has not established whether, or to what extent, the structure one wants toexplain is not the result of its history.On the basis of this observation, the following conclusions can be drawn:Quite a number of properties characterizing possessive constructions can beaccounted for satisfactorily only by relating these constructions to certainphenomena that are non-possessive in nature. This claim need not be elaboratedhere any further since it surfaces in some way or other in several chaptersof this book. A synchronic explanation on the basis of, say, syntactic orsemantic facts may be redundant if it has already been shown that these factscan be explained exhaustively with reference to their history. On the basis of apurely synchronic (ahistorical) analysis it would be hard to account for the factthat the pattern of predicative possession in Kxoe discussed here is characterizedby optional variation in the use of the clause-final particle, and it wouldbe even harder to understand the exact nature of this variation. Similarly, onewould probably not get far if one were to explain the presence of a distinctionbetween proximal and distal possession without reference to the forces thatgave rise to it.However, as this volume amply illustrates, there are always many differentexplanations for one and the same phenomenon. Which kind of explanationwill emerge depends most of all on the questions one wants to answer, and inthe present chapter we were concerned only with one particular set ofquestions.Abbreviationsccomdatffutliecommon gendercomitativedativefeminine genderfuturepostural marker for lying positionlocmnegobjoblpllocativemasculine gendernegation markerobject (case)oblique (case)plural

Ways of explaining possession 327pastposspressgsitpast tensepossessive markerpresent tensesingularpostural marker for sitting positionstand postural marker for standing position1, 2, 3 first, second, third personi ‘‘present tense’’ junctureii ‘‘past tense’’ junctureNotes1. This does not mean that the other alternatives have not been considered in previousworks on possession. It would seem, for example, that Taylor (1996) favors a position thatcomes close to (v), in that he views possession as offering itself as a paradigm for theessential character of the English construction of attributive possession, while nonpossessiveuses of this construction would be deviations from this character.2. The data presented below have been collected during a field research trip to Namibia inSeptember/October, 1997. I wish to express my gratitude to the Kxoe people, especially toDavid Naude and Andrew Gashongo, for their understanding and cooperation. My thanksare also due to my colleagues Matthias Brenzinger, Christa Kilian-Hatz, Christa König, andMathias Schladt for sharing with me the experience of field work on the Kxoe language.Finally, I feel deeptly indebted to the Deutsche Forschungsgemeinschaft (German ResearchSociety) for having sponsored this research (as part of the work on the Sonderforschungsbereich389).The following conventions are used in this chapter: High tones are marked with an accentaigu (e.g. á), low tones with an accent grave (à), while mid tones are unmarked (a). Like allother Khoisan languages, Kxoe is a click language, distinguishing a dental (¦), a palatal (=¦ ),an alveolar (!), and a lateral click (¦¦). For more details on the language, see especiallyKöhler (1973, 1981, 1989, 1991). Abbreviations are found at the end of the chapter.3. There is some variation in the tonal shape of the clause-final locative copula. In spite ofthe pioneering tonological work carried out by Köhler (e.g. 1973, 1981), the nature ofcontext-related tonal variation is still largely unclear.4. The association between noun and postural marker is on the whole not really “exotic’’,considering e.g. the fact that also in English or other European languages trees normally‘stand’ while birds ‘sit’ and countries ‘lie’ somewhere.ReferencesBaron, I. 1997. “Possession in noun phrases: A functional analysis.’’ In I. Baron and M.Herslund (eds), 115–30.Baron, I. and Herslund, M. (eds). 1997. Possessive structures in Danish [KLIMT 3]. CopenhagenBusiness School.Brenzinger, M. 1997. “Moving to survive: Kxoe communities in arid lands.’’ Khoisan Forum(Cologne) 2.

328 Bernd HeineCohen, D. (ed.) 1974. Explaining linguistic phenomena. New York, London, Sydney,Toronto: John Wiley & Sons.Creissels, D. 1996. “Remarques sur l’émergence de verbes avoir au cours de l’histoire deslangues.’’ La relation d’appartenance [Faits de langues 7], 149–158. Paris: Ophrys.Heine, B. 1994. “Grammaticalization as an explanatory parameter.’’ In W. Pagliuca (ed.),255–287.Heine, B. 1997a. Possession: Cognitive sources, forces, and grammaticalization. Cambridge:Cambridge University Press.Heine, B. 1997b. Cognitive foundations of grammar. New York, Oxford: Oxford UniversityPress.Heine, B. Claudi, U. and Hünnemeyer, F. 1991. Grammaticalization: A conceptual framework.Chicago: Chicago University Press.Herslund, M. 1997. “Partitivity and inalienable possession.’’ In I. Baron and M. Herslund(eds), 1–44.Köhler, O. 1973. Grundzüge der Grammatik der Kxoe-Sprache. Typescript, Cologne.Köhler, O. 1981. “Les langues khoisan.’’ In G. Manessy (ed.), 55–615.Köhler, O. 1989. Die Welt der Kxoé-Buschleute im südlichen Afrika: Eine Selbstdarstellung inihrer eigenen Sprache. Volume I: Die Kxoé-Buschleute und ihre ethnische Umgebung.Berlin: Dietrich Reimer.Köhler, O. 1991. Die Welt der Kxoé-Buschleute im südlichen Afrika: Eine Selbstdarstellung inihrer eigenen Sprache. Volume II: Grundlagen des Lebens. Berlin: Dietrich Reimer.Lass, R. 1980. On explaining language change. Cambridge: Cambridge University Press.Manessy, G. (ed.) 1981. Les langues de l’Afrique subsaharienne. Paris: Editions du CNRS.Miller, G. A. and Johnson-Laird, P. N. 1976. Language and perception. Cambridge, Mass.:Harvard University Press.Pagliuca, W. (ed.). 1994. Perspectives on grammaticalization [Amsterdam Studies in theTheory and History of Linguistic Science, 109.] Amsterdam and Philadelphia: JohnBenjamins.Sanders, G. A. 1974. “Introduction.’’ In D. Cohen (ed.), 1–20.Seiler, H. 1983. Possession as an operational dimension of language [Language UniversalsSeries, 2]. Tübingen: Gunter Narr.Taylor, J. R. 1996. Possessives in English: An exploration in cognitive grammar. Oxford:Clarendon Press.Vossen, R. 1997. Die Khoe-Sprachen: Ein Beitrag zur Erforschung der Sprachgeschichte Afrikas[Quellen zur Khoisan-Forschung, 12]. Cologne: Rüdiger Köppe.

Index of languagesAAlgonquian 250Altaic 243, 249, 250, 254, 261Amharic 249, 259Arabic 247Classical Arabic 31Aramaic 221Arawakan 217Armenian 249Arrernte 82Athabascan 211Attic 138Austronesian 32, 206BBaakandji 82Balkan languages 16Bantu 244, 318Bardi 67, 68, 69, 72, 73, 74, 79, 81, 82, 83CCahuilla 35, 36, 37, 212Catalan 210, 215Cayuga 212Chadic 215, 263Chinook 285Cushitic 318DDalecarlian 207, 208, 215Danish 1, 6, 7, 9, 10, 12, 14, 16, 17, 19, 21, 41,43, 44, 45, 47, 48, 51, 52, 54, 57, 58, 59, 61,64, 87, 97, 115, 116, 117, 118, 119, 120,121, 123, 124, 125, 126, 129, 131, 132, 133,134, 135, 136, 137, 139, 142, 143, 222Old Danish 115, 141EEnglish 3, 8, 10, 14, 22, 33, 34, 35, 38, 45, 52,74, 80, 81, 97, 101, 110, 119, 121, 131,133, 134, 149, 152, 174, 188, 198, 199,206, 211, 213, 216, 217, 219, 223, 244,245, 246, 247, 248, 250, 251, 253, 259,260, 261, 312, 327FFaroese 208Finnish 255Finno-Ugric 255French 5, 9, 17, 18, 19, 20, 22, 33, 34, 80, 81,85, 97, 109, 121, 147, 149, 151, 152, 154,155, 158, 163, 187, 188, 192, 193, 198,217, 218, 219, 220, 228, 229, 230, 239,247, 286, 304GGabu 318, 319Georgian 215, 216German 33, 34, 121, 188, 198, 199, 208, 248Old High German 35Germanic 97, 134, 207, 219, 244Old Germanic 134Gooniyandi 76, 82Gothic 35, 219Greek 16, 136, 137, 139, 142Ancient Greek 32Classical Greek 115, 117, 134, 138Homeric Greek 138Old Greek 140, 143Guaraní 211HHausa 264Hebrew 221Hungarian 188, 206, 248IIcelandic 223Ik 318, 319Ilgar 211Indo-European 9, 15, 80, 81, 83, 116, 117,133, 143, 219, 223, 227, 251, 286Iroquoian 212, 285, 286, 287, 298Italian 18, 151, 205, 210JJabirrjabirr 67, 73, 79, 81, 83Jaminjung 82Jaru 82Jawi 67, 71Jawoñ 217Jukun 67, 70, 71, 82KKarajarri 82Kartvelian 215

330 Index of languagesKathlamet 285, 288, 289, 290, 291, 292, 293,294, 296, 297, 299, 300, 303, 304, 305, 308Kera 215Khoisan 21, 313, 319, 327Klamath 217Komi 247, 248, 249, 251, 252, 254, 255, 258,259, 260, 262, 263Kpelle 208Kxoe 21, 313, 319, 320, 321, 323, 324, 325,326, 327LLakhota 285, 288, 289, 292, 293, 295, 299,300, 301, 303, 305, 306, 307, 308Lango 318, 319Latin 33, 34, 35, 134, 137, 222, 269, 270, 271,272, 274, 275, 276, 281, 282Lezgian 207Lithuanian 35MMande 208, 318Manding 318Mangala 82Mangarrayi 82Mari 247, 248Melanesian Pidgin English 38Mohawk 17, 286, 288, 290, 291, 298, 299, 300,301, 302, 303, 305, 306, 308Montagnais 250Mordvin 248Murriny-Patha 82NNahuatl 212Navajo 211Ndjébbana 82Ngumbarl 67, 71Niger-Congo 208, 244, 318Nilotic 318Nimanburru 67, 83Nordic 141Nyikina 67, 68, 69, 73, 79, 82Nyulnyul 67, 68, 69, 73, 77, 79, 83Nyulnyulan 67, 68, 70, 73, 74, 75, 76, 77, 78,80, 81, 82, 83OOld Norse 1, 139PPama-Nyungan 82Permic 249, 253, 257Persian 32Polish 207, 213, 215, 217Polynesian 206, 216Portuguese 18, 227, 228, 229, 230, 231, 232,233, 239, 240Brazilian 228, 231, 232, 233, 239RRomance 150, 227, 228, 244Rumanian (Romanian) 16, 20, 22, 287, 305Russian 12, 21, 22, 48, 99, 100, 101, 102, 103,107, 108, 111, 112, 208, 216, 217, 219SSamoan 206, 217ScanicMiddle Scanic 140, 141Old Scanic 140, 141, 142Scandinavian 205, 206, 222Old Scandinavian 115, 117, 134, 139, 140,141, 142, 143Semitic 259Seneca 212Serbo-Croat 15, 16Siouan 285, 287, 295, 298, 306, 307Sirionó 211, 217, 218Slave 217, 218Slavic 32, 219Old Slavic 35Somali 215Spanish 151, 169, 172, 223, 227, 228, 229,230, 239Swahili 318, 319Swedish 16, 17, 202, 203, 206, 211, 215, 217,219, 220, 221, 222, 223, 247, 251, 257, 259TTok Pisin 249Tupí-Guaraní 211Turkic 248, 249Turkish 219, 243, 248, 249, 250, 251, 252,253, 254, 257, 258, 259, 261UUbangi 318Udmurt 249, 251, 252, 255, 259, 260Uralic 243, 244, 247, 248, 249, 250, 254, 260, 261Uto-Aztecan 35, 212WWaata 318Wambaya 82Warray 82Warrwa 67, 68, 69, 72, 73, 75, 76, 77, 80, 82YYawuru 67, 71, 83Yucatec Maya 248, 249, 251, 252, 253, 257,258, 259Yuma 212Yuman 212

Index of authorsAAklif, G. 73, 79, 82Alcover, A. 210Anschutz, A. 33Anscombre, J.-C. 152Apothéloz, D. 247Armbruster, C. H. 259Aziz, Y. Y. 247BBabby, L. H. 100Baker, M. 254, 298, 300Bally, C. 30, 192, 227, 238, 240, 286, 287, 304Barker, C. 65, 174Barnes, M. P. 208Baron, I. 2, 3, 5, 8, 9, 14, 27, 63, 65, 85, 88,94, 97, 147, 148, 152, 158, 178, 238, 316Barshi, Y. 270Bartning, I. 11, 13, 147, 148, 149, 151, 152,155, 158, 159, 160, 161, 162, 164Baskakov, N. A. 250Bates, D. M. 70Bauer, W. 216Bavin, E. L. 22Bechert, J. 248, 262Bennett, C. 271, 277Benveniste, E. 9, 13, 33, 81, 83, 85, 86, 97,121Bilý, M. 101Blake, B. J. 133, 287, 299Boas, F. 285, 288, 289, 290, 292, 293, 294,295, 296, 297, 300, 304, 305, 306, 307Bolkestein, A. M. 17, 272, 275, 276, 280, 282Bonnard, H. 199Brenzinger, M. 319, 327Brès, J. 239Broschart, J. 40CCadiot, P. 147, 153Cain, B. 257, 264Carlson, G. N. 177Chafe, W. L. 240Chappell, H. 22, 27, 67, 201, 209, 238, 286,287, 299Christiansen, H. B. 222Christie, J. J. 5Christophersen, P. 240Clark, E. V. 5, 6, 85Claudi, U. 117, 133, 134, 135, 138, 249, 250,255, 260, 323Collinder, B. 248Comrie, B. 249Corblin, F. 148, 240Cristea, T. 22Cyr, D. 250DDahl, Ö. 12, 13, 202, 203, 204, 208Décsy, G. 250, 254Deloria, E. 292, 293, 295, 300, 306, 307Dench, A. 223Diderichsen, P. 51, 120, 125Dik, S. C. 271, 281, 282Domokos, P. 248Dryer, M. S. 262Ducrot, O. 239Durieux, F. 149Durst-Andersen, P. 12, 51, 99, 100, 104Dyen, I. 223EEbert, K. 215Eisenberg, P. 118Engberg-Pedersen, E. 132, 143Englebert, A. 148Englund, J. 271Engwall, G. 152, 164Epstein, R. 244, 246Ergin, M. 254Eriksson, O. 148Ernout, A. 272Escandell Vidal, M. V. 169Evans, N. 17, 211, 212FFillmore, C. 118, 119, 128, 131Flaux, N. 154Fónagy, I. 188Formiga, M. A. 239Fox, B. 238Fradin, B. 153, 157Frajzyngier, Z. 244, 263Franks, S. 100Fraurud, K. 12, 148, 151, 243, 245, 263Freeze, R. 5, 85

332 Index of authorsGGaatone, D. 152, 199Geeraerts, D. 172Gili Gaya, S. 169Givón, T. 231, 240, 256Godard, D. 158, 160, 161Goodman, N. 200Greenberg, J. H. 213, 214, 244, 254, 263Greimas, A. J. 86Grimes, J. E. 264Grimshaw, J. 95, 118, 135Grönbech, K. 248Gross, G. 152, 160, 162Gruber, J. S. 118, 119Guillaume, G. 152Guillet, A. 5, 88HHaiman, J. 240Hajdú, P. 248Halliday, M. A. K. 119, 129Hannay, M. 280, 282Hansen, E. 116, 126Harder, P. 132, 143Haspelmath, M. 16, 18, 20, 207, 209Havers, W. 270Hawkins, J. A. 245, 252Heine, B. 2, 3, 4, 5, 9, 11, 21, 27, 38, 65, 70,87, 117, 133, 134, 135, 138, 150, 158, 169,238, 249, 250, 255, 260, 263, 269, 313,316, 317, 318, 323Heltoft, L. 5, 8, 12, 119, 129, 143, 144Herslund, M. 2, 3, 5, 8, 14, 18, 20, 22, 27, 51,63, 65, 85, 86, 88, 95, 97, 103, 121, 125,135, 144, 147, 148, 152, 158, 176, 200,238, 287, 316Himmelmann, N. P. 244, 247Hjelmslev, L. 133Hoffmann, J. B. 271, 272Hopper, P. J. 249Hosokawa, K. 67, 82Howe,C.J. 174Høysgaard, J. P. 116Hünnemeyer, F. 117, 133, 134, 135, 138, 249,250, 255, 260, 323Hymes, D. 290, 296, 304IIsačenko, A. V. 5, 9, 10, 22JJackendoff, R. M. 65, 118, 119Jakobsen, L. F. 143Janhunen, J. 251, 255Jansen, L. 101Johanson, L. 248, 263Johnson-Laird, P. N. 317Jonasson, K. 247Jonsson, N. 206, 217Josefsson, G. 17Junker, M.-O. 200KKapeliuk, O. 259Kleiber, G. 148, 151, 152, 153, 154, 157, 163,240, 247Köhler, O. 327König, E. 16, 18, 20, 209Koptjevskaja-Tamm, M. 12, 13, 208Kotschi, T. 9Krámský, J. 248, 249Kuehner, R. 271, 272LLagunilla, M. F. 169Langacker, R. W. 45, 50, 52, 152Lass, R. 314Laury, R. 244, 246Leclère, C. 5, 22, 88Lehmann, C. 27, 247, 248, 253, 254, 260Leinonen, M. 248Leonard, H. S. 200Leslau, W. 259Levander, L. 207Levi, J. 152Lévy-Bruhl, L. 227, 238Liedtke, F. 282Lockwood, W. B. 208Loefstedt, B. 271, 282Loefstedt, E. 271Löfström, J. 203Lucchesi, D. 239Luraghi, S. 138Luxt, L. I. 109Lyons, J. 5, 74, 85, 97, 246MMcGregor, W. B. 11, 22, 27, 67, 74, 75, 76, 83,201, 209, 238, 286, 287, 299Manoliu-Manea, M. 20, 22, 287, 305Martineau, F. 200Martinet, A. 239Mazzie, C. 249Menges, K. H. 254, 264Merlan, F. 217Metcalfe, C. D. 82Miller, G. A. 317Milner, J.-C. 154, 155, 158, 239Mithun, M. 12, 17, 20, 295Moll, F. 210Müller, H. H. 13, 94, 178NNekes, H. 67, 68, 69, 72, 73, 74, 79, 83Nichols, J. 101, 210

Index of authors 333Nilsson, B. 249Nygaard, M. 1, 139OOennerfors, A. 271PPartridge, M. 15Payne, D. L. 217, 270Petersmann, H. 271Plank, F. 246Popescu-Ramírez, L. 22Pustejovsky, J. 86, 152RRaun, A. 250Rebollo, A. A. 169Reboul, A. 153Rédei, K. 248, 252Reichler-Béguelin, M.-J. 247Rice, K. 217, 218Riegel, M. 12, 149, 152, 154, 156, 158, 196,198, 199, 200SSanders, G. A. 314Sankoff, G. 249Sasse, H. J. 116Schapira, C. 162Schroeder, C. 252, 258, 264Seiler, H. 2, 5, 9, 10, 12, 13, 27, 28, 33, 35, 39,85, 86, 97, 147, 149, 238, 240, 312Serzisko, F. 215Sinor, D. 261Slobin, D. 174Smith, J. C. 281Sørensen, F. 1, 5, 13, 85, 94, 95, 144Spang-Hanssen, E. 152Spanoghe, A.-M. 5, 12, 14, 18, 200, 238, 239,240Stage, L. 152, 178Steensland, L. 207Stegmann, C. 271, 272Stokes, B. 68, 69, 73, 75, 79, 82Suihkonen, P. 251Szantyr, A. 271, 272TTasmowski-De Ryck L. 22Taylor, J. R. 27, 47, 327Teleman, U. 263Teyssier, P. 239Thomas, F. 271Timoc-Bardy, R. 22Togeby, O. 1, 21, 47Traugott, E. C. 249UUibopuu, V. 248Uhlenbeck, C. C. 228, 238VVan Hoecke, W. 271Velázquez-Castillo, M. 14, 20, 211Vet, C. 271, 281, 282Villiers, J. G. de 174Villiers, P. A. de 174Vossen, R. 319WWalsh, M. 17Warren, B. 152Weinrich, H. 27Weyhe, E. 208Whorf, B. L. 101Wichmann, Y. 255Widell, P. 47Wierzbicka, A. 7, 97Winston, M. E. 152Worms, E. A. 67, 68, 69, 72, 73, 74, 79,83ZŽurinskaja, M. A. 103

Index of subjectsAabsolutive 211, 296, 297, 300, 304, 306, 308accompaniment 32, 47, 54, 317, 325accusative 34, 68, 75, 76, 77, 83, 100, 102, 107,108, 112, 140, 143, 273, 276, 279, 282affectedness 46, 51, 52, 54, 271, 279, 291, 292,296, 297, 299, 301, 302, 303, 304, 305,306, 308affix 211, 244, 250, 259, 287agent 4, 5, 7, 8, 9, 10, 12, 34, 70, 75, 92, 118,119, 136, 154, 156, 160, 162, 295, 296,297, 298, 299, 307alienable possession 2, 12, 57, 64, 227, 286,288, 289, 299, 301, 308anaphora 151, 161, 245, 246, 247, 249, 252,253, 259, 260, 261anchor 201, 202, 203, 204, 205, 206, 209, 211,214, 215, 219, 221, 245, 253, 258, 260,261animacy 288, 304, 305, 308applicative 298, 301argument 5, 15, 22, 58, 64, 88, 89, 90, 95, 96,126, 149, 152, 178, 180, 185, 195, 212,279, 286, 287, 296, 297, 300, 308associative (relation, anaphora) 70, 73, 78,240, 245, 247, 249, 252, 253, 259, 260,261attributive possession 4, 12, 13, 14, 15, 151,243, 255, 263, 269, 285, 287, 300, 308,312, 313, 315, 316, 317, 327auxiliary (verb) 3, 32, 52, 67, 80Bbackground 203, 231body part (terms) 30, 54, 180, 201, 208, 209,211, 213, 215, 227, 228, 233, 235, 237,240, 277, 278, 279, 286, 301Cclassifier 32, 35clitic 16, 22, 174, 215, 217, 243, 244, 250, 259co-subject 116, 126, 127comitative 27, 81, 83, 115, 116, 117, 126, 131,132, 134, 135, 136, 138, 140, 141, 142,143, 317, 319control 29, 44, 46, 51, 52, 54, 72, 73, 76, 78,99, 129, 130, 171, 187, 196, 295copula 68, 74, 75, 80, 83, 320, 321, 323, 327Ddative 16, 17, 19, 22, 27, 28, 31, 32, 33, 34, 38,43, 50, 54, 83, 100, 101, 102, 104, 105,106, 107, 109, 111, 112, 115, 117, 134,136, 137, 138, 139, 140, 141, 142, 143,198, 269, 270, 271, 272, 274, 275, 276,277, 278, 279, 281, 282, 286, 287, 295,296, 297, 298, 300, 301, 303, 304, 306,307, 308, 317, 319, 325definite article 28, 34, 97, 187, 188, 197, 205,210, 213, 220, 221, 222, 223, 228, 231,232, 233, 235, 239, 240, 243, 244, 245,246, 247, 248, 249, 250, 257, 259, 260,262, 263demonstrative 244, 246, 248, 257, 263determiner 22, 147, 159, 160, 161, 162,163, 188, 189, 198, 204, 222, 228, 230,231, 232, 233, 235, 239, 244, 246, 263,286Eergative 68, 75, 76, 83, 212, 296established possession 39existence 34, 85, 101, 316experience 12, 18, 21, 22, 51, 52, 54, 109,316experiencer 4, 5, 12, 17, 18, 19, 20, 22, 34, 49,143, 275, 276, 277, 278, 279, 281, 282experient 118, 143external possession 4, 15, 17, 18, 19, 20, 67,209, 270Ffocus 79, 260, 261, 282, 292foreground 231Ggenitive 1, 2, 4, 14, 16, 28, 31, 32, 33, 34, 38,43, 44, 45, 46, 47, 54, 59, 60, 61, 62, 63,87, 97, 100, 101, 102, 103, 104, 106, 112,128, 129, 139, 154, 155, 156, 159, 170,204, 207, 208, 269, 270, 271, 272, 273,275, 276, 277, 278, 279, 282, 316, 317,318, 319grammaticalization 117, 119, 133, 134, 135,150, 209, 210, 211, 213, 221, 222, 244,245, 247, 249, 250, 254, 255, 258, 259,260, 313, 321, 323

Index of subjects 335Iinalienable possession 2, 11, 12, 57, 64, 76,103, 123, 157, 227, 232, 286, 287, 288,289, 299, 301, 308, 313, 316inclusion 8, 63, 85, 86, 87, 88, 89, 90, 91, 92,93, 94, 95, 96, 97, 103, 105, 107, 112incorporation 16, 209, 286, 298, 299, 303indefinite article 228, 231, 232, 233, 234, 235,237, 239, 240, 256indirect object 30, 43, 47, 49, 50, 51, 54, 107,116, 124, 125, 228, 229inherent possession 34, 39instrument 49, 118, 131, 134, 135, 136, 138, 141instrumental 27, 32, 38, 100, 101, 102, 104,105, 106, 107, 108, 109, 110, 112, 115,117, 123, 130, 134, 136, 139, 140inversion 29, 30, 34, 36, 37, 38, 39isotopic chain 86, 92Kkin terms 30, 36, 201, 202, 203, 204, 205, 206,207, 208, 209, 210, 211, 212, 213, 214,215, 216, 217, 218, 219, 221, 222, 223kinship 3, 11, 13, 16, 35, 45, 97, 152, 157, 170,180, 182, 184, 190, 191, 194, 195, 196,197, 201, 207, 208, 211, 212, 213, 214,216, 227, 240Llocation 4, 5, 8, 9, 10, 12, 18, 21, 22, 32, 35,44, 46, 49, 51, 52, 54, 57, 58, 59, 60, 63,64, 65, 85, 89, 90, 96, 123, 137, 157, 158,161, 170, 172, 269, 316, 317, 318, 319,320, 323, 324, 325locative 5, 6, 7, 8, 9, 11, 14, 19, 20, 21, 22, 38,83, 85, 86, 97, 100, 123, 129, 136, 137,138, 140, 142, 151, 153, 154, 157, 158,159, 160, 161, 162, 163, 174, 179, 183,184, 185, 198, 206, 274, 317, 318, 320,321, 323, 324, 325, 327Mmanner 134, 135, 136, 143mereological 188, 198meronymy 44, 45, 46, 51, 52, 54, 120, 122,127, 128, 130, 131, 132modifier 188, 189, 269, 276, 277, 287, 317, 325Nnominalization 149, 178, 185nominative 75, 83, 100, 105, 112, 141, 143,207, 213Ooblique object 105, 109, 112, 228Ppart–whole 8, 11, 31, 86, 89, 90, 97, 123, 151,159, 161, 163, 227, 238, 253, 303partitivity 288, 303, 308, 316passive 6, 7, 10, 11, 50, 111, 112, 120, 124,126, 279patient 5, 8, 10, 159, 278, 295, 298, 300, 307, 308polysemy 134, 147, 155, 158possessee 2, 59, 65, 216, 258, 269, 270, 272,274, 277, 281, 312, 317, 318, 319, 320,321, 323, 324, 325possessor 2, 4, 5, 6, 8, 9, 10, 11, 14, 15, 16, 17,18, 19, 20, 21, 28, 30, 32, 33, 34, 35, 37,39, 47, 49, 58, 59, 65, 67, 68, 74, 75, 76,79, 81, 83, 85, 95, 96, 99, 100, 105, 107,108, 109, 110, 112, 118, 201, 204, 209,211, 216, 218, 229, 230, 232, 233, 234,235, 236, 237, 239, 240, 258, 269, 270,271, 272, 274, 277, 278, 279, 281, 286,287, 291, 297, 299, 300, 301, 304, 305,312, 317, 318, 319, 320, 321, 324possessor ascension 4, 14, 201, 209, 270, 287possessum 2, 3, 4, 5, 8, 9, 10, 11, 13, 14, 15,17, 19, 20, 21, 28, 30, 32, 33, 34, 35, 37,39, 68, 74, 75, 76, 81, 83, 95, 96, 97, 99,100, 104, 105, 107, 111, 112, 312, 320predicative possession 2, 4, 9, 15, 19, 67, 81, 82,243, 269, 312, 315, 316, 317, 319, 325, 326prefix 9, 36, 68, 75, 217, 251, 295, 296, 297,298, 299, 300, 301, 302, 304, 306, 307preposition 7, 8, 10, 20, 22, 43, 47, 54, 58, 85,87, 89, 100, 110, 116, 126, 135, 143, 147,150, 169, 172, 173, 175, 177, 178, 179,184, 206pronoun 22, 34, 36, 43, 125, 204, 208, 209,211, 217, 250, 251, 253, 260, 280, 282proprial 206, 207, 213prototypicality 171, 172, 173, 184, 199QQualia structure 86, 152Rreflexive 34, 68, 125, 251, 296, 297, 300, 303,304, 307relational noun 3, 13, 86, 88, 157, 181, 182,183, 184Sstimulus 5, 22sub-place 8, 19, 87, 88, 89, 90, 91, 92, 93, 94,95, 96, 97suffix 36, 83, 215, 217, 227, 248, 251, 264,297, 298, 300, 301Ttopic 9, 10, 12, 14, 76, 101, 312, 316, 317, 318,319topicality 258, 282, 288, 305transitivity 22, 34, 68, 81

In the series TYPOLOGICAL STUDIES IN LANGUAGE (TSL) the following titles havebeen published thus far:1. HOPPER, Paul J. (ed.): Tense-Aspect: Between semantics & pragmatics. 1982.2. HAIMAN, John & Pamela MUNRO (eds): Switch Reference and Universal Grammar.Proceedings of a symposium on switch reference and universal grammar, Winnipeg, May1981. 1983.3. GIVÓN, T.: Topic Continuity in Discourse. A quantitative cross-language study. 1983.4. CHISHOLM, William, Louis T. MILIC & John A.C. GREPPIN (eds): Interrogativity: Acolloquium on the grammar, typology and pragmatics of questions in seven diverse languages,Cleveland, Ohio, October 5th 1981-May 3rd 1982. 1984.5. RUTHERFORD, William E. (ed.): Language Universals and Second Language Acquisition.1984 (2nd ed. 1987).6. HAIMAN, John (Ed.): Iconicity in Syntax. Proceedings of a symposium on iconicity insyntax, Stanford, June 24-26, 1983. 1985.7. CRAIG, Colette (ed.): Noun Classes and Categorization. Proceedings of a symposium oncategorization and noun classification, Eugene, Oregon, October 1983. 1986.8. SLOBIN, Dan I. & Karl ZIMMER (eds): Studies in Turkish Linguistics. 1986.9. BYBEE, Joan L.: Morphology. A Study of the Relation between Meaning and Form. 1985.10. RANSOM, Evelyn: Complementation: its Meaning and Forms. 1986.11. TOMLIN, Russel S.: Coherence and Grounding in Discourse. Outcome of a Symposium,Eugene, Oregon, June 1984. 1987.12. NEDJALKOV, Vladimir (ed.): Typology of Resultative Constructions. Translated from theoriginal Russian edition (1983). English translation edited by Bernard Comrie. 1988.14. HINDS, John, Shoichi IWASAKI & Senko K. MAYNARD (eds): Perspectives onTopicalization. The case of Japanese WA. 1987.15. AUSTIN, Peter (ed.): Complex Sentence Constructions in Australian Languages. 1988.16. SHIBATANI, Masayoshi (ed.): Passive and Voice. 1988.17. HAMMOND, Michael, Edith A. MORAVCSIK and Jessica WIRTH (eds): Studies inSyntactic Typology. 1988.18. HAIMAN, John & Sandra A. THOMPSON (eds): Clause Combining in Grammar andDiscourse. 1988.19. TRAUGOTT, Elizabeth C. and Bernd HEINE (eds): Approaches to Grammaticalization,2 volumes (set) 199120. CROFT, William, Suzanne KEMMER and Keith DENNING (eds): Studies in Typologyand Diachrony. Papers presented to Joseph H. Greenberg on his 75th birthday. 1990.21. DOWNING, Pamela, Susan D. LIMA and Michael NOONAN (eds): The Linguistics ofLiteracy. 1992.22. PAYNE, Doris (ed.): Pragmatics of Word Order Flexibility. 1992.23. KEMMER, Suzanne: The Middle Voice. 1993.24. PERKINS, Revere D.: Deixis, Grammar, and Culture. 1992.25. SVOROU, Soteria: The Grammar of Space. 1994.26. LORD, Carol: Historical Change in Serial Verb Constructions. 1993.27. FOX, Barbara and Paul J. Hopper (eds): Voice: Form and Function. 1994.28. GIVÓN, T. (ed.) : Voice and Inversion. 1994.29. KAHREL, Peter and René van den BERG (eds): Typological Studies in Negation. 1994.

30. DOWNING, Pamela and Michael NOONAN: Word Order in Discourse. 1995.31. GERNSBACHER, M. A. and T. GIVÓN (eds): Coherence in Spontaneous Text. 1995.32. BYBEE, Joan and Suzanne FLEISCHMAN (eds): Modality in Grammar and Discourse.1995.33. FOX, Barbara (ed.): Studies in Anaphora. 1996.34. GIVÓN, T. (ed.): Conversation. Cognitive, communicative and social perspectives. 1997.35. GIVÓN, T. (ed.): Grammatical Relations. A functionalist perspective. 1997.36. NEWMAN, John (ed.): The Linguistics of Giving. 1998.37. RAMAT, Anna Giacalone and Paul J. HOPPER (eds): The Limits of Grammaticalization.1998.38. SIEWIERSKA, Anna and Jae Jung SONG (eds): Case, Typology and Grammar. In honorof Barry J. Blake. 1998.39. PAYNE, Doris L. and Immanuel BARSHI (eds.): External Possession. 1999.40. FRAJZYNGIER, Zygmunt and Traci S. CURL (eds.): Reflexives. Forms and functions.2000.41. FRAJZYNGIER, Zygmunt and Traci S. CURL (eds): Reciprocals. Forms and functions.2000.42. DIESSEL, Holger: Demonstratives. Form, function and grammaticalization. 1999.43. GILDEA, Spike (ed.): Reconstructing Grammar. Comparative Linguistics and Grammaticalization.2000.44. VOELTZ, F.K. Erhard and Christa KILIAN-HATZ (eds.): Ideophones. 2001.45. BYBEE, Joan and Paul HOPPER (eds.): Frequency and the Emergence of LinguisticStructure. 2001.46. AIKHENVALD, Alexandra Y., R.M.W. DIXON and Masayuki ONISHI (eds.): NoncanonicalMarking of Subjects and Objects. 2001.47. BARON, Irene, Michael HERSLUND and Finn SORENSEN (eds.): Dimensions of Possession.2001.48. SHIBATANI, Masayoshi (ed.): The Grammar of Causation and Interpersonal Manipulation.n.y.p.49. WISCHER, Ilse and Gabriele DIEWALD (eds.): New Reflections on Grammaticalization.n.y.p.50. FEIGENBAUM, Susanne and Dennis KURZON (eds.): Prepositions in their Syntactic,Semantic and Pragmatic Context. n.y.p.