(The Expression of Cognitive Categories) William B. McGregor, William B. McGregor-The Expression of Possession - de Gruyter Mouton (2009)

The Expression of Possession
≥
The Expression of Cognitive Categories
ECC 2
Editors
Wolfgang Klein
Stephen Levinson
Mouton de Gruyter
Berlin · New York
The Expression
of Possession
edited by
William B. McGregor
Mouton de Gruyter
Berlin · New York
Mouton de Gruyter (formerly Mouton, The Hague)
is a Division of Walter de Gruyter GmbH & Co. KG, Berlin.
앪
앝 Printed on acid-free paper which falls within the guidelines
of the ANSI to ensure permanence and durability.
Library of Congress Cataloging-in-Publication Data
The expression of possession / edited by William B. McGregor.

p. cm. ⫺ (The expression of cognitive categories ; 2)
Includes bibliographical references and index.
ISBN 978-3-11-018437-2 (hardcover : alk. paper)
1. Grammar, Comparative and general ⫺ Possessives. I. McGre-
gor, William, 1952⫺
P299.P67E925 2009
415⫺dc22
2009038372
Bibliographic information published by the Deutsche Nationalbibliothek

The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie;
detailed bibliographic data are available in the Internet at http://dnb.d-nb.de.
ISBN 978-3-11-018437-2 hb
ISBN 978-3-11-018438-9 pb
쑔 Copyright 2009 by Walter de Gruyter GmbH & Co. KG, D-10785 Berlin.
All rights reserved, including those of translation into foreign languages. No part of this
book may be reproduced in any form or by any means, electronic or mechanical, including
photocopy, recording, or any information storage and retrieval system, without permission
in writing from the publisher.
Cover design: Frank Benno Junghanns, Berlin.
Printed in Germany.
Contents
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
William B. McGregor
English possessives as reference-point constructions and their

function in the discourse . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
Peter Willemse, Kristin Davidse and Liesbet Heyvaert
On the co-variation between form and function of adnominal

possessive modifiers in Dutch and English . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
Jan Rijkhoff
Is possession mere location? Contrary evidence from Maa . . . . . . . . . . . . 107

Doris Payne
Learning to encode possession . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143

Sonja Eisenbeiß, Ayumi Matsuo and Ingrid Sonnenstuhl
Plain vs. situated possession in Czech: A constructional account . . . . . . 213

Mirjam Fried
Attributive possessive constructions in Oceanic . . . . . . . . . . . . . . . . . . . . . . 249

Frantisek Lichtenberk
Possessive clauses in East Nusantara, the case of Tidore . . . . . . . . . . . . . . 293

Miriam van Staden
Possessive expressions in the Southwestern Amazon . . . . . . . . . . . . . . . . . 343

Hein van der Voort
Possession in the visual-gestural modality: How possession is

expressed in British Sign Language . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 389
Kearsy Cormier and Jordan Fenlon
Contributors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 423
Index of subjects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 425
Index of languages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 430
Index of persons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 432
Introduction
William B. McGregor
The nine papers constituting The Expression of Possession deal with a

range of issues concerning the expression of the relation of possession in
human languages. It is not intended to present a comprehensive overview
of what is known about the topic, but rather to provide some flavour of
what is interesting descriptively and theoretically about the expression of
possession, to present fresh perspectives on this well-researched topic, and
to suggest viable prospects for future research. Thus most of the papers
show a strong descriptive orientation, and present detailed and nuanced ac-
counts of possessive constructions in particular languages or areally/geneti-
cally constituted groups of languages. Many of them also address issues of
current relevance, and/or question widely presumed knowledge.
If we are to investigate the ways of expressing possession we are imme-
diately confronted with the question: What is the possessive relation? A
good deal has been written on this topic – see for instance, Seiler (1983);
Taylor (1989: 202–203); Tsunoda (1995); Heine (1997: 3–6, 33–41); Hers-
lund and Baron (2001: 1–4). For present purposes it is sufficient to say that
it is a relational concept that potentially covers a wide range of conceptual
relations between entities, including, for human beings, between persons
and their body-parts and products, between persons and their kin, between
persons and their representations (e.g. names, photographs), between per-
sons and their material belongings (animate and inanimate items they own),
between persons and things that they have usership-rights to or control
over, between persons and cultural and intellectual products, and so on. For
other animates and inanimates a more restricted range of conceptual rela-
tions is generally available.
Most linguists – including the present author and the contributors to this
volume – would probably agree that the definition should be couched in
terms of linguistic factors, rather than purely conceptually. If this approach
is adopted, it becomes apparent that different notions of possession are
generally invoked in different linguistic constructions. For instance, a num-
ber of languages show different constructions depending on how ‘close’ the
possessive relation is. Many languages distinguish between inalienable
possession and alienable possession, where the former is associated with
2 William B. McGregor
the ‘closest’ and most inherent relations (e.g. body parts, kin), the latter
with less close, less inherent relations (e.g. owned material objects). (See
further Chappell and McGregor 1995.)
To assist the the reader track their way through the volume and to iden-
tify the main issues, we now provide detailed summaries of each of the pa-
pers. It is hoped that these summaries will whet the reader’s apetite for the
detailed discussions and arguments in the individual papers, which of ne-
cessity can be at best hinted at here.
But before we begin, it is necessary to establish some basic notions and
terminology. Throughout the book the term possessum (abbreviated PM) is
used in reference to that which is possessed; it is also sometimes used of
the linguistic expression that denotes this item. Correspondingly, the term
possessor (abbreviated PR) is used in reference to the person, animal, or
whatever, that possesses the PM.
Three primary and general types of possessive construction are usually
distinguished, attributive, predicative, and external. As it is usually used,
the term attributive possession refers to constructions in which the PM and
PR expressions form an NP, as in my dog, the king of France’s bald pate,
and Cliff’s ankle. These constructions are also termed adnominal possession.
By contrast, predicative possession is used of constructions in which the
possessive relation is expressed in a predicate, often by a possessive verb, as
in I have a dog and The king of France has a bald pate. External possession
constructions (EPCs) are constructions in which the possessive relation is
not specified either by the lexical verb or within the NP – the PM and PR
expressions, that is, do not belong to an NP – but rather at the level of a
clausal construction, as in The dog bit Cliff on the ankle (see further Payne
and Barshi 1999). Sometimes the term internal possession construction
(IPC) is used instead of adnominal possession, especially when a contrast is
being drawn with EPCs.
In the first paper, English possessives as reference-point constructions
and their function in the discourse, Peter Willemse, Kristin Davidse and
Liesbet Heyvaert enquire into the information status of the PM referent in
English adnominal possessive constructions with prenominal possessors
(PRs), as in Greta Garbo’s knickers. Based on corpus data, they argue that
the standard analyses of possessives as mere definite NPs (as per e.g. Quirk
et al. 1985; Lyons 1999; Rosenbach 2002), or as NPs presupposing the
identifiability of their referent (Du Bois 1980; Martin 1992) are problematic.
At the same time, Taylor’s (1996) contrary prediction – based on the theory
of possessive NPs as reference-point constructions (as per e.g. Langacker
1995; Taylor 1996) – that PM referents will be overwhelmingly discourse-
Introduction 3
new and anchored to a given PR, is also problematic: in most instances PM

referents turn out to be discourse-given at least to some extent. Willemse,
Davidse and Heyvaert argue that a taxonomy of discourse statuses must be
recognised ranging from fully discourse-given, through text reference, in-
ferable, ‘anchoring’, ultimately to fully discourse-new. They show that PM
referents of adnominal possessive NPs in English may have discourse
statuses at any point on this taxonomy.
This paper adds a needed discourse dimension to the reference-point
theory of adnominal possessive NPs. It demonstrates that it is not sufficient
to study these constructions in isolation; account must be taken of the dis-
course context in which they occur. More generally, the need for discourse
studies of adnominal possessive constructions in other languages is indi-
cated.
Jan Rijkhoff’s On the co-variation between form and function of ad-
nominal possessive modifiers in Dutch and English deals with adnominal
possessive modifiers of common nouns denoting concrete objects that are
introduced by van ‘of’ in Dutch and of in English. He argues that these ad-
nominal possessives can fill three of the five modifier functions distin-
guished in his Functional Grammar-inspired layered model of the NP (Rijk-
hoff 2002). In particular, they can serve a classifying function (indicating
the kind of entity being referred to, as in a man of prayer), a qualifying
function (indicating a property of the entity, as in a woman of great
beauty), and a localizing function (indicating a location of the entity either
in physical or conversational space, as in the bicycle of his father). Possibly
in some languages adnominal possessives can serve discourse referential
functions as well (e.g. if the third person singular possessive pronoun is
used as a definiteness marker); however, it seems that in no language do
they fill the fifth, quantifying function.
Rijkhoff identifies three grammatical parameters with respect to which
van ‘of’ and of adnominal possessives vary: modification (whether or not
the possessive can be modified), predication (whether the possessive can
occur in predicate position), and reference (whether the possessive can be
referential). He argues, on the basis of usage data garnered from the internet,
that the variation in grammatical and semantic properties found in expres-
sions involving these markers correlates with the grammatical role of the
modifier in the NP.
It emerges clearly from Rijkhoff’s contribution that adnominal possessive
constructions marked by van ‘of’ in Dutch and by of in English do not serve
a unique modifying function in either language. An important consequence
is that attributive possession is not – at least in Dutch and English – a unitary
category, but embraces a range of emically distinct subtypes. The extent to

which this applies in other languages as well demands investigation – cf.
below on van Staden’s chapter for a different problem.
In her contribution Is possession mere location? Contrary evidence from
Maa, Doris Payne addresses the question of whether possession is concep-
tually identified with location, as has been widely presumed in typological
and cognitive linguistic literature: the “possession-is-location” view, as she
dubs it. She argues – on the basis of elicited and corpus examples in the
Nilo-Saharan language Maa (Tanzania and Kenya) – that (at least in Maa)
possession is cognitively distinct from location, and consequently that Pos-
sessor and Locative are distinct roles. Payne demonstrates that the verb tii
‘be at’ has locational and existential uses in particular constructions, while
the verb ata ‘have’ has possessive and existential uses in particular con-
structions. However, there is no evidence that tii ‘be at’ can be used to
predicate possession, or that ata ‘have’ can be used to predicate location.
The non-overlap in locative and possessive senses for these two roots would
be surprising if possession and location were conceptually indistinguishable.
What the Maa verbs tii ‘be at’ and ata ‘have’ share is use in existential
constructions. This leads one to suspect that in languages in which posses-
sion and location are represented by the same verb or construction the cog-
nitive connection might be indirect, via the existential sense. This indicates
the need for careful typological and historical investigations to test whether
there might be closer connections between possession and existence, and
location and existence, than between possession and location. Recent re-
search on Nyulnyul (non-Pama-Nyungan, Australia) lends plausibility to
the hypothesis. Nyulnyul has two formally similar negative constructions, a
negative possessive and a negative existential/presentative; there is no evi-
dence of any diachronic link via a locative construction, sense or use.
Payne accepts that the fact that various languages use a single lexical
item to express both possession and location is indicative of the existence
of conceptual connections between the two domains; this does not, how-
ever, imply that they are conceptually identical. Indeed, in many languages
where the same lexical item is used for expressing both location and pos-
session, different constructions are employed, as Payne shows for Jakaltec
and Amharic. This observation further underlines the need to go beyond
mere lexical identity, and to recognize the relevance of constructions, a
point made in other contributions to this volume.
Learning to encode possession, by Sonja Eisenbeiß, Ayumi Matsuo, and
Ingrid Sonnenstuhl deals with the acquisition of the expression of posses-
sion. Its main focus is on the acquisition of adnominal, predicative, and ex-
Introduction 5
ternal possession in German, although English and other languages are also
mentioned. In contrast to many acquisition studies, this one focusses on the
encoding of possessive relations, rather than merely uses examples of pos-
sessive constructions as instances of morpho-syntactic phenomena to be
acquired. It is based on a corpus of children acquiring German monolin-
gually, which includes data elicited utterances by stimuli designed to elicit
possessive relations; child-directed speech is also included.
It is shown that the target constructions emerge step by step, and that
similarly the range of possessive functions encoded in the constructions in-
creases over time. In the earliest stages, adnominal possession constructions
appear not to exist. In the first stages, possessive relations need not be spo-
ken of at all; when they subsequently are, it may begin with single word ut-
terances that just identify the PR, even if the child is in the two-word stage.
Prepositional possessive constructions involving von ‘of’ emerge later, the
–s genitive construction even later. One of the interesting issues in the ac-
quisition of the genitive –s concerns the acquisition of language specific
constraints: in German, its restriction to unmodified PR nouns. The child
acquiring German appears not to generalize the genitive to nominals with
modifiers; instead, they adopt a strategy such as the omission of a required
modifier (even where otherwise they would use the modifier), or omission
of the marker itself. The developmental stages may overlap: constructions
involving target morphemes obligatory in the adult language may alternate
with constructions lacking them in the child’s speech. Moreover, there are
often lexical restrictions, whereby the morphemes are initially restricted to
particular lexical nouns, and only later generalize.
Less well studied is the acquisition of predicative possession construc-
tions and EPCs, especially the latter, and comparatively little data is avail-
able. This paper thus presents significant new data on the acquisition of
these constructions. Eisenbeiß, Matsuo, and Sonnenstuhl show that HAVE-
constructions emerge prior to BELONG-constructions, and show fewer de-
viations from the adult norm, consistent with the notion that HAVE-
constructions are less marked. The German child language data shows
EPCs only rarely, and quite late. In many circumstances in which adult
speakers prefer a dative EPC, children often employ IPCs. Where they do
produce something different, it often differs from the target dative EPC. In-
terestingly, Japanese children show no tendency to extend the Japanese da-
tive construction to EPCs. How and when the double subject and double
object EPCs of Japanese (e.g. Tsunoda 1995) are acquired is not known.
More generally, research on the acquisition of predicative possession and
EPCs in a wider sample of languages is called for.
In Plain vs. situated possession in Czech: a constructional account, Mir-

jam Fried discusses the motivation for use of IPCs rather than the dative
EPC in Czech. Two genitive IPCs are distinguished in Czech which show
possessive marking of the PR, but differ in the relative order of the PM and
PR expressions. In the EPC, the PM and PR are denoted by syntactically
separate NPs, with dative marking of the PR NP.
Fried argues that the genitive IPCs and the dative EPC contrast not just
formally, but also semantically and pragmatically. She shows, using corpus
data, that IPCs can be used for virtually any possessive relation, while the
possessive relation in EPCs is more constrained. EPCs, she argues, are
strongly associated with the inalienablility of the PM and the affectedness
of the PR (hence her label “affected PR”), as is widespread cross-
linguistically (e.g. Payne and Barshi 1999). More precisely, Fried shows
that the PM in an EPC is an entity associated with the personal domain of
the PR, as per Bally (1926/1995). The possessive relation between the PM
and PR in EPCs cannot be pinned down precisely in terms of the semantic
features of the PM and PR; rather, what is relvant is the way each is con-
strued in context – hence Fried’s label “situated possession”, in contrast
with “plain possession” of the IPCs. Genitive IPCs and dative EPC, Fried
argues, encode distinct conceptualizations of possession.
The formal, semantic and pragmatic differences between the genitive
IPCs and the dative EPC indicate that they represent separate constructions
in the Construction Grammar sense (e.g. Goldberg 1995). Fried proposes
Construction Grammar analyses of the constructions which represent their
formal characteristics and meanings. The dative EPC is situated with re-
spect to other nearby constructions, including the genitive IPCs and the Da-
tive of Interest construction, thus demonstrating that they occupy overlap-
ping domains in the functional space of attributive possession. The para-
digmatic relations among the constructions is relevant to their meanings
and uses.
Frank Lichtenberk’s Attributive possessive constructions in Oceanic
shows that Oceanic languages typically distinguish two main types of at-
tributive possession construction which are semantically and pragmatically
distinct. In one type, the direct type, the PM carries an affix cross-refer-
encing the PR; in the other type, the indirect type – which falls into a num-
ber of subtypes – the cross-referencing affix is carried by a possessive clas-
sifier element. In both types the affix cross-referencing the PR is most
commonly a suffix, and the PR may be also realized by a full NP in addi-
tion; in some languages cross-referencing is reduced to an invariant third
person singular affix if the PR is also represented by a full NP. (Further
Introduction 7
complexities are identified and discussed in Lichtenberk’s article; these

need not concern us here.)
As expected given its morphological simplicity (see e.g. Chappell and
McGregor 1989, 1995), the direct type of possession construction strongly
tends to be used in expressing inalienable possession; the indirect types
strongly tend to be used for alienable possession. In the indirect types, the
possessive classifier indicates what category of possession the PM repre-
sents. As Lichtenberk observes, the possessive classifier categorises the PM
in respect to the relation of possession it exhibits to the PR. Possessive
classifier systems show properties characteristic of classifier systems gen-
erally, including fluidity (the possibility of assignment of items to more
than one category with accompanying meaning differences), and the exis-
tence of unpredictable classifications (exceptions to regularity in meaning,
where choice of possessive classifier is semantically and/or pragmatically
unexpected or irregular).
A system of three possessive classifiers can be reconstructed for proto-
Oceanic, distinguishing PMs for eating, drinking, vs. general. In some lan-
guages (particularly Micronesian languages) the system has been elaborated
to make more distinctions. In other languages it has remained a ternary sys-
tem, or has been reduced to a binary system. In a few languages, the pos-
sessive classification system has eroded completely, leaving a single pos-
sessive construction marked by a reflex of a former classifier (which may
or may not contrast with a direct construction). Lichtenberk argues that the
classifier system arose in the context of alienable possession due to their
wide range of possible interpretations. For inalienable possession, by con-
trast, the relation between PR and PM is far more predictable, and there is
thus less motivation to distinguish categories.
In Possessive clauses in East Nusantara, the case of Tidore, Miriam van
Staden shows that the Papuan language Tidore is typologically unusual in
that it does not distinguish between attributive and predicative possession,
contra the widespread belief that the distinction is universally maintained
(e.g. Heine 1997: 26). All five possessive constructions in Tidore behave
not as NPs, but as clauses. Four of them express predicative possession ex-
clusively, while one construction, the focus of the paper, expresses etically
both predicative and attributive possession. This rather simple construction
consists of a bare noun expressing the PM and a prefix specifying the PR in
terms of person and number; an NP denoting the PR may also occur ini-
tially.
Tidore is fairly typical of the languages of the East Nusantara region: in
both Austronesian and Papuan languages the possessive constructions in-
volve verbal elements and retain some clausal characteristics. Thus in some
languages the possessive construction involves a ligature deriving from a
‘have’ verb, while in others the possessive construction involves attach-
ment of a marker cross-referencing the PR to the PM nominal, this marker
being identical or almost identical with the corresponding verbal subject or
object marker. Languages differ, however, in the extent to which the con-
struction is grammaticalised; in some languages the possessive marker is
clearly verbal, in others it is a ‘ligature’. The word order in possessive con-
structions is PR–PM rather than PM–PR as elsewhere in Indonesia and
throughout the Austronesian world; this is also a reflection of the verbal/
clausal nature of the construction. Showing as it does that the distinction
between predicative and attributive possession need not always be system-
atically maintained, van Staden’s paper nicely complements Jan Rijkhoff’s
paper, which, it will be recalled, argues that adnominal attributive posses-
sion in Dutch and English do not constitute a single (emic) construction
type.
In Possessive expressions in the Southwestern Amazon, Hein van der
Voort discusses adnominal possession constructions in eight unrelated lan-
guages of the linguistically diverse southwest Amazon region: three lan-
guage isolates and five languages belonging to different genetic families.
van der Voort identifies two general types of adnominal possessive con-
struction (each of which occur in a range of subtypes): Type I in which a
general possessive marker is added to the PR; and Type II in which the PM
hosts PR agreement or person marking morphemes. In Kwaza and Aikanã
(both isolates), both types of marking are found, though Type I predomi-
nates, and Type II is reduced and used only with third person singular PRs.
Baure is also unusual in that it has a full set of PR marking morphemes, as
well as a general possessive marker that is attached to optionally possessed
nouns (but not to obligatorily possessed or unpossessable nouns), and not to
the PR as in Type I.
The main focus of van der Voort’s paper is on Kwaza. Kwaza is interest-
ing not just because it shows both types of possessive construction, but also
because of the way Type I possession is marked, by the form -dyh. The
analysis of this form is somewhat uncertain. The balance of evidence seems
to indicate that it is synchronically unanalysable. However, it is possible
that it is analysable, at least diachronically, into a possessive marker -dy-
(homophonous with a range of morphemes with related meanings, includ-
ing causative/benefactive) and a nominalising derivational morpheme -h,
which is also used as a neutral classifier (see van der Voort 2004 for dis-
cussion of the classifiers). Interestingly, in possessive constructions -h can
Introduction 9
be replaced by a specific classifier. (This situation is not to be confused

with the type of possessive classifiers found in Oceanic languages which
categorise the PM in regard to the possessive relation itself, as discussed in
Lichtenberk’s contribution to this volume – see above. In Kwaza, the classi-
fier categorises the PM independently of the possessive relation, and is not
restricted to possessive constructions.)
Another unusual feature of Kwaza is that there is a semantically empty
dummy lexeme e- which is either homophonous with a ‘have’ verb, or can
be used used to express predicative possession (van der Voort inclines to the
second possibility). Such semantically empty lexemes are found in other
languages of the region, though not with the ‘have’ sense.
The final paper deals with a topic that has not received a great deal of
attention in the literature on possession, the expression of possession in
sign languages. In this paper, Possession in the visual-gestural modality:
how possession is expressed in British Sign Language, Kearsy Cormier and
Jordan Fenlon focus on the expression of attributive and predicative pos-
session in British Sign Language (BSL), and make some comparative re-
marks on similarities to and differences from other sign languages.
Attributive possession is expressed in two main ways. In pronominal
possession it is indicated by a set of possessive pronouns that differ in form
from the cardinal possessive pronouns, and which precede the PM expres-
sion. In nominal possession PR and PM NPs are linked by a free possessive
pronoun serving as a possessive copula, a pattern also found in spoken lan-
guages (see e.g. McGregor 2001). Interestingly, the PR may be omitted if it
is inferable (see also Fried’s paper in this volume). Older signers also use a
finger-spelled S, borrowed from signed English systems, as a clitic at the
end of nominal PR expressions. This mode of expression, obsolescent in
BSL, is still viable in Australian Sign Language (Auslan) and American
Sign Language (ALS).
Cormier and Fenlon also examine whether the expression of possession
in sign languages is comparable with its expression in spoken languages.
They draw out a number of commonalities, including the distinction be-
tween attributive and predicative possession, the existence of an optional
alienable-inalienable contrast in BSL, and the existence of both HAVE and
BELONG verbs. They also observe that there are connections between pos-
sessive expressions and locative expressions, and a tendency for possessive
morphemes to grammaticalise, which has happened to a limited extent to
possessive pronouns in BSL. As this paper makes clear, sign languages
must be taken into account in serious typological investigations of posses-
sion.
As the above summaries indicate, a diverse range of genetically and ty-

pologically different languages are discussed in the volume; these languages
are also quite widely spread geographically, languages from Europe, South
America, the Pacific region, and Africa being represented. Familiar lan-
guages such as English, Dutch, and Czech are included as well as less well
known languages such as BSL and Kwaza. The papers dealing with the
former set reveal that we still have a lot to learn about possession even in
the best described languages.
The papers of this volume are all strongly empirically oriented, and many
are based squarely on usage data, including (where possible) instances from
corpora. Constructed examples also play an important role in many of the
papers, in some cases by necessity (e.g. where the bulk of the available data
consists of elicited utterances), in some cases to fill gaps in usage data or to
identify what is grammatically impossible. The orientation to usage is aug-
mented by deep concern with meaning. Thus contributions are on the whole
not content with just identifying formal construction types, but also make
serious attempts to determine what differences in meaning might be encoded
or implicated by the alternative modes of expression.
The empirical orientation of the contributions does not mean that theory
is eschewed. Indeed, the majority of papers in the volume are concerned in
one way or another with linking empirical observations with theoretical
concerns. Thus some papers explicitly confront theory with empirical data;
some use theory in an attempt to explain empirical observations; and some
use theory to account for the semantics and/or pragmatics of possessive
constructions. Granted the concern with meaning and usage, it is not sur-
prising that theories from the functionalist end of the spectrum are most
strongly represented; however, the contribution by Eisenbeiß, Matsuo, and
Sonnenstuhl shows that there may be some chance of approachment be-
tween formalist and functionalist theories – and that in some domains at
least, similar predictions are being made.
The need for further study of the means used by languages to encode re-
lations of possession is manifest: we are a long way from having a complete
knowledge and understanding of the formal modes of expression. In his
contribution van der Voort mentions that the Amazonian isolate Movima
expresses inalienable and predicative possession by reduplication (Haude
2006: 238ff.); one wonders what other unusual modes of expression might
remain undetected in the thousands of undescribed languages of the world.
It is hoped that the papers in this volume will stimulate readers to investi-
gate possession in more languages, spoken and signed. It is also hoped that
they will stimulate further research on the well described languages.
Introduction 11
References
Bally, Charles
1926 / The expression of concepts of the personal domain and indivisibility
1995 in Indo-European languages. In The Grammar of Inalienability: A
Typological Perspective on Body Part Terms and the Part-Whole
Relation, Hilary Chappell and William B. McGregor (eds.), 31–61.
Berlin /New York: Mouton de Gruyter.
Chappell, Hilary and William B. McGregor
1989 Alienability, inalienability and nominal classification. Berkeley Lin-
guistics Society Proceedings 15: 24 –36.
1995 Prolegomena to a theory of inalienability. In The Grammar of Inal-
ienability: A Typological Perspective on Body Part Terms and the
Part-Whole Relation, Hilary Chappell and William B. McGregor
(eds.), 3–30. Berlin /New York: Mouton de Gruyter.
Goldberg, Adele E.
1995 Constructions: A Construction Grammar Approach to Argument
Structure. Chicago and London: University of Chicago Press.
Du Bois, John W.
1980 Beyond definiteness: The trace of identity in discourse. In The Pear
Stories: Cognitive, Cultural and Linguistic Aspects of Narrative Pro-
duction, Wallace L. Chafe (ed.), 203–274. Norwood: Ablex.
Heine, Bernd
1997 Possession: Cognitive Sources, Forces and Grammaticalization.
Cambridge: Cambridge University Press.
Haude, Katharina
2006 A Grammar of Movima. PhD thesis, Radboud Universiteit Nijmegen.
(available on http://webdoc.ubn.ru.nl/mono/h/haude_k/gramofmo.pdf)
Herslund, Michael and Irène Baron
2001 Introduction: Dimensions of possession. In Dimensions of Posses-
sion, Irène Baron and Michael Herslund (eds.), 1–25. Amsterdam:
John Benjamins.
Langacker, Ronald W.
1995 Possession and possessive constructions. In Language and the Cog-
nitive Construal of the World, John R. Taylor and R. E. MacLaury
(eds.), 51–79. Berlin: Mouton de Gruyter.
Lyons, Christopher
1999 Definiteness. Cambridge: Cambridge University Press.
Martin, James R.
1992 English Text: System and Structure. Amsterdam: John Benjamins.
McGregor, William B.
2001 Non-verbal predicative possession in Nyulnyulan languages. In Forty
Years on: Ken Hale and Australian Languages, Jane Simpson, David
Nash, Mary Laughren, Peter Austin and Barry Alpher (eds.), 337–
352. Canberra: Pacific Linguistics.
Payne, Doris L. and Immanuel Barshi (eds.)
1999 External Possession. Typological Studies in Language, 39. Amster-
dam: John Benjamins.
Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech and Jan Svartvik
1985 A Comprehensive Grammar of the English Language. London: Long-
man.
Rijkhoff, Jan
2002 The Noun Phrase. Oxford: Oxford University Press.
Rosenbach, Annette
2002 Genitive Variation in English. Berlin /New York: Mouton de Gruyter.
Seiler, Hansjakob
1983 Possession as an Operational Dimension of Language. Tübingen:
Narr.
Taylor, John R.
1989 Linguistic Categorization: Prototypes in Linguistic Theory. Oxford:
Clarendon Press.
Taylor, John R.
1996 Possessives in English. Oxford: Oxford University Press.
Tsunoda, Tasaku
1995 The possession cline in Japanese and other languages. In The Gram-
mar of Inalienability: A Typological Perspective on Body Part Terms
and the Part-Whole Relation, Hilary Chappell and William B.
McGregor, (eds.), 565–630. Berlin /New York: Mouton de Gruyter.
Voort, Hein van der
2004 A Grammar of Kwaza. Mouton Grammar Library 29. Berlin /New
York: Mouton de Gruyter.
English possessives as reference-point constructions
and their function in the discourse
Peter Willemse, Kristin Davidse and Liesbet Heyvaert
1. Introduction
In this article we will be concerned with the discourse status of the posses-
sum (PM) referent of prenominal 1 possessive NPs such as his car, John’s
car.2 With regard to this issue we find two opposed, indeed paradoxical,
claims in the literature: (a) possessive NPs are definite NPs (e.g. Quirk et
al. 1985; Lyons 1999; Rosenbach 2002) or they presume the identifiability
of the PM referent (e.g. Du Bois 1980; Martin 1992), and (b) possessive
NPs introduce overwhelmingly new, previously unknown, PM referents into
the discourse by linking them to typically given possessor (PR) referents
(Taylor 1996). In this article we will argue, basing ourselves on real usage
data,3 that the analysis of possessive NPs as either mere definite NPs, or as
NPs which typically introduce new referents in the discourse, cannot be
maintained. We will propose that (a) possessive NPs have an identification
mechanism different from that found in NPs with definite articles or de-
monstratives, and (b) the question of the discourse status of PM referents of
possessive NPs cannot be reduced to a binary distinction between new or
given in the discourse. On the basis of a qualitative and quantitative analy-
sis of a corpus of possessive NPs in extensive discourse contexts, we will
argue that many PM referents have a discourse status in between fully
given and fully new. For this range of discourse statuses we will propose a
continuum-like classification.
1
As per Taylor (1996: 2), we use the term prenominal possessives to refer to pos-
sessive NPs in which a genitive or a possessive determiner precedes the head
noun, as opposed to ‘postnominal’ possessives such as a friend of John’s.
2
This excludes NPs with ‘classifying’ or other non-determining genitives like the
lion’s share, a mother’s boy.
3
The data we used were extracted from the COBUILD Bank of English corpus
(examples are marked ‘CB’) and the Lancaster-Oslo-Bergen corpus (examples
are marked ‘LOB’). The data from the COBUILD corpus are reproduced with
the kind permission of HarperCollins Publishers Ltd.
14 Peter Willemse, Kristin Davidse and Liesbet Heyvaert
2. English prenominal possessives in the literature
2.1. Alleged definiteness of possessive NPs and presupposed

identifiability of their referents
Possessive NPs have predominantly been associated with definiteness in

the literature. Abbott (2004: 122), in her chapter on definiteness in the
Handbook of Pragmatics, states that possessive NPs “are almost universally
considered to be definite”. Specifically with regard to English, many de-
scriptive grammars classify possessive NPs as belonging to the paradigm of
definite NPs, implying a far-reaching parallelism between possessive de-
terminers and genitives on the one hand, and definite determiners such as
the definite article on the other. Quirk et al. (1985: 326) analyze “the geni-
tive construction [i.e. the prenominal possessive, PW/KD/LH] as a noun
phrase embedded as a definite determinative within another noun phrase”.
Biber et al. (1999: 271) claim that “possessive determiners make NPs defi-
nite”. Huddleston (1988: 90–91) classifies possessives as “determiners that
mark the NP as definite”. Besides grammars of English, other accounts
have classified possessive NPs as definite. Lyons (1999: 23ff.) also holds
that in English, “possessives render the noun phrase which contains them
definite”. He supports this claim with the argument that it is generally pos-
sible to paraphrase possessive NPs with NPs marked by the definite article,
for instance:
(1) my cousin the son/daughter of my aunt and uncle

the man next door’s car the car belonging to the man next door
Rosenbach (2002: 14) proposes a similar analysis of possessive NPs as

definite on the basis of definite paraphrases, and claims that this holds true
even if the genitive has itself indefinite marking, as illustrated by the fol-
lowing examples:
(2) the king’s daughter the daughter of the king

a king’s daughter the daughter of a king
In addition to claims classifying possessive NPs paradigmatically with

definite NPs, some scholars have linked possessives to the concept of ‘pre-
supposed/presumed identifiability’ of referents, which is generally assumed
to be the meaning signalled by the definite article (see e.g. Chafe 1976;
Gundel, Hedberg and Zacharski 1993, 2001; Langacker 2002: 33). Du Bois
(1980: 218), whose early analysis of possessive NPs has been rather influ-
English possessives as reference-point constructions 15
ential, proposes that “[p]ossessive noun phrases (…) presuppose identifi-

ability”.4 In his characterization of NPs in the ‘Pear Stories’ data, Du Bois
characterizes several examples of possessive NPs as definite. The NP his
hat in the following example is described as a definite initial mention of a
referent:
(3) … when he turns around his hat flies off. (Du Bois 1980: 243)
It is added that “[h]is is similar to the in that it demands (presupposes) iden-

tifiability, but different in that it supplies some extra information that may
help make the identification possible” (Du Bois 1980: 243). Martin (1992:
132) situates possessive NPs within the general class of ‘phoric’ NPs, i.e.
NPs coding their referent as in some way retrievable. He bases his claims
about possessive NPs on Du Bois (1980) as well as on Halliday and Hasan
(1976: 70), who analyze possessives as realizing a type of specific deixis
similar to that construed by the definite article and the demonstratives
this/that/these/those. Importantly, the identifiability of the referent is con-
sidered by Martin (1992) to be coded by the possessive NP; in other words,
there is something about the use of a possessive NP to refer to a referent
that signals its identifiability. The key to the identification of the PM refer-
ent is said to be the PR: “[t]his is after all literally what the grammar of the
English nominal group argues: ‘recover the identity of the possessed par-
ticipant here through its possessor’” (Martin 1992: 133). Martin also pays
explicit attention to the questions how the referents of possessive NPs are
embedded in the discourse and how they participate in the reference chains
which are construed by NPs with anaphoric and cataphoric deictics. He
recognizes that possessive NPs have two discourse referents, the possessed
and the PR, and he considers the question whether the possessed has an
identifiability status of its own. However, following Du Bois’ (1980: 243–
245) claim that “a frog of his or a friend of John’s do not alternate with his
frog or John’s friend to introduce participants” (Martin 1992: 132–133), he
concludes that “possessive deictics are the deixis of the participants they
possess” (Martin 1992: 132). Consequently he (1992: 133) proposes that
“possessive nominal groups … only be coded once for phoricity”. In his
actual text analyses of phoric relations, he systematically analyzes posses-
sive deictics as referring back to the PR, leading to reference chains such
4
It is not fully clear whether Du Bois uses “presupposed identifiable” in the tech-
nical sense of ‘already available in the discourse context as a singled-out entity’
(Langacker 2002: 33).
as: the cat she her dinner (1992: 143), the boy his frog (1992:
144). His phoric chains, in other words, track the identifiability status of the
PR, not of the PM referent.
We can now sum up the main elements in the tradition that views pos-
sessive NPs as definite and their referents as identifiable, and formulate our
main criticisms with regard to them. Firstly, linguists adhering to this
analysis of possessive NPs often invoke systematic alternation of posses-
sive NPs with definite NPs as a grammatical argument for according
definiteness to the former. However, there is textual evidence that this al-
ternation is not as systematic as claimed. Possessive NPs in predicative
copular sentences, for instance, do alternate with indefinite NPs, e.g. You’re
my friend – You’re a friend of mine (see also Declerck 1986: 32). Likewise,
NPs with indefinite genitives such as a friend’s friend in (4) can, against
Rosenbach’s (2002: 14) claim, alternate with indefinite NPs, e.g. a friend of
a friend in (5).
(4) It is not to be wondered at [...] that a friend’s friend, described by letter,

should turn out an unrecognizable stranger. (CB)
(5) Jimmy was a referral, a friend of a friend. (CB)
Our data contained other examples in which possessive NPs alternate with
indefinite NPs, such as the following:
(6) On Monday, Christie’s in New York is to sell Greta Garbo’s knickers.

(CB)
In this example, the prenominal possessive Greta Garbo’s knickers alter-
nates with indefinite (a pair of) knickers of Greta Garbo. In view of this, it
is hard to maintain that English prenominal possessives code definiteness,
since a subset, viz. those alternating with indefinite article + noun + of +
PR, are functionally indefinite. In this context, it can be recalled that other
languages, such as Italian and Spanish, code (in-)definiteness and prenomi-
nal possession separately, making a distinction between, for instance, il mio
libro (‘the my book’) and un mio libro (‘a my book’) (Lyons 1999: 24). In
English prenominal possessives, the contrast definite – indefinite remains
covert, but it can be made explicit in the corresponding complex NPs in
which the PR is expressed by postmodifier of + NP.
Secondly, if a functional definition of possessive NPs is given, it is ob-
served that the PM is retrievable through the PR (Martin 1992: 133). This
explanation refers to the identification mechanism internal to possessive
NPs. However, this NP-internal identifying relation has to be distinguished

from the external relations which the two discourse referents may maintain
with other elements in the surrounding discourse. If we look more closely
at the latter, two distinct perspectives can be taken. On the one hand, as il-
lustrated by Martin’s (1992) text analyses, possessive NPs can be viewed as
partaking in reference chains keeping track of the PR referent. As the PR is
typically given, co-reference to the PR will account for a large part of the
reference chains construed by possessive NPs. However, in some – admit-
tedly fewer – cases, the PM referent may also be coreferential with a dis-
course referent, as in the following example, where Fleming’s cardboard
booby refers back to ‘Bond’.
(7) Goldfinger, the third Bond movie, was released in December of that
year, and with it was founded an industry that would turn Fleming’s
cardboard booby into a product that 30 years later rivals Mickey
Mouse in terms of global penetration. (CB)
In other words, the PM referent can already be present as a singled-out in-

stance in the discourse, and need not be discourse-new. This shows that the
two referents of possessive NPs insert themselves into the discourse with
distinct given-new statuses, which have to be studied in actual discourse.
Martin’s (1992) observation that the PM referent is recoverable through
the PR hints at the internal identifying relation present in possessive NPs,
but he does not explore it in more detail. The internal identification mecha-
nism of possessive NPs and its ‘anchoring’ of referents has been at the core
of the analysis of possessives as reference-point constructions, developed
within the framework of Cognitive Linguistics by Langacker (1993, 1995)
and Taylor (1996). Taylor (1996) also discusses certain properties of both
the PR and the PM referent from a discourse perspective. We will turn to
this account in the next section.
2.2. Alleged newness of PM referents as targets of reference-point

constructions
Langacker (1993, 1995) proposes that possessive NPs should be analyzed as

reference-point constructions, i.e. as constructions motivated by the relation
between two entities, one of which functions as the ‘reference point’ giving
mental access to the other. More specifically, in possessive NPs, the PR
functions as a reference point for the identification of the PM, which func-
tions as the ‘target’ needing identification. For instance, in the possessive

NP Sarah’s car, the PR, ‘Sarah’, functions as the reference point for the
identification of the ‘car’ which is ultimately being referred to; when proc-
essing this NP, the addressee will initially make mental contact with the
entity ‘Sarah’ and subsequently identify the ‘car’ in question as the one as-
sociated with Sarah (e.g. the one she owns or drives). Since the reference
point serves to give mental access to the target, a reference point is nor-
mally cognitively more easily accessible than the target that is being linked
to it; an entity is, then, chosen as a reference point for another on the basis
of the fact that it has “a certain cognitive salience, either intrinsic or contex-
tually determined” (Langacker 1993: 6).
The reference-point analysis thus offers a detailed account of the identi-
fication mechanism set up within possessive NPs. However, the PR (refer-
ence point) and PM (target) are themselves also discourse referents embed-
ded in the discourse in which the possessive NP is used. Taylor (1996), who
further develops Langacker’s (1993, 1995) reference-point analysis specifi-
cally in relation to prenominal possessives, formulates a number of predic-
tions for PR referents as well as for PMs with regard to their givenness or
newness in the discourse. In order to make these predictions, he is led by
what he perceives to be the inherent logic of the reference-point relation.
Since the aim of using a reference-point construction is to make a target
entity more accessible by tying it to a reference point, Taylor argues, it is to
be expected that the reference point should be more easily mentally acces-
sible than the target, as “it would be perverse indeed to invoke a less acces-
sible entity to aid the identification of a more accessible entity” (Taylor
1996: 210). Conversely, “were the target as easily accessible as the refer-
ence point, there would be no point in using the reference point for its iden-
tification” (ibid.). This line of reasoning leads Taylor (1996: 218) to posit
that PRs and PMs are “maximally differentiated” in terms of their (typical)
discourse properties. PRs, on the one hand, overwhelmingly have ‘given’
status in the discourse, since given entities (i.e., entities already present in
the discourse and known to the addressee) are more cognitively accessible
than newly introduced entities. More specifically, Taylor (1996: 212) pre-
dicts that PRs will typically be “entities mentioned in recently preceding
discourse”, that discourse or text topics will frequently occur as PRs and
that PR nominals will frequently be definite. By contrast, as entities that
need anchoring to a reference point for identification, “possessees over-
whelmingly introduce new, previously unnamed entities into the discourse”
(Taylor 1996: 217).
2.3. Overview and research questions
To sum up, existing accounts of English prenominal possessives have

tended not to fully recognize their complexity as constructions referring to
two discourse referents between which an NP-internal relation of identifica-
tion is set up and which at the same time maintain external relations with
the surrounding discourse context. The discourse status of the PR is rela-
tively straightforward, since it either (in the case of a genitive) has explicit
definite or indefinite marking or (in the case of a possessive determiner) is
realized by an inherently definite pronominal form. New PR referents are
therefore normally realized by an indefinite genitive,5 whereas given PRs
have definite marking or a pronominal form. What is more, the PR referent
has convincingly been established to be overwhelmingly discourse-given
(see Taylor 1991, 1996). By contrast, the discourse status of the PM referent
is controversial and underresearched – both conceptually and empirically.
On the one hand, within the tradition that includes possessive NPs in the
paradigm of definite NPs, the PM referent tends to be viewed as ‘presumed
identifiable’. On the other hand, within the reference-point approach to
possessive NPs, Taylor (1996) assumes that PM referents, as the target of
identification, are overwhelmingly new. Given such opposed claims, closer
investigation of the discourse status of PM referents imposes itself.
Since the PM referent does not have direct definite or indefinite mark-
ing, its discourse status cannot simply be deduced from the form of the NP.
In order to uncover it, possessive NPs must be studied in extensive dis-
course contexts, which make it possible to trace the PM referent’s given-
ness or newness throughout the discourse context preceding the possessive
NP (Willemse 2005: 106–133). For the present study, 400 instances of pos-
sessive NPs have therefore been studied in extensive discourse contexts.
The aim of the analysis was to investigate the ‘external’ relations of the PM
referent to the surrounding discourse context, i.e. to determine whether the
possessive NP was used to introduce a new referent into the discourse or to
refer to a given discourse referent. Carrying out this analysis involved two
steps. Firstly, a theoretical-descriptive framework had to be set up to ac-
count in a precise and systematic way for the various degrees of discourse
givenness/newness displayed by the PM referents. Secondly, this description
had to be applied to the data-base and the results quantified. Analytically,
these are two distinct steps, but in practice the descriptive categories were
5
It should be noted that indefinite grounding genitives are by no means anoma-
lous; see Willemse (2005: 183–200).
set up not only with reference to the literature, but also by shunting between
the description and the data to ensure optimal coverage of the patterns
emerging from the latter.
The remainder of this article is structured as follows. In section 3, we will
look more closely at functional equivalents of (in)definiteness discussed in
the literature. We will focus in particular on the important distinction be-
tween recoverability (relevant to the reference point analysis of possessive
NPs) and that of first – subsequent mention (relevant to the discourse status
of the PM referent). In section 4, we will present the qualitative and quanti-
tative results of the data analysis. In the first place, this involves setting out
our descriptive classification of the different discourse statuses, ranging
from fully given (coreferential), over a number of statuses in between given
and new, to fully new. Secondly, we also report on the quantitative results
of Willemse’s (2005) corpus study, i.e. on the relative frequency with which
these different statuses occur in the data-base. Finally, we will also situate
our descriptive analysis in a broader discourse perspective: possessive NPs
turn out to serve a variety of specific discourse functions such as reiteration
of the PR as a central discourse participant or reclassification of the PM
referent, which cautions against viewing them only as referential identifica-
tion mechanisms along a strict reference point logic. In section 5, we will
summarize our main findings and point to their implications for the analysis
of possessive NPs as reference point constructions and as so-called ‘defi-
nite’ NPs.
3. The functional dimensions of definiteness and possessive NPs
As observed by Lyons (1999), there is less unanimity about the functional

definition of definiteness than is often assumed. In his formally-oriented
study of definiteness, Lyons (1999: chapter 1) includes two main compo-
nents in its semantic definition. The first one is identifiability, understood
as an extended version of the so-called ‘familiarity hypothesis’. The use of
a definite article signals that the hearer should be able to identify the refer-
ent of the NP it occurs in. This may mean that the hearer is familiar with
the actual referent or that he can be directed to it via other mechanisms
such as anaphora and bridging or associative uses, as in a taxi – the driver.
The second meaning component is uniqueness: the definite article signals
that there is just one entity corresponding to the description given, relative
to a particular context. Following Hawkins (1978), uniqueness is extended
to inclusiveness to include plural and mass NPs, for which the definite article
signals that reference is made to the totality of objects or mass in the context
satisfying the description. ‘Inclusive’ reference is treated as secondary to
identifiability by Lyons and as an implicature, rather than a meaning com-
ponent, of definiteness by Hawkins (1978) and Declerck (1986).
Martin (1992) uses the notion of recoverability, rather than identifiabil-
ity, which more clearly suggests that other processes than actual knowledge
of the referent may be involved. According to Martin (1992: 98) “every time
a participant is mentioned, English codes the identity of that participant as
recoverable from the context or not”. When the grammatical resources used
in a specific NP signal that the identity of its referent is in some way recov-
erable, the NP is said to be phoric. Phoric NPs are, consequently, different
types of definite NPs (proper names, pronouns, NPs grounded by the defi-
nite article or by a demonstrative) which all embody “directives indicating
that information is to be retrieved from elsewhere” (Halliday and Hasan
1976: 31). Like Lyons, Martin also stresses that ‘this information’ may per-
tain to the referent itself or to another referent with which it is indirectly
associated. It is in accordance with this logic that Martin (1992: 133), de-
spite recognizing that possessive NPs refer to two referents, codes them
only once for phoricity, as the PM referent is recoverable through the PR
referent. Whilst this is reminiscent of the reference point analysis, it has to
be pointed out that Martin’s recoverability analysis is restricted to function-
ally definite possessive NPs, and does not apply to examples such as (4)
and (6) above. By contrast, the reference point analysis is broader – at least
as a theoretical concept, rather than in Taylor’s (1996) interpretation of it in
relation to possessive NPs. It foregrounds the conceptual relation between
reference point and target and can also be applied to functionally indefinite
possessive NPs.
As recoverability of a referent via association or bridging is not dis-
cussed much in general reference works, we will consider the specialized
literature on it more closely. When a bridging relation holds between two
elements, it may be motivated by one of several types of conceptual rela-
tions, i.e., the ‘source’ of bridging may differ. Generally speaking, the basis
for bridging is some sort of strong associational relationship between enti-
ties, one which is strong enough to allow the identification of one entity on
the basis of an earlier mention of the other, associated entity. This translates
to several more specific relation types, the most important one of which is
the part-whole relation, e.g. (8):
(8) Peter has bought a new car. There is much more room in the boot than
there was in his old car.
In example (8) the part ‘boot’ can be marked as recoverable on the basis of
its relation with the whole that it forms part of, ‘car’. The processing of
bridging reference thus requires some inferencing on the part of the ad-
dressee, who needs to retrieve the conceptual relation which forms the basis
for the identification of the referent from background knowledge (cf. Ariel
1990: 185), while the role of the immediate discourse context is to provide
an element which activates this inference (a ‘trigger’, Hawkins 1978: 123).
Besides part-whole relations, other relations of strong and habitual asso-
ciation between two entities may be strong enough to motivate bridging, as
in (9).
(9) He was very interested in buying that old house up the road, but the
owner wouldn’t sell.
Although an owner is of course not a part of a house, it is an entity typically

associated with a house, which can hence be introduced with the definite
NP the owner. Brown and Yule (1983: 257) emphasize the similarities be-
tween part-whole relations and relations of strong, stereotypical association
by classifying both under the heading of relations of ‘having’ (compare: ‘a
car has a boot’ – ‘a house has an owner’).
Both part-whole relations and relations of strong association are rela-
tions between entities. Definite reference to an entity based on its relation
to an event, activity or situation described in the preceding context occurs
frequently as well. For instance, in the following example, the referent ‘kil-
ler’ is realized with a definite NP, since the event of a murder involves a
killer as the agent of the described action:
(10) It was dark and stormy the night the millionaire was murdered. The
killer left no clues for the police to trace. (Brown and Yule 1983: 258)
In the literature, the invoking of events to make a referent recoverable has

been discussed in terms of notions such as frames, scenarios and schemata.
With reference to Chafe (1972), Du Bois (1980: 235ff.) develops the con-
cept of event-frames, which are “composed of a network of related actions,
along with the people and objects involved in those actions” (Du Bois 1980:
246), and are therefore rather hard to delineate. He notes that definiteness
can be used as an indication to decide what elements make up a particular
frame. The notion of schema is used by Chafe (1996) to capture the way in
which a referent may be inferred from contextual information about events.
He gives the example of someone who is describing the advantages of eating
cream cheese out of a carton and introduces the referent ‘spoon’ with a
definite NP.
(11) and if you just sort of rinse the spoon off afterwards, you don’t really
have to wash dishes […]. (Chafe 1996: 39)
The treatment of ‘spoon’ as an identifiable referent, Chafe argues, is due to

the fact that it was “indirectly shared because of its association with the eat-
ing-out-of-the-carton schema” (Chafe 1996: 39). The notion of scenarios,
finally, was developed by Sanford and Garrod (1981, 1998). Scenarios are
defined as representations of situations and of the roles involved in them,
which are activated by text input (i.e. by explicit ‘clues’ in the text) and
retrieved from long-term memory. Entity tokens (i.e. discourse referents)
are mapped onto the roles which are part of the invoked scenario in a spe-
cific context. Thus, when an appropriate scenario has been activated in a
specific context, discourse referents filling role ‘slots’ in the scenario can
be referred to with a definite NP. Sanford and Garrod (1981: 112ff.) support
this analysis with psycholinguistic evidence that reading times for sentences
containing a definite NP which introduces an entity in the discourse are not
longer when an appropriate scenario has been invoked than when the entity
has been introduced explicitly in the preceding discourse (i.e. when the
definite NP is an anaphor).
The question of which types of relations and associations are apparently
strong enough to motivate bridging has been speculated on a lot in the lit-
erature. Chafe (1972: 63), for instance, suggests that in the case of part-
whole relations, the obligatoriness of a part influences the possibility of
bridging – i.e., the more optional the part is, the less likely it is to be coded
definitely. Ariel (1990: 184 –185) points to a number of factors which may
govern the ‘inferability’ of an entity, such as the stereotypical nature of the
entity to another entity in the context and the prominence of the ‘antece-
dent’ in the discourse. However, she also remarks that much of what de-
termines the inferability of specific (types of) entities is probably governed
by language-specific conventions.
Returning to the functional definition of definiteness, then, a third basic
functional dimension of definiteness has been identified in the text-based
studies of Du Bois (1980) and Fraurud (1990), viz. that of first versus sub-
sequent mentions. Du Bois (1980: 220ff.) and Fraurud (1990: 413ff.) both
point out that it is traditionally assumed that indefinite NPs involve the first
mention of a referent: they are said to establish a new discourse referent and
to instruct the hearer to open a new file in their consciousness. Definite NPs,
by contrast, are thought of as involving the retrieval of previously estab-

lished discourse referents, i.e. non-initial mentions, and as instructing the
hearer to update an old file. Du Bois and Fraurud both also show that this
assumed correlation between indefiniteness – first mention and definiteness –
subsequent mention is confirmed only to a certain extent by actual text anal-
ysis. In their data, there are considerable portions of definite first mentions
and (smaller) sets of indefinite subsequent mentions. The first-subsequent
mention distinction applies only to NPs designating specific referents (Du
Bois 1980: 207); it does not apply to generic NPs and to proper names.
Du Bois discusses three main types of definite first mentions: (i) ones
marked by an unstressed demonstrative; (ii) ones containing identifying
information in a postmodifying relative clause; (iii) ones due to association
with a frame.
In Du Bois’ spoken data unstressed demonstratives this/these were com-
monly used for initial mentions, and by most speakers even restricted to
first mentions, as in he could possibly see this little boy [1st] coming on a
bicycle (Du Bois 1980: 219). The clash that we find here between the defi-
nite determiner and the introduction of a new discourse referent has gener-
ally been noted.
NPs containing specific and new information in the presupposed format
of a restrictive relative clause may also make a definite initial mention “ac-
ceptable”, as Du Bois (1980: 223) puts it, e.g. she knocks the hat that he’s
wearing [1st] off on the ground (Du Bois 1980: 222). That the identifying
information provided by defining relative clauses may motivate the use of a
definite article for the whole NP is, again, well-established.
Finally, a larger whole or a specific activity may serve as the ‘frame’
enabling the definite initial mention of referents typically associated with
them, as with living room – the wall (Du Bois 1980: 233), sell – the money
(Du Bois 1980: 215). Here we are up against the mechanism of bridging
again. However, whereas bridging has generally been discussed in the con-
text of explaining the definite form of the NP and glossing its meaning as
‘recoverable’, Du Bois also stresses that, in spite of the definite form, we
have initial mentions here.
Given Du Bois’ incisive comments about the first mention status of ref-
erents of bridged NPs, it is perhaps surprising that he does not extend this
observation to PM referents of possessive NPs, which, in a considerable
number of cases, are also first mentions with inferential relations to the con-
text. As we will see in section 4.4, possessive NPs may designate instances
that, in a strict sense, are mentioned for the first time in the discourse, even
though they are indirectly related to other elements in the discourse, from
which they are inferable such as the handbrake to the Mitsubishi Starwagon
in (12).
(12) Police prosecutor Snr-Sgt Geoff Jackson told the court Pizzino was
one of eight passengers in the Mitsubishi Starwagon which crashed
on Robina Parkway at the Gold Coast about 5.10 am on Wednesday.
Snr-Sgt Jackson said Pizzino activated the vehicle’s handbrake,
causing the driver to lose control. (CB)
In fact, in many cases the prenominal possessive can be replaced by the

definite article, showing that the conceptual relation between the frame
evoked by the context and the inferred referent is the same as that underly-
ing bridging anaphora, e.g.
(12’) […] Snr-Sgt Jackson said Pizzino activated the handbrake, causing
the driver to lose control.
Moreover, PM referents may also be ‘more new’ to the discourse than in-
ferable first mentions: they may have some link to a conceptual ‘anchor’ in
the discourse that makes their occurrence not entirely unpredictable, but is
not strong enough to intrinsically convey uniqueness/inclusiveness on them
so as to make them ‘recoverable’. This can be illustrated with possessive
NPs referring to clothes and other alienable possessions, which have a cer-
tain conceptual link to their ‘owners’. If the possessions referred to are con-
textually unique or include all the instances in the given context, then such
NPs are functionally definite. In (13) the detective is wearing only one hat,
and (14) is concerned with all of Paxman’s ties.
(13) An Irish detective arrested a wanted criminal in a Dublin street. Just

as he was about to slap the handcuffs on him, a gust of wind blew the
detective’s hat down the street. “Shall I go and fetch it for you?”
asked the criminal. (CB)
(14) Most of Jeremy Paxman’s ties don’t go with his shirts. (CB)
However, if this is not the case, then they are functionally indefinite, as in
example (6) above, Christie’s to sell Garbo’s knickers, and example (15), in
which Ruth has been stealing some, but not all (*the), clothes, jewellery
and accessories of Elizabeth’s. These examples show that the conceptual
relation of alienable possession is not in itself strong enough to make a ref-
erent inferable.
(15) Raven-haired Ruth is a statuesque woman consumed with envy for

the blonde Elizabeth, who Ruth’s parents took in after she was left
orphaned. All her life Ruth has been secretly stealing Elizabeth’s
clothes, jewellery and accessories and dressing up as the girl she al-
ternately idolises and hates. (CB)
Finally, some PM referents are ‘fully new’ in the discourse, unpredictable

from any elements in the context. This is, for instance, the case with actions
(and aspects of actions such as results) expressed by nominalizations or de-
verbal nouns, which can scarcely be predicted from the link to their agent
or patient in the preceding text, as in (16), in which Milner’s prints refers to
‘prints of Milner’s’.
(16) Point two, Milner was the only person who entered Stevens’s apart-
ment the night of the murder – he was positively identified by the
hallman. Point three, we found Milner’s prints [sic] four places in
the apartment including the library. (CB)
From this brief discussion of first-mention PM referents it transpires that

different degrees of discourse newness can be distinguished for them. As
we will see, we can also observe a difference in the degree of discourse
givenness for PM referents that are subsequent mentions. The typology of
discourse statuses of PM referents that we will set out in the next section
will therefore be couched in terms of degrees of discourse-newness and
discourse-givenness.
4. Data analysis
We pointed out in section 2 that one of the most important remaining ques-
tions in the literature on possessive NPs pertains to the discourse status of
their PM referent. In section 3, we linked up the notion of discourse status
with the discussion of initial-subsequent mentions, but we differentiated
this contrast into degrees of discourse newness and givenness. Thus, we are
now in a position to tackle the discourse status of PM referents in real us-
age. We compiled a set of 400 instances of possessive NPs, composed as
listed in Table 1.
Table 1. Overview of the data base
NP type # of instances Source

Genitive + N 200 COBUILD Bank of English (CB)
Total Genitive 200
its + N 50 Lancaster-Oslo-Bergen
corpus (LOB)
their + N [30] + [20] [LOB] + [CB]
my + N 50 CB
her + N 50 CB
Total Possessive Det. 200
Total 400
The central question of our analysis was whether the PM entity is a subse-
quent mention of a discourse referent or a first mention. Seeing the impor-
tance of distinguishing a range of discourse statutes, we propose five cate-
gories which form a continuum from discourse-given (coreferential posses-
sive NP) to discourse-new (possessive NP introducing a new referent).
Table 2. Discourse statuses of the PM referent
the PM referent has been mentioned in the preceding

COREFERENTIAL
discourse and is referred back to
the PM referent is a text referent which is construed on the
TEXT REFERENCE
basis of the preceding discourse
the PM referent is inferable from an associated referent or
INFERABLE
a scenario in the preceding context
the PM referent is ‘anchored’ to (an) element(s) in the
ANCHORED
preceding discourse, which reduces its ‘newness’
NEW the PM referent is newly introduced by the possessive NP
Table 2 gives an overview of these five discourse statuses, which, in the

rest of this section, will be discussed in more detail along with the quantita-
tive results of the corpus study for each of these categories.6
6
From the genitive sample, a few tokens had to be removed because the PM was a
proper name, e.g. The electro sound, which grew out of Bambaataa’s pioneering
4.1. The possessive NP is COREFERENTIAL with another NP in the

preceding discourse
As we saw in section 1, Taylor (1996: 217) proposed that PM referents are

overwhelmingly new entities that are being introduced into the discourse.
Our data show, however, that in a relatively small, though not insignificant,
number of cases, the PM is a referent which has been mentioned in the pre-
ceding discourse and is referred back to by the possessive NP. In such
cases, the PM is, in Chafe’s (1994) terminology, an ‘active’ referent, i.e. a
referent which is given in the discourse and therefore readily available in
the addressee’s consciousness. The possessive NP is thus, in such cases,
coreferential with another NP and may even form part of a longer reference
chain. In addition to the PR referent, which, in case of a definite genitive or
a possessive pronoun, is explicitly marked as being retrievable, the PM may
thus be a subsequent mention as well. Table 3 shows how frequent corefer-
ential or given PM referents were in our data.
‘Planet Rock’, kept the street end of rap cooking (CB). As noted in the previous
section, proper names do not construe the contrast between initial – subsequent
mention. Another small set of tokens of the dataset was left out of the classifica-
tion, because the possessive NP in these cases functioned as part of a fixed ex-
pression (often a prepositional phrase), e.g.
(i) If Laurie noticed she chose not to comment, but she was curious in her way.
(CB)
(ii) In a stern reply, he reminded him that ‘Our troops are elated and confident;
those on the enemy’s side cannot but be depressed.’ (CB)
These expressions are not processed compositionally and, as a result, they do not
have a distinct PM referent with a discourse status of its own. For each category,
the number of data left out of the classification was the following: genitive (15);
its (2); their (4); my (4); her (4). The percentages in the tables are calculated on
the basis of the number of tokens included in the classification.
Table 3. Results for the category coreferential
data set # of tokens percentage

Genitive [CB] 28 /185 15.14 %
Total Genitive 28 /185 15.14 %
its + N [LOB] 1 / 48 2.08 %
their + N [LOB] 4 / 46 8.70 %
my + N [CB] 8 / 46 17.39 %
her + N [CB] 3 / 46 6.52 %
Total Poss. Det. 16 / 186 8.60 %
Grand total 44 / 371 11.86 %
Coreferential possessives account for about 10% of the total data. Still, in a
comprehensive account of the discourse status of possessive NPs, this
makes them a non-negligible category. Let us consider some examples:
(17) Mr Ashby, a former name who suffered substantial losses at Lloyd’s,

had sought damages over an article in January last year alleging
that he shared a double bed with another man on holiday in Goa.
And when the jury found against him in a majority verdict, he put his
head in his hands and wept. […] After the verdict, the newspaper’s
solicitor, Alistair Brett, said he would expect the present editor, John
Witherow, to see the case as ‘a tragic family problem’ and be sensi-
ble about what to do now. […] Senior Tories expressed their deter-
mination to help him after the verdict and launched a campaign ‘to
keep him buoyant’ that was immediately evident in his reception in
the Commons. […] He has 28 days to appeal against the jury’s ver-
dict and it is then likely to take up to nine months for his costs to be
determined by taxation proceedings. (CB)
The referent ‘verdict’ is first introduced by the NP a majority verdict and

subsequently referred back to twice with the definite NP the verdict, before
finally being taken up again by the possessive NP the jury’s verdict. The
possessive NP is in this case clearly part of a reference chain in the dis-
course and the PM referent is not a new referent.
The relation between the PM and the PR is usually already established
in the preceding context, before the two are actually constructionally linked
up in the possessive NP. This is the case in (17), where the link between
‘verdict’ and ‘jury’ is made clear in the preceding context (the jury found
against him in a majority verdict), as well as in the following example:
(18) Britain’s leading arms manufacturers secretly liaised with ministers,

civil servants and the CIA on ways to silence the Saudi opposition
leader Muhammed al-Masari, it was claimed last night. […] Lawyers
acting for Dr. al-Masari were preparing yesterday to appeal against
his removal. […] Dr. al-Masari’s lawyers also allege that the home
secretary decided to push ahead with the removal of the dissident in
spite of a written pledge by the Home Office that his application to
stay in Britain would be considered substantively. (CB)
The indefinite NP marked in bold type introduces the referent ‘lawyers’

while the postmodifier describes the link with the PR ‘Dr. al-Masari’. In
some cases, the direct antecedent of the possessive NP is itself a possessive
NP, as in the following example:
(19) […] Her salvation was also a cause for celebration among the 60
volunteers and 12 staff who run Britain’s only national charity set up
to find missing people. This Christmas, the runaway was one of
14,000 people on the charity’s computer database, which is housed
in a spartan, donated office above a supermarket in East Sheen,
southwest London. […] A lot of adolescent girls aged around 14 and
15 do not get on with their parents. […] Some fall prey to prostitution
and others, among the most urgent on the charity’s database, become
caught up in paedophile rings. (CB)
In such cases, the reference chain which the possessive NP forms part of
consists of several possessive NPs, in which not only the PR (in this case
‘charity’) is linked up with all its previous mentions, but the PM (in this case
‘(computer) database’) as well (this use of possessive NPs is mentioned by
Barker (2000: 214)).
The coreferential use of possessive NPs has important implications for
the reference-point model in its application to possessive constructions, and
underlines the importance of studying their discourse context, instead of
treating them as isolated syntagms and looking at their internal identifica-
tion mechanism only. In actual usage, the reference-point mechanism can
be adapted to specific discourse needs and used for particular ‘rhetorical’
purposes. Thus, the use of a possessive NP to refer to a given referent ap-
pears to be often motivated by the desire to ensure non-ambiguity of the
referent in contexts where there is potential confusion. For example, in (17),
the possessive construal makes it clear that reference is to those lawyers

acting for al-Masari, and not to lawyers working for someone else.
Another ‘rhetorical’ use of coreferential possessive NPs involves em-
ploying a different lexical classification to refer to a given referent. This
may be a synonymous classification or it may (and often does) entail a
more or less drastic recategorisation of the referent. Such cases are treated
as coreferential because, despite the use of a different lexical classification,
or ‘type specification’ as Langacker (1991: 144–148) calls it, the same ref-
erent is referred to. Blanche-Benveniste and Chervel (1966) have described
such cases as anaphore infidèle (‘unfaithful anaphora’), as opposed to ana-
phore fidèle (‘faithful anaphora’), the latter being restricted to cases in
which the referent is not recategorized. It is interesting to note that when a
different type specification is used, it often incorporates additional contex-
tually specified information about the referent. For instance, in (20), the
underwear being referred to is first described as knickers whereas in the
possessive NP, the type specification silk panties is used, which incorpo-
rates the information that the garment in question is made of silk, as indi-
cated in the preceding context.
(20) On Monday, Christie’s in New York is to sell Greta Garbo’s knickers.

They are described with proper dignity. ‘A pair of silk, cream-
coloured ladies’ briefs […] In fact they are a souvenir of what
Christie’s delicately call ‘a night of romance’ between Garbo and
the Mexican star Roland Gilbert. […] When they parted, Roland gave
Garbo the gold ring he was wearing and was given the actress’s silk
panties in exchange. (CB)
Such incorporation of contextual information confirms the idea that refer-

ence is not only about referring ‘back’ to the previous textual mention of a
referent, but rather about activating a mental representation of the referent,
which naturally evolves and is enriched as the discourse progresses and
new information is added (see, among others, Emmott (1997: chapter 7)
and Brown and Yule (1983: 201–204)).7 Additionally, the use of a different
7
Brown and Yule (1983: 202) give the following example (involving pronominal
reference):
(i) Kill an active, plump chicken. Prepare it for the oven, cut it into four pieces
and roast it with thyme for 1 hour.
In such cases, they argue, although the identity of the referent ‘chicken’ does not
change, its description does: “A reader who simply went back up the endophoric
type specification to refer to a referent which is already present in the dis-

course can also be used to categorize that referent according to the subjec-
tive evaluation of the speaker. This is the case in (21), where the referent
‘Bond’ is classified as a cardboard booby:
(21) With the Bond books, as his friend Ernest L. Cuneo wrote, Fleming’s
‘objective was the making of money’ and he succeeded. But it wasn’t
until after his death in August 1964 that Bondmania erupted and the
money really began to flow. Film was responsible. Goldfinger, the
third Bond movie, was released in December of that year, and with it
was founded an industry that would turn Fleming’s cardboard booby
into a product that 30 years later rivals Mickey Mouse in terms of
global penetration. (CB)
4.2. The PM referent is a TEXT REFERENT construed from the preceding

discourse context
As observed by Halliday and Hasan (1976: 52), anaphoric retrieval relations

are not restricted to “an entity that is encoded linguistically as a participant”
but may also involve “any identifiable portion of text”. Possessive NPs
naming a referent which has been construed over a preceding stretch of text
can also realize text reference, as in the following examples:
(22) One bae insider said last week that there could be no formal discus-
sions until Daimler and the Dutch government had resolved the
problems of Fokker, the ailing short-haul aircraft maker that will
collapse unless it receives an emergency cash injection of almost £
1.4 billion. […] But Fokker’s crisis is only one contributor to the
problems of Daimler-Benz Aerospace […] (CB)
(23) Students may soon be offered American-style loans at cheap rates by

leading banks and building societies, after the government promised
lenders generous subsidies to enter the student-loan market. The plan
[…] aims to shift most of the cost of financing student loans on to
high-street lenders. The Department for Education and Employment
chain and substituted the expression an active plump chicken for it in the last
clause would, in a significant sense, have failed to understand the text” (Brown
and Yule 1983: 202).
is urging financial institutions to tender ‘up to four’ licences to lend

to students in return for loan finance at discounted rates of interest.
The government would also pay the lenders a share of running costs
and a percentage of unrecoverable debts. […] But the banks and
building societies are unimpressed. None has said it would take up
the government’s offer to lend to students. (CB)
Examples (22) and (23) illustrate a distinction between two kinds of text
reference in relation to possessive NPs. In one type, illustrated by (23), the
possessive NP contains a metatextual noun, that is, the possessive NP
summarizes and categorizes something which has been described in the
preceding discourse as a semiotic phenomenon, i.e. a symbolically proc-
essed phenomenon. The second type, illustrated by (22), summarizes and
categorizes something which has been described in the preceding discourse
as a non-semiotic phenomenon, i.e. as a phenomenon (event, state, activity,
etc.) in reality. Following Takahashi (1997: 63), we will call this the sum-
mative type. Table 4 shows how many of the possessive NPs in our data-
base have text reference, either metatextual or summative.
Table 4. Results for the category text reference
data set # of tokens Percentage

metatextual summative metatextual summative
T.S. T.S. T.S. T.S.
Genitive [CB] 10 / 185 12 / 185 5.41 % 6.49 %
Total Genitive 10 / 185 12 / 185 5.41 % 6.49 %
its + N [LOB] 0 / 48 0 / 48 0% 0%
their + N [LOB] 1 / 46 3 / 46 2.17 % 6.52 %
my + N [CB] 1 / 46 0 / 46 2.17 % 0%
her + N [CB] 4 / 46 1 / 46 8.70 % 2.17 %
Total Poss. Det. 6 / 186 4 / 186 3.22 % 2.10 %
Grand total 16 / 371 16 / 371 4.31 % 4.31 %
While the number of tokens for this category is relatively small in our cor-
pus, the cases which occur shed an interesting light on possessive NPs, as
they reveal their potential to categorize complex referents built up over
longer stretches of text. Again, this demonstrates the rhetorical versatility
of possessive NPs in interaction with the surrounding discourse.
Possessive NPs containing a metatextual type specification typically con-

tain ‘semiotic nouns’, i.e. nouns which designate linguistically processed
phenomena, such as decision, claim, remark, etc. Some examples are the
following:
(24) The Government has called on housing associations to raise more of

their funds through the private sector and announced that housing
association grants will be cut next year. John Battle, Labour’s
Shadow Housing Minister, warned that Black pensioners will be
badly hit by increasing housing association rents if Government
grant cuts go ahead. […] Mr Battle, who addressed the National
Conference on Housing Black and Minority Ethnic Elders last week,
said added rents will soar, hitting pensioners already facing huge
hikes in their fuel bills as a result of the VAT increases already an-
nounced in the Budget. A National Federation of Black Housing Or-
ganisations spokeswoman echoed Mr Battle’s sentiments. (CB)
(25) Thanks to the federal prosecutor in Munich, Compuserve subscribers

no longer have access to 200 dubious and distinctly sad Internet
newsgroups catering for people who think sex is an activity that can
be pursued through a mouse and a modem. The Germans, in a fit of
prudishness, told Compuserve it would be prosecuted if it did not
stop allowing their citizens to leapfrog through the commercial serv-
ice into the internet which is quite beyond the company’s control and
read the poison on their pcs. […] The prosecutor’s decision was
plain stupid. (CB)
In (23), offer in the possessive NP the government’s offer refers back to the
special advantages and conditions offered by the British government to
banks and building societies willing to grant loans to students, which have
been described in the preceding discourse. In (24), sentiments in the posses-
sive NP Mr Battle’s sentiments refers back to the opinions and ideas of
John Battle which have been represented explicitly through indirect speech
in the preceding discourse. Note that the possessive NP not only refers back
to what has been stated in the text, but also categorizes (parts of) the pre-
ceding text as a text referent. Thus, a possessive NP realizing text reference
not only refers back to a preceding stretch of discourse, but also actively
construes a text referent in the sense that it categorizes and ‘labels’ the text
it refers to. This is also the case in (25), where more inferencing is required
from the addressee than in (23). The decision referred to by the possessive
NP the prosecutor’s decision is not directly represented, but rather indi-

rectly rendered through the description thanks to the federal prosecutor in
Munich, Compuserve subscribers no longer have access (…). From this it
can be inferred that what the prosecutor decided was to order Compuserve
to remove offensive newsgroups from their network, and it is this decision
which is referred to by the possessive NP.
Possessive NPs with a summative type specification generally require
some inferencing on the part of the addressee as well. Consider examples
(26), (27) and (22) above:
(26) A little over a month ago the Federal Government announced a $ 17.8
million grant to Indonesia to help combat a HIV/AIDS epidemic
which may infect 2.5 million people by 2000. […] I spent a few
weeks in Thailand in 1989 and am convinced bureaucrats misled the
population into thinking the virus did not infect Asians, largely to
protect their rich international sex-tour industry. If Thailand was
lulling the population and the rather stupid sex tourists into a false
sense of security, why not other Asian countries? So now we have to
cough up $ 17 million to Indonesia alone trying to hold back the
scourge. The irony is that a month after our generous grant was an-
nounced, Indonesian President Suharto’s son, Mr Tommy, bought a
majority stake in Italy’s glamour sports car maker, Lamborghini. […]
What he paid is anyone’s guess, a trifle more than our $ 17 million, I
would think. Sounds mighty like he is in a position to match our gen-
erosity in his country’s fight against AIDS, does it not? (CB)
(27) Among the regions, London heads the list as the centre with the lion’s
share of venture-capital deals; 35 of companies involved in raising
money last year were based in the capital. Scotland showed a high
level of activity with 13 of the deals, closely followed by northwest
England. The northeast, despite its high levels of inward investment,
is not, on these statistics, generating much fresh entrepreneurial ac-
tivity; it accounted for 2 of the deals. The capital’s dominance is
confirmed by the regional breakdown for flotations. (CB)
In (26), fight against AIDS summarizes in a relatively straightforward man-

ner information given in the preceding discourse, indicated in bold type in
the example. In (27), more inferencing is needed: the preceding text con-
tains the statement that London heads the list [among the regions] and ex-
plicitly elaborates on the number of companies involved in venture-capital
deals based in London in comparison with such companies based elsewhere

in Britain. This information is summarized by dominance, the interpretation
of which at the same time requires the addressee to infer that the information
given earlier implies that London is dominant in this particular area. In (22),
finally, the information that Fokker needs a vital cash injection in order not
to collapse, is categorized as a crisis, which requires the addressee to make
the inference that for a business, such a situation qualifies as a crisis.
As has become clear from the examples we have discussed, text refer-
ence involves a fair amount of inferencing on the part of the addressee. It is
therefore a separate category in the classification of discourse statuses, dis-
tinct from coreferential possessives in which the PM refers back to a pre-
ceding nominal realization of a referent. On the continuum of discourse
statuses text reference is right next to coreferential possessives, however,
since it involves a PM referent corresponding directly to a stretch of text in
the preceding discourse.
4.3. The PM referent is INFERABLE from the preceding discourse
Table 5 lists the number of cases in which the PM referent is inferable from
a referent or another element in the preceding discourse.
Table 5. Results for the category inferable

Genitive [CB] 37 /185 20.00 %
Total Genitive 37 /185 20.00 %
its + N [LOB] 23 / 48 47.92 %
their + N [LOB] 12 / 46 26.09 %
my + N [CB] 12 / 46 26.09 %
her + N [CB] 16 / 46 34.78 %
Total Poss. Det. 63 /186 33.87 %
Grand total 100 /371 26.95 %
The conceptual relations causing inferability of a PM referent overlap to a

considerable degree with those enabling bridging or associative anaphora,
coded by NPs with definite articles (see section 3). The latter involve defi-
nite reference to a referent (signalling its recoverability) based on an indi-
rect anaphoric relationship, i.e. a relation with another element (and not a
previous mention of the same referent) in the preceding discourse. This

preceding element is thus an ‘indirect antecedent’, three subtypes of which
were discussed in section 3. Firstly, the antecedent can be another entity
referent (i.e. a ‘thing’), which allows the inference of another entity on the
basis of a part-whole relation (e.g. a car the steering wheel) or a more
general association (e.g. a house the owner). The antecedent may also be
an event or activity, from which an entity typically involved in it can be
inferred (e.g. he has been murdered the police haven’t found the killer
yet). Finally, the ‘antecedent’ may be a scenario or frame, evoking entities
fulfilling specific roles in it (e.g. [eating at a restaurant] the waiter for-
got to bring us the menu).
In the case of bridging, the definite article signals that the referent of the
NP is presumed to be recoverable by the addressee, and for the addressee to
retrieve it, the link with the indirect antecedent needs to be grasped. The same
conceptual relations of inferability between referents and discourse contexts
can also motivate the use of a possessive NP, which, however, makes ex-
plicit that there is a link between two entities (the PR and the PM referent),
and is thus in a sense an alternative for the construal with a NP with defi-
nite determiner.8 Thus, while in (28a), the relationship between ‘book’ and
‘pages’ is made explicit through the use of the possessive determiner its
(referring back to the book), in (28b) the connection has to be inferred by
the addressee in order to resolve the reference of the definite NP the pages.
(28) a. As she sprang to her feet and ran to Alistair, the book fell to the
floor, face downwards, its pages doubling up in disorder. (CB)
b. […] the book fell to the floor, face downwards, the pages doubling
up in disorder.
If we want to recognize different degrees of discourse-newness of posses-
sive NPs, PM referents inferable from other entity referents, activities or
larger scenarios evoked by the context, can be viewed as constituting the
first, lowest degree of discourse newness. The possibility of designating
such inferable referents with NPs containing a definite article is evidence of
this relatively low degree of newness, and it can also be used as a formal
indication of which referents to include in this category. This can be illus-
trated with the following example:
8
We are not claiming that there is free variation between these two possible reali-
zations of the inferable referent. Different semantic and pragmatic restrictions
apply to NPs with prenominal possessives and definite determiners, and many
instances of bridging have no possessive alternative and vice versa.
(29) The Leviathan of Parsonstown, said to be the world’s largest tele-

scope, was built in 1845 by the 3rd Earl of Rosse at Birr Castle, Co
Offaly. It will form the centrepiece of Ireland’s new Historic Science
Centre which will be based at the castle. […] The rest of the scien-
tific centre, which will include an exhibition of the astronomical
work of the 3rd and 4th Earls and scientific galleries in the castle’s
moat, will be completed by the end of the century. (CB)
The PM referent of the possessive NP, ‘moat’, even though strictly speaking
new, has a relation of strong association with the referent ‘castle’ which is
mentioned in the preceding context (a castle is very often surrounded by a
moat). On the basis of this conceptual relation the PM referent ‘moat’ is
immediately recoverable in the sense that it can be inferred from the referent
‘castle’. In fact, in this example, the ‘moat’ could also have been referred to
with a definite NP:
(29’) […] an exhibition of the astronomical work of the 3rd and 4th Earls
and scientific galleries in the moat.
In cases like these, the possibility of replacing the prenominal possessive

by a definite article can be used as formal test to establish the inferability of
the PM referent.
A question which immediately springs to mind when one is confronted
with such examples is why the speaker chooses to use a possessive NP,
rather than an NP with a definite article, to refer to a referent which is infer-
able from the discourse context. There are a number of possible pragmatic
factors which motivate the use of a possessive in such contexts. In general,
these have to do with the speaker aiming at more clarity: the inferential re-
lation or ‘bridge’ that is there in the discourse is made explicit by the pos-
sessive NP. Speakers thus seem to choose the possessive in order to avoid
potential confusion about the antecedent, as illustrated by (29), in which
there are four clauses between the nearest mention of the referent ‘castle’
and the introduction of the referent ‘moat’, so that the link between the two
might not be sufficiently salient at the moment when ‘moat’ is mentioned.
Moreover, the referent ‘castle’ is not topical at the moment when ‘moat’ is
introduced, since the four intervening clauses deal with the Victorian tele-
scope located at the castle. Non-topicality of the indirect antecedent is in-
deed a second factor which influences the choice of a possessive rather than
a definite NP in cases where bridging is possible. This also seems to be the
main reason for the use of a possessive in (30):
(30) Josephina Zacaroli-Walker is a businesswoman who became the star

of a comic strip in a Japanese magazine. The strip’s originator sim-
ply fell for the blonde and blue-eyed charms of the woman who has
been the public relations face of Land Rover abroad, and who is now
promoting Rover’s saloons and hatchbacks as well as the new MGF
around the world. (CB)
Distance between antecedent and associated entity is in this case clearly not
the problem, as the referent ‘(comic) strip’ is mentioned right before the
associated referent ‘originator’. However, the referent ‘comic strip’ is syn-
tactically embedded in an of-phrase in the preceding clause, which reduces
its accessibility. Consequently, although an NP with definite article is pos-
sible (The originator simply fell for the blonde and blue-eyed charms…),
the speaker opts for a possessive realization of the referent.
However, the data show that a clear motivation for the use of a posses-
sive NP to refer to a referent which could have been realized by a bridging
NP is certainly not always present. Consider, for instance, the following
example:
(31) But even this hardly prepared one for the spectacle that the house
itself presented on closer view. It stood, as it were, knee-deep in
weeds – like some forlorn prehistoric creature in an edible pasture.
Its grey surfaces were flaked and cracked; its woodwork was de-
nuded of paint; many of the lower windows showed tattered curtains
pulled awry, and some of the upper ones lacked entire panes of glass.
[…] If challenged to date it, Appleby would have said 1718; if chal-
lenged to the name of the builder, he would have said James Gibbs.
But now it spoke either of madness – which, indeed, was what was
attributed to its owner – or of penury. (LOB)
The inanimate referent ‘house’ is clearly the topic of this passage. Moreover,
all of the underlined NPs can be analyzed as realizing bridging reference,
since they name parts (surfaces, woodwork, windows) or associated human
referents (builder, owner). However, two of them are anaphorically bridged,
whereas the others are realized by means of a possessive NP. The antecedent
is clear for all of these referents, and no obvious other differences between
e.g. lower windows (referred to by means of bridging) and woodwork (re-
ferred to with a possessive NP) or between builder (bridging) and owner
(possessive) can account for their different realization. It seems, then, to be
a matter of personal choice on the part of the speaker for one construal or
the other in many cases.
The possibility of replacing the prenominal possessive by a definite arti-

cle is an important indication of the inferability of the referent but, as a
formal test, it has to be handled with some caution. It seems to apply pretty
systematically when the indirect antecedents are inanimate ‘wholes’. For
instance, its (example 31) and their referring to inanimate PRs (the books
their covers) can generally be replaced by the. However, the picture is
somewhat more complicated with possessive NPs referring to body parts of
human PRs. In English, body parts of humans cannot generally be intro-
duced with a definite article-NP based on the preceding mention of the
‘possessor’ (cf. Du Bois 1980: 241); realization with a possessive NP is of-
ten obligatory. This is illustrated by the following examples:
(32) a. She is on sick leave because she has broken her leg .
b. *She is on sick leave because she has broken the leg .
(33) a. During cross-examination the woman said her eyes were not open
all the time, but she was sure she did not fall asleep. (CB)
b. *During cross-examination the woman said the eyes were not open
all the time, but she was sure she did not fall asleep.
However, certain constructions – namely external possession constructions –

do allow for a body part to be introduced with a definite NP:
(34) The burglar hit John on the head with a baseball bat.
(35) She took me by the hand and led me to the study.
In these examples, reference to the body parts by means of a definite NP is

constructionally determined and a considerable degree of idiomaticity is
involved, especially in (35). One is reminded here of Ariel’s (1990: 184)
point that the coding of inferable referents is often governed by “language-
specific conventions”. Still, such examples indicate that body part-relations
are formally treated as inferable (allowing bridging reference) in English in
a number of cases. This is not surprising, since body parts stand in a concep-
tually close, ‘inalienable’ relation to their PRs (see Chapell and McGregor
1996). Another indication of the inferable nature of body parts is that when
the ‘possessor’ is non-human, and particularly when the ‘possessor’ is a
dead animal, bridging to a body part is possible, as illustrated by the fol-
lowing examples:
(36) When hounds had run a stag to a standstill, the stag would turn to
defend itself. And few dogs fancied first toss on the antlers, so would
stand baying for the huntsmen to come in for the kill. (CB)
(37) Gut the mackerel, slit down the belly and open them out kipper-style,
or skewer fillets, season and grill. (CB)
Because of the conceptually inalienable nature of the relation between

body-parts and their PRs and the fact that they can be referred to by NPs
with the in a number of cases, we have classified them with the inferable
PM referents.
What is the import of the analysis proposed here to the overall analysis
of the discourse status of PM referents? We argue that when the PM referent
is inferable from the surrounding discourse, it cannot be analyzed as entirely
new to the discourse. For instance, although strictly speaking, the referent
‘moat’ in (29), for instance, is mentioned for the first time, it is inferable on
the basis of its conceptual (part-whole) relation with ‘castle’. However,
since it has no direct antecedent but only an indirect one, it is not strictly
discourse-given (i.e. coreferential) either. Moreover, in the hierarchy of dis-
course statuses it is below text reference (i.e., closer to ‘newness’ than text
reference). The latter type of reference involves a referent which directly
corresponds to and categorizes a preceding stretch of text. Inferable refer-
ents, by contrast, involve a relation that holds between two different refer-
ents, between an event or activity and a referent which it is involved in it or
between a scenario or frame and a referent which fulfils a role in it.
4.4. The PM referent is ‘ANCHORED’ to an element in the preceding

discourse
This fourth category is transitional between the previous category of infer-

able PM referents and the next one of new PM referents. It contains cases
of first mention PM referents that have a conceptual relation to other refer-
ents in the discourse but one that is not strong enough to make them pre-
cisely inferable. Looking at these conceptual relations prospectively, one
could say that the element to which the PM referents are related does not
predict them as a unique instance or inclusive set. As a consequence, PM
referents of this category do not systematically alternate with definite NPs
with possessive postmodifiers, but can also alternate with indefinite NPs +
of + PR. In other words, in this category we find functionally indefinite
possessive NPs, as illustrated by examples (4), (6) and (15) above. We will
refer to this class as ‘anchored’ PM referents: they are conceptually an-
chored to other elements in the discourse – to which, indeed, possessive
NPs ‘anchor’ them via their internal reference point mechanism – and, as a
consequence, cannot be classified as entirely new. At the same time, they
differ from the inferable PM referents. As we saw in section 4.3, these have
a stronger conceptual relation to their antecedents in the discourse, which
inherently guides the person processing the text to unique instances or to
inclusive sets of instances, and excludes functionally indefinite NPs. Spe-
cifically, two main kinds of anchoring relations can be distinguished: on the
one hand, kinship and other interhuman relations and, on the other hand,
alienable possessions. Table 6 gives an overview of the number of cases in
which the PM maintains a relation of ‘anchoring’ with the discourse context.
Table 6. Results for the category anchoring relation
‘interhuman’ ‘alienable possession’ ‘interhuman’ ‘alienable possession’

11 / 185 26 / 185 5.94 % 14.05 %
11 / 185 26 / 185 5.94 % 14.05 %
0 / 48 14 / 48 0.00 % 29.17 %
2 / 46 9 / 46 4.35 % 19.57 %
13 / 46 5 / 46 28.26 % 10.87 %
8 / 46 3 / 46 17.39 % 6.52 %
23 / 186 31 / 186 12.37 % 16.67 %
34 / 371 57 / 371 9.16 % 15.36 %
The first main type of anchoring relation is formed by kinship and similar
interhuman relations. They are evoked by certain nouns with an inherently
relational meaning (see Barker 1991 [1995], 2000; Fraurud 1996). Taylor
(1989: 675) gives a list of nouns which in addition to kinship nouns “in-
voke, in their semantic structure, various other kinds of interpersonal rela-
tionships”, including friend, fiancée, colleague, guest, fellow student, com-
petitor, confidant, etc. Taylor (1989) discusses such relational nouns in
terms of how they ‘steer’ the interpretation of the relationship between PR
and PM by evoking an unprofiled relationship in their semantic structure,
an element of which is elaborated by the PR nominal.
Possessive NPs whose PM is a relative, friend or associate can be func-
tionally indefinite, as illustrated by (4), a friend’s friend, described by let-
ter, should turn out an unrecognizable stranger, above, (38) and (39).
(38) Created in 1968 and based on a true story, Ashton depicts Elgar’s
friends visiting him after he had sent his score of his Variations to the
famous Viennese composer Ritcher, to try and interest him in the
work. (CB)
(39) Dwight wrote Victor Serge that subscriptions were pouring in and
“all kinds of people were being stimulated to write for it […]”. […]
Dwight’s associates and contributors wrote long and constructive
criticisms. (CB)
The covert indefiniteness of the possessive NPs in these examples is made

overt if we recode them in their periphrastic form with of: a friend of a
friend, friends of Elgar, associates and contributors of Dwight (in these last
two examples, reference does not appear to be made to all of Elgar’s
friends or all of Dwight’s associates and contributors).
However, such NPs are functionally definite if there is only one instance
(or only one in the given context) corresponding to the relation, e.g. (40), or
if all contextually available instances are referred to, as in (41).
(40) In the months leading up to the killing, Avent’s wife became pregnant
and he lost his job as a supermarket shelf stacker because he stole
$300. (CB)
(41) Mack’s friends were all sixteen or seventeen. (CB)
In informal spoken English, unique or inclusive reference to close of kin

can even be realized by bridging NPs, as illustrated by (42) and (43). This
suggests that the distinction inferable – anchored referent being proposed
here is of a continuum-like nature.
(42) <ZF1> H h <ZF0> how do you get to the Fab Club? <M01> Well
FX c erm <ZF1> the d <ZF0> the daughter comes and picks us up
or we get taxis.9 (CB)
9
The codes in this example are part of the transcription of spoken language used
in the COBUILD corpus. <F01> and <M01> indicate the two participants in the
dialogue; <ZF1> and <ZF0> indicate a repetition of (parts of) words, and ‘FX’
and ‘MX’ replace, for privacy reasons, female and male proper names mentioned
in the dialogue.
(43) When a man loses his job, it deals a sickening blow to his ego and
sense of identity. Feelings of failure and rejection strain family rela-
tionships to breaking point and it’s often the wife who’s left to pick up
the pieces. (CB)
Alienable possessions form the second main type of anchored PM referents.

Possessive NPs designating such referents can also be either functionally
indefinite or definite. Thus, (44) and (45) are functionally indefinite: they
do not refer to a uniquely identified possession or to an inclusive set of be-
longings.
(44) No one checked up on us. At 14, I was cutting school in the lunch
hour and getting drunk in the afternoons usually at a friend’s house.
(CB)
(45) She smiles with some regret when the fans rip Brett’s shirts to shreds
during The Drowners. He’s lost so many beautiful clothes; […] (CB)
By contrast, (46) and (47) are functionally definite, as they refer to the one
car driven by Rachelle in (46), and to all contextually given instances of
garments of Christ in (47).
(46) […] “I saw Rachelle’s car slow in the final and figured she must have
struck problems,” Cowin said. (CB)
(47) There is a long-established anecdote concerning the two angels who

are holding Christ’s garments in Andrea del Verrochio’s painting,
The Baptism of Christ. (CB)
In sum, possessive NPs referring to kinship and interpersonal relations, and

to alienable possessions mark the point on the continuum at which the PM
referent is no longer intrinsically inferable from the preceding discourse.
Even though this category is not very frequently attested by possessive
NPs, they still should not be brushed under the carpet. The occurrence of
functionally indefinite possessive NPs in this category is one of the main
arguments against the analysis of possessive NPs as definite NPs.
4.5. The PM referent is a NEW referent
The quantitative results for the category of new PM referents in our data are
given in Table 7.
Table 7. Results for the category NEW

Genitive [CB] 61 / 185 32.97 %
Total Genitive 61 / 185 32.97 %
its + N [LOB] 10 / 48 20.83 %
their + N [LOB] 15 / 46 32.61 %
my + N [CB] 7 / 46 15.22 %
her + N [CB] 11 / 46 23.92 %
Total Poss. Det. 43 / 186 23.12 %
Grand total 104 / 371 28.03 %
These results show that although it is true that possessive NPs newly intro-
duce referents into the discourse in a number of cases (28% of our corpus),
new referents certainly do not represent the majority of PM referents.
This last category forms the ‘new’ end of the continuum of discourse
statuses that we propose. It contains cases in which the PM is a new refer-
ent in the discourse, not only in the sense of not having been mentioned as
such in the preceding discourse, but also in the sense of not being inferable
or predictable from elements in the preceding discourse. In many of these
cases, the PM is realized by a nominalization or deverbal nouns, as in (48)
or (49) below.
(48) [about the Turnable Emergency Non-capsizable Triangular System] It

can survive punctures in two of its surfaces and still remain afloat.
Hunter has produced two prototypes and is in talks with a lifeboat
manufacturer that could lead to the system’s launch in the spring of
next year. (CB)
In these examples, the PR nominal designates the subject of the action de-
scribed by the nominalization. While the PR referent typically maintains an
anaphoric relation with the preceding discourse, the action designated by
the ‘PM noun’ is not inferable from the preceding text. In some of these
examples, there is some sort of general semantic relation between the PM
referent and the preceding discourse context, but this relation is not strong
enough to confer (even partial) givenness on the referent. This is for in-
stance the case in example (49), in which the PM referent fits in with the
scenario that is being evoked in the context, without being a predictable
part of this scenario. That no ‘givenness’ of the PM referent is present in
such cases is confirmed by the fact that an introduction of the referent by

means of a definite article-NP is excluded.
(49) Queensland cricket captain Stuart Law yesterday warmed up for this
week’s Sheffield Shield battle against New South Wales with a mighty
double century for Valley in a XXXX Brisbane club cricket game yes-
terday. […] In a punishing display, Law smacked 32 boundaries in-
cluding two sixes, and faced 266 balls. He came to the wicket when
the score was eight after opener Michael Ephraims departed, and im-
mediately went on the attack. […] There was little respite for Univer-
sity’s bowling attack which included Sheffield Shield bowlers Michael
Kasprowicz and Peter Jackson. (CB)
In (49), the context evokes the scenario of a cricket game, in which a

‘bowling attack’ is possible, though by no means necessary; consequently,
the scenario does not lead to identifiability of the referent ‘bowling attack’,
which is being introduced (for the sake of clarity, the bowling attack re-
ferred to is not the attack implied earlier (Law… immediately went on the
attack), since it is being executed by the team playing against Law’s team).
With regard to this category of the ‘newest’ PM referents, the reflection
imposes itself that such nominalizations, despite being commonly used
with specific textual goals in certain registers such as newspaper and scien-
tific language, are not usually considered very typical possessive NPs.
5. Conclusion
In this paper we set out to investigate the underresearched issue of the dis-
course status of the PM referent of possessive NPs in English. We investi-
gated the ways in which this referent can be an initial or non-initial men-
tion, paying attention to how it interacts with previously given information
in extensive discourse contexts. We demonstrated that a binary distinction
between given and new does not suffice to capture its discourse status, and
proposed instead a continuum of discourse statuses, ranging from fully dis-
course-given to fully discourse-new over a number of in-between statuses
in which the PM is only partially given. Our analysis has the following ma-
jor implications.
Firstly, our analysis reviews and refines the description of English pos-
sessive NPs in a number of ways. We have shown that possessive NPs can-
not be analyzed as mere definite NPs which presuppose the identifiability
of their referent. A number of possessive NPs are functionally indefinite,

which argues for a covert definite-indefinite contrast in English. Building on
earlier analyses of possessive NPs as complex NPs with two distinct refer-
ents among which an NP-internal relation of identification exists (cf. Lang-
acker 1993, 1995; Taylor 1996; Martin 1992), we have shown that this NP-
internal identification mechanism is distinct from the external discourse
functioning of each of the two referents of the possessive NP.
Whereas the discourse status of the PR referent is straightforward (it is
given with possessive determiners and with definite genitives, and new with
indefinite genitives), the discourse status of the PM referent is controversial
and has remained underresearched empirically. Against the view of Taylor
(1996) (who, projecting the asymmetric reference-point relation internal to
NPs onto their discourse behaviour, predicts that PMs are overwhelmingly
discourse-new and anchored to a typically given PR), we have shown that
fully discourse-new PM referents represent only a relatively small portion
(28%) of the data. PM referents may also be ‘anchored’ to elements in the
preceding text, which reduce its newness but do not make it recoverable in
a strict sense. In the majority of the cases, however, the PM is given at least
to some extent. It may be retrievable from extended descriptions, quotes or
reports in the surrounding discourse, and it may be inferable from a given
referent or other element in the preceding discourse. In a small but signifi-
cant number of cases (11% of our data), finally, the PM turned out to be
fully given, in the sense of being coreferential with a given discourse refer-
ent.
The analysis of possessive NPs proposed in this paper has theoretical
implications as well. In particular, it adds a systematic discourse perspec-
tive to the theory of reference-point constructions (Langacker 1993, 1995;
Taylor 1996), which has always focused on the NP-internal reference-point
relation. We have shown that possessive NPs cannot be studied in isolation,
without taking account of the discourse context in which they occur. By
studying possessives in extensive discourse contexts, we have revealed that
the reference-point mechanism may be employed in various ways, adapted
to specific discourse purposes. For instance, possessive NPs may be used to
guarantee referential clarity by their explicit link to the PR or to reactivate
the PR as topic, and they may be employed to recategorize existing refer-
ents, or to label text referents. In this way, our analysis has shown that the
possessive is a versatile and complex NP type, whose identificatory poten-
tial as a reference-point construction can only be fully uncovered by taking
its discourse functions into account.
Acknowledgements
We would like to thank the Research Council of the University of Leuven

for the postdoctoral grant PDM/05/43 given to Peter Willemse for the aca-
demic year 2006–2007. We also greatfully acknowledge the support by
grant no. HUM2007-60706/FILO of the Spanish Ministry of Education and
Science and the European Regional Development Fund. This article is based
on a chapter from Peter Willemse’s doctoral dissertation (Willemse 2005:
59–136), which was supervised by Kristin Davidse and Liesbet Heyvaert.
We owe a debt of gratitude to John Taylor, Theo Janssen and Kurt Feyaerts,
who acted as members of the doctoral examination board, for useful com-
ments and suggestions. Our particular thanks go to Bill McGregor for his
very generous and incisive comments which identified some crucial points
to be addressed. Needless to say, we are the only ones responsible for re-
maining errors of thought in the final version.
References
Abbott, Barbara
2004 Definiteness and indefiniteness. In The Handbook of Pragmatics,
Laurence R. Horn and Gregory Ward (eds.), 122–149. Oxford: Black-
well.
Ariel, Mira
1990 Accessing Noun-Phrase Antecedents. London: Routledge.
Barker, Chris
1991 Possessive descriptions. PhD thesis, University of California Santa
Cruz [published 1995, Stanford: CSLI].
Biber, Douglas, Stig Johansson, Geoffrey Leech and Randolph Quirk
1999 Longman Grammar of Spoken and Written English. London: Long-
man.
Blanche-Benveniste, Claire and André Chervel
1966 Recherches sur le syntagme substantif. Cahiers de Lexicologie IX (2):
3–33.
Brown, Gillian and George Yule
1983 Discourse Analysis. Cambridge: Cambridge University Press.
Chafe, Wallace L.
1972 Discourse structure and human knowledge. In Language Compre-
hension and the Acquisition of Knowledge, Roy O. Freedle and John
B. Carroll (eds.), 41– 69. Washington: Winston.
Chafe, Wallace L.
1976 Givenness, contrastiveness, definiteness, subjects and topics. In Sub-
ject and Topic, Charles N. Li (ed.), 25–55. New York: Academic Press.
Chafe, Wallace L.
1994 Discourse, Consciousness, and Time. Chicago: The University of
Chicago Press.
Chafe, Wallace L.
1996 Inferring identifiability and accessibility. In Reference and Referent
Accessibility, Thorstein Fretheim and Jeannette K. Gundel (eds.),
37–46. Amsterdam: John Benjamins.
Chappell, Hillary and William B. McGregor
1996 The Grammar of Inalienability: A Typological Perspective on Body
Part Terms and the Part-Whole Relation. Berlin /New York: Mouton
de Gruyter.
Declerck, Renaat
1986 Two notes on the theory of definiteness. Journal of Linguistics 22:
25–39.
Du Bois, John W.
1980 Beyond definiteness: the trace of identity in discourse. In The Pear
Stories: Cognitive, Cultural and Linguistic Aspects of Narrative
Production, Wallace L. Chafe (ed.), 203–274. Norwood: Ablex.
Emmott, Catherine
1997 Narrative Comprehension: A Discourse Perspective. Oxford: Oxford
University Press.
Fraurud, Kari
1990 Definiteness and the processing of noun phrases in natural discourse.
Journal of Semantics 7: 395–433.
Fraurud, Kari
1996 Cognitive ontology and NP form. In Reference and Referent Acces-
sibility, Thorstein Fretheim and Jeannette K. Gundel (eds.), 65–87.
Amsterdam: John Benjamins.
Gundel, Jeannette K., Nancy Hedberg and Ron Zacharski
1993 Cognitive status and the form of referring expressions in discourse.
Language 69: 274–307.
Gundel, Jeannette K., Nancy Hedberg and Ron Zacharski
2001 Cognitive status and definite descriptions in English: why accommo-
dation is unnecessary. Journal of English Language and Linguistics
5: 273–295.
Halliday, Michael A.K. and Ruqaiya Hasan
1976 Cohesion in English. London: Longman.
Hawkins, John A.
1978 Definiteness and Indefiniteness: A Study in Reference and Gram-
maticality Prediction. London: Croom Helm.
Huddleston, Rodney
1988 English Grammar: An Outline. Cambridge: Cambridge University
Press.
1991 Foundations of Cognitive Grammar, Volume 2: Descriptive Applica-
tion. Stanford: Stanford University Press.
1993 Reference-point constructions. Cognitive Linguistics 4: 1–38.
nitive Construal of the World, John R. Taylor and R. E. MacLaury
(eds.), 51–79. Berlin /New York: Mouton de Gruyter.
2002 Remarks on the English grounding systems. In Grounding: The Epis-
temic Footing of Deixis and Reference, Frank Brisard (ed.), 29–38.
Berlin /New York: Mouton de Gruyter.
Lyons, Christopher
1999 Definiteness. Cambridge: Cambridge University Press.
Martin, James R.
1992 English Text: System and Structure. Amsterdam: John Benjamins.
1985 A Comprehensive Grammar of the English Language. London: Long-
man.
Rosenbach, Annette
2002 Genitive Variation in English. Berlin /New York: Mouton de Gruyter.
Sanford, Anthony J. and Simon C. Garrod
1981 Understanding Written Language: Explorations of Comprehension
beyond the Sentence. Chichester: John Wiley and Sons.
Sanford, Anthony J. and Simon C. Garrod
1998 The role of scenario mapping in text comprehension. Discourse
Processes 26 (2 /3): 159–190.
Takahashi, Hidemitsu
1997 Indirect anaphors: definiteness and inference. Leuvense Bijdragen-
Leuven Contributions in Linguistics and Philology 86: 53–80.
Taylor, John R.
1989 Possessive genitives in English. Linguistics 27: 663–686.
Taylor, John R.
1991 Possessive genitives in English: a discourse perspective. South Afri-
can Journal of Linguistics 9 (3): 59–63.
Taylor, John R.
Willemse, Peter
2005 Nominal reference-point constructions: possessive and esphoric NPs
in English. PhD thesis, University of Leuven.
On the co-variation between form and function
of adnominal possessive modifiers in Dutch and
English
Jan Rijkhoff
1. Introduction
This contribution is concerned with Dutch and (to lesser extent) English
possessive modifiers introduced by the preposition of (Dutch van), as in a
woman OF INFLUENCE or (Dutch) de auto VAN MIJN BROER (the car OF MY
BROTHER) ‘my brother’s car’. The main goal of this paper is to demonstrate
that the remarkable variation in the grammatical properties of this posses-
sive construction directly correlates with the kind of modifier function it
has in the noun phrase.
It is first shown that lexical possessive modifiers with van /‘of’ (“ad-
nominal possessives” for short) are used to express most of the modifier
functions recognized in a semantic, five-layered model of the noun phrase
(section 2). I will then argue that the values for certain grammatical pa-
rameters (here subsumed under the labels MODIFICATION, PREDICATION,
REFERENCE) are determined by the kind of modifier function the adnominal
possessive has in the noun phrase (section 3 and 4); a tentative explanation
is given in section 5.
The more general point this paper wants to make is that functional modi-
fier categories like CLASSIFYING MODIFIER, QUALIFYING MODIFIER, and
LOCALIZING MODIFIER can be characterized in grammatical terms, which
makes it possible to capture grammatical differences between members of
the same form class (such as prepositional phrases, as shown in Table 3)
and grammatical similarities between members of different form classes
(e.g. adjectives and prepositional phrases, see Tables 1a–b) within and
across languages. This paper is restricted to possessive modifiers of nouns
that denote concrete objects.
52 Jan Rijkhoff
2. Layering and adnominal possessives
2.1. The layered organization of linguistic structures
Layered representations of clausal structures, which are designed to reflect

scopal differences amongst modifiers categories, have been proposed in
various theories of grammar (for a short overview, see Butler 2003, Part 1:
239–249). In Dik’s Functional Grammar (FG), layering was introduced by
Hengeveld (1988, 1989, 1990), whose proposals were immediately incorpo-
rated into the general FG framework (Dik 1989, 1997). Subsequently
Rijkhoff (1988, 1990, 1992, 2002) put forward a layered model of the noun
phrase (NP), arguing that, to some extent, NPs and clauses can be analyzed
in a similar fashion. The most recent version of this layered NP/clause
model is shown in Figure 1.
Grammatical modifiers (e.g. cardinal numeral or number marker/ω2,
demonstrative/ω3, (in)definite article/4) are also called OPERATORS and in
the NP they are symbolized by ω (Representational Level) or (Interper-
sonal Level); clausal operators (such as aspect/ o, tense/ 3, (ir)realis/4 ,
mood/5 markers) are symbolized by or . Lexical or phrasal modifiers,
which are also called SATELLITES, typically involve content words (Verb,
Noun, Adjective, Adverb). NP satellites are represented by the symbols τ
(Representational Level) or (Interpersonal Level), which may stand for
adnominal adjectives (a BIG car), NPs (with or without an adposition or
case marker – the vase ON THE TABLE), or relative clauses (the man WHO
STOLE MY BICYCLE), clausal satellites take the form of various kinds of ad-
verbs and adverbials and are symbolized by σ or .1
Subscripts 0–6 indicate the various layers at which modifiers are speci-
fied. The layers are organized hierarchically, which means that illocution-
ary modifiers 6/6 have the widest scope (the whole clause), whereas the
scope of classifying modifiers (ω0/τ0 in the NP, 0/σ0 in the clause), which
only further specify the kind of thing or event referred to, is restricted to the
head constituent of the NP or clause (typically the head noun and the verb
respectively). Figure 2 shows more clearly that (up to a point, at least) NPs
and clauses can be analyzed along the same lines. For example, adnominal
demonstratives (ω ω 3) specify the location of an entity (thing) in space,
whereas tense markers (3) specify the location of an entity (event) in the
temporal dimension. The various boxes indicate differences in semantic
scope.
1
See Rijkhoff (2002) for details; the most recent accounts are Rijkhoff (2008a, b, d).
On the co-variation between form and function of adnominal possessives 53
INTERPERSONAL LEVEL (‘LANGUAGE AS EXCHANGE’)

At the Interpersonal Level, modifiers are concerned with the Interpersonal
Status of four kinds of entities in the World of Discourse:
[i] clauses (or rather the messages contained in the clauses), [ii] propositions,
[iii] events and [iv] things.
MODIFIERS IN THE NOUN PHRASE MODIFIERS IN THE CLAUSE

6. Illocutionary modifiers (6, 6)
inform Addressee about the illocu-
tionary status of the clause (Decl,
Int, …).
5. Proposition modifiers (5, 5)
inform A about S’s personal as-
sessment of / attitude towards a
proposition Xi as regards the prob-
Scope increase
ability, possibility or desirability of
the actual occurrence of event ei.
4. Discourse-Ref. modifiers (4, 4) 4. Discourse-Ref. modifiers (4, 4)
specify the existential status of thing xi or event ei in the World of Discourse.
REPRESENTATIONAL LEVEL (‘LANGUAGE AS CARRIER OF CONTENT’)

At the Representational Level, modifiers specify properties of spatio-
temporal entities (things, events) in the World of Discourse in terms of the
notions Kind (Class), Quality, Quantity, and Location.
3. Localizing modifiers (ω3, τ3) 3. Localizing modifiers ( 3, σ3)

2. Quantifying modifiers (ω2, τ2) 2. Quantifying modifiers ( 2, σ2)
1. Qualifying modifiers (τ1) 1. Qualifying modifiers (σ1)
0. Classifying modifiers (ω0, τ0) 0. Classifying modifiers ( 0, σ0)
Figure 1. Layers of modification in the noun phrase and in the clause (Discourse-
Ref. = Discourse-Referential; Greek characters symbolize grammatical
[ω/ , /] and lexical [τ /, σ /] modifiers in the NP and the clause).
54 Jan Rijkhoff
Clause operators ( /) Clause satellites ( /)

4. Discourse-Referential
3. Location
2. Quantity
1. Quality
0. Kind
EVENT
clause (head: verb/main predicate
NP (head: noun)
THING
0. Kind
1. Quality
2. Quantity
3. Location
4. Discourse-Referential
NP operators (/ ) NP satellites (/ )
Figure 2. Layers of modification: parallels between the layered organization of NPs

and clauses
As can be seen in Figures 1 and 2, there is a certain asymmetry in the dis-

tribution of operators and satellites in that QUALIFYING OPERATORS (ω1, 1)
are not deemed to exist (Rijkhoff 2008a,d). By definition, members of a
grammatical word class (OPERATORS) constitute a smallish, closed set of
items, capturing only a limited number of crucial, absolute distinctions (Dik
1997 Part 1: 160). For example, if Definiteness (4) is a grammatical cate-
gory in some language, the only two available choices are +Definite or –
Definite (indefinite). Since qualities are gradable properties (e.g. big/bigger/
biggest; fairly/rather/very big), they can only be expressed by satellites
(lexical modifiers): a house can be rather big but not rather SINGULAR.
Even though there can be certain form-function correlations (for in-
stance, if a language has a distinct class of adjectives, they are typically
used as qualifying satellites), there is no strict one-to-one relationship be-
tween the formal category of a noun modifier (e.g. adjective, prepositional
phrase) and its modifier function in the noun phrase (Figure 2). This holds
in particular for satellites (cf. Dik 1997 Part 1: 208). CLASSIFYING SATEL-
LITES in the NP (τ0) further specify what KIND of entity (thing, event) is
being referred to by the speaker (e.g. a corporate lawyer, a man of prayer),
whereas QUALIFYING SATELLITES (τ1) specify more or less inherent proper-
ties such as size (a big house, telescopes of enormous size), value or quality
(a cheap suit, wine of an incredible richness), age (a young child, youths

under age 16), or color (blue curtains, a Jovian moon of incredible redness).
These examples illustrate, once again, that the same function (CLASSIFYING
MODIFIER) can be performed by members of different form classes (here:
adjectives and prepositional phrases or PPs with ‘of’).
QUANTIFYING SATELLITES (τ2) in the NP specify the cardinality of an
entity through phrasal or lexical means (as in example (2) from Samoan)
and LOCALIZING SATELLITES (τ3) are concerned with (broadly speaking)
locative properties of the referent in conversational space, as in the book on
the table, but also the book that you just gave me or my book. A LOCALIZING
SATELLITE basically serves to make the referent of the matrix NP (book)
locatable and consequently identifiable for the addressee. DISCOURSE-
REFERENTIAL SATELLITES (4), finally, are not so much concerned with no-
tions like Class, Quality, Quantity or Location, but rather with the pragmatic
or interpersonal status of the entity, specifying whether or not some entity
already exists in the world of discourse (Definite vs. Indefinite), or whether
the speaker refers to the same or a different token of a certain type (the
same book, another book).2
The fact that members of the same formal modifier category (adjective,
NP, PP, relative clause) can be used in different modifier functions in the
NP is also exemplified below, where a relative clause serves as a qualifying,
a quantifying or a localizing satellite. Localizing satellites typically contain
reference to a familiar entity, which makes it possible to identify the referent
of the matrix NP in the shared world of discourse, as in the following exam-
ple (with the referential anchor underscored; more on this in section 3.4).
(1) Could you pass me the book that is lying next to you?
In Samoan a special kind of relative clause is used to specify cardinality. In

(2) the numeral appears as the head of a relative clause construction that is
introduced by the general tense-aspect-mood marker [GENR] e if the NP
has specific reference.
2
Synchronic and diachronic relationships between modifiers of the clause and the
NP are discussed in Rijkhoff and Seibt (2005) and Rijkhoff (2008a, d).
56 Jan Rijkhoff
Samoan
(2) Sa fau=siae e Tagaloaalagi fale e tolu …
PAST build=ES ERG Tagaloaalagi house GENR three …
‘Tagaloaalagi built three houses …’
(Mosel and Hovdhaugen 1992: 318)
Galela (a Papuan language spoken in Halmahera, Indonesia) uses participial

forms of stative verbs to express quality concepts.
Galela
(3) awi dohu i lalamo
his foot it be.big.PRT
‘his big foot’ (van Baarda 1908: 35)
Possessives, too, can be used in a variety of modifier functions. This is dis-

cussed in the remaining sections.
2.2. Adnominal possessives
Possessive constructions or “genitives” are notorious for the wide range of

meanings they cover (Williams 1981: 89). The paperback edition of Long-
man’s Dictionary of Contemporary English alone lists twenty-five different
senses of the possessive preposition of. The Dutch possessive preposition
van ‘of’ is at least as polysemous as its English counterpart and is currently
even used as a quotative marker (Hengeveld 1994):
(4) Ik dacht van ga hem eens huren.

I thought of go him once rent
‘I thought: let’s just rent it [i.e. a video – JR]’
(cf. Ik dacht: “Ga hem eens huren”)
There is no consensus as to why the possessive construction is so extremely

versatile within and across languages, but it seems this is at least partly due
to the fact that the wide ‘belonging to’ relation covered by possessives
plays a central and fundamental role in human cognition (Seiler 1983, Heine
1997).
So as to restrict to domain of research to a manageable size, I will mostly
be concerned with lexical adnominal possessives in Dutch and English that
are introduced by the preposition van /‘of’ and that modify a common noun
denoting a concrete physical object (such as car, dog or tree). Here are
some examples (all non-asterisked examples were found on the Internet
using Google or Ask):3
CLASS (see section 3.2)

(5) een man van God
a man of God
‘a man of God’ (i.e. a priest, a clergyman)
(6) een man van het toneel
a man of the stage
‘an actor’ (especially a dedicated actor)
(7) een man van de geest (the article is optional; see note 4)
a man of the mind
‘A philosopher’ (or simply an intellectual or religious person)4
(8) een meisje van plezier
a girl of pleasure
‘a girl of pleasure’ (i.e. a prostitute)
QUALITY (see section 3.3)

(9) een huis van steen (material)
a house of stone
‘a stone house’
(10) een kind van twee [jaar oud] (age)
a child of two [year old]
‘a two-year old child’
(11) een biljet van 10 [euro] (value)
a banknote of 10 euro
‘a ten-euro note’
3
Rosenbach (2006) is a recent study of genitives in English.
4
The same text also contains an example without a definite article: Julius Stinde
is niet alleen een man van geest, maar ook een man van het ware midden. Hij
weet maat te houden. ‘Julius Stinde is not just a man of mind/spirit, but also a
man of the true middle. He knows how to contain himself’.
58 Jan Rijkhoff
(12) een man van contradicties (human propensity)

a man of contradictions
‘a man of contradictions’
(13) een meloen van twee kilo [zwaar] (weight)
a melon of two kilo [heavy]
‘a melon that weighs two kilos’
(14) een muur van twee meter hoog (height)
a wall of two meter high
‘a two-meter high wall’
(15) een berg van de eerste categorie (height)
a mountain of the first category
‘a mountain of the first category’
(16) een fraaie gevel van 2.44 meter breed (width)
a nice front of 2.44 meter wide
‘a nice front that is 2.44 meters wide’
LOCATION (see section 3.4 as to why these possessives are regarded as

localizing satellites)
(17) de hoed van opa
the hat of grandfather
‘grandfather’s hat’
(18) de trein van 8 uur
the train of 8 o’clock
‘the 8 o’clock train’
(19) de krant van gisteren
the newspaper of yesterday
‘yesterday’s paper’
(20) de fiets van mijn vader
the bicycle of my father
‘my father’s bicycle’
Examples (5)–(20) illustrate that adnominal lexical possessives can occur

as CLASSIFYING (τ0), QUALIFYING (τ1), OR LOCALIZING MODIFIERS (τ3) in
the Dutch NP (Figure 1). As to QUANTIFYING MODIFIERS (τ2), we do find
NPs such as kinderen van drie ‘children of three’ but here the numeral does
not specify the number of children (a quantity) but their age (i.e. a quality:
‘three year old children’). As a matter of fact, I am not aware of any lan-
guage that uses adnominal possessives to express the cardinality of the ref-
erent of the NP (Rijkhoff 2002: chapter 5).5 What we find instead are in-
stances of so-called dependency reversal (Malchukov 2000): the numeral
word is formally expressed as the head of the construction with the noun
denoting the counted entity displaying features of a possessive dependent.
We will return to this gap in the functions of the adnominal possessive in
section 3.2 below.
Dutch does not use lexical possessives as DISCOURSE-REFERENTIAL
SATELLITES (T4) either, but notice that in some languages a (bound) posses-
sive pronoun has been claimed to mark definiteness. If this is indeed the
case, the (erstwhile) possessive pronoun would serve as a grammatical DIS-
COURSE-REFERENTIAL GRAMMATICAL MODIFIER (cf. Siewierska et al. 1998:
811–812, notes 14 and 33). For example, Comrie (1988: 465), referring to
Tauli (1966: 148), writes that some Uralic languages “use the third person
singular possessive suffix as a general marker of definiteness” (see also
Englebretson 2003 on Indonesian –nya and Guillemin 2007 on Mauritian
Creole so). Similarly, in his monograph on the Turkic languages Menges
(1968: 113) states that “the possessive significance of the suffix of the 3rd
person can completely recede when it defines or determines a noun”, in
which case the suffix conveys “a determinative idea as expressed by the
article in Indo-European or Semitic languages”.
Komi, Southern Permyak dialect

(21) et-piri
see woktis ru. ru-is ig.
once then came fox. fox-POS.3SG hungry
‘Once the fox came that way. The fox was hungry’
(Fraurud 2001: 252; taken from Rédei 1978: 474)
To what extent the possessive affix in these languages has actually gram-
maticalized into a bound marker of definiteness is still a matter of debate,
as it seems that the change in meaning has not (yet) clearly resulted in a
category switch (from suffixed possessive pronoun to suffixed definite arti-
cle; Fraurud 2001). Furthermore, in Turkish the possessive suffix does not
seem to occur on nouns.
5
Cardinality is indicated by an adnominal possessive in e.g. een group van 20
(mensen) ‘a group of 20 (persons)’ but this seems only possible when the phrase
is headed by a collective noun like ‘group’ (i.e. not a noun denoting an individ-
ual).
60 Jan Rijkhoff
Turkish
(22) Bun-u iste-mi-yor-um, baka-sın-ı ver
this-ACC want-NEG-PRES-1S other-POS.3SG-ACC give
‘I don’t want this (one), give me the other one’
(Gerjan van Schaaik, pers. comm.)
As this contribution is mainly concerned with lexical possessives in Dutch,

the issue will not be pursued here.
Before we proceed, it is perhaps worth emphasizing, once again, that
this chapter is restricted to possessive modifiers of common nouns denoting
a concrete physical object (cf. section 1). Hence I will ignore adnominal
possessives headed by (with the head noun underscored):
(a) a proper name (Anna’s Peter, Rembrandt van Rijn)

(b) a mass noun (the emperor’s gold) or a collective noun (the victim’s
family)
(c) a noun denoting a location (the place of destination, the city of Amster-
dam), an event (the celebration of her birthday, the trial of the century)
or some other higher order entity (the fairy tale of the princess and the
pea)
(d) an abstract noun (the theory of Functional Grammar, metaphors of
power)
(e) NPs expressing a part-whole relation or inalienable possession (the top
of the mountain; see also e.g. Willemse 2006).
Furthermore I have only taken into account instances in which the adnomi-
nal possessive modifier is marked by a special form (such as the -s suffix or
the preposition van ‘of’) or is a member of the special set of possessive
pronouns), thus ignoring possessives that are unmarked (as in the Nasioi ex-
ample below) or not formally distinguished from other modifier categories.
Nasioi
(23) Máteasi bauran
Mateasi daughter
‘Mateasi’s daughter’ (Rausch 1912: 120)
The results of this investigation indicate that adnominal possessives can

occur in most of the modifier functions recognized in the layered model of
the NP outlined above (examples (5)–(20)). The functional modifier cate-
gories mentioned in Figures 1 and 2 must, however, be regarded as focal
points on a scale of scope increase (from CLASSIFYING SATELLITE/τ0 to

DISCOURSE-REFERENTIAL SATELLITE /T4) rather than discrete classes of
modifiers with sharp boundaries between them (section 4). Thus Dutch in
Dutch cheese can refer to the country of origin (location), but also to a cer-
tain kind of cheese. Similarly, depending on the context, in the Dutch NP
een huis van steen ‘a house of stone’ the possessive van steen ‘of stone’ can
be interpreted as a classifying or as a qualifying satellite (see also note 8
and section 3.4 on localizing satellites).6
2.3. A note on the grammaticalization of of: scope increase or scope

decrease
According to the LOCALIST HYPOTHESIS, language is built on a spatial

metaphor in that spatial expressions are thought to be more basic than vari-
ous kinds of non-spatial expressions (Lyons 1977: 718; Lakoff and Johnson
1980: 14f., 56f.; but cf. Heine et al. 1991: 114–118). On the assumption that
the locative/possessive meaning precedes all other (non-spatial) meanings
in the diachronic development of a polysemous item like van ‘of’, one
could hypothesize the following scenario (Heine and Kuteva 2007: 280–
283). When the Dutch preposition van ‘of’ became more and more polyse-
mous, acquiring new non-spatial meanings along the way, it also appeared
in qualifying and classifying adnominal modifiers. The scope of qualifying
and classifying satellites is, however, more restricted than the scope of lo-
calizing satellites (Figures 1 and 2).
(24) SEMANTIC INCREASE AND SCOPE DECREASE: THE DIACHRONIC DE-

VELOPMENT OF NON-SPATIAL MEANINGS OF DUTCH VAN ‘OF’ AND THE
SCOPE OF LEXICAL ADNOMINAL POSSESSIVES IN DUTCH (cf. Figures 1
and 2):
narrow scope wide scope
HEAD CLASSIFYING QUALIFYING [Quantifying LOCALIZING
NOUN SATELLITE τ0 SATELLITE τ1 satellite τ2] SATELLITE τ3
historic development
• PREPOSITION van ‘of’ (the part): SEMANTIC INCREASE (polysemy)
• ADNOMINAL MODIFIER introduced by van ‘of’ (the whole): SCOPE DECREASE
6
On gradience, see e.g. Aarts (2004a/b, 2006), Rijkhoff (2008c), Sorace and
Keller (2005).
62 Jan Rijkhoff
Thus, paradoxically, according to this scenario INCREASE in meaning of the

preposition van ‘of’ must have gone hand in hand with DECREASE in se-
mantic scope of the new modifier category it could occur in (from outer to
inner layer or from periphery to center).
It may be interesting to add that grammatical modifiers in the NP (opera-
tors ω/ ) tend to develop in the opposite direction diachronically, i.e. from
inner to outer layer or from center to periphery. For example, a demonstra-
tive (Location) may turn into a definite article (Discourse-Referential), and
the numeral one (Quantity) may become an indefinite article (Discourse-
Referential). Similarly, at the level of the clause a perfective marker (Kind)
may ultimately become a past tense marker (Location), and a future tense
marker (Location) may become a mood marker, which is represented at the
Interpersonal Level (Figure 1). The only counterexamples seem to involve
ambiguous categories such as the PERFECT (between tense and aspect),
which can be source of a past tense marker (Location), as in the Germanic
languages, or a perfective marker (Kind), as in the Romance languages
(Heine and Kuteva 2002: 231–232; Rijkhoff 2008d).7
3. Adnominal possessive modifiers in a layered model of the noun

phrase
This section investigates properties of adnominal possessives with van/‘of’,

showing that each modifier function of this construction discussed in sec-
tion 2 (Classifying, Qualifying, Localizing) correlates with a different set of
values for the parameters MODIFICATION, PREDICATION, and REFERENCE.
Special attention is given to the fact that adnominal possessives do not
seem to occur as quantifying satellites (section 3.3).
3.1. Possessives as classifying satellites
Classifying modifiers specify what KIND of entity is being denoted by the

head noun; some common examples of classifying satellites in English are
the adjectives annual in annual report, presidential in presidential election,
electric in electric train, or social in social security. They differ from quali-
7
See Hengeveld (1989: 142) on the diachronic development of clausal operators
and Song (2005) for a recent discussion of scope increase and scope decrease in
grammaticalization.
fying satellites (section 3.2.) in that they do not specify some more or less
objective property of an entity (round table) or the speaker’s subjective atti-
tude towards the entity (beautiful picture), but rather a particular subclass of
the category denoted by the head noun. The semantic range of classifying
satellites is rather broad and may include such categories as material, pur-
pose and function, status and rank, origin, and mode of operation (Halliday
2004: 320). Essentially they relate to any feature that may serve to classify
entities into a system of smaller sets. In English classifying satellites may
take the form of an adjective (25a), but also of a possessive construction
(25b–c):
(25) a. a corporate lawyer

b. a boy’s shirt 8
c. a house of worship
Classifying satellites seem to be grammatically restricted in various ways.

For example, a classifying adjective does not admit an intensifier (26), com-
parison (27), or predicative position (28):9
(26) a. an electric train

b. *a very electric train [intensifier]10
(27) a. a medical examination
b. *a more medical examination [comparison]
(28) a. the presidential election
b. *the election is presidential [predicative position]
An adnominal possessive that serves as a classifying satellite is also called a

NON-REFERENTIAL or NON-DETERMINER GENITIVE, to distinguish it from its
referential counterpart (see also section 3.2). Thus the classifying possessive
8
Actually a boy’s shirt is potentially ambiguous, meaning either ‘the shirt of an
unidentified boy’ (non-classifying) or ‘a particular kind of shirt’ (classifying), see
on this difference e.g. Taylor (1996: 665) and Willemse (2005); see also section
3.4.
9
Cf. Quirk et al. (1985: 1339); on non-predicative adjectives, see also e.g. Farsi
(1968) and Levi (1973).
10
One does find e.g. quick divorce lawyer, where quick modifies divorce, but in
such cases we seem to be dealing with a fixed expression (for example, divorce
cannot be modified by slow or short).
64 Jan Rijkhoff
woman’s in a woman’s hat does not refer to any particular woman but
merely serves to specify for the hearer that the speaker is talking about the
kind of hat that is worn by women. Furthermore, as in the case of classifying
adjectives, classifying possessives such as woman’s in (29a) and (30a) can-
not be modified (29b) or separated (30b) from the head noun by other modi-
fiers, at least not without changing its function:11
CLASSIFYING POSSESSIVE NON-CLASSIFYING POSSESSIVE

(29) a. the pretty [woman’s hat] b. the [pretty woman’s] hat
CLASSIFYING POSSESSIVE NON-CLASSIFYING POSSESSIVE

(30) a. the blue [woman’s hat] b. the [woman’s] blue hat
Classifying possessives cannot be used as predicates either, compare:
CLASSIFYING POSSESSIVE
(31) a. a woman’s hat b. *that hat is a woman’s12
CLASSIFYING POSSESSIVE
(32) a. a man of prayer b. * this man is of prayer
The same goes for these Dutch examples, compare:
(33) a. De priester is dus een man God-s, een man van gebed, een man
the priest is thus a man God-GEN, a man of prayer a man
van de Kerk, een herder voor de mensen.
of the Church, a shepherd for the people
‘so the priest is a man of God, a man of prayer, a man of the
Church, a shepherd for the people’
b. *deze man is van gebed 13
this man is of prayer
11
Gunkel and Zifonun (2008) is a detailed study of one particular subcategory of
classifying satellites (‘relational adjectives’) in English, German and French.
12
Notice that it is possible to have a possessive predicate in e.g. That hat is Mary’s
(That hat belongs to Mary).
13
The same goes for een man Gods ‘a man of God’ and a man van de Kerk ‘a man
of the Church’.
(34) a. Een mooi citaat is een diamant aan de vinger van een man
a nice quotation is a diamond on the finger of a man
van geest en een kei in de hand van een dwaas
of mind/spirit and a rock in the hand of a fool
‘a nice quotation is a diamond on the finger of a wise man and a
rock in the hand of a fool’
b. *die man is van geest
that man is of mind/spirit
(35) a. Kijk in de agenda van een meisje van plezier

look in the calender of a girl of pleasure
‘Look in the calendar of a girl of pleasure’
b. *dit meisje is van plezier
this girl is of pleasure
Classifying possessives can also take the form of a case marked modifier,
as in these examples from Swedish and Lithuanian:
Swedish
(36) en folk-et-s teater
a:C people-DEF.C-GEN theatre
‘a theatre for the people’ (Koptjevskaja-Tamm 2003a: 539–540)
Lithuanian
(37) duon-os peilis
bread-GEN knife
‘a bread knife’ (Koptjevskaja-Tamm 2002: 155)
The difference between classifying and non-classifying modification may

also manifest itself in the presence or absence of a possessive suffix:
Roviana
(38) a. mamalaengi barikaleqe
voice woman
‘a woman’s voice/a female voice’ (classifying/non-referential)
b. mamalaengi-na [barikaleqe hoi]
voice-3SG.POSS [woman that]
‘that woman’s voice’ (referential/non-classifying)
(Corston-Oliver 2002)
66 Jan Rijkhoff
The English translation (‘bread knife’) of Lithuanian duonos peilis

(bread.GEN knife) shows that it may be difficult to draw the line between a
noun plus classifying satellite and compounds or quasi-compounds, espe-
cially when (as in ‘bread knife’) there is no sign of a dependency relation
between the two elements (on this problem see e.g. Bauer 1998, Giegerich
2005).14
3.1.1. Classifying possessives: modification, predication, reference
It turns out that prepositional phrases with van/‘of’ behave differently with
respect to the following three grammatical tests, depending on the kind of
modifier function they perform in the NP:
(a) whether or not the noun can be modified (internal MODIFICATION);

(b) whether or not the phrase can occur in predicate position (PREDICATION);
(c) whether or not the phrase is referential (REFERENCE).
The data presented above indicate that classifying possessives have negative
scores on all counts:
Table 1a. Classifying adnominal possessives
POSSESSIVES MODIFICATION PREDICATION REFERENCE
CLASSIFYING — — —
Een meisje *een meisje *het meisje is *het meisje
van plezier van veel plezier van plezier van zijn plezier
‘a girl of pleasure’ ‘a girl of much ‘the girl is of ‘the girl of his
pleasure’ pleasure’ pleasure’
As we saw earlier, the same restrictions hold for classifying adjectives:
14
Apart from the close connection between compounding and classification, there
is also a relation between inalienability and classification (as in a mountain top
vs. the top of a mountain; for discussion see e.g. Chappell and McGregor 1989,
1996).
Table 1b. Classifying adjectives
ADJECTIVES MODIFICATION PREDICATION REFERENCE
CLASSIFYING — —
een medisch *een erg medisch *het onderzoek is
onderzoek onderzoek medisch (not applicable)
‘a medical ‘a very medical ‘the investigation
investigation’ investigation’ is medical’
3.2. Possessives as qualifying satellites
Qualifying modifiers relate to qualitative, more or less inherent properties

of the entity designed by the noun. Dixon (1982: 16) recognized seven core
semantic domains that are typically covered by adjectives (if a language has
them): DIMENSION (big, long, wide, thin etc.), PHYSICAL PROPERTY (hard,
heavy, smooth, hot, sweet etc.), COLOR (black, white, green etc.), HUMAN
PROPENSITY (jealous, sad, happy, crazy, rude etc.), AGE (new, young, old
etc.), VALUE (good, bad, excellent, atrocious etc.) and SPEED (slow, fast etc.).
These properties can also be specified by possessives headed by abstract
nouns (denoting shape, size, age, volume, weight etc.):
Hausa
(39) mutum mài alheri / arzaki / hankali
person having kindness / prosperity / intelligence
‘a kind/prosperous/intelligent person’ 15 (Schachter 1985: 15)
Here are some adnominal possessives with van ‘of’ exemplifying some of
the major adjectival categories distinguished above.
DIMENSION
(40) kabeljauwen van enorme grootte
cod:fish(PL) of enormous size
‘cod fish of enormous size’
(41) een oudere man van geringe lengte
an older man of short length
‘an elderly man of short stature’
15
mài (also: màasú) is glossed as ‘owner, possessor of’ in Newman (1987: 721).
68 Jan Rijkhoff
(42) een tank van 50 of 100 liter (inhoud)

a tank of 50 or 100 liter (volume)
‘a tank with a capacity of fifty or a hundred liters’
(43) een kerstkraam van 3 meter lengte en 1 meter diepte
a Christmas:stand of 3 meter length and 1 meter depth
‘a Christmas stand that is 3 meters long and 1 meter deep’
PHYSICAL PROPERTY
(44) Dymfna was … een christelijke moeder van grote schoonheid
Dymfna was … a Christian mother of great beauty
‘Dymfna was … a Christian mother of great beauty’
COLOR16
(45) een Ferrari F40 van rode kleur
a Ferrari F40 of red color
‘a red Ferrari F40’
HUMAN PROPENSITY
(46) Hij was een man van grote charme17
he was a man of great charm
‘He was a very charming man’
AGE
(47) mensen van middelbare leeftijd
people of medium age
‘middle-aged persons’
16
In English the color term can occur by itself, i.e. without the abstract noun kleur
‘color’ (as in a Jovian moon of incredible redness, from the book Cosmos by
Carl Sagan), but in Dutch this seem to be restricted to NPs headed by nouns
denoting non-spatial entities: het effect van rood ‘the effect of red’, het karakter
van rood ‘the character of red’, de richting van de snelheid ‘the direction of the
speed’, de effecten van snelheid ‘the effects of velocity’.
17
There is a wide variety of qualities that can be expressed this way (here all with
the adjective grote ‘great’): een man van grote spontaniteit en een ontwapenende
openhartigheid ‘a man of great spontaneity and disarming sincerity’, een man
van grote gaven ‘a man of great gifts’, een man van grote wijsheid, geduld en
begrip ‘a man of great wisdom, patience and understanding’, een man van
grote kracht ‘a man of great power’, een man van grote, bombastische theo-
rieën ‘a man of big, bombastic theories’.
(48) een man van 40

a man of 40
‘a 40-year old man’
VALUE
(49) een parel van grote waarde
a pearl of great value
‘a very valuable pearl’
(50) een huis van 4,5 ton of meer
a house of 4.5 TON or more (a ‘ton’ = 100.000 euros)
‘a house that is worth 450.000 euros or more’
SPEED
(51) The Boots of Great Speed (title of a story by R. P. Barnett)
Thus it appears that the core adjectival categories are also fairly well covered
by adnominal possessives, but the Internet search with Google and Ask also
seems to point to certain gaps, at least in Dutch and English. For example,
it turned out to be extremely difficult to find examples of adnominal pos-
sessives expressing notions concerning certain physical properties (e.g. of
softness, of smoothness), speed (e.g. of slowness/of slow speed), or taste
(e.g. of sweetness/of sweet taste). Such properties tend rather to be ex-
pressed with the entity denoting noun in the dependent (possessive) con-
struction as in the seductive softness of the Sheridan Ultrasoft towel, the
redness of berries, the slowness of the Concorde or the sweetness of honey
(see also below on Dependency Reversal) or in combination with a noun
denoting a higher order entity (de perceptie van bittere smaak ‘the percep-
tion of bitter taste’, de bron van bittere smaak ‘the source/cause of bitter
taste’ – see also note 16). Although one could construe more or less accept-
able Dutch NPs like een appel van bittere smaak ‘an apple of bitter taste’,
such adnominal possessives do not seem to occur in the actual (written)
language.18 The precise reason for the scarcity of certain adnominal posses-
18
One can say een man van goede smaak ‘a man of good taste’, but here smaak
‘taste’ is used metaphorically, i.e. it does not mean that the man tastes good but
rather that he HAS good taste. The fact that adnominal possessives are not used
to express the notion of speed is probably at least partly due to the fact that
SPEED is not an inherent property but rather a potential property of certain mov-
able objects, like cars and trains.
70 Jan Rijkhoff
sives remains obscure, but there is little doubt that the observed gaps are at
least partly due to the fact that the current study is limited to properties of
concrete objects, since we saw above that one can easily find examples out-
side this restricted ontological domain: tekenen van roodheid ‘signs of red-
ness’, de weg van zachtheid ‘the road of softness’, oorzaken van zuurheid
‘causes of sourness’ (cf. also English the benefits of softness, a touch of
softness, areas of softness, the concept of softness, a feeling of softness; a
touch of green, shades of green; the Well of Sweetness, acts of sweetness,
scoops of sweetness, the perils of sweetness; the taste of sweetness; the dis-
covery of slowness, in praise of slowness, a philosophy of Slowness, etc.).
3.2.1. Qualifying possessives: modification, predication, reference
Qualifying possessives appear to display certain differences in the gram-

matical behaviour with respect to the three parameters mentioned in section
3.1.1: MODIFICATION, PREDICATION, and REFERENCE. We saw in section
3.1 that classifying possessives cannot be used as predicates:
(52) a. een kerel van stavast

a fellow of stand.firm
‘a hefty/plucky fellow’
b. *die kerel is van stavast
that fellow is of stand.firm
(53) a. een man van de wereld

a man of the world
‘a man of the world’
b. *die man is van de wereld 19
that man is of the world
In either case we are dealing with non-referential possessives, whose head

constituent cannot be modified:
(54) *een kerel van enorme stavast

a fellow of enormous stand.firm
19
This sentence is acceptable in Dutch if there is a qualifier such as totaal ‘totally’
or helemaal ‘completely’, but this results in an idiomatic expression with the
meaning ‘to be strange, confused’ or even ‘unconscious’.
Certain qualifying possessives do not appear as predicates either, but they

can be modified (QUALIFYING TYPE A): 20
(55) a. een man van (groot exegetisch) gezag 21

a man of (great exegetical) authority
‘a man of (great exegetical) authority’
b. *die man is van (groot exegetisch) gezag
that man is of (great exegetical) authority
(56) a. een man van (grote) tegenstrijdigheden

a man of (great) contradictions
‘a man of great contradictions’ (see also example (12) above)
b. *die man is van (grote) tegenstrijdigheden
that man is of (great) contradictions
With some qualifying possessives of type A, however, the modifier is com-

pulsory (QUALIFYING A+):
(57) a. Prokofiev: een man van meerdere werelden

Prokofiev: a man of multiple worlds
‘Prokofiev: a man of many worlds’
b. *Prokofiev: een man van werelden
Prokofiev: a man of worlds
c. *die man is van (meerdere) werelden
that man is of (multiple) worlds
We also find examples of qualifying possessives that can be modified and

occur as predicates (QUALIFYING B):
(58) a. een kroon van (zuiver) goud

a crown of (pure) gold
‘a crown of pure gold’
b. het hoofd van het beeld was van (zuiver) goud
the head of the statue was of (pure) gold
‘the head of the statue was made of (pure) gold’
20
In some cases (e.g. with possessives of age or measure, i.e. height, length,
weight) the predicative variant is only possible without the possessive marker
in Dutch: een muur van twee meter hoog (lit. ‘a wall of two meter high’) vs. die
muur is twee meter hoog (lit. ‘that wall is two meter high’); see also note 22.
21
I will say more on the classifying flavor of some of these examples in section 4.1.
72 Jan Rijkhoff
As in the case of type A qualifying possessives, some qualifying possessives

of type B are only acceptable with a modifier, both with the possessive in
adnominal and predicative position (QUALIFYING B+):
(59) a. deze Egyptische beelden (zijn) van grote kwaliteit

these Egyptian statues (are) of great quality
‘these Egyptian statues (are) of high quality’
b. *deze Egyptische beelden van kwaliteit
these Egyptian statues of quality
c. *deze Egyptische beelden zijn van kwaliteit
these Egyptian statues are of quality
Apart from possessives like van grote kwaliteit ‘of high quality’, this group
seems to be confined to possessives specifying some measure (of age, size,
length, width, height, depth, volume, weight etc.).22
(60) a. twee zusjes van geringe lengte

two sisters of short length
‘two sisters of short stature’
b. veel nieuwkomers zijn van geringe lengte en omvang
many newcomers are of short length and size
‘many newcomers are of short stature and size’
Apparently the same holds for English:
(61) a. Sperm whale: a very large toothed whale (Physeter macrocepha-

lus), having a head of enormous size
b. The head of the sperm whale is of (enormous) size [constructed –
JR]
Note, however, that the predicative variant is not possible when the meas-
ure is given a specific value. In such cases, a non-possessive predicate is
used (headed by an adjective rather than a noun), as in (62c).
22
In Dutch, there is also the variant with an adjectival form (instead of an abstract
noun), as in een boot van zeventig meter langA (a boat of seventy meter long) ‘a
boat of seventy meters’; see also example (62c).
(62) a. twee koperdraden van 1 meter lengteN

two copper.wires of 1 meter length
‘two 1-meter copper wires’
b. *twee koperdraden zijn van 1 meter lengteN
two copper.wires are of 1 meter length
c. twee koperdraden zijn 1 meter lang A [constructed – JR]
two copper.wires are 1 meter long
‘two copper wires are 1 meter long’
Qualifying possessives of type B+ can also be used referentially, but notice

that in such cases they do not refer to an entity but rather to a PROPERTY of
an entity, typically specifying the degree to which some property (quality,
size, color, etc.) applies to that entity.
(63) Kinderen van deze leeftijd zijn uitgesproken nieuwsgierig

children of this age are particularly curious
‘Children of this age are particularly curious’
(64) Munten van deze kwaliteit zijn zeer kwetsbaar

coins of this quality are very vulnerable
‘Coins of this quality are very vulnerable’
(65) Vissen van deze grootte hebben nauwelijks vijanden

fish(PL) of this size have hardly enemies
‘Fish of this size hardly have enemies’
(66) Ik vind het dus veel geld voor een lens van zulke kwaliteit
I find it thus much money for a lens of such quality
‘So I think it is a lot of money for a lens of such quality’
These adnominal possessives can also occur with an adjectival modifier or

as a predicate:
(67) een van de laatste Attische grafmonumenten van deze hoge kwaliteit
one of the last Attic burial monuments of this high quality
‘one of the last Attic burial monuments of this high quality’
(68) De meeste stenen … waren van deze kwaliteit.

the most stones … were of this quality
‘Most of the (precious) stones … were of this quality’
74 Jan Rijkhoff
At this point the classification looks as follows:
Table 2. Properties of classifying and qualifying possessives
ADNOMINAL
QUALIFYING B+
(modifier compulsory)
beelden van beelden van de beelden zijn beelden van
grote kwaliteit grote kwaliteit van grote kwaliteit deze kwaliteit
‘statues of great ‘statues of high ‘the statues are of ‘statues of this
quality’ quality’ high quality’ quality’
(REFERENCE TO
A PROPERTY)
QUALIFYING B
een kroon van goud een kroon van de kroon is van —
‘a crown of gold’ (zuiver) goud (zuiver) goud
‘a crown of the crown is of
(pure) gold’ (pure) gold
‘the crown is made
of (pure) gold’
QUALIFYING A+
(modifier een man van — —
compulsory) vele g ezichten
een man van ‘a man of many
vele gezichten faces’
‘a man of many
faces’
QUALIFYING A
een man van gezag een man van — —
‘a man of authority’ (groot) gezag
‘a man of (great)
authority’
CLASSIFYING
een man van de — — —
wereld
Table 2 shows that both classifying and qualifying possessives of type

A/A+ and B are characterized by the fact that they have no referring poten-
tial. Thus one could say that nouns heading these adnominal possessives
are deprived of what may be regarded as the most characteristic feature of
the prototypical noun: its potential to be used in a referring expression (us-
ing the same nouns as in the possessives in Table 2: our world, Hillary’s
authority, John’s face, the king’s gold, Lone’s qualities. The same is true
for these Swedish examples of non-determiner swear genitives, whose
nouns appear to have lost most or all of their nominal properties
(Koptjevskaja-Tamm 2003a: 519).
Swedish
(69) a. en satan-s kärring
a:C Satan-GEN crone
‘a damned crone’
b. ett herra-n-s oväsen
a:N lord-DEF.C-GEN noise
‘a hell of a noise’
In some languages, such as English, we see that the same property (e.g. ‘(to
be) rich, richness’) can be expressed in the form of an adjective or in the
form of a possessive construction (Quirk et al. 1987: 1278, 1286):
(70) ADNOMINAL ADJECTIVE vs. ADNOMINAL POSSESSIVE

a. a rich man a man of riches
b. an influential woman a woman of influence
c. a courageous woman a woman of courage / valor
Now compare: 23
(71) a. A Man of Riches Built His Team From a Trust Fund

b. A Rich Man Built His Team From a Trust Fund (constructed ex-
ample – JR)
There is no consensus on the exact difference between sentences like (71a)

and (71b), but the a-sentence seems to create some kind of distancing ef-
fect, which is iconically reflected in the fact that in a man of riches the
23
“A Man of Riches Built His Team From a Trust Fund” is the title of an article by
Richard Justice about Jack Kent Cooke (The Washington Post, April 7, 1997).
76 Jan Rijkhoff
preposition of appears between the head noun (man) and the abstract noun
(riches). More specifically, whereas rich seems to refer to an actual prop-
erty, of riches seems to specify a property of a man for whom being rich is
a more characteristic or permanent feature, which may come with a set of
other features (the money was acquired in a proper manner, the man has
class, etc.). One could even say that the NP containing the possessive of
riches has a distinct classificatory flavor in that it refers to a certain kind of
man (as opposed to rich in a rich man). This is confirmed by the way NPs
that contain a qualifying adnominal possessive are used in actual speech
(see section 4.1. below).
Both in English and in Dutch we also find instances of adnominal pos-
sessives (sometimes referred to as METAPHORICAL CONSTRUCTIONS in
Dutch grammar) in which the ‘qualifying’ lexeme seems to serve as the
head of the construction:24
Dutch
(72) a. een boom van een vent
a tree of a man
‘a very strong man’
b. een schat van een kind
a treasure of a child
‘a very sweet child’
(73) a. a hell of a guy

b. a whopper of a flash drive
At least in Dutch these are rather idiomatic expressions, with limited possi-
bilities for morphosyntactic variation (*een hoge boom van een aardige vent
lit. ‘a tall tree of a kind man’ or *een grote schat van een Nederlands kind
lit. ‘a big treasure of a Dutch child’ are ungrammatical). Semantically, such
metaphorical constructions typically have a rather strong emotional value
and indicate a property in excess.
A similar construction is attested in a number of Austronesian languages
of the Western Oceanic group. In these languages an NP such as the big
house is ostensibly expressed as ‘the house’s big(ness)’. Ross (1996) ex-
24
Compare on this construction Paardekooper (1956), Quirk et al. (1987: 1279,
1284–1285), Everaert (1992), Foolen (2003), Plank (2003); also e.g. Ross (1998
a,b), Koptjevskaja-Tamm (2003a,b) and Lichtenberg (2005) and in particular
Keizer (2007: chapter 5). See Malchukov (2000) on DEPENDENCY REVERSAL.
plicitly compares such possessive constructions with the ones exemplified

in (72) and (73):
[…] I have come to the conclusion that Western Oceanic adjective structures
do not reflect historically e.g. “the bigness of the house” (which would be
exclamatory) but rather “a/the big one of a house” (in my dialect of English
“a whopper of a house”), where the “possessor” is coded as non-specific.
3.3. Possessives as quantifying satellites
Possessive modifiers are also found in NPs that contain quantifying expres-
sions, but such NPs typically seem to involve constructions in which the
quantifying expression displays at least some nominal properties (such as
plural marking) and takes the quantified entity as its argument.
(74) a. hundreds of people

b. millions of dollars
This kind of construction is common in the Slavic languages. For example,

in Russian numerals higher than one in the nominative or the so-called in-
animate accusative case appear with the NP referring to the quantified entity
in the genitive of quantification (for details see e.g. Comrie 1981: 101–104):
Russian
(75) pjat’ main
five:NOM car:GEN.PL
‘five cars’
It is not clear why there should be this gap in the modifier functions of a
possessive, but it may be relevant to point out that cardinal numerals have
been found to display certain properties that sets them apart form other
noun modifiers.
Firstly, whereas other grammatical modifier categories of the noun have
a strong tendency to (further) grammaticalize and increase their scope
(‘from inner to outer layer’), numerals tend to be rather resistant to such
diachronic processes. For example, collective markers (classifying opera-
tors/ωω 0) are a common source of plural markers (quantity/ω 2) and in many
languages demonstratives (ω 3) have grammaticalized into definite articles
(4). But as far as I know, only the lowest numerals are involved in gram-
maticalization processes. For example, in many languages the numeral ‘one’
78 Jan Rijkhoff
is the diachronic source of the indefinite article (Givón 1981) and it has
been established that at least in some languages the numeral ‘two’ is the
diachronic source of the dual marker (Dixon 1980: 323).
Secondly, whereas in many languages certain words from the spatial
domain may come to be used for temporal distinctions (see e.g. Haspelmath
1996), this is not so for numerals (Rijkhoff 2008d). For example, von Garnier
(1909) showed that there is clear diachronic relation between markers of
(spatial) collectivity and (temporal) perfectivity in languages of the Indo-
European family. Thus in Dutch and other Germanic languages the prefix
ge- still has a collectivizing meaning in nouns like gebroeders ‘(collection
of) brothers’ and gebergte ‘(collection of) mountains’ (modern German
Gebrüder and Gebirge respectively). At some point in the history of Ger-
manic this prefix came to be used with certain imperfective verbs to express
the notion of completeness (i.e. perfectivity) and ultimately it became asso-
ciated with the past participle form of the verb (Kirk 1923: 65), as in:
Dutch
(76) Heb jij dat ge-daan?
have you that done
‘Did you do that?’
To give another example of the change from Space to Time, in Panare (a

Cariban language) two tense-marking auxiliaries are derived etymologically
from demonstrative pronouns (Gildea 1993: 53).
In the case of number markers there is, however, no consensus in the di-
rection of change.25 For example, Newman (1990: 118) has argued that verbal
plural markers in Chadic developed from the nominal plural markers,
whereas Frajzyngier (1977) has claimed that verbal plural markers are the
source of nominal plural markers in the Chadic languages (see also Mithun
(1988) on this development in some North American languages).26
In sum, whereas in the case of classifying, qualifying and localizing op-
erators (Figure 1) there appears to be clear evidence for unidirectionality in
language change (‘from space to time’; Rijkhoff 2008d), there is no clear
25
Gil (1993: 281) claims that there is asymmetry between nominal and verbal
quantification: “…, while nominal quantifiers can show up on the verb, …, ver-
bal quantifiers can never show up on a noun.”
26
As to the origin of number markers, Frajzyngier (1997) argued that both nomi-
nal and verbal plural markers developed from the same source: a set of deictics,
determiners and anaphors (see on this subject also Lehmann 1982).
evidence for such a development in the case of quantifying operators, which

might be at least partly due to the fact that numeral concepts (singular/plural
number, cardinality) are beyond the dimensions of space and time.27 Notice
furthermore that all languages, in one way or other, seem to allow their
speakers to verbally specify properties concerning class, quality, location, or
discourse status of an entity, but that not all languages allow their speakers
to specify the exact number of entities, which also points to the special
status of quantifying modifiers compared to other modifier categories. Re-
cent studies investigating the relationship between language and arithmetic
in human cognition indicate that there is a distinction between a nonverbal
system of number approximation and a language-based counting system for
exact number and arithmetic.28 The latter system seems to be absent in
speakers of Australian and Amazonian languages (Bill McGregor pers.
comm.; Everett 2005). Evidence for the nonverbal representation of number
in human cognition is also offered in recent studies of infants (Jordan and
Brannon 2006; Berger et al. 2006).
3.4. Possessives as localizing satellites
Localizing modifiers specify the location of the referent in time or space in

the world of discourse. In definite NPs these modifiers typically provide
addressees (hearers or readers) with an entity through which they can iden-
tify the otherwise unidentifiable referent of the matrix NP (Prince 1981,
Rijkhoff 1989, Hawkins 1991, Haspelmath 1999). In Dutch the clearest ex-
amples of localizing satellites are probably adpositional NPs with a spatial
prepositions such as op ‘on’, onder ‘under, below’ or in ‘in’ (as before all
examples were collected with Google or Ask):
(77) De man op de fiets is Dirk Mol.

the man on the bicycle is Dirk Mol.
Op de derde foto zien we …
On the third photo see we …
‘The man on the bicycle is Dirk Mol. In the third picture we see …’
27
Notice that ordinal numerals are often expressed in terms of spatial concepts
(Rijkhoff 2004: 167).
28
Evidence for the nonverbal representation of number in human cognition is also
offered in recent studies of infants (Jordan and Brannon 2006; Berger et al.
2006).
80 Jan Rijkhoff
(78) Daar riep de taxichauffeur naar de voorbijgangers dat ze

there called the taxi driver to the passers-by that they
de man op de fiets moesten tegenhouden.
the man on the bicycle must stop
‘There the taxi driver called to the passers-by that they had to stop the
man on the bicycle’
One could say that an adpositional modifier such as op de fiets ‘on the bi-
cycle’ makes the referent of the matrix NP de man ‘the man’ locatable and
hence identifiable for the addressee in the world of discourse by establishing
a semantic relationship between referent of the matrix NP (the man) and the
referent of the adnominal possessive phrase (the bicycle). Notice that the
notion LOCATION is also closely related to the notion EXISTENCE: if an en-
tity has a location in the world of discourse it exists it that world of dis-
course, and vice versa (cf. Lyons 1977: 718–724; Bugenhagen 1986: 127).
The reason for treating (definite) possessives as localizing satellites has
to do with the fact that they, too, typically serve to ‘ground’ the (otherwise
unidentifiable) referent of the matrix NP: they license the definiteness of
the matrix NP by providing a referential anchor for the referent of that ma-
trix NP. In other words, without this modifier the addressee could not lo-
cate or identify the referent of the matrix in the shared world of discourse.
This can even be illustrated with these more or less isolated sentences. The
first example concerns a 71-year old man who in his younger days used to
participate in bicycle races):
(79) hij verzorgt zijn lichaam heel goed. Geen wonder dat men
he takes.care.of his body very well. No wonder that people
hem vroeg of dat de fiets van zijn vader was!
him asked if that the bicycle of his father was
‘He takes very good care of his body. No wonder that they asked him
if that was his father’s bicycle!’
Since de fiets ‘the bicycle’ is linked up with referent of the localizing satel-
lite (i.e. an identifiable entity: the father of the 71-year old man), the ad-
dressee has no problem identifying the referent of the bicycle in the shared
world of discourse. By contrast, without the adnominal possessive van zijn
vader ‘of his father’ the sentence would be pragmatically marked, as there
is no way in which the addressee could infer the existence or location of de
fiets ‘the bicycle’ in conversational space. The same goes for the next sen-
tences, which also contain the phrase de fiets van zijn vader ‘the bicycle of
his father’.
(80) Als we er voorbij liepen, wist hij zonder nadenken

when we there past walked, was.able he without thinking
de fiets van zijn vader aan te wijzen.
the bicycle of his father to.point.out
‘When we walked past it, he could point out his father’s bicycle im-
mediately’
The following example is taken from an account of an event that happened

during World War II, when a man called Tinus Veenstra and his companion
were forced to give their bicycles to some German soldiers.
(81) Beiden konden lopend terug, terwijl Tinus drie dagen later ook
both could walking back, while Tinus three days later also
de fiets van zijn vader diende af te staan.
the bicycle of his father must give.up
‘Both were allowed to walk back, but three days later Tinus also had
to give up his father’s bicycle’
Restrictive relative clauses are also typically used to provide the addressee
with a referential anchor for the identification of the referent of the matrix
NP. In fact there are languages in which relative clauses are almost exclu-
sively attested in definite NPs. Consider in this context, for example, these
remarks taken from Lehmann’s monograph on relative clauses (see also
Moravcsik 1969: 167, 170; Bach 1974: 192, 272; Givón 1990: 645ff.).
Mit einem Relativsatz kann man leicht einen bestimmten Gegenstand durch
Spezifikation der Situation, an der er teilhat, identifizieren. So erklärt es
sich, daß die typische Relativkonstruktion von einem Definitum begleitet ist,
wiewohl das natürlich prinzipiell nicht notwendig ist [With a relative clause
one can easily identify a certain object by specifying the situation in which
it is involved. This explains that the typical relative construction co-occurs
with a determiner, although in principle this is not necessary, of course – JR].
(Lehmann 1984: 402).
Das Adjektiv dient mehr der Begriffsbildung, der Relativsatz mehr der Ge-
genstandsidentifikation [The adjective primarily adds to the meaning, the
relative clause typically serves to identify an object – JR]
(Lehmann 1984: 405).
82 Jan Rijkhoff
As in the case of possessives, reference must be made to an identifiable en-

tity in the modifier. Compare these constructed examples:
(82) a. The police have arrested the man. [What man?]

b. The police have arrested the man who stole my car a couple of
days ago.
In (82a) the man has not been mentioned before and one is inclined to ask
what man the speaker is referring to. The response to the b-sentence is
probably different, for example “I did not know you had a car” or “I had no
idea your car was stolen”. The definiteness marking of man in (82b) does
not depend on whether or not the addressee knows the referent of the NP
the man, but on the hearer’s ability to identify the speaker’s car referred to
in the relative clause. Due to presuppositions contained in this clause (‘the
speaker has a car and it was stolen a couple of days ago’), the addressee has
no problem identifying the referent of the NP the man, whose existence and
location in the world of discourse can be inferred on the basis of its semantic
relation to the identifiable entity in that world of discourse referred to in the
adnominal modifier (‘my car’ i.e. the speaker’s car; in fact, ultimately just
the referent of ‘my’, i.e. the speaker).
Relationships between locative, possessive, as well as existential con-
structions have been observed and discussed in many studies. Lyons (1967)
was among the first to point out that these constructions are related, both
synchronically and diachronically. Clark (1970, 1978) investigated the na-
ture of this relation in more detail and demonstrated on the basis of a sam-
ple of 65 languages that these constructions are systematically connected in
terms of word order and patterns of verb use. She interprets possessors as
locations, saying that, cognitively, possessed items can be argued to be lo-
cated ‘at’ the possessor (Clark 1970: 3):
psychologically it would appear quite plausible to argue that if an object is
in some place, and the ‘place’ is actually an animate being, then the object is
possessed by the ‘place’. In other words, it is the [+Animate] feature added
to the locative phrase that transforms it into a Possessor-nominal.
Lyons and Clark were only concerned with sentential constructions, but
obviously the localist account of possession (see section 2.3) also holds for
adnominal possessive modifiers. An example of a locative element that has
developed into a marker of possessorship can be found in the following ex-
ample from Ewe (Niger-Kordofanian), where “the relational noun φé ‘place’
was used as a vehicle to denote possession and developed into a general
marker of nominal possession” (Claudi and Heine 1986; Heine and Kuteva
2002):29
Ewe
(83) fofó nye φé x
father my place house
‘my father’s house’ (‘the house at my father’s place’)
(Claudi and Heine 1986: 316)
3.4.1. Definiteness and indefinites in NPs with localizing possessives
So far we have been dealing with adnominal localizing possessives that

provide a referential anchor to ground the referent of the matrix NP, i.e. we
have been concerned with definite, referential possessives as adnominal
modifiers in a definite NP. This is in fact how localizing possessives are
commonly used. Haspelmath (1999), who did a text count in several lan-
guages (English, Italian, Modern Greek), found that adnominal possessives
serve as (what I would call) localizing satellites in approximately 95% of
the definite NPs. His results are corroborated by a random Internet search,
in which I checked all definite/indefinite combinations of the matrix NP
headed by the noun fiets (de fiets ‘the bicycle’ and een fiets ‘a bicycle’) and
definite and indefinite adnominal possessives headed by the nouns father,
mother, husband, wife, child, son, daughter, man, woman, boy, or girl. So
as to put an upper limit on the number of NPs, I only searched for examples
with an indefinite (Dutch een ‘a/an’) or definite article (de ‘the:C’ is the
common article, het ‘the:N’ its neuter counterpart) or the possessive pro-
noun mijn ‘my’. Consequently I ignored NPs with other determiners (such
as the Dutch equivalents of this, that, your or his) or indefinite quantifiers
(like enige ‘some’).
It turns out that localizing (definite) adnominal possessives are almost
exclusively attested in definite matrix NPs: 23,957 out of 23,988 cases (i.e.
99.87%).30
29
The locative sense of Dutch van ‘of ’ is still present in expressions specifying a
source, as in Hij komt van buiten ‘He comes from outside’ or De jongens stalen
een fiets van het meisje ‘The boys stole a bike from the girl’ (where van het
meisje ‘from the girl’ is an oblique argument of the verb stelen (van) ‘to steal
(from)’.
84 Jan Rijkhoff
DEFINITE + DEFINITE number of hits*

(84) a. de fiets van mijn vader (lit. ‘the bicycle of my father’) 1,310
b. de fiets van mijn man (lit. ‘the bicycle of my husband’) 821
c. de fiets van mijn moeder (lit. ‘the bicycle of my mother’) 1,380
d. de fiets van mijn vrouw (lit. ‘the bicycle of my wife’) 1,050
e. de fiets van mijn kind (lit. ‘the bicycle of my child’) 247
f. de fiets van mijn zoon (lit. ‘the bicycle of my son’) 1,040
g. de fiets van mijn dochter (lit. ‘the bicycle of my daughter’) 1,070
h. de fiets van de man (lit. ‘the bicycle of the man’) 4,250
i. de fiets van de vrouw (lit. ‘the bicycle of the woman’) 3,830
j. de fiets van het kind (lit. ‘the bicycle of the child’) 649
k. de fiets van de jongen (lit. ‘the bicycle of the boy’) 4,030
l. de fiets van het meisje (lit. ‘the bicycle of the girl’) 4,280
*Google, Sept. 06 23,957
Notice that the figures in (84) are slightly exaggerated because I did not
quite catch the odd doublet and all the (rare) instances in which the van-
phrase happened to follow the NP de fiets ‘the bicycle’ as a prepositional
phrase at the level of the clause (i.e. not as a modifier) specifying a source,
as in this example from a published police report (see also note 29):
(85) Een 31-jarige fietsendief en een 23-jarige heler zijn donderdagmid-

dag in de Oude Ebbingestraat aangehouden
‘A 31-year old bicycle thief and a 23-year old fence were arrested in
the Oude Ebbingestraat on Thursday afternoon’
De 23-jarige man kocht voor f 12,– een fiets van de man
the 23-year.old man bought for f 12,– a bicycle from the man
‘The 23-year old man bought a bicycle from the man for 12 guilders’
In this sentence, the buying event involves three parties (and the price of the
bicycle, of course: 12 Dutch guilders): the buyer (Agent), the item bought
(Patient), and the seller (Source), who in (85) is being referred to in the
prepositional phrase van de man ‘from the man’.
By contrast, the Internet search produced only 31 examples (31/ 23988 =
0.13% of the total number of instances) in which the matrix NP or the lo-
calizing possessive modifier is INDEFINITE.
30
A detailed study on definite and indefinite genitives in English (such as the/a
girl’s name) is Willemse (2005); see also Willemse et al. (this volume).
INDEFINITE + DEFINITE [4 instances]

(86) a. een fiets van mijn vader/man/moeder/vrouw/kind/zoon[2]/dochter
a bicycle of my father/husband/mother/wife/child/son/daughter
‘a bicycle belonging to my father/husband/mother/wife/child/
son/daughter’
b. een fiets van de man[1] / de vrouw[1] / het kind / de jongen /
a bicycle of the man / the woman / the child / the boy /
het meisje
the girl
‘a bicycle belonging to the man/the woman/the child/the boy/the
girl’
DEFINITE + INDEFINITE [20 instances]

(87) a. de fiets van een vader / moeder / kind/zoon/dochter
the bicycle of a father/husband/mother/wife/child/son/daughter
‘the bicycle of a father/husband/mother/wife/child/son/daughter’
b. de fiets van een man[4]/vrouw[5]/kind[3]/jongen[4]/meisje[4]
the bicycle of a man/woman/child/boy/girl
‘the bicycle of a man/woman/child/boy/girl’
INDEFINITE + INDEFINITE [7 instances]

(88) a. een fiets van een vader/ moeder /kind[2]/zoon /dochter
a bicycle of a father/husband/mother/wife/child/son/daughter
‘a bicycle belonging to a father/husband/mother/wife/child/son/
daughter’
b. een fiets van een man / vrouw[1]/kind/jongen[2]/ meisje[2]
a bicycle of a man / woman/child/boy/girl
‘a bicycle belonging to a man / woman/child/boy/girl’
Interestingly, 17 of the 31 cases were found in published police reports [14;

but see also (94)], film scripts [2] or jokes [1], discourse genres in which
persons and things are often referred to in general terms, because (i) they
must remain anonymous, (ii) their identity does not matter as they are
merely non-specific props figuring in the background:
86 Jan Rijkhoff
DEFINITE + INDEFINITE
(de fiets van een man / vrouw/ kind/ jongen /meisje)
From police reports:
(89) Op maandag werd de fiets van een vrouw gestolen.
on Monday became the bicycle of a woman stolen
‘Last Monday the bicycle of a woman was stolen’ (first sentence of
the report)
(90) Bijna twee jaar nadat de fiets van een vrouw uit
almost two year after the bicycle of a woman from
Veenendaal werd gestolen, heeft zij haar fiets terug.
Veenendaal was stolen, has she her bicycle back
‘Nearly two years after the bicycle of a woman from Veendendaal
was stolen, she has got her bicycle back’
From film scripts:

(91) en een tiener-meisje dat achterop de fiets van een man zit
and a teenage-girl who on.the.back the bicycle of a man sits
‘and a teenage girl who is sitting on the back of the bicycle of a man’
(92) Richard ziet hoe Wesley bijna in elkaar getrapt
Richard sees how Wesley almost in each.other kicked
wordt omdat hij naar de fiets van een jongen kijkt
becomes because he at the bicycle of a boy looks
‘Richard sees how Wesley almost gets beaten up because he is look-
ing at the bicycle of a boy’
INDEFINITE +INDEFINITE
(een fiets van een man / vrouw/kind/ jongen/meisje).
From police reports:
(93) Daders pakten tevens een fiets van een jongen af
perpetrators took also a bicycle of a boy away
‘(the) perpetrators also stole a bicycle of a boy’
(94) Een politieman mocht echter een fiets van een meisje
a policeman was.allowed however a bicycle of a girl
gebruiken
use
‘a policeman was, however, allowed to use a bicycle of a girl’
INDEFINITE +INDEFINITE
(een fiets van een man /vrouw/ kind/ jongen/meisje).
From a joke
(95) snel pakte de man een fiets van een vrouw
quickly grabbed the man a bicycle of a woman
‘Quickly the man grabbed a bicycle of a woman’
Notice also that a NP like a bicycle of a woman can also be interpreted as ‘a

woman’s bicycle’, i.e. with the possessive modifier woman’s as a classifying
modifier (see also note 8 and the end of section 3.4.1).
Let us now turn to the remaining 14 cases where we do not find a
DEFINITE + DEFINITE combination involving (what seems to be) a local-
izing possessive modifier of the noun. The only instance of an INDEFINITE
+ DEFINITE combination in a police report is used when damaged items
are listed:
INDEFINITE + DEFINITE
(een fiets van de man /de vrouw/ het kind/de jongen /het meisje).
From a police report:
(96) De twee, een 13-jarige jongen en een 34-jarige man, hadden
the two a 13-year.old boy and a 34-year.old man had
een tuinverlichting en een fiets van de vrouw vernield en
a garden_lights and a bicycle of the woman destroyed and
haar verbaal bedreigd. …
her verbally threatened
‘the two, a 13-year old boy and a 34-year old man, had destroyed gar-
den lights and a bicycle of the woman and verbally threatened her’
Emphasis falls naturally on the two indefinite NPs een tuinverlichting and
een fiets (van de vrouw), which indicates that this particular construction
(INDEFINITE + DEFINITE) is used to highlight the listed items. This is also
true for one of the other three instances of the combination INDEFINITE +
DEFINITE. This example occurs in a book report from a teenage boy, after
he mentioned other items the man was not supposed to have in his posses-
sion:
88 Jan Rijkhoff
INDEFINITE+DEFINITE
(een fiets van de man / de vrouw/ het kind/de jongen /het meisje).
(97) Ze vonden ook nog een fiets van de man
they found also still a bicycle of the man
‘Additionally they found a bicycle of the man’
The remaining 12 cases can be divided in two main groups. In one group,
the adnominal possessive contains a modifier that specifies a property that
is relevant for the further development of the story:
(98) De jongen die de fiets van een man zonder rechterbeen

the boy who the bicycle of a man without right.leg
had gejat,
had stolen
‘The boy who had stolen the bicycle of a man whose right leg was
missing’ (About a boy who steals a bicycle of a one-legged man,
which only has a pedal for the left foot; ironically in his attempt to
steal this bicycle, the boy breaks his right leg.)
(99) … de fiets van een man met een vies gele broek.
… the bicycle of a man with a dirty yellow trousers
Hij kent die broek …
He knows those trousers
‘… the bicycle of a man with dirty yellow trousers. He has seen those
trousers before.’
The second group has adnominal possessives that mention highly non-
specific entities (like the cases attested in the police reports). The NPs in
question typically occur in texts from chat sites or web-logs:
From a chat site:

(100) een vriend van me heeft eens, toen hij 6 was ofzo, …
a friend of me has once, when he 6 was or.so …
een fiets van een meisje in elkaar getrapt
a bicycle of a girl in each.other kicked
‘when he was around 6 years old, a friend of mine once destroyed a
bicycle of a girl’
Thus, there appears to be a rather strong connection between the value of

the discourse-referential operator ±Definite and the function of certain
modifiers (cf. Quirk et al. 1987: 1276). Whereas in a definite NP like the
ship in the ship’s funnel or the funnel of the ship serves as a localizing satel-
lite, in an indefinite NP such as a ship’s funnel (or the funnel of a ship) the
possessive serves as a classifying satellite (or is at least ambiguous between
the two modifying functions; see also note 8 on a boy’s shirt).
3.4.2. Localizing possessives: modification, predication, reference
Localizing possessives are the most unrestricted NP satellites: they can be

modified (101b), used as predicates (101c) and are fully referential (101d):
(101) a. de fiets van mijn vader

the bicycle of my father
‘my father’s bicycle’
b. de fiets van mijn oude vader [+ MODIFICATION]
the bicycle of my old father
‘my old father’s bicycle’
c. die fiets is van mijn (oude) vader [+ PREDICATION]
the bicycle is of my (old) father
‘that bicycle belongs to my (old) father’
d. de fiets van mijn/jouw/etc. (oude) vader [+ REFERENCE]
the bicycle of my/your/ etc. (old) father
‘the bicycle of my/your/etc. (old) father’
This is summarized in Table 3, which also shows that the layered model
(Figures 1 and 2) seems to iconically reflect the degree of grammatical in-
dependence of NP satellites (‘iconicity of distance’): the further away the
modifier is from the head noun in the model (i.e. the wider the scope of the
modifier) the less restricted it is with respect to MODIFICATION, PREDICA-
TION and REFERENCE.
90 Jan Rijkhoff
Table 3. Properties of classifying, qualifying and localizing possessives in Dutch
ADNOMINAL
LOCALIZING
de fiets van mijn de fiets van mijn die fiets is van mijn de fiets van
vader (oude) vader (oude) vader Peters vader
‘the bike of my ‘the bike of my the bike is of my ‘the bike of
father’ (old) father’ (old) father Peter’s father’
‘the bike belongs to (REFERENCE TO
my (old) father’ AN ENTITY)
QUALIFYING B+
beelden van beelden van de beelden zijn van beelden van
grote kwaliteit grote kwaliteit verschillende kwaliteit deze kwaliteit
‘statues of great ‘statues of high ‘the statues are of ‘statues of this
quality’ quality’ varying quality’ quality’
(REFERENCE TO A
PROPERTY OF AN
ENTITY; SECTION
4.2)
QUALIFYING B
een kroon van een kroon van de kroon is van —
goud (zuiver) goud (zuiver) goud
‘a crown of gold’ ‘a crown of (pure) the crown is of (pure)
gold’ gold
‘the crown is made of
(pure) gold’
QUALIFYING A+
een man van vele een man van vele — —
gezichten gezichten
‘a man of many ‘a man of many faces’
faces’
QUALIFYING A
een man van een man van — —
gezag (groot) gezag
‘a man of ‘a man of (great)
authority’ authority’
(SEE SECTION 4.1)
CLASSIFYING
een man van de — — —
wereld
4. On the gradual nature of modifier categories
It was mentioned earlier that it is sometimes rather difficult to draw a hard

and fast line between the various types of adnominal possessives (section
3.2.1). For that reason it seems better to regard the types distinguished in
Table 3 as reference points on a scale of noun modification rather than dis-
tinct modifier classes. Section 4.1 discusses the fuzzy boundary between
classifying and qualifying satellites of type A; section 4.2 is concerned with
qualifying possessives of type B, which closely resemble localizing posses-
sives (see Table 3).
4.1. Between Classifying and type A Qualifying adnominal possessives
We saw above that adnominal QUALIFYING possessives like of (great) stat-

ure or of many faces resemble adnominal CLASSIFYING possessives such as
of pleasure, the only difference being that qualifying possessives may or
must contain a modifier (selected from a rather small paradigm, e.g. small,
great) which is not true for classifying possessives, which resist internal
modification (Table 3). The close relationship between classifying posses-
sives and qualifying possessives of type A also manifests itself in the fact
that NPs with a qualifying possessive tend to be used in grammatical con-
texts where they have a strong classifying flavor.
Qualifying possessives of type A can occur as a modifier in subject or
object NPs, as in (102)–(104), but this does not happen very often.
Subject
(102) There is photographic evidence that native Australians of small stature
[…] lived in northern Queensland during the 19th and early 20th …
(103) Man of many pursuits insists arms trade isn’t among them.
Object
(104) The past couple of weeks we’ve been honored to have had riders of
such incredible stature racing their bikes once again in T-town.31
31
The phrase of stature is particularly frequent in biblical texts. For more exam-
ples, see: http://concordance.biblebrowser.com/s/stature.htm.
92 Jan Rijkhoff
It turns out that type A qualifying possessives are almost exclusively at-
tested in indefinite NPs that occur in predicate position as well as in cap-
tions, headings and book titles. From a grammatical perspective, titles are
rather similar to non-verbal clauses, except for the fact that non-verbal
clauses must contain a copula (at least in Dutch and English), which is
typically absent in book titles or article headings (instead we tend to find a
colon). Below are examples of the string ‘… man of many …’ in the contexts
mentioned above (all found on the Internet using the Ask search engine).
Fridtjof Nansen: Man of many facets. … His gentler qualities … came per-
haps from his quieter, more ascetic father, a lawyer of repute and a man of
unswerving integrity; Benjamin Franklin: A Man of Many Talents; David A.
Poulsen: A Man Of Many Hats; Alec Guinness – A Man of Many Parts;
Bryan Adams is a man of many faces; Nansen: man of many talents; A man
of many strong opinions … Carl Everett is a man of conviction; Maseda is
a man of many talents; Russell B. Farr: a man of many colors; Hip-Hop
Artist Madlib, Man of Many Names; Thomas Merton: Man of Many Jour-
neys; Man Of Many Voices – The Official Johnny Vallis Web Site; A Man of
Many Firsts – Otellini Finds New Ways to Reach Consumers; A man of
many parts; Ebbers is a man of many contradictions; Bruno Kirby: A Man of
Many Scenes; Ed Case is a Man of Integrity; Gideon Omnibus, a collection
of fictional stories about a Scotland Yard senior superintendent of enor-
mous stature and presence, …; Mhanda is a man of stature. … This is a
man of stature; And he slew an Egyptian, a man of great stature, five cubits
high.
The string ‘… woman of many …’ did not yield as many results, but here are
some examples:
A woman of many parts; Woman of many faces: Isabelle Huppert; (on

Christina Ricci) She is a woman of many talents; Maria Skobtsov, Woman
of Many Faces; Sheila McKenna, woman of many faces, is PG’s Performer
of the Year; Kate: woman of many names; Cassie aka Abigail Miller, a
woman of many faces. … Cassie is a woman of many facets, not unlike any
other gem; New queen a woman of many talents; A woman of stature: artist
Aleta Hayes portrays Sophocles heroine in solo piece.
These examples involve NPs with a type A qualifying possessive, but notice
that in each case it is the whole NP has a strong classifying flavor, not just
the possessive modifier. This ties in nicely with the fact that indefinite NPs
in predicate position are typically used to express class inclusion (Shelley is
a communist – Dik 1997 Part 1: 205; cf. Hengeveld 1992: 89, Keizer 1992:
ch. 6).32
(105) Aaron Sorkin is [a man of many words] CLASSIFYING PREDICATE NP
Thus, whereas in these examples above the adnominal possessive by itself

can be categorized as a QUALIFYING satellite of type A, the NP in which it
occurs has a CLASSIFYING function.
4.2. Between qualifying adnominal (Type B) and localizing satellites
We saw in section 3.4 that referential possessives typically occur as localiz-

ing satellites, but that there is also one kind of qualifying possessive modi-
fier that can be used referentially. This involves qualifying possessives of
type B+, which refer to a property (rather than an entity, as in the case of
localizing possessives):
(106) Wie een paar keer koffie van deze kwaliteit gedronken heeft, …
Who a couple.of time coffee of this quality drunk has, …
‘Someone who has tasted coffee of this quality a couple of times, …’
(107) Heb je geen kaarten van deze kleur …

Have you no cards of this color …
‘If you have no cards of this color …’
This may seem as if we cannot really distinguish between localizing pos-

sessives and qualifying possessives of type B+, but there is an important
difference. Whereas localizing possessives typically occur in a definite NP,
qualifying satellites of type B+ are normally attested in indefinite NPs,
where they do not serve to make the referent of the matrix NP identifiable
but specify the degree to which some property (QUALITY, COLOR, AGE,
32
By contrast, when a DEFINITE NP occurs in predicate position, it is typically
employed for identification purposes (Suzy is the singer). This is also true for
predicate NPs with a qualifying adnominal possessive of Type A, as in:
Dutch
(i) Huub is [de man van de computers]IDENTIFYING PREDICATE NP in ons kantoor
Huub is the man of the computers in our office
‘In our office, Huub is the man of the computers’
94 Jan Rijkhoff
SIZE, etc.) applies to an entity. One could say that type B+ qualifying pos-
sessives, which refer to PROPERTIES, form a bridge between the more typical
qualifying possessives, which cannot be used referentially, and localizing
possessives, which refer to ENTITIES and which makes it possible for the
addressee to identify the referent of the matrix NP.
These data indicate that the boundaries between the various modifier
categories (CLASSIFYING, QUALIFYING, LOCALIZING) are gradual rather
than distinct. Qualifying possessives of type A occur between other types
of qualifying possessives and classifying possessives, because the NPs in
which they occur often signal class inclusion, as in Ed Case is [a Man of
Integrity], i.e. Ed Case belongs to the class of men who are honest and
strong about what they believe to be right; section 4.1). Qualifying posses-
sives of type B+ are situated between other types of qualifying possessives
and localizing possessives because they are the only qualifying possessives
that can be used referentially (but recall that they refer to properties rather
than entities).
5. A tentative explanation
How can we explain the distribution of values for the three parameters in
Table 3? The short answer is: the different values reflect the functional re-
quirements of the various modifier functions in which the adnominal pos-
sessive is used. To put it differently, the possessive construction has a wide
range of semantic and syntactic possibilities that are exploited in different
degrees (depending on the modifier function it serves in the NP) by the lan-
guage user. From a formal perspective the possessive construction is simply
a phrasal structure with (potentially at least) the full array of grammatical
features associated with the three major lexical word class Adjective
(MODIFICATION), Verb (PREDICATION), and Noun (REFERENCE):
– MODIFICATION (i.e. internal modification): the head of the possessive

phrase can be modified itself (… van mijn/jouw/etc. vader ‘… of my/
your/ etc. father’);
– PREDICATION: the possessive phrase can occur in predicate position (…
is van mijn vader, lit. ‘… is of my father’, meaning ‘… (that) belongs to
my father’);
– REFERENCE: the possessive phrase can be used to pick out an individual
in the world of discourse (… van die man daar ‘… of that man over
there’).
The values for the three parameters in Table 3 seem to suggest that there are
no grammatical restrictions when the adnominal possessive with van ‘of’
serves as a LOCALIZING SATELLITE (section 3.4), providing the addressee
with an entity (a referential anchor) that makes the referent of the matrix NP
identifiable in the shared world of discourse. There is, however, one prag-
matic constraint: since the referential anchor must be an identifiable entity,
the localizing adnominal possessive phrase must be definite (a marginal num-
ber of apparent counter examples were discussed in the previous section).
When the possessive phrase is used as a QUALIFYING SATELLITE (section
3.2), it specifies a more or less characteristic property of the referent (color,
value, size, etc.), a function typically associated with adjectives (if a lan-
guage has them). Therefore there is no need to fully exploit the referring
potential of the possessive phrase. In fact, there is only one type of qualify-
ing possessives (type B+) which has restricted referential potential in that it
can be used to refer to the extent to which some property (size, age etc.)
applies to an entity (as in munten van DEZE kwaliteit ‘coins of this quality’).
Since qualities are gradable properties (a rather/very /incredibly small ta-
ble), it is possible to modify the head of the qualifying adnominal posses-
sive (i.e. modification), as in een parel van g rote waarde ‘a pearl of great
value’. Qualities are also properties that can be predicated of an entity,
therefore possessives used as qualifying satellites may also appear as predi-
cates (just like adnominal possessives used as localizing satellites), as in een
paspoort is van g rote waarde ‘a passport is of great value’. The fact that
type A qualifying possessives cannot occur as predicates (which puts them
in between classifying possessives and qualifying possessives of type B)
nicely illustrates the fuzzy boundary between qualifying and classifying
satellites (see note 8 and section 4).
CLASSIFYING POSSESSIVES further specify what kind of entity is being
denoted by the head noun (section 3.1), e.g. steam train, electric train, ex-
press train all specify a particular kind of train. It seems, however, that one
can only linguistically predicate a property of an entity that is conceptually
complete. Thus, the reason why classifying satellites cannot serve as a sen-
tence predicate is precisely that the subject entity does not constitute a con-
ceptually complete entity yet to predicate a property of, as the classifying
satellite actually helps to define that entity.33 In other words, one cannot say
33
Alternatively one could say (Bill McGregor, pers. comm.): the reason why clas-
sifying possessives do not occur as predicates is that a classifying modifier
highlights a (proto)typical feature of a set of entities (a TYPE), and such features
cannot be predicated of any particular instance (a TOKEN).
96 Jan Rijkhoff
that *a hero is urban or that *a man is of the cloth, because there first has
to be ‘an urban hero’ or ‘a man of the cloth’ before one can predicate some
property of these entities (as in, for example, the urban hero is a myth).34
It is basically for the same reason that classifying possessives cannot be
modified themselves. Since the classifying modifier and the head noun con-
stitute a tight unit, not just conceptually (as we just saw) but also syntacti-
cally, one can only modify the combination of classifying modifier-plus-
noun. In other words, the strong conceptual and syntactic bond between
classifying modifier and head makes it impossible to single out the classify-
ing possessive for modification or predicative purposes.
6. Conclusion
We have seen that adnominal possessives with van/‘of’ can occur in three
of the four descriptive modifier functions distinguished in the layered model
of the NP presented in Figures 1 and 2: they can be used as CLASSIFYING,
QUALIFYING, and LOCALIZING NP satellites. They are, however, not attested
as quantifying NP satellites, which may have to do with the special role of
numerical concepts in human cognition (section 3.3). Furthermore there is a
positive correlation between the kind of modifier function of the adnominal
possessive in the NP and its grammatical properties. CLASSIFYING AD-
NOMINAL POSSESSIVES (section 3.1), which only have the noun in their
scope, are severely restricted with respect to the three grammatical parame-
ters Modification, Predication, and Reference. By contrast, LOCALIZING
ADNOMINAL POSSESSIVES, which have the widest scope, are not restricted
with respect to these parameters, and the various types of QUALIFYING AD-
NOMINAL POSSESSIVES occupy intermediate positions between the two ex-
tremes (section 3.2):
34
Notice that of the cloth does not refer to a particular cloth, i.e. the presence of
the definite article in of the cloth does not mean that the classifying possessive
is used referentially here.
Table 4. Classification of adnominal possessive modifiers with van ‘of’ (A+/ B+ =

internal modifier obligatory, as in ‘of enormous length’).
FUNCTION MODIFICATION PREDICATION REFERENCE
LOCALIZING MODIFIER + + +
(reference to an
entity)
QUALIFYING MODIFIER B+ + + +
(reference to a
property)
QUALIFYING MODIFIER B + + —
QUALIFYING MODIFIER A/A+ + — —
CLASSIFYING MODIFIER — — —
The more general conclusions to be drawn from this contribution are,

firstly, that the term “attributive possession” (cf. Heine 1997 and many
others) is too blunt and secondly, that there is no such thing as ‘the func-
tion’ of a particular form or construction: members of the same formal
category (here: the adnominal possessive phrase with van/’of’) can be used
in different modifier functions, and, vice versa, the same modifier function
can be fulfilled by members of different form categories (illustrated, for
example, in section 3.1). Thirdly and perhaps more importantly, the results
of this study indicate that functional categories can be characterized in
grammatical terms, making it possible to capture grammatical differences
between members of the same form class (Table 3) and grammatical simi-
larities between members of different form classes (e.g. adjectives and
prepositional phrases; section 3.1) within and across languages.35
35
See McCawley (1985: 675) and Dik (1986) for a critical discussion of the ‘one
form – one function’ approach in linguistics (Newmeyer 1983).
98 Jan Rijkhoff
Acknowledgements
I am grateful to Bill McGregor and Lærke Munkholm Molbech for helpful

comments. Earlier versions of this paper were presented on three occasions:
[i] the 40th Annual Meeting of the Societas Linguistica Europaea (Septem-
ber 2007; University of Joensuu, Finland), [ii] the symposium ‘Nouns –
Cross-linguistically’ (June 2007, Campobasso, Italy), organized by the Max-
Planck-Institute for Evolutionary Anthropology (Leipzig, Germany) and the
Università del Molise (Italy), [iii] the workshop ‘Classification in Cognition’
(October 2007; University of Aarhus, Denmark), part of COST A31 (“Sta-
bility and adaptation of classification systems in a cross-cultural perspec-
tive”) financed by the EU Research Framework Program. I am grateful to
the audiences for interesting discussion, in particular Kristin Davidse and
Lutz Gunkel.
References
Aarts, Bas
2004a Modelling linguistic gradience. Studies in Language 28 (1): 1–49.
Aarts, Bas
2004b Conceptions of gradience in the history of linguistics. Language Sci-
ences 26 (4): 343–389.
Aarts, Bas
2006 Conceptions of categorization in the history of linguistics. Language
Sciences 28 (4): 361–385.
Aertsen, Henk, Mike Hannay and Rod Lyall (eds.)
2004 Words in their Places: A Festschrift for J. Lachlan Mackenzie. Ams-
terdam: Free University, Faculty of Arts.
Bach, Emmon
1974 Syntactic Theory. New York: Holt, Rinehart & Winston.
Barton, Irène, Michael Herslund and Finn Sørensen (eds.)
2001 Dimensions of Possession. Amsterdam: Benjamins.
Bauer, Laurie
1998 When is a sequence of two nouns a compound? English Language
and Linguistics 2: 65–86.
Bennis, Hans and Jan W. de Vries (eds.)
1992 De Binnenbouw van het Nederlands: Een Bundel Artikelen voor Piet
Paardekooper. Dordrecht: Foris.
Berger, Andrea, Gabriel Tzur and Michael I. Posner
2006 Infant brains detect arithmatic errors. Proceedings of the National
Academy of Sciences 103 (33): 12649–12653.
Bugenhagen, Robert D.
1986 Possession in Mangap-Mbula: its syntax and semantics. Oceanic
Linguistics 25: 124 –166.
Butler, Christopher S.
2003 Structure and Function: A Guide to Three Major Structural-Function-
al Theories. Part 1: Approaches to the Simplex Clause; Part 2: From
Clause to Discourse and Beyond. Amsterdam/Philadelphia: Benjamins.
Chappell, Hillary and William B. McGregor
1989 Alienability, inalienability and nominal classification. In Hall et al.
(eds.), 24–36.
Chappell, Hilary and William B. McGregor (eds.)
de Gruyter.
Clark, Eve V.
1970 Locationals: a study of the relation between ‘existential’, ‘locative’,
and ‘possessive’ constructions. Working Papers in Language Univer-
sals 3 (Stanford University), L1– L36 + xiii.
Clark, Eve V.
1978 Locationals: existential, locative, and possessive constructions. In
Greenberg et al. (eds.), 85 –126.
Claudi, Ulrike and Bernd Heine
1986 On the metaphorical basis of grammar. Studies in Language 10 (2):
297–335.
Cole, Peter (ed.).
1981 Radical Pragmatics. New York: Academic Press.
Comrie, Bernard
1981 Language Universals and Linguistic Typology. Oxford: Blackwell.
Comrie, Bernard
1988 General features of the Uralic languages. In Sinor (ed.), 451–477.
Comrie, Bernard (ed.)
1987 The World’s Major Languages. London: Croom Helm.
Connolly, John H. and Simon C. Dik (eds.)
1989 Functional Grammar and the Computer (Functional Grammar Series
10). Dordrecht: Foris.
Corston-Oliver, Simon
2002 Roviana. In Lynch et al. (eds.), 467–497.
Corum, Claudia, T. Cedric Smith-Stark and Ann Weiser (eds.)
1973 You Take the High Node and I’ll Take the Low Node. Papers from
the 9th Regional Meeting of the Chicago Linguistic Society, April
13–15, 1973. Chicago: Chicago Linguistic Society.
Dik, Simon C.
1986 On the notion ‘functional explanation’. Belgian Journal of Linguistics
1: 11–52.
100 Jan Rijkhoff
Dik, Simon C.
1989 The Theory of Functional Grammar (Functional Grammar Series 9).
Dordrecht: Foris.
Dik, Simon C.
1997 The Theory of Functional Grammar (2nd revised edition, edited by
Kees Hengeveld). Part 1: The Structure of the Clause. Part 2: Com-
plex and Derived Constructions. Berlin/New York: Mouton de Gruyter.
Dixon, Robert M.W.
1980 The Languages of Australia. Cambridge: Cambridge University Press.
Dixon, Robert M.W.
1982 Where have all the Adjectives Gone? And other Essays in Semantics
and Syntax. Berlin /New York: Mouton de Gruyter.
Englebretson, Robert
2003 Searching for Structure: The Problem of Complementation in Collo-
quial Indonesian Conversation [Studies in Discourse and Grammar
13]. Amsterdam /Philadelphia: Benjamins.
Everaert, Martin
1992 Nogmaals: Een schat van een kind. In Bennis and de Vries (eds.),
45–54.
Everett, Daniel L.
2005 Cultural constraints on grammar and cognition in Pirahã: another
look at the design features of human language. Current Anthropology
46 (4): 621–634 (plus comments and a reply from the author, 635–
646).
Foolen, Ad
2004 Expressive binominal NPs in Germanic and Romance languages. In
Radden and Panther (eds.), 75–100.
Farsi, A. A.
1968 Classification of adjectives. Language Learning 18: 45–60.
Frajzyngier, Zygmunt
1997 Grammaticalization of number: from demonstratives to nominal and
verbal plural. Linguistic Typology 1 (2): 193–242.
Fraurud, Kari
2001 Possessives with extensive use: a source of definite articles? In Bar-
ton et al. (eds.), 243–267.
García Velasco, Daniel and Jan Rijkhoff (eds.)
2008 The Noun Phrase in Functional Discourse Grammar [Trends in Lin-
guistics. Studies and Monographs 195]. Berlin /New York: Mouton
de Gruyter.
Giegerich, Heinz J.
2005 Associative adjectives in English and the lexicon-syntax interface.
Journal of Linguistics 41 (3): 571–591.
Gil, David
1993 Nominal and verbal quantification. Sprachtypologie und Universa-
lienforschung (STUF) 46 (4): 275–317.
Gildea, Spike
1993 The development of tense markers from demonstrative pronouns in
Panare (Cariban). Studies in Language 17 (1): 53–73.
Givón, Talmy
1981 On the development of the numeral ‘one’ as an indefinite marker.
Folia Linguistica Historica 2 (1): 35–55.
Givón, Talmy
1990 Syntax: A Functional-Typological Introduction. Volume II. Amster-
dam: Benjamins.
Greenberg, Joseph H., Charles A. Ferguson and Edith A. Moravcsik (eds.)
1978 Universals of Human Language. Volume 4: Syntax. Stanford: Stanford
University Press.
Guillemin, Diana
2007 A look at so in Mauritian Creole: From possessive pronoun to em-
phatic determiner. In Huber and Velupillai (eds.), 279–296. Amster-
dam /Philadelphia: Benjamins.
Gunkel, Lutz and Gisela Zifonun
2008 Constraints on relational-adjective noun constructions: a comparative
view on English, German and French. Zeitschrift für Anglistik und
Amerikanistik 56: 283–302.
Hall, Kiria, Michael Meacham and Richard Shapiro (eds.).
1989 Proceedings of the Fifteenth Annual Meeting of the Berkeley Lin-
guistics Society, February 18–20, 1989. General Session and Para-
session on Theoretical Issues in Language Reconstruction. Berkeley:
Berkeley Linguistics Society.
Halliday, Michael Alexander Kirkwood
2004 An Introduction to Functional Grammar (third edition, revised by
Christian M.I. M. Matthiessen). London: Arnold.
Hammond, Michael and Michael Noonan (eds.).
1988 Theoretical Morphology: Approaches in Modern Linguistics. New
York: Academic Press.
Haspelmath, Martin
1996 From Space to Time: Temporal Adverbials in the World’s Languages
(LINCOM Studies in Theoretical Linguistics 03). München: Lincom
Europa.
Haspelmath, Martin
1999 Explaining article-possessor complementarity: economic motivation
in noun phrase syntax. Language 75 (2): 227–243.
Hawkins, John A.
1991 On (in)definite articles: implicatures and (un)grammaticality predic-
tion. Journal of Linguistics 27: 405–442.
Heine, Bernd
1997 Possession: Cognitive Sources, Forces, and Grammaticalization
(Cambridge Studies in Linguistics 83). Cambridge: Cambridge Uni-
versity Press.
102 Jan Rijkhoff
Heine, Bernd, Ulrike Claudi and Friederike Hünnemeyer

1991 Grammaticalization: A Conceptual Framework. Chicago: The Uni-
versity of Chicago Press.
Heine, Bernd and Tania Kuteva
2002 World Lexicon of Grammaticalization. Cambridge: Cambridge Uni-
versity Press.
2007 The Genesis of Grammar: A Reconstruction. Oxford: Oxford Uni-
versity Press.
Hengeveld, Kees
1988 Layers and operators. Working Papers in Functional Grammar 27.
Hengeveld, Kees
1989 Layers and operators in Functional Grammar. Journal of Linguistics
25 (1): 127–157.
Hengeveld, Kees
1990 The hierarchical structure of utterances. In Nuyts et al. (eds.), 1–23.
Hengeveld, Kees
1992 Non-Verbal Predication: Theory, Ttypology, Diachrony. Berlin /New
Hengeveld, Kees
1994 Ik heb zoiets van: “Ze bekijken het maar!”. In F.G. van Werkgem
(ed.), 8–12.
Huber, Magnus and Viveka Velupillai (eds.)
2007 Synchronic and Diachronic Perspectives on Contact Languages
[Creole Language Library 32]. Amsterdam /Philadelphia: Benjamins.
Jordan, Kerry E. and Elizabeth M. Brannon
2006 The multisensory representation of number in infancy. Proceedings
of the National Academy of Sciences 103 (9): 3586 –3489.
Josephson, Folke and Ingmar Söhrman (eds.).
2008 Interdependence of Diachronic and Synchronic Analyses [Studies in
Language Companion Series 103]. Amsterdam /Philadelphia: Ben-
jamins.
Keizer, M. Evelien
1992 Reference, predication and (in)definiteness in functional grammar. A
functional approach to English copular sentences. Ph.D. dissertation,
Vrije Universiteit Amsterdam.
Keizer, M. Evelien
2007 The English Noun Phrase – The Nature of Linguistic Categorization.
Cambridge: Cambridge Universtity Press.
Kirk, Arthur
1923 An Introduction to the Historical Study of New High German. Man-
chester: The University of Manchester Press.
Koptjevskaja-Tamm, Maria
2002 Adnominal possession in the European languages: form and function.
Sprachtypologie und Universalienforschung (STUF) 55 (2): 141–172.
2003a A woman of sin, a man of duty, and a hell of a mess: non-determiner
genitives in Swedish. In Plank (ed.), 515–558.
2003b Possessive noun phrases in the languages of Europe. In Plank (ed.),
621–722.
Lehmann, Christian
1984 Der Relativsatz: Typologie seiner Strukturen, Theorie seiner Funk-
tionen, Kompendium seiner Grammatik. Tübingen: Narr.
Levi, Judith N.
1973 Where do all those other adjectives come from? Papers from the
Ninth Regional Meeting of the Chicago Linguistic Society, April 13–
15. In Corum et al. (eds.), 332–345.
Lichtenberg, Frantisek
2005 On the notion “adjective” in Toqabaqita. Oceanic Linguistics 44 (1):
113–144.
Lynch, John, Malcolm Ross and Terry Crowley (eds.).
2002 The Oceanic Languages. Richmond: Curzon.
Lyons, John
1967 A note on possessive, existential and locative sentences. Foundations
of Language 3: 390–396.
Lyons, John
1977 Semantics (2 volumes). Cambridge: Cambridge University Press.
Malchukov, Andrej L.
2000 Dependency Reversal in Noun-Attribute Constructions: Towards a
Typology (LINCOM Studies in Language Typology 03). München:
Lincom Europa.
McCawley, James
1985 Review of Newmeyer 1983. Language 61: 668–679.
Menges, Karl Heinrich
1968 The Turkic Languages and Peoples: An Introduction to Turkic Studies.
Wiesbaden: Harrassowitz.
Mithun, Marianne
1988 Lexical categories and the evolution of number marking. In
Hammond and Noonan (eds.), 211–234.
Moravcsik, Edith A.
1969 Determination. Working Papers on Language Universals 1: 64 –130.
Mosel, Ulrike and Even Hovdhaugen
1992 Samoan Reference Grammar. Oslo: Universitetsforlaget AS.
Newman, Paul
1987 Hausa and the Chadic languages. In Comrie (ed.), 705–723.
Newman, Paul
1990 Nominal and Verbal Plurality in Chadic (Publications in African
Languages and Linguistics 12). Dordrecht: Foris.
104 Jan Rijkhoff
Newman, Paul and Roxana Ma Newman (eds.).

1977 Papers in Chadic Linguistics. Leiden: Afrika-Studiecentrum.
Newmeyer, Frederik J.
1983 Grammatical Theory: its Limits and Possibilities. Chicago: Univer-
sity of Chicago Press.
Nuyts, Jan, A. Machtelt Bolkestein and Co Vet (eds.).
1990 Layers and Levels of Representation in Language Theory: A Func-
tional View. Amsterdam: Benjamins.
Paardekooper, Petrus C.
1956 Een schat van een kind. De Nieuwe Taalgids 49: 93–99.
Plank, Frans
2003 Double articulation. In Plank (ed.), 337–395.
Plank, Frans (ed.).
2003 Noun Phrase Structure in the Languages of Europe (Empirical Ap-
proaches to Language Typology/Eurotyp 20-7). Berlin /New York:
Mouton de Gruyter.
Prince, Ellen F.
1981 Toward a taxonomy of Given-New information. In Cole (ed.), 223–
255.
1985 A Comprehensive Grammar of the English Language. London:
Longman.
Radden, Günter and Klaus-Uwe Panther (eds.).
2004 Studies in Linguistic Motivation. Berlin / New York: Mouton de
Gruyter.
Rausch, P. J.
1912 Die Sprache von Südost-Bougainville, Deutsche Salomonsinseln.
Anthropos 7: 105 –134, 585 – 616, 964 – 994.
Rédei, Károly
1978 Zyrian Folklore Texts (translated by I. Gombos). Budapest: Akadémia
Kiodó.
Rijkhoff, Jan
1988 A typology of operators: toward a unified analysis of terms and
predications. Working Papers in Functional Grammar 29.
Rijkhoff, Jan
1989 The identification of referents. In Connolly and Dik (eds.), 229–246.
Rijkhoff, Jan
1990 Toward a unified analysis of terms and predications. In Nuyts et al.
(eds.), 165–191.
Rijkhoff, Jan
1992 The noun phrase: a typological study of its form and structure. Ph.D.
dissertation, University of Amsterdam.
Rijkhoff, Jan
2002 The Noun Phrase. Oxford: Oxford University Press.
Rijkhoff, Jan
2003 When can a language have nouns and verbs? Acta Linguistica Haf-
niensia 35: 7–38.
Rijkhoff, Jan
2004 Iconic and non-iconic word order patterns: on symmetry in the NP
and counter examples to Universal 20’. In Aertsen et al. (eds.), 169–
180.
Rijkhoff, Jan
2008a Layers, levels and contexts in functional discourse grammar. In Gar-
cía Velasco and Rijkhoff (eds.), 63–115.
Rijkhoff, Jan
2008b Layering and iconicity in the noun phrase: descriptive and interper-
sonal modifiers. Linguistics 46 (4): 789–829.
Rijkhoff, Jan
2008c On flexible and rigid nouns. Studies in Language 32 (3): 727–752
(Special issue, U. Ansaldo, J. Don and R. Pfau (eds.), Parts of Speech:
Descriptive Tools, Theoretical Constructs).
Rijkhoff, Jan
2008d Synchronic and diachronic evidence for parallels between noun
phrases and sentences. In Josephson and Söhrman (eds.), 13–42.
Rijkhoff, Jan and Johanna Seibt
2005 Mood, definiteness and specificity: a linguistic and a philosophical ac-
count of their similarities and differences. Tidsskrift for Sprogforskning
3 (2): 85–132. http://ojs.statsbiblioteket.dk/index.php/tfs/issue/archive.
Rosenbach, Anette
2006 Descriptive genitives in English. English Language and Linguistics
10 (1): 77–118.
Ross, Malcolm
1996 Adjectives with possessor nouns. LINGUIST List: Vol-7-1678, Sub-
ject 7.1678.
Ross, Malcolm
1998a Proto-Oceanic adjectival categories and their morphosyntax. Oce-
anic Linguistics 37: 85–119.
Ross, Malcolm
1998b Possessive-like attribute constructions in the Oceanic languages of
northwest Melanesia. Oceanic Linguistics 37: 234 –276.
Schachter, Paul
1985 Parts-of-speech systems. In Shopen (ed.), 3–61.
Seiler, Hansjakob
Narr.
Shopen, Timothy (ed.).
1985 Language Typology and Syntactic Description. Volume I: Clause
Structure. Cambridge: Cambridge University Press.
106 Jan Rijkhoff
Siewierska, Anna, Jan Rijkhoff and Dik Bakker

1998 Appendix: 12 word order variables in the languages of Europe. In
Siewierska (ed.), 783–812.
Siewierska, Anna (ed.)
1998 Constituent Order in the Languages of Europe (Empirical Ap-
proaches to Language Typology/Eurotyp 20-1). Berlin /New York:
Mouton de Gruyter.
Sinclair, John (ed.)
1990 Collins COBUILD English Grammar. London: Harper Collins.
Sinor, Denis (ed.).
1988 Handbuch der Orientalistik, Achte Abteilung; V-1: The Uralic Lan-
guages – Description, History and Foreign Influences. Leiden: Brill.
Song, Jae Jung
2005 Grammaticalization and structural scope increase: possessive-classi-
fier-based benefactive marking in Oceanic languages. Linguistics 43
(4): 695–838.
Sorace, Antonella and Frank Keller
2005 Gradience in linguistic data. Lingua 115: 1497–1524.
Tauli, Valter
1966 Structural Tendencies in the Uralic Languages (Indiana University
Publications, Uralic and Altaic Series 17). Bloomington: Indiana Uni-
versity.
Taylor, John R.
van Baarda, M. J.
1908 Leiddraad bij het bestuderen van ’t Galela’sch dialekt, op het eiland
Halmaheira [Manual for the study of the Galela dialect, on the island
of Halmahera]. The Hague: Nijhoff.
van Werkgem, Fienie van (ed.).
1994 Dubbel Nederlands – 23 Opstellen voor Simon C. Dik. Amsterdam:
IFOTT / Dept. of Linguistics, University of Amsterdam.
von Garnier, Katharine
1909 COM– als perfektierendes Praefix bei Plautus, SAM– im Rigveda,
CYN– bei Homer. Indogermanische Forschungen 25: 86–109.
Williams, Edward
1981 Argument structure and morphology. The Linguistic Review 1: 81–
114.
Willemse, Peter
2006 Esphoric the N of a(n) N-nominals: forward bridging to an indefinite
reference point. Folia Linguistica 40 (3–4): 319–364.
Willemse, Peter, Kristin Davidse and Liesbet Heyvaert
this vol. English possessives as reference-point constructions and their func-
tion in the discourse.
Is possession mere location?
Contrary evidence from Maa
Doris L. Payne
1. Introduction
A claim sometimes found in typological and cognitive linguistic literature

is that the concept of predicative “possession” is essentially no different
than the concept of location,1 and hence that “Possessor” as a semantic role
does not exist (Baron and Herslund 2001; DeLancey 2002; Freeze 2001;
Sørensen 2001, inter alia and section 2 below). An opposing view is that
Possessor is a bona fide semantic role, distinct from Locative. My goal in
this paper is empirical as well as conceptual: I will examine Maa (Maasai
and Il-Chamus) elicited and corpus data to evaluate the extent to which cer-
tain verb roots, and certain constructions involving those roots, are used for
both location and possession. To the extent that there may be overlap in
senses for a single root or root-plus-construction, it supports the view that
possession is merely location with something like an animate Locative. To
the extent that this does not happen, it provides support for the cognitive
distinctness of possession from location. I will conclude that there is solid
evidence for the latter position, at least for Maa.
This paper is concerned with the type of cognitive category or catego-
ries expressed in predicative possession. By “predicative possession” I have
in mind the type of construction in which possession is related to the mean-
ing of the predicate (e.g. have) or to the meaning of an entire clausal con-
struction (e.g. certain copular constructions). However, I exclude here Ex-
ternal Possession which is also clausal in nature (cf. Payne 1997, Payne and
Barshi 1999), and exclude the type of predicative possession that involves a
1
I capitalize semantic role labels (Locative, Theme, etc.), but I use lower case
(location, possession, etc.) when representing the view that there is not necessar-
ily a theoretically relevant semantic role involved. When in quotes, “possession”
indicates the view that “Possessor” is a spurious semantic role and no different
from Locative or the predication of location. I will use the abbreviations PR
(possessor) and PM (possessum) only when I am not particularly concerned one
way or the other with the theoretical status of such notions as semantic roles.
108 Doris L. Payne
genitive construction in the predicate, i.e. constructions like This book is

mine/John’s). The use of NP-internal devices to express possession is only
tangentially mentioned. I basically adopt a Construction Grammar approach,
in part because it is too simplistic to say that all semantic role information
can be accounted for solely by properties of lexical predicates (cf. Goldberg
1995). It is also too simplistic to address verb senses without simultaneously
considering the constructions that verbs occur in.
I will first present two views on the relationship between possession and
location (sections 2 and 3), and lay out definitions of key terms as used in
this paper (section 4). Basic morphological and clause-level constructions
of Maa are then introduced (section 5). Against this background, we can
see that the Maa verb tii ‘be at’ appears to have locational and existential
uses in particular constructions, while the verb ata ‘have’ has possessive
and existential uses in particular constructions (sections 6 and 7). What is
less clear from simple elicited data is whether the roots tii and ata overlap
at all in predicating location and possession. Thus, a corpus study is under-
taken (section 8). The elicited and text data converge to support the conclu-
sion that speakers specialize root+constructional combinations for predicat-
ing possession versus location. The only plausible conclusion must be that
human cognizers can and do recognize sufficiently-important conceptual
distinctions between possession and location, both of which are relational –
hence, semantic role-like – conceptualizations obtaining between two enti-
ties. On the other hand, the fact that tii is used for both locational and exis-
tential predications and that ata is used for both possession and existential
predications suggests that where languages do show single forms or par-
tially shared constructions for predicating both location and possession, we
should further explore whether the conceptual link is not via an existential
stage. That is, if we have a root that historically first predicates location, is
this then extended to predicating existence; and once its use for predicating
existence is established, is it then perhaps the existential which is extended
for predicating possession? 2
2
William McGregor (pers. comm.) notes that in Nyulnyul and related Australian
languages, a negative existential and a negative possessive construction share
highly similar syntax that must be related historically. Significantly, there is no
formally related negative locative construction to provide any link between the
two – i.e. the possessive and existential constructions are directly connected.
Is possession mere location? Contrary evidence from Maa 109
2. The “possession is location” view
The view that “possession” is just location with a (likely animate) Locative
is well attested in the literature (though held to varying degrees of strength).
I will refer to this as the “possession-is-location” view. For just a few ex-
amples, Baron and Herslund (2001) argue that the fundamental meaning of
have is that one object is simply located with respect to another, though
additional part-whole or other meanings may be superimposed onto this
fundamental meaning. DeLancey (2002: 8) suggests that constructions which
predicate possession and existence/location have the “same underlying
structure, and differ only in the relative salience, inherent or contextually-
determined, of the two arguments”. Additional claims in this direction are
found in Gruber (1965), Lyons (1967, 1968), Clark (1978), Jackendoff (1983),
inter alia. These sorts of claims are generally grounded in the view that the
semantic role(s) of both putative “Possessor” and Locative derive from
one-and-the-same conceptual or cognitive model of [x BE.AT y], where x is
a Theme and y is Locative. In general, Theme and Locative derive only
from either BE.AT or GO.TO types of abstract predicate notions, which
may be understood in either literal or metaphorical ways. Agent derives
from a CAUSE (or DO) predicate notion. All more lexically-specific predi-
cates can be subsumed to one of these, and hence all core semantic roles in
any clause whatsoever can be subsumed to Agent, Theme or Locative.3
DeLancey (2002) argues that Theme and Locative are grounded in – if
not exactly equivalent to – the cognitive distinction between Figure and
Ground (cf. Wertheimer 1923). In cognitive studies, Figure designates an
entity which perceptually stands out against a Ground, it is the focus of at-
tention, and it is seen in detail. A Ground is that part of the perceptual field
against which the Figure stands out, it is in the periphery of attention, lacks
3
In this paper I cannot attempt to address the vast literature on theories or systems
of semantic roles. DeLancey (2002) is concerned with developing a constrained
theory of semantic roles that can account for surface case marking of core ar-
guments of predicates, and specifically not with oblique cases. He does not deny
the existence of semantic categories like Beneficiary or Instrument for obliques,
but denies they are ever warranted as roles for core arguments. My own approach
is to posit whatever is empirically driven, and believe the cross-linguistic evi-
dence supports a somewhat larger inventory than DeLancey posits (cf. Comrie
and van den Berg (2006) on case marking evidence for core Experiencier distinct
from core or any type of oblique Locative).
110 Doris L. Payne
detail and, according to perceptual psychologists, is not usually even per-

ceived as an object.
Particular lexemes may combine CAUSE and GO.TO/BECOME predi-
cate notions, resulting in all three roles being present in a sentence. For ex-
ample, with the predicate provide, as in This document provides you with
advice, the Agent is this document, the Locative is you, and the Theme is
advice. Also, a core role may sometimes be expressed in the verbal lexeme
itself. In Mary baked the cookies, Mary is Agent, the cookies is Theme, and
the metaphorical Locative, i.e. the condition of being baked, is lexicalized
into the verb. In Susan kissed him, Susan is Agent, him is Locative, and the
Theme “kiss”, which is transferred from one participant to the other, is
lexicalized into the verb.
A strength of the general framework surrounding the possession-is-
location view is that it is highly constrained. Faced with a new sentence
with unique verb and novel NPs, one cannot lightly suggest there must be
new roles present just because it somehow “feels” different; properties of
particular referents cannot be misconstrued as necessarily implying a dif-
ferent semantic role. Thus, in The storm blew the house down, the storm
does not have a distinct role from Agent merely because it is inanimate.
Rather, because it is linguistically presented as the CAUSE of a change, it is
necessarily the Agent. With regard to possession, if y happens to be human
or is otherwise well-individuated, and has potential power over the disposi-
tion of x, one might be tempted to say – naively, the possession-is-location
model would hold – that the relationship between x and y is one of “posses-
sion”. But analogously to the case of the storm, the naiveté turns on mistak-
ing features inherent in the nature of particular referents for semantic roles,
when the latter must derive strictly from the nature of predicates.
Returning to our focus on possession versus location, in this model
meaning differences between predicates like have/own versus be (at)/reside
(in) are considered irrelevant for determining semantic roles, as all these
predicates subsume to the [x BE.AT y] cognitive model. This interpretation
is argued for “Primarily on the basis of gross parallels in the organization of
the syntactic and lexical expression of possession and location” in multiple
languages (DeLancey 2002: 8, drawing on Gruber, Lyons and Jackendoff;
see especially Clark 1978). Consider The garden has lots of bees and I have
ten dollars, both of which employ the lexical verb have. Under the posses-
sion-is-location view, the garden and I are both Locatives, while lots of
bees and ten dollars are Themes conceptualized as being at those Loca-
tives. The fact that one of the places happens to be a well-individuated
animate entity that can control and even dispose of the money, while the
other is an extended physical space, are irrelevant conceptual differences

for an inventory of semantic roles. Similarly, French Ce livre est a Marie
‘Marie has a book’ and Pierre est a Paris ‘Pierre is in Paris’, share the
structure [NP est a NP]. Such pairs transparently suggest that a PR (like
Marie) may be conceptualized as a Locative (like Paris) where a Theme
(PM) is located. In Jakaltek (Jacaltec; Craig 1977: 19–21), existence, loca-
tion, and possession can all be predicated with the same copula ay: 4
Existential
(1) Ay anma yul conob mach scuy yuninal yin abxubal.
exist people in town not teach their.children in Jacaltec
‘There are people in town who do not teach their children Jacaltec.’
Location
(2) Ay-c’oj ha mam.
exist-DIR your father
‘Is your father here?’
Possession
(3) Ay no’ hin txitam.
exist CL my pig
‘I have a pig.’
In some very important historical and cognitive work, Heine (1997, 2001)
argues that there are eight conceptual sources of predicative possession
constructions, only one of which he calls locational. An ACTION conceptual
4
Abbreviations are as follows: A – most agent-like argument of canonical transi-
tive verb, ACC – accusative, ANTIP – antipassive, ASSOC – associative, CL –
noun class, CN – discourse connective, COMP – complementizer, COP – copula,
DAT – dative, DEF – definite, DIR – directional, DSCN – discontinuous discourse
connective, EP – epenthetic, F – feminine, IMPF – imperfective, INF – infinitive,
LOC – locative, MID – middle, NEG – negative, NMLZ – nominalizer, NOM –
nominative, NPF – non-perfect/non-perfective, OBJ – object, OBL – oblique, P –
most patient-like argument of canonical transitive verb, PASS – (impersonal)
passive, PF – perfect/perfective aspect, PL – plural, PM – possessum, POSS –
possessive, PR – possessor, PRF – perfect, PRT – particle, REL – relative, S –
single argument of canonical intransitive verb, SBJ – subject, SBJV – subjunc-
tive, SG – singular, VENT – ventive. Maa examples taken from texts are tagged
with a text name and line number.
112 Doris L. Payne
source would be something like She grabbed/took/kept/held the book where

a verb like grab/take historically develops a specifically possessive sense
(which is the historical story of English have). The other seven are LOCA-
TION, GOAL, COMPANION, GENITIVE, SOURCE , TOPIC(-comment), and
EQUATION; these are illustrated in (4). (Word order in (4) is irrelevant and
artificially falls out as it does only because I am trying to express these
conceptual sources in sensible English.)
Possession-is-location adherents would aver that essentially all of the
sources in (4) are locational, as they can be subsumed to a BE.AT predicate
type, as indicated by the headings in (4). Further, in many languages a
“Possessor” requires an adposition that doubles as a Locative marker in
clear locative phrases (Clark 1978) – which would seem to be true in at
least Heine’s LOCATION, GOAL , and SOURCE scenarios.
(4) Conceptual sources of predicative possession constructions, viewed

according to the [Theme BE.AT Locative] model
THEME BE.AT LOCATIVE
LOCATION The book is at her.
GOAL The book is for/to her.
GENITIVE Her book exists (at her).
SOURCE The book exists from her.
EQUATION The book is hers.
LOCATIVE THEME BE.AT
COMPANION She with the book is.
TOPIC As for her, the book exists.
The only one of Heine’s schemas that at first glance might seem problematic
to the possession-is-location framework is ACTION because she would be
AGENT in She grabbed/took the book (i.e. these could well stand as an an-
swer to What did Mary do? She grabbed the book.) However, the fully-
specified event chain underlying Mary grabbed the book would be [Mary
CAUSE [ [book GO.TO Mary ] & [book BE.AT Mary] ] ]. Furthermore,
since the preceding schemas are presented as historical sources of posses-
sion constructions, the possession-is-location adherent simply notes that as
CAUSE or “doing” semantics is eventually bleached out of a verb like grab
and it comes to have an inherently stative meaning, the descendent verb is
no longer a CAUSE type of predicate (nor even a GO.TO predicate), but
simply a BE.AT predicate.
3. The “possession is more-than-location” view
In some contrast to the preceding view is, I believe, the view of possession
articulated by Langacker (1993) and Velazquez-Castillo (1996), as well as
scholars who treat Posessor and Locative as distinct semantic roles (e.g.
Kemmer 2002; Tham 2005; Van Valin and LaPolla 1997: 125–127).5 Lang-
acker and Velazquez-Castillo point out how possession involves the cogni-
tive notion of singling out one entity, the “target” (or PM), for individual
conscious awareness in terms of a more salient entity which is the “refer-
ence point” (or PR). According to Langacker (1993: 6), a reference point
construction includes not only the “target” and the “reference point”, but
the relevant construction is applied within a particular contextually-deter-
mined “dominion” (or set of entities) which the reference point enables ac-
cess into.
Reference points are determined by ease of identifiability on the part of
the conceptualizer (e.g. speaker). Ease of identifiability can be context de-
pendent, but in some way the reference-point entity has greater perceptibil-
ity – whether because it is inherently more salient, or because it is an entity
for which the conceptualizer has greater empathy, familiarity, acquaintance,
etc. Strikingly, such features correspond to cognitive Figures, as opposed to
their Grounds; i.e., a Figure is conceptually the most salient entity. Thus,
conceptually the salient PR is naturally a primary Figure, which stands out
against its background. (If anything corresponds to Ground in Langacker’s
scenario, it would seem to be the “dominion”.) Thus, if Theme and Locative
are grounded in or equivalent to the conceptual notions of Figure and
Ground, overall this should lead to the designation of PRs as being tanta-
mount to Themes – contrary to what is assumed in the possession-is-loca-
tion view.
Now, of course, there can be hierarchical layers of what is Figure and
what is Ground – one Figure could be a Ground for yet another Figure as
the mind’s eye ranges across a conceptual scene – and this is doubtless part
of what is conceptually operative in possessive constructions. But very im-
portantly, this potential nesting of a Figure against Figure-as-relative-
Ground does not destroy the insight that PRs are highly perceptible and sa-
5
Van Valin and LaPolla’s (1997) multiple semantic roles are differentiated from
their small set of macro-roles, and their Locative and Possessor are claimed to
pattern together in various ways along with Experiencier, Attributant, etc.
Kemmer (2002) presents typological evidence for distinct conceptualization of
Possessor from Locative.
114 Doris L. Payne
lient. Crucially, they may be highly perceptible and salient relative to the
PM. Thus, predicating possession is different from just predicating loca-
tion. If possession is “location” at all, it is localizing something relative to a
highly perceptible, typically individuated, Figure-like entity. We might say
that the PR is a “big” Figure, relative to a “small” Figure that is the PM.
Note that these are relational (hence semantic or cognitive role-like) notions,
regardless of whether they are instantiated in a verbal or clausal predicate
or within an NP structure. That is, Figure and Ground mutually co-define
each other, and cannot be reduced to the features of solitary referents in-and-
of themselves.6 I will refer to this as the “possession-is-more-than-location”
view.
All this leads to the question: Is there empirical lexical and/or morpho-
syntactic evidence that languages treat predicative possession as a signifi-
cant, even basic, concept relevant to the structure of human language? That
is, is there evidence supporting a distinct Possessor role (and possibly also
Possessum role) related to certain predicates, different from Locative and
Theme – at least in certain languages? Or can lexical and constructional
behaviour be fully accounted for just in terms of the three core roles posited
by the possession-is-location theory?
My answer is “I don’t think that possession-is-location accords with all
the linguistic data.” Indeed, this answer should not be surprising given the
prima facie ramifications that understandings of possession have in ordi-
nary social relations.7 The most obvious empirical evidence is that many
languages do have distinct syntactic structures for predicating location ver-
sus possession. Indeed, even in many languages where a single predicate
may be used for possessive as well as locative and/or existential meanings,
there are additional required grammatical features, besides just the predi-
cate morpheme, which distinguish the meanings. This can be seen in the
6
One reason a Possessor may seem different from a Locative is because a Posses-
sor may often be topical. Proponents of the possession-is-location view would
presumably say that participant topicality has no direct bearing on the existence
of any particular semantic role in a clause, as topicality is a discourse or other
cognitive property not based in the relational nature of the predicate.
7
Indeed, referring to Lyons (1967: 392), DeLancey (2002: 9) suggests that physi-
cal possession is the more primitive concept in a developmental sense. Neither
DeLancey nor Lyons pursues empirical acquisition research; but if DeLancey’s
suggestion is true, it would seem reasonable to hypothesize that Possessor ought
to be the more basic role, and “location” just a derivative or metaphorical exten-
sion of that.
Jakaltek examples (1)–(3) above: though they share the same copular root
ay, they require distinct constructions – and something about human cogni-
tion is certainly driving the development of distinct constructions taken as a
whole.
In her focus on what is typologically common across constructions that
predicate possession, existence, and location, Clark’s (1978) exposition
does not particularly highlight how common differences are among these
constructions, even if they might happen to share a particular predicate, a
locative phrase at some point in history, or some other morphosyntactic
element. However, differences are evident from a closer examination of the
grammars of at least some languages on which she originally based her
conclusions. For just one example, Amharic is one of the 30 languages in
her sample. In Amharic, the same root may be used in existential (5) and
predicative possessive (6) clauses, but this hardly means that the construc-
tions are the same: the existential is intransitive, while the possessive is
transitive, as shown by argument marking on the verb (examples are from
Michael Ahland, pers. com.; see also Ahland 2009).
Existential
(5) a. (Be-t’erapeza laj) ms’ haf-ott all-u.
LOC-table top book-PL exist.PRF-3PL.SBJ
‘There are books/some specific books (on the table).’
Existential/Locative (vague)
b. Ms’ haf-ott be-t’erapeza laj all-u.
book-PL LOC-table top exist.PRF-3PL.SBJ
‘The/Some specific books are on the table.’
Possession
(6) L d -itwa ms’ haf-ott all-u-at.
child-DEF.F book-PL exist.PRF-3PL.SBJ-3F.OBJ
‘The girl has books/some specific books.’
Example (5b) is technically vague between predicating location and exis-

tence, depending on particular discourse context for clarification and on
how the predication of location versus the predication of existence is de-
fined – and it is not the case that these can be differentiated just on the basis
of identifiability of participants (contra Clark’s somewhat circular method-
ology). However, according to both Ahland (pers. comm.) and Clark (1978),
the predication of location in Amharic is primarily done via a different root,
not shared with the predication of possession:
116 Doris L. Payne
Location
(7) Ms’haf-ott b-t’rapeza laj n-attw.
book-PL LOC-table top COP.IMPF-3PL
‘The/some specific books are on the table.’
If one were to argue that the “Possessor” in (6) is somehow a Locative, akin
to the Locative in (7), it certainly is not the case that the constructions as a
whole are the same: first, (6) is transitive while (7) is intransitive; second,
the order of Themes and putative Locatives in (6) and (7) are reversed; and
third, the predicates are not the same.8
Clark (1978) goes on to argue that a single root (nw) is shared both for
predicating location (7) – in (7) this is shortened to n – and for what she
calls “Possession 2”, i.e. a construction that involves a genitive NP in a
copular construction akin to The book is John’s. But when we compare the
Locative in (7) with Clark’s Possession-2 structure, the constructions are
still not the same because the Possession-2 structure necessarily requires a
genitive NP, i.e. something like John’s or John’s thing – the possessive
sense does not emerge simply from the copular construction alone.9
In sum, I do not dismiss the clear linguistic evidence that human beings
can and very often do see conceptual connections between predicating loca-
tion of an object, possession of an object, and existence of an object – else
why would a Jakaltek or an Amharic root used at one point in time to
predicate one of these notions be adapted to predicate another of these no-
tions? I do, however, believe it is too reductionist to say that they are sim-
ply the same.
My concern in this paper is to explore usage patterns of lexical and con-
structional forms for expressing location and possession, so as to bring fur-
ther empirical evidence to bear on our understanding of how distinct these
conceptions may be for human cognizers. I conclude that the Maa evidence
points towards location being closely linked conceptually to existence; and
of possession being somewhat closely connected to existence. But the data
examined here present no evidence that possession and location are directly
linked conceptually. In Maa, they do not share any lexical or constructional
8
Ahland (2009) presents arguments that the PR in the Amharic structure in (6)
developed out of a formal locative phrase at some point in history. This does not
destroy the clear fact that speakers perceived the meaning of possession to be
sufficiently distinct from location that they did develop distinct constructions.
9
Clark (1978) does not give a full Amharic clause illustrating what she calls Posses-
sion-2, as she is concerned with reporting results from a large language sample.
forms in common beyond general clause-level constructions that all types

of predicates employ. Thus, it is problematic to say that Possessor is (just)
Locative.
4. Key definitions
Further definitions and elaboration of key terms are necessary in order to

proceed. These definitions are generally worded in terms of speech acts on
the supposition that, in making predications, speakers are primarily per-
forming actions.
Predicative possession. In this paper, I basically assume the reference-
point arguments presented by Langacker (1993). However, as there are mul-
tiple reference-point constructions (in Langacker’s treatment, e.g., posses-
sive constructions, topic-comment constructions, metonymy), possession
involves more than just identifying one participant with reference to another
salient participant. I suggest that the central meaning components of predi-
cative possession are (a) to predicate an intimate relationship between the
primary reference point and another participant; and/or (b) to predicate con-
trol by the reference point participant over another participant. The first
underlies the phenomenon of inalienable possession, and the second under-
lies alienable possession in languages which grammaticalize the difference –
a difference which can show up in predicative as well as NP possession.10, 11
The meaning components in (a) and (b) just above cover a range of
more specific subcases (or subsenses). The senses in (8) through (13) ap-
10
Though there is important conceptual overlap, Kuna (2003) observes that
predicative possession is more precise or constrained semantically, while NP
possession is polysemous.
11
In a corpus study of Guaraní, Velazquez-Castillo (1996: 70–77) compared two
predicative structures used for expressing possession. She found that the primary
usage of a non-verbal predicate construction was to express intimate relation-
ships like ‘I have arms’; she also found that no instances of the construction
were used to designate spatial relations. A second verbal predicate construction
with reko ‘have’ was used to express that a PR is consciously in charge of or in
control of the PM. She argues that the latter structure can sometimes also place
emphasis on an unusual location when used with a body part, but this is not a
consistent meaning to the structure; rather, conscious control is the consistent
meaning. See McGregor (2001a, b) for similar distinctions in Nyulnyul. The
Maa ata Possessive construction described further below apparently covers both
types of meanings.
118 Doris L. Payne
pear to be characteristic (based on a limited Google search for English

have).12 I am not suggesting that the subsenses surveyed below are discrete,
nor that particular examples might not express more than one subsense at a
time. (Also, sometimes examples appear to be ambiguous between a pos-
session sense, and an existential function.)
(8) Predicate control over an entity; ownership

a. I have a set of Will Rogers bookends.
b. The car I have now has a lot of things wrong with it.
c. Do I have the right team? [directed to business managers]
d. I have the job I want.
e. I cursed my fate because I had no shoes … [perhaps also existential]
(9) Predicate potentially legal rights over something (e.g. intellectual

property)
a. If you have intellectual property to protect, but you lack the re-
sources of a major company, then The Patent Process is for you.
b. Who has the most effective way for consumers to reduce the a-
mount of energy used by industry?
c. Many, perhaps most, of the people who read my site have web-
logs of their own.
(10) Predicate a mental or other experience of a participant

a. How can I have the most memorable pan-American journey pos-
sible?
b. I have the feeling about 60 percent of what you say is crap.
c. Should I have the maternal serum triple or quadruple test?
d. I have had on occassion [sic] a few calls stating that the client has
stated that the information I have provided has been incorrect.
e. I have three boys all about to have birthdays.
12
Other scholars have come up with other lists, with differing numbers, of pos-
sessive subsenses (e.g. Langacker 1995; Heine 1997). My point here is just to
sketch some of the typical semantics as a basis for later evaluating the hypothe-
sis that Maa ata codes possession. The English predicate own is semantically
much narrower than have, being mostly restricted to subsenses (8) and (9). I have
excluded aspectual and modal auxiliary uses of have. Though the modal and
aspectual senses are historically connected to the possessive sense, the former
are not central to the core possession meaning.
(11) Predicate a characteristic of a participant

a. Do I have the ability to connect an external antenna to my eTrex?
b. Do I have the right kind of faith?
c. I have the A Negative blood type.
d. I have the fear of being afraid.
f. Do I have the right to receive information about how my child is
doing in school?
g. I have the honour to introduce the combined second, third, fourth
and fifth periodic report on the implementation of the Convention
on the Elimination of All Forms of Discrimination against Women
on behalf of the Malawi Government.
h. The car I have now has a lot of things wrong with it.
(12) Predicate a part-whole relationship. (In their context of use, all the fol-
lowing are likely more concerned with predicating characteristics of
the possessors, than on the part-whole relationship.)
a. I have a herniated disk.
b. … and then I met a man who had no feet.
c. The better half thinks I have awful feet.
d. I have eyes with two different colors.
e. Do I have a big nose?
(13) Predicate kinship (or other close human relationship)

a) I have the most fantastic daughter who ever lived.
b) I have three boys all about to have birthdays.
What holds all the preceding together is the predication of a control rela-
tionship (in (8)–(9)) or particularly intimate relationship (in (10)–(13)) be-
tween two entities. In English predicative possession with have, these two
entities are generally expressed as Subject (PR) and Direct Object (PM).
Possessor. Given the preceding exposition, I define Possessor (for predi-
cative possession) as the argument of a predicate or clause (a) presented as
a salient reference point relative to which the speaker identifies another par-
ticipant, and (b) which predicates it as having an intimate and/or controlling
relationship to another participant.
Agent is the argument of a predicate or clause which the speaker pre-
sents as doing or causing something; it corresponds to x in “What did x
do?” (Chafe 1970; Van Valin and LaPolla 1997).
120 Doris L. Payne
Theme is the argument of a predicate or clause which the speaker pre-

sents as existing (or not), being in a location or state, or changing from one
location or state to another. In this paper, I will not focus on the possible
distinction between PM and Theme, taking the PM to be a kind of subordi-
nate and subsidiary Figure, relative to the Possessor as primary Figure.
Location is the (often diffuse and not particularly-well individuated) ar-
gument of a predicate or clause at which the speaker predicates the Theme
as being. We may note that English have can participate in predicating lo-
cation – but typically this requires a locative preposition, as in How would I
be able to turn on the laptop if I have it under the seat? I consider this ex-
ample to conform to a distinct construction from the have-NP construction
illustrated in all the example sets (8)–(13) above.
Finally, an Existential predication is a predicate or clause which the
speaker primarily uses to assert the existence of an argument.
5. A primer on Maa morphosyntax
Maa (Eastern Nilotic, Nilo-Saharan) is the language of the Maasai, Samburu,

il-Chamus, and possibly other ethnicities in Kenya and Tanzania. There is
dialect variation across the Maa-speaking territory, but it is not clear that
any dialect variation affects the issues explored here. In order to understand
the examples and corpus data which follow, I first introduce basic facts of
Maa grammar.
Maa is a verb-initial, mixed head/dependent-marking language. As a
head-marking language, an inflected verb usually can constitute a complete
sentence. Table 1 presents basic allomorphs of singular person-number pre-
fixes on both Transitive and Intransitive verbs.13
13
See Payne, Hamaya and Jacobs (1994) for further discussion of the Inverse-
Direct nature of the person prefixes, and Rasmussen (2002) for tonal behaviour
of these prefixes.
Table 1. Maa singular person prefixes for verbs
Transitive Intransitive
INVERSE DIRECT
P A 1SG 2SG 3 S
1SG áá- a- a-
2SG k
-
-
-
3 áà- k
- - -
Maa uses tone to mark two case distinctions on nouns and modifiers of
nouns in a marked-nominative typological pattern (König 2006). There are
multiple tone classes for nouns and adjectives, but each lexeme has two
tonal case forms. Postverbal subjects, objects of the oblique preposition t,
and vocatives occur in what Tucker and Mpaayei (1955) called the “Nomi-
native” tone. Other post-verbal NPs, all preverbal NPs regardless of gram-
matical relation, objects of the oblique Associative preposition , nouns and
adjectives in their citation form, and NP-internal PRs occur in their “Accu-
sative” tone pattern.
Though phrases can occur before the verb for specific pragmatic reasons,
the most neutral order is for clauses to begin with a verb. Thus, a basic
transitive or ditransitive clause will have the following structure (numeric
indices simply refer to relative order): [V (NP1) (NP2) (NP3)]. Because NPs
are marked for case in postverbal position, subject and object may vary in
order. In this structure, at most one NP can be Nominative, and this could
be either either NP1 or NP2/3. Invariably, NP1 is more discourse-topical than
NP2/3 (Payne, Hamaya and Jacobs 1994). Text data show that order of nom-
inals does not correspond to identifiability (definiteness) per se. Compare:
(14) -wá nk-àjí nk-ár.

3-take.PF FSG-house.ACC FSG-water.NOM
‘The house has been consumed by fire.’ (lit: “Water took the house”;
an idiomatic expression)
(Mol 1996: 31 and confirmed by additional speakers)
(15) -wá nk-ár n-cán -s

nyáí pk
.
3-take.PF FSG-water.NOM F.PR.PRT F-rain MSG-sand.ACC all.ACC
‘The rainwater has washed off (taken) all the sand.’ (lit: ‘Water of
rain took all the sand.’)
122 Doris L. Payne
Tests for transitivity of a stem include its ability to take an Inverse prefix,
and Antipassive and Middle suffixes (which then derive an intransitive
stem). Thus, the following show that the root du is lexically transitive.
(16) Áà-dû.
3>1SG-cut
‘He/she/ they will cut me.’
(17) Á-dú-ò (nàn ).

1SG-cut-MID.NPF 1SG.NOM
‘I am cut’ (in my flesh).
(18) È-dú- íshó nkráí (*nkn ).

3-cut-ANTIP FSG-child.NOM strap.ACC
‘The child is cutting something’/‘The child is able to cut without us-
ing anything.’
Basic intransitive clauses have the structure [V (NP)]. Evidence that a stem
is intransitive includes its inability to take an Inverse prefix, a Middle, or an
Antipassive suffix. Compare:
(19) K--d ò-sòít.

DSCN-3-be.red MSG-rock.NOM
‘The rock is red.’
(20) *Áà-d.
3>1SG-be.red
(21) *È-du-ísh ór-ó.

3-cut-ANTIP-MID.NPF
Another construction that will figure in our discussion is the Impersonal (or
“Impersonal Passive”). As Greenberg (1959) demonstrated, the Maa Imper-
sonal developed historically out of a third person plural suffix -
. This ap-
parently satisfied the subject requirement of the verb, and hence no free
Nominative NP could (nor can) co-occur in the sentence. Also, an Agent
cannot be expressed in the clause in an oblique phrase. Certainly in the
Maasai variety of Modern Maa, the Impersonal can occur with both transi-
tive and intransitive stems.
(22) *Áà-dù-í t n

nc.
3>1SG-cut-PASS OBL 3PL.NOM
(‘I will be cut by them.’ / They will cut me.’)
(23) Áà-dù-í (nán).

3>1SG-cut-PASS 1SG.ACC
‘I will be cut.’
(24) -d-
.
3-be.red-PASS
‘People [in general, nonspecific] are red’ / ‘Being red happens’.
Whether in transitive or intransitive clauses, NPNOM and NPACC can each

express a variety of semantic roles. In post-verbal position, I know of no
instances where NPACC expresses Agent. But beyond that (and certainly
once the Instrumental and Dative applicatives – each of which has a range
of meaning – are considered), postverbal NPACC can express a wide variety
of roles including Theme/ Patient, Recipient, Benefactive, Instrument,
Causee, Locative, Associative, and Source (cf. also Lamoureaux 2004).
The same is true for NPNOM, even though there is no promotional passive
(there is a promotional Middle). Briefly consider the following; note that
out of context any of the NPs below could be equally well translated with
an English definite or indefinite article:
(25) È-dú nk-áyíóní èn-kíné.

3-cut FSG-boy.NOM FSG-goat.ACC
NOMINATIVE = AGENT ACCUSATIVE = THEME
‘The boy cut the goat.’
(26) E-tí í nk-áyíóní ò-rèyíét .

3-be.at FSG-boy.NOM MSG-river.ACC
NOMINATIVE = THEME ACCUSATIVE = LOCATIVE
‘The boy is at the river.’
(27) -átá nk-áyíóní èn-kíné.

3-have FSG-boy.NOM FST-goat.ACC
NOMINATIVE = POSSESSOR/LOCATIVE ACCUSATIVE = THEME
‘The boy has the goat.’
We will return below to what is the appropriate semantic role of ‘boy’ in (27).
124 Doris L. Payne
With this background, we now turn to the senses of tii and ata, which can-
not really be considered apart from the constructions that the roots occur in.
Thus, I will review senses of each of these roots according to construction.
6. Senses of Maa tii
We will see below that tii can occur in both the simple transitive and in-
transitive syntactic constructions. However, it cannot take the Antipassive
(28a), Middle (28b), or Inverse (29) affixes. These properties argue that it is
not a lexically transitive root.
(28) a. *È-tíi-íshó. b. *È-tíí-ò.

3-be.at-ANTIP 3-be.at-MID.NPF
(29) a. *Áà-tìì. b. *Áà-tìì èn-kínè.

3>1SG-be.at 3>1SG-be.at FSG-goat.NOM
(‘It is at/ near/ on me.’) (‘The goat is near/ at me.’)
6.1. tii in the [V NP1 NP2] construction
In the [V NP1 NP2] construction, tii primarily has a locative sense, possibly
an existential sense, and in some contexts is possibly in the process of
grammaticalization as an imperfective aspect marker.
Tii most commonly predicates location of objects at literal physical
places. In (30), the river is Ground (or Locative) and the boy is Figure (or
Theme). This is a prototypical example of the predication of location. Cer-
tainly neither the river nor the boy have discretionary control over each
other. Additional examples follow, demonstrating a range of locative situa-
tions including metaphorical places. Note how relative order of Nominative
and Accusative phrases can vary.
(30) È-tí í nk-áyíóní l-kj.

3-be.at F.SG-boy.NOM MSG-river.ACC
‘The boy is at the river.’
(31) nk-l n-á-tìí -mányátá

FSG-day.NOM F.REL-FSG.REL.NOM-be.at FSG-settlement.ACC
ìl-mórûâk .
MPL-elders.NOM
‘a day when the elders are in the settlement’ (murrano.0105a)
(32) È-tí í òl-tóò nk-ár à-bòré.

3-be.at MSG-barrel.ACC FSG-water.NOM INF.SG-be.full
‘The barrel is full of water.’ (lit: ‘Water is at the barrel to be full.’)
Tii is strongly dispreferred, if not ungrammatical, if something like 1st or

2nd person is the Locative (33). Additionally, the best kind of Theme with
tii is one that is capable of moving or being moved from one location to
another. Thus, ‘mountain’ is not a preferred Theme with this verb (e.g., as
in something like ‘The mountain is in the forest.’).
(33) *Áà-tí í èn-kínè.

3>1-be.at FSG-goat.NOM
(‘The goat is at/near me.’)
An infrequent use of tii in the [V NP1 NP2] construction is to predicate exis-

tence. Though (34) does have a locative
nà-kp ‘that land’, the primary
purpose of the clause in its original context was probably to posit the exis-
tence of òl-tíkàná ‘East Coast Fever’.
(34) Náà k-é-tí í àpá

nà-kp
and DSCN-3-be.at formerly that.F.ACC-land.ACC
òl-tíkàná.
MSG-East.Coast.Fever.NOM
‘And in that place there is the East Coast Fever.’ (emutata.012)
The following is similar. Here the protagonist has traveled for some distance
to see some cattle, eventually locating the herd. The last line of this excerpt
states that a striped ox (which had consumed multiple people) is located /
exists within this herd:
126 Doris L. Payne
(35) ‘…until he has gone to see bulls far away. There he is, until he has
gone where they are,’
e-tí í knâ kíshú l-k
t
3-be.at these.FPL.ACC cattle.ACC MSG-cow.NOM
ó-írìm-ò …
MSG.REL.NOM-make.spots-MID.NPF
‘in this herd of cows there is a striped ox …’ (enamuke1.0012)
Another shade of meaning for tii in the [V NP1 NP2] construction is the
predication of metaphorical location. This is infrequent in the corpus data –
and it is in this point only, it seems, where the use of tii comes closest to
anything that could be thought of as possession. For example, in (36) one
might opine that the Maasai character “has” the good feature of generosity.
Note that the free translation below, using an English locative (not an English
possessive) was provided by a highly fluent bilingual speaker, perhaps sug-
gesting that he conceptualized it as closer to a locative than to English has.
(36) Náà órè -l-ón náà nàbô bá

FOCUS now FSG-be.generous-NMLZ FOCUS one.F.ACC matter.ACC
sídáí àpá n-à-tí í àtûâ
good.ACC formerly REL.F-FSG.REL.ACC-be.at inside
l-kúààk l- l-máásâ
.
MSG-character.ACC M.PM-PL.PR .ACC MPL-Maasai.PL.ACC
‘And generosity was one of the good things that is within the Maasai
character.’ (elengon2.013)
Finally, an incipient (infrequent) use of tii in the transitive [V NP1 NP2]

construction may involve imperfective aspect. In the following the discon-
tinuous expression “education with the elders” is perhaps a metaphorical
Locative, and involves a nominalization of the verb root
sma ‘read’.
(37) n-k
smá è-tí í ó l-mórùàk.
FSG-education.ACC 3-be.at ASSOC MPL-elders.ACC
‘They are learning with the elder.’
(lit: ‘They are at education with the elders.’) (bulunoto.070b)
If the Accusative referent is completely identifiable and very topical in dis-

course, then it may be possible to omit the Accusative NP (i.e. the reference
to the location or Locative is via a “definite null” zero form). For example,
in one written story, the writer recounted a famous event of when a British
District Commissioner (DC) seized cattle from a warrior and fought over a
particular bull that belonged to the warrior:
(38) a. -shm
l dìsí à-
b
l-móí
3-go.PF that.M.NOM DC.NOM SG.INF-seize MPL-oxen.ACC
l-nyná t èn-gólòn à-lòt-ú
M-3SG.ACC.PR.of.PL.PM OBL FSG-strength.NOM SG.INF-go-VENT
‘that DC went to take his oxen by force’
b. à-m
r t l-mnàndà t Mórìjò
SG.INF-sell OBL MSG-market.place.NOM OBL Morijo.NOM
Lóítà.
Loita.NOM
‘to come to sell them in the market place in Morijo Loita.’
c. N-é-m-é-tí í àpá
l mrràní
nâ
CN-NEG-3-be.at long.ago that.M.NOM warrior.NOM that.F.ACC
kátá
time.ACC
‘And that warrior was not (there) [in the market] at that time.’
(DC 6–8)
In line (c) of the preceding excerpt, there is no pronominal demonstrative or

lexical NP reference to ‘there’ or to ‘the market’; but this location is com-
pletely known from the context. It is unlikely that line (c) of the preceding
has an existential function, as the warrior was introduced two lines prior to
the beginning of this excerpt. Thus, I still treat line (c) as expressing a loca-
tion, and as a transitive construction with a definite null accusative phrase.
6.2. Locative in an oblique phrase
With the root tii, it is rare to express the Locative in an oblique t phrase,
rather than in a simple Accusative NP; indeed, this has been rejected in
elicited material. However, it is attested and three such instances occurred
in the 130-item sample from the corpus (section 8). In at least the following
example, the oblique t with ‘our home’ is conceivably motivated by the
fact that a Comitative oblique occurred before the Locative phrase, making
it more difficult for a comprehender to know how to interpret an otherwise
bare NP so structurally distant from the verb tii.
128 Doris L. Payne
(39) è-yíéún-òtó nk-áí pèê kì-tìí

FSG-want-NMLZ F.PR.PRT FSG-God.ACC so.that SBJV.1PL-be.at
táàtá t nébò
l-mn t -ná
now OBL together.NOM MPL-visitors.ACC OBL-this.FSG.NOM
â á
home.NOM our.F.NOM
‘it is God’s will that we are together with our visitors here in our
home’ (camus1.080)
6.3. tii in the [V NPNOM] construction
We now turn to use of tii in what I believe is a genuinely intransitive con-

struction [V NPNOM]. That is, the absence of an Accusative NP cannot be
interpreted as a “definite null” form. The first sense of tii in this construc-
tion is to posit the existence of the referent coded as the Nominative NP.14
(40) K-é-tí í dòí

n-krâ n-áà-bìk
DSCN-3-be.at indeed FPL-children.NOM REL.F-3FPL.REL.NOM-wait
-m--shm.
until-SBJV-3-go.SBJV
‘There are children that stay (i.e. remain without being shaved) until
they are able to walk.’ (eishoi.024a)
(41) A-é-m-é-tí í è-òkí.

CN-EP-NEG-3-exist FSG-sin.NOM
‘And there is no sin/offence.’ (embul.198)
(42) a. N-é-tìí àpá l tm ó ènkìtòjó,

CN-3-be.at long.ago this.M.NOM elephant.NOM ASSOC FSG-hare.ACC
‘Long time ago there was this elephant and/with a hare,’
b. órè l tm na á n
ny
DSCN this.M.ACC elephant.ACC FOCUS 3SG.ACC
-sàpk t n-kítòjó
MSG.REL.ACC-big.NOM OBL FSG-hare.NOM
‘and/now this elephant was bigger than the hare.’ (elephare.001–002)
14
Contra Gutierrez (2006) and others, existential constructions do not always “re-
quire locative complementation” in any overt way – clearly they do not in Maa.
There is a second closely associated construction which “inherits” the pre-

ceding construction, but is invariably in third person form, carries the dis-
course connective n- at the beginning of the verb form, includes àpá ‘for-
merly, before’, and most typically also has a pause after àpá. Consider the
following example:
(43) a. N-é-tìí àpá, l-mrràní ób ó,

CN-3-be.at long.ago MSG-warrior.NOM one.NOM
‘Long ago, there was a warrior,’
b. órè l mrránì, náà k--p
àpá òlê
DISCN this.M.ACC warrior.ACC FOCUS DSCN-3-brave long.ago very
‘and/now this warrior, he was very brave.’ (arinkoi.001– 002)
The comma in (43a) reflects a pause in the original recording. In contrast,

one native speaker listening to the following recorded story with no pause
after apá commented that it did not sound like the beginning of the story,
but rather as if we were in the middle of a story. He commented that if the
story instead began with [nét] (which is the phonological tone pattern that
could occur just before a pause), it would sound more like the beginning.
(44) N-é-tìí àpá l-páyìàn -yàm-á

CN-3-exist long.ago MSG-elder.NOM MSG.REL.NOM-marry-PF
è-sìànkí kì.
FSG-bride.ACC
‘A long time ago there was an old man who married a young woman.
(divorce 001)
A short time later, the same consultant listened to the following, and said
that this was a good way to begin a story, involving repetition of nétìí. With
tone sandhi and vowel coalesence, the first phrase sounds like [nétìápà].
(45) N-é-tìí àpá, n-é-tìí ná kítòjó

CN-3-exist long.ago CN-3-exist this.FSG.NOM hare.NOM ASSOC
l-j
n.
MSG-hyena.ACC
‘A long time ago, there was this hare and a hyena.’ (enkitojo.001)
In the corpus study reported further below, if the sequence nétìí àpá ap-
peared to have strictly a formulaic story-opening function as in (45), it was
130 Doris L. Payne
counted as such for the study. Otherwise, as in (43) and (44) above, it was
counted as an existential. Clearly, however, these two functions and con-
structions shade into each other historically.
7. Senses of Maa a ta
Unlike tii, ata is completely grammatical with Inverse prefixes and certainly
can take 1st and 2nd person objects. But like tii it does not take Antipassive
or Middle suffixes (46). The fact that it can take Inverse prefixes suggests it
is lexically (more) transitive. One hypothesis as to the reason why the Anti-
passive and Middle suffixes are not acceptable with ata may be because the
Aktionsart is stative, but this needs further investigation.
(46) a. *-át á-

sh b. *-átà-à
3-have-ANTIP 3-have-MID.NPF
7.1. ata in the [V NP1 NP2] construction
In contrast to tii, ata does predicate possession in the [V NP1 NP2] construc-
tion. It is attested with a range of possessive subsenses, as described in Sec-
tion 4 above. In the following examples, the woman is an individuated hu-
man who can exercise control over the Accusative NPs, to the extent of
selling or killing them.
(47) -átà èn-kít ók èn-kíné.

3.have FSG-woman.NOM FSG-goat.ACC
‘The woman has a goat.’
(48) N-é-l ót-ú n-tásât à-nà-

CN-3-go-VENT FSG-old.person.NOM INF.SG-throw.down-VENT
ò-rèyíét -àtá l-k
d
ó
MSG-river.ACC 3-have.while MSG-tail.ACC ASSOC
è-sòsiân.
FSG-cleaning.stick.ACC
‘The woman came down the valley having a fly whisk and a cleaning
stick.’ (emutata.038b)
The construction can express part-whole relationships; in (49) ‘ritual hair’ is

the hair on the child’s head which has been allowed to grow without cutting
or shaving.
(49) N--tn n-kráí -átâ l-màsí …

CN-3-sit FSG-child.NOM SBJV.3-have MSG-ritual.hair.ACC
‘the child will stay with the ritual hair …’ (eishoi.021b)
The construction predicates characteristics of participants:
(50) N--átá -smàsh òlê.

CN-3-have FSG-hunger.ACC very
‘And they were very hungry.’ (divorce 009b)
(51) -átà ná tìtó n-c

pâî.
3-have this.NOM girl.NOM FSG-happiness.ACC
‘This girl is happy.’
(52) … áà-jò-kì k

-àtà dòí -nyàmal í
INF.PL-tell-DAT 1PL-have indeed FSG-problem.ACC
tèné-wúéjî.
in.this-place.NOM
‘… tell that we have a problem in this place.’ (ilomon.0297)
The construction can also be used for predicating the experience of an event
or situation, relative to a participant.
(53) K--átà
l-k
shr l-shr.
DSCN-3-have MPL-Kishuru.NOM MSG-senior.elder.initiation.ACC
‘The Ilkishuru age-set has a senior-elder ceremony.’ (i.e., to make
them senior elders)
(54) Orè òshî -tn m--t a-

DSCN normally SBJV.3-sit NEG-3-have-PL
n-k
páátá k--d ár àké
FSG-circumcision.ceremony.ACC DSCN-3-shout just
l-áyíóní ób ó.
MSG-boy.NOM one.M.NOM
‘Now normally before the initial ceremony/they have the ceremony,
one boy first shouts. (bulunoto.049)
132 Doris L. Payne
The closest example I have encountered of where the predicate ata could be
argued to express location of an object is the following elicited clause,
where the barrel could not (in most worlds) be said to have control over the
water; though ‘having water’ could certainly be a way of predicating a par-
ticular characteristic of a barrel. However, out of any particular discourse
context it is hard to say whether the speaker’s communicative intent would
be predication of a characteristic or of a location.
(55) -átà òl-tóô nk-ár.

3-have MSG-barrel.NOM FSG-water.ACC
‘The barrel has (contains) water.’
Finally, marginal uses of ata may include indicating something like value
or need, i.e. the possible beginning of modal sense. The data are so sparse
on this point that I have nothing further to say about it here.
As with tii, the Accusative argument of ata can be expressed with a
definite null when it is clearly known from the context. In the following,
the understood Accusative is ‘the story about the strap to heaven’, which
was mentioned two clauses previously. The person posing the question
comes from a different region of the Maa-speaking territory than the ad-
dressee, and is asking whether people in the il-Chamus region also ‘have’
this story:
(56) M-à-yíólò tnáà

-átá-tà òshî
n t á
.
NEG-1SG-know if 2-have-PL normally you.PL.NOM
‘I do not know if you have it yourselves.’ (camus1.030a)
For an example like (57), both arguments would be known from context. I
consider all such definite-null examples to also be instances of a transitive
construction.
(57) K-áà-àtà.
DSCN-3>1SG-have
‘He owns me.’
7.2. ata in the [V-

NPACC] Impersonal construction
The Impersonal [V-

NPACC] construction with ata predicates existence of an
Accusative argument.
(58) K--át á-

òshî ìn-tòkìtín
DSCN-3-have-PASS always FPL-things.ACC
n-àá-j-î nk-áshê nk-án y
t …
REL.F-FPL.REL.ACC-say-PASS FSG-calf.ACC FSG-respect.ACC
‘There is something called the heifer of manners …’ (enkashe.001a)
(59) N--àtá-
aké n
ny òl-òtùnó.
CN-3-exist-PASS still emphatic MSG-leader.of.age.set.ACC
‘There is still the leader of the age-set.’ [called Olotuno, the-one-
who-planted]. (aibartisho.004b)
(60) a m k-é-yíólò áà-jò k--a t á-

nk-áí …
because DSCN-3-know INF.PL-say DSCN-3-have-PASS FSG-God.ACC
‘because they know there is God …’ (enkai.004)
7.3. ata in the intransitive [V NPACC] Existential construction
The corpus study reveals a construction which is unusual for Maa, and to
my knowledge unique to the verb root ata. This is an intransitive clause
with a single Accusative argument, but without the Impersonal suffix -
(which otherwise is the only known way of creating an intransitive clause

with a single Accusative NP). This construction exclusively has the sense
of predicating existence. Impressionistically, it is common to use this con-
struction when the existence of something is negated; but corpus data show
that it is not at all exclusive to negative existence.
(61) àm m--tà l-tánì ò-yíólò à-jó

because NEG-3-exist MSG-person.ACC MSG.REL.ACC-know COMP
è-tù-búl-ú-á dúóó,
3-PF-grow-VENT-PF RELEVANT.INFORMATION
‘because there is no person who knows when he grows …’
(embul.005–006)
134 Doris L. Payne
(62) àm k--ák m--tà [l-tánì

because DSCN-3-become NEG-3-exist MSG-person.ACC
ò-ìtókì á
kàtà à-ló nk-á
M.REL.SG.ACC -repeat other.time INF.SG-go FSG-home.ACC
ny.]
3SG.POSS.ACC
‘because there will not be anybody who will go to (his) home again.’
(elengon2.009)
This construction is unusual because it is the one and only morphologically

underived SO construction found in Maa (i.e. where the single argument of
an intransitive construction is coded like the object of a transitive construc-
tion).
8. Corpus results
The table incorporated into Figure 1 reports raw numbers based on a sample
of 130 tokens of ata and 130 tokens of tii, extracted from a larger text corpus
of approximately 20,000 clauses (containing mostly spoken but also some
written material, and a variety of genres). The original methodological intent
was to simply consider the absolute first 130 instances of each of these
roots; but due to some false starts or other lack of clarity among the first
130 instances, some of those examples were discarded and subsequent in-
stances were considered until a data set of 130 instances of each was com-
piled. I will henceforth refer to these 260 clauses as “the corpus”.
In the corpus, 47% of tii instances expressed existence, 47% expressed
location, and 6% expressed other senses including stative aspect; see Figure
1. There is no evidence whatsoever from this corpus that tii predicates pos-
session.
In the corpus, 65% of ata instances expressed existence, 34% expressed
possession, and 1% arguably expressed modal senses; see Figure 1. There is
no evidence from the corpus study that ata is used for predicating location.15
There is also essentially no evidence from elicitation that ata predicates loca-
15
tion, but recall the discussion of (55) above.

Figure 1. Corpus distribution of ata and tii by sense (N = 130 for each root)
Figure 2 shows the constructions used for predicating location. The vast ma-
jority are the [V NP1 NP2] transitive construction with tii. The few instances
of the [V NPNOM] construction have definite null locations.
tii-TRANSITIVE tii-NOM tii-OBL
Figure 2. Corpus distribution of predicative location constructions
Figure 3 shows that the dominant construction for predicating possession is

ata in a transitive clause (ata -Transitive). Depending on discourse context,
this could be a [V NP1 NP2] construction where both Theme and PR are ex-
pressed by overt NPs, with PR in the Nominative and Theme in the Accusa-
136 Doris L. Payne
ata-TRANSITIVE ata-NOM ata-
Figure 3. Corpus distribution of predicative possession constructions
tii-TRANSITIVE tii-NOM ata- ata-ACC
Figure 4. Corpus distribution of Existential constructions
tive case; or it could be that the Nominative PR was discourse-identifiable

and sufficiently topical that it was mentioned only via a bound pronominal
element, without additional use of an NP; in this case, it was counted to-
gether with the ata-Transitive category. Similarly, if the Accusative referent
was referred to with a definite null because it is so clearly identified in dis-
course, it was counted in the ata-Transitive category.
More rarely an ata clause with overt Nominative NP but no overt Accu-
sative NP occurs with a possession meaning, as in (56) above. For the sake
of consistency across surface constructional patterns with the two different

roots (tii vs. ata), these were counted separately from the ata -Transitive con-
structions. It is my clear understanding, however, that clauses like the (56)
have “definite null” accusative arguments and are also transitive.
Only twice did the Impersonal/Plural -
suffix occur with ata to predicate
possession, in contiguous clauses:
(63) N--ítàyù dúóó n-kráí tànáà k--a t á-

tànáà
CN-3-give indeed FSG-child.NOM if DSCN-3-have-PL if
k--át á-
dúóó nnâ krà n-é-ítàyì-tuô …
DSCN-3-have-PL indeed those children.ACC CN-3-give-PL.PF
‘The child gives, if they have any, if they have those children they
have already given …’ (ilomon 0258)
Here the -
suffix might be interpreted as referring to a plural possessive
subject, but the sense is almost existential. In sum, my conclusion is that
there is really only one predicative construction for indicating possession:
the ata-Transitive clause construction.
Where the two roots tii and ata do overlap is for predicating existence.
Figure 4 shows that the intransitive [tii NPNOM] (with no definite null locative
referent), and the intransitive Impersonal [ata -
NPACC] are the constructions
of choice for predicating existence. Though the intransitive [ata NPACC] con-
struction without the Impersonal suffix -
is robust, it is clearly less frequent.
Strikingly, for predicating existence the use of a surface transitive-like
construction with overt expression of a location (even as a placeholder of
sorts) is almost non-existent in the Maa corpus. This is notable, given the
clear attestation in so many other languages of the use of a locative phrase
in existential constructions – even if it is a semantic dummy, as in There is
an idea that democracy is the preferred political system, where there does
not refer to any location whatsoever. In Maa the near absence of any such
phrase is possibly because an abstract notion of Locative is lexicalized into
the root tii; so this language may ultimately not be particularly exceptional.
9. Conclusions
Under the possession-is-location view, ata and tii have to be analyzed as

having mirror image patterns of how semantic roles map to grammatical
cases. With two-participant ata, Locative would be Nominative and Theme
would be Accusative; while with two-participant tii, Theme is Nominative
and Locative is Accusative.
138 Doris L. Payne
The corpus study, however, argues that ata and tii are in complementary
distribution for predicating possession versus location – unlike English have
or the French locative copular construction. The corpus data are supported
by the general study of senses-by-construction which also included elicited
material. If “Possessor” were always just a metaphorical extension of Loca-
tive, the non-overlap of these two roots for predicating location and posses-
sion would be surprising. In particular, if all “possession” is cognitively the
predication of (e.g., animate) Locative (Clark 1978: 89), then why would ata
have zero attestation for spatial locational uses? And if all “possession” is
cognitively the predication of location, then why would tii have zero at-
testation for possessive uses? The wheels of grammatical and lexical
change are by all accounts slow, so finding doubling of a form across two
functions (or senses) is normal and expected in the process of change. The
lack of doubling – either for the roots or for constructions – suggests the
lack of cognitive “sameness” or identity about possession and location.
The apparent non-overlap between these two roots suggests that Posses-
sor is indeed a salient relation that Maa speakers distinguish from Locative.
Even where there are conceptual linkages that motivate sharing certain
lexical roots or pieces of the morphosyntax (in constructional terms, “in-
heritance” of subconstructions; cf. Heine 1997), distinct constructions do
emerge via the accretion and stablization of extra morphemes, and become
grammaticalized. The concept of possession (and hence Possessor) is es-
sential in explaining the specialization of both dedicated linguistic con-
structions and senses of forms.
Finally, what merits closer cross-linguistic and historical investigation is
whether human beings in general do not find closer conceptual links be-
tween possession and existence, and between location and existence, than
between possession and location. Clark (1978: 113) herself noted that loca-
tion and existence more often share the same verb root, than does either
existence with possession, or location with possession. With reference to
Maa, it could be that further usage-based research might show overlap in
sense between tii and ata. However, since the [tii NPNOM] existential con-
struction is intransitive, I would not expect it to directly develop into a pos-
sessive construction since possession necessarily involves predicating a
relationship between two entities. Similarly, given that both the [ata-
NPACC] and [ata NPACC] existential constructions are effectively one-argu-

ment predicates, I would not expect them to develop into new locationals
unless they are incorporated into larger constructions which somehow add a
further argument.
Acknowledgements
I am grateful to Leonard Ole-Kotikash, Keswe Ole-Mapena, Sarah Tukuoo,

and multiple other Maa speakers for work with Maa texts, on which this
study is based. The data in this study come primarily from Maasai and il-
Chamus varieties of Maa. I am also grateful to Tom Payne and Michael
Ahland for discussion of issues in this paper, though they are not responsible
for any of my shortcomings in understanding or interpretation. Data collec-
tion for the current study was partially supported by NSF grant Grant SBR-
9616482, and under Kenya research permit #OP/13/001/23C28.
References
Ahland, Michael
2009 From topic to subject: grammatical change in the Amharic possessive
construction. Studies in Language 33 (3): 685–717.
Baron, Irene and Michael Herslund
2001 Semantics of the verb HAVE. In Dimensions of Possession, Irene
Baron, Michael Herslund and Finn Sørensen (eds.), 85–98. Amster-
Clark, Eve V.
1978 Locationals: existential, locative and possessive constructions. In
Universals of Human Language, Vol. 4: Syntax, Joseph H. Greenberg
(ed.), 85–126. Stanford: Stanford University Press.
Chafe, Wallace
1970 Meaning and the Structure of Language. Chicago: University of Chi-
cago Press.
Comrie, Bernard and Helma van den Berg
2006 Experiencer constructions in Daghestanian languages. In Semantic
Role Universals and Argument Linking: Theoretical, Typological, and
Psycholinguistic Perspectives, Ina Bornkessel, Matthias Schlesewsky,
Bernard Comrie and Angela Friederici (eds.), 127–154. Berlin /New
Craig, Colette Grinevald
1977 Jacaltec: The Structure of Jacaltec. Austin: University of Texas Press.
DeLancey, Scott
2002 The universal basis of case. Logos and Language 1 (2): 1–15.
Freeze, Ray
2001 Existential constructions. In Language Typology and Language Uni-
versals: An International Handbook, Martin Haspelmath (ed.), 941–
953. Berlin /New York: Mouton de Gruyter.
140 Doris L. Payne
Goldberg, Adele
1995 Argument Structure Constructions. Chicago: University of Chicago
Press.
Greenberg, Joseph
1959 The origin of the Masai passive. Africa 29: 171–176.
Gruber, Jeffrey
1965 Studies in lexical relations. Ph.D. dissertation. MIT. [Reprinted 1976,
as part of Lexical Structures in Syntax and Semantics. Amsterdam:
North Holland.]
Gutierrez Morales, Salome
2006 Morphosyntactic expressions of possession and existence in Sinhala.
In Proceedings from the Workshop on Sinhala Linguistics (Santa
Barbara Papers in Linguistics 17), Robert Englebretson and Carol
Genetti (eds.), 20 –28.
Heine, Bernd
1997 Possession: Cognitive Sources, Forces, and Grammaticalization.
Cambridge Studies in Linguistics. Cambridge: Cambridge University
Press.
Heine, Bernd
2001 Ways of explaining possession. In Dimensions of Possession, Irene
Baron, Michael Herslund and Finn Sørensen (eds.), 311–328. Amster-
Jackendoff, Ray
1983 Semantics and Cognition. Cambridge, MA: MIT Press.
Kemmer, Suzanne
2002 Human cognition and the elaboration of events: some universal con-
ceptual categories. In The New Psychology of Language: Cognitive
and Functional Approaches to Language Structures, Michael
Tomasello (ed.), 89 –118. Mahwah, NJ: Lawrence Earlbaum.
König, Christa
2006 Marked nominative in Africa. Studies in Language 30: 655 –732.
Kuna, Branko
2003 Izmedu atributne i predikatne posvojnosti. (Between attributive and
predicative possession.) Rasprave Instituta za Hrvatski Jezik i Jezi-
koslovlje 29: 157–171.
Lamoureaux, Siri van dorn
2004 Applicative constructions in Maasai. MA thesis, University of Oregon.
Langacker, Ronald
1993 Reference-point constructions. Cognitive Linguistics 4: 1–38.
Langacker, Ronald
nitive Construal of the World, John R. Taylor and E. MacLaury (eds.),
51–79. Berlin /New York: Mouton de Gruyter.
Lyons, John
1967 A note on possessive, existential, and locative sentences. Foundations
of Language 3: 390 –396.
Lyons, John
1968 Introduction to Theoretical Linguistics. Cambridge: Cambridge Uni-
versity Press.
2001a Non-verbal predicative possession in Nyulnyulan languages. In
Forty Years On: Ken Hale and Australian Languages, Jane Simpson,
David Nash, Mary Laughren, Peter Austin and Barry Alpher (eds.),
337–352. Canberra: Pacific Linguistics.
2001b The verb HAVE in Nyulnyulan languages. In Dimensions of Posses-
sion, ed. by Irène Baron and Michael Herslund, 67–84. Amsterdam:
John Benjamins.
Mol, Fr. Frans
1996 Maasai Language and Culture Dictionary. Limuru, Kenya: Kolbe
Press / Maasai Centre Lemek.
Payne, Doris L.
1997 The Maasai external possessor construction. In Essays on Language
Function and Language Type, Joan Bybee, John Haiman and Sandra
Thompson (eds.), 395–422. Amsterdam: John Benjamins.
Payne, Doris L. and Immanuel Barshi (eds.)
1999 External Possession. Amsterdam: John Benjamins.
Payne, Doris L., Mitsuyo Hamaya and Peter Jacobs
1994 Active, passive, and inverse in Maasai. In Voice in Discourse, T. Givón
(ed.), 283–315. Amsterdam: John Benjamins.
Rasmussen, Kent
2002 Tone in il-Keekonyokie Maa. MA Thesis, University of Oregon.
Sørensen, Finn
2001 Possession spaces in Danish. In Dimensions of Possession, Irene
Baron, Michael Herslund and Finn Sørensen (ed.), 57– 66. Amster-
Tham, Shiao Wei
2005 Representing possessive predication: semantic dimensions and
pragmatic bases. Ph.D. dissertation, Stanford University.
Tucker, Archibald N. and John T. Mpaayei
1955 Maasai Grammar with Vocabulary. London: Longmans, Green & Co.
Van Valin, R. and R. LaPolla
1997 Syntax: Structure, Meaning and Function. Cambridge: Cambridge
University Press.
Velazquez-Castillo, Maura
1996 The Grammar of Possession: Inalienability, Incorporation and Pos-
sessor Ascension in Guarani. Amsterdam: John Benjamins.
142 Doris L. Payne
Wertheimer, Max
1923 Untersuchungen zur Lehre von der Gestalt II. Psycologische For-
schung 4: 301–350. [Translation published in W. Ellis. 1938. A Source
Book of Gestalt Psychology, 71–88. London: Routledge and Kegan
Paul.]
Learning to encode possession
Sonja Eisenbeiß, Ayumi Matsuo and Ingrid

Sonnenstuhl
1. Introduction
Learning and talking about their own possessions and the possessions of
their peers and caretakers plays a central role in children’s daily life. It is un-
surprising then that relationships between possessors and their possessions
are amongst the first relationships that children encode when they start to
string words together (see e.g. Brown 1973); and it is no wonder that many
psycholinguists have made use of this rich data source to address questions
about the mechanisms that drive children’s linguistic development.
However, most of the available studies of the acquisition of possessive
constructions that we will discuss have investigated only one or two posses-
sion-encoding constructions in an individual language. Moreover, the focus
has typically not been on the encoding of the possessive relation itself, but
on other aspects of the respective possessive construction. For instance,
possessive -s markers in German and English (e.g. Susi-s Huhn ‘Sue’s
chicken’) were analysed in studies that investigated whether the syntactic
categories of the target language were already present in early child gram-
mars (e.g. Eisenbeiß 2000; Marinis 2002, 2003; Radford 1990). In these
studies, possessive markers were simply treated as morpho-syntactic reali-
sations of syntactic categories; and semantic aspects were largely ignored.
Similarly, possessive constructions with two-place verbs like have and be-
long were investigated in studies of the acquisition of syntax-semantic
mappings, but these constructions were just treated as one type of two-
argument construction and not compared to other constructions encoding
possession (see e.g. Bowerman 1985; Pinker 1984). To our knowledge, no
study has yet provided a comprehensive cross-linguistic overview that fo-
cuses on the different ways in which possessive relationships are encoded
linguistically.
In order to fill this gap, we will provide a cross-linguistic overview of
studies of children’s acquisition of the constructions that their target lan-
guage employs to encode possession. In addition, we will present new data
from German child language and child-directed speech, and discuss the im-
144 Sonja Eisenbeiß, Ayumi Matsuo and Ingrid Sonnenstuhl
plications for theoretical linguistics and language acquisition research. Our

focus will be on three ways of encoding relationships between PRs and
possessed entities (see Heine 1997; Baron, Herslund and Sørensen 2001 for
overviews):
– adnominal possession: Both Possessor (PR) and Possessum (PM) are

encoded within the same noun phrase (e.g. my/daddy’s chickens, the
chickens of our neighbours, … );
– predicative possession: The possessive relationship is encoded by a two-
place predicate such as have, own or belong or by be (e.g. I have a dog.
The dog belongs to me. This dog is mine);
– “external possession”: the PR and the PM are realised as arguments of
a verb whose lexical meaning does not involve the notion of possession
(e.g. I tapped him-PR on the shoulder-PM).
We will first show how studies of children’s possession constructions can

help us to evaluate models of children’s linguistic development. Against
this background, we will present studies of the acquisition of adnominal,
predicative and external possession constructions (EPCs). For each of these
construction types, we will provide a brief overview of possession con-
structions in adult German and contrast them with possessive constructions
in other languages for which acquisition studies are available. This will al-
low us to discuss empirical findings from earlier studies and our own
analysis of German child data. Finally, we will compare the development of
the three types of possession constructions and discuss the implications of
our findings for theoretical linguistics and models of children’s linguistic
development. In particular, we will show how the available empirical find-
ings about the acquisition of possession constructions can be captured in
approaches that try to integrate core insights from current generative and
usage-based approaches.
2. Theoretical issues in acquisition research
While concepts of ownership and possession seem to be part of all cultures,

there is considerable variation with respect to (i) the legal norms for estab-
lishing, maintaining and negotiating ownership and (ii) the linguistic means
to encode ownership and other types of possessive relationships. Hence,
when they learn to talk about ownership and possession children have to
acquire both cultural and linguistic knowledge. In this study, we will focus
Learning to encode possession 145
on linguistic aspects, i.e., we will investigate how children acquire the pos-
sessive constructions of their target languages – and what this can tell us
about the mechanisms that drive children’s linguistic development.
Current research on the mechanisms underlying children’s language ac-
quisition is characterised by an opposition between generative approaches
(see Eisenbeiß 2009 for overview) and functionalist or usage-based ap-
proaches (Bates and MacWhinney 1987; Tomasello 2003, 2006; Goldberg
2006). Moreover, acquisition researchers have provided competing ac-
counts for the time course of linguistic development and for the orders in
which children acquire the properties of their target language. In the fol-
lowing, we will first provide an overview of these debates before we pre-
sent data on the acquisition of possession constructions that can help us
evaluate competing models.
2.1. The logical problem of language acquisition
The opposition between generative and usage-based accounts of language

acquisition arose from discussions about the so-called “logical problem of
language acquisition” (Baker 1979; Bertolo 2001; Pinker 1989, and the spe-
cial issue of The Linguistic Review 19, 1–2). Children only hear a limited
sample of their target language, and hence have to generalise over individual
input utterances in order to comprehend and produce new utterances. If
children’s hypothesis space for these generalisations were completely un-
constrained, children might make incorrect generalizations and one would
have to explain how they would ultimately reject them. For example, Ger-
man children frequently hear the possessive marker -s on nouns that encode
PRs (e.g. Susi-s Huhn ‘Susi’s chicken’, Oma-s Haus ‘granny’s house’).
This might lead them to use this marker with any PR noun or noun phrase.
However, this would not be the appropriate generalisation for German: -s
can only be affixed to proper names like Susi and a few kinship terms such
as Mama ‘mommy’ that can function as proper names (see Harbert 2007:
161ff. and the discussion below). Moreover, even with this restricted set of
nouns, -s cannot appear when the noun is modified, e.g. by a determiner or
possessive pronoun (*meine Mamas Huhn ‘my mommy’s chicken’). As we
will show below German children do not always restrict the use of -s to
unmodified proper names and kinship terms, but overgeneralise it to un-
modified count nouns such as Affe ‘monkey’, which cannot be combined
with -s in the target language (Mills 1985; Eisenbeiß 2000). Thus, one has
to explain why children produce such non-target-like combinations, how
they overcome these “errors”, and how they learn to use morpho-syntactic
forms appropriately.
One could argue that children “unlearn” errors on the basis of negative
evidence, i.e. information about the ungrammaticality of their utterances.
However, many studies have shown that explicit corrections are not system-
atically available to all children at all developmental stages (Marcus 1993).
Moreover, even for explicit corrections such as You can’t say that, it is not
obvious whether the correction refers to the phonological or morpho-
syntactic structure, the use of lexical elements, or the appropriateness of the
utterance in the social context. In addition, children do not always take up
corrections – and even if they do seem to take them up, they might later go
back to their non-target like structures; see Marcus (1993) for an overview
and the following example from Simone Miller (2;4,1 see Miller 1976; Eisen-
beiß 2003: 45):2
(1) Father: Wem gehört der Löffel?

whom-DAT belongs the-NOM spoon?
‘To whom does the spoon belong?’
Simone: Ich. (correct: Mir)
I-NOM (correct: me-DAT )
‘I’ (correct: ‘to me’)
Father: Wem gehört der Löffel?
Simone: Ich. Ja.
I-NOM. Yes
1
Age information is provided in the following format: Year; Month.
2
We have used the following abbreviations and glosses: [#] – pause; [ /] – inter-
ruption; ACC – accusative; DAT – dative; D-elements – case /gender /number-
marked articles, possessive pronouns, demonstratives, quantifiers; EPC – external
possession construction (I hit him on the head); FEM – feminine; GEN – genitive;
IPC – internal possession construction (I hit his head); MASC – masculine;
NEUT – neuter; NOM – nominative; PART – particle (note that as it is difficult
to provide exact translations for German focus and other particles, particles will
simply be glossed as PART, without further information); PM – Possesssum; PR –
Possessor; SG – singular; and TAG – tag question.

Simone: Ich.
I-NOM
Father: Mir. Wem gehört der Löffel?
me-DAT . Whom-DAT belongs the-NOM spoon?
‘To me. To whom does the spoon belong?’
Simone: Mir.
me-DAT
‘To me’
Simone: Mir.
me-DAT
‘To me’
Father: Mir. Und das bist Du. ne?
me-DAT . And that is you. TAG?
Simone: Ja. Gehört mir. [...]
yes. belongs me-DAT
Simone: Ich.
I-NOM.
Other types of negative evidence do not seem to provide sufficient substi-

tutes for explicit corrections: indirect negative evidence, i.e. the lack of a
particular form or construction in the input, might be informative, but taken
on its own, it is not reliable enough. There are many types of structures that
children are never exposed to (e.g. sentences with long extractions), but are
still not considered ungrammatical by adults. This suggests that children do
not simply stop adding possessive -s to common nouns like Affe ‘monkey’
because they never hear possessive constructions like that. Some acquisi-
tion researchers have argued that certain types of parental responses – e.g.
repetitions with reformulations – occur more frequently after errors than

after correct utterances (see Marcus 1993; Cowie 1999 for references and
discussion). However, even such “noisy” negative evidence is not provided
systematically enough and it is not clear how exactly children could make
use of it (Marcus 1993). Hence, acquisition researchers have to explain how
children can generalise beyond individual input utterances, but recover
from non-target-like generalisations even though they cannot rely on ex-
plicit, indirect or noisy negative evidence.
Faced with this “logical problem of language acquisition”, generative
linguists have postulated that children’s language acquisition is guided by
innate well-formedness constraints that apply to all grammatical structures
of human languages. These innate constraints are assumed to restrict chil-
dren’s hypothesis space so that they can only make target-like generaliza-
tions or generalizations that can be rejected without explicit correction. In
the early days of generative grammar, the universal well-formedness con-
straints were formulated as domain-specific principles; i.e. they were con-
sidered to be specifically targeted to the domain of language (Chomsky
1965, 1981). Thus, the child’s language acquisition mechanism was viewed
as a domain-specific mechanism that cannot be derived from other cognitive
or social skills.
Recently, generative linguists in the minimalist framework (e.g. Chomsky
1995, 2001) have started to derive the domain-specific principles of early
generative grammar from more general cognitive principles, for instance,
economy principles according to which grammatical operations (e.g. move-
ment) are only allowed if they are required to fulfill other wellformedness
constraints (see Chomsky 1995, 2001; Eisenbeiß 2003, 2009 for overviews).
This shows some convergence with the functionalist or usage-based ap-
proaches to language acquisition, which assume that general cognitive and
socio-pragmatic principles suffice to constrain children’s hypothesis space
because children’s input provides rich, structured information about the tar-
get language (see e.g. Bates and MacWhinney 1987; Tomasello 2003, 2006).
In particular, most generative and usage-based approaches to human
languages and their acquisition now assume some version of the Specificity
Principle, a general principle that requires operations with more specific
outputs or inputs to have precedence over operations with less specific out-
puts or inputs (see Eisenbeiß 2003, 2009 for discussion). This principle has
been formulated as a domain-specific Elsewhere Principle (Kiparsky 1982)
or Blocking Mechanism (Marcus et al. 1992) and as a pragmatically based
general Principle of Contrast (Clark 1987); and the function of the Specific-
ity Principle has been investigated in connectionist simulations (Corina
1994). Whatever its precise formulation, such a general Specificity Principle

can, for instance, explain why an adult who has acquired the specific irregu-
lar past tense form went would not use overgeneralised forms like *go-ed:
the presence of the specific irregular form would block or pre-empt the ap-
plication of the general past tense -ed-affixation.
Moreover, acquisition researchers have suggested that a Specificity
Principle could help children overcome morphological overgeneralisations
such as the overgeneralisation of the plural -s to irregularly inflected nouns
(e.g. *mouses instead of mice; see e.g. Bates and MacWhinney 1987; Braine
and Brooks 1995; Clark 1987; Eisenbeiß 2003, 2009; Marcus et al., 1992;
Tomasello 2003, 2006). For such morphological overgeneralisations, where
a specific form is competing with an overgeneralised form, any use of the
appropriate specific adult form in the child’s input could provide a competi-
tor for the child’s non-target-like general form – and thus lead to the child
abandoning the non-target-like overgeneralised form.
Recognising adult morphological forms and their contrast to the non-
target-like forms is supported by several characteristic properties of chil-
dren’s input. In particular, cross-linguistic and cross-cultural studies have
shown that child-directed speech is rich in so-called “variation sets”, i.e.
sequences of utterances with a constant communicative intention, but
minimal variations in word order, lexical choice or inflectional form (Eis-
enbeiß 2003; Küntay and Slobin 1996, 2002; Slobin et al. submitted). For
instance, the variation set in (2a) contains both the singular and the plural
form of mouse, which highlights these forms and the paradigmatic contrast
between them. In this way, variation sets with appropriate adult forms
might make competitors for the child’s non-target-like forms more salient
and support their acquisition.
Other properties of child-directed speech that highlight contrasts be-
tween adult and non-target-like child forms are reformulations such as (2b),
where the adult form is directly contrasted with the child’s incorrect form.
Such reformulations seem to be frequent in Western cultures, but it is not
clear whether they are common across the world (see Chouinard and Clark
2003 for discussion). Nevertheless, where they occur, reformulations might
offer additional support for children’s acquisition of morphological forms.
(2) a. Look there are so many mice in this picture: There is one mouse
under the table and one mouse under the chair and two mice un-
der the bed and three mice in the corner.
b. Child: There are the two mouses again!
Adult: Yes, we have seen these two mice before.
Note, however, that reformulations do not necessarily indicate that the

child’s utterance is incorrect. For instance, if children use my to refer to
their own possession and the parent replies using your, this should not be
taken as a correction of the child’s use of my, which was appropriate, but
spoken from a different perspective than the mother’s utterance.
Thus, in sum, any use of the appropriate adult form that occurs in the
child’s input could provide a competitor for a child’s overgeneralised in-
flected form. In addition, variation sets or reformulations might highlight
these competitors and their contrast to the non-target-like form and thus
provide further support. Appendix A provides a transcript from a German
family that shows how a short everyday conversation can provide children
with many models and contrasts for adnominal possessive constructions
(A1, A2, A10, A11, A12, A14, A16), predicative possessive constructions
(A18, A20, A21) and EPCs (A22, A23). However, more work on the role of
contrasts in the child’s input is needed.
Note, however, that the mechanisms that might help children to drop
overgeneralisations such as *mouses from their language cannot help them
to overcome overgeneralisations of German possessive -s to common nouns
such as Affe ‘monkey’: in contrast to the domain of morphology, the domain
of syntax allows more variation. For instance, in adult German, there are
several alternative adnominal possessive constructions: the possessive -s
construction (Mama-s Huhn ‘mommy’s chicken’), the prepositional con-
struction (das Huhn von der Mama [the chicken of the mommy] ‘mommy’s
chicken’) and the rather formal and much rarer genitive construction (das
Huhn der/meinerGEN Mama [the chicken the/my-GEN mommy] ‘the/my
mommy’s chicken’). Thus, when children hear prepositional possessive con-
structions like das Huhn von der Mama, they should not stop using Mamas
Huhn ‘Mommy’s chicken’ as this -s possessive is actually correct in adult
German. Consequently, hearing die Banane von dem Affen ‘the banana of
the monkey’ should not directly drive out an error like *Affes Banane
‘monkey’s banana’. Thus, we will have to determine how children can learn
how to use alternative possessive constructions and handle the constraints
for their use in the target language.
At the same time, we will have to explain why some errors appear in
children’s data, while other potential errors have not been observed. For
instance, we will show that though German children overgeneralise -s to
unmodified common nouns that are incompatible with this marker, children
do not seem to overgeneralise -s to full phrasal PRs (e.g. *meine Mamas
Huhn ‘my mother’s chicken’). This requires an explanation as -s overgener-
alisations are documented in German child language and phrasal possessive
markers are not uncommon in natural languages – what is more, they occur
in English, which has adnominal possessive constructions that are otherwise
very similar to the corresponding German constructions.
Thus, when we discuss the acquisition of possessive constructions in the
following sections, we will address the logical problem and explain how
children manage to avoid or overcome deviations from the target language
without recourse to reliable negative evidence. In addition, we also have to
capture the time-course of children’s linguistic development, i.e. explain
why children acquire particular properties of their target language at a par-
ticular time and in a particular order.
2.2. The time-course of linguistic development
Most acquisition researchers today, whether they subscribe to a generative

or a functionalist view, agree that the mechanisms that drive language ac-
quisition do not change qualitatively over time (see e.g. Pinker 1984; Eisen-
beiß 2009; Tomasello 2003 for overviews and discussion). Moreover, there
seems to be an emerging consensus that children acquire the most basic
generalisations of their target language very early in their linguistic develop-
ment (Guasti 2001; Tomasello 2006). This has for instance been shown for
the basic word order patterns or the ways in which motion events are en-
coded (see e.g. Slobin et al. submitted). The early adaptation to the target
language can be captured in generative and in functionalist approaches
alike, though the explanations differ: generative authors highlight the role
of innate constraints (e.g. Crain 1991), others point out children’s powerful
predispositions for pattern detection and analogy formation (e.g. Tomasello
2006), while others refer to “helpful” input properties such as variation sets
(Slobin et al. submitted). Thus, if our investigations should show early adap-
tations to the target language for the domain of possession constructions,
this would be compatible with a broad range of theoretical approaches to
language acquisition and provide further evidence for them.
In addition, investigating children’s “errors” in the acquisition of pos-
session constructions can contribute to the ongoing debate about children’s
grammatical representations at different stages of development. Proponents
of generative Full-Competence approaches claim that children’s morpho-
syntactic representations are adult-like as soon as they start to combine
words (e.g. Hyams 1996; Lust 1994; Rizzi 1993/1994, 2000; Wexler 1998).
Thus, they cannot view omissions of possessive markers or other deviations
from the target language as results of non-adult grammatical representations.
Rather, any deviations from the target must be attributed to other factors;
and Full-Competence proponents have suggested a number of such factors,
e.g. the late maturation of general cognitive constraints, problems with the
morphological or phonological realization of unstressed morphemes and
underdeveloped pragmatic knowledge (see Guasti 2002; Eisenbeiß 2009 for
overviews).
In contrast to such Full-Competence approaches, Structure-Building ap-
proaches do not assume adult-like representations for the early two-word
stage. Hence, they must account for children’s early grammatical represen-
tations and explain how children acquire adult-like representations. Most
current Structure-Building approaches adopt versions of the Lexical Learn-
ing Hypothesis (see Pinker 1984; Eisenbeiß 2007, 2009 for overviews; see
Radford 1990 for an earlier maturational approach). According to this hy-
pothesis, children have adult-like categorisation abilities, but they still need
to determine the grammatical features and properties of the input elements
they encounter. Recent versions of Lexical-Learning approaches assume
that this process is incremental: grammatical distinctions are acquired one
by one, lexeme by lexeme, and with initial restrictions of inflections to in-
dividual lexemes that frequently appear with these markers in the input (see
e.g. Eisenbeiß 2003, 2007). For instance, possessive markers should initially
be restricted to individual words – and only later generalised to all words or
phrases that can carry this marker in the target language (Eisenbeiß 2000).
Usage-based approaches make similar predictions as Lexical-Learning
approaches, though on the basis of slightly different assumptions. Accord-
ing to them, adults grammars are based on schemas or constructions, i.e.
interrelated form/meaning pairs that are characterized by various degrees of
abstractness, ranging from idioms with concrete lexical items (e.g. kick the
bucket ‘die’), to abstract templates characterized by grammatical roles such
as subject-predicate (e.g. Goldberg 1995, 2006). Children are assumed to
acquire such templates step by step; beginning with limited generalisations
that are centred on individual words or phrases and then gradually extending
these generalizations by analogy. For instance, Tomasello (2003, 2006) ar-
gues that grammatical morphemes such as case markers and agreement
markers are initially associated with individual (high frequency) verbs.
Only when a critical mass of such “verb-islands” is learned do children ac-
quire more general constructions, e.g. the transitive construction.
Thus, taken together, both generative and usage-based approaches pre-
dict that children acquire the general properties of the possessive construc-
tions that they hear in their input already in the two-word stage. Moreover,
structure-building approaches as well as usage-based approaches specifi-
cally predict an incremental acquisition of the specific properties of indi-

vidual possessive constructions and markers, with initial lexical restrictions
of possessive markers. By contrast, capturing such a developmental path
would require additional assumptions in a full-competence approach, for
instance, reference to the interaction between linguistic development and
the learning of cultural norms for possessive relations and negotiations
about them.
With respect to the order in which linguistic constructions are acquired,
most current approaches to language acquisition only make very general
predictions that are not directly relevant for the acquisition of possessive
constructions.3 However, most acquisition researchers agree that conceptual
complexity may influence acquisition orders (see Eisenbeiß 2006, 2009 for
discussion). In the following, we will discuss several factors that affect
conceptual complexity and could thus influence developmental orders: in
the section about attributive possession, we will explore the conceptual dif-
ferences between possessive relations that involve ownership, kinship rela-
tions, body-part relations, etc. Here, we will argue that more prototypical
types of possession, which involve physical control and proximity (Heine
1997; Seiler 1983) are acquired earlier than more abstract notions of pos-
session.
In the section on predicative possesion, we will discuss the observation
that possession constructions have often developed on the basis of older
locative constructions that encode a more concrete and “visible” relation-
ship; and we will investigate whether we can observe a similar pattern in
children’s linguistic development. Moreover, we will study when children
start to produce utterances that involve different types of possessive rela-
tions, e.g. ownership vs. current physical proximity and control in an utter-
ance like I have my mommy’s glasses now.
In the section on external possession, we will investigate EPCs like I hit
him on the head and internal possession constructions (IPC) like I hit his
head. Both constructions involve an Agent-Patient relation between the hit-
ter and the hittee as well as a possessive relation between the Patient and
3
For instance, Radford (1996) argues that morpho-syntactic realizations of the
functional category COMP (complementizers, wh-elements, etc.) are acquired
later than realizations of the functional category INFL (tense and agreement in-
flections, etc.). Moreover, some generative psycholinguists argue that the ability
to produce adult-like passive sentences only develops around the fourth birthday,
due to neural maturation of the underlying mechanisms (e.g. Borer and Wexler
1987). See Eisenbeiß (2009) and Tomasello (2003) for critical discussion.
their body part. However, in the EPC, both relations have to be mapped
onto a single argument hierarchy as both PR and PM are realised as verb
arguments. By contrast, in IPCs, the Agent-Patient relation is realised on
the sentential level while the possessive relationship is encoded within the
Patient noun phrase. We will argue that this is conceptually simpler than the
integration of agentive and possessive relations in the EPC – which should
be reflected in acquisition orders and possibly also in deviations from the
target language.
3. The German data
Our study of German child language is based on 64 recordings from 7

monolingual German children, aged 1;11 to 3;6. The data come from the
Clahsen corpus, the LEXLERN corpus (see Clahsen, Vainikka, and Young-
Scholten 1990) and the Wagner corpus (1985; see Appendix B for details).
Some of the LEXLERN recordings involved games in which possessions
were compared or exchanged (see Eisenbeiß 1994, 2003). Therefore, they
are comparatively rich in possessive constructions and some aspects of pos-
sessive constructions have already been studied in these corpora (Eisenbeiß
2000).
In order to analyse the course of development and to create comparable
data sets from children at similar stages of development, we assigned the
individual recordings to four developmental stages. We did not use chrono-
logical age to group the data as age is not considered to be a good indicator
of linguistic development. MLU (mean length of utterance) is also not suf-
ficient as some of the later recordings have MLUs of more than 3 words per
utterance; and MLU-values in that range are not as indicative of linguistic
development as lower values. Thus, we had to find another way of grouping
the recordings for comparisons.
Noun-phrase development is obviously crucial for adnominal possession
constructions; and the case markers in noun phrases are central for predica-
tive possession constructions and EPCs. Therefore, we decided to group
our recordings on the basis of the four-stage model of noun-phrase devel-
opment that Eisenbeiß (2000, 2003) had proposed in her analysis of Ger-
man child language data and applied to our data set (see Appendix B).
STAGE I
Children frequently omit possessive markers and other morphological
markers. They also omit D-elements, i.e. function words in the noun
phrase (e.g. case/gender/number-marked articles, possessive pronouns,

demonstratives, quantifiers). At the same time, the occurrence of D-
elements is restricted to potentially formulaic combinations with a small
set of high-frequency predicates and nouns (e.g. das-is-ein-X ‘that-is-a-
X’, die-mama ‘the mommy’).
STAGE II
The percentage of D-elements falls to even lower levels.
STAGE III
The percentage of D-elements increases again. Formulaic utterances be-
come less frequent and children start to combine D-elements more freely
with a broader range of predicates and nouns.
STAGE IV
Children typically produce D-elements whenever they are required.
Taken together, this U-shaped development in the provision of function

words and the early distributional restrictions for these elements suggest
that these elements are initially part of unanalysed chunks that are later re-
analysed. Hence, it is crucial not to mix data from the different stages for
analysis.
4. Adnominal possession
In the following, we will first give an overview of adnominal possession

constructions in adult German and contrast them with the corresponding
constructions in other languages for which acquisition studies are available –
English, Japanese, Greek, and Hebrew. Based on this, we will discuss em-
pirical findings from available acquisition studies and our own study.
4.1. Adnominal possession in the adult language
In order to illustrate the typological differences with respect to adnominal

possession constructions and their implications for language acquisition, we
will first look at two closely related languages, namely German and English.
In both languages, PRs can either be referred to by pronominal elements as
in (3a, b) or by non-pronominal noun phrases as in (3c–i). Moreover, both
pronominal and non-pronominal PRs can be combined with a preposition as
in (3b, d–f) or they can appear without it as in (3a–c, g–i). The prepositional
PR phrase has to follow the PR in English as in (3b, d), and it tends to ap-
pear in the same position in German as in (3b, d). However, in German, at
least some speakers find it acceptable to position the prepositional PR
phrase to the left of the PR for emphasis as in (3e) or to extract the preposi-
tional PR phrase in a question – see (3f). Note, however, that structures
where the prepositional phrase does not follow the PR are highly marked –
if not unacceptable.
In addition to the adnominal possession constructions in (3a) to (3f),
German has two constructions that do not have an equivalent in English.
The first one is a genitive construction, where the PR noun phrase typically
follows the PM as in (3g), but may also precede it as in (3h). This genitive
construction is more characteristic of formal and written German, has so far
not been observed in the speech of pre-school children (Clahsen et al. 1994;
Eisenbeiß 2000; Mills 1985), and does not occur in the corpora we analysed.
Note that the genitive construction is different from the possessive -s con-
struction though this is not immediately obvious for masculine nouns that
take -s as their genitive ending (see (3g, h)). The difference between the -s
possessive and the “real” genitive can be seen when one looks at masculine
nouns with a different genitive ending or femine nouns, which do not carry
any overt marker in the genitive (see (3i)).
The second German possessive construction without an English equiva-
lent appears in some spoken variants of German and involves a PM that is
preceded by a dative-marked PR and a resumptive possessive pronoun, as
in (3j). Due to its regional character, this construction is rarely discussed in
the acquisition literature (see Penner and Weissenborn 1994 for some initial
observations), and we only found three instances of this type in the data of
the German boy Carsten that we analysed. Therefore, we will not discuss
adnominal genitive and dative constructions in a lot of detail.
(3) a. sein Freund

his friend
‘his friend’
b. ein Freund von ihm
a friend of his
‘a friend of his’
c. Pauls Freund
Paul’s friend
‘Paul’s friend’
d. ein Freund von Paul/seinem Vater

a friend of Paul /his father
‘a friend of Paul’s’ / ‘a friend of his father’
e. Das ist bestimmt VON PAUL der Freund
that is surely OF PAUL the friend
‘That is surely Paul’s friend’
f. Von wem hast du den Vater gesehen?
of whom have you the father seen?
‘Whose father have you seen?’
g. ein Freund seines Vaters
a friend his-GEN father’s
‘a friend of his father’
h. Pauls / seines Vaters Freund
Paul’s / his-GEN father’s friend
‘Paul’s friend’ / ‘his father’s friend’
i. eine Freundin meiner Mutter / des Jungen
a friend my-GEN mother’s / the-GEN boy’s
‘a friend of my mother’ / ‘a friend of the boy’
j. dem Paul / Vater sein Freund
the-DAT Paul / father his friend
‘Paul’s friend’ / ‘the father’s friend’
Given the range of constructions mentioned above, both German and Eng-
lish children have to learn (i) when to use pronominal or a non-pronominal
constructions and (ii) when to use a construction with a non-prepositional
PR rather than a construction with a PR-PP. Moreover, German children
have to acquire adnominal constructions with genitive PRs or a combination
of dative PRs and possessive pronouns – though we won’t be able to inves-
tigate this aspect of the acquisition process due to a lack of relevant data.
With respect to the choice of pronominal vs. non-pronominal construc-
tions, German and English are similar: possessive pronouns are preferred
when the PRs can be identified on the basis of contextual or discourse in-
formation. As this is particularly easy for speakers and hearers, adults tend
to use 1st and 2nd person possessive pronouns rather than names when they
talk to other adults (e.g. mein-e/dein-e Henne ‘my/your chicken’). However,
in children’s early language and in language directed at young children, we
can often observe the use of names and kinship terms instead of 1st and 2nd
person possessive or personal pronouns (e.g. Ruff 2000). For instance, a
mother might tell her daughter Jane: Das ist Jane-s Auto; und das ist
Mama-s Auto ‘this is Jane’s car; and this is mommy’s car’. Thus, we will
have to investigate when and for which types of PRs children use posses-
sive pronouns.
Whether a prepositional or a non-prepositional construction is chosen,
depends on a range of factors. The first factor is the syntactic status of the
PM: if it is an unmodified noun such as the proper name Paul, it can either
appear with a prepositional PR as in (3d) or with a non-prepositional PR as
in (3c). However, possessive pronouns as well as -s-marked PRs induce a
definite reading and cannot co-occur with determiners or wh-elements
(*ein / dieser/welcher Pauls/ mein Freund ‘a/ this / which Paul’s / my friend).
Thus, the use of a wh-question or an intended indefinite reading may require
a prepositional construction. We will now look at factors that play a role if
the choice of construction is not already determined by the type of PR.
For pronominal PRs, English or German speakers have a choice between
a possessive pronoun construction as in (3a) and a prepositional construction
with a pronoun as in (3b); the possessive pronoun construction is typically
preferred as PR phrases tend to proceed the PR, especially when they are
short, animate and topical and the PR is not used contrastively, see e.g. my
friend vs. ?the friend of mine.
For non-pronominal PRs, English and German differ with respect to the
factors that determine their choice of construction. In English, -s can attach
to non-pronominal PR phrases of any syntactic complexity (e.g. Jack’s/the
old farmer’s chicken farm). Similarly, prepositional constructions can in-
volve simple PR nouns or complex PR noun phrases (e.g. a teacher of Jane/
my little daughter). Thus, for non-pronominal PR phrases, English speakers
have a choice between -s and of. This choice is determined by the animacy,
topicality and syntactic weight of the PR and by the type of possessive rela-
tionship: -s is preferred for prototypical and inalienable possessive relations
and when the PR is animate, topical and short (e.g. Sue’s eyes) – whereas of
is preferred when the PR is inanimate, not topical and syntactically modi-
fied and the relationship between PR and possesum is not a close and proto-
typical possessive relation (e.g. the fumes of a shabby old car; see e.g.
Rosenbach 2002, 2005, 2008; Jäger and Rosenbach 2006; Denison, Scott,
and Börjars 2008 and references cited there for the discussion of these fac-
tors and their interaction). Note, however, that while these factors can con-
spire to make one construction highly preferable over the other, none of
these factors on its own can determine the choice of construction.
By contrast, German exhibits a constraint for the use of possessive -s
that cannot be violated and hence can uniquely determine the choice of
non-pronominal possessive construction: -s can only be combined with PR
nominals that lack an article or any other modifiers. Typically, these nouns
are proper names like Susi, but some kinship terms can also be used with-
out a determiner and can thus be combined with -s (e.g. Mamas/Papas Auto
‘mommy’s / daddy’s car’). This has led some linguists to distinguish posses-
sive -s markers from genitive markers by calling them “proper name posses-
sive markers” (see e.g. Harbert 2007: 161ff.). We will argue that the con-
straint on the use of -s is a syntactic constraint and not a restriction of -s to
a particular semantic class of nouns. In many regional variants of colloquial
German, kinship terms or proper names appear with determiners (e.g. die/
eine Mama ‘the/a daddy’ or die Emma ‘the Emma’). However, when such
phrases are used as PRs, -s cannot be used (e.g. *der Mamas/ Emmas Auto
‘the mommy’s/Emma’s car’). Rather, a prepositional construction is chosen
(e.g. das Auto von der Mama/ Emma ‘the car of the mommy/ Emma). Thus,
it is not the type of noun per se that determines whether -s can appear, but
the lack vs. presence of determiners or other modifiers.
Given the syntactic constraints for -s, German speakers can only choose
between -s and the prepositional von-construction when the PR is an un-
modified proper name such as Emma or name-like kinship term such as
Mama ‘mommy’. As the referents of these nouns are all animate and the
length of the PR phrase is limited to one word, animacy or syntactic weight
cannot determine the choice between an -s construction like Annas Auto
‘Anna’s car’ and a prepositional construction like das Auto von Anna ‘the
car of Anna’. However, the type of possessive relationship might play a
role, with -s being preferred for closer and more prototypical possessive
relationships. To our knowledge, this has not yet been investigated thor-
oughly.
To summarise, German and English both have possessive constructions
with pronominal and non-pronomial PRs; and the choice of pronominal vs.
non-pronominal PRs is determined by similar pragmatic factors. Moreover,
in both languages, possessive pronouns are preferred to prepositional
phrases with pronominal PRs – unless the PR requires a modifier. However,
in German and English non-pronominal possession constructions, additional
factors play a role: the choice between -s and prepositional constructions
for non-pronominal PRs is determined by semantic and discourse factors in
English, whereas the use of the German possessive marker is restricted to
particular syntactic environments – i.e. unmodified nominals. Thus, a choice
between -s and the prepositional construction is only available for proper
names and a few kinship terms that can appear without a determiner.
Not all languages show such a competition between a prepositional con-
struction and a construction with a possessive marker. For instance, in Japa-
nese, all PR phrases, whether they are personal pronouns, unmodified nouns
or more complex noun phrases, are marked by the postposition no and pre-
cede the PM (e.g. watashi/Toshiko no kuruma ‘I/Toshiko’s car). In Hebrew,
the PR is marked by the preposition shel and follows the possesum (e.g. pe
shel buba ‘(the) mouth of (the) doll’). The PR < PR order can also be ob-
served in constructions with possessive pronouns (e.g. hasefer sheli ‘the
book (of) my/mine’; see e.g. Armon-Lotem, Crain, and Varlokosta 2005).
Even in a language with only one possessive marker for non-pronominal
constructions, children may have to acquire different word order patterns:
For instance, in Standard Modern Greek, the genitive-marked PR is com-
patible with determiners for the PR and can either precede the PM as in
(4a) or it can follow it as in (4b); see Marinis (2002, 2003):
(4) a. Pira tu nikous to vivlio.

took the-GEN Nikos-GEN the-ACC book-ACC
‘I took Niko’s book.’
b. Pira to vivlio tu Niku.
took the-ACC book-ACC the-GEN Nikos-GEN
‘I took Niko’s book.’
Thus, when children acquire the adnominal possession construction of their

target language, they have to learn the different markers and the syntactic,
semantic, and discourse constraints on their use. They also have to deter-
mine the word order options of their target language.
4.2. Adnominal possession in child language
In the following, we will first focus on the developmental problem and in-
vestigate when children start to produce the adnominal possessive construc-
tions of their target language and whether their earliest uses of these con-
structions are restricted to particular lexical items – as predicted by
structure-building and usage-base approaches. Then, we will study whether
children show early sensitivity to the language-specific constraints that
govern the choice of construction. Against this background, we will then
try to provide an account for the order in which adnominal possessive con-
structions are acquired and used to encode different types of possessive re-
lations.
4.2.1. The course of development
Table 1 gives an overview of adnominal possession constructions in our

German child language data.
Table 1. Adnominal possesssive constructions
Pronominal PR Non-Pronominal PR
Stage Child possessive von + PR+PR PR+PR von + NP Total
pronoun pronoun w/o -s w. -s
Ann 6 – 1 – – 7
Han – – – – – 0
I Leo 0 – 17 – – 17
Mat 3 – 3 – – 6
Total 9 (30%) 0 (0%) 21 (70%) 0 (0%) 0 (0%) 30
Ann 3 – 1 – – 4
Han – – – – –
II Leo 1 – 6 1 – 8
Mat – – – – – –
Total 4 (33%) 0 (0%) 7 (58%) 1 (8%) 0 (0%) 12
And 63 – 2 2 – 67
Ann 15 – 3 – – 18
Han – – – 1 – 1
III
Leo 17 – 2 20 – 39
Mat 7 – – – – 7
Total 102 (77%) 0 (0%) 7 (5%) 23 (17%) 0 (0%) 132
Ann 25 – – – – 25
Car 105 1 – – 4 110
Han 4 – – 1 1 6
IV Leo 26 – – 7 – 33
Mat 31 – – 5 1 37
Sve 75 1 – 17 5 98
Total 266 (86%) 2 (<1%) 0 (0%) 30 (10) 11 (4%) 309
Total 381 (79%) 2 (<1%) 35 (7%) 54 (11%) 11 (2%) 483
This tabulation suggests that there might be an early stage without adnomi-
nal possessive constructions: Hannah does not produce any of these con-
structions in stages I and II and only one proper name possessive construc-
tion in stage II. In stage I, she mostly labels or points out objects and
actions or asks for things, but does not talk about possessions. From stage
II on, possessive relations become a talking point for her and the possessive
pronoun mein(s) ‘my/mine’ appears, but only in one-word utterances, not in

combination with a PM noun. This seems to be a precursor for adnominal
possessive constructions that emerge in stage III (see Ruff 2000 for similar
observations).
When adnominal possessive constructions emerge, they seem to do so
incrementally: prepositional constructions can only be found in data from
stage IV – and even then only Carsten and Svenja produce more than one of
these constructions, see e.g. (5a,b). This suggests that prepositional con-
structions are a late acquisition, though they seem to be acquired earlier
than the more formal genitive construction, which does not occur in our
data at all. Similar observations have been made by Ruff (2000) in her lon-
gitudinal study with 7 German children (age range at beginning of record-
ings: 2;0 –2;9; age at the end: 2;4 –3;2).
(5) a. mama und wo is der kopp hier (Carsten / 3;6)4

mommy and where is the head here
von den junge von diesen?
of the boy of this-one?
‘Mommy, and where is the head of this boy?’
b. ich bin der hund von Sewenja (= Svenja). (Svenja/14 / 3;2)
I am the dog of Svenja.
‘I am Svenja’s dog.’
Both proper noun possessive constructions as in (6 a) and possessive pro-

noun constructions as in (6b) can already be found in stage I, though proper
noun possessives seem to emerge earlier. In stage I, Leonie only produces 17
proper noun possessives, but no possessive pronoun, and in stage II she uses
7 proper noun PRs, but only one possessive pronoun. And when Hannah
starts to use adnominal possessive constructions, her first construction con-
tains a proper noun PR. Moreover, we can observe a shift towards pro-
nominal PRs: In stages I and II, only 31% (13 /42) of all adnominal posses-
sive constructions involve a pronominal PR, compared to 84% (370 /441) in
stages III and IV. In stage IV, each of the children uses predominantly pro-
nominal PRs. Carsten does not use proper name possessive constructions at
all, but this is probably due to the fact that he uses the dative possessive
construction, which is a regional variant, instead (see (6c)). However, even
this construction only occurs three times.
4
Recording information is provided in the following format: (Name/number of
recording /age).
(6) a. S.E.: das is mamas koffer. ne? (Leonie/ 01/01;11)

this is mommy’s suitcase. TAG?
und das da?
and that there?
‘This is mommy’s suitcase, isn’t it? And that one over
there (is)?’
Leonie: mann koffer.
man(’s) suitcase
‘(the) man(’s) suitcase’
b. mein puppe i weg. (Annelie/02 /2;5)
my doll is gone.
‘My doll is gone.’
c. das is oma ihr lesezeichen mama (Carsten /3;6)
that is granny her bookmark, mommy
‘That is granny’s bookmark, mommy.’
The observed shift towards pronominal PRs cannot simply be due to chil-
dren initially preferring their own name for self-reference: of the seven
children, only Annelie, Hannah, and Leonie ever employ their own name as
PRs; and Annelie and Leonie use both the possessive pronoun and their
own name as soon as they talk about themselves as PRs. Hannah produces
her name only once in an adnominal possessive construction (stage II) be-
fore she starts using both the possessive pronoun and her own name (stage
III). But recall that she already used the possessive pronoun in one-word
utterances in stage II. Thus, those children who use their own name for self-
reference in adnominal possessive constructions do not seem to do so be-
cause they have not yet acquired pronouns. Rather, they exhibit an alterna-
tion between pronouns and their names in these constructions.
As we do not have video data available to study the details of the situ-
ational context, we cannot investigate all factors that determine when a pos-
sessive pronoun is chosen. However, an alternation between pronominal
and non-pronominal references to the speaker as PR has also been observed
by Ruff (2000) who argues that possessive pronouns are initially used in
demands and conflict situations, whereas non-pronominal PRs appear in
descriptive utterances. Moreover, Ruff reports that German children use pro-
nouns more frequently when the speaker is the PR than when the addressee
is the PR.
Taken together, we observe an incremental development of possessive
constructions, with a late acquisition of prepositional constructions and a
shift towards pronominal constructions. While capturing these observations

would require additional assumptions in full-competence approaches, it is
explicitly predicted by both usage-based approaches and generative struc-
ture-building approaches. These approaches are also compatible with an
early stage during which possessive markers are omitted; and they predict
initial restrictions to particular lexical elements once they do occur. For
possessive -s markers, Eisenbeiß (2000, 2003) has already demonstrated
that the German children we are investigating go through an early stage
without -s (see also Table 1). Note that in stages I and II, -s is even omitted
when the target form occurred in the preceding discourse, see e.g.:
(7) S.E.: Und Papas Hose brauchen wir noch

and daddy’s trousers need we still
‘and we still need daddy’s trousers’ (Leonie/ 03 / 2;1)
Leonie: da papa hose
there daddy(’s) trousers.
‘there (are) daddy(’s) trousers.’
An early stage with omissions of possessive markers has also been ob-
served for other German children and in studies of the corresponding pos-
sessive markers in English, Greek, Hebrew and Japanese (Clahsen, Eisen-
beiß and Vainikka 1994; Penner and Weissenborn 1996; Brown 1973;
Radford 1990; Radford and Galasso 1998; Marinis 2002, 2003; Berman
1985; Armon-Lotem 1998; Clancy 1985). Proponents of full-competence
approaches have tried to show that possessive markers appear early even if
they are not always realised. For instance, Bohnacker (1997) argued that
possessive markers appeared in the data of a young Swedish child (Embla;
age: 1;8–2;1). However, the entire corpus only contained 14 possessive
markers; and none of these occur at the beginning of the recording period
(Eisenbeiß 2003). Still, it is not clear whether all children in all languages go
through a stage in which possessive markers are omitted. Note, however,
that structure-building and usage-based approaches do not in principle rule
out the possibility that possessive markers appear early – they simply allow
for the gradual and incremental acquisition of such markers and the possi-
bility of an early, marker-less stage. In fact, for languages where the posses-
sive marker is frequent, obligatory, salient (e.g. syllabic) and not homony-
mous with any other marker, proponents of structure-building and usage-
based approaches would expect this marker to be acquired quite early.
4.2.2. Lexical restrictions of possessive markers and pronouns
With respect to the early use of possessive markers, studies of German,

English, Japanese, and Hebrew have demonstrated that adnominal posses-
sive constructions with target-like markers can temporarily co-occur with
constructions that lack such markers (e.g. Brown 1973; Berman 1985;
Clancy 1985; Peters/Menn 1993; Clahsen, Eisenbeiß and Vainikka 1994,
2000; Penner and Weissenborn 1996; Armon-Lotem 1998; Radford and
Galasso 1998; Ruff 2000). For instance, Clahsen et al. (1994) showed that
the German child Simone provided the -s only in 33 of the 49 proper noun
possessives she produced between 2;0,25 and 2;2,21.; and the English child
Nicholas used ’s only in 14 of 60 obligatory contexts between 3;2 and 3;6
(Radford and Galasso 1998). A similar stage can be observed in the data of
Leonie and Andreas who produced -s only in 8 out of 29 and 2 out of 4
obligatory contexts in stage IV (see Table 1 and Eisenbeiß 2000).
Independently of the language under investigation, the early examples
of possessive markers that are reported in the literature typically involve
kinship terms like dad or the name of the child or close friends and rela-
tives (e.g. Clancy 1985: 458; Mills 1985: 185; Radford 1990: 89; Peters and
Menn 1993: 757ff.; Clahsen, Eisenbeiß and Vainikka 1994: 97 ff.; Stenzel
1994: 196f.; Radford and Galasso 1998: 37). This is compatible with the as-
sumption that the use of possessive markers is initially lexically restricted
to individual nouns that children hear frequently with this marker. Further
evidence for this hypothesis comes from the observation that the English-
speaking child Daniel combined -s with the name Mike, then with the
kinshp term mommy and only later with other PRs (Peters and Menn 1993).
Similarly, Leonie goes through a stage where she uses -s with familiar
names and kinship terms as in (8a), but not with the name Sonja, which she
had not been familiar with before the start of recordings as in (8b). Even
though she heard the form Sonjas during the recordings, she does not use
this marker herself until the last recording in stage II. Similarly, in the re-
cording we investigated Andreas consistently uses -s with papa ‘daddy’ as
in (8c), but always omits it with mama ‘mommy’ as in (8d).
(8) a. is mamis (Leonie/ 05/2;3)

is mommy’s
‘(This) is mommy’s.’
b. S.E.: Und welches ist Sonjas Auto? (Leonie/ 06 / 2;3)
and which-one is Sonja’s car?
‘and which one is Sonja’s car?’
Leonie: Sonja autos

Sonja cars
‘Sonja(’s) cars’ (correct: ‘car’)
c. hm papas gürtel (Andreas /2;1)
hm papa’s belt
‘papa’s belt’
d. e mama ticktack is(t) das (Andreas/2;1)
e mommy ticktack is that
‘That is mommy’s clock.’
One could attribute the early omissions of possessive markers and the initial
lexical restrictions to non-syntactic factors, e.g. to the fact that possessive
markers are typically unstressed and that some combinations of PRs and
possessive markers might be more difficult to hear or pronounce. However,
Eisenbeiß (2000, 2003) has argued that phonological factors alone cannot
account for the distribution of possessive markers in the corpora we ana-
lysed. For instance, the word Papa ‘daddy’, which Andreas uses with -s, and
the word Mama ‘mommy’, which appears without -s, have the same syllable
structure and the same final vowel.
Moreover, a distributional analysis of the possessive pronouns in our
corpora offers further support for the assumption of initial lexical restric-
tions. In stages I and II, only Annelie and Mathias produce more than one
possessive pronoun in an adnominal possessive construction; and the 9 in-
stances of possessive pronouns in Annelie’s early data only involve 5 high-
frequency nouns (puppe ‘doll’, mama ‘mommy’, schuhe ‘shoes’, zimmer
‘room’, and bilder ‘pictures’). Similarly, in stages I and II, Mathias only
produces 3 types of adnominal possessive constructions with the possessive
pronouns. Finally, all possessive pronouns in stages I and II are forms of
mein ‘my’; forms of dein ‘your’, unser ‘our’ or sein ‘his’ as in (9) do not
occur until stages III and IV.
(9) da da is doch seine mama. (Svenja/ 08 / 3;0)

there there is PART his mommy
‘there is his mommy’
At the same time, agreement with the PR is not target-like: except for one
use of the correct plural form meine in stage II (meine bilder ‘my pictures’),
Annelie only uses the uninflected form mein, which leads to agreement er-
rors for feminine and plural nouns that require the form meine (e.g. *mein
mama ‘my mommy’). Mathias produces the uninflected form mein once, in
an appropriate context. The remaining two possessive pronouns are incor-

rectly inflected (*meiner buch ‘my book’, *meiner lappen ‘my rag’). Thus,
both the range of possessive pronouns and their combinations are limited
initially. This is in line with studies that observed that English-speaking and
German children initially only used possessive pronouns when the PR was
the speaker (Tomasello 1998; Ruff 2000).
4.2.3. Sensitivity to language-specific constraints
While the previous section showed quite a few studies of the time course of
the acquisition process in the domain of adnominal possessive constructions,
there are fewer studies of the constraints that govern the use and choice of
these constructions, which is somewhat surprising given the rich theoretical
literature on the topic. We have already discussed the choice between pro-
nominal and non-pronominal constructions above (see Ruff 2000 for more
details).
With respect to non-pronominal possessives, we observed earlier that
the possessive marker -s is restricted to unmodified PR nouns in German –
typically proper names like Susi, and some kinship terms (e.g. Mamas/Papas
Auto ‘mommy’s/daddy’s car’). Thus, for modified PR nouns (e.g. das
kleine Huhn ‘the (little) chicken’), German speakers have to use a preposi-
tional or genitive construction (e.g. der Kopf von dem kleinen Huhn or der
Kopf desGEN kleinenGEN HuhnsGEN ‘the head of the little chicken’). This
raises questions with respect to German child language: as we observed
earlier, prepositional and genitive constructions do not appear in our data in
stages I– III and we only found a few prepositional constructions in some
files from late stage IV. At the same time, noun-modifying determiners
emerge in stage III and appear in nearly all obligatory contexts in stage IV.
Thus, what do young German children do when the PR is a common noun
that requires a determiner or some other modifier – and is thus incompatible
with -s? In principle, they could refrain from using any non-pronominal ad-
nominal possessive constructions with PRs that require a modifier until
they have acquired the prepositional or genitive constructions of their target
language, but – as we will see – that does not seem to be what happens.
Thus, children could do one of three things. The first option is to use an -s
possessive but to omit the required noun modifier. This would not violate
the German constraint for -s, though it would mean treating a common noun
like a proper noun. As reported in Eisenbeiß (2000), two of the children we
investigated exhibited this option: Svenja produced one and Leonie pro-
duced two overgeneralizations to an unmodified PR in an adnominal pos-

sessive construction, as in (10a–c).
(10) a. S.E.: Das is? (Leonie/ 07/ 2;4)

that is?
‘That is (what)?’
Leonie: affes banane
monkey’s banana
‘the monkey’s banana’
b. clowns hut (Leonie/11/ 2;7)
clown’s hat
‘the clown’s hat’
c. das is junges gürtel (Svenja/13 / 3;2)
this is boy’s belt
‘this is the boy’s belt’
While this is clearly a very small set of examples, more can be found in the
literature: see (11) for -s overgeneralizations reported by Mills (1985: 185).
What is striking about this set of examples is that it contains a plural noun,
which can definitely not be interpreted as a proper name. Thus, children do
not seem to violate the constraint on the affixation of -s to modified nouns.
They DO violate the requirement to use determiners with count nouns.
Note, however, that some German speakers who we showed the examples
to commented that adnominal possessive constructions such as affes ba-
nane are acceptable if one analyses the respective PR noun as a proper
name, which would not require a determiner in German.
(11) a. das is männers wagen (Scupin 3;1)

that is men’s car
‘That is the men’s car’
b. hier is männers wohnung (Scupin 4;3)
here is men’s apartment
‘Here is the men’s apartment’
c. da tut männers bauch weh (Scupin 4;4)
there does men’s tummy ache
‘The men’s tummies are aching here.’
d. an elefantes zähne (Scupin 5;8)
on elephant’s teeth
‘on the elephant’s teeth’
The second option for children faced with the dilemma of either violating
the -s or the determiner constraint would be to use a modified PR and omit
-s, thus producing a construction that exhibits a modifier, but does not show
the required morphological marking. We do not have a lot of evidence for
this option, only one example, which we did not include in our counts as
the interruption in it makes it difficult to interpret:
(12) darf ich mein mama [/ ] mein mama bademütze haben?

may I my mommy [/ ] my mommy swimming.cap have?
‘May I have my mommy’s swimming cap?’ (Annelie/04/2;7)
What is striking, however, is that we did not find any evidence for the third
option: overgeneralisation of -s to a modified PR noun, i.e. a violation of
the constraint on -s that would avoid treating the PR noun like a proper
noun. That is, we did not observe any “English-style” constructions such as
mein Mamas bademütze ‘my mommy’s bathing cap’. This is quite surpris-
ing as children DO obviously overgeneralise -s. Moreover, most of the
nouns that the children in our sample correctly use as -s-PRs could be
modified in the target language because kinship terms and even proper
nouns may be combined with determiners or some other modifiers in most
varieties of spoken German – even though they do not HAVE TO. And in-
deed all children in our sample that produce -s-PRs provide modifiers for
the nouns they use as -s PRs when they appear in other contexts: Andreas
affixes -s to papa ‘daddy’ when it is an unmodified PR (as in (8c) above),
but he combines this noun with a possessive pronoun in other contexts (as
in (13a)). Hannah only uses her own name in an -s possessive construction,
but at the same time she mostly combines her name with determiners in
other contexts, see e.g. (13b). Mathias only uses unmodified proper nouns
with -s, but he produces combinations of these nouns with determiners in
other contexts (e.g. (13c)). Leonie produces -s overgeneralisations from
stage III on, but the PR nouns she uses with -s appear without a determiner,
though they are used with modifiers in other contexts, as in (13d). The PR
nouns that appear with an overgeneralized -s marker in stage III and IV,
occur with determiners during the same stage, as in (13e). Svenja also com-
bines -s with nouns that she otherwise uses with articles and other modifiers,
see e.g. (13f), including the common noun junge ‘boy’, which she incor-
rectly affixes with -s. Moreover, Svenja is one of those children that already
show a contrast between the -s construction and the possessive construction;
and we find minimal pairs with an unmodified proper name or kinship term
in the -s construction and the same name with a modifier in the preposi-
tional construction of (13g, h).
(13) a. mein Papa macht wieder drauf. (Andreas/2;1)

my daddy makes again onto-this.
‘My daddy puts (it) on that again.’
b. da is die Hannah. (Hannah/06/2;6)
there is the Hannah.
‘There is Hannah.’
c. der Daniel hat das puttmach. (Mathias / 21/ 3;0)
the Daniel has that broken.
‘Daniel has broken that.’
d. un(d) de (/) die große Sonja. (Leonie/ 11/ 2;7)
and the (/) the tall Sonja
‘And the tall Sonja.’
e. da ist wieder der clown mit den kalten füßen, ne?
there is again the clown with the cold feet, TAG?
‘There is the clown with the cold feet again, isn’t he?’
(Leonie/ 09 / 2;6)
f. nein das is doch kein papa! (Svenja/ 05 / 2,10)
no that is PART no daddy!
‘No, that isn’t a daddy!’
g. Saschas hut (Svenja/13 /3;2)
Sascha’s hat
‘Sascha’s hat’
h. von die (corr: dem) Sascha.
of the-NOM /ACC.FEM.SG (corr:DAT.MASC.SG.) Sascha
‘Sascha’s’ (Svenja/13 / 3;2)
Taken together, these findings suggest that children may overgeneralise -s

to common nouns, but they do not violate the constraint that prohibits the
combination of -s with modified nouns. They do not even use a modifier
for proper nouns and kinship terms in possessive -s constructions when they
use these nouns with modifiers in other contexts. This supports the assump-
tion that children are sensitive to a syntactic constraint on the use of -s as
soon as we find this marker in their data, while they do not show a general
restriction of -s to proper nouns and kinship terms. This is in line with our
claim that the constraint on -s is a syntactic and not a semantic constraint in
the target language.
Children’s early sensitivity to the syntactic constraint on -s raises the
question why they do not make the incorrect assumption that the constraint
on the use of -s is semantic. That is, we have to explain why they do not
restrict the use of -s to proper names and kinship terms and overgeneralise -s
to modified proper names and kinship terms. We argue that children who
hear -s with single nouns would only assume that -s can also be combined
with more complex noun phrases if they found positive evidence for this in
the input. Thus, children acquiring English would learn to use -s with com-
plex noun phrases because they hear such combinations, whereas German
children would never have any evidence that would lead them to extend the
use of -s in this way. This means that German children do not have the
means to produce any adnominal possessive constructions with modified
PR nouns before they have acquired prepositional or genitive constructions.
This would leave them only one option: to treat the common PR nouns they
want to use as if they were proper names. Note that treating a common
noun like a proper name is not that uncommon for nouns with animate ref-
erents. For instance, we know quite a few adults who use words such as
baby and cat as proper names. Thus, children’s overgeneralisations are not
completely outside the limits of the target language, children are simply
stretching these limits when they are faced with a dilemma: violating the
syntactic constraint on -s or violating the requirement to use a determiner or
other modifier with a count noun. Once prepositional or genitive construc-
tions are acquired, there is no pressure to affix -s to common nouns that
require a modifier.
The idea that children need positive evidence to assume that -s can be
used for modified PRs can be captured straightforwardly: one could simply
assume that children’s unmarked expectation would be that bound mor-
phemes will concern the word to which they are attached, rather than the
phrase. One could also capture this idea in feature-based underspecification
models of morpho-syntactic development – such as the one assumed by
Eisenbeiß (2003; see Eisenbeiß 2009 for a summary). In such models,
grammatical features are only integrated into morphological representations
if there is positive evidence in the input that requires them. For instance,
children only integrate number features into lexical entries for nouns when
they are confronted with contrasts between singular and plural forms. The
difference between a simple noun and a complex noun phrase is captured by
additional features for phrasal projections in different syntactic frameworks
(see e.g. Grimshaw 1994). Thus, one could assume that English children
include such a phrase-feature in their input specification for ’s because they
find positive evidence in their input that supports this. In contrast, German
children do not find such positive evidence and thus do not extend the range
of elements that can be affixed by -s.
Interestingly, there is some initial evidence for such a conservative ap-

proach in Carsten’s data. The three adnominal possessive constructions
with a dative PR that Carsten produces all involve an unmodified noun,
even though the noun Kind ‘child’ requires a determiner in adult German
and Carsten uses it with modifiers in other contexts, as in (14a), and pro-
duces determiners in almost all obligatory contexts. This is compatible with
the assumption that he initially restricts the use of this construction to un-
modified nouns. Similarly, Penner and Weissenborn (1996) provide a few
examples for the emergence of dative possessive constructions in Swiss
German and the earliest examples with an overt possessive pronoun also
involve unmodified PR nouns, as in (14b). However, the data base is cur-
rently too limited to warrant a stronger conclusion. Moreover, children hear
adult-like constructions with modified dative PRs when they learn a variety
of German that allows them. Thus, they might acquire the appropriate gen-
eralisation early on.
(14) a. wi(ll) das kind viele autos? (Carsten / 3;6)

wants the child many cars?
‘Does the child want many cars?’
b. Nadaw sis Ue (=uhr) (J 1;10 ;19)
Nadaw his watch
‘Nadaw’s watch’
We already observed that the syntactic factors that determine the choice of
possessive construction in German are not relevant for English, where fac-
tors such as animacy play a crucial role. So far, no published study seems
to have systematically varied all the factors discussed in the literature on
adult English, but Armon-Lotem et al. (2005) carried out an elicited pro-
duction study with English-speaking children (3;2–6;3). These children used
’s with count nouns but not with mass nouns; and they used ’s with animate
PRs 90% of the time to encode a part-whole relation (the cowboy’s arm),
but less than 50% of the time for inanimate PRs (the tractor’s wheel).
With respect to the acquisition of the target word order, some studies
explicitly mention that possessive pronouns are correctly positioned (see
e.g. Ruff 2000 for German) and we are not aware of any reports that chil-
dren who acquire the languages under study ever incorrectly position the
possessive pronoun after the PR noun. Similarly, studies of German -s pos-
sessives, English ’s-possessives, and Japanese no-possessives report that
children show the PR-initial target word order from the beginning (Brown
1973; Clahsen, Eisenbeiß and Vainikka 1994; Clancy 1985; Eisenbeiß 2000;
Radford 1990; Radford and Galasso 1998; Ruff 2000). Both observations
also hold for our German data. As we mentioned above, the PR typically
follows the PM in German prepositional possessive constructions; and pre-
positional constructions where the PR precedes the PM are highly marked.
However, of the 13 prepositional possessive constructions that we have
found in our German data, 3 exhibit the marked PR-initial order (15a–c)
and one has a fronted prepositional phrase with a wh-element (15d):
(15) a. aber das is nich von wurst die pelle. (Carsten / 3;6)
but that is not of sausage the skin.
‘But that is not the skin of the saussage’
b. ob deiner auch von diese malers das anspitzt.
whether yours also of these crayons that sharpen
‘(Let’s see) whether yours also sharpens that (tip) of these crayons?’
(Carsten / 3;6)
c. das is vo von de von de Sascha fahrrad. (Svenja/13 / 3;2)
that is of of the of the Sascha bike.
‘That is Sascha’s bike.’
d. von wem hast du die schuhe da an? (Svenja/08 /3;0)
of whom have you the shoes there on?
‘Whose shoes have you got on, there?’
Note that the examples where the prepositional PR phrase precedes the PM
come from different children and corpora and similar examples can easily
be found even in a cursory search of the German child corpora in the
CHILDES database (MacWhinney 2000); see e.g. the following examples
from the corpus provided by Wagner (1985):
(16) aber das sind von Pfe[#] von der Reitschule Pferde.
but that are of hor[#] of the riding:school horses.
‘But that are the horses of the riding school.’ (Roman /9;2)
For Greek, which exhibits two different orders for PM and PR for non-pro-
nominal adnominal possessive constructions, we cannot draw any firm con-
clusions about the early stages, due to a scarcity of early data (see Eisenbeiß
2003 for a discussion of Marinis 2002, 2003). For Hebrew, where the PR
follows the PM in constructions with the marker shel, Berman (1985) and
Armon-Lotem (1998) argue that children produce possessive constructions
with the non-target-like order PR < PM before they start to use the correct
order and then finally the posssssive marker shel.
Taken together, these observations tentatively suggest a tendency to
place the PM before the PR, independently of the target language and con-
struction. However, further research is required in this domain as the find-
ings for Hebrew are based on a rather small set of examples and thus have
to be taken with caution. Moreover, more systematic corpus searches for
prepositional constructions in German child language are required.
4.2.4. Acquisition orders and conceptual complexity
Our discussion so far suggests that children acquire the adnominal posses-
sive constructions of their target language incrementally, but show early
sensitivity to the constraints on the use of these constructions in their target
language. Now, we will have a closer look at the order in which children
start to encode different types of possessive relationships. Table 2 gives an
overview of the the types of PRs in adnominal possessive constructions (in-
cluding Carsten’s three dative possessive constructions, which were not
included in Table 1).
As can be seen in Table 2, all PR (pro)nouns in stages I and II refer to
the children themselves or to other people. In stage III, we find a non-
human, but animate PR as in (10a), but inanimate PRs only appear once in
Svenja’s data and four times in Carsten’s data from stage IV as in (17) and
(18). Note, however, that in example (18), where the referent of the posses-
sive pronoun is a vehicle, this PR is construed as having “animate” proper-
ties.
(17) die wand von’n (corr: ’m) fenster seh ich.

the wall of’the-ACC (corr: DAT) window see I.
‘I see the wall of the window.’ (Svenja/15 /3;2)
(18) dann schleudert er doch nich mehr sein räder

then flings he PART not anymore his wheels
‘Then he doesn’t fling his wheels anymore.’ (Carsten / 3;6)
The observed incremental extension in the range of PR types is consistent

with the assumption that children start out with more prototypical adnomi-
nal possessive constructions that involve a human PR and move on to less
prototypical ones with inanimate PR. In order to further evaluate this claim,
Table 2. The animacy of PRs in adnominal possessive constructions
Stage Child Self Human Other Animate Inanimate Total

Ann 7 – – – 7
Han – – – – –
I Leo – 17 – – 17
Mat 3 3 – – 6
Total 10 (33%) 20 (67%) 0 (0%) 0 (0%) 30
Ann 4 – – – 4
Han – – – – 0
II Leo 2 6 – – 8
Mat – – – – 0
Total 6 (100%) 6 (100%) 0 (0%) 0 (0%) 12
And 55 12 – – 67
Ann 11 7 – – 18
Han 1 – – – 1
III
Leo 20 18 1 – 39
Mat 5 2 – – 7
Total 92 (70%) 39 (30%) 1 (<1%) 0 (0%) 132
Ann 20 5 – – 25
Car 77 32 – 4 113
Han 5 1 – – 6
IV Leo 24 9 – – 33
Mat 24 13 – – 37
Sve 47 50 – 1 1
Total 197 (63%) 110 (35%) 0 (0%) 5 (2%) 312
Total 305 (63%) 175 (36%) 1 (<1%) 5 (1%) 486
we will look at the types of (possessive) relations involved in adnominal

possessive constructions. Table 3 shows the types of relationships that ad-
nominal possessive constructions encode in our German child data. Note
that we have used a wide sense of “ownership” here that covers temporary
possessions as well as legal and habitual possession relations. We will try
to distinguish between these later.
As can be seen in Table 3, children’s adnominal possessive construc-
tions in stages I and II only involve ownership (as in (6a)) or kinship rela-
tions as in (19a)). Constructions encoding relationships between a body part
and its owner emerge in stage III, as in (19b), and relationships between ob-
jects and their parts only appear in the stage IV data of Svenja and Carsten
(see e.g. (15a,b) above). None of the children ever uses possessive pro
Table 3. Types of possessive relations in adnominal possessive constructions
Stage Child Ownership Kinship Body Part Part of Object Total

Ann 6 1 – – 7
Han – – – – –
I Leo 17 – – – 17
Mat 6 – – – 6
Total 29 1 0 (0%) 0 (0%) 30
Ann 3 1 – – 4
Han – 0 – – 0
II Leo 8 0 – – 8
Mat – 0 – – 0
Total 11 (97%) 1 (3%) 0 (0%) 0 (0%) 12
And 50 12 5 – 67
Ann 12 1 5 – 18
Han 1 – – – 1
III
Leo 36 1 2 – 39
Mat 7 – – – 7
Total 106 (80%) 14 (11%) 12 (9%) 0 (0%) 132
Ann 18 5 2 – 25
Car 91 7 11 4 113
Han 6 – – – 6
IV Leo 27 3 3 – 33
Mat 34 – 3 – 37
Sve 71 10 16 1 1
Total 247 (79%) 25 (8%) 35 (11%) 5 (2%) 312
Total 393 (81%) 41(8%) 47 (10%) 5 (1%) 486
nouns, -s possessives or prepositional constructions for non-prototypical

possessive relations (the dog’s shadow, the size of the block, the state of the
car, etc.) Thus, the earliest adnominal possessive constructions only involve
prototypical possessive relations.
(19) a. mein mama (Annelie/02/2;5)

my mommy
‘my mommy’
b. mein nase läuft noch mehr (Annelie/ 04 / 2;7)
my nose runs even more
‘My nose is running even more.’
So far, we have not distinguished between legal ownership and temporary

possession, which is simply based on current physical control and proxim-
ity. This is difficult to do on the basis of spontaneous speech samples be-
cause, in everyday situations, the person who has physical control over
something or is close to it, is often also the owner. Thus, we did not attempt
to distinguish between legal ownership and physical control for all adnomi-
nal possessive constructions. Instead, we searched for utterances in which
two possessive relations were encoded in the same utterance – one by a
possessive predicate and one within an adnominal possessive construction.
We found 10 interpretable utterances of this type in the data of Andreas,
Carsten, Leonie, Mathias, and Svenja. In these examples, we can observe a
clear distinction between the legal or habitual ownership relation, which is
encoded noun-phrase internally, and a temporary ownership or physical
control relation, which is encoded at the sentential level. For instance, An-
dreas has a water ball. In (20), he states that another person, namely Anette
has temporarily taken over control of the ball.
(20) da Annette hat mei(nen) Wasserball da. (Andreas/2;1)

there Anette has my water.ball there.
‘Anette has got my water ball over there’
Note that 2 of the 10 utterances that encode two different possessive rela-
tions come from Andreas (stage III) and all the others occur in stage IV.
One might attribute this to the fact that the combination of two possessive
relations in the same utterance requires a certain sentence length. However,
some of these sentences are actually only 4 or 5 words long; and in each of
stages I and II, we find more than a hundred utterances that are longer than
3 words (see e.g. (6a)). Thus, the appearance of combinations such as (20)
in stage III might not be an artifact of increasing sentence length. It is at
least in line with the general observation that children in stages III and IV
extend the range of PRs and possessive relations from ownership relations
with human PRs via body part relations to part-whole relations for inani-
mate objects.
4.3. Summary
Our analysis of the German child data and our literature review lead to the
following generalizations:
– At least some children do not seem to produce adnominal possessive

constructions in the early two-word stage, but only precursors, such as
single-word utterances that consist of the PR’s name or a possessive
pronoun.
– German adnominal possessive constructions emerge incrementally, in the
order proper name possessives possessive pronouns < prepositional
constructions < genitive constructions.
– As children get older, the proportion of pronominal PRs increases.
– An early stage where possessive markers are omitted could be observed
in our own data and in studies of the corresponding possessive markers
in English, Greek, Hebrew and Japanese.
– Studies of German, English, Japanese, and Hebrew child language have
demonstrated that adnominal possessive constructions with target-like
markers can temporarily co-occur with constructions that lack such
markers.
– In our German data, we observed initial lexical restrictions of possessive
pronouns and possessive markers. To our knowledge, detailed analyses
of such lexical restrictions have not been carried out for other languages,
but the available empirical results we discussed for English possessive ’s
are compatible with the assumption of initial lexical restrictions.
– Children show early sensitivity to the constraints on the use of adnominal
possessive constructions in their target language: German children over-
generalise the -s marker to common nouns that do not take this marker in
the target language. However, children do not violate the syntactic con-
straint that prohibits the use of -s for modified PRs – not even for those
PRs that they themselves use with modifiers in other contexts. An ex-
perimental study of English child language also showed early sensitivity
to some of the factors that determine the choice of adnominal possessive
construction, in particular animacy and the count/mass distinction.
– Children also seem to adapt to the target word order early on, though we
found a few fronted prepositional PRs in our own data and there are some
reports about non-target-like PR-initial constructions in child Hebrew.
– The German children we observed showed an incremental extension in
the range of PRs and possessive relations that were encoded in adnomi-
nal possessive constructions – from ownership relations with human PRs
via body part relations to part-whole relations for inanimate objects.
Taken together, these empirical generalisations suggest that children show
early sensitivity to the constraints of their target language. Thus, the analysis
of adnominal possessive constructions provides additional evidence for the
emerging consensus about early adaptations to the target language. How-

ever, our observations tentatively suggest a tendency to place the PM be-
fore the PR, independent of the target language and construction – though
further research is required to establish whether this is a reliable finding.
With respect to the time-course of development, we found evidence for an
incremental extension both in the range of constructions and in the range of
possessive relations that are encoded by these constructions. This can be
captured straightforwardly in lexical learning and usage-based approaches
as they explicitly predict that possessive constructions are acquired incre-
mentally. By contrast, full-competence approaches would need to make
additional assumptions to account for these observations. In addition, full-
competence approaches do not directly predict the initial lexical restrictions
that we observed for possessive markers or function words such as posses-
sive pronouns. Such initial lexical restrictions are, however, to be expected
if one adopts a Lexical learning or usage-based approach as these ap-
proaches assume that children’s early generalisations are limited and linked
to particular lexical elements.
5. Predicative possession
In contrast to the acquisition of adnominal possessive constructions, the

acquisition of predicative possession constructions has not been studied
extensively. We are only aware of a few studies which have investigated
the acquisition of possessive constructions with have and belong to evaluate
approaches to syntax-semantic mappings. We will discuss these studies
briefly below, though they are inconclusive as the data base is quite limited
and there is considerable controversy about the availability of constraints
on the mapping of PR and PM to argument structure positions (see e.g.
Bowerman 1990; Pinker 1984). Therefore, we will focus on empirical gen-
eralisations about predicative possessive constructions that were established
in cross-linguistic studies and can inform acquisition studies of predicative
possessive constructions. In what follows, we first overview these generali-
zations and their application to German before we discuss our German
child language data and relevant findings for English child language.
5.1. Predicative possession in adult language
The background for all discussions of predicative possessive constructions

is the realization that the possessive relation they encode is a binary and
“symbiotic” relationship between two entities that only receive their seman-
tic interpretation by virtue of one another: a possessive relation is “a relation
between two entities, a PR and a PM, such that one, the PR, is seen as being
in some way related to the other, the PM, as having it near or controlling it”
(Baron, Herslund, and Sørensen 2001: 2). However, there is no PR without a
PM and a PM is not a PM without a PR. Possessive relations share this
property with locative constructions and experiencer constructions: some-
thing is only conceived of as a Location if something is located there and
this element is only considered to be a Locatum if it has a Location. Simi-
larly, an Experiencer requires a Stimulus to experience something, just as a
stimulus cannot be construed as a Stimulus unless it causes a sensation in
an Experiencer. By contrast, Agents do not necessarily need Patients for
their activities – we can talk about people who kick, dance, and write with-
out ever specifying who they kick or what they dance or write.
Given this inherent co-dependency between the entities involved in a
possessive, locative or experience relationship, it is difficult to rank these
entities according to their agentivity or their control over the relationship –
i.e. with respect to the variables that (co-)determine which of them is lin-
guistically realized as the topic or grammatical subject of a construction.
Hence, it is unsurprising that we can observe a large amount of inter- and
intra-language variability in the syntactic realization of Experiencers and
Stimuli, Locata and Locations, and PRs and PMs (see e.g. Chappel and
McGregor 1996; Heine 1997; Seiler 1983). An Experiencer can be realized
as topic or subject (Most farmers fear angry roosters) or it can be encoded
as an object (Angry roosters frighten most farmers) – and the same is true
for the Locatum (The pencils are (lying) in the box vs. The box contains
some pencils).
Similarly, cross-linguistic studies of predicative possession construc-
tions frequently make a distinction between HAVE-constructions, where
the PR is realized as topic and subject (I have/own a car), and BELONG-
constructions, where the PM appears in this role (The car belongs to me.
The car is mine/ Peter’s; see e.g. Baron, Herslund, and Sørensen 2001 for
overview). Some authors have claimed that this distinction can be observed
in all human languages (Heine 1997: 33), though this is controversial.
HAVE and BELONG constructions do not just differ with respect to the
mapping of arguments onto grammatical roles; they can also differ with
respect to the definiteness of their arguments. In particular, the PM is typi-
cally defninite in German BELONG constructions, but indefinite in German
HAVE constructions (Heine 1997: 30f., see e.g. Das Huhn gehört mir ‘The
chicken belongs to me’ vs. Ich habe ein Huhn). Note, however, that it is
possible to combine a definite PM with haben ‘have’, especially when a

contrastive reading is intended (Ich habe den linken Schuh, aber nicht den
rechten ‘I have the left shoe, but not the right one’). By contrast, the restric-
tion against combining gehören ‘belong’ with an indefinite PM seems quite
strict. Whether children are sensitive to the definiteness contrast between
HAVE and BELONG constructions has – to our knowledge – not been in-
vestigated so far and we try to fill this gap.
Both HAVE- and BELONG-constructions are not restricted to encoding
ownership; they can also be used to express locative relations (Boston has
many great buildings. This district has always belonged to London, etc.).
However, HAVE constructions can express a wider set of meanings (kin-
ship, abstract possession as in She has a cold, etc.), while BELONG con-
structions tend to encode more concrete relations. Therefore, many linguists
consider HAVE constructions the unmarked option (see Baron, Herslund,
and Sørensen 2001 for discussion). However, in his discussion of learnability
issues in first language acquisition, Pinker (1984) argues that the unmarked
or “canonical” mapping is the one where the Locatum Argument becomes
the subject. This would make HAVE-constructions the marked alternative.
Bowerman (1990) argues that this would lead to the prediction that children
should initially make “default marking errors” and produce HAVE-construc-
tions with PM subjects and PR objects. However, such errors have not been
observed yet – and in later work, Pinker no longer makes the assumption of
canonical mappings (Pinker 1989). In the following, we will investigate
whether we find any evidence for non-target-like mappings of PR and Pos-
sessum onto grammatical functions – and whether they follow the HAVE-
or the BELONG-pattern.
Moreover, we will investigate the relationship between the encoding of
locative and possessive relations. According to Heine, possession exhibits
various structural similarities with domains such as location, experience and
existence and predicative possessive constructions draw on different pre-
existing constructions or “source schemata” (see Baron et al. 2001 for dis-
cussion). The most commonly cited ones are the action schema, which un-
derlies HAVE-constructions, and the location schema, which is the basis for
BELONG-constructions.
These links raise the question of whether children show any evidence
for the primacy of location for verbs that are taken to be derived from the
locative source schema – i.e., whether they use verbs such as belong or be
first with a locative or existential reading (Das gehört hier hin ‘This be-
longs here’) and only later with a possessive reading.
Note, however, that some linguists argue that possession is independent

of location – hence they would not assume that locative constructions form
the source of possessive constructions (see e.g. Payne this volume).
5.2. Predicative possession in child language
In the following, we will look at HAVE- and BELONG-constructions with

the words haben ‘have’, gehören ‘belong’ and sein ‘be’ in our German
child language data. Other possession verbs – such as besitzen ‘own’ – do
not occur in our data. We will investigate (i) when the different types of
predicative possessive constructions appear in our data, (ii) whether children
are sensitive to the fact that HAVE- and BELONG-constructions impose
different constraints on the definiteness of the PM, (iii) whether we can find
any mapping errors in children’s predicative possessive constructions, and
(iv) whether children show any evidence for the primacy of location for
verbs that are taken to be derived from the locative source schema.
In Table 4, we can see how often children used haben ‘have’ and ge-
hören ‘belong’ with a definite PM or an indefinite PM. Moreover, as dis-
cussed above, we found 10 instances of haben with a possessive noun
phrase (see (20) above) and we listed these utterances in a separate column.
As Table 4 demonstrates, predicative constructions with haben ‘have’ ap-
pear before constructions with gehören ‘belong’. However, we can see dif-
ferences between the individual children: Svenja uses both verbs, whereas
Hannah, Carsten and Andreas use haben, but not gehören. Annelie used
haben from stage I, but gehören ‘belong’ only appears from stage II. Leo-
nie and Mathias do not exhibit any predicative possessive constructions in
stages I and II. In stage III, haben occurs quite frequently and we find one
instance of locative gehören in Leonie’s data. In stage IV, both children use
gehören.
Table 4 also shows that children are sensitive to the definiteness con-
straint for gehören and do not use any indefinite noun phrases as PM for
this verb. Instead, they produce definite articles – either used as pronouns
or in combination with nouns as in (21a, b, c) and demonstrative pronouns
as in (21d–g). In addition, we found one unclassifiable utterance with a
self-interruption that we excluded from the analysis. By contrast, haben
predominantly occurred with indefinite noun phrases (342 / 480 =71%, see
e.g. (22a), though we also found PM noun phrases with definite articles
(22b) and the 10 noun phrases with possessive pronouns that we discussed
above – see (20).
Table 4. Types of PM Noun Phrases co-occurring with haben ‘have’ and gehören
‘belong’
gehören possessive gehören locative

haben
Child (‘this belongs to me’) (‘this belongs here’)
a
Indef Def Poss Total Indef Def Total Indef Def Total
Ann 1 – – 1 – – – – – –
Han 4 1 – 5 – – – – – –
Leo – – – – – – – – – –
I
Mat – – – – – – – – – –
Total 5 1 0 6 0 0 0 0 0 0
(83%) (17%)
Ann 7 – – 7 – – – – 4 4
Han 2 – – 2 – – – – – –
Leo – – – – – – – – – –
II
Mat – – – – – – – – – –
Total 9 0 0 9 0 0 0 0 4 4
(100%) (100%)
And 54 18 2 74 – – – – – –
Ann 33 2 – 35 – – – – – –
Han – – – – – – – – – –
III Leo 12 11 – 23 – – – – 1 1
Mat 7 – – 7 – – – – – –
Total 106 31 2 139 0 0 0 0 1 1
(76%) (22%) (1%) (100%)
Ann 12 11 – 23 – – – – 1 1
Car 45 9 2 56 – – – – – –
Han 9 4 – 13 – – – – – –
Leo 47 11 2 60 – 1 1 – 6 6
IV
Mat 28 12 2 42 – 1 1 – – –
Sve 81 48 2 131 – 14 14 – 7 7
Total 222 95 8 325 0 16 16 0 14 14
(68%) (29%) (6%) (100%) (100%)
Total 342 127 10 480 0 16 16 0 19 19
(71%) (26%) (2%) (100%) (100%)
a
Def: Definite NP, Indef: Indefinite NP, Poss: adnominal possessive construction
Note that all of the utterances with haben are target-like with respect to
definiteness and with respect to the mapping of arguments onto grammati-
cal roles. By contrast, we find mapping problems with gehören. Svenja
never uses target-like dative marking for the PR. In example (21b), she
produces a prepositional phrase with von ‘of’ instead of a dative PR noun

phrase. In (21c–g), she produces nominative forms. Similar non-target-like
mappings have been observed in the example by Simone that we discussed
in the introduction. Simone’s resistance to corrections for these deviations
from the target suggests that such “errors” are not simply slips of the
tongue. Rather, at least some children seem to initially overgeneralise the
form-meaning mapping of the HAVE-construction to the BELONG-
construction. This is in line with the assumption that HAVE is the default,
but it would surprising if the HAVE-pattern were marked (Pinker 1984).
(21) a. der (ge)hört mir. (Mathias / 26 / 3;5)

this-one belongs me-DAT
‘This one belongs to me.’
b. der gehört von den (corr: dem) (\)
this-one belongs of this-one-ACC (corr: DAT) (Svenja/13 / 3;2)
‘This one belongs to this one’
c. wer (ge)horn die füße? (Svenja/16 / 3;3)
who-NOM belong the-NOM/ACC feet?
‘To whom do the feet belong?’
d. denn kuck mal hier wer das d(=g)ehört.
then look PART here who-NOM that-NOM/ACC belongs
‘Then just look here, to whom that belongs.’ (Svenja/13 / 3;2)
e. der (corr: dem) (ge)hört das ne?
this-one-NOM (corr: DAT ) belongs that-NOM/ACC TAG?
‘That belongs to this one, doesn’t it?’ (Svenja/16 / 3;3)
f. der (corr: dem) (ge)hört das. (Svenja/16 / 3;3)
this-one-NOM (corr:DAT) belongs that-NOM/ACC
‘That belongs to this one.’
g. dann (g)ehört das (Svenja/ 04 / 2;9)
then belongs that-NOM/ACC
niemand [#] keiner.
nobody-NOM (corr:DAT) [#] no-one-NOM (corr:DAT).
‘Then that belongs to nobody.’
(22) a. jetz ham (= haben) wir eine. (Andreas/2;1)

now have we one.
‘Now, we have one.’
b. die schere hat Julia. (Mathias/17/2;9)

the scissors has Julia
‘Julia has the scissors.’
For gehören, we observe more than just mapping problems. The data in
Table 4 also suggest that locative BELONG is a pre-cursor for possessive
BELONG: of the 3 children who use gehören more than once, only Svenja,
the most advanced child, uses both type of constructions. Recall, however,
that she shows mapping problems. She also uses the proximity preposition
bei ‘by’ in locative BELONG-constructions and not the target preposition
zu ‘to’, which is directional, as shown by (23a,b). These utterances are
somewhere between a locative and a possessive construction as they seem
to focus on placing an object in the proximity of an animate PR. In Anne-
lie’s data we only find 5 cases of the locative variant (24); and Leonie uses
the locative variant once in stage III and 6 times in stage IV (25a), but only
produces one instance of possessive gehören (25b).
(23) a. das schiff gehört bei dir. (Svenja/05 /2;10)

the ship belongs at you.
‘The ship belongs to you.’
b. die (ge)hört [#] ei [#] die (ge)hört bei (?) dir.
this-one belongs [#] ??? [#] this-one belongs at (?) you.
‘She belongs to you.’ (Svenja/04 /2;9)
(24) so (ge)hört das hin. (Annelie/03/2;6)

so belongs that hither.
‘That belongs here, like that.’
(25) a. das (ge)hört da drauf. (Leonie/15 / 2;11)

that belongs there onto-this.
‘That belongs up there.’
b. Sassa (ge)hört das. (Leonie/ 09 / 2;6)
Sascha belongs that.
‘That belongs to Sascha.’
In order to further explore the relationship between possession and other

semantic domains, we have compared different uses of sein ‘be’: utterances
that focus on location (and existence), as in (26); predication, as in (27);
combinations of location and possession, as in (28); and utterances with an
adnominal possessive as a subject and a further predication, i.e. utterances
where something is predicated about something that is said to belong to

someone, as in (29). We have also looked at sein in utterances that were used
to establish possession, as in (30). In these utterances a PR was either used
on its own or in a noun phrase with a PM. We have not distinguished between
these two types any further as they fulfill a similar function and appeared
with very similar frequencies in each stage. Table 5 provides an overview.
(26) da is die mama (Annelie/02 /2;5)

there is the mommy
‘There is mommy’
(27) is en löwe. (Hannah/01/2;0)

is a lion.
‘(This) is a lion.’
(28) da is Klaras. (Leonie/ 09 / 2;6)

there is Klara’s.
‘There is Klara’s.’
(29) daniels kopf is härter. (Mathias/26 /3;5)

Daniel’s head is harder.
‘Daniel’s head is harder.’
(30) das is doch Julias schiff. (Mathias/22 /3;1)

that is PART Julia’s ship.
‘But that is Julia’s ship.’
As Table 5 shows, sein is predominantly used for predictation and location,

though we also find one possessive use and a few utterances where the
child talks about the location of a possession. In stage III, possessive uses
become more frequent – except for Hannah. Thus, we can see a shift to-
wards more sein utterances involving possessive constructions. However,
in contrast to gehören ‘belong’, utterances with sein involve expressions of
possession from early on. But only in stage IV do we find more than one
utterance where the subject involves an adnominal possessive construction
and a further predication is added. This is in line with the observation that
combinations of haben ‘have’ with adnominal possessive constructions
only appear in stages III (Andreas) and IV (Carsten, Leonie, Mathias, and
Svenja). The only child to combine both haben and sein with adnominal
possessive constructions in stage III is Andreas, who is quite advanced for
this stage and already on the verge to stage IV (see Eisenbeiß 2000, 2003).
Table 5. Types of sein ‘be’
Location + Predication
Child Location Predication Possession Total
Possession + Possession
Ann 24 7 1 – – 32
Han – 1 – – – 1
I Leo – 2 1 – – 3
Mat 10 8 2 – – 20
Total 34 (61%) 18 (32%) 4 (7%) 0 (0%) 0 (0%) 56
Ann 29 15 1 – – 45
Han 2 – – – – 2
II Leo 1 1 – 1 – 3
Mat 12 3 – – – 15
Total 44 (68%) 19 (29%) 1 (2%) 1 (2%) 0 (0%) 65
And 82 34 10 3 1 130
Ann 31 22 1 4 – 58
Han 1 1 – – – 2
III
Leo 5 3 2 2 – 12
Mat 6 16 – 1 – 23
Total 125 (56%) 76 (34%) 13 (6%) 10 (4%) 1 (<1%) 225
Ann 42 43 6 16 – 107
Car 64 135 21 15 1 236
Han 20 29 1 4 – 54
IV Leo 83 83 6 4 3 179
Mat 34 53 9 8 1 105
Sve 275 227 7 34 1 544
Total 518 (42%) 570 (47%) 50 (4%) 81 (7%) 6 (<1%%) 1225
Total 721 (46%) 683 (43%) 68 (4%) 92 (6%) 7 (<1%) 1571
5.3. Summary
Our analysis of predicative possessive constructions in the German child

data has led to the following generalizations:
– HAVE-constructions appear before BELONG-constructions.

– As soon as children start to use HAVE- and BELONG-constructions,
they seem to be sensitive to the different definiteness constraints for
these constructions: they only use gehören ‘belong’ with a definite PM,
but combine haben ‘have’ with both definite and indefinite phrases.
– While HAVE-constructions show target-like mappings of arguments

onto grammatical roles, systematic overgeneralizations of the HAVE-
pattern to the BELONG-verb gehören are observed.
– When gehören emerges, it is initially restricted to its locative use (This
belongs here). Moreover, sein ‘be’ is initially mostly used for predica-
tion and location, though the possessive use of this verb precedes the
use of gehören.
– Combinations of haben and and sein with adnominal possessive con-
structions (X has Y’s Z, Y’s Z is X) only appear in data from late stage III
and stage IV.
Similar to the empirical generalisations about adnominal possessive con-

structions, these generalisations suggest that children become sensitive to
the constraints of their target language very early on. Thus, they are also in
line with the emerging consensus about early adaptation to the target lan-
guage. Moreover, just as in the adnominal domain, we observe an incre-
mental extension of forms and functions. In the case of predicative posses-
sive constructions, we can observe a primacy of locative relations, which is
in line with some of the typological literature on possession (see e.g. Baron,
Herslund, and Sørensen 2001). Such an incremental developmental path
can be captured straightforwardly in lexical learning and usage-based ap-
proaches as they explicitly predict that possessive constructions are ac-
quired incrementally. By contrast, full-competence approaches would need
to make additional assumptions to account for these observations.
Our findings about the acquisition of predicative possessive construc-
tions cannot only contribute to acquisition research; they can also help us
evaluate claims made in theoretical linguistics. In particular, the early and
error-free acquisition of HAVE constructions and the observed overgener-
alisaton of the HAVE-pattern to the BELONG-verb gehören support the
assumption that the HAVE-construction is the unmarked construction. This
is in line with a common view in the typological literature on possession
(see Baron, Herslund, and Sørensen 2001 for discussion), but it is incom-
patible with Pinker’s earlier analysis.
Recall that the observation that combinations of haben ‘have’ and sein
‘be’ with adnominal possessive constructions (X has Y’s Z, Y’s Z is X) ap-
pear comparatively cannot be simply attributed to the required sentence
length. Note that some of the utterances of this type actually only consist of
4 or 5 words, which is an utterance length that can be observed quite fre-
quently in stages I and II. As we argued above, the combination of different
types of possessive relations (temporary physical control vs. long-term
ownership) might be the reason for the late acquisition of combinations of

adnominal possessive constructions with haben. The observation that ad-
nominal possessive constructions also appear late in combination with sein,
could suggest a more general explanation: both the sein and the haben con-
structions that appear late involve adnominal possessive constructions that
encode a presupposition about possession, which is then combined with
another predication. We will further explore this idea below, where we dis-
cuss constructions that encode both possessive relations and Agent-Patient
relations.
6. External possession
In this section we focus on constructions that involve both possessive rela-

tions and Agent-Patient relations. This can be achieved by realizing the
possessive relation within a noun phrase and encoding the Agent-Patient
relation at the sentential level. For instance, in Jack tapped Frank’s shoul-
der, there is a possessive relationship between Frank and his shoulder,
which is captured within the adnominal possessive construction. In addi-
tion, there is an Agent-Patient relation between Jack on the one side and
Frank and his body part on the other. This relation is mapped onto the sub-
ject-object relation. Alternatively, the PR can be realized as an independent
verb argument in a so-called EPC (Neumann 1995; Payne and Barshi 1999;
Shibatani 1994). For example, in the sentence I tapped Frank on the shoul-
der; the PR Frank and its PM shoulder are realized as independent phrases.
In what follows, we focus on the realisation of possessive relations and
Agent-Patient relations in EPCs or adnominal possessive constructions in
German and Japanese child language data.
6.1. External possession in the adult language
EPCs are very common in German and occur in everyday language ad-
dressed to young children (see Neumann 1995; Appendix A, A23). Here,
the PR is typically realized as a dative-marked “extra” argument that is not
subcategorized by the verb. This extra dative argument refers to the PR of
the entity that is the Patient of the Action, while this PM is encoded as the
direct accusative object (die Haare ‘the hair’). This possessive relation
makes it possible that the dative noun phrase is integrated into the construc-
tion as an argument with all the morpho-syntactic characteristics of a sub-
categorised indirect object. The dative PR can be combined with a large

variety of intransitive and transitive verbs, but not with verbs that already
have a dative argument; see e.g.:
(31) a. Das Zebra beißt dem Hasen in den Schwanz

the zebra bites the-DAT hare- into the tail
‘The zebra bites the hare’s tail’
b. Der Mann waescht dem Baeren die Pfote
the man washes the-DAT bear the-ACC paw
‘The man washes the bear’s paw’
c. Der Junge legt dem Affen den Hut auf den Ruecken
the boy puts the-DAT monkey the-ACC hat on the back
‘The boy puts the hat on the monkey’s back’
d. *Der Junge gibt dem Vater dem Kind das Buch
the boy gives the-DAT father the-DAT child the-ACC book’
‘The boy gives the child’s father the book’
The dative PRs in these constructions are not adjuncts with a semantic case
that is solely determined by their thematic role and unaffected by any syn-
tactic processes. Rather, we observe the same syntactically determined case
alternations for dative PRs (as in (32)) as for indirect dative objects of
three-place verbs (as in (33)). In so-called recipient passive sentences with
the auxiliary kriegen/bekommen ‘get/become’ (e.g. (32b, 33b)), the dative
noun phrase of the corresponding active sentence (e.g. (32a, 33a)) carries a
nominative marker:
(32) a. Der Mann wäscht dem Hund den Kopf.

the-NOM man washes the-DAT dog the-ACC head
‘The man washes the dog’s head.’
b. Der Hund kriegt den Kopf gewaschen
the-NOM dog gets the-ACC head washed
‘The dog gets its head washed.’
(33) a. Der Mann gibt dem Baeren den Honig.

the-NOM man gives the-DAT bear the-ACC honey
‘The man gives the bear the honey.’
b. Der Baer kriegt den Honig gegeben
the-NOM bear gets the-ACC honey given
‘The bear is given the honey.’
Thus dative PRs behave differently from dative adjuncts with semantic
case, e.g. the so-called ethical datives in (34), which express the attitude the
referent of this noun phrase has to the action encoded in the verb. In con-
trast to dative PRs, ethical dative adjuncts do not exhibit any case alternation
in recipient passives. That is, the ethical dative pronouns keep their dative
marking, as in (34b, 35b) while external PRs appear with nominative mark-
ing in recipient passives, as in (35b) – just as indirect dative objects, as in
(34b). Moreover, ethical datives can be combined with both indirect objects,
as in (34), and dative PRs, as in (35), which suggests that they occupy a
different syntactic position, while indirect objects and dative PRs have the
same status.
(34) a. Dass du mir der-DAT Oma

that you-NOM me-DAT the-DAT granny
nicht wieder die falschen Tabletten gibst!
not again the-ACC wrong-ACC tablets give!
‘I hope that you won’t give granny the wrong tablets again!’
b. Dass mir die Oma nicht wieder
that me-DAT the-NOM granny not again
die falschen Tabletten gegeben kriegt!
the-ACC wrong-ACC tablets given gets!
‘I hope that granny won’t be given the wrong tablets again!’
(35) a. Dass du mir nicht wieder

that you-NOM me-DAT not again
jemandem auf die Füe trittst!
someone-DAT on the feet step!
‘I hope that you won’t step on someone’s feet again!’
b. Dass mir nicht wieder jemand
that me-DAT not again someone-NOM
auf die Füe getreten kriegt!
on the feet step gets!
‘I hope that nobody gets stepped on their feet again!’
While German exhibits external PR constructions with dative PRs as in

(36a), it is also possible to realize the possessive relation within an ad-
nominal possessive construction and the Agent-Patient relation at the sen-
tential level as in (36b):
(36) a. Sue legt dem Pferd den Sattel auf den Ruecken.
Sue lays the-DAT horse the-ACC saddle on the back.
‘Sue puts the saddle on the horse’s back.’
b. Sue legt den Sattel auf den Ruecken von dem Pferd.
Sue lays the-ACC saddle on the back of the horse
‘Sue puts the saddle on the horse’s back.’
Which of these constructions is selected is determined by semantic factors

(see e.g. Neumann 1995; Payne and Barshi 1999). For instance, the PR in
EPCs must be perceived as beneficially/adversely affected by an action by
virtue of being the PR of the entity which is directly affected by this action.
Moreover, the stronger the effect on the PR and the more intimate the pos-
session relation is, the more likely it is that the PR will be encoded as an
extra argument. Thus the inalienable possession relation between a person
and a body part is more likely to be encoded in an EPC than the relation
between a person and an alienable possession (e.g. a car).
While German shows parallels between indirect dative objects and ex-
ternal dative PRs, the presence of indirect dative objects and other dative
arguments (e.g. Experiencers) in a language does not imply that this lan-
guage will also exhibit dative-marked external PRs. For instance, Japanese
has indirect dative experiencers and a range of dative constructions that are
very similar to German. It also has “double subject” and “double object”
EPCs (Tsunoda 1995). However, Japanese lacks external dative PRs. In the
translation equivalents of the German examples (31a–c), both PM and PR
have to be encoded within the same noun phrase (e.g. Watashi-wa kuma-no
te-o aratta ‘(I) the bear’s the paw washed’). Note that the presence of such
an IPC does not rule out external PRs per se as EPCs and IPCs can co-exist
in languages such as German. Thus, if Japanese children ever started to use
datives for external PRs, it would be difficult to overcome this deviation
from the target language without recourse to reliable negative evidence.
6.2. External possession in child language
As far as we are aware, there are no systematic studies of EPCs in child

language. We will therefore first look at our German corpus data and then
present some preliminary resuls from elicitation games with German and
Japanese children. In our German data, we searched for all utterances that
involved an Agent-Patient relationship where the PR of the Patient was
mentioned: (i) as an extra argument, (ii) within the same noun phrase as the
Patient or (iii) both. Table 6 provides an overview.
Table 6. External and adnominal possessive constructions
Ownership Body-Part Total
Total
Total
Total
Both
Both
Both
EPC
EPC
EPC
IPC
IPC
IPC
And – 2 – 2 – 2 – 2 – 4 – 4
Ann – – – – – – – – – – – –
Han – – – – – – – – – – – –
III
Leo – – – – – – – – – – – –
Mat – – – – – – – – – – – –
Total – 2 – 2 – 2 – 2 – 4 – 4
Ann – 2 – 2 – 1 – 1 – 3 – 3
Car 1 12 – 13 – 4 1 5 1 16 1 18
Han – – – – – – – – – – – –
IV Leo – 2 – 2 1 1 – 2 1 3 – 4
Mat – 5 – 5 – – – – – 5 – 5
Sve – 13 1 14 1 1 2 1 14 1 16
Total 1 34 1 36 2 7 1 10 3 41 2 46
Total 1 36 1 38 2 9 1 12 3 45 2 50
As indicated in Table 6, Hannah does not produce any combinations of pos-

sessive relations and Agent-Patient relations. For the other children, such
combinations only appear in stage III (Andreas) or IV (for the other
children; see the examples below) – i.e. at the same time when we can ob-
serve the first combinations of haben ‘have’ and adnominal possessive con-
structions (see the previous section) and the first combinations of adnomi-
nal possessive constructions with a further prediction introduced by sein
‘be’. This is in line with our idea that the combination of possessive rela-
tions and other relations is a later development. Again, the fact that combi-
nations of possessive relations with another relation appear compartiavely
late cannot simply be attributed to sentence length as four or five words
would be sufficient for such constructions.
Moreover, EPCs are very rare in our child data. We only found three ex-
amples; see e.g. (37). Children tend to use constructions where possessive
relations are encoded within a noun-phrase – even when most adults would
probably prefer an EPC to encode events of taking something away from
someone, grooming and getting dressed, see e.g. (38). This suggests that the
integration of both possessive and Agent-Patient relations into one single

representation for verb arguments poses particular problems. This is sup-
ported by the fact that we also found two utterances where the PR appeared
both as a pronominal dative argument (mir ‘to-me’) and as a pronominal
possessive pronoun (meine ‘my’) – as if the occurrence of the extra dative
argument was insufficient to encode both the fact that the PR is affected by
the action and the fact that the speaker is the PR of the shoes/hands.
(37) jetz hau ich dir aber aber ganz xxx den popo.
now hit I you-DAT PART PART really xxx the-ACC botty.
‘Now I will really hit you on the botty.’ (Svenja/15 / 3;2)
(38) a. mama hat meine hose weggemommt (=weggenommen)

mommy has my trousers away-taken.
‘Mommy has taken away my trousers.’ (Annelie/05 /2;8)
b. jetz zieh i(ch) wieder mein(e) schuhe an. (Svenja/14 / 3;2)
now put I again my shoes on.
‘Now I am putting on my shoes again.’
(39) a. mama du solls(t) mir jetz meine hände aber waschen.

mommy you shall me-DAT now my hands PART wash.
‘Mommy, you ought to wash my hands now.’ (Carsten / 3;6)
b. dann dann zieh ich mir (Svenja/10 / 3;1)
then then put I me-DAT
meine meine schusche (=schuhe) an.
my my shoes on.
‘Then I will put on my shoes.’
Clearly, further studies of external possession with older children are nec-
essary. Moreover, in order to raise the number of potential contexts for ex-
ternal possession contexts, elicitation stimuli or games might be needed
(see e.g. the picture book stimulus by Eisenbeiß and McGregor 1999).
We have obtained some preliminary evidence from an elicitation study
with Japanese and 20 German and Japanese children (2–6 yrs; Eisenbeiß
and Matsuo 2003, 2005). This study made use of the puzzle tasks where
children are asked to describe events depicted on a puzzle board to obtain
puzzle pieces with matching pictures. The individual pictures on the puzzle
board differ minimally from each other, so that children have to express the
differences verbally in order to clearly identify the puzzle piece they want.
For the external possession study, three different types of events with af-
fected body parts were depicted and the pictures differed with respect to the
participants and body-parts, see Figure 1.
Figure 1. Sample elicitation materials for External Possessive Constructions

(Eisenbeiß and Matsuo 2003, 2005)
In order to obtain the desired puzzle piece, children had to mention Agents,
Patients and those body parts that were affected by the action. The German
target sentences with a dative PR involved three different constructions,
one with a nominative subject and a prepositional phrase for the body part
as in (40), one with a nominative subject and a direct accusative object as in
(41), and one with a subject, a direct accusative object and a prepositional
phrase for the body part as in (42).
(40) Das Zebra beißt dem Hasen in den Schwanz

the zebra bites the-DAT hare into the tail
‘The zebra bites the hare’s tail.’
(41) Der Mann waescht dem Hund den Schwanz

the man washes the-DAT dog the-ACC tail
‘The man washes the dog’s tail.’
(42) Der Junge legt dem Pferd den Sattel auf den Ruecken
the boy puts the-DAT horse the-ACC saddle on the back
‘The boy puts the saddle on the horse’s back.’
An initial analysis of the data showed early adaptations to the target lan-
guage: Japanese children consistently realized the possessive relationship
noun-phrase-internally, whereas German children used EPCs in more than
80% of the utterances that referred to a body part and its PR. However,
many utterances produced in the elicitation game were not target-like. When
children produced adnominal possessive constructions instead of external
PR constructions (which are not strictly ungrammatical, just dispreferred),
they sometimes exhibited the -s overgeneralizations to single PR nouns that
we described above, leading to non-target-like utterances such as (43).
When they attempted to produce external PR constructions, children often
do not manage to realize the PR as a case-marked dative noun phrase, but
use a prepositional phrase instead, as in (44). Note that these utterances are
not always target-like. If they use the preposition von ‘of’ for the PR, they
should use it within an adnominal possessive construction, following the
PM. Instead, they sometimes position the von-phrase to the left of the PM,
sometimes with other elements in between. This leads to utterances that are
not acceptable for the adult speakers we consulted, as in the case of (44a,b).
We also found utterances where the child used two prepositional phrases
with directional prepositions – one for the body part and one for its PR, as
in (44c). Thus, both were mentioned as the endpoint of the caused motion
event, but in a parallel fashion. This is also not an option in adult German.
Note that the deviations from the target we observed in the elicitation
games do not seem to be an artifact of the elicitation technique. Recall that
we found -s overgeneralizations and fronted prepositional phrases for PRs
in different sets of corpus data. We will argue that the deviations from the
target in these utterances result from difficulties in the intergration of pos-
sessive and Agent-Patient relations. This is supported by the observation that
they were more likely to use adult-like external possessive constructions for
two-place verbs such as bite than for three-place verbs such as put.
(43) a. Da legt der junge die leine auf katzes hals

there puts the boy the-ACC leash on cat’s neck
‘There the boy puts the leash around the cat’s neck.’
b. Da legt der junge den hut auf affens bauch
there puts the boy the-ACC hat on monkey’s tummy
‘There the boy puts the hat on the monkey’s tummy.’
(44) a. da v [/] legt der junge vom pferd auf’n ruecken

there [/] puts the boy of:the horse on:the back
den sattel
the-ACC saddle
‘There the boy put the saddle on the back of the horse.’
b. von kuh der junge in den (corr: das) bein beissen
of cow the boy into the-ACC.MASC (corr:NEUT) leg bite
‘The boy bites the cow’s leg.’
c. da legt der junge auf den pferd auf den kopf
there puts the-NOM boy on the horse on the head
den sattel.
the-ACC saddle.
‘There, the boy puts the saddle on the horse’s back.’
6.3. Summary
Clearly, more studies are required for EPCs. However, our analysis of predi-
cative possessive constructions in the German child data has led to some
preliminary generalizations and insights:
– In spontaneous speech from German children, we did not observe any

EPCs in stages I– III and only a few examples in stage IV.
– The German children from our corpus study sometimes produced con-
structions where possessive relations are encoded within a noun-phrase –
even when most adults would probably prefer an EPC to encode events
of taking something away from someone, grooming and getting dressed.
The older children in our elicitation study preferred EPCs to adnominal
possessive constructions for events where the body part of an animate PR
is affected physically, but also sometimes chose to realize the possessive
relation noun-phrase internally where an adult would produce an extra
dative PR.
– We found a few external possessive constructions where the PR was
realized both as a dative argument and as a possessive pronoun within
the possessum noun phrase.
– External PR constructions contained more deviations from the target-
language for three-place verbs like legen ‘put’ than for two-place verbs
like beissen ‘bite’.
– Common error types were the use of prepositional PR phrases instead of

dative PRs; and some of these prepositional phrases were fronted, which
is highly marked (or even non-target-like for many native speakers).
– By contrast, Japanese children do not overextend the use of the datives
they hear in their input to produce EPCs with dative PRs, which they do
not encounter in their target language.
Taken together, these observations suggest children adapt to the properties

of their target language quite early, but the integration of both possessive
and Agent-Patient relations into one single representation for verb argu-
ments poses problems. Moreover, we found additional evidence for a ten-
dency to position the PR before the PM – even if this leads to non-target-
like utterances.
7. Implications for theoretical linguistics and models of language

acquisition
When we compare the results for adnominal, predicative and external pos-
sessive constructions, we observe incremental development in all three do-
mains: the constructions of the target language emerge step by step. At the
same time, the range of functions that are encoded by possessive construc-
tions increases over time. For adnominal possessive constructions, the range
of relations expands from prototypical possessive relations via body part
relations to relationships between objects and their parts. In the case of
predicative possessive constructions, we can observe a primacy of locative
relations in early stages, which is in line with the typological literature on
possession.
In addition to incremental development, we found initial lexical restric-
tions for both possessive -s-markers and possessive pronouns in adnominal
possessive constructions. As discussed above, both incremental develop-
ment and initial lexical restrictions are explicitly predicted by structure-
building approaches as well as usage-based approaches. By contrast, cap-
turing such a developmental path would require additional assumptions in a
full-competence approach, for instance, reference to the interaction between
linguistic development and the learning of cultural norms for possessive
relations and negotiations about them.
While children may not use the full range of forms and functions in the
early two-word stage, they seem to adapt to the core properties of their target
language early on. For instance, children overgeneralise the German -s-
marker, which is restricted to unmodified proper names and kinship terms,

to count nouns that cannot be used in this way. However, they do not over-
generalise -s to modified nouns. Thus, they respect the syntactic constraint
of the adult language and do not adopt the incorrect generalization that -s
can be affixed to any names or kinship terms, whether they are modified or
not. We have argued that children do not make the non-target-like semantic
generalization because they would need positive evidence to assume that -s
can also be combined with more complex noun phrases – and they do not
find any such evidence in their input. In particular, we suggested that this
could be captured in feature-based underspecification models of morpho-
syntactic development – such as the one proposed by Eisenbeiß (2003; see
Eisenbeiß 2007, 2009 for summaries). In such a model, the difference be-
tween a simple noun and a complex noun phrase could be captured by addi-
tional features for phrasal projections and children would only include such
a phrase-feature in their input specification for -s if they find positive evi-
dence for it in their input – as they do in English – but not in a language
like German.
We also observed some overgeneralisations of the HAVE-pattern of
mapping arguments onto grammatical functions to BELONG-verbs. This is
surprising if one assumes that the BELONG-pattern is a “canonical” pattern,
as the analysis by Pinker (1984) suggests. However, if one assumes that the
HAVE-pattern is the default pattern (see Baron, Herslund and Sørensen 2001
for discussion), such overgeneralisations are to be expected. However, fur-
ther cross-linguistic research is required to determine whether children ever
produce the reverse overgeneralisations and whether these overgeneralisa-
tions could simply be attributed to the high frequency of the HAVE-pattern.
Similarly, we would like to see more studies of the acquisition of word order
in possession constructions to determine whether the observed tendency to
position the PR before the PM is a more general phenomenon.
BELONG-constructions and PM<PR-orders were not the only areas of
difficulty that we identified. In particular, we observed that children initially
did not use constructions that combine possessive relations with other pos-
sessive or non-possessive relations: combinations of haben ‘have’ and sein
‘be’ with adnominal possessive constructions (X has Y’s Z, Y’s Z is X) and
EPCs that integrate Agent-Patient relations into one single representation
for verb arguments. However, in order to evaluate how general this obser-
vation is, we would need more cross-linguistic studies of the encoding of
possession, in particular for predicative and external possessive construc-
tions.
Appendix A: A family discussion of possession
(A1) Mother: du isst noch dein eis?

you eat still your ice-cream?
‘You are still eating your ice-cream?’
(A2) Liam: issn [?] ich noch mein eis.
eat I still my ice-cream
‘I am still eating my ice-cream’
(A3) Mother: ja, iss mal schoen weiter.
yes eat PART nicely further.
‘Yes, just go on eating’
(A4) Mother: lass dir ruhig zeit.
leave yourself quietly time
‘Give yourself some time’
(A5) Mother: womit isst du das eis denn?
what-with eat you the icecream PART?
‘With what are you eating the icecream?’
(A6) Liam: xxx mit Ole, hm mit Leon, mit xxx.
with Ole, hm with Leon, with xxx
‘With Ole, hm with Leon, with xxx.’
(A7) Mother: aha.
Ah
‘Ah’
(A8) Mother: womit [?].
what-with
‘With what?’
(A9) Liam: mit du [?] (=dir)
with you-NOM (corr: DAT )
‘With you.’
(A10) Mother: womit [?] du [?] und isst du das [//]
what-with [?] du [?] and eat you that
ist das dein messer,
is that your knife
oder womit isst du das?
or what-with eat you that?
‘With what [?] you [?] and you eat that [//] is that your
knife or with what are you eating that?’
(A11) Liam: mein loeffel.

my spoon
‘My spoon.’
(A12) Mother: das is(t) dein loeffel?
that is your spoon
‘That is your spoon?
(A13) Liam: ja.
yes
‘Yes.’
(A14) Mother: nein, das is(t) mamas loeffel.
no, that is mommy’s spoon
‘No, that is mommy’s spoon.’
(A15) Liam: na.
na
‘No.’
(A16) Mother: aber mamas eis.
but mommy’s ice-cream
‘but (it is) mommy’s ice-cream.’
(A17) Liam: na.
na
‘No.’
(A18) Mother: mir gehoert das eis.
me-DAT belongs the ice-cream
‘The ice-cream belongs to me.’
(A19) Liam: nea.
nea
‘No.’
(A20) Mother: ach dir gehoert das eis.
oh you-DAT belongs the-NOM ice-cream.
‘oh, the ice-cream belongs to you.’
(A20) Mother: na gut, dann is(t) gut.
oh well, then is good
‘That is ok, then.’
(A21) Mother: oder gehoert das der Luna?
or belongs that the-DAT Luna?
‘Or does it belong to Luna?’
(A22) Lenny: der Leon hat sich nich(t) die zaehne

the Leon has himself not the-ACC teeth
geputzt, mama.
brushed, mommy
‘Leon has not brushed his teeth, mommy.’
(A23) Mother: Leon, putz dir bitte die zaehne, ja?
Leon, brush you-DAT please the teeth, yes?
‘Leon, brush your teeth, yes?’
This transcript comes from the L-family corpus, a corpus collected by

Sonja Eisenbeiß (http://corpus1.mpi.nl/ds/imdi_browser/, see MPI corpora >
L1 Acquisition). The collection and archiving of this corpus has been
funded by the Max-Planck-Society; and we would like to thank the family
for their participation in the project and all the members of the Max-Planck-
Institute who were involved in the collection, digitization and archiving of
the data. The transcription of the recordings was funded by a Research
Promotion Grant of the University of Essex, awarded to Sonja Eisenbeiß.
The video recordings were transcribed by two native speakers and conflicts
were either resolved or – if this was not possible – the relevant (part of the)
utterance was classified as unintelligible (Eisenbeiß and Sonnenstuhl 2007).
The ages at the time of recording are: Lenny (7;9), Leon (4;7), Liam (2;5).
Appendix B: Stages of Noun-Phrase Development
predicate remaining
utterances D-contexts +D- D+N
formula combinations
% of
child Filea age MLU n anal. stage n +D % n tokens types per file
+D
2;4– 2,01–
1–2 651 473 I 71 33 46 16 48 17 #1: 7, #2: 7
2;5 2,11
3 2;6 2,53 438 340 II 83 35 42 5 14 30 #3: 25
Ann
4 2;7 2,61 589 490 III 93 70 75 2 3 68 #4: 56
2;8– 2,54–
5–6 744 674 IV 133 125 94 9 7 116 #5: 37, #6: 47
2;9 3,07
2;0– 1,18–
1–2 552 442 I 33 19 58 14 74 5 #1: 3, #2: 2
2;1 1,23
2;2– 1,23–
3–4 499 355 II 50 10 20 6 60 1 #3: 1, #4: 3
Han 2;3 1,38
5 2;4 1,65 96 54 III 10 7 70 0 0 7 #5: 5
2;6– 2,45– #6: 23, #7: 47,
6–8 831 538 IV 136 131 96 0 0 131
2;8 2,85 #8: 31
predicate remaining
utterances D-contexts +D- D+N
formula combinations
% of
child Filea age MLU n anal. stage n +D % n tokens types per file
+D
1;11 1,57–
1–2 576 341 I 27 12 44 8 67 4 #1: 2, #2: 2
–2;0 1,67
2;1– 1,60–
3–4 659 414 II 32 11 34 7 64 4 #3: 3, #4: 1
2;2 1,66
Leo 2;3– 1,86– #5: 10, #6: 26,
5–8 1587 1166 III 183 131 72 21 16 110
2;5 2,08 #7: 26, #8: 17
9– 2;6– 2,12– 3157 2462 IV 545 499 92 33 7 466 #9: 36, #10: 38,
15 2;11 3,06 #11: 58, #12: 54,
#13: 42,
#14: 47, #15: 52
#9: 0, #10: 2,
#11: 0, #12: 0,
9– 2;3– 1,25–
1210 901 I 178 22 12 0 0 22 #13: 2, #14: 0,
17 2;9 2,11
#15: 3, #16: 2,
#17: 10
18 2;11 2,62 123 115 II 18 4 22 0 0 4 #18: 4,
Mat
19– 2;11 2,62– #19: 21, #20:
264 243 III 67 58 87 0 0 58
21 –3;0 2,65 n.a., #21: 23
#22: 37, #23:
22– 3;1– 2,24– 6, #24: 23,
804 719 IV 204 184 90 0 0 184
27 3;6 3,51 #25: 27, #26:
35, #27: 28
Andb 1 2,1 2,44 2344 1450 III
b
Car 1 3,6 4,22 2314 1795 IV
a
The column “file” provides the number of the respective recording. The column MLU
shows the mean length of utterance for this recording. The columns under “utterances”
provide the total number of utterances and the total number of analyzable utterances (i.e.
utterances that were intelligible and not simple yes/no answers or formulas such as hallo
‘hello’). Under “D-contexts” are: (i) the number of contexts in which an adult native
speaker would have produced a D-element, i.e. a nominal function word (determiner, pos-
sessive pronoun or quantifier), (ii) the number of overt D-elements and (iii) the percentage
of D-contexts where a D-element was used. The column predicate+D-formula shows how
many of the D-elements that children produced were found in potentially formulaic com-
binations with a small set of high-frequency predicates (e.g. das-is-ein-X ‘that-is-a-X’, die-
mama ‘the mommy’). The following columns show how many tokens and different types
of D-element-noun combinations remained after we excluded the predicate+D formulas
from the total number of D-elements in D-contexts.
b
For the cross-sectional data from the advanced stages no quantitative analyses of predi-
cate+D formula and D+N combinations were carried out as both Andreas and Carsten
used D-elements with broad ranges of nouns and predicates in obligatory contexts.
Acknowledgements
We would like to thank the children and the families who provided us with
data and Harald Clahsen for letting us use the Lexlern data. We are grateful
to Bill McGregor, Martin Henson and Bettina Landgraf for valuable com-
ments on earlier drafts and to Anette Rosenbach and Nikola Koch for many
discussions about possessive constructions. The research reported in this
article was supported by a grant from the Research Promotion Fund, Uni-
versity of Essex.
References
Ambridge, Ben, Julian M. Pine, Caroline F. Rowland and Chris R. Young

2008 The effect of verb semantic class and verb frequency (entrenchment)
on children’s and adults’ graded judgments of argument structure
overgeneralization errors. Cognition 106: 87–129.
Armon-Lotem, Sharon
1998 Mommy sock in a minimalist eye: On the acquisition of DP in He-
brew. In Issues in the Theory of Language Acquisition, Norbert Ditt-
mar and Zvi Penner (eds.), 15–36. Bern: Lang.
Armon-Lotem Sharon, Stephen Crain and Spyridoula Varlokosta
2005 Interface conditions in child language: a crosslinguistic look at some
aspects of possession. Language Acquisition: A Journal of Develop-
mental Linguistics 12 (2): 171–217.
Avrutin, Sergey
1999 Development of the Syntax-Discourse Interface. Dordrecht: Kluwer.
Baker, Carl L.
1979 Syntactic theory and the projection problem. Linguistic Inquiry 10:
533–581.
Baron, Irene, Michael Herslund and Finn Sørensen (ed.)
2001 Dimensions of Possession. Amsterdam /Philadelphia: Benjamins.
Bates, Elizabeth and Brian MacWhinney
1987 Competition, variation and language learning. In Mechanisms of Lan-
guage Acquisition, Brian MacWhinney (ed.), 157–194. Hillsdale, NJ:
Erlbaum.
Berman, Ruth A.
1985 The acquisition of Hebrew. In The Crosslinguistic Study of Language
Acquisition. Vol. 1, Dan I. Slobin (ed.), 255–371. Hillsdale, NJ: Erlbaum.
Berman, Ruth A.
1988 Word class distinctions in developing grammars. In Categories and
Processes in Language Acquisition, Yonata Levy, Izchak M. Schle-
singer, and Martin D. S. Braine (eds.), 45–72. Hillsdale, NJ: Erlbaum.
Bertolo, Stefano (ed.)

2001 Language Acquisition and Learnability. Cambridge: Cambridge Uni-
versity Press.
Bohnacker, Ute
1997 Determiner phrases and the debate on functional categories in early
child language. Language Acquisition 6: 49–90.
Borer, Hagit and Kenneth Wexler
1987 The maturation of syntax. In Parameter Setting, Thomas Roeper and
Edwin Williams (eds.), 123–172. Dordrecht: Reidel.
Bowerman, Melissa
1985 Mapping thematic roles onto syntactic functions: are children helped
by innate linking rules? Linguistics 28: 1253–1289.
Braine, Martin D. S. and Patricia J. Brooks
1995 Verb argument structure and the problem of avoiding an overgeneral
grammar. In Beyond Names for Things: Young Children’s Acquisition
of Verbs, Michael Tomasello and Walter E. Merriman (eds.), 352–
376. Hillsdale, NJ: Erlbaum.
Brown, Roger
1973 A First Language: the Early Stages. London: Allen and Unwin.
Chappell, Hilary and William B. McGregor, (eds.)
Part Terms and the Part-Whole Relation. Berlin and New York:
Mouton de Gruyter.
Chomsky, Noam
1965 Aspects of the Theory of Syntax. Cambridge, MA: MIT Press.
Chomsky, Noam
1981 Lectures on Government and Binding. Dordrecht: Foris.
Chomsky, Noam
1995 The Minimalist Program. Cambridge, MA: MIT Press.
Chomsky, Noam
2001 Beyond explanatory adequacy. MIT Occasional Papers in Linguistics
20. Cambridge, MA: MIT Press.
Chouinard, M. M. and Eve V. Clark
2003 Adult reformulation of childerrors as negative evidence. Journal of
Child Language 30, 637–669.
Clahsen, Harald and Martina Penke
1992 The acquisition of agreement morphology and its syntactic conse-
quences: new evidence on German child language from the Simone-
Corpus. In The Acquisition of Verb Placement: Functional Categories
and V2 Phenomena in Language Acquisition, Jürgen M. Meisel (ed.),
181–224. Dordrecht: Kluwer.
Clahsen, Harald, Sonja Eisenbeiß and Anne Vainikka
1994 The seeds of structure. A syntactic analysis of the acquisition of case
marking. In Language Acquisition Studies in Generative Grammar,
Teun Hoekstra and Bonnie D. Schwartz (eds.), 85–118. Amsterdam /

Philadelphia: Benjamins.
Clahsen, Harald, Anne Vainikka, and Martha Young-Scholten
1990 Lernbarkeitstheorie und Lexikalisches Lernen. Eine kurze Darstel-
lung des LEXLERN-Projekts. Linguistische Berichte 130: 466–477.
Clancy, Patricia M.
1985 The acquisition of Japanese. In The Crosslinguistic Study of Lan-
guage Acquisition. Vol. 1, Dan I. Slobin (ed.), 373–524. Hillsdale, NJ:
Erlbaum.
Clark, Eve V.
1978 Ergativity. In Syntactic Typology: Studies in the Phenomenology of
Language, Winfred P. Lehmann (ed.), 329–394. Hassocks, Sussex:
Harvester Press.
Clark, Eve V.
1985 The acquisition of Romance with special reference to French. In The
Crosslinguistic Study of Language Acquisition. Vol. 1, Dan I. Slobin
(ed.), 687–782. Hillsdale, NJ: Erlbaum.
Clark, Eve V.
1987 The principle of contrast: a constraint on language acquisition. In
Mechanisms of Language Acquisition, Brian MacWhinney (ed.), 1–33.
Hillsdale, NJ: Erlbaum.
Corina, David P.
1994 The induction of prosodic constraints: implications for phonological
theory and mental representation. In The Reality of Linguistic Rules,
Susan D. Lima, Roberta L. Corrigan and Gregory K. Iverson (eds.),
115–145. Amsterdam /Philadelphia: Benjamins.
Cowie, Fiona
1999 What’s Within: Nativism Reconsidered. New York: Oxford University
Press.
Crain, Stephen
1991 Language acquisition in the absence of experience. Behavioral and
Brain Sciences 14: 597– 611.
Denison, David, Alan Scott and Kersti Börjars
2008 What’s wrong with possessive ’s? Presentation at ISLE1, University
of Freiburg, 8 –11 October 2008.
(Available online at http://llc.stage.manchester.ac.uk/subjects/lel/staff/
david-denison/papers/thefile,143616,en.pdf.)
Eisenbeiß, Sonja
1994 Elizitation von Nominalphrasen und Kasusmarkierungen. In Elizi-
tationsverfahren in der Spracherwerbsforschung: Nominalphrasen,
Kasus, Plural, Partizipien. (Arbeiten des Sonderforschungsbereichs
282, Nr. 57), Sonja Eisenbeiß, Susanne Bartke, Helga Weyerts and
Harald Clahsen (eds.), 1–38. Düsseldorf: Heinrich-Heine-Universität.
Eisenbeiß, Sonja
2000 The acquisition of the Determiner Phrase in German child language.
In The Acquisition of Syntax: Studies in Comparative Developmental
Linguistics, Marc-Ariel Friedemann and Luigi Rizzi (eds.), 26–62.
London: Longman.
Eisenbeiß, Sonja
2003 Merkmalsgesteuerter Grammatikerwerb. Doctoral dissertation, Uni-
versity of Düsseldorf. (Available online at http://docserv.uni-
duesseldorf.de/servlets/DerivateServlet/Derivate-3185/1185.pdf.)
Eisenbeiß, Sonja
2006 Documenting child language. In Language Documentation and De-
scription. Vol. 3, Peter K. Austin (ed.), 106 –140. London: SOAS, The
Hans Rausing Endangered Languages Project.
Eisenbeiß, Sonja
2007 The lexical learning hypothesis. Essex Research Reports in Linguis-
tics 54: 1– 4. (To appear in: The Cambridge Encyclopedia of the Lan-
guage Sciences.)
Eisenbeiß, Sonja
2009 Generative approaches to language learning. Linguistics 47 (2): 273–
310.
Eisenbeiß, Sonja and Ayumi Matsuo
2003 External and internal possession – a comparative study of German
and Japanese child language. Poster presented at the 28th Annual
Boston University Conference on Language Development, Boston
University, USA.
Eisenbeiß, Sonja and Ayumi Matsuo
2005 Eliciting language production data from young children. Presentation
at the Xth International Congress for the Study of Child Language,
Berlin, Germany.
Eisenbeiß, Sonja and William B. McGregor
1999 The circle of dirt. Ms. Max-Planck-Institute for Psycholinguistics,
Nijmegen.
Eisenbeiß, Sonja and Ingrid Sonnenstuhl
2007 Transcription conventions for the L-family corpus. Ms. University of
Essex.
Goldberg, Adele
1995 Constructions. Chicago: University of Chicago Press.
Goldberg, Adele
2006 Constructions at Work: the Nature of Generalization in Language.
Oxford: Oxford University Press.
Grimshaw, Jane
1994 Minimal projection and clause structure. In Syntactic Theory and
First Language Acquisition: Cross-linguistic Perspectives, Barbara
Lust, M. Suner and J. Whitman (eds.), 75–83. Hillsdale, N.J.: Law-
rence Erlbaum Associates.
Guasti, Maria Teresa, Anne Christophe, Brit van Ooyen and Marina Nespor
2001 Pre-lexical setting of the head: Complement parameter through pros-
ody. In Approaches to Bootstrapping: Phonological, Lexical, Syntac-
tic and Neurophysiological Aspects of Early Language Acquisition,
Vol. 1, Jürgen Weissenborn and Barbara Höhle (eds.), 231–248. Amster-
dam /Philadelphia: Benjamins.
Harbert, Wayne
2007 The Germanic Languages. Illustrated Edition. Cambridge: Cambridge
University Press.
Heine, Bernd
1997 Possession: Cognitive Sources, Forces, and Grammaticalization. New
York: Cambridge University Press.
Hyams, Nina M.
1996 The underspecification of functional categories in early grammar. In
Generative Perspectives on Language Acquisition. Empirical Find-
ings, Theoretical Considerations and Crosslinguistic Comparisons,
Harald Clahsen (ed.), 91–127. Amsterdam/ Philadelphia: Benjamins.
Gerhard Jäger and Annette Rosenbach
2006 The winner takes it all – almost. Cumulativity in grammatical varia-
tion. Linguistics 44 (5): 937–971.
Kiparsky, Paul
1982 From cyclic phonology to lexical phonology. In The Structure of
Phonological Representations. Part 1, Harry van der Hulst and Norval
Smith (eds.), 131–175. Dordrecht: Foris.
König, Ekkehard and Martin Haspelmath
1998 Les constructions à possesseur externe dans les langues d’Europe. In
Actance et Valence dans les Langues d’Europe, Jack Feuillet (ed.),
525– 606. Berlin and New York: Mouton de Gruyter.
Küntay, Aylin and Dan I. Slobin
1996 Listening to a Turkish mother: some puzzles for acquisition. In Social
Interaction, Social Context, and Language, Dan I. Slobin, Julie Ger-
hardt, Amy Kyratzis and Jiansheng Guo (eds.), 265–286. Mahwah,
NJ: Erlbaum.
Lust, Barbara
1994 Functional projection of CP and phrase structure parametrization: an
argument for the strong continuity hypothesis. In Syntactic Theory
and First Language Acquisition: Cross-Linguistic Perspectives. Vol. 1,
Barbara Lust, Margarita Suner and John Whitman (eds.), 85–118.
Hillsdale, NJ: Erlbaum.
MacWhinney, Brian
2000 The CHILDES Project: Tools for Analyzing Talk. 3rd Edition. Vol. 2:
The Database. Mahwah, NJ: Erlbaum.
Marcus, Gary F.
1993 Negative evidence in language acquisition? Cognition 46: 53–85.
Marcus, Gary F., Steven Pinker, Michael Ullman, Michelle Hollander, John T. Rosen
and Fei Xu
1992 Overregularization in language acquisition. Monographs of the Society
for Research in Child Development. Vol. 57 (Serial No. 228), 1–165.
Marinis, Theodoros
2002 Acquiring the possessive construction in Modern Greek. In Proceed-
ings of the 1999 GALA Conference, Ingeborg Lasser (ed.), 57–80.
Frankfurt: Lang.
Marinis, Theodoros
2003 The Acquisition of the DP in Modern Greek. Amsterdam /Phila-
delphia: Benjamins.
Miller, Max
1976 Zur Logik der frühkindlichen Sprachentwicklung. Stuttgart: Klett.
Mills, Anne E.
1985 The acquisition of German. In The Crosslinguistic Study of Language
Acquisition. Vol. 1, Dan I. Slobin (ed.), 141–254. Hillsdale, NJ: Erl-
baum.
Neumann, Dorothea
1995 The dative and the grammar of body parts in German. In The Gram-
McGregor (eds.), 745–779. Berlin /New York: Mouton de Gruyter.
Payne, Doris L and Immanuel Barshi (eds.)
1999 External Possession. Amsterdam /Philadelphia: Benjamins.
Penner, Zvi, Manuela Schönenberger and Jürgen Weissenborn
1994 The acquisition of object placement in early German and Swiss
German. Linguistics in Potsdam 1: 93–108.
Penner, Zvi and Jürgen Weissenborn
1996 Strong continuity, parameter setting and the trigger hierarchy: On the
acquisition of the DP in Bernese Swiss German and High German. In
Generative Perspectives on Language Acquisition. Empirical Find-
ings, Theoretical Considerations and Crosslinguistic Comparisons,
Harald Clahsen (ed.), 161–200. Amsterdam/Philadelphia: Benjamins.
Peters, Ann M. and Lise Menn
1993 False starts and filler syllables: ways to learn grammatical mor-
phemes. Language 69: 742–777.
Phillips, Colin
1996 Order and structure. PhD thesis, Cambridge, MA: MIT.
Pinker, Steven
1984 Language Learnability and Language Development. Cambridge, MA:
Harvard University Press.
Pinker, Steven
1989 Learnability and Cognition: The Acquisition of Argument Structure.
Cambridge, MA: MIT Press.
Radford, Andrew
1990 Syntactic Theory and the Acquisition of English Syntax: The Nature
of Early Child Grammars of English. Oxford: Blackwell.
Radford, Andrew
1996 Towards a structure-building model of acquisition. In Generative
Perspectives on Language Acquisition. Empirical Findings, Theoreti-
cal Considerations and Crosslinguistic Comparisons, Harald Clahsen
(ed.), 43–90. Amsterdam /Philadelphia: Benjamins.
Radford, Andrew and Joseph Galasso

1998 Children’s possessive structures: a case study. Essex Research Re-
ports in Linguistics 19. Essex: University of Essex, 37–45.
Rizzi, Luigi
1993/94 Some notes on linguistic theory and language development: The case
of root infinitives. Language Acquisition 3: 371–393.
Rizzi, Luigi
2000 Remarks on early null subjects. In The Acquisition of Syntax: Studies
in Comparative Developmental Linguistics, Marc-Ariel Friedemann
and Luigi Rizzi (eds.), 269–292. London: Longman.
Rosenbach, Anette
2002 Genitive Variation in English: Conceptual Factors in Synchronic and
Diachronic Studies (Topics in English Linguistics 42.) Berlin /New
Rosenbach, Anette
2003 Aspects of iconicity and economy in the choice between the sgeni-
tive and the of-genitive in English. In Determinants of Grammatical
Variation in English, Günter Rohdenburg and Britta Mondorf (eds.),
379–411. (Topics in English Linguistics 43) Berlin/New York: Mouton
de Gruyter.
Rosenbach, Anette
2005 Animacy versus weight as determinants of grammatical variation in
English. Language 81: 613–644.
Rosenbach, Anette
2008 Animacy and grammatical variation – findings from English genitive
variation. Lingua 118: 151–171.
Ruff, Claudia
2000 Besitz und Besitzer in der Sprachentwicklung von deutschen und
italienischen Kindern. PhD thesis, TU Braunschweig, Biowissen-
schaften, Psychologie. (Available online at http://deposit.ddb.de/cgi-
bin/dokserv?idn=961129425.)
Scott, Alan, David Denison and Kersti Börjars
2007 Is the English possessive ’s truly a right edge phenomenon? Handout
of paper presented at The Second International Conference on the
Linguistics of Contemporary English (ICLCE2), University of Tou-
louse-le Mirail, 2nd July 2007.
Seiler, Hansjakob
Narr.
Shibatani, Masayoshi
1994 An integrational approach to PR raising, ethical datives, and adversa-
tive passives. BLS 20: 461– 486.
Slobin, Dan I., Melissa Bowerman, Penelope Brown, Sonja Eisenbeiß and Bhuvana
Narasimhan
in prep. Putting things in places: developmental consequences of linguistic
typology.
Stenzel, Achim
1994 Case assignment and functional categories in bilingual children:
Routes of development and implications for linguistic theory. In Bi-
lingual First Language Acquisition: French and German Grammati-
cal Development, Jürgen M. Meisel (ed.), 161–208. Amster-
dam/Philadelphia: Benjamins.
Stephany, Ursula
1997 The acquisition of Greek. In The Crosslinguistic Study of Language
Acquisition. Volume 4, Dan I. Slobin (ed.), 183–333. Mahwah, NJ:
Erlbaum.
Stephany, Ursula
1998 A crosslinguistic perspective on the category of nominal number and
its acquisition. In Studies in the Acquisition of Number and Diminu-
tive Marking, Steven Gillis (ed.), 1–23. (Antwerp Papers in Linguis-
tics 95) Antwerpen: Universitaire Istelling Antwerpen.
Tomasello, Michael
1998 One child’s early talk about possession. In The Linguistics of Giving,
John Newman (ed.). Amsterdam/Philadelphia: John Benjamins.
Tomasello, Michael
2003 Constructing a Language: A Usage-Based Theory of Language Ac-
quisition. Cambridge, MA: Harvard University Press.
Tomasello, Michael
2006 Acquiring linguistic constructions. In Handbook of Child Psychology,
D. Kuhn and R. Siegler (eds.), 1–101. New York: Wiley.
Tsunoda, Tasaku
1995 The possession cline in Japanese and other languages. In The Gram-
McGregor (eds.), 565–630. Berlin /New York: Mouton de Gruyter.
van der Linden, E. and A. Blok-Boas
2005 Exploring possession in simultaneous bilingualism: Dutch/French
and Dutch/Italian. In EUROSLA Yearbook, S. Foster-Cohen, M. d. P.
García-Mayo and J. Cenoz (eds.), 103–135. Amsterdam /Philadelphia:
Benjamins.
Wagner, Klaus R.
1985 How much do children say in a day? Journal of Child Language 12:
475– 487.
Wexler, Kenneth
1998 Very early parameter setting and the unique checking constraint: a
new explanation of the optional infinitive stage. Lingua 106: 23–79.
Plain vs. situated possession in Czech:
A constructional account
Mirjam Fried
1. Introduction
The focus of this paper is the kind of possession that can be identified
broadly as a time-stable relation that is presupposed (not asserted) between
two entities, a possessor (PR) and a possessum (PM). It is well known that
this relation can be expressed in a number of ways even in a single lan-
guage, let alone cross-linguistically; what still remains to be worked out in
sufficient detail is the exact nature of the variation and the relationships
among the variants. My purpose here is (i) to identify the factors that help
differentiate between syntactically distinct expressions of this kind of pos-
session, and (ii) to propose a way of representing the patterns in a network
of grammatical constructions organized around their shared features. The
illustrative material comes from authentic Czech usage, both written and
spoken, as attested in the Czech National Corpus (CNK).1
Heine (1997: 143) labels the presupposed, time-stable possession as “at-
tributive possession”, so categorized in contrast to “predicative possession”,
and this categorization includes characteristic differences in syntax as a
crucial criterion: phrasal syntax for the former, clausal for the latter. This
distinction, though, is too general to give us a realistic picture of the variety
of possessive expressions. The non-clausal expressions tend to come in
several distinct formal variants which represent a rather diverse set, both in
their form and their function. I will examine a subset of such variants in
Czech and offer a functionally and cognitively oriented analysis that is mo-
tivated by three general questions: (i) What kind of linguistic knowledge is
necessary for native-like production and comprehension of these possessive
patterns? (ii) How can the speakers’ knowledge and understanding be ade-
quately represented? And (iii) can the representation help us make more
precise claims about attributive possession as a linguistic category, with
implications beyond accounting for the Czech facts?
1
The material used here comes from roughly 400 million-word written corpora
(SYN2000, SYN 2006PUB), supplemented by a sample from Czech language
and literature (LITERA, SYNEK), and a 2.5 million-word spoken corpora (PMK,
BMK, ORAL2006).
214 Mirjam Fried
Some attributive possessive expressions are exemplified in (1)–(3); all

of these patterns are familiar and found in many languages. The examples
in (1) are two variants of an adnominal PR in a NP headed by the PM.2 (1a)
is an agreeing modification structure, in which the PR is marked by posses-
sive morphology and agrees with the noun in case, gender, and number,
like any pre-nominal modifier in Czech. In (1b), the PR is in the genitive
and obligatorily follows the head noun. Conceptually, both of them could
be treated as cases of “possessor specification” (Heine 1997: 167), but I will
refer to them collectively as Genitive PR (GP), choosing a more traditional
and familiar label. In (2), both the PM and the PR are syntactically inde-
pendent NPs and the PR status is associated with the dative marking on the
PR. This pattern is known as “external possessor” (e.g. Payne and Barshi
1999) and is typically associated with PR affectedness; I will, therefore, call
it Affected PR (AP). The example in (3) shows a strategy in which the PR
is not explicitly identified.3 For easier orientation, the possessive expres-
sions will be in bold throughout the paper.4
(1) Genitive PRs (GP)5

a. Pre-nominal
Natali-in-y rodie ty maj
Natalie-PR.F-NOM.PL parents.NOM they have.PRES.3PL
furt {ákej státní svátek , tak sou doma poád.}
at.all.times
‘Natalie’s parents, they have {state holidays all the time, so they’re
always at home}’ [PMK137;148304]
2
The distribution of these two forms is conditioned morphologically: the post-
nominal genitive is obligatory if the PR is morphologically neuter (1b), or a
multiword NP, or a plural noun, or is modified by a relative clause. In other
cases, the agreeing pre-nominal form is the neutral choice.
3
There are additional variants, including a morphological adjective as an expres-
sion of PR status. I will not be concerned with these marginal cases here but the
general approach provides ways for incorporating such patterns as well.
4
A note on presenting the data: when the surrounding text is necessary for clearer
understanding, it will be enclosed in curly brackets {} and left without interlinear
glossing.
5
Abbreviations in examples: ACC ‘accusative’, AP ‘affected PR’, AUX ‘auxiliary’,
DAT ‘dative’, F ‘feminine’, GEN ‘genitive’, GP ‘genitive PR’, IMP ‘imperative’,
INF ‘infinitive’, INS ‘instrumental’, LOC ‘locative’, M ‘masculine’, N ‘neuter’,
NEG ‘negative’, NOM ‘nominative’, PRES ‘present tense’, PST ‘past tense’,
RF ‘reflexive’, PR ‘possessor’, SG/PL ‘singular/plural’, and SPRL ‘superlative’.
Plain vs. situated possession in Czech: A constructional account 215
b. Post-nominal
Ohneme noiky dítte v kolenou
bend.PRES.1PL leg.ACC.F.PL child.GEN.SG.N in knee.LOC.PL
‘We’ll bend the baby’s legs at [the] knees’ [SYN2000;34388742]
(2) Affected PR (AP)

spravil nám i auta
fix.PST.SG.M 1.PL.DAT also car.ACC.PL
‘he even fixed our cars’ [SYNEK;1598227]
(3) Implicit PR:

ten pán ztratil brejle {a nemohl je najít}
that man.NOM.SG.M lose.PST.SG.M glasses.ACC
‘that man lost [his] glasses {and he couldn’t find them}’
[LITERA;1680536]
My primary concern will be the relationship between (1) and (2); reference
to the Implicit PR variant is necessary mainly as a background against
which the explicit patterns can be studied. The example in (3) is simple:
even the minimal context given here is enough to suggest that the owner of
the lost glasses is the subject referent, rather than some other person. The
preferred interpretation has to do with what we know about glasses as
common personal possessions. Other times the implicitness allows a greater
range of possible interpretations, especially if the sentence contains more
than one potential candidate for the PR status, including direct discourse
participants, or if the PM is something other than a body part. Speakers
know to infer the appropriate configuration based on conventional expecta-
tions about possible possessive relationships vis-à-vis particular context.
However, if it is the case that a possessive relation can be (and in Czech
very often is) left implicit, then we have to ask what reasons speakers might
have for choosing the explicit options and particularly, what – if anything –
conditions the choice of (1) vs. (2).
I will be concerned only with the second (easier) question here, pursuing
the hypothesis that the differences should revolve primarily around the
(in)alienability of the PM and the affectedness of the PR as two independ-
ent, competing factors; the hypothesis is motivated by the patterning found
in many other languages. However, the analysis will demonstrate that the
interaction between affectedness and (in)alienability is systematically more
complex than this: specifically, in cases of conflict, affectedness takes
precedence over inalienability in licensing the AP form. This leads, among
216 Mirjam Fried
other things, toward an absolute prohibition on PM (of any kind) as a tran-

sitive subject and almost equally strong prohibition on subjects of active
intransitive verbs. Given that similar syntactic restrictions on the grammati-
cal role of the PM are cross-linguistically common in external possessor
constructions (Payne and Barshi 1999), the present analysis highlights a
semantic explanation that may apply well beyond the Czech facts. Finally,
mapping out the interaction between affectedness and (in)alienability also
provides a generally applicable approach to incorporating peripheral cases
in which a non-possessive expression may acquire a possessive construal.
In order to keep the analysis reasonably focused, I will take as my start-
ing point a relatively narrow definition of the concept of possession. I ex-
clude the broad sense of belonging in part-whole relations in which the
whole is not an animate entity (Heine’s 1997: 35 “inanimate possession”),
but my conception is not quite as narrow as Taylor’s (1989: 202–203) no-
tion of prototypical ownership. Like Taylor, I will take possession to be an
“experiential gestalt”, which presupposes that the PR is a sentient being and
has specific reference. However, I will consider a broader range of possessa
than his prototype allows (“specific concrete things”). Finally, I will also
work with the notion of possessibility hierarchy. There seems to be no firm
consensus about the exact shape of a universally applicable hierarchy; the
main points of variation seem to be the relative placement of kin relations,
inherent attributes, and clothing, and their status evidently depends on par-
ticular languages. For the purposes of this study, I find it sufficient to as-
sume a very general scale along the lines of Payne and Barshi 1999: body
parts > kin > close alienable > distant alienable. However, a close analysis
of the Czech AP (section 3) will suggest a more refined scale, partially
overlapping with Tsunoda’s (1995: 576) Possession Cline.
By examining the range of factors that motivate the distribution of the
GP and AP patterns, I will argue that the distinction goes far beyond reduc-
ing the issue to treating the two strategies as two formal variants of a con-
ceptual schema called “possessive specification” (Heine 1997: 167), each
motivated by a particular instantiation of an existential event schema – Goal
vs. Genitive (p. 47).6 Both of these variants merely assert the existence of a
presupposed possessive relation, each in a different syntactic form. The
Genitive schema is associated with the pattern [PR’s PM]), while the Goal
schema, here corresponding to the AP pattern, takes the form [PM (exists)
6
Only these two schemas in Heine’s typology are relevant here. Unlike, for ex-
ample, Russian or Romanian, Czech does not have any locative-type patterns
[PM (is) at PR] expressing possession.
for/to PR]. Both schemas are said to correlate with permanent inalienable
possession. Neither of these broad schemas thus helps explain the fact that
the Czech GP and AP have very different distributions in actual discourse
and that the difference in form between (1) and (2), as well as some special
syntactic constraints associated with (2) but not with (1), follow from a
number of co-occurring properties: the inherent meaning of the PM and the
PR, the involvement of the PR in the event expressed by the clause, verb
semantics, and information flow. I will also suggest that the presence or
absence of some of these properties correlates with particular speech situa-
tions.
The two forms will thus be best treated as grammatical constructions in
the sense of Construction Grammar (see especially Fillmore 1989; Croft
2001; Fried and Östman 2004), and I will also draw on the notion “cognitive
frame” (Fillmore 1982) for incorporating the possessive relation in the con-
structional representations. I will argue that each possessive variant consti-
tutes a semantically and pragmatically distinct pattern – a conventional
cluster of semantic, pragmatic, and morpho-syntactic properties – and that
each pattern represents a functional prototype within a network of possibili-
ties for expressing attributive possession.
The paper is organized as follows. The semantic and pragmatic proper-
ties of the PR and the PM are discussed in sections 2 and 3, respectively,
addressing the issue of a possessive prototype. Section 4 studies the interac-
tion between the possessive relation and the structure and meaning of the
sentence in which it occurs. All of this comes together in section 5, in which
I present the patterns (1) and (2) as two grammatical constructions that oc-
cupy overlapping domains within the functional space of attributive posses-
sion. I illustrate the ways in which the constructions in the network may
interact both with each other and with the possessive prototype, suggesting
a more systematic account of the ways in which the prototype can be ex-
tended to more peripheral instances. Section 6 briefly concludes the paper.
2. Semantic and pragmatic features of the PR
The inherent semantics of a quintessential PR does not predict, by itself, any-

thing about the linguistic form and behavior of the noun phrase that encodes
the PR; the referents in (1)–(3) are all specific human beings. Moreover,
both AP and GP are sometimes possible for a given possessive relation, as
shown in (4); note, however, that they are not quite synonymous.
218 Mirjam Fried
(4) a. Ped pti lety nám zemel otec.

before 5 years 1PL.DAT die.PST.SG.M father.NOM.SG.M
{Matka se zhroutila. Z nemocnice ji propustili, ale není v poádku.}
‘Five years ago, father died. {Mother fell apart. She’s back from
the hospital, but she’s not in good shape.}’ [SYNEK;2279741]
b. Ped pti lety zemel ná otec.
before 5 years die.PST.SG.M 1PL.PR.NOM.SG.M father.NOM.SG.M
‘Five years ago, our father died.’
Other times, though, only one form is available. For example, the AP in
(5a) cannot be replaced by GP (5b) and the examples in (6) demonstrate the
reverse:
(5) a. V uích nám hvízdal vítr

in ear.LOC.PL 1PL.DAT whistle.PST.SG.M wind.NOM.SG.M
‘Wind was whistling in our ears.’ [SYN2000;2062266]
b. *V naich uích hvízdal
in 1PL.PR.LOC.PL ear.LOC.PL whistle.PST.SG.M
vítr
wind.NOM.SG.M
(6) a. Z njakého […] dvodu ho

from some.GEN reason.GEN 3SG.M.ACC
její vlasy nesmírn vzruovaly.
3SG.F.PR.NOM.PL hair.NOM.PL immensely excite.PST.PL
‘For some [unknown] reason, her hair excited him enormously.’
[SYNEK;941353]
b. *… vlasy jí ho nesmírn vzruovaly
hair.NOM.PL 3SG.F.DAT 3SG.M.ACC immensely excite.PST.PL
Similar effects have been noted for various languages (e.g. O’Connor 1994;
Croft 1985; Berman 1982; Chappell and McGregor 1995; Manoliu-Manea
1995; Payne and Barshi 1999), including Czech (e.g. Zimek 1960; Pit’ha
1992; Fried 1999a, 2008), and the subsequent discussion will show that the
contrast has to do both with the involvement of the PR in the reported event
and the relative communicative prominence of the PR and PM in a given
piece of discourse.
2.1. Affectedness of PR
The relationship exemplified in (5) illustrates the fact that inalienable pos-
session overwhelmingly prefers AP in Czech, to the point of leaving it as
the only expressive option in some cases (i.e., if the speaker has a reason to
mention the PR explicitly in the first place, instead of opting for Implicit
PR). The case in (6), then, appears to be an ‘exception’ since the only pos-
sibility there is GP, even though the PM is a body part like in (5).7 How-
ever, this apparent contradiction is fully motivated and I will revisit it in
sections 4 and 5. First let us consider a less categorical situation: examples
where AP appears to be replaceable by a GP form. The expressions in (a)
are corpus attestations of AP, the (b) examples are their GP counterparts
(constructed).
(7) a. {Pak pila prvodí} a protípla nám

and punch.PST.SG.F 1PL.DAT
lístky {a cesta doplynula ve ví pohod}
ticket.ACC.PL [SYNEK;2261787]
‘{Then the conductor came} and punched our tickets {and the rest
of the trip went all smoothly}.
b. protípla nae lístky
punch.PST.SG.F 1PL.PR.ACC.PL ticket.ACC.PL
(8) a. “vsate se, e se tomu ulpas-ovi policajtsk-ému

bet.IMP RF that RF that.DAT nitwit-DAT cop.ADJ-DAT
vysmolím ped dvee na schody
poop.PRES.1SG in.front.of door onto staircase
{a jet si s ním pitom budu povídat!}” [SYN2000;87610]
‘“you can bet that I’ll take a dump on the steps in front of the nitwit
cop’s door {and keep chatting with him in the process, too!}”’
b. … e se vysmolím ped dvee toho ulpas-a policajtsk-ého
door that.GEN nitwit-GEN cop.ADJ-GEN
7
Whatever observations are made in this paper about the behavior of body parts
as the PM, they apply to all specific candidates for body part status. Czech does
not make any grammar-coded distinctions between different types of body parts.
220 Mirjam Fried
(9) a. Kdy jim vzali peníze mnovou

when 3PL.DAT take.PST.PL money.ACC currency.ADJ.INS .SG
reformou,
reform.INS.SG
{prodávali koberce, perky a obrazy, jen aby mohla zstat doma.}
‘When [the government] took their money in the currency reform,
{they kept selling rugs, jewelry and paintings, anything [to make
it] possible for her to stay at home.} [SYNEK;653077]
b. Kdy vzali jejich peníze
when take.PST.PL 3PL.PR money.ACC
The issue in (7)–(9) is not the incompatibility between the PM and the GP
form, as is the case in (5). The problem in (7)–(9) is the degree to which
the PR is involved in the depicted events. The example in (7) is the most
flexible one: GP (7b) implies that the conductor punched some tickets that
belonged to us but were not necessarily related to our riding the train. The
AP (7a), in contrast, presents the tickets as necessary for the ride, we were
holding them in our hands, handed them over, and then took them back
from the conductor, to hang onto them. The GP form is in principle possi-
ble but somewhat odd, given what we know about the usual ways in which
passengers and conductors interact. A GP form in (8) would be even more
problematic since its implication would be contradicted by the subsequent
coordinated clause: GP (8b) would be felicitous whether the cop is present
or not when the speaker is squatting in front of the cop’s door; the automatic
interpretation actually would be that the cop is not around. Yet the point of
the speaker’s bet is that he will be simultaneously having a conversation
with the cop. The contextual incompatibility is still stronger in (9): GP (9b)
would be possible whether or not the PR was alive during the currency re-
form, while AP (9a) is felicitous only if the PR was alive. The AP empha-
sizes this state of affairs, which makes it the only coherent choice for the
follow-up about the consequences the PR suffered (selling off property to
make ends meet).
To summarize, the two forms display a systematic division of labor that
can be related to Bally’s (1995/1926) notion of personal domain or (subjec-
tively applied) indivisibility. GP expresses plain possession in the broadest
sense, where the concept of indivisibility plays no role. By contrast, AP casts
the possession relation as something that is relevant to the PR in a particular
way, as something in his sphere of interest beyond just the fact of being
owned. AP signals that the PR is being affected (positively or negatively)
by something that affects the PM (cf. also Chappell and McGregor 1995).
In this light, it is not surprising that certain types of possessa strongly prefer
AP, in contrast to GP: the tighter the possessive relationship, the greater the
chance that manipulating the PM will directly affect the PR. Body parts are
the extreme on the continuum since they are truly inseparable from the PR,
and hence the unacceptability of (5b). The GP form can only be interpreted
in one way: the ears are not attached to the PR’s body (are not, that is, bona
fide body parts) but are some ear-like objects, physically detached from any
bodies.
This extreme restriction applies only to the pre-nominal GP, though; as
shown in (1b), the post-nominal genitive is sometimes attested with body
parts and the form maintains the true body part reading. We thus have to
conclude that the relative flexibility between allowing both AP and GP ex-
tends all the way to the top of the possessibility hierarchy. Nevertheless, a
difference in form (AP vs. GP) always correlates with a shift in interpreta-
tion, no matter what the PM. For example, the choice of GP with a body
part in (1b) is motivated by the type of discourse and the focus of attention.
This token is taken from instructions about how to use a rectal thermometer
to take a baby’s temperature. The communicatively relevant issue here is
how to manipulate the baby’s position so that the person succeeds in what
needs to be done; the author of the instructions is not concerned with what
(dis)comfort it may bring to the baby, which would be the only possible
reading if an AP form were chosen. A similar analysis applies to example
(4), in which a kinship relation is cast in a different light depending on
whether AP or GP is used. The GP in (4b) would imply that the speaker
was estranged from, or at least not very close to, his father and therefore
unaffected by his death, in contrast to (4a), which unequivocally implies that
the father’s death had tangible consequences, as is clear from the subsequent
text.
Two generalizations emerge from these facts: (i) inherent (in)alienability
is not a good predictor of AP vs. GP encoding (as also noted explicitly by
Bally 1926/1995; Chappell and McGregor 1995) and (ii) the PR in the AP
form plays a special role in the event expressed by the predicate, while no
special status is associated with the GP variant.
2.2. Information structure
As a syntactically independent NP, the dative PR is free with respect to its

position in the clause relative to the PM. It can, therefore, participate in ar-
222 Mirjam Fried
ticulating topic-focus relations within the possessive relation itself, and

thereby accommodate the nuances of structuring information flow, which in
Czech is expressed through word order (Firbas 1966; Dane
1974; Sgall et
al. 1986; Grepl and Karlík 1998). On the other hand, the PR position within
GP is restricted: the pre-nominal PR cannot be very easily separated from
the PM (except under very special contextual circumstances) and the geni-
tive is obligatorily fixed in the post-nominal position. Consequently, GP is
dispreferred or outright impossible in certain contexts simply for reasons of
information structure, even if it might be appropriate on semantic and syn-
tactic grounds. Consider (10):
(10) a. {Za nco se stydl,} bál se podívat lidem

fear.PST.SG.M RF look.INF people.DAT
do tváe, {kdy jim nco nabízel}
in face.GEN.SG.F
‘{He felt embarrassed about something,} he was afraid to look
people in their FACES {when he was selling them something}’
[LITERA; otapavel]
b. bál se podívat do tváí lidí,
fear.PST.SG.M RF look.INF in face.GEN.SG.PL people.GEN
‘…he was afraid to look in the faces of PEOPLE’
The example in (10a) gives discourse prominence to the faces, as indicated

by the small caps in the English translation. The point of the sentence is to
describe the subject’s reluctance to face people. Replacing AP by a GP ex-
pression (10b) would shift the focal status onto the genitive PR, simply by
virtue of its phrase-final position, and this reconfiguration would sound
very odd in the context of this passage. (10b) creates the expectation that
lidí ‘of people’ is to be understood as contrastive focus: either in contrast to
the faces of other entities mentioned earlier, or as the head of a restrictive
relative clause that would have to follow (‘he was afraid to look in the faces
of [those] people who he was selling something to, but not other folks’).
Such readings are clearly unintended here, and using AP is a general strategy
for making the PR and the PM available for structuring information flow
independently of each other.
The infelicitous character of (10b) cannot be easily attributed to the dis-
embodiment reading associated with pre-nominal GP, for two reasons.
First, the GP pattern as such does not imply disembodiment, as we see in
(5a); in that sentence, the PR’s hair is unambiguously attached to her head.
Moreover, since the post-nominal GP is obligatory with certain grammati-

cal types of PRs, listed in footnote 1 (here the issue is the plural of the PR
noun), the inherent semantics of the PM becomes irrelevant. And second,
since the information structure value of the PR can be freely manipulated in
the AP, it is possible to form a discourse-structure equivalent of (10b) as
well, shown in (10c) below:
(10) c. bál se podívat do tváe lidem

fear.PST.SG.M RF look.INF in face.GEN.SG.F people.DAT
‘…he was afraid to look in the faces of PEOPLE’
This AP form is subject to exactly the same interpretive restrictions as

(10b), namely, putting the people into a focus role, whether contrastive or
plain. The only difference is that the post-nominal GP cannot be preposed,
under any circumstances.
To summarize, AP gives the PR a certain prominence that GP does not
provide. However, this kind of prominence must be understood as an event-
based prominence (i.e., treating the PR as an affected event participant, as
opposed to just an owner), independent of a discourse-based prominence
(i.e., assigning information structure roles within the possessive relation-
ship), which can be manipulated separately. I will revisit the implications of
this feature in section 4.
3. Semantics of the possessum
As illustrated in (11), the head noun in the GP pattern can be semantically

anything: concrete, abstract, animate, inanimate.
(11) Han-in-y ruce/sestry/kamarádky/knihy/názory/povinnosti

Hana-POSS.F-NOM.PL hands/sisters/friends/books/opinions/duties
This semantic freedom suggests that the GP pattern simply marks a very
general relationship between two entities, which can be conceptualized as
denoting a unit of sorts. This conceptualization accommodates possession
(including inalienable) in the narrow, experiential sense as well, but the
form evidently is not restricted to encoding truly possessive relations. The
semantic and functional breadth is corroborated by the fact that the same
form is found not only in combinations such as (11), all expressing (at least
loosely understood) possession, but also in various common ‘genitive’
224 Mirjam Fried
functions (not discussed here) or in the ‘syntactic’ function, i.e. encoding

verbal arguments in nominalized expressions. The latter is exemplified in
(12), where the pre-nominal possessive form Janovu ‘Jan’s’ marks the
agent role of the noun pomoc ‘help’, not any PR of help.
(12) dkoval mi za Jan-ov-u pomoc

thank.PST.SG.M 1SG.DAT for Jan-POSS.M-ACC.SG.F help.ACC.SG.F
‘he thanked me for Jan’s help’ [LITERA;3101771]
The AP pattern is clearly different from GP in that AP does not cover the
syntactic function shown in (12); we cannot create an AP paraphrase of
(12). However, paraphrasing the expressions in (11) is possible, at least in
principle. Thus, considering the AP examples in (7)–(9) above and (13) be-
low, AP might not appear to present a dramatically different picture from
GP with respect to the kinds of possessa it permits, along the full possessi-
bility hierarchy:
(13) a. Komolil jména lidem pi nejrznjích

distort.PST.SG.M name.ACC.PL people.DAT at SPRL.various
píle itostech [LITERA;375950]
occasion.LOC.PL
‘He mangled people’s names on all kinds of occasions’
b. kdy jsme byly tomu strejkovi na tom
when AUX.1PL be.PST.PL that.DAT uncle.DAT on that.LOC
pohbu
funeral.LOC
‘when we were at that uncle’s funeral’ [PMK441;786749]
c. jak se nám rozpadá spolen
as RF 1PL.DAT fall.apart.PRES.3SG common.NOM.SG.M
stát
state.NOM.SG.M
‘as our shared country is splitting [into two]’ [SYNEK;1953603]
It has been noted in the external PR research that external PRs tend to co-
occur with possessa at the high end of the possessibility hierarchy, with dif-
ferent cut-off points in different languages (Payne and Barshi 1999; Payne
1997a, b). We have now seen that the Czech AP roughly follows the same
scale of preferences but it is worth checking this general tendency against
the semantic range attested in actual discourse. The sample in (2), (7)–(9),
and (13) can be easily extended by additional possessa found with AP in

CNK: pytel ‘sack’, hraky ‘toys’, vodovod ‘water-pipe’, svt ‘world’, cesta
‘journey’, sebevdomí ‘self-confidence’, smlouva ‘contract’, ivot ‘life’,
svatba ‘wedding’, reforma ‘reform’, králík ‘rabbit’, k ‘horse’, etc. We
can see that even this fairly randomly assembled list covers the whole spec-
trum of the possessibility hierarchy and contains both concrete and abstract
entities. At the same time, it must be acknowledged that abstract nouns
generally do have limited currency in the AP pattern; it is impossible to ei-
ther find or construct coherent examples with nouns such as hodnota
‘value’, krása ‘beauty’, kvalita ‘quality’, chyba ‘mistake’, moudrost ‘wis-
dom’, vmluva ‘excuse’, nadazenost ‘superiority’, while any of these
would be perfectly compatible with the GP pattern.
On a closer look, however, the possessa that are found in the AP pattern
fall into several conceptual categories, which can be properly identified in
relation to conventional cultural expectations, rather than in terms of general
linguistic categories such as alienability, concrete/abstract, animate/inani-
mate, etc. The corpus material suggests a number of salient semantic
classes of items possessible either individually or collectively by human
beings and construable as such within the speakers’ cultural experience.
The list in (14) does not necessarily represent a real hierarchy in Czech but
it is of course possible to trace certain parallels with the linguistically de-
fined possessibility hierarchy that is more explicitly attested in other lan-
guages (cf. Tsunoda 1995):
(14) a. things that are part or features of self (body parts; name, title;
speech; life; doubt, memory, intention, self-confidence, right to
decide, etc.)8
b. members of ‘family’, understood broadly as a culturally established
unit of shared domestic life (kinship relations; pets and other do-
mestic animals)
c. garments and their parts
8
This group can also be conceptually further decomposed into more specific
categories, some of which are also suggested by Tsunoda (1995): parts of body,
representations of self, cognitive and speech qualities, manifestations of self-
awareness, etc. Such subcategorization, however, leaves the main point of the
generalization in (14a) for Czech unaffected, since these additional conceptual
distinctions seem to have no grammatical reflexes.
226 Mirjam Fried
d. environment perceived as essential to our existence, including

dwellings and their parts (world; house, door, plumbing, bathroom;
prison cell; backyard)
e. objects useful in an individual’s daily life (cars, toys, flashlights,
money, tickets, guitar strings)
f. common activities and established rituals (journey, wedding, fu-
neral, graduation, education, vacation, holiday)
g. social and / or political organization (state/country, constitution,
reform)
This list makes it evident that there is no blanket prohibition on the usage of
abstract nouns. Moreover, the attested combinations cannot be simply re-
placed with GP without changing their meaning, as we have seen in section
2. It is important to note, though, that the abstract concepts that co-occur
with AP tend to come from particular semantic domains and are experien-
tially based: they have to do with mental or physical states and cognitive
capacities typical of human beings (14a) or with personal and social rituals
(14f). The latter can be easily extended further into more specialized con-
texts, in which a possessive construal can apply to abstract concepts that are
inherent in various types of social institutions in general, such as political
organization and public life (14g), and where the PR is, therefore, a collec-
tive, not an individual. Granted, the categories in (14f–g) do not represent
the same sense of ownership/possession as the classes in (14a–e) and as
such are somewhat removed from the central, prototypical definition of
possession. Nonetheless, they are not the only source of abstract nouns in
the AP pattern and so they do not invalidate the general observation about
the possessibility of abstract entities.
On the other hand, we also find a number of both abstract and, espe-
cially, concrete entities that do not seem to fit easily into any of the catego-
ries in (14) at all: branka ‘goal/score [in soccer]’, branká ‘goalie’, pacient
‘[hospital] patient’, tramvaj ‘street car’, devo ‘wood’, smlouva ‘contract’,
etc. Taken out of context, these items seem entirely random and might sug-
gest, yet again, that the AP pattern is not that different from GP in the range
of possessa it permits. However, ‘context’ is the operative word here; these
items are always invoked and interpreted within specialized contexts (e.g.,
medical care, transportation, commerce), sometimes even in distinct types
of discourse (sports reporting), which frame possessive relations in ways
specific to those contexts and establish the conditions that allow the PR to
be cast as affected in particular ways. No such contextual framing is re-
quired for the use of the GP variants. And it is also worth pointing out that
these context-sensitive possessive relations are, therefore, different from

the possessa in (14f–g): those two categories are relatively independent of
special contexts and can be understood as being inherent (and in some ways
universal) parts of common human experience. By contrast, the context-
dependent possessa are more transient and non-essential.
We thus must conclude that the distribution of GP and AP in actual us-
age is subject to various preferences and sometimes even inviolable con-
straints with respect to the type of PM. Relevant generalizations can be
formulated as follows: (i) body parts cannot occur in the pre-nominal GP
pattern and, instead, require AP; (ii) the range of possessa that can naturally
co-occur with AP is semantically more restricted than with GP; (iii) the re-
strictions can be meaningfully articulated only in terms of culturally based
clusters of concepts, sometimes in combination with specific types of dis-
course contexts, not in terms of purely linguistic distinctions, as has been
traditionally accepted; and (iv) the distribution of GP is more restricted in
terms of information structure.
4. Constraints on verb semantics and clause structure
4.1. Verb semantics
The semantic and pragmatic differences between AP and GP outlined in the

preceding sections have different consequences for the types of predicates
each form co-occurs with. Predictably, the distribution of GP is entirely in-
dependent of any verb meaning, since GP expresses pure possession. By
contrast, the dative PR’s involvement in the event depicted by the verb
makes AP sensitive to the semantic type of the verb. The affectedness of
the PR, which is tied to the affectedness of the PM, presupposes events that
are semantically compatible with affectedness. This is, indeed, a correlation
well attested with the AP pattern. AP strongly prefers ‘contact’ predicates,
such as spravit ‘fix’ (2), protípnout ‘punch through’ (7a), istit ‘clean’,
postílet ‘shoot down’, narovnat ‘straighten’, roztrhnout ‘tear’, pivázat ‘tie
to st.’, hodit ‘toss’, umt ‘wash’, as well as other affective verbs, such as
komolit ‘distort’ (13a), niit ‘ruin’, osv it ‘refresh’, etc.
The corpus material shows that another very well-represented class con-
cerns verbs of removing, such as vzít ‘take away’ in (9), se rat ‘gobble up’,
vylouit ‘expel’, krást ‘steal’, ztratit ‘lose’; these verbs presuppose a pos-
sessive relation and are thus particularly good candidates for accommodat-
ing affected PRs, by expressing possession-removal situations. Among in-
transitives, verbs of states or spontaneous processes without any identifiable
228 Mirjam Fried
instigator are highly compatible with AP semantics and richly attested in

the corpus as well: zemít ‘die’ in (4), rozpadat se ‘fall apart’ in (13b), zítit
se ‘collapse’, rozbít se ‘break down’, padat ‘fall’, zmizet ‘disappear’,
mrznout ‘freeze’, zeílet ‘go crazy’, cukat ‘twitch’, tást se ‘shiver’, smrdt
‘stink’, etc.
Predicates that do not fit these semantic profiles are very hard to come
by. The following, showing the syntactically transitive verb vidt ‘see’, is
only an apparent counterexample to this generalization:
(15) {není to proto, e ti drbani jsou nadáni mimoádnou jasnozivostí a}

vidí lidem do kapes
see.PRES.3PL people.DAT into pocket.GEN.PL.F
‘{[if a businessman gets robbed…] it’s not because these lowlifes are
endowed with extraordinary clairvoyance and} see into people’s
pockets’ [LITERA; frybort2]
First of all, the use of AP in (15) can be explained by appealing to informa-

tion structure, just like we saw in (10): the use of the post-nominal GP
kapes lidí ‘pockets of people’ (the only GP possibility here due to the plural
of the PR noun) would shift the focus away from the pockets, which would
make the form contextually problematic. The context clearly centers on
what happens to the owners of the pockets; the passage gives explanations
about the circumstances under which people get robbed. But equally impor-
tant is the verb itself: it is not used here in its purely perceptual meaning of
‘having a visual experience’, which would require an accusative-marked
perceptum. Instead, the directional phrase do kapes ‘into the pockets’ indi-
cates a more active reading along the lines of ‘look inside’. The sense of
pure perception is completely incompatible with AP, as shown in (16); this
combination is incomprehensible and is equally impossible even with body
parts as PM:
(16) *vidí lidem kapsy

see.PRES.3PL people.DAT pocket.ACC.PL.F
‘(s)he sees people’s pockets’
The corpus data thus confirm the conclusions of previous studies, namely,
that non-affective verbs generally require a GP form. The following exam-
ples further illustrate the prohibition on APs with such verbs, in contrast to
the attested GP, both with transitive (17) and active intransitive (18) predi-
cates:
(17) a. ted’ znám et její rodie

now know.PRES.1SG also 1SG.F.PR parents.ACC
‘now I know her parents too’ [PMK143;519767]
b. *ted’ jí znám et rodie
now 3SG.F.DAT know.PRES.1SG also parents.ACC
(18) a. {N.P. mla dokonce kurzy v jazykovce zadarmo,}

kdy ta její máma tam
when that.NOM.SG.F 1SG.F.PR mom.NOM.SG.F there
pracovala.
work.PST.SG.F
‘{N.P. could even take courses in the language school for free}
since her mom was working there’ [oral2006; 175302]
b. *kdy jí tam ta máma
when 3SG.DAT.F there that.NOM.SG.F mom.NOM.SG.F
pracovala.
work.PST.SG.F
The impossibility of the (b) forms cannot be explained by the semantics of

the PM; recall that kinship relations preferentially select AP (examples (4a),
(13b)). Instead, it must be attributed to the AP’s semantic incompatibility
with predicates that do not allow the PR’s affected role in the depicted event
to be properly integrated with the meaning of the predicate. GP, on the other
hand, does not place any such requirement on the verb meaning and be-
comes, therefore, the only available alternative. Notice also that the affect-
edness must be facilitated by the predicate semantics, not just a potentiality
motivated by the inferences suggested by the broad context. This difference
is illustrated in (15) and (18) above. In (15), the relevant verb meaning is
signaled overtly by the morphology of the second argument. The case in
(18), with the intransitive verb pracovat ‘work’, is more subtle. An AP
variant (18b) cannot be substituted even though the general context clearly
supports the inference that the PR (i.e., the daughter, identified as N.P.) is
positively affected by the fact that her mother works in a particular place;
N.P.’s benefit is the central concern of this utterance. I will revisit this
problem in section 5, to consider potential counterexamples.
The co-occurrence patterns attested with AP vs. GP thus lead to two gen-
eralizations: GP is fully independent of verb semantics, while AP can only
occur with predicates that are semantically compatible with affectedness.
230 Mirjam Fried
The latter can be formulated in terms of a hierarchy of preferences as in

(19), with a clear cut-off point (indicated by the double arrow >>) between
active intransitive and weakly transitive predicates; the latter are fully ex-
cluded from the AP pattern:
(19) {strongly transitive, non-active intransitive} > active intransitive >>

weakly transitive (Vs of perception and cognitition)
4.2. Syntactic constraints on the possessum
These verb-related semantic preferences alone cannot account for certain

additional facts about the AP pattern; we still need to examine the cases
exemplified by (6), here repeated as (20), which involve a particular syntac-
tic constraint on the PM. In (20), the main predicate, vzruovat ‘cause ex-
citement’, is affective and the PR is a pre-nominal GP, which should not be
possible with a body part PM; recall the discussion of (5). Yet, (20a) is the
only grammatical way of expressing the possessive relation in this sentence
and cannot be replaced by an AP under any circumstances (20b):
(20) a. Z njakého […] dvodu ho její

from some.GEN reason.GEN 3SG.M.ACC 3SG.F.PR.NOM
vlasy nesmírn vzruovaly. [SYNEK;941353]
hair.NOM.PL immensely excite.PST.PL
‘For some [unknown] reason, her hair excited him enormously.’
b. *… vlasy jí ho nesmírn vzruovaly
hair.NOM.PL 3SG.F.DAT 3SG.M.ACC immensely excite.PST.PL
Since GP is inside a self-contained NP, we can expect its distribution to be

quite free across all syntactic slots. The AP, on the other hand, follows the
cross-linguistically commonly attested pattern of excluding the PM of an
external PR from certain syntactic functions. In Czech, the PM can be a
transitive object (as in (2), (7a), (9a), (13a)), an oblique complement (as in
(5a), (8a), (10a), (15)), and an intransitive subject (as in (4a), (13c)). How-
ever, we see in (20b) that AP cannot appear as a transitive subject, even if it
is a body part. The same prohibition is evidenced in (1a) with a kinship PM,
here repeated as (21a): GP is the only way to express the possessive relation
here, while AP (21b) is ungrammatical. The same restriction extends to the
subjects of active intransitive verbs, as we saw in (18) above.
(21) a. Natali-in-y rodie ty maj furt

Natalie-PR.F-NOM.PL parents they have.PRES.3PL all.the.time
ákej státní svátek,
some state holiday.ACC.SG.M [PMK137;148304]
‘Natalie’s parents, they have some kinda state holidays all the time’
b. *Natali-i maj rodie furt ákej
Natalie-DAT have.PRES.3PL parents.NOM all.the.time some
státní svátek
state holiday.ACC.SG.M
In order to motivate the syntactic restrictions on the PM, we have to appeal

again to the affectedness requirement. The PM cannot appear in event roles
that presuppose agentive participants.
The conclusion we can draw from the syntactic behavior and the com-
binability with different predicate types is the following. There are two com-
peting factors for choosing the AP vs. the GP strategy – PR affectedness
and the (in)alienability of the PM. Generally, AP preferentially attracts in-
alienable possessa and for these, it is the only expressive possibility vis-à-
vis pre-nominal GP forms. However, the (verb-determined) affectedness of
the PR evidently takes precedence over inalienability, leading toward abso-
lute prohibition on possessa (of any kind) as transitive subjects and almost
equally strong prohibition on subjects of active intransitive verbs.
4.3. Phrasal syntax and the attributive possession
We have already noted the readily obvious syntactic difference between GP

and AP: GP is so called because the PR is a modifier syntactically depend-
ent on the head noun denoting the PM. The GP is thus a particular (seman-
tic) variant of a more general Modification construction, which is a cluster
of syntactic, morphological, and semantic properties defining a particular
type of attributive relationship between a noun and its modifier, following a
simple phrasal template: [ModPR NPM]NP for the pre-nominal form and
[NPM ModPR,GEN]NP for the post-nominal GP.
In contrast, it is hard to see on what definition of constituency we could
argue that the Czech dative-marked PR forms a single syntactic constituent
with the PM. The example in (22) makes it clear that the Czech AP is not
an adnominal structure. The corpus example in (22a) shows GP used in an
adposition to a noun phrase and (22b) demonstrates that the dative form is
prohibited; (22b) is severely ungrammatical:
232 Mirjam Fried
(22) a. {Ve védském Stokholmu ije od dtství, kdy tam}

jeho maminka, sestra Jiiny,
his mom.NOM.SG.F sister.NOM.SG.F Jiina.GEN.SG.F
emigrovala.
emigrate.PST.SG.F
‘{[He]’s been living in Stockholm, Sweden, since childhood,}
when his mom, Jiina’s sister, emigrated.’ [SYNEK;11289608]
b. *jeho maminka, sestra Jiin,
his mom.NOM.SG.F sister.NOM.SG.F Jiina.DAT.SG.F
emigrovala.
emigrate.PST.SG.F
The PR and the PM have to be analyzed as two autonomous NPs, both of

them incorporated as ordinary complements into the structure and the
meaning of a sentence. In fact, both the form (dative) and the semantics of
affectedness place AP in the family of other dative complements.9 These
include ‘thematic’ datives (required by the predicate as one of its argu-
ments, namely one that is in some non-manipulative, perhaps ‘mental’ way
affected by the event in which it participates) or adjuncts that are added to
verbs that do not require a dative-marked argument but can accommodate
an extra participant that can be construed as a beneficiary or a maleficiary
of the expressed event. A thematic dative is exemplified in (23) with the
verb pomoci ‘help’, showing also the fact that not all of these datives are
restricted to animate referents. An adjunct dative is in (24), with the verb
otevít ‘open’, and for these datives, the referent is necessarily human; the
datives are in bold.
(23) a. pomohl mnoha lidem a dtem

help.PST.SG.M many.DAT people.DAT and child.DAT.PL
‘he helped lots of people and children’ [SYNEK;10733898]
9
In terms of grammatical roles, the dative complements could be classified as
indirect objects, with the understanding that a two-place predicate in Czech can
encode its arguments using the pattern [subject – indirect object], e.g. pomoci
‘help’ in (23). Since it is more informative to refer to these complements either
through their case form (always dative) or their semantic role status, I choose
not to label them in terms of grammatical roles here. The point is that Czech da-
tive NPs are primarily motivated semantically, not syntactically in terms of
grammatical roles (cf. also Fried 1994).
b. trend, kterému pomohl pedevím

trend.NOM.SG.M which.DAT.SG.M help.PST.SG.M especially
{siln píliv zahraniního kapitálu}
‘a trend which was helped particularly {by a strong flow of foreign
investment} [SYNEK;8751696]
(24) Tento balón otevel lidstvu oblohu.

this balloon.NOM.SG.M open.PST.SG.M mankind.DAT.SG sky.ACC.SG.F
‘This balloon opened up the sky for mankind.’ [SYN2000;137864]
This connection has been explored in detail elsewhere (Fried 1999a, b) and
I will not revisit it here beyond summarizing that the AP shares the special
affectedness with other dative-marked roles (experiencer, recipient, benefi-
ciary/maleficiary) and particularly with the usage shown in (24), which I
label Dative of Interest (DI). Both DI and AP require the referent to be hu-
man and in both cases, the dative NP is an extra element not required by the
valence of the verb. However, the datives are incorporated into the sentence
as if they were full-fledged arguments. But AP adds a particular feature that
is absent in the other dative roles, including DI: the presupposed possessive
relation between the referent of the dative and something else in the sen-
tence. This is an important point that bears on the issue of framing posses-
sive relations in terms of broad cognitive schemas, such as the Genitive vs.
Goal-based schemas suggested in Heine’s (1997) typology. It is of course
true that at some very abstract level, all of the datives, including AP, relate
to the concept of goal-ness. But it is also very saliently the case that AP is
distinct from other goal-expressing roles and that the distinction is not just
a matter of interpretation but has systematic reflexes in semantic constraints
and syntactic behavior. In the following section, I will suggest a way of re-
organizing the conceptual space of attributive possession so that we can
capture more accurately the relationships – commonalities as well as con-
trasts – between the relevant semantic categories.
5. Czech GP and AP as distinct functional patterns
5.1. Plain possession vs. situated possession
The syntactic behavior only confirms that GP and AP cannot be taken simply
as two alternative and fully comparable expressions of attributive possession.
To capture the essence of the distinction, we can label the patterns as Plain
234 Mirjam Fried
Possession and Situated Possession, respectively. What they both share,

necessarily, is a human PR and a pre-existing possessive relation. We could
formalize this common background in the form of an interpretive frame,
POSSESSION, roughly outlined in (25). The frame represents a conven-
tionally established knowledge structure that schematizes the speakers’ un-
derstanding of prototypical ownership as a particular relation between two
entities (PR and PM), each of which can be associated with various proper-
ties that are motivated by both individual and generally cultural experience.
Minimally, the prototypical PR is schematized as a human being that is
highly individuated and referentially specific, and the PM as fitting some-
where along the possessibility hierarchy (FE stands for ‘frame element’):
(25) frame POSSESSION: FE#1 Possessor [+human]

FE#2 Possessum [on possessibility hierarchy]
For our purposes here, it is not crucial to dwell on all the additional details
of this general frame (such as, perhaps, listing a set of preferred properties
for each frame element). I will simply take (25) as a minimal way of repre-
senting speaker’s conceptual understanding of prototypical possession,
which then is shaped into different instantiations by elaborating on the spe-
cific characteristics of the PR and/or the PM. Crucially, though, this shaping
requires reference to both form and meaning, not just one or the other. The
conceptual prototype organized in the background frame does not, by itself,
say anything about the morphological or syntactic requirements associated
with the morphosyntactic strategies for expressing the possessive relation.
Those involve additional layers of constraints and I will argue that the best
way to capture the nature of those expressive strategies is to treat each pattern
as a grammatical construction in the sense of Construction Grammar, i.e. as
a conventionally expected association between the elements of this frame
and their linguistic expression.
5.2. Constructional organization of AP and GP properties
Let us start by summarizing the properties of the two frame participants

when used in the AP pattern, which is clearly the more constrained of the
two forms; the items in the list below can be read as being in contrast with
the corresponding features of the GP pattern:
(26) Situated Possession (AP):

Possessor
– is a participant in the depicted event (i.e., sensitive to V semantics);
– is (indirectly) affected by the event;
– is distinct from other datives (semantically, syntactically);
– can fully participate in information structure relations.
Possessum
– comes from a semantically defined and restricted class of items;
– must be affected by the depicted event;
– is prohibited in certain syntactic functions;
– can fully participate in information structure relations.
Much of the information summarized in (26) can be attributed to the affect-

edness requirement. For example, the semantics of the PM is predicted to
include entities that can be manipulated and that are inherently relevant to
human beings and the routines of their daily existence. It also follows that
AP attracts particularly strongly items high on the possessibility hierarchy.
At the same time, what we know about possessive relations, dative-marked
affectedness, or the possessibility hierarchy does not lead to the prediction
that any competition between the preference for inalienable possessa vis-à-
vis affectedness must be resolved, as a rule, in favor of maintaining a con-
sistent affectedness status (hence the prohibition on possessa as transitive
subjects and intransitive agentive subjects). Nor does it account for two ad-
ditional features: the fact that both the PR and the PM can participate in
information structure relations and that the dative form does not cast the PR
in a role simply identical with the role of other dative nominals.
There are thus properties of the AP pattern that have to be captured in
some other way, not just as simple consequences of the PR affectedness.
Which brings us to positing AP and GP as complex conventionalized clus-
ters of specific syntactic, morphological, semantic, and pragmatic proper-
ties, i.e., as distinct grammatical constructions. Using the Construction
Grammar formalism of Fried and Östman (2004), we can represent what
speakers have to know about AP as in Figure 1. The inner box represents
the head verb, indicating that AP is dependent on certain features of the
verb: the verb is specified as coming from a semantic class broadly charac-
terizable as ‘affective’ and as having at least one argument (labeled FE #2)
that will be expressed in syntax (as the PM); the latter is indicated by the
val(ence) statement. The rest of the information (i.e. everything in the out-
side, larger box) represents properties that are idiosyncratic to the AP pattern,
236 Mirjam Fried
and these have to do with integrating the possessive relation with the se-
mantic and syntactic structure contributed by the head verb.
First off, it is the AP construction as a whole, not the head predicate,
that supplies the link to the possessive relation, through the statement
[frame POSSESSION]), which brings along the PR and the PM and all the
background knowledge associated with the representation in (25). Second,
the constructional val(ence) statement in the outer box captures the fact that
the syntactic and semantic properties of the argument supplied by the head
verb are constrained by the requirement that its event role be semantically
non-agentive (thus ensuring the prohibition on the PM as a transitive or ac-
tive intransitive subject). The AP construction does not specify anything
about the PM’s semantic or syntactic role (both of these are determined by
the semantics of the head verb). However, the co-indexing (#2) between the
head verb, the constructional valence, and the POSSESSION frame indicates
that whatever this argument is in terms of its semantic and syntactic role in
the sentence, it will be interpreted as the PM. Third, the construction also
gives the PR an independent information-structure status relative to the
PM; this must be specified as the construction’s prag(matic) property. It
follows from this feature that both the PR and the PM are subject to articu-
lating regular information structure relations, which operate independently
of this particular construction (cf. the discussion in section 2.2); put differ-
ently, both the PR and the PM can appear as either a topical or a focal ele-
ment, independently of each other.
The rest of the representation contains features that are shared across AP
and DI, as is indicated by the inherit statement at the top; all the features
that come from this relationship are printed in gray, to show that these
specifications are not unique to the AP construction (strictly speaking, the
inherit statement would be sufficient and all the remaining gray-colored
information need not be spelled out). These features include the following.
First, the construction is syntactically a verb-based pattern (the syn(tax)
statement at the top). Second, the construction itself has additional valence
requirements, namely, the PR (#1) is in the dative (case DAT) and is inter-
preted as other datives of interest (expressed through the rel(ation) statement
specifying the semantic role, labeled ). And finally, it is the construction
as a whole that carries the overall meaning of Situated Possession, spelled
out in the sem(antics) statement, as a combination of the inherited DI se-
mantics and the POSSESSION frame.
Affected Possessor
inherit Dative-of-Interest
syn [ cat v, lex + ]
prag [ 'greater discourse independence of Possessor vis--vis Possessum' ]
sem [ 'circumstances described by the predicate have significant consequences
for the interested party (#1), whose referent is not in control of the event' ]
frame POSSESSION
FE #1 Possessor [human]
FE #2 Possessum [ ]
val { #1 [ rel [ 'interest' ]] , #2 [ rel [ 'non-agentive']]}

[ syn [ case DAT ]]
syn [ cat v, lex + ]

sem frame [affective]
FE #2[ ]
val {...#2[ ]...}
Figure 1. Affected PR construction
When we try to represent GP as a construction in its own right, its differ-

ences from AP come into sharp relief. To save space, I will focus on the
pre-nominal variant only. Figure 2 shows that the GP construction is a sub-
type of general Modification construction (through an inheritance link).
Again, all the inherited information is in gray and is only included for
clearer exposition: the GP construction is a phrasal structure of the category
NP (indicated by the syn(tax) statement at the top of the constructional box)
and consists of two constituents (the inner boxes), with the modifier pre-
ceding the head. The information specific to the GP construction concerns
two properties. One is the mapping between the POSSESSION frame partici-
pants onto the two structural daughters, through the co-indexing: the PR is
the modifier, the PM is the head. The other feature is the cat(egory) of the
modifier; I have abbreviated this by simply giving a list of morphological
classes (enclosed in curly brackets): the filler will be either a possessive
pronoun or the special nominal form derived by the possessive suffixes –v
and –in. Nothing more needs to be specified.
238 Mirjam Fried
Genitive Possessor
syn [cat n] inherit Modification
prag ['restrict reference of the noun (#2) by property exrpessed in #1']
sem frame POSSESSION
FE #1 Possessor [human]
FE #2 Possessum [ ]
#1 cat [ poss {pro; -UV / -IN }] #2 cat n

role modification role head
morphol. case #i [ ] morphol. case #i [ ]
number #j [ ] number #j [ ]
gender #k [ ] gender #k [ ]
Figure 2. Genitive PR construction (pre-nominal)
These representations express a hypothesis about the kind of knowledge

speakers of Czech must possess in order to produce and interpret a variety
of concrete linguistic expressions of attributive possession. One part of the
hypothesis is the generalization that the speakers’ native-like understanding
of such expressions involves the understanding of a rather intricate inter-
play between several layers of information, as shown in the figures. But the
constructional analysis and representation can enhance also our insight into
the way these grammatical patterns may be organized in larger networks of
distinct but partially overlapping patterns. Such an organization, in turn,
should allow us to be more precise about the circumstances under which
the possessive prototype can be extended in various directions. I will sug-
gest such a network and its implications in the next section.
5.3. Constructional network
If we take the concept of attributive possession as a type of functional space

that can be occupied by various expressions of this general possessive rela-
tion, we can organize all the features we have identified as relevant (seman-
tic as well as grammatical) in a network that shows precisely which features
are shared across individual patterns and which are specific to each pattern.
This is what we see in Figure 3. The shaded area in the middle represents
the frame POSSESSION, with its two crucial participants and the minimal
constraints on their referents listed under each participant (human PR and the
relevance of the possessibility hierarchy). This general possessive relation
can be expressed as Plain Posession, in the upper part of the plane, or as

Situated Possession, in the lower half; we could think of both as more spe-
cialized sub-frames, each of which is conventionally associated with a par-
ticular form – the GP and AP grammatical constructions, respectively.
The solid-line rectangle in the upper part of the diagram represents the
GP construction and its relationship to the background frame: the PR must
be in the genitive form(s) and for the PM, we have to note that inalienable
possessa are generally incompatible with, or at least strongly dispreferred in,
this construction (the symbols ‘<’, ‘>’ indicate the direction of preference).
The dashed-line rectangle represents the AP construction, which modulates
the background frame by adding a number of features to both the PR and
the PM, listed in the columns under each frame participant. With respect to
the PM, the most important change consists in delimiting the semantic
classes of permissible possessa (in the diagram indicated by reference to the
classes indentified in (14) and in noting that the AP construction preferen-
tially selects items from the inalienable end of the possessibility hierarchy,
in contrast to GP.
The dotted-line, rounded rectangles are included to indicate that both GP
and AP constructions overlap in specific ways with other, non-possessive
constructions. In the case of the Syntactic Genitive, the common feature
with GP is just the Genitive form; the construction does not include any
part of the possessive frame and also differs from GP in that the genitive is
a participant in the event denoted by the head noun (agent or patient in no-
minalizations). AP, on the other hand, shares all four characteristics of the
PR (animacy, event participant status, affectedness, and dative marking) with
the DI construction, which excludes the possessive semantics of the dative
nominal and the AP’s constraint on verb semantics. Instead, DI is related to
other non-possessive dative constructions (the space labeled Affected da-
tives). I will return in a moment to the significance of the three items that
are underlined.10
10
Additional constructions expressing attributive possessive relations (such as
adjectival, coming-into-possession patterns based on transfer, etc.) can be incor-
porated into this network, once their features are properly worked out.
240 Mirjam Fried
Synt. genitive (ex. 12)
event partic.
Plain possession (GP)
GEN/poss. morphol. inalienable < alienable
frame POSSESSION: [ Possessor, Possessum ]

human possessibility hiearchy
DAT inalienable > alienable
Situated
possession (AP) event partic. 'salient in daily routine' (14)
'affective' verbs affected affected
Dat. of interest (ex. 24)
Affected datives (ex. 23)
Figure 3. (Partial) ‘attributive possessive’ network
The network thus captures important facts about the way constructions may
interact and the ways in which linguistic expressions may stretch the prop-
erties of those constructions. I will now briefly comment on three such ex-
amples, all involving the AP construction.
As already noted, a systematic account of possession always faces the
question of what should count as possession and how inclusive or non-
inclusive we should be in defining the PR and the PM. Even with the rela-
tively permissive prototype assumed in this paper (compared to Taylor’s),
we still have to account for examples such as (27), in which the PM does
not easily fit the categories suggested for AP in (14):
(27) Ale jednou se nám ztratil jeden

but once RF 1PL.DAT get.lost.PST.SG.M one.NOM.SG.M
pacient
patient.NOM.SG.M
‘But once one of our patients disappeared’ [SYNEK; Hrabal 1993]
It is of course hard to argue that there is an ownership relation between the

hospital personnel (here the speaker) and the patients. As discussed in sec-
tion 3, the speaker’s choice to employ AP will be motivated by the dis-
course context and/or genre: one that sets up the plausibility of an affective
possessive reading which involves the relationship between a human being

and another entity that is in a contextually salient relation to it. Such cases,
then, receive the reading of situated possession by virtue of being used in
the AP construction, and we are not forced to relax the possessive prototype
itself to account for them. Since the criteria for a plausible PM are evalu-
ated on a sliding scale to begin with, there is perhaps always room for
stretching the scale in a motivated way.
A more dramatic departure from the possessive prototype is shown in
(28), where an inanimate entity seems to be cast in a PR-like relation to an-
other inanimate entity. Since the PR is defined as human (or at least ani-
mate), admitting (28) as a case of possession is a much more serious matter.
(28) {Pokud by pda poklesla jen o padesát centrimetr,}

uhnijí vem stromm koeny.
rot.away.PRES.3PL all.DAT tree.DAT .PL.M root.NOM.PL.M
‘{Should the soil sink even by as little as 50 cm,} the trees will lose
all their roots to rot’ (lit. ‘the roots of all the trees will rot away on
them’) [syn2006publ; Respect 26 /1993]
We could declare that (28) is not a case of possession but a simple part-
whole relation. The advantage of such an analysis would be that it would
preserve the concept of possession as an experiential gestalt, specific to
human beings. This way we would also avoid the danger inherent in any
prototype-based analysis: in order to account for every new deviation, we
could, in principle, keep relaxing the prototype ad infinitum, which then
amounts to justifying just about anything as an instance of the same con-
cept, and the prototype loses its coherence as a tool of systematic analysis.
However, a categorical exclusion of examples such as (28) leaves unan-
swered an obvious similarity between them and the AP pattern, both for-
mally (the ‘whole’ being in the dative) and in the overall affective interpre-
tation: when the roots die, the tree is certain to die as well. In order to
account for these and similar extensions, the constructional approach offers
an alternative that allows us to incorporate the full range of deviations that
may arise in actual discourse, while at the same time preserving the posses-
sive prototype as the conventional semantic basis. The AP construction ex-
pects the prototype – represented in the frame – to hold, but of course the
match between the prototype and the lexical fillers of the constructional
slots will not be always perfect, stretching the prototype to varying degrees.
In (28), the stretching concerns the semantics of the PR, but at the same
time, it is very close to the prototype in two ways: (i) the PM is (construable
242 Mirjam Fried
as) inalienable in the same way body parts are and (ii) the mutual relation-
ship between the whole and its part is fully compatible with the affective
meaning of the AP construction: the whole is affected because its constitu-
tive part is affected. It is also important to note that GP, shown in (29), does
not evoke the situated possession reading but stays purely at the level of a
part-whole relationship. In (29), the implication may very well be that the
trees will somehow make it anyhow; in any case, (29) is not concerned with
the fate of the trees, it is about the fate of the roots only.
(29) uhnijí koeny vech strom

rot.away.PRES.3PL root.NOM.PL.M all.GEN tree.GEN.PL.M
‘the roots of all the trees will rot away’
I would argue that this difference is due precisely to the fact that the GP
construction is not associated with a special meaning; the form can cover
both possessive and non-possessive relations, and consequently cannot im-
pose a possessive reading so easily on combinations that deviate from the
prototype in a radical way (such as presenting an inanimate entity as an
owner of anything). By contrast, if the semantic and pragmatic properties of
the AP construction as a whole (especially the affectedness of the PR) are
invoked we have a principled way of explaining what allows the stretch
into domains in which we do not have real PRs but only a very specific
(and tight) part-whole relationship. In other words, it is the use of the AP
construction in encoding a (close) part-whole relation that allows a personi-
fication reading, i.e. a conceptualization which mimics a relationship be-
tween an animate PR and a body part.
We could say that both of these cases (27)–(28) illustrate scenarios in
which the AP construction, as a conventional grammatical pattern in its
own right, facilitates manipulations of the possessive frame, sometimes also
in an interaction with the closely related, but more general, PART-WHOLE
frame.11 But the AP construction may also attract other patterns in the net-
work and pull them into an AP reading because of certain shared construc-
tional properties. Here I have in mind the issue of intransitive agents and
their potential for compatibility with an AP interpretation, in an apparent
11
I have not provided the details of this frame or its place in the network, mostly
for reasons of keeping the representation uncluttered and easily readable, given
the focus of this paper (pure possession). But it is obvious that it must be part of
a more complete representation of this functional space, particularly in working
out the Genitive-related domain.
contradiction to the participant hierarchy established in section 4.2. It might

seem that cases such as (30), with the active intransitive verb utéct ‘run
away’, are simply evidence that Czech does not, after all, restrict the pos-
sessa to non-agentive referents and instead excludes only agents/subjects of
transitive verbs.
(30) vaemu tat’kovi utek ten

your.DAT dad.DAT run.away.PST.SG.M that.NOM.SG.M
keek taky,
hamster.NOM.SG.M also
{to byl první jeho keek. von byl stran z toho smutnej. a tak sme
mu museli koupit hned dalího}
‘your dad’s hamster also ran away [in our backyard], {that was his
first hamster and he was all so sad about it and so we had to buy a
new one for him right away}’ [oral2006;1195644]
Such an analysis, though, would be an oversimplification since this pattern-

ing, attested very rarely to begin with, is also highly restricted. First, the
utterance has to be available for an affective construal, as is the case here;
replacing the dative in (30) with GP would be incoherent in the light of the
speaker’s subsequent elaboration about what effect the event had on the
PR. And as noted in section 4.2, this availability has to be accommodated
by the verb meaning. The fact is that only very few verbs appear to work
semantically: recall that the quintessential active verb pracovat ‘work’ re-
sists an AP reading (18). A better fit seems to come with verbs of removal
(utéct ‘run away’, schovat se ‘hide’), which is not surprising, given the se-
mantics of the AP construction as affecting pre-existing possession. Also
verbs of eating or drinking are possible candidates, presumably because the
result of such acts has the potential of entering the sphere of interest of
people other than the eaters/drinkers.12 And second, the PM (necessarily
animate) must be very high on the possessibility hierarchy. Examples such
as (31) cannot be interpreted possessively, even though there are two ani-
12
While I have not run across such an example in the corpus, I can think of usages
such as (i). The verb semantics found in these combinations is still waiting for
more careful research.
(i) Syn se jim opíjel.
son.NOM.SG.M RF 3PL.DAT drink.to.excess.PST.SG.M
‘Their son drank heavily [to their worry/embarrassment].’
244 Mirjam Fried
mate participants and the action of one (the police) clearly is intended to
have consequences for the other (ostensibly mice, figuratively for the in-
habitants):
(31) {…obklopila nai vilu vozidla Státní bezpenosti a postupovali tak,}

aby jim neutekla ani my
so.that 3PL.DAT NEG.run.away.PST.SG.F not.even mouse.NOM.SG.F
‘{…cars of the Secret Police surrounded our house and they [=secret
agents] proceeded [carefully]} so that not even a mouse [could] escape
from them’ (lit. ‘to them’) [syn2006pub;LN 33/1992]
We have to conclude that examples such as (30) are more plausibly analyzed
as instances of the DI construction, such as exemplified in (24) and (31), in
which any kind of verb, including all semantic types of intransitives, can be
used, and which expresses a situation with an indirect effect on the ‘inter-
ests’ of an animate entity. DI of course overlaps with the AP construction
to a great extent; outside of not having the possessive dimension, the DI
only differs in that it places absolutely no constraints on the verb semantics.
It is not a stretch, then, for a DI token to invite an AP reading, provided that
certain features of that token coincide with a particular narrow set of fea-
tures of the AP construction.
This finally brings us to the significance of the three items that are under-
lined in Figure 3. The features human on the PR and inalienability on the
PM are central to the notion of experientially defined possession. We can,
therefore, expect that if an inherently non-possessive expression (such as
DI) invites a possessive reading, it can be only at the level of the core pos-
sessive properties. But satisfying these two features of the possessive frame
does not, by itself, guarantee a successful AP interpretation. The PR must
also be construable as affected by the event expressed by the verb, which is
a central property of the AP construction. This is, of course, related to the
lexical meaning of the verb, which is constrained in AP, but that restriction
is evidently not as rigid as the requirement that, whatever the verb, the PR
must come out as an affected entity. Some verbs with active semantics are
inherently better equipped for such a stretch (e.g. verbs of removal) than
others (e.g. pracovat ‘work’), and that is what accounts for the relative
(un)availability of DI tokens for an AP interpretation.
The important point is that all these shifts, whether they involve exten-
sions of the possessive prototype into broader semantic domains or, on the
contrary, attracting tokens of non-possessive constructions, can all be ex-
plained by appealing to the same cluster of properties (the core features of
the possessive prototype associated with AP) and to the AP pattern as a con-
structional gestalt: the situated possessive meaning of the whole construc-
tion is the ‘glue’ that holds all these seemingly disparate uses together. We
could thus think of the three underlined items as having a privileged status
within this network: they constitute the set of features that are instrumental
in various partial shifts, giving rise to and at the same time constraining
novel usages.
6. Conclusions
Perhaps the central – and inherently thorny – issue in sorting out possessive
expressions in any language is the question of how we define the concep-
tual category to begin with. The present analysis is based on the notion of
possessive prototype understood as an experiential gestalt, which takes the
PR to be necessarily an animate (human) entity and the PM to be placed
somewhere along the possessibility hierarchy, without stating categorically
what may or may not count (universally) as a PM. Based on the corpus at-
testations of the Czech patterns, we can draw at least two conclusions about
possessibility. (i) Rather than relying on purely linguistic categories, such
as (in)animacy, concreteness, control, etc., possessibility is best defined in
terms of culturally determined clusters of concepts and expectations about
what is conventionally construed as possessible; the conventional under-
standing can be then extended to cases where possessive construal is condi-
tioned by the type of discourse or genre. And (ii) different grammatical
forms expressing possession may interact with the possessibility hierarchy
in different ways.
These generalizations are based on a close study of two syntactic patterns
that both express possession as a time-stable and presupposed relation. The
analysis has established that the patterns are not equivalent semantically or
pragmatically and, therefore, cannot be treated simply as structural variants
of a single possessive schema. Each pattern encodes a distinct conceptuali-
zation of a possessive relation, compatible with different communicative
contexts: plain possession is expressed by a Genitive PR (GP), which shares
certain features with non-possessive genitives; situated possession – an idio-
syncratic combination of possession and affectedness – is expressed by da-
tive-marked Affected PR (AP), which shares certain features with non-
possessive dative-marked roles. It follows that an adequate representation of
the speakers’ understanding of these patterns requires reference to several
layers of information: semantic and structural limits on the PM (AP), affect-
246 Mirjam Fried
edness of the PR (AP), semantic and morphological constraints on the PR

(GP), relative discourse prominence of the PR vis-à-vis the PM (AP, GP),
verb semantics (AP), and contextual compatibility (AP, GP).
The analysis thus makes an argument for taking a Construction Grammar
approach as a particularly useful way of framing our understanding of all
the relevant issues. First, the clusters of conventionally co-occuring features
are naturally captured through the notion of ‘grammatical construction’.
Second, the two constructions constitute distinct pieces of a larger network
of grammatical entities organized around shared features, both formal and
semantic: AP and GP can be shown to occupy partially overlapping domains
within the general functional space of attributive possession. And finally,
organizing our knowledge about individual constructions in such a network
provides us with a more refined map of criteria that can play a role in ex-
pressing possessive relations in general. Based on such a map, we can start
articulating more systematic hypotheses about the paths along which the
prototype might be extended into more peripheral instances; specifically,
the extensions can be systematically motivated (and also constrained) by
the ways in which the constructions in the network may interact both with
each other and with the possessive prototype itself.
Acknowledgements
I wish to thank all those colleagues whose questions, at various conference

gatherings, have helped me articulate my thoughts about this topic. And I’m
especially grateful to Bill McGregor and Elizabeth Traugott for very careful
reading of the manuscript and helpful comments on the work.
Data
esk národní korpus (spoken corpora PMK, BMK, ORAL2006; written

corpora SYN2000, SYN2006PUBL). Ústav eského národního korpusu FF
UK, Praha. <http://ucnk.ff.cuni.cz>
esk národní korpus – A sample of Czech language and literature
(SYNEK, LITERA) [CD-ROM]. Ústav eského národního korpusu FF
UK, Praha. <http://ucnk.ff.cuni.cz>
References
Bally, Charles
1995 The expression of concepts of the personal domain and indivisibility
in Indo-European languages. In Chappell, Hilary and William B.
McGregor (eds.), 31–61. [First puplished in 1926.]
Berman, Ruth A.
1982 Dative marking of the affectee role: data from Modern Hebrew. He-
brew Annual Review 6: 35–59.
de Gruyter.
Croft, William
1985 Indirect Object ‘Lowering’. Berkeley Linguistic Society 11: 39–51.
Croft, William
2001 Radical Construction Grammar. Syntactic theory in typological per-
spective. Oxford: Oxford University Press.
Dane
, Franti
ek
1974 Functional sentence perspective and the organization of the text. In
Papers on Functional Sentence Perspective, F. Dane
(ed.), 106 –128.
Praha: Academia.
Fillmore, Charles J.
1982 Frame Semantics. In Linguistics in the Morning Calm, The Linguistic
Society of Korea (ed.), 111–137. Seoul: Hanshin.
Fillmore, Charles J.
1989 Grammatical Construction Theory and the familiar dichotomies. In
Language Processing in Social Context, R. Dietrich and C. F. Grau-
mann (eds.), 17–38. Amsterdam: North-Holland /Elsevier.
Firbas, Jan
1966 On defining the theme in functional sentence perspective. Travaux
linguistiques de Prague 2: 267–280.
Fried, Mirjam
1994 Grammatical functions in case languages. Berkeley Linguistic Society
20: 184 –193.
Fried, Mirjam
1999a From interest to ownership: a constructional view of external posses-
sors. In External Possession, Doris L. Payne and Immanuel Barshi
(eds.), 473–504. Amsterdam /Philadelphia: John Benjamins.
Fried, Mirjam
1999b The ‘free’ datives in Czech as a linking problem. In Annual Work-
shop on Formal Approaches to Slavic Linguistics 7, K. Dziwirek, H.
Coats, and C. Vakareliyska (eds.), 145–166. Ann Arbor: Michigan
Slavic Publications.
248 Mirjam Fried
Fried, Mirjam
2008 Dative possessor: delimiting a grammatical category based on usage.
In Gramatika a Korpus / Grammar and Corpora, F. tícha and M.
Fried (eds.), 51–64. Praha: Academia.
Fried, Mirjam and Jan-Ola Östman
2004 Construction Grammar: a thumbnail sketch. In Construction Grammar
in a Cross-Language Perspective, Mirjam Fried and Jan-Ola Östman
(eds.), 11–86. Amsterdam: John Benjamins.
Grepl, Miroslav and Petr Karlík
1998 Skladba spisovné etiny [Syntax of literary Czech]. Praha: Votobia.
Heine, Bernd
Manoliu-Manea, Maria
1995 Inalienability and topicality in Romanian: pragma-semantics of syn-
tax. In Chappell, Hilary and William B. McGregor (eds.), 711–743.
O’Connor, Mary Catherine
1994 The marking of possession in Northern Pomo: Privative opposition
and pragmatic inference. BLS 20: 387–401.
Payne, Doris L.
1997a The Maasai external possessor construction. In Essays on Language
Function and Language Type, Joan Bybee, John Haiman and Sandra
A. Thompson (eds.), 395–422. Amsterdam/ Philadelphia: John Ben-
jamins.
Payne, Doris L.
1997b Argument structure and locus of affect in the Maasai External Pos-
session Construction. BLS 23 [Special Session on Syntax and Se-
mantics in Africa]: 98–115.
Payne, Doris L. and Immanuel Barshi (Eds.)
1999 External Possession. Amsterdam /Philadelphia: John Benjamins.
Pit’ha, Petr
1992 Posesivní vztah v etin. Praha: AVED.
Sgall, Petr, Eva Haji ová and Jarmila Panevová
1986 The Meaning of the Sentence in its Semantic and Pragmatic Aspects.
D. Reidel Publishing Company.
Taylor, John R.
1989 Linguistic Categorization: Prototypes in Linguistic Theory. Oxford:
Clarendon Press.
Tsunoda, Tasaku
1995 The possession cline in Japanese and other languages. In Hilary
Chappell, and William B. McGregor (eds.), 565–630.
Zimek, Rudolf
1960 K chápání posesivnosti. Rusko- eské studie, Sborník Vysoké
koly
pedagogické v Praze, Jazyk a literatura II, 131–156.
Attributive possessive constructions in Oceanic
Frantisek Lichtenberk
1. Introduction
A feature characteristic of the Oceanic languages 1 is the existence of more

than one type of attributive possessive construction. As a rule (there are a
few exceptions), Oceanic languages have two basic types of possessive
construction, one of which usually has two or more subtypes. Examples
(1)–(3) from Manam illustrate. In (1) the possessive suffix, which indexes
the PR, is attached to the PM noun, while in (2) and (3) the possessive suf-
fixes are attached to two different possessive classifiers:2
(1) ara-gu
name-1SG:POSS
‘my name’ (Lichtenberk 1983a: 282)
(2) pera ana-gu

house POSS.CLF-1SG:POSS
‘my house’ (Lichtenberk 1983a: 292)
1
Oceanic is a subgroup within Austronesian. Oceanic languages are spoken in
mainland New Guinea and neighbouring islands, Island Melanesia, Polynesia
and Micronesia, but not all of the indigenous languages of New Guinea, Island
Melanesia and Micronesia are Oceanic. For a detailed overview of the Oceanic
family see Lynch et al. (2002).
2
Besides the Leipzig Glossing Rules, the following abbreviations are used in
glossing the examples: CONSTR – construct; NONSG – non-singular; NUM –
numeral marker; PC – paucal; REAL – realis; SV – stem vowel; THC – thematic
consonant.
The glossing conventions are – by and large – those of the sources. In some
cases the glosses have been adjusted for the sake of uniformity. The inclusive
forms are not considered here a subtype of the first person but a category of its
own (Daniel 2005; Lichtenberk 2005a), hence the absence of specification of
person. Stress marking has been omitted from the Manam examples. The Toqa-
baqita data come from my own field notes.
250 Frantisek Lichtenberk
(3) asi ne-gu

bushknife POSS.CLF-1SG:POSS
‘my bushknife’ (Lichtenberk 1983a: 401)
As discussed in section 3, the formal differences among the possessive con-

struction types are associated with semantic/pragmatic differences: they
signal different types of relation between the PM and the PR, although there
are language-specific exceptions.
The terms “PM” and “PR” will be used here in two senses: first, they
will signify the two constituents in a possessive construction: the PM is the
head and the PR its modifier; and second, they will signify their referents.
Context will make it clear which of the two senses is intended on a given
occasion.
The paper is concerned primarily with the semantic/pragmatic aspects of
the Oceanic possessive systems. However, to set the stage, some formal
aspects will have to be considered first. This is the subject of section 2.
Section 3 is concerned with the kinds of semantic/pragmatic distinctions
expressed by the different types and subtypes of possessive construction.
The phenomenon of fluidity, the ability of one and the same noun to occur
in the PM position of more than one type or subtype of possessive construc-
tion, will be discussed in section 4. The existence of exceptions to the gen-
eral patterns will be considered in section 5. Section 6 is concerned with
two views of the Oceanic possessive systems, as a noun-class system and as
a relational system, and it will be argued there that the Oceanic possessive
systems are basically relational in the sense that, by and large, the choice of
a possessive construction depends on the nature of the relation between the
PM and the PR entities. Section 7 will offer some remarks on the motiva-
tion behind the development of a system with multiple types and subtypes
of possessive construction, and the last section provides a summary.
The languages referred to here and their (approximate) locations are
given in Table 1; cf. Map 1.
Attributive possessive constructions in Oceanic 251
Table 1. Languages referred to and their locations
Language Location
’Ala’ala Papua New Guinea
Anejom Vanuatu
Araki Vanuatu
Bali-Vitu Papua New Guinea
Banoni Papua New Guinea
Cèmuhî New Caledonia
Erromangan Vanuatu
Fijian (Standard) Fiji
Gapapaiwa Papua New Guinea
Hawaiian Hawai’i (Polynesia)
Hoava Solomon Islands
Houaïlou New Caledonia
Iaai New Caledonia
Kairiru Papua New Guinea
Kilivila Papua New Guinea
Kokota Solomon Islands
Kosraean Kosrae (Federated States of Micronesia)
Kwaio Solomon Islands
Labu Papua New Guinea
Lenakel Vanuatu
Lenkau Papua New Guinea
Lolovoli Vanuatu
Lou Papua New Guinea
Manam Papua New Guinea
Mussau Papua New Guinea
Nalik Papua New Guinea
Niuean Niue (Polynesia)
Paamese Vanuatu
Pohnpeian Pohnpei (Federated States of Micronesia)
Pukapukan Pukapuka (Polynesia)
Puluwatese Puluwat (Federated States of Micronesia)
Rotuman Rotuma (Fiji)
Tamambo Vanuatu
Tobati Papua (western New Guinea, Indonesia)
Toqabaqita Solomon Islands
Ulithian Ulithi, Fais (Federated States of Micronesia)
Vinmavis Vanuatu
Wayan Fiji
Zabana Solomon Islands
Map 1. The boundaries of the Oceanic subgroup. (From: The Oceanic languages, John Lynch, Malcolm Ross and Terry Crowley, 2002, p. 5
[Map 1.2 there], Richmond, Surrey: Curzon. Reproduced by permission of Taylor & Francis Books UK.)
2. The formal aspects of Oceanic possessive constructions
With over 400 languages in the Oceanic group, it is impossible to do full

justice here to the possessive systems found there, and the discussion will
be restricted to the major patterns.
2.1. The typical Oceanic patterns
2.1.1. Direct and indirect possessive constructions
The typical system consists of two basic types of possessive construction,

direct and indirect. In the direct type, the PM noun carries a possessive affix
encoding or cross-referencing the PR:
Lolovoli
(4) hava-da
family-NONSG(INCL):POSS
‘our family’ (Hyslop 2001: 169)
See also (1) from Manam in section 1.

In the indirect type, it is a possessive classifier 3 rather than the PM noun
that carries a possessive affix:
Lolovoli
(5) no-da hala
POSS.CLF-NONSG(INCL):POSS visitor
‘our visitor’ (Hyslop 2001: 180)
See also (2) and (3) from Manam in section 1.
3
Considering the elements that carry possessive suffixes in indirect possessive
constructions to be (possessive) classifiers is the usual approach adopted in recent
descriptive work. However, Palmer and Brown (2007) argue that in the Kokota
language and possibly some other languages those elements are, in fact, “generic
nouns” (p. 208), and that it is these nouns that head possessive constructions.
However, this analysis is not without problems. Here, the standard analysis is
retained, considering those elements to be classifiers. Since the present study is
concerned primarily with the semantic/pragmatic properties of Oceanic posses-
sive constructions, the question of whether in a given language those elements
form a morphosyntactic category of their own or whether they are perhaps a
subcategory within the category of nouns is not of primary importance here.
With very few exceptions, the possessive affixes are suffixes, as in (1)–
(5) above. A few languages have possessive prefixes. However, the prefixes
normally exist in addition to possessive suffixes and their use is restricted
in various ways. For example, in Western Fijian dialects the possessive pre-
fixes are used only in direct possessive constructions and only when the
PM–PR relation is other than kinship, such as ‘my blood’ (Geraghty 1983).
It should also be noted at this point that the “possessive” affixes have, in
some languages, functions other than indexing the PR in a possessive con-
struction. For example, in Toqabaqita the same set of suffixes is used with
one class of transitive verbs to index the direct object, and with certain
verb-phrase internal particles to index the subject.
The PR may be encoded by a noun phrase, in which case the typical pat-
tern is for the PR noun phrase to be cross-referenced by means of a posses-
sive affix either on the PM noun if the possessive construction is of the di-
rect type, or on the possessive classifier if the possessive construction is of
the indirect type (see section 2.1.2 for discussion of cross-referencing). Ex-
amples (6) and (7) from Hoava illustrate:
(6) sa bele-na sa boko

ART:SG tail-3SG:POSS ART:SG pig
‘the pig’s tail’ (Davis 2003: 98)
(7) a-na napo sa koburu

POSS.CLF-3SG:POSS drink ART:SG child
‘the child’s drink’ (Davis 2003: 102)
The ordering of the expressions of the PM, the PR and the possessive clas-
sifier varies from language to language: for example, (classifier) PM (PR)
(Hoava), (PR) PM (classifier) (Manam), PM (classifier) (PR) (Anejom).
2.1.2. Cross-referencing of the PR
With respect to PR phrases that are lexical (rather than pronominal; but see
the Vinmavis example (24) in section 2.1.3), three basic types of cross-
referencing can be distinguished: full, partial and construct. In full cross-
referencing the possessive affix cross-references the PR both for person and
for number. Besides singular and plural, many Oceanic languages also have
a dual number, and some also have a trial or paucal number. A singular/
plural/dual/paucal system with full cross-referencing is found in (Standard)
Fijian. This is illustrated in the two pairs of examples in (8)–(11) for the
singular and the paucal numbers. In each pair, the same possessive suffix is
used whether or not there is a PR phrase present.
(8) na no-na vale

ART POSS.CLF-3SG:POSS house
‘his house’ (Milner 1972: 22)
(9) na no-na vale na traga

ART POSS.CLF-3SG:POSS house ART chief
‘the chief’s house’ (Milner 1972: 23)
(10) na no-dratou waqa

ART POSS.CLF-3PC:POSS canoe
‘their (few) canoe’ (Milner 1972: 22)
(11) na no-dratou waqa na cauravou

ART POSS.CLF-3PC:POSS canoe ART young.man
‘the (few) young men’s canoe’ (Milner 1972: 22)
Of the three types of cross-referencing systems, the full cross-referencing

system is the most common one in Oceanic.
In partial cross-referencing, the PR is cross-referenced only for person,
not for number. Partial cross-referencing is found in, for example, Kwaio.
When there is no PR noun phrase, there is a singular–plural (and dual) dis-
tinction in the third person:
(12) ’i’i-na
tail-3SG:POSS
‘its tail’ (Keesing 1985: 113)
(13) falai-ga
head-3PL:POSS
‘their heads’ (Keesing 1985: 113)
However, when there is a PR noun phrase present, the “singular” possessive

suffix must be used even if the PR is plural, as explicitly stated by Keesing
(1985: 107):
(14) lata-na wela

name-3SG:POSS child
‘the child’s name’ (Keesing 1985: 107)
(15) lata-na ta’a

name-3SG:POSS people
‘the people’s names’ (Keesing 1985: 107)
The system of partial cross-referencing is not common. It is found in a

group of closely related languages spoken in the southeast Solomon Islands;
but see also further below for Erromangan.
In construct cross-referencing, a special “construct” affix is used on the
PM noun or on the possessive classifier to cross-reference the PR, but only
if there is a PR phrase present. This is illustrated by the following set of
examples from Anejom. In (16), without a PR phrase, the possessive classi-
fier carries the third person singular possessive suffix -n, while in (17), with
a PR phrase, the possessive classifier carries the construct suffix -i:
(16) neto lida-n

sugarcane POSS.CLF-3SG:POSS
‘his/her sugarcane’ (Lynch 2000: 60)
(17) nade-n lida-i inhalav
breast-3SG:POSS POSS.CLF-CONSTR baby
‘the baby’s breast’ (Lynch 2000: 60)
(In [17] the possessive suffix -n on nade ‘breast’ indexes the baby’s mother,
not the baby; the breast is “a source of milk to be sucked by the baby”,
hence the possessive classifier lida, “used with nouns whose referents are
things one sucks the juice out of, but without consuming the flesh in any
way” [Lynch 2000: 60].)
The same construct suffix is used when the PR noun phrase is plural.
(18) injap uma-i elpu-Uje

sea POSS.CLF-CONSTR PL-Uje
‘the Uje people’s sea’ (Lynch 2000: 61)
The construct suffix is also used on the PM noun in direct possessive con-
structions:
(19) risi-i di?

mother-CONSTR who?
‘whose mother?’ (Lynch 2000: 58)
Construct cross-referencing systems are common in languages of Vanuatu

and Micronesia. It should be noted, however, that the term “construct” pos-
sessive affix is interpreted in different ways by different analysts. Thus, for

example, Crowley (1998) uses the term to refer to a possessive suffix in
Erromangan that functions as the third person singular possessive suffix
when there is no PR phrase, as in (20) below, but which is also used to index
the PR when there is a PR phrase, regardless of the grammatical number, as
in (21) and (22). This is comparable to the system referred to further above
as partial cross-referencing with respect to Kwaio.
(20) retpo-n
wife-3SG:POSS
‘his wife’ (Crowley 1998: 52)
(21) nompu-n natmonuc

head-CONTR/3SG:POSS chief
‘the chief’s head’ (Crowley 1998: 172)
(22) nompu-n ovatmonuc

head-CONSTR/3SG:POSS PL:chief
‘the chiefs’ heads’ (Crowley 1998: 172)
In the present study, the term “construct” possessive affix is restricted to

cases either where the form of the affix is unique, unlike that of any of the
other possessive affixes, as in Anejom ((17)–(19) above), or where the suf-
fix need not index any of the features of the PR phrase, for which see (24)
from Vinmavis in section 2.1.3.
Finally, in at least one language PR phrases are not cross-referenced at
all, even though possessive affixes are used on PM nouns and on possessive
classifiers in the absence of a PR phrase. This is the case in Cèmuhî (Riv-
ierre 1980).
2.1.3. Type of PR phrase
The type of possessive construction required may depend on the type of PR

phrase. In a number of languages possessive constructions with pronominal
PR phrases exhibit idiosyncratic properties, although this need not apply to
pronouns of all numbers and persons. For example, Vinmavis has a posses-
sive suffix that functions specifically to index third person singular PRs,
but it also functions as a construct suffix with all the other pronouns as
PRs:
(23) netal-n
leg-3SG:POSS
‘his/her leg’ (Crowley 2002: 642)
(24) netal-n get

leg-CONSTR 1PL(INCL)
‘our legs’ (Crowley 2002: 643)
The same suffix is also used with lexical PR phrases:
(25) netal-n matoro

leg-CONSTR old.man
‘the old man’s leg’ (Crowley 2002: 642)
Similar use of the third person singular possessive suffix as a construct suffix
with plural independent pronouns as PR phrases is found in Lou and Lenkau
(Ross 1988: 332).
In some languages independent personal pronouns cannot form a PR
phrase. The PR can be indexed only by means of a possessive affix. Bali-
Vitu4 is one such language: example (26a) without a pronominal PR phrase
is grammatical, while (26b) with a pronominal PR is not:
(26) a. a lima-ma b. *a lima-ma oho

ART hand-2SG:POSS ART hand-2SG:POSS 2SG
‘your hand’ (‘your hand’) (Ross 2002a: 370)
In a few languages there are differences in the structure of possessive con-

structions depending on whether the PR noun phrase is common or proper.
In Iaai, a common-noun PR in a direct possessive construction is cross-
referenced on the PM noun by means of a possessive suffix, but there is no
cross-referencing of proper-noun PRs (Ozanne-Rivierre 1976). And in Fijian,
constructions with singular proper-noun PRs are different from those used
with common-noun and plural proper-noun PRs (Milner 1972).
This concludes the survey of the typical pattern of Oceanic possessives
constructions with a distinction between direct and indirect constructions,
with relatively minor variations on the basic pattern in some languages.
4
Ross (2002a) calls the language “Bali-Vitu”, but van den Berg and Bachet say
in a grammar of Vitu that “Ross’s sketch is primarily a description of the Bali
variety” (2006: 2).
Next we will briefly consider some possessive systems that are different
from the typical pattern in more significant respects.
2.2. The Polynesian pattern
In the Polynesian languages the direct-indirect system of possessive con-

structions has been replaced by a different binary pattern. This is also, to
some degree, true of the Rotuman language, a sister of the Polynesian group.
For convenience, I will refer to the pattern discussed in what follows as the
“Polynesian pattern”. In the Polynesian pattern, the direct possessive con-
struction type is absent, apart from some lexical exceptions. Instead, there
is a binary system that is customarily referred to in Polynesian linguistics as
A-possession and O-possession on the basis of the vowels found in the two
sets of possessive marking. Discussions of possessive constructions in sev-
eral Polynesian languages (and in Rotuman) can be found in Fischer (2000),
together with an overview of the Polynesian pattern by Clark (2000). The
semantic/pragmatic aspects of the Polynesian pattern are briefly discussed
in section 3.4 below.
Examples (27) and (28) from Hawaiian illustrate A-possession and O-
possession, respectively:
(27) ka pahi a Kimo

ART knife POSS Kimo
‘Kimo’s knife’ (Wilson 1976: 29)
(28) ka hale o Kimo

ART house POSS Kimo
‘Kimo’s house’ (Wilson 1976: 29)
Within the basic A-O constrast, each Polynesian language has a number of
subtypes of possessive construction. In the Hawaiian examples above, the
possessive markers are formally prepositions. In another type of construc-
tion, the possessive markers are fused with an article; see (29) and (30) from
Pukapukan:
(29) t-a-ku tama

ART-POSS-1SG:POSS child
‘my child’ (Salisbury 2002: 172)
(30) t-o-ku vaka
ART-POSS-1SG:POSS canoe
‘my canoe’ (Salisbury 2002: 171)
The article t- signifies definiteness and singular number of the PM.

Another subtype of possessive construction is usually referred to as “ir-
realis possession”: “the intention or anticipation that something will be pos-
sessed” (Clark 2000: 262). The possessive marker is added to an element m-,
which, according to Clark, continues an earlier irrealis or optative marker.
The variety of possessive constructions even in a single Polynesian lan-
guage can be quite considerable, and the discussion above is no more than
basic. The collection of articles edited by Fischer (2000) is a useful refer-
ence where more detail on some of the Polynesian languages can be found.
2.3. Absence of the indirect system of possessive classifiers
In the Polynesian languages it is the direct possessive type that is absent.

There are also languages where it is the indirect type that is absent. This is
the case in Toqabaqita and its close relatives. The possessive construction
in (31) corresponds to the direct type in other languages, but for reasons to
become clear presently, it can also be referred to as suffixing, because the
PR is cross-referenced on the PM noun by means of a possessive suffix:
(31) qaba-na wela

hand-3SG:POSS child
‘the child’s hand(s)’
In the other type, there is no possessive classifier and no indexing of the PR

on the PM noun. This type of construction can be referred to as bare.
(32) fanga wela

food child
‘the child’s food’
The semantic/pragmatic aspects of Toqabaqita possessive constructions are

discussed in sections 3.4 and 5.
A possessive system with a contrast between a direct/suffixing construc-
tion and a bare construction is also found in Kairiru (Papua New Guinea)
(Wivell 1981). For some remarks on Kairiru see section 5 below.
2.4. Some other types of possessive system
There are some other types of possessive system found in Oceanic, but as
these will not figure in the discussion of the semantics/pragmatics of pos-
sessive constructions, they are mentioned here only very briefly. In some
languages prepositions are used in possessive constructions, the comple-
ment of the preposition being the PR noun phrase. Discussion of possessive
prepositions in some Oceanic languages can be found in Hooper (1985).
The usual case is for possessive prepositions to be part of a larger system of
possessive constructions. Possessive prepositions are found in Polynesian
languages (see section 2.2 for Hawaiian), where they are part of the overall
A–O possessive contrast and have a classificatory function, not unlike the
possessive classifiers found in the typical Oceanic pattern.
Finally, there are a few Oceanic languages that have only one basic type
of possessive construction: there is no contrast comparable to those between
direct and indirect, suffixing and bare, or A and O types of construction.
Labu is one such language (Siegel 1984), and Tobati is another (Donohue
2002). Niuean, a Polynesian language, also has basically just one type of
possessive construction, although, according to Massam and Sperlich
(2000), traces of the Polynesian A–O possessive contrast do exist there.
3. The semantics/pragmatics of possessive systems with multiple

types of construction
3.1. Introduction
We can now turn our attention to the vast majority of Oceanic languages
that have possessive systems of multiple construction types and subtypes,
and consider their use. Is their use governed strictly lexically, each noun
having to be specified for which type of possessive construction it occurs in
the PM position? Or are there some general patterns that determine the use
of the various types? Once again, given the large number of languages in-
volved, the answer is not simple. There are clear patterns, but there are also
exceptions to these patterns. In this section and in section 4, the focus will
be on the patterns; the existence of exceptions will be considered in section
5. The basic, overall pattern is that the choice of a possessive construction
depends on the relation between the referents of the PM and the PR
phrases. Since the notion of relation is central to the understanding of how
the systems of Oceanic possessive constructions (normally) operate, some
discussion of the notion is in order.
3.2. Different kinds of the notion of relation with respect to possessive

constructions
There are (at least) three different types of the notions of “relation” and “re-
lational” that are relevant to the present discussion. One is the idea that
possessive constructions are relational. There are two entities which stand
or are put in a certain relation to each other, one as PM and the other as PR.
As is well known, the number of possible relations expressed by possessive
constructions, while not open-ended, is quite large. Besides ownership (PR
owns PM: ‘my money’, the money that belongs to me), some other relations
are: use (PR uses PM without necessarily owning it: ‘my bus’, the bus I
will take), control (PR has control over PM without owning it or using it:
‘my office’, the office I am in charge of), manufacture (PM is made by PR:
‘my cake’, the cake I baked), kinship (PM is a kin of PR’s: ‘my sister’), part
of a whole (PM is part of PR: ‘my head’, the head which is part of my own
body), and many others (see, for example, Langacker 1995: 56–57). This I
take to be an uncontroversial sense of the notion of relation.
Another sense of the terms “relation” and “relational” has to do with the
fact that certain concepts are inherently relational. Nouns that express in-
herently relational concepts are sometimes referred to as “relational nouns”,
such as mother (see, for example, Barker 1995; Partee 1997; Partee and
Borschev 2003). When inherently relational nouns occur in the PM position
in a possessive construction, the type of relation usually expressed involves
inalienable possession, as in my mother. When a noun that is not inherently
relational occurs in the PM position, the type of relation usually expressed
involves alienable possession, as in my knife. Relational nouns tend to
strongly favour a certain kind of relation between the PM and the PR. The
relation is intrinsic to the meaning of a relational noun. Barker (1995) also
uses the term “lexical possession” for this. Thus with my mother (with the
core meaning of mother) there is an intrinsic relation of kinship. With non-
relational nouns, on the other hand, there is typically no intrinsic relation
between the PM and the PR, and a variety of relations are freely available.
Thus with my knife the relation may be one of ownership (the knife I own),
use (the knife I use without owning it), manufacture (the knife I made), etc.
In such cases, the relations can be said to be “extrinsic” (Barker 1995). The
notions of inalienable and alienable possession and intrinsic and extrinsic
relations will be relevant in later discussion. I take the notions of inherently
relational concepts and relational nouns also to be uncontroversial.
It is the third sense of “relation(al)” that has enjoyed some controversy
in Oceanic linguistics with respect to possessive constructions. This is the
idea that the choice of a possessive construction is determined by, or sensi-

tive to, the nature of the relation between a PM entity and its PR. In princi-
ple, it is possible to recognize two views concerning the choice of posses-
sive constructions in Oceanic, which will be referred to as “noun-class
based” and “relation-based”, respectively. A useful discussion of the two
views can be found in Pawley and Sayaba (1990), and I will consider their
views in more detail in section 6. On the noun-class based view, the nouns
of a language with multiple types of possessive construction fall into a
number of classes depending on which of the possessive constructions they
select when occurring in the PM position. For example, Milner (1972: 65)
speaks of “gender” in Fijian, “a grammatical category of four classes”, which
he terms “neutral”, “edible”, “drinkable” and “familiar”: a given noun (or
“base” in Milner’s system of morphosyntactic categories) belongs in a cer-
tain class depending on which type of possessive constructions it selects. At
the same time, however, Milner makes it clear that the classes have a se-
mantic underpinning. For example, of the “edible” class he says (p. 66) that
it includes nouns that denote “[a]rticles of solid food considered from the
point of view of consumption (i.e. as distinguished from planting, selling,
etc.)”. Furthermore, he notes that there are nouns that belong in more than
one class, or in our terminology that they exhibit fluidity (section 4). François
(2002) says about the Araki language that it has two basic categories, inalien-
able nouns and alienable nouns, although his classification has also partly
to do with whether a given noun takes a possessive suffix (in a direct pos-
sessive construction) or not. However, François also notes that the assign-
ment of nouns to classes is not always straightforward and that “most nouns
can almost freely shift from one pattern to the other” (François 2002: 48).
And for Gapapaiwa, McGuckin (2002: 303) identifies three classes of nouns,
“each with its own set of possession markers”, but then goes on to say (p.
304) that “[t]hese possession classes do not correspond to fixed noun
classes, as the same noun can occur in more than one possessive category”.
A common theme that emerges from those studies that postulate noun
classes on the basis of possessive constructions is that the languages in
question exhibit fluidity, the ability of a noun to occur in the PM position of
more than one type of possessive construction.
On the relation-based view, the choice of a possessive construction de-
pends on the kind of relation that holds between the PM and the PR. (For
discussions of the relation-based view see Pawley 1973; Pawley and Sayaba
1990; Lynch 1973, 1982; Lichtenberk 1983b, 1985.) For example, if the PM
is part of the PR’s own body, the direct possessive construction is (typi-
cally) used; if the PM is an article of food for the PR, an indirect possessive
construction with a certain possessive classifier is used; if the PM is an item

of drink for the PR, an indirect possessive construction with a different pos-
sessive classifier is used; and so on. (It is for this reason that the possessive
classifiers are termed “relational classifiers” in Lichtenberk 1983b.) Some
of the detailed studies of possessive constructions in individual languages
that adopt the relation-based view do, however, mention the existence of
exceptions.
We can now consider the semantic/pragmatic properties of the different
types of possessive construction.
3.3. Direct possessive constructions
In direct possessive constructions, the PM noun carries an affix that indexes

the PR. Direct possessive constructions are overwhelmingly used to express
inalienable possession, where the PM noun is inherently relational. This
does not mean, however, that all types of inalienable possession are ex-
pressed by means of the direct construction (see section 5). There are several
subtypes of inalienable possession that are normally expressed by the direct
construction. These are discussed in A–I below.
A. Parts of a whole, body parts. Included here are also concepts such as
body and integral contents of a PR, such as blood (in the PR’s body) and
juice (e.g. of fruit):
Paamese
(33) vati-n
head-3SG:POSS
‘his/her head’ (Crowley 1996: 389)
Toqabaqita
(34) suul-a fa qota
juice-3SG:POSS CLF areca.nut
‘juice of an areca nut (being chewed)’
(fa is a “numeral” classifier used in noun phrases referring to fruit
and certain other entities, not a possessive classifier)5
5
As (34) shows, the use of the classifier fa is not dependent on the presence of a
numeral in the noun phrase.
B. Natural bodily products that emanate from the PR’s body, and other
products of physical bodies: for example, ‘tears’, ‘sweat’, ‘urine’, ‘faeces’,
‘semen’, ‘voice, sound’ (produced by the PR), ‘breath’, ‘smell/scent’ (ex-
uded by the PR); and also ‘shadow, shade’ (cast by the PR), ‘reflection (of
the PR, e.g. in water), ‘picture’ or some other representation of the PR, all
of which are often part of a polysemy:
Manam
(35) boro tae-di
pig faeces-3PL:POSS
‘pigs’ excrements’ (Lichtenberk 1983a: 279)
Tamambo
(36) nunu-ku
photo/reflection/picture/shadow-1SG:POSS
‘my photo/reflection/picture/shadow’ (a likeness of me)
(Jauncey 1997: 229)
C. Entities, matter on the surface of the PR’s body. Included here are
concepts such as sores, dirt, tattoes, clothing (especially, though not
necessarily, when being worn by the PR), and parasites such as lice:
Banoni
(37) kipi-na-i moono
dirt-3SG:POSS-ART girl (Lynch and Ross 2002: 445,
‘the girl’s dirt’ from Lincoln 1976)
Lolovoli
(38) tatai-ne
tattoo-3SG:POSS
‘her tattooes’ (Hyslop 2001: 171)
D. Mental organs, states, and products of mental processes:
Kilivila
(39) nano-gu
mind-1SG:POSS
‘my mind’ (Senft 1986: 45)
Lolovoli
(40) domi-mu
thought-2SG:POSS
‘your thoughts’ (Hyslop 2001: 172)
E. Attributes such as the PR’s shape, size, name and age:
Nalik
(41) a nounau-naande
ART shape-3PL:POSS
‘their shape’ (also ‘their interest in something’) (Volker 1998: 130)
Puluwatese
(42) yii-e-mw
age-SV-2SG:POSS
‘your age’ (Elbert 1972: 283)
F. Spatial and temporal relations. The possessive affixes are added to

spatial (and temporal) prepositions or to what is sometimes referred to as
relator nouns. Such prepositions and relator nouns usually derive historically
from nouns that designate spatial aspects of objects, especially human bodies,
and certain of their parts (such as the face or the back) (Bowden 1992).
Toqabaqita
(43) qi ninima-ku
at beside-1SG:POSS
‘beside me’
Lolovoli
(44) Lo tagu-i bongi gai-vesi ….
LOC behind-CONSTR day NUM-four
‘After four days ….’ (Hyslop 2001: 176)
G. Kinship categories and certain other categories of social/cultural re-

lations. The latter categories include concepts such as ‘friend’ and ‘partner’
(e.g. trading partner).
Kilivila
(45) ina-si
mother-3PL:POSS
‘their mother’ (Senft 1986: 140)
Iaai
(46) ihumwi-
friend-3SG:POSS
‘his/her friend’ (Ozanne-Rivierre 1976: 157)
H. The PR is a Patient, Theme or Stimulus (of emotion or of sensory per-

ception). In a number of languages the direct construction is used when the
PR has one of these roles in a transitive situation: Patient or Theme acted
on by Agent, or Stimulus perceived by Experiencer. Often in such cases the
PM phrase is, or contains, a nominalization of the corresponding verb. Such
possession is sometimes referred to as “passive” (see, for example, Lynch
2001), as opposed to “active” possession (section 4 below).
Manam
(47) udi tanom-a-di
banana plant-NMLZ-3PL:POSS
‘the planting of the bananas’ (Lichtenberk 1983a: 251)
Toqabaqita
(48) riki-la-na wane baa
look.at-NMLZ-3SG:POSS man that
‘that man’s appearance/look/mien’ (i.e., the way the man appears to
be to others, how others see him) (lit.: ‘that man’s looking-at’)
The PM need not be a nominalization; nevertheless, the PR has the role of a

Patient, Theme or Stimulus in the associated event:
Kokota
(49) mereseni-na mheke
medicine-3SG:POSS dog
‘medicine for dogs’ (Palmer 2002: 506)
In some other languages passive possession is expressed by means of an

indirect possessive construction, using a classifier (see [53] in section 3.4).
I. Emphatic pronominal forms. Such forms often carry the significance

of ‘by oneself’, ‘on one’s own’:
Paamese
(50) Inau nakanian s so-k.
1SG 1SG:REAL:eat self-1SG:POSS
‘I ate by myself.’ (Crowley 1996: 407)
Manam
(51) Rube-gu u-yalale.
alone-1SG:POSS 1S:REAL-go
‘I went alone, by myself.’ (Lichtenberk 1983a: 307)
3.4. Classifier systems (indirect possessive constructions)
As also mentioned in section 7, it is generally assumed that Proto Oceanic

had (at least) three formally distinct possessive classifiers, and we can take
the tripartite system as our starting point. Some languages have a tripartite
system; in some, the number of classifiers has been reduced to two, or one,
or none; and in some others it has been expanded beyond three, sometimes
considerably. (Standard) Fijian has three possessive classifiers: a “food /
passive” classifier ke-, used when the PM is an item of food for the PR (but
see below) or another entity metonymically related to food (such as places
where food is grown, and containers of food for the PR to eat), and it also
signals passive possession, where the PR is a Patient, Theme or Stimulus,
rather than an Agent or Experiencer in the associated situation (see category
H in section 3.3, and examples (85)–(88) and the accompanying discussion
in section 4); a “drink” classifier me-, used when the PM is an item of drink
for the PR (but see below) or another entity metonymically related to drink
(such as containers of drink for the PR to drink); and a “general” classifier
no-/ne-, used when none of the other classifiers nor the direct construction
are called for. Note that the food category subsumes only solid food; food
that is runny, mushy, juicy, suckable is included in the drink category (see
Pawley and Sayaba 1990 for Wayan, one of the Fijian languages). However,
tobacco (for smoking) is in the food category. Examples (52) and (53) illus-
trate food and passive possession, respectively:
(52) na ke-da vei-niu

ART POSS.CLF-PL(INCL):POSS group-coconut
‘our coconut plantation (which we eat from)’ (Pawley 1973: 162)
(53) ke-mu i-roba6

POSS.CLF-2SG:POSS NMLZ-slap
‘your slap (you are slapped)’ (Geraghty 1983: 249)
6
Geraghty (1983) calls the i- prefix a “preformative”.
Example (54) illustrates the drink classifier and (55) the general one:
(54) na me-mun
t
ART POSS.CLF-2PL:POSS tea

‘your tea’ (Milner 1972: 66)
(55) na no-mu waqa

ART POSS.CLF-2SG:POSS canoe
‘your canoe’ (Milner 1972: 65)
In Manam the number of possessive classifiers has been reduced to two.

This is due to the merger of the food and the drink categories into one “ali-
mentary” category. The classifier ana- (also an- and anan-) is used when
the PM is an item of food or drink for the PR or an entity metonymically
related to such, for example, gardens where food is grown, implements used
to obtain or eat food, containers of food and drink for the PR (including
personal baskets used to hold tobacco, areca nuts, lime and betel pepper for
chewing, and bottles). The other classifier is a general one, used in all cases
other than those that call for the alimentary one or for the direct construc-
tion. Examples (56) and (57) illustrate the alimentary classifier, and (58) the
general one:
(56) ulu ana-mi

breadfruit POSS.CLF-2PL:POSS
‘your breadfruit (to eat)’ (Lichtenberk 1983a: 291)
(57) botoli ana-gu

bottle POSS.CLF-1SG:POSS
‘my bottle’ (drink container) (Lichtenberk 1983a: 293)
(58) tamoata asi ne-ø

man bushknife POSS.CLF-3SG:POSS
‘the man’s bushknife’ (Lichtenberk 1983a: 294)
There are languages that have only one possessive classifier. In one respect,
such forms are unlike possessive classifiers in other languages, because
they do not contrast with other classifiers. Nevertheless, the term “classi-
fier” is retained here for two reasons. First, systems with single classifiers
are historical reductions of systems with multiple classifiers. And second,
constructions with a simple classifier do contrast with direct possessive
constructions, without a classifier.
Houaïlou is one language with a contrast between a direct possessive

construction and a single indirect one. The sole possessive classifier is used
whenever the direct construction is not appropriate (and vice versa). It is
used even with some (but not all) kinship terms. Examples (59) and (60)
illustrate the classifier:
(59) d v i-vu 7

garden POSS.CLF-1DU(EXCL):POSS
‘our garden’ (Leenhardt 1932: 192)
(60) p vaa i- a

father POSS.CLF-1SG:POSS
‘my father’ (La Fontinelle 1976: 300)
A possessive system with an opposition between a direct construction and

an indirect construction with only one possessive classifier is also found in
’Ala’ala (Ross 2002b).
A sole possessive classifier contrasts only with the direct possessive
construction, and so, strictly speaking, is not necessary, provided the con-
trast is made in some other way. This is the case in Toqabaqita, which has
no possessive classifiers whatsoever. The contrast is between a direct/suf-
fixing construction, where the PM noun carries a possessive suffix, and a
bare construction, where there is no indexing of the PR on the PM noun. In
some of its aspects the bare construction is analogous to the system of indi-
rect constructions in other languages (but see section 5).
(61) fanga nia

food 3SG
‘his/her food’
(62) biqu wane baa

house man that
‘that man’s house’
(Nia in [61] is the third person singular independent pronoun.)

A different kind of reduction in the number of possessive classifiers has
taken place in the Polynesian languages, with their binary systems of A-
possession and O-possession (and no direct possessive construction). The
Leenhardt (1932) gives the form as dov xivu; the representation d v i-vu is
7
in accordance with La Fontinelle (1976).

details of the use of the two constructions vary from language to language.
Wilson (1982) has put forward an “Initial Control Theory” to account for
the uses of A- and O- possession in Polynesian: the choice of a possessive
construction is determined by whether or not the PR has control over the
initiation of the relation. If the PR does have control, A-possession is used;
if the PR does not have control, O-possession is used. This is illustrated for
Hawaiian in the next pair of examples. In (63) the PR initiates the relation
by having the child, and so A-possession is used. On the other hand, in (64)
the PR does not initiate the relation to his parent, and so O-possession is
used.
(63) k- -na keiki

ART-POSS.CLF-3SG:POSS child
‘his child’ (Wilson 1982: 19)
(64) k-o-na makua

ART-POSS.CLF-3SG:POSS parent
‘his parent’ (Wilson 1982: 19)
We can now turn to languages in which the number of possessive classifiers

is greater than the three-member system with a food–drink–general con-
trast. Moderate expansions are found in some of the languages of Vanuatu.
Lolovoli has a food classifier, a drink classifier, a general classifier and a
classifier for “natural or valued object possession” (Hyslop 2001: 176). The
latter classifier, whose form is bula, is mainly used to express ownership of
animals and crops:
(65) bula-na boe

POSS.CLF-3SG:POSS pig
‘his pig’ (Hyslop 2001: 178)
A more complex classifier system exists in Lenakel, which has a food classi-
fier, a drink classifier, a plant classifier (for plants planted by the PR), a
general classifier, and a classifier for locations (occupied by the PR). The
latter classifier is only optional, and the general one may be used instead.
The location classifier is illustrated in (66):
(66) tn iimwa-nil-lau

land POSS.CLF-3NONSG-DU
‘their homeland’ (Lynch 1978: 81)
In some languages the original classifier system has undergone great ex-
pansion. This is the case in most Micronesian languages. (Lee 1975: 111)
gives a list of 19 “commonly used classifiers” for Kosraean. The list includes
classifiers for transportation; land and shelters; plants; tools, pets and toys;
drink; several classifiers for food; several classifiers for kinship relations;
and several classifiers for decorations. And there is a general classifier. Ex-
ample (67) contains the tool/pet/toy classifier:
(67) mos nuhti-k

breadfruit POSS.CLF-1SG:POSS
‘my breadfruit for toy’ (Lee 1975: 117)
For Pohnpeian, Rehg (1981) gives a list of 21 classifiers, which is not ex-
haustive. In fact, Rehg (1981: 179) says that “how many [possessive classi-
fiers] there are in Ponapean [Pohnpeian] is difficult to determine”.
Quite a few of the classifiers in the Micronesian languages are transpar-
ently related to nouns, and one finds cases of repeaters (Aikhenvald 2000),
where a classifier is used with a noun from which it has developed through
grammaticalization. The classifier and the source noun have identical or
very similar forms. For example, Ulithian has a vehicle classifier of the
form waa, which is used when the PM serves as a means of transportation,
such as a ship, a bicycle, a plane or a canoe, for the PR. The word for ‘ca-
noe’ too is waa:
(68) waa-yire waa

POSS.CLF-3PL:POSS canoe
‘their canoe’ (Sohn and Bender 1973: 268)
Outside of Micronesia a large system of possessive classifiers (including

repeaters) is found in Iaai (New Caledonia). Besides a general classifier,
there are classifiers for food, drink, chewable food, game killed or caught in
hunting or fishing, voice and sounds, land and various products, boats, and
so on. According to Ozanne-Rivierre (1976: 189), the set of Iaai possessive
classifiers is open, and a comprehensive list is difficult, if not impossible, to
establish.
Mussau (Papua New Guinea) too has a relatively large set of possessive
classifiers. Ross (2002c) lists nine of them, but the list is not necessarily
exhaustive.
4. Fluidity in the possessive systems
A feature of Oceanic possessive systems regularly commented on in gram-

mars is the fact that some nouns can occur in the PM position of more than
one type of possessive construction. The phenomenon is sometime referred
to as “overlap” (following Lynch 1973). Here the term “fluidity” will be
used (pace Nichols [1992]). How fluid a possessive system is varies from
language to language, but fluidity is by no means uncommon. There is flu-
idity between the direct possessive construction type and an indirect/classi-
fier construction type, and there is fluidity between different indirect/classi-
fier constructions. Sometimes the fluidity has to do with different senses of
a noun. Such cases are illustrated first, starting with fluidity between the
direct and the indirect construction types. In (69), from Tamambo, nunu has,
among others, the sense of ‘pictorial representation of (the PR)’, while in
(70) it has the sense of ‘object that carries a pictorial representation of some-
thing (not necessarily of the PR)’:
(69) nunu-ku
photo/reflection/picture/shadow-1SG:POSS
‘my photo/reflection/picture/shadow’ (a likeness of me)
(Jauncey 1997: 229)
(70) no-ku nunu

POSS.CLF-1SG:POSS photo/picture
‘my photo(s)/picture(s) that belong(s) to me’ (Jauncey 1997: 229)
The different senses of a noun may also call for the use of different posses-
sive classifiers, as in (71) and (72) from Araki. In (71), where the sense of
the PM noun is ‘pig’, it is the “economic possession” classifier pula- that is
used, while in (72), where the sense of the same PM noun is ‘pork’, it is the
food classifier ha- that is used:
(71) pula-ku po
‘my pig (I breed)’ (François 2002: 100)
(72) ha-ku po
‘my piece of pork (to eat)’ (François 2002: 100)
And in (73) and (74) from Lolovoli there is a contrast between the food and
the drink possessive constructions, depending on whether the PM noun has
the sense of citrus fruit to eat or citrus juice to drink:
(73) ga-ku moli

POSS.CLF-1SG:POSS citrus
‘my orange/pomelo (for me to eat)’ (Hyslop 2001: 185)
(74) me-ku moli

POSS.CLF-1SG:POSS citrus
‘my orange/pomelo juice’ (Hyslop 2001: 185)
However, there is fluidity even when polysemy is not involved. In some

cases, the use of different possessive constructions has to do with different
referents or different kinds of referents. Examples (75) and (76) from Fijian
illustrate. In both cases the PM noun has the sense of mango as fruit, but
when a mango is green, unripe, it is “eaten” and the food classifier is called
for, but when a mango is ripe and juicy, it is “sucked” when being con-
sumed, and the drink classifier is called for:
(75) na ke-na maqo

ART POSS.CLF-3SG:POSS mango
‘his mango for eating (i.e. green mango)’ (Pawley 1973: 168)
(76) na me-na maqo

‘his mango for sucking (i.e. ripe, juicy mango)’ (Pawley 1973: 168)
And in (77)–(79) from Manam there is a three way contrast with the noun
‘head’ in the PM position, and in each case the sense of the noun is that of a
body part. However, in (77) the head is part of the PR’s own body and so
the direct construction is used; in (78) the head is food for the PR and so
the alimentary classifier construction is used; and in (79) the head is neither
part of the PR’s own body nor food for him/her, and it is the general classi-
fier construction that is used:
(77) paana-gu
head-1SG:POSS
‘my head (part of my body)’ (Lichtenberk 1983a: 302)
(78) paana ana-gu

head POSS.CLF-1SG:POSS
‘my head to eat (e.g. a fish head)’ (Lichtenberk 1983a: 302)
(79) paana ne-gu

head POSS.CLF-1SG:POSS
‘my head (e.g. a head I found, cut off, or a head I will give my dog to
eat)’ (Lichtenberk 1983a: 302)
Fluidity is found even if the PM referent is one and the same entity but is
conceptualized differently. As shown in (75) and (76) above from Fijian,
the noun ‘mango’ can occur in the PM position with the food or the drink
classifier. However, one and the same mango can serve a purpose other
than being for the PR’s consumption, for example as something to be sold,
in which case the general possessive classifier is required, and the distinc-
tion between a mango as food and a mango as “drink” disappears:
(80) na no-na maqo

‘his mango (as property, e.g., which he is selling)’ (Pawley 1973: 168)
In Manam the noun for ‘grass-skirt’ occurs as PM in the direct construction

when a grass-skirt is being worn by the PR at the relevant time, but when
the same grass-skirt is not being worn, it is the indirect construction with
the general classifier that is called for:
(81) baligo-gu
grass.skirt-1SG:POSS
‘my grass-skirt (when I am wearing it)’ (Lichtenberk 1983a: 301)
(82) baligo ne-gu

grass.skirt POSS.CLF-1SG:POSS
‘my grass-skirt (when I am not wearing it)’ (Lichtenberk 1983a: 301)
In the examples just given, the PR is (or may be) one and the same person.
There is a different kind of fluidity, where one and the same entity is re-
ferred to by means of different possessive constructions because of different
perspectives due to different PRs. For example, one and the same woman
may be one person’s wife and another person’s sister. In Kosraean, two dif-
ferent classifiers are used:
(83) muhtwacn kiyuh-k

woman POSS.CLF-1SG:POSS
‘my wife’ (Lee 1975: 118)
(84) muhtwacn wiyuh-k

woman POSS.CLF-1SG:POSS
‘my sister’ (Lee 1975: 118)
Kosraean, which has a large set of possessive classifiers (section 3.4), has
several kinship classifiers. Example (83) contains the classifier for mothers
and wives, and (84) one of the classifiers for siblings.
Fluidity having to do with PR perspective is common in the opposition
between passive and active possession. In passive possession the PR is a
Patient, Theme or Stimulus in the relevant situation (section 3.3), and cor-
respondingly in active possession the PR is an Agent or Experiencer. One
and the same state of affairs can be encoded from the perspective of the
Patient/Theme/Stimulus or that of the Agent/Experiencer. This is the case in
(85) and (86) from Fijian, with the food/passive and the general classifiers,
respectively:
(85) ke-mu i-vacu

POSS.CLF-2SG:POSS NMLZ-punch
‘your punch (you receive)’ (Schütz 1985: 462)
(86) no-mu i-vacu

POSS.CLF-2SG:POSS NMLZ-punch
‘your punch (you give)’ (Schütz 1985: 462)
In the Fijian examples the formal contrast is between two indirect construc-
tions. In (87) and (88) from Manam the contrast is between a direct con-
struction and an indirect construction. The PM in the direct construction is
a nominalization of the verb nanari-t-a ‘tell a story about’.
(87) nanari-t-a-a-gu
tell.story-THC-TRANS-NMLZ-1SG:POSS
‘my story (story about me)’ (Lichtenberk 1983a: 303)
(88) nanari ne-gu

story POSS.CLF-1SG:POSS
‘my story (e.g. one that I invented, told or like)’
(Lichtenberk 1983a: 303)
5. Exceptions in the use of possessive constructions
As discussed in sections 3 and 4, there are clear semantically/pragmatically

based patterns in the use of the possessive constructions. However, descrip-
tions of some languages, especially more detailed descriptions, also mention
cases where the choice of a possessive construction for a given PM does
not follow the general patterns in that language. (Exceptions, or apparent
exceptions, to the use of classifiers are not unusual in classifier systems in
general; see the quote from Aikhenvald 2000: 82 in section 6.) Some of
these cases are genuine exceptions, in the sense that there is no explanation
available for their existence. In other cases, however, the exceptions are
systematic, and their existence can be accounted for. In the discussion that
follows, the focus will be on cases that involve distinctions between the
direct and the indirect possessive constructions, specifically cases where
the direct construction would be expected but is not used. The cases will be
divided into two broad categories, one having to do with kinship terms, and
one having to do with body-part terms and terms for concepts having close
association with the body.
In Anejom some kinship terms, such as ‘father’, ‘grandparent’ and ‘wife’,
occur in the direct construction, while others occur in one of two indirect
constructions: for example, ‘husband’ takes the general classifier, and ‘sister’
and ‘brother’ take the passive-possession classifier. Lynch (2000) does not
attempt to provide an explanation for the use of the indirect constructions.
In Houaïlou the term for ‘mother’ occurs in the direct construction, but the
term for ‘father’ occurs in the only type of indirect construction. La Fonti-
nelle (1976) offers no explanatory comment.
On the other hand, in some languages the use of more than one type of
possessive construction is said to be (at least partially) pragmatically moti-
vated. In Kairiru, of the 18 kinship terms recorded by Wivell (1981) one half
occurs in the direct/suffixing construction and the other half in another,
bare construction, where the PM carries no indexing of the PR. The latter
construction is also used to express alienable possession. Wivell (1981: 54)
characterizes the distinction thus: “Those terms that were inalienable were
the ones that expressed important kin relations within the social structure,
while the alienable ones were those that expressed fairly unimportant
roles.” The former category includes the terms for ‘father’, ‘mother’ and
‘child’, and the latter category the terms for ‘mother’s brother’s wife, hus-
band’s sister’s daughter’ and ‘great grandparent, great grandchild’.
In Gapapaiwa, according to McGuckin (2002: 304), “[k]in terms for
persons who are peers or subordinates are possessed in direct constructions,
while terms for persons in authority over ego require an indirect alienable
construction”, for example, ‘spouse’ and ‘mother’, respectively.
In some languages the exceptional treatment of some kinship terms is
due to lexical replacement. In such cases, terms that are relatively new to the
language do not (necessarily) select the direct construction. For example, in
Toqabaqita the term for ‘mother’ occurs in the direct construction but the
term for ‘father’ does not:
(89) thaina-ku
mother-1SG:POSS
‘my mother’
(90) maka nau

father 1SG
‘my father’
Thaina continues Proto Oceanic *tina ‘mother’,8 while maka is a lexical

innovation. There is an archaic term for ‘father’, thaama, which continues
Proto Oceanic *tama ‘father’, and it occurs in the direct construction. Simi-
larly, the term for ‘child’, wela, is a lexical replacement (cf. Proto Oceanic
*natu) and occurs in the bare/non-suffixing construction, not in the direct /
suffixing construction.
In Wayan there are two types of possessive construction in which kin-
ship terms occur as PMs, and, according to Pawley and Sayaba (1990: 158),
“[t]here are no clear semantic grounds for the split”. However, Pawley and
Sayaba point out that one of the two structures is an innovation unique to
Western Fijian. It was originally used for part-of-whole terms, some of
which later acquired kinship meanings. It is this class that is open, admitting
new kinship terms, while the other class is closed.
Exceptions are also found in some languages with body-part terms and
terms having to do with the body. In a detailed study of inalienable posses-
sion in Paamese, Crowley (1996) concludes that the choice of a possessive
construction is not fully predictable on semantic grounds. Most nouns that
refer to internal organs occur in an indirect construction, and Crowley
(1996: 398) offers an explanation: “Internal organs are the kind of things
that would normally only be directly observed when there is a dead body
that has been opened up. The possession of these items by the butcherer of
the animal is clearly transferrable and so there is an alienable relationship
8
The Proto Oceanic reconstructions are from Lynch et al. (2002).
between him and the body parts, in contrast to the formerly inalienable rela-
tionship between the animal and the body parts when it was still alive.”
Note, however, that the indirect construction is also used when the internal
organ is part of the PR’s, for example, an animal’s, own body, and so the
implication in what Crowley says is that the use of the indirect construction
expressing the alienable nature of the relation between an animal’s internal
organ and the butcherer has been extended to the relation when the organ is
part of the PR’s own body. Be that as it may, the fact remains that the noun
that refers to a bird’s crop occurs in the direct construction, not the indirect
one, and there is no obvious explanation for that. Outside of the area of in-
ternal organs, there is no explanation for the fact that the noun for ‘shin’
occurs in the indirect construction although the noun for ‘lower leg’ occurs
in the expected direct construction.
A different kind of exceptional behaviour of body-part terms is found in
Toqabaqita. Such nouns occur in the direct/suffixing construction unless they
are in the scope of a modifier other than the PR, such as a verb,9 a numeral
or a demonstrative. Compare (91), where the noun ‘eye’ occurs in the direct/
suffixing construction, and (92) and (93), where the same noun occurs in the
bare/non-suffixing construction:
(91) maa-ku
eye-1SG.POSS
‘my eye(s)’
(92) maa mauli nau

eye be.on.left.side 1SG
‘my left eye’
(93) maa nau naqi

eye 1SG this
‘this eye of mine’
The bare/non-suffixing construction is also used to express alienable pos-

session; see (61) and (62) in section 3.4.
There is no difference in inalienability between ‘my eye(s)’ on the one
hand and ‘my left eye’ and ‘this eye of mine’ on the other. As argued in
Lichtenberk (2005b), the relevant factor is that of PM individuation. When
the PM is not in the scope of a modifier other than the PR, it is not individu-
9
In Toqabaqita, verbs can directly modify nouns.
ated vis-à-vis the PR; it is viewed basically as an aspect of the PR. On the
other hand, through specification the PM is individuated, given more iden-
tity with respect to the PR, and there the same construction is used that
serves to express alienable possession, where the PM is always individu-
ated with respect to the PR.
6. The relational basis of the Oceanic possessive systems
In a detailed study of the system of attributive possessive constructions in

Wayan (a Western Fijian language) Pawley and Sayaba (1990) ask the fol-
lowing question: with respect to possessive marking, is the system one of
noun classes or is it relational? The conclusion they reach is that it is a mix-
ture of both: “Certain nouns belong to strict and semi-arbitrary noun classes,
for purposes of possessive-marking, others show marking consistently fol-
lowing semantic principles.” (p. 168). They suggest that their findings apply
in their basics also to (Standard) Fijian. And the conclusions they reach with
respect to Wayan and Fijian may be of relevance to Oceanic in general.
In the present study, on the other hand, it has been argued that the Oce-
anic systems are, on the whole, based on semantic/pragmatic principles, the
crucial factor being the nature of the relation between the PM and the PR,
even though there are also genuine exceptions. There is then some arbi-
trariness, but such cases are exceptions against the backdrop of semanti-
cally/pragmatically motivated systems.
While certain of Pawley and Sayaba’s conclusions concerning Wayan
are applicable to other Oceanic languages, there is some clarification that is
in order. They use the term “relational analysis” (or “hypothesis”) to iden-
tify an approach to Oceanic possessive constructions in which the choice of
a construction type is viewed as having to do with the nature of the relation
between the PM and the PR. (They mention the existence of several variants
of the relational analysis.) They contrast the relational analysis with a noun-
class analysis, according to which each noun belongs to a certain class de-
pending on the type of possessive construction it selects, and characterize it
in this fashion:
We take the relational hypothesis to entail not only the claim that: (i) posses-
sive marking is determined by the semantic relation holding between pos-
sessed and possessor, but that (ii) this relation is not constant for all situations.
That is, for any noun, speakers have some choice of possessive marker,
constrained only by their imaginations or belief systems. Any constraints
which can not be readily accounted for in these terms must be regarded as
grammatical not semantic constraints. (Pawley and Sayaba 1990: 169)
In the present study, the Oceanic possessive systems have been character-
ized as basically relational in nature: exceptions apart, the choice of the
type of possessive construction depends on the nature of the PM–PR rela-
tion. This corresponds to point (i) in the quote from Pawley and Sayaba.
Their point (ii) has to do with fluidity. Fluidity is, of course, evidence of the
relational nature of the system: one and the same noun occurs in the PM
position of different types of possessive construction depending on the na-
ture of the relation between the PM and the PR. Pawley and Sayaba say that
according to the relational hypothesis “for any noun, speakers have some
choice of possessive marker, constrained only by their imaginations or be-
lief systems” (see the quote above; emphasis added here). I am not aware of
any such strong version of the relational analysis, according to which any
noun in a given language exhibits, in principle, fluidity. (See also further
below on fluidity and rigidity.)
Importantly, Pawley and Sayaba’s point (i) is independent of their point
(ii). Fluidity is not necessary for a system to be relational. Even if no noun
exhibited fluidity, where each noun occurred in only one type of possessive
construction and so could be said to belong in a certain noun class, the over-
all system could still be relational if each noun class was defined by the na-
ture of the relation between the PM and the PR. An unstated assumption in
Pawley and Sayaba’s approach is that noun classes are essentially arbitrary.
This, however, is not the case. As Aikhenvald (2000: 21) points out: “There
is always some semantic basis to the grouping of nouns into classes, but
languages vary in how much semantic transparency there is.” And: “In lan-
guages with purely semantic assignment the class of a noun can be inferred
from its meaning.” (p. 22). Thus, noun classes are semantically motivated to
various degrees, even though in a noun-class system “[e]ach noun … belongs
to one (or occasionally more than one) class(es)” (Aikhenvald 2000: 21).
Classifier systems (for example, numeral-classifier systems) too are based
on semantic/pragmatic principles. Sometimes there is fluidity, but there are
also exceptions: “The choice of a classifier is usually semantically trans-
parent; in some cases, however, the semantic link between a noun classifier
and a noun is not obvious.” (Aikhenvald 2000: 82).
It is not lack of fluidity that counts as an exception. Only those nouns
that require a possessive construction that goes against the general pattern
are exceptional. Such exceptions apart, the Oceanic possessive systems are
basically relation-based, regardless of the degree of fluidity they permit.
One more remark on fluidity is in order. Descriptions of Oceanic lan-
guages that do comment on fluidity (and most of them do), say explicitly or
imply that certain nouns do not have fluidity, that they can occur only in one
type of possessive construction. The question that needs to be asked is to

what extent such lack of fluidity is lexical/grammatical and to what extent
it is semantic/pragmatic. It may be that certain kinds of fluidity do not occur
because the kind of semantic/pragmatic relation that a given type of posses-
sive construction expresses would be highly implausible or even absurd for
a certain pairing of a PM and its PR. Thus, it is unlikely that the noun for
‘father’ would occur in the PM position in the food or the drink possessive
construction. But is this a grammatical constraint or a matter of pragmatics?
On the other hand, less than one-hundred-percent fluidity does not in itself
mean that (genuine exceptions apart) the possessive system in a given lan-
guage does not have a relational basis.
At present, we have mostly only brief comments on creativity with re-
spect to possessive fluidity in individual languages; what is needed is in-
depth studies. It may turn out that there is more fluidity than meets the eye.
7. Emergence of the possessive classifier system
It is generally agreed that Proto Oceanic had a direct possessive construction

and (at least) three possessive classifiers used in indirect constructions: a
food, a drink and a general one (Lichtenberk 1985, Lynch et al. 2002).
Lynch et al. (2002) suggest that Proto Oceanic may have had other posses-
sive classifiers, including another, large-member set of classifiers, not unlike
those found in most of the Micronesian languages, Iaai and Mussau (section
3.4). However, although such large-member systems are found in different
primary subgroups of Oceanic, it is quite likely that those are later, inde-
pendent developments from the more restricted, probably tripartite, system.
This kind of development is quite natural.
Possessive classifiers may have begun to develop shortly before the
Proto Oceanic stage, but the historical evidence is unequivocal that earlier
there had been only one basic type of possessive construction, which was of
the direct/suffixing type. The following question then arises: why did a sys-
tem of possessive classifiers develop for alienable possession but not for
inalienable possession? I have discussed this issue elsewhere (Lichtenberk
2005b) and so only a brief synopsis will be given here, together with new,
albeit indirect evidence. In inalienable possession there is typically a highly
salient relation between a PM and its PR: a kinship relation, a part-whole
relation, or some other kind of intrinsic relation (for example, Barker 1995).
The interpretation of the relation as the salient one is quite stable across
different contexts (although it can be overridden). In alienable possession,
on the other hand, there is often no such strongly salient, context-stable re-
lation, and correspondingly the referents of one and the same PM phrase
can stand in various relations to their PRs. This was illustrated in the dis-
cussion of fluidity in section 4, in particular by the “mango” examples (75),
(76) and (80) from Fijian: a mango as an item of food, or as an item of
“drink” or as property to be sold. In alienable possession, the interpretation
of the relation between a PM and its PR is highly variable across contexts.
Possessive classifiers specify more closely the nature of the PM–PR re-
lation. The development of possessive classifiers for alienable possession is
well motivated because of the variability in the PM–PR relation. On the
other hand, in inalienable possession there is no such strong motivation be-
cause of the presence of a highly salient, default relation between the PM
and the PR. The default interpretation of linguistic constructions needs no
overt marking (cf. Haiman 1985; Dixon 1994; Croft 2001 [2002]). On the
other hand, there is motivation for there being an overt marker of a non-
default interpretation. For possessives, this is illustrated by the next pair of
examples from Manam. The default interpretation of ‘X’s skin’ is for the
skin to be part of the PR’s own body, as in (94), where the direct construc-
tion is used. However, if the intended interpretation is not the default one,
this is signalled by means of a classifier, as in (95), in this case the alimen-
tary classifier:
(94) usi-gu
skin-1SG:POSS
‘my skin (the skin of my body)’ Lichtenberk, field notes
(95) usi ana-gu

skin POSS.CLF-1SG:POSS
‘my skin (for me to eat, e.g. chicken skin)’ Lichtenberk, field notes
To test the hypothesis that there is a difference in the presence of a highly

salient, context-stable interpretation in inalienable possession, with intrinsi-
cally relational nouns, on the one hand, and in alienable possession with
nouns that are not intrinsically relational, on the other, a set of experiments
was performed with native speakers of English (Lichtenberk et al. 2004).
Early in the history of Austronesian there was only one type of attributive
possessive construction, without a distinction between inalienable and alien-
able possession. Similarly, English does not have a formal distinction be-
tween inalienable and alineable possession in its attributive possessive con-
structions, and so provides a good testing ground for the hypothesis.
In the experiment the subjects were presented with several sets of pos-
sessive noun phrases, some of which had inherently relational nouns as
PMs, for example his children, and others had nouns as PMs that are not
inherently relational, such as her cookies. For each stimulus the subjects
were asked to give one interpretation of the relation between the PM and
the PR. The results of the study convincingly demonstrated the existence
both of a PM effect and of a PR effect.
The PM effect has to do with the fact that the relational nouns as PMs
elicited a restricted range of interpretations of the PM–PR relations, while
the non-relational nouns elicited a broader range of interpretations. Fur-
thermore, with the relational nouns there was always one interpretation that
was clearly dominant, while such strong dominance was not found with the
non-relational nouns. Thus, the interpretation of his children was in terms
of a kinship relation, while her cookies elicited a variety of interpretations:
the cookies she owns, the cookies she made, the cookies she bought, the
cookies she will eat.
The PR effect has to do with the fact that the nature of the PR had a
greater effect on the interpretation of the PM–PR relation with PMs that are
not inherently relational, while with the inherently relational nouns as PMs
the interpretations were quite stable. Thus, for example, both for the sol-
dier’s legs and for the general’s legs the interpretation was uniformly that
of the legs being part of the soldier’s or the general’s own body. On the
other hand, for the soldier’s regiment the dominant interpretation was that
of the regiment the soldier is a member of, while for the general’s regiment
the dominant interpretation was that of the regiment the general is in charge
of. The PR effect is stronger with non-relational nouns than with relational
nouns because with non-relational nouns there is typically no intrinsic, sali-
ent relation between the PM and the PR. It is true that a certain kind of PR
is likely to favour a certain kind of interpretation, but such contextual fac-
tors are absent or attenuated with “neutral” PRs, such as possessive deter-
miners, for example, her cookies.10
Although the study was done on English, the assumption is that very
much the same cognitive factors operate in other languages and that they
operated in the history of Austronesian, when the system of possessive
classifiers began to develop. Obviously, a system of possessive classifiers
does not have to develop (after all, they are not common in the languages of
10
William B. McGregor has suggested (pers. comm., 11 August 2007) that the
gender of the PR might be relevant in some cases; cf. her cookies and his coo-
kies. The study discussed here did not take gender into account.
the world), but if they do develop they are more likely to develop, at least
initially, for alienable possession, with PM nouns that are not inherently
relational, than for inalienable possession, with PM nouns that are inher-
ently relational. (Some Micronesian languages do have possessive classifiers
for different kinship categories, as mentioned and illustrated for Kosraean
in section 4 above), but those languages have large sets of possessive clas-
sifiers, which are later developments, postdating the emergence of the first,
restricted set.)
8. Summary and conclusion
With very few exceptions, Oceanic languages have more than one type of
attributive possessive construction. In the typical system, there is a distinc-
tion between a direct possessive construction, where the PM noun carries
affixes that index the PR, and more than one subtype of indirect construc-
tion, where the possessive affixes are attached to a possessive classifier.
The direct construction type is strongly associated with inalienable posses-
sion, where there is an intrinsic link between the PM and the PR. There are,
however, also language specific exceptions where certain PMs take an indi-
rect construction rather than the direct one.
With some exceptions, the choice of a possessive construction depends
on the nature of the relation between the PM and the PR. This relational
nature of the choice of possessive construction is particularly strongly evi-
denced by fluidity in the possessive systems. Sometimes such fluidity is due
to the polysemy of a noun, but in some cases there is fluidity without poly-
semy, and the choice of a possessive construction depends on the pragmatics
of the situation (for example, whether or not an article of clothing is being
worn at the relevant time).
When there is no fluidity, one can say that the given noun belongs in a
certain class (the classes being established on the basis of the type of pos-
sessive construction used), but such classes are themselves by and large
semantically/pragmatically grounded in the nature of the PM–PR relation.
To say that the Oceanic possessive systems are relational in nature and
semantically/pragmatically motivated does not mean that the choice of a
possessive construction is always predictable, even disregarding genuine
exceptions. As pointed out in section 3.4, the type of possessive construc-
tion that a noun selects may be based on metonymy. Thus, in Manam the
noun for ‘garden’ selects the alimentary (food and drink) possessive classi-
fier, because gardens are places where food is grown (Lichtenberk 1983a).
However, not in all languages does the noun ‘garden’ select the food or ali-
mentary classifier. For example, in Anejom it selects the general classifier
or the customary-possession classifier, even though the language has a food
classifier (Lynch 2000); and in Zabana it selects the general classifier, even
though the language has an alimentary classifier (Fitzsimons 1989). Meton-
ymy is language/culture specific. A system may be sematically/pragmati-
cally motivated, but that does not mean that everything is predictable.
In Oceanic, possessive classifiers are used (again with some exceptions)
to express alienable possession. The development of possessive classifiers
for alienable possession was motivated by the fact that with nouns that are
not inherently relational there is typically no highly salient, context-stable,
relation between the PM and the PR. A classifier specifies more closely the
type of the relation. There are specific classifiers, such as food and drink.
There is also a general classifier (provided a language has more than one
classifier), which only signifies that the relation is not any one of the more
specific types. In inalienable possession, where the PM noun is intrinsically
relational, there is normally a highly salient, context-stable kind of relation
between the PM and the PR, and the development of classifiers there is
much less motivated.
While the development of a system of possessive classifiers for alien-
able possession was motivated, it is also a fact that the original system of
(at least) three classifiers has been simplified in some languages or has dis-
appeared altogether. While cognitive factors may motivate the existence of
a grammatical construction or contrast, they do not determine their exis-
tence, and such factors may be overridden by other kinds of development.
Acknowledgement
I am grateful to William B. McGregor for detailed comments on an earlier

version of this paper.
References
Aikhenvald, Alexandra Y.
2000 Classifiers: A Typology of Noun Categorization Devices. Oxford:
Oxford University Press.
Barker, Chris
1995 Possessive Descriptions. Stanford: CSLI Publications.
Benthem, Johan van and Alice ter Meulen

1997 Handbook of Logic and Language. Amsterdam: Elsevier; and Cam-
bridge, MA: MIT Press.
Bowden, John
1992 Behind the Preposition: Grammaticalisation of Locatives in Oceanic
Languages. Canberra: Pacific Linguistics.
Chappell, Hilary and William B. McGregor (eds.).
de Gruyter.
Clark, Ross
2000 Possessive markers in Polynesian languages. In Steven R. Fischer
(ed.), 258–268.
Croft, William
2001 Radical Construction Grammar: Syntactic Theory in Typological
Perspective. [Reprinted in 2002 with corrections]. Oxford: Oxford
University Press.
Crowley, Terry
1996 Inalienable possession in Paamese grammar. In Hilary Chappell and
William B. McGregor (eds.), 383–432.
Crowley, Terry
1998 An Erromangan (Sye) Grammar. Honolulu: University of Hawai‘i
Press.
Crowley, Terry
2002 Vinmavis. In John Lynch, Malcolm Ross and Terry Crowley (eds.),
638–649.
Daniel, Michael
2005 Understanding inclusives. In Elena Filimonova (ed.), 3–48.
Davidson, Jeremy H.C.S. (ed.).
1990 Pacific Island Languages: Essays in Honour of G. B. Milner. London:
School of Oriental and African Studies, University of London; and
Honolulu: University of Hawai‘i Press.
Davis, Karen
2003 A Grammar of the Hoava Language, Western Solomons. Canberra:
Pacific Linguistics.
Dixon, R. M. W.
1994 Ergativity. Cambridge: Cambridge University Press.
Donohue, Mark
2002 Tobati. In John Lynch, Malcolm Ross and Terry Crowley (eds.), 186–
203.
Elbert, Samuel H.
1972 Puluwat Dictionary. Canberra: Pacific Linguistics.
Filimonova, Elena (ed.).
2005 Clusivity: Typology and Case Studies of the Inclusive–Exclusive
Distinction. Amsterdam /Philadelphia: John Benjamins.
Fischer, Steven R. (ed.).

2000 Possessive Markers in Central Pacific Languages. Special issue of
Sprachtypologie und Universalienforschung, Language Typology
and Universals, 53.
Fitzsimons, Matthew
1989 Zabana: a grammar of a Solomon Islands language. Unpublished MA
thesis, University of Auckland.
Frajzyngier, Zygmunt, Adam Hodges and David S. Rood (eds.)
2005 Linguistic Diversity and Language Theories. Amsterdam /Phila-
delphia: John Benjamins.
François, Alexander
2002 Araki: A Disappearing Language of Vanuatu. Canberra: Pacific Lin-
guistics.
Geraghty, Paul A.
1983 The History of the Fijian Languages. Honolulu: University of Hawai‘i
Press.
Haiman, John
1985 Natural Syntax: Iconicity and Erosion. Cambridge: Cambridge Uni-
versity Press.
Halim, Amran, Lois Carrington and S. A. Wurm (eds.)
1982 Papers from the Third International Conference on Austronesian Lin-
guistics, Vol. 1: Currents in Oceanic. Canberra: Pacific Linguistics.
Hooper, Robin
1985 Proto-Oceanic *qi. In Andrew Pawley and Lois Carrington (eds.),
141–167.
Hyslop, Catriona
2001 The Lolovoli Dialect of the North-East Ambae Language, Vanuatu.
Canberra: Pacific Linguistics.
Jauncey, Dorothy
1997 A grammar of Tamambo, the language of western Malo, Vanuatu.
Unpublished PhD thesis, Australian National University.
Keesing, Roger M.
1985 Kwaio Grammar. Canberra: Pacific Linguistics.
La Fontinelle, Jacqueline de.
1976 La Langue de Houaïlou (Nouvelle-Calédonie): Description Phono-
logique et Description Syntaxique. Paris: Société d’Études Linguisti-
ques et Anthropologiques de France.
Lang, Ewald, Claudia Maienborn and Cathrine Fabricius-Hansen (eds.).
2003 Modifying Adjuncts. Berlin/New York: Mouton de Gruyter.
1995 Possession and possessive constructions. In John R. Taylor and Robert
E. MacLaury (eds.), 51–79.
Lee, Kee-dong (with the assistance of Lyndon Cornelius and Elmer Asher)
1975 Kusaiean Reference Grammar. Honolulu: University Press of Hawai‘i.
Leenhardt, Maurice
1932 Documents Néo-Calédoniens. Paris: Institut d’Ethnologie, Université
de Paris.
Lichtenberk, Frantisek
1983a A Grammar of Manam. Honolulu: University of Hawai‘i Press.
1983b Relational classifiers. Lingua 60: 147–176.
1985 Possessive constructions in Oceanic languages and in Proto-Oceanic.
In Andrew Pawley and Lois Carrington (eds.), 93 –140.
2005a Inclusive–exclusive in Austronesian: an opposition of unequals. In
Elena Filimonova (ed.), 261–289.
2005b Inalienability and possessum individuation. In Zygmunt Frajzyngier,
Adam Hodges and David S. Rood (eds.), 339–362.
Lichtenberk, Frantisek, Jyotsna Vaid and Hsin-Chin Chen
2004 Alienable/inalienable possession in grammar and in the lexicon: evi-
dence from possessive phrase interpretation in English. Paper pre-
sented at the 4th International Conference on the Mental Lexicon,
University of Windsor, 30 June – 3 July 2004.
Lincoln, Peter C.
1976 Describing Banoni, an Austronesian language of Southwest Bougain-
ville. Unpublished PhD thesis, University of Hawai‘i.
Lynch, John
1973 Verbal aspects of possession in Melanesian languages. Oceanic
Linguistics 12: 69–102.
Lynch, John
1978 A Grammar of Lenakel. Canberra: Pacific Linguistics.
Lynch, John
1982 Towards a theory of the origin of the Oceanic possessive construc-
tions. In Amran Halim, Lois Corrington and S.A. Wurm (eds.), 243–
268.
Lynch, John
2000 A Grammar of Anejom. Canberra: Pacific Linguistics.
Lynch, John
2001 Passive and food possession in Oceanic languages. In Andrew
Pawley, Malcolm Ross and Darrell Tryon (eds.), 193–214.
Lynch, John and Malcolm Ross
2002 Banoni. In John Lynch, Malcolm Ross and Terry Crowley (eds.),
440–455.
Lynch, John, Malcolm Ross and Terry Crowley (eds.)
2002 The Oceanic Languages. Richmond, Surrey: Curzon.
Massam, Diane and Wolfang Sperlich

2000 Possession in Niuean. In Steven R. Fischer (ed.), 281–292.
McGuckin, Catherine
2002 Gapapaiwa. In John Lynch, Malcolm Ross and Terry Crowley (eds.),
297–321.
Milner, George B.
1972 Fijian Grammar. (3rd ed.). Suva, Fiji: Government Press.
Nichols, Johanna
1992 Linguistic Diversity in Space and Time. Chicago and London: Uni-
versity of Chicago Press.
Ozanne-Rivierre, Françoise
1976 Le Iaai: Langue Mélanésienne d’Ouvéa (Nouvelle-Calédonie). Pho-
nologie, Morphologie, Esquisse Syntaxique. Paris: Société d’Études
Linguistiques et Anthropologiques de France.
Palmer, Bill
2002 Kokota. In John Lynch, Malcolm Ross and Terry Crowley (eds.),
498–524.
Palmer, Bill and Dunstan Brown
2007 Heads in Oceanic indirect possession. Oceanic Linguistics 46: 199 –
209.
Partee, Barbara H.
1997 Genitives – a case study. Appendix B to Theo M. V. Janssen (1997),
Compositionality. In Johan van Benthem and Alice ter Meulen (eds.),
464–470.
Partee, Barbara H. and Vladimir Borschev
2003 Genitives, relational nouns, and argument-modifier ambiguity. In
Ewald Lang, Claudia Maienborn and Cathrine Fabricius-Hansen
(eds.), 67–112.
Pawley, Andrew
1973 Some problems in Proto-Oceanic grammar. Oceanic Linguistics 12:
103–188.
Pawley, Andrew and Lois Carrington (eds.).
1985 Austronesian Linguistics at the 15th Pacific Science Congress. Can-
berra: Pacific Linguistics.
Pawley, Andrew, Malcolm Ross and Darrell Tryon (eds.).
2001 The Boy from Bundaberg: Studies in Melanesian Linguistics in Hon-
our of Tom Dutton. Canberra: Pacific Linguistics.
Pawley, Andrew and Timoci Sayaba
1990 Possessive-marking in Wayan, a Western Fijian language: noun class
or relational system? In Jeremy H. C.S. Davidson (ed.), 147–171.
Rehg, Kenneth L. (with the assistance of Damian G. Sohl)
1981 Ponapean Reference Grammar. Honolulu: University of Hawai‘i
Press.
Rivierre, Jean-Claude
1980 La Langue de Touho: Phonologie et Grammaire du Ce muhi
(Nou-
velle-Calédonie). Paris: Société d’Études Linguistiques et Anthro-
pologiques de France.
Ross, Malcolm
1988 Proto Oceanic and the Austronesian Languages of Western Melane-
sia. Canberra: Pacific Linguistics.
Ross, Malcolm
2002a Bali-Vitu. In John Lynch, Malcolm Ross and Terry Crowley (eds.),
362–386.
Ross, Malcolm
2002b ’Ala’ala. In John Lynch, Malcolm Ross and Terry Crowley (eds.),
347–361.
Ross, Malcolm
2002c Mussau. In John Lynch, Malcolm Ross and Terry Crowley (eds.),
148–166.
Ross, Malcolm, Jeff Siegel, Robert Blust, Michael A. Colburn and W. Seiler (eds.).
1984 Papers in New Guinea Linguistics no. 23. Canberra: Pacific Linguis-
tics.
Salisbury, Mary C.
2002 A grammar of Pukapukan. Unpublished PhD thesis, University of
Auckland.
Schütz, Albert J.
1985 The Fijian Language. Honolulu: University of Hawai‘i Press.
Senft, Gunter
1986 Kilivila: The Language of the Trobriand Islanders. Berlin / New York:
Mouton de Gruyter.
Siegel, Jeff
1984 Introduction to the Labu language. In Malcolm Ross, Jeff Siegel,
Robert Blust, Michael A. Colburn and W. Seiler (eds.), 83–157.
Sohn, Ho-min and B. W. Bender
1973 A Ulithian Grammar. Canberra: Pacific Linguistics.
Taylor, John R. and Robert E. MacLaury (eds.).
1995 Language and the Cognitive Construal of the World. Berlin /New
van den Berg, René and Peter Bachet
2006 Vitu Grammar Sketch. Ukarumpa, Papua New Guinea: Summer In-
stitute of Linguistics.
Volker, Craig A.
1998 The Nalik Language of New Ireland, Papua New Guinea. New York:
Peter Lang.
Wilson, William H.
1976 The O and A possessive markers in Hawaiian. Unpublished MA the-
sis, University of Hawai‘i.
Wilson, William H.
1982 Proto-Polynesian Possessive Marking. Canberra: Pacific Linguistics.
Wivell, Richard
1981 Kairiru grammar. Unpublished MA thesis, University of Auckland.
Possessive clauses in East Nusantara, the case of
Tidore
Miriam van Staden
1. Introduction
Discussion of the origin of attributive possessive constructions have so far

typically focussed on a number of different possible underlying schemas
(Heine 1997). One possible schema that has been understudied is one where
the attributive construction is directly derived from the predicative posses-
sive construction, involving a lexical predicate (verb) whose meaning may
be glossed as ‘have’ or ‘belong’. Heine (1997: 29) notes in passing that Tok
Pisin, the national language of Papua New Guinea, is an example of this
type (cf. Verhaar 1995), but apparently the schema is deemed to be cross-
linguistically uncommon enough to warrant exclusion from the discussion.
Possessive constructions that involve markers that are related to either sub-
ject or object markers in the clause are likewise ignored, although this is
found more commonly cross-linguistically (e.g. in Yaqui Jelinek and Esca-
lante 1988: 412ff., and in Hungarian de Groot 1983, see also Chappell and
McGregor 1996; Allen 1964).
In East Nusantara, a linguistic area in eastern Indonesia and Timor
(Klamer et al. 2008), we find both possibilities. In several languages, in-
cluding the local Malay varieties, the possessive construction involves a
ligature that is derived from a verb meaning ‘have’. And in various cases
these constructions have retained some clausal characteristics, e.g. in Buru
(Grimes 1991). In other languages, and these include Papuan languages
such as Abui (Kratochvíl 2007), Maybrat (Dol 1999) and Inanwatan (de
Vries 2004), the possessive construction is made up of a possessum head
with a subject or object marker cross-referencing the possessor. The Papuan
language Tidore is an apparent mix between these two types. Like Aus-
tronesian Buru and the local Malay variety North Moluccan Malay it has a
construction that displays clausal properties. But it is rather different in that
it uses a possessor prefix that does not correspond to the subject prefix, and
at the same time there is no traceable verbal element, like in various other
Papuan languages. The result is a basic split in the system of argument
cross-referencing with two types of first argument markers on the predicate:
294 Miriam van Staden
“actor”, corresponding to the usual S and A roles in an accusative language,

as in (1), and “possessor”, as in (2): 1
(1) Ngori to-fayaa.

1SG.N 1SG.F.A-woman
‘I am a woman.’
(2) Ngori ri-fayaa.

1SG.N 1SG.N.POS-woman
‘I have a wife; my wife.’
The former occur typically on verbs, and the latter typically on nouns. But
neither do so exclusively. It is not the case, therefore, that the distribution
of the prefixes is given by the word class of the host. Of course, the likeli-
hood that the semantics of a verb call for an agent or undergoer participant
rather than a possessor participant may be greater and at the same time pos-
sessor participants will be linked most typically with possessed goods ex-
pressed in nominal predicates (cf. Taylor 1989, 1996 on the prototypical
properties of possession as an experiential gestalt, which typically involves
a possessum that is a ‘specific concrete thing (usually inaninmate)’.) Yet
these are facts of life rather than language, and indeed both types occur, as
will be demonstrated. Clearly the gloss ‘A’ for ‘actor’ may be somewhat
misleading as it would appear to exclude undergoer arguments in intransi-
tive predications, but these, too, align with transitive A:
(3) piga yo-peka tora si suka

plate 3NH.A-fall downwards first break
‘The plate falls and breaks.’
Here it is a label only to distinguish it from the ‘possessor’ participant.

Tidore is furthermore interesting because it appears to lack a distinction
between attributive and predicative possession. In attributive possession the
possessor functions as an attribute in the noun phrase, whereas in predica-
1
The following abbreviations are used: A – actor, first argument of actor predica-
tion; CAUS – causative; CTRF – centrifugal; – CTRP – centripetal; DIST – distal;
EMPH – emphatic; F – feminine; GEN – generic; H – human; INAL – inalienable;
LOC –locative; M – masculine; N – neutral; NH – non-human; NM – noun marker;
NOM – nominalisation; O – object; PL – plural; POS – possessor, first argument of
possessive predication; PRED – predicate; PROX – proximal; S – subject; SG – sin-
gular; U – unmarked (feminine+non-human); and <> – unsegmented complex.
Possessive clauses in East Nusantara, the case of Tidore 295
tive possession the possessum and the possessor are in a predicational con-
figuration in which either may be the predicate or the argument (see also
the introduction to this volume). It has been claimed that all languages dis-
tinguish these two configurational relations (Heine 1997: 26). Tidore appears
to be a counter-example.
2. Word classes and structural properties of Tidore
Before embarking on the study of the possessive construction, it is impor-

tant to first detail some aspects of Tidore grammar. One of the most notable
properties of Tidore is that although word classes may be established, these
classes play only a minor role in the syntax. A morphological nominalisa-
tion process distinguishes verbs on the one hand from nouns and adjectives
on the other. And adjectives are a separate category because they can be
noun attributes, unlike verbs. This is then also the only syntactic process
where category appears relevant. Otherwise, nouns and verbs can be used
to refer when they are arguments and to predicate. Take for instance a cate-
gorial noun like kolano ‘king’. This noun may be inserted in an argument
position, it may have a noun attribute, and may be quantified, as in (4). But
at the same time, it can also function as the predicate in a clause, in which
case it will have the first argument ‘actor’ cross-reference prefix, as in (5).
Here kolano predicates a property (a function) of an entity; it does not serve
to identify a referent. Referential nominal predications are also possible,
but then the predicate is a noun phrase, as in (6), rather than a bare noun.
Now the noun cannot host a cross-reference marker:
(4) Kolano simo=ge i-gonyihi toma Kota ma-Bopo.

king old=GEN 3SG.M.POS-place LOC Kota ma-Bopo
‘The old king’s place is at Kota ma-Bopo.’
(5) Una wo-kolano toma Arab.

3SG.M 3SG.M.A-king LOC Arabia
‘He is king in Arabia.’
(6) Una=ge kolano rimoi toma Arab isa.

3SG.M king one LOC Arabia landwards
‘He is a king from Arabia.’
One of the most curious examples showing that virtually anything may be
the predicate requires a little explanation. In example (7), the interjection
joo is used as the main predicate in a sentence. There are two forms for
‘yes’, formal joo and informal oe. During conversations, listeners are ex-
pected to signal to the speaker that they are still listening by saying joo! or
oe! at regular intervals. In the example below, the speaker comments on my
own acquisition of this discursive practice. The modifier saki ‘tasty’ is also
used to express satisfaction with non-food items and bahaya ‘dangerous’ is
a loan from Malay that is used to express ‘extreme’ qualities:
(7) Mina mo-joo=ge saki bahaya!

3SG.F 3SG.F.A-polite.yes=GEN tasty dangerous
‘She says polite ‘yes’ in a very good way. (lit. ‘she yesses dangerously
tasty!’)’
This flexibility is a property of the North Halmahera language family. In

fact, among its relatives, Tidore is one of the least flexible. In various de-
scriptions we find statements like the following:
The question, what constitutes a verb in Tabaru, can only be determined
through use, from which it must appear whether they can occur with the
personal pronouns [i.e. the prefixes, mvs]. (Fortgens 1928: 357)
Among the words that occur as nouns and verbs are the following from
Tabaru: toimi ‘bow, to shoot with bow’; gumaa ‘fish hook, to angle’;
ngowaka ‘child, give birth’; singina ‘heart, think’; mede ‘moon, walk in the
moonlight (for hunting)’. And on Tobelo Hueting writes:
Almost all nouns, adjectives, stative verbs, even adverbs can occur as verbs
or in a verbal form. This gives the language as a whole a verby character.
(Hueting 1936: 361)
In Galela adjectives take person marking also when used attributively. At-
tributive adjectives show reduplication of the first syllable, but otherwise
are inflected like predicatively used adjectives, as in the b. example (van
der Veen 1915: 82; van Baarda 1908: 35–36): 2
Galela
(8) a. A-wi-dohu i-lamo
3N.S-3SG.M.O-foot 3N.S+3SG.M.O-big
‘His foot is big.’
2
Van Baarda calls these forms “participles”.
b. A-wi-dohu i-la-lamo
3N.S-3SG.M.O-foot 3N.S+3SG.M.O-RED-big
‘His big foot.’
The Northeast Halmahera languages not only have inflected nouns and ad-
jectives, but even number expressions with human referents require a
predicative construction with an inflected numeral:
(9) Yanau (i-)wi-moi.

man (3PL.SU-)3SG.M.OBJ-one
‘One man.’
Word classes, including (sub)classes of verbs and adjectives are nevertheless

distinguishable, cf. Hueting (1936: 366 ff.) for Tobelo, chiefly by looking at
the category changing morphology. Only adjectives and verbs may be
nominalised by means of the N-prefix.
Conversely, verbs and adjectives are not commonly used as referring
expressions. But we do find examples like (10) and (11) in which a verb is
embedded in a possessive construction, or in which a clause (12) functions
as an argument in another clause:
(10) Ona gahi ena ma-kia.

3PL make 3NH 3NH.POS-marry
‘They prepared the wedding.’
(11) Una i-karja duga gahi gii jira ifa jira

3SG.M 3SG.M.POS-work only make ‘person’ bad don’t bad
wako koliho.
return go.back
‘His only occupation was (saying) “don’t do anything evil to anyone,
or the evil will come back (to you)”.’
(12) Difutu bolo modiri si ngona no-manyasal haro.

tomorrow or day.after.tomorrow first 2SG 2SG.A-regret arrive
‘Tomorrow or the day after you will regret this. (lit. “you regret” ar-
rives)’
Because categories are flexible in Tidore, it is important to keep functional

labels distinct from categorial ones. The label predicate will therefore be
used to refer to the functional notion of expressing a property or relation,
i.e. the typical function of a verb category, and the term ‘referring expres-
sion’ is used to refer to the functional property of referring to some entity,

i.e. the typical function of a noun (phrase). Similarly, ‘attribute’ is the func-
tional label for the adjective in its typical function modifying the noun in
the noun phrase (Hengeveld 1992).
3. The expression of possession in Tidore
3.1. The possessive construction
There exists a vast literature on what constitutes ‘possession’, and as a con-

sequence, what should be typed a ‘possessive construction’. Possessing fin-
gers is different from possessing brothers and different again from possess-
ing boats or houses or books. The linguistic categories in a language reveal
how different kinds of possessive relationships are grouped and distin-
guished from other kinds of possessive relationships, or ‘relationships of
association’ as they may also be called. In Tidore there is a single construc-
tion that expresses relations of ownership, kinship, association and certain
topological relations such as body parts, parts of objects, and relational
nouns. This construction will be referred to as the ‘possessive construction’.
The possessive construction consists minimally of a ‘possessum’ entity
that is typically, but not exclusively, a noun, with a possessive prefix that
cross-references the possessor. The possessive prefix is obligatory. It is very
different in this respect from the actor prefix that is optional in most con-
texts. The possessor may be further expressed in a noun phrase preceding
the possessum:
(13) Una i-fola yo-ruba rai.

3SG.M 3SG.M.POS-house 3NH-fall.apart already
‘His house is falling apart.’
(14) Cole ma-giba yo-foluji.

bra 3NH.POS-strap 3N.A-come.loose
‘The bra strap has come loose.’
The possessor noun phrase may be internally complex. It may, for instance,
consist of a possessive construction:
(15) Ica mi-ngofa na-guru.

Ica SG.F.POS-child 3PL.POS-teacher
‘The teacher of Ica’s children.’
As common in the world’s languages, in Tidore, too, the possessive con-

struction is used to express benefactive/malefactive relations (Lichtenberk
1985, 2002; Margetts 2004). The possessive construction then follows the
transitive verb and the possessor is the beneficiary of the action described
by the main predicate and the possessum.
(16) Mina mo-foli Ica mi-saloi rimoi.

3SG.F 3SG.F.A-buy Ica 3SG.F.POS-basket one
‘She bought Ica a basket.’
Note that for the expression of a beneficiary it is also possible to have a

double object construction (see van Staden 2000: 222–223 for details):
(17) Mina mo-foli Ica saloi rimoi.

3SG.F 3SG.F.A-buy Ica basket one
‘She bought Ica a basket.’
Like most East Nusantara languages, but unlike some of the related North
Halmahera languages, e.g. Tobelo (Holton 2003), Tidore distinguishes alien-
able from inalienable possession. The latter unsurprisingly includes the ex-
pression of body parts, kinship terms, locative nouns and expressions of
part-whole relations (Chappell and McGregor 1996), but also items that are
‘baptized’, i.e. ceremonially attached to a possessor, such as ‘boats’,
‘houses’, and ‘names’ (van Staden 2006). The group of alienables include
notions of non-permanent or incidental ownership as in ‘my bag’, and asso-
ciation, such as ‘his photograph’ (the photograph in which he appears). The
difference between the two is that in the case of inalienable possession the
possessive marker may in all cases be reduced to the 3NH marker ma-
whereas in alienable possession, the possessive prefix must cross-reference
person and number of the possessor. The formal difference between alien-
able and inalienable possession is thus slight and I consider them to be
variations on the possessive construction, rather than different construction
types.
The possessive construction as a whole is typically an argument in a
clause. It may then be preceded by a preposition:
(18) Ngoto lahi [se mansia miskin romoi ena=ge

1SG.N.A ask.for ADD people poor one 3NH=there
na-ngofa rimoi=ge] la ngoto kia.
3PL.POS-child one=there in.order.to 1SG.N.A marry
‘I will ask a poor person’s child to marry me (lit. ‘so that I marry’).’
As this example shows it can be quantified also. Whether the quantifier has
scope over the possessum only or over the entire construction is ambigu-
ous. Tidore has no opposition between ‘their three children’ and ‘three of
their children’. Adjectival modification of the possessive construction has
so far not been attested in discourse; although in elicitation it is deemed
possible:
(19) ?Ona na-ngofa jang rimoi.
3PL 3PL.POS-child beautiful one
‘Their beautiful child; a beautiful child of theirs.’
We will return to modification and quantification in section 5.1, in which
implications for the analysis of possessive constructions are discussed.
4. Other means of expressing relationships of possession
Although the focus of this chapter is on the possessive construction, this is

not the only way to express relationships of possession or association.
There are four additional constructions that may express relations of pos-
session. All four are predicative possessive constructions. Three are con-
ceptually grounded in other domains: the ‘Topic’ construction may be ren-
dered as ‘(as for) X, Y’, the ‘existential’ locative construction expresses “X
exists (at) Y” and the company construction gives “X is with Y” (Heine
1997: 47). The fourth construction is rather interesting, involving the
dummy possessum element due and may be seen as a variation on the Topic
construction combined with the possessive construction.
The Topic construction is a juxtaposition of a possessor noun phrase fol-
lowed by a possessum noun phrase, as in (20). The existential construction
is rather similar but has the existential verb sema ‘to be, exist’ (21). The
companion construction is again superficially highly similar but now the
possessum noun phrase is preceded by preposition soma ‘additional partici-
pant’ as shown by (22):
(20) Fola=ge/ ngora biasa ua.
house=there door normal NEG
‘That house does not have a normal door.’ (lit. ‘The house, no normal
doors’)
(21) Mina sema ngofa rai.
3SG.F be child already
‘She already has children.’ (lit. ‘She, children already exist’)
(22) Mina soma raa yang.

3SG.F ADD husband not.yet
‘She does not have a husband yet.’ (lit. ‘She not yet with husband’)
An important difference between the Topic construction and the other two
is that the prosodic contour of the Topic construction shows a rise in pitch
on the Topic followed by a sharp drop, indicated by the slash in the lan-
guage line, whereas the other two have a regular declarative clause contour.
Although they are similar in form and appear to have a similar meaning,
sema ‘be’ and soma ‘additional participant’ must be distinguished. In most
contexts, they are in complementary distribution:
(23) Mina sema/ *soma bolo ua?

3SG.F be ADD or NEG
‘Is she (here/there)?’
(24) Oyo hula soma / *sema igo ma-ake.

eat sago ADD be coconut 3NH.POS-water
‘Eat sago with coconut juice.’
Preposition soma is part of a paradigm of four adpositions: soma and se pre-

cede company or instrument phrases. The former is used before non-human
denoting nouns, the latter before human denoting nouns. Its etymology is
unclear. It is possible that the collocation of te and se before non-human
denoting nouns preceded by ma- led to this reanalysis.3 But it is equally
possible that soma is derived from coma ‘to add’, e.g. for adding spices to a
dish. Existential predicate sema similarly lacks obvious cognates in the
3
The other North Halmahera languages have a morpheme re (Sahu, Visser and
Voorhoeve 1987) or de (e.g. in Galela, van Baarda 1895) which, like se in Tidore,
functions both as a conjunction and as a preposition on (some) oblique argu-
ments. It is also found in expressions of possession, for example in Galela where
it may only be used with inanimate possessors ‘to indicate that it has something’
(van Baarda 1895: 246):
(i) O kurisi de ma gogocoho-ka
NM chair with 3SG.POS armrest-DIR
‘The chair with armrests.’ or ‘The chair has armrests.’
In this use it resembles the Tidore preposition soma. However, there is no regular
sound correspondence between Sahu r, Galela d and Tidore s. In fact, according
to Wada (1980: 503), Tidore should have had y instead.
other North Halmahera languages. Instead, some of the related North Hal-
mahera languages have an existential predicate ka, which in Tidore is found
only as a bound morpheme in the locative predicates (e.g. ka-tai ‘be located
seawards’, ka-tau ‘be located upwards’, etc.) and in the negative existential
predicates ka-ua ‘be not’, ka-yang ‘be not yet’ and ka-rewa ‘be no more’.
The fourth construction type involves the bound morpheme due ‘posses-
sion(s)’ that takes the position of the possessum. It is considered a ‘dummy’
possessum element since it does not occur outside the possessive construc-
tion. In function it resembles the predicative possessor constructions, as
found in English when the possessor is the predicate and the possessum its
argument (‘the red one is mine’) or vice versa (‘mine is the red one’). This
construction type will be referred to as ‘generic possession’, after the ‘ge-
neric’ element due. But Tidore has no set of independent possessive pro-
nouns that could function as the predicate. Instead, the equivalent in Tidore
involves non-verbal predication with two NPs, one of which is a possessive
construction built on a ‘dummy’ possessum due. This possessive construc-
tion may be the argument of the construction, as in (25), where the predicate
gives the identity of the possessed entity, or it may be the predicate, as in
(26):
(25) Mina mi-due ena ma-bulo.

3SG.F 3SG.F.POS-possess 3NH NOM-white
‘Hers is the white one.’
(26) Ena=re laba ma-due se ena=re saihuu

3NH=PROX monkey 3NH.POS-possess and 3NH.here captain
ma-due.
3NH.POS-possess
‘This is the monkey’s and this is the captain’s.’
The two sentences answer different questions. In the first example, (27), the
utterance is a response to the question which one she owns, while in the
second, (28), it is an answer to the question who owns (or should own, as in
this case) the entity:
(27) Mina mi-due mbe?

3SG.F 3SF.POS-possess which.one
‘Which one is hers?’
(28) Ngge nage na-due?

this.one who 3PL.POS-possess
‘Whose is this one?’
These alternative ways of expressing possessive relations are illustrated

here with a purpose. Not only do they give a more complete picture of the
ways of expressing possession in Tidore, they also show that all alternative
strategies to the possessive construction described in §3.1 involve predica-
tions. In the following section I will show that the possessive construction
itself is also a case of predicative possession, and that in fact, Tidore has no
true attributive possessive construction. In the remainder of the paper I will
then try to find explanations for this highly unusual situation.
5. The possessive construction as a predication
Heine’s study of possession led him to conclude that in all known lan-
guages clear differences show up between ‘attributive possession’ and
‘predicative possession’ (1997: 26). In the former, the possessor and pos-
sessum are generally contained in a noun phrase and the possessor func-
tions as an ‘attribute’ of the possessum; in the latter, the relation between
the possessum and the possessor is expressed in a predication. In the former
possession would be presupposed, in the latter possession is asserted. In
terms of Hengeveld and Mackenzie’s (2008) framework, following Dik
(1989, 1997), it could be said that in the former case the possessor further
restricts the potential set of referents designated by the possessum (from
‘all possible entities X’ to ‘the entities X belonging to Y’) while in the latter
a relation of possession is predicated of the possessor and possessum argu-
ments. The expression of both types may rely on the same conceptual struc-
ture and even on the same linguistic form, but ‘clear differences’, according
to Heine, may always be found. Tidore is an exception. The evidence pre-
sented here shows that one construction is used whether possession is pre-
supposed or not. The possessive construction in Tidore is a type of predica-
tive possession. But this predication is regularly embedded as an argument
of another predicate in which case the entire construction may used referen-
tially. This is the typical case where possession is presupposed, i.e. where
an attributive possessive construction would be expected. In brief, the
structure is: [Y pos-X] for ‘Y has X’, which is embedded as an argument
into [[Y has-X] Verb] ‘Y’s X Verbs’.
5.1. The arguments
The arguments in favour of the analysis of possessive constructions as

predications are the following:
1) Possessive constructions can constitute independent utterances;

2) Embedded possessive constructions may have either a referential or a
predicative reading;
3) The heads of possessive constructions can be preceded by focus marker
kama;
4) Possessive constructions and other clauses, but not noun phrases, can
follow conjunction untuk;
5) Adjectives and verbs otherwise do not occur as heads of referring ex-
pressions; the only time they occur in a typically nominal position is as
heads of possessive constructions.
The first argument is more appropriately a condition for a clausal analysis

rather than evidence, since non-clausal utterances also exist. The second,
third and fourth arguments constitute evidence, however, while the last ar-
gument may be considered supporting evidence. Each will be discussed in
turn.
5.1.1. Independent utterances
If possessive constructions are predications then they should be able to con-

stitute independent clauses. And indeed they are. Possessive constructions
need not be embedded in a larger structure but they may predicate posses-
sion on their own. As such they may have their own temporal modal and
aspectual modifiers, and they may be questioned like other clauses, as in
(33):4
(29) Ua si, ngona yuke dadi ngona ni-due.

NEG first 2SG earlier so 2SG 2SG.POS-possess
‘No, you were first so it is yours.’
4
Yes /no questions in Tidore are formed by adding ‘or no’ to the questioned
clause. In conjunction with aspectual modifiers rai ‘already’ and moju ‘still’,
their negative counterparts are used, so rai bolo yang ‘yet or not yet’ (as in ex-
ample (33)), moju bolo rewa ‘X still or no more’.
(30) Fayaa rimoi ngge toma fola ma-doya ngge, mina

woman one 3NH.GEN LOC house INAL-inside 3NH.GEN 3SG.F
mi-cahaya.
3SG.F.POS-shine
‘There is a woman in the house, she is a shining beauty (lit. has bright-
ness).’
(31) Ino Jailolo ena-re i-sultan rewa rai, […]

CTRP Jailolo 3NH-PROX 3SG.M.POS-sultan not.anymore already
‘Then later (lit. towards the present) Jailolo no longer had a sultan, […].’
(32) una wo-rasa paha rea karna una

3SG.M 3SG.M.A-feel bear not.anymore because 3SG.M
i-daera ua
3SG.M.POS-area NEG
‘He felt he couldn’t take it any more because he did not have a piece
of the land.’
(33) Mina mi-ngofa rai bolo yang?

3SG.F 3SG.F.POS-child already or not.yet
‘Has she got children? (lit. she has child already or not yet?)’
In discussions with native speakers, these sentences are never felt to be in-
complete or to require a context as is often the case with non-clausal utter-
ances.
It is notable that when used as an independent utterance the possessive
construction typically has a modal or aspectual modifier. In the context of
Tidore this may be an important fact since historically the modal and aspec-
tual modifiers are intransitive verbs, taking as their only argument the event
description that is in their scope. In some of the related North Halmahera
languages they still carry person marking like other verbs, and even in Ti-
dore some of the modal and aspectual modifiers regularly function as main
predicates, e.g. rewa ‘not anymore’ (cf. example (31) above):
(34) Gula=re yo-rewa rai.

sugar=PROX 3NH.A-not.anymore already
‘The sugar is all gone.’
In present-day Tidore, the placement of modal and aspectual modifiers is

always immediately after the unit they scope over, e.g. the following exam-
ples in which the predicate is underlined and the scope unit in brackets:
(35) Turus [gahi ena ma-mammam] dadi.

then make 3NH 3NH.POS-sweets able.to
lit.: ‘Then (make its pudding) is possible/happened.’
(36) [[Mina mo-wako] maya] rewa.

3SG.F 3SG.F.A-return may not.anymore
lit.: ‘((She return) may) not anymore.’
Possessive constructions are no different. What is interesting is that in many

cases there are two readings when the possessive construction has a modal
or aspectual modifier:
(37) una i-ngofa nau-nau duga rimoi

3SG.M 3SG.M.POS-child RED-male only one
1. ‘He has one son only.’ 2. ‘His male children number just one.’
In the first reading it is the possessive relation that is predicated, but in the
second possession of sons is presupposed and it is just number that is predi-
cated. The former reading, without the presupposition, is expected when the
predicate is the possessum; the latter reading is expected if the possessive
construction is a referential phrase and the predicate is duga rimoi ‘only
one’.5 A similar situation arises with numerals:
(38) Mina mi-ngofa rimoi reke.

3SG.F 3SG.F.POS-child one cry
1. ‘One of her children is crying.’ 2. ‘Her one child is crying.’
Tidore has no distinction between ‘one of her children’ and ‘her one child’.
In the first reading the numeral has scope over the entire possessive con-
struction (cf. the paraphrase ‘as for her children, one is crying’), and in the
second it scopes over just the possessum (cf. she has one child; it cries).
5
A reading in with naunau as the predicate is excluded here on prosodic grounds
because there is no drop in pitch after i-ngofa and on syntactic grounds because
in that case it could not have be followed by duga rimoi ‘only one’. But it is in
theory possible to say:
(ii) Una i-ngofa [nau-nau.]pred
3SG.M 3SG.M.POS-child RED-male
‘His child is a boy.’
The former is therefore compatible with a non-presupposed or predicative

reading and the latter with a presupposed or attributive one.
5.1.2. Complement taking predicates
When possessive constructions are embedded as complements of verbs that

take both clausal and noun phrase complements, different readings are often
available:
(39) Una wo-hoda mina mi-ngofa.

3SG.M 3SG.M.A-see 3SG.F 3SG.F.POS-child
1. ‘He sees her children.’ 2. ‘He sees (that) she has children.’
It is doubtful that this points at a ‘clear distinction’ between an attributive

and predicative construction in Tidore. Instead of structural ambiguity, the
different readings are the result of differences in the informational status of
possessive relation as either presupposed or asserted. What it does show
clearly is that predicative readings are available.
5.1.3. Negative focus with kama
Perhaps the most striking property of the possessive construction is that the
possessum behaves like a verbal predicate with respect to the placement of
the negative focus particle kama. Tidore negative particles ua ‘not, no’,
rewa ‘not anymore’, ifa ‘don’t’, yang ‘not yet’ all occur in clause final po-
sition. When focus is on the main predicate, a particle kama may directly
precede it. This applies equally to verbal and nominal predicates:
(40) Ngom kama hoda mina rewa.

1PL.EX NEG.FOC see 3SG.F not.anymore
‘We don’t see her anymore.’
(41) Una kama kolano ua.

3SG.M NEG king NEG
‘He was not a king.’
In this last example, the noun kolano ‘king’ predicates a property (function,
profession) of some person (recall example 5). Because the negative parti-
cles always occur clause finally, negation in complex clauses is ambiguous.

An important use of kama is to disambiguate:
(42) Ngom kama baso mina mo-nyanyi yang.

3SG.M NEG.FOC hear 3SG.F 3SG.F.A-sing not.yet
‘We have not heard her singing yet.’
(43) Ngom baso mina kama mo-nyanyi yang.

3SG.M hear 3SG.F NEG.FOC 3SG.F.A-sing not.yet
‘We hear that she hasn’t sung yet.’
The focus particle cannot precede noun phrase arguments, as (44) shows,
but in clauses containing a possessive construction kama may occur before
the possessum, as in (45) to (47):
(44) *Ngom baso kama mina yang.

3SG.M hear NEG.FOC 3SG.F not.yet
intended reading: ‘We have not heard her yet.’
(45) Ngoto hoda mina kama mi-ngofa ua.

1SG.N.A see 3SG.F NEG 3SG.F.POS-child NEG
‘I saw she has no children.’ (lit. ‘I saw she was without children’)
(46) Mina kama mi-oli fugo ua.

3SG.F NEG 3SG.F.POS-voice come.out NEG
‘Her voice did not come out.’ (lit. ‘she does not have a voice that
comes out) i.e. she did not speak’)
(47) Una kama i-yohu gola ua.

3SG.M NEG 3SG.M.POS-leg hurt NEG
‘His leg does not hurt.’ (lit. ‘he does not have a leg that hurts’)
There is one restriction on this type of negation. If the semantics of the

matrix predicate presupposes the existence of an entity as, for instance the
undergoer, and if this entity is is the possessum, internal negation is im-
possible. In other words, it is nonsensical to express an action affecting an
undergoer, when this undergoer is then claimed to be non-existent:
(48) *Ngoto cako mina kama mi-ngofa ua.

1SG.N.A hit 3SG.F NEG 3SG.F.POS-child NEG
??? ‘I am hitting the child she does not have.’
(49) *Ngoto jau mina kama mi-gia ua.

1SG.N.A hold 3SG.F NEG 3SG.F.POS-hand NEG
???‘I hold her non-existing hand.’
This is, however, a semantic and not a structural restriction. Informants

stated that these utterances were nonsense as it would be impossible to ‘hit
non-existing children’, or ‘hold non-existing hands’. These facts all point to
an analysis of the possessed noun as a clausal predicate with a possessor
argument.
5.1.4. Complements of untuk
Similar evidence comes from the type of complement that loan conjunction
untuk takes. As we saw earlier, beneficiaries may be expressed through the
possessive construction. An alternative to this construction is the use of a
purpose clause introduced by the Malay loan word untuk ‘for’. In Standard
Indonesian this is a preposition directly marking the beneficiary, as in (50).
In Tidore, the complement of untuk cannot be a noun phrase, hence the un-
grammaticality of (51), but must be clausal, as in (52):
Indonesian
(50) Apa surat ini untuk saya?
what letter this for 1SG
‘Is this letter for me?’
(51) *Mina foli saloi rimoi untuk Ica.

3SG.F buy basket one in.order.to Ica
(52) Mina isa rasi untuk gahi

3SG.F landwards first in.order.to make
mina mi-raa=ge ngam.
3SG.F 3SG.F.POS-husband=GEN food
‘She goes home (lit. ‘landwards’) now to make her husband a meal.’
Yet, possessive constructions are possible complements, as shown in (53).

In fact, generic -due must be used if the possessed item is not further speci-
fied, compare (54) to (51) above. Interestingly, even in the North Moluccan
Malay spoken on Tidore a clausal complement is preferred to the nominal
phrase, as in (55) (see section 6 for more cases of analogy in structure be-
tween Tidore and North Moluccan Malay): 6
(53) Ua si ngona ngofa no-nau, ge ngan

NEG first 2SG child C.RED-male there 2SG
sa-dia untuk ngori ri-diali.
CAUS-NOM.leave.behind for 1SG.N 1SG.POS-NOM.replace
‘Otherwise, if you have a boy, then you leave it for my replacement
(i.e. to replace me).’
(54) Mina foli saloi rimoi untuk Ica mi-due.

3SG.F buy basket one in.order.to Ica 3SG.F.POS-possess.
‘She bought a basket for Ica (to have).’
(55) Ini untuk kita punya!

this for 1SG have
‘This one is for me (to have)!’
This demonstrates that the possessive construction groups with clauses and
not with noun phrases.
5.1.5. Verbs and adjectives as heads
If possessive constructions are treated as clausal predications with a ‘pos-

sessum’ main predicate, it is not surprising that the possessum could also be
a verb or adjective. While verbs and adjectives are extremely uncommon as
heads of referring expressions (i.e. noun phrases with ‘non-nominal heads’
are virtually non-existent) they do occur as heads of possessive construc-
tions.
(56) Mina mi-jang.

3SG.F 3SG.F.POS-beautiful
‘Her beauty.’
6
In formal contexts Standard Indonesian may interfere with North Moluccan Malay.
(57) Una tagi pana namo nde una i-oyo se

3SG.M go shoot bird 3NH.PROX 3SG.M 3SG.M.POS-eat and
yuru ake […] te simo nange.
drink water LOC old just.now
‘He (who) went shooting birds, he had food and drink with this old
man just mentioned.’
The fact that adjectives and verbs occur in the possessive construction in a
syntactic position typically associated with nouns, but not as heads of noun
phrases, may be explained by the predicative rather than referential function
that the head of the possessive construction has. The reason that the vast
majority of possessums are nominal is semantic rather than structural: what
is possessed is typically a physical object expressed by a noun.
There is no need to assume nominalisation in these cases. In actor
clauses, the main predicate is typically verbal, although nominal and adjec-
tival predicates are also attested, while in possessor-clauses the main predi-
cate is typically nominal, although both adjectival and verbal predicates
also occur. Contrast this to the nominalisation process involving the prefix
ma (also for 3NH.POS): 7
(58) Mina ma-jang.

3SG.F NOM-beautiful
‘She is a beauty.’
(59) Mina ma-din kabaya.

3SG.F NOM-sew dress
‘She is a seamstress.’
Here, there is evidence of derivation, since the ma-prefixed stem takes actor
marking:
(60) Mina mo-ma-din kabaya.

3SG.F 3SG.F.A-NOM-sew dress
‘She sews dresses for a living.’
7
Again in North Moluccan Malay we find the literal equivalent:
(iii) Mina ma- jang sado! Tidore
de pe cantik sampe! North Moluccan Malay
3SG(.F) POS beautiful until
‘She is such a beauty!’
The difference in meaning between the actor and possessive constructions

is that with ma- the property is no longer transient but permanent. Thus
(62) is unacceptable.
(61) Ngire mina jang sodooo!

last.night 3SG.F beautiful until
‘Last night she was very beautiful!’
(62) *?Ngire mina ma-jang sodooo!

last.night 3SG.F NOM-beautiful until
5.2. Against a clausal analysis
So far the arguments supported the analysis of the Tidore possessive con-
struction as basically ‘clausal’. However, although historically this may be
the case, and synchronically also the construction shows clausal properties,
there are several ways in which possessive constructions differ from actor-
clauses and these could be taken as counter-arguments against this clausal
analysis.
First of all, unlike typical actor clauses, possessive constructions do not
take aspectual or temporal modifiers when functioning as arguments of
verbs. And as independent utterances they are never modified by modal aux-
iliaries such as dadi ‘be able to’, or maya ‘want to’:
(63) *Mina mi-ngofa dadi ua.

3SG.F 3SG.F.POS-child able.to NEG
intended reading: ‘She is unable to have children.’
Instead, a verbal predicate is called on:
(64) Mina mo-dahe ngofa dadi ua.

3SG.F 3SG.F.A-find child able.to NEG
‘She cannot have children.’
Furthermore, the distribution of actor and possessor prefixes is not perfectly

parallel. Actor prefixes are always optional. There are situations in which
actor prefixes cannot occur on an auxiliary, but wherever they are found in
texts native speakers consider the utterances equally good without the pre-
fix. Possessor prefixes, on the other hand, are obligatory. And although ac-
tor prefixes occur also on nominal predicates, as was illustrated in section 2,

this is only possible when the predicate is a bare noun. Once the predicate
is a full noun phrase it can no longer have an actor prefix, recall examples
(5) and (6), repeated here for convenience:
(5) Una wo-kolano toma Arab.

3SG.M 3SG.M.A-king LOC Arabia
‘He is king in Arabia.’
(6) Una=ge kolano rimoi toma Arab isa.

3SG.M king one LOC Arabia landwards
‘He is a king from Arabia.’
However, in possessive constructions, the possessum is a noun phrase that

has the possessor prefix. This lack of parallelism could be seen as counter-
evidence.
Finally, if the possessive is analysed as a clause, one might expect that
the assertion reading (i.e. the predicative reading) should be primary in
complementation constructions of the type discussed in § 5.1.2 above. But
in fact it is the presupposition reading that is the unmarked one, perhaps
suggesting that the construction is a noun phrase. But this could also be the
result of the typical use of the construction expressing presupposed posses-
sion in conjunction with the likelihood that in cases in which the possessive
construction follows a verbal predicate that may take both clausal and
nominal complements the second interpretation is typically intendend.
5.3. Evaluation
The various arguments that were presented show that Tidore lacks a dis-
tinction between an attributive and predicative possessive construction. The
possessive construction has clausal properties also when possession is pre-
supposed. The result is that in Tidore there is a split in the system of argu-
ment marking on the predicate: transitive predicates cross-reference the
first argument by means of a ‘subject’ or ‘actor’ prefix, as do intransitive
predicates that describe actions, states, experiences, etc., except those
predicates that give a possessive relation, because these take a ‘possessor’
first argument. This gives the regular oppositions illustrated in the introduc-
tion, and repeated here:
(65) Ngori to-fayaa.

1SG.N 1SG.N.A-woman
‘I am a woman.’
(66) Ngori ri-fayaa.

1SG.N 1SG.N.POS-woman
‘I have a wife; my wife.’
Typologically, the Tidore system is rather unusual. Possessive markers that

are identical to subject markers are found in some languages, but to find
what appears to be a split in the system of predicate cross-referencing de-
pending on whether the relation between the argument and the predicate is
one of actor or undergoer cross-referencing on the one hand versus posses-
sor marking on the other appears truly rare. Also, as pointed out, not to find
a clear distinction between attributive and predicative possessive construc-
tions is exceptional. In the next sections, possible historical explanations
for this synchronically ‘odd’ possessive construction are explored. One
possibility that is examined is that the Tidore system may be copied from
North Moluccan Malay through a process called ‘metatypy’ (Ross 2006). I
also investigate whether other languages in the same linguistic area display
constructions that are anything like what we find in Tidore and how Tidore
compares to genealogically related languages of the North Halmahera Family
and the Papuan languages of the Bird’s Head. It is shown that this last com-
parison gives one possible account for the synchronically rather unusual
Tidore construction.
6. North Moluccan Malay
Throughout Eastern Indonesia, different varieties of Malay have served as

the lingua franca for many centuries and despite education in Standard
Indonesian, their importance has increased rather than decreased during the
past sixty years. The specificities differ from place to place so that ‘Ambo-
nese Malay’ is distinguished from ‘Kupang Malay’ (spoken on Timor),
‘North Moluccan Malay’ and ‘Irian Malay’, but the general patterns are the
same in all these varieties that together are referred to as ‘Eastern Indonesian
Malay’. The possessive construction shows the same pattern throughout the
area. Unlike Malay varieties in other parts of Indonesia and Malaysia, and
unlike the standardized national variety ‘Indonesian’, the order in the pos-
sessive construction is possessor-possessum. Furthermore, the construction
contains a possessive ligature that is derived from the verb punya ‘have’:
pu in Ambonese Malay (Tjia 1997; van Minde 1997), pe in Irian and North
Moluccan Malay. Where the varieties of Eastern Indonesian Malay differ is
whether punya may be used in full in attributive possessive constructions,
and whether this construction is also used to express benefaction. The
claims in this section apply only the variety spoken on Tidore, although
following native speaker practice it is referred to as North Moluccan Malay
which would cover all of the North Moluccas.
Where languages are in prolonged contact, the systems may converge to
a greater or lesser extent. The diachronic process underlying such syntactic
convergence of systems has been referred to as ‘metatypy’ (Ross 1996,
1997, 2001, 2006). In metatypy ‘the syntax of one of the languages of a bi-
lingual speech community is restructured on the model of the syntax of the
speaker’s other language’ (Ross 2006). The result is very similar ways of
speaking in languages that are in contact. In the North Moluccas, in particu-
lar in Ternate, Tidore and Taba, this has occurred on a large scale. We find
it for instance in the use of spatially loaded terms in the regional Malay va-
rieties (Bowden 2005). The North Moluccan Malay expressions appear to
be literal copies from the indigenous varieties:
(67) Dong pi lao di=lao.8 North Moluccan Malay

Ona tagi hoo ka-tai. Tidore
3PL go sea(wards) ‘locative’9=sea(ward)
‘They went seawards to a location seawards.’
In the expression of possession the same similarities in patterning are

found:
(68) Dong pe anak so tiga. North Moluccan Malay

Ona na- ngofa range rai. Tidore
3PL 3PL.POS-child already three already
‘They already have three children.’
8
Note that Tidore has distinct morphemes for ‘sea’ (ngolo), moving seawards
(hoo) and seaward location (=tai), where Malay has only lao ‘sea’, showing that
Tidore was the model for North Moluccan Malay rather than vice versa.
9
In North Moluccan Malay this locative corresponds to the locative preposition
in Standard Indonesian; in Tidore the locative is a predicate deriving prefix.
(69) Turus potong de =pe kapala. North Moluccan Malay

Turus tola ena ma- dofolo. Tidore
then cut/chop 3NH 3NH.POS head
‘Then cut off its head.’
Depending on the focus of the construction possessor third person pronoun

dia can be reduced to de to which the ligature pe cliticises in (69), further
obscuring the original verbal element and the original verbal construction.
In attributive possession, the possessive ligature in North Moluccan Malay
is typically the reduced form pe, but at least on Tidore, the full form of the
verb punya is also found in attributive possessive expressions. In example
(70) the North Moluccan Malay possessive construction is a code switch
used to translate the Tidore word dano ‘grandchild’. The remainder of the
sentence is also in North Moluccan Malay:
North Moluccan Malay

(70) Dano-dano raja-raja punya ana cucu itu kalau fam
RED-grandchild RED-king have child grandchild that if family
Alting itu Blanda.
Alting that Dutch
‘(As for) the grandchildren, the grandchildren of the kings, if they are
called Alting then they are Dutch.’
In the beneficiary expressions, too, the choice is between the full verb and
the reduced form:
North Moluccan Malay

(71) Ekal, ambil ci Miriam punya/pe kursi ka=mari.
Ekal fetch sister Miriam have chair DIR =here
‘Ekal, fetch Miriam a chair.’
In predicative possession only the verb may not be reduced, giving a formal
distinction between the two functions. And in generic possession also the
full form of the verb is obligatory, but the similarity in structure to Tidore is
again striking:
(72) Yang kita punya puti. North Moluccan Malay

Ngori ri- due bulo. Tidore
REL 1SG 1SG.POS-have white
‘Mine is/are white.’
The question that this parallel raises is whether perhaps the Tidore con-
struction was reanalysed as clausal on the basis of its similarity to the Malay
possessive construction with its obvious clausal history. But the opposite is
also possible. Perhaps the North Moluccan Malay construction evolved un-
der the influence of the clausal constructions in the indigenous languages of
the area. A third possibility is of course that the two happen to be similar
synchronically, but have independent histories. There are two arguments
against the first scenario. First, clausal possessive constructions are a char-
acteristic of all North Halmahera languages as will be demonstrated in the
next sections. Furthermore, such constructions are found throughout East
Nusantara. Second, the North Moluccan Malay possessive construction has
the possessor-possessum order which is unusual for an Austronesian lan-
guage. Malay dialects spoken in western Indonesia have the usual posses-
sum-possessor order. But this order is common for both Papuan and Austro-
nesian languages in East Nusantara. It appears that the indigenous Papuan
languages influenced the Austronesian languages in this area. If the clausal
analysis in Tidore is a case of metatypy, we would need to assume that
North Moluccan Malay first took the order possessor-possessum from lan-
guages like Tidore, and then introduced a clausal construction to express this
relation of possession. And this construction was then subsequently bor-
rowed back into Tidore. Complex scenarios like these should not always be
ruled out, but it would seem more likely that the influence has been the other
way round. The North Moluccan Malay construction is based on the indige-
nous languages of the area in which it is spoken. For the same reason it is
unlikely that the Malay construction developed completely independently
from the indigenous languages of the area. In the following two sections the
expression of possession in East Nusantara as the linguistic area that Tidore
forms part of, and in the North Halmahera Family will be explored in order
to examine to what extent the Tidore construction is a product of its neigh-
bours or ancestors.
7. East Nusantara
East Nusantara is roughly the area east of Bali and Sulawesi up to and in-
cluding the Bird’s Head of Papua and perhaps even further East along the
North coast of New Guinea (van Staden and Reesink 2008; Klamer et al.
2008; Klamer 2003; Baird 2002). This area is characterized historically by
high mobility of people and intense contact between people and languages.
It is a complex patchwork of languages. Here different branches of Aus-
tronesian meet as well as a large number of Papuan languages for which

genetic relations cannot or only with great reservations be established (Wurm
1982; Foley 1986; Reesink 2000, 2002, 1998). A West Papuan Phylum has
been proposed (Voorhoeve 1987, 1987) relating the North Halmahera lan-
guages, including Tidore, to several of the languages of the Bird’s Head,
Papua. But Reesink (2000) has proposed that this phylum is made up of six
language families: 1. West Bird’s Head with North Halmahera, three iso-
lates: 2. Abun, 3. Maybrat, 4. Mpur, and two small families in the Eastern
Bird’s Head: 5. Meyah and Sougb, and 6. Hatam and Mansim. In addition,
the South Bird’s Head languages again form a separate family, as do the
Papuan languages of Alor-Timor-Pantar, such as Abui and Adang. In addi-
tion to these Papuan languages, we also find languages belonging to two
different branches of Austronesian: Central Malayo-Polynesian (e.g. Buru)
(Blust 1993) and South Halmahera West New Guinea (Biak), which to-
gether with the Oceanic languages make up the Eastern Malayo-Polynesian
subgroup (e.g.Tryon 1995: 34).
Figure 1. East Nusantara. Papuan languages are found in Papua and in the encircled
areas; Austronesian languages are spoken elsewhere in the archipellago.
Map adapted from Klamer, et al. (2008)
Possessive constructions in East Nusantara are relatively well-studied be-

cause they exhibit a number of cross-linguistically interesting features. First,
the Austronesian languages in East Nusantara show an unexpected word
order in attributive possessive constructions. In spite of their general SVO
typology, the order in the possessive construction is typically possessor-
possessum. This is most likely the result of language contact with neigh-
bouring Papuan languages that are at least originally SOV and also have
possessor-possessum order (Klamer et al. 2008). Another characteristic of
East Nusantara is that many languages formally distinguish alienable and
inalienable possession. Furthermore it is a striking feature that for the ex-
pression of at least one of these two possessive relations, the marker of pos-
session is often derived from a subject or object prefix. Again, this is found
alike in Austronesian and Papuan languages, and again there is evidence
that the latter have influenced the former in this respect.
Several languages have a verbal element in at least one of the attributive
possessive constructions. For example in Biak, an Austronesian language of
Papua, the construction that expresses alienable possession involves a com-
plex ‘ligature’ that comprises of an element ve ‘POS’ traceable to a verb ‘to
possess’, which is inflected for subject (possessor) and followed by a com-
plex ‘article’ that expresses the possessum (Van den Heuvel 2006: 229ff.):
Biak
(73) … ko-swar min ko-ve=s-i
1PL.INC-love member 1PL.INC-POS=3PL .ANIM-SPC
‘… we love our neighbours.’ (Van den Heuvel 2006: 437)
Buru has a possessive construction which ‘mirrors the structure of an active

transitive clause with a pre-verbal Actor, the possessive word ( verb) and
a post-verbal Undergoer’ (Grimes 1991: 278). The possessive word has an
applicative suffx -k to indicate ‘a definite pronominal object’ (p. 229). The
possessive word even accepts certain valence changing verbal prefixes:
Buru
(74) Kawasan p-em-nake-k geba rua ute tinge […]
head CAUS-STAT-3SG.POSS-k person two DAT 3SG […].
‘The village head put two people at his disposal […]’
As in Tidore, this one construction may be interpreted predicatively and

attributively. Grimes summarises this as follows:
When the possessive construction functions predicatively, the possessive

word is the nucleus of the clause […], behaving like a transitive verb with a
subject and object. Used predicatively, the possessive word is the head of
the construction. When the possessive construction functions as a nominal
argument in a clause, the possessive word functions as a possessive pronoun,
being a pre-head modifier to the head noun. Thus, whether the possessive
word is behaving verbally or nominally, or whether it is functioning as the
head of a predicative possessive construction or as the modifier of a posses-
sive NP depends on its distribution. (Grimes 1991: 282)
More generally the possessor affixes correspond to either the subject or the
object markers, also when no verbal element is found. For instance in May-
brat the possessor is expressed by a subject prefix in inalienable possessive
constructions:
Maybrat
(75) Fnia m-ao.
woman 3U-foot
‘The woman’s foot.’ (Dol 1999: 149)
And other Papuan languages of the Bird’s Head such as Meyah (Gravelle
2004: 171), Moi (Menick 1995: 65), Mpur (Odé 2002: 62) and Hatam
(Reesink 1999: 49) are identical.
The Southern Bird’s Head language Inanwatan has two highly similar sets
of prefixes for inalienable possession (de Vries 2004: 29) and object marking
(de Vries 2004: 36) although the two differ in certain morpho-phonological
properties. As in the North Halmahera languages, the possessive markers in
Inanwatan have optional arguments that the object markers lack.
In Adang and Abui, two Alor-Timor-Pantar languages, we find that per-
son prefixes marking possessor on nouns also cross-reference objects, though
not subjects. In Adang only a particular subset of verbs is inflected:
Adang
(76) N-e na-fel mi habu.
1SG-GEN 1SG-ear COMP wide
‘My ears are wider (e.g. than yours)’ (Haan 2001: 134)
Adang
(77) Sa na-tan.
3SG.NOM 1SG-ask
‘S/he asked me.’ (Haan 2001: 46)
But in Abui there are even two sets of possessor prefixes that correspond to
two sets of second argument markers. The prefix for inalienables in (78a)
corresponds to the ‘patient’ marker in (78b), and the prefix for alienables in
(79a) corresponds to the ‘locative’ category in (79b), which includes bene-
ficiaries and ‘themes’ and which may be thought of as locations (Kratochvíl
in progress).
Abui
(78) a. Na-min
1SG.INAL-nose
‘My nose.’
b. Me na-dak-e.
come 1SG.PAT-clutch-IPFV
‘Come and hold me tight.’
Abui
(79) a. Ne-wil.
1SG.AL-child
‘My child.’
b. Simon ne-l to-ha-loi.
Simon 1SG.LOC-give DISTR.REC-3ii.PAT-chase
‘Simon chases me.’
And further east in the Cenderawasih Bay inalienable nouns in Yawa have a
prefix identical to the undergoer prefixes used on transitive verbs and stative
verbs.
Yawa
(80) In-aneme.
1SG.OBJ-hand
‘My hand.’ (Jones 1986; see also Klamer et al. 2008)
Summarising, in a number of East Nusantara languages, the possessive

construction involves an (erstwhile) verb, but in the majority the possessor
is marked directly on the possessum by means of an affix that is also used
to mark either subject or object in the clause. Where Tidore differs is that
the possessor marker does not correspond to the subject marker (and there is
no object marker) and at the same time the possessive construction displays
clausal properties even though there is no trace of an erstwhile verb, as in
Buru and Biak.
8. The North Halmahera family
In the final part of this chapter I compare the Tidore possessive construc-
tion to the closely related languages of the North Halmahera family. The
first obvious question is whether a clear distinction between attributive and
predicative possession can be made in these languages, whether the posses-
sive constructions in these languages, too, appear to be clausal. But the
other North Halmahera languages are also interesting because they appear
to be rather conservative compared to Tidore in a number of respects, in-
cluding the systems of person cross-referencing on both verbal predicates
and possessum noun phrases. It is argued here that these systems of cross-
referencing have some important clues to a possible account of the present-
day rather unique distinction between actor prefixes and possessor prefixes
in Tidore.
The North Halmahera family is usually presented as consisting of ten
languages.10 However, six of these are so closely related that they may be
considered dialects of one language ‘Northeast Halmaheran’ (Voorhoeve
1988): Galela, Tobelo, Tobaru, Pagu, Loda, Modole, and similarly Ternate
and Tidore are dialects of one language (cf. Figure 2). The Northeast Hal-
mahera languages are the most archaic in terms of retention of sounds from
proto-Halmaheran (Wada 1980), SOV word order and postpositions, use of
articles and noun class markers, the spatial deictic system, and also the oc-
currence of object prefixes, which in Tidore as well as in Ternate and West-
10
The data for this section were taken from various sources, most of them dating
from the early 20th century, but some more recent studies are also available.
Where the reports are in disagreement I give specific references. For details on
Galela see Van Baarda (1908) and Shelden (1998, 1991), on Tobelo see Hueting
(1936) and Holton (2006, 2003), on Pagu Wimbish (1991) and on Tabaru Fortgens
(1928). For comparative studies see in particular Van der Veen (1915); Capell
(1975); Voorhoeve (1988, 1987, 1987); and Wada (1980). I have omitted from
the presentation some details on phonetic realization, e.g. differences between
long and short vowels and stress where they are non-pertinent to the present
discussion. Galela has a distinction between a dental /d/ and a retroflex //
which are represented by the single grapheme d in this paper. I have changed
the glosses of some of the examples to align the terminology or I have added
glosses where the original sources (usually the older ones) do not give them.
Also, most older sources as well as some of the more recent ones (e.g. Wimbish)
give the verbal markers and the possessive markers as separate words, but there
is enough morpho-phonological evidence to warrant their presentation as pre-
fixes in all North Halmahera languages.
Makian (van der Veen 1915) have been lost. The possessive constructions in
all the North Halmahera languages show strong resemblance, although
again the Northeast Halmahera languages appear to be more conservative
than Tidore.
Figure 2. The North Halmahera languages
8.1. Clausal analysis in the North Halmahera languages
The first question is to what extent the possessive constructions in the other
North Halmahera languages have clausal properties like Tidore. Although
this issue has not been addressed for any of the languages in the literature,
text material shows that the possessive construction may be used predica-
tively in at least several different North Halmahera languages, e.g. in Tobelo:
(81) Ma-koano ngo-i-hekata ya-tumidi,

NM-king NM-3SG.M.POS-wife 3N.SU+3PL.OBJ-seven
ya-butanga manga ngohaka o nauru ya-butanga.
3N.SU+3PL.OBJ-six MANGA child NM male 3N.SU+3PL.OBJ-six
‘The king had seven wives, six (of them) had six sons.’
(Hueting 1936: 373)
Because the possessed item is female, the possessive pronoun is ngo-i

rather than a-i cf. also (Hueting 1936: 353). Like other clausal arguments
the possessor noun phrase may be postposed as in the Pagu example (82):
Pagu
(82) ma kolan awi-ngoak mosoles dauk
ART king 3M.POSS- child bachelorette there:south
m-o-matetengo, ami-penjagaan o macan de o garuda …
3SG.F-SU-one 3SG.F.POSS-guard ART tiger and ART eagle
‘The king’s only daughter is there, and she has a tiger and an eagle
for guards.’ (Wimbish 1991: 136)
Furthermore, the possessive constructions are frequently translated as having

two readings: a predicative and an attributive one, as in Tobelo o tau ma-
hailoa ‘a beautiful house (the house has beauty), the beauty of the house’
and o gakana ma-doto ‘a sharp knife, the sharpness of the knife’ (Hueting
1936: 368).
Also in support of a clausal analysis is the occurrence of adjectives in
this construction and the ensuing ambiguity between a phrasal and a clausal
reading as in the following Galela example:
(83) Awi-lamo.
3SG.M.POS-big
‘His bigness; how big he is!’ (van Baarda 1908: 38)
There is, however, no data on possessive constructions in complementation

or on the extent to which possessive constructions pattern with clauses or
noun phrases of the kind described for Tidore (e.g. the complementation of
untuk). This means that the data so far show that as in Tidore there is no
clear distinction between a predicative and attributive possessive construc-
tion, but whether the constructions in the other North Halmahera languages
also pattern with clauses rather than noun phrases is still unclear.
8.2. Personal and possessive pronouns
The Northeast Halmaheran languages and Sahu all have pronoun invento-
ries that are very much like the pronouns in Tidore and Ternate. The per-
sonal pronouns are without exception historically morphologically complex
as shown in Table 1 where the roots of the pronouns are underlined:
Table 1. Personal pronouns in Tidore and Tobelo (Northeast Halmahera)
singular plural
Tidore Tobelo Tidore Tobelo
1 EXC ngo-ri ngo-hi ngo-m ngo-mi
1 INCL – – ngo-ne ngo-ne
2 ngo-na ngo-na ngo-n ngi-ni
3M u-na u-nanga
3F mi-na mi-nanga o-na o-nanga
3 NH e-na e-nanga e-na e-nanga
The suffix na(nga) found with the third person pronouns has its origin in the
demonstratives, cf. re-na(nga) ‘here’ and ge-na(nga) ‘there’. It may be typed
as a ‘deictic pointer’. First and second person have a prefixing element ngo-
that also occurs before women’s names or kin terms in all North Halmahera
languages, e.g. ngo-Desi ‘Desi’ (see e.g. van der Veen 1915: 194; van Staden
2000: 92; Visser and Voorhoeve 1987: 53). In Sahu, ngo is in opposition
with a marker a before nouns referring to men, but this is not found in the
other languages. Its presence on first and second person pronouns could be
explained if ngo- is considered a kind of vocative element used to refer to
those present in the discourse. However, the noun markers ngo- (and a- in
Sahu) is not used before proper names or kinship terms when addressing a
person (Visser and Voorhoeve 1987: 36).
Where Tidore and Ternate (cf. Hayami-Allen 2001: 136–137) are rather
different from the other languages is that they lack a set of possessive pro-
nouns, which can be formed morpho-syntactically in the other members of
the North Halmahera family.
In the Northeast Halmahera languages as well as in Sahu the personal
pronouns may be preceded by a specific marker to give what appear to be
possessive pronouns:
Pagu
(84) Oli to-una ai-kadu dumoi, w-a-gogon-oka
so POS-3SG.M 3SG.M.POS-sack one 3SG.M-NH-hide-NFUT
‘So he hid his sack (lit. so his sack, he hid it).’ (Wimbish 1991: 158)
Tobelo
(85) To-munanga ami-gakana nenanga?
POS-3SG.F 3SG.F.POS-knife this
‘Is this her knife?’ (Hueting 1936: 353)
In the Northeast Halmahera languages, this marker is to for all persons,

with the exception of the 2PL form in Tobelo which takes ti. In Sahu the
marker is tV, where V assimilates to the first vowel of the pronoun, e.g. to
ngoi ‘my, mine’, ta ngana ‘your(s)’, ti ngini ‘your(s) (plural)’, but is /o/
before /u/, i.e. to mungana ‘her(s)’ (Visser and Voorhoeve 1987: 53). This
possibility is completely absent in Tidore and Ternate.
Van Baarda (1908: 69–70) considers to a preposition ‘of, for’, e.g. ton-
gohi ‘mine or for me’, but most authors treat it as a prefix, e.g. Wimbish on
Pagu (1991: 139–140), or give it as part of the base form as Hueting does
for Tobelo (1936: 353). In Sahu the possessor marker preceeds the noun
markers a and ngo, as in example (86), supporting rather the prepositional
analysis, in which case it may be considered a semantic marker. It indicates
that the noun phrase has the role of possessor:
Sahu
(86) Wala ge t-a Salaka
house that POS-NM Salaka
‘That house is Salaka’s’ (Visser and Voorhoeve 1987: 54)
This particular prefix or preposition is rather interesting for two reasons.

First, it raises the question whether perhaps this reveals an underlying loca-
tive schema for the possessive constructions in North Halmahera. Recall
that Tidore has the locative prepositions te/toma and these could perhaps be
related to this possessive to. But the second reason is the more intrigueing
one. Perhaps these constructions with to reveal something about the nature
of the possessive construction in the North Halmahera languages. If they
form possessive pronouns, do these occur attributively or predicatively? Or
is there no distinction here, as in the Tidore possessive construction? And if
Van Baarda is correct and to is a semantic role marker rather than a posses-
sive pronoun deriving prefix, does this support the analysis of possessive
constructions as predications?
To start with the first question on the relation between locative preposi-
tions te and toma in Tidore and possessive to, it is indeed tempting to see
the two as related. This would then suggest an instance of locative-to-
possessive grammaticalisation (Heine 1997). However, the status and his-
tory of such locative prepositions in the North Halmahera languages is un-
clear. Sahu has a preposition toma11 but its distribution has not been de-
scribed and it is not in opposition with te but with ra (John Severn, pers.
comm.). A form to occurs infrequently in Tabaru (Fortgens 1928: 389–390)
to derive verbs with a locative component, e.g. toobiri to arrive by night,
and Fortgens suggests that the many names of languages beginning in to
(Togutil, Tobelo, Tabaru and even Tidore, regularly pronounced as Todore)
may also find their origin in a to- prefixed root. But signs of a locative
preposition to are completely absent in other Northeast Halmahera lan-
guages such as Pagu, and are generally unexpected in SOV languages.
In relation to the second question, pronouns preceded by to may occur
dependently and independently. In other words, they may occur before
noun phrases as apparent attributes (my, your, etc.), or they may constitute
noun phrases on their own (mine, yours, etc.) as in the following example:
Tobelo
(87) to-ngohi ahi-tiwi nenanga? koali, to-ngohi.
POS-1SG 1SG.POS-money this no POS-1SG
‘Is this my money? No, (it’s) mine.’ (Hueting 1936: 353)
In all the North Halmahera languages that have to/tV it also occurs on the
question word particle, as in (88):
Tobelo
(88) to-nago-nanga manga-tau nenanga?
pos-who-3P 3P.POS-house this
‘Whose house is this?’ (Hueting 1936: 355)
The marker is, however, not obligatory – at least not in all languages and in
all contexts. What precisely determines its presence is still not entirely clear.
In Galela it is absent in contrastive expressions and after conjunctions or
adverbs (van Baarda 1908: 70), but not in Pagu and Loda. And in Pagu it is
11
Visser and Voorhoeve (1987) do not mention it in their sketch grammar although
it occurs in one of the examples in contrast to re the preposition for company /
instrument (cf. Tidore se) (p.56).
absent when the possessor pronoun is preceded by the company/instrument

preposition de ‘and, with’, as in (89):
Pagu
(89) … wa-make nage de ami lomang-oka.
3M.SG.A:3NH.O-see that with 3SG.F.POS name-N.FUT
‘… he saw that it had her name (lit. he saw it already with her name).’
(Wimbish 1991: 133)
Also in assertion of possession the pronouns again do not always take to

(90), although they may (91):
Pagu
(90) Ei ngoi ai-ngoak iwa.
hey 1SG 1SG.POS-child gone
‘Hey, I don’t have any children.’ (Wimbish 1991: 135)
Pagu
(91) To-ngoi naga alat-oka.
POS-1SG ‘is’ tool-NFUT
‘my tool is already here.’ (lit. mine is tool-already)
(Wimbish 1991: 135)
Another question is whether this prefix or preposition is part of the pro-

nominal paradigm or whether it is an element that occurs in other environ-
ments as well. In Sahu, it occurs with pronouns as well as nouns to ‘empha-
size the possessorship’ (Visser and Voorhoeve 1987: 53), and although the
other studies do not make explicit reference to the occurrence of to outside
the pronominal paradigm, textual examples show that this is possible in at
least several of the Northeast Halmahera languages, as e.g. in the following
example from Tobelo:
Tobelo
(92) Nenanga ma-hangaji ai-ja? Koali,
this 3NH.POS-district.head 3SG.M.POS-fishing.net NEG,
to ma-uku.
POS 3NH.POS-hukum
‘Is this the district head’s fishing net? No, the hukum’s.’
(Hueting 1936: 353)
Clearly, on noun phrases to is not obligatory as (93) shows:

Pagu
(93) Ma kolan awi-ngoak mosoles dauk
ART king 3SG.M.POS-child bachelorette there:south
m-o-matetengo, ami-penjagaan o macan de o garuda.
3SG.F-S-one 3SG.F.POS-guard ART tiger and ART eagle
‘The king’s only daughter is there, and she has a tiger and an eagle
for guards.’ (Wimbish 1991: 136)
Summarising, all North Halmahera languages except for Tidore and Ternate
have a marker to that occurs before the personal pronouns, and at least in
some languages it also occurs before possessor noun phrases. In most lan-
guages the distribution of to is restricted at least in part by the occurrence
of adpositions and adverbs, and in Sahu discourse considerations also play
a role in its distribution. It appears then that to does not so much derive
possessive pronouns as mark the semantic role of the possessor in the pos-
sessive construction. This ties in with a predicative analysis for these lan-
guages also. The marker distinguishes possessor ‘first arguments’ from
other first arguments as a distinct semantic role in the clause, giving the
possessive relation as a special event type.
But is there further support for predicative possession also in the other
North Halmahera languages? And how does the possessive construction in
these languages shed any light on the Tidore situation? In order to address
this question it is necessary to examine the possessive prefixes and the sub-
ject and object markers in the North Halmahera languages in some detail.
8.3. Possessive and other person prefixes in the North Halmahera languages
The possessive prefixes in the North Halmahera languages all bear close
resemblance. The differences that we do find are, however, rather signifi-
cant. Also the differences in the systems of verbal cross-reference marking
between Tidore on the one hand and the Northeast Halmaheran languages
on the other is important and may shed light on the curious synchronic
situation in Tidore. In this section I explore the possible origin of the pos-
sessive markers in Tidore as related to the object cross-reference markers
that are found in the Northeast Halmaheran languages but not in Tidore.
Table 2 gives an overview of the possessive prefixes in a number of North
Halmahera languages. The bracketed elements are optional, often charac-
teristic of informal, spoken language.
Table 2. Possessive marking in several North Halmahera languages
possessive prefixes
Tidore Tobelo Tabaru Galela Sahu
singular
1 ri ahi ai (a)i (a)ri
2 ni ani (a)ni (a)ni (a)ni
3M i ai (ani) wi (a)wi ai
3F mi ami (a)mi (a)mi (a)mi
3 NH maa ma ma ma ma
plural
1 EXC mi mia mi(a) minga amia minga
1 INCL na nanga nanga nanga nanga
2 ni nia ni(a) ninga ania ninga
3 H.PL na manga manga manga ma(nga)
3 NH ma ma ma ma ma
a
ma is incongruous in all aspects. It is likely that it has its origins outside the pos-
sessive paradigm, possibly as a stative verb marker (see Holton 2006) and then
saw its distribution widened to include relations of associated and relational
possession (and reflexivity). This use then developed into the distinction be-
tween human plural and non-human possessors, which is otherwise absent in the
paradigms of person affixation.
Table 3. Pronouns, subject and object markers in the North Halmahera languages
Pronouna subject prefix object prefixb

singular
1 ri/hi to- (h)i-
2 na no- ni-
3M u wo- (w)i-
3F mi mo- mi-
3 NH e - a/ia/ya/yo-
plural
1 EXC m/mi mi- mi-
1 INCL ne p/h/w/fo- na-
2 n/ni ni- ni-
3 H.PL o/a yo- a-/yo-/ki-
3 NH e yo- a/ ia/ya/yo-
a
This column gives the roots from the forms in Table 1. Slashes indicate different
forms in different languages.
b
Not in Tidore, Ternate and West-Makian.
8.3.1. Prefix a-
If we compare the forms in Table 2 to the pronominal roots, subject and

object prefixes in Table 3, it transpires that it is in particular the addition of
the prefix a- that distinguishes the possessive prefixes from the object pre-
fixes and to a lesser extent from the roots of the personal pronouns. Van der
Veen (1915: 195) following Schmidt (1895: 134) indeed suggests that the
possessive forms have been derived from the object markers, which would
in turn have been derived from the personal pronouns (van der Veen 1915:
47; Schmidt 1895: 138). We return to this issue below.
Clearly suffixing -a on the plural forms is not related to the prefix a-, as
the Galela paradigm shows where the two co-occur, e.g. 1EXC a-mi-a. Van
der Veen (1915: 195–196) and Schmidt (1895) suggest that the suffix nga
on the third person plural derives from an ‘old plural marker’. But on the
other plural forms minga ‘1PL.EXC’ and ninga ‘2PL’ it would occur by ‘false
analogy’ because the root morphemes are already distinct from the singular
forms and do not need the nga to disambiguate. But, it is as likely that if
nga is indeed an old pluralizer, this has been reduced to a in most of the
forms, including mia and nia, but survived in those forms where reduction
would yield homonymy.
In a number of languages the prefix a- in the possessive prefixes is op-
tional, as indicated by the brackets, and in some languages it is omitted al-
together. Its origin is unclear. Van der Veen (1915: 195) again follows
Schmidt (1895) treating it as deriving from the masculine and neuter noun
marker o because the two would not co-occur, and because whenever the
possessor is female, a- is replaced by ngo, which is otherwise the feminine
counterpart of o. This analysis cannot be correct. First, the position of the
article is before the feminine noun marker, while the possessive prefix fol-
lows it. Compare Tobelo ani ‘2SG.POS’ before a masculine possessum ver-
sus ngoni ‘2SG.POS’ before a feminine possessum (Hueting 1936: 353). Fur-
thermore, in several languages, including Galela and Tobelo, the article o
and the female noun marker ngo do co-occur, although the article and a do
not, e.g. o ngo Luri ‘the woman Luri’ in Tobelo (Hueting 1936: 353) and o
ngo-meme ma-roka ‘mother’s husband’ in Galela (van Baarda 1908: 30).
And finally there is no further support for a sound change from a to o. The
absence of the article before possessum noun phrases is then more likely
the result of phonological elision. It is notable, however, that this a- is re-
lated to the -a- that is found in the object.
8.3.2. Possessor and object markers
Returning to the similarity of the possessor prefixes to the object prefixes,

the analogy between the forms suggests a historical derivation of the former
from the latter. This would tie in with the general typology of the languages
of East Nusantara. As we have seen, in many languages in this area the
possessive construction involves markers derived from either the subject or
the object verbal cross-reference markers. At the same time, this historical
derivation would provide an explanation for why the possessive construc-
tions in these languages have clausal properties. However, the evidence for
derivation from object prefixes is not immediately straightforward. The
possible origin of the possessive marker proposed in this section is tentative
and requires a closer study of the history of the languages of North Hal-
mahera as well as a closer examination of the systems of verbal cross-
referencing in the Northeast Halmahera languages to substantiate the hy-
pothesis put forward here.
First, we need to examine what it is about the object markers that makes
them resemble possessive markers more than the other markers. Indeed, if
we compare just the roots of the personal pronouns, then all verbal cross-
reference prefixes and the possessive markers resemble each other since
they all derive from the personal pronouns. The reason the possessor pre-
fixes resemble the object markers more than the subject markers is because
of the irregular subject markers for 1SG (to) and 1PL.INCL (po), where the
object markers are not irregular, and because the subject prefixes end in o
rather than i. But the inclusive/exclusive opposition in the North Halmahera
languages is generally attributed to language contact with surrounding Aus-
tronesian languages (Klamer et al. 2008). The irregularity of the 1PL.INCL
form may then be the result of a later introduction of the forms into the sys-
tem. The irregular 1SG form is still unexplained, but it is notable that the
same form is found throughout the Papuan languages of the West Papuan
Phylum, including Maybrat, Moi, and Abun. The final vowel of the subject
prefix is likely a separate morpheme. This has been argued, for instance for
Pagu (Wimbish 1991). The forms in -o- indicate that there is only a subject
or that the object is either unspecified or human. As such it is in opposition
with a prefix ending in -a- , which indicates that the object is non-human,
directly affected by the verb and specified or definite (i.e. it occurs with
canonical transitive verbs).
Pagu
(94) t-o-sano
1SG-SU-ask
‘I ask (something)’ (Wimbish 1991: 46)
Pagu
(95) t-a-sano
1SG-NH-ask
‘I ask a specific question’ (Wimbish 1991: 46)
This means that if we disregard the inclusive form and consider the endings
in -o- to be separate morphemes, only the aberrant first person singular sub-
ject prefix is different from the object prefix and the possessive marker. So
then the question is whether the possessive prefix would be derived from
the object cross-reference marker or whether it could have derived directly
from the pronominal roots.
Yet there is another link to the object cross-referencing paradigm. In the
North Halmahera languages, the possessive markers typically had a prefix a-
and it is possible that this prefix is the same a- as the one occurring before
the object markers when these refer to referents that are non-human, directly
affected by the verb and specified or definite. One reason to assume this is
that at least according to Wimbish’ account, the Pagu language does not
allow for the possibility of having human denoting objects that are affected
by the event, and the other Northeast Halmaheran languages appear to be
similar in this respect. In the following paragraphs, I will try to make a
plausible case that the semantic role of possessor could be just this: a human
object that is affected by the event. If that is the case, then the generalisation
would be that all affected objects take a and all unaffected objects have o,
but how does that explain that there are no transitive verbs that have af-
fected human objects? The hypothesis is then that there is a restriction on
participant structure which stipulates that each predicate may have only one
argument that is actively involved in the event, either as an agent or as a
patient. This means that if there is a human agent, the patient cannot be an
affected human. But if there is no agent, then this leaves the possibility of
having an affected human object. And this is the possessive construction: a
predication without a subject, but with an affected human object.
The argumentation is as follows. First, we follow Wimbish and assume
that the subject prefixes are indeed historically bi-morphemic, consisting of
the pronominal roots followed by o in intransitives of when the second argu-
ment is either human or not affected by the predicate (“oblique”) as in Table 4.
Table 4. Morphological analysis of subject markers
singular plural
person ‘intransitive’ fused person pluralizer ‘intransitive’ fused
root marker form root marker form
1EXC h/ri o to mi ngaa o mi/miyo
1INC n/f/w/p/h o fo
2 ni o no ni nga o ni/niyo
3M u/w o wo
i/y nga o yo/i/ya
3F mi o mo
3NH e o yo
a
nga is the form of the pluralizer that survives today, but it is actually more likely
that it was ka or /a, as evidenced by the various languages that have a /k/ in the
3pl object prefixes.
For the irregular plural forms we assume that the old pluralizer, now lost in
the subject prefixes but still visible in the possessive markers in several
languages, prevented the kind of fusion that took place with the singular
forms, explaining their irregular endings in –i, as in Tidore, and the forms
miyo and niyo in the Northeast Halmahera languages. These subject markers
followed by o are again followed by the pronominal roots cross-referencing
the object when this is human. Fusion of subject and object prefix takes
place in a number of cases, reducing for instance the 3NH object pronoun e
to ‘zero’. In the case of a non-human affected object (direct object), the
marker is not o but a. Otherwise the process is identical, giving an identical
fused paradigm with the one presented in Table 4. This account gives a ‘de-
fective paradigm’ as in Table 5 as it appears to exclude the possibility of
affected human arguments:
Table 5. -o- and -a- marking in Pagu
-o- -a-
single argument + –
unaffected non-human object + –
unaffected human object + –
affected non-human object – +
affected human object ? ?
But what if the initial a- on the possessive markers is reinterpreted as a

marker of affected human objects? In that case the constructions are like
experiential constructions to the extent that there is no subject. This could
be explained as a constraint on the number of verbal arguments that may be
directly involved in the event. Only one argument may be involved either as
an agent (subject) or as a possessor (object). If there is a subject, the inter-
pretation of the predication is ‘eventive’ with possibly an indirectly affected
human object, or a non-human object which is never as involved as a human
argument. In experiential constructions, the subject is a third person plural
‘dummy’ and the non-affected object is the experiencer. When there is no
subject, the interpretation of the predication will be ‘possessive’, with the
object expressing the possessor.
There is one small hitch in Galela. In Tabaru and in Pagu, the subject
markers for 1EXC and 2PL are always miyo and niyo, regardless of whether
there is an unaffected non-human object or not. However, in Galela there is
an opposition, so that mi and ni are found with intransitive verbs and miyo
and niyo with transitive verbs that take unaffected non-human objects (and
of course miya and niya with affected non-human objects).12 One possibility
is that this is evidence for different paradigms for transitive and intransitive
verbs, whereby neutralisation of intransitive subject and transitive subject
followed by an unaffected 3NH object has taken place for all persons and
numbers, except for these two. However, I propose that the opposite has oc-
curred: in Galela the intransitive forms miyo and niyo have been reanalysed
as bimorphemic mi-yo and ni-yo in which the first element expresses subject
and the last object. In intransitive predications, the ‘object’ prefix was then
deleted. Tabaru and Pagu then give evidence of the original system.
What this means is that there is a way in which the similarity between
object prefixes and possessor markers in the North Halmahera languages
can be accounted for historically. This links the North Halmahera languages
to many other languages of East Nusantara, and it may explain the clausal
properties that possessive constructions in Tidore have. In various ways the
possessor markers are more archaic: they have retained the affected object
marker a-, but also the pluralizer -a still shows in most North Halmahera
languages, though not in Tidore.
12
Sheldon (1991) discusses only miya and niya, but Van Baarda also mentions the
other two forms.
9. Conclusion
Tidore like all North Halmahera languages is ‘verby’ in nature. Although

categories may be established the language allows virtually all categories to
fulfil the function of main predicate in a clause. What are categorially nouns
or adjectives or even interjections may occur as predicates in clauses, in
which case they receive first argument cross-referencing. There is a ten-
dency to see such processes as instances of nominalisation or denominalisa-
tion, but this is unnecessary. In fact, it obscures the observation that syntac-
tic templates are not defined in terms of categories but in terms of func-
tions. Possessive constructions predicate a possessive relation, regardless of
whether the construction as a whole is used to refer or to ascribe a relation.
This is deferred to pragmatics as the way in which a speaker employs the
construction in discourse. The result in Tidore is a split in the system of
argument cross-referencing distinguishing actor arguments from possessive
arguments. The first type comprises all the traditional first argument types
from controller of activity to experiencer or theme (cf. Dik 1997: 56); the
second type is the possessor first argument. A second result of this analysis
is that there is no clear distinction between attributive and predicative pos-
session in Tidore. Finally, the present analysis of the Tidore system ex-
plains the clausal properties that the construction has. Although this con-
struction is typologically rather unique, in this paper it was shown that
Tidore does not stand alone among its relatives, but that it is in fact an areal
pattern to have possessor markers that correspond to clausal argument
markers and even the clausal properties of possessive constructions are
shared in other unrelated languages of the same linguistic area.
References
Allen, W. Sidney
1964 Transitivity and possession. Language 40 (3): 337–343.
van Baarda, M. J.
1895 Woordenlijst Galelareesch-Hollandsch. Met Ethnologische Aanteek-
eningen, op de Woorden, die Daartoe Aanleiding Gaven. Den Haag:
Koninklijk Instituut voor Taal-, Land- en Volkenkunde van Neder-
lands Indië.
van Baarda, M. J.
1908 Leidraad bij het Bestudeeren van ’t Galela’sch Dialect, op het Eiland
Halmaheira. Den Haag: Nijhof.
Baird, Louise
2002 A grammar of Keo: an Austronesian language of East Nusantara.
PhD thesis, Australian National University.
Bowden, John
2005 Language contact and metatypic restructuring in the directional sys-
tem of North Maluku Malay. Concentric: Studies in Linguistics 31 (2):
133–158.
Capell, Arthur
1975 The West Papua Phylum: General, and Timor and the areas further
West. In New Guinea Area Languages and Language Study. Volume
1: Papuan languages and the New Guinea Linguistic Scene, Stephen
A. Wurm (ed.), 667–716. Canberra: Pacific Linguistics.
de Gruyter.
Dik, Simon C.
1989 The Theory of Functional Grammar. Part I: The Structure of the
Clause. Dordrecht: Foris.
Dik, Simon C.
1997 The Theory of Functional Grammar. Part I: The Structure of the
Clause. Second, revised edition, edited by Kees Hengeveld. Berlin /
New York: Mouton de Gruyter.
Dol, Philomena H.
1999 A Grammar of Maybrat: A Language of the Bird’s Head, Irian Jaya,
Indonesia. PhD thesis, Leiden University.
Foley, Willam A.
1986 The Papuan Languages of New Guinea: Cambridge Language Sur-
veys. Cambridge: Cambridge University Press.
Fortgens, J.
1928 Grammaticale aantekeningen van het Tabaroesch, Tabaroesche volks-
verhalen en raadsels. Bijdragen tot de Taal-, Land- en Volkenkunde
84: 300 –544.
Gravelle, Gilles
2004 Meyah: an east Bird’s Head language of Papua, Indonesia. PhD thesis,
Vrije Universiteit.
Grimes, Charles E.
1991 A grammar of Buru. PhD thesis, Australian National University.
de Groot, Casper
1983 On non-verbal predicates in Functional Grammar: the case of pos-
sessives in Hungarian. In Advances in Functional Grammar, Simon
C. Dik (ed.), 93–122. Dordrecht: Foris.
Haan, Johnson Welem
2001 The grammar of Adang: a Papuan language spoken on the Island of
Alor East Nusa Tenggara – Indonesia. PhD thesis, University of Sydney.
Hayami-Allen, Rika
2001 A descriptive study of the language of Ternate, the Northern Moluc-
cas, Indonesia. PhD thesis, University of Pittsburgh.
Heine, Bernd
1997 Possession: Cognitive Sources, Forces and Grammaticalization.
Hengeveld, Kees
1992 Parts of speech. In Layered Structure and Reference in a Functional
Perspective, Michael Fortescue, Peter Harder and Lars Kristoffersen
(eds.), 29–55. Amsterdam: John Benjamins.
Hengeveld, Kees and J. Lachlan Mackenzie
2008 Functional Discourse Grammar. Oxford: Oxford University Press.
Holton, Gary
2003 Tobelo. Languages of the World /Materials, Vol. 328. Munich: Lincom
Europa.
Holton, Gary
2006 The relational noun marker in Tobelo (Northeast Halmaheran). Tenth
International Conference on Austronesian Linguistics (10-ICAL).
Puerto Princesa City, Palawan, Philippines.
Hueting, A.
1936 Iets over de spraakkunst der Tobeloreesche taal. Bijdragen tot de
Taal-, Land- en Volkenkunde 94 (3): 295–407.
Jelinek, Eloise and Fernando Escalante
1988 ‘Verbless’ possessive sentences in Yaqui. In In Honor of Mary Haas:
From the Haas Festival Conference on Native American Linguistics,
William Shipley (ed.), 411–429. Berlin/New York: Mouton de Gruyter.
Jones, Linda K.
1986 The question of ergativity in Yawa, a Papuan language. Australian
Journal of Linguistics 6: 37–55.
Klamer, Marian A. F.
2003 ‘Report’ constructions in Kambera (Austronesian). In Reported Dis-
course: A Meeting Ground for Different Linguistic Domains, Tom
Güldemann and Manfred von Roncadorm (eds.), 323–340. Amster-
Klamer, Marian A. F., Ger P. Reesink and Miriam van Staden
2008 East Nusantara as a Linguistic Area. In From Linguistic Areas to
Areal Linguistics, Pieter Muysken (ed.), 95–149. Amsterdam /Phila-
delphia: Benjamins.
Kratochvíl, Frantiek
2007 A grammar of Abui: A Papuan Language of Alor. PhD Thesis, Leiden
University.
1985 Possessive constructions in Oceanic languages and in Proto-Oceanic.
In Austronesian Linguistics at the 15th Pacific Science Congress,
Andrew Pawley and Lois Carrington (eds.), 93–140. Canberra: Pacific
Linguistics.
2002 The possessive-benefactive connection. Oceanic Linguistics 41 (2):
339–474.
Margetts, Anna
2004 From implicature to construction: emergence of a benefactive con-
struction in Oceanic. Oceanic Linguistics 43 (2): 445–468.
Menick, Raymond
1995 Moi: a language of the West Papuan Phylum. a preview. In Tales from
a Concave World: Liber Amicorum Bert Voorhoeve, Connie Baak,
Dick van der Meij and Mary Bakker (eds.), 55 –73. Leiden: Projects
Division, Department of Languages and Cultures of South-East Asia
and Oceania.
van Minde, Don
1997 Malayu Ambong: Phonology, Morphology, Syntax. Leiden: CNWS.
Odé, Cecilia
2002 A sketch of Mpur. In Languages of the Eastern Bird’s Head, Ger P.
Reesink (ed.), 45–107. Canberra: Pacific Linguistics.
Reesink, Ger P.
1998 The Bird’s Head as a Sprachbund. In Perspectives on the Bird’s
Head of Irian Jaya, Indonesia, Jelle Miedema; Cecilia Odé and Rien
A.C. Dam (eds.), 603–642. Amsterdam: Rodopi.
Reesink, Ger P.
1999 A Grammar of Hatam. C–146. Pacific Linguistics. Canberra: ANU.
Reesink, Ger P.
2000 West Papuan languages: roots and development. Paper presented at
Papuan Pasts, Canberra.
Reesink, Ger P. (ed.)
2002 Languages of the Eastern Bird’s Head. Canberra: Pacific Linguistics.
Ross, Malcolm
1996 Contact-induced change and the comparative method: cases from
Papua New Guinea. In The Comparative Method Reviewed: Regular-
ity and Irregularity in Language Change, Mark Durie and Malcolm
Ross (eds.), 180–217. New York: Oxford University Press.
Ross, Malcolm
1997 Social networks and kinds of speech community events. In Archae-
ology and Language, Roger M. Blench and Matthew Spriggs (eds.),
209–261. London: Routledge.
Ross, Malcolm
2001 Contact-induced change in Oceanic languages in North-West Mela-
nesia. In Areal Diffusion and Genetic Inheritance, Alexandra Y.
Aikhenwald and Robert M.W. Dixon (eds.), 134–166. Oxford: Ox-
ford University Press.
Ross, Malcolm
2006 Metatypy. In The Encyclopaedia of Language and Linguistics, Vol. 8,
Keith Brown (ed.), 95–99. New York: Elsevier.
Schmidt, Pater Wilhelm
1895 Die Sprachlichen Verhältnesse von Deutsch-Neuguinea. Zeitschrift
für Afrikansiche, Ozeanische und Ostasiatische Sprachen II (4).
Shelden, Howard
1991 Galela pronominal verb prefixes. In Papers in Papuan Linguistics,
Tom Dutton (ed.), 161–175. Canberra: Pacific Linguistics.
Shelden, Howard
1998 Transitivity and Galela pronominal reference. In SIL Electronic
Working Papers SILEWP 1998-005. (http://www.sil.org/silewp/
1998/005/) Dallas, Texas: SIL International.
van Staden, Miriam
2000 Tidore: A linguistic description of a language of the North Moluccas.
PhD thesis, Leiden University.
van Staden, Miriam
2006 The body and its parts in Tidore, a Papuan language of Eastern Indo-
nesia. Language Sciences 28: 323–343.
van Staden, Miriam and Ger P. Reesink
2008 Serial verb constructions in a linguistic area. In Serial Verb Construc-
tions in Austronesian and Papuan Languages, Gunter Senft (ed.),
17–54. Canberra: Pacific Linguistics.
Taylor, John R.
1989 Possessive genitives in English. Linguistics 27: 663–686.
Taylor, John R.
1996 Possessives in English: An Exploration in Cognitive Grammar. Ox-
ford: Oxford University Press.
Tjia, Johnny
1997 Verb serialisation in Ambonese Malay. MA thesis, University of
Oregon.
Van den Heuvel, Wilco
2006 Biak. Description of an Austronesian language of Papua. PhD thesis,
Vrije Universiteit.
van der Veen, Hendrik
1915 De Noord-Halmahera'se taalgroep tegenover de Austronesiese talen.
Leiden: van Nifterik.
Verhaar, John W. M.
1995 Toward a Reference Grammar of Tok Pisin: An Experiment in Corpus
Linguistics. Oceanic Linguistics Special Publication, Vol. 26. Hono-
lulu: University of Hawai’i Press.
Visser, Leontine E. and C. Lambertus Voorhoeve
1987 Sahu-Indonesian-English Dictionary and Sahu Grammar Sketch.
Verhandelingen van het Koninklijk Instituut voor Taal-, Land- en
Volkenkunde, Vol. 126. Dordrecht: Foris.
Voorhoeve, C. Lambertus
1987 The Non-Austronesian languages in the North Moluccas. In Hal-
mahera dan Raja Empat sebagai Kesatuan majemuk, E.K. M. Masi-
nambow (ed.). Jakarta: Lembaga Ekonomi dan Kemasyarakatan Na-
sional, Lembaga Ilmu Pengetahuan Indonesia.
1987 The masked bird: Linguistic relations in the Bird’s Head area. In
Peoples on the Move, Paul Haenen and Jan Power (eds.), 78–101.
Nijmegen: Centre for Australian and Oceanic Studies.
1988 The languages of the North Halmahera Stock. Papers in New Guinea
Linguistics 26: 181–209. Canberra: Pacific Linguistics.
de Vries, Lourens J.
2004 A Short Grammar of Inanwatan, an Endangered Language of the
Bird’s Head of Papua, Indonesia. Canberra: Pacific Linguistics.
Wada, Yuiti
1980 Correspondence of consonants in North Halmahera languages and
the conservation of archaic sounds in Galela. In The Galela of Hal-
mahera: A Preliminary Survey, Naomichi Ishige (ed.), 497–529.
Senri Ethnological Series, Vol. 7. Osaka: National Museum of Eth-
nology.
Wimbish, Sandra Gay
1991 An Introduction to Galela through the Analysis of Narrative Dis-
course. Ann Arbor: University of Michigan Press.
Wurm, Stephen A.
1982 Papuan Languages of Oceania. Ars Linguistica, Vol. 7. Tübingen:
Gunter Narr Verlag.
Possessive expressions in the Southwestern Amazon
Hein van der Voort
1. Introduction
The Southwestern region of the Amazon river basin, which in the present
article is understood to be the region that covers the Bolivian and Brazilian
sides of the Guaporé, Mamoré and Upper Madeira rivers, is one of the
world’s linguistically most diverse places. The Southwestern Amazon, here
also referred to as the Guaporé region, harbours representatives of seven
different linguistic stocks, including Arawak, Chapacura, Macro-Jê, Nam-
bikwara, Pano, Tacana and the majority of the branches of the Tupi linguis-
tic stock. Furthermore twelve language isolates or unclassified languages
are spoken in the region. The majority of the approximately 50 languages of
the region are highly endangered with extinction, about half of them having
fewer than 50 speakers and a third ten or fewer. During the past two dec-
ades, a considerable number of initiatives to document and describe these
languages have fortunately been taken.
This article is meant to give a preliminary survey of the different ways
adnominal possession is expressed across a sample set of languages from the
Southwestern Amazon (see Map 1). In section 2 I will sketch the types of
adnominal possessive expression found in Kwaza, a language isolate spoken
by approximately 25 people. In section 3 I will discuss the similarities with
possessive expressions in other languages of the region: Aikanã (isolate),
Arikapú (Macro-Jê), Baure (Arawak), Kanoê (isolate), Latundê (Nambik-
wara), Mekens (Tupí) and Wari’ (Chapacura). The final section 4 contains a
table summarising the different possessive strategies. In view of the fact
that the languages discussed here belong to different families, and even in-
clude three isolates, explanations for certain similarities are probably found
in areal diffusion.
This article deals mainly with adnominal possession, which can be de-
scribed as a grammatical construction involving a modifying noun and a
head noun that expresses a possessive relationship between a possessor
(PR) and a possessum (PM), respectively. It contrasts with predicative pos-
session and external possession. Predicative possession constructions in-
volve the verb ‘to have’, or another verbal copula, and express a possessive
344 Hein van der Voort
relationship between a possessor argument and a possessum argument (e.g.

Stassen 2006, 2009). External possession concerns an understood posses-
sive relationship between entities that represent distinct arguments of a
(any) verb that do not (necessarily) carry possessive marking, as in ‘He
kissed her on the cheek.’ (e.g. Payne and Barshi 1999). Predicative posses-
sion will be dealt with only briefly, when relevant (sections 2.4, 3.1.2,
3.3.2), and external possession not at all, since they are not of central im-
portance here and merit discussion in separate articles.
Map 1. Location of languages discussed (© Willem Doelman).
The possessive relationship is prototypically one of ownership, but it can

also metaphorically express body part, kinship, availability, location, con-
trol and other more abstract relationships (e.g. Herslund and Baron 2001;
Koptjevskaja-Tamm 2006; Lichtenberk this volume). Adnominal posses-
sion is often called ‘attributive’ possession (e.g. Baron et al. 2001), but in
this article the term ‘attributive’ will for practical reasons be reserved for a
modifying relationship between a head noun and a dependent modifying
noun. Although the term ‘genitive’ can be used in a wider sense than just
case marking (Dryer 2007), it will be avoided here.
Possessive expressions in the Southwestern Amazon 345
According to Nichols (1992: 71) South American languages show a

preference for ‘head-marking’, which is confirmed for the Guaporé region
(Crevels and van der Voort 2008: 171). However, the Guaporé languages do
not unambiguously reflect Dixon and Aikhenvald’s (1999: 8) observation
regarding Amazonian languages that possession is typically marked on the
PM. As shown in section 4, three of the eight languages discussed in the
present article canonically mark possession on the PR rather than on the
PM. The issue of head- and dependent-marking is not a central concern in
this article and it will be touched upon only peripherally, in sections 2.2.1
and 2.3.
In this article I will approach possession primarily from the perspective
of form rather than of function. Hence, I will focus on the forms that are
dedicated to the expression of the possessive function, rather than elaborate
on the range of different functions that possessive forms and constructions
might have. Nevertheless, in order to analyse the origin and development of
specific possessive forms in the little-studied languages under discussion, it
will now and then be necessary to expand on alternative functions, usually
in separate subsections (e.g. § 3.1.1, § 3.2.2, § 3.4.2). Also, when languages
have no “dedicated” possessive expressions, such as Mekens and Arikapú,
it will be necessary to explain how the possessive function is fulfilled.
2. Possessive constructions in Kwaza
Kwaza, which has been studied and described since the mid 1990s by the
author (e.g. van der Voort 2004), is a morphologically complex language
with a variable word order, although there is a clear tendency towards head-
final structures. The grammatical categories of Kwaza are verbs, nouns,
adverbs and particles. Most of the grammatical complexity is found in ver-
bal morphology, whereas nominal morphology is relatively simple. How-
ever, the highly productive nominalisation possibilities enable the creation
of morphologically very complex nouns. The morphology is mainly suffix-
ing. The verb is obligatorily inflected for person and mood, although mor-
phological ellipsis is attested under specific circumstances. The morpho-
logical structure of the verb is: root + optional derivation + obligatory
person inflexion + obligatory mood inflexion, as shown in this example: 1, 2
1
The Kwaza consonants /c/ and /x/ are pronounced as IPA retracted [t] and [s]
respectively. The vowel /y/ is pronounced somewhere between IPA [] and [].
Main word stress is marked by an apostrophe ['] preceding the stressed syllable.
(1) cari-'nã-da-ki
shoot-FUT-1S-DEC
‘I will kill (it, e.g. the game animal).’
The verb that is unmarked for person is interpreted as having a third person
subject. Overt expression of arguments as pronouns or other independent
constituents is not obligatory. Nouns are not obligatorily inflected, there is
no nominal number marking, and the animate object case is not always ap-
plied. The distinction between inclusive and exclusive person reference is
found in both the pronominal and the bound person marking paradigms.
There are several basic moods, including declarative, interrogative, volitive,
imperative, negative imperative, etc. There are several moods for subordi-
nated adverbial clauses and there is a cosubordinated mood for clause
chains. Different nominalisers are used for attributive and complement
clauses. In the following subsections I will sketch the different possessive
strategies in Kwaza and analyse their morphosyntactic properties.
2.1. The canonical adnominal possessive construction
There are no specific possessive pronouns or inflexions in Kwaza. The

usual way to express an adnominal possessive relationship in Kwaza re-
quires derivation of the PR noun by the morpheme -dy-, followed by the
nominalising morpheme -h . This latter morpheme is probably the most
frequently occurring morpheme of Kwaza. As will be seen in § 2.2.1 and
§ 2.2.2, -h functions among others in attributive clause formation and as a
2
The following abbreviations are used: ART – article, AUX – auxiliary, BEN –
benefactive (verbal derivation), BER – beneficiary, C – coreferent, CAU – causa-
tive, CD – directional classifier, CL – classifier, COL – collective, COMIT –
comitative, COP – copula, CSO – cosubordination, DEC – declarative, DEM –
demonstrative, DET – detrimental, DR – directional, DS – different subject, F –
feminine, FUT – future, IMPF – imperfective, INCL – inclusive, INFL – Wari’
‘inflection’, INT – interrogative, LOC – locative case, M – masculine, N – neuter,
NOM – nominaliser, OBL – oblique case, P – plural, PM – possessum; POS –
possessive, POT – potential, PR – possessor; PROX – proximate, REF – referen-
tial, REFL – reflexive, S – singular, VOL – volitive. The numerals 1, 2, and 3
indicate respectively, first person singular, second person, and third person. The
hyphen (-) indicates a morphemic boundary; the equal sign (=) indicates compo-
sition or a clitic boundary; and the period (.) separates semantic units in a port-
manteau morpheme.
semantically neutral classifier. Other ways to express possessives in Kwaza

are less productive. The most frequently attested constituent order in pos-
sessive constructions is PR–PM:
(2) tawi'wi-dy-h a'xy

Tavivi-POS-NOM house
‘Tavivi’s house’
Personal pronouns can also occur as PRs:
(3) 'xyi-dy-h 'kopu

you-POS-NOM cup
‘your cup’
The only exception to the pattern described here is the first person inclusive
pronoun txana, in the possessive use of which omission of the morpheme
-dy- is preferred, i.e. txana-h ‘ours (INCL)’ rather than txana-dy-h .
Headless possessive constructions, in which an explicit PM is absent,
are frequently attested in Kwaza:
(4) na-'ay-h -dy-h i'si-ki

PROX-that-NOM-POS-NOM die-DEC
‘That person’s (son) has died.’
Note that the PR in the above example is a demonstrative expression. Since

demonstrative roots are verbal, the structure of demonstrative expressions
requires a nominaliser, hence the repeated occurrence of -h in a demon-
strative PR.
The use of bare or derived nominal stems as predicates (called ‘zero-
verbalisation’ in van der Voort 2004) is very productive in Kwaza. It may
also involve (headless) possessive constructions in different ways:
(5) 'si-dy-h -ki ('kopu)

I-POS-NOM-DEC cup
‘(The cup) is mine/it is my (cup).’ 3
(6) 'si-dy-h 'kopu-ki

I-POS-NOM cup-DEC
‘The cup is mine./It is my cup.’
3
Reverse constituent order is also possible.
2.2. Morphological characteristics of the canonical possessive construction
It is not so clear whether the morpheme -dy- should be considered deriva-

tional or inflexional. The morpheme is inflexional according to the criterion
that it does not change the semantics of the word to which it attaches.
Rather, it indicates the syntactic function of the word in the clause. Also the
characteristic position of stress, on the ultimate syllable of the (derived)
root, suggests that -dy- belongs to the inflexional layer of the word. How-
ever, the fact that the morpheme -dy- does not form part of a paradigm sug-
gests that it may be a derivational affix. In addition, the obligatory presence
of the nominaliser -h suggests that -dy- has caused a category change of
the preceding noun. This adds further support to the derivational analysis.
The nominalising morpheme -h is one of the most productive and fre-
quently occurring morphemes of Kwaza. It typically creates deverbal nouns
that can function as arguments and attributes. Besides functioning as a
straightforward nominaliser, it occurs with temporal and aspectual effects
in recursively derived verbal constructions. Also it is often used as a se-
mantically neutral classifier, in case a specific classifier is not employed
(this will be explained in §2.2.2). The possessive construction may actually
be regarded as one of the least expected habitats of -h . The obligatory oc-
currence of this element in possessive constructions may be a reason to
consider the combination -dy-h as lexicalised. Consider the following ex-
amples:
(7) 'si-dy-h
I-POS-NOM
‘mine’
(8) *si-dy-ki
I-POS-DEC
‘It is mine.’
If the element -h really does have a nominalising function under all cir-
cumstances, example (8) would be grammatical. In the following subsec-
tions, I will discuss the elements -h (§ 2.2.1, § 2.2.2) and -dy- (§2.2.3), in
order to better understand the combination -dy-h .
2.2.1. Similarities with the attributive construction
The possessive construction has similarities with the attributive construc-

tion. On the structural level, attributive modification of nouns is expressed
by juxtaposition of nominal constituents, as shown by the following exam-
ple:4
(9) ywy'nw duma'ru

tree ladle
‘big wooden spoon’
In terms of parts of speech membership, attributively modifying lexemes in

Kwaza are mostly verbs, for example: awy- ‘to be cold’, arwa- ‘to be new’,
atsile- ‘to be heavy’, wotsu- ‘to be skinny’. Even quantitative expressions
such as aky- ‘to be two’ (as in example (13)) are verbs. Hence, on the mor-
phological level, attributive modification of nouns usually involves nomi-
nalisation:
(10) a'xy arwa-'h

house new-NOM
‘new house’
In fact, the nominalised verb can be morphologically very complex and

contain derivational morphemes such as classifiers, and person inflexion.
Therefore, no fundamental distinction can be drawn between relative
clauses and attributive constructions. The contrast between the following
examples shows that replacement of the word-final verbal mood marker by
a nominaliser is enough to create a relative clause (see also van der Voort
2006):
(11) atxitxi'n bar -'ri-da-ki

pancake heat-CL:flat-1S-DEC
‘I baked the maize pancake.’
4
Even though the relatively rare modifier-noun word order could help support an
analysis of (9) as a (classifying) compound, I chose to regard it as an attributive
construction since, in contrast to verbal compounds, full nominal compounds
are very rare in Kwaza. Furthermore, Kwaza classifiers are suffixes and cannot
occur as free elements (van der Voort 2004: 128ff.).
(12) atxitxi'n bar -'ri-da-h

pancake heat-CL:flat-1S-NOM
‘maize pancake which I baked’
Also, headless attributive constructions, including headless relative clauses,

are frequently attested:
(13) (a'le) aky-'h

axe two-NOM
‘the two (axes)’
Since possessive constructions also contain a nominaliser, one might won-

der whether they can be distinguished from attributive constructions at all.
There are good reasons to assume there is a fundamental distinction, how-
ever. In the first place, there is a morphological difference between posses-
sive and attributive constructions. Even though the possessive construction
requires the use of a nominaliser, the preceding possessive morpheme -dy-
does not appear to produce verb roots. As examples (7) and (8) above
show, the nominaliser cannot be replaced by a mood marker. By contrast,
morphologically complex attributive constructions such as (12) are probably
derived from predicative constructions such as (11) in which a mood marker
occurs instead of a nominaliser.
Second, there is a marked difference in constituent order. As mentioned
in section 2, in spite of the attested syntactic variability, the language tends
to prefer head-final structures. Consequently, the great majority of the pos-
sessive constructions are head-final, whereas attributive constructions tend
to be head-initial. An explanation of this unexpected attributive order is
probably found in the predicational origin of the relative clause. Its forma-
tion is a purely morphological operation: a nominaliser is applied to a head-
final predicative expression like (11), which, without change of constituent
order, results in a head-initial attributive expression like (12). Only in the
case of underived noun-noun attributive constructions, as in example (9),
head-final order is attested with some frequency.
For the reasons discussed in this and the previous section, it is doubtful
whether the possessive construction is a subtype of the attributive construc-
tion; moreover, it is possible that the element -h in the combination -dy-h
does not function as a nominaliser. Indeed it may well be synchronically
unanalysable.
2.2.2. The occurrence of classifiers
As mentioned above, the nominaliser -h can function as a neutral classifier.

Conversely, classifiers can also have nominalising properties and can be
considered as bound derivational morphemes. Kwaza classifiers are de-
scribed and analysed exhaustively in van der Voort (2004). In this subsec-
tion I will discuss their behaviour in possessive constructions and their con-
sequences for the analysis of -dy-h .
Kwaza has an enormous arsenal of classifiers (over a hundred), that
have a very wide distribution and that express semantic characteristics in-
cluding shape, substance, texture, part-whole relations, direction and sex.
Their actual meanings can range from highly abstract, such as ‘thing’ to
uniquely specific, such as ‘spoon’. In principle, all classifiers have the same
distribution. Moreover, in almost every place where the nominaliser -h can
be found, a classifier can replace it. Hence, classifiers can occur in attribu-
tive, numeral, demonstrative, interrogative and adverbial expressions, and
they can furthermore be attached to nouns and incorporated in verbs. They
can also occur in possessive expressions. In Kwaza, classifiers describe
properties of their referents and they may be used anaphorically. Note the
following examples that contrast the use of a specific classifier with the use
of -h as a neutral classifier:
(14) ha-ha-'ro-ki
clean-clean-CL:vessel-DEC
‘He washes pans.’
(15) ha-ha-'h -ki

clean-clean-NOM-DEC
‘He washes things.’
In attributive constructions, a specific classifier can occur instead of -h :
(16) (a'xy) aky-'xy

(house) two-CL:house
‘two houses’
(17) aky-'h
two-NOM
‘two (ones)’
The following examples show the contrast between a predicate and a head-
less relative clause that is nominalised by a specific classifier:
(18) a'xy-dy-'xa-tsy-tse
house-CAU-2S-POT-DEC
‘You are going to make a house.’
(19) a'xy-dy-'xa-tsy-ka'n a'w i-da-ki

house-CAU-2S-POT-CL:oblong see-1S-DEC
‘I saw boards for you to make a house with.’
Also in possessive constructions classifiers can occur instead of the element

-h . In principle all classifiers can occur in this position and there is no spe-
cific subcategory of classifiers restricted to possessive constructions exclu-
sively. In addition, the possessive constructions in which classifiers occur
are canonical and do not represent a separate strategy to express possession.
Thus, Kwaza does not have the type of possessive classifiers found in many
Austronesian languages – see e.g. Lichtenberk (this volume) and Crowley
(1995). Consider the following example:
(20) (a'xy) 'si-dy-xy

house I-POS-CL:house
‘my house’
Since the classifier -xy in this example has the specific semantic content
‘house’ (it is indeed etymologically related to the noun a'xy ‘house’), its
referent can be omitted without loss of semantic content, similar to headless
relative clauses and the pro-drop possibility of predicates.5 Compare the
following to example (2), with which no difference in meaning was attested:
(21) tawi'wi-dy-xy
Tavivi-POS-CL:house
‘Tavivi’s house’
The next example contains a headless possessive construction modified by

a relative clause, with the specific classifier -xy acting almost like a cross-
reference morpheme:
5
Overt pronominal expression of arguments was attested to have a disambiguating
or emphatic effect. The pragmatic consequences of full versus reduced expression
of the PM have not been investigated.
(22) 'si-dy-xy haka-xy-'na 'a-xa-ki

I-POS-CL:house old-CL:house-LOC exist-2S-DEC
‘You live in my old house.’
Many classifiers characterise or refer to specific entities for which no inde-

pendent monomorphemic noun is available. Such classifiers can be attached
to the noun formative root e- to form an independent noun (see also § 2.4).
They can also occur in other expressions and replace the element -h . No
difference in meaning was attested between the following examples:
(23) o'lu-dy-h e-'kai

curassow-POS-NOM Ø-CL:leg
‘leg of a curassow bird’
(24) o'lu-dy-'kai
curassow-POS-CL:leg
‘leg of a curassow bird’
Many body-part classifiers are of this type and probably just as many are
not. Furthermore, a number of classifiers that have a very general referential
scope and that are not etymologically related to a full noun can also occur
with the empty root e-. The following examples illustrate the classifiers -ro
‘container, vessel, cup’ and -ri ‘flat object’:
(25) 'si-dy-ro 'dai-xa-ki

I-POS-CL:vessel grab-2S-DEC
‘You took my cup.’
(26) e-'ro
Ø-CL:vessel
‘container’, ‘radio’ 6
(27) a'ri-dy-h e-'ri

tapir-POS-NOM Ø-CL:flat
‘the tapir’s liver’
6
The second interpretation of this example is an illustration of (the frequently
attested) lexicalisation of classifiers in combination with other elements.
(28) a'ri-dy-'ri

tapir-POS-CL:flat
‘the tapir’s liver’7
The most general classifier is the semantically neutral nominaliser -h , both

in possessive constructions and in any other construction that can feature a
classifier. As shown above, classifiers can have nominalising properties and
they can also replace -h in possessive constructions. For these reasons, the
combination -dy-h (discussed in §2.1 and §2.2) should not be regarded as
a single fossilised possessive morpheme. It is a partially lexicalised combi-
nation in the sense that it functions as an idiom on the morphological level.
But it is also still transparent in that the constituting elements have pre-
served their original functions and permutability to a certain extent. This
kind of semiproductive morphological patterning represents a phenomenon
also attested in other parts of the grammar.
2.2.3. Homophonous forms of -dy-
In view of the special position that possessive -dy- occupies in the general
make-up of Kwazá nominal morphology, it is necessary to consider other
morphemes with which it is homophonous. There is another form -dy- that
most likely originates in the possessive morpheme. It is used in a fixed
combination with the adverbial nominaliser -nãi, with the specific meaning
‘language of’. This construction is adverbial and no possessed head can be
identified on the semantic level, as the following examples show:
(29) 'tyka-dy-nãi jã'si-ta

Mekens-POS-NOM hear-CSO
‘they understood the Mekens language’
(30) 'xyi-dy-nãi are'ta-da-m

you-POS-NOM know-1S-VOL
‘I want to learn your language.’
7
In some rare cases the omission of the possessive morpheme may not seem to
change the meaning, as in ari-'ri ‘the tapir’s liver’. The structure of ari-'ri is
fundamentally different from (28), however, since it (ari-'ri) represents a clas-
sified nominal comparable to a nominal compound. A similar alternative for ex-
ample (24), *olu-'kai, was rejected by the consultant.
One of the reasons to consider -dy- as originally possessive here is its simi-
lar behaviour with regard to the first person inclusive pronoun. The expres-
sion txana-nãi ‘our language’ is preferred to txana-dy-nãi. Remember that a
similar preference with regard to the canonical possessive was mentioned
in the paragraph under example (3).
An additional morpheme that looks suspiciously like the canonical pos-
sessive is the comitative case suffix -dyn . Consider example (31):
(31) jere'xwa-dyn hyhy'rwa-da-ki

jaguar-COMIT move-1S-DEC
‘I’m walking with the dog8 (by my side).’
The comitative belongs with the beneficiary -du, instrumental -ko and loca-
tive -na to the semantic cases of Kwaza; these do not mark obligatory
grammatical arguments but optional “oblique” arguments. There are both
formal and semantic similarities with the canonical possessive construction,
but there are also important differences. Even though there could be an
etymological relationship with the possessive morpheme -dy-, and perhaps
with the verbal derivational reflexive morpheme -n -, the comitative mor-
pheme -dyn is not further analysable synchronically. And even though in
certain languages the comitative can mark a PM (e.g. McGregor 2001: 81;
Stassen 2006: 771) nothing like that has been observed in Kwaza.
Finally, the possessive morpheme -dy- is homophonous with the causa-
tive/benefactive morpheme -dy- and with the different subject marker -dy-.
It is possible that they are etymologically related, but it is difficult to dem-
onstrate this empirically. Being mainly verbal suffixes, their distribution is
to a large extent complementary with that of possessive -dy-. However,
their very different specific functions, combined with their very limited and
specific distributions are sufficient reason to consider of each of these items
synchronically as separate morphemes.
2.3. An alternative adnominal possessive
Kwaza has no clear inflexional paradigms for possession according to person

or other categories. Instead, person and number are distinguished through
pronouns, as seen in the canonical possessive construction presented above.
8
In Kwaza as well as in various other languages of the region, the words for ‘jag-
uar’ and ‘dog’ are identical.
Nevertheless, Kwaza has an alternative possessive construction that is used

frequently, although its distribution is restricted. Whereas the canonical
possessive is marked on the PR, with (neutral or specific) reference to the
semantic class characteristics of the PM, the alternative possessive is
marked on the PM, with reference to a third person PR. It is created by at-
taching the suffix -tjate directly to a nominal PM. Note the following ex-
ample, which is based on the independent noun mã ‘mother’:
(32) mã-tja'te
mother-3.POS
‘his mother’
The possessive morpheme -tjate does not form part of a paradigm and it
refers exclusively to a third person PR. It has no other uses and there are no
other head-marking possessive strategies to express different persons. The
morpheme -tjate is not further analysable. It may be a borrowed morpheme,
as will be discussed in (§3.1.3).
Overt expression of the dependent PR is rare in the alternative posses-
sive construction, but it is attested:
(33) hu'ri en-tja'te-(na) ojabu'ru-da-h -ki

paca foraging.place-3.POS-LOC arrive.thither-1S-NOM-DEC
‘I arrived at the foraging place of the pacas.’
Herslund and Baron (2001: 14) discuss the phenomenon that languages make
use of different adnominal possessive constructions to distinguish different
kinds of possessive relationships, such as alienable vs. inalienable. Rijkhoff
(this volume) mentions a special definiteness effect for third person PR
markers in Uralic and Turkic languages. In Kwaza, however, no such dis-
tinctions were attested between the alternative and the canonical possessive
in the third person. The following contrasted examples are semantically
identical:
(34) kanwã-tja'te
canoe-3.POS
‘his canoe’
(35) '
-dy-h ka'nwã
he-POS-NOM canoe
‘his canoe’
No productive strategies are attested to distinguish between reflexive and

non-reflexive possession. The different adnominal possessive constructions
do not seem to contribute to such a distinction at any rate:
(36) tsu'ty-si'ki
-dy-'h 'bu-ki
head-CL:skin he-POS-NOM put-DEC
‘He uses his own hat.’
(37) tsu'ty-si'ki-tja'te 'bu-ki

head-CL:skin-3.POS put-DEC
‘He uses his own hat.’
In the following example, the object can refer either to the wife of the third
person subject or to the wife of another third person:
(38) etay-tja'te 'ta-ta

woman-3.POS talk-CSO
‘hei said to hisi/j wife’
Recursive embedding of possessives is attested, though very rarely. There is

one example of an alternative possessive construction inside a canonical
possessive construction (see also van der Voort 2004: 202):
(39) h
dj-tja'te-dy-h eto'hoi
older.brother.female-3.POS-POS-NOM child
‘child of older brother of female’9 (lit. ‘her older brother’s child’)
Furthermore, there are a handful of attestations of a canonical possessive

inside an (redundant) alternative possessive construction:
(40) dodo'txi-dy-'h e'tay-tja'te

latex-POS-NOM woman-3.POS
‘Latex-man’s wife’
The next example is rather exceptional because the head of a canonical

possessive has two referents:
9
This expression was equated by the informant with the more general kinship
terms kore ‘son of brother or sister’ and koretay ‘daughter of brother or sister’.
(41) 'xyi-dy-h a'ha-tjate

you-POS-NOM father-3.POS
‘father of your father (i.e. your grandfather)’ (lit. ‘your father’s’)
2.4. The dummy root e- and the verb ‘to have’
As mentioned in the previous section, many languages have different ad-

nominal possessive constructions with different semantic/pragmatic uses.
In Kwaza, no productive strategies are encountered that express a distinc-
tion between alienable and inalienable possession. The semantically empty
noun-formative dummy root e- discussed towards the end of § 2.2.2 might
represent a historical remnant of an inalienable construction, since it occurs
often with part-whole expressions that may involve obligatory possession.
This must probably be regarded today as a regularity in the lexicon and not
as a productive morphosyntactic strategy (assuming e- ever had anything to
do with alienability distinctions). Even granted that alienability distinctions
are culture- or language-specific (Chappell and McGregor 1995), no consis-
tently observed criteria for such a distinction in Kwaza could be identified.
Besides expected inalienable PMs such as body parts and kinship relations,
the e- root occurs also with expressions for items that one would not expect
to count as inalienable. Furthermore, not all entities that one would expect
to count as inalienable PMs require e-. As an example, the body part noun
tsuty ‘head’ does represent a likely candidate for inalienable concepts even
though it does not contain the empty root. By contrast, the noun e-m ‘liq-
uid’ does not seem to represent an inalienable or obligatorily possessed
concept in spite of the empty root e-. The following example refers to a leg
that may be cut off or not:
(42) e-'kai-tja'te
Ø-CL:leg-3.POS
‘his leg’
Another analysis of the dummy root e- relates it to the verb e- ‘to have’,
with which it is homophonous. The verb e- ‘to have’ is used to express
predicative possession in Kwaza, as in:
(43) si eto'hoi aky-'h e'mã 'e-da-ki

I child two-NOM three have-1S-DEC
‘I have three children.’
This verb occurs also in specific habitual constructions, where it can be lit-
erally interpreted as an abstract predicative possessive:
(44) duture-xwa'nã ja-'nãi e-'re

pig-CL:flesh eat-NOM have-INT
‘Does she eat pork?’ (lit. ‘Does she have the eating of pork?’)
The relationship between the verb ‘to have’, possession and aspect has been
discussed in typological work by among others Benveniste (1960), Heine
(1997) and Seiler (1977).
To a very limited extent, the verb e- ‘to have’ is also used as an existen-
tial verb. The normal existential verb in Kwaza is a- ‘to be’, which in addi-
tion functions in another specific habitual construction. As it happens, there
is also a semantically empty adverb formative root a-. This dummy root is
used to create adverbs out of bound verbal derivational directional markers.
Since the two roots e- and a- both can function as semantically empty for-
mative dummy roots and as semantically very abstract verb roots, their par-
allel distributions suggest a relationship between their different functions.
Further discussion of the dummy roots would take us too far from the pre-
sent topic and merits a separate article. The reason for mentioning their
parallelism here is to show that the empty noun formative e- and the verb
root e- ‘to have’ should probably be considered as the same element.
The dummy root e- or i- is also encountered (without the meaning ‘to
have’) in Kanoê (§3.2), Latundê (§3.3), Baure (§ 3.4) and other languages
of the Guaporé region. It was discussed as an areal feature in Crevels and
van der Voort (2008: 167–168) and van der Voort (2005: 397–398).
3. Possessive constructions in neighbouring languages
The previous section is a relatively detailed case study of Kwaza (isolate). In

the present section I will discuss similarities and differences with possessive
constructions in Aikanã (isolate), Kanoê (isolate), Latundê (Nambikwara),
Baure (Arawak), Wari’ (Chapacura), Mekens (Tupi) and Arikapú (Macro-Jê),
in that order.
3.1. The canonical possessive construction in Aikanã
The Aikanã (also referred to as Massaká, Tubarão or Huari) are the tradi-
tional neighbours of the Kwaza. Their language is probably also an isolate,
although there are some lexical and structural similarities with its neigh-
bouring languages. Aikanã is spoken by about 150 persons. Since the mid
1980s the language was studied and was partially described by the Brazilian
linguist Ione Vasconcelos (2003) and by an American team of linguists led
by Leanne Hinton (ed. 1993), and it is presently being studied by the author.
Aikanã is a morphologically highly complex language. Like Kwaza, most
of the morphological complexity resides in the verb. Furthermore it has
similar classifying and directional suffixes, valency-changing suffixes, a
distinction between future and non-future tense, person and mood inflexion,
animate object case marking, and no nominal number inflexion. Unlike
Kwaza, but similar to another Rondônian isolate Kanoê (§ 3.2), Aikanã has
several different verbal inflexion classes, some of which require argument
prefixes rather than suffixes. Also, it lacks an inclusive-exclusive distinc-
tion.
With respect to possessive expressions, Aikanã shows both differences
and similarities to Kwaza. Unlike Kwaza, Aikanã has a (partially transpar-
ent) set of specific possessive pronouns. Some of the possessive pronouns
contain the element -z, which probably originates from the productive
beneficiary case suffix that also has a genitive function.10 Furthermore, a
specific verbal benefactive inflexional paradigm exists for beneficiary and
possessive objects. In the following table the different Aikanã pronouns and
benefactive inflexions are listed:
Table 1. Aikanã pronouns and benefactive inflexions
personal possessive benefactive

1S hi'sa txü'txü -ku-, -k-
2S h
'zã h
'z -u(a)-
3S kaj'ne, ka'ria kajne'z, kari'z -w(e)-
1P sa'te sate'z -kjã-
2P h
zã('za) h
z'za -uãjã-
3P kaj'ne?ene, ka'ri?ene kari?ene'z -we?eje-
a
The first form kaj'ne ‘he, she, it’ is a third person pronoun. The second form ka'ri
‘that, that one’ is a demonstrative pronoun that is often used as a third person
pronoun.
10
Aikanã /z/ is usually pronounced as IPA [] and in nasal environments some-
times as [n]. The vowel /ü/ is pronounced as IPA [y].
The possessive and benefactive are complementary constructions in Aikanã.

Both involve the suffix -z. As illustrated in § 3.1.1 below, -z marks the
beneficiary object in Aikanã verb phrases. Inside noun phrases it marks a
possessive relationship. The following example illustrates its possessive use:
(45) 'masio-z ha'ne kãwã-'

Marcio-BER water that’s.it-DEC
‘It is Marcio’s river.’
The suffix -z can also be recognised in the possessive pronominal forms in

Table 1, where it is a fossilised element in the second person and transparent
in the third person and first person plural. In possessive constructions, the
possessive pronouns do not receive an extra beneficiary suffix -z, whether
they contain the (fossilised) element -z or not:
(46) txü'txü ha'ne kãwã-'

my water it’s-DEC
‘It is my river.’
(47) 'txütxü 'kapedi'ka

my shoulder
‘my shoulders’
(48) 'h
z 'kapedi'ka
your shoulder
‘your shoulders’
Application of -z to the first person singular possessive pronoun is charac-

teristic of children’s speech: txütxü-z ‘mine’ (lit. ‘of my’).
3.1.1. The benefactive and possessive pronouns
Beneficiary objects are marked both by benefactive object inflexion (cf.

Table 1) of the verb, the beneficiary case marker on the relevant noun, and,
optionally, an oblique case marker:
(49) aikanã-'ene-z 'pürü-'ka-pa-we?eje-'

Aikanã-COL-BER work-1S-CL:big-3P.BEN-DEC
‘I worked for the Aikanã.’
(50) ma'new-z-(je) 'pürü-'txa-pa-we-'

Manoel-BER-OBL work-1P-CL:big-3S.BEN-DEC
‘We worked for Manoel.’
When the beneficiary object is a first or second person, possessive pronouns

with obligatory oblique marking (rather than personal pronouns with bene-
ficiary marking) are (optionally) employed:
(51) h
z-'je pürü-'ka-pa-ua-'
your-OBL work-1S-CL:big-2S.BEN-DEC
‘I worked for you.’
(52) ü'?ü-ka-ua- h
z-'je kaw-me-i-'za
keep-1S-2S.BEN-DEC your-OBL eat-2S-NOM-COL
‘I kept for you things for you to eat.’
(53) txütxü-'je pürü-me-'pa-k-

my-OBL work-2S-CL:big-1S.BEN-DEC
‘You worked for me.’
Instead of a benefactive construction involving possessive pronouns, one

might rather expect a more transparent benefactive construction with bene-
ficiary-marked personal pronouns. The fact that the latter is not attested is
an additional indication that the possessive is related to or identical with the
benefactive.
The morpheme -z ‘BENEFICIARY’ might just as well be glossed as
‘GENITIVE’, or, conversely, the possessive pronouns as beneficiary pro-
nouns. An even better descriptive solution may be to use a term that covers
both the concepts possessive and beneficiary. This would be in accordance
with the spirit of Gil’s (2001) ‘macrofunctional’ approach, which seeks to
eliminate artificial ambiguity between Eurocentrically defined functions
that are not actually distinguished from one another in the language. For the
time being a compromise is found in the consistent use of ‘BENEFICIARY’
as a morphological gloss and ‘possessive’ to characterise the pronouns.11
11
It is unclear whether the suffix -z is related either with the Kwaza possessive
suffix -dy- or with the Kwaza beneficiary suffix -du.
3.1.2. Predicative possession and the benefactive
In predicative possessive constructions, the expression of the relationships

may involve both possessive and personal or demonstrative pronouns and
benefactive verb inflexions:
(54) 'hiba 'kari-we-

this that-3S.BEN-DEC
‘This is his.’
(55) 'kari-kjã-

that-1P.BEN-DEC
‘It’s ours.’
Some possessive relations can be expressed in alternative ways. The seman-

tic difference between them is not yet clear:
(56) 'hiba txütxü-ku-

this my-1S.BEN-DEC
‘This is mine.’
(57) 'hiba 'kari-ku-

‘This is mine.’
(58) 'hiba 'h

z-ua-
this your-2S.BEN-DEC
‘This is yours.’
(59) 'hiba 'kari-ua-

‘This is yours.’
In example (56) the possessive pronoun is used as a verb root. The example
could be interpreted literally as: ‘This is mine for me.’. In (57) a personal
pronoun forms the root of the predicate: ‘This is (it) for me.’. There is a
similar difference between (58) and (59). In the following set of examples,
the first expression is less common, since, as the consultant explained, a
wife is a human being:
(60) deti'a 'txütxü-ku-

woman my-1S.BEN-DEC
‘The woman is mine.’
(61) deti'a 'kari-ku-

woman that-1S.BEN-DEC
‘That is my wife.’
A similar distiction is suggested by the translations of the next set of exam-

ples:
(62) wãwã'?
'txütxü-ku-
child my-1S.BEN-DEC
‘The child is mine.’
(63) wãwã'?
'kari-ku-
child that-1S.BEN-DEC
‘That is my son/daughter.’
It could be that the alternative versions of these examples involve animacy

distinctions. Additional field research in the future will hopefully improve
our understanding of the semantic or pragmatic distinction between these
expressions.
3.1.3. The alternative possessive construction in Aikanã
In addition to the canonical possessive constructions, Aikanã also has an

alternative way to express the possessive. This alternative possessive is
similar to the Kwaza alternative possessive discussed in § 2.3. Even the
morpheme involved, -deri ‘third person PR’, is similar both in form and
meaning to the equivalent Kwaza morpheme -tjate. Note the following con-
struction:
(64) kura-de'ri
husband-3.POS
‘her husband’
This alternative possessive is attested very frequently, but again, it only exists
for a third person PR and it does not seem to be determined by features of
alienability or animacy. Among the languages of the Guaporé region, only
Aikanã and Kwaza were found to have such an extra-paradigmatic third

person PR morpheme. The construction could be considered a shared ir-
regularity on the structural level. As was briefly discussed in van der Voort
(2005: 401), this is a possible indicator of a genealogical relationship be-
tween the languages. It is perhaps even more likely that it represents an in-
stance of (ancient) borrowing.
3.2. Possessive constructions in Kanoê
Kanoê (also referred to as Kapixana) is a language isolate whose speakers

used to be traditional neighbours of the Kwaza, Aikanã and Mekens. With
five remaining speakers Kanoê is on the brink of extinction. The language
has been studied and described since the early 1990s by Bacelar (e.g. 2004),
on whose work this section is based. It is a morphologically very complex
language, especially with regard to verbs. It shows some lexical and struc-
tural similarities with the languages of the region, in particular with Kwaza
and Aikanã, as discussed in van der Voort (2005). Kanoê differs from
Kwaza and is similar to Aikanã in that it has different inflexional verb
classes and that it employs both person prefixes and suffixes. Another re-
semblance with Aikanã is that Kanoê has a set of possessive pronouns in
addition to personal pronouns and that it lacks an inclusive-exclusive dis-
tinction. The Kanoê pronouns are listed in Table 2.
Table 2. Kanoê pronouns (from Bacelar 2004: 143, 145)
personal possessive
1S aj ña, jã
2S m
pja
3S oj ojo
1P ajte jato
2P m
te pjato
3P ojte ojoto
As in Aikanã, some of the possessive pronouns in Kanoê are transparent.

They contain the possessive morpheme -o, which in the plural pronouns
seems to have changed or replaced the vowel of the collective morpheme
-te. The PR in a noun phrase may be represented by a possessive pronoun: 12
12
The Kanoê vowel /y/ is pronounced as IPA [].
(65) pja tyj

your house
‘your house’ (Bacelar 2004: 146)
The possessive morpheme -o itself is productively used in full noun phrases,

such as:
(66) nañu-o kyj

wasp-POS sting
‘the wasp’s sting’ (Bacelar 2004: 99)
Just as in Aikanã, PRs that contain the possessive morpheme (-o) represent
morphologically independent nouns and no further marking is required of
either the PR or the PM.
Kanoê does not have an alternative possessive construction like the
Kwaza (cf. §2.3) and Aikanã (cf. §3.1.3) expressions that involve just the
third person singular.
3.2.1. Kanoê classifiers in possessive constructions
Similar to Kwaza (as in examples (23), (26), (27) and (42)), Kanoê has a
large subset of classifiers that can be turned into full nouns by the semanti-
cally empty noun formative root i-. The empty root occurs mainly with ‘in-
alienable’ items such as body part classifiers. In possessive constructions,
these classifiers build equivalent alternative expressions. Either the classi-
fier combined with the empty root functions as an independent PM, as in
example (67), or the empty root is omitted and the classifier is attached as a
bound morpheme to the PR, as in example (68).
(67) ytse-o i-katsi

tree-POS Ø-CL:root
‘root of tree’ (Bacelar 2004: 100)
(68) ytse-o-katsi
tree-POS-CL:root
‘root of tree’ (Bacelar 2004: 100)
This set of alternative expressions is strikingly similar to the set of alterna-

tive espressions in Kwaza, illustrated by examples (23) and (24). An impor-
tant difference from Kwaza, however, is that in Kanoê the classifier does not
attach to an incomplete possessive stem, replacing a nominaliser. Instead, it

is affixed to a complete PR form, which does not require a classifier. In
fact, the Kanoê construction looks rather like a compound, since a third al-
ternative is the following:
(69) ytse-katsi
tree-CL:root
‘tree root’ (Bacelar 2004: 100)
This form resembles Kwaza noun-classifier combinations, such as tsu'ty-si'ki

‘hat’ in examples (36) and (37), which, however, are not very productive.
It may be that the subset of classifiers which combine with the empty
root should instead be analysed as bound noun roots and the empty root i- as
a noun-formative prefix.
3.2.2. The benefactive and possessive marking
When used in combination with the oblique case suffix -ni, the Kanoê pos-
sessive suffix -o has a beneficiary (70) or comitative (71) sense. This was
attested only with the first and second person singular personal pronouns
and it notably does not involve possessive pronouns:
(70) mi pi-tej-ja-ry-nu mi-tsi piña aj-o-ni

you 2-transport-DR-REFL-FUT 2-INT pineapple I-POS-OBL
‘Will you bring pineapple with you for me?’ (Bacelar 2004: 235)
(71) aj uræ ipæ õ-e-re mi-o-ni

I pig kill 1-DEC-AUX you-POS-OBL
‘I hunt pigs with you.’ (Bacelar 2004: 242)
There are some similarities between the benefactive constructions in Kanoê

(70) and Aikanã (§3.1.1). Both may involve a combination of a possessive
element and an oblique case marker. Then again, they are different in that
in Aikanã a possessive pronoun is used, whereas in Kanoê a more analytic
combination of a possessively marked personal pronoun is used in the
benefactive construction. Furthermore, Aikanã and Kanoê differ also with
respect to beneficiary arguments in general: in Aikanã the PR and the bene-
ficiary argument are marked by the same morpheme -z; in Kanoê benefi-
ciary objects are usually unmarked, just like dative, recipient and goal ob-
jects (Bacelar 2004: 234).
3.3. Possessive constructions in Latundê
Lakondê and Latundê are dialects of a language that belongs to the northern
branch of the small Nambikwara language family. The language, hence-
forth referred to as Latundê, has about 20 speakers (there is only one sur-
viving speaker of the Lakondê dialect). It has been described by Telles
(2002a), on whose work this section is based. Like many languages of the
Guaporé region, Latundê is a morphologically complex language. The pos-
sessive constructions of Latundê, which are also discussed in a separate
article by Telles (2002b), are different from those in the language isolates
discussed above. Latundê has a full paradigm of possessive prefixes, which
are attached to the nominal PM. The prefixes are reduced versions of the
personal pronouns, listed in Table 3 below:
Table 3. Latundê personal pronouns and possessive prefixes (from Telles 2002a:
150, 156 and 2002b: 158)
personal possessive
1S ta ja ta -
1P nh nh-
2 wa ja wa -
3S hãja hãj-, nã-, ã-
3P a a wja a w-, ã-
a
The possibility of distinguishing between the singular and plural of the third per-
son possessive in Latundê does not exist in Lakondê (Telles 2002a: 157).
Independent nouns in isolation in Latundê, including possessed nouns, have

to end in the referential suffix -te or -tu, which produces a citation form:13
(72) wa-sih-te
2.POS-house-REF
‘your house’ (Telles 2002a: 156)
In Latundê, possession can be expressed either in a synthetic manner, em-

ploying a possessive prefix, or in an analytic manner, using a full personal
pronoun.
13
In spontaneous speech the referential suffixes also play a not yet fully understood
role in the marking of pragmatic status (Telles: 2002a: 215).
(73) ta-najn-kin
n-te
1S.POS-head-CL:round-REF
‘my head’ (Telles 2002a: 163)
(74) taja najn-kin

n-te
I head-CL:round-REF
‘my head’ (Telles 2002a: 163)
The semantic value of the alternatives is identical, although the analytic

expression tends to occur more frequently in predicative constructions than
the synthetic one.
Compounding is a productive word-formation process in Latundê. In
Telles (2002a,b), a distiction is made between ‘non-genitive’ and ‘genitive’
compounds, as in the following contrasting examples and their translations:
(75) kejajn-tawn-te
peccary-tail-REF
‘peccary tail’ (a kind of tail) (Telles 2002a: 159)
(76) kejajn-ã-tawn-te
peccary-3.POS-tail-REF
‘the tail of the peccary’ (Telles 2002a: 159)
In the ‘genitive’ compound, the third person PR prefix ã- occurs in inter-

mediate position, coincidentally resembling the Kanoê PR morpheme -o in
example (68). The referential suffix indicates that the preceding roots rep-
resent a single (composite) stem.
Latundê also has a noun-formative empty root
-, that can be used to turn
classifiers into independent nouns (Telles 2002a: 96), and which is similar
to Kwaza e- (see §2.4) and Kanoê i- (see § 3.2.1).
Possessive constructions in other Nambikwara languages, with which
there are certain differences, are discussed in Araujo (2004) and Kroeker
(2001).
3.3.1. Alternative possessive
Again, there are alternative ways to express possession, since there may be
a reason to consider the third person possessive morpheme ã- as a special
case, different from the others. Note that, in addition to full nouns, Latundê
classifiers can also be directly preceded by a possessive prefix.14 In this

type of construction the third person possessive morpheme ã- can be in-
serted between the possessive prefix and the classifier, without any known
consequences for the meaning. Compare the following examples:
(77) ta-ni-tu
1S.POS-CL:hemispheric-REF
‘my pan’ (Telles 2002a: 194)
(78) ta-ã-ni-tu
1S.POS-3.POS-CL:hemispheric-REF
‘my pan’ (Telles 2002a: 194)
The canonical possessive construction in Latundê is of the same type as the

alternative construction in Kwaza and Aikanã. These constructions are
characterised by PR marking on the PM. The difference is that in Latundê
there is a full paradigm, whereas in Kwaza and Aikanã it features only a
third person morpheme that does not belong to a paradigm. However, also
in Latundê the third person possessive behaves differently from the other
possessive morphemes.
3.3.2. A dummy possessum
In predicative possessive constructions, a special semantically empty root

nãn (Telles 2002a: 160) may substitute an overt noun if its referent is under-
stood from the context:
(79) ta-hu -kah-tãn

1.POS-bow-CL:long-IMPF
‘it was my bow’ (Telles 2002a: 161)
(80) ta-nãn-kah-tãn
1.POS-Ø-CL:long-IMPF
‘it was mine (my bow)’ (Telles 2002a: 161)
14
Although Latundê has a small set of classifiers, they have a similar distribution
and nominalising potential as in Kwaza. Some classifiers are even identical in
form with those in Kwaza, Kanoê and Aikanã, which is probably due to areal
diffusion (van der Voort 2005: 395–397).
The empty root nãn was encountered only in possessive constructions. It

should probably be regarded as verbal rather than nominal, because the
nominal referential suffix -te cannot be attached to it unless a classifier is
inserted:
(81) ta-nãn-saw-te
1.POS-Ø-CL:liquid-REF
‘my liquid (e.g. coffee)’ (Telles 2002a: 162)
(82) *ã-nãn-te
3.POS-Ø-REF
‘his (e.g. hand)’ (Telles 2002a: 162)
3.4. Possessive constructions in Baure
Baure is a language that belongs to the southern branch of the Arawak lan-
guage family. It is spoken by about 60 elderly people in the Bolivian low-
lands. It has been described by Danielsen (2007), on whose work this section
is based. Like most languages of the region, the morphology of Baure is
rather complex and shows polysynthetic traits. It has several characteristics
in common with other non-Arawak languages, such as classifiers and direc-
tionals. There is no inclusive-exclusive distinction in the person reference
system. There is a gender distinction for the third person singular. The sys-
tem is based on person proclitics that are attached both to verbs and nouns.
When attached to verbs, the proclitics refer to a subject argument, when
attached to nouns they refer to a PR. In addition, there is a possessive suffix
-no that is homophonous with a nominaliser.
As Danielsen (2007: 119) explains, two different kinds of nouns are dis-
tinguished with regard to the expression of possession. Optionally possessed
nouns receive a person proclitic and usually the derivational possessive suf-
fix -no, as shown by the following examples:
(83) yaki
fire
‘fire’ (Danielsen 2007: 87)
(84) ni=yaki-no
1S=fire-POS
‘my fire’ (Danielsen 2007: 87)
Obligatorily possessed nouns receive only a person proclitic:
(85) e-ser
Ø-tooth
‘a tooth of someone’ (Danielsen 2007: 120)
(86) ni=ser
1S=tooth
‘my tooth’ (Danielsen 2007: 120)
The element e- is analysed by Danielsen (2007: 119) as an ‘unspecified PR’

prefix, because its application to obligatorily possessed noun roots is re-
quired when the PR is unknown. Note, however, that it is not fully produc-
tive, and with some roots it occurs lexicalised with a specific unpredictable
meaning.15 A similar element occurs in the other languages discussed
above, where it is often, but not always, attested with body parts, and where
its distribution is not strictly determined by ‘inalienability’ distinctions. Just
as in these languages, the element e- in Baure should perhaps be regarded
as a semantically empty noun-formative root, which is an areal feature of
the region (see also § 2.4). In many respects, the bound noun roots behave
like other classifiers in Baure (Danielsen 2007: 142 ff.), and similar to the
body part classifiers in Kwaza (§ 2.2.2) and Kanoê (§ 3.2.1).
In addition to the above two kinds of nouns, there is a set of nouns that
cannot be possessed, which includes astronomical bodies and animals.16
Danielsen (2007: 124) nevertheless points out that it is possible to express
possession in the case of domestic animals. This involves a construction
where the PM is preceded by an ‘inalienably possessed noun’:
(87) kove’
dog
‘dog’ (Danielsen 2007: 124)
(88) ni=per kove’

1S=domestic.animal dog
‘my dog’ (Danielsen 2007: 124)
15
Danielsen (2007: 120) mentions the existence of several lexicalised combinations,
such as -waki ‘palm of the hand’, ewaki ‘forked branch of tree’.
16
Consequently, these nouns are not attested with person proclitics, except in certain
lexicalised combinations.
As a bound noun root, the generic noun per ‘domestic animal’ can function
as a classifier. Consequently, the above construction is very similar to the
possessive classifier constructions discussed for Gran Chaco languages by
Fabre (2007) and Messineo and Gerzenstein (2007) and the ‘indirect’ pos-
sessive construction described by Lichtenberk (this volume) for Oceanic
languages. The generic noun per also occurs in other constructions than the
above.
3.4.1. Expression of the possessor
Possessive NPs may consist of an inflected PM and a PR:17
(89) te ro=wer18 to kotis

DEM.M 3SM=house ART lizard
‘the house of the lizard’ (Danielsen 2007: 125)
In order to form personal pronouns, person proclitics are attached to the

element -ti’, which is etymologically related to the feminine demonstrative
root ti. Possessive pronouns are subsequently created by the attachment of
an additional suffix -ro (or its phonotactically determined variant -r), which
may be related to the possessive suffix -no (Danielsen pers.comm.). Hence,
the structure of the first person possessive pronoun, for example, is ni-ti-ro
‘my’. The segmentation of possessive pronouns is ignored in the examples:
(90) ritiro-wo na’

her-COP egg
‘She has got eggs.’ (lit. ‘The eggs are hers.’) (Danielsen 2007: 203)
Note that possessive NPs may also involve possessive pronouns, albeit
rarely, as shown in example (91).
(91) ritir na’

her egg
‘her eggs’ (Danielsen 2007: 203)
17
Baure // represents IPA []. The glide /y/ is pronounced as IPA [j].
18
The root -wer ‘house’ belongs to a small set of nouns that cannot undergo deriva-
tion. Therefore, the possessive suffix -no cannot be applied here, even though the
referent might be regarded as alienable.
3.4.2. Possessive pronouns in benefactive constructions
It is noteworthy that possessive pronouns are used in analytic renderings of

benefactive expressions. The synthetic benefactive construction involves a
benefactive morpheme on the verb and extra person inflexion for the bene-
ficiary argument, as in (92). The analytic benefactive construction is not
realised through verbal morphology but by an extra personal pronoun pre-
ceded by a possessive pronoun, both referring to the beneficiary argument,
as in (93).
(92) ni=ay-ino-wo=pi=ro
1S =desire-BEN-COP =2S =3SM
‘I wish it for you.’ (Danielsen 2007: 16)
(93) ni=aya-wo=ro pitir piti’

1S =desire-COP =3SM your you
‘I wish it for you.’ (Danielsen 2007: 16)
According to Danielsen (2007: 16, 320), the analytic benefactive construc-

tion may be a recent development caused by language obsolescence. She
notes that the speakers of Baure do not regard the construction as the result
of Spanish influence.
3.5. Possessive constructions in Wari’
Wari’ (also known as Pakaa Nova) is a language that belongs to the small
Chapacura language family of western Brazil and northern Bolivia. It has
close to 2000 speakers and it is described in Everett and Kern (1997), on
whose work this section is based. Wari’ has a grammaticalised gender sys-
tem, like certain other languages of the region. The language is syntacti-
cally rather than morphologically complex. Person marking is realised
through particles that Everett and Kern refer to as nominal and verbal in-
flexional clitics (NICs and VICs). Inclusive and exclusive first person plural
and third person feminine and masculine are distinguished. According to
Everett and Kern (1997: 311) there are no possessive pronouns. The prag-
matically unmarked constituent order of Wari’ is VOS.
Possessive constructions in Wari’ involve PR-marking on the PM, like
in Latundê. The possessed noun agrees through an inflexional clitic with
person, number and third person gender of the PR. There are three different
types of possessive marking, corresponding to different classes of nouns;

(a) inalienably possessed nouns are inflected by attachment of special pos-
sessive suffixes;19 (b) possessable nouns are inflected by NICs, i.e. particles
that consist of the element ne- combined with variants of the possessive
suffixes; (c) inalienably possessed kinship terms are inflected by VICs. The
following examples show the respective different possessive expressions:20
(94) xiri-con Xijam

house-3SM.POS Xijam
‘Xijam’s house’ (Everett and Kern 1997: 150)
(95) pije’ nequem Hatem

child POS.3SF Hatem
‘Hatem’s child’ (Everett and Kern 1997: 150)
(96) co-xa’ ca Xijam

INFL-younger.sibling.my 3SM Xijam
‘Xijam’s younger sibling’ (Everett and Kern 1997: 376)
The third construction is limited to a specific set of kinship terms that re-
quire the verbal ‘inflection’ prefix co-. The kinship term always has a first
person singular possessive interpretation. It is followed by a VIC because,
as explained by Everett and Kern (1997: 375–377), it has undergone zero-
derivation from an original verbal clause. The literal interpretation of (96)
is therefore quotative: ‘Xijam he (says) “my younger sibling”’. The con-
struction is grammaticalised and not transparent to the speakers of Wari’.
Also in other parts of Wari’ grammar, the quotative is used in other func-
tions than to quote direct speech, such as future tense.
Note that in all constructions an independent noun referring to the PR
can be omitted if understood from context, as in e.g. xiri-con ‘his house’
(cf. example 94), and in:
19
Wari’ has two noun classes. The class of inalienable nouns is characterised by
the fact that the citation form ends in -xi ‘1P.INCL’. The class of possessable nouns
does not have this characteristic (Everett and Kern 1997: 3). From the examples
below it appears that the noun ‘house’ belongs to the inalienable class, whereas
the noun ‘child’ belongs to the possessable class.
20
Wari’ /x/ represents IPA [(t)∫].
(97) wina-hu’
head-2P.POS
‘your head’ (Everett and Kern 1997: 236)
In the absence of adjectives in Wari’, attributive concepts are often expressed

by nouns in possessive constructions:
(98) wijima-in xirim

smallness-3N.POS house
‘small house’ (lit. ‘the house’s smallness’)
(Everett and Kern 1997: 152)
3.6. Possessive constructions in Mekens (Sakurabiat)
The language of the Mekens (also known as Sakurabiat) belongs to the

Tuparí branch of the Tupí language family. It has about 25 speakers. The
language is described in Galucio (2001), on which work this section is
based. Mekens is not a morphologically highly complex language, although
in general, verbs are more complex than nouns. Mekens morphology in-
volves inflexional argument prefixes and derivational valency changing
prefixes, whereas for tense, mood, aspect and category change, suffixes are
employed. Inclusive, exclusive first person plural and coreferential third
persons are distinguished both in the pronominal and person marking sys-
tems. There are personal and reflexive pronouns, but no possessive pro-
nouns.
No special possessive morphemes are used in Mekens, and possessive
full noun phrases consist of juxtaposed nouns. Anaphorically expressed
PRs are indicated by inflexional person prefixes on the PM. A PR can never
be expressed by a full pronoun. Somewhat similar to Baure and Wari’, pos-
sessive constructions are organised in different ways, depending on the
(language-specific) alienability of the PM. In Mekens, inalienable nouns
and kinship terms are always possessed, which is indicated by a person pre-
fix or by a preposed noun:
(99) e-pisa
2S-liver
‘your liver’ (Galucio 2001: 32)
(100) sakrap okway

spider.monkey tail
‘spider monkey’s tail’ (Galucio 2001: 32)
If there is no overt PR noun, third person PRs are expressed by the prefix i-
before consonants and by the prefix s- before vowels:
(101) i-piso
3S-foot
‘his foot’ (Galucio 2001: 35)
(102) s-anp
3S-head
‘his head’ (Galucio 2001: 36)
When alienable nouns occur possessed, this is indicated in the same way as
the inalienable situation, except for a subclass of nouns which require pre-
fixation of a t.21 The word ek ‘house’ belongs to this subclass:
(103) o-tek
1S-house
‘my house’ (Galucio 2001: 33)
(104) o-top tek

1S-father house
‘my father’s house’ (Galucio 2001: 33)
A third set of nouns does not accept person prefixes, but needs to be juxta-
posed to an inflected obligatorily possessed (inalienable) noun of general
semantic content that functions as a classifier. The word apara ‘banana’
belongs to this subclass:
(105) o-iko apara

1S-food banana
‘my banana’ (Galucio 2001: 33)
21
This t is not an affix. As Galucio (2001: 32) observes, its occurrence is neither
phonologically nor morphologically predictable, and its prefixation to certain
nouns under possession represents a regularity in the lexicon.
(106) *o-apara
1S-banana
‘my banana’ (Galucio 2001: 33)
The latter construction reminds one of the ‘indirect’ possessive construction

described by Lichtenberk (this volume) for Oceanic languages. The differ-
ence is that the Oceanic construction involves a special possessive classifier,
whereas in Mekens a general classifying noun with a wider functional and
distributional domain is used. The construction is also very similar to the
one described for Baure in §3.4.
In Mekens, a reflexive PR, i.e. a PR that is coreferent with the subject of
the sentence, is expressed for the third person singular and plural, by distinct
person markers. The distinction is a consequence of the switch-reference
marking system of Mekens. The following examples contrast two third per-
son singular forms:
(107) i-tek
3S-house
‘his house’ (Galucio 2001: 76)
(108) se-tek
3S.C-house
‘his (own) house’ (Galucio 2001: 76)
3.7. Possessive constructions in Arikapú
Arikapú is a language that belongs to the small Jabutí branch of the Macro-
Jê language family. It is spoken by only one elderly person. The language
has been studied by the author since 2001, and it was recently classified in
cooperation with Eduardo Ribeiro (Ribeiro and van der Voort 2010). Its
traditional neighbouring languages belong mainly to the Tuparí family. Like
the Chapacura and Tupí languages, Arikapú is morphologically a rather
simple language although there are some nominal and verbal suffixes. Per-
son marking is realised by agreement prefixes, that can be used in addition
to personal pronouns. The distribution of person marking obeys an ergative
pattern.
Arguments are indicated through the application of person prefixes on
the verb. Possession is indicated by the same prefixes, when applied to PM
nouns, referring to the PR:
(109) a-kan
2-hat
‘your hat’
(110) i-n
'r
1S-house
‘my house’
Juxtaposed nouns and (rare) pronoun-noun combinations have a possessive

interpretation. The constituent order is PR–PM:
(111) 'nw 'kaj

tapir penis
‘tapir’s penis’
(112) i'h pr 'm

I plate
‘(It is) my plate.’
Some examples suggest that a PR that is coreferential with the subject of a

sentence can be distinguished by the third person marker ta-. This person
marker is otherwise applied only to intransitive verbs:
(113) ta-n
'r-n ta-n't
3-place-LOC 3-sleep
‘He’s sleeping in his own house.’
(114) paku' ta-'txi ã'rã-wiro

woman 3-mother see-FUT
‘The woman went to see her mother.’
The Arikapú coreferential feature does not seem to be part of a switch-

reference marking system as in Mekens. Arikapú shares the coreferential
feature with other Macro-Jê languages (Ribeiro and van der Voort in press,
Rodrigues 1986: 55).
3.7.1. Impersonal reference in possessive constructions
The prefixes for first person singular (i-) and first person plural (txi-)22 are
also used as third person or impersonal prefixes, especially in citation forms.
These two prefixes seem to function often as a dummy PR with inalienable
concepts as well as a dummy argument with verbs. When the PR or the
third person object is expressed by a full noun, the impersonal prefixes are
absent. The following examples contrast these constructions:
(115) i-'k
1S-skin
‘(its) paper’, ‘my skin’
(116) patxi='k
tobacco=skin
‘cigarette paper’
(117) 'txi-ku'ju
1P-wing
‘our/one’s/its wing’
(118) 'aro='kuju
guan=wing
‘wing of a guan bird’
Example (115) and (117) show that the absence of a (pro-)noun may allow for
grammatical ambiguity, since in principle, various interpretations of the per-
son prefixes are possible.
Alternatively, examples (116) and (118) could perhaps be analysed as at-
tributive compounds rather than as possessive constructions, or maybe even
as classifying derivations, since the forms k ‘skin’ and kuju ‘wing’ were
never attested as free nouns.
Unlike many other Macro-Jê languages (Rodrigues 1999: 191) and unlike
Baure (§ 3.4) and Mekens (§ 3.6), Arikapú does not have generic classifying
nouns to distinguish between alienable and inalienable possession.
22
Arikapú /x/ represents IPA [].
4. Final remarks
The languages discussed in this article form only a small subset of the entire
number of languages of the Southwestern Amazon, and they mainly repre-
sent the language diversity on the Brazilian side of the Guaporé River.
Kwaza is taken as a standard of comparison and is presented in more depth
than the other languages, because of the sources, data and experiences
available to me. The purpose of this article is to provide a first impression
of the different structural possibilities for the expression of adnominal pos-
session in the Southwestern Amazon region, and to explore possible areal
linguistic relationships.
Two main types of adnominal possessive strategies were encountered,
both expressed in a range of different varieties. Type I is characterised by
the occurrence of a general possessive element that is attached to the PR. It
forms the standard possessive construction in Aikanã, Kanoê and Kwaza. It
could be considered as a ‘genitive’ construction, but I have not used this
term in order to avoid possible association with case marking, which is dif-
ferent in these languages. Type II is characterised by the attachment of PR
agreement morphemes or person markers to the PM. This is the standard
possessive construction in Arikapú, Baure, Latundê, Mekens and Wari’. An
overview of the different characteristics of possessive expressions is pre-
sented in the Table 4:
Table 4. Possessive features in the Southwestern Amazon

Latundê
Arikapú
Mekens
Aikanã
Kwaza
Kanoê
Baure
Wari’
a possessive morpheme (I) (+) + + +

b PR agreement (II) (+) (+) + +
c person agreement (II) + + +
d possessive pronoun + + +
e personal pronoun + +
f juxtaposed nouns + +
g PR-PM order + + + + + + +
h (in)alienability + + (+) (+) +
i PM reference by classifier + + +
j possessive classifier + +
k benefactive effects + + +
l coreferential possessive + +
Some observations should be made about these characteristics. With regard

to characteristics (a), (b) and (c), both Aikanã and Kwaza have an additional
type II construction. This alternative possessive has limited applicability in
that it only agrees with the third person singular. Note that in Latundê the
canonical possessive is of type II, although the third person singular pos-
sessive behaves in a special way too. Furthermore there is partial overlap in
Baure because a possessive morpheme can be attached to the PM in addition
to PR agreement.
In Table 4 a distinction is made between two varieties of PR marking on
the PM (i.e. type II): characteristic (b) concerns agreement marking that is
used exclusively in possessive constructions, whereas characteristic (c) con-
cerns person marking that is also used in other constructions, such as verbal
cross-reference.
With regard to (e), both Arikapú and Latundê have an additional third
type of possessive, in which juxtaposition of a personal pronoun and a noun
can have a PR-PM interpretation. In Arikapú this is correlated with the pos-
sibility of juxtaposing nouns with a possessive interpretation, as mentioned
in (f). Only Aikanã, Baure and Kanoê have a special set of possessive pro-
nouns that differs from their personal pronouns, as mentioned in (d).
Wari’ is the only exception to the predominant constituent order PR–
PM, mentioned in (g). Its constituent order is PM–PR.
With regard to (h), half of the languages display (in)alienability effects.
However, only Baure, Mekens and Wari’ make a real distinction between
obligatorily possessed (in the sources often called ‘inalienable’) and non-
obligatorily possessed (often ‘alienable’) nouns. In those languages, the in-
alienable possessive constructions are less complex than the alienable, as is
typologically expected (e.g. Chappell and McGregor 1995: 4, Heine 1997:
172ff., Nichols 1992: 117). Arikapú and Kanoê do not make a clear distinc-
tion in the sense that it is not expressed through different possessive con-
structions.
Characteristic (i) concerns several languages in which the PM can be
expressed by a classifier.
Characteristic (j) is about a specific construction of limited applicability
in Baure and Mekens, in which a PR can be expressed by an inflected clas-
sifying noun that is juxtaposed to the PM noun. The construction strongly
resembles the possessive classifier construction described for Austronesian
languages by Lichtenberk (this volume). Elsewhere in South America, the
possessive classifier construction was attested in the languages of the Gran
Chaco region of Argentine, Bolivia and Paraguay. It is discussed by Fabre
(2007), who considers it an areal feature. Interestingly, he includes Chiqui-
tano (possibly a Macro-Jê language) in his discussion. Chiquitano not being

remote from Baure, the areal feature discussed by Fabre may have spread
all the way into Rondônia.
Characteristic (k) concerns three languages of the region where there is
a relationship between possessive and benefactive expressions: Aikanã,
Baure and Kanoê. Although the constructions differ widely between these
languages, their existence might be the result of areal diffusion. Also the
close formal resemblance between the possessive and beneficiary suffixes
in Kwaza is evident, as is the fact that the genitive and dative in Cavineña
are expressed by the same morpheme (Guillaume 2008: 521). Universal
tendencies of grammaticalisation may have reinforced a relationship be-
tween possessive and benefactive expressions. Heine and Kuteva (2002:
55) list several languages (Modern Arabic, Baka of Cameroon, Nigerian
Pidgin English, Tayo Creole French of New Caledonia) in which a benefac-
tive preposition has developed into an adnominal possessive marker. The
significance of the parallel distribution of feature (d) concerning the exis-
tence of dedicated possessive pronouns is unclear.
Characteristic (l) is about the grammaticalised distinction of a PR that is
coreferent with the subject of the sentence. The Arikapú coreferential feature
is limited to possessive constructions, whereas in Mekens it follows from a
general switch-reference marking faculty of the language. The parallel dis-
tribution of feature (f) concerning the possessive interpretation of juxta-
posed nouns is probably coincidence.
As far as we know, the languages discussed in this article are all gene-
alogically unrelated. Three are language isolates and five represent different
linguistic families (or ‘stocks’). This diversity seems to be reflected in Table
4 above, although the languages are too few to draw firm conclusions.
Nevertheless, Aikanã, Kanoê and Kwaza are clearly similar with regard to
characteristic (a). Since these languages, being isolates, are otherwise very
different, this similarity may be due to areal diffusion. One could further-
more imagine that the alternative possessive construction of Aikanã and
Kwazá in (b) represents an older canonical construction that was replaced.
Also the feature (j) may have been subject to areal diffusion, involving
Baure and Mekens. Feature (k) shared by Aikanã, Baure and Kanoê could
reflect a universal tendency, although it could also be the result of areal dif-
fusion, possibly reinforced by universal tendencies.
Possession in the other languages of the region deserves to be investi-
gated as well. Several other thorough descriptions of previously under-
documented languages, besides the ones cited here, have appeared recently,
such as e.g. van Gijn (2006) on Yurakare, Haude (2006) on Movima and
Sakel (2004) on Mosetén, which all are Bolivian isolates, and Guillaume
(2008) on Cavineña, a Tacanan language. Furthermore, work on other
members of the language families represented here, in particular the Tupí
languages, needs to be considered. The study of this corpus may shed addi-
tional light on the extent of areal diffusion of specific possessive strategies
recurring in this article, or may lead to the discovery of others.
Finally, possession in other parts of South America deserves investiga-
tion. Especially the Amazonian languages still suffer from a lack of atten-
tion in general typological work, whereas their impressive genealogical di-
versity makes them an important testing ground for universal aspects of
grammar. Some of these languages employ rare or unique possessive
strategies, like the abovementioned isolate Movima, in which inalienable
and predicative possession are expressed by reduplication (Haude 2006:
238ff.).
Acknowledgements
The Kwaza, Aikanã and Arikapú data in this article are from my own field
work in Rondônia during extended periods between 1995 and 2009. With-
out the help of my language consultants Mario Kwazá, Luiz Aikanã, Manoel
Aikanã and Nazaré Arikapú this work would not have been possible. I also
want to acknowledge the Netherlands Organization for Scientific Research
(NWO) for having generously financed my study of Kwaza, Aikanã and
Arikapú during many years. In addition, I would like to express my deep
gratitude to the indigenous communities in Rondônia for their hospitality. I
am furthermore very grateful to Willem Doelman for providing an excellent
map. Finally, I owe many thanks to Swintha Danielsen, Mily Crevels, Rik
van Gijn, Olga Krasnoukhova, Bill McGregor, Pieter Muysken, Jan Rijkhoff
and Leon Stassen for their valuable comments. Data from other languages
cited here are from the work by colleagues, who do not necessarily agree
with my analyses, and I apologise for sometimes having strayed from theirs.
All errors are mine.
References
Araujo, Gabriel Antunes de

2004 A grammar of Sabanê, a Nambikwaran language. PhD thesis. Utrecht:
Netherlands Graduate School of Linguistics (LOT). (also available on
http://www.lotpublications.nl/publish/articles/000919/bookpart.pdf)
Bacelar, Laércio
2004 Gramática da língua Kanoê. PhD thesis, Nijmegen: Katholieke Uni-
versiteit Nijmegen. (also available on http://webdoc.ubn.kun.nl/mono/
b/bacelar_l/gramdalik.pdf)
Baron, Irène, Michael Herslund and Finn Sørensen (eds.)
2001 Dimensions of Possession, Typological Studies in Language 47. Am-
sterdam /Philadelphia: John Benjamins.
Benveniste, Émile
1960 “‘Être’ et ‘avoir’ dans leurs fonctions linguistiques”. Bulletin de la
Société de Linguistique, 55 (1): 113–134. [Reprinted in: Problèmes de
linguistique générale, Paris: Gallimard, 1966, 187–207.]
Brown, Keith (ed.)
2006 Encyclopedia of Languages and Linguistics, Second edition, Vol. 9.
Oxford: Elsevier.
1995 Prolegomena to a theory of inalienability. In Hilary Chappell and
Part Terms and the Part-Whole Relation, Empirical Approaches to
Language Typology 14. Berlin / New York: Mouton de Gruyter.
Crevels, Mily and Hein van der Voort
2008 The Guaporé-Mamoré region as a linguistic area. In From Linguistic
Areas to Areal Linguistics, Pieter Muysken (ed.), 151–179. Studies in
Language Companion Series 90. Amsterdam/ Philadelphia: John Ben-
jamins.
Crowley, Terry
1995 Inalienable possession in Paamese grammar. In Hilary Chappell and
Danielsen, Swintha
2007 A Grammar of Baure. Indigenous Languages of Latin America
(ILLA), Vol. 6. Leiden: CNWS Publications.
Dixon, Robert M. W. and Alexandra Y. Aikhenvald
1999 Introduction. In Robert M.W. Dixon and Alexandra Y. Aikhenvald
(eds.), 1–21.
Dixon, Robert M. W. and Alexandra Y. Aikhenvald (eds.)
1999 The Amazonian Languages, Cambridge Language Surveys. Cam-
bridge: Cambridge University Press.
Dryer, Matthew S.
2007 Noun Phrase structure. In Language Typology and Syntactic De-
scription, Volume II: Complex Constructions, Timothy Shopen (ed.),
151–205. Second edition. Cambridge: Cambridge University Press.
Everett, Daniel L. and Barbara Kern
1997 Wari’: The Pacaas Novos Language of Western Brazil, London:
Routledge.
Fabre, Alain
2007 Morfosintaxis de los clasificadores posesivos en las lenguas del Gran
Chaco (Argentina, Bolivia y Paraguay). UniverSOS. Revista de Len-
guas Indígenas y Universos Culturales 4: 67–85.
Galucio, Ana Vilacy
2001 The morphosyntax of Mekens (Tupi). PhD thesis, University of Chi-
cago.
Gijn, Rik van
2006 A grammar of Yurakaré. PhD thesis, Radboud Universiteit Nijmegen.
(Avail. on http://webdoc.ubn.ru.nl/mono/g/gijn_e_van/gramofyu.pdf)
Gil, David
2001 Escaping Eurocentrism: fieldwork as a process of unlearning. In Lin-
guistic Fieldwork, Paul Newman and Martha Ratliff (eds.), 102–132.
Guillaume, Antoine
2008 A Grammar of Cavineña. Mouton Grammar Library 44. Berlin /New
Haude, Katharina
2006 A Grammar of Movima. PhD thesis, Radboud Universiteit Nijmegen.
(Avail. on http://webdoc.ubn.ru.nl/mono/h/ haude_k/gramofmo.pdf)
Heine, Bernd
Cambridge Studies in Linguistics 83. Cambridge: Cambridge Uni-
versity Press.
2002 World Lexicon of Grammaticalization. Cambridge: Cambridge Uni-
versity Press.
2001 Introduction: Dimensions of possession. In Baron et al. (eds.), 1–25.
Hinton, Leanne (ed.)
1993 Aikana Modules: A class report on the fieldnotes of Harvey Carlson.
Typescript, 158 pp. Berkeley: University of California.
2006 Possession, Adnominal. In Keith Brown (ed.), 765–769.
Kroeker, Menno
2001 A descriptive grammar of Nambikuara. International Journal of
American Linguistics, 67 (1): 1–87.
this vol. Attributive possessive constructions in Oceanic.
2001 The verb HAVE in Nyulnyulan languages. In Baron et al. (eds.), 67–98.
Messineo, Cristina and Ana Gerzenstein
2007 La posesión en dos lenguas indígenas del Gran Chaco: toba (guay-
curú) y maká (mataguayo). Línguas Indígenas Americanas (LIA-
MES) 7: 61–79.
Nichols, Johanna
1992 Linguistic Diversity in Space and Time, Chicago: University of Chi-
cago Press.
Payne, Doris L. and Immanuel Barshi (eds)
1999 External Possession. Typological Studies in Language 39. Amster-
dam / Philadelphia: John Benjamins.
Ribeiro, Eduardo and Hein van der Voort
in press Nimuendajú was right: The inclusion of the Jabuti language family
in the Macro-Jê stock. International Journal of American Linguistics
76 (4).
Rijkhoff, Jan.
this vol. On the co-variation between form and function of adnominal posses-
sive modifiers in Dutch and English.
Rodrigues, Aryon Dall’Igna
1986 Línguas brasileiras: Para o conhecimento das línguas indígenas, São
Paulo: Edições Loyola.
Rodrigues, Aryon Dall’Igna
1999 Macro-Jê. In Robert M.W. Dixon and Alexandra Y. Aikhenvald (eds.),
165–206.
Sakel, Jeanette
2004 A Grammar of Mosetén. Mouton Grammar Library 33. Berlin /New
Seiler, Hansjakob
1977 On the semantico-syntactic configuration ‘Possessor of an act’. In
Sprache und Sprachen: Gesammelte Aufsätze, Hansjakob Seiler (ed.),
169–186. München: Wilhelm Fink.
Stassen, Leon
2006 Possession, Predicative. In Keith Brown (ed.), 769 –773.
Stassen, Leon
2009 Predicative Possession. Oxford Studies in Typology and Linguistic
Theory. Oxford: Oxford University Press.
Telles, Stella
2002a Fonologia e gramática Latundê/ Lakondê. PhD thesis, Vrije Univer-
siteit Amsterdam.
Telles, Stella
2002b Construções possessivas em Latundê. In Línguas Indígenas Brasilei-
ras: Fonologia, gramática e história, Ana Suelly Arruda Câmara Cab-
ral and Aryon Dall’Igna Rodrigues (org.), 157–163. Atas do I. Encontro
Internacional do GTLI, tomo II. Belém: Editora Universitária UFPA.
Vasconcelos, Ione
2003 Aspectos fonológicos e morfofonológicos da língua Aikanã. PhD
thesis, Maceió: Universidade Federal de Alagoas.
Voort, Hein van der
2004 A Grammar of Kwaza. Mouton Grammar Library 29. Berlin /New
Voort, Hein van der
2005 Kwaza in a comparative perspective. International Journal of Ameri-
can Linguistics, 71 (4): 365–412.
Voort, Hein van der
2006 Construções atributivas em Kwazá. Boletim do Museu Paraense
Emílio Goeldi (Ciências humanas), 1 (1): 87–104. (Also available on
http://marte.museu-goeldi.br/seer/index.php/boletimhumanas/article/
viewFile/5/30)
Possession in the visual-gestural modality:
How possession is expressed in British Sign Language
Kearsy Cormier and Jordan Fenlon
1. Introduction
All known languages have some way of expressing possession, and signed
languages are no exception. The primary question we pose in this chapter
is: Is possession expressed differently in signed languages due to the use of
the visual-gestural (rather than the aural-oral) modality, or are patterns re-
lating to possession essentially the same for signed and spoken languages?
Our chapter begins with background about British Sign Language (BSL),
followed by an overview of the pronominal system of BSL. We then move
on to look at attributive and predicative possession in BSL and finally an
adjectival predicate of predisposition in BSL closely related to the posses-
sive pronoun. Although this chapter is primarily an overview of possession
in BSL, we will also include observations on other signed languages, such
as American Sign Language (ASL), where applicable.
2. British Sign Language
BSL is the natural signed language used by the deaf community in the
United Kingdom. BSL has its own phonological, morphological and syn-
tactic structure. Phonologically, lexical signs in BSL are made up of four
primary parameters: handshape, movement, location (i.e. place of articula-
tion) and palm/finger orientation. Every sign is specified for each of these
parameters, which are phonologically contrastive (i.e. minimal pairs can be
identified with each parameter).
Clearly, BSL is neither simply an elaborated gestural system nor a manual
code based on English. Nonetheless, because BSL is a minority language
within the United Kingdom, there is strong and constant contact between
BSL and English and thus BSL does borrow elements from English. Having
said that, we will point out instances where there are possessive construc-
tions within signed languages which are known borrowings from the sur-
rounding spoken language.
390 Kearsy Cormier and Jordan Fenlon
The examples in this chapter come from a variety of sources. Some are
from or based on elicited data, some are from broadcast television pro-
grammes, some are from an annotated corpus of BSL stories and fables
(Woll, Sutton-Spence, and Waters 2004), and some are from naturalistic
video-recorded conversation. In all cases, examples are from or based on
signing from both native and non-native signers. To be representative of the
community, it is important to include data from both native and non-native
signers, since only a very small percentage of the British Deaf community,
generally considered to be 5–10%, are native signers born to deaf parents,
and thus non-native signers (deaf with hearing and usually non-signing par-
ents) make up the vast majority of the typical deaf population.1
3. Overview of pronouns and determiners
One of the striking characteristics about signed languages is the way that
signers use the space around them for referential purposes. Signers refer to
a person/object physically present within a discourse situation simply by
pointing to him/her/it. For referents who are not present, the signer estab-
lishes a location in space for the referent by pointing to that location. These
pointing signs acts as pronouns and are glossed here as PRO-1 (first person)
or PRO-non1 (for non-first person) – see Figures 1a and 1b.2 The first person
singular pronoun consists of a point to the signer’s own chest.
1
The proportion of native signers in the UK is generally considered to be ap-
proximately 5–10%, following statistics reported in Australia and the United
States (Deaf Society of New South Wales 1998; Mitchell and Karchmer 2004).
2
Following conventions in the sign language literature, glosses are given for signs
using all caps. Signs which require more than one English word for translation
are glossed with English words separated by hyphens (e.g. ICE-CREAM).
Fingerspelled words are indicated with hyphens in between letters (e.g. J-O-E) –
see also footnote 9. Pronouns are glossed as PRO-1 for first person and PRO-
non1 for non-first person. Possessive pronouns are glossed as POSS-1 for first
person and POSS-non1 for non-first person. For ease of exposition, the dual and
number-incorporated pronouns will be notated with a simple English gloss in-
stead (e.g. TWO-OF-US, THREE-OF-THEM). Within examples, when indexing
different non-first person locations, PRO-2/POSS-2 is used for forms which index
the location of the addressee. Indices such as a, b, c (e.g. PRO-a, POSS-a) are
used to index non-addressed participants. Unless otherwise noted, where indices
occur with nouns (e.g. BOX-a), the noun sign may have been established in that
locus previously in the discourse, or some sign within the example establishes
Possession in the visual-gestural modality 391
Figure 1a. PRO-1 ‘me’ Figure 1b. PRO-non1 ‘him/her/it’
3.1. Person
There has been a fair amount of debate within the sign language literature
regarding the issue of person marking on pronouns in signed languages. In
the early days of sign language research, researchers assumed a three-
person system analogous to those found in spoken languages (Friedman
1975; Klima and Bellugi 1979; Padden 1983, 1990). A three-person system
is problematic, however, because there is no listable set of location values
in the signing space to which a non-first person pronoun may point, for ad-
dressee or non-addressed participants. That is, there is no single location
that may act as a morpheme to indicate second or third person. To address
this issue, some researchers have taken the view that sign language pro-
nouns do not exhibit person marking at all and that locations associated
with pronouns instead act as variables (‘loci’) whose content comes from
discourse (Cormier, Wechsler, and Meier 1999; Lillo-Martin and Klima
1990). More recently, most sign linguists subscribe to a two-person system
as proposed by Meier (1990) for ASL. Such an analysis recognises the ‘lis-
tability problem’ (Rathmann and Mathur 2002) of multiple second / third
person location values while at the same time recognising the special status
the noun at that locus. Where repetitions of movement occur, the number of path
movements is indicated with ‘+’ (e.g. POSS-non1++ has two movement paths).
In English translations of examples, for ease of exposition, different gendered
pronouns (e.g. he vs. she) are used to distinguish different referents, though BSL
pronouns do not mark gender.
of first person, for which there is only one location (the signer’s chest). This
two-person system has been assumed by other researchers for ASL and
other signed languages (Engberg-Pedersen 1993; Farris 1994, 1998; Liddell
2003), including BSL (Cormier 2007), and will be assumed in this chapter
as well.3
3.2. Number
In terms of number marking, BSL has singular, dual, number-incorporated

and plural pronouns, as shown in Table 1.
Table 1. Personal pronouns in BSL
First person Non-first person

Singular PRO-1 (‘I’) PRO-non1 (‘you/he/she/it’)
Dual TWO-OF-US TWO-OF-YOU/THEM
Number-incorporated (3/4/5) 3/4/5-OF-US 3/4/5-OF-YOU/THEM
Plural PRO-1pl (‘we’) PRO-non1pl (‘they’)
The primary reason to distinguish dual pronouns from number-incorporated

pronouns that represent three, four or five referents in BSL is the way these
pronouns index their referents. The dual pronoun behaves as a singular
pronoun by indexing the location associated with each referent. Specifically,
the dual pronoun moves between the two locations associated with each
referent. If one of the indexed locations is the location of the signer’s own
body, then this form may be glossed as TWO-OF-US (shown in Figure 2); if
not then it may be glossed as TWO-OF-YOU (if the location of the addressee
is included) or TWO-OF-THEM. The number-incorporated pronouns (e.g.
THREE-OF-US shown in Figure 3), on the other hand, only index a single
general location for all three, four or five referents; these pronouns have a
circular movement which may or may not encompass the locations associ-
ated with each individual referent.
3
Not all sign language researchers subscribe to a two-person analysis for signed
languages – e.g. Berenz (2002) argues for a three-person system for Brazilian
and American Sign Languages, as do Alibasic, Ciciliani and Wilbur (2006) for
Croatian Sign Language.
Figure 2. TWO-OF-US Figure 3. THREE-OF-US
The non-first person plural pronoun (PRO-non1pl) is similar to the number-

incorporated pronouns in that it only generally indexes the locations of its
referents. This pronoun is a pointing sign which has a sweeping movement
across a span of space associated with the referents.
To index the locations associated with each of more than two referents,
signers use a sequence of singular and/or dual (indexing) pronouns, as in (1).
(1) PRO-a PRO-b PRO-c ALL LIKE CAT

‘That person, that one and that one all like cats.’
The first person plural pronoun PRO-1pl is produced with an index hand-
shape at the signer’s chest but, instead of pointing directly at the signer’s
chest, in the plural form the index finger traces a small circular movement
just in front of the chest. This form is the least indexic of all the pronouns –
i.e. the other pronouns point to the location(s) associated with their refer-
ents. The first person plural pronoun only indexes (points to) the signer’s
chest and does not index the other referents.
3.3. Exclusive pronouns
Cormier (2007) has investigated whether BSL has an inclusive/exclusive

distinction within first person plural pronouns.4 The findings indicated that
within first person plurals in BSL, the citation forms of these pronouns can
be used in either inclusive or exclusive contexts. The citation form for the
first person plural pronoun as shown in Figure 4a is produced at the centre
of the signer’s chest (though the location can change to the right or left of
the signer’s chest to index location of the group, as described above). BSL
also has variants of plurals which are exclusive forms. When a referent is
excluded, the signer displaces the plural pronoun to the right or left side of
his/her chest as shown in Figure 4b. The excluded referent(s) may be the
addressee(s) or any other non-first person referent(s) which is/are salient in
the discourse. The location of the pronoun gives no information about the
perceived location of the excluded referents, only that they are in a location
other than that of the included referents.
Figure 4a. PRO-1pl (citation form) ‘we’ Figure 4b. PRO-1pl (displaced) ‘we’
(excluding someone salient in the
discourse)
4
A similar system of exclusive pronouns has also been identified in ASL – for
more see Cormier (2005, 2007).
Exclusive forms were identified with first person plurals (PRO-1pl) and
with first person number-incorporated plurals (3/4/5-OF-US). Because the
first-person dual pronoun indexes the location of each referent, no gram-
matical exclusive form is proposed for the dual pronoun, only plurals and
number-incorporated forms.
3.4. Word order and pro-drop
Syntactically pronouns in BSL occur in situ in the position of the subject or

the object, as in (2) and (3).5 Morphological case systems have not been
identified for any known signed languages (Alibasic Ciciliani and Wilbur
2006), so pronouns do not vary in form based on whether they occur in sub-
ject or object position.
(2) PRO-1 LIKE ICE-CREAM

‘I like ice-cream.’
(3) BOSS LIKE PRO-1

‘The boss likes me.’
Pro-drop (with either the subject or object pronoun) is common in BSL and
other signed languages, particularly with singular pronouns, when the ref-
erent is retrievable from context, as in (4). With plural pronouns, studies of
ASL and Australian Sign Language (Auslan) have shown that first person
plural pronouns are dropped more often than first person singular pronouns
(Schembri and Johnston 2006; Wulf, Dudis, Bayley, and Lucas 2002). With
double agreement verbs (a type of verb in which the start and end location
of the verb or the direction in which the verb is facing reflects the subject /
source and object/goal of the verb, respectively), pro-drop is usually the rule
rather then the exception, particularly with objects, as in (5).
5
Word order in BSL like other signed languages is quite variable on the surface.
Grammatical processes like topic marking make many different word orders
possible. There have been attempts by sign language researchers to posit a basic
default order from which other orders are derived. Such researchers have
claimed that the basic word order for ASL is SVO (Fischer 1975; Neidle, Kegl,
MacLaughlin, Bahan, and Lee 2000). There has not been as much work on word
order in BSL but it appears that BSL is also, underlyingly at least, SVO, although
see Deuchar (1983) for a different perspective.
(4) LIKE ICE-CREAM

‘I like ice-cream.’
(5) BEEN a-HELP-1

‘He helped me.’
Pronoun copy is also common, whereby the pronoun in situ is then copied
and produced again at the end of the clause as in (6).
(6) PRO-1 LIKE ICE-CREAM PRO-1

‘I like ice cream.’
This section has provided an overview of the pronominal system of BSL,

specifically personal pronouns. This is relevant for considering possessive
pronouns in BSL, to which we turn next.
4. Possession in BSL
As in spoken languages, signed languages exhibit both attributive and

predicative possession. We first look at attributive possession, expressed
primarily by possessive pronouns.
4.1. Attributive possession
The singular and plural possessive pronouns in BSL behave very similarly
to the singular and plural personal pronouns, respectively. The primary dif-
ference is the handshape. While the personal pronoun PRO uses an ‘index’
handshape (extended index finger similar to the handshape used in a point-
ing gesture), the possessive pronoun POSS uses a fist handshape with the
thumb alongside the index finger or wrapped across the other fingers. For
the first person form, the back of the fingers contact the signer’s chest, and
for the non-first person form the back of the fingers face outward away
from the signer, toward the location associated with the referent.
4.1.1. Number
Table 2 shows the BSL possessive pronouns by person and number.

Table 2. Possessive pronouns in BSL
First person Non-first person

Singular or collective plural POSS-1 ‘my’ POSS-non1 ‘your/his/her/its’
Plural POSS-1pl ‘our’ POSS-non1pl ‘their’
The singular forms are shown in Figures 5 and 6. Singular possessive pro-
nouns representing the possessor (PR) may occur on their own with the
possessum (PM) noun (e.g. POSS-a CAR ‘his car’) or in apposition with a
personal pronoun (e.g. PRO-a POSS-a CAR ‘his car’). (See § 4.1.2 for the
use of POSS with nominal PRs.)
Figure 5. POSS-1 ‘my’ Figure 6. POSS-non1 ‘your’
Plural possessive pronouns use the same handshape as the singular posses-
sive forms. The non-first person plural possessive pronoun POSS-non1pl
‘their’ is very similar to the pronoun PRO-non1pl ‘them’ (with a sweeping
motion across the locations associated with the referents) but uses a fist
handshape, as shown in Figure 8. The first person plural possessive pro-
noun POSS-1pl ‘our’ as shown in Figure 7 is the same as the first person
plural pronoun (PRO-1pl ‘we’: circular motion at the chest) but has a fist
handshape; the back of the fingers remain facing the signer during the
production of the sign.
Figure 7. BSL POSS-1pl ‘our’
Figure 8. BSL POSS-non1pl ‘their’
With plurals, signers may use a combination of possessive pronoun(s) with

multiple singular pronouns, in which case the personal pronouns typically
precede the possessive pronoun(s) as in (7).
(7) PRO-a PRO-b PRO-1 POSS-1pl PARTY TONIGHT

‘Our party (the party of him, her and me) will be tonight.’
To specify distribution of possession, signers may use a sequence of singular

possessive pronouns as in (8) – cf. collective possession as in (12a).
(8) POSS-a POSS-b POSS-1 TELEVISON BUY SAME SHOP

‘We (he, she and I) bought our televisions from the same shop.’
Although these plural forms do exist and are used, it is also possible (and
quite common) for BSL signers to use the singular possessive instead of the
plural with plural referents, both in first and non-first person, for collective
plurals, as shown in (9a). If the number of referents is not already known or
not clear in context, then a plural personal pronoun or other quantifier may
accompany the singular possessive; in this case, again, the pronoun/quanti-
fier generally precedes the possessive but can occur afterwards. In (9b) and
(9c), the non-first person possessive is directed towards the same general
location as the quantifier/pronoun which occurs in the same utterance.6 In
(9d) and (9e), the singular first person possessive is, as always, directed
towards the signer’s chest. If a plural pronoun or quantifier is used in com-
bination with a possessive, the data and intuitions from deaf informants in-
dicate that the possessive must be singular and cannot be plural, as shown
in (9f).
(9) a. WHAT ABOUT PARENTS MAKE POSS-non1-emp DECISION7

‘What about parents making their own decision?’
b. THERE PEOPLE ALL-OF-THEM POSS-non1 BELIEF DIFFERENT
‘There, people have different beliefs.’
c. NOW WANT FOCUS POSS-non1 PERSONALITY TWO-OF-YOU
‘Now I want to focus on your (both of your) personalities.’
d. THREE-OF-US POSS-1 HOUSE SELL
‘The house belonging to the three of us is for sale.’
e. TWO-OF-US HOUSE POSS-1 SELL
‘The house belonging to the two of us is for sale.’
f. *THREE-OF-US POSS-1pl HOUSE SELL
‘The house belonging to the three of us is for sale.’
6
With non-first person forms, singular collective possessive pronouns are directed
towards a location in space that appears approximately equidistant between the
multiple locations associated with the referents (if dual or plural), or the same
general location of the pronoun (if a number-incorporated pronoun).
7
This particular example is an emphatic use of the possessive, as described in
§ 4.1.5.
The use of singular possessive pronouns as a way of marking plural posses-

sion in a collective context should not be surprising, as the nature of collec-
tive plurals crosslinguistically is that they act as singulars. However, these
singulars used as collectives are more interesting when considering their
indexicality (i.e. the extent to which they point to or index the locations as-
sociated with their referents). While the personal pronouns in these combi-
nations may be providing indexic information (e.g. the dual and plural forms
which point to more than one location), the collective plural possessive
forms in these constructions have lost much of their indexicality (compared
to truly plural possessives) in that they only index a single location. As
noted above, the first person plural personal pronoun PRO-1pl (as well as the
first person plural possessive pronoun POSS-1pl) are already less indexic
than their singular counterparts. Because the collective plural form is the
same as the singular form with the same degree of indexicality, these col-
lective plurals are in fact not marking number at all but only person (first or
non-first). Thus when a plural personal pronoun occurs which marks num-
ber, as in (9f), an additional plural possessive form is not acceptable.
For distributive readings (e.g. multiple objects that are each possessed
separately by different people, as opposed to the default collective reading
– the single object that is possessed together by more than one person),
signers use multiple possessive pronouns directed to the location of each
referent, either juxtaposed (10a) or with multiple PM nouns (10b).
(10) a. POSS-2 POSS-1 LAND CONTINUE SELL

‘Your land and my land are still for sale.’
b. POSS-2 LAND POSS-a LAND POSS-1 LAND CONTINUE SELL
‘Your land, his land and my land are still for sale.’
Personal pronouns may also be used in addition to the possessives as in (11).

These lists are not necessarily exhaustive, so that (11) may indicate either
precisely three land owners or it may indicate a number larger than three.
(11) POSS-a POSS-b POSS-1 LAND PRO-a PRO-b PRO-1 WILL SELL
‘His land, her land and my land’/‘The land that each of us owns will
be sold.’
4.1.2. Word order and pro-drop
As seen in the examples above, possessive pronouns in BSL occur before

the noun they modify (12a), and optionally may be copied immediately after
the noun as well, as in (12b). Possessive pronouns may also occur only after
the nouns they modify (e.g. TELEVISION POSS-1pl NEW), though the
patterns in (12a) and (12b) appear to be preferred.
(12) a. POSS-1pl TELEVISION NEW

‘Our television is new.’
b. POSS-1pl TELEVISION POSS-1pl NEW
‘Our television is new.’
With a nominal PR, the PR, possessive pronoun and PM are placed in ap-
position as in (13). The possessive pronoun usually follows the PR, such
that the typical order is PR POSS PM.8 This same ordering is identified for
nominal PRs in ASL, Croatian Sign Language, and Austrian Sign Lan-
guage (Pichler et al. 2008).
(13) JOHN POSS-non1 CAR NICE

‘John’s car is nice’
Similarly to pro-drop with personal pronouns, possessive pronouns are not

required if the PR can be inferred from context as shown in (14).
(14) NAME WHAT?

‘What is your/his/her name?’
8
It is possible for the possessive pronoun to precede the possessor in both BSL
and ASL (e.g., POSS-non1 JOHN CAR NICE ‘John’s car is nice’). MacLaughlin
(1997) shows that this construction (POSS PR PM) can only be used in ASL
when the possessor in question has been established previously (i.e. it cannot be
used with indefinite possessors) and that this construction is strongly preferred in
topic position. According to Neidle, Kegl, MacLaughlin, Bahan and Lee (2000),
“the proper analysis of this construction remains something of a mystery” (p.
182). We only identified one token of this ordering in our corpus of BSL data;
by far the most frequent ordering was PR POSS PM.
4.1.3. Abstract PMs
Attributive possessive pronouns can be used with either concrete entities as

in many of the examples above, or with abstract concepts, such as ‘time’,
‘view’, or ‘fault’, as in (15) below.
(15) a. COMPARE POSS-1 TIME NOTHING

‘Compared to my time there was nothing.’
b. POSS-1 VIEW
‘That’s my view.’
c. BUT POSS-non1 FAULT
‘But it was her fault.’
4.1.4. Exclusive possessive pronouns
First person plural possessive pronouns in BSL also have an exclusive form,
similar to the exclusive personal pronouns described in § 3.3. The exclusive
form for BSL shown in Figure 9 below is the same as the citation form as
in Figure 7 above but is displaced to the signer’s left or right side, and this
form may exclude any salient referent.
Figure 9. POSS-1pl (displaced) ‘our’

(belonging to us, excluding someone salient in the discourse)
4.1.5. Emphatic possessive pronouns
Possessive pronouns may be made emphatic by producing the possessive

with a sharp sweeping movement towards the location associated with the
referent, with either one or two hands, as shown in Figures 10a and 10b and
in (16).
Figure 10a. POSS-non1-emp (one hand)
Figure 10b. POSS-non1-emp (two hands)

(16) a. BUY CAR YESTERDAY. NICE POSS-1-emp CAR. FREE DRIVE-

ABOUT.
‘I bought a car yesterday. It’s nice to have my own car. I’m free to
drive anywhere.’
b. WHAT ABOUT PARENTS MAKE POSS-non1-emp DECISION
‘What about parents making their own decision?’
c. POSS-non1-emp BUSINESS MEDIA, WOW
‘Your own business in media – wow!’
d. PRO-1 POSS-1-emp ROOM
‘I had my own room.’
4.1.6. Morphological marking of possession
Possession in signed languages that have been documented to date is gener-

ally marked using possessive pronouns rather than morphologically. BSL
does not have an equivalent to the English -’s possessive marker in wide-
spread usage today. However, it is reported that older BSL signers used (and
some still use) a form which consists of a modified form of the finger-
spelled letter -S- cliticised to the end of the noun PR. Auslan, a language
historically related to BSL, has retained this form in both older and younger
signers, shown below in Figure 11 (Branson, Toms, Bernal, and Miller
1995; Johnston and Schembri 2007).9
9
Fingerspelling systems within signed languages are based on the written alphabet
of the surrounding hearing community and are used for various purposes, in-
cluding proper names and other concepts for which a native lexical sign may not
exist. Auslan is historically related to BSL; Auslan, BSL and New Zealand Sign
Language are generally considered to be dialects of the same language (Johnston
2003; McKee and Kennedy 2000), and all three languages share a common two-
handed fingerspelling system.
Figure 11. Auslan MOTHER’S (MOTHER + ’S)

(Reprinted with permission from T. Johnston & A. Schembri (2007).
Australian Sign Language: An introduction to sign language linguistics.
Cambridge: Cambridge University Press.)
ASL also has a similar affixal ’S possessive marker. Pichler et al. (2008)
note than although this form was likely borrowed from Signed English sys-
tems (sign systems which are not natural signed languages but are based on
English morphology and syntax), it is judged acceptable by ASL signers,
particularly in utterances with multiple possessors, as in FATHER ’S
BROTHER ’S WIFE. This marker (glossed by Pichler et al. as APOS-
TROPHE-S) consists of a modified form of the letter -S- from the finger-
spelling system used by ASL signers.
Unlike English, these affixal possessive markers are not obligatory for
expressing possession with nominal PRs in Auslan and ASL, and are used
along with other types of possession marking as described in this chapter.
4.1.7. Inalienable possession
It has been suggested that BSL systematically encodes inalienable posses-

sion (Fenlon and Cormier 2006; Sutton-Spence and Woll 1999). Inalienable
possession refers to nouns that are inherently linked to the PR (such as kin-
ship terms and body-part nouns) as opposed to alienable nouns which are
separable from the PR (e.g., a book). This distinction is encoded in BSL by

using a form identical to the personal pronoun PRO in place of the posses-
sive pronoun POSS with inalienable nouns; see (17).
(17) a. PRO-1 MOTHER

‘my mother’
b. PRO-1 LEG
‘my leg’
c. POSS-non1 BOOK
‘your book’
Inalienable possession in spoken languages is strongly associated with two

semantic categories: kinship terms and body-part nouns. However, this fea-
ture is not exclusive to these categories and has been shown in some lan-
guages to extend to personal possessions, clothing and physical and mental
characteristics (Cooper 2002). In keeping with this pattern, BSL uses a form
identical to the personal pronoun PRO in place of the possessive pronoun
POSS with kinship terms and body part nouns as well as NAME as in (18a)
and other personal possessions such as HOUSE as in (18b).
(18) a. PRO-1 NAME J-O-E

‘My name is Joe.’
b. FRIEND STAY PRO-1 HOUSE PRO-1
‘Friends stay at my house.’
However, the issue of whether BSL encodes inalienable possession gram-

matically can be called into question. Firstly, although personal pronouns
are used with inalienable phenomena such as body-part nouns, kinship terms
and NAME, it appears that signers show a preference for the possessive
pronoun POSS (rather than the personal pronoun PRO) with kinship terms.
Secondly, it is unclear in examples such as (18a) whether NAME can be
classed as a noun or a verb. If it is a verb, then the PRO is functioning as a
personal pronoun (e.g. I am called Joe) as opposed to a possessive pronoun
(e.g. My name is Joe).
Despite problems with claiming that BSL grammaticises inalienability,
patterns of inalienability in BSL are consistent with patterns across spoken
languages. Specifically, BSL fits the crosslinguistic pattern in which kinship
terms and body-part nouns are inalienable in languages that have the dis-
tinction. It has also been shown crosslinguistically that the existence of
other semantic categories marked as inalienable in a given language (beyond

kinship and body part terms) implies that body parts and/or kinship terms
will also be marked as inalienable (Kliffer 1996; Nichols 1988, 1992); this
generalisation fits BSL as well. Finally, in some spoken languages (such as
Navajo) the alienable/inalienable distinction is optional as it appears to be
in BSL as well (Nichols 1992).
In addition, our BSL data indicates that it is possible to express inalien-
able possession with overt nominal PRs through the juxtaposition of the PR
and PM, as in (19). Similar observations have been made for ASL, Croatian
Sign Language, and Austrian Sign Language in Pichler et al. (2008) where
the construction PR PM is strongly preferred over PR POSS PM for inalien-
able constructions.
(19) a. BABY EAR

‘the baby’s ear’
b. WOMAN SON
‘the woman’s son’
To our knowledge, the use of two distinct forms marking possession (where
one encodes inalienable possession and the other marks alienable posses-
sion, as in (17) above) has not been identified in any other signed language
in a possessive construction without an overt nominal PR. Use of the per-
sonal pronoun instead of the possessive pronoun has been noted for some
other signed languages – specifically, ASL, Croatian Sign Language, and
Austrian Sign Language – but the claim with these signed languages is that
the personal pronoun PRO may optionally be used instead of POSS in any
possessive construction, not just with inalienable nouns (Pichler et al. 2008).
4.1.8. Spatial marking
In addition to the use of possessive pronouns, BSL can also associate PMs
with PRs by changing the location of some noun signs directly within the
signing space. The primary criterion for such spatial marking of noun signs
is that the place of articulation of the sign must be the neutral space in front
of the signer, not a location on the body; the sign may then be located at a
particular location in space associated with some referent. It is not entirely
clear if this spatially marks PMs only for possession (‘my house’, ‘your
house’) or if the marking is at particular locations which just so happen to
correspond to locations associated with PRs (‘the house here/associated
with me’, ‘the house there/associated with you’). In (20a) and (20b), the
sign HOUSE is produced at two different locations. Note the first person
pronoun PRO-1 is used as a possessive in (20a), as in (18b) above, and thus
the spatial marking of HOUSE at the first person locus is redundant, while
in (20b) both tokens of the pronoun PRO-1 refer to the subject of the clause,
not to the PR, so the only indication of the PR here is the non-first person
spatial marking of the noun HOUSE.
(20) a. FRIEND STAY PRO-1 HOUSE PRO-1 …

‘Friends would stay at my house …’
(HOUSE is signed close to the signer chest.)
b. INVITE SAME CHANGE STAY PRO-1 GO HOUSE PRO-1 GO
‘Then they would invite me to go to their house.’
(HOUSE is signed away from the signer chest)
The grammaticisation of locative markers into possession markers is well

established for many spoken languages (Heine 1997), and it may be that this
same process is occurring/has occurred in BSL. Further evidence for the
strong connection between locatives and possession comes from Pichler
and colleagues (2008), who have found that the predicative possessive in
Austrian Sign Language is expressed by using a sign which appears to be
derived from the sign DA meaning ‘here’. We now turn to predicative pos-
session in BSL.
4.2. Predicative possession
There are two primary ways to mark predicative possession in BSL –

HAVE and BELONG. These two types are attested crosslinguistically
amongst spoken languages and as in other languages differ in whether the
PR (as with HAVE) or the PM (as with BELONG) is the subject or topic of
the sentence (Heine 1997; Herslund and Baron 2001).
4.2.1. Predicative expression of possession and existence
The lexical verb HAVE in BSL (shown below in Figure 12) can be used to
indicate possession. The PR, if overtly expressed, acts as subject of the sen-
tence. HAVE (glossed here as HAVEposs) can be used with any type of al-
ienable possession, as in (21).
Figure 12. HAVEposs/ex
(21) a. JOHN HAVEposs CAR

‘John has a car.’
b. BOOK HAVEposs
‘I have a book.’
As with many other languages, the same lexical item HAVE which is used
for possession can also be used for existence (Heine 1997), glossed in (22)
as HAVEex.
(22) a. GOD HAVEex

‘There is a God.’
b. HAVEex GOSSIP
‘There was gossip.’
BSL also has signs for negative possession and negative existence. For ne-
gation of possession, the sign HAVE-NEGposs is phonologically related to
the sign HAVE, with what Brennan (1992) described as a negative affix
occurring on some BSL signs (forearm rotation accompanied by an opening
of the hand; see Figure 13). This sign is used for possession only (23a), not
existence (23b). There are other variants of this sign also meaning ‘not
have’, including one which begins with two 5-hands palm down and ends
with both hands pronated so that the palms are facing upwards.
Figure 13. HAVE-NEGposs
(23) a. CAR HAVE-NEGposs

‘I/someone doesn’t have a car.’
b. *GOD HAVE-NEGposs
‘There is no God.’
The sign NOT-HAVE, a sign formationally unrelated to HAVE (shown in

Figure 14) can be used to negate either possession (examples (24a) and (24b))
or existence ((24b) and (24c)). A more general negation sign NOTHING (as
shown in Figure 15) can also be used predicatively to negate either posses-
sion (25a) or existence (25b), in addition to its (pro)nominal usage (25c).
Figure 14. NOT-HAVEposs/ex

(24) a. PRO-1 CAR NOT-HAVEposs

‘I don’t have a car.’
b. CAR NOT-HAVEposs/ex
‘There is no car.’/‘I/someone doesn’t have a car.’
c. GOD NOT-HAVEex
‘There is no God.’
Figure 15. NOTHING
(25) a. WILLIAM NOTHING HUSBAND

‘William didn’t have a husband.’
b. BISCUIT NOTHING
‘There are no biscuits’
c. OPEN-CUPBOARD BUT NOTHING
‘(I) opened the cupboard but there was nothing there.’
Interestingly, according to Pichler et al. (2008), in ASL, Croatian Sign

Language, and Austrian Sign Language, the lexical sign which is used for
predicative possession is the same as the lexical sign used for existence
(HAVE, IMATI, and DA, respectively), and the same is true for negative
possession and negative existence (NONE, NEMATI, and KEIN (DA), re-
spectively). Kristoffersen (2003) and Arik (2008) note the same for Danish
Sign Language and Turkish Sign Language respectively – i.e. that the same
lexical items are used both for possession and existence. The fact that ex-
pression of possession and existence patterns differently in BSL from these
other signed languages reminds us that, although there may be an underlying
and historical relationship between possession, existence, and locatives

(Freeze 1992), these constructions do surface differently in some languages.
We return to the relationship between possession and locatives in § 6.
4.2.2. BELONG
BSL has another predicative possessive form which foregrounds the PM

rather than the PR. The sign BELONG makes reference to the location of
not only the PR (as possessive pronouns do) but also the PM. This form uses
the same fist handshape as the possessive pronouns. But rather than simply
directing the hand towards the PR referent, BELONG begins at the location
associated with the PM noun (the person or object being possessed, L1 in
Figure 16) and ends with the location associated with the PR (L2 in Figure
16). Thus whether the PM is named before BELONG as in (26a) or after
BELONG as in (26b), in each case, the location of BELONG begins at the
location associated with the PM. It is not acceptable for BELONG to begin
with the location of the PR and end with the location of the PM, as in (26d).
Figure 16. BSL a-BELONG-b
(26) a. BOOK-a a-BELONG-b TEACHER-a

‘That book is the teacher’s.’
b. a-BELONG-2 BEDROOM-a?
‘Is that your bedroom?
c. a-BELONG-1 (where a is the location already associated with a box)
‘That box is mine’
d. *1-BELONG-a
‘That box is mine’
In (26), the referents ‘book’, ‘teacher’, and ‘bedroom’ may have been set
up in space already within the discourse, or they may not have been. If they
have not, the sign BELONG establishes the loci for each. In example (26c),
either the location of the box has already been established in space previ-
ously in the discourse, or the box is physically present during the utterance.
The verb BELONG falls into the category of verbs in BSL known as
agreement verbs, described in § 3.4 above, in which the verb begins at the
location associated with the subject and ends at the location associated with
the object. Agreement verbs exist in all known signed languages. However,
a predicative possession form like BELONG in BSL which acts as an agree-
ment verb has not to our knowledge been identified in any other sign lan-
guage outside of the British Sign Language family.
4.2.3. Possessive pronouns
In addition to attributive possession, possessive pronouns may also be used

predicatively as below in (27a) and (27b). The short, sharp movement of
predicative possessive pronouns towards the PR is often reiterated, possibly
due to its phrase final position toward the location associated with the ref-
erent (cf. Nespor and Sandler 1999). This form may be one-handed or two-
handed, as shown in Figures 17a and 17b.
Figure 17a. POSS-non1++

Figure 17 b. POSS-non1 (2hands)++
(27) a. SAW SOMEONE TAKE CUP. BEEN TOLD POSS-1++

‘I saw someone take my cup. I had told him it was mine.’
b. CAR POSS-non1++
‘The car is yours.’
5. Adjectival predicate of predisposition
The predicative possessive pronoun has also evolved into an adjectival

predicate indicating that the ‘possessor’ has a predisposition to act a certain
way, or that a certain characteristic is typical of the ‘possessor’. The term
‘predispositional’ follows the use of the term predispositional aspect
coined by Klima and Bellugi (1979) to describe aspectual modulation on
some ASL adjectival predicates to indicate a predisposition to a certain
characteristic (e.g. ASL MISCHEVIOUS vs. MISCHIEF-PRONE, where
the sign MISCHIEF-PRONE is the same as MISCHEVIOUS but is two-
handed and has a reduplicated circular path movement instead of only the
hand-internal movement of MISCHEVIOUS). Examples of the BSL adjec-
tival predicate of predisposition BE-TYPICAL-OF, identical in form to the
predicative possessive pronouns shown above in Figure 17, are given in
(28a) and (28b).
(28) a. RABBIT BOUNCE-ALONG LAUGH, SLOW IS-TYPICAL-OF-

non1++ PRO-1 FAST
‘The rabbit bounced along laughing and said “It’s your way to be
slow, I am fast.” ’
b. WHY IS-TYPICAL-OF-non1(2hands)++ ALWAYS FAR STILL
CRAWL
‘Why is it always your way, even though it is far, to still crawl
along?’
BSL also has a compound sign which has combined the adjectival predicate
of predisposition with the sign ALWAYS to resulting in an adverbial form
meaning ‘typically’, glossed in (29) as ALWAYS-TYPICAL and shown in
Figure 18. In ALWAYS-TYPICAL, the repeated movement of the predis-
positional marker is lost. This pattern of phonological reduction of move-
ment in the surface form is common in compound signs in signed languages
(Klima and Bellugi 1979; Sutton-Spence and Woll 1999).
Figure 18. ALWAYS-TYPICAL
(29) JONATHAN ALWAYS-TYPICAL-a LATE

‘Jonathan is always late.’
The predicative possessive pronoun in ASL can also be used predisposi-

tionally, either alone (BE-TYPICAL-OF, formationally identical to POSS++)
or compounded with the verb TEND (TEND-POSS). This form TEND-
POSS, glossed by Klima and Bellugi (1979) as TEND^(HIS), is used with
predicates to indicate predisposition to a certain activity or characteristic.

As with BSL ALWAYS-TYPICAL, the repeated movement of the pre-
dispositional marker BE-TYPICAL-OF in the ASL compound TEND-POSS
is lost.
These predispositional markers in both BSL and ASL are very likely ex-
amples of grammaticisation of a possessive pronoun to an aspectual marker
denoting predisposition. Typologically grammaticisation of possession into
aspect marking in spoken languages is very common (Heine 1997). However,
this grammaticisation path typically occurs with verbal markers of posses-
sion which then become verbal aspectual markers (e.g. English ‘have’). With
BSL (and ASL), what we see instead is pronominal markers of possession
(POSS ‘his/her/its’) becoming adjectival predicates of predisposition (BE-
TYPICAL-OF). The BSL and ASL verbs HAVE do not have any aspectual
function at all and only mark possession or existence (as described in
§4.2.1).
6. Discussion and conclusion
In this chapter we have seen that British Sign Language exhibits many of
the same patterns of possession as spoken languages and in similar ways.
BSL has both attributive and predicative possession. BSL exhibits patterns
consistent with those in spoken languages with inalienable possession. The
lexical item HAVE is used for both possession and existence, which is an-
other pattern seen in many spoken languages. BSL also has a split between
HAVE and BELONG (foregrounded PR vs. foregrounded PM) shared by
many spoken languages.
When we turn to looking at differences across signed languages regard-
ing possession, these are generally the types of differences that we find
across spoken languages – i.e. those features which are subject to parametric
variation. Some languages like Australian Sign Language (Auslan), ASL and
English have a possessive clitic which attaches to nouns; some languages
like Danish (and, it seems, present-day BSL) do not. Some spoken lan-
guages mark inalienability while others do not; there is evidence that BSL
and some other signed languages mark inalienability (optionally at least)
though it seems in different ways.
So it is clear there are many similarities between possession in signed
languages and spoken languages. The differences seem to be strongest when
considering the locative nature of many BSL signs. Across spoken lan-
guages, there is a strong link between possession, existentials and location
(Freeze 1992; Heine 1997; Herslund and Baron 2001).10 Specifically, in many
spoken languages, the PM in a possessive construction is encoded as subject
while the PR is encoded as a locative (e.g. dative) argument. In signed lan-
guages, signs such as pronouns, some nouns, agreement verbs, demonstra-
tives, etc. are inherently locative. As we have seen in this chapter, many
signs denoting (and/or grammaticised from) possession – e.g. possessive pro-
nouns, nouns which can be spatially located, BELONG as an agreement verb,
the predispositional markers BE-TYPICAL-OF and ALWAYS-TYPICAL –
are spatially modified. This use of space is a natural product of the visual-
gestural modality and is perhaps the most striking difference between signed
and spoken languages when it comes to possession (and also pronominal/
nominal reference and agreement in general), and is a good reason to con-
sider signed languages when looking at the marking of possession across
different languages.
Indeed, typological studies which aim to cover phenomena across a
wide variety of languages often neglect to include signed languages. This
oversight is unfortunate because signed languages have much to offer lan-
guage typologists. The features and categories that all known spoken and
signed languages have in common are candidates for universal grammar,
while those categories that differ across signed and spoken languages – par-
ticularly those which are thought to be universal amongst spoken languages –
can help us tease apart those features which are truly inherent to human lan-
guage versus those which may not be universal after all. In this chapter we
have seen that while possession is likely a semantic category that is ex-
pressed in all human languages, language modality certainly does shape the
way in which it is expressed.
Acknowledgements
We would like to acknowledge the Possession in Sign Languages project

headed by Ulrike Zeshan; some of the elicitation materials used to collect
some of the data given in this chapter were created by Zeshan specifically
for that project, originally based at the Max Planck Institute for Psycholin-
guistics. We are also grateful to the participants from whom we collected
data, and to Mark Wheatley, June Fenlon, Helen Earis, and Sandra Smith
for modeling for the example figures in the chapter. We thank Rachel Sutton-
10
Though see Payne (this volume) for arguments against the “possession is loca-
tion” view.
Spence for assistance in collecting, analysing and discussing data presented

in this chapter, Bencie Woll for her input particularly on inalienable posses-
sion, and Frances Elton and Clark Denmark for helpful discussions of some
of the signs and patterns described in this chapter. This work was supported
by the Economic and Social Research Council of Great Britain (Grant RES-
620-28-6001) Deafness, Cognition and Language Research Centre (DCAL)
and also by the Centre for Deaf Studies (University of Bristol). Finally, we
thank William McGregor and Adam Schembri for helpful comments on
drafts of this chapter. Any errors that may remain are, of course, our own.
References
Alibasic Ciciliani, Tamara and Ronnie B. Wilbur

2006 Pronominal system in Croatian Sign Language. Sign Language and
Linguistics 9 (1–2): 95–132.
Arik, Engin
2008 Locatives, existentials and possessives in Turkish Sign Language
(TID). Paper presented at the Annual Meeting of the Linguistic So-
ciety of America, Chicago, IL.
Berenz, Norine
2002 Insights into person deixis. Sign Language and Linguistics 5 (2):
203–227.
Branson, Jan, Jennifer Toms, Brian Bernal and Don Miller
1995 The history and role of fingerspelling in Auslan. In Sign Language
Research 1994: Proceedings of the Fourth European Congress on
Sign Language Research, Munich, September 1–3, 1994, Heleen Bos
and Trude Schermer (eds.), 53–67. Hamburg: Signum Press.
Brennan, Mary
1992 An introduction to the visual world of BSL. In Dictionary of British
Sign Language/English, David Brien (ed.), 1–133. London: Faber and
Faber.
Cooper, W. Robert
2002 Inalienable possession in Finnish and English: the use of possessive
pronouns/suffixes with nouns designating part of the body. Helsinki
English Studies 2.
Cormier, Kearsy
2005 Exclusive pronouns in American Sign Language. In Clusivity: Typo-
logy and Case Studies of Inclusive-Exclusive Distinction, Elena Fili-
monova (ed.), 241–268. Amsterdam: John Benjamins.
Cormier, Kearsy
2007 Do all pronouns point? Indexicality of first person plural pronouns in
BSL and ASL. In Visible Variation: Comparative Studies on Sign
Language Structure, Pamela M. Perniss, Roland Pfau and Markus
Steinbach (eds.), 63–101. Berlin /New York: Mouton de Gruyter.
Cormier, Kearsy, Stephen Wechsler and Richard P. Meier
1999 Locus agreement in American Sign Language. In Lexical and Con-
structional Aspects of Linguistic Explanation, Gert Webelhuth, Jean-
Pierre Koenig and Andreas Kathol (eds.), 215–229. Stanford, CA:
CSLI Press.
Deaf Society of New South Wales
1998 Hands up NSW: A Profile of the Deaf Community of NSW. Sydney:
Deaf Society of New South Wales.
Deuchar, Margaret
1983 Is British Sign Language an SVO language? In Language in Sign: An
International Perspective on Sign Language, Jim Kyle and Bencie
Woll (eds.), 69 –76. London: Croom Helm.
Engberg-Pedersen, Elisabeth
1993 Space in Danish Sign Language. Hamburg: Signum Press.
Farris, Michael A.
1994 Sign language research and Polish Sign Language. Lingua Posna-
niensis 36: 13–36.
Farris, Michael A.
1998 Models of person in sign languages. Lingua Posnaniensis 40: 47–59.
Fenlon, Jordan and Kearsy Cormier
2006 Inalienable possession in British Sign Language. Paper presented at
the Ninth International Conference on Theoretical Issues in Sign Lan-
guage Research, Universidade Federale de Santa Catarina, Floriano-
polis, SC, Brazil, 6–9 December 2006.
Fischer, Susan D.
1975 Influences on word order change in American Sign Language. In
Word Order and Word Order Change, Charles N. Li (ed.), 1–25.
Austin: University of Texas Press.
Freeze, Ray
1992 Existentials and other locatives. Language 68: 553–595.
Friedman, Lynn
1975 Space and time reference in American Sign Language. Language 51
(4): 940–961.
Heine, Bernd

2001 Introduction: Dimensions of possession. In Dimensions of Possession,
Irène Baron, Michael Herslund and Finn Sørensen (eds.), 1–25. Ams-
terdam: John Benjamins.
Johnston, Trevor
2003 BSL, Auslan and NZSL: Three signed languages or one? In Procee-
dings of the Seventh International Conference on Theoretical Issues
in Sign Language Research, Anne Baker, Beppie van den Bogaerde
and Onno Crasborn (eds.), 47–69. Hamburg: Signum Verlag.
Johnston, Trevor and Adam Schembri
2007 Australian Sign Language: An Introduction to Sign Language Lingu-
istics. Cambridge: Cambridge University Press.
Kliffer, Michael D.
1996 Commonalities of French and Mandarin inalienable possession. Lan-
guage Sciences 18 (1–2): 53–69.
Klima, Edward and Ursula Bellugi
1979 The Signs of Language. Cambridge, MA: Harvard University Press.
Kristoffersen, Jette Hedegaard
2003 Existence, location and possession and the order of constituents in
Danish Sign Language. In Crosslinguistic Perspectives in Sign Lan-
guage Research: Selected Papers from TISLR 2000, Anne Baker,
Beppie van den Bogaerde and Onno Crasborn (eds.), 131–139. Ham-
burg: Signum Verlag.
Liddell, Scott K.
2003 Grammar, Gesture and Meaning in American Sign Language. Cam-
bridge: Cambridge University Press.
Lillo-Martin, Diane, and Edward Klima
1990 Pointing out differences: ASL pronouns in syntactic theory. In Theo-
retical Issues in Sign Language Research, Vol. 1: Linguistics, Susan
D. Fischer and Patricia Siple (eds.), 191–210. Chicago: University of
Chicago Press.
MacLaughlin, Dawn
1997 The structure of determiner phrases: evidence from American Sign
Language. PhD thesis, Department of Linguistics, Boston University.
McKee, David and Graeme Kennedy
2000 Lexical comparisons of signs from American, Australian, British and
New Zealand Sign Languages. In The Signs of Language Revisited:
An Anthology to Honor Ursula Bellugi and Edward Klima, Karen
Emmorey and Harlan Lane (eds.), 49–76. Mahwah, NJ: Lawrence
Erlbaum Associates.
Meier, Richard P.
1990 Person deixis in ASL. In Theoretical Issues in Sign Language Re-
search, volume 1: Linguistics, Susan D. Fischer and Patricia Siple
(eds.), 175–190. Chicago: University of Chicago Press.
Mitchell, Ross E. and Michael Karchmer

2004 Chasing the mythical ten percent: parental hearing status of deaf and
hard of hearing students in the United States. Sign Language Studies
4 (2): 138–163.
Neidle, Carol, Judy Kegl, Dawn MacLaughlin, Ben Bahan and Robert Lee
2000 The Syntax of American Sign Language. Cambridge, MA: MIT Press.
Nespor, Marina and Wendy Sandler
1999 Prosody in Israeli Sign Language. Language and Speech 42 (2–3):
143–176.
Nichols, Johanna
1988 On alienable and inalienable possession. In In Honor of Mary Haas:
From the Haas Festival Conference on Native American Linguistics,
W. Shipley (ed.), 557–609. Berlin: Mouton de Gruyter.
Nichols, Johanna
1992 Linguistic Diversity in Space and Time. Chicago: University of Chi-
cago Press.
Padden, Carol A.
1983 Interaction of morphology and syntax in American Sign Language.
PhD thesis, Department of Linguistics, University of California at
San Diego.
Padden, Carol A.
1990 The relation between space and grammar in ASL verb morphology.
In Sign Language Research: Theoretical Issues, Ceil Lucas (ed.),
118–132. Washington, DC: Gallaudet University Press.
Payne, Doris
this vol. Is possession mere location? Contrary evidence from Maa.
Pichler, Deborah Chen, Katharina Schalber, Julie Hochgesang, Marina Milkovic,
Ronnie B. Wilbur, Martina Vulje et al.
2008 Possession and existence in three sign languages. In Sign Languages:
Spinning and Unraveling the Past, Present and Future. TISLR9, Forty-
five Papers and Three Posters from the 9th Theoretical Issues in Sign
Language Research Conference, Ronice Muller de Quadros (ed.),
440–458. Petropolis /RJ. Brazil: Editora Arara Azul.
Rathmann, Christian and Gaurav Mathur
2002 Is verb agreement the same cross-modally? In Modality and Structure
in Signed and Spoken Languages, Richard P. Meier, Kearsy Cormier
and David Quinto-Pozos (eds.), 370–404. Cambridge: Cambridge Uni-
versity Press.
Schembri, Adam and Trevor Johnston
2006 Sociolinguistic variation in Australian Sign Language Project: Gram-
matical and lexical variation. Paper presented at the 9th International
Conference on Theoretical Issues in Sign Language Research, Uni-
versidade Federale de Santa Catarina, Florianopolis, SC, Brazil, 6–9
December 2006.
Sutton-Spence, Rachel and Bencie Woll

1999 The Linguistics of British Sign Language. Cambridge: Cambridge
University Press.
Woll, Bencie, Rachel Sutton-Spence and Dafydd Waters
2004 ECHO data set for British Sign Language (BSL). London: Depart-
ment of Language and Communication Science, City University.
http://www.let.ru.nl/sign-lang/echo.
Wulf, Alyssa, Paul Dudis, Robert Bayley and Ceil Lucas
2002 Variable subject presence in ASL narratives. Sign Language Studies
3 (1): 54 –76.
Contributors
Kearsy Cormier Mirjam Fried

Deafness Cognition and Language Ústav pro jazyk esk AV R
Research Centre (DCAL) Letenská 4
University College London 118 51 Praha 1 – Malá Strana
49 Gordon Square Czech Republic
London WC1H 0PD fried@ujc.cas.cz
UK
k.cormier@ucl.ac.uk Liesbet Heyvaert
Department of Linguistics
Kristin Davidse University of Leuven
Department of Linguistics Blijde-Inkomststraat 21
University of Leuven PO Box 3308
Blijde-Inkomststraat 21 3000 Leuven
PO Box 3308 Belgium
3000 Leuven liesbet.heyvaert@arts.kuleuven.be
Belgium
kristin.davidse@arts.kuleuven.be Frantisek Lichtenberk
Department of Applied Language
Sonja Eisenbeiß Studies and Linguistics
Department of Language and Linguistics University of Auckland
University of Essex Private Bag 92019
Colchester C04 3SQ Auckland
UK New Zealand
seisen@essex.ac.uk f.lichtenberk@auckland.ac.nz
Jordan Fenlon Ayumi Matsuo

Deafness Cognition and Language The School of English Literature,
Research Centre (DCAL) Language and Linguistics
University College London Jessop West
49 Gordon Square 1 Upper Hanover Street
London WC1H 0PD Sheffield S3 7RA
UK UK
j.fenlon@ucl.ac.uk a.matsuo@sheffield.ac.uk
424 Contributors
William B. McGregor Miriam van Staden

Afdeling for Lingvistik – Institut for Theoretical Linguistics
Antropologi, Arkæologi og Lingvistik Faculty of Humanities
Aarhus Universitet University of Amsterdam
Building 1410 Spuistraat 210
Ndr. Ringgade 1012 VT Amsterdam
8000 Århus C The Netherlands
Denmark m.vanstaden@uva.nl
linwmg@hum.au.dk m.vanstaden@minaz.nl
Doris Payne Hein van der Voort

Department of Linguistics Afdeling Taalwetenschap
University of Oregon Radboud Universiteit Nijmegen
Eugene, OR 97403 Postbus 9103
USA 6500 HD, Nijmegen
dlpayne@uoregon.edu The Netherlands
hvoort@xs4all.nl
Jan Rijkhoff Peter Willemse

Afdeling for Lingvistik – Institut for UFR des Langues Etrangères
Antropologi, Arkæologi og Lingvistik Appliquées
Aarhus Universitet Université Lille 3 Charles-de-Gaulle
Building 1410 14 Place Bodart Timal
Ndr. Ringgade BP 447
8000 Århus C 59058 ROUBAIX cedex 01
Denmark France
linjr@hum.au.dk peter.willemse@univ-lille3.fr
Ingrid Sonnenstuhl
Institute for Language and Information
University of Düsseldorf
Universitaetsstrasse 1
40225 Düsseldorf
Germany
ish@phil-fak.uni-duesseldorf.de
Index of subjects
abstract nouns, 67, 225–226 applicative, 123, 319

abstract possession, 181, 359 areal diffusion, 343, 370n
active possession, 276 areal features, 359, 372, 383–384
actor prefixes, 295, 298, 312–313, 322 areal patterns, 336, 381
adjectival predicates of predisposition, association (see also bridging), 22–24,
389, 414–416 37–38, 298–300
adjectives, 54–55, 94, 121, 214n, 295– attributive compound, 380
298, 304, 310–312, 336 attributive construction, 344, 349–351
classifying, 63–64, 67 attributive possession (see also adnomi-
relational, 64n nal possession), 2–3, 6, 9, 97, 213–
adjuncts, see satellites 214, 217, 231–234, 238, 239n, 246,
adnominal possession (see also attribu- 249–286, 293, 316, 319, 344, 396–
tive possession), 2–5, 8, 51–97, 144, 408, 416
150, 343–384 non-universality of contrast with
in child language, 155–179, 185–186, predicative possession, 7–8, 294,
188–189, 193, 196–199 303–314, 324, 336
adverbs, 52, 345, 359
adverb formatives, 359 benefactive, 8, 123, 299, 355, 360–364,
affectedness, 6, 192, 197, 214–216, 219– 367, 374, 381, 383
221, 227, 229–233, 235, 239–240, beneficiary, 109n, 232, 233, 299, 309,
242, 244–245 316, 355, 360–362, 367, 374
affected objects, 333–335 body part classifiers, 353, 366, 372
affected possessors, 214–246 passim body part nouns/terms, 40–41, 117n, 153,
affective construal, 243 175, 195–196, 197, 216, 219, 221,
affective verbs, 227, 230, 236 225n, 227, 230, 242, 264, 274, 277,
Agent – see agent role 278–279, 299, 358, 405, 406–407
agent role, 100, 110, 112, 119, 123, 180, body product terms, 265
224, 243, 267–268, 276, 294, 333 bridging, 20–25, 36, 37, 39–41
agentive participant, 231
alienable noun, 377, 382 causative (see also benefactive), 8, 355
alienable possession, 7, 262, 277, 286 classifiers (see also modifier, classifying
vs. inalienable possession, 9, 262– and possessive classifiers), 8–9, 264,
263, 282–285, 299, 319, 320, 356, 281, 347–348, 349n, 351–354, 366–
358, 375, 381–382, 406–407 367, 370–373, 377
alienable possessum, 25, 42, 44, 192, 216 comitative, 355, 367
alienability, 215–216, 221, 225, 231, 364, compounds, 66, 349n, 369, 380, 415
376 Construction Grammar, 6, 108, 217, 234–
anchor – see referential anchor 235, 246
animate, 1, 82, 107, 109, 158, 197, 223 constructional gestalt, 245
426 Index of subjects
constructional networks, 213, 217, 238– existence, 3–4, 80, 109, 111, 115–116,
240, 246 133, 138, 181, 409–411, 416
context, 6, 20–21, 23, 113, 226–227, 250, existential constructions, 4, 82, 108n,
283, 286 115, 128n, 133–134, 136–138, 300
coreference (see also discourse status, existential predicates/verbs, 108, 120,
co-referential), 17 301–302, 359, 409–411
corpora, 13n, 27, 44n, 45, 134, 154, existential schema, 216
202, 213, 390 experiential gestalt, 216, 241, 245, 294
extension
dative arguments, 192, 194, 197, 232 in child language, 174, 178–179, 188
dative constructions, 192, 239 metaphorical, 114n, 138
dative marking, 156–157, 183, 189– prototype, 244, 246
191, 195–196, 198, 214, 222, 227, external possession constructions (EPCs)
231–-232, 236, 239, 245 (see also affected possessors), 2, 5–
dative object, 190–192 6, 144, 153–154, 189–192, 199, 214,
Dative of Interest (DI), 6, 233, 236–237, 216, 230, 344
239 in child language, 192–198
dative possessor, 214, 222, 227, 231–232, external possessor – see external posses-
235–236, 245 sion construction
definite, 13–28 passim, 37–47 passim,
54–55, 57, 59, 79–89 passim, 93, flexible parts-of-speech, 296–297
181–182, 187, 332 fluidity (of classifiers), 7, 250, 263, 273–
dependency reversal, 59, 76n 276, 281–282, 285
direct possessive construction, 6–7, 253– focus, 222–223
254, 256, 258–260, 263–267, 269– focus markers, 304, 307–308
270, 273, 276–277, 279, 282, 285 frame, 22, 24, 234
direct prefixes, 121 cognitive, 217
discourse context, 3, 15n, 19, 22, 30–32, interpretive, 234
37, 42, 47–48, 240 frame participant, 235, 237, 239
discourse status Functional Grammar, 52
anchored, 27, 41–44, 47 functional prototype, 217
coreferential, 27, 28 –31, 36, 379, functional space, 217, 238, 246
382–383
given, 3, 13, 17–20, 26–28, 41, 47 genitive
inferable, 25–27, 36–47 compound, 369
initial mention, 15, 24, 26, 28n, 46 construction, 6, 14, 56, 108, 150, 156,
new, 3, 13, 17–20, 24–28, 31, 37–38, 162, 167, 171, 178, 381
41–42, 45–47 non-determiner/non-referential, 63, 75
non-initial, 24, 46 possessor, 214, 222, 238, 245
text referent, 27, 32–36, 41 of quantification, 77
discourse functions, 20, 47–48 -s, see -s genitive
double object construction, 299 schema, 217, 233
dummy, 9, 300, 302, 335, 358–359, 370– syntactic – see syntactic genitive
371, 380 goal schema, 217, 233
Index of subjects 427
grammatical construction, 213, 217, 234, status, 307

239, 246 structure, 222–223, 227–228, 235–
grammatical role, 152, 180, 183, 188, 236
216, 232n interpersonal level, 52, 53, 62
grammaticalization, 59, 61–62, 77, 124, intrinsic relation, 262, 282, 285
138, 272, 327, 374–375, 383 inverse prefixes, 121–122, 130
involvement, 217, 219, 227
habitual possession relation, 175, 177
homophony, 354–355, 358, 371 kin terms, 44, 145, 147, 159, 165, 167,
169–171, 199, 270, 277–278, 299,
identifiable (referent), 15n, 16, 19, 23, 325, 357n, 375–376, 405–407
55, 80, 82, 93, 95, 136 kinship, 119, 153, 176, 221, 226, 229, 262,
identification, 15–22, 47, 81, 93n 266, 272, 282, 284, 298, 344, 358
Impersonal passive, 122, 133, 137 as anchoring relation, 42–43
impersonal reference, 280 classifier, 276, 285
inalienable noun, 321, 375n, 376–377,
382, 407 language contact, 315, 317, 319, 332, 389
inalienable possession, 7, 10, 40–41, 60, layering
117, 158, 192, 217, 219, 262, 264, in the noun phrase, 51–56, 60, 62–
277–278, 284, 405–407, 416 90, 96
inalienable possessum, 235, 239, 285, in constructions, 238, 245
321, 366 levels – see interpersonal level, represen-
inalienability, 6, 66n, 215–216, 221, 231, tational level
244, 279, 416 linguistic areas, 293, 314, 317
inanimate, 1, 110, 158, 216, 223, 301n location, 4, 52–55, 58, 62, 80, 120, 132,
inclusive-exclusive contrast, 332, 346, 134–135, 137, 180, 185–188, 321
394 relation to possession, 82, 107–108,
inclusive pronouns, 249n, 347, 355, 374, 109–117, 137–138, 181, 416–417
376 predication of, 124–127
indefinite, 14, 16, 19, 21, 24–25, 30, 42– locative schema, 326
44, 47, 54–55, 83–89, 92–93, 182–
183, 187 malefactive, 299
indirect object, 190–192, 232n metaphor, 61, 109, 114n, 126, 138, 344
indirect possession, 6–7, 253–254, 258, metaphorical constructions, 76
263–264, 267–273, 275, 276–279, metaphorical Locative, 110, 126, 138
282, 285, 373, 378 metatypy, 314–315, 317
individuation, 279 modification relation, 51, 62, 66–67, 70,
indivisibility, 220 74, 89–90, 94, 96
inflexion, 345–346, 348–349, 360–361, Modification construction, 231, 237–238
363, 374–375 modifiers (see also satellites)
information adnominal, 60–61, 82–83, 97
flow, 217, 222 aspectual, 305–306, 312
given, 47 classifying, 51, 53, 55, 58, 61–62, 87,
new, 24 96–97
428 Index of subjects
discourse-referential, 53, 59 possessive pronouns, 3, 9, 28, 59, 155–

grammatical, 52–53 163, 165–167, 169, 172, 178, 197–
illocutionary, 53 198, 237, 320, 324–329, 360–363,
localizing, 51, 53, 58, 79 365, 367, 373–374, 381–383, 389,
proposition, 53 390n, 396–408, 413–417
qualifying, 51, 53, 58, 61, 67, 71, 74, possessive prototype, 217, 238, 240–241,
93, 97 244–246
quantifying, 53, 56, 79 possessive -s – see -s genitive
temporal, 312 possessive schema, 216, 245
modifying construction – see attributive predication, 3, 51, 62, 66–67, 70, 74, 89–
construction 90, 94, 96–97, 117, 185–188, 295,
302–304, 310, 326, 333, 335
nominalizations, 26, 45–46, 126, 224, of existence, 115, 120
239, 267, 276, 295, 297, 311, 336, of location, 115, 124, 126, 132, 138
345–352, 354, 367, 370n, 371 predicative possession, 2, 5, 7, 9–10,
non-agentive, 236–237, 243 107–108, 111–112, 114, 117, 119,
non-topicality, 38–39 136, 144, 154, 179–189, 213, 294,
noun formative, 353, 358–359, 366–367, 303, 316, 322, 329, 336, 343–344,
369, 372 358–359, 363–364, 384, 389, 408–
noun phrase 414, 416
in child language, 154–155, 202–203 in child language, 182–189
layered model of – see layering prenominal possessive, 13–20, 25, 37n,
38, 40
object prefixes, 319, 322, 330–335 prepositional phrases, 66, 84, 159, 195–
operators, 52, 89 196, 198
clause operators, 52, 54 prepositional possessive constructions,
NP operators, 54, 62, 77–79 5, 150, 173
pronouns in British Sign Language
part-whole relations, 21–22, 37, 41, 60, inalienable possession, 405–407
107, 119, 131, 172, 177–178, 216, exclusivity, 394–395, 402
242, 282, 299, 351, 358 number, 392–393, 397–400
passive possession, 276 person, 390–392, 397–400
possessibility hierarchy, 126, 221, 224– spatial marking, 390, 407–408
225, 234–235, 238–239, 243, 245 word order, 395–396, 400–402
possessive classifiers, 6–7, 9, 249, 253– prototypical ownership, 216, 234
254, 256–257, 261, 264, 267–277, prototypical possession/possessive
281–286, 352, 373, 378, 381–382 relation, 158–159, 176, 198, 234
possessive construal, 31, 216, 226, 245
possessive determiner, 13n, 14, 37, 47, quotative markers, 56
284
possessive frame, 236 –244 passim reduplication, 10, 296, 384
possessive prefixes, 254, 293, 298–299, reference point (see also referential
312–313, 320–321, 322, 329–333, anchor), 17–18, 21, 113, 117, 119
368–370, 372, 377–378, 380
Index of subjects 429
reference point construction, 2–3, 13, syntactic constraints, 159, 170–171,

17–21, 30, 47, 48, 113 178, 199, 216–217, 230–231
referential anchor (see also reference syntactic functions, 224, 230, 235, 348
point), 55, 80–81, 83, 95 syntactic roles, 236
relational classifier – see possessive syntactic genitives, 239–240
classifier
relative clauses, 24, 52, 55–56, 81–82, ‘thematic’ dative, 232
222, 349–350, 352 thematic role, 190
representational level, 52–53 Theme, 109–114, 116, 120, 123, 125,
135, 137, 267–268, 276, 321, 336
satellites, 52, 54–55, 58–59, 61–64, 66– topic, 48, 112, 180, 408
70, 77–83, 89, 91, 93–96 topic construction, 117, 300–301
schemas, 23, 112, 152, 181–182, 293 topic marking, 395n
scope, 52–53, 61–62, 77, 89, 96, 279, topical item, 38, 114n, 158, 236
300, 305–306 topicality, 121, 126
semantic roles, 107–111, 113, 114n, 123, topic-focus relations, 222
137, 232n, 236, 326, 329, 333 transitivity hierarchy, 230
-s genitive, 5, 13n 27, 116, 143, 145, 147,
149–150, 156, 158–159, 164–172, unaffected object, 333–335
178, 196, 198–199, 404–405 universals, 383–384, 417
situated possession, 6, 234–237, 239–
242, 245 verb semantics, 217, 227–230, 239, 243–
subject prefixes, 293, 319–320, 333–334 244, 246
switch-reference, 378–379, 383 verbal ligature, 8, 293, 315–316, 319
Index of languages
Abui, 293, 318, 320–321 East Nusantara languages, 7, 293–336

Adang, 318, 320 English, 2–3, 13–48, 51–97, 112, 118–
’Ala’ala, 270 120, 138, 143, 155–156, 157–159
Amazonian languages, 79, 343–384 passim, 164–172 passim, 178, 199,
American Sign Language (ASL), 9, 283–284, 389, 390n, 404–405, 416
389, 391, 392n, 394n, 395, 401, Erromangan, 256–257
405, 407, 411, 414–416 Ewe, 82-83
Amharic, 115–116
Anejom), 254, 256–257, 277, 286 Fijian, 254–255, 258, 263, 268, 274,
Araki, 263, 273 276, 278, 280, 283
Aikanã, 8, 359–361, 364–367, 370, French, 111, 138
381–383
Arikapú, 378–383 Galela, 56, 296–297, 301n, 322, 324,
Australian languages, 79 327, 330–331, 335
Australian Sign Language (Auslan), 9, Gapapaiwa, 263, 277–278
395, 404–405, 416 German, 5, 143–203
Austrian Sign Language, 401, 407–408, Germanic languages, 62, 78
411 Greek (Modern), 160, 164, 173, 178
Austronesian languages, 76, 249–286,
293, 317–319, 332, 352, 382 Hausa, 67
Hawaiian, 259, 271
Bali-Vitu, 258 Hebrew, 160, 164–165, 173–174, 178
Banoni, 265 Hoava, 254
Baure, 8, 371–374, 381–383 Houaïlou, 270, 277
Biak, 319, 321
Buru, 293, 319, 321 Iaai, 258, 267, 272, 282
British Sign Language (BSL), 9, 389– il-Chamus, 107; see also Maa
417 Indonesian, 59, 309, 314–315
Italian, 16
Cavineña, 383
Cèmuhî, 257 Jabutí languages, 378
Chadic languages, 78 Jakaltek, 111, 115–116
Chiquitano, 383 Japanese, 5, 164, 172, 178, 192, 194,
Croatian Sign Language, 401, 407, 411 196, 198
Czech, 6, 213–246
Kairiru, 260, 277
Danish Sign Language, 411 Kanoê, 359, 365–367, 369, 381–383
Dutch, 3–4, 51–97 Kilivila, 265–266
Index of languages 431
Kokota, 253n, 267 Pohnpeian, 272

Komi (Southern Permyak dialect), 59 Polynesian languages, 259–260, 261,
Kosraean, 272, 275–276 270–271, 318
Kwaio, 255 Pukapukan, 259
Kwazá, 8–9, 345–360, 362n, 364–372 Puluwatese, 266
passim, 381–383 passim
Romance languages, 62
Lakondê, 368 Rotuman, 259
Labu, 261 Roviana, 65
Latundê, 368–371, 381–382 Russian, 77
Lenakel, 271
Lenkau, 258 Sahu, 301n, 325–330 passim
Lithuanian, 65–66 Sakurabiat – see Mekens
Loda, 322, 327 Samoan, 55–56
Lolovoli, 253, 265–266, 271, 274 Signed English, 405
Lou, 258 Swedish, 65, 75, 164
Maa, 4, 107-138 Tamambo, 265, 273

Maasai, 107, 122; see also Maa Tidore, 7–8, 293–330 passim, 334–336
Macro-Jê languages, 343, 378–380 Tobati, 261
Malay/North Moluccan Malay, 293, Tobelo, 296–297, 299, 323–331 passim
296, 309–310, 311n, 314–317 Tok Pisin, 293
Manam, 249–250, 254, 265, 267–269, Toqabaqita, 254, 260, 266–267, 270,
274–276, 283, 285 278–279
Mauritian Creole, 59 Tuparí languages, 376, 378
Maybrat, 293, 320, 332 Turkic languages, 59, 356
Mekens, 345, 376–378, 381–383 Turkish, 59–60
Movima, 10, 384 Turkish Sign Language, 411
Mussau, 272, 282
Ulithian, 272
Nalik, 266 Uralic languages, 59, 356
Nambikwara languages, 343, 368, 369
Nasioi, 60 Vinmavis, 254, 257–258
Niuean, 261
Nyulnyul, 4, 108n, 117n Wari’, 374–376, 381–382
Wayan, 268, 278, 280
Oceanic languages, 6–7, 249–286, 318, Western Oceanic languages, 76–77
373, 378 West-Makian, 322–323
Paamese, 264, 267, 278–279 Yawa, 321

Pagu, 322, 324, 326–329, 332–335
Panare, 78 Zabana, 286
Index of persons
Abbott, Barbara, 14 Brennan, Mary, 409

Ahland, Micahel, 115, 116n Brooks, Patricia, 149
Aikhenvald, Alexandra, 272, 277, 281, Brown, Dunstan, 253
345 Brown, Gillian, 22, 31
Alibasic Cicilliani, Tamara, 392n, 395 Brown, Roger, 143, 164–165, 172–173
Araujo, Gabriel de, 369 Bugenhagen, Robert, 80
Ariel, Mira, 22–23, 40 Butler, Christopher, 52
Arik, Engin, 411
Armon-Lotem, Sharon, 160, 164–165, Chafe, Wallace, 14, 22–23, 28, 119
172–173 Chappell, Hillary, 2, 7, 66n, 218, 221,
293, 299, 358, 382
Bacelar, Laércio, 365–367 Chomsky, Noam, 148
Baker, Carl, 145 Chouinard, M. M., 149
Barker, Chris, 30, 42, 262, 282 Croft, William, 217–218, 283
Baron, Irène, 1, 107, 109, 144, 180–181, Crowley, Terry, 257–258, 264, 267, 278–
188, 199, 344, 356, 408, 417 279, 352
Barshi, Immanuel, 2, 6, 107, 189, 192, Chervel, André, 31
214, 216, 218, 225, 344 Clahsen, Harald, 154, 156, 164–165, 173
Bates, Elizabeth, 145, 148–149 Clancy, Patricia, 164–165, 173
Bauer, Laurie, 66 Clark, Eve, 82, 109–110, 112, 115–116,
Bayley, Robert, 395 138, 148–149
Bellugi, Ursula, 391, 414–415 Clark, Ross, 259–260
Berenz, Norine, 392n Claudi, Ulrike, 83
Berman, Ruth, 164–165, 173 Comrie, Bernard, 59, 77, 109n
Bernal, Brian, 404 Cooper, W. Robert, 406
Bertolo, Stefano, 145 Corina, David, 148
Benveniste, Émile, 359 Cormier, Kearsy, 9, 389, 391–392, 394,
Berger, Andrea, 79 405, 423
Biber, Douglas, 14 Cowie, F., 148
Blanche-Benveniste, Claire, 31 Craig, Colette, 111
Blust, Robert, 318 Crain, Stephen, 151, 160
Bohnacker, Ute, 164 Crevels, Mily, 345, 359
Börjars, Kersti, 158
Borschev, Vladimir, 262 Danielsen, Swintha, 371–374
Bowden, John, 266, 315 Davidse, Kristin, 2–3, 13, 423
Bowerman, Melissa, 143, 179, 181 Declerck, Renaat, 16, 21
Braine, Martin, 149 DeLancey, Scott, 107, 109–110, 114
Brannon, Elizabeth, 79 Denison, David, 158
Branson, Jan, 404 Dik, Simon, 52, 54, 93, 97n, 303, 336
Index of persons 433
Dixon, Robert, 67, 78, 283, 345 Haiman, John, 283

Donohue, Mark, 261 Halliday, Michael, 15, 21, 32, 63
Du Bois, John, 2, 13–15, 22–24, 40 Hamaya, Mitsuyo, 120n, 121
Dudis, Paul, 395 Harbert, Wayne, 145, 159
Hasan, Ruqaiya, 15, 21, 32
Eisenbeiß, Sonja, 4–5, 10, 143–202 Haspelmath, Martin, 78–79, 83, 144
passim, 423 Haude, Katharina, 10, 383–384
Emmott, Catherine, 31 Hawkins, John, 20–22, 79
Engberg-Pedersen, Elisabeth, 392 Hedberg, Nancy, 14
Englebretson, Robert, 59 Heine, Bernd, 1, 7, 56, 61–62, 83, 97,
Everett, Daniel, 79, 374 –375 111–112, 118n, 138, 144, 153, 180–
181, 213–214, 216, 233, 293, 295,
Fabre, Alain, 373, 382–383 300, 303, 327, 359, 382–383, 408–
Farris, Michael, 392 409, 416–417
Fenlon, Jordan, 9, 389, 405, 423 Hengeveld, Kees, 52, 56, 62n, 93, 298,
Fischer, Steven, 259 –260 303
Fischer, Susan, 395n Herslund, Michael, 1, 107, 109, 144, 180–
Fitzsimons, Matthew, 286 181, 188, 199, 344, 356, 408, 417
François, Alexander, 263 Hinton, Leanne, 360
Frajzyngier, Zygmunt, 78 Hooper, Robin, 261
Fraurud, Kari, 23 –24, 42, 59 Huddleston, Rodney, 14
Freeze, Ray, 107, 412, 417 Hyams, Nina, 151
Friedman, Lynn, 391 Hyslop, Catriona, 271
Galasso, Joseph, 164–165, 173 Jackendoff, Ray, 109

Galucio, Ana, 376–378 Jacobs, Peter, 120n, 121
Garrod, Simon, 23 Jäger, Gerhard, 158
Geraghty, Paul, 254, 268n Johnston, Trevor, 395, 404
Gerzenstein, Ana, 373 Jordan, Kerry, 79
Giegerich, Heinz, 66
Gil, David, 78n, 362 Karchmer, Michael, 390n
Gildea, Spike, 78 Keesing, Roger, 255
Givón, Talmy, 78, 81 Kegl, Judy, 395n, 401n
Goldberg, Adele, 6, 108, 145, 152 Keizer, Evelien, 93
Greenberg, Joseph, 122 Kemmer, Suzanne, 113
Grinevald, Collette, see Craig, Collette Kennedy, Graeme, 404n
Gruber, Jeffrey, 109–110 Kern, Barbara, 374–375
Guasti, Maria, 151–152 Kiparsky, Paul, 148
Guillaume, Antoine, 383–384 Kirk, Arthur, 78
Guillemin, Diana, 59 Kliffer, Michael, 407
Gundel, Jeannette, 14 Klima, Edward, 391, 414–415
Gunkel, Lutz, 64n König, Ekkehard, 144
Gutierrez, Salome, 128n König, Christa, 121
434 Index of persons
Koptjevskaja-Tamm, Maria, 75, 344 Meier, Richard, 391

Kristoffersen, Jette Hedegaard, 411 Menges, Karl Heinrich, 59
Kroeker, Menno, 369 Menn, Lise, 165
Kuna, Branko, 117n Messineo, Cristina, 373
Küntay, Aylin, 149 Miller, Don, 404
Kuteva, Tania, 61–62, 83, 383 Miller, Max, 146
Mills, Anne, 145, 156, 165, 168
La Fontinelle, Jacqueline de, 270n, 277 Milner, George, 255, 258, 263
Langacker, Ronald, 2, 14, 15n, 17–18, Mitchell, Ross, 390
31, 47, 113, 117, 118n, 262 Mithun, Marianne, 78
LaPolla, Randy, 113, 119 Moravcsik, Edith, 81
Lee, Kee-dong, 272 Mpaayei, John, 121
Lee, Robert, 395, 401n
Leenhardt, Maurice, 270 Neidle, Carol, 395n, 401n
Lehmann, Christian, 81 Nespor, Marina, 413
Lichtenberk, Frantisek, 7, 249, 263–264, Neumann, Dorothea, 189, 192
269, 275, 279, 282–283, 285, 299, Newman, Paul, 67n, 78
344, 352, 373, 378, 382, 423 Newmeyer, Frederik, 97n
Liddell, Scott, 392 Nichols, Johanna, 273, 345, 382, 407
Lillo-Martin, Diane, 391
Lucas, Ceil, 395 Ozanne-Rivierre, Françoise, 258, 272
Lust, Barbara, 151
Lynch, John, 249n, 256, 263, 267, 273, Padden, Carol, 391
277, 282, 286 Palmer, Bill, 253n
Lyons, Christopher, 2, 13–14, 16, 20–21 Partee, Barbara, 262
Lyons, John, 61, 80, 82, 109, 114n Pawley, Andrew, 263, 268, 278, 280–
281
MacLaughlin, Dawn, 395, 401n Payne, Doris, 2, 4, 6, 107, 120n, 121,
MacWhinney, Brian, 145, 148 –149, 173 182, 189, 192, 214, 216, 218, 225,
Malchukov, Andrej, 59 344, 417n, 424
Marcus, Gary, 146, 148–149 Penner, Zvi, 156, 164–165, 172
Marinis, Theodoros, 143, 160, 164, 173 Peters, Ann, 165
Martin, James, 2, 13, 15–16, 21, 47 Pichler, Deborah, 401, 405, 407–408, 411
Massam, Diane, 261 Pinker, Steven, 143, 145, 151–152, 179,
Mathur, Gaurav, 391 181, 184, 199
Matsuo, Ayumi, 4–5, 10, 143, 194, 423 Prince, Ellen, 79
McCawley, James, 97n
McGregor, William, 1–2, 7, 9, 40, 66n, Quirk, Randolph, 2, 13–14, 75, 89
79, 95n, 108n, 117n, 180, 194, 218,
221, 284n, 293, 299, 355, 358, 382, Radford, Andrew, 143, 152, 153n, 164–
424 165, 173
McGuckin, Catherine, 263, 277 Rasmussen, Kent, 120n
McKee, David, 404n Rathmann, Christian, 391
Index of persons 435
Rehg, Kenneth, 272 Tham, Shiao Wei, 113

Ribeiro, Eduardo, 378–379 Tomasello, Michael, 145, 148–149, 151–
Rijkhoff, Jan, 3, 51–52, 54, 55n, 59, 62, 152, 153n, 167
78–79, 356, 424 Tryon, Darrell, 318
Rivierre, Jean-Claude, 257 Tsunoda, Tasaku, 1, 5, 192, 216, 225
Rizzi, Luigi, 151 Tucker, Archibald, 121
Rodrigues, Aryon, 379–380
Rosenbach, Anette, 2, 13–14, 57n, 158 Vainikka, Anne, 154, 164 –165, 173
Ross, Malcolm, 76, 258, 270, 272, van Baarda, M. J., 296, 301n, 322n, 326,
314–315 327, 331
Ruff, Claudia, 157, 162–163, 165, 167, van den Berg, Helma, 109n
172–173 van den Berg, René, 258n
van der Voort, Hein, 8–9, 10, 343, 345,
Sakel, Jeanette, 384 347, 349, 351, 357, 359, 365, 370n,
Sandler, Wendy, 413 378–379, 424
Sanford, Anthony, 23 Van Valin, Robert, 113, 119
Sayaba, Timocy, 263, 268, 278, 280–281 Varlokosta, Spyridoula, 160
Scott, Alan, 158 Vasconcelos, Ione, 360
Schembri, Adam, 395, 404 Velazquez-Castillo, Maura, 113, 117n
Seibt, Johanna, 55n von Garnier, Katharine, 78
Seiler, Hansjakob, 1, 56, 153, 180, 359
Shibatani, Masayoshi, 189 Wagner, Klaus, 154, 173
Siegel, Jeff, 261 Waters, Dafydd, 390
Siewierska, Anna, 59 Wechsler, Stephen, 391
Slobin, Dan, 149, 151 Weissenborn, Jürgen, 156, 164–165, 172
Song, Jae Jung, 62n Wertheimer, Max, 109
Sonnenstuhl, Ingrid, 4 –5, 10, 143, 202, Wexler, Kenneth, 151, 153n
424 Wilbur, Ronnie, 392n, 395
Sørensen, Finn, 107, 144, 180 –181, 188, Willemse, Peter, 2–3, 13, 19–20, 63n,
199 84n, 424
Sperlich, Wolfgang, 261 Williams, Edward, 56
Stassen, Leon, 344, 355 Wilson, William, 271
Stenzel, Achim, 165 Wivell, Richard, 260, 277
Stephany, Ursula, 160 Woll, Bencie, 390, 405, 415
Sutton-Spence, Rachel, 390, 405, 415 Wulf, Alyssa, 395
Wurm, Stephen, 318
Takahashi, Hidemitsu, 33
Tauli, Valter, 59 Young-Scholten, Martha, 154
Taylor, John, 1–2, 13, 17–19, 21, 28, 42, Yule, George, 22, 31
47, 63n, 216, 240, 294
Telles, Stella, 368 –371 Zacharski, Ron, 14

(The Expression of Cognitive Categories) William B. McGregor, William B. McGregor-The Expression of Possession - de Gruyter Mouton (2009)

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

(The Expression of Cognitive Categories) William B. McGregor, William B. McGregor-The Expression of Possession - de Gruyter Mouton (2009)

Uploaded by

Copyright:

Available Formats

The Expression of Possession

Library of Congress Cataloging-in-Publication Data

The expression of possession / edited by William B. McGregor.

Bibliographic information published by the Deutsche Nationalbibliothek

English possessives as reference-point constructions and their

On the co-variation between form and function of adnominal

Is possession mere location? Contrary evidence from Maa . . . . . . . . . . . . 107

Learning to encode possession . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143

Plain vs. situated possession in Czech: A constructional account . . . . . . 213

Attributive possessive constructions in Oceanic . . . . . . . . . . . . . . . . . . . . . . 249

Possessive clauses in East Nusantara, the case of Tidore . . . . . . . . . . . . . . 293

Possessive expressions in the Southwestern Amazon . . . . . . . . . . . . . . . . . 343

Possession in the visual-gestural modality: How possession is

The nine papers constituting The Expression of Possession deal with a

new and anchored to a given PR, is also problematic: in most instances PM

category, but embraces a range of emically distinct subtypes. The extent to

In Plain vs. situated possession in Czech: a constructional account, Mir-

complexities are identified and discussed in Lichtenberk’s article; these

be replaced by a specific classifier. (This situation is not to be confused

As the above summaries indicate, a diverse range of genetically and ty-

Peter Willemse, Kristin Davidse and Liesbet Heyvaert

2. English prenominal possessives in the literature

2.1. Alleged definiteness of possessive NPs and presupposed

Possessive NPs have predominantly been associated with definiteness in

(1) my cousin the son/daughter of my aunt and uncle

Rosenbach (2002: 14) proposes a similar analysis of possessive NPs as

(2) the king’s daughter the daughter of the king

In addition to claims classifying possessive NPs paradigmatically with

ential, proposes that “[p]ossessive noun phrases (…) presuppose identifi-

It is added that “[h]is is similar to the in that it demands (presupposes) iden-

(4) It is not to be wondered at [...] that a friend’s friend, described by letter,

(5) Jimmy was a referral, a friend of a friend. (CB)

(6) On Monday, Christie’s in New York is to sell Greta Garbo’s knickers.

NPs. However, this NP-internal identifying relation has to be distinguished

In other words, the PM referent can already be present as a singled-out in-

2.2. Alleged newness of PM referents as targets of reference-point

Langacker (1993, 1995) proposes that possessive NPs should be analyzed as

tions as the ‘target’ needing identification. For instance, in the possessive

2.3. Overview and research questions

To sum up, existing accounts of English prenominal possessives have

3. The functional dimensions of definiteness and possessive NPs

As observed by Lyons (1999), there is less unanimity about the functional

Although an owner is of course not a part of a house, it is an entity typically

In the literature, the invoking of events to make a referent recoverable has

The treatment of ‘spoon’ as an identifiable referent, Chafe argues, is due to

by contrast, are thought of as involving the retrieval of previously estab-

In fact, in many cases the prenominal possessive can be replaced by the

(13) An Irish detective arrested a wanted criminal in a Dublin street. Just

(15) Raven-haired Ruth is a statuesque woman consumed with envy for

Finally, some PM referents are ‘fully new’ in the discourse, unpredictable

From this brief discussion of first-mention PM referents it transpires that

Table 1. Overview of the data base

NP type # of instances Source

Table 2. Discourse statuses of the PM referent

the PM referent has been mentioned in the preceding

Table 2 gives an overview of these five discourse statuses, which, in the

4.1. The possessive NP is COREFERENTIAL with another NP in the

As we saw in section 1, Taylor (1996: 217) proposed that PM referents are

Table 3. Results for the category coreferential

data set # of tokens percentage

(17) Mr Ashby, a former name who suffered substantial losses at Lloyd’s,

The referent ‘verdict’ is first introduced by the NP a majority verdict and

(18) Britain’s leading arms manufacturers secretly liaised with ministers,

Clause operators ( /) Clause satellites ( /)