This textbook investigates definiteness both from a comparative and a theoretical point of view, showing how languages express definiteness and
what definiteness is. It surveys a large number of languages to discover
the range of variation in relation to definiteness and related grammatical
phenomena: demonstratives, possessives, personal pronouns. It outlines
work done on the nature of definiteness in semantics, pragmatics and syntax, and develops an account on which definiteness is a grammatical category represented in syntax as a functional head (the widely discussed
D). Consideration is also given to the origins and evolution of definite
articles in the light of the comparative and theoretical findings. Among
the claims advanced are that definiteness does not occur in all languages
though the pragmatic concept which it grammaticalizes probably does,
that many languages have definiteness in their pronoun system but not
elsewhere, that definiteness is not inherent in possessives, and that
definiteness is to be assimilated to the grammatical category of person.

General editors: s. r. anderson, j. bresnan, b. comrie, w. dressle
c. ewen, r. huddleston, r. lass, d. lightfoot, j. lyons,
p. h. matthews, r. posner, s. romaine, n. v. smith, n. vincent


In this series
p. h. matthews Morphology Second edition
b. comrie Aspect
r. m. kempson Semantic Theory
t. bynon Historical Linguistics
j. allwood, l.-g. anderson and ö. dahl Logic in Linguistics
d. b. fry The Physics of Speech
r. a. hudson Sociolinguistics Second edition
a. j. elliott Child Language
p. h. matthews Syntax
a. radford Transformational Syntax
l. bauer English Word-Formation
s. c. levinson Pragmatics
g. brown and g. yule Discourse Analysis
r. huddleston Introduction to the Grammar of English
r. lass Phonology
b. comrie Tense
w. klein Second Language Acquisition
a. j. woods, p. fletcher and a. hughes Statistics in Language Studies
d. a. cruse Lexical Semantics
f. r. palmer Mood and Modality
a. radford Transformational Grammar
m. garman Psycholinguistics
w. croft Typology and Universals
g. g. corbett Gender
h. j. giegerich English Phonology
r. cann Formal Semantics
p. j. hopper and e. c. traugott Grammaticalization
j. laver Principles of Phonetics
f. r. palmer Grammatical Roles and Relations
b. blake Case
m. a. jones Foundations of French Syntax
a. radford Syntactic Theory and the Structure of English: a Minimalist Approach
r. d. van valin, jr. and r. j. lapolla Syntax: Structure, Meaning and Function
a. duranti Linguistic Anthropology
a. cruttenden Intonation Second edition
j. k. chambers and p. trudgill Dialectology Second edition
c. lyons Definiteness

lecturer in linguistics
university of cambridge

The Pitt Building, Trumpington Street, Cambridge CB2 IRP
40 West 20th Street, New York, NY 10011-4211, USA
477 Williamstown Road, Port Melbourne, VIC 3207, Australia
© Cambridge University Press 1999
This edition © Cambridge University Press (Virtual Publishing) 2003
First published in printed format 1999

A catalogue record for the original printed book is available
from the British Library and from the Library of Congress
Original ISBN 0 521 36282 2 hardback
Original ISBN 0 521 36835 9 paperback

ISBN 0 511 00998 4 virtual (netLibrary Edition)

To the memory of my parents,
Edith and Patrick Lyons

3 Simple and complex indefinites 1.5 A unified account? 1.4 The noun phrase Definites: semantic content and behaviour 2.2 Familiarity and identifiability Possessives Proper nouns 1.6 Universal quantifiers 1.2.2 The distribution of definites and indefinites 2.5 Definiteness beyond the noun phrase 2 Simple definites and indefinites 2.1.2 Anaphoric articles page x .2.4 Identifiability.1 Simple definites and indefinites 1.5 Personal pronouns 1.3 Uniqueness and inclusiveness Indefiniteness and cardinality 1.1 Definiteness and indefiniteness marking 2.2.1 Optional definite articles 2.1 What is definiteness? 1.1.CONTENTS Preface Abbreviations 1 Basic observations 1.1 The occurrence of articles in languages 2.3.1. inclusiveness and indefinites 1.1 Testing for definiteness 1.3.2 Complex indefinites 1.2 Types of definite noun phrase 1.2 Demonstratives 1.2.

3.1 Demonstratives 3.5.1 Indefinite article 2.3 Forms and positions 3.3.2 Proper nouns 3.1 Pronoun and full noun phrase possessives 3.3 Pronominal marking 2.1 Semantic and grammatical content 3.3 Types of cardinal article Nominalizing and other functions Definite article types 2.4.4 Definite and demonstrative personal forms Quasi-indefinite article 2.2 Non-deictic distinctions 3.4.5 Definite adjectives Other ways of expressing definiteness 2.5 2.4 Personal pronouns 3.3.5 Universals and other quantifiers x .2.3.3 Alienable and inalienable 3.2 Bound articles 2.3 2.5. weak and null pronouns 3.3.4 Personal determiners 3.5.2 Strong.5 Bare indefinites General remarks 3 Complex definites and indefinites 3.1 Adpositional marking 2.2 Agreement 2.1 Free-form articles 2.4 Word order Indefinites 2.4.4 Mixed systems and double determination 2.4 Possessives and definiteness 3.3.4 Partitive indefinites 2.6 Phrasal clitics 2.4 Definite and specific Deictic distinctions 3.4.3 Possessives 3.2 Aaxal possessives 3.

2 Indirect object agreement 5.3 Definiteness and animacy 5.3 PRO 5.4 Implicit arguments .3 Subject agreement Concluding comments 5 Interaction with other grammatical phenomena 5.2 Pronouns 3.2 Verb agreement 5.2 Transparent contexts 4.3.1 Generics in English 4.2.3 Oblique direct objects 5.5 Null and implicit noun phrases 5.2 Definite object marking 5.1 Direct object marking 5.6.2 Specificity and referentiality 4.3 Non-singular generics 4.9 Non-configurational languages 4 Some semantic and pragmatic distinctions 4.1 Identifiability and inclusiveness Vocatives 3.3 A unified account of specificity? Dicerential case marking 5.2 Singular generics 4.3 Generics Traces 5.4 Articles and nominal feature marking 5.2.4 Generics and proper nouns 4.7 Indefinite demonstratives One definiteness or several? 4.2 Anaphoric and non-anaphoric definites 4.1 Opacity and scope ambiguities 4.3 “Deictic” and non-“deictic” definites 4.1 Pro 5.1 Full noun phrases 3.2.1 Direct object agreement 5.

2 Grammaticalization and definiteness 7.2.1 Grammaticalization Other definiteness ecects Superlatives 6.3.2 Aaxal articles 8.4 Internal-head relatives 6.3 Property predication 6.2 Generalized quantifiers 7.1 The grammatical.3 Weaknesses in the DP analysis xii .3.2 Hawkins: inclusiveness and location 7.3.5 Well.3.1.4 Relevance theory 7.3 The definite constraint The DP hypothesis 8.1 Milsark and the quantification restriction 7.7 Concluding point 6 Definiteness effects 6.1 Movement processes in DP 8. logical and pragmatic traditions 7.1 The organization of information Existential sentences 6.2 PP-extraposition and any opacity 6.3 Definiteness and quantification The specifier position and definite Dets 8.2 Explanatory accounts 6. what is definiteness then? 7.3.1. inclusiveness and exclusiveness 7.3.1 Discourse structure 6.1.3 Simple and complex definiteness 8 Definiteness and noun phrase structure 8.1 The phenomenon 6.2.2 Information structure and definiteness 6.5.5 Concluding remarks 7 Defining definiteness 7.1 The noun phrase as NP 8.2 Subsequent developments Other functional categories in the noun phrase 8.1 Shared sets.3 Discourse semantics: Heim 7.

3 Indefinites and CardP 9.1 Personal determiners and pronouns 8.3 The nature of person-definiteness 8.5.3 Numeral to cardinal article Demonstrative to definite article 9.1 Expansion in article use 9.2 A definite article cycle Definiteness and person 8.2.1 D as definiteness 8.2 The origin of articles Configurationality and definiteness 8.3 The analysis of personal pronouns The content of determiners 8.4 In conclusion References Index .2 Person and definite determiners 8.1 The emergence of functional structure 9.2 Articles in competition 9.1 The development of DP 9.6 Summary of proposals 9 Diachronic aspects 9.3 The longer perspective 9.2 The life cycle of definite articles 9.1.2.

I have not itated to advance far-reaching claims anticipating this approach. it surveys two areas. Second. where I have had to omit m which I see as important. This is not just a survey. but. First. Both the cross-linguistic survey and the theoret survey are introductory and far from complete. there is considerable variation in the ways in which these expressi relate to definiteness. some of which (like semantics and syntax) are highly techn While I assume some familiarity on the part of the reader with the principles . of course. My aim in this book is not to present a set of facts and analyse be learned. and in those that do t vary strikingly in both their form and their range of use. contain much discussion of the appro I believe to be the most promising. and other expressions which ei seem to be inherently definite or to interact in interesting ways with definiten but again. This literature is vast. unlike some other topic-ba books in this series. but to ocer a body of ideas to be thought about and improved up The investigation of definiteness necessarily takes the reader into sev domains of inquiry. Chapters 7 to 9. the book gives a (very selective) outline of the oretical literature on definiteness. Most languages do not have “articles”. possessives. and many of the choices I h made in reducing the material to manageable proportions are no doubt arbitr This is true particularly as regards the literature. All languages have dem stratives. I am much too interested in the topic no want to present my own view of what definiteness is. The view of defin ness I propose may be wrong. in particular. it ocers an account of the ra of variation displayed by languages in relation to definiteness and related gr matical concepts. But in the earlier chapters too. personal pronouns.P R E FAC E This book is primarily a survey. consisting both of di accounts of definiteness and of work mainly concerned with other phenomena which definiteness impinges. and it is essential that the reader follow up further erences given in the works I do refer to. but it will have achieved its purpos a student reading my proposals is spurred to investigate further and show t inadequacy. and I believe the work g in coherence from the aim of reaching and defending (if in outline) a prefe account. however.

The “new comparative guistics” in generative work indicates a recognition among theorists of the va of cross-linguistic investigation. Many people have helped me in various ways in the course of my writing book. to repeat a familiar point. I fir believe that descriptive breadth and analytical depth benefit one another. the wide-ranging descriptive work of typological studies. not as fully reliable data. K Jaszczolt. our understanding of the way guage works is deepened by bringing to bear serious analyses of languages. mere observational facts. But there is little in the syntax discussed here which cannot be easily rec in this paradigm. partly because the typologist looking at familiar languages in pursuit of a generalization is prone to the same inaccur as the theorist aiming to prove a point of theory. Most special thanks to Ricarda Schmidt for constant intellectual moral support. or discussed p ticular points with me: Nigel Vincent. partly beca the descriptive grammars on which they are based are often unclear on cru points or analytically unsophisticated. and I hope that the gulf between these approaches to language is narrower now than it was. on the ot the deep analysis of a smaller range of languages done in theoretical work. In my own cross-linguistic s vey here.and I have given at appropriate points brief outlines of essentials and referen to further reading. but the latter must be the ultimate goal. on the hand. Most of the of this book was written at a time when the current “minimalist” version of approach was in its infancy and there were few accessible accounts available this framework to refer the reader to. even the descriptive observations and gene izations made in typological work must be treated with great caution. And even the best descriptive grammars are rarely a quate by themselves to provide the basis for an analysis of any depth of a spec aspect of linguistic structure. There has been much debate over the years on the relative merits of. Noel Burton-Roberts. and I urge the reade treat it as a guide and starting point. sometime a fairly advanced point. But. in order to come fully to grips with the issues in qu tion. except as regards syntax. and I wish to thank in particular the following friends and colleagues w have read and commented on the manuscript or sections of it. so I have taken little account of minim ism. where I assume the principles-and-parame approach which is the most highly developed and best known. I have in general maintained neutrality between dicerent theoretical fram works. where possible at an elementary level. though some of it can be criticized as too se tive in scope. But it must be stres that the interested reader would need to follow up these references. I too will certainly have included inaccuracies. Deirdre Wilson. Indeed. xvi . and.

In the case of doctoral theses subseque published this means the formal publication. female. Finally. un otherwise stated. and the hearer or addressee male. . a note on my use of gender-marked personal pronouns in desc ing conversational exchanges: I follow the convention that the speaker is.Where items of literature discussed exist in dicerent versions. The ecect is sometimes that my erence is to a version dated several years later than the version most commo cited. I have tried to r to the most easily accessible version.

Dem DIR DIST DU ELAT xviii first person first person exclusive first person inclusive second person third person ablative case absolutive or absolute case accusative case adessive case anaphoric article aspect associated with first person associated with first person exclusive associated with first person inclusive associated with second person associated with third person auxiliary common gender clitic classifier. Def DEM. Ass2 ASS3 AUX C CL CLASS CONT DAT DECL DEF. Ass1 ASS1EXC ASS1INC ASS2. ana ART ASP ASS1. class marker continuous aspect dative case declarative definite demonstrative direction distal dual number elative case .ABBREVIATIONS 1 1EXC 1INC 2 3 ABL ABS ACC ADESS ANA.

imperfective aspect impersonal inanimate indefinite inessive case instrumental case intransitive irrealis linker locative case masculine gender neuter gender negative nominative case non-past tense nunation object oblique case partitive case passive voice past tense person plural number possessive present tense perfective aspect primary case proximal particle reflexive relative marker .EXP F FUT GEN GENR HAB HON IMP IMPF IMPRS INAN INDEF INESS INST INTR IRR LINK LOC M N NEG NOM NONPAST NUN OBJ OBL PART PASS PAST PERS PL POSS PRES PRF PRI PROX. Prox PRT REFL REL experiential aspect feminine gender future tense genitive case generic aspect habitual aspect honorific imperative imperfect.

D and Det. Agr et Any idiosyncratic or non-standard labels used are explained at the appropriate p in the text. V.SUBJ TNS TOP WH subject tense topic interrogative The standard labels are used for syntactic categories (N. xx .

This is followed by consideration of the vari types of noun phrase which are generally regarded as definite or indefinite – s definiteness and indefiniteness are not limited to noun phrases introduced by or a.1 What is definiteness? I begin in this section by attempting to establish in informal.and indefinite suax -n. with compara observations on other languages where appropriate. I shall refer to such elements by the traditional la article. without commitment at this stage to what their grammatical status actu . 1. English is taken as the starting point. It begins examining the concept of definiteness itself. to establish a preliminary accoun what this concept amounts to. since it has readily identifiable lexical artic which make definite and indefinite noun phrases on the whole easy to distingu It is important to bear in mind that the discussion in this chapter is prelimin and aims at a tentative and provisional account of the points examined. This element may be a lexical item like the definite and indefi articles of English (the. Finally.1. 1. For this purpose. Eng serves as well as any language. a). than to hop from one language to another. or an aax of some kind like the Arabic definite pr al.1 Simple definites and indefinites In many languages a noun phrase may contain an element which se to have as its sole or principal role to indicate the definiteness or indefiniten of the noun phrase. because it is easier and confusing to outline basic issues as they are instantiated in one language.1 Basic observations This chapter sets the scene by presenting some basic issues and id which will be investigated in greater depth in the rest of the study. some basic ideas concerning the syntactic structure of noun phr are presented in outline. theoretical terms what the intuitions about meaning are that correspond to our term a noun phrase “definite” or “indefinite”. Man the proposals made here and solutions suggested to problems of analysis wil refined as the study progresses. wh this can be done.

whereas in the case of an indefinite noun phr the speaker may be aware of what is being referred to and the hearer proba 1 2 We will see. This is clear from the fact that in English this house would u ally be judged (at least by linguists and grammarians) to be definite and seve houses indefinite.1. and I introduce these in 1 and 1. but. betw the hibiscus I planted last summer and a hibiscus I planted last summer? Many ditional grammars would give answers like the following: The indicates that speaker or writer is referring to a definite or particular car etc. “particular”. and is distinguished in my mind from all others. The car here is in some sense more “definite”. compare the follow two sentences: (1) I bought a car this morning. that articles can encode more than definiteness or indefiniteness. and they have been argued to have a quite dicerent principal function. apart from being rather vague.. the car I bought is a parti lar one. If I say I bough car this morning. So the question we are concerned with is: What is the dicerence in mean between the car and a car. which has as its essential semantic function to express category. Yet a car is indefin There is in fact no general agreement on what the correct answer is. however.1 I shall refer to such noun phrases as simple definites and sim indefinites. (2) I bought the car this morning. judgments would probably be more hesitant over every hou Noun phrases with the and a and their semantic equivalents (or near-equivale in other languages can be thought of as the basic instantiations of definite indefinite noun phrases. than a car.2 Familiarity and identifiability Continuing with the example just considered.1. a car certainly denotes a particu or specific car as far as the speaker is concerned. I am not referring to just any car. but two ma components of meaning have been much discussed. not just any.3 in relation to some illustrative English data. and I limit the discussion to them in this section to avoid any po bility of disagreement over the definite or indefinite status of example n phrases. “specific”. between the greedy child and a greedy child. 1. “indiv ualized” etc. this answer is quite inaccurate. in that the definiteness or indefiniteness stems from presence of the article.– though the definite–indefinite distinction is never thought of as applying o to those that do. as noted above. The dicerence is that the erence of the car in (2) is assumed to be clear to the hearer as well as the spea This is the first crucial insight. at least in some langua .

I re to this in Chapter 4. and its expression in terms of speaker’s hearer’s familiarity with a referent is standard in at least the less technical literature. (9) They’ve just got in from New York. dark-haired woman. The plane was five hours la (10) The president of Ghana is visiting tomorrow. a work which has gre influenced much subsequent writing on the subject. (5) I hear the prime minister behaved outrageously again today. before I put this vase it. Second. will you. skirting a number of issues subject to debate. We were rather worried by the prospect of having to cook for for two weeks. The signals that the entity denoted by the noun phrase is fam to both speaker and hearer. (12) a. a well-dressed man with d glasses. First. The children also looked vaguely familiar (8) I had to get a taxi from the station. b. (6) The moon was very bright last night. consider (3) –(12): (3) Just give the shelf a quick wipe. (7) An elegant. or they may be standing looking at it together in her dr or it may be that the hearer has not yet seen the car in the speaker’s possess but was aware that she had been looking over a particular car in a showroom recen Examples like these have led to a view of definiteness known as the famil ity hypothesis. 2 This is something of a simplification. and first full presentation is in Christophersen (1939). and the discussion I give here owes much to this acco As further illustration. (4) Put these clean towels in the bathroom please. or whether it is ra speakers who sometimes refer using them. if definites can refer or be used to ref is less clear that reference is involved in the case of indefinites like a car here. and a is used where the speaker does not want to nal such shared familiarity. I immediately ognized the woman. Neverthe the distinction drawn captures a clear intuition. The familiarity hypothesis has a long history. (11) The bloke Ann went out with last night phoned a minute ago. The fact that you’ve known them for years is no excuse. and is being newly introduced to it. The major recent work in tradition is Hawkins (1978). On the way the driver told there was a bus strike. (2) wo be used where the hearer knows or has seen the speaker’s new car.both participants. . and two children entered the compartment. One would typically utter (1) where the car in question has place yet in the hearer’s experience. t is dispute over whether definite noun phrases can be referring expressions. She may at the wheel right now.

the reference to the prime mi ter would normally be taken to be to the prime minister of that country. and any of these thi can then be referred to by means of a definite noun phrase. and it is part of our general knowledge that taxis have drivers. But travelling from New York to most places necessa involves some form of conveyance. as in the following exchan (13) A: B: An old man. In (3) the situation is the immediate. but is familiar in the sense being known to exist and probably known by report. but it could equally w occur in part of the discourse spoken by another person. or as a use in which fam iarity stems from general knowledge. with an aircraft being the most likely if present conversation is taking place in. the fare etc. in a particular country. Manchester. 4 . the hearer would m naturally take it that the reference is to the bathroom of that house. The ide that the mention of a taxi conjures up for the hearer all the things that are as ciated with taxis (a driver. In (7) we have examples of anaphoric the. and be thought of as a combination of the anaphoric and general knowledge types (8) the driver has not been mentioned before.). new referents introduced into the discourse in this form because they are so far unfamilia the hearer. (6) can be regarded as a uational use in which the situation is the whole world. though the referent of the defi noun phrase is probably not visible. wheels. vis one. In (5) the evant situation is wider. or to a unique entity forming part of the hear general knowledge. the c dren and the old man take the form of indefinite noun phrases. So the referent of driver is familiar through association with the antecedent a taxi. (9) is parti larly interesting because the antecedent which warrants the definite the plan not even a noun phrase. but there has been mention o taxi. seats. the individ concerned is not personally known to the hearer. Thus the moon is taken to refer to the partic moon associated with this planet. Did you recognize the old man? It is significant that in (7) and (13) the earlier mentions of the woman. say. the shelf is familiar to speaker and hearer in that it is before their eyes (4) the situation is still relatively immediate. they have been mentioned before. The referents of the woman and children are familiar not from the physical situation but from the linguistic c text. two women and several children were already there w I arrived. Examples (8) and (9) are bridging cross-reference or associative uses. in a particular house. In this example the previous mention ta place in an earlier sentence uttered by the same which the speaker and hearer are located contributes to the familiarity of referent of the definite noun phrase.

then it is possible for an associative use of the to be based following as well as preceding information. the familiarity of the bloke depends on the foll ing relative clause. so that of Ghana is a phrase added to prov clarifying information and not itself within the scope of the. where previous mention makes the referent fa iar (by report rather than direct acquaintance). but that is not the same as knowing this person. the a ciation appealed to in (9) is certainly real. and also informs him she went out with someone. many linguists basically sympatheti the familiarity thesis prefer to see definiteness as being about identifiability. in f the bloke can be replaced by a bloke. again following rather than preceding the definite noun phrase. The familiarity of the bloke then consists of its a ciation with this succeeding information. It may be already clear from this presentation that the concept of familiarit an explanation for the definite–indefinite distinction is not unproblematic. Finally. Assume that the hearer did not even know that Ann had g out last night. So th clauses. (7). consider (12). the hearer would normally be prepared to accept Ghana has a president. On this view. Here. It is fa straightforward for examples like (3)–(6). which are therefore anticipa anaphoric (or “cataphoric”) uses. In ( where cataphoric information is appealed to. prepositional phrase has the same function as the previous mention of a tax (8): to provide a trigger for the association that familiarizes the definite noun phr If this is correct. A similar treatment seems appro ate for (11). But ting from New York to Manchester does not necessarily involve flying. without changing the referent. The al native characterization involves taking the definite article to be modifying not p ident of Ghana. and even (8). In this example. that you’ve known them for years is the fac question. ac “antecedent” for the fact and the prospect. fact that Ann went out with a man is not expressed in the relative clause. which se to make it clear that the information in the relative clause is not such as to es lish the familiarity that would make the obligatory. where the fact taxis always have drivers acords the same sort of familiarity as in (7). nor even have heard of him. and having to cook for six for two weeks is the prospect. but just president. idea is that the use of the definite article directs the hearer to the referent of . but in (11). Because of considerations like these. The relative clause informs him of this. one can claim that the necessary fa iarity is established after the utterance of the definite noun phrase. but will know f his knowledge of the world that there probably is such an individual. but can one really say that the p was in any sense known to the hearer before the utterance of the second sente of this example? In (10). where the hearer is genuin acquainted with the referent.the president of Ghana personally.

ness does not altogether reject familiarity. The dicerence betw (14) and (3) is that. In the examples discussed above where familiarity seems rather forced. and the most straightforward identification is wit plane the travellers probably came on from New York. familiarity. sure enough. for example). and it is with this probable in vidual that the reference of the president is identified. where it is pres is what enables the hearer to identify the referent. In (9). Rather. Ann went out last night with someone. It does this by indicating that its referent be identified by the hearer. but this is inaccurate. He has to l 3 6 Note that the article itself does not identify the referent. In such cases the hearer is inv to match the referent of the definite noun phrase with some real-world entity wh he knows to exist because he can see it. it is g erally the case that the definiteness of the noun phrase confirms an associat which is only probable or possible rather than known. the journey m tioned makes the involvement of an aircraft likely. In (11) the relative clause provides a context in which a referent for bloke can be found. will you? Joe looks around and. without turning round. Back in the sitting-room which was the setting for (3). and. or infers its existe from something else he has heard. The article has been said by many writers to “ out” an entity. phrase which provides the probable referent occurs after the definite noun phr and is attached to it in such a way as to make the association certain rather t probable. and the referent of bloke is that someone. sees a hammer on a chair. and then the definite noun phr the plane authorizes the hearer to associate its referent with this journ confirming the possible association. So while on the familiarity account the tells the hearer that he knows wh on the identifiability account it tells him that he knows or can work out wh Let us now consider a case where an explanation in terms of familiarity wo be impossible. the may be about identifiability. but not identificatio . Ann is ing to put up a picture on the wall. A similar association involved in (10). the is a “grammatical word” wit descriptive lexical content. even though the relative does not provide any informat about the person (that it was a man. says to Joe w has just entered: (14) Pass me the hammer. whereas the hearer in (3) knows there is a shelf in the ro which provides an obvious referent for the definite noun phrase. But in this example. has heard of it. Joe does not kn at the time of Ann’s utterance that there is a hammer in the room. and therefore contains nothing which can itself identify a refe The most it can do is invite the hearer to exploit clues in the linguistic or extralinguistic text to establish the identity of the referent. Ghana probably has a president.

he can identify the hammer Ann is talking about, and the verb pass (which tend
take things immediately available as complement, by contrast with fetch, get, b
makes it almost certain that he will find it in the room. The referent of the defi
noun phrase is unfamiliar to the hearer, but he is able to find a referent for i

1.1.3 Uniqueness and inclusiveness
Identifiability certainly ocers a more comprehensive picture t
does familiarity, but there are also cases of definites for which an account in te
of identifiability is either not fully convincing or simply inadequate.
Associative uses of the definite article in general are problematic for identifia
ity; consider the following example:

I’ve just been to a wedding. The bride wore blue.

The definite reference the bride in (15) is successful because the hearer kn
that weddings involve brides, and makes the natural inference that the refere
is to the bride at the particular wedding just mentioned. But is it accurate to
that the hearer identifies the referent in any real sense? He still does not kn
who she is or anything about her. If asked later who got married that morning
would be in no position to say on the basis of (15), and if he passes the new
wed in the street the next day he will not recognize her as the person referred
Many situational uses are also associative; they work because the hearer is
to associate a definite noun phrase with some entity which he expects to fin
or associates with the situation. This is the case with the following:

[Nurse entering operating theatre]
I wonder who the anaesthetist is today.

A definite is possible because we take it for granted that operations involve an
thetists. But it is clear from what is said in (16) that the speaker cannot iden
the referent of the definite noun phrase, and does not necessarily expect the he
to be able to. Both participants know there is or will be such an individual,
that is not identification. The point becomes all the clearer if we replace the defi
article in (16) by a demonstrative:


I wonder who that anaesthetist is.

In the semantics literature, the term “description” is used of all material that ascribes pro
ties to entities – including nouns as well as, more obviously, adjectives. A particularly im
tant use of the word, especially in the philosophical literature, is in the term defi
description, meaning an expression which ascribes a property or properties to a particular e
– in other words, a definite noun phrase. The hammer, then, is a definite description.

is referring to a particular individual and expects the hearer to be able to pick
precisely which individual she means. This is not the case with (16), indica
that while demonstratives may require identifiability, definites do not.
Consider also a cataphoric case, where the definite article is sanctioned b
relative clause following the noun:

Mary’s gone for a spin in the car she just bought.

In (18) the relative tells the hearer something about the car (the fact that M
just bought it), but it does not help him identify it. He still would not know
car in question if he saw it (unless Mary was driving it).
What can be claimed about all these examples is that they involve the idea
uniqueness: the definite article signals that there is just one entity satisfying
description used. This uniqueness is generally not absolute, but is to be und
stood relative to a particular context. Thus in (15) there is just one bride at
wedding which triggers the association. In (16) the assumption is that there is
one anaesthetist taking part in the operation about to begin, but who it is is
known. And in (20) the conveys that Mary bought one car.
In the associative examples an indefinite article would seem unnatural, for v
ious reasons; in (15), for example, the general knowledge on which the asso
tion is based includes an assumed normal pattern of one bride per wedding.
in (18) it is perfectly possible to substitute a for the:

Mary’s gone for a spin in a car she just bought.

The most natural interpretation is still that only one car is involved, but the p
sibility is left open that Mary may have just bought more than one car. So
indefinite article does not signal non-uniqueness; rather it does not signal uniq
ness. Indefinites are neutral with respect to uniqueness (though this will
qualified below).
As observed, the uniqueness of the definite article is usually relative to a p
ticular context, but it can be absolute. This is the case with nouns which are inh
ently unique, denoting something of which there is only one. We can speak of
sun and the universe, but not normally of a sun or a universe; the qualificatio
important, because although for most purposes we think of our sun and our u
verse as the only entities to which those names apply, there are situations in wh
we might speak of our sun as one of many or entertain the possibility of th
existing another universe. Nouns like Pope are also often thought of as inher
uniques, because there is usually only one at any given time; but of course if
looks across history there have been many Popes, and with this perspective i
reasonable to speak of a Pope. The fact that one can always find a context in wh

possible to claim that the count–mass distinction is basically valid despite the p
sibility of recategorizing any noun (as in the count use of the basically mass m
in He strode up to the bar and ordered three milks), so there is a class of in
ently unique nouns. And such nouns, used as uniquely denoting, require the defi
Consider also the following immediate situation definite:

Beware of the dog.

This is intended to inform the reader that there is a dog in the vicinity, and
he is likely to meet it if he waits long enough or proceeds any further. One co
argue that identifiability is involved, in that if he sees a dog nearby he is likel
connect it with the one mentioned in the notice. But there is no expectation
he will seek a referent for the dog; rather, (20) is equivalent to There is a d
Uniqueness, on the other hand, does seem to ocer an adequate account here, si
an intrepid intruder could reasonably claim to have been misled if he found
had to deal with two dogs.
The uniqueness criterion is particularly attractive in cases where the referen
hypothetical, potential, or in the future:

The winner of this competition will get a week in the Bahamas


The man who comes with me will not regret it.

Assuming the competition in (21) is not yet over and no one has yet agree
accompany the speaker in (22), the winner and the man are certainly not
identifiable. But they are unique, in that a single winner and a single male c
panion are clearly implied.
Finally, there are certain other modifying constituents of the noun phrase wh
are incompatible with the indefinite article; among these are superlatives, fi
same, only and next:

Janet is the/(*a) cleverest child in the class.


You are the/(*a) first visitor to our new house.


I’ve got the/(*a) same problem as you.


He is the/(*an) only student who dislikes phonology.


I ocered a discount to the/(*a) next customer.

Uniqueness ocers an explanation for these facts, according to Hawkins (19
since the unacceptability of the indefinite article seems likely to stem from a sema
incompatibility between an element of uniqueness in the meaning of the mod

is neutral with respect to uniqueness, there are cases where choosing a rather t
the implies non-uniqueness; this is a point I will return to. For the moment it w
suace to look at it in this way: if the descriptive material in the noun phrase in
cates that the referent is unique, then the only appropriate article is the one
encodes uniqueness. This is the case with inherently unique nouns, and noun phra
containing superlatives etc. Cleverest means ‘cleverer than all the others’, and
means ‘before all the others’; so uniqueness can be argued to be involved here
it obviously is with only. In (25), if the hearer has a single problem, or a sin
salient problem, as seems to be implied, then the speaker can have only one pr
lem which is the same as the hearer’s. Next means ‘immediately following’,
given that customers are generally dealt with one by one, there can be only
customer who immediately follows the preceding one.
All the examples so far considered in this section have involved count no
in the singular. But the definite article can occur equally well with plural count no
and mass nouns, and the obvious question is: How can a definite noun phrase wh
is plural or mass have a referent which is unique (in the context)? The noun phra
the pens and the butter (the latter occurring with its usual mass value and
recategorized as count) cannot refer to just one pen and just one butter. Let us l
at examples corresponding to those examined above, but with plural (the (a) s
tences) and mass (the (b) ones) definite noun phrases:
(28) a.

We’ve just been to see John race. The Queen gave out the prize
We went to the local pub this lunch time. They’ve started chilling

(29) a.

[Nurse about to enter operating theatre]
I wonder who the anaesthetists are.
[Examining restaurant menu]
I wonder what the pâté is like.

(30) a.

We’re looking for the vandals who broke into the oace yesterda
I can’t find the shampoo I put here this morning.

(31) a.

Beware of the dogs.
Beware of the electrified wire.

(32) a.

We’re ocering several prizes, and the winners will be invited to Lon
for the presentation.
Fred’s decided to take up home brewing. He plans to sell the bee
his friends.

(33) a.

Janet and John are the cleverest children in the class.
This is the best muesli I’ve ever tasted.

(34) a.

You are the first visitors to our new house.
This is the first rain to be seen here for five months.



All the family used to take their bath in the same water.

(36) a.

They are the only students who dislike phonology.
This is the only water you’re likely to see for miles.

(37) a.

I ocered a discount to the next three customers.
The next water is beyond those hills.

As a first attempt at a solution, one might propose that uniqueness still app
but to sets and masses rather than to individuals. Thus the set or mass referre
by a definite noun phrase is the only set or mass in the context satisfying the desc
tion. But this does not work. In (28a), suppose there are three prizes. These f
a set, but there is also the set consisting of the second and third prizes, that c
sisting of the first and second prizes, and so on; these, of course, are subset
the set of three – and this is the point. Our intuition about (28a) is that the Qu
gave out all the prizes, not some subset of the total; similarly in (28b), we assu
that all the beer at this pub is now served chilled.
This points us to the proposal that definiteness, at least with plural and m
noun phrases, involves not uniqueness but inclusiveness (a term due to Hawk
(1978) ). What this means is that the reference is to the totality of the object
mass in the context which satisfy the description. So in the (a) examples of (29)–
the reference would be taken to be to all the anaesthetists about to take par
the operation, all the vandals involved in the break-in, all the dogs guarding
property, and all the winners in the competition. In the (b) examples, it is to
the pâté on ocer in the restaurant, all the shampoo left there, the electrified w
surrounding the property as a whole, and all the beer Fred brews. In (33), J
and John are the only children in the class meriting the description cleverest,
the muesli praised is the totality of the muesli in the speaker’s experience des
ing to be called the best. I leave the reader to work out how inclusiven
accounts for (34)–(37).
It appears, then, that with plural and mass nouns the is a universal quanti
similar in meaning to all. As support for this position, consider the following
(38) a.

I’ve washed the dishes.
I’ve washed all the dishes.


No you haven’t, you’ve only washed some of them.

Our intuition is that (38a) and (38b) are equally false if there are still some dis
unwashed, and in that case (39) would be a reasonable retort to either. So the
some uses) and all are very close in meaning, and the dicerence between th
may be that all is simply more emphatic. But it seems unsatisfactory to say
the signals uniqueness with singular noun phrases and inclusiveness with pl

When the noun phrase is singular, inclusiveness turns out to be the same as uniq
ness, because the totality of the objects satisfying the description is just one.
example, the speaker in (16) is assuming there will be only one anaesthetist
the total number of anaesthetists assumed to be involved is one.5

1.1.4 Identifiability, inclusiveness and indefinites
The relationship of the indefinite article to identifiability and inc
siveness is rather complex. We saw in relation to (18)–(19) that a is neutral w
respect to uniqueness rather than signalling non-uniqueness. Where the is u
the referent has to be unique: Mary bought one car. A allows this same interp
tation, while also permitting an interpretation in which the car referred to is
of several. This picture is reinforced by sentences like the following:

I went to the surgery this afternoon and saw a doctor.

The doctor the speaker saw may have been one of several in the surgery, but
necessarily; (40) is perfectly compatible with there having been only one doc
But there are other cases where a signals non-uniqueness and the choice o
rather than the makes a significant dicerence:

Pass me a hammer.


Janet ran well and won a prize.

These sentences clearly imply that there is more than one hammer in the si
tion to choose from, and Janet won one of a number of prizes.
It appears that, while the logically entails uniqueness with singular n
phrases, a is logically neutral with respect to this. But it carries a weaker im
cation of non-uniqueness. How this can be formalized will be discussed more f
in Chapter 7, but for the moment the point is that a may imply that the refer
is non-unique, but this implication may be overridden. This can be illustrated
relation to (42), where it is possible to add material indicating that the referen
after all, unique:

Janet ran well and won a prize – the only prize in fact.

It should be noted, however, that with singulars the is not (near-)synonymous with all. We
the table to refer to the only table in a certain context, but all the table denotes every pa
the table, and *all table is not a well-formed noun phrase.
There have been various reformulations of the basic insight behind inclusiveness, for exam
as “maximality” (Sharvy 1980, Kadmon 1987): the reference of a definite description is to
maximal set satisfying the description.


in (42) the indefinite a prize strongly implies non-uniqueness, this implicatio
less central to the meaning of a than is uniqueness to that of the: (44) is m
less acceptable than (43):

?Janet ran well and won the prize – one of several in fact.

We can summarize by saying that when a referent is inclusive in the cont
the is normally used rather than a, because a implies non-inclusiveness (equ
lent to non-uniqueness, since a only occurs with singulars). This implication
be overridden, however, making it clear that non-inclusiveness is not an en
ment of the indefinite article. When the referent is not inclusive in its contex
must be used and the may not be, as (44) makes clear.
But there is a further use of a. This is when the entity referred to is not a
ciated with what I have been calling a “context” – a physical situation or the
vious discourse. First-mention uses exemplify this:

I met a lion-tamer this morning.

Where there is no contextual set (other than perhaps the whole world) within wh
inclusiveness may or may not apply, a is used. And in such cases it seems to
a matter of non-identifiability rather than non-inclusiveness. The referent is ta
to be unfamiliar to the hearer because it has not been mentioned before. Thi
probably also the best explanation for the indefinite in (19), where the refe
may well be unique. This use is probably not to be seen as cataphoric (the rela
postmodifier establishing a domain for uniqueness or familiarity); rather, the en
noun phrase, including head noun and relative, is treated as non-identifiable,
is therefore indefinite.

1.1.5 A unified account?
We have seen that familiarity can be subsumed under identifiabi
and that uniqueness is merely a special case of inclusiveness, resulting from
singularity of the noun phrase. So the question now is: Can we make a ch
between identifiability and inclusiveness? Is one of them right and the other wro
or are there two kinds of definiteness?
In section 1.1.3 we examined a number of uses of the which can be accoun
for by inclusiveness but not (or not very convincingly) by identifiability. Ind
the inclusiveness account could be extended to many other examples that are
problematic for identifiability. If certain uses can be handled by inclusiveness
not by identifiability, and other uses can be handled by either hypothesis, then in
siveness must be preferred. Consider for example (14), repeated here:

We have. But inclusiveness works equally well. (47 similar. So both identifiability and inclusiveness work here as explanati for the appropriateness of the. and has a suitcase in each hand] Open the door for me. And in (49). and they cannot be adequa accounted for by inclusiveness. with all the other students contributing w In (46). please. will you? (49) A: B: [Two academics] How did the seminar go? Fine. the nature of a sem makes it clear that one student in the group stood out as having a special ta and this individual will be taken to be the referent of the student. definite reference would have failed. If on looking around hearer saw three hammers. is examining a large nut. but the speaker’s state of preparedness for a jo ney makes it obvious that the street-door is meant. All these exa ples represent perfectly normal uses of the. (48) [Ann. The speaker is dres in coat and hat. But for the moment let us settle for a v 14 . However. please. The student gave an excellent presentation. (47) [In a hallway where all four doors are closed. an shall return to the matter in Chapter 7. fixing her motorbike. then. This is not a satisfying conclusion. some which can only be accounted for by inclusiveness. the door is not unique.This example is unproblematic for the identifiability account. The hearer was previously aware of the presence of a hammer in the room. and some wh both theories account for equally well. but takes it from definite noun phrase that there is one which he can identify. he might have to ask back which one was meant. but i easily identified because of the verb – you can only close an open door. are three spanners. the door referred to is not unique. because the hearer could not identify the erent. In (48) the hearer is expec to be able to work out that only one of the three spanners can possibly fit the and identify that as the intended referent. Behind her. some usage types which can only be accounted for by ide fiability. an immediate situation use. two of them obviously far too sm for the nut] Pass me the spanner. one of which is open] Close the door. out of reach. there are cases where identifiability works and inclusiveness does Consider the following: (46) [In a room with three doors. Identification is thus only possible in (14) if the desired hammer is uni in the context. which generate really good discussion. he looks around finds the referent.

This is by no means an isolated instance of confusion of identifiability uniqueness. A way to test this possibility would be to look languages in which dicerent kinds of “definiteness” are expressed in dicerent w by dicerent articles for example.2. that the two properties are in princ independent of one another. to cover non-adjectival noun phrase modifi such as this. 1. Some of these are definite determiners. it either represents a failure to appreciate the dicerence between unique and identifiability. but that neither identifiability nor inclusiveness is the correct cha terization. Ann. the two theories make quite distinct claim It is also possible that what we are calling definiteness is in fact two or m distinct semantic categories. every shop. several. One should be misled by this. . Leech says tha use of the conveys that there is a referent that can be uniquely identified by speaker and he He adds that “uniquely” means that “we should be able to select the one X concerned from other X’s”.2 Types of definite noun phrase This section surveys.reference of a noun phrase is characterized by either property. The following noun phrases all have much in common sem tically with definite noun phrases containing the: that man. then that noun ph should be definite. however. these houses. and still using mainly Eng data. they. our. 1. without commitment to the grammat category of the items concerned. We will come back to these speculations. even if in many examples the presence of one lows from the presence of the other. which happen to have the same lexical or morp logical realization in English. Ven my car. characterizing definites as “uniquely identifying”. all writers. still informally. or “identifying unambiguously good example is the account of definiteness given by Leech (1983: 90–3). dicering f the in that they combine definiteness with some other semantic content. There is nothing here about the referent being unique (in the context). Bear in mind. But he refers back to this characterization as “the uniqueness implicature associated with which ensures that the postcard I got from Helen last week implies that there exists one such postcard. the range of noun phrase types which have definiteness as part of their me ing. Including the definite article in a noun phrase is not the only way of mak definite reference to some entity. There are several other kinds of noun phrase wh appear either to express the inclusiveness of the referent or to indicate that referent is identifiable. Or it may be that definiteness is a unified p nomenon. a friend’s house. assuming for the present definiteness is a single category. all. or is merely equivalent to “identifying”. 6 A number of writers present definiteness in terms which suggest an attempt to combin reconcile the two accounts. us.1 Testing for definiteness Articles like the and a are part of the larger class of determine another term used here informally.

For example. and not too much reliance should be pla on them. 16 . Is there a dictionary in the house? ?Is there the dictionary in the house? These diagnostics are not foolproof. b. so ot definite noun phrases. in which [+ Def ] is present as a consequence of. (50) a. It has long been recognized that a number of syntactic environments either not admit or admit only with some diaculty a definite. which/some/all of the women *which/some/all of (some) women (53) a. I will in fact argue in later chapters that definiteness (though not necessarily dem strativeness) is a grammatical category. b. or conversely an indefin noun phrase. The house is mine. to determine whe a given noun phrase type is definite or not. b. happen to be joi encoded by a portmanteau morpheme. These environments have been used as diagnostics. he couldn’t lift it. (52) a. and similar counterexamples be devised to some of the other tests: 7 Although writers are commonly noncommittal as to whether features are grammatical (m phosyntactic) or semantic. For the moment. or oth wise in combination with. not a semantic one (though it is related semantic/pragmatic concept). Indeed nearly all discussion in the litera of definiteness and related concepts like demonstrativeness takes these to be semantic or p matic. the structure in (50) admits an indefinite noun phrase in preted generically (as referring to an entire class). e if neither inclusiveness nor identifiability seems to be involved in the interpretation. another feature. in principle independent of one another. so that a noun phrase with one of these determiners is definite. I shall follow the established practice representing the grammatical and semantic content of lexical items and morphem by features. I assume that [+ Def] is a defining charact tic of certain determiners.definiteness or is incompatible with indefiniteness. b. the constructions in (50)–(53) were first exploited by Postal (19 and that in (53) is discussed in detail by Milsark (1979).4 below. my discussion so far (with examination of usage types and the atte to characterize definiteness in terms of notions like inclusiveness and identifiability) imp that I take the features I introduce to be semantic. Other determiners may be ch acterized as [+ Def ] along with other feature specifications. ?Big as a boy was. ?A house is mine. This assu tion will be important in 1. are complex definites.7 I have adopted term “simple definites/indefinites” to describe noun phrases with the or a. (51) a. So the is [+ Def ] and a is [− Def ]. The following examples show h some of these “definiteness ecects” distinguish between simple definites and sim indefinites. Big as the boy was.2. he couldn’t lift it. Or it may be that two sem tic properties.

2 Demonstratives Demonstratives are generally considered to be definite. Clever as you are. and noun phrases with all do not readily occur at all in titives. c. c. which/some/all of us *Is there him here yet? (60) a. d. d. d. which of cannot be followed by a proper noun because th are always singular. Strong as every contestant is. Fido is mine. and th diagnostics ocer a rough guide. Nevertheless. I bet you won’t solve it. c. d. but it is c that their definiteness is not a matter of inclusiveness. proper nouns. b. possessives. which/some/all of those women ?Is there that dictionary in the house? (57) a. c. Big as my cousin is. I am yours. (?which)/(?some)/(*all) of all the men ?Is there every visitor here? The partitive structure in the (c) examples shows some deviations from the tern seen with the. c. she won’t be able to reach it. b. A sentence like Pass that book is likely to be used in a context where there is more than one poss referent corresponding to the description book. That house is mine. Make yourself at home – my house is yours. In the following sections we will look more closely at dem stratives.(55) A house is mine if I pay for it.2. they range these clearly with the: (56) a. b. he couldn’t lift it. Big as that boy was. for reasons which are not obvious. Tall as Nuala is. b. personal pronouns. and determiners like and every. d. b. Otherwise the only point of note is personal pronouns are even less acceptable in the there is/are construction t are other definites. (*which)/some/all of Paris ?Is there Peter in the house? (58) a. which/some/all of the students’ essays ?Is there Rachel’s racket in here? (59) a. 1. If applied to a range of putative comp definites. they’ll never shift it. the examples (50)–(53) do draw a recognizable distinction. and the utterance may wel . he can’t lift it. All hats are yours.

between the actual referent and other potential referents. This tance is not necessarily spatial. it may be temporal (that day referring to so past or future occasion. meaning the present we or emotional (There’s that awful man here again. This is used refer to some entity which is close to or associated in some way with the spea or with a set of individuals which includes the speaker. others. the hearer. On view so far adopted here. N this concept of a set of one or more individuals which includes the speaker is definition of “first person”. I and they are interested in’. But identifiability is only part of the semantic content of demonstratives. But in 18 . or I’ve read this book. This that are often termed proximal and distal demonstratives. it corresponds to the pronouns I (set of one) and (set of more than one). But i possible to relate this distance contrast to the category of person. the letter may have already been the subject of discussion betw speaker and hearer. Demonstratives like this and that are deictic because they locate the en referred to relative to some reference point in the extralinguistic context. for example. That is used where the referent is associated with a including hearer but not speaker (second person) or a set including neit speaker nor hearer (third person): (61) Show me that (?this) letter you have in your pocket. In uttering Pass me that book. and the speaker may have been previously thinking about amazing drill she has recently heard one of her friends has acquired. The c trast between this/these and that/those is to do with distance from the spea this book denotes something closer to the speaker than does that book. the speaker assum that the hearer can determine which book is intended. (62) Tell her to bring that (?this) drill she has. T are often grouped with the varied class of words which express deixis. as opposed to this week. by contrast with Pass a book or I’ve read a book (the former not involving any intended referent). respectively. among other possibilities. but it would imply that the lette drill is in some way associated with the speaker (or a set including the speak Thus. ‘the article which you and I are discussing’ ‘the article which you. Deixi the property of certain expressions and categories (including tense and gramm ical person) of relating things talked about to the spatio-temporal context. identifiability is what links demonstratives with the defi article. or that between the speaker. This is certainly possible in these examples. and particular to contextual distinctions like that between the moment or place of ut ance and other moments or places. What about this present promised me?).Hawkins (1978) argues that demonstrative reference always involves a contr clear or implied. For discussion of these possibilities see Lakoc (1974). so this article could ‘the article which I am reading’.

that is the appropriate demonstrative: in (61) the letter is in the possession of
hearer (second person), and in (62) it is someone not present in the discourse
uation (third person) who has the drill. So it would be reasonable to speak of
as a first-person demonstrative and that as a non-first-person demonstrative.
An important question is whether this deictic element is the defining cha
teristic of demonstratives, distinguishing them from the definite article. An a
mative answer is given by many writers. Sommerstein (1972) and J. Lyons (19
1977) assume that the distance component is the only thing that distinguishes
and that from the (so that this marks proximity, that marks distance or non-p
imity, and the is neutral with respect to distance); and Anderson and Keenan (19
comment that a demonstrative system with only one term “would be little dic
ent from a definite article”. But there are good reasons for rejecting this view
English that is sometimes neutral with respect to distance or person, as w
used pronominally in relative constructions:8

She prefers her biscuits to those I make.


I want a coat like that described in the book.

Taking the deictic opposition to be expressed by a feature [± Prox], then ei
these demonstrative occurrences do not carry this feature, or the negative v
[− Prox] characterizing that and those can include neutrality with respect to
tance. On the assumption that the too is either unmarked for [± Prox] or red
dantly [− Prox], there must be some other feature distinguishing the and that
More striking evidence is acorded by languages which have a demonstra
unmarked for any deictic contrast which is either the only form in the system o
distinct in form from the terms which are deictically marked. Egyptian Arabic
basically a one-demonstrative system with no deictic contrast: da ‘this’ or ‘t

Stockwell, Schachter and Partee (1973) regard that and those in such sentences as forms o
definite article rather than demonstratives, because of a supposed complementary distribu
between these forms and the usual simple definite form – that and those occurring with m
and plural value before a post-nominal modifier, and the one occurring with singular count no
They would disallow the ones for those in (63) and that (as opposed to the one) in (64)
they are simply mistaken in this. The claim of complementary distribution fails particu
clearly in the case of plurals like (63), where the ones and those are equally acceptable. Stock
Schachter and Partee do have a point as regards singulars like (64), for which the distribu
they present does represent the general tendency; and there is no obviously simple definite a
native to that relating to a mass noun phrase (as in I prefer this butter to that you got from
market), unless we propose that the “free relative” what is this alternative (I prefer this b
to what you got from the market). But the fact that the complementary distribution claimed is
partial and optional indicates that an analysis in which the demonstrative forms really are dem
stratives cannot be ruled out. But if we take the forms in (63)–(64) to be true demonstrat
it is evident that they do not express any degree of distance or association with person.

from the speaker, but such information is not lexicalized in the demonstrative s
tem. But da is not a definite article; Egyptian Arabic has a definite article hil,
tinct from the demonstrative. Another case is French, where distance is expres
by a suax -ci or -là on the noun, so that the information usually thought o
belonging to the demonstrative is divided between two morphemes: ce bateau
‘this boat’, ce bateau-là ‘that boat’. But this suax can be freely omitted, so
no information about distance is conveyed: ce bateau ‘this/that boat’. But ag
ce is still distinct, in form and meaning, from the definite article le.9
So deixis is a usual but not invariant property of demonstratives. There m
be some other property, then, that distinguishes them from the definite article.
us call it [± Dem].10 So this/these is [+ Def, + Dem, + Prox]. Views dicer v
widely, however, on what the [± Dem] distinction amounts to; put diceren
writers who agree that demonstratives have some distinctive property apart fr
deixis are divided on what it is.11 Hawkins (1978) claims that demonstratives
distinguished by a “matching constraint”, which instructs the hearer to match
referent with some identifiable object, that is, some object which is visible in
context or known on the basis of previous discourse. This constraint captures
observation that, whereas with a simple definite the referent may be inferrable
the basis of knowledge of the world (as in (65), cars being known to h
engines), with a demonstrative the referent must be given in the linguistic or n
linguistic context – thus the impossibility of (66):

I got into the car and turned on the engine.
*I got into the car and turned on this/that engine.

Hawkins’s matching constraint looks rather like identifiability, though a more
stricted notion of identifiability than is involved in definiteness. See also Macla



Harris (1977, 1980) argues that ce is rivalling le as definite article, the latter form tendin
become an unmarked determiner with the function of carrying agreement features (num
and gender), but this process is certainly not complete yet. The sentences Passe-mo
marteau ‘Pass me the hammer’ and Passe-moi un marteau ‘Pass me a hammer’, in imm
ate situation use, show the same contrast as the English glosses.
The conclusion that the deictic content of demonstratives that have it is not what distingui
them from the definite article becomes inescapable in view of the fact, which will emerg
Chapter 2, that there are languages in which non-demonstrative definite articles show these d
tic distinctions.
It is important to be aware of a certain variation in the literature in the use of the term “deix
I have been using it here to denote the distinctions of proximity to the speaker or associa
with a particular person which make it possible to locate entities in the context of uttera
relative to others. But the term is also often used to denote the basic property common t
demonstratives, the property expressed here by the feature [+ Dem]. On this use, “deictic
equivalent to “demonstrative”, being applied to forms which “point out”.


demonstrative signals that the identity of the referent is immediately accessibl
the hearer, without the inferencing often involved in interpreting simple defin
This may be because the work of referent identification is being done for the he
by the speaker, for example by pointing to the referent. The deictic feature t
cally expressed on a demonstrative plays a similar role to pointing, guiding
hearer’s attention to the referent. This suggests a necessary connection betw
[+ Dem] and [+ Def ], the former implying the latter. I take demonstratives, th
to be necessarily definite.

1.2.3 Proper nouns
The term “proper noun”, or “proper name”, is applied to a very
erogeneous set of expressions, including John, The Arc de Triomphe, South F
Road, some of which have internal grammatical structure and contain descrip
elements, and some of which do not. I limit my attention here to names like J
and Paris, which (though they have an etymology) are not generally thought o
having any descriptive semantic content, or as having any meaning independ
of the entity they name. The name Paris is applied to a particular city, but t
us nothing about that city; of course we would normally expect the bearer of
name John to be male, but it is argued that this is not part of the meaning of
name. By contrast, the common noun man, in being applicable to a particular i
vidual, is so in virtue of the fact that that individual satisfies certain descrip
criteria (being human, male, adult). Proper nouns are often said to be refer
expressions but to have no sense. They are also sometimes said to be logic
equivalent to definite descriptions, in being uniquely referring expressions.
There may be millions of people called John and there are several towns ca
Paris, so context is important for the identification of the referent, as with defi
descriptions. But a common view is that we use proper nouns as if they w
absolutely unique, corresponding more closely to inherently unique definites (
the sun) than to possibly contextually unique definites (like the man). When
are conscious of there being more than one possible referent for the name J
we can either expand it to a fuller proper noun (John Smith) or recategorize i
a common noun and add some descriptive material (the John I introduced yo
last night). I shall assume this to be correct as an outline of how proper nouns
used; for a detailed and clear discussion of the complexities of proper nouns
J. Lyons (1977: 177–229).
It is clear that the uniqueness of reference of proper nouns is what aligns th
with definites, though it may be added that this very uniqueness will gener
ensure the identifiability of their referent. But a number of questions arise at
point. How do proper nouns dicer from inherently unique nouns like sun? T

entity, but they dicer grammatically: sun behaves like a common noun in tha
takes the article, or some other definite determiner (the sun, that lucky old su
John, unless recategorized, generally does not, and in fact is not only a noun,
also a complete noun phrase. One answer is that nouns like sun denote single
sets, while proper nouns denote individuals; this would be in keeping with
view that proper nouns have reference but not sense. Another, implying that pro
nouns do have sense, is that both types of noun denote singleton sets, but in
case of sun the set just happens to have only one member, while the set satis
ing John is by definition a single-member set.
This latter proposal goes some way towards answering the question why pro
nouns in English do not take the definite article. If by definition they denot
singleton set, there is no need to signal the uniqueness of their referent. But
they then in fact definite, or merely semantically similar in some way to defi
noun phrases? In other words, is the feature [+ Def ] present in proper nouns,
if so, where? If proper nouns are [+ Def ], this feature would appear to be on
noun, given the lack of a determiner. But it seems to be clearly a determiner
ture in common noun phrases. It would be preferable to be able to say that
definiteness feature occurs in one place only, and in general the determiner see
the most probable locus (unless we say a grammatical feature can have its lo
in a phrasal category, so that it is the noun phrase, not the noun or the determi
which carries [± Def ] ). If we assume that the feature [+ Def ] pertains only
determiners, it may be that proper nouns are accompanied by a phonetically n
determiner, or that the feature does not after all appear on proper nouns. One p
posal along these lines is Lyons (1995c), where it is argued that proper no
in English are in fact indefinites, and their apparently definite behaviour com
from their being generics – generic noun phrases anyway (whether definite
indefinite) showing similar distributional behaviour to definite non-generics. T
idea will be taken up again in Chapter 4.
In a number of languages, at least some types of proper noun (most typic
personal names) do regularly take the definite article. Examples are Classical Gr
(ho S$krat@s ‘Socrates’, h@ Hellas ‘Greece’), Catalan (l’Eduard, la Maria). It m
be that these names should be treated as no dicerent from uniques like Eng
the sun – and thus unlike English proper nouns. It is clear that they are defi
noun phrases, with [+ Def ] encoded in the determiner.

1.2.4 Possessives
Under this heading I include determiners like my, their (together w
their pronominal forms mine, theirs), and also “genitive” forms like Fred’s,
woman’s, that man next door’s. These genitives are clearly full noun phrases,

other possibilities) appears right at the end of the whole phrase. Possessives
my are also formed from noun phrases, since they are derived from personal p
nouns. These possessive forms of noun phrases occur as modifying express
within other noun phrases, as illustrated in (67), where each of the bracketed exp
sions is a noun phrase (or a derivative of a noun phrase):
(67) a.

[ [my] cousin]
[ [Fred’s] only friend]
[ [that man next door’s] car]

Now, in English at least, possessives render the noun phrase which contains th
definite, as shown by the diagnostics introduced in 1.2.1. And the phrases in (
could be roughly paraphrased by the expressions the son/daughter of my aunt
uncle, the only friend Fred has, the car belonging to that man next door – cle
definite noun phrases, beginning indeed with the definite article.12 The s
applies to many other languages; as a further example, with a dicerent word-o

It might be supposed that what makes these noun phrases definite is that the noun phrase un
lying the possessive expression (me, Fred, that man next door in (67) ) is definite – the defi
ness of the embedded noun phrase is somehow transferred to the matrix noun ph
Precisely this explanation has been ocered by a number of writers for the definiteness o
corresponding structures in Semitic languages; we will return to this in Chapter 3. This w
mean that if an indefinite noun phrase (such as a woman) were made the basis of the po
sive (thus a woman’s), the matrix noun phrase (a woman’s drink for example) would be indefi
And this may seem to be the case, since such noun phrases occur fairly readily after exi
tial there is/are:

There’s a woman’s drink on the shelf here.

But this impression is mistaken. Noun phrases like a visitor’s hat, a friend of mine’s co
containing an indefinite possessive, are definite, and are naturally paraphrased by the cle
definite the hat of a visitor, the cousin of a friend of mine. The impression of indefinite
comes from the fact that examples like a woman’s drink are structurally ambiguous, as
cated in (ii)–(iii), where the possessive phrase is bracketed:

a [woman’s] drink


[a woman’s] drink

The structure we are interested in here is the one in (iii), where the possessive is the indefi
a woman’s. The matrix noun phrase is definite, natural paraphrases being the drink belon
to a woman, the drink left behind by a woman, etc. In (ii) the possessive expression wom
is probably not a full noun phrase, and is therefore not indefinite. The indefinite article
not part of the possessive expression, but is a modifier of the matrix noun phrase and acco
for its indefiniteness. The sense of (ii) is something like ‘a drink suitable for women’. F
detailed discussion of this distinction between “inner genitives” (as in (ii) ) and “outer g
tives” (as in (iii) ) see Woisetschlaeger (1983). The observation stands, then, that in Engl
possessive noun phrase, whether itself definite or indefinite, renders its matrix noun ph

nominally (mo/do hata ‘my/your hat’), and those based on “full noun phras
are genitive case forms occurring to the right of the head noun (hata an fhir (
the+GEN man+GEN) ‘the man’s hat’). In neither case can the head noun be modi
by the definite article (*an mo hata, *an hata an fhir), yet the matrix noun phr
must be understood as definite; these examples are equivalent to ‘the hat belo
ing to me/you/the man’. But in many other languages possessives do not imp
a definite interpretation on the matrix noun phrase. Italian has only pronoun-deri
possessive determiners (mio ‘my’ etc.); possession with full noun phrases
expressed prepositionally (il libro di Carlo (the book of Carlo) ‘Carlo’s book
a structure which does not seem to impose definiteness in any language. Now
Italian translation of the definite my book is il mio libro (the my book), in wh
the definiteness is conveyed by the article il; this article may indeed be repla
by the indefinite article un, and the matrix is then indefinite: un mio libro (a
book) ‘a book of mine’. The point is that the presence of the possessive mio
no bearing on whether the matrix noun phrase is definite or indefinite. The s
ation is similar in Classical Greek, which uses genitive case forms for full n
phrase possessives. To express the definite ‘the man’s horse’, the possessive n
phrase ‘the man’s’ (itself definite and therefore having the definite article) is p
ceded by the article which renders the matrix noun phrase definite, with the re
that two articles occur in sequence: ho tou andros hippos (the+NOM the+G
man+GEN horse+NOM).
This dicerence between English and Irish on the one hand and Italian and Gr
on the other is discussed by Lyons (1985, 1986a), in terms of a typological
tinction between “DG languages” (English, Irish) and “AG languages” (Ital
Greek).13 The dicerence between the two types is claimed to reside in the str
tural position occupied by the possessive; what it amounts to essentially is tha
DG languages possessives appear in a position reserved for the definite article
other definite determiners, but in AG languages they are in adjectival or some o
position. This claim is controversial, and I shall return to a more detailed disc
sion of it. For the moment we can say that in some languages a possessive indu
definiteness in the matrix noun phrase while in other languages it does not. T
traditional assumption that possessives are definite determiners, stated without
ther comment in many descriptive grammars and in much recent theoretical w
– presumably because possession is assumed to entail definiteness – is misguid
It reflects a lack of awareness of the AG phenomenon. We have seen above

“DG” and “AG” stand for “determiner-genitive” and “adjectival-genitive”, respectively (tho
it is not claimed that possessives are necessarily determiners in the first type and adjective
the second).


is semantically incompatible with indefiniteness. But this is not the case w
possession.14 We shall return to the “definite constraint” on possessives in
languages in Chapters 3 and 8.
There is one circumstance in which this definite constraint can be suspended
least in some DG languages; this is when the noun phrase is in predicative posit

Mary is Ann’s friend.


I was once Professor Laserbeam’s student.

The noun phrases indicated may be understood as either definite or indefin
There is not necessarily any implication that Ann has only one friend
Professor Laserbeam has only had one student; rather, the bold phrases in
and (69) are likely to be interpreted as equivalent to a friend of Ann and a
dent of Professor Laserbeam. But notice that this same interpretation is poss
even where the predicative noun phrase with possessive modifier is marked as defi
by the article: I was once the student of Professor Laserbeam.15
There is also a use of non-predicative noun phrases with possessive modifi
tion which cannot be characterized as either inclusive or identifiable:



This is easily demonstrated, even limiting the discussion to DG data. Note first that po
sives do not necessarily express possession in the sense of ownership. The phrases John’s ho
John’s club, John’s annoyance, John’s son may mean ‘the house John lives in’, ‘the club Jo
a member of’, ‘the annoyance John feels’, ‘the boy John has fathered’. Possessives exp
merely that there is a relationship of some kind between two entities (represented by the
sessive noun phrase and the “possessum” – the matrix noun phrase). The nature of that
tionship may be dictated lexically by the head noun, if it is relational or a body part for exam
or it may be determined pragmatically, the context and background determining whether Jo
team is the team John owns, the team John supports, or the team John plays in. Now, ass
for the sake of argument that John’s car is ‘the car belonging to John’ and John’s team is
team John supports’. If “possession” is semantically incompatible with indefiniteness in
possessum, the phrases a car belonging to John and a team John supports should be sem
tically anomalous. But they are not. In fact English has a possessive construction which ca
indefinite, the “postposed possessive”: a car of John’s, a team of John’s. So the fact that a
head possessive in English imposes a definite interpretation on the matrix noun phrase
nothing to do with the semantics of possession.
This phenomenon is probably to be distinguished from the ambiguity as regards definite
of predicative possessives with no overt possessum. In That pen is mine, mine is either m
the predicative form of my or a pronoun. If it is the latter, it is a definite noun phrase (‘
pen is the one belonging to me’) because the definite constraint applies to possessive prono
If it is the former, then the definite constraint does not apply, because the possessive is
part of a noun phrase (‘That pen belongs to me’). In Spanish these two readings woul
clearly distinguished, and indeed the definiteness of the pronoun clearly marked by the ar
Esa pluma es mía (predicative non-pronominal possessive), Esa pluma es la mía (pron
nal possessive).

though unique or identifiable. but they have also long been recognized as defi and are often referred to as “definite pronouns” (by contrast with indefinite p nouns like one and someone). 1. some. I’ve just torn my sleeve on that bramble. There is no implication in these sentences that the speaker has only one bro and Joe only one leg. but w makes them interesting is that they are not specific to English. nevertheless.(71) Joe has broken his leg. that the “possessed” body part. several and num ous other items can occur both pre-nominally (as determiners) and independe 26 .5 Personal pronouns The personal pronouns are traditionally so called because t express grammatical person. and generally a limited number of brothers).2. and i perhaps unimportant which of this number is at issue. given the definite article. is referred to by a definite noun phrase. inalienable and ot intimate possession. The referent is in each case one of a small number (people h only two legs or sleeves. The point is made all the m clearly here.’ In fact this example does not contain a possessive. Lyons (1977). Since true indefinites can be used. that. Compare the lowing French example: (73) Jacques s’est cassé la jambe. there is reason to consider the bold noun phrases in (70)–(72) to be indefinite. though with some modifi tions first proposed in C. Such an account has obvious attractions. and family and other personal relationships – thus. (72) Oh. all. or that the hearer is in a position to tell which of the speak sleeves has been torn. It is not at all clear what lies behind these facts. Postal (1970) proposes to account for the defin ness of personal pronouns by deriving them transformationally from defi articles. each. one. and I will adopt the essence of this account. so this. neither inc siveness nor identifiability applies. In all these cases. It see rather that what we are dealing with here is definite noun phrases used where conditions for definiteness do not strictly hold. an indefinite (a brother of mine) o partitive (one of my sleeves) could have been used. French generally using the defi article and a dative pronominal form to express possession with body parts (t literally ‘Jacques has broken to himself the leg’). but the definite structur usually preferred. This use is limited to noun phra where the head noun is one of a small lexical class denoting mainly body pa clothing. In general English pronouns are p of determiner–pronoun pairs. ‘Jacques has broken his leg.

. The fact that normal third-person pronouns are (at least in colloquial stand excluded with such modifiers (?him who/that Lucille divorced. I know the one with brown hair. but surfaces in the third person when the noun phrase contain restrictive modifier: (77) a. dicering from only in respect to person. but *Pass me the). which are always pronominal (Pass me it. it and they are the pronominal correlatives of the determiner the (which is th fore third person). I bought the green one. and third-per personal pronouns. but occurring both as determiners and as pronouns. b. but *Pass me it bo A neat solution to this oddity would be to pair these together. you does not oc freely as a determiner (*You linguist are relying on some pretty odd data). . German permits both the exclama use and the more general use with ich ‘I’: (75) Ich Esel! I donkey ‘Silly me!’ (76) Ich Vogelfänger bin bekannt bei Alt und Jung . I met the one who/that Lucille divorced.and second-person articles in the presence of a preposed modi us clever ones. I don’t trust you politicians an inch. . at least in part. I birdcatcher am known to old and young . s pletive variation is limited to the third person. The latter is g erally deleted. b. In the singular. *I idiot!). a language-specific fact. Postal derives pronouns from an underlying full noun phrase consisting o article plus a minimal noun head (with the form one in English). . I can suggest no reason for this restriction. The first-person gular pronoun does not occur at all in English pre-nominally (*I wish you’d le me foreigner alone. is further support for the analysis. So we and you can be regarded as forms of the definite article. *it green). and say that he. This idea is strongly supported by the observation that the fi and second-person plural pronouns we and you also occur pre-nominally: (74) a. We Europeans are experiencing some strange weather patterns. bu is. b is so used in exclamations: You idiot!.occurs as a determiner (Pass me the book. You lucky bastard!. The hypothetical noun one shows up with first. ?her with br hair. Sommerstein (1972) points out that while we and can be stressed or unstressed (with a reduced vowel) when used pronominally ( . c. Contesting this account.

and I’m glad we now this one. because of distributions like the followin (78) a. In fact the identity of sense example (80) is also situational that the demonstrative gets its reference from the immediate situation. and bought t (one). and absent in non-anaphoric use. This ties in with the point m by Hawkins (1974) that pre-nominal we and you have as their third-person c relative. consistin a pronoun with a full noun phrase in apposition with it. optional in strict anaphoric use. (80) I could never follow our first coursebook. the strongest/only soldiers in the army *we/you/these strongest/only soldiers in the army The point that personal pronouns often have more in common with demonstrat than with the is taken up by C. This claim is based on a case of complem tary distribution between demonstratives and third-person pronouns. There are two kinds of anaphoric use. Thus the redu form [ja] may occur in You fool!. the only di ence being that the appositive noun phrase is indefinite. (79) Jim finally found a flat with a view over the park. who argues that Postal is basic right but that the English personal pronouns are forms both of the definite art and of the [− Prox] demonstrative. but demonstratives. 28 . but not in (74b). which falls out naturally on their analysis since boy is not a possible singular indefi noun phrase.stressed and with a full vowel. b. who see we men as identical in structure with we. N 16 Hawkins’s claim is that noun phrases like we soldiers are appositive structures. where it is near-obligatory in identity sense use. The various uses and forms of English demonstrative pronouns must first distinguished. strict anaphora as in (7 where the demonstrative is coreferential with a previously occurring noun phr and identity of sense anaphora as in (80). Then there is the non-anaphoric use. This is also the position of Delo and Dougherty (1972). except in the exclamatory use. But their own analysis fails equally to account for the impossibility (in stan English) of they men. but we n to distinguish this from non-anaphoric situational uses like the following: (81) Bring that along with you. with the zero determiner norma plural indefinites. the men. They point out that Postal’s account fails to explain the impossibility of gular *I boy. typically indicating something in the im diate situation. not the. Lyons (1977). One only appears in singular count use. where there is no coreference but previously occurring noun phrase supplies the understood descriptive conten the demonstrative.

2. and the cap ized forms as stressed: (82) a. THE and IT here are replaced by that (or this). they. e. these demons tive forms can be regarded as belonging to the deictically unmarked. or perh [− Prox]. SHE lo fed up. What is impossible is the kind of demonstrative stress l to be accompanied by pointing or some equivalent gesture appropriate to (83). it is natural to c clude that the stressed forms of these are also demonstrative – first. and for non-human reference would be replaced by th (or these). in the third person: singular pre-nominal that. *Take a look at it. There is thus a neat complementary distribution between third-per personal pronouns and demonstratives. also represent forms of a demonstrative.use can only have non-human reference (while these/those can be human in id tity of sense use). b. *Take a look at the car/teacher. plural pronominal they (human).and seco 17 The and it are both capable of bearing stress. Take a look at her. she. she. The forms of demonstrative are. when stressed. in standard Engl THEM must be human. not one of the owners and I just met Julia walking her poodle. Moreover. to express contrast. sug ing that a demonstrative used in this way is neuter: Who is it?. her. singular pron inal he. represent pronominal forms of the definite article. c. It’s Jill’s husband. Human reference in sentences like (79) and (81) would be m by stressed him. Among other qualifications to the pattern observed is the fact that this and that do o with human reference in subject position in copular sentences: Who’s this?. d. her or them. The claim is that these pronouns used in this are forms of a demonstrative. as in I’m owner of this house. for example. th (non-human). Take a look at him. but in addition so of them. d.2. plural pre-nominal those. him.17 The conclusion is that the third-person pronouns he. demonstrative already identified for English in 1. c. Further evidence comes from the following p digms. Since they not encode any degree of distance or proximity to any person. the noun phrases are to be understood as non-anaphoric. Take Take Take Take Take a a a a a look look look look look at at at at at the car/teacher. e. but IT was chirpy enough. the latter filling certain gaps in stres occurrences of the former. b. Take a look at them. that. . it. it. That’s Jill’s band.and second-person pronouns. then. w unstressed. (83) a. But this exception may be only apparent because it is used in the same way. them. Turning to first.

When unstressed they are forms of a defi article. involving some of the most cen issues in syntactic theory. which s that these expressions (labelled anaphors) must be “bound” by an antecedent. for these plural forms as well as for I and singular you. and these are not subject to binding.18 I take personal prono generally. is straightforwardly accounted for. th selves) and reciprocals (like each other) are handled by Principle A. And in non-exclamatory uses the nominal simple definite article only occurs in the third person: The linguist(s) but ( unstressed pronunciation of the determiner) *I/you/we linguist(s). and in some. to be determiners. and can appear with full or reduced pronunciation fairly freely. but *T gangster is afraid the police will shoot himself ).simple definite. and if they are the antecedent cannot be in th “governing category”. you and we show the full range of occurrence available to the third-pe forms. What distinguishes personal pronouns from full definite (or demonstrative) n phrases is their lack of descriptive content (beyond partly descriptive gramm cal features like gender in some forms). We cannot o in the exclamatory use: You fools! but *We fools!. then. Thus. stressed occurrences they are form a demonstrative. in The gangster shot him. in the simplest cases. in gene where the associated descriptive content can be readily recovered from the course or the non-linguistic context. dividing these into three catego accounted for by three binding principles. but it is important to realize binding is a technical concept defined partly in terms of structural configurati 18 There are gaps in the paradigms of the deictically neutral or [− Prox] demonstrative an the definite article. The theory of binding sets out the anaphoric pos ilities of dicerent types of noun phrase. but probably not all. the pronoun him may t its reference from some noun phrase occurring earlier in the discourse. 30 . called the anaphor’s “governing category”. But there is much more to pronouns than t and their grammar is enormously complex. which is in the same minimal clause. but it may sometimes be a noun phrase or other phr and can be a complex clause. Reflexive pronouns (like myself. that the anaphor and its antecedent must both appear within a particular syn tic domain. Ordinary personal pronouns the second category of nominal expression. And the demonstrative-like behaviour of pre-nominal we and y shown in (78). in their nominal use I. an anaphor and its anteced must be in the same minimal clause (thus The gangster shot himself. We have seen. these too are demonstrat the corresponding third-person form being those. cannot function as its anteced Principle C is concerned with all remaining referential noun phrases (full n phrases essentially). By contrast. and Principle B says they may be bo by an antecedent but need not be. The governing categor typically the minimal clause. of cou that definite full noun phrases may be anaphoric. but gangster. A pronoun is therefore used.

Hausser (1979).binding theory. Bound variable pronouns are dependent on (or “bou by) a quantifying expression (such as one expressible in terms of the logici universal quantification). (85) Every student thinks they have passed the exam. Lazy pronouns are so called because they relate somewhat sloppily to an a cedent. and you could see he was pleased with himself. see Haegeman (1994: chapter 4). agreeing with it in descriptive content rather than referential identity 19 It is nevertheless important not to overstate the dicerence between personal pronouns full definite noun phrases. though singular. Work on the semantics of pronouns has identified further complexities in t behaviour which might be taken to argue against their being merely determin without the descriptive content found in full noun phrases. These examples are anaph in that the pronoun has an antecedent. Reinhart (1983). with behaviour resembling that of third-person personal pronouns: (i) Finally John arrived. every girl and every student. becoming increasingly comm nowadays where the antecedent is of mixed gender. and the configurational relationship between pronoun antecedent is not the one defining binding (because the antecedent does c-command the pronoun: for explanations of c-command see Radford (19 chapter 3) and Haegeman (1994: chapter 2) ): (86) Every man who bought a car crashed it. be an important factor in behaviour as regards binding. As well as the strai forward referential uses of pronouns in which they get their reference ei from a linguistic antecedent or from the situation. and. (ii) Finally John arrived. we can distinguish the bou variable use. and the lazy use. but the antecedent of the pronoun is itself a vari bound by a quantifier. do not have a specific referent rather denote a range of individuals. the E-type use. For example: (84) Every girl thinks she should learn to drive. There is in fact a class of full definite noun phrases sometimes ter “epithets”. . but this antece defines a range of entities and the pronoun refers to each of these individually type pronouns are similar. and you could see the lad/bastard/little darling pleased with himself. in fact. Note that they in (85) is the vague singular use. For discussion of these and of semantics of pronouns more generally see Cooper (1979). The inherent definiteness of personal pronouns may. Ev (1980). typically with acective content showing something of the speaker’s attitude to referent.

2. The claim stands that pronouns are definite noun phrases minus description – thus determiners.6 Universal quantifiers Determiners like all. in terms of description. it was subseque reported that the house was sucering from subsidence problems (89) Everyone who bought a house discovered too late that the house riddled with damp. being a pronou also needs a full noun phrase “antecedent” to permit recovery of the description to be assig to its referent. 20 Strong as most contestants are. Most hats are yours.1. but need not concern us here. in at least some of the diagnostic environme (90) a. It gets its reference from association with the bathroom. to my neighbour’s car. Kempson does not suggest this.1. they were quite clean. For discussion see McCawley (1981: 98–101). my car serves this function. that some other determiners which do not express tota show similar behaviour to all etc. however. either within a context or absolutely. The lazy use seems to be a purely pronominal phenomenon. But while they refers to windows. But m of these uses are not. In fact these determiners dicer in imp tant ways from the logician’s universal quantifier. since they express to ity. and from each other. on the basis of the preceding noun ph the windows. ali ing them with definites. The pronoun it here refers. The suggestion is. the bridging cross-reference or associative anaphoric discussed above in 1.20 but otherwise p sonal pronouns are strongly parallel in behaviour and range of use to definite noun phrases. they can’t shift it. it does not refer to the same windows. restricted to pronouns. 1. is paralleled by a peculiar to full definite noun phrases. in fact. is to be related to inclusiveness. in Kempson’s discussion this use. The pronoun is interpreted.his drive. as the windows gets its reference from association with the kitchen. on the other hand. t that it in (87) is similarly an associate of my next-door neighbour. It is to be noticed. but it is tempting to argue that pronouns are the pronominal equivalent of bridging cross-reference. b.2.2. Examples like the lowing suggest a case can be made for this view: (i) I glanced into the kitchen and saw that the windows were filthy. not mine. Kempson (1988) shows full definite noun phrases can have the bound-variable and E-type uses: (88) Of every house in the area that was inspected. every and each can be thought of as appro mating to universal quantification in logic. peculiar to pronouns. th what makes it a lazy pronoun. ?Are there most visitors here? Interestingly. of course. 32 . in the b room. It see obvious that the behaviour of these expressions in the diagnostics in 1. But. c.

together with a. But the object noun phrases in the following sentences also indefinite. some and eno are examples.1 Indefiniteness and cardinality An important question is whether indefiniteness is a function of presence of certain indefinite determiners in a noun phrase. 1. and of course a and are mutually exclusive. These indefinite noun phrases do not contain the indefinite article however. But it is clear they do not encode [− Def ]. (92) I wonder if Helen has read many books. or whether it is sim a matter of the absence of definite determiners. and look more closely at the indefinite article it and its role in the expression of indefiniteness. and are then (at least where interpreted generically) indefinite: . These cardinality terms are obviously neutral with respect to (in)definiteness. This has led to the suggestion that the in siveness of the definite article is one case of a broader concept of quantificat characterizing determiners which denote a proportion.3. 1. so there are indefin which do not contain the indefinite article. The indefinite article a sugg the former. as opposed to a p portion – have been characterized as indefinite determiners. involving neither identifiability nor inclusiveness: (91) I bought three books this morning. with indefiniten being signalled by some other determiner.3 Simple and complex indefinites The article a is the obvious signal that a noun phrase is indefin but just as definites need not involve the definite article. as indefi determiners. they can co-occur with definite determin (93) Pass me those three books. Th are some determiners which only appear in indefinite noun phrases. We here consider the range indefinite noun phrase types. But what makes the noun phrases in (91)–(92) indefinite is absence of any definite determiner. a number or an amount. Indeed determiners like three and m which denote cardinality – that is. on the assumption that a encodes [− Def ]. perhaps they should be considered to be complex indefinites. This is an issue I will up in Chapters 6 and 7.– as indeed does the if it is inclusive. and these could reasonably be categorized. Notice also that count nouns in the plural mass nouns can occur without any determiner. (94) I’ve only read a few of the many books she’s written. Unlike a.

can appear in definite noun phra the one orange. A is also numeral-lik only occurring with count nouns. with a full vowel. in singular count noun phrases – odd limitation – and this has led many grammarians to suppose that its plac taken in examples like (95)–(96) by a “zero” variant of the indefinite article. not a/one bit. Th is in fact good reason to classify a and sm as cardinality expressions. In view of these facts. Another determiner traditionally regarded as being a plural and m variant of a. In fact it is necessary to tinguish stressed some. it is clear that there is a particularly close relationship between them. I prefer to regard a and one as distinct items. plural and mass noun phrases do not need to take sm to be indefin as seen from (95)–(96) (and (91)–(92) ). is some. and it is which partially complements a and is reasonably taken to be an indefinite arti But again. or at least as complementing a. in fact a does not co-occur with eit the definite article or with any numeral. In these languages. 34 . one must ask whether a and sm really are indefinite a cles. who arg that a is derived from one by a phonological rule of reduction operating in absence of stress. perhaps even num als). notice that even singular indefinite count noun phrases do not necessarily tak one orange. in fact. whether or not it has an indefi determiner.21 This position is strengthened by the fact that in many languages the sema equivalent of a is identical to the equivalent of one: German ein. So the idea that there is in English a zero pl and mass indefinite article finds little support. The indefinite article a only occurs. a/one mile away. French un. A is deri historically from the same ancestral form as the numeral one.(96) I’ve already put spoons on the table. although descriptive grammars do tend to make a distinc between an indefinite article and a singular numeral. from a form which is usually (tho probably not always) unstressed and with reduced vowel. and. it is a cardin ity term like three and many. which aligns with vague dinality terms like much/many rather than numerals. this one ticket. si we can have noun phrases with neither. the reality seems to be that a n phrase is indefinite if it has no definite determiner. So can occur with singular count nouns (There’s some man at the door). a/one week from now. This is not because one is an indefinite determiner. They are not in full complementary distribution with definite determiners. though this is not true of sm. with little dicerence of mean a/one fifth. and while few wri now recognize a synchronic link. but if they assigned to the same lexical class (both being cardinality terms. they are mainly distinguis 21 There are many contexts in which they are interchangeable. though the and numerals can co-occur. thus pronounced [sa – and often represented for convenience as sm (a convention I shall adopt). an exception is Perlmutter (1970). Turk bir. like them.

unstressed variant of the singular numeral. is obliga 22 The distribution of sm in relation to plural and mass noun phrases such as those in (95)– is a complex and controversial matter. For analysis the problem is to explain why a particular cardinality term. red dantly. is either unavailable or at least diacult to get. Consider the following: (i) I’m looking for a record. the specific reading. where a can be contrastively stressed and is then realized as not as one. commonly available with indefinites. three etc. (iii) I’m looking for records. are termed specific and non-specific. while has some additional content besides. A thus encodes no more than [+ Sg]. Sm can be thought of as expressing a va quantity. with a as the default form (the one that must occur if no other do So for the Perlmutter analysis the problem is to explain why some cardinality exp sion is required if there is no definite determiner but optional otherwise.22 The treatment of a as a cardinality expression. that the speaker has particular records in mind she hopes to find. it may therefore stand in op sition to a. to be a cardinality term. while a i contrast with ‘more than one’. But whereas a obligatorily accompanies sing indefinites in the absence of some other determiner. the same ambiguity appears in (ii). this one car. Pairs of reading this sort. We will return to a more detailed discussion of this in Cha 4. a. . while attr ive. I suggest that the semantic dic ence between them is that one is in contrast with two. but (if the noun is count) more than one. that sm i plural correlate of a. mass noun phrases). and for the present it will suace to note that in some contexts a plural or mass noun ph may occur either in bare form or with sm with little dicerence in meaning. Turning to the plurals. It has been claimed. thus the one house. this is apparently not so sm and non-singular indefinites. by Carlson (1977). I take a. does leave unexplained the fact that it cannot co-occur with the or other defi determiners. whe in the absence of a definite determiner a cardinality expression is obligator the singular. Sentence (i) is ambiguous between a reading on which the speaker is looking for a partic record and one on which she does not have a particular record in mind. closely related to one. On Perlmutter’s view that a is simply the unstressed f of one. encoding [− Sg] (this specification applying both to plurals and. But one is purely optional here. one might claim that this numeral can only be unstressed and reduced w initial in the noun phrase. But (iii) is dicerent. whereas in o contexts there may be a more substantial dicerence. and the “bare plural” (as in (96) ) is something radically dicerent – an alternative indefinite plural. perhaps even a kind numeral. with the bare plural. so that after another determiner the full form is ap priate. therefore. but a distinct lexical item from one. *the a house or * a car should be possible. But the dicerences are more subs tial in English. res tively. Since numerals can follow these determiners. (ii) I’m looking for sm records..

A definite noun phr must normally contain a definite determiner. and I adopt the label cardinal article for a and sm. an expression of cardi ity. The basic. The and a are both phonologically weak forms.2 Complex indefinites I use the term “complex indefinite”. cardinality. Indefinite noun phrases characterized by the non-occurrence of a definite determiner. The is. No though that the suggested phonological constraint blocking the appearance o and sm after another determiner ensures that these items only appear in indefi noun phrases. reflected in their phonological weakness default behaviour. unmarked nature of the and a. and cardinality is intrinsic less central. si the noun carries no number inflection. th the definite article. to mass expressions. But I suggest the issue is complicated b phonological constraint that weak forms can only occur initially in the phras The result is that in definites the initial definite determiner blocks the expec appearance of a in second. and this shared charac istic can be taken to reflect their both being the unmarked or basic members their respective classes. while not itself encoding [− Def ]. and a is the default form.3. 1. I shall describe the further label quasi-indefinite article. The fact that this can be dispensed with much more easily in plural and m noun phrases may be related to the fact that plurals do have an expression cardinality in the plural inflection on the noun. if indeed relevant at all. but such sequences are not impossible an 23 The major exception to this is that “pre-determiners” such as all and half can occur be articles: all the way. For this reason they can be said to signal indefiniteness indirec An article that does this. half an hour. to denote n phrases in which some determiner other than one of the quasi-indefinite artic seems to compel indefiniteness and render a definite determiner impossi Cardinality terms other than a and sm tend to be compatible with definitenes well as indefiniteness. the is the basic definite determiner and a is the basic c dinality determiner. in principle. Several is rarely used in company with a definite determiner ( several trees. but both defin and indefinites should normally contain. I shall take to be what defines the term article. These basic items act as default forms. position. those several visitors). and the is the one that occurs in absence of some other with more semantic content. 36 .miner. a cardinality determiner is in princ required. With singulars. But there are some cardinality expressions which show ei a strong preference for indefiniteness (several) or incompatibility with defin ness (enough). with their minimal semantic con ( [+ Def ] and [+ Sg] respectively). informally.

and we will discuss some below in relation to this tinction. the adjective cer compels the specific interpretation available anyway to a. The pronoun corresponding to a is one illustrated by the following:25 24 25 It is not obvious why this is so. We shall see that determiners corresponding closely to th are identifiable cross-linguistically. wenig-e Freude or wenig Freude ‘little joy’. such as viel ‘much’. like any count noun. But if the one in (97) were this same min noun we would expect it to be preceded. Here a pink one is a complete noun phrase. *der viel Kum die wenige Freude. by the article a. There are some o cardinality determiners in German. Now genug (which is equally excl from definite noun phrases) also has the peculiarity of being indefinites. but does not care what is taken to satisfy the predic To express specificity unambiguously English has a certain.2. *die wenig Freude. Rather. which suggests that the incompatibility is not entirely sem tic. w can occur pre-nominally either with or without the normal inflection: viel-er Kummer or Kummer ‘a lot of trouble’. Note that the one in (97) is distinct from that in (i): (i) I’m looking for a white hydrangea but I can only find a pink one. We have observed that English a and sm can specific or non-specific. But enough is incompatible with definiteness (*the enough su *those enough cups). and th impossible. we have seen in 1. A synonym such as sudcient does seem to be acceptab a definite context (though barely). . Determiners in English can be used pronominally as well as pre-nominally. but there seem be no reason to see this as a separate determiner. Most prominent in this gr are some (as distinct from sm) and any. many languages showing indefinite express forming a cline of specificity. discu above. ‘few’.5 that there can be substantial dicerence in shape betw pronominal and pre-nominal forms. These determiners are tradition labelled “indefinite”. But these cardin terms must be inflected if preceded by a definite determiner: der viele Kummer. ‘many’ and wenig ‘little’. and clearly has nothing to do with a.24 There are other determiners which are not self-evidently cardinality expressi and which occur only in indefinite noun phrases. This is speculation. And this suggests the incompatib with definiteness of enough may be an idiosyncratic syntactic property of this lexical i left over from a constraint (similar to that in German) applying in an earlier stage of En when adjectives were inflected. and I will not pursue the point. There may be some syntactic constraint peculiar to the item enough behind the restric This determiner is unique in being able to follow the noun (enough money or money enou a characteristic shared by its German equivalent genug. and indeed must be serious candidates for consideratio encoding [− Def ]. So the impossibility of genug after a definite d miner may be linked to its inability to take inflection. with a as determiner and one as the head n This is the same minimal noun as in the blue one or the one who Lucille divorced. But it is clear that any expresses a kind of extreme non-specificity which the speaker does not merely have no particular entity in mind correspo ing to the description.

I’ll have some or I don’t want any.Turning to some and any. please? I don’t want any noise. b. but they cannot be regarde full. 38 . it is not obvious whether someone and something correspond to so or sm. except that the pronom form of sm is unreduced. Pronominal some and any are restricted identity of sense anaphoric use. These complex forms combine the determiner with a noun.27 But while anyone and anything clearly correspond as pronouns to the de miner any. as when Mary is silently oc cake and says Yes. The anaphoric expression relates to an antecedent with which it sh descriptive content. Pronominal some can only be mass or plural. which is a major of indefinite pronouns. like thus the occurrence in (99a) cannot be interpreted as singular count. thanks. as the following make clear: (100) a. not some impor one (which would only occur as an identity of sense anaphoric use of some. noun phrases which happen to be represented orthographically as si words. You haven’t taken any. together with sm. implies that the referent is unfamiliar to the spea thus non-specific. however. The fact that someone is singular count suggests it includes some. but rather a kind of immediate situation use. non-pronominal. and something big as distinct from the equally possible but not synonymous full n phrase some big thing. (99) a. So with a singular count noun. Joan met some man today. b. but the two are not coreferential. But there are restrictions. Could you pass me sm broccoli. pronom any also tends to be mass and plural. these determiners. continue to use term loosely to cover such cases. Note that with identity of sense anap the “antecedent” can be provided by the non-linguistic context.26 In other functions they are replaced by fuller for someone/somebody and anyone/anybody for human reference (singular on something and anything for non-human reference. ?I’ve spent the afternoon reading some book. can function p nominally and pronominally with no change in form. c.2.5 above). thus some: (98) a. I will. This is clear from the position of adjectives: someone important. But someone and something show the same ambiguity with regard to specifi as a and sm: 26 27 Sentence (97) illustrates identity of sense anaphora (see 1. b. not but there is a striking dicerence in interpretation between someone and some. of course. Give me some. There’s some book lying on the table. This. not necess human). is not re anaphora.

These articles between them co the range singular. T seem to correspond in fact to sm and a jointly. plural and mass. This seems to indicate that only non-assertive is indefinite. ??There is/are any book(s) in this library. but in posi declarative sentences. b. c. (103) a.). here. You can borrow any book you wish. d. I’ll eat anything. I’m not choosey. and. c. The fact that someone and something cannot be plural can attributed to the incorporated noun. Yet while sm is mass or plural. indicating an u stricted choice. I’ve drunk enough. Have you bought any vegetables? I don’t want any wine. in particular. b. The main uses of any and the pronouns based on it are illustrated by following: (102) a. where it tend be preferred to sm. where it must be given the random interpretat it is of very low acceptability. I illustrate from the there is/are construction: (104) (105) a. and random any definite.b. Have you had anything to eat? I don’t think John’s in his room. like (104). while rand any has to have some degree of stress and is phonologically full. at least. Anyone could paint a picture like that. Any is perfectly good in non-assertive sentences with there is/are. which would presumably mean we have . b. So these pronouns do not correspond closely to either some or sm. and they show the ambiguity between spec and non-specific. On the other hand the two uses distinguished behave dic ently with regard to the diagnostic environments for definiteness. Sentences (102) illustrate what I shall term the “random” sense. I’m looking for someone/something – but I can’t find anyone/anyth I like. Examples (103) show the use of any in interrogative and nega contexts (which we can term jointly “non-assertive” contexts). Non-assertive any tends to be unstressed and may be redu (with pronunciations like [nP b] or [nb]. Are there any books in this library? There aren’t any books in this library. This suggests that n assertive any may represent no more than a less emphatic use of random any ra than a distinct sense. someone is singular. are perfe good in interrogative sentences like (103a) and (103c). But sm and rel forms are also possible in non-assertive contexts. and similarly [nP bθbi] etc. and something is singula mass. I can’t hear anyone.

b. having originated in normal post-article adjective position. come to be in pre-article position as a re of a movement process. I don’t like those boots. three such people). and a complete adjective phrase. but in this p tion (which could be definite determiner position) it can only have the random in pretation. so/how/too clever a m The fact that a appears here in second position. in that it never modifies definite noun phrases. We need consider only the second use. b. and. any one man). Mary is such a clever girl. b. Such has two p cipal uses. and that indefiniteness amounts to absence of a [+ Def ] determi still plausible. It follows most ca nality expressions (one such person. th sentences could be used to refer to a type of car and a style of boots exempli by the entity indicated (thus equivalent to that type of car or a car of that ty and that style of boot or boots of that kind ). Let us examine one further determiner which is a candidate indefinite status. a variant of a and sm restricted to negative and interrogative conte occupying the same position in the noun phrase as other cardinals. occurring the same “slot” as definites like the and that. 40 . I was given such lovely flowers. meaning something like ‘of that ki or ‘like that’. so that demonstrativeness is one element in its mean But it is interesting to observe at this point that noun phrases generally can h a “variety interpretation”. But in (107) it modifies the n (an interpretation also possible in (106) ). it will be notic involves a demonstrative element. and modifies the adjective. I’d love to have such colleagues. Any seems to dicer in position from cardinal being able to precede numerals (any three books. Now notice that a car of that t and boots of that kind are indefinite. suggests strongly that such and so clever etc. Whatever the syntactic status of such in (106). I wish I could acord to buy that car. which. (107) a.relates with another observation. Consider the following sentences: (108) a. The discussion so far seems to leave the hypothesis that no determiner is speci [− Def ]. moreover. which we have seen it is generally unab do.28 We could simply say that such means ‘of kind’ or ‘of this kind’. So it may well be that random any is a definite determiner. it is (on one interpretation) sem tically a variant of so. As well as comments about a particular car and a particular pair of boots. illustrated in the following sentences: (106) a. while non-assertive any is a ca nality term. We don’t need such a man here. very close in meaning to s 28 Such is almost certainly an adjective. but behaves like adjectives acc panied by degree modifiers in preceding the article a: such a man.

XP specifier YP X′ X complement ZP In addition. Interesting as this suggestion is. according to which a phr category. This would mean that demonstrativeness is not. to adjoined to some or all of the projections of X (X itself. Between these two is an interm ate level X′. The question of how noun phrases are structured will be examine more depth in Chapter 8. inherently definite. This gives the following schema. the position of specifier (wh is therefore sister to X′). as I h so far assumed.interpretation of a demonstrative. we will in Chapter 3 that it does not seem to be supported by cross-linguistic evide It may. and superficial (S-structure) order may dicer from the underlying orde D(eep)-structure. but this and occur only in definite noun phrases while such is the demonstrative form use indefinite noun phrases. and prop that such is an indefinite demonstrative. The specifier and complement posit are occupied by phrasal categories. depending on its lexical properties. but it will be useful to establish some preliminary not at this point. an XP. it is attached to it in this way: . such is synonymous w this/that (though not showing the proximal–distal distinction). XP immediately dominates. is projected from a head X. 1. X′. The general view within generative syntactic theory is that the structure phrases is determined by the principles of X-bar theory. But perhaps one can go yet further.4 The noun phrase No investigation into the nature of definiteness can proceed far w out consideration of the place of articles and other determiners within noun ph structure. merit further investigation. besides X′. it is possible for other expressions. On this proposal. If A adjoined to B. typically phrasal categories. however. And the head may take one or more complement sister. in the lin order generally appropriate for English. and XP). linear order varies from language to guage.

with determin in specifier position. and corresponds more or less to N′ on the older analysis. but of the determiner (Det or D). but this is within DP. modifying expressions such as adjective or adverb phrases have b taken to be adjoined to X′ or XP in the phrases they modify. It is therefore DP. Given this general framework. the reader is referred to Radford (1988: chapte and 5). This is that the “noun phrase” is a phrasal projection. adjectival expressions adjoined mainly to N′. not of noun. the usual view until recently was that the n phrase is a maximal (that is. A dicerent account has recently come to prominence. and any co plements typically expressed by prepositional phrases. phrasal) projection of the head N. not NP. and achieved almost g eral acceptance. Some examples will m this clear: NP Det N′ N the car NP Det N′ AP N′ N the treacherous murder PP of the prisoners For a clear and much fuller exposition of the general principles of X-bar theory NP structure as just outlined. as complement of the h D. projected from N. There still ex a category NP. The general princip 42 .A B B A For example.

The following exemplify DP structure: DP (specifier) D′ D NP (specifier) N′ N car this DP (specifier) D′ D NP (specifier) this N′ N PP murder of the prisone The position of adjectives has been the subject of much debate. each projecting a phrase. They have b claimed to be adjoined to NP. These grammatical categories as opposed to “lexical” or “substantive” categories N(oun). and the DP analysis extends this tren the “noun phrase”. states. relations etc. Other functional X′ is not permitted. A(djective). properties.). In clause struct functional categories like T(ense) and Agr(eement) (representing subject–verb ag ment) are treated as heads of phrases. but a currently influential view is that they are speci of some phrasal category between NP and DP. P(reposition) (which can be thought of as deno real-world entities. activities. examples are Num(ber) and K (for case . taken to be a projection of the functional category of D(e miner). V(erb). have also been claim to be involved in nominal structure. The DP analysis arises from an interest in “functional” categories.

when in this position it allows the noun phrase to be defi without the. Given the absence of the. which is formally either NP DP depending on the framework. the king’s every whim. All and both are sometimes termed “pre-determiners” beca they may appear before definite Dets. somewhat marginally. Every clearly oc pies the same position as the. *the every chair. every three chairs. *most those men. both your friends. For the present I will use the label Det to denote the mal category to which the English definite article and other determiners belo but will leave open the question of whether it is the head or the specifier of “noun phrase”. though it could be claimed to be in a cardinality posit deeper inside the noun phrase. the definiteness of domain of quantification must come from all. *all a potato. and it can precede numerals: *ev the chair. every one chair. and the definite quantifiers taken up in Chapter 8 (where I will propose a substantially modified vers of the DP hypothesis). and indeed an overtly partitive construction is av able as an alternative: all of the girls. to express a proportion of the wh denoted by the definite noun phrase: all the girls. but c not be a pre-determiner: most of those men. most tend to oc in this position too. I shall continue to term informally “noun phra English demonstratives and possessives apparently occupy the same slot in noun phrase as the. every. and it is probably in position in most men. The simplest explanation is that is in Det position. The overall phrase. All here is probably in Det. All three de dants does not denote the totality of an indefinite group of three defendants. The phr is equivalent to a partitive. 44 . rather than pre-determiner. follow possessives: his ev whim. position. But the two are equivalent. For reasons which far from clear. it cannot occur with it either as pre-determine (normally) in some more interior position. Most has been argued to be a definite Det. *both sm boys. This is also possible after the but only with a posses 29 The convention has grown up of using the label D within the DP framework. while Det is ferred within the older NP framework.29 The position of the definite article in English I will term n committally “Det position” – which is either specifier of NP or head of DP depe ing on the framework chosen. But t may occur with no following determiner: (109) All three defendants were acquitted. Most can also head a partitive construction. both. of a definite group. all and both (at least in the pre-determiner construction) do quantify over indefinites: *all sm girls. both of your friends. it is synonymous with all the three defendants. The only co plication here is that it can. since it co-occurs neither with definite Dets with cardinality expressions. not all of th defendants (if this is possible).

in English. I will assume. possible QP Dets are as. enough good books. as well as a strong tendency for these expressions to stand closer to noun than definite determiners. the margina of this construction lies in its being available only with a limited range of h nouns: *his every car. closely related to one. they are probably heads of egory Q. . She defines much. The tense–aspect distinction between past historic or preterite I read book and perfect I have read that book has sometimes been described in tr tional grammars in terms of a distinction between “definite” and “indefinite” p These may be mere labels. I shall assume the English noun phrase has a position for cardinality expressions between definite Det position and the position of adjectival modifiers. Det position is the locus not only for definite article but for definite determiners more generally. tho Bresnan does not include numerals in her treatment. If numerals and a are QPs. but oc more widely. but it is arguable that the preterite does make a defi time reference. In this they are like enough and sev On the idea of a cardinality position less peripheral than definite Det position see also Ver (1981).30 A general similarity in behaviour and distribu between numerals and other cardinality determiners is to be found in many guages. this label is taken from Bresnan’s (1973) study of the comparative construction. and that all the ab words appear in this position. it is natural to suppose that it occupies (o least originates in) the same position as numerals. *the every carriage of the k So it is possible to say that. Most writers assume that a occupies the same position in noun phrases as but this is because they take a to be an indefinite article. *the king’s every castle. which also contains a Det position. In the absence of a time adverbial which identifies the time of 30 It may also be that they belong to the same category – possibly a phrasal category QP (quan phrase). many. dicering from many in not admitting a Det.5 Definiteness beyond the noun phrase I end this chapter with the suggestion (which I will not pursue in rest of the study) that definiteness is not only a feature of noun phrases. but only in the presence of a possessive expression. as possible heads of QP. that syntactic theory ma available to languages a cardinality position which is more internal to the n phrase than Det position.only in Det position (the normal case) but also in some position deeper in noun phrase. little. the three/many/few good books. 1. Let us consider two possible instances of it. Numerals agree closely in tribution with various other cardinality expressions – all being mutually exclu and apparently occupying the same place relative to other constituents such as defi determiners (where they are compatible with these) and adjectives: the one g book. so (-er much/many underlying more). therefore. If it is in reality a ca nal article.

and its use in examples the following is close to that of the: (111) a. If this is correct. th is no implication that the hearer knows or can work out (or needs to) when event occurred. and in (111b) and (111c) it is provided i relative-like modifier. and their more formal counterpart is so. but. while the speaker may know when she read the book. with identification of time of event substituted for refer identification. the degree they convey of the property expressed by the adjective is acces anaphorically in (110a) and communicated by means of an ostensive gesture (110b). lacking only the deictic tinction. not exactly between tense and definiteness. so must also be a demonstrative Deg. Joe is as bright as Ann. the distinction is closely parallel to that betw the car and a car. As th near-synonym.event temporally on the basis of contextual knowledge.31 The structural position of determiners in the noun phrase is paralleled in adjective phrase (and in adverb and quantifier phrases) by “degree modifiers” (D as/so/that/too big. just as pronouns can relate to a referent introd in the previous discourse or to a referent understood on the basis of the context. but between t and (definite) pronouns. . I believe the Deg to consider for this characterization is as. and that this also occurs in both uses: (110) a. b. so tense relate to an antecedent time or to an understood time. a like this. T and that as Degs are colloquial. c. Tom is stupid but not that stupid. 31 46 Partee (1984) discusses parallels. These words can be treated as being also of category Det. as being either specifiers of AP or heads of a functional phrase contain AP. The fish I almost caught was this big! There is little reason to doubt that this and that have demonstrative meaning this use. Note in particular that that operates both as a definite Det in the noun phr and as a Deg in the adjective phrase. Joe is as bright. Joe is not as bright as you think. The perfect also prese the event as past. I phonologically weak. the obvious question is whether there is then also a s ple definite Deg. In (111a) the degree of brightness referred to is accessed by the hearer from context or the preceding discourse. She observes that. Put this way. b. with a normally reduced vowel. exactly parallel to what happens with noun phrase demonstratives.

to be obligatory (except perh under certain generally specifiable conditions) in the absence of other s expressions. and with the general behaviou the markers identified. We wil fact see that these expectations do carry over to many languages. On the basis of what we have s in English. one of the articles the sm. This chapter is concer then. In fact it turns out that semantic distinctions very c to those identified for English do occur widely. Of course it cannot be assumed that what resem definiteness in certain languages is exactly the same semantic category as obser in other languages. It thus covers aaxal defin ness markers as well as free-form determiners. In the next chapter we will extend this survey to “complex” definites indefinites. syntactic and morphological devices used by guages to encode the distinctions in question.1 Definiteness and indefiniteness marking Not all languages have definite or indefinite articles. we are now a position to survey the languages of the world to see how the definite–indefi distinction is expressed. we may expect articles more widely to act as default members of la categories of definite or indefinite expressions. at most. 2. but that t cannot all be taken as universal properties of articles. These are n phrases which correspond in terms of what they express. We begin this survey in this chapter by examining w I have called “simple” definite and indefinite noun phrases. to Eng noun phrases in which [± Def ] is signalled by. Bear in mind that the term “article” is being used here informally mean any linguistic form which has as its central function to encode a valu [± Def ] (or [± Sg] in the case of cardinal articles). and with a picture of the kinds of determ and noun phrase that are central to an understanding of definiteness. and to be unstressed and perhaps phonologically weak. and we must examine the semantic content as well as the f of the articles considered. but distinctions are also found wh divide the general semantic area in question dicerently.2 Simple definites and indefinites Having established a provisional (and clearly less than satisfacto conception of what definiteness is. if approximately. with the range of lexical. and we here c sider the extent to which the definite–indefinite distinction is overtly expres .

2. comparing Latin (which had no article) with Gr (which did have one). and two aspects of this situation are particularly strik One is that Bulgarian and Macedonian are almost alone among the Slavo 1 The minority of languages showing articles is. that between languages which have them and languages wh do not.1 The marking of simple definiteness is often an “areal feature”.1. O smaller scale. a classic example of a “Sprachbund” (an areal grouping of languag is the Balkan region. The first-century Rom grammarian Quintilian. may develop common characteristics. of course. and it may pure coincidence that Norwegian and Arabic both have a definite article. The greatest concentration of guages marking definiteness today is in Western Europe and the lands around Mediterranean (as well as parts of the world where languages from this reg have been planted through colonization in recent centuries – for example Engl Spanish and Portuguese in the Americas). there is an impress that the distribution between the two can vary. It is well kno that languages which are geographically contiguous. however. Yet many languages w do not mark simple definiteness can be argued to compensate by having other distinctions a similar function.1 The occurrence of articles in languages The most fundamental cross-linguistic distinction relating to the a cles is. incl ing features not especially common outside this region. Helle and Albanian) have been shown to share features of syntax and morphology. even genetically unrelated guages. This will be discussed further in Chapt 48 . not a small minority. Romance. All languages have demonstratives and personal pronouns. languages marking it are in a distinct minority. and one probably only speculate on why many languages express a distinction which many other not. so that some languages seem make more use of apparently definite noun phrases than others. This is debatable. Several Balkan langua belonging to dicerent Indo-European sub-families (Slavonic. I shall sugg some reasons for this impression in this section. and indeed most of the world gets along quite w without being obliged to distinguish consistently the article from an article. and here the situation is clearer. commented: Noster sermo articulos non desiderat ‘Our l guage does not need articles’. which are p haps inherently definite. so it could be claimed that the feature [± Def ] is rep sented in some form in all languages.which do consistently distinguish definites and indefinites. This area is a big one. Definiteness may be thought of as one of a number of categories which serv guide the hearer in working out how the discourse is structured and how entities referred t into it: markers of topic and focus come to mind here. One of the shared f tures is a definite article. Definiteness marking is obviously not essential to communication. But the encoding of simple (in)definiteness is far from univer indeed. and is a question I return to.

all the guages involved – Bulgarian and Macedonian. for it is not certain that what is represen is. Another example of the involvement of definiteness in areal facts concerns combined representation of definiteness and the direct object relation. leading to a strong intui . In languages which distinguish simple definites and indefinites. definiteness.Russian dialects). A good example is ocered by Erzya-Mordva and Moksha-Mordva. This be discussed in detail in Chapter 5. which are distinguished from their neighbours in having a defi article. w geographical areas where the marking of definiteness is absent. it is clear that languages develop a definiteness marker spontaneously. It is by no means the case. And given that this article is thought to be a Mordva deve ment rather than an inherited Uralic characteristic. closely related Uralic languages spoken in a wide area to the west of the U (Krámskm 1972. I have argued that English a and sm are cardinality words. The converse. there are th obvious ways in which the distinction can be expressed: (a) marker of definiteness only (b) marker of indefiniteness only (c) markers of both definiteness and indefiniteness But the matter is complicated by the phenomenon of quasi-indefinite card articles. Romanian (a Romance langua and Albanian – have a postposed article. and in showing this (as dist from definite object marking. but the phenomenon is concentrated in a geograph area (again a huge one) covering mainly the Middle East and Central Asia. The other is that. not indefi articles. strictly speaking. mentioned above) the Mordva languages dicer f all their neighbours. Uralic and Indo-European languages being included. There are numerous examples of languages. apart from Greek in the south. however. Australia and South America (as regards their older-established guages) are such areas. or small group related languages. Tur Semitic. appearing as a suax on the head n or on some other word in the noun phrase. For the moment I shall merely observe that marking of something akin to definiteness on direct object noun phrases onl found in many languages. but they do indirectly signal indefiniteness while not encoding it: a is ob atory in singular indefinite noun phrases in the absence of any other determi and neither a nor sm ever appears in definite noun phrases. This “indirect nalling” of indefiniteness by a cardinality determiner. Comrie 1981b). that definiteness marking only occur an areal feature. or almost abs is also found. in form of inflectional material on the head noun. In case too the languages involved are not always related to each other. These languages have a definite article.

If we restrict “marker of indefiniteness” to forms which dire encode [− Def ]. b. Boas and Deloria 1939) c‘a kc ‘the stick’ c‘a wa ‘a stick’ (6) (7) 50 ‘the book’ ‘a book’ . but have either a true indefinite or a quasi-indefinite article. b. Lakhota (Buechel 1939. with a vague cardinality term like sm sometimes occurring w plural and mass indefinites. b. Standard Arabic (Tritton 1977) albaytu (DEF+house) ‘the house’ baytun (house+INDEF) ‘a house’.examples below. Classical Greek to dendron ‘the tree’ dendron ‘a tree’ (2) ‘the table’ ‘a table’ Let us take pattern (b) to be satisfied by languages which do not have a defi article. Mam (England xiinaq jun xiinaq kab’ xiinaq (4) 1983) ‘man’. then possibility (a) above is by far the most common. possibili (b) and (c) are also frequent. c. b. b. a. Turkish (Lewis 1967) ev ‘house’.) Pattern (c). Danish boggen (book+DEF) en bog a. an item closely related to the num ‘one’ occurring. But if allow quasi-indefinite articles to count as markers of indefiniteness. indefinite n phrases being recognized simply by the absence of the definite article. Irish an bord bord a. ‘the house’ bir ev ‘a house’ bazı ev ‘sm houses’ a. b. are: (1) a. is illustrated by the following (as well as English): (5) a. with both definite and indefinite signalled by a cles. where there is form like a which could even be argued to be an indefinite article. Examples are: (3) a. Some examples of pattern (a). This art is found most commonly in the singular. c. b. ‘the man’ ‘a man’ ‘sm men’ (also ‘two men’) (In fact definiteness is indicated in Turkish when the noun phrase is a direct obj but not otherwise.

1. But there are. they do not always coincide with the English ones. 2. then on a strict interpretation of the terms pattern (a) is only one occurring. English French Cotton is easy to wash. where a Spanish definite se to correspond to an English indefinite. and it is arguable that since bed here is referential it is not indefinite in the way that a bed would be. .2 The distribution of definites and indefinites There are considerable dicerences across languages in the relative tributions of what look like definite and indefinite noun phrases. so that the pattern of only the p tive value of the feature [± Def ] being directly encoded is very general. Another striking observation is that the definite article seems to have a m wider range of use in French than in English. translatin bed into Spanish gives en la cama (in the bed). a plurality of members. for example). For example: (8) a. This is a major dicerence between languages. I will adopt the simpler view generics are (typically) definite in some languages and indefinite in others. but sub masZ (under table). English French She loves detective novels. Elle adore les romans policiers. not * masa (under table-DEF). Rather than assuming that the article occur in generics in languages such as French is a special generic article that happ to be homophonous with the definite article. Minor diceren are to be expected.dinal articles rather than true indefinite articles. for exam languages in which the definite article is systematically omitted after a position. ‘under the table’. This is the case in Romanian (Murrell and jtefznescu-Drzgzneiti 19 sub o masZ (under a table) ‘under a table’. Le coton est facile à laver. If completely general. For the moment it will suace to note that there are m languages like French in which generics take definite form. and will be discussed in detail w we come to examine generics in Chapter 4. (9) a. b. Until then I will confine my atten largely to non-generics. b. and it is not particularly odd that. for example. The main reason this impression is that generic noun phrases (those denoting a whole class or m rather than an individual member. for instance. Spanish also some expressions of this kind (en casa (in house) ‘at home’. The reason for this pattern is q mysterious. or a part) tend to the definite article in the former language but not in the latter. and many like Eng in which they are typically bare. English has a number of set prepositio expressions of this kind in which a state or direction is expressed rather t a relationship to a particular entity.

’ Yaro ya zo. emir AUX get-up AUX enter house all mind at spoil ‘The emir got up. T is the case in Hausa (data from Kraft and Kirk-Greene 1973. he had left his cap where the collision had happened.’ The definite suax is principally used anaphorically. a boy AUX come ‘A (certain) boy has come. as in the follow example: (11) To. hu ‘the cap’ has its antecedent hularsa ‘his cap’ in the same sentence. But th are numerous languages in which the definite article can be omitted where the uational or discourse conditions for definiteness do apply (and therefore wh translation into a language such as English would require a definite article). boy AUX come ‘The/A boy has come. boy-DEF AUX come ‘The boy has come. where the referent has p viously been mentioned in the discourse. is definite because it is an associate of the trig karon.’ Two of the definite noun phrases indicated in (11) are second mentions. as the following shows: (12) 52 Sarki ya tashi. while the p vious mention of karon ‘the collision’ is considerably further back in the discou The third.1 Optional definite articles Among languages that have a definite article.’ . . ya shiga gida duk rai a bace . ashe ya bar hula-r-sa a wuri-n da aka yi OK really AUX leave cap-DEF-his at place-DEF REL AUX do karo-n. Wani yaro ya zo. But the definite suax is optio even in such clearly anaphoric or associative cases. wurin ‘the place’. . b. Jaggar 1985): (10) a. entered the house most distressed .2. collision-DEF then a boy AUX see cap-DEF ‘OK. Thu “simple” noun phrase with no definite article will normally be indefinite. . as in English.2. the dominant patter for this article to act as a default form that must occur in a definite noun phr in the absence of a semantically fuller definite determiner. where the referen related to something previously (or subsequently) mentioned. . c. sai wani yaro ya ga hula-r.’ Yaro-n ya zo. making this a case of associative cataphora. then a saw the cap. and associatively.

AUX bring bicycle the-2SGM ‘I brought your bicycle (previously mentioned). free-form. 2. article wa. while the appearance or omission of article cannot be fully predicted. Jaggar (1985) argues that. In such circumstances the speaker tend use the “heavier coding” of an article-marked noun phrase as a way of aler the hearer to the need to find the referent and thus helping him in the task other words. Hausa has another. min. restricted to anaphoric use. bu the article is essentially anaphoric rather than general definite. see Ariel (1990). Other languages have definite articles dividing the overall field of definiteness into anaphoric and n . -s.’ (14) Na kawo keke gin-ka. a demonstrative is u It may not be obvious how the article (or demonstrative) helps the hearer.2. which acts as a generally obligatory default f with indefinites. it is largely determined by the accessibility the referent.are appropriate. similar. and if other. but both nouns occur in bare form in the Hausa. Hidatsa also has an indefin or perhaps cardinal. then one can that it serves to direct the hearer to the preceding discourse rather than seekin situational referent for a description which does not immediately activate a re ent. definite article. refer have occurred in the intervening discourse. Interestingly. AUX want pencil the ‘I want the pencil. If the previous mention of a referent is considerably far back in discourse it is less easily activated by the hearer. these can interfere with the hear identification of the intended referent.’ This phenomenon is not unusual. Hidatsa (Matthews 1965) has a single defi article. For a detailed study of accessibility. a bare noun phrase is used when the referent is judged to be eas access. and a definite-marked noun phrase when more ecort seems to be requi If the task of referent identification appears yet greater. In a detailed study of this phenomenon in Hausa. based on a statistical an sis of texts. and in most cases occurrin an alternative to -n/-r: (13) Ina son fensir gin. used specific when the referent has been previously mentioned. the result is that noun phrases with no article or other determ are usually to be understood as non-anaphoric definite. which are closely rel to anaphora).2 Anaphoric articles I have suggested that the Hausa definite article suax -n/-r is u principally for anaphoric definites (and associative definites.

that man the wise ‘That man is wise. unstressed – a normal characteristic of articles but no demonstratives. appear to function as anaphoric artic 2 Some descriptive grammars treat this suax as a mere “linker” (a meaningless morpheme fo in some languages linking a noun to its modifiers) rather than a definiteness marker. But I little justification for this view. He wic‘a-a kc ksape’. gona-r can ‘that farm’. Now may co-occur with min. boy the this NEG-AUX return NEG ‘That boy (previously referred to) has not returned. The article kC appears to be totally excluded from anaphoric use. and we will examine this phenomenon more closely in Cha 3. Second. therefore. wh has -n/-r suaxed to it:2 abinci-n nan ‘this food’. In Lakhota attributive demonstratives must be accompanied in noun phrase by the definite article.’ A very similar argument works for Hausa min. and it is not nor cross-linguistically for distinct demonstratives to appear together modifying a sin noun phrase. 54 . it can co-occur with demonstratives. which suggests this form is gen definite rather than non-anaphoric. The attribu demonstratives nan ‘this/these’ and can ‘that/those’ normally follow the noun. which would involve a linker. despite the complication that is a free-form determiner by contrast with the definite aax -n/-r. occurring precisely construction in which many languages do require the definite article to occur. and this requirement can be met by either general definite kC or the anaphoric k’E: (15) a. the definite article. But this means that it may be unclear whether a determiner specialized in anaph use is an article or a demonstrative. the aforesaid as renderings). But there is rather strong evidence some anaphoric determiners are definite articles rather than demonstratives.form k’E used when the referent has already been mentioned: wowapi kC ‘the bo (situational). Indeed Buechel (1939) says that the Lakh anaphoric form k’E corresponds to English that (as well as giving the abo mentioned. wowapi k’E ‘the book (mentioned before)’.’ Lakhota k’E and Hausa min do. in which case min replaces the article suax: (16) Yaro gin nan bai dawo ba. and showing the same allomorphy as. b. and that the two forms overlap in meaning It is possible in most languages to refer to something previously mentioned us a demonstrative. like kC. which happens to be homo nous with.’ He wic‘a-a k’e ksape’. F Lakhota k’E is. that man the wise ‘That man (previously mentioned) is wise.

Minassian 1976): definite article demonstrative -s -d -n ays ayd ayn ‘the (near me)’ ‘the (near you)’ ‘the (near him etc. The demonstrative system is as follows: . d. preceding the noun. So Mod Armenian. features like [± Prox] postulated for English can appear independently of [± Dem] – a p which gives further support for the existence of this latter feature in demons tives. and demonstratives also formed by suaxation to the class marker. Malherbe and Sall 1989). e. de Bray 1980). usually following the noun. That is. has lost it in the definite article.The kind of deictic distinctions which typically occur in demons tives (distinctions of distance from the speaker. Thus. the point is that the two-way proximity contrast found in the demonstratives shows in the definiteness marker: (17) a. indefiniteness is mar by a. is fo in the suaxed definite article as well as the demonstrative of Classical Armen (Meillet 1936. c. Another language showing a threedeictic contrast in both demonstratives and the definite article is Macedonian (L 1952. This class marker forms basis of any accompanying articles or demonstratives. particularly since in some languages the same deictic features appear b on demonstratives and on the definite article. b. common in demonstrative systems. The third form. the Modern Armenian forms descended from the article of the c sical language have been partly reinterpreted as possessive suaxes: -s ‘my ‘your’ (Feydit 1969).prefixed to the class marker.)’ ‘this (near me)’ ‘that (near you)’ ‘that (near him e Interestingly. while maintaining a three-way person-based deictic contrast in dem stratives. modern -n/-[. Jensen 1959. ab xale xale bii xale bee xale bi xale ba ‘a child’ ‘this child’ ‘that child’ ‘the child (nearby)’ ‘the child (further oc )’ A three-way person-based distinction. association with dicerent pers for example) are occasionally found in simple definites. An example is Wolof (Mbassy N 1982. following the noun. in which nouns are accompanied by a marker i cating the noun’s membership of a particular class. descended from classical serves both as the third-person possessive and as the definite article. definiteness is marked -i or -a suaxed to the class marker.

Nouns are accompanied by a proclitic or prefixed particle. toj. it can. Rovek-ov ‘the man here’. occurring much more commonly t the others and with no “second-person” or other deictic value. Rovek-on ‘the man over the A particularly interesting example of this phenomenon is Bella Co (Newman 1969. since it appe 56 . The prefix is not itself an indefinite article. Nater 1984). “Proxim can be interpreted to mean close to the speaker. not expre ing a particular degree of distance. absent from the spe situation. so that -ov is ‘the (near me)’ and -on is ‘the (at a tance)’. -ot is the unmarked article form. and location ative to the speaker: singular proximal distal plura female non-female tsilha-/filh- tita- wa-/a tu-/ta A noun marked only with one of these prefixes is generally understood indefinite: tsi-cnas ‘a woman (proximal)’. and then would express lesser proximity in contrast with ovoj and lesser tance in contrast with onoj. “distal” can mean remote from the speaker.M SG F SG N SG PL ovoj toj onoj ovaa taa onaa ova toa ona ovie tie onie ‘this (near m ‘that’ ‘that (yonde The middle term of this system. however. The definite articles are bound forms. contrast with either of the ot two. which encodes number. present in the speech situat or simply visible. derived fr these demonstratives with the reduction typical of articles (though exhibitin gender contrast in the plural not found in the demonstratives): M SG F SG N SG M/F PL N -ov -ot -on -va -ta -na -vo -to -no -ve -te -ne -v -t -n The three terms of this system are close in meaning to the morphologically c responding demonstratives. or invisible. lha-cnas ‘a woman (distal)’. perh best regarded as a class marker. is unmarked or general in value. in which indefinites as well as definites are dei cally marked. gender. it is simply ‘t Thus Rovek-ot ‘the man’.

definiteness. but not very closely. which is either distinct from definiteness yet cuts ac it.2. A categor expressed. The category in question responds in part to the concept of specificity which we have discussed inform It is found in a number of Polynesian languages. and relative location. Samoan distinguishes a “specific” article le and a “non-specific” article se (b showing several allomorphic variants). c. Languages with anaphoric article divide the range of uses covered by [+ Def ] into two parts. ti-fimlk-t’ayc ti-fimlk-tc lha-cnas-filhafilh lha-cnas-filh ‘this man’ ‘the man (close to speaker)’ ‘that woman’ ‘the woman (remote)’ 2. with N representing the noun s (and omitting prefix and suax variant forms): PROX DEF PROX DEM DIST DEF DIST DEM female non-female plural tsi-N-tsc tsi-N-ts’ayc lha-N-filh lha-N-filhafilh ti-N-tc ti-N-t’ayc ta-N-tx ta-N-t’ax wa-N-t wa-N-f tu-N-tx tu-N-t’ So definite article and demonstrative are both suaxal and involve the same d tic distinctions: (18) a. typically by a particle with the positional and morphophonological c acteristics of an article. Definiteness is marked by a set of enclitic or suaxed particles encoding same distinctions of number. d.4 Definite and specific The material discussed in the last two sections is compatible with a c ception of definiteness close to that outlined in Chapter or encode only one part of the fi And where a language expresses deictic distinctions in the definite article. b. these simply additional to [+ Def ]. gender. These sua also encode a distinction of simple definite versus demonstrative. which I exemplify here f Samoan (Mosel and Hovdhaugen 1992) and Maori (Bauer 1993). which can occur in a tion to the prefixes and must agree with these in the features encoded. A noun phrase introduced by le may definite or indefinite. in that its reference need not be familiar to or identifi by the hearer: . either use a separate article for each part. or is broader than. We now come to the phenomenon of a language enc ing something which resembles definiteness. and inclusive of. The combi prefixal and suaxal marking works as follows.

is specific indefinite.’ Se indicates that no specific referent is intended. but any member or part of w is denoted by the noun and associated descriptive material: (20) ‘Au-mai se niu. The following examples illustrate the of these articles: (24) 58 Kei te ruku raaua i te kooura. invariable for number. GENR not cry ART angel ‘Angels do not cry.’ The non-specific article is required in a negative or interrogative context. the wife. The article te. appe to be non-specific indefinite. teetahi (made up of te and tahi ‘on with plural eetahi.’ Bauer (1993) discusses the meaning and use of articles in Maori in some de and is doubtful whether definiteness or specificity ocers a consistent basis for distinctions occurring. Papa. GENR edible ART snake ‘Snakes are edible. He. plural ngaa. and applies equally to generics. Eleele ART woman ‘There was a couple. the husband. which are usually expressed in positive statements w the specific article: (22) E ai-na le gata.’ It is also used where there is a particular referent but the identity of this is ei not known exactly to the speaker or considered unimportant or uninteresting: (21) Sa fesili mai se tamaitai po-o ai l-o ma PAST ask DIR ART lady WH PRT who ART POSS 1EXC tama. father ‘A lady asked us who our father was. though some writers regard it as specific. A third item.PAST exist ART couple PRES Papa ART husband but PRES Eleele le fafine. and Eleele. TNS-ASP dive 3DU OBJ ART crayfish ‘They are diving for crayfish.’ . take DIR ART coconut ‘Bring me a coconut. is standardly described definite.’ (23) E le tagi se agelu.

. occurs: (27) E tuhi-tuhi reta ana ia. u when the particular senses of the other articles are not appropriate. can co-occur with demonstratives. and tends especially to occur instead of the third-person p noun for inanimate reference. and other things. paper.TNS-ASP buy PASS by 3SG ART cigarette ART paper ART wa reka. Sango has a post-n inal particle ní. used “to identify and single out a particular object”. the descriptive content of the noun phrase) is crucial. it is replaced with direct obj by the “incorporation” structure in which a bare noun. with no determiner or c marker (and consequently no number marking). he aha. Two further probable instances are Shus (Kuipers 1974) and Sango (Samarin 1967). and the article system seems to relate at least partly to the distinction betw specific and non-specific – though this too may not be fully adequate to desc it. sweet ART what ART what ‘He bought cigarettes. TNS-ASP gaze ART Mere OBJ ART coat ‘Mere . is looking at some coats. to repla personal pronoun. It is clear the meanings and patterns of use of the Maori articles are not yet established call for further investigation. It can be u anaphorically in reference to an object already mentioned in the discourse. It cannot follow a preposit and in fact only occurs in intransitive subjects. car had-killed certain child ART ‘A car had killed a child. i eetahi koti. The former indic that the referent is “uniquely determined for the speaker”. he aha. But examples like the following. She sees te as a “default” article. Articles marking specificity. in which it occurs with mbéni ‘some’. soft drinks. make it clear that ní need not be defin (28) Auto afáa mbéni m0r0ng0 ní. . It can also be used pronominally. ‘a certain’. definiteness–indefiniteness does not seem to be a consistent distinction in this guage. Shuswap distinguishes “actual-de minate” and “hypothetical-indeterminate” articles J/l and k.’ The article he is severely limited in distribution. and tahi/eetahi when the number is significant. TNS-ASP write write letter TNS-ASP 3SG ‘She is writing letters/a letter. or something close to specificity. rather t definiteness are fairly widespread.’ (26) Kei-te maatakitaki a Mere . and the reader is referred to Bauer’s discussion.’ Bauer suggests that he (which does not distinguish number) is used when the k of entity (that is. .’ . .

expres by variation in a class prefix.5 Nominalizing and other functions An important aspect of the behaviour of definite articles is their other than with nouns. Suárez 1983): (29) Ivenav ti vkaf ti tef. see Baker (1996). More strikingly.non-specific indefinite. 60 . can occur with the definite article: Spanish El hacer e fue fácil (the to-do this was easy) ‘Doing this was easy’. a definite article can sometimes serve to introduce an en finite clause. to occur with adjectives (and other noun modifiers). a form of the definite article unmar for gender).’ Consider also the following example from Lakhota (Buechel 1939) in which complementizer-like article (occurring post-clausally. as observed by Mithun (1984). She gives this example from Mohawk (30) Teionatonhwentsó:ni ne aontakontiráthen. it-appeared the is-rotting the wood ‘It appeared that the wood is rotting. there is reason to believe that the Mohawk particle ne is not a definite cle. see the discussion of Dzamba in Bokamba (197 2.2. defi and indefinite. This applies particularly to subordinate clau with an argument function – that is. in keeping with the p nominal position of articles in the noun phrase in Lakhota) is reinforced by demonstrative he: 3 On the other hand. It is common for articles and other determiners. Verbs can also be nominalized. ‘It was easy to do th Lakhota ktepi kC wAyake (kill the saw) ‘He saw the killing’. they-want(-it) the they-would-climb-up-here ‘They want to climb up here. that there is no definiteness marking in p synthetic languages. clauses behaving like noun phrases. so that the la constitute the descriptive core of a noun phrase: English the rich. thus functioning somewhat as a complementizer – and this may w be the correct analysis of such uses. Spanish un po (a poor) ‘a poor person’. furthermore. and in many languages an infinit or other basic verb form. who argues.’ This phenomenon may be particularly characteristic of polysynthetic langua (in which the verb can incorporate all the information necessary for a comp clause). as s ject or object of a higher sentence for instance – as in this example from Huix Tzotzil (Cowan 1969. lo difícil (the diacult) ‘the diacult thing/part’ or ‘w is diacult’ (where lo is the “neuter article”. For a complex interaction of definite and specific.

Thus two possible replies to NC kta ‘Are you going?’ are: (33) a. either familiar to the hearer or relatable by the hearer to wha familiar). In German the relative pronoun is in most forms identical to the definite article (w can also have demonstrative value. In the genitive forms in the dative plural the relative pronoun is morphologically fuller than the article. On Lakhota see Van Valin (1985). otherwise the indefinite article is used.nckta wa-.nckte-ni kc-. The indefinite wA expresses approval or agreement. In the following examples the relative element or p noun (indicated as REL) is identical morphologically to the definite article of language the that how killed-REFL the that anyone know NEG ‘Nobody knows how that man killed himself. By the time of Classical Greek. m. they said. depe ing on the language’s word-order pattern) relative clauses is very common. the questioner should have known better than to ask. ‘Yes. and the definit and definite anaphoric k’E express disapproval or disagreement. in some cases is probably to be explained in terms of the close relationship betw determiners and pronouns.4 4 The Homeric Greek example is complicated by the fact that at this stage of the language “article” forms tended to have demonstrative value. b. The use of definite articles. In Lakhota the anaphoric article k’E is used at the end of direct reported spe (Boas and Deloria 1939): (32) “Oyakapi k’e”.’ Articles. can also be appended to a sentence in Lakh with the conversational function of expressing the speaker’s attitude to some of the content of her utterance or to a preceding utterance to which sh responding. The definite articl used if the fact expressed in the clause is “related to previous experience” (me ing. tell-3PL the said-3PL ‘ “They tell it”. definite and indefinite. I am not going!’ To-. to introduce (or close. the relative pronoun was clearly distinguished this. ‘No.’ Andrade (1933) reports that in Quileute both definite and indefinite articles occur before a subordinate clause in an argument function. . an extra lable -en being added. Hiya. ’eyapi’. and can be used pronominally). when t forms were unambiguously a definite article. WA and kC may intensified by the addition of a particle -_. presumably. of course I am going!’ The wA_ in (33b) can be interpreted as implying that the question was unne sary. m. and also demonstratives.

A typological tinction has often been drawn in traditional descriptive work between prepo articles. and postposed articles.4 below. see also 6.’ See Comrie (1981a) for discussion of the internal-head relative and other ty of relative structure.3 Definite article types The form and position of definite articles show considerable varia across the world’s languages. that man came today REL said so ‘The man that came today said so.d4ra ta hoi xeinos d4ke gifts REL-NPL him-DAT stranger gave ‘gifts which a stranger gave him’ (35) German Der Mantel. however.3. the coat REL-ACC he wears is too big ‘The coat he is wearing is too big. appearing before the head noun. This being so. (36) baha waikna naiwa balan ba baku win. Suárez 19 illustrates the “internal-head relative” construction. 2. den er trägt.’ The following example from Miskito (Thaeler and Thaeler undated. the relative part ba is probably to be seen as simply the article modifying the entire relative “n phrase”. ist zu groß. The article in Miskito is post-nominal anyway. in the relative constr tion it appears at the end of the phrase and a demonstrative (here baha) may app at the beginning. This variation does. fall within certain lim and a small handful of types of expression occurs repeatedly. in which the noun modified the relative occurs within the relative clause. following the no Preposed article Hungarian a n6 Arabic albaytu Catalan l’ampolla la noia (the woman) (the-house) (the-bottle) (the girl) Postposed article Wolof nenne bi Romanian cartea Hausa kujerar fensir Uin 62 (child the) (book-the) (chair-the) (pencil the) .

The Hausa examples above show that this guage has both a free-form and a bound article. since it is often far from clear. but l’ is almost certa not bound but merely a reduced independent form required before a vowel. The is a member of the category Det. Among bound forms there is an important distinction between inflectional aa and clitics.3. Quite apart from the issue of whether an aaxal form is an inflection or a cl it is not always a simple matter to decide even whether the article occurring particular language is an independent lexical item or an aax. has been used to denote a variety of possibly distinct phenomena. attaching enclitically to the last word of a noun phrase in [ [the man I spoke to’s] car] ). Cat looks similar.3. Much more important is the question of whether an articl an independent word or a bound morpheme. though both are “postposed”. an art I here class as lexical could be argued to be aaxal. perhaps depending on other tors present in particular languages. Many of the bo articles I shall consider are attached to the lexical head of the appropriate phr the noun. Klavans (1985). 2. but I shall not attempt to draw this distinction systematically for cases discussed. for example. for discus see Zwicky (1977). an independent lexical item. along with other “determiners” (a t I continue to use informally). It may well be that. In addition to this both independent. with free and bound “preposed” articles. Detailed analysis of the fact a particular language are usually required to determine the status of a given fo and writers often disagree. an example migh English possessive ’s.1 Structural position In the great majority of languages where a lexical article occur is closely comparable to English the in terms of its position in the noun phr . Zwicky and Pullum (1983). It is primarily this distinction wh is to be observed in contrasting Spanish el hombre (the man) with Romanian o (man-the) ‘the man’. The term “clitic” is itself a pre-theoretical one. free-form articles bound articles may precede or follow the head. They are proclitic or encliti either the first or the last word or constituent of the phrase.found significance.1 Free-form articles One of the most common ways of expressing simple definitenes by means of an article which is a free form.1. not necessarily to a head or to any specified c stituent. 2. like Eng the. These “head aaxes” are to be distinguished from “phrasal cliti a type of clitic which attaches. some of these are inflections and some clitics – I shall use the t “aax” for both. but to one of the boundaries of a phrase. given that all art types tend to be phonologically weak.

while the definite article is p nominal: Irish hata an fhir (hat the-GEN man-GEN) ‘the man’s hat’. Dutch de [da]. and I limit myself here the observation that the generalization of phrase-initial position for the defi article is a rather strong one.2 Some characteristics Definite articles show a strong tendency to be unstressed (though t are often capable of being stressed for emphasis or contrast. moreover. whatever this turns out to be structurally. with adjectival modifiers following the head noun and the article lá ( of two article forms.3. the French masculine singular fo le [la] is the result of the same kind of vowel reduction in Old French. to occur initially in the noun phrase. but an rot (the bicycle). Sam le aso muamua o le v!iaso (ART day first POSS ART week) ‘the first day of week’ (Mosel and Hovdhaugen 1992). for exam to show noun phrase specifiers also in post-head position. there is a strong tendency for definite article. Ewe shows the mirror image of the Eng pattern. one might expect VSO languages. independe of a language’s general constituent-order pattern. that a f form definite article is usually. moreover. in the article’s being a weak fo This is exemplified in the reduced vowels occurring in English the ( [2a] or [2 Irish an ( [an] or [a] ). het [at].1.cardinality expressions (but frequently not the equivalent of all. Kekchi li cha. This fact shows itself in many langua which have phonological reduction processes. 2.nk (the good man) (Eachus and Carlson 1966). with its “strong” pronunciation [2iT] ). if a free form. it is perfectly possible for a language to show one struct pattern in clauses but a dicerent one in noun phrases. This can be taken as evide against the assumption that Det position is NP specifier. It does seem possible. Cases like Ewe and Lakhota notwithstanding. A few examples are: Portuguese as duas casas (the houses). as English the be. Lakhota is similar: wowapi wa kC (book good the) ‘the good book’ (Buechel 1939). To take a dicerent word-order pattern. in accordance with the strong tende for unstressed words to be monosyllabic and for stress reduction to be accom nied by reduction of polysyllabic items to monosyllabicity (see Selkirk 1984). almost always monosyllabic. since subjects of clau are generated in a specifier position. T 64 . if not always. the other being an aax) coming last (Westermann 196 atí nyuí lá (tree beautiful the) ‘the beautiful tree’. Artic are.b’il wi. This expectation is co monly borne out as regards genitive expressions. therefore. which in m languages takes the whole noun phrase as sister). to advance the str hypothesis that lexical definite articles are generated universally in some “Det p tion”. Malagasy is a VOS langua yet has the article preceding the head noun: ny zazavavy (the girl) (Keenan 197 On the other hand. a Det appearing in the same position as in English. It is likely.

A great deal of vowel elision and contraction occurs in Ewondo. le. stressless nature can be accounted for in terms of its being a monosyllabic function (particularly one consisting of an open syllable with a lax vowel). without the need to ap to cliticization. to ’s [z]. A “simple clitic” is a weak or reduced i which is phonologically dependent on an adjacent “host”. On the other hand. This kind of assimila occurs not only within words. the of the low tone of the first syllable of oXgóla then leads to downstepping of next high tone to mid. A rule of nasal deletion. but which nevertheless occupies the same position in the sente or phrase as the corresponding full form (if there is one). but not across boundaries. that is. Now the definite article is é. by processes which are idiosyncratic to the item question rather than instantiating general phonological rules. lo. a North-West Bantu tone language (Red 1979). Mary’s fo it. illa. the reduced form of both is and has. which may itself already be unstressed and with a reduced vo [haz]. since it is normally unstressed. in expressions like John’s ill. varying in degree of reduction. however. But apart from cases like these. Nespor and Vogel (1986) give one argument for treatin definite article of Modern Greek as a clitic. la. il yielding Romance article forms el. though. . where an item occurs in dice forms.5 An interesting manifestation of art reduction is exhibited by Ewondo.derived historically from demonstratives and the development being accompan by syllable reduction – for example Latin demonstrative forms ille. which applies under tain segmental conditions within words and within host-clitic combinations. and this wo usually be the case with the definite article. Not all simple clitics have a corresponding full form. being further redu in particular environments. Thus á oXgóla (to fence) ‘to the fence’ contracts to óXg$la. the h tone of the preposition á transferring to the assimilating syllable following. which is likewise elide 5 Selkirk. the two together form a word-like unit. It may be that many instances of the definite article fall into the class of w Zwicky (1977) calls “simple clitics”. but l’ampolla ‘the bot Apart from its obligatoriness. A clear example is Eng ’s. discussing the English definite article along with other function words. of articles. and the t of an elided segment may pass to the adjacent syllable. il. some monosyllabic function words can be comple elided by it. We saw an instance of same phenomenon above from Catalan: la noia ‘the girl’. claims th weak. it is often far from obvious whether a w item is a clitic or whether its weakness follows merely from its being a “func word” (in the sense of Selkirk 1984). but l’autre femme ‘the other woman’. applies within the group consisting of the article and a following noun: [tin → [ti θéa] (the-ACC view-ACC). An example is reduction of the French feminine singular article la to l’ pre-vocalically: la fem ‘the woman’. illum. already weak and unstressed. this can perhaps be compared with the reductio English has. a “grammatical word” as opposed member of one of the major lexical classes. Th are instances.

howe because it depends on the suprasegmental phonological structure of the langua A distinction between fully stressed “lexical” words and frequently unstressed “fu tion” words can only exist in languages which have word stress. moreover. German in dem → im ‘in the’. le blé. but it is particularly striking because it is operating across a phrase bou ary. there seems to be more phonological material remaining from preposition than from the article. ‘All our food products are imported: meat. the preposition i ‘in’ and the article an yield the combined fo sa [sa] (via a still extant insan). This may seem odd in view of the fact. as in the following. in Irish.’ The downstepping is due to Ewondo being a “terraced-tone” language. downstepp of the second high again occurring under the influence of the suppressed low A very commonly occurring process is for noun phrase-initial definite arti to combine with a preceding preposition. vegetables. zu der → zur ‘to the’. where it would be normal the articles to carry the main stress of their noun phrase:7 (37) 6 7 Tous nos produits alimentaires sont importés: la viande. One s context is that of enumerations. that the French article exhibits clitic behaviour in attaching in a reduced form to a following vowel-initial word (l’autre femme) in entering into contractions with prepositions (du. the second high still being lo than the first by the same interval as a mid – it thus becomes superficially a mid. les légum les fruits. In some instances the changes following the cliticizat have been so far-reaching that neither component is recognizable in the com nation. So é oXgóla ‘the fence’ becomes óXg$la. les produits laitiers – enfin tout. au) – behaviour associated with reductio weak forms. Examples are: Fre de le → du ‘of the’. such reductions do not occur synchro ally in Modern French. A common an sis within the DP framework is that D. fruit. This is particularly clear from the fact that the orthographic u of du and au [o] is the result of a process of vocalization of [l] which occurred in the medieval per 66 . it is. à le → au ‘to the’. In many of th “contractions”. The tendency for definite articles to be unstressed is not universal. moves by a transformatio process of “Head Movement” to attach to the higher P head. but its ecect remains. when a low interve The intervening low may be elided. noted above. not immediately obvious which is the clitic and which is host. in which a hig normally lower than a preceding high by the same interval as a mid is. An ecect of this is that articles are stressed in some contexts. the article. But these cliticized and contracted articles are the result of reductions which occu at a time when French had a very dicerent stress system. since both articles and prepositions tend to be weak items. wh dairy products – in fact everything. Italian in il → nel ‘in the’. which suggests that the latter is phonologic enclitic to the former.passing to the next syllable. French has phr rather than word stress. This is presumably a kind of clitic tion. the stress tending to fall on the last syllable of phra but sometimes on the first syllable – whatever the status of the word occurrin that point. con il → ‘with the’.

. encoding tures such as number. Lakhota kC k’E. this point will be discusse detail in Chapter 5. the article is an inflected form. This is the case particularly in Indo-Europ languages. English the show no agreement with the head noun or other c stituents of the noun phrase. Hausa min. In many other languages. as illustrated in the following paradigms: Classical Greek singular NOM ACC GEN DAT dual plural M F N M F N M F ho ton tou t4 h2 t2n t2s t2 to to tou t4 t4 t4 toin toin t4 t4 toin toin t4 t4 toin toin hoi tous t4n tois hai t1s t4n tais German singular NOM ACC GEN DAT pl M F N der den des dem die die der der das das des dem di di de d Catalan SG PL M F el els la les In many languages the inflection of the article assumes a particular importa since some nominal features are only encoded. does not. derived historically from it. Tzotzil ti (Suárez 1983). a. the demonstrative az takes case inflections as does the h noun. gender and case. in which a related demonstrative does show inflect Thus in Hungarian.sense that it is not inflected (though it may undergo some allomorphic variation with Hungarian az pre-vocalically and a pre-consonantally). This is often the case even in languages with a fa high degree of inflection. or only consistently encoded the article (or on the determiner more generally). but the article az. however. For this reason often labelled in descriptive grammars a “particle”. Akan no (Schachter 1985). Thus Pocomchi re (Mayers Mayers 1966).

A second commonly encountered device for representing definiten is aaxation. but the indep dent determiner and demonstrative from which it derives no longer exists. A particularly clear example is that of Icelandic (Einars 1949. w being attached to a noun which is also separately inflected for these categor Consider the paradigms for the definite forms of hestur ‘horse’. still used also as a demonstrative (w contrastive value: ‘the/that other’). I exemplify from Danish: singular common neuter den det plural -(e)n -(e)t de de -(e -(e A particularly interesting feature of the Icelandic suaxal article is that it is it inflected for number. moreover. Many definite aaxes are known derive historically from lexical determiners (and ultimately. otherwise the suax occurs: hestur-inn ‘the horse’. where the determiner from which the definite sua probably (though not certainly) derived still exists as a free-form definite arti and the two articles dicer from each other only slightly. but is used in writ and elevated spoken style when the head noun is modified by a preceding ad tive. fr demonstratives). hinn sterki hestur ‘the str horse’. The determiner hinn is. as can be seen in the lowing paradigms: masculine feminine neuter NOM SG ACC SG DAT SG GEN SG hinn hinn hinum hins -(i)nn -(i)nn -num -(i)ns hin hina hinni hinnar -(i)n -(i)na -(i)nni -(i)nnar hi2 hi2 hinu hins -( -( -n -( NOM PL ACC PL DAT PL GEN PL hinir hina hinum hinna -nir -na -num -nna hinar hinar hinum hinna -nar -nar -num -nna hin hin hinum hinna -( -( -n -n The free-standing article rarely occurs in spoken Icelandic. hav been replaced by a rival form. dicering in form from the article only in the neuter nominative singular is hitt rather than hi@. Glendening 1961). case and gender (like the determiner it derives from). The suaxed article of other Scandinavian languages is cognate with that of Icelandic. to the head noun or elsewhere. like these. exemplifying each gender: 68 . borg ‘town’ f jall ‘mountain’.

with the result that number. The only dicerence is in the dative plural.correspond to unsuaxed hestum. in fact. The two are diachronically distinct and synchronically not particul similar in form. I exemplify again from Danish: singular mand-en kone-n barn-et ‘the man’ ‘the women’ ‘the child’ plural mænd-ene koner-ne børn-ene ‘the men’ ‘the wom ‘the child Again the form of the noun to which the article suax is added. borgum. The article suax is not simply one inflectional morpheme. the quasi-indefinite article en/et. In fact the suaxed definite article closely resembles. with some reduction.3 above in Bella Coola. in both sing and plural. among others on the noun. The o Scandinavian languages present the same phenomenon. is identical to the corresponding form of the noun without article su It is clear that the noun and the article suax are separately inflected items. but can hardly be taken to be a clitici version of this! We saw similar facts in 2. though with a m reduced inflectional system. It seems plausible to treat it ra as an independently inflected lexical item – probably. where . case and gender are e encoded twice.masculine feminine neuter NOM SG ACC SG DAT SG GEN SG hestur-inn hest-inn hesti-num hests-ins borg-in borg-ina borg-inni borgar-innar fjall-i2 fjall-i2 fjalli-n fjalls-i NOM PL ACC PL DAT PL GEN PL hestar-nir hesta-na hestu-num hesta-nna borgir-nar borgir-nar borgu-num borga-nna fjöll-in fjöll-in fjöllu-n fjalla-n The form of the noun to which the article suax is attached is in fact almost id tical to the corresponding unsuaxed form (the form used when indefinite or w a free-form determiner). is added to inflected form of the noun. the definite hinn – which under certain circumstances is cliticized onto the noun. The point is the suaxed article. f jöllu. f jöllum – clearly a sim case of allomorphic reduction triggered by the article suax -num. in the gular. where hes borgu-. enc ing definiteness. The main dicerence betw Icelandic and Danish is that in the latter language the suaxal article canno plausibly be argued to be a cliticized form. of the free-sta ing article. so in these definite forms number is encoded twice.2. itself fully inflected to agree with the noun.

(Hewitt 1979) and the Ewe suax -á. this is because these categories are only encoded on the noun when definite article is attached. varying for the plural. the preterit can be reasonably argued that the tense–aspect morpheme also varies for person. it can be claim to express number. and person–number. 70 . Another instance is Vai (Welmers 1976). but these features are also encoded on the definite suax. There are also aa articles which do encode more than [+ Def ]. fairly readily segmented into discrete morphemes exp ing verb class. but without duplication of the encoding of these additional egories. it is worth considering Spanish verb morphology. -S (after -e and -o) and -w otherwise.8 Aaxal definite articles are also frequently invariable. gender and case for example. given in Spencer 1992). for the most part. But in one tense paradigm. and -n’e. where article is -[ after a consonant (kirk’-[ ‘the book’) and -n after a vowel (gadou-n cat’). in which the suaxal article has forms -Q (after -a). tense–aspect. on the grounds tha inflection cannot itself be inflected. as in Modern Armenian (Feydit 1969).in the singular (excep the nominative where it takes the form -s’). which encodes plu 8 Before an inflection analysis of these definiteness markers is dismissed. show allomorphic the speaker. so that pe is encoded twice. The plural variant is always preceded by -t-. In this l guage definiteness is encoded suaxally along with other inflectional categor but the indefinite form of nouns only encodes number in the nominative wher the definite form distinguishes singular and plural throughout the paradigm. I g here a partial paradigm (Erzya-Mordva has a system of ten cases in all): kudo ‘house’ indefinite definite NOM SG GEN SG DAT SG ABL SG INESS SG kudo kudo-n’ kudo-n’en’ kudo-do kudo-so kudo-s’ kudo-nt’ kudo-nt’-en’ kudo-do-nt’ kudo-so-nt’ NOM PL GEN PL DAT PL ABL PL INESS PL kudo-t kudo-n’ kudo-n’en’ kudo-do kudo-so kudo-t-n’e kudo-t-n’e-n kudo-t-n’e-n kudo-t-n’e-d kudo-t-n’e-se Since the definiteness morpheme varies between -nt’. Examples are the Abk prefix a. See also the discussion of Erzya-Mordva below. Uninflected aaxal articles may course. and articles of the sort discussed assumed to be clitic some kind. This is partially the case for the category of num in Erzya-Mordva (data from Feoktistov 1966. in which uncontrovers inflectional material is.

any morphological variation shown by al.9 9 Another article of uncertain morphological status is the Standard Arabic definite prefix This is clitic-like in that there are no idiosyncrasies displayed by any combination of it w host noun. there is good reason to tre as a phrasal clitic (or perhaps a “phrasal inflection”). Albanian ocers a good example inflectional definiteness marking in which it is not possible to identify a disc morpheme representing [+ Def ] – definiteness. Hubbard and Prifti 1982): Masculine ‘boy’ INDEF DEF Feminine ‘sister’ INDEF DEF NOM SG ACC SG DAT SG ABL SG djalë djalë djali djali djali djalin djalit djalit motër motër motre motre motra motrën motrës motrës NOM PL ACC PL DAT PL ABL PL djem djem djemve djemsh djemtë djemtë djemve(t) djemve(t) motra motra motrave motrash motrat motrat motrav motrav It seems impossible to further segment the inflection -in.3. is predictable on phonological grounds. it is a p manteau morpheme encoding simultaneously definite. will. for example. return to the Albanian definiteness marker in 2. feminine -a/-ja. and definiteness agreement takes the form of the article prefix appearing on the ad tive as well as on the noun: al-bustan-u l-kabir-u (DEF-garden-NOM DEF-big-NOM) ‘the garden’. gender and case being inextr bly combined. neuter (which is rare) -t. masculine determined by general rules.only appears with the definiteness morpheme (and could haps be argued to form a single morphological element with this). But in the other pl case forms the -t. Attributive adjectives agree with their noun in definiteness. and general plural -t/-të. Such inseparability of categories is common in inflectional systems. so that nouns have distinct definite and indefinite declensio these are illustrated here for a masculine and a feminine noun (Pipa unda Newmark. as well as in number.Scandinavian languages in that number is doubly expressed. however.) But the su also varies for case. des the inflection-like characteristics just pointed out. . There is considerable variation in the relationship between definite aaxes other aaxal material occurring on nouns.3. with mascu -i/-u. (The v ation between -i and -u etc. But t is one respect in which definiteness behaves in a way normally associated with inflectional egories. since. ge and case. accusative. The article suax encodes the gender of the noun.

the definite article haq is a clitic appearing in phrase-second positi that is. such as tense-mood “auxiliary”. together with case morphemes. The behaviour of the Kwakw’ala article is obviously of great importance to theory of cliticization if it indeed shows cliticization (of dicerent kinds) in directions. in that clause-level clitics. In some languages. like Nootka. Two examples of phrasal clitics have achieved some prominence the literature (for example Anderson 1985. it is enclitic to the first word of the noun phrase. It is not certain whether they are a tinct phenomenon from the clitics already discussed.Most of the bound definite articles discussed in the last section att necessarily to the head noun. it encliticizes pho logically to the preceding word: (38) kwi9zfid-i-da bagwanama-9-a q’asa-s-is t’alwagayu clubbed SUBJ DEF man OBJ DEF otter INST his club ‘The man clubbed the otter with his club’ The article is thus. they may be simply cli whose position of attachment makes them more clearly distinct from inflecti – the lexical head of a phrase being merely one of the phrase “positions” that m host clitics. not necessarily to the head of the phrase. b. occurs at the initial boundary of noun phrase. a. In others attachmen 72 . Klavans 1985). the defi article da. whatever that word (39) a. The Nootka article is particularly interesting because it represen very common pattern of phrasal cliticization. Wackernagel clitics att suaxally strictly to the first word of the phrase or clause. are similarly enclitic to the first word of the clause. In Kwakw’ala. after the nineteen century linguist who investigated such clitics in the Indo-European langua (Wackernagel 1892). as argued by Klavans. bowatc faq deer the ‘the deer’ fi9 faq bowatc big the deer ‘the big deer’ This pattern is more general in Nootka. and not obligatorily to one specific host category. but to a host specified by its position in the phras term these bound forms “phrasal clitics”. But while it immediately precedes the first word of the noun phr to which it relates (and thus the noun phrase as a whole). that of second-position attachm The second position in a phrase or clause is sometimes known as “Wackerna position” and second-position clitics as “Wackernagel clitics”. syntactically proclitic to the follow noun phrase and phonologically enclitic to whatever word precedes the noun phr In Nootka. Another type of bound item consists of eleme which attach.

recent work by Adams on weak pron position in Classical Latin (Adams 1994) indicates that some cases of what h been thought to be second-position clitics actually involve attachment. But when the noun is modified by a ceding adjective or numeral. tallaq-u bet big DEF house ‘the big house’ dähna-wa set good DEF woman ‘the good woman’ ammast-u kafl-orr five DEF room-PL ‘the five rooms’ Bulgarian has a definite article varying for gender and number (but without deictic distinctions of the closely related Macedonian. Bulgarian Romanian. c.3). Further instances of languages showing them are Amharic. the article attaches to the first element: (41) Viydam golemi-ja xubav grad. the impression second-position attachment is due to the tendency of focussed expressions to initial. geta-w ‘the master’). Scatton 19 NOM SG OBL SG PL masculine feminine neute ‘city’ grad-Et grad-a gradove-te ‘earth’ zemja-ta zemja-ta zemi-te ‘field polepolepolet If the noun is pre-modified. see-1SG big DEF fine city ‘I see the big. fine city.of this constituent. doro-wa ‘the hen’). in a tion it distinguishes nominative and oblique case forms in the masculine singu a distinction not made in the absence of the article (de Bray 1980. Leslau 1968. The Amharic article (Armbruster 1908. n[g oRR-u (queen-PL-DEF) ‘the queens’). but to a focussed or prominent host. b. And some languages allow a choice between these two in pretations of second position. it is clear that many cases of phrasal cliticization do invo Wackernagel position. Second-position clitic definite articles are in fact rather c mon. discussed in 2.2. Hartmann 1980) va for gender and number. feminine singular -wa (gäräd-wa maid’. the article is attached to this: (40) a.’ . the most common forms being: masculine singular -u ( färäs-u ‘the horse’. In fact. Nevertheless. not to first word or constituent. plural -u (n[gus-oRR-u (king-PL-DEF) ‘the kings’.

not the first word. The Romanian definite article. too. The Romanian article is. gender and case. this initial word cannot host the article. possible hosts being alm exclusively nouns and adjectives. -le -(u)lui -a -i PRI PL OBL PL -i -lor -le -lor -le -lor The noun or adjective hosts to which these are added are separately inflected number. not always allowing easy segmentation. and being added a separately inflected host). If a pre-nominal adjective is itself pre-modified an adverb or degree word. This marking is shown for both nouns and adjectives in following paradigms: 74 . while it is clearly clitic in character (being strictly a second-posi form and attaching to more than one possible host category. -le -(u)lui -(u)l. which m attach to the adjective: (42) mnogo xubavi-jat grad very fine DEF city ‘the very fine city’ In other words. The article varies for number. Case in par ular is only marked in the feminine singular. with an oblique form consiste identical to the plural.first word of the noun phrase. Romanian sho rather severe limitations on what can host the article. gender and case (“primary” case be a combined nominative–accusative form and “oblique” a combined geniti dative form – these finer distinctions being made only in personal pronouns) follows (data from Murrell and jtefznescu-Drzgzneiti 1970): masculine neuter femin PRI SG OBL SG -(u)l. by contrast with Bulgarian. But. is a Wackernagel form enclitic to the fi constituent of the noun phrase. the article is enclitic to the whole adjective phrase. particul interesting since. moreover. but in a much more impoverished way. it is at the same time relatively well integrated m phologically with its host. indicating Wackernagel position in the Bulgarian noun phrase is after the first constituen this phrase. and (in comm with free-form determiners like demonstratives and the quasi-indefinite article carries the main burden in the noun phrase of expressing grammatical informat especially case.

they will be discussed in 2. të.3. without and with definiteness marking. and both nouns and adjectives in pendently take the inflections shown above. But a noun following a numeral cannot take the article either this would violate the strict second-position rule. and fairly common with some adjectives. I illustrate for a feminine noun: (43) a.4. Consider the paradigms for a feminine noun followed by ad tive. and then it is the adjec not the noun. and commonly do when in the superlative degree. despite commonly being the first noun phrase constituent. But they can. unmarked order is for adjectives to follow the noun. in which the article ocers a good example of so thing which is not straightforwardly classified as inflectional aax or phrasal cl and perhaps gives reason to call into question the validity of the distinction shows very definite Wackernagel behaviour. Note also that I conflate the da and ablative cases distinguished in 2. c. as “oblique”. can host the article. Let us return to Albanian. The definite art attaches to whichever element comes first. for emphasis or c trast. indefinite definite NOM SG ACC SG OBL SG ‘good girl’ vajzë e mirë vajzë të mirë vajze të mirë ‘the good girl vajza e mirë vajzën e mirë vajzës së mirë NOM PL ACC PL OBL PL vajza të mira vajza të mira vajzave të mira vajzat e mira vajzat e mira vajzave(t) të m . since the separate abla forms are becoming obsolete.2. but adjective–n order is also possible. Instead a free-form determ cel is used: cei patru prieteni ‘the four friends’. cartea bunz czr]ii bune czr]ile bune czr]ilor bune or or or or buna carte bunei czr]i bunele czr]i bunelor czr]i ‘the good book’ ‘of/to the good book’ ‘the good books’ ‘of/to the good books’ Most numerals.3. disguised in part by the fact that ad tives rarely occur pre-nominally in Albanian. which carries any definiteness marking (which is necessarily pre with the superlative). së can be igno for the moment. The particles e. d. b.masculine neuter feminine PRI SG OBL SG ‘good dog’ cîine bun cîine bun ‘good train’ tren bun tren bun ‘good bo carte bu czr]i bun PRI PL OBL PL cîini buni cîini buni trenuri bune trenuri bune czr]i bun czr]i bun The usual.

is reversed. in a definite noun phrase numeral is. it may be that case and defin ness constitute together a single formative. in which the defi article is attached to the right boundary of the noun phrase. wh is enclitic to the first constituent of the noun phrase. katër vajza të katër vajzat ‘four girls’ ‘the four girls’ So the Albanian definiteness–case morpheme does not always show clear phra clitic behaviour. an enclitic on the word (de Arrigaray 1971. the noun becomes invariable except for number marking. b. As a final example of a phrasal clitic article I take Basque. given the inseparability of the definiteness morphe from the case morpheme pointed out in 2. The ad tive.3. Saltarelli 1988): 76 . in addition. The distinction between clitic and inflection is not a simple o and the reader is referred to Börjars and Vincent (1993) for discussion of a nu ber of borderline phenomena and a proposal for accounting for them in term distinctions within the concept of host. But the Albanian definite article does not always appear in second position may appear on the noun following demonstratives and numerals. now in initial position. Demonstrati may be followed by nouns in either definite or indefinite form. carries the case endings as well as the definiten marking: indefinite definite NOM SG ACC SG OBL SG e mirë vajzë të mirë vajzë të mire vajzë e mira vajzë të mirën vajzë së mirës vajzë NOM PL ACC PL OBL PL të mira vajza të mira vajza të mirave vajza të mirat vajza të mirat vajza të mirave(t) v So both definiteness and case in Albanian are encoded as second-position for Or. the former be more likely when the reference is anaphoric or when the noun is followed by identifying modifier: (44) këto vajzat e Agimit these daughters-DEF PRT Agim-GEN ‘these daughters of Agim’s’ Cardinal numerals cannot host the definite article or case marking. preceded by a particle të (about which more below): (45) a.2. more accurately perhaps. but these egories can appear on a noun following a numeral. a kind of noun phrase “auxiliary”.

and the suax otherwise: (47) a. only -á is possible in the plural. -ek.snow white cold and thick-DEF ‘the white. 2. In most cases the two are complementary to each ot either one or the other occurs in a given definite noun phrase. contrary to the normal Hausa pattern. depending on var factors. and preceding the plural a atí-a ‘the tree’. case. It is clear that placement cha teristic of phrasal clitics is compatible with other aspects of behaviour which inflection-like – giving rise to what are sometimes called “phrasal aaxes” or “phr inflections”. this article is not always identifiable as a discrete element. in others it is a matter of free choice or stylistic pre ence. b. especially th ending in a consonant. are inseparably encoded example. One is that the suax expresses agreement with a null articl Det position. tho the free-standing article is preferred with nouns of foreign origin. b. In so forms. This is the case in Hausa. Icelandic báturinn hinn fallegi bátur hinn eini bátur ‘the boat’ ‘the beautiful boat’ ‘the one boat’ (48) The explanation for this distribution is not obvious. Danish huset det gamle hus de tre huse ‘the house’ ‘the old house’ ‘the three houses’ a. c.3. cold and thick snow’ But.4 Mixed systems and double determination We have seen that a number of languages have both a bound defi article and a free-form one. and possibly number. so that it is blocked by any intervening modi . definiteness. The free-form article is used whenever the noun is ceded by modifying material. occurring finally in the noun phrase: atí lá ‘the tr There is also an aax -á. Ewe is another guage in which two articles are. In some languages there are grammatical rules governing the distribu of the two article forms. again. But while -á and lá are in variation in the singular. attaching to the noun. In Danish and wri Icelandic the choice between the suaxal article and the free-form determine grammatically conditioned. and that this agreement. atí-a-wo (tree-DEF-PL) ‘the trees’. in part. which licenses the null definite determi is subject to an adjacency condition. where -n/-r and min are simply alternatives. c. simply alternatives (Westermann 19 Ewe has a lexical article lá. though some possibilities s gest themselves. encoding definite ergative plural).

perhaps representing agreement with determiner. a suax and a free form. one might argue that den is the head D and an aax on the noun. and is of considerable importance for the theory of defin ness and determiners. b. otherwise the alternative free-form article must used. in both Swedish and Dan when it is a matter of a commonly occurring adjective–noun combination or a fixed exp sion. in the literature. resan den långa resan de fyra resorna ‘the journey’ ‘the long journey’ ‘the four journeys’ This phenomenon is termed ‘double determination’ or ‘double definition’. But the principle wid accepted within this framework that inflectional categories generally originate independent functional heads.10 but the aax also appears in this case: (49) a. combining with lexical items through movemen one or the other. Swedish also has two definite articles. and the reader is referred to Börjars (1994) for m detailed discussion of it.possibility is that the bound article is actually a Wackernagel clitic. that definite articles are to be found cross-linguistic originating in one or both of two positions. Since they express the same category. This is a less than desira conclusion. implicitly at le by a number of writers. am other labels. with the co plication that only the noun can host it. and I take double determination to be evidence for the analysis I w propose in a later chapter. as in Danish. In the DP analysis. ‘big’. Some of these have (near-)proper name status or can be argued to function as single ical units. though in others it seems to be merely the fact that the adjective is one of a se frequently occurring ones (‘old’. in the presence of a p nominal modifier. c. definiteness. are realizations of a D he while free-form articles tend to stand in a DP specifier position. Specifica Swedish -n here. including Taraldsen (1990). again expressing agreement. ‘young’. Double determination is reminiscent of clitic doubl in which a clitic object pronoun appears as well as an overt noun phrase ob in some languages. a head and a specifier. one would p sumably have to say that in den långa resan the determiner den is in NP speci position and -n is an aax on the head N. This means it can only appear when noun is initial in the phrase. stora flickan ‘the big g 78 . it seems need two head D positions – and this is the analysis adopted. Within the NP analysis of noun phrases. ‘small’ etc. but dicers in that the two can co-occur (Holmes and Hinchl 1994). and suaxal articles more generally. The free-form article is used. cognate w those of Danish. 10 In fact the suaxal article can occur alone on a pre-modified noun. would lead rather to the assumption that both den and -n functional heads.) that licenses the constructio exemplify from Swedish: katolska kyrkan ‘the Catholic Church’.

b. të miri-t djalë ‘the good boy (OBL)’. gender and case of the head noun. though the situation in these languages is much less strai forward than in the Scandinavian languages. including p sessives and oblique (or “dative”) case nouns used attributively. Recall from 2. and it appears with most. Thus: .perhaps of double determination. The similarity is evident in phrases such as: i m i djalë ‘the good boy (NOM)’. Consider its parad (omitting the rare neuter): NOM ACC OBL M SG F SG P i të (e) të e të (e) të (së) të të të The bracketted forms are used only with an attributive modifier when the nom head is definite and immediately precedes. To avoid confusion and to avoid the of prejudicing the issue. gen and case. adjectives: Albanian has “art lated” and “non-articulated” adjectives. But point is that this particle is cognate with the definite article suax and derived torically from a definite article. In fact this part accompanies not only adjectives but a wider range of modifiers. The form of the particle is de mined by the number. because it appears also in indefinite noun phrases: (50) a.3 that adjective Albanian are accompanied by a particle which varies in form for number. are ocered by the Balkan languages Alban and Romanian. një djalë i mirë a boy PRT good ‘a good boy’ disa të mira vajza sm PRT good girls ‘sm good girls’ The adjectival particle is probably best analysed as an agreement morpheme a ciated with the adjective – though it is interesting for diachronic linguistics a definite determiner should develop into such a morpheme. It therefore expresses agreement of a modifier with grammatical features of the expression modifi But the adjectival particle is sensitive to definiteness.2. But this particle is quite clearly not a definite art today. I shall prefer to call it the “adjectival particle”. This particle is termed in some descriptive grammars the “adjectival a cle”. së mi vajzë ‘the good girl (OBL)’. This last use ma particularly clear the agreement function of the adjectival particle: drejto shkollës ‘the principal (M NOM) of the school’. no the modifying attributive noun to which it is attached. though not all. the possessum.

’ E gjeta shtëpin të pastër. and also of definite article ori though occurring now in indefinites. expresses partial agr ment for definiteness as well as other categories. CL found-1SG house-DEF PRT clean ‘I found the house clean. in (51d) të is used because the adjective is not attributive (thus not in the n phrase) but predicative. It is simply that the adject particle. It is used to link a genitival or possess modifier to its head noun when the latter is either indefinite or has anot modifier: (52) a. b. a expressing agreement with the noun modified. In (51c) the s ond adjective takes të because it is not directly adjacent to the definite noun. and an ordinal (above ‘first’) cannot host the arti 80 . an agreement morpheme attached to modifiers. But these observations do not amount to seeing a defi article in Albanian distinct from the suaxal one. where the ordinal precedes the no no definite article appears.’ Examples (51a) and (51b) show the indefinite–definite contrast. b. CL found-1SG house-DEF PRT clean ‘I found the clean house. d. CL found-1SG a house PRT clean ‘I found a clean house. Notice that in the second alternative.b. (53) a. The noun cannot carry the article because this wo then not be in second position.’ E gjeta shtëpin e pastër e të këndshme. Romanian has a similar particle occurring with certain kinds of modifier. c. prietenul meu friend-DEF my ‘my friend’ un prieten al meu a friend PRT-M-SG my ‘a friend of mine’ cartea scriitorului book-DEF writer-DEF-OBL ‘the writer’s book’ noua carte a scriitorului new-DEF book PRT-F-SG writer-DEF-OBL ‘the writer’s new book’ It is also used with predicative possessives (Cartea e a mea ‘The book is min and with ordinal numerals: studentul al doilea or al doilea student ‘the sec student’.’ E gjeta shtëpin e pastër. CL found-1SG house-DEF PRT clean and PRT pleasant ‘I found the clean and pleasant house.

which does alternate with -(u)l etc. descended torically from determiners or pronouns. cel. (54) a. which varies for number. -n and -r. found accompanying modifiers in vari languages. In Hausa. Finally. gender and case of the expression modified. as a definite arti and indeed sometimes “doubles” the is at least as plausible to argue that there simply is no article here express the definite value which is inherent in the ordinal. possessive expressions are linked to the modified head such a particle: na/ta. a reduced form of the rem demonstrative acel ‘that’. and instead of the bound article when the adjec is pre-nominal (because an article bound enclitically to the adjective would to be in second position): (55) a. but their occurrence in indefinites makes this likely for the Albanian and Romanian particles discussed. This determiner. cel is used in place of the bo article with most numeral modifiers (which appear pre-nominally but cannot h the clitic article): cei doi bZie}i ‘the two boys’. occurs only in definite phra It is used with adjectives in the superlative. The p . recall. These Albanian and Romanian particles are comparable to forms. But Romanian has another free-standing form. It may well be that Hausa na/t a variant of the definite article. clzdirile cele mai mari buildings-DEF the-F-PL more big ‘the biggest buildings’ cele mai mari clzdiri the-F-PL more big buildings ‘the biggest buildings’ Cel is optional more generally with post-nominal adjectives in definite noun phra bulevardele cele mari ‘the big boulevards’. in addition to the bound article w the adjective is post-nominal. gida na sarki or gida-n sarki house PRT-M chief house PRT-M chief ‘the chief’s house’ riga ta Garba or riga-r Garba gown PRT-F Garba gown PRT-F Garba ‘Garba’s gown’ The variation in the form of the particle between na and ta. b. is the definite ar suax) attached to the head. again expre the gender and number of the modified head. b. The examples of cel with a p nominal modifier do look like candidates for treatment as double determinat But in these cases the free-form article appears from its position to be par the adjective phrase rather than qualifying the noun phrase as a whole. usually shortened to -n/-r (which.

mlad student ‘a young student’ (Javarek Sudjik 1963). the dicerence is one of tone (rising in the indefinite falling in the definite).however. Insofar as the distinction does still survive is made segmentally only in the masculine and neuter singular. Len 1982). indeed. first syllable for definite). The South Slavonic languages Serbo-Croat and Slovene show relics of an ol system of indicating definiteness inflectionally on adjectives. Romanian. and no longer necessarily conveying definiten many adjectives. This system is fo in Old Slavonic (a South Slav language dating back to the ninth century). For adjectives that do have two forms.3. Albanian. The tinction between the two adjective declension patterns survives in Serbo-Cr mladi student ‘the young student’.5 Definite adjectives A phrasal clitic article may appear attached to an adjective if is in the appropriate position. and indefinite form in predicative use. novyj bog[ (new-DEF god) ‘the n god’. the “definite” form being now u far more than the “indefinite”. definite and indefinite (de Bray 1980) noun phrase can be indicated as definite by the inclusion of a definite attribu adjective: nov[ bog[ (new god) ‘a new god’. by the position of the stress (sec syllable for indefinite. Here we obviously do not have a bound form and a free form joi expressing definiteness. the definite–indefinite distinction in adjectives is expressed mainly by to 82 . and. and adjectives (and other modifiers) may be m systematically marked for definiteness in agreement with the modified noun noun phrase) as in Arabic. have only a definite form. or. Th is also evidence of a distinction between definite and indefinite adjective dec sions in a number of Indo-European languages (principally Slavonic. Similarly. 2. the two articles occurring are of the same type. In Slovene (Svane 1958. The definite inflection is derived from an old demonstrative stem j-. to a limited degree. wh has two adjective declension patterns. while Romanian does have two article forms. it does not have the kind of double determinat seen in Swedish – both forms occurring together to express the definiteness o noun phrase. Baltic Germanic). on a theoretical level. the distinction has partly become grammaticalized. both (probably) fr form determiners: l’étudiant le plus doué (the student the more gifted) ‘the m gifted student’. For further discussion. and the second article must be taken to be part of the sup lative modifier. the definite fo being obligatory after demonstratives and possessives and with vocatives. Cornilescu (1992) and Giusti (1994). in some adjectives. in the femin singular and the plural. a cl and a free-form determiner. But it is falling into disuse. either in combination and agreement with a definite marker elsewh in the noun phrase or as the only marker. of these issues.

there is another. thos the personal pronoun jìs ‘he’: indefinite definite pron NOM SG ACC SG GEN SG DAT SG INST SG LOC SG báltas bálta bálto baltám báltu baltamè baltàsis báltajc báltojo baltájam baltúoju baltãjame jìs jf jõ jám juõ jamè NOM PL ACC PL GEN PL DAT PL INST PL LOC PL baltì báltus baltg baltíems baltaCs baltuosè baltíeji baltúosius balthje baltíesiems baltaCsiais baltuõsiuose jiB juõs jg jíem jaCs juosè But. for comparison. and. Klimas and Schmalstieg 1972). But the distinction between the two declensions seems to be still me ingful to a greater extent than in Serbo-Croat. as with Serbo-Croat. Macedonian -ot. Bulgarian and Macedonian have completely lost the Old Slavo definite–indefinite distinction in adjective declension. I give h the masculine forms of one adjective. -to) der from a demonstrative. Definite forms are used pronominally (‘white one’ etc. synonymous. -iot rather than -ot in Macedon This -i. -to. according to Dambri5nas. with consequent dual encoding of some grammatical feat – rather as in the addition of the article suax to the noun in Icelandic. báltas ‘white’. The definite adjective declension of Lithuanian is cognate in structure w the Slavonic one. -ta. it takes a dicerent form in the masculine singular. adjective available to the gap. In some cases where an adjec lacks a definite form. simply emphasize the characteristic denoted by the adjective. It is apparent from the foll ing paradigm that the aax is added to the inflected forms of the adjective (w some modification). the adjective stem be extended: -ijat rather than -[t in Bulgarian. But when this second-position article is attached to an ad tive. and. the Lithuanian definite adjective declension has larg if not entirely.tive use. have de oped a clitic definite article (Bulgarian -[t. -ta.). but is replaced by mali for defi function. w attributive. Kli and Schmalstieg. they . thus majhen ‘little’ is indefinite only.extension to the stem is a relic of the older South Slavonic definite ad tive ending. and. being the result of aaxation of a demonstrative with stem (Dambri5nas. lost its original definite value. This demonstrative is still extan the language as the third-person personal pronoun. as seen above.

The definite morpheme is more fully integrated into the adjec inflection.can be discerned in several cells preceding the c morpheme. m%su m#{! m!te ‘our dear mother’. b. ‘a tree’. koks liels koks lielais koks ‘tree’. Latvian no longer has an extant de miner based on a stem j-. the simple definite–indefinite distinction (or its residue) is expressed o in noun phrases containing adjectival modification. The Latvian definite ad tive declension also dicers morphologically from that of Lithuanian. and it may be possible to posit this as an underlying form in paradigm more generally). though similar origin. with no duplication of the encoding of case etc. since Latvian has no other definite arti simple definiteness is expressed only in noun phrases containing an adjective (56) a. Schmalstieg (19 presents the definite adjective declension as still expressing definiteness. the definite–indefinite distinction in adjectives is f meaningful (Budika-Lazdika 1966).other being a matter of subjective choice. the definite form of adjectives is requi after definite determiners. I take adjective balts ‘white’ in the masculine. but in these languages the earlier definiteness marking 84 . As the follow paradigm shows. c. On the other hand. because there is no other defi article. showing either fully functional defin ness marking on adjectives or the morphological relics of an earlier system of s marking. To facilitate comparison with Lithuanian. a discrete definite morpheme cannot be consistently identi (though a form -ai. and. like demonstratives and possessives or genitives: tas lie koks ‘that big tree’. Bulgarian and Macedonian show a development in which a distinct finite article has arisen.or -aj. In the related Latvian. ‘the tree’ ‘a big tree’ ‘the big tree’ As well as expressing definiteness itself. indefinite definite NOM SG ACC SG GEN SG DAT SG LOC SG balts baltu balta baltam balt1 baltais balto balt1 baltajam baltaj1 NOM PL ACC PL GEN PL DAT PL LOC PL balti baltus baltu baltiem baltos baltie baltos balto baltajiem baltajos In most of the languages so far considered.

But. d. as in gender and number.other encodings of definiteness in the noun phrase. traditionally termed either “strong” and “weak”. it can occur in the absence of other definite marking with vo tives and proper nouns and when the adjective itself is inherently definite: k Peter ‘dear Peter’. and attributively with no determiner or after the quasi-indefinite art or other indefinite determiner: kloka vänner ‘wise friends’. i sista stund ‘at the last moment’.11 Thus the strong form of the adjective is typically used dicatively. adjectives agree (in)definiteness. c. with the aaxal definite article. Definiteness marking adjectives in Swedish is essentially an agreement process. which I exemplify from Swedish. Here again. ny ‘new’ strong/indefinite C SG N SG PL klok klokt kloka ny nytt nya weak/definite kloka kloka kloka The distribution of the two declensions correlates precisely with the defin indefinite distinction. and. 11 German too distinguishes strong and weak adjective declensions. den nya bilen denna kloka vän hennes nya man svenska folket ‘the new car’ ‘this wise friend’ ‘her new husband’ ‘the Swedish people’ Less commonly. and with genitives possessives (Swedish being a DG language): (57) a. we must turn again to Scandinavian languages. the definite adjective ending alone does not suace to encode the defin ness of a noun phrase (and even here it can be argued that the definiteness essentially in the proper noun or in the lexical content of the adjective sist ‘la In general the definite adjective form must be accompanied by a definite de miner or article. “indefinite” and “definite”: klok ‘wise’. b. but they do not correlate indefiniteness and definiteness. it is clear that the adjective ending plays o a secondary role in the expression of definiteness. given that one of these can mark a noun phrase as defi in the absence of any adjective. . with whatever component of the n phrase is the principal exponent of these categories. en ny vän ‘a new frie The weak form occurs with definite determiners such as demonstratives and free-form definite article. adject show two declensions. apart from these cases.

the agr ment morpheme -ok encodes first-person singular subject. In several of the Uralic l guages there is object–verb agreement.Many languages have less direct ways of expressing definiteness t the encoding of [+ Def ] by some morpheme within the noun phrase. the obj marking postposition -ro is cliticized to the final word of the noun phrase (Com 1981b): (58) xona-i surx-ro house LINK red OBJ ‘the red house’ Hebrew has a preposition ’et as definite object marker. In Tadzhik. Thus in the indefinite olvas-ok ‘I read’.1 Adpositional marking A number of languages have a prepositional or postpositional ob marker which only occurs with “definite” object noun phrases. also encodes person and num subject agreement features. These dev will be examined in greater detail in later chapters. as well as indicating such an object. -om encodes first-person singular subject and definite object. 2. they are certainly closely associated with it and in some cases c cized to it. and the v inflection. in the definite olvas ‘I read (it)’. and only a brief survey is gi here. the object agreement and subject agreement featu are inseparably fused together. for example.4. and two examples will serve here to illustrate adpositional marking of something close to definiteness. The definite conjugat 86 . the so-called definite and indefinite conjugatio The definite conjugation is used where there is a definite direct object. shows two pa digms for each transitive verb.4. though the n phrase so marked is often “referentially prominent” rather than strictly definite pointed out by Comrie (1978). as well as a prefixal defi article on definite noun phrases generally: (59) ’et ha-mora OBJ the teacher ‘the teacher’ 2.2 Agreement While adpositional markers of definiteness are perhaps not within noun phrase. But in some languages the definiteness of a noun phrase is expres by an agreement marker elsewhere in the sentence. simil for indefinite olvas-ol and definite olvas-od ‘you read’. Hungarian. I shall return to the relationship between defin ness and the object relation.

But if subject or object n phrases are pronominal. These bo clitic pronouns cannot in general co-occur with coreferential free pronouns or n pronominal noun phrases.’ iaa-nhi-ju see PAST 1SG-ERG ‘I saw a child. In related langua other features of the object are encoded on the verb. The Australian ergative language Ngiyambaa does not generally show ag ment (Dixon 1980: 365–6. which is absolu in case (subject of an intransitive verb or object of a transitive verb). This f can co-occur with a co-referential full noun phrase. for example. and this noun phrase is t understood as definite. b. but in all the Uralic langua which have object–verb agreement. definite or indefinite.’ 3SG-ABS burraay. that is. Thus: (61) a. child-ABS burraay. they may appear as bound forms which make up a c complex attaching after the first word or constituent of the sentence. see PAST 1SG-ERG ‘I saw him. Similar facts are to be observed in some Bantu languages. iaa-nhi-ju-na. c. b. U-me-leta kitabu? 2SGSUBJ PERF bring book ‘Have you brought a book?’ U-me-ki-leta kitabu? 2SGSUBJ PERF OBJ bring book ‘Have you brought the book?’ In fact this dicerence between subject and object “agreement” may reflect a dee dicerence. Perrott 1951): (60) a. the person and number of b subject and definite direct object are encoded on the verb.’ iaa-nhi-ju-na see PAST 1SG-ERG ‘I saw the child. Donaldson 1980). 3SG-ABS child-ABS . In the Mordva languages. and with definite di objects (Ashton 1944. the object marker ki being an incorporated pronoun rather than agreement marker. this agreement only applies to defi objects.information about such an object. In this respect Hungarian shows the simples the Uralic object–verb agreement systems (Comrie 1981b). the v shows agreement with the subject. In Swahili. and t are not mere markers of agreement with noun phrases occurring in the sente The one exception to this is a third-person bound form -na. there is no “doubling” of these clitics. See Bresnan and Mchombo (1987) and 5.2 below. and the associated noun phrase being a topic rather than object.

b.4. CLASS cat walk PRF enter come ‘The cat came in. person come PRF ‘The person has come.4 Word order Definiteness can be signalled in Chinese by word order. In Yor (Rowlands 1969). aw*n mi (they book my) ‘my (various) books’. In general. since p verbal noun phrases are constrained to be definite (or generic). English There came a big spider. including Eng and French. come PRF person ASP ‘A person has come. walk PRF CLASS cat enter come ‘A/The cat came in. The p 88 . b. and therefore definite.Some languages allow non-bound personal pronoun forms to app together with overt noun phrases. Mandarin (Li and Thompson 1981) Rén lái le.’ The verbs which allow a post-verbal subject are those of appearance or locat some verbs of motion. these post-verbal subjects are normally understood as indefin (63) a. though subjects of “presentational” verbs also be post-verbal. and it is then constrained to be indefinite: French Il est arrivé d étudiants ‘There arrived two students’. subje are pre-verbal. which are then interpreted as definite. being a state or undergoing a change of state. The interpretation is also definite: (62) a. Unaccusative verbs (or a subset of the allow this argument to occur post-verbally in many languages. in fact the resulting interpretation is of a plurality of individuals opposed to a collectivity: iwe mi (book my) ‘my book’ or ‘my books’. (64) a. 8p8l8p8 enia many person ‘a lot of people’ 8p8l8p8 aw8n enia many they person ‘a lot of the people’ 2. the third-person plural pronoun aw*n ‘they’ optionally acco panies noun phrases to mark them as plural (number not being encoded nouns). These are “unaccusative” verbs. b.’ Lái-le rén le. and a few others.’ Jáu-jó jek m1au yahp-làih. with wh the subject has the thematic (semantic) role of theme rather than agent.’ Cantonese (Matthews and Yip 1994) Jek m1au jáu-jó yahp-làih.

5 Indefinites 2. English a and sm are candidates for this sta a is obligatorily present in singular count indefinite noun phrases in the abse of some other indefinite determiner. The complement m move to subject position. and have no external argument or subject at this level. demonstratives or numerals) ambiguo definite or indefinite – the context usually making it clear which: (65) Mandarin Wv mqi-le shuuguv le I buy PRF fruit ASP ‘I have bought sm/the fruit’ (66) Cantonese Ngóh t3ngyaht wúih wán go leuhts3 I tomorrow will contact CLASS lawyer ‘I’ll contact a/the lawyer tomorrow’ But there are ways of making objects pre-verbal. and in this posi they are (in the absence of. the const tions involved will be discussed in Chapter 6. or other). can take it to be an indefinite article. for example.1 Indefinite article Real indefinite articles – encoding [− Def ]. see Haegeman (1994: chapters 1 and 6). and thus definite. The unmarked position for objects in Chinese is post-verbal. and the same is likely to be true of indefinite artic But if we do find an expression which either obligatorily marks any indefinite n phrase. . English A big spider ca and the Chinese (63a) and (64a).verbs is that they take only an internal argument or complement in underlying st ture. Many languages have determiners expressing so thing like indefiniteness or arbitrariness (like English any. . The point been made in the case of definites that such tests for deciding whether a part lar determiner is an article or not are not foolproof. and sm is normal (though not obligatory .5. and this is what yields the subject–verb order in Fre Deux étudiants sont arrivés ‘Two students arrived’. and in part identifi by not being the same as or readily derivable from a cardinality word – are r if they genuinely exist at all. we have noted languages w optional definite articles. and others indicating specific in finite reference (Spanish cierto ‘a certain’). In part because of their optiona these are not what is generally understood as indefinite articles. For a simple presentation of the concepts of a ment structure and thematic roles (or ‘θ-roles’) and of the “unaccusative hy thesis”. or normally does so but is absent under certain specifiable conditions. some . wh can be optionally present in a noun phrase. 2.

if rough. He does not vary for number. with l appreciable semantic dicerence among them: (67) a. And again. -# is most commonly suaxed to the entire phrase: mard-e-kh%b-#. perhaps reduced. The result is that th are five ways of expressing ‘any good man’. discussed above in 2. to express essenti the same meaning. in which case it replaces the ez!fe: mard-# kh But the determiner yek may also be used (pre-nominally). . ‘some good man or other’. but may optionally appear on a modifier. ‘some man or other’. it is distinct from the modern yek ‘one’. A determiner which looks more lik true indefinite article is Maori he. But we have also seen that these determiners probably best treated as cardinality terms. or in addition to. . it is not unheard of fo quasi-indefinite article identical with the numeral ‘one’ to be compatible with n singular nouns. Like some of the definite aaxes we have seen. b. criterion. d. It is not obligatory. ‘a man’. e. Indefinite aaxes occur in Persian and Standard Arabic. either instead of. It does. 90 mard-e-kh5b-3 mard-3 kh5b yek mard-e-kh5b-3 yek mard-3 kh5b yek mard-e-kh5b . It m alternatively appear on the noun. Recall that Maori an article te (plural ngaa) of unclear semantic value. Interestin this aax is believed to derive historically from a form of the singular numeral. ‘any man’. he wai ‘sm water’. but perhaps definite-spec and a non-specific indefinite article he. and does not simply enc [− Def ]. as we shall see. however. and are linked to it by a particle -e. Windfuhr 1979). react in an interest way with yek. c. the suax -# does not neces ily appear on the noun. and I s examine these more closely. and is approximately equivalent to any in non-assertive contexts some . English a and qualify as articles on this test. But most do again we have here a useful. Persian has a suax -# which may be added to indefinite noun phrases (M 1971. and can used with plural and mass nouns as well as singular ones: he whare ‘a house’ tamariki ‘sm children’. words). To mark the noun phrase as non-spec indefinite.2.called ez!fe. or other in positive declarative contexts. thus: m ‘the man’. however. and need not be present for a noun phrase to be indefinite. semantically. and. -#. Adjectives n mally follow the noun in Persian.4. But he is believed to be diachronic related to the numeral tahi ‘one’. thus mard-e-kh%b ‘the/a good man’.not certain that all articles show morphological or phonological weakness (occ ring as bound forms or as unstressed. mard-# ‘a man’. It thus occurs with a subse indefinites. it marks the noun phrase as non-specific or arbitrary reference.

-# oc in plural as well as singular noun phrases: ket!b-# ‘a book’. It is largely in complemen distribution with the definite prefix and other definite modifiers. it is present by agreem both on the head noun and on any modifying adjectives. semantically weakened in this way. Thus: (70) a. in plural noun phrases. lim to non-specific use. -# is most probably to be seen as a suaxed or clitic cardinal article. unlike English a. like the definite prefix. and which probably cannot plausibly be argued to be a cardinality exp sion (Tritton 1977. c. plural noun phrase is not particularly unusual. ‘sm books’ ket1bh1-3 yek ket1bh1-3 yek ket1bh1 (69) The occurrence of the equivalent of one. though it is tematically absent with certain noun classes (some singulars and some plura and is considered to be absent with dual forms generally. Against this view it has to be observed that. as we will see below. and it may significant that modifiers and determiners in Persian do not show morpholog agreement for number. Moreover. however.greater than that between a and one. It is perh to be likened to the Spanish unos ‘sm’. also be a cardinal article combining the senses of English a and sm. ‘a book’ ket1b-3 yek ket1b-3 yek ket1b a. wh comes after the case inflection. What this means is that appearance in a plural gramm ical context does not by itself prove that a given item must be a real indefi article as opposed to a quasi-indefinite article really encoding cardinality. ket!bh!-# ‘sm boo Surprisingly. al-bust1n-u l-kab3r-u DEF garden NOM DEF big NOM ‘the big garden’ . On ance. wh does not show any such alternation with the singular numeral as is seen Persian. c. it is reasonable to speculate that -# actually represents a quasi-indefinite ca nal article. the numeral yek can also appear. and since yek can express the same ide -#. plural of un ‘a’. b. Standard Arabic has a suax usually described as marking indefiniteness. This putative articl usually presented as having a single form -n (traditionally termed “nunation”). ‘one’. with the same indefi sense. as well as being the singular numeral (‘one’). again either with or instead of -#: (68) a. b. the fact that yek is not limited to singular oc rence suggests that this form. Haywood and Nahmad 1962).

92 fayn-u-n eye NOM INDEF ‘an eye’ fayn-1ni eye NOM-DU ‘two eyes’ al-fayn-u DEF eye NOM ‘the eye’ . and sometimes by variation in case morpholo There is reason to believe. required in certain gra matical circumstances but not encoding anything. that indefiniteness is dire encoded in Arabic. In general. The forms are thus as follows: (71) a. The dua formed by adding the endings -!ni (nominative) and -ayni (oblique) to the st There is no nunation in the dual. indefinites distinguish two cases morphologic (nominative -u. however. for the most part what this means morphologically is that it does not have -n it is identical to the form occurring with al-. The construct state dicers. modification by a possessive (since Arabic is a DG language).in the dual and in masculine external plurals. for these nouns have separate definite and indefinite declensio Whereas definites show the usual three-way case distinction (nominative accusative -a. When a noun modified by a possessive it is traditionally said to be in the “construct state”. usually by -n. The evidence for this conce the distribution of the forms of nouns with and without the -n morpheme. indefiniteness markin not simply absent. genitive -i).garden NOM INDEF big NOM INDEF ‘a big garden’ In fact in those noun classes where -n does not occur. As a result. the marking of indefiniteness is abs only in the dual. and in the plural of a small class of nouns. so the indefinite form is the same as the fo occurring after al-. then. The construct state only dicers fr the form following al. in that the final -ni is dropp leaving the dual endings as nominative -! (undergoing shortening to -a bef the article of the possessor noun phrase) and oblique -ay (modified to -ayi before article). (72) a. that the indefinite form of nouns is an unmarked form. those masculi forming an “external” (non-mutational) plural (inflections: nominative -% oblique -#na). and the noun phr can be made definite in two ways: attachment of the definite article al-. b. -n occurs when the noun phrase is not definite. It appears. oblique -a). on a superficial examination. that this is not the correct conclusion. however.

is dropped in the for tion of the construct state. like the -na and -ni of external plural and dual endings. the “weaker” -n variant is dropped. and feminine external plurals). Where the noun is made definite by attachment of the art al-. Recall that nunation generally. Second. but the phonologically fuller -ni and are not. -#-. First. It is probably a semantically empty ma . dicering only in length. This morpheme is alw dropped in one type of definite context: where the noun is the possessum o genitive construction. but occurs in three variants: -n (occurring in singulars. the -n/-ni/-na ending cannot encode indefiniten since it sometimes co-occurs with al-. (76) x1dim-u-n servant NOM INDEF ‘a servant’ x1dim-5na servant NOM-PL ‘servants’ al-x1dim-u DEF servant NOM ‘the servant’ al-x1dim-5na DEF servant NOM-PL ‘the servants’ x1dim-u l-malik-i servant NOM-PL DEF king GEN ‘the servants of the king’. the second syllabl both the plural and the dual endings contains [n]. but the final -na is dropped (and resulting final -% and -# shortened be a following article) for the construct state: (74) a. (75) a. b. b. for both definites and in finites. -ni (occurring in the du and -na (occurring in masculine external plurals).DEF eye NOM-DU ‘the two eyes’ (73) fayn-a l-bint-i eye NOM-DU DEF girl GEN ‘the (two) eyes of the girl’. ‘the king’s servants’ Now two things are striking here. A plausible analysis is that the indefinite morphem not just -n. most “brok or mutational plurals. the first syllable. closely resembles the nominative or geni inflection of the singular. -#na look bim phemic. -%-. the plural endings -%na. which is formed by addi to the stem of -%na (nominative) and -#na (oblique). But if this is correct. ‘the girl’s (two) eyes’ A similar situation obtains with the external plural. thus closely resembling pho logically the “indefinite article” -n (nunation).

and can be variou rendered by a/sm. in non-singular use only. bu optional. being usual with human referents but much less frequent with non-hum ones. and by the fact that it shares a stem with demonstratives. it can therefore be characteri as an article expressing non-specific indefiniteness. It indicates that the referent is arbitrary or hypothetical.ness because of its partial complementary distribution with definite determin This analysis of nunation and the dual and plural endings dicers from the accou usually given in the manuals. . “common”. wamansu or wasu (plur Whether this is an article or a “complex indefinite” is not obvious – the latte suggested by its not being monosyllabic. it neutrali gender distinctions. . with the result that num but not gender. Interestingly. any. it is identical to the common gen form in the dual and plural: singular dual plural -’ì -rà -p/-ìp So kxòe-P means either ‘(the) persons (male and female)’ or ‘some persons or oth Hausa has a free-form determiner expressing specific indefiniteness (‘a certai wani (masculine singular). wit third gender category. Jaggar argues that its occurrence is determined by the “discourse salien 94 . applicable to gro containing individuals of more than one gender. some . The forms of number–gender morpheme are: masculine feminine common singular dual pl -p/-i -s -kxà -rà -rà -k -t -p Now the “indefinite” suax replaces these suaxes. see also Lukas (1968) and Schuh (1983). Nama has a nominal inflectional suax very similar in function to the Pers -#. or other etc. Let us briefly examine two m instances of possible indefinite articles. is distinguished in the non-specific indefinite forms. wata (feminine singular). but no real indefin article. Nama distinguishes two genders and three numbers. It is used with first-mention indefinite noun phrases. But I believe it is plausible. thus kxòe-ku ‘persons (mal kxòe-tì ‘persons (female)’. Jaggar 1985). in Nama (Hagman 1973) and Hausa (K and Kirk-Greene 1973. But while th is an indefinite form for all three numbers. and if it is corr it puts Arabic among the languages which have a definite. kxòe-P ‘persons (male and female)’..

but dicer in position and grammat status: jek mal ‘one house’. further examples are Armenian meg ‘one’.5. Ashkun aR (Krám 1972). but this is frequently omitted. 2. Tura 1986). m[ ‘a’ (Feydit 19 Samoan tasi. is suggested by the fact the item in question shows the same obligatoriness in singular indefinite noun phr in the absence of another determiner as noted for English. nu. In languages where ‘a’ and ‘one’ are the same phonologically (segmentall least). But the general point made in this section is very many instances of “indefinite articles” derive from cardinality express and can be reasonably argued to be cardinal articles. to be unstressed. whereas numerals tend to car non-zero degree of stress. Basque bat. Amharic and (Leslau 1968). I have suggested above that English a encodes [+ Sg]. Turkish. Turkish also agrees with many o . Plural indefinites may take the determ bazı ‘sm’. with a suax -[k (reduced to -k after a vowel). mal[k ‘a house’. a and one are now so dissimilar phonologically they are not obviously related. Another example of this phenomenon is Kurd (Kurdojev 1957). Many languages are like Kurdis having a cardinal article which is a reduced form of. True markers of indefin ness are rare. Tagalog isa (Matthews 1949).2 Quasi-indefinite article What I have called the indirect signalling of indefiniteness b cardinality word is extremely common and widespread. Albanian një. encodes additional lexical content expressing its contrast with the other num als. in which the qu indefinite article is segmentally identical to the numeral ‘one’. in many languages at least. and recognizably related the numeral ‘one’. like the definite article. in this case numeral and art show a significant phonological similarity. Phonological identity betw the quasi-indefinite cardinal article and the numeral ‘one’ is found in many guages: German ein. Again this is not universal. Turkish bir. Lakhota wA+i. But there is such a dicerence. it is not so easy to motivate a semantic dicerence between them. and may not occur at all. English a exemplifies a cardinal article which is a reduced f of the singular numeral.quent discourse and is likely to recur. Calabrian unu. There is. when it appears in accordance with this requ ment. I have no evidence on the question of whether the Nama and Hausa forms cussed directly encode [− Def ] (or a category hyponymous to [− Def ] ) or if t too are really cardinality terms. Aym ma (Ebbing 1965). French un. moreove strong tendency for this item. as with the definite arti it does not apply to French and other languages not having word stress. Let us examine more closely one language. with the form (Lewis 1967. Dede 1986. wA. one is also [+ Sg]. derived from numeral jek ‘one’ (which is a free lexical item). se.

we GEN guest 1PL remarkable a man is ‘Our guest is a remarkable man. -ler and perhaps bazı in the plu and from the definite. usually included when a singular noun used predicatively is modifi (77) Biz-im misafir-imiz yaman bir adam-dır. when it is a specific indefinite (as in I’ve just bought a sup car). which takes an accusative case suax: (79) a. Where there is no specific referent (for example in I’m looking for a r able car. but in non-initial position ambiguity is possible: (81) 96 Yer-de çocuk yat-ıyor-du. Yesterday letter write PAST 1SG ‘Yesterday I wrote a letter/letters. b. b.’ ‘Children were lying on the ground. ‘Yesterday I wrote sm letters. however. Dün bir mektup yaz-dı-m.’ Dün mektup-lar-ı yaz-dı-m. bir is only used when the indefinite noun phrase refers to a par ular entity. and without number marking. that is. where I do not have a particular car in mind). or where the identity the thing referred to is of no importance for the discourse.’ The speaker here is not referring to any particular letter or letters. yesterday letter PL ACC write PAST 1SG ‘Yesterday I wrote the letters. the restriction of senten initial position to topics in Turkish will ensure that a bare noun occurring initi is interpreted as definite. letter-writing. This incorporated object is clearly distinguished fr the specific indefinite (with bir in the singular. and as subject for example. (80) a. Turkish has the opt of using the bare noun. but is rep ing an activity. yesterday letter ACC write PAST 1SG ‘Yesterday I wrote the letter. ground LOC child lie CONT PAST ‘A/The child was lying on the ground.’ Dün (bazı) mektup-lar yaz-dı-m.’ .’ Dün mektub-u yaz-dı ‘Yesterday I wrote a letter.’ In general. where the non-specific bare noun is a direct object: (78) Dün mektup yaz-dı-m. noun is therefore vague as to number in this “incorporation” construct Consider (78). without determiner.’ Non-accusative noun phrases. a bare noun can be either definite or arbitrary indefin The two would usually be distinguished by word order. are not marked for definiteness. however.

It must be accompanied by amháin ‘only’: aon leabhar amh ‘one book’. Indeed the aon is regularly omitted: leabhar amháin ‘one book’. . But if she intends to convey the latte is clear that she will be satisfied on finding a single book. It is in fact fairly common for such vague indefin ness to be expressed by a form related to or identical to the singular nume English any is derived from the same ancestral form as one with the additio an adjective-forming suax. In a numb languages it has the sense ‘other’. é mod Ob ‘the other man’ (Redden 1979).’ Other “secondary” meanings can be found attaching to the singular numeral.’ (82b). aon means ‘any’. In Lezgian. permits two interpretations: the speaker may be lo ing for a particular book. Irish acords an interesting case. a book look-for CONT 1SG ‘I am looking for a book. that word is ‘one’ is shown by the fact that it is used in arithmetical contexts (A h óna deich sin a naoi ‘One from ten leaves nine’). is-not any money at-me ‘I haven’t any money.the bare noun just discussed involves neutralization of the number distinction. ‘A child was lying on the ground. . the numeral sa ‘one’ (which also occurs. optionally. to give a haon). The singular num is aon (preceded by a particle a when not pre-nominal. like English a. With this sense it can occur in the plural: é bod b[vWg other men’. and with this sense can oc with plurals as well as singulars. Aon is not used as a qu indefinite article. . For an alternative analysis to that just lined see Enç (1991). leabhar ‘a book’. like its English gloss. as a quasi-indefi article) can express ‘only’: sa za-z (one me DAT) ‘only to me’ (Haspelmath 1993). But aon cannot occur alone pre-nominally to c vey ‘one’ either. appearing mainly in non-assertive contexts w mass nouns:12 (83) 12 Níl aon airgead agam. b. who presents bir as fully ambiguous between specific non-specific. U without amháin. indefiniteness is indicated by absence of an article: an leab ‘the book’. or any book. which might not h been the case in the absence of bir. or other’). Yer-de bir çocuk yat-ıyor-du.’ Bir kitap arı-yor-um. and then bir may used: (82) a. We have observed above that in Persian the numeral yek ‘one’ can be use express a vague indefiniteness (‘some . the speaker may wish to mark the noun phrase for number. the accusative case suax is the marker of specifi not (as suggested above) of definiteness. an example is Ewondo: mod Obóág ‘one man’.

b. Examples are Hebrew (Gi 1981) and Mandarin Chinese (Li and Thompson 1981): (84) a. The Persian numeral yek article -#. have one CLASS car block CONT CLASS exit mouth ‘There’s a car blocking the exit.’ But there are also languages in which an optional cardinal article does not sh this restriction. 98 specific Yáuh (y1t) ga ch2 jó-jyuh go ch2ut-háu. and Irish aon. PRT Ying need find one CLASS lawyer ‘Ying has to find a lawyer.’ . generally with stress reduction. b. it is clear from number of the languages considered that an article derived from the numeral ‘o may be restricted to specific or to non-specific use. (85) a. illustrate the restriction of a cardinality term to non-spec vague or arbitrary value. In Cantonese the numeral y!t ‘one’ functions as an optional a cle. And. in the latter case.’ (not a particular one) Mandarin Mén-kvu zuò-zhe yi-ge nán-háizi. Hebrew Hu mexapes isha-xat. Sev writers report on what appears to be a diachronic change in progress in a num of languages previously not having a cardinal article: the increased use of the num ‘one’. either by hav a separate determiner or article for each or by having a single cardinal article wh is restricted to one or the other sense. And Turkish bir tends to be limited to specific use. In other wo these languages are gradually acquiring a quasi-indefinite cardinal article deri from the singular numeral.’ (a particular one) Hu mexapes isha. door mouth sit DUR one CLASS male child ‘In the doorway was sitting a boy. in indefinite noun phrases.’ non-specific A-Y3ng yiu wán (y1t) go leuhts3. ‘He is looking for a woman.’ Wv méi yvu qi1nbu. he look-for woman one ‘He is looking for a woman. I not exist pencil ‘I don’t have a pencil.between specific indefinite and non-specific or vague indefinite. but it seems to be possible in at least some types of non-specific indefinite well as in specific indefinites (Matthews and Yip 1994): (86) a. and in this early stage where the article is optiona is commonly restricted to specific indefinite use. b.

The qu indefinite article wA. C‘w ww ’ag. related to wA+i ‘one’.’ Finally. unas botellas ‘sm bottles’). We have seen this in Persian.’ C‘w wwxi ’ae wo. ‘Put a stick on [the fire]. occurs in plural as well as singular indefinites (una botella ‘a bottle’. with the suggestion that it is possible an article to be a cardinality term yet have no semantic content. But what do the Persian and Navaho quasi-indefinite articles encod will return to this point in Chapter 8. I see man one outside ‘I see a man outside. acting as a k of pleonastic. In languages which l number marking on nouns. is limited to specific indefinites. The article resembling numeral less (presumably because derived from it earlier). the numeral wA+i itself is used for non-specific indefiniteness: (88) a. in which a f of the same determiner. definite and cardinal. inflected for number. with generally obligatory articles. But there are m cases where a cardinal article. c‘A ’etA (wood/trees from) ‘sm wood’). c. for a language combining some of the patterns exemplified. . A major reason for analysing English a as a cardinality expression is tha only appears in singular count noun phrases. A similar pattern is seen in many other languages. alternating with sm (and zero plural and mass phrases. continues express specificity rather than having spread to cover both specific and non-spec indefinites as has occurred in the history of some languages. and another example is Nava where $a’ ‘one’ may express specific indefiniteness both singular and pl (Reichard 1951). This is case in Yoruba (Rowlands 1969): (87) Mo ri 8kunrin kan’ ‘He brought a stick. We can say un(a) encodes [+ Sg] and unos/u [− Sg].’ So here we see the determiner closest morphologically to the numeral ‘one’ expr ing non-specificity (though only in the singular).singular numeral with indefinites where this is optional. cons Lakhota. In such cases it is hardly possible to claim that the article enco [+ Sg] as proposed for English a. A language such as Spanish. is proba not problematic in this respect. c‘A-hAskaska k’eya (wood/trees sm) ‘sm tall trees’. occurs readily with plurals or m expressions or both. s as Lakhota (winE)cala wA ‘an old woman’. clearly derived from and perhaps morphopho logically identical to the numeral ‘one’. wA. the numeral may occur (in contexts where English wo use a) because singular number is a significant feature of the referent.

A dicerence between the two l guages is that Italian also permits plural and mass indefinites without any de miner. This form is particularly favoured in non-asser contexts (Hai comprato pane? ‘Have you bought bread?’). could be glossed literally as ‘so of the bread’. As the last example shows. The partitive article is almost certainly best regarded as a genuine partitive str ture. Italian delle case (of-the hous ‘sm houses’.4 Partitive indefinites Some languages in which a cardinal article is used with sing indefinites use a partitive structure for plural and mass indefinites.5. yek ke with free-form cardinal determiner. while French does not. and plural and mass n phrases take a “partitive article”: French des journaux (of-the newspapers) newspapers’. the possibility of a “genuine” partitive interpretation supports 100 . as in the Persian examples (67)–(69) abo to repeat the relevant data. du vin can mean both ‘sm wine’ and ‘some the wine’. du pain (of-the bread) ‘sm bread’. but notice that double determi tion is also found with cardinal articles. or yek ket!b-# with both. Note that “genuine” partitives. de de l’. In French Italian singular indefinites take an article identical to the singular numeral (Fre un ami ‘one/a friend’. des) are identical to sequences of the preposition de ‘of’ plus the vari forms of the definite article. but the discussion carries over to Italian too. Thus. it sho be noted that in negative contexts the partitive article tends to be replaced by preposition de alone: As-tu de la bière? ‘Have you sm/any beer?’ – Non. but is not limited such contexts (Luigi ha comprato (del) pane ‘Luigi has bought (sm) bread’). and not as an indefinite article. W French does not in general admit such “bare” plural and mass indefinites. by which I mean pa tives over a non-generic domain. 2. I haven’t any beer’. where ‘some’ is expressed by a null element. and ‘the bread’ (r resenting the domain of the partitive) is generic – generics normally being of definite form in French. Further examples are unnecessary here. je n pas de bière ‘No. are also expressed in French by means of de may also have a null head: J’ai bu de ce vin ‘I have drunk some of this wi J’ai bu du vin que tu m’as apporté ‘I have drunk some of the wine that you brou me’. It can be analysed as a prepositional partitive str ture with a null head and with the noun phrase following de interpreted generic. The former type is by far the more co mon. representing de+le pain.The morphological types of quasi-indefinite articles are essentially same as the more commonly occurring definite article types: free-form lexical ite and bound forms (inflectional or clitic). ‘a book’ may be ket!b-# with bound article. Italian una casa ‘one/a house’). In the following remarks I limit mysel French data. because the various forms it takes (du. del vino (of-the wine) ‘sm wine’. du pain.

indicating definiteness: (91) Poika osti kirjat. get-2SG piece ACC cheese ELAT ‘You will get a piece of the cheese. One of these cases is the partitive. For plural and mass direct objects. the partitive case. The French partitive indefinite construction is probably to be related to the of a partitive case for plural and mass indefinites in some languages. The partitive case is o used where the whole of which some part is being picked out is generic (as in above example).’ The situation is not so simple.’ Thus the partitive is well suited to express the same as French de. in its parti indefinite article use. occurs after expressions of quantity: litra maito-a (litre milk-PART) ‘a of milk’.partitive. indicate that the ac . Interestingly in view of the ambiguity as regards the domain of pa tivity in French du vin. as well as indicating indefiniteness in the object. boy-NOM bought book-PL-PART ‘The boy bought sm books. this need not be complemented by any obligatory (or regular) marking of defin and singular indefinites. among o uses.’ (92) Poika osti kirjoja. An important dicerence between these two langua however. is that Finnish has neither a definite article nor a cardinal article. indicating indefiniten stands in opposition to the accusative. Finnis an agglutinating language with a large number of grammatical cases realized sua ally. The two interpretations of du vin match the two interpretations of le as ‘the wine’ (simple definite) and ‘wine’ (generic).’ (90) Saat palase-n juusto-sta. which. The use of the partitive for the ob may. Comp (89) and (90): (89) Saat palase-n juusto-a. however. boy-NOM bought book-PL-ACC ‘The boy bought the books. expressing much of what in such languages as French is expressed by positions (Whitney 1956). Where the whole is a definite set or mass smaller than the gene a dicerent case is used – the elative (typically expressing “extraction”). Finnish makes a clear distinction in such expression partitivity between a generic and a restricted domain. get-2SG piece ACC cheese PART ‘You will get a piece of cheese. Wh languages mark indefiniteness in plural and mass noun phrases by a partitive exp sion.

An intr sitive verb with partitive subject does not show subject agreement as it does w a nominative subject. flower-PL-NOM are vase-INESS ‘The flowers are in the vase. This restrict would be hard to account for if partitive case were assigned within the subj 102 .’ Maljakossa on kukkia. the object noun phrase is not nec sarily understood as indefinite. as [− Sg].’ The fact that the partitivity expressed by the Finnish partitive case does not n essarily take as its domain the noun phrase which bears this case. the “subject” being non-subject position (as in English There entered three strangers) – this being doubt what permits it to take a case other than nominative: (94) a. to indic indefiniteness.’ Plural and mass subject noun phrases may also be in the partitive.the use of a progressive verb form in English. makes it clear that there is no encod of [− Def ] here.’ Tyttö lakaisi lattiaa. so (92) can mean ‘The boy buying sm/the books’. these too must involve a null head. Tyttö lakaisi lattian. but only if the verb is intransitive. the partitive noun phrase bein complement to this head. to ensure that des voitures means ‘sm cars’ (some of the cars there are) and not ‘a car’ (one of all the cars there are). T structure is evidently an impersonal or unaccusative one. On this interpretation the object can also singular count: (93) a. If the analys have suggested for the French partitive article is correct also for Finnish parti indefinites. Kukat ovat maljakossa. But the fact that partitive case is apparently restric to noun phrases not in subject position argues against this view. Note that when the partitive is interpreted as having whole predicate as its domain in this way. but appears in an unmarked third-person singular form. girl-NOM swept floor-PART ‘The girl was sweeping the floor. girl-NOM swept floor-ACC ‘The girl swept the floor. vase-INESS is flower-PL-PART ‘There are sm flowers in the vase. b. subjects of transitives must nominative (and may in principle be understood as definite or indefinite). In fact for French too there is no reason to suppose that the p titive structure encodes [− Def ] any more than the “indefinite article” un do The null head of the partitive (if this is the correct analysis) need only be mar for cardinality. but may be in preted relative to the predicate as a whole. b.

The material discussed here relates to a proposal in the syntax literature unaccusative verbs (those like English enter which permit a post-verbal “subj with pleonastic there in subject position.assigner. with the result that many languages show a dinal article in specific contexts and in some. since partitive case (it is claimed) en indefiniteness. is only available to indefinite complements. as above) are able. English Three strangers entered or French Des ét ants sont arrivés ‘Sm students arrived’). Haegeman (1994: chapter But this hypothesis dicers from the present account of the French and Finn data discussed in that it assumes singular indefinites (un ami) and those with num determiners (trois étudiants) to be abstractly partitive. but not all. These are usually less specific or less referential types. which underlyingly complement. This complement may then advance to the vacant subject p tion (giving. Or it may stay in complement posit subject position then being filled by a pleonastic (There entered three strang Il est arrivé des étudiants ‘There arrived sm students’). I leave this question open. for example. The p is that these verbs take a single noun phrase argument. the verb. agent. 2. because it typically requires a theme or patient θ-role rather th for example. see Chesterman (1991) and Kiparsky (1996). optionally. to as an abstract partitive case to their complement in a number of languages. This latter option is m possible by the option of partitive case assignment to complement position. For discussion see Belletti (1988). non-specific conte Let us exemplify this from Spanish (data partly from Butt and Benjamin 199 Spanish has a cardinal article varying for gender and number: singular plural masculine feminine un unos una unas . We have seen that there are language which the article only appears in non-specific or vague indefinite noun phra but the more common case is for the article to occur only in specific indefin There is then a diachronic tendency for such specific quasi-indefinite article spread into non-specific contexts. The unaccusative facts their relationship to indefiniteness will be further discussed in Chapter 6. For deta discussion of definiteness and indefiniteness and of the range of functions of partitive in Finnish.5 Bare indefinites In languages that have a quasi-indefinite article. there are almost inv ably some types of indefinite in which it does not occur. not just those with a titive article (French) or overt partitive case (Finnish).5.

sex etc. which are commonly therefore “ba Hay vino en la mesa ‘There’s (sm) wine on the table’. in spec as well as non-specific indefinites. ‘Juan is a bachelor. social status. ‘They are repentant conservatives. ‘María became a dentist.’ Juan es soltero. b.’ Son unos conservadores arrepentidos. Trae clavos ‘Bring (sm) na Unos/unas tends to be included where some emphasis on the cardinality is desir Tomamos unas cervezas ‘We had sm/a few beers’.’ Es un ladrón. And it can indicate non-literal use the noun: Son payasos ‘They are clowns’. and equ in many non-specifics. The articl used in singular count indefinite noun phrases with specific reference. Son unos payasos ‘They are (like) clow The article cannot be used with mass nouns.Sg] being encoded in the desinence rather than in the lexical stem. ‘He is a genius. often with little if any dicerence of meani Nos dieron flores ‘They gave us (sm) flowers’. simple indefinites tend to be bare in two major circu stances. b. ‘He is an actor who never finds work. Todavía tenía unos restos fe ‘He still had sm/a few vestiges of faith’. María se hizo dentista. Turning to singulars. Predicative noun phra expressing the profession. in both of which the emphasis is on the descriptive content of the no The first is when the noun phrase is predicative. Es un genio. In the plural. of human beings particularly t to be bare: (95) a.’ But if the noun is modified it usually takes the article (and this applies also plurals): (96) a. b.’ (not a professional one) Omission of the cardinal article in predication is also normal (though not ob atory) with a number of other nouns not easy to characterize in general ter 104 . the article is commonly omitted. Bare plural and mass n phrases will be considered more closely in Chapter 4. ‘He is a thief. Es un actor que nunca encuentra trabajo. where the article is in general obligatory in the abse of another determiner.’ And nouns denoting personal qualities rather than membership of a particular gr or category take the article (compare (unos) payasos ‘clowns’ above): (97) a. and here Spanish is display a characteristic shared by a great many languages.

‘It is necessary to ask for an appointment. defin ness being signalled by absence of the cardinal article. omission of the article is usual under negation: No es problema ‘It i a problem’. the article is omitted after certain prepositions. But on the other hand una pena ‘It is a pity’. while indefinite a cles generally derive from the singular numeral. A complication for this cl is presented by languages with a cardinal article but no definite article. ‘Ana has a car. While definite articles. taking articles to be typically weak fo with limited semantic content. como ‘as’: una casa con jardí house with a garden’. the article tends to be included if the noun is modified. In addition. With this last gro however. 2. Es un problema ‘It is a problem’. and Spanish ocers a good example of a middle-of-the-road guage in this respect.’ Hay que pedir hora. ‘carry’.’ Again. along with other definite determin are associated syntactically with some Det position (yet to be given a pre definition). See also Van Petegh (1989). ‘instead of’. ‘produce’. The reader is invited to compare other languages kno with Spanish to see whether the list of contexts allowing bare indefinites in language is a subset of the corresponding list in the other. The second circumstance inducing bare simple indefinites is when cer nouns (again diacult to characterize. I will address this pr lem in Chapter 8. ‘seek’ among othe (98) a. but said by Butt and Benjamin to typic denote things of which one normally has only one at a time) head the com ment of a particular range of verbs (‘have’. por ‘by way of’. Ana tiene coche.6 General remarks This chapter has surveyed the forms and uses of articles wh express definiteness and indefiniteness. in some more interior cardinality position in the n . articles have t locus.circunstancias ‘He/She is a victim of circumstances’. such as con ‘w sin ‘without’. along with numerals. or quasi-indefinite. Definite articles derive historically overwhelmin from demonstratives (which are also definite determiners). b. which express indefiniteness only indirectly result of never co-occurring with definite determiners. my proposal is that cardinal. and I have claimed that the ter are in fact cardinal articles. Languages with cardinal articles vary greatly in the extent to which b indefinites occur. utilizar su zapato como martillo ‘to use one’s shoe a hammer’.

beyond the scope of survey. It is true that in English both articles are pre-nominal free fo and that in Lakhota both are (probably) post-nominal free forms. I want to maintain the distinction. 106 . or between either of these and salient typological features the language. A point which emerges indirectly from the discussion is that there seems to no necessary relationship between definite article type and cardinal article typ a given language. Moreover. A particularly interesting observation is that so second-position phrasal-clitic articles form a clitic complex with other n phrase features. with the Wackernagel type being parti larly common. In fact we will see in Chapter 5 that the adpositional and agreement ty very often do not involve direct reference to [+ Def ].). is this really “simple” defin ness or indefiniteness? And is there then any significance in the term “articl This term is a traditional one. between articles and other (“co plex”) determiners. And I will argue in Chapter 8 that this intuition can be gi content by treating articles as having a pleonastic function. I think our intuition points to a real distinction here. may be restricted to. Cardinal articles tend to be free-standing determiners. This last point raises the question whether an anaphoric definite article is dic ent in kind from a demonstrative. more importantly. suggesting that what we have here is the nom counterpart of the clause-level auxiliary constituent. agreem etc.(English a) or [− Sg] (English sm). instance. So any impression that anything goes in the way definitenes expressed would not be accurate. and the rough definition I have given – weak fo and minimal semantic content – is one that most traditional grammarians wo be prepared to go along with. anaphoric definite use or specific indefinite use. In other words. Free forms and bound forms are both frequent among defi articles. But both may also carry agreement featu (for gender.). And it is not clear that the post-nominal position of the Lakhota articles lows from the SOV word order of this language. bound artic turn out to be usually phrasal clitics. and “minimal semantic content” is a l vague. such as case. and. howe diacult to pin down. Issues such as this call for deta investigation of the structures of individual languages. however. number etc. but occasionally oc as bound forms. but Roman and Amharic have pre-nominal free-form cardinal articles and suaxal definite a cles. and these two types taken together are overwhelmingly more comm than the other forms of expression discussed (adpositional marking. if a definite article carries a ture [+ Ana] or a cardinal article a feature [+ Spec]. But for some languages it is diacult to demonst phonological weakness in putative articles.

This possibility will be examined below. structures behaviour more widely. it seems to m that definiteness exists in some form in all languages. Some of these expressions are central to an understa ing of what definiteness is and how it works. the person analysis represents the contrast as association with first person or . and the discussion here will be limited to pointing out the most sal aspects. and noun phrases cont ing a demonstrative or possessive modifier) and indefinites was outlined Chapter 1 in relation to English. and rejected this section I anticipate this finding. personal pronouns. On both accounts the speaker fo the deictic centre. 3. that in Chapter 1 I entertained the possibility that this assu tion of inherent definiteness could be mistaken. presumably by some process of semantic weaken Bear in mind. and continue to assume that demonstrat are universally definite in meaning. and will play an important par the discussion in subsequent chapters. and here we look at their forms.1 Demonstratives Demonstratives are probably to be found in all languages. The grammar of these noun phrase ty is complex. Note that if this is correct. an indefinite demonstrative ex ing in the form of such.1 Deictic distinctions The point was made earlier that the two-way deictic contrast exhib in English between this and that can in principle be characterized in terms of tance or in relation to the category of person. and t seem to be inherently definite – which is in part why definite articles almost alw arise from them historically. however.1. 3.3 Complex definites and indefinites The cross-linguistic survey continues in this chapter. in which we our attention to noun phrases whose definiteness or indefiniteness is due to so thing other than presence or absence of an article. [± Prox] is understood in terms of proximity to the speaker. enough to enable us to consider how they fit into the general system definite and indefinite noun phrases. The range of these “comp definites (including proper nouns.

with a two-term contrast. SWlU ‘th something near to or connected with the hearer. this kind of demonstrative system. But languages vary on whether person or distance from the speaker is organizing principle. and the third something quite distant. ano (Dunn Yanada 1958). Each class has a distinct root from wh demonstratives are formed by the addition of suaxes (Redden 1979). But many languages distinguish th demonstratives. onaj (Javarek and Su 1963). ayn (third person) (Feydit 1969). is basic. 2 person assoc. though Frei (1944) claims this system is person-based. ka (distal) (Buechel 1939. he (medial). since the presence of the speaker in a set of individu defines that set as first person. being found in almost all guages. taj. other examples being Modern Armenian ays (first person). but the c of it consists of the following three-way contrast: proximal medial distal i a at’a The first term is used to indicate something which is close to the speaker. Cross-linguistically. a singular–plural pair of mostly human nouns: assoc. This contrasts with the three-term system of Ewondo. and then the situation may be clearer. B and Deloria 1939). 3 person class 1 class 2 kV kVlT kVlí bá bál1 bálí SW ‘this’ indicates something near to or connected with the speaker. second something not particularly close (but not necessarily close to the he either). probably Lithuanian Wìs. I illust from classes 1 and 2. wh the speaker is the primary reference point. and possibly Serbo-Croat ovaj. Other three-term distance-ba systems are: Lakhota le (proximal). 108 . a Bantu language in wh nouns are grouped into a number of classes (some of these classes complem ing each other as singular–plural pairs). 1 person assoc. anàs (Dambri5nas. For English. naa. it is not obvious whe distance or person is the relevant dimension. sono. Such three-term person-based systems are v common. tàs. regardless of whether or not the hearer(s) or oth are also present. ayt (sec person). Lezgian has a rather rich demonstrative system (Haspelmath 1993). and Maori nei. Klimas Schmalstieg 1972). Japanese kono. and SWlí ‘that’ something rem from or unconnected with both.entire category of person. raa (Bauer 1993).

but Anderson and Keenan (1985) claim that distanc the basis.tance or person is the principle involved. the second term of the latter series rel to second person rather than being merely negatively specified with respect to person: proximal medial distal cy-fén(e) y-cíne cy-lúne ‘this’ ‘that’ ‘yonder’ assoc. but Bastuji (1976) argues that Iu is merely an emphatic variant of Spanish ocers an example of a possibly mixed system. but Frei (19 argues that iste does not necessarily relate to second person. aque Languages may even have more than one demonstrative system. Lewis (19 describes the Turkish system bu. But an answer to E:r maY? ‘Do you want this one (near me)?’ cannot make reference to the same ob . The participant-related set appears to have inanim forms only. and another showing a two-way person-rela contrast (Kuipers 1974). and in this make use of both the distance and the person basis. and some languages may mingle the t The Latin demonstratives hic. With many language is simply unclear from descriptions what the basis is. u khaya/khuyu. Shuswap has one series expr ing a three-way distance contrast. iste. but with two series each of two terms (Hoc 1968). as with Aymará aka. like those of Latin. I want this one here’. and has mo (animate) and e:nï (inanimate) to pick out something in the immediate vicinit the speech event (thus apparently corresponding to both proximal and media standard three-term systems). 1 person assoc. said by Ebbing (1965) to correspond to Spanish este. relating to or concerning the spea and mo:ro for something not associated with the speaker. distance-based set distinguishes animate from inanimate reference. with e:ro for something near. e:nï se wa ‘Yes. Thus the question nï se maY? ‘Do you want this one here?’ may receive an answer in which is repeated: A:a. Kennedy 1962). and mo:kï (animate) and mo:nï (inanimate) to p out something more distant. though. o as based on three degrees of distance f the speaker. regardless of distance from hearer (¿No ves ese coche? ‘Can’t you see that car?’). Iu. and can also contrast with aq as indicating lesser versus greater remoteness. traditionally regarded as based on person (see example Ramsden 1959). ille are traditionally considered to relate to son (for example Gildersleeve and Lodge 1895. curiously. ese. aquel are. The three forms este. It seems in fact that ese can contrast with este as indicating connec with the hearer as opposed to connection with the speaker (Ojalá que yo tuv ese talento ‘I wish I had that talent (of yours)’). 2 person cy-cif cy-cey ‘this (by me)’ ‘that (by you)’ Surinam Carib is similar.

aqu (second person). with the merging of the first. I give demon tives and personal pronouns together for comparison: demonstratives ASS1EXC ASS1INC ASS2 ASS3 e:ro e:nï (mo:se) mo:ro mo:nï (mo:kï) personal pronouns singular plura 1EXC au 1INC kïxko 2 amo:ro 3 moxko afna kïxka amïiy moxk The distinction between inclusive and exclusive first person plural. on Ho analysis. Thus aquesta galleda may be ‘this bucket’ (which I have) or ‘that buc (which you have). correspond). kïxko. would have the same kind of contrast in its distance-based ser between distal and non-distal. the “[Ass1]” form e:ro expresses association with the speaker o first-person singular. has larg fallen into disuse. Catalan exemplifies a rather common diachronic phenomenon. Carib seems to have this distinction in singular as well as plural. though Carib. especially in the spoken language. wi “singular” inclusive pronoun form. glossed by Hoc as ‘you and I’. and its function has b taken over by aquest. the reduction three-term demonstrative systems to systems of two terms. that this language may have a single demonstrative system based on more elaborate person tinctions. so that Catalan has ecectively a two-term system.and second-person functions un one form. 110 . made by many langua amounts to having two expressions corresponding to we. or which n ther you nor I have). but shows an interesting pe liarity. But the middle term. or distal non-distal. the contrast is between third person and non-third person.1 The Catalan system resembles the Spanish one.4.So. I suggest the correct analysis distinguishes association with exclusive and inclu first person as well as with second and third person.1. not with first person in general as is standardly the case person-related systems. Such a system is extremely unusual. where “proximal” is defined in terms of proximity to the spea But in Catalan. aqueix. Now l guages with two-way demonstrative contrasts nearly always follow the Eng pattern of a first-person or proximal form contrasting with a non-first-person non-proximal one. kïxka presumably refers to a set larger than two which includes speaker and hearer. on this analysis. Old Fren 1 This oddity suggests that Hoc’s analysis of the Carib demonstratives may be mistaken. its plural. one of which includes the he and the other not. For fur discussion see 3. The evidence is that distinctions of kind are made in the personal pronoun system (with some minimal similarities of f between pronouns and demonstratives which. aquell (third person). In principle there are three demonstratives: aquest (first person). The three terms of L were reduced at a much earlier period in some other Romance varieties. and aquella galleda is ‘that bucket’ (over there. unusually.

too. ’éí ‘remote and invisible’. suggesting that the same process is at work here. preferred for crete. these are supplemen 2 3 Interestingly. three demonstratives: proximal dieser. the others being formed by the addition of suaxes. with den här (used in speech and informal writing) and denna/denne (essent restricted to writing) for ‘this’. Navaho has a similar system. distal jener. Malagasy has a pu distance-based system. yôo ‘less close to the speaker. Lléí ‘ tant from both speaker and hearer’. . iroa. but distinguishing only two degrees of tance: proximal houtos and hode. and the latter for referring to something about to be mentione something approaching or coming into view. medial der. shows more forms than de distinctions. Wolof bii or bile. the outcome of this reduction in spoken German is a system in which der f tions as a general demonstrative without deictic content (like French ce). Malherbe and Sall 1989). iny. but with terms (Young and Morgan 1987): díí ‘near the speaker’. Classical Greek has three principal demonstratives. ai (Newmark. with six degrees of distance in relation to the spea (Anderson and Keenan 1985): ity. there are languages with f five or even six demonstratives organized in the ways discussed. by far the most commonly occurring type. preferred for abstract or present referents) for ‘that’ (Holmes and Hinchlice 1994). Persian #n. io. The distinction betw houtos and hode is not clear cut. iry. Latvian Wis. in for varieties. Some further instances Mandarin zhèi ‘this’. here medial form appears to have largely taken over the distal earlier three-term system in the distal forms yon. yôno. perceptible referents) and den (stylistically unrestricted. yonder (archaic in mod spoken English though still in use in some dialects). which in careful speech is i tical phonologically apart from its lack of stress (Durrell 1991). !n (Mace 19 Biblical Hebrew zeh. Hubbard and Prifti 1982). the term. is relatively infrequent. German too has. distal ekeinos (Goodwin 1992). mentio above as distinguishing three degrees of distance. I suggested ab that Carib may have a four-term person-related system. with the following forms be those used for nouns in class 1: yôno ‘immediately next to the speaker’. itsy. nagháí ‘near the hearer or at some distance from the speaker’.2 In Lezgian. yô ‘ atively near the speaker’. the distal form at’a is (Haspelmath 1993). Swedish. but the former is sometimes preferred for referring to so thing already mentioned. But last has largely fallen out of use in the spoken language. especial contexts where der might be confused with the definite article. ’eii ‘relatively near speaker’. and usually near the hea yôlê ‘remote from speaker and hearer’. Hidatsa héo. Khalkha Mongolian en[. Notice that the second term is morp logically basic. be bale (Mbassy Njie 1982. ter[ (Po 1970). tas (Budika-Lazdika 19 Albanian ky.3 Despite this predominance of two-term systems. and den där (speech and informal writing. in fact. Two-term syst are. hû’ (Lambdin 1973). nèi ‘that’ (Li and Thompson 1981). while dieser is both as a proximal form and as a second general form without deictic restriction. Tswana combines b distance and person in its system (Cole 1955). ka (Matthews 1965).

see below. Thus tamten is distal and ten d tically unmarked (Kryk 1987). without deictic content. and the latter form. the most commonly occ ring (Lee and Lee 1959). tamten ‘th Polish has in theory a two-term system. The deictic distinctions illustrated in this section are not essential to dem stratives. ten ‘this’ and tamten ‘that’. generally in addition to distance or person (or both). the general demonstrative is ce. ce jardin-là ‘that garden’. ce jardin-ci ‘this garden’. Czech a general demonstrative ten. but has in addition a distal form ace used when it is necessary to make a deictic contrast (Murrell and jtefznes Drzgzneiti 1970). make distinctions of height. dukha. ya ‘close to both spea and hearer’. Latin has. French is a case in point. Some languages. as seen in some of the examples above. in ad tion to the three degrees of distance mentioned (largely reduced to two). Other languages have a deictically unmarked demonstrative wh is not basic in the sense of being the form generally used except in contrastive otherwise special circumstances. wa ‘close enough to point at’. In the same way Romanian uses acest(a) as a g eral deictically unmarked demonstrative.tance and association with person: va ‘close to the speaker’. this form too is devoid of deictic content. there is another form to be discussed below. and a much more common situation is wh such a general demonstrative coexists with others which do make deictic dist tions. Bender and Yamamoto 19 There are many languages in which the demonstratives involve other dei categories. hic. In such cases the unmarked demonstrative m be partly or fully restricted to particular functions. are largely limited to contrastive contexts. nyu ‘close to the hearer’. iste. But where one needs to be more exact there are fo available incorporating the deictic particles -to and tam-: tento ‘this’. This kind of system is unusual. but this can reinforced by the deictic enclitics (on the noun) -ci and -là: ce jardin ‘this/ garden’. has demonstratives distinguishing “superiority” and “inferiority”: wini ‘that up the and aTa ‘that down there’ – though these too are falling into disuse. in addition to series expressing three degrees of deixis. Certainly one of the common uses of is is to pick out 112 . rather it is additional to a fully specified dei system such as those discussed. of labelled “anaphoric”. otherw ten is used with the deictic contrast neutralized. ha ‘far aw tha ‘very remote and perhaps invisible’ (Watahomigie. and there are demonstrative systems in which no distance or person c trast is expressed. Dyirbal (Dixon 1972). I have mentioned Egyptian Arabic. an unmarked form is. see Frei (1944) Anderson and Keenan (1985). Similarly. with a single demonstra da. A fairly comm one is visibility. ille. expressing contrast. Lezgian. with further examples. A similar general purpose demonstra is found in Supyire: Zké kàn-he (DEM village-DEF) ‘this/that village’ (Carl 1994). For use surveys of demonstrative systems. But the tinction.

And Finnish supplements its two-degree series. b (distal visible). Nouns in Dyirbal are usually accompanied by a c marker which encodes deixis: yala.ative clause. general demonstrative occurs more freely. it is not accompanied in the noun ph by the definite suax on the noun. However. medium and l distance uphill.2 Non-deictic distinctions Demonstrative systems may involve other. the unmarked. with unmarked demonstrative se (Whitney 1956). Thus proximal forms ten serve for anaphoric reference to things very recently mentioned and distal fo to things mentioned further back in the discourse. serves simply as an alternative to the deictically marked forms. A language with a single demonstrative which is not. balbalu for medium and long distance downriver. dayi. Person-based systems too al the same anaphoric use. which can be added to this demonstrative or to any non-demonstra noun marker. The first consists of gala ‘vertically up’. Dyirbal also has several sets of secondary de markers.(invisible). and distinguished from the free-form defi article only by being fully stressed. with three-term distance-based system mentioned above. Lakhota. tuo ‘that’. The most obvious of these is anaphora. daya. non-deictic categorie addition to a deictic basis. g ‘out in front’. it seems to lose this sense in combination with the second markers.1. bayd¸a. however. Then there are the sets bayd¸i. But yala. In addition there are guya ‘ac the river’ and bawal ‘a long way’. as in is vir quem vidisti ‘that man whom you saw’. and non-first-person or (often) third-person form similar distal use. while the demonstrative giSa itself is proximal. gali ‘vertically down’. also has a general demonstra ’e. balbala. Xala. also encoding [+ Prox] and the only true demonstrative – and only oc ring in absolutive case. deictically n tral is Dyirbal (Dixon 1972). It is nearly alw the case that the distance distinctions discussed can relate to temporal and e tional distance as well as spatial distance. so giSa-bawal (this-long+way) expresses something like ‘that over the 3. dayu indicating short. Interestingly.(proximal visible. This is the case in Latin (generally taken to have a person-ba . As well as the four forms expressing two degrees of de discussed. daw dawalu for medium and long distance upriver. or deictically neutral). This demonstrative only occurs as anteced to restrictive relatives: de bilar som jag gillar bäst ‘those/the cars that I like be In other languages. unlike this. with first-person forms admitting what is in ecec extended use as proximal. tämä ‘this’. and that demonstratives expressing th can pick out referents in the preceding generally replaced in absolutive c by giSa-. Swedish has a deictically unmarked form den – identical to the di demonstrative except that. med and long distance downhill. bayd¸u indicating short. A similar si tion obtains in Swedish.

has an “immediate anaphoric” demonstra ino:ro (animate). wit frequently appearing anaphorically. The basic h.8 ti mo ra lana yi naa cloth which I bought yesterday this ANA ‘this (already mentioned) cloth which I bought yesterday’ Demonstratives cannot usually co-occur.and -le demonstratives can be used anaphorically. The anaphoric fo can therefore be comparable in use with the anaphoric function of the general dem strative in Latin. of course. yYn ‘that’. and non-proximal. the forms with -o can only be so used. plus an anapho form naa (Rowlands 1969). Swahili has a two-way distance-based deictic syst formed by attachment to the class marker of the proximal prefix h. Yoruba also has a two-term deictic series. It is also common for pr imal or first-person forms to be used anaphorically to pick out something pr ously mentioned by the speaker. hivyo. Now. as more versus less recently m tioned) for anaphoric reference. This is so of Surinam Carib. and indeed I used such co-occurrenc Chapter 2 as evidence for anaphoric determiners in some languages being arti 114 . though in some cases it seems to have the more specific sense ‘just mentioned’. Latin again illustrates this. the anaphoric demonstrative is the proximal form but with a suax -o (which triggers some changes to the ro hicho. we have proximal hiki ‘this’. non-first-person second-person forms to pick out something mentioned by the person currently be addressed – and. Unusually. which. especially where the distance back into previous discourse of the earlier mention of the referent is not an issue. Typic this does not prevent the deictically marked forms occurring anaphorically. to take two classes representing singular and plural forms of inanimate nou with the class markers ki and vi respectively. distal or third-person forms for a referent mentio by some third party. naa can co-occur with the dei forms: (1) as. i:ro (inanimate). For a not necessarily immediate anaphoric fo consider Swahili. in addition to the comp deictic system discussed above. though. Languages with a general. yi ‘this’. or the distal suax -le (Ashton 1944. a Bantu language like Ewondo. ‘these’. with demonstratives formed the root of class markers. deictically unmarked demonstra may also use this in the anaphoric function. But many languages have a special demonstrative for anaphoric use. distal and proximal demonstrati frequently have the senses ‘the former’ and ‘the latter’.followed reduplication of the root vowel.mally used (often in contrast with each other. Perrott 195 Thus. vile ‘those’. es cially where there is reason to indicate that the earlier mention was recent or rem or to make a recent–remote contrast between two referents. In this use. and distal kile ‘that’. medial.

since a topic is likely to have been just mentioned. rìgá-n nàn gown DEF this ‘this gown’ rìgâ-n nán gown DEF this-ANA ‘this gown (previously mentioned)’ kújèrá-r càn chair DEF that ‘that chair’ kújèrâ-r cán chair DEF that-ANA ‘that chair (previously mentioned)’ A final observation on anaphoric demonstratives is that. and then they indicate the referent is the current topic. with senses such as ‘also’. particularly for th limited to immediately preceding mention. where the proximal form dìní can be used in reference to so thing under discussion and the distal form nùcú for something mentioned ea (Cook 1984). hii haifutiki ‘That [sc. and anaphoric refere quite dicerently. Ashton (1944) and Perrott (1951) comment this use approximates to that of a definite article. this cannot be’. This is case in Sarcee.since it does have some adverbial uses. b. Some languages have a special demonstrative to express contr Egyptian Arabic has been mentioned as having a basic one-demonstrative sys with no deictic distinctions: da ‘this’ or ‘that’.and -le forms can appear pre-nominally. Contrast between referents is expressed in a variety of ways cross-linguistic Swahili does not permit the use of the proximal and distal forms together contrast (as in English this and/or that). ‘indeed’. The tone on these forms is usually falling or low. b. But there is in addition a form du . the deictic h. But changing tone to high (with a preceding high-tone syllable changing to falling) signals vious mention: (2) a. In Hausa. This is the case with Lezgian ha. These forms occur post-nominally. But a particularly clear instance is provided by Swahili. (3) a. they often serve also as the express for ‘same’. We have seen above that Swahili makes a two-way distance c trast and has an anaphoric demonstrative. has been said’. in wh topic is expressed by the position of the demonstrative. anaphoric demonstrative reference is indicated by t (Kraft and Kirk-Greene 1973). nan ‘th can ‘that’. The tendency is rather to use the proxi form twice: Hii yafutika. It is often diacult to distinguish such topic demonstratives f anaphoric ones. Hausa too has a two-way deictic system. sin] can be blotted out. and Nama //xaá (Hagman 1973 Demonstratives can also encode the fact that a referent is the current to of the discourse.

Nama has a basic two-way distance contrast: nee ‘this’, //nãá ‘that’. But it is
possible to use these two forms contrastively; instead a third form, náú, is u
and can contrast with either of the two basic terms. Náú expresses the further m
ber of a contrasted pair, and so naturally occurs frequently with the proximal dem
strative: nee kxòep tsHí náú kxòep ‘this male person and that male person’. Bu
neither referent is near, náú can be contrasted with //nãá.
Finally, many languages have emphatic demonstratives. The process of reinforc
demonstratives morphologically, for example by the addition of further deictic p
ticles, is common, and a reinforced form sometimes becomes the basic one w
a previous basic form weakens semantically to become a definite article. But b
and reinforced demonstratives co-occur widely. Lithuanian has an emph
demonstrative Wìtas formed by combination of the basic proximal and medial for
Swahili uses reduplication for emphasis: hiki hiki ‘this’, kile kile ‘that’.
Swahili has another, more complex, emphatic structure, based on the anapho
demonstrative; this form (hivyo in the plural class illustrated above) is prece
by a double reduplication of the root, but with the two vowels of the anapho
form. Thus vivyo hivyo ‘these very same’.

3.1.3 Forms and positions
Demonstratives overwhelmingly take the form of lexical items, a
English, and they are almost invariably stressed.4 This is a major point of cont
between demonstratives and definite articles (for languages which have word stre
and is important in diachronic considerations, because definite articles in ne
all languages that have them are descended historically from demonstratives. I
in fact usually a deictically unmarked demonstrative, or a non-proximal or n
first-person one, which provides the source of a definite article. One conseque
of this is that in many languages the definite article is segmentally identica
very similar to one of the demonstratives (though dicering in stress). An exam
of this is German, where unstressed der is the article, and the same form with st
is a demonstrative unmarked for distance. In Danish, similarly, den is both a dem
strative and (with reduced vowel: [dan] ) the free-form definite article. Recall, h
ever, that definite articles occasionally show deictic distinctions, and then they m
well derive from first-person or proximal demonstratives as well as others.
Not all demonstratives are free-form lexical items. It was observed in Chap
2 that demonstratives in Bella Coola are enclitic on the noun, like the defi

English provides a clear illustration of this point. The complementizer that, which serve
introduce certain types of subordinate clause, is normally unstressed and pronounced wi
reduced vowel: [2at]. But the demonstrative that never has this pronunciation; it always h
full vowel, because it is in principle a stressed form: [2æt].


nomenon of three of its six demonstratives (those representing the second, fo
and sixth deictic degrees) being pre-nominal lexical forms and the remaining th
being suaxes on the noun. Lango has phrasal-clitic demonstratives, suaxal to
final word of the noun phrase (Noonan 1992): gwôkk à dwóX-Xì (dog LINK big+
this) ‘this big dog’.
A number of languages have discontinuous demonstratives in which part of
information is conveyed by an aax or particle of some kind. In French the dem
strative ce is a pre-nominal determiner, but the optional deictic element is a su
on the noun: ce livre-ci ‘this book’, cette voiture-là ‘that car’. The two elem
come together in the corresponding pronominal demonstrative, but with a dic
ent form for the determiner element: celui-ci ‘this one’ (masculine), celle-là ‘
one’ (feminine), and also ceci ‘this’, cela ‘that’ (forms not marked for gender
somewhat similar situation is found in Irish, but the post-nominal particle is ob
atory, and the pre-nominal determiner element is the definite article: an leab
‘the book’, an leabhar seo ‘this book’, an leabhar sin ‘that book’, an leabhar
‘yonder book’. The pro-nominal forms consist (not surprisingly in view of the r
tionship assumed between pronouns and definite articles) of the appropriate th
person personal pronoun plus a form of the same particles: sé seo ‘this o
(masculine), sí sin ‘that one’ (feminine), siad siúd ‘those’, ‘yonder ones’. Th
Irish facts look at first sight like counterevidence to the claim that demonstrat
involve (universally) more than a combination of the definite article and a dei
element. But the suaxal element can in certain circumstances occur independe
of the article or personal pronoun, as a complete demonstrative equivalent to
French genderless forms ceci, cela, referring to a pro-position or an entity wh
cannot be assigned a gender: Is trua sin (is pity that) ‘That’s a pity’. This se
to indicate that seo, sin and siúd, unlike French -ci and -là, are not merely dei
particles (variants of anseo ‘here’, ansin ‘there’, and ansiúd ‘over there’, as -ci
-là probably are of ici and là). Rather they are complete demonstratives, mar
[+ Dem] as well as with an appropriate deixis feature, but generally requirin
be accompanied by an expression of [+ Def ]; this is not an unusual requirem
as we shall shortly see. It should not be ruled out, however, that there may be
guages with expressions of deixis not marked [+ Dem] – perhaps, therefore,
having true demonstratives. We have seen that Dyirbal has only one genu
demonstrative, a proximal form, with no demonstratives expressing other dei
degrees. This is presumably possible in Dyirbal because three degrees of de
are expressed anyway on most noun phrases, and because there is a large bat
of secondary deixis markers which can be added. I suggested above that
anaphoric “demonstrative” of Yoruba could be an adverbial particle rather tha
determiner; in fact the other forms too have adverbial uses: ‘here’ or ‘now’ as w

or something similar rather than determiners, it may be that they express only de
and are not true demonstratives.
Dyirbal and Yoruba are among the languages which do not express simple defin
ness, and it would not be surprising if in such languages demonstratives (or
nearest thing to demonstratives) were not of the same syntactic category
articles (in other languages) and occupying a “Det position”, as is the case
English. In Dyirbal the one demonstrative occupies the same position, and occ
in place of, the class marker which generally appears in noun phrases. In ot
languages demonstratives are accompanied by a classifier while themselves be
in what may be a determiner position. This is the case in Chinese:

Cantonese (Matthews and Yip 1994)
n3 go
this CLASS problem
‘this problem’


Mandarin (Li and Thompson 1981)
that six CLASS book
‘those six books’

In Cantonese a linker morpheme ge may also occur: n# go ge jitmuhk (
CLASS LINK programme) ‘this programme’. In Latin, another language with
expression of simple definiteness, demonstratives show considerable freedom
position, like adjectives, and there is probably little reason to treat them as c
gorially dicerent from adjectives.
Turning now to languages which do have simple definiteness marking, th
can be grouped informally, and roughly, into two types as regards the relations
of demonstratives to the definite article, and, I will suggest, the position
demonstratives in the noun phrase. There are languages like English in which dem
stratives replace the definite article, and there are languages in which the two
occur in the noun phrase. In fact the situation is complicated by the existence
(at least) two dicerent types of definite article, free-form determiners and aax
which appear to be of dicerent syntactic categories and in dicerent structural p
tions. But on the current assumption that demonstratives are [+ Def ], it is not s
prising that they should be in a paradigmatic relationship with the definite art
in many languages, such as English. In these languages no co-occurrence is p
sible because demonstratives are of the category Det, like the definite article,
occupy the same structural position as the definite article. If there is a posit
what I have termed “Det position”, which free-form definite articles univers
or generally occupy, it seems natural that demonstratives too, as definite de
miners, should also appear there.

miners co-occur with the definite article. Examples are Standard Arabic h
al-bust!nu (this DEF-garden) ‘this garden’; Ewondo é mvú S# (the dog t
‘this dog’; Swedish den här bil-en (the/that here car-DEF) ‘this car’; Armen
ajt gadou-n (that cat-DEF) ‘that cat’; and Irish, mentioned above. Care is nee
in assessing the significance of these observations, because this second group
languages at least is very heterogeneous. In some of them, Arabic, Swedish
Armenian for example, the definite article is an aax, either on the head noun
as a phrasal clitic, attached to the noun or to some other element. These bo
articles do not look like determiners in a structural position comparable to tha
free-form articles (though they may be associated with such a position, perh
via agreement). This Det position should then be available to be filled by a dem
strative. In other words, in some languages where a demonstrative co-oc
with a definite aax, it may be that the demonstrative is a determiner in the s
Det position as in English (where the demonstrative replaces the article). Th
probably the case in Swedish, where there is no obvious reason for drawing a dic
ence of position between the free-standing definite article den and the demonstra
to which it is closely related. And since both co-occur with the definite aax,
must be in some other position. The only dicerence between these languages
the first type is that the article, or one of its variants, is something other tha
determiner. On the view that this aaxal article is a morpheme representing ag
ment with a Det (which may be phonologically null), a demonstrative in Det p
tion would be the definite Det with which the aax displays agreement.5
But there are also clear cases of demonstratives co-occurring with a lexical defi
article, which is itself probably in Det position. Irish and Ewondo exemplify
situation. In these languages the demonstrative is probably not of category De
Ewondo it is probably deeper inside the noun phrase than the article and may
adjectival, but not in Irish, where it can be shown to be external to the noun phr
Noun phrases like Ewondo é mvú S# (the dog this) and Irish an madra seo (the
this) ‘this dog’ look similar, but the formation of pronominal demonstratives is q
dicerent. In Ewondo the pronominal form corresponding to é mvú S# is é S#,
article remaining in combination with the demonstrative. But the correspond
form in Irish involves a personal pronoun rather than the article: sé seo (it/he th

It is important to note that there is not always complete uniformity within a language. Whe
the demonstrative determiner of Armenian must be accompanied by the article suax
Swedish this varies from one demonstrative to another. The forms den, den här and den
are accompanied by the article suax on the noun, but denna is not. This form, also pres
ably a Det, follows the pattern of Danish, in which the article suax does not co-occur
any demonstrative. In Albanian too, demonstratives (also probably determiners) usually o
without the article suax appearing on the noun; perhaps here and in Danish the article se
to license a null Det. But the suax can appear in Albanian, as noted in 2.3.3.

here must be outside the noun phrase.6
In some languages demonstratives co-occurring with the definite article
more clearly adjectival. In Spanish and Catalan this is the case optionally –
other option being that the demonstrative occur without the article, in typical p
nominal Det position: Spanish este país (this country) or el país este (the co
try this) ‘this country’, Catalan aquella ciutat (that city) or la ciutat aquella
city that) ‘that city’. When the article is present the demonstrative must follow
head noun, in typical adjective position. A similar situation obtains in Maori
other Polynesian languages (Krupa 1982, Bauer 1993). Maori has a three-way dem
strative system based on the roots nei, naa, raa, which combine with a prefix
in the singular, ee- in the plural; tee- is a form of te, the singular definite-spec
article, though the normal plural of this is ngaa. Thus: teenei whare ‘this hou
eenei whare ‘these houses’. But there is an alternative, in which the demonstrat
without prefix, follows the noun, and the noun is preceded by the article (in its s
dard form): te whare nei ‘this house’, ngaa whare nei ‘these houses’.
Adjectival demonstratives also occur in languages with aaxal definite artic
Romanian shows the same kind of optionality as Spanish and Catalan, though
demonstrative dicers in form as well as position according to the pattern us
The unmarked position is again before the noun, probably in Det position,
then the article suax does not appear: acest apartament ‘this flat’. The emph
option is to place the demonstrative post-nominally, in adjectival position; it t
takes a final -a (as in pronominal use) and the article suax appears: apartame
ul acesta. Unlike ordinary adjectives, however, the demonstrative in this posit
takes case markers as it does when in Det position: acestui domn or domnu
acestuia ‘of/to this gentleman’. Hausa is another language in which demons
tives can be pre-nominal, in a fuller form in this case, without the article su
appearing, or post-nominal, possibly in an adjectival position, accompanied by
suaxal article: wancan gona (that farm) or gona-r can (farm-DEF that) ‘that far
Let us summarize. Demonstratives, as inherently definite expressions,
occur in Det position. Here they normally replace a free-form definite article, thou
as in Maori, they may occasionally be incorporated with it. In languages with
aaxal article, this can co-occur with a demonstrative Det, as in Swedish
Armenian, but more commonly it is omitted when the demonstrative is in Det p
tion (Danish, Albanian, Romanian, Hausa). Demonstratives can also occur in ad
tive position, or in other modifier positions such as the noun-phrase-external posi
of Irish. These are positions which have no particular association with definiten

Irish demonstratives are in fact to be grouped with some other items, including the emph
particle sa/se and féin ‘self’, ‘even’, of uncertain category status, and perhaps adjoined to
noun phrase (see McCloskey and Hale 1984, Lyons 1992b).


which have articles. For languages which do not, a similar range of positions se
to be available, though it may be that what can be interpreted as Det positio
not. Finally, demonstratives occasionally occur as aaxes, both in languages wh
express simple definiteness and in languages which do not.
These observations suggest that, in languages in which definiteness clearly ex
as a category, as evidenced by encoding of simple definiteness, demonstratives m
be associated syntactically with it. This association can take the form of the dem
strative occupying a position specialized in the expression of definiteness: Det p
tion or the aax position where definite articles appear. Or it can take the form
the demonstrative occurring in some modifier position not associated with defin
ness, but then with the noun phrase being marked as definite by the article.
might suppose that if demonstratives are inherently [+ Def ] there should be no n
for them to be accompanied by the definite article or otherwise to show a s
tactic association with definiteness.7 And indeed, at least some languages with
simple definite markers manage without this association. A solution to this ap
ent redundancy is to propose that demonstratives are not lexically [+ Def ], in
language, as so far assumed. Instead, they are merely constrained to appear onl
definite noun phrases through being semantically incompatible with indefiniten
This constraint may be met either by the demonstrative being in a structural p
tion which expresses definiteness or by there being in the noun phrase a sepa
encoding of definiteness. This speculation will be developed in Chapters 7 an

Proper nouns
Proper nouns show the general behaviour of definite noun phrases, tho
as suggested in Chapter 1, it does not necessarily follow that they carry the specifi
tion [+ Def ] in languages such as English in which they do not take the defi
article. Most languages follow this pattern, but there are also languages in wh
proper nouns are obligatorily accompanied by the definite article, and in these c
it cannot be doubted that they are [+ Def ]. One example is Modern Greek
already noted: ho Georgos ‘George’. Another is Albanian, in which it will be reca
that the definite article is a Wackernagel aax. Proper nouns, including name
persons and places, almost always occur in definite form: Agim-i e pa Dritë-n ‘A
saw Drita’. They only occur without the definite aax in a few contexts, in m
of which common nouns too would be “bare”: when complement to certain, mo

One might propose that those demonstratives which co-occur with the article are discontin
items, with [+ Def] encoded separately from the main body of the demonstrative where the o
features are expressed. But this is not an attractive idea. It would remain to explain why t
are so many discontinuous demonstratives, and why non-discontinuous ones cannot occu
adjectival positions.

definite noun phrase (në fshatin Dushk ‘in the village of Dushk’); when used
a vocative (Prit, Lumtë! ‘Wait, Lumtë!’). Of course they are also “bare”, tho
clearly not indefinite, when modified by a preceding adjective, which must ca
the second-position article aax: i zi-u Petrit (PRT poor-DEF Petrit) ‘poor Pet
There is considerable language-internal variation, many languages using the defi
article with proper nouns either optionally or in certain circumstances. In Mod
Western Armenian, proper nouns in the accusative, dative and ablative must t
the suaxal definite article (and it is used with genitives in speech), but the a
cle is not permitted with proper nouns in the nominative and instrumental (
genitive in writing). In German, it is common in colloquial usage to use the a
cle with first names (die Claudia, der Hans), and this usually conveys famil
ity. With surnames the use of the article is not necessarily colloquial, and car
complicated pragmatic and socio-linguistic connotations. It is common, even
formal contexts, with the names of celebrities (actors, writers etc.), but tend
be limited to female ones (die Dietrich, die Droste, but Schiller, Goethe) – a po
which many now find ocensive. The colloquial use of the article with surnam
more generally carries a connotation of assumed familiarity, which can be used
denigrate or praise (Der Brandt ist doch ein Säufer ‘Brandt drinks’, Der Brand
ein ganz kluger Kopf ‘Brandt is very clever’). In Italian too it is common co
quially to use the article with names, but only female first names (la Maria, l’An
but Luigi); with surnames the article is used for prominent individuals both m
and female. In Sissala the article is optional with names, and simply serves
emphasize the fact that the person named is known to the hearer (Blass 1990
An important observation is that, even in languages like English where pro
nouns are almost invariably bare, they can take determiners, as in this Peter,
Annie. As already noted, this may be a matter of the proper noun being rec
gorized as a common noun, no longer treated as unique; this would be the c
if Peter and Annie are being picked out from others of the same names. But
is not necessarily the function here of this and my, which can be non-restric
– with acective value for instance.8 On this interpretation, Peter and Annie
probably still proper nouns. Longobardi (1994) points out that ‘my John’ can
expressed in three ways in Italian: il mio Gianni, il Gianni mio, Gianni mio; *
Gianni is impossible. Of these possibilities, il Gianni mio is strongly contrastive,
therefore probably involves recategorization; il mio Gianni need not be contrast
so could correspond to the acective, non-restrictive English interpretation. T
dicerence is expected, because post-nominal possessives in Italian tend to be c
trastive, pre-nominal position being the unmarked one. But the most interest

This is particularly clear in the northern British usage our Annie, your Peter.


t is no doubt that lui is a special form of the definite article. however Ana ‘to/for/of Ana’. I will return to t An interesting point is that many languages use a special article form w proper nouns. but a special f (en rather than el) is used before masculine proper nouns beginning with a c sonant: la Maria. lix Rosa ‘Rosa’ (Eachus and Carlson 1966) Catalan the definite article is optional with proper nouns. such as Taga (Krámskm 1972): ang gur$ (the-NOM teacher). whic pre-nominal underlyingly. Longobardi also claims that where a defi article does appear overtly with a proper noun (or with a generic). This point. and almost identical to the oblique mascu singular definite article suax of common nouns (un prieten ‘a friend’. Maria. the article is omitted lowing a preposition: din BucureIti ‘from Bucharest’. but only in the nominat Kekchi has li as the definite article with all common nouns (li ixk ‘the woman wi. and despite its only occurring in oblique functions. which happens to be identical to feminine definite article suax: Ana. so that the phrase is structurally equ lent to one containing a definite article. gara ‘the station’. This is the case in many Austronesian languages. prietenul ‘the frie prietenului ‘to/for/of the friend’).9 A final. . not used with com nouns (though it does occur with the names of months and years): Sandu va trimite o scris lui Petre ‘Sandu will send a letter to Petre’.need not be understood contrastively. in this way noun itself plays the role of the article. but si Marya (the-NOM Ma In Maori the “personal article” ko (distinct from the usual definite article te used with proper nouns and personal pronouns. But most feminine place names keep definite suax -a: Constan}a. feminine names ending in -a may optionally change the -a to -ei (on the oblique forms of the feminine definite article suax) instead of being preceded by lu seems that the final -a of such names is optionally interpreted as the nominative–accus definite suax. and not on oblique functions: BucureItiul ‘Bucharest’. which normally take the definite article – the common article. N in oblique functions. din Constan}a. it is a m pleonastic without substantive content – not [+ Def ] therefore. As with common nouns. together with the unaccepta ity of a pre-nominal possessive without article. The behaviour of feminine pr nouns makes this clear. compare o garQ ‘a station’. Many feminine names end in -a. But despite this. Generalizing from this. This determiner lui is identical in form to the culine oblique third-person singular pronoun.nk ‘the man’). but distinct male and female article forms with personal pro names: laj Manu’ ‘Manuel’. It is not restricted to masculine proper nouns. The ambiguous status of final -a in Romanian is supported by the behaviou place names. en Prat (Yates 1975). Personal na occurring as subject and direct object do not require an article. curi 9 Romanian presents a complex picture (Murrell and jtefznescu-Drzgzneiti 1970). leads Longobardi to propose (wi a DP framework) that in Gianni mio the noun has moved into the Det posi which the article would otherwise occupy – passing over the possessive. but en Pere. it has the same interpretation as the nominal possessive of il mio Gianni. he argues that all articleless pro nouns in Italian involve movement of the noun to Det position. l’Eduard. but in “oblique” functions (es tially indirect object and possessive) they are preceded by a form lui.

though w it is now (case aax. and their pronominal and predicative forms mine. since their syntactic behaviour is ess tially the same (though for some minor dicerences see Lyons (1986a) ).case by what seems to be the indefinite article (Andrade 1933). The main reason for supposing my etc. T same historical development has occurred in some other Germanic languages. yours. The dichotomy is far from watertight. may warrant a dicerent analysis fr the ’s forms is that a common pattern. gender and case do determiners and adjectives. postposition. however.. theirs) and full noun phrases (Peter’s. even Det) is a matter of some debate. the pronoun possessives my etc. yo their etc. Russian. wher may be attached to a word of any category (the man I was talking to’s dog). It is longer tied to the noun. not Denmark.3. since it is common for pronouns to be highly irregular in th morphology. th is little reason to draw a major distinction between possessives based on prono and possessives based on full noun phrases. s as Danish: Kongen af Danmarks hat (king-DEF of Denmark-POSS hat) ‘the k of Denmark’s hat’ – where the king. pronoun possessives. Chom (1981) characterizes both as genitive case forms. ben-im (me-GE ‘mine’. Examples are German. is the possessor of the ha ’s is treated as a case aax.3 Possessives 3. is for possessi derived from pronouns to be adjectives or determiners and those derived from noun phrases to be genitive case forms. limited to the pronoun-based possessives.1 Pronoun and full noun phrase possessives English has possessives corresponding to personal pronouns (my. may well also be g itive case forms. Greek permits a genitive case form mou as an alternativ 124 . while possessives based on full noun phrases in an invariable genitive form. La Greek. where the pronoun-derived possessives for the most part inflect to ag with the noun (or other noun phrase constituents) in number. In Turkish personal pronouns nouns inflect in essentially the same way (though see the discussion of Turk possessive aaxes below): Ahmed’-in (Ahmet-GEN) ‘Ahmet’s’. Apart from the distinction between attri tive and pronominal/predicative. gakusei no nim (student POSS luggage) ‘the student’s luggage’. this woman’s). Japanese uses g erally a structure which has been variously considered to involve a postposition a case suax: watasi no nimotu (me POSS luggage) ‘my luggage’. The ’s of full noun phrase p sessives derives historically from a genitive inflection on the noun. Many other languages make no distinction between personal prono and full noun phrases as regards the formation of possessives. 3. in inflecting languages. but appears right at the end of the noun phrase.

3. except that this takes a dicerent form from the n possessive personal pronoun. sv but genitive case forms are used in the third-person non-reflexive (Latin eius ‘his/h Russian jevo ‘his’. Latin and Rus have adjectival/determiner forms in the first and second persons and third-per reflexive (Latin meus ‘my’.or hoi adelphoi mou (the brothers me+GEN) ‘my brothers’. sáá ’oms ‘your house’. these aaxes are often phonologically identical to personal inflection pronominal clitics appearing on the verb and representing a subject or object a ment of the verb. and fuses with the particle -a producing so morphophonemic modifications: kiti ch-a-ngu ‘my chair’. This is the tern found in most of the Romance languages. kiti ch-e-tu ‘our ch Another common pattern distinguishing personal pronouns and full n phrases is for possessive determiners or adjectives to exist corresponding onl the former. Possession is expressed in Swahili b particle -a which attaches suaxally to a class marker determined by the pos sum noun. 3. the person and number of the subject are adequately signa . and is followed by the possessor expression: kiti ch-a Hamisi (c CLASS-POSS Hamisi) ‘Hamisi’s chair’. Examples from Turkish are: köy-üm (village-1SG) ‘my villa el-in (hand-2SG) ‘your hand’. jejo ‘her’). but often dicerin position (superficially at least) from the pronoun-derived forms. mojeho bratrowe dNSKi (my+M+SG+G brother’s+PL+NOM children+PL+NOM). çocuk-lar-ımız (child-PL-1PL) ‘our childr Turkish is a “null-subject language”. nearly always appearing on the head noun. instead a pre sitional construction is used – still with possessive meaning. with pre-nominal possessive determiner. It is exactly the same w a pronoun possessive. the possessum. subject personal pronouns only being u for emphasis. In estingly. but l’ami de cet homme (the fri of this man) ‘this man’s friend’. A tern which may be related is found in Nama. Russian moj.2 Affixal possessives The possessives relating to personal pronouns are in many langua realized as aaxes. tvoj. suus ‘his/her own’. Moreover. and for the latter to have no genitive case form either. tuus ‘your’. several Slavonic languages have posses adjectives derived from full noun phrases (Corbett 1987): Upper Sorbian Jan kniha (Jan’s+F+SG book+F+SG). ’áop tì ’oms ‘the man’s hou The prepositional construction is sometimes an optional alternative to geni case. where full noun phrase possess take a particle tì but are not obviously in a dicerent position from pronoun p sessives: tíí ’oms ‘my house’. So the possessive agrees with the p sessum through its association with the class marker. for example French son ami friend’. as in German: das Haus des Mannes and das Haus von dem Mann ‘the m house’ (the prepositional option being more colloquial). with post-nominal prepositional possessive.

In current syntactic theory. e:tï ‘name’. Returning to Turkish. the possess personal aaxes can be reinforced. Carib has personal prefixes. by a personal pronoun in genitive case: siz-in sokaT-ınız (you-GEN street-2PL) ‘your street’. and we can take it that Turkish possessive suaxes are also associated w pro in the absence of an overt free-form genitive form. v closely related morphologically to the possessive aaxes. to:pu/to:p ‘stone’. for emphasis. In other languages. while in others they can. is phonologically null. and Jaeggli and Safir (1989) for a m advanced treatment. and is parallel to the use of a subject pronoun in null-sub languages to reinforce the personal marking on the verb. So we can assume that the personal aa are not themselves pronoun possessives. In some languages p sessive aaxes are purely pronominal in sense. attac to verbs (and prepositions). This is u ally for emphasis. the phonologically null p noun. but markers of agreement with lex 126 . Khal Mongolian has no personal morphemes on verbs relating to either subject or obj but does have possessive aaxes. which can be used either instead of or in addition to the aaxes. while showing a lot of allomorphic variation. -tan ‘your (plural)’). Languages with aaxal possessives almost always have free-form possessi also. The Turkish personal aaxes occur not only with pronom possessors. y-e:tï ‘my name’. Egyptian Arabic a has personal suaxes. the empty subjec null-subject languages is usually identified with pro. See Haegeman (1994. such as Finnish. un emphatic. uses the same forms for nouns. Examples are: garabiyyit-na (car-2PL) ‘our car’. again. but also with full noun phrase possessors: çoban-ın kız-ı (shephe GEN girl-3SG) ‘the shepherd’s daughter’.subject–verb agreement are. -man ‘our’. but. in this case occurring equally on nouns. these personal suaxes represent the object rat than the subject. ve and prepositions. A peculiarity is that they attach to an extended form of the no if the noun has such a form. wayyaa-na (with-2PL) ‘w us’. the poss sive suaxes are distinct from the personal aaxes appearing on verbs. a-to:puru ‘your stone’. and the third-person aax (-n/-in/is of demonstrative origin. -Rin ‘y (singular)’. A major dicerence from Turkish is that. except that the fi person singular is -ni for verbs but -i (after a consonant) or -ya (after a vow for nouns and prepositions. verbs and p positions. Thus. ch ter 8) for an elementary discussion of pro. fihmuu (understand+PERF+3PL-2PL) ‘they understood us’. The para with the use of nominative pronouns to emphasize a clausal pronominal sub suggests that the possessive aaxes are to be analysed as agreement morphem encoding the agreement of the head noun with a genitive pronoun which. The suaxes are the same for all three categories. For the first and second persons they are redu forms of the genitive of the corresponding personal pronoun (-min ‘my’. or m accompany full noun phrase possessives.

Notice that. -nin ‘their’. thus Tikrani zk\sd-[ ‘Tigrane’s garment’. it is unclear whether it occurs here to double possessor noun phrase. These forms may be reinforced b personal pronoun: miye mit‘awa ‘mine’. incident is zero. as wel being used pronominally and predicatively. beca the third-person aax is zero. It is also possible for the plural pronouns in the genitiv occur without the corresponding personal suax. There are also fuller forms which are f standing. both hiapió kaiíi (him house) and hi hi kaiíi (him 3SG house) are possible. Th follow the noun: _Ekak‘A mi-t‘awa kC (horse 1SG-PRT the) ‘my horse’. but the possessive aaxes ma accompanied by a genitive pronoun for emphasis: hattu-nsa (hat-3SG) ‘his hat’. where the possessive structure is expres some relationship other than ownership: meidän kylä (us-GEN village) ‘our village’. The s set of clitics can represent both objects and possessors. If a free-form possessive occurs. is Pirahã (Everett 1987). Doubling is optiona that for ‘his house’ with emphasis on the possessor. and this is prefixed to the no mi-t‘a-_Eka kC (1SG-PRT-dog the) ‘my dog’. the form relating to each person in Class Armenian (-s. like the Egyptian Arabic suaxe hi xibáobá (I 3SG hit) ‘I hit him’. . it must be doub by a suax. as opposed to the possessive particle having no prefix attac Modern Armenian has a particularly interesting system of possessive sua which are descended historically from a definite article distinguishing th degrees of person-related deixis. The third-person form. 10 The same applies to Finnish as regards full noun phrases. Corresponding plurals have been formed with a morpheme -nis ‘our’. agreement between the two possessive elements is limited to the singular. both free and bound posse expressions being possible either alone or in combination. ‘her’). -nit ‘your’. A full noun phrase possessor also oc with one of the particles t‘a and t‘awa: Peter t‘a-_Ekak‘A kC (Peter PRT-horse or _Ekak‘A Peter t‘awa kC (horse Peter PRT the) ‘Peter’s horse’. so ‘his dog’ is t‘a-_Eka kC. im s\can-s ‘my table’. in the f of a full noun phrase or a personal pronoun in the genitive. *m\r doun-nis.on the other hand. thus if the possessor is a full noun phrase a suax does not occ The Lakhota possessive prefixes are attached directly to the noun only cases of inalienable possession: ni-c‘iye (2SG-elder+brother) ‘your elder broth Otherwise they are attached to a particle t‘a. -d. One guage which allows a great deal of freedom in these respects. formed by attaching the personal prefix to a fuller particle t‘awa. but -n/-[ is use double all plural possessors as well as third-person singular: thus m\r doun-[. -n) has given the possessive of that person (-s ‘my’. It may be that what is appearing on the noun in s instances with a plural free-form possessive is not a possessive suax but the defi article (which is also -n/-[). the possessive suaxes do not in general “double” overt l cal possessives.persons singular.and second. -t ‘yo -n/-[ ‘his’. ‘our house’. A full noun phrase possessor too may be doubled: x kaiíi (foreigner house) or xaoói hi kaiíi (foreigner 3SG house) ‘the foreigner’s house’. -s -t are the suax forms for first. hi kaiíi (3SG house) ‘his house’. hän-en h nsa (him-GEN hat-3SG) ‘his hat’.

Dyirbal dr a distinction between present and past possession. Most of these distinctions have no bear on the question of definiteness in possessives. family relations etc. 3. T illustrates the typical pattern. and that the possessive is enc to this initial noun/adjective–article combination. and that the term is informal one which may cover a number of distinct categories of relation. as shown for Finnish above. items of clothing. We have seen that in English and some other languages nouns of this type that permit a definite possessive structure with apparently indefi sense (my brother. Bauer 1993). expressed by two dicerent g itive inflections. ‘my house’ is la mia casa (the my house) but ‘my mother’ is 128 . in terms of the nature of the relation holding between possessor possessum. but not necessarily. The distinction between alienable and inalienable possession.pronoun-based possessives are Wackernagel forms: maIina mea cea n (car+DEF my the new) or noua mea maIinQ (new+DEF my car) ‘my new c Notice that the Wackernagel article (which must be present in this definite str ture) attaches to the first word of the noun phrase. and I limit my attention to one them (though a rather generally defined one) which may have some bearing. In Ital for example. the latter invo ing possessa which are more intimately or intrinsically tied to the possessor – b parts. a and o. your ear. and the sec otherwise (Biggs 1969. Mary’s sleeve with no implication of uniqueness) many languages nouns in this class involve dicerent possessive structures fr those with “ordinary” noun heads. Past p session may involve an implication that the object denoted by the head still belo to the past possessor. ownership may be distinguished from mere physical p session or from a general association. the first used where the poss sor is in a position of dominance or control over the possessum. is central to the possessive sys of many languages. For example.. The Lakhota examples above show a posses particle to which personal prefixes are attached appearing in alienable possess but the prefix being attached directly to the noun in inalienable possession. -Xu for the former and -mi for the latter (Dixon 1972). having been lost or temporarily left. the ecect again being closer integration of the possessive with noun: mwenzi wako (companion your) ‘your companion’ reducing to mwenzio is important to bear in mind that what counts as inalienable for linguistic p poses varies considerably from one language to another. Swahili shows an optional process of possessive reduction in inal able possession. M has two possessive or genitive particles.3. which is that in inalienable possession the struct is morphologically simpler or the possessive is in some way structurally close the head noun.3 Alienable and inalienable Numerous distinctions are made by languages between types of p session.

incidentally. but the head nouns allowing omission of the a cle in this way are limited in Italian to a small number of kinship terms. as well as having two genitive morphemes for present and past p session (both alienable). also distinguishes an inalienable possessive construc with no genitive marker. but instead the head can be ceded by an adjectival particle (with no possessive following). in the t person the prefix is always zero. I illustrate the two constructions in (6) (alienable) and (7) (inalienabl (6) bayi waial baiul yaoaiu CLASS-ABS boomerang-ABS CLASS-GEN man-GEN ‘the man’s boomerang’ (7) balan d¸ugumbil mambu CLASS-ABS woman-ABS back-ABS ‘the woman’s back’ Kinship terms. But when a posses precedes there is no definiteness marking: im atë ‘my father’. which conv the sense of a third-person possessive: e ëm-a (PRT mother-DEF) ‘his/her/t mother’. are treated as alienable in Dyirbal.involving inalienable possession. But whereas nouns are generally acc panied by a class marker in Dyirbal. John hE-ku kC ‘John’s mother’. This pattern of the definite article tending not to occur in inalien . Possessor and possessum are simply juxtaposed in a n phrase. When a possessive follows the head the latter must carry the defi inflection unless there is a preceding indefinite determiner: tren-i im (train-D my) ‘my train’. In Lakhota the possessive prefix is attached directly to an inalienably posses noun. tem-a jote (theme-DEF our) ‘our theme’. To see the r vance (though not the precise significance) of the distinction for definiteness us turn to two languages which have articles. and kinship heads with a fi or second-person prefix generally take no definite article: ni-hE ‘your mother’ Albanian pronoun possessives (which are adjectives or determiners) follow the h noun except in the case of inalienably possessed heads (mainly kinship terms some terms of social relationship). mi. when first.respectively. Kinship terms and body-part terms dicerent prefixes for the first-person singular. without an intervening possessive particle: ma-p‘oge kC (1SG-nose the) nose’. Lakhota and Albanian. with both nouns carrying the case inflection determined by the synta function of the noun phrase as a whole. a full noun phrase possessive is simply placed before the head noun: J p‘oge kC (John nose the) ‘John’s nose’. third-person possessives cannot precede. Dyirbal. jot g jyshe ‘our gra mother’. But there are some interesting distinct within the class of inalienable possession. but kinship heads usually take a suax -ku: hE kC ‘his/her mother’.and second-person forms gener precede.and ma. the possessum noun in this construction c not be.

French ma bicyclette. In DG languages possessives do not co-occur with articles. so that an article is also required: la bicyclette de Jeanne (the bicy of Jeanne) ‘Jeanne’s bicycle’. German ein Freund von mir (a friend of me).Italian. and a definite article cannot also appear the second type a possessive does not have this ecect. In languages like English in which noun phrase possessives have essentially the same distribution and behaviou pronoun possessives. the reason tra tionally assumed for this being that they are themselves definite determin Examples are English my bicycle. are not inherently definite or semantically incompatible with indefiniten in the noun phrase they modify. 11 It has been claimed that the morpheme ’s is a definite determiner to account for the defi ness of a girl next door’s bicycle. and this is neutral with respec definiteness. and is widespread. In Fre only pronoun possessives follow the DG pattern. German mein Fahrr all meaning ‘the bicycle belonging to me’. they too compel definiteness in the matrix phrase. Re that in the first type a possessive has the ecect of inducing a definite interpr tion in the noun phrase it modifies. Full noun phrases can only used as possessives in a prepositional structure. This will be discussed in Chapter 8. ‘the wife of a friend’.4 Possessives and definiteness The most important issue arising from possessives for our concer the (informal) distinction between “DG languages” and “AG languages”. Irish cara liom (friend with+ ‘a friend of mine’. In fact such a prepositional construction is the most common strat used in DG languages to obtain the indefinite reading which is unavailable w the “basic” possessive structure. despite the second having an indefinite possessor expressi a girl next door’s is clearly not a definite determiner. Danish: en vens kone ‘a friend’s wife’. une bicyclette à Jeanne (a bicycle to Jeanne) ‘a bicy of Jeanne’s’. example. The phrases the girl next door’s b cle and a girl next door’s bicycle are both definite (‘the bicycle belonging to th girl next door’). and the article must occur with it to get a definite interpretation (in languages that have an artic The central observation in this discussion is that possessives. And fact makes it very clear that the DG phenomenon cannot be accounted for in te of DG possessives being definite determiners.3. 130 .11 The same applies in. For further examples and discussion of the syntax of alienable and inaliena possession see Haiman (1985). Some examples are French un ami à moi (a fri to me). 3. unlike demons tives.

b.1). No article is associated with the possessum. see Lyons (1992b). with possessive aaxes doubling free-form possess (at least in the first. The Egyp Arabic example above. Another ex ple is Modern Armenian. The possessum is obligatorily b unmarked for definiteness or indefiniteness. is definite (‘the car belo ing to us’) though the article prefix cannot appear: *hil-garabiyyit-na. But it is understood as definite fact it is stated in much of the literature that the possessum (or.5. Arabic layn-a l-bint-i eye NOM-DU DEF girl GEN ‘the eyes of the girl’. defin ness is marked by a prefixal article in Arabic and a free-form article in Irish. this is almost certainly incorrect.the traditionally termed “construct” of the Semitic languages (usually mislabe “construct state” in the recent syntax literature. The AG–DG parameter applies as much to aaxal possessives as to lex ones. Essentially the same p sessive construction characterizes also the Celtic languages. apart from the dicerence in word order. the descriptive facts in relatio definiteness are exactly the same as for the DG structures of English and o languages. and the latter cannot the usual definiteness marker. the matrix being in preted as definite without the appearance of a definite determiner. ‘a bird’s neck’ In these examples the possessor phrase may be definite or indefinite. indefiniteness by nunation in Arabic (see 2. Many possessives of this kind are of the DG type. If the language has articles.and second-persons singular): im kirk’-s (me+GEN book-1 ‘my book’. ‘a girl’s eyes’ Irish muineál an éin neck the-GEN bird-GEN ‘the neck of the bird’.1 above). garabiyyit-na (car-2PL) ‘our car’. ‘the girl’s eyes’ layn-a bint-i-n eye NOM-DU girl GEN NUN ‘the eyes of a girl’. but it is definite.5. with a dicerent meaning. also from Sem philology. The construct is a construction in which the possessive follows the head. (9) a. see 2. I illustrate from Standard Arabic and Irish: (8) a. these can co-occur with possessives to indi . ‘the bird’s neck’ muineál éin neck bird-GEN ‘the neck of a bird’. b. For present purposes the p is that. using a term. AG languages are those in which a possessive does not induce a definite in pretation. more accura the matrix phrase) is interpreted as definite only if the possessor is definite.

një libër i saj (a book PRT h ‘a book of hers’. një libër i vajzë-s (a book PRT girl-GEN+DEF) ‘a book of the girl’s’. Krámskm 1972). Peter t‘a-_Ekak‘ (Peter PRT-horse the) or _Ekak‘A Peter t‘awa kC (horse Peter PRT the) ‘Peter’s hor Peter t‘a-_Ekak‘A wA (Peter PRT-horse a) or _Ekak‘A Peter t‘awa wA (horse P PRT a) ‘a horse of Peter’s’. but with an aaxal definite article and a (free-form) cardinal a cle: libr-i i saj (book-DEF PRT her) ‘her book’. Chinese lacks articles (though hav optional incipient ones in the unstressed use of distal demonstrative and singu numeral). uma nossa casa (a our house) ‘a house of ou Other determiners than the articles can equally appear: esta nossa casa (this house) ‘this house of ours’. Mam has only a cardinal article. A in Mandarin Chinese. tou and koin$nos (the-GEN man-GEN partner-NOM) ‘a partner of the man’s’. so these noun phrases are simply ambiguous between definite and indefi (Huang 1987). nèi-ge háizi-de péngyou (that-CLASS child-GEN friend) ‘that chi friend’ or ‘a friend of that child(’s)’. This makes it clear that the AG type is not limited to langua which have articles (though the DG type probably is). à otherwise). One language w aaxal possessives and no articles which shows evidence of being AG is Na 132 . They are in Chamorro (Ch 1987. This is the case with English possessives (that/a friend of Mary(’s) ). I have already observed that prepositional possessi are generally AG in ecect. and have to be combined with an appropriate de miner to express definiteness or indefiniteness. sos koin$nos (your partner) ‘a partner of yours’. ho tou andros koin$ (the-NOM the-GEN man-GEN partner-NOM) ‘the man’s partner’. libr-i i vajzë-s (book-DEF PRT girl-GEN+DEF) ‘the g book’. only pronoun possessives are non-prepositional): a no casa (the our house) ‘our house’. Alban works similarly. AG languages in which full noun phrase possessors are expres by a genitive case form and appear in the same position as pronoun possessi include Classical Greek (Goodwin 1992): ho sos koin$nos (the your partner) ‘y partner’. and is AG (England 1983): n-jaa-ya (1SG-house-1SG) ‘ house’. Lakhota too is AG. where there is a t dency to use a dicerent preposition depending on the determiner (de with the defi article. both in its prefi possessives and free-form ones: mi-t‘a-_Ekak‘A kC (me-PRT-horse the) or _Eka mi-t‘awa kC (horse me-PRT the) ‘my horse’. its absence imp ing definiteness. juun n-jaa-ya (a 1SG-house-1SG) ‘a house of mine’. and even in French. which has a definite article while simple indefinites characterized by absence of an article: i paine-kku (the comb-1SG) ‘my com paine-kku (comb-1SG) ‘a comb of mine’. pronoun and full noun phrase possessives are formed exactly the same way: wV-de péngyou (me-GEN friend) ‘my friend’ or ‘a fri of mine’.languages more generally. mi-t‘a-_Ekak‘A wA (me-PRT-hors or _Ekak‘A mi-t‘awa wA (horse me-PRT a) ‘a horse of mine’. Aaxal possessives too can be of AG type.

As well as being nor and unmarked with the cardinal article. the dicering somewhat in form. and this is odd. It is therefore constructions. The essen dicerence between DG and AG possessives. so w it seems odd that some demonstratives should appear with the definite article the case of possessives this co-occurrence involves no redundancy. Spanish has both DG possessives (mi casa ‘my house’. wh the AG possessive is in a typical adjective position and shows full adjectival ag ment behaviour. But dem stratives apparently dicer from possessives in being inherently [+ Def ]. as with the two types of demons tive. historically. the AG forms. seems to be one of position. There are also mixed languages – in fact a great many languages combine ments of both patterns. this being why they appea Det position. and their definiteness explains wh quasi-indefinite article cannot co-occur with them. except that the latter rarely take the form of aaxes. since we have established that there is no log 12 See Lyons (1986b) for discussion of the phonological reduction of DG possessives in e Romance resulting from their being unstressed. the AG possessives also occur with o determiners (as seen above in Portuguese): aquella casa mía ‘that house of mi They also occur in predicative and pronominal use: Aquella casa es mía ‘that hous mine’. the DG structure is unmarked one for expressing ‘my house’ and the like. definite) and possessives (la casa mía ‘my house’. una casa mía ‘a house of mine’). while the DG possessive is in a paradigmatic relationship w the definite article and is. Free-form AG possessives typically adjectival and are positioned deeper within the noun phrase. La mía es blanca ‘Mine is white’. Free-form DG possessives behave like defi determiners. which are always post-nominal. a reduced form like the articles. Una mía es blanca ‘One of mine is wh The distinctions drawn in this section are remarkably similar to those seen am demonstratives. This characterization is supported the dicerence of position and form of the two possessive types in Spanish. their position accounts for a free-form defi article not being able to accompany them. infl like adjectives for number and gender.which should be impossible if these prefixes induced definiteness: y-o:tïrï (1 something) ‘something of mine’. they are marked [+ Def ]. while most of the DG forms are morp logically reduced (like many determiners) and inflect for number only. that are or AG. not languages. and since they leave Det position free this can be filled by the defi article or some other definite determiner – and has to be if the language has an cle and a definite interpretation is required. and it is reasonable to suppose that they occur in Det position. standard position for such determiners. The forms are emphatic when used with the definite article. .12 This still takes DG possessives to be [+ Def ].

3. The discussion of inalienable possession above. mon germà ‘my brother’. showed a str tendency for the definite article not to occur with this. but with the possessive t ing a dicerent form: el meu amic ‘my friend’. but meu irmão ‘my brother’. Italian la tua casa ‘your house’. despite the fact that the sense is probably mostly definite. in the AG cases. The same applies to Catalan. to give distinctions which make possible reference to the various participants and non-participants in the discourse situation. they also freque resemble both definite articles and demonstratives phonologically. but figlia ‘your daughter’. even in strongly AG l guages. Person combines invariably with number. second per 134 . and for the moment it is enough to note that the facts relating to possessi (and demonstratives) support the idea that there is a special relationship betw definiteness and some “Det position”. aaxal possessives is rea accounted for in the same way as for lexical possessives. Thus first person singular denotes speaker. Typical AG langua frequently also have DG forms. Given this. and so times of demonstratives. an want to adopt the position that possessives are never lexically specified as [+ D The definiteness of the DG structure is the consequence of the possessive be in Det position rather than the other way round. The AG–DG distinction for some. Examples are Portuguese and Ital in which possessa denoting family relationships (when in the singular and w out modification) usually take a possessive with no article: Portuguese o meu am ‘my friend’. for languages which distingu it from possession more generally.anyway by the many AG languages). the empty or overt possessor triggering agr ment is in Det position. In other wo inalienable possession constructions tend to be of DG type.4. This point will be argued in Cha 8.1 Semantic and grammatical content The basic category encoded by personal pronouns is grammat person – whence their name. on the assumption the aaxes are markers of agreement with a lexical possessive. In the DG cases. if not all. first person plural a plurality which includes the speaker. But it is not a necessary assumption. like Lakhota and Albanian. like definite artic they are often descended historically from demonstratives. the possessor phrase is in some ot position within the matrix. it is not surprising that.4 Personal pronouns The discussion here will focus on the claim developed in Chapte that personal pronouns are the pronominal counterpart of definite articles. and these are often used with head nouns wh fall into the inalienable possession category. 3. which may be pho logically empty.

since it does not refer to speaker or hearer (though its referent m be human). Take for example Warrgamay (Di 1980). and it is fairly common for singular.It is fairly standard to distinguish three persons. In addition. Dicerent non-participants single sentence or limited discourse unit must be dicerently marked as reg person. This category seems to be in a sense an e third person. dual plural to be distinguished in each person. a rather common phenomeno the distinction between inclusive and exclusive in non-singular first person. and sometimes not all sons distinguish the same numbers. a “fourth person” has been identified for some languages. shown here in intransitive subject forms. as in most European languages. and any non-participant. Examples are Maori and Nam Maori first singular dual plural exclusive inclusive au maaua maatou taaua taatou second thi koe koorua koutou ia raa raa . and Sanskrit (Coulson 1976 Warrgamay singular dual plural first second th iayba iali iana iinba nyubula nyurra ny bu ja first second th aham 1v1m vayam tvam yuv1m y5yam sa ta te Sanskrit singular dual plural Some languages distinguish more than three numbers. in the latter not. and the general pattern is to use obviative for the entity less central to current focus of concern. its use is determined by complex grammatical and pragmatic fact A similar distinction is that between “proximate” and “obviative” third-person nouns in Algonquian languages (Hockett 1966). Of course number systems vary. based on the two participa speaker and hearer. Navajo for exam (Akmajian and Anderson 1970). In former the hearer is included.

is expressed by the corresponding plural form. In Ewondo this is done by simply adding a second third-person pronoun (singular or plural) to the first-person plural pronoun: bí (we you+SG) ‘you and I’. Notice also that whereas Carib distinguishes exclusive and inclus also in the first person plural. ’Ekiy and ahna. bí m#n! (we you+PL) ‘you (plural) and I’.1. these combinations take a verb with the nor first-person plural marking. in other words exclusive first p son dual (‘he and I’ etc. The pattern shown by these p adigms and the morphology of plural formation certainly suggests strongly these languages do not in fact have a dual number.first singular dual plural exclusive inclusive tií-ta sií-kxo sií-ke saá-kxo saá-ke second third saá-ts saá-kxò saá-ko //’Ci//’Ci//’Ci- An interesting variant of this found in a number of languages. In some languages the inclusive–exclusive distinction m be expressed optionally. ‘you and I’: Lakhota first exclusive singular dual plural second th niye ’iy niyepi ’iy inclusive miye ’ekiye ’ekiyepi Carib first exclusive singular dual plural au afna second third amo:ro moxk amïiyaro moxk inclusive kïxko kïxka:ro A set of two including speaker but not hearer. is that dual number appears to exist only for inclusive first person denote just the speaker and hearer. including Lakh and Carib. See the analysis of the Carib pronouns s gested above in 3. and that the “inclusive du pronouns are grammatically singular. bí bU they) ‘they and I’. Used as subjects. Lakhota does not. 136 .).1.

singular and plural. manifested in a number of dicerent ways.2 that possessives der from personal pronouns show the same strong–weak distinction. Most languages dicerentiate al one or other of these axes.2 Strong. Reflexive pronouns are commonly distinguished from non-reflexive perso pronouns. Distinctions are widely made. otherwise the ordinary personal pronouns are u reflexively. -sja. or between overt and null forms. between free-standing forms and tics or aaxes.4. in third person only (French). 3. German has the third-per plural Sie for singular and plural address. It may be a distinction betw stressed and unstressed occurrence. in all persons. . therefore. And Spanish has special forms for po address.(English. Classical Greek is like English in having distinct forms for all pers number combinations: emauton ‘myself’. heauton ‘himse etc. and Yoruba In Nama the second-person plural masculine form saáko may be used by a m to a single male equal to indicate closeness. and it will be recalled from 3. to express respect versus familiarity or otherwise to encode the r tive social status of referents. especially but not exclusively in second-per pronouns. except first (Standard Arab Gender marking is very rare for first person singular.3. seauton ‘yourself’. between augmented and simple forms. for all persons. which are grammatically third person. sebja. Russian has a single reflexive pronoun. among other reasons. in third person and plural g erally (Spanish). Cooke (1968) for a detailed study of the extremely elaborate systems opera in Thai. Yoruba too uses a plural form (nw*n) for po third-person singular reference. when emphatic. Russian (vy). Italian uses the third-person feminine Lei for the sin lar and the general third-person Loro for the plural. while the common gender plural p noun saátú is used as a highly respectful form of address to persons of either Carib uses the third-person plural pronoun with an extra suax. Many languages use third-person forms for po second-person reference. as Swahili (-ji-). weak and null pronouns There is a general tendency for pronominal reference. German). German sich). Burmese and Vietnamese. between phonologically full and redu forms. plural ustedes. to exp emphasis or contrast. But these forms can be strengthened in various ways. to take forms which are phonologically or morphologically weak – the definite article. It is possible to say. The second-person plural form is used for po ness to address a single individual in French (vous). In French and German a distinct form occurs in the third per only (French se. nearly all languages exhibit a strong–weak contrast in their personal pronoun tems. moxka:ro-koY refer respectfully to a third person. singular usted.

but it is not. since human di objects require a preposition. which means they normally occur after verbs only with human reference. ellos. Object position is normally post-verbal. consider Spanish object forms (direct object forms only for the clitics): clitic first second third singular plural singular me te lo. sí sinn sibh siad mise tusa seisean. and they have c responding augmented forms used for emphasis. the subject of a very extensive literature ( Lyons (1990) and references there). In fact a strong pronoun here would be acco panied by the clitic rather than replacing it: La veo a ella ‘I see her’. las mí ti él. ‘they’ are frequently reduced in speech when unstressed to [ and [za]. ‘I see it’ (referring to a table) in Span is La veo (CL see+1SG) – since mesa ‘table’ is feminine – with the clitic attac pre-verbally. la nos os los. Whether the clitic is generated in clitic posi 138 . The clitic serves to “identify” (indicate the content this phonologically empty pronoun. vosotros. a widely held assumption is that the object posit (where a full noun phrase or strong pronoun object would occur) is occupied pro. as follows (subject forms): weak first second third strong singular plural singular plur mé tú sé. T strong object pronouns are purely prepositional object forms. The unstressed forms are (like those English) subject to reductions not indicated orthographically.tory and may be stressed or unstressed. ella It is not always the case that the strong forms simply replace the weak or cl forms in emphatic or contrastive contexts. ella full pronoun plural nosotros. the so-called “personal a”: Veo a la mujer (see+1 to the woman) ‘I see the woman’. one might suppose that the contrastive ‘I see would be *Veo ella (parallel to Veo la mesa ‘I see the table’). Where a clitic pronoun is not reinforced by a strong form but is the only exp sion of the object argument. sise sinn sibh siad As examples of clitic pronouns versus full. since strong prono are free-standing noun phrases. Thus du [duT] ‘y and sie [ziT] ‘she’. so. This is “clitic doubling” construction. The unmarked pronouns of Irish are also free forms. free-form pronouns. the null pronominal.

in each t paradigm. like t corresponding to overt weak pronouns. in part. where the ending -mos. in most cases. and in ecect renders an overt subject redundant. the agreement morphology identifies pro by encoding its per and number. a null-subject language. Langua permitting phonologically empty pronouns in subject position are actually substantial majority. Otherwise “anal (non-agreeing) forms occur. Swahili also has a series of stro free-standing pronouns. The Irish verb is defective. But there are strong similari between object clitic structures and the null subject phenomenon. tugann siad (give+PRES they) ‘they give’. and indeed. the third-person prefi take the form of the appropriate class marker. but tu tú (give+PRES you+SG) ‘you give’. that is. while actually being unstressed. A typical exampl Spanish comemos ‘we eat’. encoding first person plu makes it possible to omit the overt pronoun nosotros ‘we’. Lyons 19 treat object clitics as representing precisely the same phenomenon syntacticall subject–verb agreement. and a pronominal subject must be overt. and only with these is pro possible. which is licensed agreement. Now in t parts of a tense paradigm where the availability of a synthetic verb form permits a pronom subject to be null. tugaimid (give+PRES+1PL) ‘we give’.sis may be appropriate for dicerent languages). in which the two seem v similar. phonologically strong in the sense of being fully stre and perhaps polysyllabic. emphasis on this subject is expressed by suaxing the same reinfor . The “null subject parameter” is usually linked to agreem morphology on the verb (or elsewhere: many languages show an “auxiliary” c stituent which hosts agreement) in the form of aaxal or clitic material. I noted above that Irish free-standing weak pronouns. there is cross-linguistically no consis morphological distinction between the agreement material identifying object opposed to subjects. “synthetic” verb forms (displaying agreement morphology) are available for some person–number combinations. The subject is taken to be. Thus. tugaim (give+PRES+1SG) ‘I give’. But it is possible for the “strong” form to be stronger than pro in having some overt material. Let us look briefly at Swahili. A number of studies (for example Suñer 1988. and strong pronouns consisting of augmented forms of th But Irish is also. the null pronominal pro. which would only included for emphasis. for others. used only for emphasis:13 13 In the examples so far considered. Swahili represents pronominal subjects and objects as prefixes on ve the subject forms being more peripheral: subject first second third object singular plural singular pl niua- tumwa- -ni-ku-m- -t -w -w These are the forms used for human referents. to take the sent tense. the strong pronouns corresponding to pro are.

but only w pro objects. respectively. for example. The clitic may well represent agreement with a object. Where such “doubling” occurs. perhaps some other factors. but *Mi vede me (1SG sees me) ‘He/She sees m Cross-linguistically. First. In m Romance languages. emphatic forms of tugaim ‘I give’ and tugaimid ‘we give’ are. the pronom clitic or inflection on the verb can be treated as encoding agreement with the fr standing pronoun. where there is an overt object no clitic appears (Italian Io lav (I work+1SG) ‘I am working’. th are varying constraints on the noun phrase types that trigger agreement. because it contains pro. number. whether or not there is agreement may depend on whet the subject or object is overt or null. definite or indefinite. For example. pronominal or full. where the overt object has to be associated w pre-position (La veo a ella (3SG+F see+1SG to her) ‘I see her’). 140 . the so-called “φ-features”. the correct analysis of this form is alm certainly that the augment is attached to pro just as it attaches to overt pronouns: tugaim se. though wit overt suax. and case. the strong forms reinforce rather than replace both subject object prefixes: Mimi nimekwisha ‘I have finished’. the verb agrees with any subject. so the whole pronominal form. Second. Language which prepositions show agreement morphology and may take a null pronom complement include Welsh (amdano (about+3SG+M) or amdano ef (about+3SG him) ‘about him’) (Sadler 1988) and Berber (zg-s (from-3SG) ‘from him’) (Ren 1932). Nilimwona yeye na Ham pia ‘I saw both him and Hamisi’. it is far from certain tha verb is agreeing with the strong pronoun. though “stro in function. For m detailed discussion of these points see Lyons (1990). But this suax is itself unstressed. with some phonologically conditioned allomorphy) to the verb.14 All the points made here about pro in subject or ob position can apply also to the position of object of a preposition. gender. tugaimid pro-ne (see McCloskey and Hale 1984). clitic doubling occurs in Macedon but only when the object is definite: Ja vidov maRka-ta (3SG+F saw+1 cat-DEF) ‘I saw the cat’. is unstressed. 14 particle which is used to augment the free-form pronouns (in a form determined by pe and number. va they are typically some or all of person. This means that the strong pronoun oc ring with a synthetic verb form is still partly null. Thus.first second third singular plural mimi wewe yeye sisi ninyi wao In emphatic use. tugaim-se and tugai ne. ella being in a prepositional phrase in some position other than that of object. Given that the subject is post-verbal in Irish. In the kind of doubling noted in Spanish. The process of agreement which licenses null arguments may be constrai in a number of ways. the features involved.

4. Keki-se sensayng-nim-ul Minca TOPhiking go PAST DECL there teacher HON ACC manna-ss-ta. I will take this point up again in Chapters 8 and 9. Since personal pronouns are assumed to be the pronominal counterpart of definite article. There she met her teacher. 3. and since both these entities can take the form of lexical item aaxes. show no s correlation. German).3 Personal determiners By “personal determiners” I mean forms related to personal prono occurring within full noun phrases. and that therefore the null arguments in sentences like (10) are not a ally pronominal. Swedish and Lakhota (if the article in this language is a free form). may represent a radically dicerent phenomenon f that of null arguments identified by agreement. meet PAST DECL ‘Minca went hiking. Certainly many languages with a f form definite article also have free-form pronouns (English. where an argument of a sentence which would naturally be dered pronominally in other languages is not overtly expressed by either a pron or by any agreement material. We h seen that these determiners are subject in English to two kinds of constraint their non-exclamatory use they are restricted (except in the third person) to stro unreduced form (which I have claimed represents a demonstrative). Huang (1994).understood than the discussion so far implies.’ Cases like this. The second constraint at least is langua specific. as can be seen from comparison of the following paradigms in Eng and German: . you students. But many other languages. It is not necessarily dependen identification by rich agreement morphology. an interesting question is whether there is a correlation within langua between the forms of pronouns and articles. Arabic). and m languages with an aaxal or clitic article also have null subjects and/or obj identified by inflection or clitics (Albanian. It is likely that pro is not involved in th languages. A number of languages permit pronominal subjects or objects in the total absence of any agreement morpholo an example is Korean (Sohn 1994): (10) Minca-nun haikhing ka-ss-ta. For discussion see Huang (19 Jaeggli and Safir (1989). And t are not available for all person–number values (being completely excluded f the first person singular: *I student). of the type we teachers.

’ On the other hand. person is expressed inflectionally in the noun phrase. The/That boy disgraced himself. man we AUX 1PL shout ‘We men are shouting. though these can h second-person reference: Le salaud! ‘The bastard!’. We boys disgraced ourselves.’ Ngarka nganimpa ka-rnalu purlami. And the ar ment use. *vosotros estudiantes (you students) are all impossible. Another language completely unconstrained in this respect is Warlpiri (H 1973). In Nama. I lazybones must myself hurry ‘*I lazybones must hurry’. man you-SG AUX 2SG shout ‘*You man are shouting. *nosotros e diantes (we students). Die Faulpelze müssen sich beeilen. Wir Faulpelze müssen uns beeilen. The/Those boys disgraced themselves. Fre permits only third-person forms in the exclamatory use. vous autres hommes ‘you men’.’ Du Faulpelz mußt dich beeilen. there are languages much more constrained than English Spanish *yo estudiante (I student).exclamatory use argument use *I fool! You fool! The fool! *We fools! You fools! The fools! *I boy disgraced myself. c. Der Faulpelz muß sich beeilen. lazybones I am. man I AUX 1SG shout ‘*I man am shouting. is only poss (except in the third person) with the addition of autre ‘other’: nous autres hom ‘we men’. You boys disgraced yourselves. limited to the same person–number values as English. ‘You bastard!’. b. ‘I must hurry. *tú estudiante (you student). Ich Esel! I donkey ‘Silly me!’ Du Esel! Der Esel! Wir Esel! Ihr Esel! Die Esel! Ich Faulpelz muß mich beeilen. as a suax the noun. Ngarka ngatju ka-rna purlami. Nama is as unrestricted as German and Warlpiri as regards pers number: 142 .’ Ngarka njuntu ka-npa purlami. a language with no definite article and apparently only strong personal de miners (identical to personal pronouns): (11) a. Ihr Faulpelze müßt euch beeilen. *You boy disgraced yourself.

which means the suax agrees with the determ in person but not in number (perhaps because the noun’s own plural morpheme makes agreement redundant): \s ousoutsich-s (I teacher-1) ‘*I teacher’. Assuming that the doubling in the Nama and second persons represents morphophonological strength. But this inflecti marking of person in the noun phrase is extremely rare.2 that the Armenian possessive s does not agree with a plural free-form possessive even in person. . This constraint may. so this suax does see be distinct. though a str form (this or that) is also possible. third person -n/-[) the same forms are used for the plural. second person -t. That these constructions are genuinely possessive. Similarly. as in Danish Din idiot! (your idiot) idiot!’ (Haugen 1976).16 15 16 Incidentally.saá kxòe-ts kxòe-p sií kxòe-ke saá kxòe-kò kxòe-ku (you person-2SG+M) (person-3SG+M) (we person-1PL+M) (you person-2PL+M) (person-3PL+M) ‘*you man’ ‘the man’ ‘we men’ ‘you men’ ‘the men’ Notice that in first and second persons a free-standing personal determiner (id tical to the corresponding pronoun) obligatorily co-occurs with the pers number–gender aax. Despite the dicerence in st ture. This personal suax occurs only to double a free-standing personal determiner (i tical to the personal pronoun) in first and second person – exactly as in Nama. And in some languages t possessive aaxes can be used as if they were aaxal “personal articles”. free-standing demonstrative determiner could occur.15 but not in the third person. as doubling norm does. where it was seen that personal determiner can be weak only in the third person (the). the Scandinavian languages exemplify this. as we would expect given the constraint just discussed. person–num aaxes cross-referencing or identifying a possessor are common. this is remarkably similar to the English pattern. On the other hand. In the sing it is identical to the possessive suax (first person -s. m\nk’ ousoutsich-n (we teacher-PL-1) ‘we teachers’. Instead. This recalls the strategy pointed out for Ewondo in the cla Another language displaying aaxal personal determiners like Nama is Classical Nah n-oquich-tli (1SG-man-SG) ‘*I man’. apparently on exclamatory use. therefore a very general one. both languages are constrained to have strong determiners only (claime be demonstrative in English) in these persons. is suggested by fact that Turkish has what looks like the same construction with an overt full noun ph “possessor” and third-person possessive aax: Bekir hırsız-ı (Bekir thief-3SG) ‘that Bekir’. for discussion see Andrews (1975). Nama can express inclusive first person in this construction by having a sec person determiner doubled by a first-person dual or plural suax: saá tará-se (you wom 1PL+F) ‘we (inclusive) women’. The third-person pronoun m not occur in this construction. But one language in which there is a case for claiming that there is an aaxal pe marker homophonous with the possessive marker (and not limited to the exclamatory us Armenian. But recall from 3. rather than sh ing a person morpheme homophonous with the possessive morpheme.3. the noun can occur with aax only. some guages permit free-form possessives to have the sense of personal determiners in the exc atory use.

some students work-1PL much Tres estudiantes trabajáis mucho. the students work-2PL much ‘You students work hard. man AUX 1SG shout ‘*I man am shouting. Algunos estudiantes trabajamos mucho. like personal pronouns. it is clear that the verb’s inflection need not agree in person with the s ject in Spanish. Spanish compensates by perm ting noun phrases third person in form to have first.’ . as is well known.or second-person referen this being shown by the appropriate agreement morpheme on the verb (so the p nomenon is limited to subject position): (12) a. which has a full paradigm of personal determiners. c. it is presumably constrained merely not to disagree.’ Los estudiantes trabajamos mucho. three students work-2PL much The sense here is a number of students including the speaker. But. Interestin this possibility is only available in the plural. A further important point is t whereas personal determiners are. 144 Ngarka ka-rna purlami. structure can be indefinite: (13) a. b. man AUX 2SG shout ‘*You man are shouting. the students work-1PL much ‘We students work hard. el estudiante (the stude cannot express ‘*I student’ or ‘*you student’.’ Los estudiantes trabajáis mucho. invariably definite. I return to this in Chapte Warlpiri.nominal determiners at all. b. and without the number constraint of Span (14) a. also has the Span construction (though without articles). b.’ Ngarka ka-rnalu purlami. Los estudiantes trabajan mucho. the students work-3PL much ‘The students work hard.’ Ngarka ka-npa purlami. and a group of th students excluding the speaker but including the hearer. c. or even if it is a default form unmarked person. man AUX 1PL shout ‘We men are shouting. which means that the range references possible in terms of person–number values with full noun phrases the argument use) is exactly the same as in English.’ If the definite article is third person.

4.’ Ngarka njampu ka-rna purlami. ultra-distal yinja. that supplies the personal pronoun. (16) a. for the non-human pronoun. In Warlpiri these exh four degrees of deixis. Now the first two of these. On this view these sentences show a variant of the construction ( I think it is more likely that what we see in (15b) and (16b) is a variant of the “n agreement” structure (14). Similarly in Finnish. on a mixed distance-person basis: proximal or associa with first person njampu. In Turkish too it is the third (distal) term of the demonstrative syst o. Often it is a demonstrative unmarked for de that fills this function. the general demonstrative is is u especially in non-nominative cases where there is more need for an overt f (Latin being null subject).4 Definite and demonstrative personal forms It is very common for demonstratives to serve as third-person sonal pronouns. man this AUX shout ‘This man (near me) is shouting. frequently oc rather than is. ille. ‘s Sometimes one of the deictically marked demonstratives supplies the personal p noun.’ Ngarka yalumpu ka purlami.and second-person determiners. as well as picking out so thing near speaker or hearer.’ It is arguable that the demonstratives in the (b) sentences here are functionin first. b. An interesting variant involves demonstratives. man that AUX shout ‘That man (near you) is shouting.’ Ngarka yalumpu ka-npa purlami. 3. and for languages to have no personal pronoun in the third son distinct from a demonstrative. can be used with reference to speaker or hearer. !n. and so trigger the appropriate agreemen the auxiliary. the demonstrative related to third person. man that AUX 2SG shout ‘*You man are shouting. b. distal or third son yali. man this AUX 1SG shout ‘*I man am shouting. with the non-person-marked subject noun phrase tak a demonstrative appropriate to its reference. Persian uses the distal form of a two-t system. for example. In Latin. medial or second person yalumpu. the general demonstrative se serves as inanimate third-person pronoun. Again in Latin. Ngarka njampu ka purlami. In Lezgian the medial demonstrative . and sometimes replaces hän ‘he’.in (11). agreement morphology on the auxiliary makes the sense clear: (15) a.

the two-term demonstrative series supplies the third-per pronoun: en[ ‘he’. Japanese uses a full noun phrase for the third-per pronoun: 1SG 2SG 3SG 1PL 2PL 3PL watasi anata kono/sono/ano hito watasitati anatatati kono/sono/ano hitotati ‘I’ ‘you’ ‘he’. emphatic use. ‘she’ ‘we’ ‘you’ ‘they’ The third-person forms consist of the noun hito ‘person’ preceded by an app priate choice from the three-term person-based demonstrative system. ano. ed[ ‘th ‘these’. which is non-demonstrative. and whatever deictic choices apply to the demons tives apply equally to the personal pronoun. ted[ ‘they’. The corresponding weak pron has the form of pro or an aax or clitic. ‘those’. in strong. ‘those’ Equally in Khalkha. ‘that one’ ‘we’ ‘you’ ‘they’. ‘these’ ‘they’. ‘that one’. it follows the pronominal pattern of t ing the number marker -tati in the plural. The fo related to third person. But this p tern is not limited to null-subject languages. ‘she’. ‘these’ ‘they’. ‘those’ ‘they’.the demonstrative system that has this function (presumably becoming unmar for deixis in the process). ‘it’. In languages in which personal p nouns are normally overt and free-standing. ‘it’. while nouns are otherwise invaria for number. ‘she’. Though ano hito ‘that pers looks like a non-pronominal noun phrase. which means that the demonstratives used for third person are functioning as strong pronouns. ‘it’. ‘it’. ‘it’. Thus Albanian has the following p sonal pronoun system: 1SG 2SG 3SG-M-PROX 3SG-F-PROX 3SG-M-DIST 3SG-F-DIST 1PL 2PL 3PL-M-PROX 3PL-F-PROX 3PL-M-DIST 3PL-F-DIST unë ti ky kyo ai ajo ne ju këta këto ata ato ‘I’ ‘you’ ‘he’. This use of demonstratives for the third-person pronoun is particularly evid in null-subject languages. ‘this one’ ‘she’. is the most common. In these languages personal pronouns are only overt course. ter[ ‘he’. ‘this one’. ‘it’. strong use may simply involve add 146 . ‘this one’ ‘he’. ‘that one’ ‘she’.

This treatment is supported by the evidence presented Chapter 1 that strong occurrences of personal pronouns and determiners English are [+ Dem]. aq already discussed. to be repla by the demonstrative forms der/die/das/die. .and second-person forms too. for example). when emphatic. What this suggests is that. where the perso determiner structure can be weak in the third person). In French. ‘it’ ‘it’ 1PL 2PL wir ‘we’ ihr ‘you’ 3PL sie ‘they’ These can all be stressed for emphatic or contrastive use. discussed ab Recall that. Consistency s gests that first.often a demonstrative. deictic unmarked demonstrative. in this construction. ese. But. and no general demonstrative so far identified. so that personal pronouns in subject position are ge ally emphatic (except that they may also appear non-emphatically to avoid am guity over the gender of a referent. ella ‘she’ etc. A somewhat similar picture is s gested by the facts of the personal determiner system of Nama. as in English. él ‘he’. Consider the German system:17 1SG 2SG 3SG-M 3SG-F 3SG-N ich du er sie es ‘I’ ‘you’ ‘he’. for insta has the three-term deictically marked demonstrative series este. ‘it’ ‘she’. h ever. the person marker on the noun mus reinforced by a free-form personal determiner. in first and second persons. are demonstrat though the deictic distinctions familiar in demonstratives are generally lim to third-person ones. the third-person strong f is expressed by a demonstrative. The generalization that these observations point to is that strong occurren of personal pronouns and personal determiners are demonstrative. Spanish. I took this to suggest that the c struction is strong. A demonstrative can. and in the same way es occasionally be ‘she’. But Spanis a null-subject language. but there is a tende in colloquial German for the third-person forms. But in the third person the aax does not req this reinforcement (again recalling the English situation. and indeed a determine the form of the third-person pronoun stem cannot occur. An interesting consequence of this analysis is that deictically unmar demonstratives are more common than otherwise appears. when strong. er and sie correspond to English it when a non-human refere denoted by a noun which is grammatically masculine or feminine. Whether t must express deixis too varies from one language to another. on the analysis propo emphatic yo ‘I’. are forms of a general. the emphatic third-person pronouns 17 The glosses are simplified. tú ‘you’.

though it does not follow that the com nation is indefinite. In Germ der eine Mann (the one man) would normally mean ‘one man’ (out of a gro 148 . English most of a town). Hungarian a t^l-ed kapott valamennyi levél (the from-2 received each letter) ‘each letter received from you’ (Szabolcsi 1994). but then with the defi article present (in languages that have one) to ensure the definiteness with wh the quantifier must be associated: German der ganze Wein in Kalifornien (the wine in California) ‘all (the) wine in California’. cette . The French equivalent of most.5 Universals and other quantifiers It was noted in Chapter 1 that quantificational expressions like and most share some of the behaviour and distribution of definites. many of th seem to be to do with degrees of specificity. celle-là etc. is definite in form: la plupart des clients (the most+part of+the c tomers) ‘most customers’. die meisten Menschen (the m people) ‘most people’. Within the noun phrase. despite being definite in form. The num one. are not central to this stu and will not be investigated in detail. Note that in the first pattern the n phrase quantified over can sometimes be indefinite (French toute une semain whole week’. ‘most of the customers’.unmarked demonstrative determiner ce. m of these cars. . not necessarily themselves [− Def ]. . 3. In English th determiners can stand outside a complete noun phrase. and that cardinality does not entail indefiniteness. as is their occurrence in what looks like Det position inside noun phrase (German alle Autos ‘all cars’). in acc dance with their apparent definite determiner status. the most interesting cross-linguistic observation is quantifiers corresponding to all and most show the same pattern of distribut seen with demonstratives. They may occupy Det position as in English. -là etc. for example. appears in definites and indefinites. while the deictically mar pronouns celui-ci. We have seen that many so-called “indefinite” determiners are in fact ca nality expressions. I am taking the determiners involved to incompatible with definiteness. This is a common pattern cross-linguistically (French tous les arb ‘all the trees’). though linked to a noun phr partitively.6 Indefinites Complex indefinites. noun phrases whose indefiniteness is due to so expression which is not an article. . ces. cette. 3. . with something like adjectival status. correspond to ce . possibly (necessarily the case of most) linked to it by a partitive construction: all (of) the trees. But it is appropriate to m tion here one type of noun phrase containing the singular numeral which is numerous languages indefinite in sense. -ci. not typically weak. or they may be in some m internal position.

Latin ullus and I aon for example. It was also seen in Chapter 1 that some is in c plementary distribution with sm. The same phenomenon is p ably at the root of the Albanian structure in which një ‘a’. This is refle in considerable variety in structure and source. Some forms are clearly determin but others (like Irish ar bith ‘any’. Thus Spanish cierto. T situation is partly mirrored in Spanish. the sense being ag indefinite: një motr-a ime (a/one sister-DEF my) ‘a sister of mine’. correspond approximately to strong some. Latin quidam. French aucun. ‘one of sisters’. tho occurring in strong and weak varieties. and similar conte Other languages distinguish a special form for this function. interrogative. where algún/algunos. 3. more literally ‘on earth’. the Irish negative polarity aon ‘any’ for example. The strong form. éigin. has this function: sans faire aucun b ‘without making any noise’.implied with some other man or men. it c not in the singular. Recall that English random any shows some grounds for being treated as defin and it is common for a determiner with this random sense to function also the equivalent of every. quivis other languages the cline makes fewer distinctions. any can be reproduced for many languages. ar bith (the a prepositional phrase rather than a determiner). l’autre est resté ‘One left. French tout.prefi . so they could be the same determiner. this is true of Russian vsjakij. can occur in singular and plural (though not m contexts. English any German einig ‘some’ are derived from ‘one’ with an adjectival suax added. Spa algún consists of the singular numeral with an indefinite morpheme alg. These indefinites are not only semantically extremely complicated and po understood. Russian has specific (‘certa and general (‘some’) njekij/njekotoryj and random (‘any’) vsjakij/ljuboj. Irish áirithe. German je Hausa kowane. but also in all likelihood semantically very heterogeneous. is the only one possible in singular count use. but whereas in the plural it tends to express a vague cardinality. aliqui. The same applies to the French L’un parti. ‘one’ is used wi noun in definite form accompanied by a possessive. as well as occurring with purely negative se as ‘no’ (en aucune façon ‘in no way’). and others more complex (French n’importe quel ‘no matter whi ‘any’). cualquier. algún. In English the random determiner serves also as a non-asser or “negative polarity” form used in negative. ‘in the world’) are positional. which may have so additional semantic content.6. if the correspondences are app imate. Compare German mein einer Sohn (my one son) ‘one of my sons’.1 Full noun phrases The cline of specificity illustrated by the English series (a) cert some. Many indefinite determiners are (like most cardinal articles) of sing numeral origin. the other stayed’.

und-a Najit’ (which time-INESS any) ‘at any time’. the original sense thus being som thing along the lines of ‘which . Span cualquier. and the la is illustrated by Maltese xi ‘what’. indefinite pronouns merit separate treatment because of special formations wh are widely found. or inc porated with a determiner. me ‘what’. but the determiner can be drop 150 . Rto ‘what’. and in Sarcee !dácá is ‘who’ and ‘someo dìt’á is ‘what’ and ‘something’. t may be simply identical to interrogatives. and. Albanian dikush ‘someone’. respectiv duine éigin (person some). o:tï is ‘what’ and ‘something’. in Carib no:kï is ‘who’ and ‘som one’. Hausa kowane ‘a (from wane ‘which’). Dixon (1980) observes that m Australian languages have a set of forms that have indefinite or interrogative sen or both together. with sa w jat’ani (one who some) ‘someone’. you wish’. to also capable of pronominal occurrence. like other determiners. an interrogative modi and an indefinite particle Najit’ani ‘any’ or jat’ani ‘some’: hi waxt. and the Albanian pronominal kush Indefinite determiners containing an interrogative element.6. ‘thing’). . Rto-to and njeRto ‘som thing’. A common pattern in indefinite pronouns is the use of a noun of somewhat g eral sense (like ‘person’. in wh up to three separate elements co-occur. from k ‘who’.in combination with an interrogative element. or simply identica an interrogative. so a sentence containing a form of the root Xaan. wuR Najit’ani (what any) ‘anything’. sa wuR jat’ani (one what some) ‘somethin and wuY Najit’ani (who any) ‘anyone’. are very common. fr wa ‘who’. more commonly than the determiners. kome ‘anything’. sometimes with slight variation in fo This is essentially the case with the complex indefinites of Lezgian. Like the determiners they are frequently based on interrogativ often with an additional marker. modified by a determiner. kushdo ‘anyone’. ‘some’ (Aquilina 1965) and Korean enu ‘whi ‘some’. Khalkha alibaa ‘any’ (from ali ‘which’). the numeral sa ‘one’. was is ‘what’ and ‘something’. . A particularly interesting case is Lezgian (Haspelmath 1993). For ‘someone’ and ‘something’ Irish has. Hausa kowa ‘anyone’. Latin quivis and quilibet. either alone. See also Ultan (1978). rud éigin (thing some). Examples are Russian lyuboj. 3. Examples of augmen interrogatives are: Russian kto-to and nikto ‘someone’. in German wer is ‘w and ‘someone’. Similarly. and all the L forms mentioned containing the element qui exemplify the former. sa hiNtin jat’ani Xalabulux (one which so confusion) ‘some confusion’. The Spanish cualquier.2 Pronouns It is common for indefinite determiners. from kto ‘who’.‘someo ‘who’ may be understood as simultaneously predicating something of some in vidual and asking for this individual’s identity.

which is something like ‘of this/that k or ‘like this/that’. somebody big. some unexpected thing. which. French quelqu ‘someone’. Irish again uses the noun duine ‘person’. Of course it is debat whether these are pronouns or simply indefinite noun phrases. Spanish. ‘one’ often occurs in place of a noun in human forms. German and Old English have man. The first is it is an indefinite demonstrative. is grammatically masculine (that is. that is. The a native analysis is that such is not a demonstrative.has kxòe-’ì (person-INDEF). Khalkha has a two-way demonstrative contr en[ ‘this’. one such language. something unexpected are not synonymous with s nice one. English someone. Languages with three-term demonstra 18 That the noun is incorporated into the determiner to form a complex pronoun is clear for lish and French. which is duplicated in iim[ ‘such as this’. if uncertain for other languages. ‘she’. someone nice. ‘of that kind’. An English. though b on a feminine noun. For an excellent and much more detailed cross-linguistic survey of indefi pronouns see Haspelmath (1997). beh like a pronoun in taking number marking. as in the case of French on. For comparison. though it contains a dem strative element as part of its meaning. permitting indefinite as as arbitrary reference (On frappe à la porte ‘Someone is knocking at the doo Many languages prefer an impersonal verb construction to express this idea.7 Indefinite demonstratives In Chapter 1 we entertained two analyses of such. being therefore of the “general demonstrative” ty dicering only in that it is constrained to occur in indefinite noun phrases. though sometimes. as show past-participle agreement (quelque chose que j’ai fait/*faite ‘something I’ve done’). . of unmarked gender). and French on also has its or in the word for ‘man’. and this also consists of the singular nume Other languages use a pronoun derived from (or identical to) a noun such as ‘ son’ or ‘man’. and Turkish uses in ‘person’. ter[ ‘that’. it is synonymous with this/that (tho lacking the deictic contrast. English someb and something and French quelque chose ‘something’ exemplify the incorpor noun type. Persian yek-# (one-a) ‘someone’. and Navaho $a’ (one) ‘someone’ illustrate the fact that. The French quelque chose.18 The English “arbitrary” human pronoun one has its equivalents in many guages. ‘of this ki tiim[ ‘such as that’. some big body. The close relationship of such to the demonstratives is evident in the fact the corresponding term in other languages often does show the same deictic tinctions as the demonstratives do. while looking like a full noun phrase. recall the Japanese defi pronoun ano hito (that person) ‘he’. 3. xuu-’ì (thing-INDEF). also has a pronoun used mainly in circumstances wh the impersonal construction is excluded. in all th types.

anna). were correct. and then they are typically bare in English (as in Hello. ai ‘that’) and makes a corresponding contrast between i kët ‘such as this’ and i atillë ‘such as that’. for both English and Fren may be that the noun phrases involved are not formally vocative. tàs. can be definite indefinite. follow me). les mecs ‘Hello. Since vocatives are forms of direct address. The answer to these diaculties. th is perhaps no grammatical category of vocative in these languages. Since personal pronouns and personal determiners are inh ently definite. T is. konna. there should be languages with a single word for ‘that’ and ‘such’. that such etc. in vocative function. that is. Wìtoks. while the correspond demonstratives are definite. but has in addition a deictically neu form i tillë ‘such’ not corresponding directly to a specific demonstrative. Albanian has a two-way contrast in dem stratives (ky ‘this’. ano. vocatives involving these must be definite. sonna.o. we might exp them to be second person. general’). chaps’) or a possessive (Bonjour. 3. are simply indefi variants of the demonstratives. The problem in French is that at least the first of th looks third person. Lithuan (Wìs/Wìtas. but since indefiniteness matter of absence of definite marking rather than presence of some other ma ing. definite or cardinal. come here!). Often this takes the form a particle accompanying the noun (obligatorily. but n phrases of dicering structure can be used in address function. öyle). This means that the assumption demonstratives are inherently definite is maintained. In French common noun vocatives are always definite. chaps or Childr what are you doing?). also arguably indefinite if bare. we would expect to find langua in which the complementary distribution were reflected in identity of form. Iöyle. optionally. There appear to be no such languages. anàs. böyle. mon gén ‘Good morning. tóks. Many languages do have special vocative forms. but they are interesting in being of uncer status in relation to definiteness. anóks). But what about the b noun type? It has no determiner. Proper names of persons are undoubtedly the n phrase type most commonly used this way. But if the first analysis. as can full noun phrases with second-person determin (You boys.8 Vocatives We have not so far given any consideration to noun phrases used direct address. taking eithe definite article (Salut. sono. but common nouns too can be vo tive. English vocatives. and indeed second-person pronouns can be used way (You. then. the tra lation being determined by whether the noun phrase in which the word occur marked definite or indefinite. All these forms are limited to indefinite occurrence. these expressions should be indefinite. or only with cer 152 . Japanese (kono. They may be of a similar status to pro nouns. and I t this to point decisively to the second analysis.

lower. Perhaps to be relate this last observation is the fact that in Serbo-Croat an adjective modifying a vo tive takes “definite” form. with a particular structural position. where most vo tive forms which are distinct from the nominative consist of the mere stem (wh may include a theme vowel indicating declensional class). Cantonese A-baak ‘Uncle’ (addressing an older man). in some cases at least. This is so in classical Indo-European languages Greek. Sanskrit and Latin. These ideas will be clarifie Chapter 8. Aaxal definite m ers seem. . but it is no more t a tendency. but some fe nine names do. tha to be subject to the X-bar schema. 3. For a clear discus of the issues see Hale (1983. But there is reason to believe that many guages are “non-configurational”. position. In general. In Albanian. Sometimes the vocative particle is additional to a special vo ¯ andres Ath@naioi ‘Men of Athen tive case form of the noun: Classical Greek O Irish A Sheáin ‘John’. subject to limited variation. to be in a position distinct from that of free-f definite determiners. and the cussion so far has implicitly taken all languages to be “configurational”. vocatives are not consistently definite indefinite. which I have termed “ position”. In this morphological minimality the vocative resembles the perative in verb conjugation.9 Non-configurational languages This survey of the expression of definiteness and indefiniten reveals variety which is compatible with the idea that definiteness is associa perhaps universally. and a great deal of work in recent years been investigating a possible configurationality parameter. The postulation of these generalizations exploits and depends on hypothesis of a universal X-bar schema. There is a strong tendency for vocatives to be bare. which is the s with no case aax. and I suspect that (in terms of the analysis of gr matical categories as functional heads) the vocative is the nominal counterpar the imperative. often no more than the noun stem. where vocatives are identical in form to the nomina but take a particle. but it is arguable that they are in some position which en into an agreement relationship with Det position. they do not normally take the definite aax. appearing also before a modifier (as d the definite article). but where nouns do have a spe form for the vocative (with or without a particle) there is a strong tendency this to be a minimal form. and all do when restrictively modified. The vocative function is traditionally dealt with under rubric of case (though see Blake (1994: 9).‘Boy!’. In Turkish the f used for the vocative is the same as the nominative–absolutive. 1989). and cardinality with a dicerent. In Egyptian Ar the particle is involved in agreement.

The results of investigation al this path are still indecisive. Languages are not necessarily configurational or non-configurational ov all. Mohanan 1983. a subject can act as anteced for a reflexive in object position. but not vice versa (John shot himself. which he terms A and B. and the reader is referred to Marácz and Muys (1989) for some good examples of the debate. and identifies two language types. but *Him shot John). discontinuous expressio extensive use of null arguments. but may vary in dicerent parts of their grammars. and arguably in part made possible. by a rich auxil system identifying the arguments of the sentence. it turns out they are. respectively. Austin and Bresnan 1996). An alternative account of languages appearing to require characterization is that they are configurational in the underlying D-structure subject to a process of “scrambling”. that languages can be non-configu tional. also Baker (1995) for detailed investigation of the polysynthetic subtype of n configurationality. I will assume here. It follows fr the proposed absence of VP in syntactic structure that all arguments are on same level. however. absence of syntactic movement rules and (p haps) of empty categories. For exam Hungarian and Lakhota have been argued to be non-configurational in clause str ture (Kiss 1987. there are numerous dic ences in behaviour between the two. and to Speas (1990: 123–201) a strong argument against the existence of non-configurational languages. In a guage like English where subjects are specifiers and objects are complements that the former c-command the latter. the use of numeral classifiers (type B but not type A). configurational and non-configurational languages. but not vice versa. for example.order. degrees of configurationality. “flat” (as opposed to hierarchical) phrase structure. langua showing some of the properties of each type or showing them to a limited ext Gil (1987) surveys a range of noun phrase properties for which languages m vary. for expository purposes. and 154 . Jelinek 1984. Th properties include the obligatory marking of (in)definiteness (characterizing t A but not type B). In some putatively non-configurational languages th properties are accompanied. Williamson 1984). and a rich case morpholo The phenomenon has been accounted for as consisting of a mismatch betw a fully configurational level of “lexical structure” (expressing the predica argument structure of a clause) and a flat phrase structure with no VP node (H 1983. Such asymmetries (in so far as they are due to the dicerence betw specifier and complement positions) should be absent from languages in wh subjects and objects are structurally equivalent. moreover. There are probably. and this idea has led to an emphasis more rece in the literature on the investigation of “subject–object asymmetries” as ocer more decisive criteria for configurational or non-configurational status. sisters to the verb. but both are highly configurational in noun ph structure.

introducing new referents to discourse: y!t ga 19 The claim that definiteness marking does not occur in languages which are non-configurati in the noun phrase because they lack the appropriate structure makes an interesting predic that is easily tested: that these languages cannot have DG possessives. for exam noun phrases with possessive modifiers are neutral with respect to definiteness: anata no (you POSS book) ‘your book’ or ‘a book of yours’. because “indefinite articles”. but the empirical evidence points to such a reduction. and classifi are necessary with numerals to establish units for enumeration. since they are in ecect w numerals. But then why should a quasi-indefinite article not take the form weak ‘one’ plus a classifier? In fact there is evidence for this possibility. but of the hypothesized definite Det position. that configurational languages have obligatory marking of definiteness. the prediction is borne out. m over. Whatever account is adopted. the ques is important. however. DG possessives are associated with the same structural positions as defi articles. but it does have number marking and does not use classifiers. This is because. which I have claimed are cardinal. and. It seems likely. In Canton classifiers are obligatory with numerals. not of determiners g erally. One po bility is that non-configurational languages have no cardinality position and so can have numerals except with classifiers. Gil’s language sample is limited: he examines in detail only English (as a p digm type A language) and Japanese (illustrating type B). on present account. In type B guages all nouns are mass. that the count–mass parameter is to be subsumed under the configurationa parameter. Within the present investigation. sensei no hon (teacher POSS book) teacher’s book’ or ‘a book of the teacher’s’. But the properties obser for Japanese do seem to typify the most strongly non-configurational langua (such as Warlpiri and Korean). on a limited investigation. Type B languages do not mark (in)defin ness because they lack determiners. In Japanese. for whatever rea cardinal articles should be equally excluded. it seems unlikely. The other properties he explains in terms of the count–mass distinction. Russian ha articles. And. If non-configuratio languages do not have numerals (except with classifiers). . and so se to be basically configurational. the temptation i account for this in terms of absence in these languages. appear to be a tenable g eralization that languages that are non-configurational in the noun phrase alw lack definiteness marking. There does. for example. determiners require hierarchical structure cause they have the function of mapping one bar-level into a higher the configurationality parameter.19 It is not clear how to reduce the count–mass parameter to the configuratio ity parameter. so there can be no marking of plurality. the numeral is used optionally as an article. as already observed. since the positive values of the two appear to coincide. And languages may show a continuum of cha teristics relating to configurationality. absent in non-configurational languages according to Gil.

this is additional evidence for the claim that articles like a really cardinal rather than indefinite. 156 .configurational languages cannot have definiteness marking but can have card articles.

we return to the analysis of proper no which we have seen resemble generics in being overtly definite in form in so languages but not in others. specific and n specific. though shared by a number of languages. Generics are typically definite in form in some langua but not in others. which appea be independent of that between definite and indefinite.1 One definiteness or several? Here we consider the question whether it is merely a language-spec fact. wh can be expected to be separately encoded in some languages? There are semantic distinctions. 4. is the definite article. situational and anaphoric. but which interact w the latter distinction. It will be suggested that proper nouns are a kind generic. ambiguous? Or is it rather simply polysemous or vague in mean . that the range of noun phrase marked as definite shares this marking. su ordinate category embracing a number of distinct but related categories. coul be that English the expresses two or more separate semantic categories. In other words. misle ing us into failing to see them as distinct? Or that “definite” is a broad. We will examine these distinctions more closely in this chapter. so the question arises: are they also a kind of semantically defi expression which does not necessarily appear in definite form in certain langua (like English) in which the encoding of [+ Def ] by an article is limited to a m restricted version of definiteness? Finally. wi view to determining whether they warrant splitting the concepts of definite indefinite into a number of independent parameters of meaning. like that between generic and non-generic. vari distinctions apparently subsidiary to that between definite and indefinite w made: identifiable and inclusive.4 Some semantic and pragmatic distinctions In the discussion of the nature of definiteness in Chapter 1. in English other languages. But generics do have a lot in common with definites in te of behaviour. That is. which just h pen not to have distinct encodings in certain languages.

but there seems to be no language in which they are separately enco by articles corresponding to Art1 and Art2 above. and that we must seek to unify them. Pustejovsky (1995). Of the first three groups. The conc sion must be that the dichotomy between identifiability and inclusiveness does reflect the reality. But the evidence for identifiability and inclusiveness being distinct featu is lacking. oth correspond ambiguously to either Art1 or Art2. is linguistic.1. in that the context within which the referent is to found. and it seems improbable that either one can reduced to the other. situational.1 Identifiability and inclusiveness It appears that identifiability and inclusiveness are both required to acco for the facts of the use of the. especially. Perhaps they are distinct features which overlap. for example. so that Art1 and Art2 are both translatable by the in Engl but whereas some occurrences of the correspond to Art1 and some to Art2. and general knowledge uses. The other groups of uses relate to the extralinguistic context. cuts across the first three (so that. 158 . Numerous languages are like English in expressing both by a sin article. One might imagine a language in which they are enco by dicerent articles. expl ing encyclopedic knowledge or knowledge of the immediate or wider situatio The discussion in Chapter 2 showed that there is evidence from langua for drawing a line between anaphoric and other definites. which happen have the same encoding in English (and many other languages). so th number of uses of the can be characterized by either.2 Anaphoric and non-anaphoric definites The various uses of the identified in Chapter 1 (whether most re ily accounted for in terms of identifiability or inclusiveness) fall into three m groups: anaphoric. with little obvious dicere in what is conveyed. or within which the reference is inclusive. the anaphoric defin stand apart from the others. consisting of discourse. nor in which there is only article encoding identifiability but not inclusiveness.1. Some langua have a definite article which is only used anaphorically (Hidatsa -s. and for more recent and detailed work polysemy see Ruhl (1989) and. situatio uses may be associative or not). or vice versa. Ewe á 1 For a clear discussion of the distinction between ambiguity and vagueness (and betw homonymy and polysemy). 4. either by deriving from the other or by deriving both of them from some as yet undiscovered c cept of which they are manifestations. treating them eithe separate categories or as distinct sub-categories of definiteness. see Kempson (1977).category of definiteness?1 4. a fourth major gro that of associative uses.

In languages displaying b an anaphoric and a non-anaphoric article. and noun phrases with either k’E or kC (or other defi determiners) are ruled out. wh relation to meaning can vary from language to language. its range in a particular language being shown by which uses req the presence of a definite article or other definiteness marker. and generally assumed). they are hyponyms. A given language need treat the full range of these uses as grammatically definite. Rather the art should be simply [+ Def ]. it would appear that only on the categories hyponymous to definiteness is encoded: [+ Def. and which need not e be semantically uniform within a single language. That Lakhota k’E and kC do share the specification [+ Def ] is clear from t behaviour in the internal head relative construction investigated by Williamson (19 and examined below in Chapter 6. But feature specification would not be justified as a description of the content of article in the absence of any contrasting [+ Def. Perhaps [+ Def ] always involves identifiability and/or inclusiven but with varying characterization of the domain within which identifiability or clusiveness may apply: the previous discourse in Hidatsa and Ewe. Thus k’E is [+ Def. the previ discourse or a range of situational and other domains in English. but merely fine. but this is not what articles encode. If definiteness is a semantic or pragmatic c cept (as so far implied. in encoding just the broader category. For languages like Lakhota it is not necessary to see the two categor anaphoric and extralinguistic. sub-catego of a broader category of definiteness. The internal head in this construction is c strained to be indefinite. because they be described in terms of identifiability or inclusiveness. + Ana]. Thus (setting a complex definites). in Eng . Eng the. as unrelated. so the feature sp fication [+ Def] can segment the semantic field at dicerent points in dice languages. it seems not to be uniform ac languages. Another possible approach is to take [± Def ] to be a grammatical feature. Of course it is important in making such ob vations on a language to make sure that the determiner used for anaphoric re ence is a definite article rather than a demonstrative. is not ambiguous. − Ana] article. − Ana] or (if it can appear in anaphoric contexts) simply [+ Def ]. an interesting question is whether latter is excluded in anaphoric uses. This is the treatment I s argue for in later chapters.(Lakhota k’E and kC respectively). rather. In languages with only an anaphoric article. + Ana] and kC is ei [+ Def. The problem then is to account for the restriction the definite article to anaphoric use. These uses represent “sem tic definiteness”. Thus for languages in general there is a range of n phrase uses which can in principle be characterized as definite. The Lakhota article kC appears not to be excluded. though this is not entirely clear. in Ewe only anaphoric noun phrases are [+ Def ].

notably Kempson (19 and Heim (1989). To k the two concepts apart. so that definite article is said to be essentially deictic (that is. 19 chapter 15). I have used it to denote distincti involving closeness to or association with some centre (typically the speaker the moment of utterance). I have already suggested something al these lines for proper nouns. The sa view is held by Clark and Marshall (1981) and others. is one which draws the hear attention to a referent.3 and 4. as grammatical rather than semantic. T point will be discussed further in 4. Ostension is present in demonstratives. fits in well with the view of some theorists. that the anaphoric use is basic and other uses derivative. it may 2 There seems less reason to interpret some of the other features introduced in Chapter 1. who claims that deixis is at the root of definite reference. then. and at the sa time deictically neutral (expressing no distinction of the proximal–distal kin I will substitute my labels in outlining his position. An ostensive expression in J.4. since I will argue in Chapter 8 that there is no feature [± Def]. Others use it to describe expressions which direct hearer’s attention towards a referent – essentially what demonstratives do. 160 . where it is ty cally complemented by deictic information such as distinguishes this and t in English. This is the position of J. unfortunately uses “deixis” in both senses. that the use of definites in the immediate situat where the referent is perceptible (often termed the “deictic” use) is basic. [± Prox]. ostensive). T view is based on the assumption that definiteness is about familiarity (and by ex sion identifiability). let us adopt the label “ostension” for the latter use. and anaphora represents the paradigm case of familiarity. Lyo in the works referred to. J. general knowledge and generic noun phrases. possibly indefinite in English and definite in Gre so that the same entity is represented in dicerent ways by dicerent languages. This means that a n phrase may be definite in French and its translation equivalent indefinite in E though the two are referentially identical.1. Lyons (1975. defi ness being represented in the grammar as a functional head rather than a feature. Recall that the term “deix is used in dicerent ways by dicerent writers. of anaphoric definites being separa encoded in some languages and being the only type of definite encoded in ot languages. Lyons’s account. and in Chapter 7 an analysis definiteness as a grammatical category will be proposed which is compatible w definiteness varying considerably across languages in the range of noun phra it characterizes. there is an opposing view.2 4. But a determiner may be ostensive without deixis (that is. and anaphoric use (and others) derived. But an inconsistency of treatment in this turn out not to matter.3 “Deictic” and non-“deictic” definites The picture just discussed.situational.

It is either directly perceive the physical surroundings or straightforwardly recalled from the preceding discou All other uses are arguably more complex in that they require some inferenc to interpret.2. two referents mentioned the preceding discourse can be distinguished by the use of a distal demonstra for the one mentioned earlier and a proximal demonstrative for the one m tioned more recently. The relationship between an anaphoric expression and antecedent is itself a temporal one. using background knowledge and exp ence. Fre ce livre ‘that book’ and le livre ‘the book’ are not synonymous. also does not oc But an interesting phenomenon does occur which bears some similarity to w we have just sought and not found. and the latte well as the former can be used situationally. is t derived from their situational. A distinction of article form corresponding ra closely to this division is found in Fering. since he does not recognize a distinction between a deictically neutral dem strative and a definite article. Suppose we set up an extended notion of ostension ( it “textual–situational ostension”) which includes anaphoric reference as wel reference to an entity perceptible in the physical situation. Demonstratives are the archetypal osten expression. what the referent is. and anaphoric reference involves the trans ence of basically spatial deictic concepts to the temporal dimension of discou The anaphoric use of demonstratives. thus which have a definite article used only for referring to something physic perceptible. of cou be temporal as well as spatial.2 that they are distinct. the “A-article” and “D-article”. unlike articles. and demonstratives occur freely in anaphoric as well as situational probably in all languages. In many languages. Deixis can. The other phenomenon we m expect to find. But I argued in 1. the hearer must calculate.and then it is likely to be unstressed. one might expect that there are languages which have only this kind of defin ness. a North Frisian dialect studied in d by Ebert (1971a. Fering has two definite articles. ostensive use. But we have seen no instances of this. Ostension in simplest form is reference to entities present in the physical situation of uttera and this is the basis of all other uses of definite determiners. with the following principal forms: . 1971b). What these have in c mon is that the referent is immediately accessible. and by extension the definite article. Moreover. one for situatio ostension and the other for anaphoric and other derivative use. deictically neutral dem stratives are phonologically strong. and Lyons might consider this meets the pose. so that a distal demonstrative is likely to be u to refer to events further into the past or future than a proximal demonstrat This is the link with anaphora. a language in which two definite articles occur. This is the definite article. If situational ostension rather than anaphora is the most basic kind of defin ness. Many languages hav demonstrative not marked for deixis.

where the referen identified in a following relative clause as in (2). referents uni or identifiable in the wider situation as in (6).’ (2) Det buk. In some cases it is because the us 162 . ‘The book that he wrote first is worthless. generics as in (5). wat a feringen snaaki. ‘The/those cats are scratching. Di hingst haaltet. ‘Cats scratch. and when we there.’ (9) Hi ded ham a hun.’ (6) Ik skal deel tu a kuupmaan. referents identified by a p nominal adjective as in (7). in associative occurrences as in (8).’ (7) A fering spriik as det spriik. the door is locked.’ (8) Jister wul wi deel an Sina bischük. cataphorically. do a dör feest. ‘Yesterday we wanted to go (down) and visit Sina.MSG FSG NSG PL A-article D-article a at at a di det det dön The D-article is used anaphorically as in (1).’ (3) Dön kaater kleesi. ‘The Fering language is the language that the Föhrer speak. ‘I must go (down) to the grocer.’ There are various contexts where both the D-article and the A-article are possi sometimes with a dicerence in interpretation. ‘He gave him his hand. ‘Oki bought a horse. and (rather t a possessive) with inalienable properties as in (9): (4) a san ‘the sun’ (5) A kaater kleesi. wat hi tuiast skrewen hee. and where the referent is vis in the physical context as in (3): (1) Oki hee an hingst keeft. dochts niks.’ The A-article is used with uniques as in (4). The horse is lame. an üüs wi diar uunkem.

while di hingst would imply a contrast with other horses. ‘At the market they wanted to palm oc a cow and a horse on The horse I naturally didn’t buy. or when a perceptible referent in the immediate situatio described with a pre-nominal adjective as in (11): (10) Dön/A foomnen. thus when a past day or today is meant. and in some cases distinguished f other entities satisfying the same description (as in (12) ). a hingst would mean the horse was being spoken of in contrast with cow. that with the A-article. ‘I had my hands full all day. given the necessary knowledge and inferenc skills. The A-article is u generically. The referent. the D-article is used where the identity of the referent is to found by searching the spatio-temporal or textual context. then. and where. wat ei mulki kön. a is always possible. th fore.’ (14) a/di prääster faan Saleraanj ‘the vicar of Süderende’ In (12). In general. fu neen maan. In (14). or a p reference to it. Vergn . In (13). di is only possible wh the conversation is not taking place in Süderende and the speaker does not there. di is o possible when a particular day in the past is meant. a is possible on all inter tations. This dic ence is reflected in the greater morphophonological substance of the D-articl The dicerence just suggested between the two articles has been explo recently as support for another putative distinction among article uses. ‘Girls who can’t milk don’t get a man. given the hearer’s general knowledge or knowledge the wider situation and of appropriate associations.’ (11) Wäl dü mi ans di/a brons dask auerda? ‘Will you pass me the purple bowl?’ Other cases are more intriguing: (12) Üüb’t markels wul’s mi an kü an an hingst üübdrei.’ (13) Ik hed a/di hiale daai a hunen fol tu dun. Di/A hingst ik natüürelk ei keeft. the descriptive part of the noun phrase plays the major role in refe identification. is there to be picked out.modifier as in (10). the description is enoug single out the referent without the need for ostension. while with the D-article the article itself has greater significa signalling the presence of identifying information in the surroundings. It might be suggested.

Recall from 1. This expletive article occurs essentially in generic and proper n phrases. using the informal notion textual–situational ostension suggested above. In several environments. the is almost alw 164 . please. eit this deictically neutral demonstrative or the definite article can be used with v little dicerence in what is conveyed: (15) [In a room where there is just one stool] Pass me the/that stool. in languages where these take the article – environments in which th is a long tradition of seeing the article as not semantically motivated. and all situational uses other t those involving the immediate situation where the referent is perceptible require A-article. while the exple is not definite and merely fills a Det position which for various reasons may be left empty. and ten years it never occurred to me to move on to anything better. (16) I bought my first car. and before an identifying relative clause. (17) The exam results came out this morning. see 5. 1. distinct in content. particularly to refer t perceptible entity. The substan article is [+ Def ] and functions semantically as an operator.2. But (as he partly acknowledges in a footnote) the distinction betw the two articles of Fering does not correspond closely to Longobardi’s distinct In particular. I o replaced the/that car when my new husband refused to be seen d in it. This does not mean that the and that/those are synonymous in these contexts. I w return to the expletive–substantive distinction in Chapter 8 and suggest reas for rejecting it.2. are claimed to be neutralized in many guages. There is another way to interpret the Fering data. an old banger. uniques like ‘sun’. with minimal dicerence in what the hearer understands). me that their closeness of meaning produces an overlap in use (somewhat as on a p ticular occasion one might freely choose either the car or the vehicle to pick the same object.substantive (meaningful) occurrences of the definite article and expletive pleonastic occurrences in which the article has no semantic content. This article cannot be characterized as expletive in these uses. and the/those students w passed are already at the pub celebrating.4 bel These two articles. associative uses.2. in anaphora. Now is striking that in the English translations of Ebert’s Fering data. I the means to get from A to B and that was all I cared about. but Longobardi points to Fering as one language which distinguishes th morphologically.5 and 3 that many languages have a demonstrative unmarked for deixis and that Eng that/those can have this value. when I was twenty.

a dem strative. like the German Scandinavian determiners with which it is cognate. the area of noun phrase use where simple definiteness and demonstrat ness overlap. the first refers to something specific (and fa iar to the speaker). it is precisely in this function definite article and demonstrative overlap. m over. On one view. a similar distinction can be made for definites. so that the c acterization of the specific–non-specific distinction just given would not be versally accepted (at least as a semantic account). hyponymous to definiteness. indefinite noun phrases being specific or non-specific fact. as will appear below. while neither invo a referent identifiable to the hearer. that Fering divides the field of definiteness al a line corresponding to a real semantic or pragmatic distinction. since understanding the sentence I bought a car does not invo . so we should take it to be a g uine article. which we h seen is separately encoded in some languages. I assume. A similar but “weak position is that definites may (but need not) refer. that the D-article can be unstressed.2 Specificity and referentiality We have seen that indefinite noun phrases like the direct object the sentences I bought a car and Pass me a book dicer in that. Wilson 1978. and I will return to them in Chapter 9. reflecting some of the stages by which langua acquire or replace definite articles. The D-article form certainly can be demonstrative. Ebert notes that the D-art can be stressed and the A-article cannot. This distinction is usually discussed under heading of specificity. The subcateg represented by my extended notion of ostension. we would have to account for why the true definite article. or to speak of an arbitrary member of class described by the noun phrase. Textual–situational ostension is.the Fering D-article represents textual–situational ostension. One possible account of the Fering facts is that the D-article is. I believe the facts of Fering have important implications for diachrony of definiteness. the extended vers of J. the A-arti is excluded from a wide range of uses normally open to a definite article. Lyons’s concept of ostension (“deixis”). There is an ongoing debate in the semantics literature on the question whether definite or indefinite descriptions can involve reference. It see moreover. therefore. while specific a car is a referring exp sion (Donnellan 1978. But if the D-article were always dem strative. but that indefinites do not re This is precisely because the “referent” of a specific indefinite is not identifi for the hearer. m well itself be superordinate to the lower subcategory of anaphoric. An indefinite singular noun ph may be used to denote a particular entity. but the second does not. Fodor and Sag 1982). with textual– uational ostension on one side and all other uses on the other. 4. in fact. non-specifi book above describes but does not refer.

if one says I bought the car. expecting hearer to know or to work out which car is meant. Thus while the pro sition representing the literal meaning of a sentence containing a definite or indefi description will involve quantification. then the definite the car is referential. Lud and Neale 1991). it is not possible to substitute for noun phrase in question a coreferential expression. potential or hy thetical.2. a c which includes proper nouns. On the other hand. modals. hope. These elements. following the acco of Russell (1905) (to be discussed in Chapter 7). For a clear and accessible discussion of the issues involved in this controve see Larson and Segal (1995: chapter 9). Neale 1990. conditionals. which h in common that they present a proposition as counterfactual. On this view. I shall continue here to speak inform of both simple definites and indefinites as potentially referring. Sim definites and indefinites describe. 4. so that refere is treated as a matter of pragmatics rather than semantics. for at least some writers. reference is limited to “singular terms”. definites are not always referential. expressi whose meaning consists essentially of picking out an individual entity. indefinites) have referential uses. questions. and denote whatever meets the descript (definite) or something which meets the description (indefinite). set up “opaque contexts” – so called because in th contexts. inten negation. But it is recognized that defin (and. his next-d neighbour or one of his neighbours could be substituted for a merchant ban 166 . so that identifying a refe is part of understanding the sentence. on the non-specific interpretation. rather than factual. B as will appear below. These contexts include th involving verbs of “propositional attitude” (such as want. future tense. I mean no more than that ther a particular object which the speaker is thinking of as motivating the choice description. demonstratives and personal pronouns. The opposing view (which is probably the more widely held) is that neither defin nor indefinites are ever semantically referring (Kripke 1977. this sentence may be used to convey a pro sition involving direct reference. For example. and may be ambigu between a referential and a non-referential interpretation.1 Opacity and scope ambiguities Certain grammatical contexts have the ecect of creating an ambi ity corresponding to the distinction drawn above. An obvious assumpt is that this is an ambiguity in the article. In charac izing indefinites like a car above as referring. in (18a) bel if the merchant banker Peter plans to marry lives next door to him. but with commitment as to whether the reference is semantic or pragmatic. They analysed as quantificational rather than directly referring. of the sentence.

b. but it does make this more probable. We can also incl here as opacity-creating expressions “intensional” verbs like look for. b. b. I give under e heading both an indefinite and a definite noun phrase acected by such ambigu to show that the phenomenon applies equally to both. Note that the continuation does not in all cases impose the appropriate spec or non-specific reading. b. Questions (22) a. In the examples that follow. could be substituted here. In each example the clause shows the relevant ambiguity. b. (23) a. Joan wants to present the prize to the winner – but he doesn’t w to receive it from her. since the descriptive content of the noun phrase is crucia the truth value of the sentence – only a synonymous expression. (19) a. I didn’t meet the professor during my visit to the philosophy dep ment yesterday morning – so I began to wonder whether that c had been filled yet. I didn’t meet the professor during my visit to the philosophy dep ment yesterday morning – but I managed to get hold of him in afternoon. Joan wants to present the prize to the winner – so she’ll have to w around till the race made in (18b). Peter intends to marry a merchant banker – even though doesn’t get on at all with her. He didn’t see a car parked at the door – until the two men got of it and asked him for directions. which is then cleared up by the contin tion. Negation (20) a. b. and in this way ma clear what the two readings are. (21) a. such as some who works in a merchant bank. Peter intends to marry a merchant banker – though he hasn’t one yet. He didn’t see a car parked at the door – so he knew the visi hadn’t arrived yet. Have you found a watch? – I’m sure I left it lying here. Verbs of propositional attitude (18) a. which duce the same kind of ambiguity. Have you found a watch? – or can’t you decide what kind you w to buy? Did Fred meet the woman of his dreams during his trip to Pol last year? – or am I mistaken in thinking that accent is Polish? Did Fred meet the woman of his dreams during his trip to Pol last year? – or is he still looking? .

I’m going to have lunch with the president tomorrow – that is. (29) a. b. (25) a. Liz is looking for a business partner – but it will have to be som one with plenty of experience in catering. The terms “specific” 168 . The readings indicated in the (a) examples of (18)–(31) are termed ext sional.(24) a. equivalent). buy it for m need one. you’d better tell Joe. it’s mine. I’m still searching for the solution to this puzzle – and I think close to finding it. b. you must know what unpleasant person she is. in contrast with the intensional. b. beca he’s convinced it doesn’t belong to anyone. that is. In each case the noun phrase in question is understood as denoting either a p ticular entity. We may visit John’s cottage soon – if he gets round to buying o Future (28) a. You should go to a film at the Odeon tonight – don’t just sit at ho (27) a. strictly. he’s such a boring man. I’m going to have lunch with the president tomorrow – I’m dre ing it. There is great var in the terminology used in the linguistic and philosophical literature for this tinction. If you’ve come across a copy of the Decameron. I’m still searching for the solution to this puzzle – though John ins it’s insoluble and I think he’s probably right. You should go to a film at the Odeon tonight – it’s superb. b. Liz is looking for a business partner – the poor fellow disappea last month and she suspects he’s been kidnapped. If you’ve met the owner of this cat. I’m going to buy a suit tomorrow – you’ll be horrified by the col I’m going to buy a suit tomorrow – even if I can’t find one I re like. de di non-specific or non-referential readings given in the (b) examples (though four terms are in each case not. I le here. as discussed) and on the other it is not. If you’ve come across a copy of the Decameron. if election takes place today and we have a president. b. Intensional verbs (30) a. b. or whatever satisfies the description. We may visit John’s cottage soon – if he invites us to see it. on one reading it is “re ring” (in some sense. Modals (26) a. If you’ve met the owner of this cat. (31) a. b. specific or referential. de re. b.

the existential quant has narrow scope relative to the negation operator. On the narrow-scope. b. ∃x (stranger(x) & met(John. x) ) ∃x (stranger(x) & ~ met(John. and this interaction. A the definite or indefinite noun phrase has an existential entailment as part o meaning. interrogation. In (33b). the modal operators of modal logic. includ negation. reading. amounting to relative sco is expressed in terms of relative order. a belief operator. ~ is the negation operator. Andersson and Dahl 1977). In (33) it is through its interaction with the quantifier that ne tion creates an opaque context. This can be illustrated in relatio the following sentences. he couldn’t have. the existential quantifier is subordinate to the operator. The amb ity is then a matter of whether the existential quantifier is in the scope of the opera or vice versa. the quantifier has w scope. The element which has scope over the other is said to have “w scope” and the element within its scope has “narrow scope”. x) ) In these formulas ∃ is the existential quantifier. n specific. If the existen quantifier has wide scope. John met a stranger. and th need not exist any individual corresponding to the quantified expression. The element which creates the opa is a logical operator. representing the specific reading on which for a particular person with wh John is not familiar. a continuation is possible in which the existence of such an indi ual is explicitly denied: (34) John didn’t meet a stranger. for which the simplified logical formulas given show essential points of dicerence: (32) a. The standard account of the availability of two readings of noun phrases in opa contexts is in terms of “scope ambiguity”. x) ) (33) a. this corresponds to the specific reading.the distinction applies equally to definites and indefinites. c. if it has row scope the non-specific reading applies. John didn’t meet a stranger. In (33c). expressed as the existential quantification of predicate logic (for an mentary account of which see Allwood. so that (32b) may be read as: ‘There is some x such that x is a stran and John met x’. he knows everybo . an is a variable. corresponding to the non-spec reading (‘It is not the case that there is some x such that x is a stranger and J met x’ – thus roughly ‘John didn’t meet any stranger’). Thus this reading. ~ ∃x (stranger(x) & met(John. John did not meet that person (‘There is some x such th is a stranger and it is not the case that John met x’). b. logicians have posited many dicerent operators.

A train passes here every half hour. or neither. I don’t know how we’ll find classroom space for these ex students – but they’re so well qualified we couldn’t refuse to acc them. b. Scope ambiguities occur with quantifiers as well as with the kinds of operators discussed ab Consider the following: (i) (ii) All the students had read a novel by Flaubert. I don’t know how we’ll find classroom space for these ex students – but the question may not arise if applications are substanti down. This reflects the dicerence between the unive quantification in these examples and the intensional operators in (18)–(31). where there is no question of scope ambiguity. b. Debbie’s still waiting for Mr Right – he’s called Mark. On the specific reading (33c) and with the non-negated (32). but need not have.2 Transparent contexts Distinctions similar to those just discussed can obtain in transpar (non-opaque) contexts. and I only count fourteen. but it. are in fact a property of all types of potentially re ring expression.3 This existence entailment is often treated as a property of the article. and it may be same train that passes each time. The students may have. I’m missing a student – there sho be fifteen. (37) a. Debbie’s still waiting for Mr Right – I wonder when she’ll fin see there’s no such thing and settle for Bob. Tom plans to bring up three children on his own – but first he ne to find a woman to bear them for him. the scope facts discussed. 170 . Tom plans to bring up three children on his own – they’re horr brats and I wish him luck. I haven’t started the class yet. the other hand. read the same Flaubert novel. the existence of a particular stranger is entailed.2. b. whether they contain a. and she’s as crazy about him as when they first met last year. she must be put up with his unpunctuality. Let us be by considering indefinites: (38) a. 4. (36) a. b. as shown by the foll ing examples: (35) a. the. but is probably not. 3 I haven’t started the class yet. The Department has decided to expand its student intake by ten year. But on both readings the existence Flaubert novel or train is entailed. I’m missing a student – Mary’s alw late. The Department has decided to expand its student intake by ten year.non-specifically.

The major discussion is Donnellan (19 whose much repeated example is (40): (40) Smith’s murderer is insane. by contrast with the scope ambiguity cases. the more overtly definite the murd of Smith would serve equally well.) On one reading of (40). On the attributive re ing. but about whoever murdered Smith – the murdering of Smith or so thing associated with that deed being the grounds for the attribution of insan If it should turn out that Smith was not murdered. On the (a) reading of (38) and (39). the existential en ment does not seem to be suppressed in these cases. But on the referential reading. (Notice the definite possessive structure.b. though it seems to be distinct from the one illustrated in (18)–(31) in there is no opacity-creating operator involved. the exp sion is vague between readings on which the speaker either has or does not h a particular referent for the indefinite noun phrase in mind. A similar distinction is also found with definites. who might just as well have b described in other ways (Joe Bloggs. The terms “specific” and “non-specific” are also used for this dist tion. the speaker has a particular individual in m On the (b) reading this is not the case. The most common account of facts exemplified in (38)–(39) is that the distinction is pragmatic rather than sem tic. Notice that. A dog was in here last night – there’s no other explanation for these hairs and scratch marks. on the other hand. the description only serves to pick out the individual. then on the attributive read there is no person to whom the utterer of (40) can be said to have attributed ins ity. That is. the fellow in the pinst suit). it sit by the fire on wet nights. (40) is a statement not about a particular individual believed to have murd Smith. The existence of a perso thing to whom the description applies is assumed by the speaker and canno denied in the continuation: . the identity of the student or dog is be the point. Annie’s boyfriend. the referential o Smith’s murderer refers to a particular individual. Rather. the description used is crucial to what is stated or c veyed. Here there is a tendenc the literature to use dicerent terminology – “referential” versus “attribut rather than “specific” versus “non-specific” – but the distinction may well be same and require the same explanation. there is no ambiguity in the sense of each such sentence having semantic representations (dicering either lexically or in structure). insanity was still attributed to someone person the speaker thought had murdered Smith).

the ambiguities discussed in 4. In the former the speaker intends to communicate something abou particular individual and intends the hearer to realize which individual is intend In the latter the speaker has a particular individual in mind corresponding to description but does not expect the hearer to pick out any individual. distinguish a “referential” use from a “specific” use. These writers thus distinguish. On this taxonomy one missing though. because in each c there may be more than two readings to distinguish. I will not pursue these questions further. or to suppose specificity is hyponymous to either definiteness or indefiniteness (as anaphori may be). that is. following ideas developed in Kripke (1977) and e where.3 A unified account of specificity? It appears from the above discussion that there is no reason to p an ambiguity relating to specificity in the articles themselves.2. it may be that both referentiality and specificity are common with defini but that the former is somewhat marginal with indefinites. for both definites indefinites. Ludlow and Neale (19 and Larson and Segal (1995). however.2 apply to all types of po tially referring expression. th is no intention to communicate something about the individual which provides grounds for the assertion. in addition to the literal meaning (alw quantificational according to Kripke). and half the class is absent. For more detailed discussion of issues involved in specificity. To distinguish the two kinds of ambigu discussed I shall henceforth use the labels wide-scope and narrow-scope for 172 . Rouchota (1994).2. We can’t start the seminar. E (1991). 4. which is purely quantificational. be more complex. because the student who’s giving presentation is absent – typical of Bill. (42) ??Smith’s murderer is insane – even though no one has murdered Sm The following example can be compared with (38) to show further the pa lelism with indefinites: (43) a. These uses are both distinct from the attributive or n specific use. and Jaszczolt (1997). because the student who’s giving presentation is absent – I’d go and find whoever it is. he’s so unreliable. and involve semantic or pragmatic distinctions wh cut across the definite–indefinite one. and the grounds the speaker has for the assertion. but nocan remember.2. the proposition the speaker wishes to co municate. The correspondence between the dicerent readings of definites and those of in finites in transparent contexts may. b. see. in addition to the references given above. involving no individual “refere at any level.1 and 4. We can’t start the seminar.

A similar observation can be made in relation to English indefinite (to be discussed below). In fact matters are less straightforward than this. In many languages which have overt en ing of the specific–non-specific distinction. thus narrow-sc readings and non-referential readings are both types of non-specificity. It is simply that cer contexts. and I shall use specific and n specific as informal cover terms to embrace both distinctions. in addit contexts in which a specific or a non-specific reading is virtually obligatory.for those appearing in transparent contexts.5 A in Pass me a book. fav a non-specific reading more than transparent contexts do. cating that the identity of the car bought is of no importance to the point being made. 5 But it is arguable that. But in fact the parameters of scope and referentiality can vary to some extent independently. where the description is in the scope of some intensional element. She not know this individual and could not identify her. a non-specific reading is unavailable because the speaker must hav referent in mind where she is reporting having had some interaction with it. not so much of the speaker. I wonder what she’s like. There are. but ra of the subject (Peter) having a particular individual in mind: Peter intends to marry a merchant banker. O might try to unify them by taking the pragmatic concept of the speaker havin particular referent in mind to apply to all cases. which signals specificity. Exam will be given below. narrow-sc noun phrases. I (1977) shows that wide-scope noun phrases can be referential or non-referen and non-referential noun phrases can have wide or narrow scope. a specific reading is practically impossible because if the spea did have a particular book in mind she would want to ensure that the hearer pic out the same book. among other things.4 On this view the vagueness betw specific and non-specific is in principle always present. in I bought a yesterday. There have been attempts to reduce both kinds of ambiguity to a single spec versus non-specific distinction. though they are generally held to be distinct. or a demonstrative). . it seems to be a matter. but she has in mind the person that P intends to marry. In the two types of context dicerence in interpretation seems similar. the non-specific form can occur in this context. glossing over. the speaker does have someone specific in mind. a pragmatic explanation can be advanced for these cases. must be non-referential (thus non-specific in b 4 In an example like the following. on the other hand. Bear in mind that this is a delibe simplification for expository purposes. I bought this car (with this understoo indefinite rather than demonstrative) invites a more specific reading than I bought a car n to receive. even here. the l lihood that there are more than two readings to be distinguished in a given c The purpose is merely to distinguish the phenomena we have observed in opa contexts and those seen in transparent contexts. Thus. this would require a definite description (probably with an id tifying relative or adjectival expression. There is in her mind a referent corresponding to the description.

senses). however. Lakhota distinguishes quasi-indefinite determiners: a “realis” one wA ‘a’ with plural k‘eya ‘sm’.’ And the subject pronoun kto-to in (45). someone NEG sings ‘Someone isn’t singing. argue that sentences with more than one quant are not ambiguous. b. See also van Deemter and Peters (1996). she wants go in-marriage to someone ‘She wants to marry someone (a particular person).6 Similar facts appear Lakhota (Williamson 1987) and Jacaltec (Craig 1986). consisting of an interrogative with a suax -to or -nibu The distinction between these two forms corresponds to the concept of specifi defined in terms of scope. typic by encoding one but not the other. because this interpretation is entailed by the wide-scope reading and is there common to both readings. Examination of these may help in id tifying linguistic distinctions which English does not.’ Ioup takes this to point to the distinctness of scope ambiguities and referentia the former being semantic and the latter pragmatic. make ove In fact a number of languages treat the two types of specificity dicerently. Dahl (1970) and Ioup (1977) refer to indefi pronouns in Russian. The availability of two interpretations is seen as a matter of va ness rather than ambiguity. This corresponds to the narr scope reading. Th (44) a. Ona khochet vyjti zamuzh za kogo-to.’ Ona khochet vyjti zamuzh za kogo-nibud’. ‘She wants to marry someone (anyone). -nibud’ is used only in opaque contexts and indica that the pronoun is to be interpreted as within the scope of the opacity-creat expression. and “irrealis” one wA+i with plural ’etA. the wide-scope (“stronger”) interpretation being a special cas the narrow-scope (“weaker”) interpretation. while -to is used both in opaque contexts when the pronoun has w scope and in transparent contexts regardless of considerations of referentiality. The latter only occurs w 6 Kempson and Cormack (1981). see 2. do have lex or morphological encodings for specificity.2 above.5. Thus: wide-scope referential narrow-scope non-referential It is clear from the survey in Chapter 2 that there are many languages wh in contrast with the English data examined so far in this chapter. or need not. 174 . is ambiguous between referential and non-referential readings: (45) Kto-to ne pojet. which cannot be interpreted as narr scope. but have a single semantic representation.

and I have found no example of a guage which encodes the distinction between referential and non-referential not that between wide and narrow scope (though there are many that enc both identically).’ C‘a ’eta aku we. plural ’etAni. uses are illustr in (46): (46) a. in which -uj is the irrealis suax. It is also the case in the three languages discussed above that row scope rather than wide scope is the morphologically marked term of this op sition. opacity-crea elements. c. started PL CLASS look-for a pot ‘They started looking for a (specific) pot. This is not universal. apple a-IRR want-1SG ‘I want an apple. and there m even not be one. it is the narrow-scope interpretation which is encoded by the addition an aax or by replacing a basic morpheme (like Russian -to) which occurs w wide-scope referential and non-referential readings. b. There is also a special form wA+ini.. T‘aspa wa=i wac‘c. occurring in the sc of the negative marker _ni. and clearly consisting of the irrealis form plus an a tional negative morpheme. The quasi-indefinite article is hu which is replaced by hun-uj. and English is one language which has a form encoding specific rather t . The situation is similar in Jacaltec. I shall return to such fo in 4. and the irrealis mood marker kte – in other words.’ Notice that in the three languages just discussed it is the distinction relatin scope which is indicated morphologically. A possible exception to this generalization is pronominal fo occurring in many languages corresponding to English one and French on. h ever.2 below. Xfoc’ heb’ ix say-af hunef munlab’al. in opaque cont for the narrow-scope reading: (47) a.imperatives.3. narrow-scope.’ Wowapi wa=i lawa kte iyececa.’ In each case the speaker is not thinking of a particular apple etc. Replacing wA+i and ’etA by wA and k‘eya would give the w scope sense. that is.’ Xfoc’ heb’ ix say-af hun-uj munlab’al. wh are probably non-referential and in some cases (English being such a c restricted to a generic subset of non-referential uses. Some irrealis. [book a-IRR read-2SG IRR] be-proper ‘You should read a book. started PL CLASS look-for a IRR pot ‘They started looking for a (non-specific) pot. b. sticks sm-IRR bring-back IMP ‘Bring back sm sticks.

(52) a. b. which is typ of a. It is typically used when the referent is going to be talked about furt and is for this reason particularly suitable for starting stories. b. (49e) shows a context largely limited indefinites. ??Peter intends to marry this merchant banker – though he ha met one yet. You should go and see this film at the Odeon tonight – it’s sup ??You should go and see this film at the Odeon tonight — don’t sit at home. (51) a. Colloquial this is a purely optional alternative to a and is of course stylistic limited. e. specific in both senses. when this man came to me and asked if I was a news announcer. I was walking to work yesterday morning. ??He didn’t see this car parked at the door – so he knew the visi hadn’t arrived yet. specific and non-specific noun phrases. c. which exp a kind of referential prominence. Liz is looking for this business partner – the poor fellow dis peared last month and she suspects he’s been kidnapped. and is most naturally understood as an instance of colloquial t It is also clear from (48) that this can be used in first mentions. It is clear from its behaviour in diagnostic environments that this use of this is indefin (49) a. This woman round the corner breeds pedigree pigs. d. 176 . he couldn’t lift it. b. anecdotes. but there is the optional possibility indicating the specific reading of indefinites by means of the colloquial use of in examples like the following: (48) a. All these sentences are acceptable. It is. moreover. b. Which of these women are you talking about? Some/All of these people know more than I do about phonetics. the indefi must be wide-scope and referential for this to be acceptable: (50) a. This house is mine. but the first four must be understood anaphoric or situational demonstrative uses of this/these – the referent has b spoken of or is being pointed out. Strong as this chap was. (53) a. Peter intends to marry this merchant banker – even though doesn’t get on at all with her. There’s this strange message on the noticeboard. He didn’t see this car parked at the door – until the two men out of it and asked him for directions. elements like object markers and classifiers. jokes In this it is comparable to a number of elements appearing in other language be discussed in Chapter 5.

distinctions of specificity are doubly encoded. ??This dog was in here last night – there’s no other explanation all these hairs and scratch marks. or a definite determiner at least.2. Is it a real indefinite article. it is not a quasi-indefinite article expressing cardinality. falsifying hypothesis that no language has such an expression? Or is it after all a dem strative. and indeed no language appears to mark either scope or referentia distinctions for definites. It is not obvious what to make of this. ‘that’. Thus bál rx ‘a man (non-specific)’. On English indefinite see also Prince (1981). Colloquial this is interesting syntactically because. In fact a number of languages of Austronesian family have an article which combines definiteness with spec indefiniteness (both wide-scope and referential). its indefiniteness notwithstand it is quite clearly in a paradigmatic class with definite determiners. by the morphology or tonol of the noun and by an accompanying marker (possibly a determiner). See the discussion of Samoan and Maori in 2. despite the fact that these distinctions apply very cle to definites just as much as to indefinites. with odd semantics? Its unusual m phology (with the irregular number alternation this–these) suggests strongly th is not a distinct lexical item from demonstrative this.someone with plenty of experience in catering. It will have been noticed that all the languages discussed which show so morphological or lexical expression corresponding to specificity do so only indefinites. though Blass (1990) notes that the Sissala specificity marker nx m be a form of the general demonstrative né ‘this’. This dog was in here last night – it’s called Lulu and Fred alw lets it sit by the fire on wet nights. ??I haven’t started the class yet. In Sissala (B 1990). which are in position. but it is also used for reference to something kno or familiar to the speaker but not necessarily either previously mentioned or fa iar to the hearer. báálW nx ‘a man (specifi . non-spec rx or specific nx. This is evid from its co-occurrence with and position in relation to numerals in examples l (56) I was just minding my own business when these two men came to me. (54) a. I’m still missing this student – Ma always late. The German demonstra dieser ‘this’ also has the same colloquial specific indefinite use as English t This could be due to borrowing. b. This determiner is usu described as a definite article. (55) a. and the phenomenon is not common crossguistically. I haven’t started the class yet. b.4. I’m still missing this student – th should be fifteen and I only count fourteen.

We h seen other examples of this: in Samoan. It is not clear which form of the noun is unmarked the non-specific form. like English t when the referent is likely to recur in the subsequent discourse. There are languages which display morphologica lexical encoding of the scope distinction but not the referentiality distinction. and is discussed in m detail and with further exemplification by Givón (1982). It also makes it very unlikely that ther a quantificational–referential ambiguity in the English definite article. there are languages which show the same encoding for both. where the non-specific article se is u where the identity of a referent may be known but is considered unimportant.5. but languages can lexi ize or grammaticalize both. There is another respect in which linguistic expressions of specific or non-spec can go beyond the distinctions of interpretation discussed in 4. 178 . though there a number of interesting mismatches between the distinctions made morpholo cally or lexically by languages and the semantic–pragmatic distinctions identi in 4.2. The phenomenon is common. and the Sissala specificity mar nx has the same function. that the Turkish cardinal article bir. ha particle (from a verb root meaning ‘exist’) which can be used to signal that a erent is to play an important part in the discourse. essentially limited to spec indefinites. I no above that English specific indefinite this is optional and tends to be used wh the referent is to be a significant topic in the ensuing discourse. To summarize. equally.1 and 4. are closely related (as they se intuitively to be) and that languages can treat specificity as a unified phenom non.2.2. Nama. contrary to what I sta at the beginning of this section. definites be neutral in this respect. Scope ambiguities may be semantic (whether involving true am guity or merely vagueness) and referentiality pragmatic.2. though dicerent. too. The fact that specificity and non-specificity (of either kind) are overtly tinguished only in indefinites is striking. if there w we would expect to find languages in which this ambiguity did not hold. This s gests that the two phenomena. where the specific indefinite wani tends to be used. if the specific fo then specific and non-specific are only distinguished in indefinites. the two types of specificity seem to be distinct.2.2.2. because it contrasts strongly with semantic–pragmatic findings. then Sissala treats definites as specific. in Hausa. tends to be omitted when the identity of the referent does not ma What this amounts to is that referentiality and non-referentiality are extended embrace instances where the speaker may be in a position to identify the refer of the noun phrase but chooses to treat its identity as significant or not. wit distinct article form for each interpretation.noun: báálW ná ‘the man’. specificity as a lexical or morphological categ may be hyponymous to indefiniteness. It suggests that linguistically. It was also observ in 2.1 and 4.

Generic noun phrases are those in which reference is made to an en
class, or, perhaps more accurately, which are used to express generalizations ab
a class as a whole – the class in question being that consisting of all the enti
satisfying the description inherent in the noun or nominal. Of course a strai
forward way to refer to the whole class is by means of a determiner such as
every, each, any, which approximate to universal quantification. In fact the t
“generic” is not used of these because of semantic dicerences; as pointed ou
Lawler (1973), Smith (1975), a single exception would usually invalidate a st
ment with all etc., whereas generics admit exceptions, since they express gen
tendencies. It has often been pointed out that no language has noun phrases
tinctively generic in form – whether with a special generic determiner, a morp
logical mark on the noun, etc. Perhaps not too much should be made of this po
given that lexical devices (such as the adverbs typically, generally) are availa
as are determiners like all which come close to generic meaning – especially s
not all generic expressions are generic in the same way, and some admit excepti
more readily than others, as we shall see. But the point is that genericity is t
cally expressed by noun phrase types which also have a non-generic use, wh
is arguably more basic. This leads to the suggestion that genericity is not a pr
itive category of semantic or syntactic description; generic noun phrases are
basically something else and are to be characterized in other terms (such as n
We have observed that in some languages generic noun phrases are typic
indefinite (if it is correct to treat “bare”, indeterminate, nominals as indefin
in others definite. But many languages show a range of available noun phrase ty
with generic value, definite and indefinite, singular and non-singular, as is sho
by the following three, more or less synonymous, English sentences:
(57) a.

A dog has four legs.
The dog has four legs.
Dogs have four legs.

In such cases where a language has several noun phrase types capable of gen
interpretation, it is not clear that they form a unified class of expressions; rat
they may get their generic value in dicerent ways from dicerent basic values. T
is particularly probable in view of the fact that the alternatives are not all f
synonymous or interchangeable.
Some of the semantic dicerences most discussed relate to the nature of
reference to a class: in particular, whether it is to the class as an entity, a seco
order individual; or to the class as the aggregate of its members, the general
tion being about the members of the class. Which of these interpretation

perse, be numerous, abound are class predicates; they require a class (or gro
expression as subject, and apply to the class as a whole, as a unit. Similarly, ve
like decimate require such an expression as object. On the other hand, predica
denoting an action or state applicable to individuals may involve a generic s
ject or object being treated as a collection of individuals. Compare:
(58) a.

Ostriches are rare these days.
Ostriches lay eggs.

Rarity is a property of the class as a whole, while egg-laying is something in
vidual ostriches do. In fact it is not quite so simple, because though individu
cannot be rare (in the relevant sense), it is not so clear that only individuals,
not a set, can be seen as laying eggs. We return to this below. But the point is
these two ways of looking at a class are not equally available to all generic exp
sions. Another important semantic dicerence between generic types is the ex
to which they admit exceptions, that is, the extent to which they approximate
universal quantification; some are closer in interpretation to all N than are oth
Indeed this distinction may relate to the first one, references to a set as a seco
order individual being more liberal as regards members not conforming.
Genericity in noun phrases interacts with aspectual distinctions in verbs, in
generics are generally – though not in all cases – accompanied by verb forms expr
ing habitual/generic/timeless aspect. This has led linguists to talk of generic s
tences rather than generic noun phrases. But a distinction between generic n
phrases and generic sentences has to be made, since non-generic noun phra
can occur perfectly well in generic or habitual contexts:
(59) a.

My best friend shaves twice a day.
This kitten frightens easily.

Also, not all generics require such a context; at least those generics which t
the class as a unit can occur with punctual aspect:
(60) a.

The dodo died out in the eighteenth century.
Dodos died out in the eighteenth century.

Also, at least some generics can occur with progressive aspect (Ostriches are/
ostrich is laying smaller eggs these days), though it is arguable that this is a s
cial habitual use of the progressive.
It is important to note that genericity applies equally to mass nouns and
count nouns. And mass nouns do not, strictly, correspond to sets or classes. J
as we can make a statement about ostriches in general, that is, about the clas

the description butter, or the quality matching the description sincerity –
opposed to specific quantities or instantiations. One might, somewhat loosely, str
the terms “class” and “set” to embrace such memberless “masses” as well as
lections of individuals. What is really needed is a concept embracing both
and the wholes with no atomic minimal parts corresponding to mass nouns;
term “ensemble” is introduced by Bunt (1979, 1985) for this idea, and the t
“kind” is used by several writers in a similar way, as we will see below. I s
adopt the former term (as well as continuing to use “set” or “class” in relatio
count nouns). On the analysis of mass nouns see also Pelletier and Schubert (19
The literature on generics is vast. Some important references are: Smith (19
Nunberg and Pan (1975), Carlson (1977, 1980), Heyer (1987), Burton-Rob
(1989a), Declerck (1986b, 1991), Schubert and Pelletier (1989), and the pa
in Carlson and Pelletier (1995), especially the introductory contribution of Kr
et al.7 For a promising recent pragmatic account of the interpretation of gener
see Papafragou (1996).

4.3.1 Generics in English
Let us briefly survey English generics, since English displays a w
range of generic noun phrase types, and they are the most intensively stud
Consider the following paradigms defined on the parameters of definiteness
(61) a.

A dog has four legs.
The dog has four legs.
Dogs have four legs.
(The dogs have four legs.)

(62) a.

I admire an intellectual when he speaks out.
I admire the intellectual when she speaks out.
I admire intellectuals when they speak out.
(I admire the intellectuals when they speak out.)

The definite plural is in general not available for generic use, though it can
used generically with some types of noun, such as nouns of nationality, and so
nouns denoting classes of classes (for example, names of animals and plants
resenting groups larger than the species):

KriAa et al. (1995) gives a very useful survey of recent thinking on genericity as well
detailed theory of the semantics of many of the phenomena involved. The reader should appr
this work with some caution, however, on the level of description and data; for example
predicate be numerous is said to require a plural subject (wrongly ruling out The family is nu
ous), and cattle is presented as a mass noun.

(64) a.

John has a soft spot for the Finns.

The dinosaurs dominated the earth for a very long time.
The cats – at least the big ones like tigers and pumas – are parti
larly fierce predators.

For further discussion of these see Lyons (1992a).
The distinction between generics which predicate something of each mem
of the class and generics which predicate something of the class as an entit
presented in Smith (1975), with the following examples:
(65) a.

The squid lives on seaweed.
Squids live on seaweed.
A squid lives on seaweed.

(66) a.

The dodo is extinct.
Dodos are extinct.
*A dodo is extinct.

(67) a.

The lion hunts the antelope.
The lion hunts antelopes.
*The lion hunts an antelope.
A lion preys on an antelope.

(68) a.

Pollutants are decimating the squid.
*Pollutants are decimating squids.
*Pollutants are decimating a squid.

The picture is that all three generic types can be used for generalizing over
members of the class. But a N is ruled out for referring to the class as a u
while both the N and Ns are generally possible here.8
This picture is criticized by Burton-Roberts (1989a), who claims there is
justification for positing a type of predicate which holds only of individuals
therefore only of members of a class; for him this kind of generic reference d
not exist. Thus the squid is not ambiguous between “class generic” and “individ
generic”. It is a class generic only, and lives on seaweed is as much predicated
the class as is is extinct. This view is supported by the claim of Kempson
Cormack (1981) that sets can have actions predicated of them which are actu
not performed by the set as a whole or are performed by only some of the memb

Example (68b) points to the impossibility of indefinite plural generics as object of a class
dicate, an unexplained restriction. But perhaps Smith’s intuition of unacceptability here is
ply wrong – I find (68b) less good than (68a), but not impossible. The impossibility of
indefinite singular generic object in (67c), by contrast with (67d), is explained by Smith in te
of a constraint on combinations of indefinite and definite singular generics in the same
tence. Smith examines in detail many other mysterious restrictions on generics.


narios as regards just who did what, but Kempson and Cormack claim there is
ambiguity, the correct semantic representation being one that simply predic
the marking of six scripts to the set of three examiners. Similarly, Six students t
five papers is true if six students took part in a raid in which only one of th
actually carried the papers away; it is true because took five papers is predic
of the whole set of students involved. As Burton-Roberts observes, if a gen
noun phrase can be ambiguous between interpretations as class qua class and c
as aggregate of members, it should not be possible for both interpretations to
present simultaneously as a result of occurrence with both types of predicat
coordination, but it is:

The dodo lived on figs and is now extinct.

So classes can “inherit” properties, actions etc. from their members, and thi
true not only of predicates like live on figs, which we might be able to conc
of the species as well as its members doing, but also of those which can only
understood as holding of individuals:
(70) a.

The dodo had two legs and is now extinct.
Blackbirds often re-use the same nest several times and are incr
ing in numbers.

Burton-Roberts introduces another distinction to explain the impossibility
*a squid is extinct. This is ruled out, he claims, because is extinct is an accide
property rather than an inherent characteristic. The indefinite singular generi
alone in being a property-generic: it denotes not a class, nor a representativ
arbitrary member of a class, but the essential property defining that class – t
not the extension of the noun (the set of entities satisfying the description),
the intension (the sense, the description itself ). Generic a dodo, for example, den
“dodohood”. And an accidental characteristic like being extinct or numerous c
not be predicated of the intension of a noun. We shall return to this.
I have suggested that Smith’s asterisk in (68b) is too strong, though this
tence is certainly less felicitous than (68a). I believe the same can be said (if
to the same degree) of generic subjects with class predicates. The dodo is ext
is considerably better than Dodos are extinct, though both are possible. On
other hand, The squid lives on seaweed and Squids live on seaweed are equ
good. This suggests there may still be something in the distinction between a c
and the members of a class. Reference to a class qua class is markedly better w
the N; Ns inclines more to interpretation as the totality of members, though
as far as a N does. The dicerent generic forms show a cline of acceptabilit
relation to the two kinds of generic interpretation.

tions. The generic type that most readily accepts a class-qua-class interpretat
admits exceptions most readily:
(71) a.

The academic likes his comfort, though I believe Profes
Laserbeam is very spartan.
Academics like their comfort, though I believe Profes
Laserbeam is very spartan.
?An academic likes his comfort, though I believe Profes
Laserbeam is very spartan.

The point is made more clearly when nationality terms (which allow definite pl
generics) are taken into account:
(72) a.

The Italian drinks rather a lot, though I must say Luigi is
*An Italian drinks rather a lot, though I must say Luigi is
The Italians drink rather a lot, though I must say Luigi is
?Italians drink rather a lot, though I must say Luigi is





Definite generics admit exceptions, then, more easily than do their indefinite co
terparts. And the definite plural certainly accepts class predicates more readily t
does the indefinite plural:
(73) a.

The Brazilians are twice as numerous as thirty years ago.
?Brazilians are twice as numerous as thirty years ago.

Again it is a matter of degree; but this is enough to suggest that the distinct
drawn by Smith between two types of generic has some validity.
How can these dicerences be accounted for, and how do the dicerent type
generic expression get their generic interpretation? Let us look at the dicerent ty
in turn, at the same time looking beyond English.

4.3.2 Singular generics
We begin with the definite and indefinite singular generic ty
exemplified by English the squid and a squid, limited to count nouns. Recall
mass nouns like butter or sincerity pattern in most respects with the plural of co
nouns; not only do they not occur with a, but in generic use they do not oc
with the either, but appear in “bare” form comparable to the bare plural of co
nouns. I am taking mass noun phrases, like plurals, to be [− Sg]. So the two sing
generic types are non-central in the sense that they are not available in princ

very common in languages that have articles, French and German for examp
(74) a.

Le/Un castor construit des barrages.
Der/Ein Biber baut Dämme.
‘The/A beaver builds dams.’

And many languages that do not have articles, like Russian, can use the bare
gular of count nouns with generic value:

nje spitsja.
rich-SG-DAT not sleep-IMPRS
‘A/The rich man cannot sleep.’

This is probably to be distinguished from the many languages in which a f
unmarked for number and without any determiner is the central one, used b
with count and mass nouns. In Turkish, for example, absence of plural mark
with count nouns does not entail singularity. Rather, such a noun (or noun phr
is neutral with respect to number, and can be used generically to denote the en
As noted above, it has been claimed (Burton-Roberts 1976, 1989a) that the in
finite singular represents a “property generic”, denoting the intension of the n
with no extension; for this reason this generic type can only take predicates wh
express inherent or defining characteristics. It is to be equated with the pred
tive use of a noun phrase, as in John is a doctor, where a doctor is not a re
ring expression; it predicates a description of the subject. (This is by contrast w
John is the doctor you met, which expresses identity of reference between
referring expressions.) Although this analysis is presented as a claim ab
English only, it would obviously gain in plausibility if supported by compara
evidence. This support is largely lacking, however, since many languages wh
have indefinite singular generics, behaving much as they do in English, do
show identity between these and the form of noun phrases used predicatively
French, Spanish and German, for example, the bare noun or nominal is the u
predicative form (Jean est médecin, Juan es médico, Johann ist Arzt), but the b
singular is not used generically; rather, the quasi-indefinite article appears as in Eng
(un médecin, un médico, ein Arzt). A possible response to this fact is that w
the predicative use of a property expression is non-argumental, the prop
generic is an argument, and that singular arguments in these languages requi
determiner, while non-arguments do not.
The more common view, which I believe is correct, is that the indefinite sing
generic is simply a special case of what I termed in 4.2 the “non-referential” indefi

French on and. Ich muß geh man wartet auf mich ‘I must go. with arbitrary human erence. exclaims: An Indian did this! Many languages have a pronominal form which can be analysed as indefinite singular generic. usually singular. non-referential indefinites (close meaning to non-referential someone). what could one say. or the us English one by some speakers as a vague first person.9 They are therefore like non-pronom indefinites in having the possibility of both generic and non-generic reference: n generic On frappe à la porte ‘Someone’s knocking at the door’. But thi to consider as non-specific only the narrow-scope interpretation. generic On ne pense jamais à tout ‘One can’t think of everythin Man muß alles selbst machen ‘One must do everything oneself ’. nightly pipe-smoking is a defining property of Indi – to be a real Indian you have to smoke a pipe every night. an Indian is in the scope of every. (1995). pipe-smoking occurs nightly. one was rather embarrassed). well. This still leaves the n referential reading. corresponding to English one. Burton-Roberts (1989a) argues aga identifying generic with non-specific on the grounds that an indefinite singu noun phrase can be ambiguous between generic and non-specific interpretatio he points out that the following sentence is three-ways ambiguous: (76) An Indian smokes a pipe every night. to a more limited degree. This can be contrasted with the non-generic n referential interpretation that results from a punctual aspectual context.Nunberg and Pan (1975). 186 . like the use of Fre on as a first-person plural pronoun (On part à six heures ‘We’re leaving at six’). On the specific reading. but the speaker does not have a particular Indian in mind and the descrip is crucial. someone’s waiting for me’ (in which the n referentiality may amount to a judgment that the identity of the person waitin unimportant). as w Tex finds his favourite ten-gallon hat shot through by an arrow and ruined. generic cannot be the same as non-specific. in the case of (76) the generic value is imposed by the habitua generic aspect of the predicate. But one is ess tially limited to the arbitrary. on the non-spec reading. form (They told about it yesterday. KriAa et al. on the sec reading described. Since the second third readings are distinct. generic use: 9 With all these pronouns it is also important to set aside definite uses. German man are dicerent in they can also be non-arbitrary (non-generic). a particular Indian is a nightly smoker. but it may be a dicerent Indian each nig and on the generic reading. in which an Indian has scope over every (as in the first re ing). It is this non-referential use which is to be equated with the gen third reading. The generic reading is in fact the non-referential use in an app priate context.

Jill says she finds this whale pretty b ing since she read about three other whales that have much m interesting habits. ther no middle construction distinct from the passive. . or plural of varieties. and I here suggest an explanation based on a proposal of Burton-Roberts (1989a). and set these. In Chapter 1 I termed this use the “var interpretation”. and she’s asked if she can work on these wh instead. But the implicit agent of the middle construction (Co washes easily or This surface doesn’t paint well ) is by definition arbitrary.(78) *One knocked at the door. but merely reference to varieties. well known that noun phrases can be used to refer to kinds or varieties. as well as to (sets of ) individuals. This possibility is avail for all determiner–noun combinations: (79) We’re studying the migration habits of a whale that may be on verge of extinction. rather than individuals. and so probably generic. Th overt and implicit items encoding non-referential and limited to generic cont are discussed in Lyons (1995a). because on the usual un standing that generic means reference to the whole class constituting the ex sion of the noun. The definite singular is the most diacult generic form to account for. These languages therefore have an indefinite which is marked as non-referen and in the case of English one also generic. and it is restricted to the same habitual or generic aspectual c texts that indefinite singular generic noun phrases are constrained to occur in. but in others (like French English) there is. We can of course use it to refer generically to the whole clas varieties just as we can refer generically to a whole class of individuals: (80) We’ve been investigating fluctuations in the populations of dice whales. Another indefinite “expression” is implicit. like Spanish. understood subject or agent of passive and passive-like constructions. is generally equivalent to an indefi pronoun like someone. eq alent to one. agent of a passive with no overt agent expression (or “by phrase”). thus a whale can denote not only an i vidual animal but also a variety or species of whale. such as was mugged on his way home last night. and we return to them in Chapter 5. Unfortunately I can’t remember the name of whale that we’re studying. I can’t remember the names of the whales she’s intere in either. Burton-Roberts refer this as the “class-generic” interpretation – misleadingly. it is not generic. I just think whales are fascinating whichever ones you All the highlighted noun phrases in (79) are most naturally interpreted as de ing species of whale (except the last one – see below). and it seems that the whale/a whale/whales increase(s numbers when the sun shines. In some languages.

The individuals are usually portions. as is the case also with whales in the last sentence (79). this value is a 188 . The dodo ate figs and The Londoner is spoilt for choice regards theatre are generalizations about dodos or Londoners. But. by contrast with the defi singular. unlike the bare plural and other forms. because. Burton-Roberts’s propo is that. it only occ with generic value with nouns or nominals denoting “well-established kinds”: t the lion or the Coke bottle. but to the class of individuals p ceived as a unique variety. And in the case of recategorization this varieties level. So what we have with the definite singular gen is entities on the varieties level which are treated as unique. The variety interpretation of count noun phrases which underlies the acco given above is presumably related to the use of normally mass nouns as co When mass nouns are recategorized in this way. and therefore defi – like the sun.while the whale and whales can be read either as generic over individual wha or over species of whale. but not *the green bottle. not about varie of dodo or kinds of Londoner. treat the definite singular as the central nominal generic (or “ki referring” expression). in which the variety is viewed as uniq standing in this case (the beer) for all beer. Another hot whiskey please) or on the varieties level (a gr wine. a bare plural like green bottles can have generic value. interestingly. to say that the reference h is not to the “super-class” or class of varieties. the set of all varieties or species of whale. a beer means a type or br of beer. The assumption is that o with such kinds can noun phrases be themselves generic (as opposed to gett generic value from being in the scope of a generic operator. What I mean by this apparent contradiction is that gen the whale is the variety use. the equator on the individuals level. For example. Since. but to the “su class”. but on the varieties level. there is apparently no possibility of a definite singular gen of the type in The squid lives on seaweed. I want to modify this account slightly. in the absence of qualifying information (such as the relative clause he to identify the variety. This is the definite sin lar generic. though Burton-Roberts considers it to be a sim specific singular reference to a unique class of classes. but the class of all individual whales is taken as c sisting of one single variety. not a kind of glass/bottle of beer. it is either on the individuals le (Give me a beer. an expensive perfume). the latter does not norm involve kinds of the former. so a beer me a glass or bottle of beer. KriAa et al. I make this modification because there are instan of the N where one would not normally think of the class having obvious s classes. Now consider the definite singular the whale in (79). the reference is taken to be not to a variety. as happens with indefi singulars in generic aspectual contexts). Thus it is like the generic-ov varieties interpretation of (80).

but this is because the verb ate is ambiguous between cha teristic and event interpretations. Plumbers stormed into the convention demanding longer lu breaks. Mark really loves puppies. 4. b. (82) a. c. The bold noun phrases in (81) admit a generic reading only. in the verb for instance. A widely accepted view is that the indefinite type.non-generic (presumably non-referential) indefinite. what he terms the “bare plur argues that the generic and non-generic uses are in complementary distribut and that their distribution is wholly determined by the linguistic context. like French. It is not certain. is semantically the same whether used genericall not. This en that bare plurals are never ambiguous in context. d. it may be that the restriction noted is a peculiarity of definite singular ge ics rather than of generics generally. definite in others. Frogs are clever. however.3 Non-singular generics Plural count noun phrases form a natural class with mass n phrases. Where the predicate reports an event or a non-essen state. c. b. The following are his examples: (81) a. the notion of well-established kind provides a valid diagnostic for nominal ge icity. Alice personally knows actresses. a predicate which posits a character or property selects the generic reading. (83) Dinosaurs ate kelp. These are the central or canonical generic fo in that they are available for both count and mass nouns in these languages. (83) is ambiguous between generic non-generic readings. it selects the non-generic reading. like English.3. Frogs are awake. Dogs bark. any ambiguity lies elsewher the sentence. Carlson (1977). The man over there believes Texans to be friendly. both dicering from singular count noun phrases in denoting non-ato parts of a totality rather than individuals and for this reason being compatible w certain quantifiers and cardinality expressions (sm/all/a lot of books/water/ *bo I assume this can be captured in terms of number by seeing both as [− Sg]. consisting of the noun nominal with no determiner. limiting himself to plurals. and those in ( only a non-generic reading. For a discussion of generics in Fren concentrating on the definite singular. see Kleiber (1990). the English simple past systematically represen . Pl and mass generics are typically indefinite in some languages.

b. it is not enough to consider the aspect of the verb. This is also true.and non-generic bare plurals. if a generic or n generic interpretation of a bare plural is selected by other elements in the s tence. Lions live in Africa – so if you want to see lions. the open air. Thus lions in (87a) is the topic of the sentence. (85) Mick traps lemmings. The following examp show that other parts of the predicate that can vary for generality are relevan (88) a. (87) a. The highlighted noun phrases in (86a) and (87a) seem to be generic. 10 Cats Cats Cats Cats mess mess mess mess in in in in loose soil. corresponding to a dicerence in information structure. while in (87b) the sa noun phrase is part of the comment. c. in fact. This appe to cast doubt on the claim that it is the predicate that imposes a generic interp tation on a bare plural. b. the second in a discussion ab Africa. that’s where have to go. of the simple present in (81a). Notice that a dicerence in intonation is likely betw (87a) and (87b). however. desc ing what happens at a certain point in the drama. these elements of discourse structure w be discussed in Chapter 6. gardens other than their own. even though he knows full well that they protected by law. As part of the text of a play. Lions live in Africa – in fact there are more lions in Africa than other continent. b. There are probably grounds for positing an ambiguity the verb know in (86). and vice versa: (84) Lemmings are protected by law. I know actresses – I’ve got a couple coming for dinner tonight. Further examples make it clear that. I know actresses – and I can tell you there’s no more neurotic gr of people. my garden. but Mick goes ahead and traps th anyway. 190 . Consider following: (86) a. d. and one might argue that there is some property–state am guity in the verb in (87). Carlson’s descriptive claims are not unproblematic. and thos (86b) and (87b) to be non-generic – though the predicate is the same. this sentence would be non-generic. It appears that the position of a bare nominal in r tion to the discourse structure of a sentence can play a role in its interpreta as generic or non-generic. the fi is likely to occur in a discussion about lions. Carlson shows that a generic instance can serve antecedent for a pronoun understood non-generically.

one stage of the kind denoted by would be a particular group of cats at a given point or period in time. . bare plurals. where dicerent senses of the word bank fail to coordin acceptably: (90) *Banks are good places to keep your money and to picnic.3. There then stage-level predicates and individual-level predicates – the latter selecting whole kinds because these are for Carlson second-order individuals.sarily generic. compare (90). The genera of the locative expression seems to play a central role. This should be impossible if the generic and non-generic uses were semantically tinct. (92) a. as seen in 4. which denote kinds. unanalysable who He gives detailed arguments to show that bare plurals pattern very much like phr such as this kind of animal. Harriet caught this kind of animal yesterday. This kind of realization or instantiation is termed a “stage” Carlson. which also can have both generic and “indefinite plu readings: (91) a. Carlson supports this treatmen pointing to similarities in behaviour between bare plurals and proper nouns. This kind of animal is likely to win the race. Despite these diaculties. like indefinite sin lars. (88d) is most naturally understood as non-generic. Carlson’s proposal is that bare plurals are pro names of sets or “kinds”. See also Burton-Rob (1976) for similar observations. So one stage of Fred migh a particular three-hour stretch of his life. And s proper names like Fred denote individuals. also proper names – proper names of kinds. Stages are instantiations not only of kinds (denoted by bare common nou but also of individuals (denoted by proper nouns). b. The non-generic indefinite plural reading arises when the predicate is one that sele not the set or kind as a timeless whole. c. Carlson’s claim that bare plurals have a cons semantic representation is supported by the possibility of coordination betw generic and non-generic instances (reinforcing his argument from anaphora): (89) Hedgehogs are shy creatures but often visit my garden. these being abstract individuals. This kind of animal is in the room. since their distribution is dicerent from that of indefinite gular generics. but a realization of the kind at a part lar time and place. This kind of animal is a vertebrate.1. b. is problematic. To treat bare plural generics as non-referential indefinites. This kind of animal is tall.

but only with the indefi plural reading. which suggests they represent generalizations over members of a class rather than over the class as an individual. one would expect that bare plurals could be used gen cally.general than definites. But they cannot. (1995) ). as in (91). 192 J’adore le vin. This nationality is musical.’ Le vin est délicieux. interesting contrast with English bare plurals is that the linguistic context d not select the interpretation of a non-singular definite noun phrase in the sa way. It can probably be analysed as a special c of the non-generic use of the same form. c. collaborative. ‘(The) wine is delicious. ‘I love (the) oysters. But both readings tend to be available in non-ev contexts: (94) a. thus in so languages the context within which inclusiveness applies may be the universe les livres (the books) may refer to the totality of books in some pragmatically de mined domain or to the totality of books without this pragmatic limitation. my impression is that kind-expressions on their generic ra than indefinite plural reading. though a few of them are tone-deaf. b. If common nouns are names of ki in these languages too. ?Italians are musical. has a lot of explanatory power (though Carlson seems to backtrack on claim in more recent. though a few of them are tone-deaf A further problem is that the analysis does not extend straightforwardly to ot languages. work. analysis of bare nominals as names of kinds. There are many languages (such as French and Spanish) in which pro nouns are typically bare but generics are not. Despite these problems. ‘I love (the) wine. though a few of them are tone-deaf. b. thus with stage-level predicates. The Italians are musical.’ . the central generic in these languages being (for count nou the definite plural. an will illustrate it here from French. see KriAa et al. J’ai pe les oeufs ‘I’ve lost the eggs’.d) (73) above. The non-singular definite is the central generic form in many languages. Moreover. ‘(The) oysters are delicious. It is true that an event context sometimes imposes a non-generic read as on the object noun phrases in J’ai bu le vin ‘I’ve drunk the wine’.’ Les huîtres sont délicieux. The obvious account would be to re the generic interpretation to the inclusiveness element in definiteness. d. in both their generic and non-gen uses. see (72c. are closer to the definite plural gen than to the bare plural: (93) a. In Spanish bare plurals do occur.’ J’adore les huîtres. c.

On ano view. in a particular scenario or version of reality) lives next door to me or m dered Smith. b. A non-generic interpretation is possible however. denoting individuals as opposed to classes. The Bolivians have a subtle sense of to use a demonstrative to convey the non-generic sense (ce vin ‘this wine’. but definites in (96) admit both generic and non-generic interpretations. Bolivians have a subtle sense of humour. on Carlson’s the due to the fact that they name the kind. The n availability of the non-generic reading with the indefinites is. Mary admires Russians. as “rigid desig tors”. And the same is true even more obviously of individual comm nouns. following Kripke (1972). No langu appears to permit both definite and indefinite non-singular generics freely – interesting contrast with the singular. In logic. Mary admires the Russians. The n generic reading is only possible when induced by a stage-level predicate. They are supposed to designate individual entity directly. T full noun phrase descriptions like my next-door neighbour or Smith’s murd may apply to one individual or another. and are thus generic in principle. No the contrast with the corresponding English bare non-singulars. expressions which denote the same individual in all possible worlds. the logici preferred label) are usually treated. which can o be understood generically: I love wine/oysters. Wine/Oysters is/are delicious. The indefinites in (95) can probably only be understood as generic.4 Generics and proper nouns Proper nouns have traditionally been viewed as almost the e opposite of generics. depending on who in a particular w (that is. N singular definites in English pattern in this respect in the same way as in Fren (95) a. rather than via a description which that entity satis (thus being a member of the class of entities satisfying it). which wo be required in (95) (probably necessarily in unreduced form) if a non-generic in pretation were intended. a given person may or may not be accurately described by the noun stud . proper nouns (or proper names. While English has non-singular definite generics for only a limited class of no French does not seem to have non-singular indefinite generics at all. huîtres ‘these oysters’). They therefore h reference but not sense. (96) a. and as be fundamentally dicerent from common nouns. 4.3. b. this constraint on the interpretation of the bare or indefinite non-singu would be linked to the availability of the cardinality quasi-article sm.

(98) Bill wants to meet Mary. never a plurality or a mass. and I outline only the main points h The familiar count–mass distinction is central to the argument. There is another logical tradition. In their generic use. they do denote ensembles (or better in this case. proper no are a kind of generic. with the lack of such ambiguity in (9 (97) a. Compare the ambiguities in (97). For one thing. never pluralitie masses. it abolishes the prop common distinction. The following traditional schema would. but this is rejected by most logicians to This Russellian treatment parallels a linguistic view of proper nouns as a pe liar kind of definite noun or noun phrase – peculiar in that in most languages t show no encoding of [+ Def ]. it is generally assumed that the count–mass distinct applies to common nouns only. that is. As pointed above. because these “ensembles” consist of only one entity. and they can be thought of as directly denoting. going back to Russell (1905). The contrast between proper and common nouns is perhaps no longer quite clear given the proposals of Carlson (1977) for treating bare plurals as proper nam of classes (“kinds”). in a generic the noun does describe a class or ensembl has descriptive semantic content. b. for this reason. see Cann (1993: chapter 9). Rather. the fore. may be understood as specific or non-specific in opaque context created by the verb want. this is supposedly not the case with proper nou I want here to advance a proposal which I believe has the merit of integrat proper nouns into the general system of nouns. rigidly designating. always generically. these being second-order individuals. as can be the c with the latter. following suggestions in the philosophical literature Kripke (1972) and others. of seeing pro names as disguised definite descriptions. This account inverts the claim of Carlson that generics names of ensembles. 194 . kinds). be generally accepted as a categorization of nouns: 11 For an outline of possible-world theory. Bill wants to meet a bishop.consequence of this is that proper nouns are said not to be subject to sc ambiguities. proper nouns only have individual entities as referents. A further distinction between proper nouns common nouns is that the referent of the former is always an individual (tho it may be a collective individual). bare plurals do resemble pro nouns in a number of ways. where common noun phra definite and indefinite. Bill wants to meet the winner. an entire class in the same way as proper nouns do an individ But there still appear to be dicerences between names of classes and names of in viduals. This anal is presented in more detail in Lyons (1995c).

The first is that proper no are not distinguished from common nouns by having reference but no sense. they do denote an ensemble – a class of car in and a substance in (100). It is reasonable to say that typical names of persons or places (like J and Paris) have no sense. (100) I bought a 10-kilo box of Persil to deal with all these dirty clot I’ve never used so much Persil for one wash. t are in principle no more devoid of sense (descriptive content) than are comm nouns. they l sense. or they can (as with detergent brand name in (100) ) be uncountable and denote a mass or substa Their syntactic behaviour is exactly that of common nouns. so that a more accurate categorization would show a three-way cou mass–proper contrast. These names have no more descriptive content than typical proper names.common count proper mass My claim is that the category of proper noun is of the same order as co and mass. But they take articles and other de miners (unlike normal proper nouns in English). while denoting an ensemble. as follows: noun count mass proper There are two main reasons for making this claim. directly. But. might claim that they are proper nouns. count in (99) and m in (100). they can (in the case of the model names in (99) ) be singular or plural (without it being a matter of the k of recategorization occurring in three beers or sm apple). then G and Persil here are common nouns. commercial brand names or model names a case in point: (99) Peter has sold his Astra and now drives a Golf. He’s owned th Golfs before. If proper nouns denote individuals. And semantically. But the same is true of names given by a deliberate of christening to kinds of things. The converse of this is that there are true proper nouns – bare one-w forms in English designating individuals – which do have sense: .

And there are no with descriptive content and nouns without it in all three categories. most modern parliaments).b. Their extension is simply trea as being reduced to one object. But most nouns are pro bly typically used as one or the other. But they h the same sense as when used as common nouns. as a result of familiar process of recategorization. we assume such a three-way distinction in nouns (at least in princ – not all languages necessarily show all three types). or to a subpart (non-gene I’ve bought sm butter). But here. and the concept of recategorization is a u ful one for expository purposes. This measure must be put before Parliament. I have been using this term to label what is happen when a noun which is normally mass is used as count and vice versa. A mass noun denotes a kind with subparts. thus mas count in three beers. But notice that this process seems to underlie uses in (101). a good wine. count and mass are categories on the same le If. and. A count noun denotes a kind with subparts (non-atomic instantiations) individual members. sm ap In fact it may well be the case that nouns are not lexically marked as count or m and that any noun can in principle be used either way. then. On the account ocered here dicerence lies in the possibility or not of cardinality modification. senate of your university. in terms of card ity. But th quite dicerent from the traditional assumption of a radical proper–common dichotomy. or as pro This suggests strongly that proper. c. but no atomic memb we can refer to the whole (generic: I love butter). they appear as proper nouns. The second reason follows from these observations and is to do with the p cess of recategorization. So common count nouns can be recategorized as mass. mass nouns dicer as much from count nouns as from proper nouns. A proper noun denotes a kind with neither subparts members. and count to mass in a piece of lamb. 196 . normally common count nouns being turned into proper nouns (assu ing the common use of these nouns is the basic one. or to a member (singular dog bit me). the argument would still wo with some modification.12 12 It will be noticed that on this picture count and mass nouns do have something in com distinguishing them from proper nouns: the possibility of reference to a subpart. we can define the three egories on the basis of the type of ensemble or kind constituting a noun’s ext sion. proper nouns regarded as marginal to the nominal system. These nouns can be used as common count nouns (a high-ceilinged hall. if the recategorization were taken to be in the oppo direction). being made up of one entity. we can refer to the whole ensemble (generic use: I love do to a subpart (plural non-generic: I like those dogs). reference is always to the whole an therefore always generic (I admire John/Paris/Parliament). denoting an in vidual entity (such as an institution) and requiring no determiner. Senate has decided to abandon the scheme.

b. b.2. is cor It is not self-evident that arguments need a Det. as said in 3. there are also languages in which this correspondence does not hold. Count and m generics and proper nouns are definite in these languages. Is there a/?the dictionary in the house? Is there a lion in Africa? Are there dolphins on the verge of extinction? (102) and (103) show indefinite generics patterning like non-generic defini in (104) the (b) and (c) examples are good but the indefinites cannot have a gen 13 Longobardi claims that proper nouns (and generics) in English too move to the Det posi but at the abstract level of “logical form” (the level of analysis at which many relations o essentially semantic nature are explicated).13 I assume the definite article or equivalent (noun in Det position) characterizing proper no and “common” generics in Italian.other generics. Surely a letter is mine if it’s addressed to me. les chiens ‘dogs’) but pro nouns typically bare. (103) a. whether [+ Def ] or [− Def ]. and this can only be null if indefinite. And this expectation is borne out for many languages. c. includ English. where the typical generic is the bare noun or nominal (plural if cou and where proper nouns too are bare. I return to this in Chapter 8. and indefinite (t cally) in languages such as English. like Greek and Cata both “common” generics and some proper nouns take definiteness marking. Catalan and Greek is the full definite art and not some empty pleonastic article restricted to these uses. In other languages. I take view. b. c. The/?A house is mine. This is because he assumes that all argumental n phrases must have a Det. we’ll find room for Nellie here. But in at least some such languages there is evidence for proper noun itself to be in Det position. The apparent definiteness of all proper nouns follows from the fact that gene generally. we saw this for Italian in 3. But both definite and indefinite proper no are generic – except that. itself taking the part of the definite a cle. Big as an elephant is. (104) a. Big as elephants are. following Longobardi (1994). c. that a unified treatment of bare nouns. he couldn’t lift it. this would apply to non-generic bare nouns.1. This can be shown by reference to some of the diagnostics u earlier. (102) a. with Carlson. ‘Vengeance is mine’ said the president. A universal principle ass an indefinite interpretation to any null Det. you’ll never get one to lift that load. In Fre generics are typically definite in form (le vin ‘wine’. . Big as the/*a boy was. generic and non-generic. show much of the behaviour and dist ution of definites. it could be that definite proper noun languages like Greek are merely a special case of uniques like the sun. w are thus seen (in contrast with Carlson’s view) as quite distinct from bare generics.

since their reference is to a whole ensemble. but it does not follow that languages must represent them gramm ically as definite. because they apparently meet both basic semantic/pragmatic criteria for definite reference (inclusiveness and ide fiability). the ensemble denoted by generic p cils is familiar to us all. They can perhaps also be said to be familiar. They behave much like definites.due to the fact that. but (strikingly in view of the sema and pragmatic research findings) never to definiteness. For while a hearer may fail to identify referent of a noun phrase like a/the pencil. can be definite or indefi in form. Noun phrases used gen ically never have a specifically generic form. I h suggested a primary division of definiteness between what I term textual–sit tional ostension on the one hand and all other uses on the other. and theref to meet the criterion of identifiability.4 Concluding comments What emerges from this chapter is that there is no evidence for am guity in the definite article relating either to the various uses of definites dis guished in Chapter 1 or to specificity. 198 . they can be ch acterized as inclusive. Both “ordinary” generics and pro nouns. There is hyponymy. There is also no reason to posit special generic articles. The dicerent types of specificity seem to be treated by some guages as hyponymous to indefiniteness. Generics are thus semantically definite. Languages m represent these finer distinctions within definiteness morphologically or lexica but need not. and even noun phrase types claim to be kind-referring also have non-generic uses. however. though not n essarily grammatically definite. which I have suggested are a kind of generic. 4. The forme these itself divides into anaphoric and immediate situation uses.

languages in which accusative c or a simple unmarked form. if ever.1 that some languages have adpositional object mar which are restricted to occurring with definite noun phrases. third. somew arbitrarily. or an ob marker. An ex ple is object marking.5 Interaction with other grammatical phenomena A number of grammatical processes appear to refer to the fea [+ Def ]. in this connection we will consider the relevance of defin ness for the theory of empty categories. definiteness is only marked in object position. Apart from restriction this kind. An alternative way to view such phenomena is definiteness is only encoded in certain grammatical contexts. is restricted to definite objects. some grammatical categories interact with definiteness in that they encoded together on the same formative – number and gender on the French defi article for example. is restricted to indefinite objects. strictly or only definiteness tha the decisive factor in their appearance. or in a c form not distinct from that of subjects. indefinite objects appearing in a dicerent (but non-subject) case. for some guages. In fact the phenomenon is not lim to such adpositional markers. There is good reason to believe the latter is the correct con sion. other objects being unmarked. . On the face o it is debatable whether these adpositions are “articles” (encoding definiteness object markers. as we have briefly observed. second. languages in which accusative case. thus. in part because it is not always.4. definites taking so oblique case or a marker basically associated with some other function. languages in which the accusative restricted to definite objects. There are also certain grammatical structures in which a defi element may be inherent. but extends to what appear to be accusative c morphemes. in the sense that they only apply when this feature is present. In this section we will examine a number of instances. which.1 Direct object marking We saw in 2. 5. grouped. only takes place in so languages for definite objects. under three headings: first.

un Finnish. It is clear that the case variat does not directly express variation for [± Def ]. and that this correlates partly with definite–indefinite distinction. the accusative can be used where the object is indefinite. But the choice of case can also indicate co pleted versus incompleted action. if the patient o transitive verb is totally involved or acected it takes the absolutive case. Indeed. What this means is that in Hungarian. the partitive if it is only partially involved (1) Ette a süteményt. even when this is defi as here. not indefin ness. a matter we will return to. Hungarian is also interesting in having a definite article.5. It can be seen that the parti construction is used in Hungarian where the domain of partitivity is definite. partitive case expresses genuine partitivity. ate-3SG the pastry-PART ‘He/She ate some of the pastry. 200 7zm q˚ppq7zr yeràqe.’ (2) Evett a süteményb6l. bu only partially involved it takes the locative: (3) a. The distinction expressed by this case variation is u ally treated as one between total and partial acectedness.’ The variation in the verb form between ette and evett is due to agreement of ve with definite direct objects. This means that it is possible to see the definiteness or indefinitenes the object noun phrase independently of its case. ate-3SG-OBJ the pastry-ACC ‘He/She ate the pastry. or even alter the time reference or express emp sis. the referent of the ob noun phrase is completely acected (accusative) or partially acected (partitive) the action expressed by the verb. This fact suggests that such partitive noun phrases are not actually di objects. Moravc also cites Kabardian (data from Knobloch 1952). Similar facts are found in Hungarian (Moravcsik 1978).We have seen in 2.’ .5.4 how in Finnish direct objects can appea the accusative or in the partitive case. an ergative language showin state of acairs very similar to that of Hungarian and Finnish.4). even m clearly than in Finnish. Finn would use the contrast between elative and partitive to distinguish definite indefinite domains (see 2. It is interesting beca agreement does not take place with a partitive object. perhaps they are rather partitive complements of a null indefinite head object position. dog-ERG bone-ABS chews ‘The dog chews up the bone. the accusative be used if the object is fully involved. see Whitney (1956) and Chesterman (1991).

and in some c also for definites.’ Ona nikogda nje vidjela morja. And no that this locative object. This may be support for the suggestion above partitive objects in Finnish and Hungarian are not true direct objects. position. the lo tive case patient. like the partitive object in Finnish and Hungarian. the pattern of accusative case alternating with an oblique cas widespread. though definiten does play a part in what is happening. This use of an oblique case for (mai indefinites within the scope of negation occurs in many languages where this c is not otherwise used for indefinite or partially acected arguments. If this is so. suggesting that construction here is an intransitive one (since the general pattern of ergative guages is that ergative is the case of subjects of transitives. a growing tendency in Russian for genitive case in nega contexts to be restricted to indefinites. not the ergative. oblique. The following examples illustrate process in Russian: (4) a.’ Notice that ‘dog’ in (3b) is in the absolutive. I not want-1SG bread-GEN ‘I don’t want (any) bread. she never not saw-F-SG sea-GEN ‘She has never seen the sea.’ There is. both this variation. n not be indefinite. Languages making use of a partitive or other oblique case to express indefin ness or partial acectedness often use this case for indefinites. q˚\p|&Zm. One such guage is Gothic (Wright 1910): (5) jah ni was im barn2 and not was them-DAT child-PL-GEN ‘and they had no children’ In summary. however. in the scope of negation. and absolutive case of objects of transitives and subjects of intransitives). . then it canno correlated directly with the definite–indefinite distinction. is not grammatically a direct bone-LOC chews ‘The dog is chewing (on) the bone. but is rathe some adjunct. b. The temptation to view it as a unified phenomenon is very stro despite variation in precisely what triggers the appearance of the oblique case. and the complexity of the trigger within individual langua make it clear that if we are dealing with a single phenomenon. Ja nje khochu khljeba.

c. realized by agglutinative suaxes. or some other marker of direct objects. b.’ Biz-i selâmladı. Turkish illustrates this phenomenon in terms of case morphology. (6b) shows an indefinite with the cardinal article. is used only w definite objects (or. is. realized by -i (and its vowel harmony variants) is limited to defi direct objects.’ Bu gazete-yi okumadım.’ Bir mavi kumai istiyor.’ Mavi kumai-ı seçti. but Turkish is archetyp 1 Nominative is in this sense the unmarked case in a great many. a blue material want-3SG ‘She wants a blue material. blue material ACC chose-3SG ‘She chose the blue material.’ Sentence (6a) illustrates the “incorporation” construction: the use of a n unmarked for number. again. with no commitment as to the cardinality of wha denoted. ticket sell-3PL ‘They are selling tickets. they are not distinguished formally from subjects. though this language shows major dic ences in overall pattern from Turkish.’ Hasan’-ı hemen tanıdım. this newspaper ACC read-NEG-1SG ‘I have not read this newspaper. The following examples are from Lewis (1967): (6) a. Both are SOV. languages 202 . distinguished by absence of a sua It is in fact sometimes termed “absolute” or “absolutive” rather than “nominati in traditional grammars. phenomenon is that of langua in which accusative case. Bilet satıyorlar. because it is the form used also for indefinite direct obje The accusative. d. a subset of objects characterized in some other w Direct objects which lack the determining feature are in an unmarked form. Hasan ACC immediately recognized-1SG ‘I recognized Hasan immediately. Both these types indefinite normally occur as direct objects without the accusative suax. us ACC greeted-3SG ‘He greeted us. though possibly related. personal pronouns. though not all. The situation is similar in Persian. this suax is used with direct object noun phrases interpreted as s ple definite. The case form used subjects is the unmarked one morphologically. (7) a. As il trated in (7). and proper nouns. Turkish a system of six cases. as well as demonstratives.A distinct. b.

person ART OBJ sent-3SG ‘He sent someone.’ Kh1ne-3-r1 1tesh zadand.’ Hasan-r1 d3dand.’ (9) a.’ . This is because they do someti appear on indefinite direct objects. b.’ b. one letter write-3SG ‘He writes a letter. c. Kas-3-r1 ferest1d. neutral w respect to number: (10) a. the status of which is unclear. (11) N1me-3 minevise. d. in particular.inflection. Persian has no morphological case system. The following ex ples are from Mace (1971): (8) Man p5l gereftam. man us OBJ saw-3SG ‘The man saw us. house ART OBJ fire struck-3PL ‘They burned a house. I money took-1SG ‘I took (sm) money. letter write-3SG ‘He writes letters/a letter. I money OBJ took-1SG ‘I took the money.’ But Turkish -i and Persian -r! must be analysed as accusative or object mar rather than definite articles restricted to object position (or morphemes encod simultaneously [+ Def ] and accusative case). Hasan OBJ saw-3PL ‘They saw Hasan.’ Kn ket1b-r1 kh1nd. b. that book OBJ read-3SG ‘He/She read that book.’ Mard m1-r1 d3d. But defi direct objects take a suax -r!. letter ART write-3SG Yek n1me minevise.’ N1me minevise. Persian also makes the distinction observed in Turkish between singular indefin and incorporated indefinites consisting of the unmarked noun. Persian examples (Windfuhr 1979) are: (12) a. Man p5l-r1 gereftam.

In many languages accusative or object marking is limited to a subset of di objects characterized by other properties.’ The correlation between appearance of ’et and definiteness of the object is v close. accusative case is restricted to animates in Thargari. notes that ’et can oc with indefinite pronouns. is important. saw-1SG OBJ DEF cat ‘I saw the cat. It wo similarly in Turkish. following Cole (1975).’ (15) Ra’iti ’et ha-xatul. Prep tions in Hebrew inflect for the features of pronominal complements. every day a newspaper ACC read-1SG ‘Every day I read a newspaper. personal pronou proper nouns and kin terms in Gumbainggir. which are then null ( l-i (to-1SG) ‘to me’).2 but Hebrew also has a prefixal defi article. example. Comrie still treats this as definiteness. they con “referential prominence” – a concept to be interpreted pragmatically.’ Here the implication is that the speaker reads one particular newspaper every d Comrie (1978. while not known to the hearer. and indefinites are typically recognized by absence of this: (14) Ra’iti xatul. but a more plausible conclus is that these morphemes do not relate directly to [+ Def ]. and ’et shows precisely the inflectional behaviour of prepositi “Object pronouns” are in fact inflected forms of ’et. wh embraces definites and some specific indefinites. though Moravcsik (1978). and serves to make the specificity of the reference prominent. and one where object marking c relates more strongly with definiteness. Hindi uses an object marker ko w 2 Borer claims ’et is a case marker rather than a preposition. but without substantiation. is Modern Hebrew (Rosen 1962. Borer 19 The direct object marker is a preposition ’et. perhaps beca it is to recur in the discourse – a concept we have encountered in several l guages.optional. 204 . saw-1SG cat ‘I saw a cat. where the accusative suax can co-occur with the quasi-indefi article bir: (13) Her gün bir gazete-yi okuyorum. such as animacy or pronominality. A third language showing the phenomenon. Rather. 1981a: 128–9) argues that the use of the “definite object mark in Persian and “definite accusative case” in Turkish can suggest that the refere of the noun phrase. while leaving room for sub tive choice.

which is otherwise associated with some “obliq function. When separated from it by other material. This case is an interesting one because. they too take accusa morphology: (16) rina barilga bariba. some book table on put ‘Put some book on the table. the one used also for subjects. and nouns modified by a geni or with a possessive suax. an inanimate direct object can also take nJ. proper nouns. Croft 1990).1. typically that of indirect object. subs tivized adjectives. nouns are unmarked for any case distinctions. Cross-linguistically. But again. that blue book to table on put ‘Put that blue book on the table. new building-ACC local people built-3SG ‘The local people built a new building. new building-ABS built-3SG ‘He built a new building.’ 5. ex in clitic pronouns. the complexity of the data makes possible any clear correlation with any one grammatical or semantic categ Spanish does not distinguish nominative and accusative cases morphologically. Croft (1990. human nouns.’ Another frequently mentioned case is that of Spanish. numerals. case form. and form is also used for direct objects which are personal pronouns or animate c mon nouns. together with features re ing to humanness and animacy. which is worth exam ing in some detail. and strong nouns simply lack direct object forms. In Khalkha Mongo (Poppe 1970) accusative case is used with personal pronouns.3 Oblique direct objects In some languages. but only if it is defin (18) Éo nili kit1b nE mezte rakkho.’ (19) Koi kit1b mez te rakkho.(Comrie 1981a. Indirect objects are generally marked . s as the subject in sentences like (17) with atypical word order. or other marker. All other object noun phrases take the unmar absolutive. though definiten does not appear to be involved at all. but only when immedia followed by the governing verb. definiteness is just one of a c plex array of properties which may trigger object marking. direct objects which are distinguished definiteness or other features dicer from others in being marked by a case fo preposition. following Shackle 19 cites Punjabi. Indirect objects are indicated by the postposition nJ ‘to’.’ (17) rina barilgaiig nutagiii araduud bariba. specificity is.

And relationship with humanness is only partial too. look-for-1SG to a policeman ‘I’m looking for a policeman (a particular one). even when the object is n specific: (23) ¿Conoces a alguien en esta ciudad? know-2SG to someone in this city ‘Do you know anyone in this city?’ (24) No veo a nadie. A is often used with animate non-human nou 206 . with human reference.’ So the correlation between appearance of a and specificity is only partial. The basic rules for the occurrence of the perso a are as follows. have-3PL two children ‘They have two children. or na ‘no one’. a is not used: (21) Busco a un policía. look-for-1SG a policeman ‘I’m looking for a policeman (any policeman). but (26) makes it clear that thi not necessarily the case. (25) Tienen dos hijos.’ Also. There is a great deal of flexi ity in the use of the personal a. the speaker’s attitude to the referent and styli factors playing a complex role.’ (26) Tengo una hermana muy lista. ‘anyone’.’ Where the direct object is non-specific. have-1SG a sister very clever ‘I have a very clever sister.’ But when the direct object is a pronoun like alguien ‘someone’.’ (22) Busco un policía. it may be that this verb most co monly takes a non-specific object. as in (25). not see-1SG to no one ‘I don’t see anyone.ditionally termed “personal a”). a is never used after tener ‘to have’. for example after opacity-creating ve like buscar ‘to look for’. a must be used. It must be used with direct objects referring to specific hum beings: (20) Conozco a ese chico. know-1SG to that lad ‘I know that lad.

2. and it is interesting to look at verb–subject agreem and agreement with indirect objects. in the light of what we see in agreem with direct objects. number . so the clitic corresponding to a direct object is sometimes dative. and o in the masculine singular.2 Verb agreement The definiteness of objects may also be indicated by marking on verb. the dative clitic is used only for human referents. that of verb showing agreement with objects which are specified for properties only of which is definiteness. But this agreement indic only that there is such an object – there is no encoding of its person. For a fuller discussion see Lyons 19 5.saw-1SG to your dog in the car ‘I saw your dog in the car. There is in fact much variation in this matter. T can be interpreted as meaning that verbs agree for definiteness with direct obje or that they agree only with definite direct objects.’ It can also be used with any direct object. B oddly. to avoid am guity. 5. the accusa clitic is used. someti accusative (though the two case forms are only distinct in the third person).1 Direct object agreement Transitive verbs in Hungarian show distinct “indefinite” “definite” conjugations. many speakers dative forms beyond this narrow limit. even an inanimate one. This too is part of a wider phenomenon. for the masculine plural and feminine. Intransitive verbs have only the indefinite conjugation. this is the phenomenon introduced in 2.’ Pronominal objects in Spanish are represented by clitics attached to the v And just as full noun phrase objects vary between those with and those with a.2 of verb–object agreem restricted to definite objects.4. resulting for example from the relatively free word order of Spanish: (28) El artículo precede al nombre. the latter being used when there is a definite direct obj the former otherwise. In the standard language (as described for ex ple in Ramsden (1959) ). the use of a with full noun phrases and the use of dative clitics for pron inal objects do not correspond. but also in many dialects the accusa form is used for the masculine singular. defined by the same sort of properties as c acterize those direct objects triggering agreement. the article precedes to-the noun ‘The article precedes the noun. What emerges is that agreement with these arguments to often limited to a subset of them. Of course the direct object is not the only argument w which verbs may agree.

In French Pierre le voit ‘Pierre sees him’. by contrast with French. This “clitic doubling” is. limited to (and ob atory with) definite objects: (32) 208 Marija go poznava nego. full or pronominal. however. This is illustrated in the follow partial paradigm (Bánhidi. and the agreement features corresponding to subject and object are fu together into a single agreement morpheme.’ (30) Olvas-om a könyvet. read 1SGSUBJ a book-ACC ‘I read a book. Let us consider the case of Macedonian (Lunt 1952. read 1SGSUBJ-OBJ the book-ACC ‘I read the book. in which. Marija 3SG-M-ACC knows him ‘Marija knows him. It is in fact often very diacult to distinguish clitics from agr ment aaxes. and the syntactic processes underlying the two are not clearly tinguished. as in French. Lyons 1990). it is a controversial qu tion whether the clitic le is a pronominal element attracted to the verb or an a generated as such.number. Marija 3SG-M-ACC knows ‘Marija knows him. Verb–object agreement is often considered to be less common than ve subject agreement. clitics can occur in associat with an empty object position – they thus arguably represent agreement with null pronominal pro: (31) Marija go poznava. clitics in Macedonian can be associated with an o object. and Szabó 1965) and examples: ‘I read’ ‘you read’ ‘he/she reads’ indefinite defini olvas-ok olvas-ol olvas olvasolvasolvas- (29) Olvas-ok egy könyvet. Jókay.’ The definite conjugation is also used with a null pronominal object: Olvassa ‘He/ reads it’. But it too may be very widespread if object clitics are ta into account. de Bray 19 Berent 1980.’ But.’ .

3SG-M saw-1SG to your son ‘I saw your son. Spanish has data superficially similar to that of Macedonian. strong personal pronoun objects: (37) La quiero a ella. 3SG-M-ACC saw-1SG Grozdan ‘I saw Grozdan.’ b. both overt and null – and thus again with a subse definites (and in fact is only obligatory with non-finite verb forms in periphra constructions): (35) Gwnaeth Emrys werthu y llyfr. 3SG-F love-1SG to her ‘I love her.’ .’ Go vidov Grozdana. however. Another language showing verb–ob agreement in the form of a clitic is Welsh (Sadler 1988). In the standard language. On this analysis Macedonian closely resembles Hungarian in having v agreement with definite direct objects only (though with much fuller encodin agreement features). principally South American. Gwnaeth Emrys ei werthu ef. and obligatory with. Here agreement is o with personal pronouns.catch-IMP 3SG-F-ACC cat-DEF ‘Catch the cat. did Emrys 3SG-M sell it Gwnaeth Emrys ei werthu.’ (34) The clitic encodes agreement in person. dialects. varying in extent from one dialect to anot but appearing in some degree in all dialects. did Emrys 3SG-M sell ‘Emrys sold it. that pro is definite (see 5. case and. so object agreement in Fre is with a subset of definite noun phrases.5 below).’ In some other. doubling is available with noun-phrase objects: (38) Lo vi a tu hijo. direct ob doubling is limited to.’ (36) a. did Emrys sell the book ‘Emrys sold the book. in the third per gender. It is o described as having clitic doubling. number. In French verb–object agreement only occurs with note.

’ In fact clitic doubling is slightly more restricted than the personal a. This pronoun is itself the object argument. shows a strong correlation between doubl and definiteness (provided proper nouns are treated as definite in Macedoni There is a further complication however. Macedonian. Since a only occ with a subset of direct objects (human. nevertheless not true agreement. 1SG PAST read book ‘I read a book.’ Ni-li-mw-ona 1SG PAST OBJ see ‘I saw one person. while a does. or clarifying the reference the null object argument. where the verb agrees with all hum objects and definite non-human ones: (39) a. 3 Ni-li-mw-ona 1SG PAST OBJ see ‘I saw this person. (40) a. – see above). that Spanish does not have clitic doubling – and t verb agreement with overt objects. Lyons (1990) claims that the clitic in (37) (38) represents agreement with pro. 210 . it may be that some apparent instance object agreement. in other languages this noun phrase m be an adjunct. clitic doubl is restricted to this same subset. the following prepositional phrase a ella tu hijo being an adjunct or modifier. this person mto mmoja. specific etc. This digression is important because many apparent instances of clitic doubl involve a preposition. where the putative agreement morpheme is not especially cli like and where no pre-position is required before the associated noun phrase.3 So for (non-standard) Spanish the same pa meters which determine object marking also determine verb–object agreemen has been argued. since it does not o with pronouns like alguien ‘someone’. This is argued to be the case in Chiwewa and o languages by Bresnan and Mchombo (1987). there are languages showing object agreement limited not just definites. person one Ni-li-ki-soma kitabu. nadie ‘no one’.’ Ni-li-soma kitabu. reinforcing. and the noun phr associated with it is a topic (see 6.1 below). while the Chiwe subject agreement marker represents true agreement. b. 1SG PAST OBJ read book ‘I read the book. Givón (1976) points to Swahili. But the fact remains that there are many clear cases of t doubling. as suggested above for the Spanish clitic doubling constructio Finally.’ yule mtu. who claim that. the object marker is an inc porated pronoun.obligatory presence of the preposition a before the direct object. b. however. One of these.

I 3SG-DAT send-FUT-1SG this document ‘I will send him/her this document. agreement of the v . which is duplicated by an overt PP. 3SG-DAT gave-1SG a book to her ‘I gave her a book. requires the preposition a. which can be taken to encode ag ment with indirect object arguments.’ (46) Le di un libro a una vecina. like the direct obj may be null. however. it may be that this is not true clitic doubling – that agreement is w the null pronominal pro. 3SG-DAT gave-1SG a book to a neighbour ‘I gave a book to a neighbour.’ Again. in addition to the di object clitics discussed above. 3SG-DAT speak-IMPF-1PL of that matter ‘We were speaking to him/her about that matter.’ (43) Macedonian Mu dadov edno penkalo.’ In Spanish a dative clitic may also co-occur with an overt expression of the i rect object. But the fact this “quasi-doubling” is less constrained for indirect objects is interesting. dative clitics. Spanish and Macedonian have. cross-linguistically. thus examples like the following are as acceptable in standard Span as in any variety: (44) Le di un libro al estudiante.’ (45) Le di un libro a ella. But more strikin the fact that indirect object doubling is possible with all noun phrase types in dialects. The indirect object. which.’ (42) Spanish Le hablábamos de ese asunto. probably not unrelated to the fact that. 3SG-DAT gave-1SG a book to-the student ‘I gave a book to the student. as in the following examples: (41) French Je lui enverrai ce document. 3SG-M-DAT give-PAST-1SG one pen ‘I gave him a pen. Indirect object a is limited in what noun phrases it may occur with as direct object (“personal”) a so as a result indirect object doubling is not restricted either.French.

and more likely than direct objects. opened-3PL they the door ‘They opened the door. 5. this is arguably not true clitic doubling. Facts like this may relate to the greater prominence given. And in most of the languages so far mentioned it is not constrained by defin ness. Macedonian has genuine indirect object doubling with strong pronouns (a s set of definites).’ With non-pronominal indirect objects. in that the noun phrase in the PP may be indefinite. as in Spanish. 3SG-M-NOM 3SG-F-DAT gave girl bread *U-si-niga intoombi isiinkwa. Subject agreement in We shows the same limitation as object agreement – to pronominal arguments (ov and null): (50) 212 Agoron (hwy) y drws. a clitic may (and in Macedonian must) also occur: (49) Mu dadov edno penkalo na eden urenik. animacy etc. In We and Irish it is restricted to a subset of definite subjects. 3SG-M-DAT gave-1SG one pen to one pupil ‘I gave a pen to a pupil. again. this duplication of a clitic by a PP adjunct is freer than di object doubling. 3SG-M-DAT gave-1SG him-DAT one pen ‘I gave him a pen. b.’ Agreement is only possible with one object. 3SG-M-NOM 3SG-INAN-ACC gave girl bread ‘He gave the girl the bread.’ As in Spanish.’ . U-yi-nige intoombi isiinkwa.(1976) observes that in Zulu. But in some languages subject agreement is limited. dative agreement ta precedence over accusative agreement: (47) a.3 Subject agreement Verb–subject agreement is more widespread than verb–object ag ment. the preposition na ‘to’ must be used.2. given the preposition. Its limitation to strong pronouns is due to the fact that these the only overt noun phrase type having dative case forms: (48) Mu dadov nemu edno penkalo. as have seen. also as in Spanish. and this must be the indirect ob if one is present. to animate and human arguments. as in other Bantu languages. since indirect objects are much m likely to be human or animate than not.

corresponding to person plural: (52) a. b. and the subject must be overt. Otherwise an “a lytic” (non-agreeing) form must be used.’ Níomar ár mbróga. there is only one synthetic form.4. washed I my shoes ‘I washed my shoes.1 and 5.3 Definiteness and animacy Facts like those discussed in 5. The constraint may refer to a subset of definites (a dice subset in each case).opened the men the door ‘The men opened the door. Com 1981a.’ Nigh sí a bróga. washed the girls their shoes ‘The girls washed their shoes. c. agreement is thus only with pro (like object ag ment in French). for example. The latter solution would fail to express the unity of the phenomenon it is one phenomenon. Croft 1990): .’ Irish has almost complete complementary distribution between agreeing (“synthet verb forms and overt subjects. or simply to pronominality (Welsh) and non-overtn (Irish). 5. The principal positions on this hierarchy it commonly appears in the literature are as follows (Silverstein 1976. washed-1PL our shoes ‘We washed our shoes. (53) Nigh mé mo bhróga.’ In these languages it is not clear whether definiteness plays a part in the l tation of agreement. The situation is complicated by the defectiveness of the Irish v as observed in 3.2. In each tense paradigm. and only with these forms is pro possible.’ Nigh na cailíní a mbróga. only some verb forms are synth (showing agreement). In past tense.2 have led to the formula of the so-called “animacy hierarchy” – a name which is somewhat mislead The idea is that dicerent kinds of noun phrase can be ranked in a hierarchy s that those in a higher place are more favoured by certain grammatical proce (like agreement and case marking). washed she her shoes ‘She washed her shoes.

by agreement. the actual cut-oc point varies from one language to another Guaraní only first. but not on non-human nouns. animate and non-anim etc. However. on the verb. Phenomena like object marking and agreement suggest that definiteness ne to be taken into account too. th person pronouns may have too. human and definite. or it needs to be combined w it to form a more general hierarchy. Mandarin Chinese only pronouns (and a few human nouns). of objects high in a macy or definiteness. in Tiwi pronouns human nouns. Hence the morphological mark either on the noun phrase or.and second-person pronouns as the link between least some of them. since they are pro-nominal. Comrie and Croft say that the animacy hierarchy is actually a complex c tering of distinct parameters: person. w first. Deviations from this natu pattern are then marked morphologically.and second-person pronouns nearly always have human referents. in Chukchee all n phrases show number in the absolutive case. Mary and sister are no more human than woman. and third-person pronouns are not in principle more human/animate/definite than simple definite full noun phrases. and in Kharia pronouns and animate nouns. proper nouns.Third-person pronouns Proper names Common nouns with human reference Non-human animate nouns Inanimate nouns Apart from the processes already discussed. In m languages only noun phrases which are relatively high in “animacy” show a nu ber distinction. since it often combines with animacy. animacy proper. number marking ocers a good il tration of the significance of the hierarchy (Comrie 1981a. Similarly. and pr ably definiteness. those between human and non-human. with first. in terms of a natural tendency for the subject or agent to be m animate and more definite than the object or patient. but in other cases it appears obl torily only on pronouns. noun phrase type. Either it complements the animacy hierarchy. It is in 214 . and some kinship terms.and second-person pronouns show number marking. and optionally other human nouns. The operation of the hierarchy in some proces (most notably object marking and verb–object agreement) has been accounted functionally. It appears the distinction between pronouns and non-pronominal noun phrases is distinct fr and cuts across. Croft 1990). humann etc. the observation that subject agreement too is som times limited to noun phrases high on the hierarchy poses a problem for this acco The concept of the hierarchy is in fact problematic in a number of respects is not obvious why proper names and kinship terms should rank higher than hum common nouns.

4 Despite the diaculties. more salien human perception than non-human ones. the notion of a hierarchy. Languages will then dicer regards what kinds of noun phrase conventionally count as prominent. in general.2. definite referents are more to the for our minds than indefinites because. + Spec] may not be a problem if [+ Def. the apparent implication of the proposed sub-hierarchy that [+ − Spec] ranks higher than [− Def. 4 The point here is that work in the semantic and pragmatic literature on noun-phrase inte tation seems to show that both definites and indefinites can reveal narrow-scope and referential readings. [± Spec] is arguably an independent para ter cutting across [± Def ]. − Spec] is n actually a category in languages. it is clear that definiteness and specifi play a major role in the phenomenon. But the that there is a considerable amount of flexibility and subjectivity suggests that largely a pragmatic matter and not fully specified in grammars. does seem express some real cross-linguistic generalizations. For more deta discussion of the concept of salience in pragmatic analysis see Lambrecht (19 Whatever may be the correct approach. of entities in the domai discourse. particularly that posed by the heterogeneity of the p erties grouped together. . but apparently n for definites. non-specificity (of one or types) is morphologically encoded for indefinites in many languages. the definite article is invariable. As we saw in 4. and in some cases seems to bear the major burde it in the noun phrase. The question is how to inter these generalizations – because the hierarchy is itself merely a presentation of th not an explanation. we have also seen that there is reason to believe guages work as if [± Spec] were hyponymous to indefiniteness. so that there are non-specific definites as well as n specific indefinites. Given this. In many languages wh show inflectional material (such as for agreement) on modifying words in the n phrase. they are familiar. My inclination is to suggest that what we are dealing wit the subjective prominence or salience. In other languages the definite article d carry inflectional material. especially languages where article inflection contrasts with a relative lack of such morp ogy elsewhere in the phrase. the refer of pronominal noun phrases are more salient because the very fact of their be pronominal means that the speaker takes their referents to be accessible to hearer without even the need for description. The latter state of acairs is worth examining. stated informally.4 Articles and nominal feature marking We observed in Chapter 2 a striking divergence among language regards the morphological behaviour of free-form articles. by definition. On the other hand. in some sense. It can be argued that human referents are. 5.definite : specific indefinite : non-specific indefinite – is not uncontroversial single continuum.

although nouns and adjectives are inflected for number. What this mean 216 . and a few others). But only determiner can encode the full range of cases (with some syncretism). Compare bons livres [bA livr] ‘good boo and bons élèves [bAz elhv] ‘good pupils’. demonstratives. and liaison is limited to certain contexts where the two wo involved are syntactically closely linked. With some other determiners (incl ing the cardinal article and possessives) which are defective for inflection the nominative or nominative–accusative singular (depending on the gender). adjective encodes these case–number combinations in addition. And in neither cas the -s of the noun likely to be directly followed by a vowel-initial word suacie closely bound to the noun phrase to trigger its pronunciation. GEN SG. and even these are not always encoded: SG. only in the second phrase. where the n is vowel-initial. genitive singular and dative plural are the o cases normally marked on the noun. DAT PL. is the -s of the adjective bons pronounced.German and French. the adjective shows o a two-way case–number distinction: -e indicating nominative singular for mas line nouns and nominative–accusative singular for feminine and neuter nouns. all other cases ‘child’ ‘eye’ ‘hand Kind-es Kind Kind-er-n Kind-er Auge-s Auge Auge-n Auge-n Hand Hand Händ Händ And adjectives only show the full range of cases in the absence of a determiner noun phrases of the form determiner–adjective–noun (for certain determiners. -en for all other cases singular and plural. inc ing the definite article. all other cases PL. An il trative pair of paradigms follows (with a masculine noun): ‘the little tree(s)’ ‘my little tree(s)’ NOM SG ACC SG GEN SG DAT SG der kleine Baum den kleinen Baum des kleinen Baumes dem kleinen Baum mein kleiner Baum meinen kleinen Baum meines kleinen Baum meinem kleinen Bau NOM PL ACC PL GEN PL DAT PL die kleinen Bäume die kleinen Bäume der kleinen Bäume den kleinen Bäumen meine kleinen Bäum meine kleinen Bäum meiner kleinen Bäum meinen kleinen Bäum In French. In German. the us orthographic plural aax -s is only pronounced when liaison occurs with a foll ing vowel-initial word.

This idea is taken up by Harris (1977. Observations like these have motivated the suggestion that. or trig some other phonological process which ecectively distinguishes masculine f feminine: masculine grand [Vr=] ‘big’. -e has no ec masculine fier ‘proud’. more accurately. it is natural that it should be the main focus of agreement. w many adjectives its addition causes a preceding “silent” consonant to be pronoun (or. les maisons blanches [le mhzA bl=v] ‘the white houses’. les livres [le livr] ‘the books’. la maison blanche [la mhzA bl=v] white house’. w out the need to claim that articles are semantically empty. number is not in general ecectively indicated on nouns and adjectives. This interpretation is consistent w demonstratives being as rich in agreement as the definite article. both [f jhr]. masculine bl [bl=] ‘white’. for reasons we will come to . in some langua what is usually taken to be a definite article is really a semantically empty be of agreement features. who claims that demonstrative ce is in the process of developing into a new definite article. It is a sim story with gender. and carried to other noun ph constituents by agreement. O determiners consistently indicate the number of the noun phrase: le livre [la l ‘the book’. accompanying all nouns except in certain fi expressions. but this aax is not normally pronounced. Adjectives typically add orthographic -e when agreement with feminine nouns. feminine blanche [bl=v]. The proposal is an interesting one. The article. a category inherent in nouns. mak clear gender distinction between masculine le [la] and feminine la [la]. 1980). attr ing the inflectional features of its dependents. however. I believe both these interpretations of the inflectional importance of the arti in some languages are mistaken. I reject the second because I do not accept the category Det is the head of the noun phrase. This proposal has a long history particularly in rela to French. plural chev [vvo] ). If the article is the h of the phrase. feminine grande [VrAd]. and I will return to discussion o more general implications in Chapter 9. feminine fière.plurals formed otherwise than with -s (cheval [vval] ‘horse’. Foulet (1958) argues that article has become in Modern French (by contrast with earlier stages of the guage) a mere nominality particle. unmar generic construction as well as in simple definites has given the impression its presence is frequently semantically unmotivated. and therefore heads. traditional article le serving now as a carrier of inflectional categories for the n phrase. it blocks deletion of an underlying consonant). since dem stratives too are Dets. But in many other cases. The c trality of determiners in the expression of inflectional categories in German French assumes a particular importance in the context of the DP analysis. where the fact that the definite article is used in the usual. There is a second way of interpreting the facts we are considering.

as in Eng In the case of douleur. but sim definite rather than generic. But the point is that the “abstr use of la douleur is simply its generic use. ces pulls blancs ‘these white pullovers’. and masculine ce is repla by cet. So what are seen as dicerent uses of le. 218 . There is in fact good reason to consider that the centrality of articles in expression of nominal inflectional features has been overstated. I would conclude that the undeniable importance of determiners the realization of inflectional categories is simply an incidental consequence the fact that. les. use. in the singular too gender is neutralized with both and ce pre-vocalically. feminine singular. pendant les grandes chaleurs ‘in hot season’. I take French le to be no less semantically motivated generally than Eng the. example. The three-way distinction between mascu singular. the “concrete” use is just as abstract. la. Second. is due entirely to the fact that the standard generic constr tion in French is definite. 5 Foulet (1958). but it is a prob for the view that le is an inflection-bearer rather than a true article. this is not always so of gender. whereas adjectives a purely optional component. Fo regards this as a “concrete” use of a normally abstract noun. Old French would have used the article when reference was being m to a particular instance of pain. But this does not justify the claim that the exp sion of number and gender is the principal function of the article. the article would only have occurred in simple definite. over and above the sim definite use.French uses its definite article in a much wider range of contexts than does. This is not a problem for the DP in pretation. makes a distinction between nouns used ge ically and abstract nouns – the article appearing with both in Modern French: le beurre ‘but la douleur ‘pain’. are really only one – the generic one. English. and it is in such use that the article in Modern French is tho of as “semantically motivated”. w it is the case that the article (or the determiner more generally) is the princ indicator of the number of a noun phrase. Gender neutralization a takes place with many adjectives pre-vocalically. Starting with Fren it should first be noted that certain other determiners carry as much inflectio material as the definite article does.5 Since definite generics are no more odd than indefi ones. whereas those ad tives that indicate gender (and they are numerous) do so also in the plural: p dant les grands froids ‘in the cold season’. not generic. le and la are both replaced by l’. thus la douleur de mon coeur ‘the pain of my heart’. which would see all Dets as possible heads of DP. expressed by le. ces maisons blanches ‘th white houses’. and general plural. also appe with the demonstrative ce. In the cas beurre. ces (as well as with the cardinal article un. In Old French the article would have been used with neither. cette. since generics are standardly definite in French. with the same pronunciation as feminine cette. Moreover. Artic and demonstratives only distinguish gender in the singular. in common with many writers. but a fair number maintain distinction. u replaced by the partitive des in the plural). it is very rare noun phrases to occur without a determiner of some kind.

they are null noun phrases. 5. which then appear furthe the right. trace. Aoun 1985). that is. is a null anap It occurs. for example. which are moved to initial posi . We have already mentioned pro. with interrogatives. Anaphors and pronominals sho rough complementary distribution. or phonologically null express is central to much work in syntax. the null pronominal assumed to appear in “empty” subject position of null-subject languages. which establishes one kind of referential dep dency: anaphors (principally reflexives and reciprocals like English ourselves each other). and referential exp sions or R-expressions (full noun phrases). voll-er Flaschen ‘full bottles’ (geni plural). the null element assumed to be left behind by NP Movement. Second the absence of a determiner. cannot accept certain inflections. not to another argument posit but to the non-argument specifier of CP (complementizer phrase) position. in that the former must be bound (have t reference determined) by an antecedent in a certain domain.determiner–adjective–noun. as well as other positions. if the determiner is the cardinal article ein. Let us first identify the various types of noun phrase which have been distinguished. The correct descriptive generalization is that the inflectional material app on the leftmost modifying element. and is bound by the mo expression. it is the adjective which shows inflection in the nom tive singular masculine (ein klein-er Baum ‘a little tree’) and in the nomina and accusative singular neuter (both ein klein-es Kind ‘a little child’). it is claimed that there is an empty category corresponding to e of these noun phrase types (Chomsky 1981. personal pronouns). An interesting ques is whether the [± Def ] distinction applies to them. or a possessive. T is the case. whether any of th are inherently [+ Def ] or [− Def ].2. This means that articles have no special status in the realization inflectional features. the adjective shows the full inflectional paradi with the same aaxes as appear on definite determiners: gut-er Stoff ‘good m rial’ (nominative singular masculine). whether it be a determiner or an adject except that ein etc. in passive structures. It is the trace left beh when a noun phrase moves through Wh-Movement. kein ‘ ‘not a’.5 Null and implicit noun phrases The concept of empty category. occupying the vacated position o underlying object which has moved to subject position. three types of noun phrase are distinguished in terms of their behav in relation to the binding theory. R-expressions are bound. and if so. As sketc in 1. typically the cla while the latter may also be bound. Now. pronominals (that is. but not in this domain. Wh-trace. for instance. is a null R-expression.5. or variable. The empty categories most intensively stud correspond to nominal expressions.

It has the interpretation of and used under the same conditions as personal pronouns. is that of implicit arguments. A phenomenon distinct from that of empty categories. so in Fred was expelled from school at the ag seven. but. for example. so pro must be defi too. number etc. The fourth empty category is PRO. 220 . in the higher clause. in Latin amo ‘I love’. the null argument identified by agreement m rial on the head governing it is understood as a definite pronoun. for example. implicit arguments only semantically present (though see Roberts (1987) for the claim that they also syntactically present).. But unident pro is not constrained to be definite. Since. pro may correspond to any overt pronoun represents the full range of pronominal feature values. definite indefinite. these features determining the form of the agr ment morpheme. unlike empty categories. but for our purposes wo considering in conjunction with them. this null element is nothing more nor less than a personal pronoun. as poss sor identified by a personal (possessive) suax on the noun in Turkish köyüm ‘ village’. which is never bound is the empty subject of non-finite clauses. typically. 5.4) that null arguments not identified by agreemen probably not pro. They correspond. and the same is true of verb–object agreement in many languages.5. C (1987) claims that some languages (including Korean) have unidentified pro. The distribution of P is accounted for by the theory of “control”.the binding theory is. t are usually assumed not to have any syntactic representation. In all these occurrences. at least in its earlier versions (though see Aoun 1985 theory only of “A-binding” – binding by an expression in an argument positio this does not count. as complement of an inflected preposition in Irish orainn ‘on us’. While Huang (1984) and Rap (1986) argue that unidentified null arguments in Chinese and Portuguese are variables.1 Pro In languages where rich agreement morphology licenses a pro s ject.6 In most null-subject languages agreement is with all overt subjects. and in pro amant ‘they love’ it is th person plural. for person. Pro can be assumed to occur as direct object in Hungarian Olvasom ‘I r it’. the implicit argument is the agent of Fred’s expulsion. pro is first person singular. though this may be an oversimplification. while empty c gories are abstract elements present in syntactic structures. as antecedent). dis guished only by being phonologically empty. to the understood ag of passives with no by phrase. These are nominal “expressions” lacking overt form. and carries the same f tures. Personal pronouns are inherently [+ Def ]. 6 I am assuming here (as suggested in 3. For discussion of all these empty egories see Haegeman (1994). as in Mary wants [PRO to go hom (where it has Mary.

’ (56) Pro pasaron la mañana en la fábrica. But in (56). since in general they do correspond to any overt form as close to them in meaning as are overt prono to pro. These pronouns are typically the same as personal pronouns. in the positions where traces appea other languages. But examples like (57) can be rescued by the use of a resumptive pronoun: (59) This is the man that I was wondering if he knew you. in agreement. a fact recognized most cle by Aoun’s (1985) “generalized binding theory” which disregards the distinc between argument binding and non-argument binding. not ‘some’. A number of languages make use of “resumptive pronouns” – overt fo appearing. especially in relative clauses. It is triggered by.2 Traces The situation is less obvious with traces.’ (55) Cincuenta hombres pasaron la mañana en la fábrica. therefore definite. most notably in contexts where a trace is ruled out.5. It seems reason that they should be. Being anaphoric in this se traces meet the criterion of familiarity. [+ Def ] is inherent in pro.’ In all three sentences the verb morphology (-ron) encodes a third-person pl subject. This makes it clear that the agreement morph on the verb is neutral with respect to definiteness. they are anaphori the traditional sense that their reference is established from the preceding discou This applies as much to wh-trace as to NP trace. by contrast with the object trace in (58): (57) *This is the man that I was wondering if t knew you. ‘These men spent the morning in the factory. it can only be interprete definite – ‘they’. since they are bound by an antecedent. includ English. ‘Fifty men spent the morning in the factory. where the subject is null. ‘They spent the morning in the factory. definite and indefinite overt subject (54) and (55). (58) This is the man that I was wondering if you knew t. This strategy exists marginally in many languages. But there is evidence that they are inherently definite. . respectively.noun. Consider the following examples of subject–verb agreement in Spanish (54) Estos hombres pasaron la mañana en la fábrica. 5. This is the case w the subject trace in (57).

in wh transitive verbs show prefixal subject marking and suaxal object marking. In many languages in which resumptive pronouns are normal. but o for definite objects: (65) la-falle-i doef-ikso i Baso.’ Topicalization of the objects in (65) and (66) gives the following results: (67) 222 doef-ikso la-alle i Baso money DEF 3 take ART Baso . in Vata (Koopman 198 resumptive pronouns only occur (and are obligatory) in subject position a wh-movement: (63) àlV z l2 sâká lâ who he eat rice WH ‘Who is eating rice?’ (64) y3 Kòfí l2 lâ what Kofi eat WH ‘What is Kofi eating?’ Further evidence is provided by languages in which verbs show object agr ment with definite objects only. they only occur are only obligatory.Resumptive pronouns are less acceptable (though they do occur) in contexts wh a trace is not ruled out: (60) This is the man that I was wondering if you knew him. Basri and Finer (1987) discuss Selayarese. 3 take 3 money DEF ART Baso ‘Baso took the money. the verb also takes what the writers term an “intransitive” pre (66) (a)i-alle-i doef i Baso. not for subjects and direct obje Persian is an example: (61) kit1b-3 ke khar3d3d book ART that bought-2 ‘the book that you bought’ (62) mard-3 ke shom1 b1 u sohbat kard3d man ART that you with him talk did-2 ‘the man that you spoke with’ But in other languages the distribution is dicerent. INTR take 3 money ART Baso ‘Baso took (sm) money. for objects of prepositions. the verb shows no object agreement and sub marking is suaxal.’ When the object is indefinite.

5. but by the context. Mary asked me [PRO to wait for her]. This is begging the question however. Jill expects [PRO to arrive late]. PRO can also be pragmatically controlled. It isn’t easy [PROarb to learn to ski]. c. its reference is fixed not by argument in the matrix clause. . This is the case in: (70) It wasn’t easy [PRO changing the wheel in the dark]. 5.b.3 PRO The empty subject of infinitival and gerundive clauses is defined a “pronominal anaphor”. but there is another use where P is not controlled and has a non-specific or generic indefinite interpretation si lar to that of the pronoun one. because it d not always have an antecedent.4 Implicit arguments Perhaps the most general feature of the passive cross-linguisticall that the agent of the event or action described is not expressed by an argumen subject position. [PROarb cycling at night without lights] is very dangerous. The conclusion is that the gap lef movement of a definite or an indefinite noun phrase is treated as a definite obj though without the person feature to trigger agreement. In this use it is known as “arbitrary PRO”. b. It may be that PRO is definite in these uses. Passive structures vary in regard to whether the agent may 7 In a discussion of these data. It does have in the following examples. but appears not to be inherently definite.5.7 5. or PRO and is limited to non-event contexts: (71) a. b. wher is “controlled” by either the subject or the object of the higher clause (depend on the verb of this clause): (69) a. but the subject agreement marker is prefi and the intransitive marker is suppressed. money INTR take 3 ART Baso doef la-alle i Baso money 3 take ART Baso The verb now shows the same marking regardless of whether the fronted ob is definite or indefinite. Hukari and Levine (1989) ask why the gap should be defi and claim this must be a language-specific fact since not all languages have a feature [+ in their syntax. that is. and this marking is that appropriate to a definite obj There is no object agreement morpheme. Joe always enjoys [PRO playing the piano].

like deliberately. That old tree was cut down last week. Thus (74) has the same interpretation as (73a): (74) The new photocopier is switched on at the side. b. but where such an adjunct is possible it is probably alw optional (Your letter was posted yesterday). similar to the English passive. In the absence of a by phrase. The police bribe easily. The police are frequently bribed. Lyons argues that all these ch acteristics follow from the arbitrary nature of the implied agent. again. (73) a. it is the expression of the agent tha suppressed. c. this indefi has the freedom of indefinites more generally to be either specific or non-spec 224 . c. involving an auxiliary and the past participle. though this has no expression in the senten overt or null. it is understood something like ‘someone/some people. by the presence (generally) of a manner adverbial like easily. The new photocopier switches on at the side. In the passive sentences an agent is implied but not identified. Consider the following examples of the English passive. In principle. The middle is characterized by an aspect constraint. th is still understood to be an agent. other involving se and its variants. this is the implicit argument. and it is und stood as predicating a property of the subject. The English pas and the Romance participial passive can also have the agent interpreted as a trary. In a study of constructions involving the clitic se in the Romance languag Lyons (1995a) argues that these languages have two passive constructions. and by impossibility of adverbs which imply an agent. it doesn’t matter who/I don’t know wh That is. Ann’s new novel is widely read. middle. if the aspectual context is appropriate. thus Spanish El hierro fue descubierto h tres mil años and El hierro se descubrió hace tres mil años ‘Iron was discove three thousand years ago’. in (72). where. Ann’s new novel reads well. it is a (possibly specific) indefinite. The point for the present discussion is that the implicit arguments of pass and middle constructions are interpreted as indefinite. in (73): (72) a.was posted by Fred). and is arg to be simply the passive with the indefinite implicit argument interpreted as a trary. But in the middle sentences the impl agent is more like ‘anyone’ or ‘whoever wishes/tries to do it’ – an arbitrary indefi very close in interpretation to one and PROarb. limiting it to non-ev contexts. and are then subject to the sa constraints as the middle. b. Another construction involving an imp argument is the middle. The Romance middle also involves se.

’ (76) Ma]tic’a ch’illax naj yu tz’ulik ch’en [ach’en. Another typ that of noun classifiers. but dicer in degree of grammaticalization. unstressed when used adnominally. because a common pattern is a classifier to be used in second and subsequent mentions. ix ‘woman’) or reduced forms of existing nouns (naj from winaj ‘man’). This is the case in Japanese: Hon no san-satu ka asita (book OBJ three-CLASS bought) ‘I bought three books’. which accompany nouns and do not depend on the p ence of a numeral. At first sight this appears plausible. and Craig considers the possibility that the Jacaltec ones enc definiteness.distinct from the passive it can be constrained to be non-specific. the most common proba being that of numeral classifiers. h ever. and in Jaca (Craig 1986). They are found. never was-seen he for small CLASS cave ‘He was never seen because the cave was small. a marked alternative is huneh + classifier + noun. in Dyirbal (Dixon 1972). While the typical sing indefinite is huneh + noun.9 C argues. that a referent introduced 8 9 The Jacaltec classifier system is evidently a fairly recent development. There is also a third indefinite structure.’ . with twenty-four classes. being discrete items in c trast with the inflectional agreement categories which frequently encode gen There are several dicerent types of classifier system.6 Classifiers Many languages have classifiers as a component of noun phrases. for example.’ But classifiers can also appear in indefinite noun phrases. so called because they occur with numerals which they may be aaxed. started PL CLASS look-for a IRR pot ‘They started looking for a (non-specific) pot. since it is lackin other languages of the Mayan family and most of the forms are either identical to existing n (e. and the cardinal article takes a suaxal irrealis marker: (i) Xfoc’ heb’ ix say-af hun-uj munlab’al. wh they distinguish four classes and also encode deictic distinctions. used for non-specific indefinites in opaque cont Here no classifier occurs. and (76) represent successive sentences in a single discourse: (75) Scawilal tuf xil naj hunef [ach’en tz’ulik. 5. near there saw he a cave small ‘Near there he saw a cave that was small. the first-mention n phrase taking the cardinal article huneh (also ‘one’) instead of the classifier. They are.g.8 In languages having noun classifiers it is generally possible for nouns to oc without them. on the basis of findings by Ramsay (1985). Th are related in function to gender systems in that they group nouns into sema classes.

This subsuming of definiteness under prominence may l especially disappointing. and definiteness apparently only comes into ve argument agreement and object marking as part of a wider. conc of “prominence”. promised to endow it with considera importance in grammar. 5. like these markers. that the referent is impor and is to be paid particular attention.7 Concluding point It will no doubt have struck the reader that this chapter has a so what negative flavour. better. and I have suggested that generics. but distinct from it – and it is this categ (expressed so far by the feature [+ Def ] ) which is probably not involved in ob marking and agreement. w be presented in Chapter 6. Further evidence for the claim that definiteness i grammatical category.course. We have also s that referential prominence or importance for the discourse may be the main f tor determining the appearance of a direct object marker in some languages. I think this significance will turn out to be greater t generally suspected. may be formally or gra matically indefinite in some languages and yet behave in certain ways definites because they are in some sense semantically (or. and vague. and function as a trigger for certain grammat choices. trace) or indefinite (implicit arguments). The use of this construction indicates. widespread as they are. the fact of the referent of a noun phrase being fami identifiable. More generally. then. and the idea will be developed in Chapter 7 subsequently. but it may play an important role in language use. A distinction is sometimes drawn between semantic and grammatical defin ness. But this finding is of major importance in clarify the nature of definiteness. noun classifiers do not encode definiteness. and will help us identify the real significance of category in grammar. Whi number of non-overt syntactic elements can be claimed to be inherently defi ( pro. This recalls the English indefinite this a number of other similar phenomena we have encountered. But then there is a grammatical category of definiteness – the gramm icalization of the pragmatic concept. articles have no special role carrying nominal inflection. since it relegates definiteness to a subsidiary role in p nomena which. for example. It has shown that definiteness plays a part in a wide ra of grammatical phenomena. but frequently in a rather imprecise manner. though based on a discourse or pragmatic concept. 226 . perhaps as part o larger concept of prominence. pragmatica definite. or inclusive may not have any formal place in the grammar of a gi language.

and with each opposition there is variation o whether the terms are taken to denote linguistic expressions or the referent these expressions. prov the diagnostics for definiteness introduced in Chapter 1.6 Definiteness ecects At various points in the discussion so far I have observed that tain positions or contexts within sentences or utterances require a noun phrase oc ring there to have a particular value of [± Def ] (or to be interpreted as hav such a value). among other labels. Many sentences (or utterances) can be said to consist of two parts. It is concerned with the w in which sentences package the message conveyed so as to express the relat ship between this message and its context or background. 6. “informa structure”. But the variation the use of these pairs of terms is considerable. given–n presupposition–focus figure prominently in this literature. termed “(in)definiteness ecects”. To a large extent they are u interchangeably. For exam . The topic. glossing over m of this variation. though for some writers one opposition closely overlaps with ano rather than being equivalent to it. represents w the sentence is about. Vallduví and Engdahl (1996). Restrictions of this kind. “thematic structure”. The following remarks represent a synthesis.1. They relate. theme–rheme. For discussion Lambrecht (1994). a topic a comment.1 Discourse structure We begin by looking at the place of definiteness in that area of p matic theory which has been variously termed “discourse structure”. The comment is what is said about the topic. moreove the suggestion that definiteness plays a role in guiding the hearer through the org ization of information in discourse. interacting therefore with other concepts distinctions in the structure of communication. typically but not exclusively a noun phrase. The behaviour of definitenes its discourse and sentence context is examined in the present chapter.1 The organization of information The oppositions topic–comment. 6.

The topic expression foregrounds something alre in the consciousness of the participants in order to make it the point of depar for some new information. Indeed the term predication is used in something like this sense in m current syntactic work. Note that givenness is distinct from the familiarity or identifiability which may und definiteness. though the two frequently coincide. the part which is not subject to interrogation is presupposed (here. the comment carries the main informative burden of sentence. is the comment. in its usual modern usage as a grammatical concep not the same as topic.1 For m detailed discussion of the notions of topic and “aboutness”. The topic is the point of departure for the message. since there can given material which is not topical. In s questions. that he climbed out through a window. see Strawson (196 Dik (1978). For this rea the topic is likely to be something the speaker can assume to be in the hear present consciousness. This presupposition–fo 1 2 Note that in this instance the topic is the subject of the sentence.This sentence may be understood as a statement about the individual identi by the description the burglar. and in the older gramm cal and philosophical tradition subject and predicate were commonly defined as I have defi topic and comment. there was a burglar and that he or she did something). then the rest of the sentence is the comm But this comment. the ob noun phrase in I saw your father yesterday almost certainly denotes a person well know the hearer. Another concept overlapping with given and topic is presupposition. Decl tive sentences such as those used so far for illustration in this section can be thou of as answers to implicit wh-questions (such as What did the burglar do?). as opposed to the comment which is new informat In fact topic–comment and given–new do not coincide exactly. having already occurred in the discourse or being par more general knowledge related to the material in the discourse. in which case this is the topic. As Strawson (19 puts it. What is said ab the burglar. Consider the following sentence: (2) The burglar climbed in through the window Jean had forgotten to cl Taking the burglar to be the topic. is topics are generally given. contains a definite noun phrase window Jean had forgotten to close which has perhaps been mentioned befor the discourse and would therefore be given. but if this person is only now being introduced into the discourse then your fa counts as new information. What is not presuppo is in focus. 228 . though again there is considerable overlap. while new information. In the usual case. The topic is therefore often ch acterized as what is given. making statements involves giving or adding information about what matter of standing current interest or concern. the wh-expression represents the focus. Reinhart (1981). As Chafe (1976) points out. but “subject”.2 The important point. however.

Indeed there m well not be an element with topic status. all sentences divide into topic and comment.So the topic. which links (or anch – a concept discussed by Prince (1979) and Löbner (1985) ) the referent to an i vidual present. Particularly in the case of senten beginning a discourse or section of discourse. it is important to observe that the topic need not be given. New in mation can be made topical. the speaker. more generally. Here the subject noun phrases are much less likely to be topics. But this is made possible by its consider descriptive content due to the relative clause. and. of the comment containing given in mation. . though this possibility is rather constrained (and ru out in many languages). provided their referents are suaciently salient to permit recovery of their descriptive cont As well as the possibility. so that new information is stres and given information less prominent. Three strangers arrived this morning. the burglar. on the other hand. the whole sentence being in focus. In fact “focus” is sometimes defined as part of a sentence which carries the main stress. (5) A man has phoned several times. noted above. Consider the following: (3) A man I work with has won the pools. is the focus. The presupposition–focu given–new partition largely determines the prosodic structure of the sente utterance in many languages. including English. increases the current interest of the referent. The subject noun phrase here may well refer to an individual completely new to discourse. climbed out thro a window for example. particularly favour a non-topic interpretation of their subjects: (6) (7) (8) John appeared at the window. If. there need not be a point of de ture expressed. Comp the following: (4) A pensioner has won the pools. This lesser prominence cha terizing topical. topical. given. content also shows up in the fact noun phrases of this kind show a strong tendency to be pronominal. is presupposed and the comment. this declarative tence is thought of as answering an implicit question Where did the burglar cl out?. Certain presentatio verbs. A surprising event then occurred. In such cases the implicit question that the sentence can thought of as an answer to would have minimal presuppositional content (thus so thing like What’s happened today? or Anything to report ?). The relative. then only through a window is in focus. yet it seems to be a topic. which can be used to express the introduction of a new referent into discourse. and the burglar climbed out is presupposed and (for some writers at least).

In some languages. serving either to pl non-subject topics in the prominent sentence-initial position or to give additio prominence to a subject topic (for example because it did not figure in the im diately preceding discourse). Synta processes which have the function of singling out the topic frequently involve front this is the case in English with topicalization (That friend of yours I really d like) and left-dislocation (That friend of yours. The disloca 230 . as we w see below. . and indeed the “topicalization” left-dislocation constructions can be used to give positional prominence to a focus element as well as a topic. is concern One might also mention here passivization. comments therefore to coincide with predicates. But point is that in the absence of some indication to the contrary (such as focal str on Mary). in which they occur finally. in which case that friend of yo would be a topic. But care is needed here. Thus. though it is unlikely to be correct that such languages lac category of subject as some linguists have claimed (Li and Thompson 1976). Second. If the sentence is taken to respond to the question (imp or explicit) Who drove over the flower bed?. topics can also be picked out by right-dislocat structures. for several reasons. It’s far too expensive for me. are “topic prominent” in the se that topic–comment structure is more evident. that friend of yours. the numer subject-second languages show the same tendency for subjects to be topics do subject-initial languages. than subje predicate structure. but linear order is the same. as regards. but if they were part of a discussion of who one liked. This is by no mean necessity. the sentence would normally represent a comment on Mary that drove over the flower bed. I really don’t like him). Some languages. This is very common in English: I rat like him.subjects. and the simple identificat of topic with initial position is untenable. it is common (in English for example) focussed constituents to be placed initially. as wel with topic-introducing frames like as for. If the topic is the poin departure for the message it is natural that it should come first. this would normally be clear from the prosodic pattern of the sentence. superficially at least. It is tempting to relate this subject–topic correlation to the statistical predo inance cross-linguistically of sentence-initial subjects. this car. . Th fronting processes are extremely common in languages. Third. the examples above might express a reaction a person recently mentioned or recently present. In the sentence Mary drove over the flower bed. this common tendency is taken further and subjects are larg required to be topics. The stress pattern would distinguish the two. that fri of yours would be the focus. First. This topic–subject identity as the unmarked situat is extremely widespread and probably universal. then Mary is comment or focus. as far as . the subject Mary co be but need not be topical. indeed. in which an underlying complem advanced to subject position is commonly the topic of the sentence.

that tree leaves big ‘That tree has big leaves. Topical tion almost certainly involves movement to an initial position. can also picked out by a range of syntactic devices which place them in a peripheral p tion. This means that the clause is gr matically complete without the dislocated expression. with initial position acording greater prominence. ?Mary. a dislocated topic is not so construed. rather tha 3 It could be improved by introducing a resumptive possessive. .’ A more-or-less literal rendering.and evidently unmarked nature of this construction for many English speakers its function is not to give the topic the degree of prominence achieved by leftlocation. The correct generalization is that subjects.3 Chafe (1976) c acterizes the function of this kind of topic as being to “limit the applicability the main predication to a certain restricted domain”. in this conn tion see Bresnan and Mchombo (1987). subject or not. making the structure confor the English pattern of left dislocation: That tree.’ ‘The leaves of that tree are big. Mandarin is usually taken as typifying this p nomenon (Li and Thompson 1976): (9) Nèike shù yèzi dà. but that topics. and the relationship between the topic and the clause it prec is by no means always one that can be expressed by a possessive. while disloca (despite this misleading label) does not. In many topic-prominent languages. But the resumptive pronoun ensures that the disloca topic is construed as an argument or other element of the clause. tempo or individual framework within which the main predication holds”. in which th is a particular structural position for the topic (or for the focus. its leaves are big. It is construed pragmatically as being related to the cla indeed as being its point of departure. a dislocated topic is picked up by a pronoun (marginally a pron possessive) in the main body of the clause: That friend of yours. h ever. the leaves are big. which stands outside it topicalization (That friend of yours I don’t like). termed “discourse configurational”. because it is not picked up in the cla by a resumptive pronoun. keeping the dislocation structure. to set “a spatial. the peripheral expression being ge ated outside the clause. See Kiss (1995) for cussion of a type of language. çiftçinin (house-3SG big farmer-GEN) ‘His house is the farmer’s’). whether or not ini show a strong tendency to be topical. or for both). In English. but the clause is semantically as wel grammatically complete without it. I’m impressed by her car. by contrast. But there is no posse in the Chinese example (9). the fronted exp sion is integral to the clause. which would be incomplete without it. I really d like him. The same is true of the similar construction used in colloquial Turk (Lewis 1967): Evi büyük. would hardly acceptable in English: ??That tree.

’ Notice the strong tendency in both Japanese and Korean for the topic to be s tence-initial.topic as well as dislocated ones and those presented in a frame like as for. here LOC TOP pen SUBJ is ‘There is a pen here. B languages also have a subject or nominative particle (Japanese ga.2 Information structure and definiteness Given that topics almost invariably represent given information. that thief TOP I SUBJ catch PAST DECL ‘I caught that thief. it thus marks non-topical subjects o and therefore tends to be associated with focus noun phrases.1. the Eiffel Tower is really spectacular has both a the (Paris) and a topic (the Eiffel Tower). Definiten and givenness are by no means the same. There is probably no need to draw a fun mental distinction between these “Chinese-style” topics and the English type as Ch and Dik do. s sentence like As for Paris. don’t cut bread.’ ‘That thief. pencil TOP there LOC is ‘The pencil is there. the subject in this structure can be a second topic. lim ing the term “topic” to expressions forming part of the main predication. a definite noun phr can represent new information in that its referent has not already figured in 232 . Many languages have morphological marking of topic noun phrases.’ Khal-lo-nun ppang-ul ssel-ci mal-ala.’ ‘With a knife. in wh a particle (wa and (n)un respectively) is postposed to the topic constituent. As already noted. I caught him. 6. knife with TOP bread OBJ cutting stop IMP ‘Don’t cut bread with a knife. Japanese (10) (11) Korean (12) (13) Koko ni wa pen ga arimasu. For subjects are as likely in Chinese as in other languages to be top and since a dislocated topic is additional to a semantically complete subje predicate structure.’ Enpitu wa asoko ni arimasu. Rather. Korean ka) wh cannot co-occur with the topic marker. this is particularly clear in the contrast between (10) and (11). I exe plify from Japanese (Dunn and Yanada 1958) and Korean (Sohn 1994). But an important point wh the Chinese dislocation structure brings out is that sentences may have more t one topic. to be expected that topic noun phrases will frequently be definite. languages vary in the looseness they tolerate in the relati ship between the dislocated topic and the sentence.’ Ku totwuk-un nay-ka cap-ass-ta.

1981. French. Tzotzil has both. the generalization is pr ably that a topic in Japanese is required to be identifiable – thus pointing to a sociation of identifiability (or “semantic/pragmatic definiteness”. as used in and definiteness. do not have definite articles. there is one type indefinite which occurs readily as topic: indefinite generics. to indicate topic–comment st ture). and the resultant tendency for topics to be definite. and these are likely to be familiar to a he even though particular instantiations of them might not be. in languages with definiteness marking. that. (14) and (16) show left-dislocated n subject topics. do not have morphological marking of topics (though t may have other devices. n phrases marked with ga. which ma it possible to say that definiteness serves partly to guide the hearer in working how the information in an utterance is organized. subjects (in the unmarked pre-verbal p tion) are always topics in the absence of left-dislocation of some non-subject n phrase. and generics are commonly grammatically indefinite (that is. as in: Would you care for a scone? – I’d love a scone. overlaps in fu tion with topic marking. The tendency for topic and subject to coincide is equally a requirement in m languages. Japanese and Kor with topic markers. Since wa-marked noun phr can be generic. thanks. indefi in form) in languages that have definiteness marking. It is important to obse however. So the overlap between definite and given is remarkably strong is this overlap. on the other hand. so that a topic noun phrase nearly alw has topic marking and a definite article (Aissen 1992). of course. and the “familiarity” of many situatio or general knowledge definites is often suacient to acord the mental salience nee for givenness. Ne theless. For this reason the two tend not to co-occur in langua a language will usually have one or the other. On the other hand. Thus definiteness marking. can in principle be construed as defi or indefinite. For example. It does not follow. In Japanese a noun ph marked with wa can only be rendered into English as definite or generic. In Mandarin and Cantonese. and topics must be definite or generic in interpretation (Li and Thomp 1976. tho probably only where it is accounted for in terms of identifiability. so many languages require topics to be definite or generic. Ara with definite articles. This is presuma because generics are identifiable and readily represent given informat Generics refer to entire ensembles. or e that a category of definiteness exists in Japanese. the identifiability which characterizes many definites is often a matte occurrence in the preceding discourse. that wa is a definite article. like prosodic structure.and an indefinite can be given in that its descriptive content has occurred in discourse. Matthews and Yip 1994). What is a strong tendency in some languages is obligatory in others. and (15) and (17) sentence-initial subjects: . while English.

normally post-verbal definite or indefinite in interpretation. have one CLASS car block CONT CLASS exit mouth ‘A car is blocking the exit.4.4. cat like drink cow milk ‘Cats like to drink milk.’ b. dog I already see EXP ASP ‘The dog I have already seen.’ (17) a. involving in Chinese the verb ‘have’: (18) Mandarin yvu rén gsi nu dq-diànhuà. Cantonese (16) N3 d3 yéh móuh yàhn s3k ge. a follow pre-verbal subject is still constrained to be definite or generic. this CLASS stuc no person know PRT ‘No one knows this stuc. that CLASS house very expensive ‘That house is very expensive. This may mean this subject is a secondary topic.’ With presentational verbs it is possible to place the subject in post-verbal p tion. CLASS car block CONT CLASS exit mouth ‘The car is blocking the exit.’ Where a left-dislocated non-subject topic occurs in these languages. Ga ch2 jó-jyuh go ch2ut-háu. Nèi-suv fángzi hqo guì. cat much like eat fish PRT ‘Cats like to eat fish. in which case an indefinite interpretation is possible (and more likely) shown in 2.’ ‘There’s a car blocking the exit. can be placed pre-verbally (yielding an S structure).(14) Gvu wv yujing kàn-guo le. direct objects in Mandarin.’ M1au hóu j5ngyi sihk yú ge.’ (15) a.’ (19) Cantonese Yáuh (y1t) ga ch2 jó-jyuh go ch2ut-háu. have person to you hit telephone ‘Someone telephoned you. b. In fact. and then show a strong tendency to be definite or generic: 234 .’ M1o xuhu1n h2 niú-nqi. To achieve an indefinite interpretation of the agent of a n presentational predicate. this predicate has to be subordinated to an existen (‘there is/are’) structure. and that the subject position generally is a to position in Chinese.

so that a subject following a fronted non-subject topic be interpreted as definite or indefinite: (23) lócz dákô òmGò búk. in the absence of left-dislocation or topicalization. with some dicerences (Noo 1992).’ wv bq sh5 mqi le. can be indefin but only if it is accompanied by the indefiniteness marker -mWrD (subject to ass ilation of the initial consonant to a noun stem-final consonant): . but any non-subject may be left-dislocated and con tute the topic. I CONT buy book ASP ‘I am buying the/a book.’ (22) dákô òtèdò rìió. lion king 3SG-think-HAB that 3SG-kill-PRF ‘The lion. woman 3SG-cook-PRF meat ‘The woman cooked (the) meat. The unmarked constituent order is SVO. The loose linking between topics and the clause found in Chin does not occur in Lango. typically a null position (possibly to be analysed as filled b resumptive pro. man woman 3SG-give-PRF book ‘The man was given a book by the/a woman. a subject.’ Moreover. though only when human is a null verbal object identified by ag ment morphology. A dislocated topic must be construed with some p tion in the clause. Again the to is constrained to be understood as definite or generic. and order is rather rigidly fi by contrast with Chinese. I OBJ book buy PRF ‘I bought the book. the subject is the topic. Where there is no dislocated non-subject. A similar state of acairs is found in Lango. and then not necessarily if it is third person). gi that sentences may in principle have more than one topic. it may be that pre-ve objects are also topics. tho not having the freedom to occur post-verbally as in Mandarin.’ (The particle bQ accompanies pre-verbal definite or specific objects which den entities acected by the action of the verb. this definiteness requirement applies only clause-initial subjects. by contrast with Chinese. and clause-initial subjects definite or generic in the absence of specific marking to the contrary: (21) àbwòr rwòt támô nH -nékò.b.) Again.’ But. the king thinks that he killed it. thus with a theme θ-role.

19 In its narrowest sense the label “existential sentence” denotes the there is/are c struction (as in There is a fly on the wall or There are sm flies on the wall) and equivalents in other languages. following the major study of them by Milsark (1977. It is important to distinguish the th of this construction from the deictic locative adverb there (meaning ‘in/to place’).2 Existential sentences It is in relation to existential sentences that the term “definiteness ece is most commonly used. demonstratives are used in many instances wh English would have the. as pointed out. obligatory use morphological topic marking. The crucial observation. For further discussion of the interaction of definites. other forms of reference with aspects of discourse structure. and the constraints on interpr tion of topics and subjects outlined relate to what I have termed “seman pragmatic definiteness”. The latter is a fully meaningful expression. It is clear that this notion plays a significant part in m languages which lack formal marking of definiteness. but it still represents demonstrative use. Mandarin. it see to be broader than the grammatical definiteness associated with lexical or m phological marking by an article. since it embraces all generically interpreted n phrases. Deictic there accompanying the verb be readily accepts a post-ve definite noun phrase or proper noun: There’s the man you’re looking for. which do not make regular. dog-INDEF 3SG-be-HAB in house ‘There’s a dog in the house.’ Even the subjects of presentational verbs must be pre-verbal. 6. see Fraurud (199 Gundel. beh 236 . More extensive use of demonstratives is normal in guages with no definite article. In the last three languages.snake-INDEF 3SG-bite-PRF child ‘A snake bit the child. Cantonese Lango. Sim definiteness is not marked in these languages.’ Note that there is no definite article in Japanese. and since they almost invariably indefinite in Lango they too usually take -mWrD: (25) gwókOkVrC tíê G zt. to which it is semantically related) a non-zero degree of str and it has a full (commonly diphthongal) vowel – a normal RP pronunciation be [2ha]. and the papers in Fretheim and Gun (1996). I shall refer to these as the “central” existen construction. is that the noun phr following there is/are is usually indefinite. it typically carries ( demonstratives. made long ago in descriptive grammars studied in detail by Milsark and other recent writers. demonstratives. Hedberg and Zacharski (1993). but.

and it does not readily admit definites or pro nouns in post-verbal position.4 The thr be construction alternates with a structure without thr in which the n phrase itself appears in subject position: thus A man is at the door correspond to Thr is a man at the door. It is typic unstressed. (27) a. and represent the existential pleonastic as thr. Thr man at the door is generally preferred to A man is at the door. follow. This construction. some major discussions Rando and Napoli (1978). though The m is at the door is perfectly good. readily accepts defin indeed it is indefinites that seem somewhat less felicitous in this case. embracing s tences with “presentational” verbs – mostly verbs which introduce an entity the discourse. A disturbance occurred soon afterwards. Lumsden (1988). . 6. emerge. There occurred a disturbance soon afterwards. and the post-verbal argument also shows the definiten ecect: (26) a.1 The phenomenon There is a certain amount of cross-linguistic variation as regards verb appearing in the central existential construction. b. can undergo phonological reduction (with pronunciations like [2az] there is and [2ara] for there are). b. remain. The analysis of existential sentences and of the definiteness ecect holdin them has been central to accounts of definiteness since Milsark’s discussion.1. by contrast. The captain appeared at the door. unlike thr be. arrive. ar ensue. To make clear the distinction between these items. even requ ment in many languages. this ecect has even been widely regarded as defining for definiteness. These verbs can also take pleonastic th subject in English. ?There appeared the captain at the door. There is an extended sense of the term “existential sentence”. a semantically empty filler of subject position. Examples of presentational verbs are: appear. Existential her. Many languages are like Eng 4 French distinguishes ‘there is/are’ and ‘thr is/are’ as voilà and il y a. is not obviously meaningful. the papers collected in Reuland and Meulen (1987). I will henceforth follow the practice already adopted for the some–sm tinction.2. come. Only a sm selection of the large literature can be examined here. discussed in 6. occur. Safir (1985). and it seems natural to relate this to the similar preference. and is usu taken to be a pleonastic. Notice that the preference for a definite argum in this construction without thr amounts to a preference for a definite in sub position.

as well as es ist/sind. has been much debated. as would be expected. The complement noun phrase is itself commonly followed by a further exp sion which seems to be predicated of it. there occurs in some languages in existential construction an additional particle of locative origin. 5 6 Russian also commonly uses a null form in the present tense. or whate predication is expressed by the post-nominal phrase in the coda. but with other. 238 . compare Es ist Mann (with nominative) and Es gibt einen Mann (with accusative) ‘There man’. but with no t element. The argument noun phrase is freque in subject or nominative form where the verb is ‘be’ (though commonly ins mental in Russian). German. It is null in null-subject guages. This phrase together with the noun phrase is term the “coda” by Milsark. among other possibilities. making structure an unaccusative one. A further contenti issue is whether the verb. the subject pleonastic is overt. be etc. so that if the predi is PP. therefore. German es gibt. Spanish uses a form of auxiliary haber. The structure a whole may express a variety of things: presence in a location. it is generally an object or accusative form. It appears. verbs. Other verbs too occur. Other factors may intervene here. The question of its structure. in a single language. especially where there predicate in the coda. Chung (19 claims that it is NP in Chamorro because this language does not have small clau while Huang (1987) argues against an NP analysis for Mandarin. with ‘give’. descended historically from the possessive verb but now oth wise mainly used to form the perfect tense.. It is usually taken to have the category status of its predicate. that the single argument is a complement. including whether it is a c stituent. as in a dog on the lawn. This may be a locative prepositional phr (Thr’s a dog on the lawn) or a non-finite verb phrase (Thr’s a man waiting to you). as in Thr is a God ). well as the possible pleonastic subject. chapte for discussion. in keeping with the fact tha copula byt’ ‘be’ is usually null in the present and that Russian does not have overt pleonas A small clause is a clause-like constituent comprising subject and predicate. S (1985) argues that it is a “small clause” and complement of the existential ve It may be that the status of the coda in fact varies cross-linguistically. apparently transitive. Williams (1984) claims the coda is an NP. has semantic content.Others use ‘have’: French il y a. the whole small clause is a PP. like Frenc ‘there’ in il y a (and the similar Spanish -y occurring only in the present te form hay). In languages like English and French wh require overt subjects. existence (es cially where the coda is a noun phrase alone. Cantonese yáuh. as in Germ where the pleonastic is only obligatorily overt where this is necessary to ens that a main clause meets the requirement that the verb be in second position. See Radford (1988.

but even in such languages it is not general. There are other specialized uses of the existential construction which accept defin and proper nouns easily. thr’s the chicken.the door and A man is at the door is probably limited to languages in which central existential verb is ‘be’. There is certainly a tendency against the use of definites in the existential c struction. and that cheese you bought. In guages using ‘have’ or ‘give’ or some other verb. Go and open the door. Thr’s that man on the phone again. But the complement noun phrase is by no means invariably indefinite. b. which presents the data as far m certain than they are. What the definiten ecect amounts to. In this sense a list can be of one item: (29) A: B: Is thr anyone still in the garden? Thr’s Fred by the pond. Thr’s John waiting at the door for you. is that a noun phrase subject of ‘be’ may be allowe be definite or indefinite (in languages that allow indefinite subjects). thr’s the postman coming up the drive. It is indeed often hard to decide whether a particular acceptable occurrence definite is a list use or not. For many speakers the following are perfectly good: (30) a. It is rec nized that definites and proper nouns are readily accepted with what is terme “list reading”: (28) A: B: What have we got to eat? Well. c. the central existential const tion may alternate with a structure in which the argument appears as subject clause with ‘be’ as verb (Spanish Hay un hombre a la puerta ‘Thr is a man at door’. for languages with articles and therefore clearly possessin category of definiteness. and this tendency (like that against indefinite subjects observed abo is stronger in some languages and varieties than in others. In fact the prohibition on definites in non-list uses in English is far from c gorical. I’m afraid thr’s only the cheese in the fridge. the bacon. but the c plement noun phrase of the existential verb is normally indefinite. is (31) a list u (31) A: B: I’m starving. since this use is not clearly defined. . and is subject to considerable variation between dialects and individ – a point neglected in most of the literature. Un hombre está a la puerta ‘A man is at the door’). such as the colloquial French Y a Jean qui t’attend ‘Je waiting for you’ (literally ‘Thr’s Jean who is waiting for you’).

see Lumsden (1988) for discussion. *We arrived three students. since even in English this definiteness ecect relates only to grammatically definite expressions. argument (by contrast with ergative verbs like break): (35) a. Thr appeared a policeman at the door. but also indefinite generics. c. and there is no use with additional. ?Thr sat a huge dog in the doorway. Thr walked into the room a tall man. c. Thr arrived three students. bare n phrases (which I claim are not definite) and singular indefinites must normally interpreted non-generically after thr is/are. though location verbs work considerably better in the inner ver structure than do movement verbs: (34) a. This means either that these languages do have feature [± Def ]. . b.guages lacking definiteness marking. Thr jumped from the wall three children. (33) a. c. 240 Three students arrived. there are several classes verbs involved in English. The clearly presentational verbs illustrated in (32) occur readily in the inner v bal structure (and in the outer verbal one: Thr appeared at the door a policema Verbs of movement and location like those in (33) are better in the outer ve structure. Milsark distinguishes an “inner verbal” structure in wh the argument noun phrase stands directly after the verb and before any predic in the coda. Thr sat in the doorway a huge dog. whether in subject or complement position. and not all show the same behaviour. ??Thr walked a tall man into the room. their single argument is n agentive. The behaviour of the dicerent classes of verb in these frames shows considera uncertainty. Thr rose a great cheer from the crowd. b. very gen in languages lacking articles. The verbs which occur ea in the inner verbal structure (32) are unaccusatives. Huang (1987) shows that in Manda proper nouns and nouns modified by a demonstrative are excluded in the co plement of the existential construction. The definiteness ecect is. or that the constraint is about semantic/pragmatic definiten The latter is more likely. not all clearly presentational. Thr arrived a bus in front of the house. Turning to the non-central existential sentences. b. agentive. and an “outer verbal” structure in which the argument follows predicate: (32) a. b. in this form. and bare noun phrases are interpreted indefinite and non-generic.

Again this is equally true of languages lacking definiteness m ing. The assumption here is that the man I had spoken to earlier is generated in s ject position (in keeping with its agentive status) and moved because of its he iness. is claimed to depend on the tence of three men. the central existential construction is also plausibly analy as unaccusative (though there may be problems with this treatment for the ‘have’ ‘give’ verbs used in some languages. So thr be makes a claim of existence about the entity described by complement noun phrase. in which thr be occurs. such as Zulu (Ziervogel. the distinctions drawn here not absolutely clear. The claim is that thr be is an exis tial quantifier which requires a variable to bind. In the unaccusative inner verbal construction. but in the unaccusative structure (37b) the verb is in imperso form and the post-verbal argument must be understood as indefinite: (37) a. given that these also have a transitive use ergatives). probably to adjoin to VP. being non-agentive. Milsark (1977.7 But definites are themse 7 This is on the basis of the following contrast: (i) (ii) Thr were three men believed to have been in the garden. The outer verbal c struction is said to admit a definite argument provided it is “heavy”: (36) Thr walked into the room the man I had spoken to earlier. . Louw and Taljaard 1967). are true intransitives. with an agentive argument. 6. In (37a) the verb sh agreement with the argument in subject position (which is ambiguously defi or indefinite). The truth of (i). is not an underlying subject but is genera in complement position. but they seem to be valid in essence. Three men were believed to have been in the garden.’ Given the close similarity. unlike that of (ii).2.structure. b. live old-woman here ‘An old woman lives here. 19 proposed an explanation in terms of semantic incompatibility between a defi complement and the existential structure. The inner verbal st ture seems to require a non-agentive verb.’ Kuhlala isalukazi lapha. This structure is much more strongly subject to the defin ness restriction. Isalukazi sihlala lapha.2 Explanatory accounts In the first major study of the phenomenon. single argument. old-woman lives here ‘The/An old woman lives here. Again.

This association is expressed by “co-indexati between the noun phrase and the pleonastic subject thr. or at least occurrences or uses of th as “strong” and “weak”. the indefinite determiner can only be interpreted as cardinal. Some students here means some of the students already mentioned. essentially non-generic indefinites. Some students wanted to move to a smaller room. res tively. is itself a configurati concept: the binder is co-indexed with and c-commands the bound expression. but a vague proportion of so already given set.5 and 5. universals like all. are not necessarily quantifi tional. are quantificational. this post-verbal noun phrase must be assume be associated with subject position. a “quantifica restriction”. But since nominative case is assigned to subject position by inflectional element of the clause. which permits strong noun phrases. Indefinites. in which they express not simply a quantity. Binding. therefore. generics. 242 . Mils classifies determiners and noun phrases. Principles A and B specify. Nominal expressions lin 8 Recall that the binding theory consists of three principles. They thus come close to the range of “semantic/pragm definites”. which says they may not be bound at all. in Thr were some/many students in the street o side. Safir (1985) ocers a syntactic account.2. personal pronouns. within the theory of binding outline 1. involve cardinality rather than quantification. behave like definites in rela to the existential construction. The defin ness ecect in existential sentences is for Milsark. excluded from existential sentences. interestingly. Milsark s gests the list itself counts as the argument of thr be for semantic purposes. The weak ones.8 Safir takes the post-verbal noun phrase in the English thr be c struction to be in nominative case (as we have seen is typical of ‘be’ existen constructions).5. The syntactic dom involved (typically the minimal clause) is termed the “governing category” of the expres in question. quantifiers like most. the determ (which in this use must be some as opposed to sm) expresses a partitive relati ship between two sets rather than simply the cardinality of one set. As for the list free for binding. R-expressions (full noun phrases essentially. they are compatible with the quantifier in thr be. For example: (38) About a dozen students and three or four lecturers turned up for discussion. and incl definites like the and that. that anaphors must be and pronominals may not be bound within a particular syn tic domain. But certain cardina determiners. on the basis of their behaviour in the existential construct The strong ones. so quantificational nature of any members of this list is irrelevant. Indefinites u in this way are therefore strong and. notably the vague ones some/sm and many. also have a quantificatio use. while accounting for referential dependence. on the other hand. As dinals. which are referentially independ are accounted for by Principle C. so may provide a variable that the existential quantifier can bind. which are acc able in existential sentences.

when normally semantically full expression should be the head. Ann is the author of book). bound wi its governing category. Safir claims that the list reading of thr be involves identificational be. this is because indefinites essentially predicative. and there is th fore no obvious reason to assume co-indexing with the pleonastic subject to p duce the crucial unbalanced chain. is c-commanded by the semantically empty ment thr. It is widely accep that English be is ambiguous between a “predicational” sense (as in John fool or Ann is very clever) and an “identificational” sense expressing identity reference between two arguments (John is the president. and “less referential” than definites. it is predicative rather than argumental. The noun phrase is in ecect bound by the pleonastic (since binding is mally defined in terms of co-indexing and c-command). the discussion in Huang (1987) of . indefinites are optionally exempt from Princ C. be bound like anaphors. In later work (S 1987) he proposes replacing Principle C of the binding theory (which says an R-expression is free. Accusative case sho be assigned directly to the complement noun phrase by the verb. like German and Spanish). but in fact it is perfectly good as long as the noun phrase is indefinite. This violation of Principle C should render the structure ungramm ical. A fur problem is the non-absoluteness of the data. There are many diaculties with this account. As Lumsden (1988) no specific indefinites are perfectly good in existential sentences. that is. at S-structure. This perm the indefinite noun phrase in the existential construction to be bound. when in fact full noun phrases are no interpreted.single discontinuous argument. This makes the pleonastic the “head” of the chain. not bound) by a “Predication Principle”. The noun phrase is therefore apparently behaving like anaphor. therefore. moreover. The account ocers no explanation for definiteness restriction in languages where the complement noun phrase of the e tential construction is accusative rather than nominative (essentially languages wh a verb other than ‘be’ is used. In this connection. for discussion see Reuland (1983). in this case the lexical elem the semantically full noun phrase. beca Safir argues. the full noun phrase is in a synta configuration which should require it to be interpreted as referentially depend on the (semantically empty) pleonastic. when in fact it is an R-expression. though they are f referential rather than predicative. with definites occurring in the c struction much more readily than is usually acknowledged. definites are often acceptable in thr be sentences that not obviously list uses. But. as noted above. They may. wh says that a potential referring expression is a predicate or else free. Sa explanation is that. subject to Principle C of the bind theory. Safir terms such a chain “un anced”. unusually. and. that only the predicational thr be construction is subject to the definiteness ec But. To put this in non-technical terms.

in which tones are not shown. w a “variety” reading. but about the presence of a copy. like indefinite generics English. any copy. not syntactic. Thus. rather than justice being the expression limits the domain of the variable bound by the quantifier. but that a fronted l tive topic such as zheli in (40) is moved into subject position. Huang assu that the subject position is underlyingly empty in both (39) and (40). variety interpretation does not suace to rescue (39). in which the determiner applies to a propositio function represented by the descriptive portion of the noun phrase. this indefinite. terms. with you ‘have’. and versions the definiteness ecect in existentials hold in many languages lacking definiten marking. The definiteness ecect constrains the occurrence of a class of noun phr characterized in semantic or pragmatic. so the quantifier mus 9 I follow Huang’s transcription. and Lumsden (1988) a ma pragmatic one. Perhaps the major diaculty facing a syntactic account is that of attempting characterize the definiteness ecect data in syntactic terms. that non-central existentials with unaccusa verbs show the definiteness ecect in main clauses and assertive contexts. Higginbotham takes the noun phrase following thr be to have interpretation of a sentence. but in non-assertive subordinate clauses. in example Thr is no justice the quantifier no combines with the propositional func provided by justice. the complement noun phrase in (40). 244 . for example.esting. There is in fact no pro sitional function specifying the domain of this variable. must be understood as semantically indefinite. Huang shows. of a particular book. But definiteness (at least simple defin ness) is not encoded morphologically or lexically in Mandarin. while formally definite in tha has a demonstrative modifier. It is the truth of this propositional function that is at issu the interpretation of the sentence. sh the definiteness ecect when the subject is null (presumably with a null pleon tic) but not when an overt topic occupies subject position:9 (39) (40) *You-mei-you zheben shu zai zheli? have not have this book at here ‘Is thr this book here?’ Zheli you-mei-you zheben shu? here have not have this book ‘Is thr this book here?’ Moreover. the definiteness ec excludes certain noun phrases indefinite in form. Central existentials. in his examples. And in languages that do have definiteness marking. Higginbotham (1987) proposes a semantic account. The acceptability (40) depends on the indefiniteness of one interpretation of a kind of noun phr that would normally be treated as definite. the question in (40) is not about the presence of a p ticular book. that is.

in which a definite is acceptable. This leads the hearer to de the list interpretation. str noun phrases in subject position are necessarily topics. if taken literally. and. This means that a str noun phrase in the subject position of the small clause would impose a pred tion interpretation. The noun phrase cannot be interpreted as a predication. and a n tral description is a sentence which does not have this structure. He argues that the thr be const tion represents existential quantification over this. She claims that this is the case with list uses. this is the definiteness ec Lumsden regards the list use as semantically the same as the standard use. The proposal the noun phrase is interpreted as a sentence rules out “singular terms” such proper names. the inclusiveness which she takes to characterize definiteness m not be relative to the preceding discourse or the immediate situation. The list use involves a coda with noun phrase rather than small cla status. She regards the e tential construction as having the function of introducing a new entity into discourse. and thus to have propositional status. Lumsden (1988) takes the coda of existential sentences to be typically a sm clause. This means that if the new entity is introduced in the form of a defi noun phrase. Provided referent is inherently unique or meets the condition of inclusiveness relativ some other pragmatic domain. definite descriptions are ruled out by the hypothesis that definite determiners can be unrestricted quantifiers. being derived by im cature. The analysis is strai forwardly extended to existentials with unaccusative verbs by treating these as being predicated of e. special interpretation. Higginbotham hypothesizes that o indefinite determiners. so that it can still count as new to the discou a definite is acceptable. that sentences c tain an “E-position” in their logical representation. rejecting any syntactic or semantic definiteness ecect. can be unrestricted. which are adjectival. where e is a variable rang over events and states. Follow Burton-Roberts (1984). But the nominal coda presupposes the existence that thr be asserts the sentence is. so its definiten does not produce a violation of any semantic constraint and the sentence is w formed. Lumsden claims that definite. . and that the small cla proposition is interpreted as a “neutral description” rather than a “predicatio by “predication” he means a sentence with topic–comment structure. more generally. pragmatic approach to the use. Holmback (1984) also takes an interpretative. This variable satisfies the single argument of the verb which is claimed to be a predicate true of everything.restricted and unrestricted quantification. since these are taken not to have a quantifier-variable structure. uninformative. moreover. which is incompatible with the requirement that the const tion consist of quantification over a neutral description. Higginbotham proposes.

Given this. applies to some semantic or pragmatic grouping of noun phrase types or u which largely includes but also overlaps with those marked as grammatically defi in some languages. count as strong for this definiteness ecect. Other writers argue that there is no restriction on the occurrence definites in the existential construction. and also involving someth broader than grammatical definiteness. only possessives among the definite determiners are poss (my cleverest student). when used with a p titive implication. and some. that too is a strong cross-lingui tendency. even if one takes “strong”. Some of these. There are even so clearly definite noun phrase types which do not always trigger the definiteness ec like the Mandarin example (40) with a variety reading and certain uses of supe tives in English to be considered below in 6. such as the diagnostics for definiteness used in Chapte have been mentioned in the present work. Cardinal determiners like many. Apart fr the definite article.3 Other definiteness effects A great many putative definiteness ecects have been pointed to in literature. excluded with superlatives (*all/every/most/some/m cleverest boy(s) ).3. quantificational determiners like all. three.3. as in cated by the obligatory presence of the definite article. grammatical sense. And it is clear what is involved here is definiteness in the narrow. 6. It has much in common with the constraint indefinite subjects or topics examined in 6. and cardinal determiners like som their quantificational use. stricter in some languages than in others. and the extent to wh what is at issue really is definiteness. but so too are demonstratives (*that cleverest boy). In fact only are strong determiners like all. nota 246 . only constraints on their interpretation st ming from the meaning of thr + verb. not the broader notion strong noun phrase nor a semantic/pragmatic concept of definiteness. presumably because the definiteness of DG possessive that of the definite article. But whatever kind of constraint is at wo it is not based on definiteness in the limited sense of a noun phrase with defin ness marking. 6. *a cleverest boy. are examined here. “simple” definiteness.and various attempts have been made to account for it in syntactic and sema terms. most. it is more likely to be a semantic or pragm constraint than a syntactic one.1. whate it is. as do bare nomin like dogs and simple indefinites like a dog used generically. ev most to be definite. The definiteness ecect. Some other modifiers.1 Superlatives It is a general fact that languages which have definiteness mark use it with superlatives: the cleverest boy.

most with this sense acc the indefinite article. and in this they can be indefinite: I met a most intriguing girl at the party. both most and -est superlative forms can be used fairly readil existential sentences: (42) a. For example. b. and a logical contradiction wo result. Superlatives are discussed in detail by Hawkins (1978). In relation to ( Elinor can be described as one of the prettiest girls at the party. As just pointed out. why not *I talked to a/one cle est student? It is well known that English superlative forms with most. Thr was the biggest frog you can imagine sitting beside the pon Even if these are not meant to be taken quite literally. but then why as *a prettiest girl at the party? Similarly. or indeed one of the strong determiners which do entail inclusiveness. But this does not explain it. who claims that a uniq ness or inclusiveness element in their meaning accounts for the definiteness ec they no accident that these forms are of superlative origin. Superlativeness means having some property to an extent to wh no other objects have it. and this suggests that the adjective phrases in ( 10 Hawkins notes that there are some non-inclusive uses of superlative forms. it is certainly fully defi Consider further the following: (43) a. Thr’s the strangest man I’ve ever met in the drawing room. apparently violating the definiteness ecect found with existen sentences. like a best buy a first course in German (which allow the existence of other best buys and first course German). they can only be interpr as superlative uses of -est. which can be indefinite. as opposed to can have a non-superlative sense. b. with a superlative would imply that there are other enti with the same degree of the property in question. and in (42a) the has been chosen over the alternative a the most intriguing in this sentence is not fully superlative. given that one can say I talked to a/ student who was cleverer than all the others. But even with definite article. is usually taken to be due to the superlative having the extremely re ing. equivalent to very or extremely. An indefinite determiner. . Thr is the most intriguing girl in the garden. (41) means that no individuals other t Sarah and Elinor satisfy the description prettiest girl at the party: (41) Sarah and Elinor were the prettiest girls at the party. This possibility. Thr is the strangest man in the drawing room.10 But this account does not seem to be fully adequate.

t closely complement the existential construction. is the following: (46) John didn’t expect X with any fresh eggs in it/them to arrive at 3 today. every basket etc. as more g erally in discussions of definiteness ecects. X may be replaced by a basket. designed to abstract away from cer complications. Safir does not explain these distributional dicerences between definite and in finite (or strong and weak) noun phrases beyond suggesting a link with the “m referential” nature of the former. Safir uses the following sente frame to illustrate: (45) X was/were sold on linguistic theory. three baskets. only behave like definite existential sentences when they are intended literally. The second test involves any in the scope of negation. Bill’s baskets.2 PP-extraposition and any opacity Safir (1985) discusses two tests for definites which are particul interesting in the context of the existential construction.. many baskets etc. these are thus opaque for polarity any. involves PP-extraposition. But these examples contrast with the following: (44) ?Thr’s the brightest student in the class leading the seminar toda It appears that superlatives. t draw exactly the same distinction among types of noun phrase. but not the book(s). It follows that “definiteness” relevant to the definiteness ecect of existential sentences is a g uine. though formally definite. the noun phr left behind by this process must be indefinite. all (the) books. many books. because. 6. If the non-literal ones are sem tically indefinite. it is clear that a distinction must be made between grammat and semantic definiteness. including one posed by a possible generic interpretation of the n phrase. more bo etc. The position X here may be filled by a book. from Guéron (1980). Safir’s illustrative sentence frame.3. John’s books. For discussion see Ra and Napoli (1978) and Lumsden (1988: 176–9). three books. is the reliability and firmness of so 248 . while indefinites are tra parent. But the tests are interesting if. and that the definiteness ecect in existential senten is about the latter. The main diaculty.. independently motivated semantic category. but not the b ket(s). The first test. Construal of any wi negative outside the noun phrase is claimed to be blocked in the case of defi noun phrases. as claimed.than extremely uses. he claims. most books etc.

like many.3 Property predication Milsark (1977) notes that indefinite or weak noun phrases can oc as the subjects of predicates that assign states.3. 6. b. b. secretaries usually ap with any ambition. The/Every man has hair. He shows that noun phrases like men w any sense. and then applies the PP-extraposition test to such generics: (47) a. must be understood ge ically. Whenever there are jobs like these available. Lumsden (1988) also discusses this material in detail. But notice that the same applies to ove indefinite noun phrases with the same any. claiming that st (and events) are stage-level predicates while properties are individual-level . Again the point is clear that the class of noun phrases characterized as definite these ecects is a semantic class. of course. but Milsark claims that this represents quantificational use of card determiners. a secretary usually app with any ambition. Again the classes of noun phrase distinguished by this ecect correspond clo to those distinguished by existential sentences. A man is injured. several. and that grammatical definiteness. He proposes that properties can only be predicated of strong n phrases. A man here must. be interpreted non-generically.tant point that bare nominals interpreted generically count as definite on these te and demonstrates the point as follows. *A man is intelligent. three. so ex position as in (47b) is not possible. Strong noun phr are not restricted in this way: The/Every man is drunk. *Whenever there are jobs like these available. A man is drunk. extraposition as in (47a) imposes the latter re ing. secretaries usually ap with good credentials. but not of predicates that as properties: (49) a. is something distinct. work perfectly well in the envi ment (50). b. *A man has red hair. also necessarily understood generic such as a secretary with any ambition: (48) *Whenever there are jobs like these available. The (apparent) problem is that m weak determiners. in which any is not in the scope of negation. While secretaries with good credentials is ambiguous between a generic an non-generic indefinite reading. But secretaries with any ambition can only be understood generically. as expres by articles. (50) a.

but many languages with internal-head relatives use no s 250 . The function of the head in the relative claus expressed in that clause by the relative pronoun which. There is considerable cross-linguistic variation in the syntactic strategies u for relative clause formation. in the position appropriate to its function th and is not separately represented outside the relative. the bulk of the noun phrase (though not necessarily determiners other grammatical markers) which it is interpreted as modifying. This noun phrase as a whole functions as direct ob of the main clause.3. But an alterna form of the relative in this sentence would be that you recommended. though the pattern just exemplified of the role of head within the relative being expressed in a reduced form (by a relative prono or a resumptive pronoun in the relevant position) or not at all predomina But another pattern is the “internal-head” type. the relative is part of the noun phrase of wh book is the lexical head. The following example is from Bambara (Bird 1968. again as direct object. generally taken to und go transformational movement from its underlying object position after the v recommended to a position at the left periphery of the clause. in which the head (book abo appears inside the relative clause. postposed to the n which is the head. man PRES I PAST house REL see build ‘The man is building the house that I saw. A third alternativ to omit even the complementizer (I’ve just bought the book you recommende making the point yet clearer that the position or role within the relative cla with which the head is construed can be unexpressed. which means that there is no express inside the relative clause of the direct object of this clause. 6. The relative clause abso as it were. while its head is also construed as having a function in relative clause. See Radford (19 chapter 9) for discussion.’ Bambara marks relative clauses with the morpheme mìn. More technically. and th is reason to believe that that is not a relative pronoun here but simply the gen complementizer or subordinating particle. which is usu termed its head. thus the book is both the thing bou and the thing recommended. The relative which you recommended modifies the noun book.of weak noun phrases as quantifying over stage-level entities. Comrie 198 (52) Tyh be [n ye so mìn ye] dyj.4 Internal-head relatives Consider the following sentence containing a relative clause: (51) I’ve just bought the book which you recommended. See Comrie (198 for discussion.

kC ‘the’ in this example.5 that articles can modify clauses in Lakhota. The higher determiner makes the phrase definite and the lo one is neutral in this regard. exemplified by (53). in this connection re from 2. Williamson regards simple definites as non-quantificatio but argues that the familiarity they express would be at variance with what is c veyed by a restrictive relative clause. and work out where they divide range of noun phrase types. This higher determ can be definite or indefinite.1. Whether or this explanation is correct. represented by the outer brackets in (53 a noun phrase. Consider (53): (53) [ [Joe wowapi wa owa] kc] wac‘c. A further striking point.2. and in this language it displays a definiteness ecect. This is very strong evidence for my thesis that “qu indefinite” articles are cardinality determiners and do not encode indefinitene 6. wowapi wA ‘a book’. rendering this uninformative. What we have seen here is that most definiteness ec identify a definite or strong class which is considerably broader than the clas noun phrases in which definiteness is encoded. Joe book a wrote the want-1SG ‘I want the book that Joe wrote.2. and determines the interpretation with respect to defin ness of the noun phrase modified by the relative11 – thus whether it is ‘the b that Joe wrote’ or ‘a book that Joe wrote’. But whatever the higher determi the internal head itself must be indefinite or weak in form. the inner brackets represent a clause structure. If it is a full n phrase with determiner as in (53) this determiner cannot be definite or strong range of indefinite pronouns are also permitted. Williamson predicts that this definiteness ecect should apply gener to languages with internal-head relatives. Williamson argues that relative clauses are treated in the semantics as pro sitional functions with a free variable. If we recognize quantifiers 11 Williamson assumes the overall relative structure.Lakhota. it is striking that the range of noun phrases ruled as internal heads is again apparently the same as that excluded from existen sentences. This makes a quantified head impossi Following Heim (1988).3.5 Concluding remarks The reader is invited to turn back to the remaining definiteness gnostics of Postal (1970) introduced in 1. But there is a second determiner modify the relative clause as a whole.’ The head inside the relative clause takes the form of a complete noun phrase inc ing determiner. examined in de by Williamson (1987). sister to the higher determ . this does not prod a semantic clash. is that where the higher de miner is definite and the lower one is the cardinal article.

252 . is always paraphrasable a noun phrase with the. These definiteness ecects clearly do identify a genuine semantic class. gramm cal sense is superlatives. that still leaves indefinite generics. tho it seems to be defined morphologically or structurally rather than semantica This is clear from the phenomenon of DG extension personal pronouns). proper no (if these are indeed indefinite bare nominals). and cardinals used quantificati ally. These two notions of definiteness will be discussed further in next chapter. which appears to dep on the possessive being in a particular “Det position”. but narrower class of noun phrases overtly marked as definite is equally real. and seems to occur only in languages which have defin ness marking. The one construction which is limited to definites in the narrow. even demonstratives being excluded. and here the restriction is in fact to a narrow subse definites.

But neither works for all uses. Identifiability is part larly attractive for referential uses. As will be seen. but ther is apparently fully adequate as the defining feature. In this chapter we survey attempts to ana definiteness within various theoretical frameworks. and p matists to prefer identifiability. uniq ness – thus limiting themselves to accounting for singular definites). Definiteness seems empiricall be a unified phenomenon. occurrin Apollonius Dyscolus (second century AD) who distinguishes presence and abse of the definite article in Greek in terms of whether or not the referent has alre . very often. I will p pose an account of definiteness as a grammatical category which. with versions of familiar identifiability at its centre.1 The grammatical. some h indeed sought to combine the two. This hypothesis goes back to ancient times. like other s categories. tho it represents the grammaticalization of some category of meaning. on the evidence of the way languages represent it. or have sim assumed one or the other.7 Defining definiteness The informal attempt in Chapter 1 and subsequently to reach a g eral definition of definiteness ran into a puzzle. 7. as the basic descriptive insight. Two characteristics are prominent. After outlining some major approaches I will argue (following up h dropped in preceding chapters) that the attempt to find a fully unified charac ization of definiteness in semantic or pragmatic terms is misguided. it is not straightforwardly characterized. and inclusiveness is particularly attractive for n referential uses. Indeed many uses are readily handled by either one of these c cepts. writers h variously argued for versions of identifiability or of inclusiveness. especially where the referent is a physical en locatable in a physical context. The general tende is for logicians and semanticists to prefer inclusiveness (or. cannot be completely defined in semantic or pragmatic terms. But this is by no means a general rule. logical and pragmatic traditions Work by descriptive grammarians on definiteness (or on the mean and use of articles) has long been dominated by the kind of discourse appro characteristic of current work in pragmatics.

Jespersen discusses sources of the “nearly complete familiarity”. the dicerence be that with the latter there is no overt mention of the thing providing the contex basis. the relevant volume of wh appeared in 1943. 1885). for him such nouns denote concepts fixed in minds of speakers. who anticipa a large part of this account. wh the referent has been introduced in the previous discourse. or the wider sit tion (in a particular town. Stage II is “nearly complete familiarity”. claiming there is no such thing in human thou as an essentially singleton class.this view in the present century is due to Christophersen (1939). and corresponds to proper nou vocatives. and here his categories are based clo on those of Christophersen. the table). Finally there is the “constant situational basis”. and correspo to the use of the with any common noun. and which is based squarely on the familiarity hypothes Jespersen’s account consists essentially of a theory of “stages of familiarity”. the gasworks). The “situational ba is where the identity of the referent is clear from the non-linguistic context – immediate situation (in a particular room. There may be an “explicit contextual basis”. and reacting the earlier work of Frege (1892) on the sense–reference distinction. and Christophersen in turn strongly influenced account appearing in Jespersen’s great grammar. who was building on. An “implicit con tual basis” would be where the reference is to something connected with an en already mentioned – that is. fam iarity being defined as “knowledge of what item of the class denoted by the w is meant in the case concerned”. The logical or formal semantic analysis of definite noun phrases (or defi descriptions) goes back to Russell (1905). Stage III is complete familiarity. which always take the definite article – though Jespersen dismis the characterization “unique”. Jespersen points to a close relations between the implicit contextual basis and the situational basis. Christophersen was certainly influenced by the descriptive work of the g nineteenth-century German grammarian Maetzner (1880. In elaborating on stage II. and a few other cases. Stage I is complete unfamiliarity. when the referent i be found in the linguistic context or the non-linguistic situation. and correspo to indefiniteness. the associative anaphoric use. in the case of uniques devil or sun. 254 . In devel 1 The sections of Jespersen’s grammar dealing with the articles were actually written Haislund after Jespersen’s death on the basis of a plan he dictated. A major failing in Jespersen’s account is that he takes fam iarity or unfamiliarity to be to the speaker – misunderstanding Christophersen w makes it clear that (un)familiarity to the hearer is what is crucial. the door. who argues the use of the in English directs the hearer to the referent of a noun phrase indicating that this referent is familiar to hearer as well as speaker. see 1. wh renders the use of the definite article redundant.

Straw (1950. Much later. There is only one King of France. individual uses of these. examining sentences such as: (1) The King of France is bald. What is not so generally agreed on is the status of th clauses in relation to the third proposition. So if (i) is false. but not at another). it is a necessary background assumption to the sente the sense of which is that the person referred to is bald. He argues that in (1) only Russ (iii) is asserted. Frege held that referring exp sions presuppose a reference to something. For Frege reference of the expression the King of France to some individual would no part of the sense of (1). Of these three conjoined propositions. the st ment (1) lacks a truth value. which is true or f (since one may correctly assert at one point in history that the King of Franc bald. All these propositions are of equal status in that all three are asserted. and which therefore carries the preconditions for t . thus if one of them is false.expressions. Thus a sentence like (1) may be used to make a par ular statement or assertion. while (i) and (ii) (taking th to be descriptively correct) are felt to be background assumptions. which Russell rejected. the first two – the existential clause the uniqueness clause – characterize the definite description the King of Fra and these two elements are assumed in nearly all logical work on definites. Ind in many accounts which replace uniqueness by familiarity or identifiability. This in was Frege’s position. discussing a version of the exis tial clause only (as have most writers on this issue). a precondition the truth or falsity of the statement presupposing it. 1952) criticizes Russell’s account and proposes in ecect a return to Fre theory. though in a more elaborated version. not the sentence. This individual is bald. because the definite description contained in it f to refer. It foll that each of them is a logical entailment. Russell’s account runs counter to a strong intuition that only cla (iii) is actually asserted by use of the sentence (1). the existential clause (i) is a presupposition. Strawson insists that it is not sentences or propositions which presuppose. the wh conjunction is false. the assertion that the individual in q tion is bald. the e tential clause is still taken to be a constituent of the meaning of definite noun phr (as also of indefinites). and it is this. and that such presuppositions m hold true for the sentence as a whole to be either true or false. Russell claims that this sentence represents a conjunction of three propositio (i) (ii) (iii) There is a King of France.

Presupposition as a semantic concept is in contrast with the lon established and fundamental concept of entailment: a sentence A entail sentence B if whenever A is true B must be true. whereas A presupposes B whenever A is true B is true and whenever A is false B is true. the negation of (1): (2) The King of France is not bald. in which the existence of the K of France as well as his baldness is denied. Negation have narrow scope. however. ∃x (Kx & ~∃y ( (y≠x) & Ky) & Bx) . deictic expressions and personal pronouns could only be asses given a specification of the time and place of utterance and the identity of spea and hearer etc. or wide sco so that the entire proposition is negated. as already recognized by Russ the existential implication of definite descriptions does not necessarily survive un negation of the sentence. The most natural reading of (2) is that there exists an individual who is King France and that this individual is not bald. Strawson’s position is not obviously distinct from that o number of semanticists working within the generative semantics paradigm the early 1970s who developed a clearly semantic concept of presupposition a property of sentences. There is a problem. This is shown in the following log representation (where K stands for King of France and B for bald ): (4) 256 The King of France is bald. As noted abo the presuppositional account of the existence implication of referring expressi (and of the uniqueness implication of definites if one accepts this) captures intuition that these implications are less to the fore than the central assertion a (like later accounts. Consider (2). applying only to the predicate (clause (iii) ). here the existential implication is p served. since for him all questions truth (generally taken to be fundamental to semantics) relate to statements ra than sentences. the contribution to the sentence’s truth value of te morphology. This interpretation can in fact be impo by adding a continuation which explicitly denies the existential implication: (3) The King of France is not bald – because there is no King of Fran This fact was part of Russell’s motivation for rejecting Frege’s presuppositio account of reference. since. By treating the existential clause as an assertion on the sa level as the foreground assertion of the sentence. But there is a second possible reading. to be discussed below). Russell was able to account the two readings of (2) in terms of dicering scope of the negation. since proponents of a truth-based theory of sentence me ing recognize that the truth or falsity of a sentence is usually relative to a par ular context of utterance.

Nonsense. where it can be argued to fall within scope of some counterfactual or hypothetical operator. It n looks as if both Russell and Strawson are right: the existential claim is an en ment because it necessarily holds in positive non-opaque sentences. But the pragmatic theory characterizes the cancella . b. Kemp 1975.a. treating the presupposition as a conversational implicature of the sort posed by Grice (1975). but may within the scope of the operator in negative and opaque contexts. In the absence of nega or some other such operator creating an opaque context the existential impl tion always holds and cannot be explicitly denied (??Mary has married the K of Paraguay. narrow-scope negation ∃x (Kx & ~∃y ( (y≠x) & Ky) & ~Bx) wide-scope negation ~(∃x (Kx & ~∃y ( (y≠x) & Ky) & Bx) ) Strawson has no account of these facts. the wide-scope interpretation works better: (6) A: B: I’ve just heard that Mary got married last week to the King Paraguay. In other instances. so that the only interpretation of (2) is the one correspondin Russell’s narrow-scope analysis. and (3) is simply odd. there being no King of France at the present t is no valid reason for asserting that the King of France is not bald. and typically doe This led a number of writers in the mid-1970s to propose that the existen implication (along with other putative presuppositional phenomena) is b entailed (in positive transparent contexts) and pragmatically presupposed – to so extent a conflation of the accounts of Russell and Strawson (Stalnaker 1974. for him. She certainly didn’t marry the King of Paragua Paraguay doesn’t have a king. and (3) – while speakers admittedly do say such things – does st us as odd. The crucial point about these pragmatic presuppositi (by contrast with earlier semantic definitions) is that they can be cancelled. the narrow-scope reading of (2) is by far more natural. and this fact suggests that this im cation may in fact be an entailment of the sentence – Russell’s position. Moreover. And ind this view has some intuitive appeal. Strawson’s position is that there is no wide-sc reading. This view combines semantic and pragmatic elements in account. for example where the definite descrip is in complement position. we saw above in Chapter 4 that the existential implication may fa hold in a wide range of opaque contexts. but it is supposition-like in that it may survive under negation etc. Wilson 1975). this is what happens in the cases where they fail to be preserved under nega and in opaque contexts. however. presuppositions are preser under negation.. though there is no such person).

this feature then being spelled out as the. whether correctly described as a uniqueness implication an identifiability implication. the distinguishing clause being (except in the senten anaphoric use) a conversational implicature. partly because for m writers the point at issue was the correct treatment of presupposition-like phenom in general rather than the analysis of definiteness. In this use Kempson a syntactic rule of definitization. which he discus in pragmatic terms: we use the to signal reference to something previously refer to. For criticism of this approach see Kleiber (1981). or to signal that features of the context or background knowledge should ena the hearer to single out the referent. because it can be fa 2 Hintikka (1970) argues that knowing who or what an object is amounts to knowing that object is the same as some existing object. behaves dicerently from the existential one. neither this identifiability implication nor the r Russellian uniqueness implication can be an entailment. this is con ered suacient to constitute knowledge of the referent on the part of the hear But. 258 .that survival of the existential implication under negation is more natural than suppression. The tr ment of clause (ii) is crucial to understanding definiteness. however. rejecting Russell’s uniqueness implication on the grou that the referent of singular indefinites is just as unique as that of singu definites. and partly because this elem Russell’s clause (ii). and argues that definites and indefinites are semantically (tru conditionally) identical. The uniqueness implication attracted less attention pa because a familiarity or identifiability account was more to the taste of lingu pursuing an analysis involving Gricean pragmatic elements. the existen implication is common to both. As the discussion so far has implied. since it is implication which is taken to distinguish definites from indefinites. but that it is “uniquely identifiable” by the hea This is because she regards the sentence-anaphoric use (where the definite n phrase picks up a previous mention of the referent in the same sentence) as ba providing the model for the interpretation of other uses. this debate concerning the status of putative elements of the meaning of definite descriptions was concentrated on existential implication. In fact Strawson (1950. Burton-Roberts (1989 Chierchia (1995). and for more rec work on presupposition see van der Sandt (1988). which assigns the feature [+ Def ] to the sec of two coreferential noun phrases. Kempson (1975) also treats this question some detail. A it is his knowledge of this rule that enables the hearer to identify the referen the definite noun phrase with the referent of the earlier mention. He sees the essence of definiteness as identifiability. She claims that the definite article imp not that the referent is unique. 1952) paid considera attention to this question. Kempson argues.

did you? (11) The King of Ruritania came to my exhibition. b don’t know which glass you mean. It does seem possible to deny a proposition on the basis of the falsity of the uniq ness or identifiability implication of a definite noun phrase within it provided article is stressed – not surprisingly since the failed implication is what defi this article. you’ve only done half of the (10) The King of Ruritania never showed up at my exhibition. No – the glass hasn’t fallen on the floor. Principles of communication ens that the hearer infers that the speaker must be intending to convey some e . maybe a glass has. which has no semantic interpretation. ?No – the glass hasn’t fallen on the floor. where the spells out an instance of [+ Def ] not inserted by rule. which show how identifiability can be cancelled b in negative and in positive non-opaque sentences: (9) A: B: I’ve washed the dishes. This distinction is reinfor by examples such as (9) (based on inclusiveness rather than uniqueness). In fact the denial in (7) seems to work considerably better t the one in (8). suggesting that a uniqueness implication would be an entailm but an identifiability implication would not be. Sentence-anaph definites get this presupposition from a syntactic rule of definitization. a glass has – there w eight. the supposition is a conversational implicature derived from the presence of this ture. let alone that there was a King of Rurit – did you? Kempson takes identifiability to be pragmatically presupposed in posi transparent sentences as well as negative and opaque ones. No. It is not cer that Kempson is right here. thoug don’t suppose you knew there was such a person. although it is unacceptable to say It is not true the glass has fallen on the floor because there were eight glasses or It is not that the glass has fallen on the floor because I don’t know which glass you me consider the following: (7) A: B: or B: (8) A: B: The glass has fallen on the floor.floor would not be rendered false either by there having been more than one g or by the hearer being unable to identify the glass in question. though I don’t supp you knew I was having one. one of the glasses has – th were eight. The glass has fallen on the floor. and Kempson’s (10) and (11). you haven’t washed the dishes. In o uses. No – the glass hasn’t fallen on the floor.

or an association 260 . and the reader is referred to his book for a more deta (and highly readable) presentation. revised condensed as his 1978 book.2 is a simpli version of Hawkins’s. But it is interesting to see that the status of a uniq ness implication and that of an identifiability implication. This paradigm is still very influential in the French-speaking wo but almost unknown beyond this (for good reasons – although it has produ many worthwhile ideas). the entities known by speaker and hearer to constitute eit the previous discourse. Finally in this section. but account in many ways foreshadows the discourse semantics of Heim (1988) well as more recent pragmatic work in relevance theory. the most substantial body of work definiteness is probably that of Hawkins. where most ot important references can be found.1 Shared sets. is Pattee (1994). coming down firmly on the side of the first – in fact the te “inclusiveness” is his. mention should be made of the grammatical tradit following Gustave Guillaume. 7.1. 7. Kempson’s syntactic mechanism would not be acceptable today. make a significant concession to the ditional criterion of familiarity in his concept of “location”. the immediate or larger situation. that is. within which a great deal of work has been devo to definiteness. Guillaume’s major work on definiteness been reissued as Guillaume (1975).2 Hawkins: inclusiveness and location In terms of descriptive breadth. What is clear more generally from the cussion in this section is that none of the three traditions in the period conside came to grips with the issue of inclusiveness (uniqueness as it then was) ver identifiability. this extra information can only be that implication of iden ability which necessarily accompanies [+ Def ] in the sentence-anaphoric use wh it is obligatory. Hawkins’s typology of uses of the points the generalization that the referent of a definite noun phrase must be part o shared set.semantically empty. and we will examine own later work under this heading. starting with his 1974 thesis. however.2. My own informal classification in 1. inclusiveness and exclusiveness Hawkins (1978) gives the most detailed account to date of the ra of uses of the and a. Hawkins does engage with the inclusivene identifiability issue. as entailment or pragm presupposition. The problem is that the Guillaumian literature tend be highly metaphorical and obscure. and a good recent study. would probably be dicerent – though I am not aware that any noticed this at the time. He does. but each chose whichever of these best suited its own purpos reasonably enough perhaps.

and bought some tyres oc anothe (18) A prime minister has just died. and tore out some pages. and this is incompatible with exclusiveness which would be implicit in an indefinite reference. Indefinites refer exclusively. In (12)–(15) the reference is to one bu or cabinet minister. larger uation. But in e case the noun phrase in bold. The point is the fact that we can associate a noun phrase with a shared set does not by it make definiteness obligatory. only and superlatives. only and superlatives is that the semantics of these modifiers impo a unique or total reading on the noun phrase. in wh certain modifiers. and associative anaphoric use. See the dis sion in 6. These look like examples of Hawkins’s visible or immediate situation. even tho they do not seem to identify an antecedent.1. which is readily related to a shared set. . We have seen with (12)– that indefinites may be locatable. (c) refers to the totality of the objects or mass within set which satisfy the referring expression.locate the referent in it (that is. but now consider the following: (16) Fred bought a car last week.3. please. It is essentially the third of these clau which distinguishes definites from indefinites. To summarize Hawkins’s account of the use of the definite article. The modifier in these cases does not serv identify the shared set.3. (b) instructs the hearer to locate the referent in so shared set of objects. situation or association to permit lo tion. it (a) in duces a referent to the hearer. require the. And the reason that definiteness is obliga with same.1. (14) [At a wedding] Have you seen any bridesmaids? (15) Fred picked up a book. but neutral with respect to the pragmatic restriction of the domain of refere embodied in the idea of location in a shared set. I n’t catch which one. which is established in some other way. respectively. including same. or a number of pages or bridesmaids. Hawkins has an additional usage type of “unexplanatory modifiers”. Consider the following: (12) Pass me a bucket. and then he sold some tyres to his frie (17) Fred sold a car to one friend. general situation. (13) Have you heard the news? A cabinet minister has just resigned. among several m But with definites the reference is to the only entity or all the entities in the sha set satisfying the description used. is indefin This is where inclusiveness comes in. these are illustrated in 1. understand the referent to be part of it).

In general. however. and likely to involve successful re ence. I begin with Ly (1980). What we have then is a st dard first-mention indefinite. 7. the speaker stands dres for a journey. And in (18) the refer of a prime minister is almost certainly not to be located in a larger situation sha set. it must therefore be definite. One of these points already been discussed in chapter 1. I repeat the relevant exampl (19) [In a room with three doors. all closed.Fred sold could be from the car he bought – but it need not be.2. The second point is that there is an important distinction to be drawn in acceptability data for definites. (20) [In a hallway with four doors. The referent of an indefinite n phrase is only locatable in a shared set if the reference can be understood as be to a proper subset (that is. But in neither case does inclusiveness apply. speaker who uses a definite description is appealing to knowledge on the hear part of a relationship between the referent and a shared set. Since the country has only one prime minister at a ti the use of a noun phrase Det prime minister to refer to this individual is nec sarily inclusive. and this is enough. In this last case the reason is that such location would violate exclusiveness requirement. one open and two closed] Close the door.2 Subsequent developments Hawkins (1978) has been the starting point for much subsequent wo including further developments in Hawkins’s own thinking. who criticizes two aspects of the above account. In (17) some ty is almost certainly not associatively anaphoric to a car. These are immediate situation uses of the. please. it is not to be understood as referring to the present prime ministe this country. 262 . For example. not clearly discerned by Hawkins. a suitcase in each hand] Open the door for me. Lyons takes this to argue for identifiability rat than inclusiveness. please. a wedding one can refer to the bridesmaids because one can assume that the he knows that weddings tend to involve bridesmaids. less than the whole) of the entities in that set satis ing the description. Hawkins notes. Some factor in the sentence the situation makes it clear which object among several satisfying the descript is intended. that is. If there are in the putative shared set no entities of the ri kind which can be understood as excluded from the reference. then the refere must be construed as not relating to that shared set. but I return to it here briefly because i central to the issue of the nature of definiteness.

But the gre looseness of this usage compared with instances where the speaker is appea to the hearer’s prior knowledge of the required connection is evident from the that it is open to the hearer not to cooperate and to reject the definite refere with a reply such as The what? or What horse-trough? or I didn’t know the I Sea had been swum. (25) Unexplanatory modifier The first person to swim the Irish Sea was a Cossack. he is being informed it. In such examples the hearer is being informed of the existence of the referen the appropriate shared set (the physical vicinity). make your way to the zócalo. and the refere yet be successful.3 All this applies too to cases where the description in definite noun phrase is new to the hearer: (26) When you arrive in Mexico City. which he will manage to find. or accepting that it is probably a major feature of Mexico City. The dog will bite you. he must son that since he is unaware of the connection implied. In each of these examples. (24) Establishing relative (Hawkins’s example) What’s wrong with Bill? Oh.lowing (where the hearer does not know of the presence of a dog): (21) Don’t go in there. the locatability of the referent in the appropriate sha set (the fact that this household has a butler. . This success depends on the hearer’s cooperation. . . and he must accept the definite reference as thus informing him. the fact that Fred went out with a woman last night. The hearer may have no idea what a zócalo is. (23) Larger situation Meet me at the horse-trough tonight. and Lyons gives examples corresponding to all the other usage ty I repeat some here: (22) Immediate situation I’ll get the butler to show you out. According to Hawkins this p sibility is only available in immediate situation uses. or of Mex cities. response like this does not necessarily imply rejection of the refere The hearer may accept the information and still indicate that it is new to him. the existence of a horse-troug this village. and has the choice of either ing so. the fact that Irish Sea has been swum) may be complete news to the hearer. but in fact it is much m pervasive. The examples of definite usage which f 3 An I didn’t know . the woman he went out with last n was nasty to him.

a fact is “manifest” at a given time to a perso he/she is capable at that time of representing it mentally and accepting its rep sentation as true or probably true. referent is commonly not in any P-set. claiming that much the meaning of the articles is a matter of pragmatic inference. for example by overt denial of their import.him to locate the referent. Implicatures are two kinds. But in addition we have a kind of use in which this knowledg quite absent. Hawkins (1991) partly backtracks on the inclusiveness claim of his earlier work. 264 . claiming however that it does hold of demonstrat and personal pronouns – implying that these are not related to the definite arti He frames his account now in neo-Gricean pragmatics. depending on the indulgence or cooperative ecort of the heare is not certain that the distinction between these two kinds of use can be explai by Hawkins’s account. Thu (27) 4 I didn’t buy the house. which remains essentially the same. the reference can be guaranteed be successful. while conversational implicatures dicer in being “defeasible” – is. and indeed in which the hearer may not even recognize the desc tive content of the definite noun phrase. Hawkins (1991) updates his theory and introduces some modifications. He again rejects identifiabi as a criterion for definiteness. But Hawkins argues there is a prefere for interpreting indefinites as involving “P-membership” where possible. A entails only existence and conventionally implic non-uniqueness.4 and carries a conventional im cature that there is some P-set accessible to speaker and hearer within which e tence and uniqueness hold. it carries in addition other conversational (cancellable) impl tures. the entails existence and uniqueness. because a window was broken. Conventional implicatures are a fixed element in the meaning of expression. as expressed in concept of implicature (Grice 1975. he now labels “pr matic set” or “P-set”. He dismisses counterexamples to inc siveness like (19)–(20) as marginal. Now. With first mentions. 1987). Levinson 1983. to account for (among other things) the fact that membership of a P-set so times holds with indefinites but not always. and characterizes it in terms borrowed from Sperber Wilson (1986) as a “mutual cognitive environment” – a set of facts manifest equ to the individuals concerned. at same time reacting to various criticisms. they can be cancelled. and limit discussion to singular count cases so he can speak of uniqueness. Hawki concept of “shared set”. are straightforward. claiming that the door in (19) is shorth for the door which is open – but see Larson and Segal (1995: chapter 9) for cussion of the weaknesses in this ellipsis treatment. in particular. And these uses may or may not succ as references.

a in case would implicate non-P-membership. We have also seen that the uniqueness clause can be reformulated as in siveness or totality. Bu definiteness is one of several phenomena important for understanding existen sentences. the two im cations can be treated in terms of the logical concept of quantification.3. representing respectively existential quantification and unive quantification (of which we have already had some discussion in 4. as seen in Chapter 6. In fact. Holmback (1984). developing the claims of Rus (1905) outlined above.3 Definiteness and quantification We saw in 7. 1979). equivalent therefore to one of the windows. Thus a can conversationally impli P-membership provided the referent is non-unique in the P-set (and the there inappropriate). however. Chesterman (19 7. Where the referent is unique in a P-set. indefinite determiners being cardinality expressi . Declerck (1986a). Behaviour in existential sentences is often taken as crite for definite or indefinite status. the must be used. The existential quantifier is traditionally taken to be part of the representa of indefinites (at least specific indefinites) as well as definites. so on this view is the universal quantifier which distinguishes definites.2 and 6 For clear introductory discussions of the predicate calculus and of quantifica the reader is referred to Allwood. to cancel this im cature by adding something like not one of the windows in the house in quest but one in an identical house on the estate. Andersson and Dahl (1977).1 Milsark and the quantification restriction The nature of definiteness is not Milsark’s central concern. this sentence type has since Milsark been taken as central to un standing definiteness. 7. Neale (1990) is a good discussion and defence of the qua ficational approach to definite descriptions. Given this. It is possible. With this interpretatio implicates non-uniqueness in the P-set. with the desirable consequence that plural and mass defi descriptions are accounted for as well as singular ones. McCawley (19 and Cann (1993).1 above that the traditional logical view of definite desc tions takes them to imply the existence of a referent and the uniqueness of referent.the house. Milsark breaks w most earlier logical analyses in claiming that only definites (and some use indefinites) are quantificational. The log notation of the predicate calculus provides the two operators usually symbol ∃ and ∀. Other work using Hawkins’s framework or sharing many of its assumptions inclu Clark and Marshall (1981). This idea has given to a huge literature on the relationship between definiteness and quantificat The starting point is the work of Milsark (1977.

most people. this is correct. If is correct.But this would mean that not only all and most. every. second. A quantifier can be thought of as making an assertion about s Thus.2 Generalized quantifiers The theory of generalized quantifiers represents a major departure fr the assumptions of the predicate calculus. the existential and the universal. Outlines are also given by C (1993) and McCawley (1981). a window in (27) above see to be quantificational too. dominant since the work of Frege. the corresponding strong form is o which can be quantificational. some/sm. the non-universal quantifier most. My feeling is that numerals belong rather in the sec group. The initial claims of the theory are. first. complex partiti like ten of the. says that the set of things satisfying φ(x). quantifiers. and. the are not quantifi but determiners. both. universal quantification. many. The second group is those which are ambiguously quantificatio or cardinal: few. words like all. that is. develop work by Mostowski (1957) and Montague (1974). The ma presentation of generalized quantifiers is Barwise and Cooper (1981). The sentence Three men were shot could imply that others were not. each. but it seems coun intuitive. In generalized quantifier theory. and often only one reading is available in a particular syntactic con (like existential sentences). wh makes available only two quantifiers. The third group is those which are cardinal only and perhaps numerals. so that all men. that quantifiers expressed by many natural language determiners (like most) canno expressed in terms of ∃ and ∀. consisting o determiner plus a set expression. ∀xφ(x) (corresponding to.3. To summarize the relevant portion of Milsark’s findings. But note that the criterion of quantification that defines “strong” d not just apply to universal quantification – most is not universal. the girl are. determiners fall i three groups with respect to the quantificational–cardinal distinction. On the other hand. are definite behaviour in existential sentences is criterial. having propert (the set of mortal things). nor are str uses of some. and this may be related to always being weak morphophonologically. for example. all. quantifiers correspond directly to noun phrases. most. The first gr is those which are clearly quantificational: the overt universal quantifying te the. but also some and many w used quantificationally (in the sense of expressing a proportion). existential quantificat 266 . contains all individuals. sem tically. the two readings can generally be distinguis by stress. All m are mortal). many. it may be that only a is purely cardinal. so perhaps there are no purely cardinal determiner 7. that the logical structure of natu language quantified sentences does not correspond to that of the predicate cal lus. it is in fact much closer to their syntactic structure.

respectively. The VP sets which. contrasting with the acceptable one of the two men – an unexpected contrast s they take both to be definite and equivalent to the two. these responding closely to Milsark’s strong and weak but defined dicerently. necessarily false. no). If it is contradictory. Many men are men is true only in a model in which there are many m Definite determiners are considered to be a subset of the strong set. therefore. when combined with a quanti yield a true assertion are taken to constitute the denotation of that quantifier. all of many men is im sible as Barwise and Cooper claim. the determiner is “ne tive strong” (neither). true in every model. If noun phrases are thus generalized quantifiers. . Thus Every man man is tautologous. Neither man is a man is false in ev model. for example. some men in those domains where there are. including proper names. The is a sentence of the form Det N is a N or Det Ns are Ns. and the denotation of Harry is the famil sets containing Harry. A quantifier which does not allow either of these possibili 5 This is not unproblematic. set of sets) for wh it yields the value true. a quantifier denotes the “family” of sets (that is. recogn by their ability to occur in the diagnostic partitive environment all/most/so many of –.5 Barwise and Cooper regard the. respectively (the set constituting the denotation of the VP in each case). men) or to be all sets (every man or most men in those domains wh there are no men). both. since it is not certain that. All noun phrase types. the. Thus the sentences All are mortal and Some men are bald will be true if the set of mortals and bald p ple. The answer is surely that both doe mean ‘the two’ but rather ‘all two’. With some quantifie is possible for the family of VP sets corresponding to the value true to be em (many men. This assertion may be true or false. If its truth is contingent on the “model” (the domain of course).ness) is not empty. thus Harry has a cold is true if the VP se things having a cold contains Harry. c tains all men or some men. three. For a given determi if the sentence is necessarily true the determiner is “positive strong” (every. that). I have said that the denotation of a quantifier is taken to be the fam of VP denotations which combine with it to yield a true sentence. another way. The quant can be said. determi are functions from noun denotations (which are sets) to noun phrase (or gene ized quantifier) denotations (families of sets). They also note with puzzlement the impossibility of of both men. demonstratives. Barwise and Cooper distinguish “strong” and “weak” determiners. the determiner is weak (many. Generalized quantifiers are sai “live on” the sets representing the denotations of their constituent nouns. are ta to be uniformly quantificational. some. they are the determiners which form quantifiers which are alw “sieves”. not m or not any. to “sift” VP denotations into those for which it gives value true and those for which it gives the value false. a. and both as defin More formally.

(29) Everybody found a cat and kept it. within the opa context set up by the verb want. But when defin the N and both N are always sieves. which should be impossible if anaphora is about hav 268 . This is a formal semantic acco incorporating many pragmatic aspects of interpretation. so that it is restricted to use in singular count noun phra This is inadequate of course. This suggests that Barwise and Cooper ( Hawkins) take uniqueness to be a special case of inclusiveness. in accorda with the logical tradition. T completion of the and both is every. closely related to the “ course representation” theory of Kamp (1984). that is. and a number of the papers in Reul and ter Meulen (1987). Barwise and Cooper treat the in terms of uniqueness. given that Barwise and Cooper assume a uniq ness account of the meaning of the.a “proper quantifier” or a “sieve”. But they also state that corresponding to any de miner which can be undefined there is always another determiner which is “completion”. is undefined unless there is a unique man the domain.3. She characterizes definiteness in terms of traditional concept of familiarity: a definite is used when the referent is fami at the current stage of the discourse. they be “undefined”. To circumvent the major problem facing accounts of definiteness based on noti like familiarity and identifiability. Heim argues that neither defin nor indefinites are quantificational.3 Discourse semantics: Heim A very dicerent approach to definiteness is proposed in the very in ential “file-change semantics” of Heim (1988). This is the definition of definite quanti intended to capture the idea that definites presuppose the existence of the re ent. Yet the definite pronoun it is able to relate b anaphorically to a fish. a fish is most naturally understood as non-specific. Consider the following: (28) Joe wants to catch a fish and eat it. both men is only defined if there are exactly two men. 7. Thus the man. and an indefinite is used to introduce a no referent. Some quantifiers can fail to denote any se a given domain (which is distinct from denoting the empty set). so the and both are definite determiners. Heim appeals to Karttunen’s (1976) concept of “discourse re ent”. which is semantically equivalent to it but is always defined. Some important studies making use of generalized quantifier theory are de J and Verkuyl (1985). namely that many definites and indefinites non-referential. It ran away. Keenan and Stavi (1986). As noted. In (28).

Heim assumes a grammatical level of logical form. representing the information generated in courses. pay considerable attention to the nature of the relationship between file c and referents in the world. t its existence does not extend beyond the scope of this quantifier. Establishing the truth of a file involves set the sequence of file cards against a sequence of actual individuals in such a that each individual matches the description given on the corresponding card. It is essentially this concept that Heim takes up and de ops. When a new disco referent is introduced into the conversation a new card is added to the file. like Karttun discourse referents. It is c that discourse reference is something quite distinct from real-world reference. Heim does. Definites and indefinites are claimed to acect the growth of the file in dic ent ways. If everybody is taken to have wide scope in rela to a cat. The sentence of (29) is similar. and update an old card for each defin But it is file cards. as expressed in an “appropriateness condition” to the ecect that the he must add a new card for each indefinite. not real referents. Heim’s file cards can fail to correspond to referents. on this card is entered whatever is said about this discourse referent. The gene ization is that if a discourse referent is set up within the scope of a quantifier. are characterized as true or false. w dicerent patterns of behaviour. a quantifier has scope over its c-command domain. whatever new is said is added to the s card. the latter is then interpreted as a variable bound by the unive quantifier and does not refer. S a sequence of real individuals is said to satisfy the file.sets up a “discourse referent” which can then act as antecedent for it. depending on whether or not they ac rately represent the real-world facts. it is at this level of representation that m . and discourse referent is familiar by the time it is picked up anaphorically by it the definite pronoun is appropriate. or file cards can correspond to the same referent. Files. The idea is that understanding a discourse is like keeping a file in which e discourse referent is represented by a numbered file card. but a cat introduces a discourse referent. This is beca the discourse referent created by a cat ceases to exist at the end of the first tence and is no longer available to be picked up by the second it. But notice that the it of the second sente of (29) cannot relate back anaphorically to a cat (or to the first it). but this abstract concept does appear to be nee to explain instances of apparent anaphora as in (28) and (29) where there is real-world reference. essentially as advance work on government-binding theory. and so on. and the file is true sequence can be found to satisfy it. that are novel or familiar/old. If the sa discourse referent is mentioned again. using the metaphor of a “file” to express the information built up in the co of a discourse. never less. in terms of s tactic configuration.

At logical form. the expressions over which they have scope. and the r clause e left (where e is the trace of the raised subject) as sisters to each other. leavi trace behind. or by a def implicit quantification which provides the clause with “existential closure” (si the truth of a file depends on the existence of referents satisfying it). example. The indeterminate noun phrase remaining is t treated as an argument in the scope of the operator every. in particular. result is a tripartite structure with the quantifier every. and indeed have no semantic content at all. the quantifier every 6 is raised out of the n phrase to become its left sister. this condition says that for a sentence with a given l ical form (which includes the referential indexing of all noun phrases) to be app priate with respect to the current file. which can subsequently be bound by a quantifier having sc over the noun phrase. But ot definite uses are more problematic. These non-quantificational noun phrases con a free variable. and the index of every indefinite n phrase must be new to the file. by a quantificational adverb like usually. as traditionally supposed by logicians. In this respect Heim’s account of defin ness has much in common with the earlier one of Kempson (1975). and other uses (p cipally situational and associative) to be derived. the rump subject man. They do quantify. discussed abo Both take discourse-anaphoric uses of definites to be central. Their role is to mark noun phrases as definite indefinite. Heim extends her treatment to include definite reference to visible entitie the immediate situation. specific indefinites are understood as exist tially quantified. the index of every definite noun phrase the sentence must be already in the file. expressions which have the status of logical operators. 270 . corresponding to the syntactic no of determiner.7 But this quantifier r ing does not apply in the case of the and a. which are not operators. the rules of semantic in pretation simply ignore them. For associative uses. are raised to an adjoined position in which t are sisters to their arguments. these are in ecect treated as being in the file. So a sentence like Every man left would involve two raising and adjoining proce the subject every man out of the clause.cated. in the case of every man. for the purposes of the appropriateness condition mentioned abo Somewhat more formally. not in the generalized quantifier sense in which it corresponds to noun phr A quantificational noun phrase is itself raised out of its clause and adjoined to it. often. An important sequel to Heim’s work is that of Diesing (1992). So H does not deny that. she introduces an acco modation mechanism which connects via “bridges” the new file cards created other cards already present in the file. Her point is that the e tential quantification is not inherent in the noun phrase (or the determiner). This will no considered closely here because it is largely limited to the analysis of indefini 6 7 Heim uses the term “quantifier” in the traditional sense. and the quantifier every out of the noun phrase. incl ing quantificational determiners.

based on Sperber and Wilson (1986) and represented in a la volume of subsequent work. in ec utterances come with a guarantee of their own optimal relevance. and utterance is said to be relevant to the extent that it achieves contextual ecects. and Wilson (1992) discu the assignment of reference to definite noun phrases. see Kamp and Reyle (19 7. Diesing claims that indefinites are not uniformly represented variables. background assumpt (the context) and combine these with the logical form encoded by the uttera to yield the proposition or propositions which constitute the intended interpr tion. The hearer’s task is to access further.4 Relevance theory The most influential development in pragmatics in the last few y is relevance theory.tactic representations play in the formation of the kind of logical representa Heim proposes (with the tripartite structure noted for quantified sentences). is o mally relevant if it leads to substantial contextual ecects and causes the heare gratuitous ecort in achieving those ecects. in the form propositions. I will look at these two s ies below. For general accounts of the theory see Sperber Wilson (1987. For discussion of definite descripti in the framework of discourse representation theory. Th claimed to be a process of construction of mental representations. the amount of ecort is determined by the c plexity of the utterance. There is not a great deal of published work on the na of definiteness in the relevance theory framework. but that some are inherently quantified. but what there is merits at tion. This pa tion of indefinites corresponds to the distinction between cardinal and quantifi tional indefinites drawn by Milsark (1979). as argued by Heim. A basic claim of the theory is that the concept of relevance is central to hum cognition. We expect relevance in utterances we have to interpret. 1995). In brief. then. These three possibilities are contextual effects. In course of this. so that. It follows f . The idea is that the grammatical form of a sentence uttered enco a logical form which considerably underdetermines what the speaker actually wis to communicate. contradicting and therefore eliminatin or combining with it to yield an implication not derivable from either contex new information alone. An utterance. The interaction of the newly given information with the context may amo to confirming an existing assumption. the accessibility of the context. Kempson (1988) gives a relevance theoretic account of a range of type simple definite noun phrase and personal pronoun. the relevance of an utterance is reduced by the mental ecort required of the he in deriving these contextual ecects. relevance theory is an account of how we interpret utterances. and the complexity of inferencing required to reach the intended implications.

that the geography tures were less boring than the linguistics ones. In bridging cases like (31). it may be that one potential antecedent rep sentation is more accessible (that is. would achieve relevance either explaining why the speaker switched subjects or as an account of what she fo after switching subjects. The lectures were boring. the two windows the room mentioned. by way of the encyclopedic knowledge that rooms can h windows) and this will be accepted as the correct referent if it meets the requ ment of relevance (as would be the case here). but the correct bridge. (32) I switched from linguistics to geography. and combined instances like (32). The bridges available are the propositions (derived from encyclopedic knowled that studying linguistics involves attending lectures and that studying geogra involves attending lectures. (30) The room had three doors. in (30) the mental representation of the open door wo certainly be more accessible because it has already been singled out and m said about it than the others. one of which was the hearer as satisfying his expectation of relevance in a way the speaker co manifestly have foreseen is to be taken as the correct interpretation. but there is more than one po tial bridge or trigger for the association. where not only does the re ent have to be inferred by bridging or association. where there is more t one potential antecedent for the definite. the hearer to find a maximally accessible context which will combine with one represe tion to yield worthwhile ecects and an acceptable interpretation at no unjus able processing cost. Assigning refere involves constructing or retrieving a mental representation of the referent. optimal relevance is achieved by taking the o door to be the antecedent of the door. Both windows were open. In the cases exemplified by (30). because the resultant interpretation. hence examined first. I closed the d (31) I walked into the room. The second of these would be the correct choice trigger for the lectures. 272 . has to be inferred. too. In (30). the m easily inferred representation is the most accessible (here. where the referent is not already given at all and to be inferred. salient or prominent) than the other(s). But whether or not this is the case. The mixed case (32) involves same process of inference. bridging cross-reference (associa anaphoric) cases like (31). Wilson looks at anaphoric uses like (30). Wilson (1992) argues that a hearer’s task of establishing the reference definite noun phrases is part of the process just described. and representation is included in the representation of the proposition expressed the utterance.

This is what is involved in bridging cr reference. and s information contributes to the context. it was subseque reported that the house was sucering from subsidence problems Quantifier–variable dependencies are subject to configurational requirements. so the definite both windows m be taken as a guarantee of accessibility of a contextual premise such as ‘The ro had two windows’. including associated information. The most strai forward cases are immediate situation and anaphoric uses. for instance. this is clearly a pragmatic matter. But the dependency between the accompanist and ev singer also depends on the construction of the additional premise ‘Every sin had an accompanist’. possibly a dicerent one for each singer). in which the relev easily accessible information consists. But Kempson observes further bound-variable anaphora can interact with bridging cross-reference: (35) Every singer complained that the accompanist played too loud (The relevant interpretation is that each singer complained about his or her o accompanist. Kempson establishes that many of the distincti found in pronoun uses apply equally to full definite noun phrases. as in (33). Kempson (1988) pres a similar view more fully. This requ ment is claimed to be met in (34) as in (33). As already noted in 1. It seems that bound-variable anaph is not always fully determined by grammatical principles. bound-variable use of pronouns. Where the situation and the explicit content of the discourse do not p vide the representation required. is paralleled by examples such as ( (33) Every boy worries that he’s inadequate. In particular. Again the configuratio requirement is met. of the hearer’s perception the scenario in which the utterance occurs and the content of the preceding course. On the basis of t . the concept ‘window’ consists in part of the in mation that windows are a feature of rooms. Definiteness amounts to a guarantee of accessibilit speaker uses a definite noun phrase to indicate that a conceptual representa corresponding to the noun phrase is easily accessible to the hearer.5.2. The concepts expressed by wo are claimed to be stores of information. In (31).definite noun phrase means retrieving or constructing a conceptual represe tion which uniquely identifies the intended referent. Specificall quantifier must c-command a variable in order to have scope over it. derived from the contextual knowledge that singers may h accompanists. (34) Of every house in the area that was inspected. the hearer must assume that this representa can be easily inferred by way of the context. wh are specified at some syntactic level (S-structure or logical form). respectively.

what is definiteness then? It is clear from the foregoing survey of accounts of definiteness fr various theoretical perspectives that versions of the two basic criteria introdu in Chapter 1. For an application of these ideas to definiteness in a particular language ot than English. This type of non-referential use of definites is thus claimed to involve same process of interpretation – the accessing of a representation – as the straig forward situational and anaphoric uses. pragmatically ad variables. To account examples like (35). These concepts h undergone considerable mutation. writers almost inv ably choose one or the other of them and claim that this one gives the cor account. and is used whenever the spea has reason to suppose the hearer is in a position to access a conceptual repres tation corresponding to the intended referent.5 Well. however. see Blass (1990) on Sissala. In this example it quantifies over ordered pairs of singer and accom nist. is that no one has shown conclusively that a v sion or mutation of either identifiability or inclusiveness accounts adequately all definite uses. in so far as we still say that identifiability and inclusiveness have persisted. is a pragmatic matter. is v dicerent from the simple version of this we started with. Blass sees determiners generally as h ing the function of guiding the hearer towards appropriate conceptual addres so as to establish the intended interpretation of an utterance. therefore. it is necessary to assume that a quantifier can bind not o the variable assigned to it in logical form. The reality. The question with which we began the chapter – identifiability or inc siveness? – is. but also additional. Some uses still seem to yield to only one or the other char terization. The Sissala defi article ná behaves essentially like English the. But. perhaps most radically in the case of the r vance theory account which. The proposal is that definiteness stricto sensu is not a semantic or pragm 274 . still unresolved. In this section I propose a solution to this problem. including the binding variables within this proposition. based on the suggest made earlier that there is a distinction to be made between grammatical defin ness and semantic/pragmatic definiteness. identifiability and inclusiveness.tion of the proposition to be conveyed by an utterance. The grammar can thus imp restrictions on the pragmatic process of proposition construction. and on the concept of grammatical tion. 7. and this contains the quantifier and an associated variable. But the representati accessible to the hearer in this process include the logical form specified by grammar. The latte also a representation. and is constrained by the logical form to be accessible o within the c-command domain of the quantifier. though obviously derived from identifiability. keep recurring.

’d.category on a par with tense. 7. so that the words acected develop into grammat forms such as aaxes or inflections. it is grammaticalization (that is. v stage at which the reflex of this verb was an auxiliary. and coming to express instead so grammatical concept. can. but related to. The second and third of these possibilities. either by other items or by the reduced item it before its grammaticalization. grammatical category or concept which comes to be expressed by a grammati ized form may already be present in the language with a dicerent exponent (wh is perhaps ousted by the new form). the representation in grammar) of some category meaning. This semantic reduction is usually accompanied by m phophonological reduction. or perhaps free-form but non-lexical wo In some cases it is a matter of open-class items coming to form a closed cl usually with accompanying phonological reduction. This typically involves lexical items undergoing “sema bleaching”. would. and then disc how this can be applied to definiteness. loss of part of their meaning. derived from Latin habeo ‘I have’. But. this is the case with the English modals mentio above. that most common in the literature. must e usually unstressed and showing further reduced forms like ’ll. see Hopper Traugott (1993) and the studies in Traugott and Heine (1991).5. therefore. number. that a concept which is 8 A recent exception is Lambrecht (1994). the evolution in so languages of nouns used in counting (like English head in three head of ca into non-lexical classifiers.1 Grammaticalization I am using the term “grammaticalization” in a sense somew dicerent from. gender etc. The term is u mainly in diachronic studies to refer to the process by which lexical items are redu to grammatical status. the development in Romance languages of a future tense paradigm with endings like Spanish -é French -ai in the first person singular. Examples are: the crea in English of a closed class of modal auxiliaries (will. Let us. and perhaps further into gender morphemes. And the crucial observation here is that the correspondence betwee grammatical category and the category of meaning it is based on is never one one. . like these. or it may be new to the language. A t possibility is that something of what the new grammatical form expresses already conveyed lexically. mood. The remaining chapters of this book then be largely devoted to developing the proposal. examine the concept of grammaticalization. from what w originally lexical verbs expressing modal concepts. whose view of definiteness is close to that adva here.

A few examples will make this clear. respectively. to continue to the prototypical value of the grammatical category. the moment of utterance. present and future ten one might expect the present to be used in describing events or states simulta ous with the utterance. It is generally the case that grammatical categories are not direct expressi of the semantic/pragmatic concepts which they can be said to be the gramm calizations of. Tense is usually thought of as the grammatical category which expresses t distinctions. present and future. in a classic study of ten Reichenbach (1947) points to the possibility of reference points distinct from. And for the most p this is likely to be the case. anterior to the utterance). This is what makes it p sible to speak of the category as synchronically the grammaticalization of concept. themselves defined in relation to. But these complications do counter the basic point. Another complication is that m languages. including English. I will therefore also use the term in a diachronic sense to denote development whereby a concept of meaning comes to be represented as a gra matical category. much more serious diaculties confront any atte to account for tense in terms of the expression of relative time distinctions many languages. In English. so that the category can be seen as expressing that concept in its central uses. I use the term essentially in a synchronic sense. But note that the situation described by this use “grammaticalization” commonly comes into being through the diachronic proc described. make link with my use of the term “grammaticalization”. with the result that the grammatical category created is not limited expressing that concept. When a concept comes to be represented grammatically it takes a new life. Thus.cally. The dicerence between meet and met has nothing to do with any tempo 276 . The original concept is likely. However. the past and the future to be used in describing event states anterior and subsequent. p and non-past rather than past. comes to be expressed by one or more grammatical morphemes. to denote the representation a grammatical form or forms (and thus with the status of a grammatical catego of some concept of meaning. for example in conditional contexts where unreality to be conveyed. to the utterance. it is possible for a past tense form to be used in describin present or future event. thus. This simple picture is complicated by what migh thought of as secondary temporal relationships. If I meet John tomorrow he’ll be surprised and met John tomorrow he’d be surprised both relate to some possible meeting ye occur. a p perfect form like I had left describes an event anterior to a reference point wh is itself past (that is. however. in a language which distinguishes past. For example. make only a two-way basic tense distinction.

1 that these languages have a grammatically singular first-person in sive pronoun. it captures. with the meaning ‘you and I’ – thus semantically dual. like trou and. as should be possible if trousers really expre ‘more than one’.9 Subjunctive mood is often said to indicate that the action or state expressed a verb is being presented as hypothetical or virtual rather than factual. consider number. we can expect that. though there may be no element of hypotheticality in the clause: B qu’il soit malade. But languages vary greatly in the use of the subjunctive. Of course the origin of this phenomenon lies in the fact that trousers and scissors have a part structure. But there are complications. Thus John’s trousers are grey could be a statement ab six garments John owns or about a single garment he is wearing at prese Moreover. he has already arrived To take a nominal grammatical category. like most grammatical categories. . the contras Spanish between aunque + indicative ‘although’ and aunque + subjunctive ‘e if’. In formal regis of German it is used for reported speech – a striking dicerence from Spanish other languages. this category grammaticalizes the distinction between ‘one’ and ‘more t one’. 7. but this does not acect the point. If nouns for garments with two legs. which are plural even when use refer to single objects. Again. of which remoteness in (past) time and factual remoteness are manifestations. such as that presented in English by plur tantum nouns of the type trousers.2 Grammaticalization and definiteness Definiteness is. 9 10 It might be argued that the English past tense has a very general core meaning of “rem ness”. il est déjà arrivé ‘Although he is ill.1. sugar occur in a morphologically sin lar form though not denoting single discrete objects. overalls (which have a substantial trunk section as well as legs) are pl why not those for garments with two sleeves (shirt. and prototypically expresses these. In singular–plural tems. This is discussed and rejected by Palmer (1974) and Huddleston (1984).’ll) expresses (something like) the view that the meeting is unlikely to occur while it may be fair to say that tense is the grammaticalization of relative t distinctions. though gr matically singular. a grammatical category like those discussed. for example. though grammatically count and plural (like oats). in other words.1 and 3. mass nouns like water. And in French the subjunctive is obligatory after bien ‘although’. I suggest. mud. Given this. they are not semantically “singular”. A good case could be made for regarding nouns of this type as semanti mass. there is clearly much more to te than this. recall from discussion of number in the personal pronoun systems of Carib and Lakhot 3. Finally. scissors.4. characterization works to a large extent. pullover)? And note that we canno *a trouser to speak of half of a pair of trousers. better still.5.

The point h is that this semantic/pragmatic concept occurs widely in languages which lack corresponding grammatical category. this article too can be treated as encoding defin ness. when identifiability comes to be grammaticalized as defin ness.4. this category will go on to develop other uses. to with whether or not a referent is familiar or already established in the discou – thus identifiability rather than inclusiveness. in so languages definiteness will be optional even in noun phrases clearly interpre as identifiable. This assumption is bolstered by observation that demonstratives. Putting this poin diachronic terms. The ecect of these possibilities is that (as with the subjunctive) th will be considerable variation between languages in the use of the category. as was seen with Hausa in 2. in this case certain types of noun phrase occurrence which in other langua are treated as indefinite are grammatically definite. Definiteness is the grammaticalization what I have informally termed “semantic/pragmatic definiteness”. where it was seen that a “definite” interpretation plays an important part in m languages which show no formal marking of definiteness. In languages where identifiability is represented grammatically.2. This is clear from the discussion in Chap 6. 278 . In languages like Maori wh as seen in 2. It may be that identifiability is an element in interpretation in languages.2. invariably treated as definite in interpretation in (in)definiteness ecect conte Taking this as the clue. Thus in Mandarin a n phrase in subject position must be a topic and therefore “definite”. But it is to be expected that there will be other uses of definiteness wh do not relate to identifiability – inclusive uses for example.1. show an article combining obviously “definite” (identifiable) u with something akin to specificity.which show overt definiteness marking. But w is a definite or indefinite interpretation? It appears that in these languages w no definiteness marking it is. while a n phrase in the existential construction must be understood as “indefinite”. but no greater than that found with many other grammatical categor What justifies us in identifying the same category cross-linguistically despite variation is that there is always a large central core of uses relatable directly identifiability. but in many languages it is not grammaticalized. T some languages will require generics to be definite while others do not. There may also be some instan of identifiability (generics for example) which are not treated in a given langu as definite. this repres tation is definiteness. but not full. and definiteness is likely to express identifiability protot ically. let us say that definiteness is the grammaticalization identifiability. which cannot be characterized as inclusive. as an element of discourse organization. a definite article of some kind – tho I will develop below the qualification that many languages have definiteness pronominal. The range of variation is c siderable. noun phrases.

But weak personal pronouns (including null pro) are the pronom correlate of simple definite noun phrases. Japanese. there is no doubt that noun phrases with a demonstrative mod are definite. m over.1. adapting the an sis of Postal (1970). despite their not occupying the same slot in the noun phrase. and DG possessives in turn.3 that dem stratives are not lexically specified as [+ Def ]. This is enough to ensure that in guages with definiteness this category is triggered by the presence of a dem strative. in many such languages. It may be. typically with bound articles. where they have considerable phonological substance. But demonstratives occur in all languages. presumably a pragmatic c cept or an element of information structure (see Lambrecht (1994) for deta discussion). it therefore itself. that the forms functioning as personal pronouns in some languages. definiteness does not. are actually noun phrases. But in other languages only the pragmatic concept of identifiabilit involved and there is no definiteness with demonstratives. by its position. where the article which enco the category is in many cases not present? Let us consider demonstratives. I have claimed that strong personal pronouns are demonstratives.3 other cases demonstrative and article do not co-occur. where I will argue that the demonstrative is in a specifier position a ciated with the head position normally occupied by the article. Indeed I have argued. Treating defin ness as a grammatical category seems fairly straightforward for simple defin where the category is encoded by aaxes. In fact. sonal pronouns. In languages showing definiteness marking (thus having the category definiteness). but the demonstrative Det like the article and takes its place. indic definiteness as well as demonstrativeness.The proposal just outlined gives an answer to the question whe definiteness is universal or not: while identifiability. the demonstrative either must or co-occur in the noun phrase with the definite article. clitics or morphophonologically w free forms – typical grammatical or “functional” morphemes. may play a role in all languages. The problem here is that w pronouns do occur in languages which otherwise lack definiteness mark . as was shown in 3. But what consequen does this treatment have for complex definites. there are languages. so demonstratives bring the pragm concept underlying definiteness into play. Suppose they are marked [+ D and that this feature is interpreted as meaning that the speaker is pointing ou otherwise providing suacient information for the hearer to pick out the refer It follows that the referent is identifiable. Th cases can only be accounted for when we come to examine the DP analysi Chapter 8.1. Less straightforwardly. including those lacking definiten and here we must come back to the tentative suggestion made in 3. in which articles and demonstratives not co-occur. that they are definite articles.

1 following Cole (1987) ). again like dem stratives. for example. like demons tives. even pragmatically. If pro is simply a null weak perso pronoun. A pro appears in a great many languages which otherwise have no definite artic To account for this. I have made the point several times that possessives. given that it has no demonstrative content. as shown in 5. but it is unusual for pronouns to dicer radically in structure from full noun phrases many Australian languages. where DG phenomenon comes from. Thus Spanish canta ( pro sings) can be ‘he is singing’. In all languages in which a n argument is identified by agreement.4.4 and in a more limited way in other languages. I propose that many languages have the category of defin ness in pronominal noun phrases only.5. it should allow an indefinite (non-identifiable) in pretation in appropriate contexts. which (as one would expect) probably only occur in langua which have definiteness. display pro identified by agreement morphology. Unlike demonstratives. commonly displace the definite article.3. I shall return to this p nomenon in Chapters 8 and 9. they are frequently aaxal. for example. they may also occupy a dicerent position from the article. since if it were not. see 3. unlike demonstratives. But. One may wonder. so may displace aa articles as well as free-form determiner articles. We can therefore set up a typology of guages as follows: Type I: Type II: Type III: no definiteness definiteness available only in pronominal noun phrases definiteness available in pronominal and full noun phrases Languages of Type II certainly represent an odd phenomenon.enon of languages. not inherently definite. ‘she is singing’ or ‘it is singing’. then it is in ecect a null definite article. as discussed by Dixon (1980. it is interpreted as definite. chapters 9–11). therefore. And this does seem to the correct characterization. Importantly. Korean (see 3. I suggest it has its origin in the phenomenon inalienable possession.5. apparently without definiteness as a category. In these languages the null argument is proba not pro (but see the reservation noted in 5. we can say that pro (at least when associated with agreement morpholo is definite. These occur freely in.1. as in object position in Latin Ha (have+1SG) ‘I have it/one/some’. Again I will return to this in connection with the DP analysis of noun phrase. but ‘one/someone is singing’. yet not occur with it. Assum this. DG possessives. this is not the case with null arguments identified by agreement. they are organized on a dicerent c system. 280 . Recall that many languages (with and without definiten make a structural distinction between alienable and inalienable possession.

Recall also that in languages with definiteness. involving either a free-f possessor or an aaxal possessor in a position which could be that of a free-f or aaxal article. This structure. with the possessor treated as be in definite Det (or a related) position. is then eventually extended to other. wh the syntactic structure of the noun phrase is examined in greater detail. h nouns. .sessum. My suggestion is that when a langu acquires the category of definiteness. thus my hand. This is the topic of the next chapter. non-inalienable. even when the usual conditions for definite occurrence do not stri apply. no of the inalienable type show a strong preference for occurring in definite pos sive structures. Mary’s sister when it is not contextually clear which of hands or Mary’s sisters is intended. If definiteness is a grammatical category it is important to consider how represented in grammatical structure. the structure used for inalienable posses comes to be interpreted as a definite structure. including English.

modify nouns. I begin by examining the s tactic representation of definiteness in this older framework. particularly from the comparative sur of Chapters 2 and 3. the treatment of definiteness as a grammatical category will be ta a step further. is part of a m general theory of “functional heads”. we will n consider the representation of definiteness in syntax.8 Definiteness and noun phrase structure I have argued that definiteness is a morphosyntactic category.1 The noun phrase as NP Before the recent challenge of the DP hypothesis. Specifically. I will propose that the relationship between definiteness person is so intimate that it justifies treating the two as a single category.4 of what this. being therefore NP. With this in mind. Almost all current work on noun phrase assumes the DP analysis. that of p son. gra maticalizing a pragmatic category of identifiability. and I shall in fact argue that the categ of definiteness is itself such a functional head. but ra themselves head noun phrases. Other questions to be considered along the way include: What is the categ status of articles and other determiners? Are there constraints on the position which definite determiners may occur in the noun phrase? Are certain determin including definite articles. the DP hypothesis. combi 282 . or does defin ness arise in a noun phrase in some other way? How can the range of article ty occurring in languages be accounted for? How does the analysis of definite p nouns relate to that of definite full noun phrases? These are fairly obvious qu tions arising from our earlier discussion. as traditionally assumed. T reader is at this point referred back to the outline in 1. This view. Finally. but since much still important less rec work on definiteness is cast within the NP analysis. by examining how it relates to another such category. it was assumed the noun phrase was projected from the category N(oun). This means discuss recent developments in the theory of phrase structure according to which defi and indefinite determiners do not. specified in the lexicon as [+ Def ]. 8. attempting to relate it to the theory of functional heads. We will also look more closely at the phenomenon of n configurationality.

4. a position from which items can m further on. out of the immediate phrasal projection. NPs. this house. not just of NP seen as having three further roles. The specifier c-commands the rest of the phr and this structural relationship is assumed to be involved in the binding of v ables by a logical operator. The third rol the specifier is to close oc the phrasal projection. for instance. with an intermed level N′ between N and NP (or N″). and fact is part of the attraction of the DP hypothesis. and that at least some determiners (of categ Det) occupy the specifier position which combines with X′ to form the maxima phrasal projection NP. since it entails the spec 1 Recent proposals by Chomsky (1995). example because they cannot receive grammatical case in their original posit moreover. however. is taken to be applicable to configurational languages generally. It follows from these two fu tions that specifiers are commonly occupied by arguments. but specifier must be daughter to the maximal projection. This is essentially the function assumed for the specifie a lot of work in formal semantics. on either the NP or the DP analysis. other determiners and indeed possess occur (the house. Second. but the ove structure given in 1.points for present purposes are that N heads the structure. move into or pass through. it may act as an “escape hatch”. which treats determiners as he It is not necessarily ruled out. First. The intermediate X′ level m on some views. the specifier position is the target for the external θ-role (tha Agent. so it completes the phra This third role relates to the semantic function referred to.1. as we have seen. it is a p tion expressions may move into from positions deeper inside the structure. Therefore determiners which express quantificat including in some treatments. recur to allow multiple modification by adjunction. the specifier is a position where determiners may be generated as well as be one which arguments may arise in. however. defined in terms of what is sister to what and what domin what. Ann’s house) – items thought of as delimiting or defin the reference of the NP. in a phrase headed by a θ-role assig VP for example. In this sec we will discuss definiteness and definiteness marking in relation to this conc tion of noun phrase structure before going on to examine the rival DP analys 8. and in more theoretical work the specifier was first concei of as the position in NP where articles. Linear order may vary across languages. both definite and indefinite artic are taken to be specifiers binding a variable in the semantic representation of In current work in syntax the specifier (of phrases generally. that is. allow multiple specifiers in a single phra . which is typically assigned to the subject).1 The specifier position and definite Dets In many traditional grammars the term “specify” is used synonymo with “determine”.

all may be in specifier position in th examples. as Lyons (1995c) argues. the quantifiers are almost certainly in specifier position. most commonly on the left. either can or typically do occur in this position With all and both there is also the pre-determiner use in which. both your friends. since they may conta demonstrative. synonym with the alternative pre-determiner structures all the three defendants and b the lawyers. It should mentioned that there are occurrences of all where the analysis is less clear: (1) All cats like milk. We saw in Chapter 2 that one of the most common devices occurring in l guages for encoding definiteness is an item appearing in a peripheral position NP. in speci position. t express a proportion of the whole denoted by the complete NP following: all girls. like a partitive. for example. . a generally taken to be definite. interpreted as non-restrictive and expressing emotional distan and demonstratives only occur in the specifier in English: (3) 284 I’m all in favour of people cycling more. but this is perfectly possible with generics. but those mountain bi are a nuisance in the country. and that NPs like this sh my car are definite. But it is arguable that cats and pupils here are complete bare plu generic NPs with all as pre-determiner. generics of this kind are not incompatible w Dets (as the description “bare” implies they should be). But in all three defendants and both lawyers. We have also seen. that in English demonstrati and possessives occupy the same slot in NP as the.tant concept in recent research is a process of Specifier–Head Agreement. Given the lack of a second determiner. both. in wh specifier expressions of various kinds enter into an agreement relationship w their head – ensuring. and that in rogatives in specifier of CP (complementizer phrase) do not co-occur with dec ative complementizers. We can hypothesize that such definiten markers will generally have the same basic analysis (though bearing in mind this treatment cannot be simply assumed for a given case without serious anal of the language in question): that they are of category Det and occupy the specifier position. most. It is frequently phonologically weak. Yet even treating these n phrases as “bare” plural generics it may still be that all is within the NP. that agreement morphology on verbs does clash with the subject (which is a specifier) in person and number. pupils in (2) represents a generalizat over a more limited domain than cats in (1) (a particular school rather than world). and may may not carry inflectional morphology. every. in Chapter 1. The quantificational determiners all. (2) All pupils must assemble in the courtyard at noon.

are often found in some posi other than the specifier: Irish an leabhar sin (the book that) ‘that book’. In all the languages just examined. and it is plausible to suppose a dicerence of c gory between the two (though I shall question this supposition below). Corresponding to English all. This generalization does not. . hold for all languages. But this use is quite marginal. in Spanish the adjectival inflection and position of the AG posses contrasts with the more limited inflection of the alternative Det-like DG pos sive (mi amigo ‘my friend’). Summarizing. die beiden Kinder ‘both (the) children’. however. And this is more generally true. L (1994a). The sam true of other definite determiners. where posses expressions (in nominative case form) can follow the definite article and be th selves followed by demonstratives or other definite determiners: a Peter minden/e kalapja (the Peter+NOM every/this hat) ‘every/this hat of Peter’s’. Meiste ‘most’ and beide ‘both’ are similar in position to ganz: meisten Leute ‘most people’.(the normal case). though not invariably. But two further facts must be conside here. This is not surprising since possession does not imply defin ness. the article occupies a peripheral posi which is likely to be the same as that of English the. and that the Hungarian definite article occupies a position responding to the complementizer. The s holds of a range of determiners in Hungarian (Szabolcsi 1994). appearing further inside the and possibly adjectival: all die Frauen or (colloquially) die ganzen Frauen the women’. the ev whim of the king. demonstratives too. in which the Hungarian and English article occupy the same position. ocers a dicerent analysis. in all languages in which the definite art 2 Szabolcsi (1994) argues that the Hungarian noun phrase has a structure closer to that o clause than has that of English.2 and cannot occur lower in NP. German has close in behaviour and position. the general ture for English is that definite Dets occur typically. in specifier of NP. First. All th observations are made more fully in Chapter 3. Facts like these seem to tell against the hypothesis that the specifier position a special relationship to definiteness. a position more peripheral than that of English the. and in many cases they are proba adjectives. but also in some more interior position in the presence possessive and with a restricted set of head nouns: the king’s every whim. which do entail definiteness. So p sessives need not be definite determiners. and indeed an indefinite interpretation is possible in AG languages if definite article is not included: unos amigos míos ‘sm friends of mine’. the definite article itself is not subject to this kind of variation in p tion. but also ganz. Recall first AG possessives are in a distinct position from the definite article: Spanish el am mío ‘my friend’. Cat (optionally) els detalls aquells (the details those) ‘those details’. however.

as well as die beiden Kinder ‘both (the) children’. which requires definite interpretation. and the one unquestionably occupy dicerent positions. Thus Cata *detalls aquells ‘those details’. the definite article (or some other definite Det) must also app in what I am taking to be specifier position. demonstratives and quantifiers replace the. Assuming cardinality position to be daughter to N′ (and ado ing the arbitrary category label Q. in the languages considered above in which vari determiners associated with definiteness occur deeper in the NP structure (or o side this structure). The likelihood that a belong numeral position is all the greater in the many languages where the cardinal a cle and the singular numeral are identical in form. That is. which m be required by the presence in it of certain determiners.4) would give the three blue cars and a blue car the following structu NP Det the NP N′ Q three N′ N′ Q AP N′ blue N cars 286 a N′ AP blue . is not itself able to impose it when not in the right posit I have argued that quasi-indefinite articles like English a are cardinality exp sions. and I shall take this to be correct analysis. German. The essen thing is not simply that the definite article must be present. depends on the express in the specifier. recalling Bresnan’s (1973) QP analysis refer to in 1. German *meiste Leute ‘most people’. whereas in English defi possessives. a is in fact just as complementary to the numeral one. the demonstrative or quantifier lower down. in these other langua where such determiners occupy a dicerent position they do not replace the defi article but co-occur with it (or with some other specifier definite Det). Second. It is likely that the definiteness of the NP. The po about some other definite Det in the specifier is important. like the numerals and much/many which clearly occur in a post-speci position. also diese beiden Kin ‘both these children’ and meine beiden Kinder ‘both my children’.a peripheral position. While a and the not co-occur. for exam permits. as shown by the one problem. but that some Det m be present in specifier position. these many letters.

The basic rule is the free-form article is used when the noun is preceded by an adjective or o modifier. position of the lexical article den/det/de in relation to other pre-nominal modifi shows it to be closely comparable to English the. huset ‘the house’. Subject–verb agreem and possessive agreement both involve a relationship between an aax and an a ment in the specifier. depending on the i vidual language and on the theoretical position adopted. the same thing. this might involve invoking Specifier–Head Agreem so that the aax is seen as a marker of agreement between N and a null (in m cases) [+ Def ] Det in the specifier. this latter phenomenon (where it has a DG interpretation) is probabl be seen as a relationship between a possessive agreement aax and a null p nominal possessive in the specifier – exactly like the relationship between agreement inflection on a verb and a null pronominal subject. where an aax occurs on a h noun. in some languages. In both cases relationship may be seen as true agreement or cliticization.1. at least some other defi Dets. Where the aa attached to the head noun. it might be a matter of mo ment. Alternatively. which give some suppor the proposal. on the other h is a relationship between an aax and a Det in the specifier. Danish has both an aaxal article and a free-form article. or itself be the only possessive expression. One poss analysis is that there are two specifier definite articles. and the obvious way to relate this type to the present discussion is in te of a relationship between the aax and the specifier position. det store hus ‘the big house’.indefinite article though many have a cardinal article. as a specifier Det: de tre b ‘the three children’. one overt and one null. This dicerence by means rules out the idea. Which of th approaches is the more appropriate could only be determined through analysi the facts of individual languages. Thus hus ‘house’. It is plausible to claim that it is a position only definite Dets may occu 8. and both processes may occur. The first pointer to a relationship between the two article fo is that they are in complete complementary distribution in simple definite N This suggests that they may well be. at some level. a specifier article cliticizing to its head or elsewhere. Given that free-form DG possessives occur in specifier. either showing agreement with an overt possessive expression. .2 Affixal articles The second commonly occurring type of definite article is the aa one. The pheno non would be parallel to possessive agreement. but it should be pointed out that a lot less is kno about agreement and cliticization processes involving categories other than Let us examine the relevant facts of one language. then not only is the spec the position of the definite article and. What I am suggesting for aaxal articles.

on which again there are two specifier a cles. 198 who suggests that definite Dets are not. a lexical item. I will not attemp choose between these analyses. and is comparable to the subject–verb agreem morphology of Irish. T other analysis is the cliticization one. but would fol from the assumption that the aaxal article is syntactically related to the speci position. The lexical article cannot co-oc with a possessive. and why is it that only defi Dets occupy this position? An answer has been ocered by Lyons (1985. which is also only available when the subject is null. This is less expected. one is den/det/de. but more generally. But the point is that for some languages there is a clear case for taking aaxal article to stand in a syntactic relationship with the specifier position. in fact.1.of “identifying” the null Det. despite the definiteness of the NP: *dine børnene ‘your childr *mandens børnene ‘the man’s children’. mandens tre børn ‘the man’s three c dren’) and imposing a definite interpretation. it is legitimate to hypothesize that a similar analysis can be extended to other l guages. as expected. occupying specifier p tion (dine tre børn ‘your three children’. which must move to attach to the noun. the important thing is that both involve a synta relationship between specifier and head. we may suppose that the proposal of a special relations between the specifier and definiteness is valid not only for languages in which definite article is a lexical item. whether or not some kind of empty category remains cannot hold if the specifier is occupied by a possessive. because they occupy the same “slot”. although NPs modified by certain determiners must be defin 288 . Given this.1. but both are overt.3 The definite constraint This generalization gives rise to a question: what is the nature of association between NP specifier and definiteness.1 that demonstrati and other definite determiners in languages where they are in some non-speci position have to be supplemented by the definite article or some other speci Det to give a well-formed definite NP (in languages having a definite artic It seems that. and definiteness or indefiniteness is determined by synta structure. but nor the aaxal article. 8. Pre-nom possessives and genitives in Danish are of the DG type. lexically specified as [+ Def ]. The starting point for this claim is the observation in 8. th is no such feature. wh there is not complete complementarity between the aaxal article and speci Dets. The evidence goes further. The account would be more complex for a language such as Swedish. the other is the cl -(e)n/-(e)t/-(e)ne. Since this relationship involves the specifier being phonologically n on the surface (either because agreement is only with a null Det or because a cl Det moves from there.

Spanish has b DG mi ‘my’. So the suggestion is that the “inherent definiteness” of such de miners lies in their having a lexico-semantic content which is incompatible w indefiniteness in the modified NP. So we would not even consider marking such possessives [+ Def ]. Rather. and therefore be argued to follow from this. is not part of their meaning or grammatical content. But if the possessive is elsewhere. So the morphological dicerence does not necessarily point dicerence of category. This sug tion accounts straightforwardly for the necessity of the article in Irish an fear (the man that) ‘that man’. showing a clear morphological dicerence. tu ‘your’. a definite in pretation can still be achieved by putting some other item. But definiteness itself. The possessive facts make it clear that what occupies the specifier to yie definite reading need not be a “definite determiner”. the definiteness or indefiniteness of the NP has nothing to do with possessive. German die meisten Leute (the most people) ‘most p ple’ etc. A better account is to say that such determiners are not mar [+ Def ] in any language. Lyons propo that English my car and Spanish mi coche are definite because the specifier is fi by the possessive as it happens. the definiteness (which determiner requires but does not itself impose) comes from the filling of the speci and it may be the demonstrative or quantificational determiner itself which this position. and not otherwise? Th is a redundancy here.definite value. or este coche (this car my) ‘this car of mine’. AG possessives dicer from demonstratives in that the modified NP can indefinite. Again. But then why are they lexically defi only in those cases where they occur in specifier position. that definiteness in an NP arises from the filling of the speci . such as the defi article. thus Spanish el coche mío ‘my car’. nor English most from German meiste. and AG mío. post-nom adjectives. moreover that the two types do not necessarily dicer in category. suggesting that th determiners are [+ Def ] in these languages. an indefinite NP with a modifier like that. in the specifier. typically the definite article. gran ‘great’. apparently definite Dets. reductions of bue grande. mi and mío may well both be adjectives. m both would be semantically anomalous. English this is not dicerent in meaning from Spa este. expressed as a ture [+ Def ]. no further marker of definiteness appears. the apparent [+ D value of these possessives correlates invariably with specifier position. In languages like English where demonstratives and definite quantifi are in the specifier. It is possible to make the attractive cl that specifier possessives do not dicer in meaning from adjectival possessives. or something else. primero. however. But several commo occurring adjectives in Spanish appear in a reduced form when pre-nomi examples are buen ‘good’. primer ‘first’. tuyo. DG possessives are like definite determiners. This points to a general p ciple of interpretation.

there is no n even for the definite article to be [+ Def ]. The major presentation of the DP hypothesis is Abney (1987). is that. or. more recently I has been split i separate T(ense) and Agr(eement) heads. and with NP. between lexica substantive categories like N. this means it can semantically and grammatically empty.4 for the essentia 290 . . A currently influential view is that phrases projected from lexical c gories occur as complements in higher phrases projected from functional hea Thus for a time the clause was taken to be headed by a category I(nflection). V. 8. Lyons leaves open the question whether indefinite NPs simply have their speci empty. argued to be headed by Det (or D). comparable to the pleonastic pronoun it in It see that . Haider (1988). since the expression inducing definiteness in the merely by its presence in the specifier need not be a definite Det. some basic discussions for a number of languages are: Hellan (1986). A just as the number of functional heads in the clause has multiplied. lower down in a place in the tree parallel to that of VP in the clause. not lexically. And since the definite article freque has no semantic or grammatical content apart from [+ Def ]. The ideas discussed in this section will be reinterpreted below in terms of DP framework. Szabo (1987).2 The DP hypothesis The idea that the noun phrase is projected from Det rather than arises out of the distinction. Ritter (1988). An important consequence of the definite c straint. It is a meaningless filler. paralleling I (or one its components T and Agr. whether they have no specifier position. A and P. expr ing tense and “subject–verb” agreement. with the role occupying NP specifier in the absence of any contentful item to fill that posit It is thus a pleonastic Det. if it is correct. . which ocers a way of explaining why the filling of a particu structural position induces definiteness. Radford (1993). or perhaps C(omplementizer) ).ness is thus seen as being determined structurally. which has grown in importance. whose r is essentially grammatical and which are often realized as inflectional m phemes. For critical discussion see Pa (1993). there are n numerous proposals in the literature for functional heads additional to D in noun phrase. more radically. whether t lack the specifier because they are not maximal projections. hea by N. and the lexical projection VP is relativ low in the tree structure. and “functional” categories. In the works refer to. This last possibi is closely comparable to the proposal of Rothstein and Reed (1984) indefinites are N″ and definites N′′′. The DP hypothesis represents the extension of such id to the noun phrase. The reader is invited to turn back at this point to 1.

and both nominalization and clause can have a subj taken to occupy the specifier of the functional head. at least in some languages. In these expressions the posi of the noun destruction corresponds to that of the verb destroy. which in both cases may unde a passivization process. exemplified by the relationship between nominalizations like the arm destruction of the city. For example. whereby a head moves. 8. since now all categories. the analysis of pronouns becomes more straightforw On the NP analysis pronouns are arguably anomalous in lacking an overt n head – unless pronouns are taken to be a kind of noun. Within DP framework pronouns can be treated as Dets (following Postal 1970) which an NP complement. have full phr projections. not just the major lexical ones. This is to account the fact that languages vary in the surface position of N relative to modifying c stituents like APs and PPs. again mostly upwa This applies particularly to possessives. A major movement process is Head M ment. in some cases all the way to in others only as far as some intermediate head position. but to move. the vary surface patterns can be accounted for by movement of N over varying distan Movement of N all the way to D would be taken to occur in languages where definite article is an aax on the noun. invariably upwards. both these l cal heads may take a complement. and that they mostly originate pre-nominally. in English APs generally precede N PPs follow. *the me). the city’s destruction and the corresponding clauses army destroyed the city. representing merging of N and D. in most Romance languages some APs precede N while most A and all PPs follow. . And thirdly. For discussio the theoretical issues involved see Haegeman (1994: chapter 2). to another head position wh it combines in some way with the head already in that position. following Fukui (1986). a problematic view gi that they cannot in general be modified by Dets (*this him.1 Movement processes in DP A central assumption of current syntactic theory is that constitu may move from their underlying position. (of ) the city. It is propose much research that N raises in many languages. A second advantage acor by the DP hypothesis is that it makes the theory of phrase structure much m general.phrase closely parallel in its structure to the clause. On the assumption that the position of most modifying exp sions is relatively fixed. The city was destroyed. Possessives (such as John’s. my) are widely assumed to originate in NP spec position.2. since it captures the many similarities in behaviour betw the two. The desirability of this long been evident. B complicating factor is that some modifiers can also move.

Let us term this process P Movement. With phrases like John’s car. But it is assumed that a head can only θ-mark arguments within its o projection (thus. A diaculty for this analysis is that this movement applies only to nominal possessives (noun phrases and pronouns in the genitive). and probably originate as complements rather than specifiers of the head nomina tion. particularly prepositional ones like in English a painting of Ann’s. These two variants can be shown by the lowing S-structures. but not necessarily in NP specifier. indicating movement. and with intermediate projections between DP and omitted: 3 Note that the possessives in John’s arrival and the town’s destruction have theme or pa θ-roles. But it may be that some factor other t case assignment underlies this process. John’s arrival is an unaccusative structure and the town’s destruction is passive-like possessives originate in NP. however. that pass through NP specifier before moving out of NP. On another (adopted by Ab (1987) among others). English and many other languages Poss Movement is taken to be to DP speci – thus. the ’s is a genitive case morpheme or postpositio case transmitter or some such particle forming part of (or attached to) the p sessive DP. her house it is less clear that possessive gets its θ-role from the noun. the town’s destruction) and therefore must receive their θ-role fr N. my friend’s is the D.3 The same reasoning is used to argue that s jects of clauses originate in VP specifier (the “VP-internal subject hypothes and then raise to the specifier of some functional projection where they can rece nominative case. it may be. in terms of linear order. On one view (most f presented in Radford (1993) ). just to the left of D. within NP). also to apparently adjectival ones. Just as a head can only move to another head position. appear superficially to the right of N despite fact that in the languages concerned the leftward movement of N is thought to rather short. the ’s which occurs with full noun phrase (D possessives like John’s. but some writers at any rate treat these cases in the same way those above. which does not seem to have an ar ment NP specifier. But some possessives surface well to the left. and the head D is null. as in English Ru first great publishing success. A further reason for believing that possessive expressions origin rather low in the structure is that some. with t representing the trace left behind in the D-struct position of the possessive. here. The reason for taking the D-structure p tion of possessives to be within the NP projection is that they frequently rep sent arguments of the head noun (in examples like John’s arrival. a specifier express (which must itself be a complete phrase) can only move to another specifier. 292 . her disgus his behaviour.

if not all. and the familiar Dets not have this property.(spec) D′ DP D John ’s (spec) NP (spec) N′ t car D′ DP D John’s 0 NP (spec) t But in many other languages. arguable that every is an excep to the generalization made here in Peter’s every whim. and no other Det can appear there. are still to right of (and therefore lower than) D. DG languages 4 5 This is on the assumption that cardinality expressions are not D in phrases like Peter’s th many pictures. for example Olsen (1989) for German. while clearly higher in S-structure than specifier of NP. *your a plate). like Italian. Let us term this “partial Poss Movement” by contrast w the “full Poss Movement” assumed for English. and the assumption that . -s is one of a range of realizations of genitive case in German. But this is the only possible excep Olsen claims that only a limited range of noun phrase types (principally proper no can occur in DP specifier in Modern German. It must be tha these languages the possessive raises only as far as some specifier between of NP and that of DP. The definite article or some o recognizable determiner may occur in Poss Movement structures in the for (Italian i miei quadri ‘my pictures’. it is. as is shown by the presence of the defi article or some other determiner before the possessive: Italian i miei tre bei qu (the my three beautiful pictures) ‘my three beautiful pictures’. but these are mo unconvincing.4 The son assumed for this is that the D head in the English DG structure has to be to assign genitive case to the possessive in its specifier. So a special Det (in form ’s or zero) which is a geni case assigner occurs in D only in the Poss Movement structure. To avoid c mitment to whether ’s is this special possessive Det. Of the definite determiners. Catalan and Portuguese. These AG languages with partial Poss Movement and DG languages with Poss Movement dicer in another respect.5 and DPOSS seems to be null in most. let us represent it as D There are arguments in the literature for DPOSS having overt form in other langua as well as English. dei/quei miei quadri ‘sm/those pictures of min but not in the latter (English *Peter(’s) the/these pictures. and that the -s inflection they take is DPOSS. Notice that this dicerence in distance of Poss Movement correlates with the AG–DG distinction. pronoun p sessives. the D position in the English structure is taken to be either null or oc pied by ’s (which is therefore a Det). As no above. moreover.

so there is nothing to prevent ordinary Det appearing in D. the possessive form personal pronouns.) and to all the grammatical categories associated with nominal exp sions. but. and DG possessives all occupy the same slot ( specifier). other determiners too) to be D heads. Abney (1987) and Kornfilt (1991) discuss a possible Agr(eement). If ’s is seen as a genitive case morpheme. while aaxal articles are in a dicerent position (typically attached N). This is essentially the distribution I have taken for granted up to now.2 Other functional categories in the noun phrase As observed above. free-form definite artic at least some other determiners.tial Poss Movement does not carry it that far. demonstrat quantifier etc. The possessive (null pro or overt biz would be in AgrP specifier and enter into Specifier–Head agreement with the h Agr. attached to the noun a result of Head Movement of N to Agr. like der Welt grö Saurierskelett ‘the world’s biggest dinosaur skeleton’. can be simply taken to involve idiosyncratic genitive morphy. his etc. Radford (1993) claims that my etc. 294 . Functional categories have been sugges corresponding to all the dicerent classes of determiner (article. Notice that this proposal means that there are both clausal and nominal A projections. 8. my. are proposed D fails to account for DG structures in which it does not occur. According to the latter. I will claim below that these two assumptions of DP paradigm are misguided. T in Turkish ev-imiz (house-1PL) or biz-im ev-imiz (us-GEN house-1PL) ‘our hou the agreement morpheme -imiz would be the Agr head. But DP analysis takes both free-form and aaxal definite articles (as well as.. Let us briefly examine a few of these. And DG possessives are taken be (after Poss Movement) DP specifiers – thus in a position distinct from tha the article and other Dets. many more functional projections have b hypothesized in the literature than were originally envisaged. as well as I being split into separate T and Agr hea other clause-level categories like negation. for m writers. The clause was for a time seen as having only IP and above the lexical VP. both in the cla and in the noun phrase. dicers from the NP anal in the following crucial respect. heading A to represent possessive agreement in languages such as Hungarian and Turkish. A similar argument for not taking be DPOSS in English is that this form does not occur when the possessive in the specifier is noun-derived: *my’s car. in its standard form. Similarly. are Ds like ’s. Notice that the DP hypothesis. your.2. Num and NumP. there h since been numerous proposals for functional categories other than D project phrases. while the DP analysis of the noun phr originally envisaged only one functional projection above the lexical NP. principally between DP and NP. representing grammatical number. aspect and mood have been claim to project functional phrases. On the other hand.

KP wo be a peripheral projection above DP. I shall take this up in Chapter 9. Unlike the other possible functional categories mentioned. which assigns the noun in its complement to a class of the k distinguished in gender systems. And the KP analysis is adopted more rece by Bittner and Hale (1996). If it does. Lyons (1995c) argues that NumP is present in Eng only in plurals. so th nominative noun phrase is caseless and therefore not KP. where the plural morpheme (-s etc. is still lo than Num. Cornile 1992). Gen. so that the overall nominal structure i category KP. then this category. Taking all these categories. anticipating discussion there. There is some evidence for such a relationship. NumP must be rather low in the overall st ture. with the dicerence that only “marked cases” considered to be realizations of K. Some other writers cl that case is realized in D rather than in a separate category (Giusti 1992. Finally Travis and Lamontagne (1992) argue a functional head K corresponding to case and showing properties in comm with clausal C. This implies a relationship of some kind between case and referential p erties like definiteness. and it been used to explain the diachronic emergence of definiteness marking in te of changes in case systems. Laughren (1989) similarly treats the noun phrase in Warlpiri as with its case morpheme as head. to be valid would give the following overall schema (omitting levels for brevity): KP K DP D AgrP Agr NumP Num GenP Gen NP . I here take case to be represented as a separate K head. though arguably a property of individual nouns therefore to be represented in their lexical entries. Even grammatical gender.writers for other languages.) is the Num head. nominative is regarded as “unmarked”. Picallo describes Ge a “word marker”. Since n ber morphology appears on the noun in many languages like English which app to have only short movement of N. and the hierarchical relationships generally assum among them. has been claimed to consti a functional head (Picallo 1991). But. given its intimate association with the noun.

This mea however.) tend to show a specific serialization cross-linguistically.2. The first has already been discussed briefly in 2.head (which is its sister). And note that each phrase has a specifier position ( shown in the above schema). Take. any variation in the ordering of the three A would be marked and probably require a marked intonation contour. Now given the clai have made at various points in the present discussion that the “indefinite artic is a cardinality term in a paradigmatic relationship with numerals and modifi like many. Note also that categories shown here are not all claimed to occur in all languages. respectively. Among these additional projections proposed by Cinq one of the highest in the overall structure is one relating to cardinality. but I believe thi incorrect. And the sa pattern. possessive agreement. This is ironic. and further from or closer to N than in D-structu Functional projections corresponding to each of these modifier classes ocer attractive account of these observations. Thus the AP little. implying Head Movement of N to D (or poss 296 . would generated in the specifier of a “SizeP” projection. 8. quality. as regards which semantic classes of modifier are generated closest to head noun. for exam languages lacking determiners. assumption is that numerals would be specifiers of this. Cinque (1995) makes the more radical proposal that adjectives and other att utive modifiers are generated in the specifiers of a series of functional projecti between NP and DP. it would be natural to take this article too to be generated here. is found in many other languages (though upward Head Movemen N may interfere. a considerable expansion of the number of functional projections assum to figure in the noun phrase. DP. usual assumption in work on DP structure is that a is a D. for instance. for exam the phrase a lovely little white cat. daughter to XP and sister to X′. and grammatical num would be taken not to have. causing some or all modifiers to be superficially post-nom rather than pre-nominal. the DP hypothesis in its standard form problematic in a number of ways.3 Weaknesses in the DP analysis Despite its attractions. col etc.3. Double determination it occurs in Swedish (den långa resan ‘the long journey’) and Norwegian ( nye boken ‘the new book’) is diacult to account for. because aa articles attached to the head noun. I will come back to this point below. AgrP and NumP. and that the dicerent classes of adjective (ordinal. It will suace here to outline just two of th problems. The evidence for this is that in various languages the nu ber of non-coordinated attributive APs occurring in a noun phrase is limited around six. which would be slightly m peripheral than the “ColourP” within which white arises as specifier.4. size.

but there should not be an additional free-f article. in the f of the free-form article. This is better.3. The second is m technical. is a head. implicit in Taraldsen (1990) for Norwegian. as discussed in 2. those people. This is unattractive because it means postu ing two co-occurring functional heads with the same semantic content. One solution. proposed by Giusti (1994). It therefore looks m more like a case of amalgamation of D and N after movement.3. in terms of the NP framework. there are two diaculties with it. is that what makes car. on this analysis. First. A widespread current view is that inflectional categories on lexical he are indeed generated there rather than resulting from movement of lexical he into functional head positions. the former in the D h position and the latter in the specifier of this head. one above the adjective with the free-form article as head one below with the aax as head. it seems to be versally the case that semantically contentful Dets like demonstratives . whether by a Det or a possessive or anything else. The reason assumed for D being already overt in this case. is that there two DP projections. On this view what ma Ann’s car or a neighbour’s car definite is that DPOSS is a definite Det – lexic specified as [+ Def ] like the. the definite aax on N is checked in D at LF. But the result is that [+ Def ] (or whatever content of the article is) will be overtly encoded twice in the same head posi at LF. is that this overt Det is necessary to license the imm ately following AP by governing it. Ann’s car. Ano way out. either overtly or at the abstract level of logical form (L for the inflectional category to be “checked”. The s gestion advanced in 8.2. the aa article of all the Scandinavian languages has the morphological structure o inflected formative attached to a separately inflected noun. But in s cases the functional head where checking takes place should normally be em prior to LF Head Movement. The second problem posed by the DP hypothesis relates to the definite c straint on full Poss Movement structures – thus the DG phenomenon. this at any rate is the case in other instances of process. The first is that. Now two questions arise. see Chomsky (1995) for discuss This is Giusti’s analysis. like N. and the complementary tribution between them is taken to be only apparent. is to say that only the free-form article head D and the aax is an agreement morpheme on the noun. a neighbour’s car definite is the mere filling particular position.the strongest evidence for the DP hypothesis – because these processes are o possible if D. But then the lexical head has to raise to the ap priate functional head.1. But the analysis has Dets and possessives in dicerent positions. the overt D in Swedish Norwegian (by contrast with Danish) shows a greater propensity to trigger ag ment on the noun than does the null D – the opposite of what one tends to in agreement systems. A further observation is that.

other determiners being associated with some lower fu tional head. identical in essential c tent to the definite article except that it can assign case to its specifier (abstr ing away from the deictic marking and agreement morphology sometim occurring with articles).fore cannot replace DPOSS in the possessive structure. that the noun phrase is really “de miner phrase”. but specifi of DP rather than NP. is. So the gramm cal category which I have claimed definiteness is has its representation in syn in the form of this functional head. the distinction English dr between the two being a language-specific peculiarity. again apparently universally. 8. But if determiners in general are not associated w D it makes little sense to see this head as representing the class of Det. Why should this be so in languages? Second. it is reasonable to suggest that only definite determiners are associated w D and its projection DP. as in the NP analysis. specifiers rather than heads.1 and earlier chapters for believing definite determiners are associated with a position higher in the noun phrase str ture than cardinality or quasi-indefinite determiners are equally valid in the framework.3. This claim fits in well with the fact that ne 298 . DPOSS is. But the definite article languages having one) and DPOSS appear to be always distinct. that free-form definite a cles like the are. Given the widely accepted assumption of multiple functional pro tions. collapses. The crucial claims of this analysis are. DPOSS (whether realized as ’ null) is completely synonymous with the. and this surpris fact is unexplained. is the place to look for the definite “ position” hypothesized earlier. and second. in the whole original point of the DP hypothesis.3 A modified DP analysis The literature of the last few years contains numerous propo modifications of the basic DP framework. Rather. dicering only in its genitive-c assigning property which guarantees that it is the only Det occurring in the P Movement construction. that D represents not the word class of D but the grammatical category of definiteness. The proposa present in this section summarize the account given in more detail in Lyons (199 1995b). The DP projection. D is definiteness (and by a happy coincide serves just as well in this role!) and DP is definiteness phrase. Given this. 8. and it is possible to amend the DP an sis outlined above so as to solve the diaculties just discussed. we would expect that in at least some guages DPOSS would be non-distinct morphologically from the definite article. the equivalent of the should also serve as DPOSS.1 D as definiteness The arguments presented in 8. first. In English for example. then.

perhaps bette is initially an aax on N which is checked in D after raising of N at LF). And other languages too may have D heads not needing morphological support of a lexical host. These assumptions also ocer a more natural account of DG possessive st tures. and aaxation is assumed to arise usually from Head Movem (or at least to be associated with Head Movement. still possible to regard definite and indefinite noun phra . I do not rule out the possibility of D heads being free-stand rather than aaxal in some languages. the two positions nee to account satisfactorily for double determination. partly to simplify the exposition. may be a h (see Ritter 1992). indefinites can have no D. surfacing after the operation of ious movement processes as inflections on N or second-position clitics. free-form articles are typically specifiers. D itself is most commo null. N may raise to to combine with the article suax there: [om-ul [batrin [t] ] ]. In particular. in languages like English. I assume here. see Lyons (1995b) for the det of the analysis. which may be abstract. in which position the art cliticizes to it: [batrin-ul [t [om] ] ]. They are specifiers. am the possibilities discussed in 2. In particular. or the AP may r from its intermediate specifier to specifier of DP.3. the special Det DPOSS with its attendant problems can be pensed with. as are noun phrase modifiers more gener following Cinque’s (1995) treatment discussed above. It is. w the free-form article is generated in DP specifier. DP speci And these structures are definite because Poss Movement to DP specifier me DP and D have been projected – and D is definiteness. however. therefore no DP jection. In a noun ph with the underlying form [-ul [batrin [om] ] ] (the old man). The aaxal article in Swe and Norwegian is the head D which combines with N (or. the post-nominal definite ar la of many French-based creole languages. a the checking account). heads if aaxal) and qu indefinite Dets appear lower in the structure.gories rather than to word classes. however. Possessives fail to co-occur with free-form articles and other defi Dets. in a rather natural way. This is a matter for investigation of i vidual languages. aaxal definite articles are realizations of D. It may be that heads in such a language must be cap of standing alone. It follows from this analysis that o definite noun phrases will be DPs. This is because it is thought that creoles have very few mo ment processes. See Giusti (1993) for an account of how Wackernagel position of the Romanian aaxal article comes about. Free-form definite articles and other definite determiners are generally no heads. again as must be true of most functional heads on Cinque’s analysis. probably a free form. This is why only defi Dets are associated with DP (as specifiers if free-form. The assumption of specifier articles and h articles immediately provides. because they occupy the same position. however.

a noun phrase may h NumP and even DP and still count as “bare”. lamay! ‘the boy’. call it Ca identified by Cinque (1995) as one of the highest functional projections. I shall say more about in 8. definite articles are normally ov forms. absence of CardP is definitional of bare nominals. The definite–indefinite distinction also be made in the plural by attaching the cardinal article to a numeral: lam denn! ‘the two boys’.6 Parallel to my treatment of the defi article. NumP must be present in bare plurals to account for their number morpholo but bare plurals cannot be accompanied by cardinality Dets. and it is not obvi how to rule these out if they are specifiers of a projection which is present. and DP therefore non-overt. one. Aaxal cardinal articles can be taken to be Card heads. As is the case with overt defi articles. the projection indefinites stops one level below that of definites. I propose that the cardinal article too. I exemplify fr Sinhalese (Masica 1986). As for the position occupied cardinality Dets like a. The usual indication that a language does have a DP projection is that it ha definite article – because the essential function of this is to express the pro tion. and probably Turkish. I more probable that they are specifiers of the cardinality projection. For m detailed discussion of this see Lyons (1995c). One is that they are specifiers of NumP. This article attaches to nouns only in the singu and a singular noun lacking it is understood as definite: pota ‘the book’. many. 300 . but by absence of a quasi-indefinite article. potak book’. but that definiteness as a grammatical categ is more limited. lamay dennek ‘two boys’. is in this speci along with numerals. Sinhalese. which has a suaxal cardinal article (a phrasal clitic fact. sm. On the hypothesis advanced here.K selects a DP complement only optionally. But this is unlikely.4 below. headed by the category of gra matical number. suaxed noun phrase-finally) -ek/-ak (derived from the numeral eka ‘o with which it can co-occur). but this does not account for its expressing indefiniteness. a point I develop below. I assume the article suax Card head. where a free form.5 that the pragmatic concept of identifiability may play a rol communication in all languages. lamayek ‘a boy’. there are two obvious possibilities to c sider. But there are languages which probably make it necessary to suppose a definite article can be null. since NumP appears to be not far above and cardinality modifiers tend to occur well to the left in terms of linear or Also. These are langua like Mam. For this reason. the presence of this in a noun phrase excludes the overt appearance 6 For Lyons (1995c). what characterizes langua in which definiteness is not found is that they lack DP. I argued in 7. in which definiteness is signalled by a definite article. Apart from KP. It see Sinhalese has a DP and a null definite article.

if they have neither (typically) nor CardP (necessarily) as suggested above.2 The content of determiners If D is definiteness and DP is definite noun phrase (under KP). again. Note that this treatment of the articles is incompatible with the claim Vergnaud and Zubizarreta (1992) and Longobardi (1994). The card article. where sm does not occur.1. 8. Lyons (1995c) argues that in English a or sm alw appears in CardP specifier in indefinites if no meaningful Det is generated th bare plural and mass indefinites. and the that definite and cardinal articles do not both appear in definites is due to so constraint (perhaps phonological) which permits only one article to introduc noun phrase. and in the latter semantically empty and has a merely grammatical function. The same analysis be extended to cardinality Dets. I believe.3. may oc in DP specifier. Aaxal articles realize D head. This pleonastic or default function is what defines articles. . either to realize the head if aaxa as a pleonastic specifier. In particular. cardinality Dets such as numerals associated with specifier of CardP. but can also appear elsewhere provided DP is projected (and sho to be projected by a default definite article). as already s gested in 8.3. What argumental nominal expressions have in comm is not D. but t are certainly argumental. Just as demonstratives. Related to the substantive–expletive claim is the further assumption made many writers that a noun phrase must have a Det in order to function as an a ment.indefiniteness. it lows that no definite Det need be lexically specified [+ Def ]. On my analysis.1. they are lexically empty pleonastic Dets. but K. f form definite articles are always expletive. this assumption cannot stand on my analysis – and is. exactly as in languages with overt definiteness marking. appears only in CardP. but we should not exclude the possibility they too can be adjectival and generated lower in the noun phrase. definite articles have no sema or other lexical content. and free-form articles are the default or minimal Det which is requ to occupy DP specifier position in the absence of any contentful expression th Apart from any agreement or deixis features. that “substantive” and “expletive” uses of definite articles must be distinguish in the former the article genuinely expresses definiteness. are determinerless. nor even a broader notion of determiner.3. This is in fact taken to be one of the main roles of the expletive article. this applies even to the definite article. but they can only appear in DP. for example. however. of wh the head is definiteness. so they always signal definiteness while not encod this. have no CardP jection. English bare plural and mass generics. with the fu tion of indicating by their presence that DP is projected. as suggested in 1. unne sary. referred to in Cha 4. And.

since h and specifier. In the Romanian example in which the demonstrative has raised to DP speci it is to be noted that the article. When a demonstrative raises to DP speci or Poss Movement carries a free-form possessive to the same position. on the present analysis. and similar filters have been sugges for other structures. Recall that possessives are frequently aaxal. by Head Movement. On this view demonstratives other definite determiners are linked to DP specifier either by being in that p tion or by being bound by it. The best-known example of this is the “doubly-filled Comp fil discussed. This is proposed by Giusti (19 for Romanian (omul acesta and acest om ‘this man’). It may be partly a matter of avoiding redundancy. a D head. This is presumably because the semantic conten demonstratives. in Chomsky (1981). for example. Lyons proposes that the fact that a dem strative requires the projection of DP should be expressed syntactically by a requ ment that unraised demonstratives be linked by co-indexing to the DP speci position. entails identifiability of the referent by the hea which is what underlies definiteness. I assume it reflects a fact frequently observed in relation to o syntactic projections for many languages: that head and specifier may not both overtly expressed. Demonstratives too are occasionally in DP specifier or in some other position (such as an adjectival position dee within the noun phrase). and an overt D head becomes unn essary. this ser to express the presence of the DP projection. But if an aaxal possessive is of the DG type. [+ Dem]. whence the definite article accompanying non-raised demonstratives. assumption is that it has raised. As pointed out in 7. it is common for bound articles to fail to co-occur with demonstratives and fr form DG possessives (which are also DP specifiers). induc a definite interpretation without the presence of any definiteness marker. thus heads 302 . and it is not obvious w this should be so. these aaxes may origin in a lower Agr head. In some languages both these possibilities are fou and this has given rise to the suggestion that demonstratives can originate i lower specifier and then (optionally in some languages. and by Lyons (1995d) Spanish (la casa esta and esta casa ‘this house’). share the function of indicating the p ence and content of the projection. obligatorily in others) r to DP specifier – somewhat like DG possessives. to D – thus again signall the projection of DP. and that in the latter case a definite article must also app (in languages which have DP). and if the demonstrative raises to DP specifier it will bind its own trace in its original position. Whether or not this raising app the noun phrase must be definite. does not appear. This amounts to binding of the demonstrative by the article in Span (and perhaps by a null pleonastic specifier in Specifier–Head Agreement with aaxal article in Romanian).

and was found compelling by many in its e days. that personal pronouns definite articles has a natural attraction in a framework that takes articles to heads. 8. typically but invariably the verb. While the DP treatm works quite elegantly.5. This captures the close morphophonolog similarity between clitic pronouns and the definite article in many languages. and Cardina draws the conclusion that pronouns cannot be Dets underlyingly. On this approach. are similar to it in many languages. C pronouns dicer from full pronouns in attaching to some host. as in their Eng equivalents. outlined in 1. in Italian as in Eng takes the form of the definite article or. a demonstrative: i lingu ‘the linguists’.2. I believe this last proposal must be rejected. him.3. Moreover. it is worth noting that the very fact that the pronouns Dets analysis dates back so far. moreover. see also Radford (19 for a similar DP account of pronouns. and Abney (1987) takes up and adapts this analysis.3 The analysis of personal pronouns Postal’s (1970) claim. because there is then need to posit a head N which is abstract or gets deleted. The impossibility of loro befo modifying PP is also paralleled in many languages. they are D he which dicer lexically from the in being subcategorized for no NP complem while we and you are specified as optionally taking an NP complement. But this is to miss the point that there is s pletion and that the “personal determiner” in the third person. Cardinaletti analyses clitic pronouns as Abney does all personal pronouns D heads with no NP complement. she. wh raise by Head Movement to D. makes it clear that it is not dependent on a view of determiners as head is. only first. though often distinct morphophonologically f the corresponding definite article. crucially. noi/voi lingu must be an apposition structure. it misses the point that e strong third-person pronouns. voi linguisti ‘you linguists’. and o identical to a demonstrative Det. It is arg that Postal’s proposal works better in the DP framework. including standard Engl and is not explained by the apposition proposal. here too loro would be repla by a demonstrative.and second-person “pronouns” are possible: *loro lingu ‘*they linguists’. ironic that some recent revisions (such as Cardinaletti (1994) ) p pose that personal pronouns (at least strong ones) are underlyingly nouns. quei linguisti ‘those linguists’.too raise to D from some lower head position. a similar constraint operates with other structu noi/voi/*loro con i capelli rossi ‘we/you/*they with the red hair’. This “c . rather than appearing in argument position. Part of Cardinaletti’s motivation is tha Italian noi linguisti ‘we linguists’.

some). If weak (in the very general sense discussed in 3. except it lacks the NP complement. Uriagereka (1995) claims pronom clitics are the head of a DP with null (but present) NP complement. mine (both pronominally predicatively). tho complex indefinite pronouns like something presumably do involve an NP of wh the head raises to a higher. The same analysis can be applied pronominal occurrences of cardinality modifiers (like one. an example is French. M languages have a free-form definite article but bound weak pronouns (thou interestingly. this h raises by Head Movement through the various phrasal projections of a clause adjoin to a special clitic-hosting head. weak free-fo or bound (clitic) has the same structure as the corresponding full DP. This article is the only overt material in the DP. the red o the three.7 But if the three types of definite pronoun distinguished correspond str turally to definite Det types. derived from my account of definite articles that weak non-clitic pronouns like English unstressed you. 304 . The analysis I want to propose. the reverse is less common). Strong defi pronouns are. whether strong. and clitic pronouns are D heads like aaxal artic Personal pronouns are DPs (under KP) like full definite noun phrases except they contain no NP. she) not carried by the non-prono nal article. my red my three. pa but not entirely accountable for in terms of the need for more content where descriptive part of the noun phrase is lacking. latter as a clitic. Where this overt m rial is in D (thus aaxal).2 in wh weakness may be manifested in a variety of ways). is that the disti tion between free-form and bound pronouns corresponds exactly to that betw free-form and bound articles. the former is realized as a free-form pronoun. functional head where it cliticizes to a determine The essential point of this account. for example my book. there is a mismatch in individual languages. but where any post-D material is present: the book.4. presumably because the lack of the descriptive content of NP gi more importance to such features in referent identification.position. Notice that in English the m phologically poorer the occurs rather than the fuller he. as described by Uriagereka. it may expr features (such as for gender in English he. where no material follows in the noun phrase. but this content is demonstrative. he are DP specifi like free-form articles. and this analysis depends on the treatment of clitics as D heads. not only in presence of NP. like weak non-clitic pronouns. for personal pronouns. it can undergo Head Movement and climb through hig heads to its eventual placement position. she etc. they contain a definite art either in DP specifier or in D. where the sa 7 Other determiner–pronoun pairs show the same distribution. but. In important recent version of this analysis. A definite pronoun. three. DPs with their overt content in specifier. There are in addition morphological dicerences.

A further com cation. and we must consider how proposals advanced apply to non-configurational languages. mismatch. which languages are to be so characterized if there are. at least in the noun phr non-configurational (Japanese and Warlpiri for example) lack definiteness m ing. more profound. are highly clitic pronouns. this is apparent from the fact t in Latin and Russian respectively for instance. pro and weak personal prono can only be interpreted as definite. I have argued that languages which do not encode definiteness lack the c gory D. already discus in 7. striking that the languages most frequently taken to be. despite the lack of definiteness marking in noun phrases. in which definiteness arises from filling of NP specifier.1.9. their noun phrases may be projections of a dicerent functional head they may be NPs. a position missing in such languages (because N is projected only as as N′). This view fits in well with the account of the relationship between defin ness and syntactic structure outlined in 8. Many languages seem to have a category of definiteness in the perso pronoun system but not in full noun phrases. is posed by the development of the the of functional categories. indefiniteness too) is expressed by a Det in specifier. where I will suggest that they are connected 8. There is a second. and whe the properties giving rise to this supposition might follow from the lack of on more functional projections. It should be bo in mind that the issues of whether there are in fact any truly non-configuratio languages. yet to be investigated in detail. I shall discuss the latter of these two mismatches further in 8. Recall Gil’s observation that the languages with . however. and what n configurationality actually amounts to.4 Configurationality and definiteness The discussion in this chapter has been based on languages wh show fully configurational noun phrase structure. As indicated in 3. It is not clear what would be meant by a “flat structu within the current highly elaborate conception of clause and noun phrase with t multiple functional heads. come back to both in Chapter 9. compatible with the DP account of n phrase structure.4. but it is important to pursue the question of configurationa a little way because it has considerable bearing on the issue of definiteness. Gil (1987) proposes that absence of definiteness m ing is a typical characteristic of languages with non-configurational noun phra because definiteness (and. It is not. either in its standard form or in the modified form proposed h It should be perfectly possible for a language which has D but only projects i far as D′ to express definiteness.5. It may be fruitful to consider whether these languages are ones that have been thought to be non-configurational in the noun phrase. The resolution of these matters is beyond the scop the present study.3. for him.

Suaxed to these roots are the case inflection and a c 306 . and numerals normally occur in speci of CardP. they occur in most noun phrases indicate the noun’s membership of one of four classes (roughly: masculine. There are l guages in which demonstratives occur with classifiers. In the absence of CardP some other device is required to enable nume to appear. Gil’s paradigm non-configurational language. Simil the lack of DP does not prevent the occurrence of demonstratives. perhaps even an overt real tion of Card.1. noun class markers rat than numeral or demonstrative classifiers. Dyirbal has a system of classifiers. two CLASS POSS book ACC bought ‘I bought (the) two books. it may be that a number of features taken to be characteristic of n configurationality in fact follow from the absence of certain functional categor Japanese. and uses numeral classifiers for enumeration: (4) a. The status of the classifier (of wh Japanese has several dozen.1. also specify the location and visibility of the refe of the noun phrase. classifiers then being needed to provide units for enumeration. But another possible explanation these characteristics is absence of such functional categories as Num and C Num is the locus for the singular–plural distinction (as well as more elabo number oppositions in many languages). which usually precede the n and agree with it in case. The absence of Num in a guage may perhaps also be taken to have the direct consequence that all no are mass. the two specifier positions typically filled definite Dets and numerals are unavailable. thus they are based on three deictic roots. fe inine. book ACC two CLASS bought Ni-satu no hon o kaimasita. and edible vegetable). or b. Another language widely assumed to be non-configurational is Dyirbal (d from Dixon 1972). the one appearing depending on the noun) is uncl it could be one of a limited class of count nouns. permitting numerals in its specifier – among other possibilities.definiteness marking. W is clear is that the classifier is a device permitting numerals to occur. Hon o ni-satu kaimasita. they are also characterized by the use of numeral classifi and absence of number marking – features he attributes to the absence of count–mass distinction in such languages. These markers. neuter. has neither defi article nor number marking. and a classifier system is one possibility. but Japanese uses the al native device of treating demonstratives as adjectives: kono tegami ‘this lett ano hako ‘that box’. M generally.’ If Japanese lacks D. the forms of wh are given above in 3. Num and Card.

may be missing in a language. But plurality (three or more) may optionally indicated by reduplication: SalXga ‘girl(s)’. etc. This sugg that the position of the possessor phrase balan d¸ugumbil in the matrix noun ph is the same as that of a noun marker. s gesting that these markers are determiners and occur in some specifier: bayi y (CLASS man) ‘the/a man’. or an adjective may occur alone. plural nayinba. But no defi article is possible in languages lacking DP. but ellipsis is so pre lent that not only may the marker be omitted. again. then. That this may be NP speci so that this position is a possible locus for demonstratives in languages lack DP. giyi yana ‘this man’. or in an adjectival position (poss ities that exist more marginally in configurational languages too). On the plausible assumption that the p sessor is in NP specifier in this construction. This may mean Dyirbal has a category Num which is not generated in the unmarked case. D. The quasi-indefi article. since when an article is not itse realization of D it has no semantic content. They can occur elsewhere: in a classifier construct in a dicerent specifier such as that of NP. SalXgaSalXga ‘girls’. usually in that order. should be unable to appear in a language lacking the categ . possessed noun cannot be accompanied by a marker. and half a dozen nouns (all with human reference) have obliga plural inflection: nayi ‘girl (past puberty)’. dual number m be expressed by a suax -d¸aran attaching to either the noun or the noun mar bayi baXguy-d¸aran (CLASS frog-DU) or bayi-d¸aran baXguy (CLASS-DU f ‘two frogs’. similarly. Determiners such as demonstratives numerals which have independent semantic content are not dependent for t appearance on these functional categories in whose specifiers they often occu configurational languages. while the possessor can: ba d¸ugumbil mambu (CLASS-F woman back) ‘the/a woman’s back’. Dyirbal has no definiteness mark but is not without number in complementary distribution with the noun class markers. The single dem strative giSa. the bare noun is the unmarked form to refer to one or m d¸ugumbil ‘woman’ or ‘women’. it follows that the class markers the demonstrative are NP specifier determiners. this does not prevent numerals occurring in such unmarked noun phra and in this language without the support of numeral classifiers. signalling definiteness only indire by filling a position which shows DP to have been projected. is suggested by the following. In the inalienable possession construction. Numerals simply adjectives: bayi yana bulay (CLASS man two) ‘two men’. that this may correlate with properties taken by Gil to indicate some deg of noun-phrase non-configurationality. though this is purely optional in non-pron inal noun phrases.and possibly one or more adjectives. but a marker may occur without noun. The hypothesis is. Num and Card. that one or more of the functional heads connected w definiteness and cardinality.

whose content is identi by agreement morphology. in (5b) pala and jana cross-refere full noun phrases – and in the second case partly identify the content of 308 . The situation with pronouns is more complex. The situation is parti larly clear in the third person. This suggests strongly that they co spond to strong rather than weak pronouns in other languages. Dyir has first. anata (2SG). where personal pronouns are equally strong phonologically: wa (1SG). then Japanese. The diaculty arises with languages in which a non-overt argument. If approach is correct. because no DP.classifiers of this language are realizations of Card. interpre pronominally. dual and plural personal pronouns (n markers serving the pronominal function in the third person).5. is identified by agreement morphology. This seems to be a very dicer phenomenon from the pro of null-subject languages. Subalad¸ingu (2DU+DAT) ). b. anatatati (2PL). But the speci concerned is that of some other projection – perhaps the same specifier that dem stratives occupy in full noun phrases in this language. Japanese may have genuine gaps in sentence structu the missing arguments being supplied pragmatically.’ wajili Malikijarrarlu ka-pala-jana marlu dog-DU-ERG IMPF 3DU-NOM 3PL-ACC kangaroo-ABS chase ‘(The) two dogs are chasing (the) kangaroos.and second-person singular. a non-overt argument has to identified pragmatically from the context. not grammatically. This is the case in m Australian languages including Warlpiri. while not DPs. I-ERG IMPF 1SG-NOM 2SG-ACC you-ABS see ‘I see you. for many linguists the paradigm n configurational language. And if strong p nouns are demonstratives. we can maintain the posi that Dyirbal has no category of definiteness. A similar situation obta in Japanese. but these not identified by any agreement morphology.3.’ Every Warlpiri sentence includes an auxiliary complex which contains a cl cross-referencing each argument of the sentence. as already noted in 7. where no real pronoun exists and ‘he’ and ‘she’ rendered by a full noun phrase with demonstrative determiner: ano hito ‘ person’. has only strong pronouns. Phonologica these pronouns are fully stressed words of between two and five syllables ( S (1SG+NOM). like Dyirbal. I assume the Dyir personal pronouns. watasitati (1PL). Japanese also permits pronominal arguments to be null. The Warlpiri system is illustrated by the follow sentences (adapted from Hale 1983 and Laughren 1989): (5) a. (Ngajulurlu) ka-rna-ngku (nyuntu) nyanyi. and these need not be definite. Thus in (5a) the clitics rna ngku cross-reference free-form pronouns. are like strong DP pronouns in other langua in that their overt content consists of a specifier demonstrative.

unless it is desired to express emphasis on a pronom argument. For recent argum against this view see Austin and Bresnan (1996). On the pro analysis. The pro-nominal interpretation resulting from omission of overt noun phrases corresponds to the weak-pronoun option. O free-form pronouns. IMPF 3DU-NOM 3PL-ACC chase ‘They (the two of them) are chasing them. or ‘Two are chasing some’. then this categ . What this means is that the grammar of Warl must make reference to definiteness – though only. whereas the noun phrases (including free-form pronouns) they “agree” w are organized on an ergative–absolutive case system. advocated for Warlpiri-type n configurational languages by Jelinek (1984). but one might reasonably expect in this case the clitic could identify a null pronoun. for pronouns. clearly represent the strong-pronoun option. An alternative. where overt noun phrases occur. the full noun phrases in (5b) could be omitted. as pointed out in 7. But otherwise the cha terization of the auxiliary as an agreement element is accurate enough. the content of which is identified by the responding clitic in the auxiliary. Clitics cross-reference all o nominal arguments. It is natural that the absence of an overt noun phrase should l to a pronominal interpretation. this is always the c with identified null arguments). these are not argum but adjuncts. is that the clitics themselves h pronominal status and represent the arguments. a ably with demonstrative value.’ A standard analysis of this phenomenon is that. tha non-demonstrative definite pronouns. including noun phrases for which an indefinite gloss wo be appropriate. or itself be construed as a pronoun.3. w out reference to definiteness. A if definiteness is universally a function of the occurrence of D. the free-form pronouns can be omitted – and t typically are omitted. with the re that the two arguments identified by the clitics would be interpreted pronomina (6) Ka-pala-jana wajilipinyi. the sub and object are realized as pro. and that in some instances an indefinite gloss co be appropriate. Bittner and Hale 1996). supplementing the arguments in the auxiliary. What does matter is the following. For our purposes it does not m ter which of these treatments is preferred. in sentences like (6).number. Similarly. the clitic is the weak pronoun. as in (5a).5. Its cr referencing of all arguments makes it possible to dispense with the noun phra themselves: as indicated in (5a). On either v the crucial point is that the interpretation is that of a weak definite pronoun (H 1983. (6) could not be glossed ‘They are chasing som ‘Two are chasing them’. but it is not so (and. Notice that the clitics are case-marked on a nominative–accusative tem. there are no null pronouns in and in sentences like (5). pro is the weak prono on the argumental clitics analysis. it seems.

This claim has been advanced Lyons (1994b and 1995e). He proposes that determiners modify. There those.5. do not involve a DP project and are neutral with respect to definiteness.3 and 8. and is in this way discharge “saturated”. the result being a weak definite pronoun in pretation. 310 .4. see 9. typically m ingless and attaching to nouns as a mere formal requirement. This R argument commonly takes the form of a free variable – the variable taken in m semantic treatments to be bound by determiners. The second group consists of langua in which either there are no null arguments (like Dyirbal) or null arguments not identified by agreement morphology (Japanese).5. Baker explains this in terms of the lack of a f tioning system of determiners in these languages. so that DP. In Dyirbal non-pronom noun phrases are neutral with respect to definiteness. Recall that I am taking p sonal pronouns and personal determiners to have the same source. This determiner is a 8 See Baker (1996) for discussion of the absence of definiteness marking in the polysynth subgroup of non-configurational languages.1 Personal determiners and pronouns The reader is at this point referred back to the cross-linguistic s vey of personal pronouns and personal determiners (as in we linguists) in 3. 8. Baker allows that s polysynthetic languages may still have determiners. they do not include genuine cles. which I have arg is definiteness phrase. These languages have no syntactic category D and he no definiteness. obligato encode number. Degenerate determiners of this kind are also discussed by Greenberg (1978) as deri historically from earlier fully functioning determiners. This weak pronoun is of category DP. a special “R(eferent)” argument of NPs (representing the entity described by the no nal). in s sense. and the “pragmatic defin ness” of personal pronouns derives from their being demonstrative. This blocks its modification by a determiner outside NP. But in the polysynthetic languages this a ment is assigned as a θ-role to the specifier position of NP. the treatment of these forms in 1.3. but these are degenerate.both Warlpiri and Dyirbal. including perh the demonstrative free-form personal pronouns. is actually person phrase. like Warlpiri. We have identified two groups of languages lacking a definite article. in which agreement morphology makes possible the om sion of argument noun phrases.5 Definiteness and person Here I develop the claim that definiteness is to be assimilated to well-established grammatical category of person. Full noun phrases.2.8 8. pronouns. and the interpretat is not necessarily definite. Expression the type we linguists are full noun phrases in which the personal form we definite – and in this case demonstrative – determiner. In Japan an unexpressed argument may be interpreted pragmatically.2 below. 8. and the discussion here is based on this work.4. unlike full noun phrases.

discrepancy between the noun phrase and Agr in the encoding of person is c siderable. where the noun phrase shows ergative–absolutive pattern and the auxiliary a nominative–accusative patt More important for person is where Agr makes more distinctions than do n pronominal noun phrases and compatibility rather than strict agreement is requi . in which a Agr makes possible a pronoun encoding nothing overtly. But if personal pronouns and personal determiners are essentially the same th there are dicerences between them. and in a number of languages demonstrative and non-demonstrative argument use excluded in first and second persons singular. which does show the constraints seen in the noun phrase (where person is often only f represented in pronouns. for example. in limited ways so that semantic compatibility is m tained. But in addition many languages sh more severe constraints. and in some cases perhaps only in weak or null prono Both personal pronouns and personal determiners express grammatical per but they are one of two principal expressions of this category. categories encoded may be organized dicerently in the noun phrase and on A this is true of case. and. to the that the relationship between a noun phrase and Agr is often not quite one of “ag ment”. Interestin these gaps are cross-linguistically rather consistent: in many languages n demonstrative argument use is only possible in the third person. first. The pronouns show complete person–num paradigms.involves a demonstrative Det sií in DP specifier and a person-marked D head In languages lacking DP structure. Their overt content typically consists o specifier Det (for free-form pronouns) or a head D (for pronominal clitics) in languages. and not then. and weak nouns are simply definite articles. to the null subject/object phenomenon. for example disallowing any first. overtly at least. in the case of pro). and is due. in Warlpiri. whereas the pre-nominal determiners tend to show gaps.or second-person nominal forms. They may disagree. Strong pronoun forms involve a demonstrative element. and may consist of a determiner in some other projection in langua lacking DP. second. Personal pronouns (in common with other type pronoun) are noun phrases of the same kind except that they lack the lexical h noun and other descriptive material. with Agr encoding either more material or dicerent features in compari with the noun phrase argument (the apparent source of the “agreement”). it may be that a demonstrative encoding son distinctions occurs as part of some other projection (functional or lexical modifier of the noun phrase. I have also argued that many languages only have DP struc in personal pronouns. In person tends to be more fully and consistently expressed in Agr. since it is also c monly encoded in an agreement morpheme (standardly analysed as a functio head Agr) appearing on the verb or as part of an Aux(iliary) constituent.

man AUX 1PL shout ‘We men are shouting. Some of us like our beer chilled. Some of us disgraced themselves last night. which is not first person. but an anaph expression dependent on it (and required in English to agree for person with antecedent) may be first or third person. It seems anaphor agreement may be ei with the head or the complement (us) of the partitive. (10) a. b. the students work-2PL much ‘You students work hard. Some of us disgraced ourselves last night. in sentences contain a partitive. man this AUX 1SG shout ‘*I man am shouting. But this anaphor-non-agreement is only possible in English wh the first-person interpretation of the partitive is rendered probable explicitly b 312 . The head of the partitive structure is some.’ The subject noun phrase in such sentences (which may be definite or indefi in Spanish. it may be a matte semantic compatibility between third-person some and the first-person possess or reflexive. Warlpiri Ngarka ka-rnalu purlami. b. Other languages req strict agreement between arguments and Agr – including English. Note that the subje are also unmarked for number in Warlpiri. with fi person agreement. here the agreement or non-agreement is not with Agr but with anaphoric expression which takes the partitive structure as its antecedent: (9) a.’ Ngarka njampu ka-rna purlami.or second-person marking on Agr. however. This suggests t as with the Spanish and Warlpiri subject–Agr relationship. this might be captured saying that the person feature of the complement can optionally percolate up the head. b. the former. where in fact “non-agreement” for person is only permitted in the plural. while Agr may encode singular or plu subject and Agr must agree for number in Spanish. This functions as the unmarked person an compatible with first. imply that it does not. with third-person agreement.(7) Spanish Los estudiantes trabajáis mucho. Some of us like their beer chilled. But the (a) and (b) variants are not synonymous. comparable to the Spanish and Warlpiri facts. while Warlpiri has no definiteness marking in non-pronominal n phrases) seems to be third person. imply that the subset defined by some includes the speaker.’ (8) a. w the latter. But English does show an intriguing agreement p nomenon. though Ag very impoverished in English.

in the creation of a DP p jection. . and second. A reasonable assumption is that this association reflects a semantic incom ibility between person and indefiniteness.2 Person and definite determiners Having established some important background facts. where identifiability comes to be grammaticalized. are found in languages. some students work-1PL much On the other hand. like demonstratives. 9 Dos hombres hablamos. I now wis argue that definiteness and person should be conflated. Some of the team disgraced ourselves/yourselves is perfectly possible. these words must then be linked to that projection. This covers personal pronouns and personal determin Languages lacking DP. Words which express participant r lexically are. a noun phrase in which person is enco must be definite – DP therefore. for ex ple. a certain complementarity between person and definiteness. would denote an indefinite plurality (such as might be expressed by so which includes the speaker. In languages that have definiteness. Th two arguments are developed here. like Dyirbal and Japanese.) just as all express demonstrat ness. and it is interesting to consider wh person-marked indefinite would mean. characterized pragmatically as identifiable. do nevertheless have str personal pronouns (or forms functioning in an equivalent manner) which exp person distinctions. but that the expression of these roles need not be grammaticalized (tha represented as a functional projection). There are two observati favouring this analysis: first. And this does not seem nonsensical. indeed preci this is conveyed by the following Spanish sentences: (11) a. two men speak-1PL Algunos estudiantes trabajamos mucho. the fact that person is incompatible with indefin ness. b. The point is that marking of person in a noun phrase is universally incompatible with indefin ness. More specifically. This is expected if person is always associated with D in languages hav DP structure.5. the person features are alw encoded either on a definite Det (which may be the definite article or a dem strative) or on D. therefore. These appear in many cases to be demonstrative. that all languages distinguish course participant roles (speaker. A first-person plural indefinite.9 8. w the complement of the partitive does not agree with the reflexive but does not clash wi This supports a pragmatic mechanism involving compatibility of features. hearer etc.graced ourselves and *Some of us disgraced yourselves. which are identifiable or pragmatically definite. recall demonstratives. It is probably the case.

And. But if it is universally the case that person cannot be encoded on an indefin the fact seems to remain that a first-person plural indefinite interpretation is p sible. as here. namely we – as if a group of cars of which one is mine could be refer to as my cars. the combination of person with indefiniteness should be inconceivable should produce a semantic anomaly. It is only the non-agreement for person between Agr subject in Spanish which makes possible the combination of first-person Agr indefinite subject. it does not encode first p son. But the point is that the fi person feature which co-occurs with indefiniteness in (11) is not encoded in subject noun phrase. the indefinite subset denoted by some (of the totality denoted by is understood. and I would maintain it is not grammatically first person. since if person and definiteness are. the noun phrase may be indefinite. despite appe ances. As observed. Some of us disgraced ourselves last night. it is perfectly possible for the speaker to be a member of indefinite set.a first or second person-marked Agr and a subject not showing this per feature. So we is not the plural of I in the same way as th books is the plural of this book. And while an encoding of the speaker role c not be indefinite. And I believe this is the case. a plurality which includes first person rather than the plural first person. (14) *Un estudiante trabajo mucho. as I claim. This is problematic. to include the spea thus the speaker likes her beer chilled or disgraced herself. The non-agreement in English between an anaphoric word the head of a partitive antecedent. but first person plural does represent a plurality of speakers. with singular subjects. The key lies in the dicerence between singular and plural person valu First person singular corresponds to the speaker. Thus (13) in English and (14) in Spanish. because of the first-person anaphoric word. as a first-person plural indefinite – where it semantically. also gives rise to the sa first-person plural indefinite interpretation: (12) a. discussed above. the sa thing. rather it expresses a plural set which inclu the speaker (or speakers). a student work-1SG much 314 . Some of us like our beer chilled. impossible: (13) *One of us disgraced myself last night. crucially. But again. where at all. b. The issue is distorted by the fact that the concept of a defi plurality of which I am a member is represented grammatically as the plura I. the first-person indefinite interpretation discus is only available. though subject is interpreted as first person plural indefinite.

that in m languages definiteness (DP structure) is limited to pronouns. a horse). and on the assump accepted here that third-person pronouns are forms of the definite article or a dem strative. it is far from clear that there is any encoding of third person in a horse. Perso one of the “φ-features” figuring widely in agreement. We have seen that person is encoded on perso pronouns and on non-pronominal noun phrases containing a personal determi though the latter are severely constrained in occurrence in many.and second-person determiners that one be sure that person is encoded in a non-pronominal noun phrase. this man.10 This contrasts strongly with number and gender. But it dicers from o φ-features like gender and number in typically not being encoded on the cr referenced noun phrase itself. while meaningful and indire available as an interpretation because of the phenomenon of non-agreement. observed at several points. definiteness is typically not cross-referenced Agr elements. it is not obvious that there is any person feature present. it is not one of the usual φ-features. Noun phr of this type are traditionally treated as third person. The second piece of evidence favouring conflation of person and definiten also involves the relationship between noun phrase arguments and Agr. probably m languages. I cla because such inclusion is treated grammatically by languages as the plural of person value.3 for details). In guages which have DP for non-pronominal as well as pronominal noun phra definiteness marking is usually obligatory in noun phrases meeting the conditi for its occurrence. curiously. whether on the verb or on an auxiliary compon to cross-reference a subject or object or other argument. On the other hand. and given restrictions found on these cross-linguistically (see 3. it seems reasonable to see the houses and this man as third person. It is only with first.including a given person) can neither be expressed nor occur as an interpretat Moreover. it is fai say that in general non-pronominal noun phrases typically do not show per marking. . And in the case of non-pronominals without a personal determiner apples. a (11) and (12). definiteness is regularly encoded in the noun phrase. a plural indefinite including person. sm ap and many people.4. many people. as nom φ-features. the houses. which. But.2 of the involvement of definiteness in agreement 10 The fact that personal pronouns do encode person while non-pronominal noun phrases erally do not seems likely to be related to the fact. it appears in numer languages on Agr elements. is never directly encoded in a noun phrase (in any language. This is not to say it neve recall the discussion in 5. tend to occur obligatorily in the noun phrase in languages which h these features. which are far more comm in occurrence.

b.mer amounts to the encoding in the verb’s Agr morpheme of the presence o definite object. ‘Some students are working. In keeping with unmarked value. definiteness tends to appear in the noun phrase but not in Agr. third person is traditionally recognized as dicering significantly fr first and second persons. It is now possible to say that in all languages that have defin ness marking.’ However. This tinction may seem improbable because most agreement systems would not tinguish between these two sets of noun phrases. b. ‘These students are working. This analysis has important consequences for the definition of “third perso since it treats only definites as personal. number.’ Algunos estudiantes trabajan. This suggests the two categories should be conflated. ‘The woman is waiting. w the person. with the result that these distributional o ities vanish. I illustrate with some partial paradigms: 316 . Person tends to appear in Agr but not in the noun phrase. Thus the house and these houses are th person. as in the following subject-agreement instances: (15) a. The object clitics of Macedonian operate in a similar fashion. of having a default function (as when it co-occurs with Agr displaying a dicer person value in the Spanish construction illustrated in (7) ).’ Spanish Estos estudiantes trabajan. Both third-person defi arguments and indefinites are generally cross-referenced by what looks like a th person Agr morpheme. gender (for third person) and case of the object encoded addition. ‘A woman is waiting. (16) a. while a house and six houses are personless. In most cases (the exception being the non-agreement phenome illustrated in (7)–(8) ) this encoding will correspond to the person feature o cross-referencing Agr. because not DPs. in that the two latter relate to participants in the spe situation whereas third person does not. third person is frequently morphologically unmarked in A taking zero form. and even in these languages subject agr ment does not take account of definiteness. person is encoded in every noun phrase that has person – tha in every DP. So the general cross-linguistic pattern is of a complementarity between per and definiteness. It is the unmarked person value. German Die Frau wartet.’ Eine Frau wartet. But this is not very common.

Marija found one cat The conventional assumption is of two distinct categories. as in the paradigms above. That is. d. the usual morphological identity betw third-person Agr and the Agr form accompanying indefinites is unremarkable But just as the morphological realization of singular number is not always z we would expect an overt realization of third person on Agr sometimes. but absent with in finite arguments: (17) a. we should find cases of encoding of first. The ec of this would be that all three persons are overtly encoded. Marija 2SG-ACC knows you Marija go poznava nego. which encode person (as wel other features) and are obligatory with all definite objects. it alm invariably encodes number. the point applies in many langua more clearly to third person plural. also frequently tak zero realization. Marija me poznava mene. sing tends to be the morphologically unmarked number value. c. person and defin ness. If this is correct. second. This is the situa with the Macedonian object-agreement clitics. In fact. So while I am suggesting that Germ -t in (15) represents singular number only. So it is plausible to suppose that what looks lik third-person singular or third-person plural morpheme is actually encoding o singular or plural. encoding other features at the same time. where the proposal is that Spa -n encodes only plural. Marija 1SG-ACC knows me Marija te poznava tebe. non-feminine forms) Turkish Warl various -s – -tî -t1 – -m/-im -n/-sin – -rna -npa – Moreover. the morpheme which expresses person on Agr is nearly always a p manteau morpheme. third person dicering from first and second in not relating to a particip . Marija 3SG-F-ACC found cat DEF Marija najde edna marka. Marija 3SG-M-ACC knows him Marija ja najde marka-ta. third person having its usual zero realization. b. In particular. e. and defi third person. contrasting with lack of marking for indefinites.1SG 2SG 3SG Spanish Biblical Hebrew (perfect aspect. with the result that the combined person–number form for t singular is zero. as in (16). and person mark should then distinguish definites and indefinites in what is traditionally seen third person.

on exp sions in their specifiers). is represented by a functio head. only the two subordinate.11 We can therefore add a feature [± Speaker]. First and second persons behave dic ently from third person in a number of ways. But we have seen that personal forms. person.3 The nature of person-definiteness I am assuming that at least some grammatical categories are rep sented syntactically as functional heads. equivalently. only third-person forms may be indefinite. definite n participant). 8. syntactically and morphologica as observed. for example.always definite. We may this category either “person” or “definiteness”. We can hypothes 11 Recall that the hearer may be included in the reference of a non-singular first-person fo except in the case of exclusive first-person forms in languages which have these.) in this head or in its projection. is the c gory D. can be considered to be simply per with no further subordinate specification. including languages like Dyirbal and Japanese with no DP structure these languages the personal forms found are strong forms. and it seems per distinctions are expressed lexically rather than grammatically. occur in languages. so let us assume a feature [+ Participant] characterizing them. What distinguishes them is that they correspond to discourse par ipants. what matters is that the traditi ally recognized person distinctions are hyponymous to definiteness. What tinguishes first from second person is that only the former may (and must) incl reference to the speaker. Thus if mood. person is a category under wh a number of subordinate distinctions must be made. third or unmarked person (or. like he and she. it will still be necessary to express sub-categories of mood (indicative. And while car is.5. and the feature represen tion of these distinctions is considered next. s junctive etc. on the account assumed. and features are the usual way expressing such distinctions. in some cases. Against this assum tion I have argued that person and definiteness are the same thing. but also that distinctions within catego can be specified by features encoded on these heads (or. as the unmarked value. 318 . indefinites like a car or someone are personless. Third person. the superordinate property. In the same way. giving following hierarchy of properties: Person Participant Speaker Note that. hyponymous pr erties are represented as features. like demonstratives.

Participant role features must be encoded in the locus of person-definiten itself because this represents their own superordinate category. This wo account for the morphophonological strength of the forms concerned in non languages. Recall my suggestion in 7. they always themselves in one or other of these positions. in an adjective p tion for example. Common patterns are for a qu indefinite article to be limited to specific indefinite use. and for an “irrealis” ma to accompany a cardinal article with non-specific indefinites.5 that in languages like Maori where article combines pragmatic definiteness (identifiability) with specificity.2. simple definites in earlier terms) tend to be weak. But. This is ap ently not possible with forms expressing participant roles. Catalan la ciutat aquella (the that) ‘that city’.3.represented as features. Some of the n phrase types argued to play a part in the animacy hierarchy (such as proper na and animates proper) will still have to be accounted for separately. a demonstrative need not be expresse D or specifier of D. than is [+ Dem]. but a con erable number of the properties involved can now be treated as forming a ch of hyponymy. It is not enough these to be accompanied by a marker of definiteness in D or its specifier. which is merely incompatible with omis of D. it may be elsewhere in the noun phrase. I suggest this is because features [+ Participant. Person and definiteness (understood in a broad se to include specificity) are two of the sub-hierarchies proposed in the literatur being involved in this complex. It may also account for an interesting contrast betw personal forms and demonstratives. see 2. It is worthwhile at this point to return to the “animacy” hierarchy (or cluste hierarchies) discussed in 5. T association of the [± Spec] distinction sometimes with definiteness and so times with indefiniteness is not problematic provided we keep in mind that pers definiteness is a grammatical category capable of considerable variation in w . can be lexically expressed. ± Speaker] are more directly subordinate to pers definiteness. Examples. in DP languages. since forms expressing merely person (that is. not person itself. Both are incompatible with indefiniten though both occur in non-DP as well as DP languages. Maori te whare nei (the house this) ‘this house’. and in languages with structure. provided there is a definite article present (to which the dem strative may perhaps be syntactically linked by coindexing). both must be associated with D at least in the sense that they req projection of DP.5 and 5. This means that specificity in such languages is also part of person. But in other languages specificity is encoded independently of definiten and indeed associated rather with indefiniteness. this be treated as grammatical definiteness with merely a broader than usual rang use. from are Ewondo é mvú S# (the dog this) ‘this dog’. and the unification I have advanced here betw these categories ocers a significant simplification of this idea. thus to D.

and the subordinate distinctions are features on some expression in (in head or specifier position). up to but not above ostension. I h gone on to claim that definiteness and person are one and the same categ Thus the familiar discourse participant distinctions traditionally associated w person. specifics are thus intermediate between indefinites and definites. This category of p son-definiteness (or whatever we choose to call it) is represented as D and DP grammar.1. We can therefore establish a hierarchy of properties. so that it would more accurate to say that noun phrases are specific or non-specific. consider the fact that in numerous languages polite pronoun forms referrin the hearer are grammatically third person. So DP is definiteness phrase rather than determiner phrase. In the former case.2 and 4. part or all of which (working up from less to more general) m be embraced by person-definiteness: Specific General definite (identifiable) Ostensive (textual–situational: see 4. indefinites are defined negatively with respect to definiteness. and I have identified this category with the wid accepted D. and speci are definite or indefinite. are finer distinctions within the larger categ of person-definiteness. in much the same way as past. 12 If further illustration is needed of the mismatches tolerated between grammatical catego and meaning.1. as first and second person. It is then open to languages to have a grammatical egory of person which embraces pragmatic definites only or which also take (at least some) pragmatically indefinite specifics. I suggest that this chain of properties may be a use input to comparative work relating to the idea discussed of an “animacy hierarch 8.1.3) Anaphoric (see 4. present and future appear as distinctions subordinate to the category of tense.and we have seen evidence that languages treat definiteness as hyponymous specificity.3) Participant Speaker The expression of these properties. [± Spec] m then be encoded for those noun phrases not covered by person. to be interpreted as a ch of hyponymy. 320 . may also combined with [+ Dem].6 Summary of proposals I have argued in this chapter that definiteness is represented syn tically as a functional category.

but the lexical element. and this will correlate with lack of card articles and number marking. to head or specifier position. such as demonstratives. and an article or other determiner in DP has to appea languages with no DP. they are probably syntactically linked to DP specifier in guages which have DP. wh is the definite article in the case of weak pronouns and a demonstrative in the c of strong pronouns. Languages with category of definiteness lack DP. . NP. may also occ DP specifier. either completely or only in tain (for example full) noun phrases. Definite pronouns have the same structure as definite full noun phrases. Head articles are typically bound fo and directly express the category (D or Card). but are not grammatically definite. grammatically or pragmatically. indicating merely by their presence that the category is p jected. while specifier articles are typic free forms and pleonastic. may similarly be absent in languages. a specifier article gives a free-form w pronoun and a head article gives a pronominal clitic. but there are many languages which have D pronouns but not in full noun phrases. Substantive definite determiners. but in many DP guages they can raise into DP. Other functional projections. They consist essentially of a definite determiner. lexically empty fillers of a position that wo otherwise be empty. A parallel an sis is probably appropriate for substantive cardinal determiners. the pragmatic c cept underlying definiteness.mately of the same category KP (which is higher than DP). though they may also appear in a range of other modifier positi In the latter case. demonstratives involve identifiability. In the former case. like Ca and NumP. Articles may be either heads or specifiers (of DP in the case of definite arti and CardP in the case of cardinal articles). and this yields DG phenomenon in which possessives give the appearance of being defi determiners. like numer Possessives are not definite.

determiners. It may be too much to say that this a of morphosyntax is more prone to historical change than others. or other determiner. the emergence of articles fr other. on the standard DP hypothesis as outlined in that this language has no class of determiner.1 The emergence of functional structure Until recently. The emergence of articles can t be linked to the development of DP structure. third. Apart fr the fact that articles can be aaxal. but research in cates that it is possible to trace quite radical shifts acecting definiteness and de miners. because a great deal has been written this topic. it is not clear why an adjectival demons tive in a non-DP language should not weaken to express merely definiteness w 322 . In this chapter I will exam three aspects of this issue: first. substantive. An explanation along these line problematic. the subsequent development of articles definiteness. typically demonstrative. It is assumed that languages vary in w functional projections they have (so that. Indeed. has made it p sible to look at the question in new ways. however. elements. most research on the appearance of definiten marking in languages which had previously lacked it concentrated on the proc of semantic weakening whereby a demonstrative. for example. in the course of the pres study. the acquisition of a syntactic category of defin ness by languages previously lacking it. sometimes over relatively short time spans. I have repeatedly made reference to the diachronic sources of article other. both from the point of view of general historical linguistics and in w on particular languages or language families. But the development in the last few years of the theory of fu tional categories. and the absence of DP in a given langu should have the consequence. 9.9 Diachronic aspects It is appropriate to devote some attention to the emergence development of markers of definiteness. second. a language with no num marking can be taken to lack NumP). including the DP analysis of the noun phrase. becam definite article. since it implies that definiteness markers must be determin and that a language with determiners will have definiteness marking.

See also Lyons (1995d) for this accoun 9. represents appearance of the category of definiteness in languages. and amounts to a cha in syntactic structure: the creation of a DP projection. but only the former op is open to the article. The modified DP analysis presented in 8. This is a desirable posi because the empirical evidence is for a much closer association with D on part of definite articles than of other “definite determiners” like demonstrati We have seen (in 3.1. to . on which D is not the word c of determiner but the category of definiteness. is the reinterpretation of an inalienable possession structure as a structure (as suggested in 7. Another possible mechanism. therefore. then.3. which might interact with this reanalysis encourage it. It is the projection of DP that makes a n phrase definite. that creation of DP entails creation of definiteness mark The obvious mechanism by which this might come about is the reanalysis by spe ers of an adjectival demonstrative as an article. and I shall concentrate the dominant pattern. cardinal determiners. may either occur in head D or DP specifier position. and even that the creation of an article is a necess concomitant of the emergence of DP structure.3 of what a defi article is: either a realization of D itself. It is clear. or.1. languages with structure must have a definite article. with no special inalienable construction. serving merel indicate that DP has been projected. ocers a clearer account. it may be that can exist without overt realization. pleonastic DP specifier. or alternatively linked with an expression in one of these positions. and the article signals this projection. a (usually obligatory) place-fi of this position in the absence of some substantive occupant.3. because there is a tendency for their semantic equ lent in DP languages like English.miners (because with DP) should not have only demonstrative.1 The development of DP The diachronic emergence of definite articles. This follows from the view given in 8.3). The idea is that inalienable constructions are re ily interpreted as definite. and.5. there can be no definite article in languages lacking DP structure. to the ex that it is obligatory to have some expression of a projection. without a marker of simple definiteness.3. but definite articles are never adjectival in position.2) that demonstratives and quantificatio determiners can be adjectival in languages with DP structure as well as in guages without. The major qualification must be added is that in languages like Turkish and Sinhalese which sig indefiniteness by a cardinal article but have no definite article. Demons tives etc. if a free fo a semantically empty. This is uncertain. With framework it is possible to maintain that DP structure is necessary for a langu to have a definite article. 3.5 and 8. particularly if aaxal. and I shall say more about below.

There is a tradition of linking it to a shift fr synthetic to analytic structure. and figures prominently in the recent literature on this topic. Given this. the alienable–inalienable distinction is pr ably easily reanalysed as an AG–DG distinction. Moreo 324 . is to carry the inflectional morphol of the noun phrase.ness) do not apply. and such facts have been used to support the claim of a close relations between definiteness and case. But it is clear from the discussion in 5.1 that case distinctions function in a number of l guages to express a variety of contrasts approximating to the definite–indefi one. at least in some languages. on the basis of Germa that case and articles are both devices for expressing the definite–indefinite tinction. Philippi (1997) argues. one might argue that. and to the l or reduction of morphological case systems. Nevertheless. rein preted in terms of current syntactic theory. The creation of articles is therefore a way of ensuring noun phrases will generally contain an expression of their case. and the fact that inali able constructions tend to be morphologically simpler than alienable ones.1 that the contrast expres by dicerent choices of case for direct objects in languages like Finnish. But the evidence is lacking for the assump that languages with articles generally used such a case strategy before. This last suggestion is particul intriguing. This suggest relates to the view discussed (and rejected) in 5. to changes in word-order pattern. but in some languages (such as the Romance and Germanic groups) two developments have coincided approximately. showing an additional mar such as a preposition. A traditional account of this coincidence is that case morphology well as other nominal feature marking) is often preserved on determiners afte has been lost on nouns. Other possible triggers for the creation of definiteness and definite articles h been discussed in the literature. A more recent account is based on the proposal that case its locus in the D position. and the idea grew out of observation. AG possessives tend similarly be more complex morphologically than DG ones. so that weakening of a case system naturally leads to the creation o definite article. so that case morphology and an article are alterna realizations of this head (Giusti 1993). But we saw in 5. as in I’ve cut my finger. the pragmatic contrast expressed by the choic direct object case in these languages is suaciently close to the definite–indefi distinction to encourage speakers to create a category of definiteness when the p sibility of the case strategy is lost.4 that the principal function of definite article. while case and defin ness are grammatically distinct. Hungar Kabardian and Turkish is not the same as that involved in definiteness. It is not at first sight obvious wh loss of case distinctions should lead to the emergence of a category of defin ness. I am assuming a K head distinct from however. thoug may come close to it. wh often involve an additional particle.

Of course it is not ruled out a language may move directly from Type I to Type III. NP specifier).do not show a particularly close correspondence cross-linguistically. and perhaps only in weak personal pronouns. a form I take to be a demonstrative. Latin and Warlpiri. occurring in a position which could be interpreted as specifier. if the possessor expression in an inalienable constructio reanalysed as being in DP specifier (rather than. with the usual phonological weakening. Fairly s stantial case systems co-exist with definite articles in Classical Greek. languages w no article in full noun phrases. perhaps additionally. both these k of weak pronoun are always definite. But at the point at which this h pens. but with either overt weak personal pronoun null pro-nominal arguments (probably pro) identified by agreement. representing definitenes DP structure: Type I: Type II: Type III: no DP pronominal DP only pronominal and full noun phrase DP Type I is exemplified by Japanese. Japanese is such a language. And Type III is represented by English all languages with definiteness marking in full noun phrases. what is ruled out is a T I language acquiring a definite article in full noun phrases only. According to Givón (1978). . and either no weak or null personal pronouns or o unidentified null arguments (interpreted pronominally but not constrained to definite).5. and Hungarian even shows the direct object case strategy discus as well as a definite article. the re is necessarily that the structure becomes DP. Re from 7. whe overt or null. If a demonstrative. Type II is exemplified by Russian. I assume that the rean sis of demonstratives (brought about by factors to be discussed below) is principal trigger for the creation of a category of definiteness. The shift from Type II to Type III consists of the extension of availability of DP structure to full noun phrases. I h 1 This strong claim would be falsified by a language with unidentified null arguments but a definite article.5.1 Interestingly. b shifts are probably typically mediated by the weakening of a demonstrative. is reanalysed as an article. in which there is no defi article in full noun phrases.3. it may well be that the language already has DP in a more limited way. I repeat the typology proposed in 7. Icelan and German.4 that many languages apparently have DP in t pronominal system only. Korean and Dyirbal.3 and 8. perhaps. T from Type I to Type II consists of the development of weak personal prono overt or null.3. Givón claims Japa has a definite article sono. 8. These three language types give us two possible major diachronic shifts.3. For these reasons I discount the claim that the em gence of definiteness is linked to the loss of case systems.

the former are pleonastic Dets in DP speci while the latter are D heads. are likely be demonstrative (except where they are full noun phrases. in Typ languages like Dyirbal as well as in Type II and Type III languages. It may be. This process is well attested. cross-referencing pro. and the latter (pro) must be identified agreement morphology (Agr). as evidenced by the numerical preponderance of null-subject language Personal pronouns may. Spanish. This means that the second major shift. with lexical encoding of discourse participant features. might come about. therefore.1. Just as agr ment clitics or inflections arise from weak pronouns. This second sub-shift. that the shift from Type I to Type II should be divi into two sub-shifts. and can be illustrated by the similarity betw 326 . This second sub-shift consists of a specifier Det evolving into head. languages may have free-form or bound definite a cles. First a demonstrative adjective com to be reanalysed as a free-form article. generally. 9. so free-form articles diachronically prior to bound articles. w not essential to the achievement of Type II status. This may. then. But Agr (whether on a verb as in the Romance guages or on an auxiliary as in many Australian languages) itself usually com into being through the cliticization of weak pronouns (see Dixon (1980: chap 11) ). of which only the first is essential. But there is a diachronic process at work here since bound forms commonly originate from free-standing items. fr Type II to Type III. And the cha to Type II consists of these demonstrative pronouns weakening semantically morphophonologically to yield weak pronouns – which are necessarily definite. need not. can also be seen as involving two sub-shifts. as may be the cas Japanese). and I have argued that. This is the formation overt weak pronouns through the weakening of strong pronouns. once ag only the first being essential to reach Type III.2 A definite article cycle Type III. with. We can spe late that this is related to the fact that pronouns exhibit discourse participant f tures (which underlie person-definiteness) much more readily and commonly t do non-pronominal noun phrases (because reference to speaker and hearer is u ally pro-nominal). and probably Warlpiri). and subsequently this may become a c or inflection. more often than not does t place.from Type I to Type II. as noted. fully DP. Russian) or n (Latin. represent the grammaticalization of the discourse participant features as pers Weak definite pronouns may. Strong personal pronouns. But we will see that this more “advanced” nature shows it again at a later stage in the diachronic progression. be either overt (English. develop DP structure before full noun phrases and are in this sense more “advanced” in relation to definiteness. be followed by the evolution of these weak pronouns into clitics eventually agreement inflections.

this then in turn becomes a head. wh revives the idea first proposed by Jespersen (1917) of a “negative cycle”. others to a dicerent host. again typically fro demonstrative. I assume these are all D heads. Analogous to this. this wo recall recent work on the syntax of negation (Zanuttini 1991.Romanian aax and the Italian free form: (1) a. The phenomeno definite adjectives seen in the Baltic and Slavonic languages (at least in relic fo may also have arisen from the Wackernagel pattern. with repeated renewa definiteness marking. is as follow negative expression in or associated with NegP specifier evolves into a Neg he this is subsequently reinforced by a new specifier negative.4 for details of these distributions. reinterpreted in terms of this analysis.3).3.3. as in the double determination structure Swedish. Ne tion in this research is analysed as a functional head Neg projecting a phrase N and the idea of the cycle. Here the combina is with an element outside the noun phrase. the head form t drops out. What is unclea this account is why the older Slavonic and Baltic definiteness marking only sisted (if it ever occurred more generally) in the presence of an adjective. Icelandic hestur-inn hinn sterki hestur ‘the horse’ ‘the strong horse’ Romanian profesor-ul Italian il professore ‘the teacher’ ‘the teacher’ In some cases. I sug there may be a diachronic “article cycle” as described. (2) a.3.3 that aaxal articles are of various kinds. Ano interesting case of cliticization of an article to something other than the nou the preposition–article contractions also mentioned in 2. Adjectival modifiers are nominal in these languages. Wackern forms are particularly common. and it is. having in c mon that they need morphological support from some host. It may be that the next in the diachronic process is for the aax to be lost and for the more recent f form article to take over as the sole exponent of definiteness. Or it may reinforce the aax. leaving the more recent specifier form as the sole exponent of ne tion. in some cases at least. m . as in Danish where two divide up the range of definite structures or in Hausa where they are a natives. This new form may complement the aax. If so. It will be recalled from 2. the evolution of a free-standing article into a bound form been followed by the creation of a new free-form article. some atta ing to the lexical head of the noun phrase. Rowlett 1994). b. See 2. and so on. as it does in Amharic and with the mod Wackernagel article of Macedonian and Bulgarian (see 2. b. and a second-position article would therefore at to such a modifier where present.

The redu forms would be bound variants requiring morphological support from a host. And the choice between the two artic is optional but involves the same stylistic and syntactic triggers mentioned abo This analysis implies a diachronic process whereby (at some point in the medie period) a head D article developed from a specifier article in the usual way ( with the typical reduction). are less common in writing. wh could be achieved by Head Movement of D to P or by cliticization to whate is immediately to the right in the noun phrase. are as lows. others. When contraction does not occur. Preposition–article contraction in German is optional. though unacected speech uses a few heavily reduced forms. the article is free-standing in written and careful spo German. together with the corresponding free forms: free clitic das den dem der s n m r The standard assumption concerning the diachrony of these clitic forms is that t are reductions of the corresponding free forms. also pr ably clitics. taking both article forms as D heads. and in contexts involving a preposition. but only one contraction occurs in writing with the feminine da (zu der → zur ‘to the’). These cl article forms. Synchronically. are normal in all styles. mainly involving the masculine–neuter singular dative form of article (like in dem → im ‘in the’). but without ousting the latter. might involve an optional reduction proc triggered by stylistic factors or presence of a governing preposition. free-form articles are specifiers and the clitics D heads (again undergoing rais to P or cliticizing to adjacent material). oth including further masculine singular datives and all instances of the mascu singular accusative ( für den → für’n ‘for the’) and dative plural (mit den → m ‘with the’). and more common w some choices of preposition and form of article than others (Durrell 1991). Let us look more closely at this phenomenon in Germ which shows some intriguing distributions of article forms. both in preposition–article contractions and elsewhere.turally higher P head. So combinations. involving neuter singular accusative ( für das → fürs). resulting in co-existe 328 . In my revised DP framework. Most prepositi take the dative. are essentially limited to the spoken language. a standard DP an sis. These are the same as those occurring in contractions. and perhaps an in esting illustration of the article cycle suggested.

and their optionality. then these forms instantiate article cycle proposed. Pronouns tend to travel more rapidly through cycle. is not derived diachronically from this. note that the English colloquial pronoun form ’em. a specifier definite article is likely to evolve into a D head ea in pronominal than in non-pronominal noun phrases. I suggest that the clitic article forms could be relics of an lier article (perhaps cognate with the A-article of Fering – see below – and e with the Scandinavian aaxal article) which was largely supplanted by the d. But while the synchronic c acterization just ocered for German may be accurate. but from earlier hem which was otherwise supplanted by them in the Middle English period. There are two striking points which may cast doubt on the standard view the relationship between the two German article forms. especially given that the loss of initial [d] or an ea dental fricative is not a process known to have occurred otherwise in e German. A second manifestation of this characteristic rel to the article cycle outlined. since they o acquire DP structure earlier. It is well established that these clitics deve diachronically from free-form pronouns. While stressing that this dicers f the standard view. especially given that the preposition–article contractions attested in very early High German texts. the change of a specifier article into a head. . fuller system. If so. the first step in the formation of agreement inflection. First. I suggested above that personal pronouns tend to be more “advanc diachronically in relation to definiteness than full noun phrases.2 Second.two articles of that language are indeed cognate. bri discussed above in relation to Type II languages. the fact that clitic forms exist corresponding to only some f standing forms. while the m phophonological reductions acecting grammatical items in weak positions ( clitic articles) can be radical and idiosyncratic. make them look like relic forms from older. In these languages it can o 2 To take a somewhat parallel case. there is reason to ques the diachronic background outlined. of course. whil doubt synchronically a reduction of them. that is. I proposed in 8. the systematic loss of the en root here is surprising.a cle after developing into a bound D head. The fact that the free-standing forms are always possible s gests that these may represent the more recent article rather than being the sou of the clitic forms.3 that dicerence between weak free-form personal pronouns and clitic pronouns is the definite article which constitutes the overt and essential content of both specifier Det in the former and a D head in the latter. so it is clear that their formation ins tiates part of the article cycle.3. This head status is what ena pronominal clitics (in the Romance languages for instance) to undergo H Movement to their placement position. T process is.

for example. but it is not obvious why they should also be more advanced in rela to the article cycle. many T III languages show a mismatch here. They nearly always develop from the semantic weakening the singular numeral. with bound definite a cles but free-form personal pronouns. most probably a free form. if in an appropriate position relative to other noun phr constituents. I have proposed that both articles are to be analysed as eit pleonastic specifiers or functional heads.1. So this article depends the language having a projection CardP. Numerals and other substantive cardinality modifi need not be CardP specifiers. the reanaly leads to its creation. there are languages (like the Scandinav group) which show the opposite of the French pattern. 330 . in the case of the cardinal article means the specifier or head of CardP rather than DP.guages both pronominal and full definite noun phrases can undergo this devel ment. Perhaps this second mismatch between pronouns and full n phrases should not be overstated. T has happened in Romanian. I suggested above that the more advanced nature of p nouns in relation to the acquisition of DP structure can be accounted for in te of their special relationship to the discourse participant features which unde person.3. In French the definite article.3. comparable to the weakening of a demonstrative to y the definite article. they can for example be adjectival. But an adje val numeral ‘one’ may. and it may be hypothesized that the em gence of a cardinal article in many cases goes hand-in-hand with the developm of this functional projection. with clitic pronouns but free-form arti The mismatch is particularly striking in languages where the two are cognate close in form. as pointed out in 8. But. and third-person direct object clitic pronouns are identical: M SG F SG PL article clitic pronoun le la les le la les It seems the article (originally a specifier) has advanced to D status only in p nominal DP in French. If the language lacks this projection. 9. so that both free-form pronouns and free-form articles become aaxal. But this pattern is less common.3 Indefinites and CardP The emergence of cardinal articles is in many ways parallel to of definite articles. undergo semantic and phonological weakening and be reanalysed a pleonastic CardP specifier.

Thus. 9.2 The origin of articles Here we look more closely at the sources of definite and cardinal a cles (again concentrating on the former. is derived from a numeral class (Masica 1986). Lyons (1977: chapter 15) for discussion of how “quasi-referential” items of behold type may ontogenetically underlie demonstratives and definiteness.5. which still has the same form (Blass 19 The development seems to be via the use of this verb in the ostensive sens French voici/voilà (expressions which in French too are based on the verb ‘se Latin ecce or archaic English behold: Ná c|lm| (see spear) ‘There is the spe In fact ná has developed a range of other uses. It is much more common. Mauri Creole has lili la (Baker 1984. though this seems to occur less readily than with definite a cles. We have seen that in some languages (like Macedon definite articles maintain the deictic distinctions displayed by demonstratives the only dicerence is that the latter are [+ Dem]. The semantic weakening or “bleaching” taken to be involved in the shift f demonstrative to article reflects a very general diachronic process of devaluatio lexical content. and in the case we are concerned with it is essentially the feature [+ D which is acected. so that their creation involves los . including that of a deictic pa cle ‘there’. Grant 1995). The French definite article has in many cases been reanalyse part of the noun stem. for articles to have no deictic content. Demonstratives are. overwhelmingly the most common source.1 further suggests the possibility of cy developments of specifier and head articles. h ever. Interestingly. and only when attached to the noun that it expresses definiteness: ek-Ri chele (o CLASS child) ‘one/a child’. chele-Ri (child-DEF) ‘the child’. This process can be described in terms of loss of lexico-sema features. Even genitive morphemes may b possible source for definite articles. h ever. among other possibilities. sometimes from verbs and classifiers. In many French-ba creoles the article is a form la derived from the French adverb là or deictic ticle -là ‘there’. 9. the same form is still in use as a classifier. see Bader (1993). The Sissala a cle ná has its source in the verb ‘see’.into a Card head. Bengali definite article. J. usually a demonstra or numeral.1 Demonstrative to definite article In fact definite articles derive not only from demonstratives. The Persian data discussed in 2. which have been much more intensiv studied) in the reanalysis of some substantive determiner.2. corresponding to French le lit ‘the bed’. aaxed to the noun. and that of a temporal demonstrative ‘that’ (t\X ná ‘that time’).

which also has two definite article forms. with definiten 332 . the D-article. occasionally. We have seen that there are many contexts in. it is its p tion in DP which makes it a definite article. though unstres morphologically fuller than the A-article – in keeping with the fact that it is clo in function to a demonstrative. example. as discussed. where form used for textual–situational ostension. in specifier of DP if free-form. As a result. therefore. that causes it to exp anything. We have s that this band of uses of the definite article can be overtly distinguished from o uses by a distinct article form. divid up the field of definiteness in almost exactly the same way as Fering. a new articl likely to be restricted to this range of uses before. indeed the D-article is.there has been an intermediate stage at which the demonstrative concerned already lost its deictic feature to become a general. indeed. Fering is in fact not an isolated instance of the p tition of article functions along this line. If this is so. the two operating simultaneou and reinforcing each other. Latin was a Type II language. generalizing to ot definite uses. This development is. though La the ancestor language. This area of over is what I have called (in 4. In this c it is the position of the item. consisting anaphoric (and cataphoric) together with immediate situation uses. deictically unmarked dem strative.1. definite articles may carry agreement features. not just a matter of feature loss (or h ever else one chooses to characterize semantic weakening). had not. Apart from dei features. I suggest the crucial fact that makes this possible is that there is a broad ov lap in function between definite article and demonstrative (particularly deictic unmarked demonstrative). see Heinrichs (1954) for discussion the German dialect of Amern. it may be that definite articles not encoding deixis typic have their immediate source in such general demonstratives. For further discussion of this point see Lyons (1995d). I beli it is in the textual–situational ostensive function that demonstratives unde phonological weakening and reanalysis as articles. All the modern Romance languages have a definite article. I have argued th demonstrative would have to be in a position in the noun phrase which perm reinterpretation as specifier (or. and I have noted that languages la ing definiteness tend to use demonstratives more extensively. even if it does have deictic or agreement content. perhaps. English where the choice of definite article or demonstrative ma little dicerence to the message conveyed. head) of DP to evolve into a defi article. It is probably a co plex interaction of reanalysis and weakening. apart from its lack of str identical to a demonstrative. For this reason. but they may have n ther of these and be grammatically and semantically completely empty. is.3) “textual–situational ostension”. One of the most intensively studied cases of definite article creation is tha Romance. This is the case in Fering.

between them. and accu-ille ing Spanish aquel ‘that’). The place of iste as second-person f was taken over by ipse. A major reason for the doubt is that texts during this long pe show a high. late spo Latin (or early Romance) largely dropped is and hic. Selig 1992) indicates that ipse was restricted to second-mention. Much has been written on the question of when ille ipse became articles. This gives a system something like following: general ille. forms from surviving only in a few pockets. at least in the earlier part of the period. while ille was less constrained but mainly u cataphorically. ipse tended to occur as general demonstrative. The point is that. . perhaps in part because t were phonetically rather insubstantial and would become more so with the so changes of the period. Aebischer (1948) coins the term “articloid” to label this incipient article. Caesar ipse ‘Caesar himself’). and many late Latin texts of that period show forms predominating. More recent research (Re 1976. frequency of occurrence of ille and ipse in noun phra but that. and increasing. Alm all modern Romance varieties have an article descended from ille. most estimates falling into the period between the third eighth centuries.a three-way person-based deictic contrast as well as a deictically unmarked fo general is ASS1 ASS2 ASS3 hic iste ille In a complex series of developments (which I greatly simplify here). while iste replaced hic. Ille took the place of is (while continuing in use as th person form). ipse ASS1 ASS2 ASS3 iste ipse ille In further developments. before a restrictive relative. this frequency is greater t would be expected for demonstratives (even given the tendency of these to be u more in non-DP languages) yet smaller than obtains later with the definite a cle. originally not a demonstrative but expressing the se of emphatic self (ego ipse ‘I myself’. wh is for some writers a demonstrative with very high frequency of occurrence for others an article with a still limited range of use. stri anaphoric (not cataphoric) use. while ille and ipse evolved into definite articles. But ipse articles were much more widespr during the first millennium AD. the deictically marked forms tended to be morphol cally reinforced (thus ecce-iste to yield Old French cist ‘this’.

do have an optional definite article. see comments on Lezgian and Nama in 3. It is fairly common emphatic ‘self’ and ‘same’ to be expressed by a single form (French même. who argues that early Romance ipse is in fact furt limited. 3 It was probably also still optional in this function.2 Articles in competition In early Romance. The widespread view that a weakening demonstrative only achieves f fledged article status when it becomes obligatory in appropriate contexts is incorrect. to marking topic continuity.2. My su cion is that. and expands from there. involving just the usual loss of feature [+ Dem]. The originally distinct article function the two forms seem to follow from their earlier senses. 334 . The demonstrative ille. informat about immediate situation reference. covering the same range of functions as Fering D-article. derives from a reinforced form of ipse. though around the end of the millenn ille expanded geographically to almost eliminate its rival. so that it would not appear with seco mention items which are not informationally prominent.1. and the further limitation on the ipse article.1 that some langua like Hausa. wo naturally weaken initially to a more general textual–situational ostensive arti See also Vincent (1997).2. Romance had a definite article (with two for ipse limited to strictly anaphoric use). its appearance being determined by com pragmatic factors.part of the period indicated. From anaphoric demonstrative anaphoric article is then a small enough step.3. And it is comm for the form for ‘same’ to function also as an anaphoric demonstrative. Given that the evidence for these early Romance developments is necessa from texts rather than speech. But the most inter ing point here for general historical linguistics is that two article forms sho have been introduced in the first place. and ipse seems to have replaced Latin idem in the sense ‘sam French même. Vincent also sugges pragmatic connection between topic marking and second-person deixis. points out that the geographical areas where ipse forms survived longest as article are also areas with an [Ass2] demonstrative derived from ipse. Thus the “articloid” is an article limited to textual–situational os sion.2. say around 400 AD. Class Greek autos). and this may distort conclusions. indeed. But recall from 2. the n article begins by being restricted to the area of overlap already available to demonstrative. not having the anaphoric feature. we cannot expect to find much. 9. both ille and ipse articles expanded from the rest tions described to become general definite articles. The point for the diachrony is a demonstrative does not immediately become a general definite article. clearly re to the chain of hyponymy discussed in 8. Dicerent Romance-speak regions came to prefer one or the other.5. if any.3 This limitation.

Fe may well be an instance of a language in which a free-form article has appea to compete with an existing free-form article. 9. Again the item yielding the article would h to be in a position which can be reanalysed as specifier or head of the relev . The younger D-article has partly replaced it. If this is correct.. and all attested forms derive from ille only. but o in textual–situational ostension. Clitic object prono are another Romance creation. There are many cases of articles coming into existence in languages alre possessing an article. objects (like complements of prep tions in many languages) often appear without an article even though identifia This may relate to the cross-linguistic generalization that subject position is to position and that topics are definite. if a language has a category of definiten it must be represented in this position but may be optional elsewhere. as discussed above in relation to the proposed article cy The Scandinavian languages. The greater age of the A-article would (perhaps as well as its n being limited to uses not available to demonstratives) also account for its wea form.2. illustrate this well.from two substantive forms. in which a free-form article was introduced at so stage to complement or double an existing aaxal article. The same process may be at a more advanced stage in Germ where I have suggested that the clitic article forms are relics of an older art (which may be cognate with the Fering A-article). and while these same two forms also survived (without the morphopho logical reduction characterizing the article. indeed in some Romance varieties w reinforcement) as demonstratives and strong personal pronouns. It is likely that the two Fering a cles are of distinct origins. it is an interesting illustration of how one article may rep another in stages. This too is discus by Vincent. This also arises through semantic weakening. the range of functions commonly covered b new article. having once covered whole range of definiteness. Vincent arg that Romance clitics. though non-singular forms like English sm often derive from so other substantive cardinality item. who relates it to the observation that definiteness marking in full n phrases was initially largely limited to those in subject position. and in many other languages. and that the A-article is older. almost always of the sin lar numeral. did not corresp to the value of the topic-marking article ipse. the two first complementing and later rivalling e other. being objects and phonologically weak (therefore inap priate for representing informationally prominent referents). there is no dence at any period of pronominal clitics derived from ipse.3 Numeral to cardinal article Much of the above is paralleled in the development of the card article. perhaps arising about the same time as the art or slightly later. In Old Fren Spanish etc.

See 2. Instead they h introduced a new one based on one – Hawaiian Creole wan. 9. 9. we now examine further possible expansions which do not yet t the article beyond the domain of definiteness. Languages with definiteness ma ing vary considerably in their use of it. But article.5.5 for exemplification these points. but no cardinal. where it was s that the main dicerence in this regard between English and French is that gen ics are typically definite in French but not in English. The expansion of a definite a cle from an initial limited range of uses to the more general definite value of Eng the does not stop at that point. It is important to note that these parallels between emergence of definite articles and that of cardinal articles do not imply that two processes are likely to occur together. where definiteness has spread into the generic function. to use the singular num more widely than other languages.‘a’ and ‘one’. argues that the “indefinite” article emer later than the definite article as a distinct development. This is clear from the many langua with a definite. p ticularly if they also lack number marking on nouns.2 and 2.1 Expansion in article use We have discussed the expansion of cardinal articles from singu specific use to non-specific and to non-singular use. usually be limited initially to specific indefinite function and gradually extending into n specific uses. For disc sion see Bickerton (1981) and Janson (1984).5. for example where it is desired to indicate non-plurality of a referent. An article may even expand in use to the po where it ceases to be a definite article. In this connection it is interesting to note that a number of Engl based creoles have not taken over the English article a. It is not unusual for languages which lack a cardinal article. is restricted to specific singular indefinites. as noted in Chapter 2. because Old French was much like Modern English in us bare plural and mass nouns for generic reference. A singular form derived from ‘o may also expand into non-singular use. Pl and mass generics can be definite in English too with a limited range of nou 336 . discussing the gins of definiteness marking in Germanic. article. The same is true of the ot Romance languages. unlike English a. We have seen that cardinal articles expand their use in stages.3 The longer perspective Let us now consider what can happen to an article once establis (apart from elimination by a new rival article). and Abraham (1997).3. for example. Limiting the discuss to definites. and that of definite artic from textual–situational ostension to wider definite use. This dicerence is in pa diachronic fact.

generic. Greenberg (1978) argues the normal development is for the definite article to end up as a mere marke nominality. the definite article does. as we see next. and indeed Italian.2). use it with definite possessives. Italian Portuguese. generic. possessive. thus c ing to resemble the articles of Samoan and Maori. and by extension a series of diachronic stages languages may p through.advantages. I can’t stand the rain). Italian and Portuguese also use the art with generics. and Greek and Catalan use it with generics and definite pos sives.4 and 7.2. a neat progression in respect of definite a cle use: 1 2 3 4 (English): (French): (Italian): (Greek): simple simple simple simple definite definite. a noun class or gender marker.5. Interestingly. But. In all the uses just discussed. I have suggested that spec reference counts as definite in these latter languages (2. His Stage I would emb all the definite article uses I have so far discussed – that is. represent definiteness. according to the an sis I have given. This gives. this is true of Greek. a guage is only likely to start using the article with proper nouns when it alre makes extensive use of it. It could expand fur to embrace what are at present in these languages specific indefinites. AG languages. he cla in particular that the gender aaxes of many languages are derived from old defi articles. possessive definite. for these languages. Other languages use the definite article yet more than French. an article can expand in use beyond this field. He posits three stages in this diachronic process. or even a case morpheme. which has l ited use of the article with names. 9.2 The life cycle of definite articles In languages like Greek and Catalan there are few potentially defi noun phrase types which do not already take the article. while Greek and Cat use it with proper nouns. where the art . Examination of further languages would make the picture much less n But there is a diachronic progression reflected here. for example. in that. so it may be that English is moving in same direction. But expansion beyond this would entail the article acquiring non-definite uses. and t becoming something other than a definite article. The point is that within the semantic/p matic field that definiteness may cover there is room for a language to incre the ground which the category actually covers in that language. proper noun It is almost certainly too much to suggest that this represents a universal im cational scale. generic definite.3.

but he gives few c crete illustrative examples and few references. 338 . has a preposed marker like Gurma. in which no take a suaxed class marker. let us consider one more of his cases. and that they have th origin in a pronoun or determiner showing class agreement. developed a new. moreover.Romance or the greatly expanded version of Modern Greek. At the intermed Stage II. Greenberg’s survey of languages is very wide ranging. One dialect of East Aramaic has. definite article. In Gurma. a further guage of this group. Thus n ba ‘men’. Since then it has becom Stage II article in Modern Western Aramaic. this is Stage III. ba niti-ba ‘the men’. By the e Christian period this had developed into a Stage II article in eastern diale but was still a Stage I article in western dialects. Other languages in the group lack this prepo article. prefixed. the article (which Greenberg terms the “non-generic article”) no lon has any function relating to reference. they do not origin as class markers. but nor does it yet accompany all nou Its presence in a noun phrase is largely or wholly “grammaticalized”. it accompanies the noun in the vast majority of occurrences. of real definiteness. His Stage III is of the gender morpheme or meaningless nominality marker. Greenberg points to many further instances of the phenomen mostly in Africa. express nothing beyond class membership. and some involving noun classification. “so that they correspond grosso modo to the combined uses of a definite and indefi article”. The Samoan and Maori definite (as I claim) articles do seem to be still at Stage I. the “non-gen article” seems to correspond to the definite-specific article of Samoan and Maori. there evidence that the prefixes are younger than the suaxes. saying at one point that it combines uses of a definite article with those of a specific indefinite article. In many these instances the historical development from Stage I to Stage III seems to h been very rapid. while in Eastern Aramaic the ear Stage II article is now a meaningless Stage III noun marker. others not. for example. He adds that Stage II articles always appear in addition in many non-spe uses.4 Greenberg illustrates from a subgroup of the Voltaic languages. Then there are other langua in the group in which nouns always carry prefixes as well as suaxes. Diachronically. His thesis is therefore hard to ass 4 Greenberg is less than clear about the Stage II article. it tends to be absent. In contrast with this. but is omit for example. but this is a Stage II a cle. with the object of negative sentences. but th not what is meant. and presumably have no category of definiteness. being de mined by the construction. Put like this. The earl recorded Aramaic (ninth century BC) had a definite article suax -!. Gangam. and this expresses definiteness. in predicative n phrases and in the objects of negative sentences. nouns may have in addition a prepo marker also encoding class membership. on which the historical development has taken several thousand years.

Stage III is then the a logical extension of the particle to all. In Serbo-Croat and Lithuanian. it is already a sem tically empty marker of nominality or has acquired some other function like tha indicating gender.4.1. descended from the definite m pheme.5. and in many cases cease to have grammatical or semantic function. . But proper nouns always appear without “article” at Stage II in the languages investigated by Greenberg. and has been generalized to most noun phrases except thos which it rarely occurred when it expressed definiteness. and include the article c ing to be used with proper nouns. It is far from obvious why a formative with important discourse function should lose it. These steps are all within Greenberg’s Stage I. An impor point to make here relates to the steps of definite article expansion discusse 9. with other grammat categories) it can vary considerably in the semantic/pragmatic ground it cov When it comes to be stretched close to its maximum possible extent. Note also that the move of an article to St II is in many instances not accompanied by the creation of a new definite arti This means that in these languages the category of definiteness. nominal contexts. The Stage II “article” is clearly not an article. the fact that (in common. where it may be that the definite adjective f no longer expresses definiteness but is either merely required in certain grammat contexts or is an emphatic alternative. The answer to this puzzle may lie in the flexibility of the category of defin ness. We can also illust the Stage II article from our discussion of definite adjective declensions in 2. this is sim to what I proposed for Arabic nunation in 2. it must be emphasized. He also shows how definite de miners can develop into linker morphemes in possessive constructions.3. as Modern Greek and Samoan. in 2.3. or most. Schuh notes that in some dialects of Bade a suax which once a definite article and is now a Stage II article actually signals indefiniten because of the range of grammatical contexts in which it occurs. It is this shift from Stage I to Stage II which is for our purposes the most triguing point in this progression. it probably represents Stage II. it has such a high frequency of occurrence in n phrases that it becomes possible for speakers to reanalyse it as a particle cha terizing nouns or noun phrases but conveying nothing. it is not clear whether we have to do with a Stage II or a Stage III fo The distinction between Greenberg’s Stage II and Stage III is small and so what arbitrary.stem extension of Macedonian. we h seen this process at work in Romanian and Hausa. DP structure lost.1.careful study of the development of definite determiners in the Chadic langua refines and gives detailed exemplification of Greenberg’s notions of Stage II Stage III articles. In the c of the relic -i. This means an article can move to Stage II and lose its function of marking definiteness be realizing its full potential as a definiteness marker.

340 . they can also lose it. which. Central to the account ocered is the claim that. Definiteness itself can expand its ra of application. and a point can come at wh its exponent is reanalysed as grammatically and semantically empty (perhaps le ing to its being pressed into service with some other function). specifics etc. Not only can languages acquire the category definiteness.. while pragmatic concept of identifiability may play a part in all languages. Articles can undergo morphological modification from free form aax (typically with concomitant syntactic change from specifier Det to D h status). When a language acquires definiteness mark it acquires definiteness. means it acquires DP structure.In this chapter I have discussed the origins and subsequent evolut of definiteness and definiteness marking in the light of the synchronic finding the rest of the study. Vari subsequent developments can then occur. the grammatical categ of person-definiteness does not. and can be replaced by new forms. At this point. formally. acecting both the category and exponent. the defin indefinite distinction collapses. and discou participant features certainly do figure in all languages. un a new article emerges with a reduced function to renew the category. taking in generics.

o sea gramática de euskera (dial guipuzcoano). Maltese. 259–308. Aoun. Topic and focus in Mayan. de 1971. Lars-Gunnar Andersson and Östen Dahl 1977. The English noun phrase in its sentential aspect. Cambridge. J. Language 68: 43– 80. Andrews. III: Grammatical categories and the lexicon. Wackernagel’s law and the position of unstressed personal nouns in Classical Latin. Austin: University of T Press. Logic in linguis Cambridge: Cambridge University Press. Cambridge: Cambridge University Press. Aissen. Inflectional morphology. Quileute. Adrian and Stephen R. Paul 1948. Stephen R. aspect and referentiality in history of German: the case of the verbal genitive. 1994. Initia amharica. C. Judith L. Arrigaray. C. J. Euskel-irakaspidea. 150–201. Stephen Paul 1987. Contribution à la protohistoire des articles ille et ipse dans langues romanes. vol. London: Routledge. Richard 1975. Keenan 1985. San Sebastian: Auñamendi. Joseph 1965. Cultura Neolatina 8: 181–203. or Navaho made harder. Jens. Allwood. ed.REFERENCES Abney. Tim Shopen. Joseph 1985. Mass. Anderson. Anderson 1970. Adams. 1908. London: English Universities Press. Introduction to Classical Nahuatl. Andrade. Akmajian. 1985. Stephen R. International Journal of American Linguis 36: 1– 8. 1933. In Parameters of morpho tactic change. and Edward L. N. Manuel J. Aquilina. Ans van Kemenade and Nigel Vincent. On the use of the fourth perso Navaho. New York: Columbia University Press. Timothy Shopen. Cambridge: Cambridge University Press. Transactions of the Philological Society 92: 103–7 Aebischer. ed. Mira 1990. Armbruster. The interdependence of case. MIT. Accessing noun-phrase antecedents. H. 1992.: MIT Press. In Language typology and tactic description. III: Grammatical categories and the lexicon. 29–61. vol. . Deixis. Anderson. PhD sis. Abraham. An introduction to spoken Amha Cambridge: Cambridge University Press. A grammar of anaphora. Ariel. In Language typology syntactic description. Cambri Cambridge University Press. Werner 1997.

Kersti 1994. Generalized quantifiers and natural langu Linguistics and Philosophy 4: 159–219. Non-configurationality in Australian aborig languages. Biggs. Jon and Robin Cooper 1981. Blass. G. The case of unaccusatives.. Z. Columbus. Hasan and Daniel L. Catherine V. A study with special refere to Sissala. Let’s learn Maori: a guide to the study of the Maori langua Wellington: Reed. Peter and Joan Bresnan 1996. London: Routledge. J. Mark C. Barry J. Bader. 1976. Basri. Parametric syntax. Regina 1990. Philip 1984. Gerh Meiser. Z. Les relations spatiales en turc contemporain. Winifred 1993. Patrick 1994. and tensed logic. ed. François 1993. ed. A textbook of the Hungarian language. 1971. Berent. Case studies in Semitic and Romance langua Dordrecht: Foris. E. Case. Borer. Blake. Journal of West African Langua 5: 35– 47. The definiteness of trace. Adriana 1988. New York: Oxford University Pr Baker. On the realization of trace: Macedonian clitic pronouns Morphosyntax in Slavic. Bokamba. Cambridge: Cambridge University Press. Lond Collet’s. Natural Language and Linguistic Theory 14: 215– 68.Austin. Journa Semantics 11: 83–101. Blackburn. Tense. Bruce 1969. Gerald P. Innsbruck: Universität Innsbruck. Ohio: Slavica. Linguistic Inquiry 19: 1–33. Chvany and Richard D. Barwise. Relevance relations in discourse. 1996. 12– 45. Ann Arbor: Karoma. Bre 150–86. 342 . Studies in Afri Linguistics 2: 217–38. Bauer. Paris: Klincksieck. Boas. Maria and Ken Hale 1996. Dakota grammar. Geburtstag. Maori. Roots of language. Derek 1981. Cambridge: Cambridge University Press. Bittner. Relative clauses in Bambara. Te Reo 27: 89–129. Jókay and D. Bastuji. Bánhidi. Bickerton. Les génitifs-adjectifs déterminés et le problème de l’arti comparaison typologique entre l’étrusque et les langues indo-européennes Indogermanica et Italica. Festschrift für Helmut Rix zum 65. Linguistic Inq 18: 141–7. Bird. Washington: National Acade of Sciences. Börjars. Agglutinated nominals in Creole French: their evolution significance. 1994. Linguistic Inquiry 27: 1– 68. Finer 1987. Charles 1968. Belletti. Hagit 1984. The polysynthesis parameter. Baker. Swedish double determination in a European typological perspec Nordic Journal of Linguistics 17: 219–52. Franz and Ella Deloria 1939. Specificity and definiteness in Dzamba. temporal reference. 1980. The structural determination of case and ag ment. Szabó 1965.

Cambridge: Cambridge University Press. Li. Anna 1994. Bunt. Butt. Francis Jeffry Pelle 249–77. Bresnan. 1995.) 1995. Noam 1981. Les paradigmes génériques en anglais. A grammar of Lakhota. Eugene 1939. On the internal structure of pronominal DPs. G. Noel 1976. Syntax of the comparative clause construction in Eng Linguistic Inquiry 4: 275–343. New York: Garland. Colchester. Cambridge: Cambr University Press. Cambridge. and Francis Jeffry Pelletier (eds. presupposition and the the of grammar. 1995. The limits to debate. Mchombo 1987. 1980. Bresnan. Topic. Givenness. pronoun and agreement in Chiw Language 63: 741–82. An introduction. Dordrecht: Foris. Ronnie 1993. Travaux de linguistique 19: 17– 1989b. ed. Wallace L. Linguistic Rev 11: 195 –219. Formal semantics. Paper presented at the Autumn Meeting of the Linguistics Associa of Great Britain. Ter2ze 1966. Latvian. In Subject and Topic. Andrew 1991. A unified analysis of the English bare plural. The generic b Chicago: University of Chicago Press. Dordrecht: Reidel. A new reference grammar of Modern Span London: Arnold. Cambri Cambridge University Press. Dynamics of meaning: anaphora. Mass terms and model-theoretic semantics. Harry C. 1977. Carlson. 1985. 25–55.: MIT Press. Cardinaletti. definiteness. Chicago: Chicago University Press. Chomsky. top and point of view. Gregory N. Carlson. Topic and the presuppositions of simple sentences. The language of the Teton Sioux Indi Rosebud: Rosebud Educational Society. Burton-Roberts. ed. . Joan W. Chierchia. Gregory N. Joan W. Berlin: Mouton de Gruyter. On definiteness. Robert 1994. Cann. 1989a. Lectures on government and binding. A study with special reference to Eng and Finnish. Charles N. Linguis and Philosophy 1: 413 – 56. Carlson. Language 52: 427– 1984. Cambridge: Cambridge Uni sity Press. Chafe. Ensembles and the formal semantic properties of mass ter In Mass terms: some philosophical problems. contrastiveness. New Y Academic Press. 1979. 1976. John and Carmen Benjamin 1994. 1973. and Sam A. A grammar of Supyire. Reference to kinds in English. A revised theory of semantic presupposition. Mass. Budika-Lazdika. Paper presented at the Aut Meeting of the Linguistics Association of Great Britain. Buechel. On the generic indefinite article. Chesterman. The minimalist program. London: English Universities Press. subjects. Bangor.tinction.

Heny and H. Cowan. Pronominal reference in Thai. Cook. On the evidence for partial N-Movement in the Roma DP. Vancouver: University of British Colum Press. Cinque. Luigi Rizzi and Raffa Zanuttini. Typology and universals. Selections from the Third Groningen Round Table. Alexandra 1992. 10–63. In Syntax and semantics. The syntax of Chamorro existential sentences. Lond Hodder and Stoughton.Copen-hagen: Munksgaard. Studies in honor of Richard S.). Clark. Guglielmo 1995. 1986. Slavonic’s closest approach to Suffixaufnahme: the possessive adjective Double case. Chung. Comrie. ed. An introduction to Tswana grammar. Cambridge: Cambridge University Press. Language universals and linguistic typology: syntax and morpholo Oxford: Blackwell. Aravind K. Cooke. Norman. Remarks on the determiner system of Rumanian: the dem stratives al and cel. 1987. and Catherine R. 1981b. Joshi. Peter 1975. Desmond T. R. London: Longman Cole. 344 . 1995. 191–225. Studies in the Linguistic Sciences 5: 1–35. In Paths towards Universal Grammar. Definite direct objects and referent identification. In Reuland ter Meulen (eds. Frans Plank. Herbert H. The interpretation of pronouns. Cole. Tzotzil grammar. Agreement by Suffixaufnahme. Michael 1976. Corbett. Bonnie Webber and Ivan A. Null objects in universal grammar. Sanskrit. 1969. Craig. Washington: Georgetown University Press. M. Cornilescu. An apparent asymmetry in the formation of relative clauses in Mod Hebrew. Cambridge: Cambridge Univer Press. Jacaltec noun classifiers. Definite reference and mutual kno edge. Burmese and Vietnamese. 265–82. Croft. 85 –110. J. New York: Academic Press. 1955. vol. Schne 61–92. ed. Jean-Yves Pollock. Coulson. Language 63: 299–345. Sandra 1987. New Y Oxford University Press. Linguistic Inquiry 18: 597–612. Kay ed. Berke University of California Press.1. Jan Koster. Sag. 1968. William 1990. Probus 4: 189–260. Robin 1979. F. ed. Bernard 1978. Marshall 1981. Cooper. Cambridge: Cambridge Univer Press. A Sarcee grammar. Oklahoma: Summer Institute Linguistics. Eung-Do 1984. Guglielmo Cinque. The languages of the Soviet Union. A study in grammaticalizat Lingua 70: 241– 84. Colette G. 1987. In Elements of discourse understanding. An introduction to the classical language.D3. The morphology–syntax interface: evidence from pos sive adjectives in Slavonic. Greville G. 1981a. Pragma Microfiche 3.

ed. etc. 1986b. M. A. Linguistic Inquiry 11: 337–62. Two notes on the theory of definiteness. 1980. Zwei Formen des bestimmten Artikels. Linguistic Inquiry 22: 1–25. ed. Journa Linguistics 22: 25–39. England. 1991. Simon C. 147– Amsterdam: Benjamins. Keith S. A grammar of Mam. 159–74. 1978. Cambridge.) 1996. Nora C. Donaldson. Martin 1991. London: English Universities Press. Definiteness and referentiality in Turkish verbal sentences Studies in Turkish linguistics. . Müserref 1986. Lingua 68: 149–88. Indefinites. The Dyirbal language of North Queensland. In Probleme und Fortschritte Transformationsgrammatik. The semantics of specificity. The origins of genericity. R. Schmalstieg 1 Introduction to Modern Lithuanian. The languages of Australia. Francis and Ruth Carlson 1966. and S. 110–24. Austin: Univer of Texas Press. Evans. texts. C. Kekchi. Karen H. G. La Paz: Don Bosc Ebert. De Bray. J. Yanada 1958. men. Appositive NP constructions: we. ed. Durrell. Philosophical Rev 75: 281–304. Reference and definite descriptions. New York: Academic Press. Dunn. Cambri Cambridge University Press. Icelandic: grammar. 1978. Cole. Dan Isaac Slobin and Karl Zimmer. ed. Guide to the South Slavonic languages. Antanas Klimas and William R. Deemter. T. Stanford: Center for the Study of Language and Information. Wunderlich. 1983. 47– 68.: MIT Press. Evelyne and Ray C. 1972. Columbus. Ma K. W. Eachus. Gramática y diccionario aymará. glossary. D. Pronouns. P. Enç. R. Molly 1992. Delorme. 1980. Sprechsituation und die bestimmten Artike einem nordfriesischen Dialect (Fering). Cambridge: Cambridge University Press. Donnellan. Leonardas. 1971b. Foundations of Language 8: 2–29. Studien und Materialen 4. Cambridge: Cambridge University Press. van and S. a man. descriptions. The Hague: Mouton. Munich: Hueber. 1966. New York: Franciscan Fathers. Stefán 1949. Linguistics 29: 79–102. 1980. Ebbing. Gareth 1980. In Languages of Guatemala. Ohio: Slav Declerck. Mass. Dik. Hammer’s German grammar and usage (second edit revised). The manifold interpretations of generic sentences. we men. Juan Enrique 1965. Baltimore: Johns Hop University Press.Dambri5nas. K. Ngiyambaa: the language of the Wangaaybuwan of New So Wales. Peters (eds. Referenz. Einarsson. Mürvet 1991. London: Arnold. Dougherty 1972. In Syntax and seman vol. Dixon. I. 1971a. Functional grammar. Diesing. Semantic ambiguity and un specification. Japanese. 9: Pragmatics. Speaker reference. Bredst Nordfriisk Instituut. and anaphora. a Mayan language. Mayers. Dede. Amsterdam: North-Holland. Renaat 1986a.

291–3 Stanford: Stanford University Press. Paris: Champion. and the count-mass tinction. Ferguson and Edith A. Foulet. How does a language acquire gender markers? In Univer of human language. 149–88. Fukui. ed. Greenberg. Gil. Lond University of Westminster Press. New York: Academic Press. 47–82. Reference and referent ac sibility. La sintassi dei sintagmi nominali quantificati. Definiteness. Jazyki Narodov SSSR 3: 177–98. Gildersleeve’s Latin Gramm London: Macmillan. F Linguistica Historica 2: 35–53. A theory of category projection and its applications. Moravcsik. Fraurud. 1978. In Subject and to ed. Charles A. 1982. Talmy 1976. Greenberg. Goodwin. Feoktistov. Icelandic. 3: Word structure. Kari 1990. Zeitschrift für Philosophie Philosophische Kritik 100: 22–50. Frege. noun phrase configurationality. T.) 1996. Feydit. Gottlob 1892. 1994. Moravcsik. Systèmes de déictiques. Frédéric 1969. 4: Syn ed. Anthony 1995. Joseph H. Henri 1944. Naoki 1986. Definiteness and the processing of noun phrases in natural course. Definiteness and referentiality. Petite syntaxe de l’ancien français. William W. with human language as the referee: toward an em ically viable epistemology. In Universals of human language.5: 245–76. Joseph H. 1978. Erzjanskij jazyk. Über Sinn und Bedeutung. On the development of the numeral ‘one’ as an indefinite marker. Journal of Pragmatics 6: 81–133. Linguistic Review 11: 241–55. Sag 1982. 1993. B. Walton-on-Thames: Nelson. Gildersleeve. A Greek grammar. Amsterdam: Benjamins. David 1987. Journal of Semantics 7: 395– 433. Referential and quantificational indefin Linguistics and Philosophy 5: 355–98. Joseph H. Fretheim. and grammatical agreement. 1981. Philip Baker. 1992. and Gonzalez Lodge 1895. vol. Janet Dean and Ivan A. vol. J. Topic. P. Grant. ed. Thornstein and Jeanette K. London: Hodder and Stoughton. Glendening. Padua: Unipress. Charles N.). La sintassi dei determinanti. L. Fodor. pragmatics. Stanford: Stanford University Pre 346 . Giuliana 1992. 254–69. P. Givón. Greenberg. Giusti. PhD the MIT. Lucien 1958. PhD the University of Padua. Li. Gundel (eds. In Reuland and ter Meulen (eds. Logic vs. Article agglutination in Creole French: a wider perspective From contact to creole and beyond. Frei. Enclitic articles and double definiteness: a comparative analysis of nom structure in Romance and Germanic. Acta Linguistica 4: 111–29. A. Manuel de langue arménienne (arménien occidental moder Paris: Klincksieck. pronoun. 1961. 149–76. 1966. Charle Ferguson and Edith A.

John A. Ken 1973. Hagman. Jacqueline 1980. New York: Holt. Nama Hottentot grammar. ‘articles’ and ‘third person pronouns French: changes in progress. 1979. Schn 93–139.. 1978. 308–44. Heny and H. A new Arabic grammar of the written guage. Colum University. Gustave 1975. Haspelmath. ed. The semantics of definite and indefinite noun phrases. The Scandinavian languages. revis Oxford: Blackwell. Definiteness and indefiniteness. Martin 1993. Jeanette K. PhD thesis. A grammar of Lezgian. Introduction to government-binding theory (2nd edition. Nahmad 1962. Liliane 1994. Cognitive status the form of referring expressions in discourse. J. An introduction to their hist London: Faber. . On (in)definite articles: implicatures and (un)grammaticality predict Journal of Linguistics 27: 405 – 42. Roy Stephen 1973. Natural syntax: iconicity and erosion. Irene R. 1989. ed. Zeitschrift für Romanische Philologie 93: 249– 1980. Step R. Haiman. Hale. Peter Cole and Jerry L. New York: Academic Press. Ja Fisiak. Lingu Inquiry 11: 637–78. Morgan. Hubert 1988. Language 69: 274 –307. Haider. Humphries. Definiteness and indefiniteness: a study in reference and grammatica prediction. M. London: Croom Helm. Zeitschrift Sprachwissenschaft 7: 32–59. Josef 1980. vol. Warlpiri and the grammar of non-configurational languages. In Historical morphology. Anderson and Paul Kiparsky. In Marácz and Muysken (eds. ed. Wiesbaden: Steiner. University Cambridge. ‘Demonstratives’. Einar 1976. Natural Langu and Linguistic Theory 1: 5– 47. and H. The marking of definiteness in Romance. Selections from the Third Groningen Round Table. Cambridge: Cambr University Press. F. Die Struktur der deutschen Nominalphrase. Oxford: Oxford University Press. Person marking in Walbiri. How do pronouns denote? In Syntax and semantics. 141–56. Haywood. PhD thesis. Berlin: Mouton.). Indefinite pronouns. Guillaume. Rinehart and Wins 1983. A. On the syntax and semantics of PP-extraposition. New York: Academic Pres Guéron. Martin 1977. John 1985. London: Lund. In A Festschrift for Morris Halle. Hartmann. 1991. Hausser. Haegeman. Gundel. New Y Garland. Haugen. 1974. 1997. Hawkins. Amharische Grammatik. Nancy Hedberg and Ron Zacharski 1993. The Hague: Mouton. On nonconfigurational structures. R.acts. Le problème de l’article et sa solution dans la langue frança Quebec: Presses de l’Université Laval. 1988. ed. Heim. 41–58. 293–3 Harris.

Giessen: Schmitz. 43–70. Heyer.chen Sprachen. Grammaticalization. Introduction to the grammar of English. Indefiniteness and predication. Amsterdam: North-Holland. Holmback. Lars 1986. The headedness of NPs in Norwegian. Zur Logik und Ontologie gen ischer Bedeutungen. ed. Lingui Inquiry 20: 506 –12. Linguistics Philosophy 1: 233– 45. Philip and Ian Hinchliffe 1994. Factors governing the morphological coding of referent Hausa narrative discourse. International Journa American Linguistics 32: 59–73. An interpretive solution to the definiteness effect pr lem. Heather K. The syntax and pragmatics of anaphora. Hewitt. Hoff. Hukari. Fred Householder and Robert Austerlitz. The syntax of Apollonius Dyscolus. In Features and projectio ed. Hockett. Yan 1994. Charles F. Pieter Muysken and Henk van Riemsdijk. Hinrichs. Jaeggli. Safir (eds. Paul J. Linguistics Philosophy 9: 63– 82. Rodney 1984. Generische Kennzeichnungen. 89–122. Translation with c mentary. J. Hintikka. Holmes. 1966. Temporal anaphora in discourses of English. Hamp. Objects of knowledge and belief: acquaintances and public figu Journal of Philosophy 67: 869– 83.) 1989. Hellan. PhD thesis. 2. Swedish: a comprehensive grammar. 1984. Roman 1966. In Reuland and Meulen (eds. On the definiteness of trace. Ioup. Levine 1989. 51–89. Gerhard 1987. Georgette 1977. B. Huang. and Elizabeth Closs Traugott 1993. Cambrid Cambridge University Press. and Robert D. Fred W. C. Lond Routledge. Hopper. On the distribution and reference of empty pronou Linguistic Inquiry 15: 531–74. Jaako 1970.-T. Huang. Linguistic Analysis 13: 195–215. A study with special erence to Chinese. Higginbotham. Munich: Philosophia. Osvaldo and Kenneth J. Jagger. Eric P. Philip John 1985. 226–53. Jakobson. Thomas E. Cambridge: Cambridge University Press. In Reuland and ter Meu (eds. 1981. Gesamtbedeutungen russischen Kasus. 1987. UCLA. Amsterdam: Benjamins. Existential sentences in Chinese and (in)definiteness. The null subject parame Dordrecht: Kluwer. James 1984. vol. What Algonquian is really like. The Carib language. Brian George 1979.). Dordrecht: Foris. Chicago: University of Chicago Pr 348 . Erhard 1986. 1968. In Readings in linguistics. Beitrag zur allgemeinen Kasuslehre.). Specificity and the interpretation of quantifiers. Abkhaz. James 1987. Householder. The Hague: Nijhoff. Cambrid Cambridge University Press. Huddleston.

Events. In Syntax and semantics. ed. The revised Latin primer. 1975. Heidelberg: Winter. Cop hagen: Munksgaard. II: Linguistic theory: extensions and implications. Hans and Uwe Reyle 1993. Partitive case and aspect. Stanford University. New Y Academic Press. T. vol. and Jonathan Stavi 1986. In Semantics from di ent points of view. Stokhof. Paul 1996. In Generalized quantifiers in natural language. . Negation in English and other languages. Urs Egli and Arnim von Stech 376– 417. Semantic theory. Grammar and conversational principles. ed. Eloise 1984. Nirit 1997. Journal of Pragmatics 28: 315–36. University of Massachusetts. Linguistics and Philosophy 9: 253–326. 21– 43. Johan Benthem and Alice G. Karttunen. Kempson. Ruth M. and Annabel Cormack 1981. Discourse referents. J. ed. 1976. Kempson. Part 7: Syntax. Janssen and M. Jong. A modern English grammar on historical principles. 7: N from the linguistic underground. Jelinek. Generalized quantifiers: the proper of their strength. Kiparsky. London: English Univers Press. V. Cambridge: Cambridge University Press. Remarkable subjects in Malagasy. ed. Hans 1979. M. Presupposition and the delimitation of semantics. Serbo-Croat. Jensen. Altarmenische Grammatik. case. New York: Academic Press. Javarek. 1984. instants and temporal reference. James D. and configurationality. Katarzyna M. Edward L. 363–86. Jespersen. McCawley. ed. Quantification and ambig Linguistics and Philosophy 4: 259–309. Kamp. Kamp. Frederick Newmeyer. Rainer Bäuerle. Keenan. Empty categories. From discourse to logic: introduction to modeltheo semantics of natural language. Charles N. A semantic characterization of nat language determiners. Edward L. London: Longman. 247–301. Kadmon. Benjamin Hall 1962. In Linguistics: the Cambridge Sur vol. 1977. Ruth M. Ms. Keenan. Berlin: Springer. Vera and Miroslava Sudjik 1963. 1– Dordrecht: Foris. A. Jaszczolt. In Subject and To ed. Cambridge: Cambridge University Press.Lingua 64: 291–323. Otto 1917. formal logic and discourse representation the Dordrecht: Kluwer. Groenendijk. Hans 1959. In Truth. A theory of truth and semantic representation. 139–63. G. Cambri Cambridge University Press. The ‘default de re’ principle for the interpretatio belief utterances. Dordrecht: Foris. 1988. ter Meulen. B. Kennedy. 1997. interpretation and in mation. Franciska de and Henk Verkuyl 1985. B. Copenhagen: Ho 1943. J. Lauri 1976. Natural Langu and Linguistic Theory 2: 39–76. On unique and non-unique reference and asymmetric quantifica PhD thesis. Li.

Carlson. Lambrecht. San Diego: Academic Press. Minneapolis: University of Minnesota Press. London: Hodder Stoughton. Lakoff. Mosc Izdatjel’stvo Akadjemii Nauk. La Galy. A guide. Barbara 1987. Hilda 1984. Remarks on This and That. 1975. Discourse configurational languages. 253–355 and 763–9. Frankfurt: Lang. French. 1985. H. Alice ter Meu Godehard Link and Gennaro Chierchia 1995. 1991. On deixis in English and Polish. Don Davidson and Gilbert Harman. V. 1977. E. T. Speaker’s reference and semantic reference. The independence of syntax and phonology in cliticizat Language 61. K. Charles H. 25: Perspectives on phrase structure: heads and licensing. ed. 345–56. Gramatika kurdskogo jazyka (kurmandzhi). In Carl and Pelletier (eds. The Hag Mouton. Manfred. ed. K. Topic. In Semantics for natural language. Georges 1981. The role of demonstrative p nouns. texts. Krifka. L’article LE générique. 350 . dictionary. Koopman. In Papers from the 10th Regional Mee of the Chicago Linguistic Society. Naming and necessity. M. Saul 1972. Information structure and sentence form. Kornfilt. Introduction to Biblical Hebrew. P. A case for emerging functional categories. Dordrecht: Foris. Susan Rothstein. 1982. La voyelle thématique -e-/-o.) Kryk.1995. and A. The Hag Mouton. Klavans. 1957. Kleiber. Judith L. Chicago: Chicago Linguistic Society. Genericity: an introduction. Jivi 1972. Thomas O. The syntax of verbs: from verb movement rules in the Kru guages to Universal Grammar. Kurdojev. Kripke. 1990. 95–120. Geneva: Droz. A. Aert H. Krupa. ed. K. Longm and Todd. In Contemporary perspective the philosophy of language. The Shuswap language. (Originally published as Polinjezijskije Jazyky. London: Routledge and Ke Paul. 1974. Knud 1994. In Syntax and semant vol. Oxford: Oxford University Press. focus and mental representations of discourse referents. Knobloch. Francis Jeffry Pelletier. Gregory N. Krámskm. Grammar. Kraft. The Polynesian languages. London: Darton. Moscow. Problèmes de référence: descriptions définies et noms prop Paris: Klincksieck. Kirk-Greene 1973. La généricité sur le mode massif. Fox Anthony Bruck. Jean 1952.). Uehling and H. Robert A. ed. Wettst 6–27. Dordrecht: Reidel. Robin 1974. 1–124. The article and the concept of definiteness in languages. Hausa. J. Cambridge: Cambridge Univer Press. Kuipers. 1973. Lambdin.serait-elle un indice d’objet in européen? Lingua 3: 407–20. Michael W. 11–35.

Thompson 1976. 1985. Nunation in afrikanischen Sprachen. University of Michigan Paper Linguistics 1. Laughren. Sebastian 1985. 81–95. A grammar of the Macedonian literary language. 1973. Mary 1989. 457–89. Li.: MIT Press. Johan Van der Auwera. 1967. Transactions of the Philolog Society 88: 1–57. 1990. A possessive parameter. 1983. In The semantics of determin ed. Lunt. New York: Academic Press. Oxford: Oxford University Press. . J. Sheffield Working Papers in Language Linguistics 2: 98–104. Mandarin Chinese: a functional reference grammar. Levinson. 319–53. 1952. Lond Croom Helm. 1– 41. Cambridge: Cambridge University Press 1987. 1981. The configurationality parameter and Warlpiri. Leslau.5. In Marácz Muysken (eds. Lewis. Giuseppe 1994. Pragmatics. Rado L. An agreement approach to clitic doubling. Lencek. and Z. Amharic textbook. Longobardi. English nationality terms: evidence for dual category membership. Christopher 1977. Charles N. Lee. Definites. G. Cambridge. Stephen C. R. The syntax of English genitive constructions.semantic theory. Charles N. Li. Journal of Linguistics 23: 379–434. Lukas. Anthropos 63: 97–11 Lumsden. The meaning of the English definite article. 1991. Ludlow. Berkeley: Universit California Press. London: Croom Helm. Colum Ohio: Slavica. 1983. Turkish grammar. Linguistic Inquiry 25: 609–65. Johannes 1968. Lyons. Czech. Peter and Stephen Neale 1991. Lawler. Berkeley: University of California Press. 1982. Skopje: Drzha Knigoizdatelstvo. Mass. Studies in English generics. Subject and topic: a new typology Subject and Topic. Horace G. London: English Universities Press. Jou of Literary Semantics 20: 97–116. On the origin of the Old French strong–weak possessive distinct Transactions of the Philological Society.). L. Indefinite descriptions: in defense of Rus Linguistics and Philosophy 14: 171–202. Reference and proper names: a theory of N-Movemen syntax and logical form. W. Wolf 1968. Pragmatics and the grammar of anaphora: a partial pragmatic reductio binding and control phenomena. Journal of Semantics 4: 279–326. Principles of pragmatics. Michael 1988. The structure and history of the Slovene language. 1986a. The demonstratives of English. Geoffrey N. and Sandra A. Journal of Linguistics 22: 123– 1986b. Leech. Belfast Working Paper Language and Linguistics 2. Existential sentences: their structure and meaning. London: Longman. ed. Lee 1959. 1980. Löbner.

John 1975.und Satzfügung. Manchester. 1995b. James and Kenneth Hale 1984. László and Pieter Muysken (eds. 77–1 Amsterdam: Benjamins. University of Salford E Working Papers in Language and Linguistics 8. Parlons wolof. generics. Hidatsa syntax. Maetzner. The Hague: Mouton. III: Die Lehre von der Wort. Englische Grammatik. Paper presented at the 21st Internatio Romance Linguistics Conference. H. 1986. Mayers. The Hague: Mouton. langue et culture. The origins of definiteness marking. Pocomchi. Paper presented at the Autumn Meeting of Linguistics Association of Great Britain. I: Die Lehre vom Worte. Maclaran. Lingua 2: 14 –31. Definiteness and person. Cambridge: Cambridge University Press 1977. Malherbe. ed. Dordrecht: Foris. Ms. John Charles Smith and Martin Maiden. Ber Weidmannsche Buchhandlung. In Formal semantics of nat language. Eduard 1880. Mace. Definiteness marking in South Asian languages. B. James D. 1995e. Paper presente the ESRI Research Seminar. Everything that linguists have always wanted to know ab logic but were ashamed to ask. ed. vol. Proper nouns. G. Marvin K. Mayers 1966. vol. Mayers. University of Salford. Lyons. convergence and diglossia. Marácz. Pa L’Harmattan. 1995c.Papers in Language and Linguistics 15. and the count–mass distinction. K. Semantics. aspect. 61–83. Natural Language and Linguistic Theory 1: 487–533. Masica. 1995a. In Linguistic theory and Romance languages. Determiners and noun phrase structure. L. Be Weidmannsche Buchhandlung. Cornell University. 1885. John 1971. 87–109. Matthews.) 1989. Movement in ‘NP’ and the DP hypothesis. Marvin K. Matthews. University of Salford. ed. and arbitrary arguments. The semantics and pragmatics of the English demonstrati PhD thesis. 352 . and Marilyn A. McCloskey. In South As languages: structure. London: Hodder and Stoughton. Krishnamurti. Configurationality: the typolog asymmetries. Paper presented at the 12th Internatio Conference on Historical Linguistics. 123– Delhi: Motilal Banarsidass. 1994b. Matthews. Oxford: Blackwell. 1995d. 1994a. ed. Deixis as the source of reference. Englische Grammatik. Cambridge: Cambridge University Press. E. On the syntax of person–number infl tion in Modern Irish. Michel and Cheikh Sall 1989. Stephen and Virginia Yip 1994. Cantonese: a comprehensive gramm London: Routledge. Colin P. 1965. 1981. London. Modern Persian. Palermo. Rose 1982. W. In Languages Guatemala. McCawley. 1949. The Polynesian articles. Keenan. Voice. Pronombres y persona en español.

Chicago: Chicago Lingu Society. On the case marking of objects. Vance. The English verb. Milsark. Gary L. Stephen 1990. Nater. Mithun. ed. Mohanan. 412–22. Grossman. Richard 1974. Meillet. Minassian. Marina and Irene Vogel 1986. In Universals of hu language. 493–509. Paris: Klincksiec Mitchell. 1983. 249–89. J. Stanford: Stanford University Press. AGR(eement) in the German noun phrase. F. Elisabeth Löbel and Claudia Schmitt. R. Susan 1989. The Bella Coola language. Thomason. International Journal of Amer Linguistics 35: 299–306. A. 247–70. On generics. J. Colloquial Arabic. vol. Newmark. Bella Coola paradigms. K. Stanford: Stanford University Press. 1978. Berkeley: Berkeley Linguistics Society. In Proceedings of the 10th An Meeting of the Berkeley Linguistics Society. Descriptions. 1984. Lexical and configurational structure. Standard Albanian reference grammar for students. Joseph H. Manuel pratique d’arménien ancien. New York: Garland. O Scandinavian University Press. A grammar of Lango. Moravcsik. Inferring quantification in generic senten In Papers from the 11th Regional Meeting of the Chicago Linguistic Society R. 1979. Anna 1996. T. Martin and Virgiliu jtefznescu-Drzgzneiti 1970. H. How to avoid subordination. London: Longman. P. Newman. Samoan reference grammar. F. 4: Syntax. Dordrecht: Foris.: MIT Press. Mosel. UCL Working Papers in Linguistics 8: 165 . Edith A. Prosodic phonology. L. Toward an explanation of certain peculiarities of the exis tial construction in English. Linguistic Analysis 3: 1–29. Leonard. New Haven: University Press. Fundamenta Mathem icae 44: 12–36. On a generalization of quantifiers. Ulrike and Even Hovdhaugen 1992. Existential sentences in English. ed. Michael 1992. Nespor. Montague. ed. 39– Amsterdam: Benjamins. The proper treatment of quantification in ordinary Eng In Formal philosophy. Olsen. Noonan. Marianne 1984. Berlin: Mouton de Gruyter. Ottawa: National Museums of Can Neale. Claudia Brugman and Mo Macaulay. London: Hodder and Stoughton. Mostowski. Nunberg. 1974. Esquisse d’une grammaire comparée de l’arménien classi Vienna: PP Mekhitharistes. Philip Hubbard and Peter Prifti 1982. Andrej 1957. F. 1962. 1936. Christa Bhatt.Nouvelles Editions Africaines. E. Murrell. Richmond H. Martiros 1976. Ferguson and E A. In Syntactic phrase st ture phenomena. Greenberg. Linguistic Review 113– 41. San and T. Mass. 1977. Stanley 1969. Papafragou. Romanian. Moravcsik. Charles A. ed. Lond Hodder and Stoughton. Palmer. Geoffrey and Chiahua Pan 1975. Cambridge.

In He in grammatical theory.: Ginn. On so-called pronouns in English. Rosenbaum. 1970. Norman M. V. Head-hunting: on the trail of the nominal Janus. Probus 279–316. 1993. Tomlin. London: Harrap. 1970. 233– 48.: MIT Press. On the inferencing of indefinite-this NPs. Fehime undated. William F. Perlmutter. In Proceedings of the F Annual Meeting of the Pacific Linguistics Conference. Manfred Bierwisch and Karl Erich Heidolph. Mass. Ramsden. Cambridge. Elementary Albanian: Filltar i shqipes. Mass expressions. Perrott. Philippi. Greville G. ed. S. Postal. Cambridge: Cambridge University Press. Transformational grammar: a first course. In Elements of discourse und standing. 327– 4 Dordrecht: Kluwer. ed. 231– Cambridge: Cambridge University Press. Francis Jeffry and Lenhart K. Pustejovsky. Picallo. Eugene: University of Oregon. Guenthner. 56– Waltham. James 1995. In Parame of morphosyntactic change. Paul R. The Hague: Mouton. Ellen F. ed. 1951. London: Hodder and Stoughton. Corbett. Washington: Center Applied Linguistics. David M. Chicago: Chicago Linguistic Society. Carme 1991. On the given/new distinction. 62– Cambridge: Cambridge University Press. Webber and Ivan A. Nicholas 1970. The generative lexicon. 1979. Hofbauer. D.7: 243– 86. Greville G. Ramsay. Prince. Fraser and S McGlashan. Payne. Nominals and nominalizations in Catalan. 289–312. vol. 267–78. Joshi. ed. Julia 1997. Jacobs and Peter S. 1981. Swahili. Sag. Rando. Mongolian language handbook. M. Pattee. Pelletier. The headedness of noun phrases: slaying the nominal hydra. The rise of the article in the Germanic languages. ed. Classifiers and referentiality in Jacaltec. On the article in English. Clyne. In Handb of philosophical logic. Language 300 –13. Emily and Donna Jo Napoli 1978. In Progress in linguistics. Cambrid Cambridge University Press. Roderick A. ed. 114 –39. Norman M. In Heads in grammat theory. Ha and Carol L. Cambridge: Cambridge University Press. H. 4. DeLancey R. 354 . Boston: Vatra. V. Pipa. Fraser and Scott McGlash 73–113. Tübing Niemeyer. 1959. Gabbay and F. Aravind K. Schubert 1989. Mass. ed. In Papers from the 15th Regio Meeting of the Chicago Linguistic Society. Radford. Corbett. 1985. Joseph 1994. Bonnie L. In Readings in English tra formational grammar. John 1993. Ans van Kemenade and Nigel Vincent. ed. Definites in there-sentences. Andrew 1988. Poppe. Le problème de l’article: sa solution en allemand. Paul M. An essential course in Modern Spanish. D.

Elements of symbolic logic. with an introduction to the c sical language. Mind 14: 479–93. Groningen. Mass. Bar D. textes et lexique. Etude sur les dialectes berbères des Beni Iznassen. Studi di Gramma Italiana 5: 5– 42. Louisa 1988. Reuland. Rouchota. Sadler. vol. Rothstein. Ruhl. Rosen. Paris: Leroux. M. Canadian Journal of Linguis 37: 197–218. Paul 1994. Reinhart. Journal of Linguistics 30: 441– Rowlands. 1979. Charles 1989. 217–36. Hans 1947. du Rif. A. San Diego: Academic Press. Villy 1994. The extended projection principle and the definiteness ef In Proceedings of the West Coast Conference on Formal Linguistics 2. New York: Augustin. Pragmatics and linguistics: an analysis of sentence top Philosophica 27: 53–93. London: Hodder and Stoughton. Southern Illinois University 4. 1969. Rowlett. Dordre Foris. ed. Tanya 1981. Reichenbach. 1992. Susan and Ann Reed 1984. A descriptive grammar of Ewondo. B. Gladys A. The Negative Cycle.: MIT Press. Stanford: Stanford Linguis Association. Eric J. James E. Eric J.linguistics. Cross-linguistic evidence for number phrase. Redden. Cambridge. London: Macmillan. The representation of implicit and dethematized subje Dordrecht: Foris. On monosemy: a study in linguistic semantics.) 1987. On indefinite descriptions. 1932. Welsh syntax: a government-binding approach. Two functional categories in noun phrases: evidence from Modern Heb In Syntax and semantics. 1987. C. Definiteness and set determination. Susan D. Grammaire. ed. London: Croom H . Yoruba. Wescoat. A textbook of Israeli Hebrew. 25: Perspectives on phrase structure: heads licensing. Reichard. Elizabeth 1988. Anaphora and semantic interpretation. et Senhaja de Sraïr. 373–90. Russell. A head-movement approach to construct-state noun phra Linguistics 26: 909–29. Chicago: Chicago University Press. Renisio. Albany: S University of New York Press. negative concord and the nature of s head agreement. E. 1962. Ian G. and Alice G. Roberts. 1951. Flickinger and M. 1983. 1983. The representation (in)definiteness. Rothstein. Ritter. Paper sented at the Fifth Groningen Round Table on (In)definiteness. ed. Osvaldo Jaeggli and Carmen Silva-Corvalán. Bertrand 1905. Renzi. 37– 62. Occasional Paper Linguistics. On denoting. 1991. Grammatica e storia dell’articolo italiano. Haiim B. ter Meulen (eds. London: Croom Helm. Reuland. University of Salford ESRI Working Papers in Language Linguistics 7. Lorenzo 1976. Navaho grammar.

vol. M. Smith.1987. V. William J. Samarin. A grammar of Sango. M. The major s tactic structures of English. Speas. On the so-called definite article in English. Relevance: communication and cognition. types and meaning. ed. Cambridge. Elisabeth O. Mario 1988. Silverstein. Sperber. Pragmatic presuppositions. Sohn. 1972. 1987. N. What explains the definiteness effect? In Reuland and ter Meulen (eds. 1990. 71– Saltarelli. Munitz and P. Columbus. A Lithuanian historical syntax. The evolution of determiners in Chadic. 1983. ed. 1974. Précis of Relevance: communication and cognition. B. Schuh. A reference grammar of Modern Bulgarian. K. Oxf Blackwell. G. 356 . London: English Universities Press. revised). I: Clause structure. Margaret J. 27– 48. Lenhart K. In Language typology and synta description. London: Routledge. Dixon. van der 1988. Ho-min 1994. Jou of Linguistics 28: 313– 41. 1967. Alan H. ed. 1972. Phonology and syntax: the relation between sound and st ture. Hamburg: Buske. Sandt. Columbus. Selkirk. Rinehart and Winston. Context and presupposition. Timothy Shopen. Stalnaker. Lingui Inquiry 3: 197–209. Paul Schachter and Barbara Hall Partee 1973. The Hague: Mouton. 1983. K. Turner. Sommerstein. Oxf Blackwell. Shackle. N York: New York University Press. Ekkehard Wolff and H. Mass. Schachter. Basque. Behavioural and B Sciences 10: 697–754. Robert P. Pa and R. Russell G. Dordrecht: Kluwer. 1975. 197–214. Ohio: Slav Schubert. Unger. Canbe Australian Institute of Aboriginal Studies. Hierarchy of features and ergativity. Stockwell. In Semantics and philosophy: stu in contemporary philosophy. 3–61. London: Croom Helm. Maria 1992. R. 1984. Paul 1985. Dordrecht: Kluwe Spencer. 193–268. In Grammatical c gories in Australian languages.. vol. R. Korean. London: Croom Helm. Nominal inflection and the nature of functional categories. Transactions of the Philological Society. and Francis Jeffry Pelletier 1989. Ernest A. Selig. ed. Phrase structure in natural language. O Slavica. In Studies in Cha and Afroasiatic linguistics. Michael 1976. Parts-of-speech systems.). Scatton. New York: Holt. C. Cambrid Cambridge University Press. Relevance: communication and cognition (2nd edition. Die Entwicklung der Nominaldeterminanten im Spätlat Tübingen: Niemeyer.: MIT Press. Andrew 1992. ed. Rob A. W. 2: Semantic issues. William R. Punjabi. Meyer-Bahlb 157–210. 1987. Dan and Deirdre Wilson 1986. 112–71. Generically speaking Properties. On generics. Schmalstieg. 1995. Chierchia.

Van Valin. ed. Case marking and the structure of the Lakhota cla In Grammar inside and outside the clause: Some approaches to theory from field. 1994. ed. Linguistic Inquiry 23: 595– . Marleen 1989. Tritton. D-projections and N-projections in Norwegian Grammar in progress: GLOW essays for Henk van Riemsdijk. Cha A. Joseph H. Travaux de Linguistique 18: 45–56. Johanna Nichols and Anthony C. 419–31. 1964. Szabolcsi. Tura. Definiteness and referentiality in Turkish nonverbal tences. A. Managua: Board of Chris Education. San Di Academic Press. Ferenc Kiefer and Katalin É. ed. vol. Traugott. Jean-Roger and María Luisa Zubizarreta 1992. Copenha Rosenkilde and Bagger. ed. Kiss. ed. Approaches to grammati ization. Miskito grammar. vol. Anna 1987. Moravcsik. Dordrecht: Foris. Vergnaud. 167–89. Taraldsen. Greenberg. Sabahat Sansa 1986. Sze JATE. Gunnar Olaf 1958. Jorge A. Juan 1995. The definite determiner the inalienable constructions in French and in English. Cambri Cambridge University Press. S. vol. Svane. 363– 413. The case filter and licensing of empty Canadian Journal of Linguistics 37: 157–74. 1983. In Syntax and semantics. The noun phrase. Enric and Elisabet Engdahl 1996. 1977. Suñer. Nat Language and Linguistic Theory 6: 391– 434. Philosophical Review 73: 439– Suárez. The role of agreement in clitic-doubled constructions. Russell 1978.) 1991. Intention and convention in speech acts. Amsterdam: Benjamins. 2: Theories and analyses. In Approache Hungarian. Thaeler undated. Joan Mas and Marina Nespor. Aspects of the syntax of clitic placement in Western Roma Linguistic Inquiry 26: 79–123. Introduction to logical theory. Ferguson and Edith A. and A. Non spécificité. Van Peteghem. Amsterdam: Benjamins. Cambridge: Cambr University Press. M. István Kenesei. Functional categories in the noun phrase. Grammatik der slowenischen Schriftsprache. In Studies in Turkish linguistics. Travis. attributivité et article indéfini dans langues romanes. London: Methuen. Margarita 1988. 211– 48. Knut Tarald 1990. Ultan. Woodbury. 4: Syntax. London: Hodder and Stoughton. Elizabeth Closs and Bernd Heine (eds. The linguistic realization of informa packaging. D. Stanford: Stanford University Pr Uriagereka. Arabic. The Mesoamerican Indian languages. Linguistics 34: 459–519.1952. 27: The syntactic struc of Hungarian. 179–274. Vallduví. Some general characteristics of interrogative systems Universals of human language. Lisa and Greg Lamontagne 1992. Dan Isaac Slobin and Karl Zimm 165–94. Robert D. D. Thaeler. 1985. ed.

Innsbruck: Institut für Sprachwissenschaft der Universität Innsbruck. Cliticization vs inflection: Eng n’t. Welmers. Deirdre 1975. Persian grammar: history and state of its study. Oxford: Oxford Univer Press. Raffaella 1991. Wilson. ed.. E. S. London: Oxford Univer Press. There-insertion. Philosophical Review 87: 48– Windfuhr. Catalan. A grammar colloquial dictionary. Alan 1975. Gernot L. Reference and relevance. C. Watahomigie. 1978. 1992. 1892. Hualapai erence grammar. A. On definite and indefinite descriptions. Arthur H. 567–99. Young. In Formal methods in the study of language. The Navaho language. A study of the Ewe language. and Geoffrey K. Louw and P. In Phonologica 1976. University of Pennsylvania. Pfei 23–39. (Originally published as Grammatik der Ewe-Sprache. Diedrich 1960. Syntactic properties of sentential negation: a compara study of Romance languages. PhD thesis. Jansen and M. Zwicky. Janis S. 1984.) Whitney. Arnold M. Studies in Lakhota grammar. Wright. V. J. 1956. J. G. and William Morgan 1987. The emergence of the D-system in Romance. London: English Universities Press. In Parameter morphosyntactic change. Arnold M. Pullum 1983. London: Hodder and Stoughton. ed. Williamson. Grammar of the Gothic language. 1976. Presuppositions and non-truth-conditional semantics. A handbook of the Zulu langua Pretoria: Van Schaik. B. J. D. University California at San Diego. Language 59: 502–13.interpretation. An indefiniteness restriction for relative clauses in Lakhota. Lucille J. Yates. Yamamoto 1982. On clitics. The Hag Mouton. Lond Academic Press. Dressler and O.. T.). Ans van Kemenade and Nigel Vincent. Groenendijk. Linguistic Inquiry 15: 131–53. W. Los Angeles: American Indian Studies Center. UCL Working Papers in Linguistics 4: 167–91. Berkeley: University of California Press Westermann. Erich 1983. Über ein Gesetz der indogermanischen Wortstellu Indogermanische Forschungen 1: 333– 436. On the question of definiteness in ‘an old man’s bo Linguistic Inquiry 14: 137–54. 1907. E. Robert W. Zwicky. M. Williams. Wackernagel. UCLA. Stokhof. Taljaard 1967. 1987. Jorigine Bender and Akira Y. Finnish. A grammar of Vai. J. PhD thesis. 1984. Wilson. 358 . 149– Cambridge: Cambridge University Press. 1979. Nigel 1997. Ziervogel. Joseph 1910. In Reuland ter Meulen (eds. Berlin. Albuquerque: University of New Mexico Press. Amsterd Mathematisch Centrum. Zanuttini. Vincent. 168–90. ed. Woisetschlaeger. 1977. A. W.

271–4 accusative case morphemes 199–207 Adams. 323 and the specifier position 287–8 affixation. 73 “adjectival article” 79 adjectival demonstratives. 242n anchor. 28.INDEX A-article in Fering 161– 4. J. 296 person expressed in 311–12. 336 absolutive case 200 –1. P. 294. referent to speaker by relative 229 . of articles see bound articles AG languages. 265 ambiguity or vagueness 158n see also scope ambiguities analytic structure. possessives and definitene in 24. 333 affix inflectional or clitic 63 use of term 63 affixal articles 287–8. N. 340 indefinite 89–95 inflection of 68 –71. 290 and AgrP to represent possessive agreement 294. Specifier-Head Agreement 284 287 agreement affixes. 217–19 articulated and non-articulated 79 definiteness-indefiniteness distinction 82–5 gender neutralization pre-vocalically 218 plus definite articles 60 position of 43 reduced form when pre-nominal 289 serialization of in APs 296 strong and weak declensions 85. 314 –18 agreement. 289. 332 anaphors 30 governing category of 30. 329. 202 abstract nouns 218n accessibility of context 271–2 of the referent and use of article 52–3. 297– 8. W. A. shift from synthetic structure to 324 anaphora and deixis 160 –1 identity of sense 28. 135 Allwood. 332. and clitics 208–9 agreement marking 86–7. J. 130 – 4. 273. 339 see also definite adjectives adjunction 42. 337 AG possessives and DG possessives 293. 324. 315–1 agreement morphology. 303 “aboutness” 227–8 Abraham. 199–207 Aebischer. 207–13. 326 –9 as realizations of D 299. 324 position of 285 agglutinating languages 101 Agr 43. 198. 233 Akmajian. 272. 290. 258 anticipatory see cataphoric use in chain of hyponymy 319 of definite suffix 52. 283 adpositional marking 86. P. 292. 158 – 60. 165. 38n strict 28 anaphoric articles 53–4 compared with demonstratives 106 anaphoric demonstratives 113–15 person-based 113–14 and tone 115 anaphoric use 4. S. and non-overt argument interpreted pronominally 308 –10 Aissen. L. 53 of definites 270. J. and affixal definite articles 120–1. 335 Abney. 323 “adjectival particles” 79–81 “adjectival-genitive” see AG languages adjective phrase degree modifiers in 46 serialization of attributive modifiers in 296 adjectives and article inflection 215–16.

204. J. 208 bare indefinites 103–5 bare nouns non-generic 197n plural count nouns 189–92 use to indicate indefiniteness 96–7 Barwise. non-referential use of generics 186 “appropriateness condition”. 109 Bauer. M. 73 Arrigaray. 177. 242–3 Principle A 30. 289 –90. F. 242n Principle C 30. bound articles. A. 84. 108 body parts. free-form articles. 122. 154. cardinal articles. 128–9 Bokamba. 124 Andrews. 205. Heim’s 269–70 Aquilina. P. definite articles. E. 128 Belletti. R. J. 274 Boas. affixal articles. 7. 208 Bickerton. H. 60–2 head or specifier 301. 2 noun phrases 32n see also associative anaphoric use Budi!a-Lazdi!a. D. M. preposed articles. 231. 154. P. 321 inflected forms 67 invariable 67 and nominal feature marking 215–19 nominalizing function 60–2 occurrence in languages xv. 243 see also “generalized binding theory” (Aoun) Bird. J. 7. indefinite article article inflection 215–19 and DP analysis 217–19 article reduction 65–6 articles “actual-determinate” and “hypothetical indeterminate” 59 category status of 282. 220. 106 see also definite article. 219. 120. 61. M. 222 Bastuji. 309 Baker. H. 106. 53 Armbruster. 310n Bánhidi. H. 78 bound articles as head 301. 285–8 in competition 334 –5 distinguished from determiners 106 expansion in use 336 –7 functions of 2n. 150 arbitrariness 89 “areal features” 48–9 argument structure 89 arguments agentive or non-agentive 240–1 missing supplied pragmatically 308 Ariel. 60 Borer. possessives 25n. C. 50. 3 bridging cross-reference use 4 –5. 26. R. 210. J. 84. 61. C. 103. B. 273 and bound-variable anaphora 269–70. 326–7 see also affixal articles bound-variable pronouns 31 brand names 195 Bresnan. 294 associative anaphoric use 32n. 111 . 272. Z. 103 Bender. C. 286. 105 Berent. G. 58 –9. 301 status in realization of inflectional features 215–19 use in Romance languages 336–7 see also “adjectival article”. 336 Biggs. M. 143n animacy and definiteness 213–15 and inanimacy 109 and object marking 204 –5. B. 295. C. 212 animacy hierarchy 213–15. 57. 68–71. 60n. see also bridging cross-reference use associative uses 4 –5. W. 112 Benjamin. 321 inflection or clitic 63.Andersson. 108. 267–8 Basri. 154. L. 158 attributive or non-specific use 172 versus referential use 171 attributive modifers. 6. 309 Blake. J. de 76 article defined 36 use of term 47. 250 Bittner. 265 Andrade.-G. 204n Börjars. 319 Aoun. 242n Principle B 30. 221 Apollonius Dyscolus 253 appropriate context. 153 Blass. 52. quasi-indefinite articles “articloid” (Aebischer) 333 360 aspect 180. J. 254. G. 158. 163–4. T. K. serialization of in AP 296 Austin. postposed articles. 128 binding theory 30. 266. 220. J. 48–50 the origin of 331–6 pleonastic or default function 99. J. J. 219–20. C. 76. 45n.

definite article as a 60 complementizer phrase (CP) 219 complements 41 complex definiteness 279–80 complex definites 16 and complex indefinites 107–56 listed 107 complex indefinites 36–41. W. 321 absence in bare nominals 300n. H. 181. 148 definite and indefinite expressions 36 numerals as specifiers of 296 cardinality determiners. 191. 181 Burton-Roberts. 238 Cinque. 123 case in D position 324 –5 and definiteness 324 loss or reduction of morphological 324 –5 marked and unmarked 295 – 6 and referential properties 295 see also K (case). 258 Butt. 111 Cole. 321. N. 189 –91. 124. 35n. 335–6 and double determination 100 free and bound forms 106 indirectly signal indefiniteness 49. D. 302 coda 238. 265. C.Bunt. 245 cognition. 197n Carlson. G. 193. 280 colonization 48 ColourP 296 comment 190. 245. 294 –5 clitic. use of term 63 clitic definite article 82–3 clitic doubling 78. and noun phrase structure 290–1. L. 303 cardinality and/or indefiniteness 33–6. 296. KP analysis case morphology. 333 Chafe. 301. 228n. 148–50 and complex definites 107–56 and simple indefinites 33–41 use of term 36 . 89–95. “new comparative linguistics comparatives. 132. G. A. 220n. 210 and definiteness 210 with a preposition 210 clitic pronouns clitic doubling 138. 8. 231 checking theory 297 Chesterman. 243. 138. 219. 215 vocative 152 comparative approach see cross-linguistic survey. see also topic-comment opposition common nouns distinguished from proper nouns 194 –8 with human reference 214. 273 Cann. 187–8. numeral classifiers as realizations of 306 –8 cardinal articles 34 –6. 300 Clark. H. 181. 306 indefinites and 330 Carlson. 266 Card. P. 208. 254 Chung. 297. N. 326 and second-position attachment 72–6 clitics and agreement affixes 208–9 and inflectional affixes 63. 104 –5 defined 36 development of 330–1. 271–2 Cole. 52. 302 Christophersen. 265 Chierchia. 160. 283n. 105 c-command 31. 185–6. R. sim clitics co-indexation 242–3. numeral classifiers clause. 332. 194. object clitics. and numerals 45 cardinality position 47. 140 clitic placement 303–4 as D heads 304. 103. 3. 183. 140. 103. P. S. 194n. T. 265 class 179–80 class predicates 179 or “variety interpretation” 187–8 class-qua-class 183 classifiers 225–6 and degree of grammaticalization 226– see also noun classifiers. Ruth 64. Robert 112 Carlson. and QP 45n complementizer. A. 300–1 in non-configurational languages 155–6 optional 98–9 with plurals and/or mass expressions 99 types of 100 see also quasi-indefinite articles Cardinaletti. 182. G. J. N. H. 299. 329 cliticization 66 or agreement 287–8. 204. 248 Chomsky. 105–6 CardP 300. preserved on determiners 324 –5 case variation 200–2 cataphoric use 5. and the concept of relevance 264 –5. 76 phrasal 72–7 pronominal representing arguments 308 –10 see also dative clitics.

A. M. 189. 184 plural. 181. theory of 220. 332. 265 Dede. G. 63. 213. R. 288–90 definite adjectives 82–5 definite article. 205. 30 321 types 62–85 as universal quantifiers 11 see also anaphoric articles. 31. 299–301 substantive or expletive uses 163– 4. 266. 183 Cornilescu. 137 Cooper. A. demonstratives in 29n Corbett. R. M. W. 332 presence or absence of 253–4 semantically unmotivated 217–18 as specifiers of DP 298. G. 174. 337–9 overlap in function with demonstratives 164 –5. 73. G. R. 243 copular sentences. and definiteness 10–11 count–mass distinction 9. 63. 323 languages lacking 309–10 D-article in Fering 161–3. 174n. 107–56 definite-indefinite distinction 47–106 D grammatical category of definiteness 298–301. Ö. K. 194 and configurationality parameter 155–6 Cowan. J. 116.-D. 125 Cormack. 265 Dambriûnas. 115 Cooke. 231. clitic defini article. in indirect object marking 211 de Bray. 335 362 Dahl. 106 inflectional function 324 languages without 236. 279–80 mixed systems and double determinatio 77– 82 in null form 300–1 optional 52–3. 323 and person 313–18 and specifier position 283–7 as a subset of the strong set 267–8 definite noun phrases logical or formal semantic analysis of 254 –5 as referring expressions 3n types of 15 –33 definite personal forms 145–8 . 331– 4 as determiners 26–7. 196 “bare plural” 189–92 genericity of 180. 164 –5. 55. G. 288–90 explained 289–90 definite description reference is to the maximal set 12n use of term 7n definite determiners demonstratives as 105. 95 Deemter.250 conditionals 168 configurational languages 154. 264 –5 Cook. 295 Coulson. 332 position of 44. 195. defined 36 definite article cycle 326–31 definite articles 1. 223 conventional implicatures 264 conversational implicatures “defeasibility” of 264 Grice’s 257–9. and mass nouns 189 in the plural without determiner 33 singular with a 34 singular or plural. 108 dative clitics. 214 cross-linguistic survey xv. L. xvi. E. 320 –1. contradiction or new information 271 contrast demonstratives 115 “control”. 331 Croft. 82. 283 configurationality parameter 154 –6 and count–mass distinction 155 and definiteness 305–10 constant situational basis 254 “construct”. R. C. 267–8 copula ‘be’. 35–6 free forms and bound 53. 285 – 6. explicit or implicit 254 contextual ecects: confirmation. 208 de dicto readings 168 de re readings 168 Declerck. 60 Craig. 334n origins and evolution of 326–30. 174. in existential sentences 237–41. of Semitic languages (possessive follows the head) 131 context accessibility of 271– 4 and logical form 271 and truth or falsity 256 contextual basis. 15 as complementizers 60–1 derived from demonstratives 105. van 174n deep analysis xvi Def feature 16. 135 count nouns 8. M. 225 creole languages English-based 336 French-based 299. free-form articles definite constraint 282. 83. A.

265–71 and salience 215 semantic categories of 15 semantic or grammatical 226 and specificity 57–60 specificity hierarchy (Comrie and Croft) 214 –15 and structural (Det) position 153 summary of conclusions 274 –81. 322 and grammaticalization 253. 107–21 degrees of deixis 18–21. 40. 289 –98. five or six 111 inherent definiteness assumption 107–2 152. 277–8. 151– 6 with no distance or person contrast 112 three-way contrast 55–6. 302 in copular sentences 29n as definite determiners 105. 21 deixis 18 –21 and anaphora 160–2 as at the root of definite reference (J. 28n demonstrative feature 20 demonstrative personal forms 145–8 demonstrative pronouns 28 anaphoric use 28 non-anaphoric use 28–9 situational uses 28 demonstrative systems anaphoric use in person-based 113–14 languages with more than one 109–10. 253 definiteness ecects 16 –17. E. 107–13. and double determination 78 diachronic aspects of 322– 40 as a functional head rather than a feature 160n grammatical. 273– 4. 227–52 in existential sentences 236–48 internal-head relatives 250–1 PP-extraposition and any opacity 248–9 property predication 249 –50 superlatives 246 –8 definiteness marking 47–51. 323 anaphoric determiners as 53–4 co-occurrence with definite article 118 279. 340 absence of 305 – 6. Kempson’s syntactic rule of 258 – 60 Deg (degree modifiers) 46 deictic distinctions 55–7 and visibility 112 deictic features of demonstratives 18–21 loss of 331–2 deictics. 110 –11. 274 –5. 320–1 testing noun phrases for 15–17 typology of languages in terms of 280 –1 a unified phenomenon? 157– 65. pointing role 20n. 274 as a grammatical category 16n. 284. 61. 323 definiteness of 17–21. 50. 302 compared with possessives 133. 107–21. 226. 108–9. 145 in Det position 118–19. 86 –8. E. 160 Deloria. 307– 8 emergence and development of 322–40 inflection analysis of 70n languages without 277–8 acceptability data for 262–3 anaphoric and non-anaphoric 158–60 “deictic” and non-“deictic” 160–5 non-referential use of 273–4 semantic content and behaviour 52–62 semantic and pragmatic distinctions 157–73. 320 identifiability and inclusiveness 13–15 and information structure 232–6 interaction with other grammatical phenomena 199–226 and the noun phrase 41–5 and noun phrase structure 282–322 other ways of expressing 86–9 and person 310–20 of personal pronouns 26–32 and possessives 22–6. 178 see also complex definites. 278 “matching constraint” on (Hawkins) 20 . 120 discontinuous 117 forms of 116–21 four. 151– demonstratives xv. logical and pragmatic traditions 253–60. simple defin definitization. 277–8. 110–1 151–2 two-term contrast 107– 8. 130–4 of proper nouns 21–2 and quantification 32.definite-indefinite distinction cross-linguistic survey 47–106 methods of expression 49 definiteness and animacy 213–15 and case 324 –5 concept of 1–15 configurationality and 305–10 defining 253–81 of demonstratives 17–21 and determiners. 309 –10 and identifiability 13–15. 253. Lyons) 160–1 defined 18 temporal and spatial 161 uses of term 20n. 108 Delorme.

113. 63 see also definite determiners. quantificational determiners DG languages 24 –5. 285 reanalysis of as trigger for category of definiteness 277–81. 321 in non-configurational languages 155n in specifier position 287–8. 275 – 6. 324 oblique direct objects 205 –7 discourse. 107– 8 temporal and emotional as well as spat 113 distribution of definites and indefinites 51 Dixon. 171 double definition see double determination double determination cardinal articles and 100 and the DP analysis 78. indefinite demonstratives. Kamp’s 2 271 discourse salience 94 –5. indefinite determiners. M. 252. 128. 232 direct object agreement. 233 discourse semantics. 326 Donaldson. S. C. 112. 165. 325 relationship to definite article 118–21 require identifiability 7. 92 possessives and definiteness in 130–4. 334 –5 and the modified DP analysis 299. hierarchy of properties 318 discourse participant features 313–20. 298 and definiteness 153. Heim’s 249. 253– 4 Det (Determiner) 44 Det position 44 –5. contrast demonstratives. distal demonstratives. 332 positions of 116–21. 85. 327 NP analysis 78 Dougherty. W. and double determination 78 definiteness markers as 322–3 degenerate 310n deictically neutral 160 as possessives 22–5 pronominal or pre-nominal use in English 37–8 proposed functional categories corresponding to 294 –6 quantificational–cardinal distinction 266 strong and weak (Barwise and Cooper) 266–7 and universal quantification 32 use of term 15. 199–207 definite object marking 202–5 dicerential case marking 200–1. 306. 215. 289 diachronic aspects of definiteness 165. 284. 270–1 Dik. 118 “determiner-genitive” see “DG languages” determiner-pronoun pairs 26. verb agreement 207–10 direct object marking 49. T. 340 discourse referent Karttunen’s 268–9 within scope of a quantifier 269 discourse representation theory. 246. proximal demonstratives. 268–71 discourse structure 190. Russell’s theory of 254 –5 descriptive grammar xvi. 280. 322– 40 diagnostics for definiteness 16–17. 320 the development of 323–6 as person phrase 310–20 DP analysis and article inflection 217–19 of demonstratives 279 . C. 135. 252 demonstratives in 116–17. 302. S.non-deictic distinctions 113–16 as origin of definite articles 331– 4 ostension in 160–1 overlap in function with definite articles 164. personal determiners. emphatic demonstratives. 330. 304n determiners the content of 289. 116 see also adjectival demonstratives. 297 364 and the modified DP analysis 299. 320 326. 228. 251–2 Diesing. 64. 323–4 DG possessives and AG possessives 293 – 4 and definite constraint on full Poss Movement 24 –5. 320–1 serving as third-person personal pronouns 145–8 stress and 29–30. R. K. 225. R. 113 –14 distance/proximity distinctions 55–7. 87 Donnellan. 301–3 definiteness. 28n DP as definiteness phrase 298–301. topic demonstratives description can it involve reference? 165–6 use of term in semantics 7n see also definite description descriptions. 87. 296 explained 78 and mixed systems 77– 82. 227–36 information structure and definiteness 232– 6 organization of information 227–32 distal demonstratives 18 distal forms 56 –7. M.

270 existential quantifiers 169 –70. 327–9 DP hypothesis 42– 4. J. 111n. G. 290 – 8. 95. 266 sense-reference distinction 254 –5. 50 “ensembles” 181. 135 – 6 pre-nominal use 27–8. 68 elative. 254 –5. 251. 245 non-central 240–1. 236 fronting. 340 Frege. 254 Jespersen’s theory of stages of 254 and location 260 traces and 221 see also identifiability Feoktistov. 64. 219–25. 95. S. 325 – 6 DP structure definiteness as 279–80. 328 “E-position” 245 E-type pronouns 31 Eachus. 161. 226. 123 Ebbing. 233 hypothesis 3–7. and singling out the topic 230 Fukui. A. J. 340 in personal pronouns 311 Dunn. 108. 70 Feydit. 70. 308 exclusive and inclusive 110n. 258 colloquial use of article in 122 defined 254 and definiteness 160. doubly-filled Comp 302 Finer. 194. 239. L. 277. P. T. H. N. 315 in the animacy hierarchy 213–14. defined 18 first-person plural indefinite 313–14 first-person pronouns 18.and configurationality 305–10 and person 310–20 of preposition combined with definite article 66. 282. 29 focus 228–9 markers of 48n see also presupposition-focus oppositio Fodor. 258 –9. 326– post-nominal 106 pre-nominal 106 as specifiers 301. 340 Enç. 127n exclusiveness 12n. 264 first person. N. 214 –15 dual meaning of inclusive 136. 321. and definiteness 255 existential closure (Heim) 270 existential entailment 169–70. 244 “inner” and “outer” verbal structures 240–1 the phenomenon 237– 41 “Predication Principle” (Safir) in 243 quantification restriction (Milsark) 241 245. G. C. 322–3 movement processes in DP 291– 4 other functional categories in the noun phrase 294 –9 weaknesses in the 296–8 DP languages 319. 109. 108. 243 – 4. D. 238. 232 Durrell. definiteness and 199. 260–1 existential clause. F. 108 “file-change” semantics (Heim) 268–71 filters. 256 Frei. 241 expletive occurrences of the definite artic 164. 301 extension of a noun 183 extensional readings 167–70 familiarity 61. D. M. C. 55. 264 epithets 31n ergative case 200–1. 171–2 existential implication 255 – 8. 236 free-form articles 53. K. 195 entailment defined 256 or presupposition 256 –7. L. 253. 27–8. 112 Fretheim. 164 –5 Einarsson. 93. 222 first mentions 262. 283. F. 165 Foulet. 172 Engdahl. 217 “fourth person” pronoun 135 Fraurud/Franrud ?. and partitive 200 emphatic demonstratives 116 empty categories. J. 268 –9 or givenness 228n. 200 Evans. D. 241 ergative languages 87. 265 – 6 use of “have” 234. 240. K. 242. E. L. 265 existential sentences 236 – 46 “central” 236 copula “be” 237– 41. E. 243 defined 236 definiteness ecect in 236 – 46 explanatory accounts 241– 6 list reading 239. 323 structural position 63. 291 function words 65 . 31 Everett. 264 existential quantification 265. M. H. 109 Ebert. 321. 227 England. 325–6. 266–7. 63–7 characteristics 63–7 morphological behaviour 215–19.

322. M. singular generics genitive case. 20. 153. E. J. H. 336 indefinite as topic 233 non-singular 182n. 332 Hellan. 274 –5. 290–305. 120n. P. W. and pragmatics 273–4 366 and category of meaning 274 –7. T.. 305 – 6. implicature 257–9. 193 – 8 semantic or grammatical definiteness 197–8. T. 294 –6 theory of 322–3 functional heads 282. R. 248 Guillaume. 103. 236 Haegeman. 226 semantic and pragmatic distinctions 179–98 use of the article with 51. and definiteness 228n. 94. M. K. R. H. 277– 8 of the pragmatic concept 226. D. 218. 111. 160n grammatical traditions. 2 264n on inclusiveness and location 260–5 Haywood. Gustave 260 Gundel. 226. 253. 320n and speaker 2–3 task to access context and logical form 271 Hedberg. ? 254n Hale. H. 62. 268–71 Heine. 331– uses of term 275 Greenberg. stages in definite article cycle 310n. 139– 40n. 189–93 and PP-extraposition 248–9 and predication 185. K. 236 Heim. 130 Haislund. 325n Glendening. on definiteness 253 – 60 grammatical words 6n. 251 discourse semantics 260. 220. L. 179 –98 admit exceptions 179 – 81. 25 pronominal reference to 318n. 275 Heinrichs. 41– 4 genericity of mass nouns 180 –1. 142. 18. J. 302. 111n. 304 hearer data new to the 263–4 familiarity to or to the speaker 254. 320n definiteness as a 159–60. 197. theory of 266–8 generative semantics 256 generative syntactic theory xvi. 307 Gildersleeve. 212. 89. 264 –5 Guéron. A. 337 see also class-generics. 65 grammaticality judgements 248–9 grammaticalization 275–7 and definiteness 253. 337–9 Grice. 189–91 and proper nouns 22. 290 Haiman. W. J. L. 210.. 185. 309 Harris. 115 Haider. 297. 228. 299. 73 Haspelmath. 318 functional structure. 20n. 31 Hawkins. N. 98. in chain of hyponymy 319 general knowledge use 4. 191 generics 157. 318 correspond to grammatical or semantic categories 298–9. 91 head 41 relative as 250 “head affixes” 63 Head Movement 291. in negation 201 genitive forms 22–3. 28. I. 108. the emergence of 322–31 future tense 168 Gen (grammatical gender) 295 gender marking in articles 218 and pronouns 137 general definite (identifiable). H. 274 –5. 151 Haugen. J. 324 given-new opposition 227. B. individual generic. J. 68 Goodwin. DP hypothesis 291. J. 290 . A. 217 Hartmann. 28n. 109 Giusti. 97n. 218. 301 in English 181– 4. 143n Hausser. J. 295. 82. 31. 242n government-binding theory 269 grammar. 284. the noun phrase. 150. 277– 8 which interacts with definiteness 199 grammatical contexts. 303. 291 Hagman. P. 301 and non-generic use 179. 229 givenness. S. 126. L. 232–3 Givón. 132 “governing category” 30. M. 11. 154. 184 and aspectual distinctions in verbs 180 bare plural 189 –73. 154 –5. R. creating ambiguity 166 –7 grammatical feature 159–60. 331 genitives. B. G. 2 308. 3. 158 “generalized binding theory” (Aoun) 221 generalized quantifiers. 9. “inner” and “outer” 23n Gil. 310 and semantic bleaching 275. J. 178.

107. 274. 226. 319 in demonstratives 18. 111n Hintikka. K. G. 334 identifiability 5– 6. 198 chain of 319. R. 212 and salience 214 –15 hyponymy 159. in verb-object agreement 210 “incorporation” structure 59. 233 grammaticalization of 277–8. 332. 89–90. 111 Huddleston. 158. 261. 253. 274. 78. 277n Hukari. 258n Hockett. 226. 109. 71. 278 generics and 192–3. 339 problems for 6–7 universality of 279 see also familiarity immediate situation use 4. 148 interrogative element in 150 pronouns 150–1 indefinite noun phrases and definite constraint 289–90 specificity of 165 indefinite pronouns 150–1 and interrogatives 150–1 indefiniteness marking. 265 Holmes. 264 –5 see also conventional implicatures. and definiteness marking 47–52 indefinites cardinal or quantificational 33–6. simple indefinites indirect object agreement. 155–6. 253 and/or identifiability 13–15. 275 Householder. 160. 321 generics and 197–8. 319 . 155 neutral with respect to uniqueness 8. 202 absence of 33–4 cardinality of 34 –6. 132. 105 in DP structure 296. 254. E. 78. 238. 233 or givenness 228n. 76 information. and clitics 63. 57. 302. J. verb agreemen 211–12 individual generic 181– 4 inflection category (I) 290 inflectional affixes. 313 inclusiveness and /or 13 –15. 224 –5 implicit noun phrases 219–25 inanimate reference 109. and polysemy 158n Hopper. 151–2 indefinite determiners 37. 32. conversational implicatures implicit arguments 220. C. partitive indefinites. 220n. B. E. 181 Higginbotham. 214 inclusiveness 11–12. 245. W. 243 Hubbard. 205. F. 173. J. 244 –5 Hinchclice. G. J. J. 15 250 –1 interrogatives 219–20 and indefinite pronouns 150–1 Ioup. 260 and definiteness 273– 4. organization of 227–32 information structure. 223n humanness 29. 89–9 271 and CardP 330 and exclusiveness 261 full noun phrases 149–50 identifiability and inclusiveness 12–13 neutral with respect to uniqueness 8. I. 259. P. 12–13 semantic and pragmatic distinctions 17 see also complex indefinites. F. C. 254 Hovdhaugen. 64 Huang. 240. P. 262– 4. 168 internal-head relative construction 62. P. 111n homonymy. 141. 12–13 non-familiarity 3 optional 89 specificity of a certain 37 superlatives incompatible with 10 zero plural in English 34 indefinite demonstratives 41.-T. 262. 260. 174 irrealis marker 225n. 273. 135 Hoc. and definiteness 22 232– 6 intension of a noun 183 intensional readings 167–8 “intensional” verbs 167. 258 can be cancelled 258–9 confusion with uniqueness 15n. 49. 335–6 derivation of 34. H. 278 inclusiveness and indefinites 12–13 pragmatic concept of 300. T. 158. 110 Holmback. 157. 334 implicature Grice’s 257–60. 157.Heyer. 197– 8 identifiability and indefinites 12–13 and location (Hawkins) 260–5 and uniqueness 7–12 incorporated pronoun. J. 300 emergence of 335 – 6 encodes singular feature 95 implies non-inclusiveness or nonidentifiability 13 languages without 94.

229 location. R. 187. 224. S. M. 50. 270 relevance theory 271. 194 kinship terms 214 possession and 129 –30 Kiparsky. 210. 325–6. 264 Laughren. P. 208. 172 Javarek. and phrase structure with VP 154 lexico-semantic features. 95 Levine. C. R. 32. 200 Koopman. 55. H. 181. 231 Lencek. 197. 15n left-dislocation. K. 202. R. 308 Lawler. 49. 295. 139. creating opacity 169 logical representation. 174n. loss of 331 Li. G. 109. 72 Kleiber. 321. 258 Klimas. 301. 98. R. 301 Louw. 279 Lamontagne. 327 Jókay. 197n. 188 –9. 82 Leslau. 94. 160. W. 231 lexical categories 43. C. 237. 94 Lumsden. 53.Jaggar/Jagger ?. 275n. Z. Z. G. 109 Kurdojev.. 193. G. 151–2 typology of definiteness 280. 192 Kripke. 166. 132 Krifka. 271–4. E. 207. 111 Lambrecht. 324 Kraft. H. lack DP 300. 257 on definitization 258–60. J. L. 95. 64. de 268 K (case) 43. E. 332 definite constraint 288–90 DP as person phrase 310–21 on Hawkins 262– 4 . 166. O. 242 243. 208 Jong. 179 lazy pronouns 31–2 Lee. 240. 301. 233 “linker” suffixes 54n list reading. 273–4 Kennedy. A. 52. 172 Lukas. 158n. 88. 284. 154. 19. 271 Karttunen. J. 26. M. G. G. C. 254. 223n Levinson. M. 122–3. 222 Kornfilt. 28. J. 278. 172. É. F. G. 118. T. 182. W. 112 Leech. S. 191. “E-position” in sentences 245 logical traditions. 112 Lee. D. proper nouns in 193–4 logical form 197n. 103 Kirk-Greene. M. B. 63. 111. J. 172. R. H. 111. 59. T. 208 Lyons. 244. 188. 120 Kryk. discourse representation theory 268. H. K. 194 Krupa. 109. 290 lexical structure. 73. 305 –9 without indefinite article 94. K. non-configurational languages Larson. 82. J. 124. 131. 285n. 166. A. and singling out the topic 230. 133 138. V. K. V. N. 52. 336 Jaszczolt.. P. A. 307 with more than one demonstrative system 109–10. O. 109 “kinds” 181. and inclusiveness (Hawkins) 260–5 locative case 201 Lodge. see also KP analysis Kadmon. H. L. 309 Jensen. 94. 164. L. 24. H. 115 Kiss. 241. S. 186. M. J. G. 294 KP analysis 295 – 6. 295. B. 231. 155 see also configurational languages. C. 112. 94 Janson. K. 112 Kuipers. 329–30 vary in functional projections 322–3 368 309 without definite article 236. 32n. 243. 154. 123. L. 279. 231 Klavans. 295 languages A-type and B-type (Gil) 154 –5. 297 explained 269 –70 logical operator. 305–6. 140n. P. 12n Kamp. 108 Jelinek. 183. 279–80 without definiteness. 264 Lewis. existential sentences 239. J. 109 logic. 300. L. 52. H. 83. H. 245 literal meaning 172 Löbner. discourse referent 268–9 Keenan. on definiteness 253–60 Longobardi. 108 Knobloch. 215. 241 Ludlow. 95 Lakoc. J. 268 Kempson. 24 Lunt. 189. A. 95. M. 55 Jespersen. 248. 227. 305 without definiteness marking 240. 18 Lambdin. R. 115 Krámsk!. 230. N. 120n.

294 any in the scope of 248 narrow or wide scope 256–7 syntax of 327 “negative cycle” (Jespersen) 327 NegP 327 neo-Gricean pragmatics 264 –5 Nespor. 19. 199 definite-indefinite distinction 47–51 definiteness 47–51. P. A. 53. 116 Mithun. 160 –1. A. 36. 304 –10 properties 153–4 non-“deictic” definites. 185. W. K. 67 Mbassy Njie. 338 –9 non-human animate nouns 214 non-human reference 28 non-opaque contexts see transparent cont . 118. C. M. D. T. 71. 265 negation 166. 237. M. E. in negation 256–7 narrow-scope ambiguity 169–70. 111 manifest facts 264 Marácz. 154 Nahmad. H. 265. G. articles and 215 nominalization by articles 60–2 and passivization 291 non-anaphoric use 28–9. 19–20. 111 nominal feature marking. 64 Mostowski. 231 meaning. 154 marking adpositional 86. 22. and “deictic” defi 160 –5 non-generic article (Greenberg) 157. W. H. 339 direct object 199–207 emergence and development of definiteness 322–40 indefiniteness 47–51 nominal feature and articles 215–19 number 214 pronominal 88 Marshall. 291–4 Murrell. J. 265 Masica. S. M. 158–9 non-configurational languages 153–6. 195. 203 Maclaran. 238. J. 166. 91 Napoli. 160. A. 254 Malherbe. 56 Newmark. D. representation by grammatical form 275–7 Mediterranean region 48 Meillet. 95 “maximality” 12n Mayers. 200. 233 Matthews. 98. A. 245. E. 60 mixed systems. J. F. K. 111 morphosyntax 322 Mosel. 16. 168 modifiers non-adjectival noun phrase 15 “unexplanatory” (Hawkins) 261 Mohanan. 156. 21. S. M. 304 –10. L. S. 310n pronominal definiteness in 308–9 non-configurationality 282. J. 248 narrow scope. 210. R. 111. C.proper nouns as generics 22. 51. 56 nationality terms 181. R. 86. 65n “new comparative linguistics” xvi new information see given-new oppositio Newman. 204 Morgan. R. 123n mutual cognitive environment 264 Muysken. 55. M. 240. P. 87. 335 modals 166. 215n Nater. 271 and the quantification restriction 265–6 Minassian. 57. and double determination 77–82. 337– 338n. S. 167. 111 Matthews. A. L. J. 196 and definiteness 10–11 genericity of 180–1. 241–2. L. 53–4. 139 – 40n Mace. 55 mental representations. M. 165 McCawley. F. 111 Mchombo. 184 natural language. 160. 88. C. 172. H. 266 mood 294. 318 Moravcsik. 55. 20 –1 Maetzner. P. 249. 301 in singular form 277 used as count 188 without determiner 33 Matthews. 86 –7. U. M. M. 55 minimal clause 30 minimalism xvi Mitchell. 154 monosyllabicity 64 Montague. 265. 266 movement 288. 112. 67 Mayers. G. and syntactic structure Neale. 236 –7. propositions as 271–2 middle construction 187 implicit arguments in 224 Middle East 49 Milsark. 74. 120n. 300 mass nouns 9. K. 193 – 8 Lyons. 32. 266 McCloskey. 172–3.

277n Pan. 46n participant roles 313–18 particle with characteristics of an article 57–8 description of invariable definite article see also “adjectival particles” “partitive article” 100–1 partitive case 101–2 and elative 200 partitive indefinites 100–3 and progressive verb form 102 partitive structure. 210–12. dual and plural 135–6 numeral classifiers 225. 169. generics. count nouns. 181 Partee. 339 Nunberg. 181. H. A. 300. 218n see also common nouns. 186 object clitics. and scope ambiguities 166–70. 131. R. scope of 269–70 optimal relevance 270–4 ostension defined 160 and development of definite articles fro verbs 331 situational 161 see also textual-situational ostension ownership. 317. 173. 290 –8 functional categories in. between a generic and a restric domain 101 passives implicit arguments in 233–4 two types in Romance languages 224 understood agents of 220 . 171 non-topic interpretation of subject 230–1 Noonan. 220–1.non-referential readings 168. C. 295. 308. 253 non-referential use of definites 274 non-specific. proper nouns NP analysis 41–2. of direct objects 205–7 Olsen. 293n opacity. 169. M. 321 degree of stress 95 development to cardinal article 335–6 quantificational-cardinal distinction 265 secondary meanings 97 NumP 294. G. 21 pluralia tantum 277. DP hypothesis 294 –6 have only one article 33–6. 235 noun classifiers 225–6 noun phrases anaphoric possibilities of 30 bridging cross-reference 32n definite-indefinite distinction 1–2 and definiteness 41–5 definiteness beyond 45–6 DP hypothesis 43 – 4. 321 nunation in Arabic 91– 4. 186 Papafragou. 37. 257. 186. 293. 316. implicit noun phrases. B. and possession 128 Palmer. 331 numerals adjectival 330 and cardinality determiners 45. 282–90 as possessives in prepositional structure 130 “simple” definite and indefinite 47 strong types 251 structure and definiteness 41–2. 146. 282–321 in vocative function 152–3 see also definite noun phrases. 306–7. 277n used generically and abstract nouns in French 218. 301 indefinites in full 148–50 modified DP analysis 298–305 as NPs (maximal projection of the head N) 42–3. 38. 282–90 double determination 78 NP trace 219 null anaphor 219 null Det. F. use of term 173 non-specificity 35n. 310 number systems. 2 opaque contexts 166. 326 Num 43 absence of 306 number generics and 181 and grammatical singular 277 370 and person in pronouns 134 –6. agreement 208–9. 117. S. 335 object marking 199 and the animacy hierarchy 213 see also direct object marking oblique function. 139n. universal principle of indefinite interpretation 197n null noun phrases 219–23 null pronouns 138 – 41 licensing of 141 null subject parameter 138–41 null-subject languages 126–7. and definiteness 102 partitives 17. mass nouns. 19n. 215n. 181. null noun phrases nouns categorizations of 194 –6 inherently unique 8. 268 operators. 152. 44 in French 100 –11 partitivity. indefinite noun phrases. 238.

303 postposed articles 49. 279–80. second person. definite and demonstrative 145–8 personal pronouns xv. 124 –34 affixal 125–8. 27. 128–9 and kinship terms 129–30 present or past 128 possessive ’s 22–3. 143n constraints on 141–2 defined 141 and personal pronouns 310–13 personal forms. 326 personal suffixes 126 Peters. 295 Pipa. 16. DG possessiv phrasal clitic possessives “possessum” 24n possible-world theory 194n Postal. 27 291. D. 205 portmanteau morpheme 16. F. D. partial or full 291– 4. J. and homonymy 158n polysynthetic languages 60.nominalizations and 291 past historic tense see preterite past tense definite and indefinite 45 habitual and punctual aspect 189–90 remoteness meaning 277n Pattee. 310n Poppe. 317 Poss Movement. 320 personal determiners 141–5 affixal 143. 26. 304 strong are demonstratives 28–30. 289 –90. 293 – 4n with theme or patient θ-roles 292n see also AG possessives. 28. 321 weak bound with a free-form definite article 304. J. as head of θ-chain 103. 71. third person person phrase. 181 perfect tense. 330–1 overt or null 325. 24. 72–7 possessives 128 second-position 106 use of term 72 “phrasal inflections” 71. 71 pleonastic. 301 polysemy. 282–321 Picallo. F. N. 29 302 possession alienable and inalienable 128–30. C. 34. 329–30 and personal determiners 26–30. 111. 280 –1. grammaticalization of 226 . 307. 315–18 conflation with definiteness 313–18 and definite determiners 313–18 and definiteness 310–20 incompatible with indefiniteness 313–15 and number in personal pronouns 134–6 see also first person. 87. 326. 284 postposed 24n pronoun and full noun phrase 124 –5 pronoun-derived 22–3. 340 hierarchy of properties 318. 163 – 4. 323 – 4 and body parts 26. S. weak and null 137– 41. P. 134. J. 321 development of 325–6. 132–3 compared with demonstratives 133 definiteness of 22–6. 62–3 PP-extraposition 248–9 and any opacity 24 pragmatic concept. M. 242–3 pleonastic function. and preterite tense 45 Perlmutter. 280 phrasal clitic 128 position of 24 –5. 320n as pronominal counterpart of definite articles 134 –48 semantic and grammatical content of 134 –7 stressed or unstressed 29–30 strong. of articles 99. 77 phrasal stress 66 phrase structure 41–2. 106. 35 Perrott. 115 person complementarity with definiteness 313. 320 the nature of 318–20. 251. M. 308. 130–4 free-form 126–7 movement in DP 291– 4 not inherently definite 130 – 4. 304. M. V. J. 260 Payne. 134 –48. DP as 310–20 person-definiteness 313–18. 310–13 polite forms 137. 63 as a definite determiner 130n status of 292–3 possessives xv. 324 phonological constraint 36 phonological reduction 64 phonologically null expressions see empty categories “phrasal affixes” 77 phrasal clitics 63. 303. 219 and binding theory 30 definiteness of 26–32 modified DP analysis of 303 –5. 174n Philippi. 290 Pelletier. 28n.

211 as weak definite pronoun 220 –1. 214 proper nouns 20. and generics 180 prominence 204. and presuppositions 257–8 pragmatic traditions. 210. 242n pronouns in appositive structures 28n bound-variable use of 31. indefi pronouns. 315 E-type use 31. 303 – 4. 251 propositions. 226 pronominal definiteness. 139–41. or comment 228n. resumptive pronouns proper names 21 in the animacy hierarchy 213. profession etc 104 –5 preposed articles 62–3 preposition-article contractions 66. 273 complex 151n correlation between forms of and articl 141 DP analysis of as Dets lacking an NP complement 290 –1. 228–9 preterite. in existential sentences 243 predicative noun phrases. reflexive pronouns. P. 249–50 property-generic 183. 195 opposed to generics 193 recategorized as common nouns 21.distinctions 157–98 pragmatic set (“P-set”) 264 –5 pragmatic theory. omission of definite article after 51 presentational verbs and existential sentences 237 and non-topic subject 229 and subject position 234 presupposition 255–6 and negation 256–7 as a semantic property of sentences 255–7 presupposition-focus opposition 217. personal pronouns. 281 pronominality and object marking 204 –5 and salience 215 pronominals 219. existential quantification of 169 predicates individual-level 183–4. 32 and gender 137 lazy use 31–2 obligatorily encode number 310 Postal’s derivation of 27 semantics of 31–2 strong and definiteness 308 strong and indirect object doubling 211 strong-weak contrast 137–8. verbs of 166 propositional meaning 172. use of term 228n “Predication Principle”. 111 Prince. hierarchy 318–19 property predication 183. E. 219. 304 –5. 44. 177. null pronouns. 12 195 – 6 as “rigid designators” 193 taking the definite article 22. 193–8 have reference but not sense 193. 185 proposition construction. on definiteness 253–60 pragmatics and grammar 273–4 neo-Gricean 264 –5 salience in 215 pre-determiners 36n. F. 121–4. 249–50 stage-level 191. 32. and perfect tense 45–6 Prifti. 223 PROarb (arbitrary PRO) 223. grammar can impose restrictions on 273–4 propositional attitude. 191. 229 principles-and-parameters approach xvi pro in null-subject languages 126. 309 PRO (empty subject of non-finite clause) 220. 249–50 predication. 224 progressive aspect. expressing status. 71. 336 –7. 339 use of special article form with 123–4 properties of discourse. 230 predicate calculus 265–6 logical notation 265 predicate logic. 304 –5 see also demonstrative pronouns. 233 proximal demonstratives 18 proximal feature 19 . 308 in subject agreement 213 in verb-object agreement 208. 244 –5. as mental representations 271–2 prosodic structure. 196 and definite adjectives 85 definite or indefinite 193–8 definiteness of 21–2 distinguished from common nouns 193 as generics 22. 280. to indicate information salience 229. in nonconfigurational languages 307–9 pronominal indefinite singular generic (one) 186 pronominal marking 88 372 definiteness in 278. 284 predicate. 327–9 prepositional expressions 51 prepositions. 121– 4. 139–40n. 160.

J. 108 reduction. 228 relative. 65. L. R. C. 207 Rando. 265 and definiteness 265–71 or reference 166. 171. van der 248 Scatton. 31. H. K. 319 as cardinality determiners 49–50. 95–9. 111 Saltarelli. and topic-final position 230 –1 Ritter. G. 290 Rouchota. 103 –5. W. 194. 299 Roberts. R. 173 referential uses of definites 165–6. 63 Pustejovsky. 339 . E. A. of proper nouns as common nouns 195– 6 reciprocals 30. 243. 158n QP analysis 45n. 286 quasi-referentiality 331 questions 167 Quintilian 48 R (Referent) argument of NPs 310n R-expressions 219. 172 Rowlands. 265 – 6 quantificational determiners 284. 95 –9. universal quantification quantification restriction. 114 Rowlett. I. 172. E. 181 Schuh. 248 –9 Sag. C. what as free 19n relative clauses. 108 Schubert. 67 Schmalstieg. Milsark on 241–2. and case 295 referential readings 168. K. E. levels of 269–71 resumptive pronouns 221–2. 141. 248 random determiner. A. 271 quantifier. 109. J. 237. 231 Reuland. 204 Rothstein. 42. L. 233 in pragmatic analysis 215 Sall. A. 271– 4 Renisio. 158n Russell. 258. 237. 76 Samarin. T. 237. 290. B. H. as a negative polarity form 149–50 Raposo. 245. 97n. S. 243. 69. E. 245–6. tripartite structure of 270n. 137. Heim’s use of term 270n quantifier phrase see QP quantifier raising 270. 219 Redden. 290 referent identification. 140 Renzi. use of definite article to introduce 61–2 relatives anchor referent to speaker 229 see also internal-head relative construc relevance theory 260. 253 referentiality and scope 172–8 and specificity 165–78 reflexive pronouns 30. 88. J. or time of event 46 referential prominence 204 referential properties. V. V. 215. H. 209 Safir. E. E. 242n null 219 Radford. 166. W. 333 representation. J. morphophonological and phonological 275 Reed. P. 242–3.distance/proximity distinctions Pullum. J. P. A. L. generalized quantifiers. 73 Schachter. 257. 220n real-world reference 269 recategorization. 243. 94. 323 quantified sentences. M. A. 290. 55. 327 Ruhl. 292. 99 Reichenbach. B. 165 salience in discourse 94. 295. 140. 250. 265 theory of descriptions 254 –5 S-structure 41. 238. 83. 219 Reichard. 273 Sadler. E. 220 Rosen. 271 rheme see theme-rheme opposition right-dislocation. 268 Reyle. G. G. 225 Ramsden. 19n. 303 Ramsay. 270n quantifiers c-command a variable to have scope over it 273 definite 44 and demonstratives 148 scope ambiguities with 170n as “sieves” 268 in specifier position 284 see also existential quantifiers. R. U. 178 see also existential quantification. C. A. I. G. 256. A. semantic. 238n. 286 quantification 32. 276 Reinhart. 293 – 4n. 251. universal quantifiers quasi-indefinite articles 36. 59 Sandt. 99. 172. 31. 126. K.

277– 8 semantics “file-change” (Heim) 268–71 literature 6n. 141. 179. 256 sets 180–1 predicated by some of the members 182 quantifiers and 267 Shackle. 19. wide scope scope ambiguities 169–70. 47–51. 47– 89. 89 –106 and complex indefinites 33– 41 defined 2 singular feature 34 374 bare nouns 184 definite 186–9 indefinite 185 –7 situational use 4. 172– 4. 273 and referentiality 173–8 see also narrow scope. 152. 29–30 vocative 152 second-position attachment. 322. 112. 331–2 semantic categories. H. 205 shared sets. 228. 333 Selkirk. 158. R. 28. 105 – 6 defined 2 deictic distinctions in 55–7 simple indefinites 1–2. H. 96. 57– 60. 37. 182. P. M. 233–6. N. J. and topic marking 334 second-person pronouns 27. 232 Sommerstein. 178 and opacity 166 –70. 64 –7 and demonstrative pronouns 28–9. R. 337 and non-specific indefinite 98–9 specific readings 168–9. 154 specific in chain of hyponymy 319–20 and non-specific 171 use of term 173 specific indefinite reference 89. G. C. and definite Dets 283–7 Specifier–Head Agreement 284.of operators 269–70. P. definiteness ecects 252 semantic features 16n semantic and pragmatic distinctions 157–98 semantic weakening see semantic bleaching semantic/pragmatic definiteness 226. D. 165 of pronouns 31–2 see also discourse semantics sense-reference distinction (Frege) 254 –5. of definite noun phrases 254 –5 semantic bleaching. 70 Sperber. 264 Selig. V. 74. 181. 287. 202–3 speaker attitude of the 61 in chain of hyponymy 319–20 familiarity to or to the hearer 254 feature 318 and hearer 2–3 see also distance/proximity distinctions Speas. 166. 157. A. 178 cline of 149 and definiteness 57–60. A. 25 wider situation 4. 11 Stockwell. E. 214 familiarity vs respect 137 pre-nominal use 27–8. 240. 264. ? 12n Silverstein. inclusiveness and exclusiveness (Hawkins) 260–2. 215 and indefiniteness 198 as part of person 319 and referentiality 165–78 in transparent contexts 171–2 a unified account? 172– 8 specifier 41 roles of the 283 specifier position. 172. 183. and questions of truth 255–6 states. 159 semantic class. 7. 51. 172 specificity 35n. of definiteness 15. 258 on presupposition 255–7 stress in definite articles 29n. 248 –9 proper nouns and 194 with quantifiers 170n scrambling 154 second-person deixis. 213 simple clitics 63–4 ’s as contraction of has or is 65 simple definiteness 279–81 simple definites 1–2. 116 . J. 271 “Sprachbund” 48 “stages” 191 Stalnaker. 254 see also constant situational basis. and grammaticalization 275. stage-level predicates 249–50 Stavi. and cliticization 72–7 Segal. 274. 302 Spencer. 18 Sohn. 19n Strawson. 268 !tef"nescu-Dr"g"ne#ti. O. 157. 264 Sharvy. M. F. 27 SOV word order 106. V. 39. 257 statements. 182n. 315 in the animacy hierarchy 213. immediate situation use SizeP 296 small clauses 238n as coda of existential sentences 245 Smith. 64 semantic analysis. 198. M.

62. 242– 4. A. M. 290 Taljaard. 297 tense and aspect distinction 45–6 and time distinctions 167. or ambiguity 158n Vallduvi. D. 227 Van Peteghem. 2 233 time. 265 definiteness of 32 Uriagereka. 62 thematic roles (θ-roles) 88. D. and origin of articles 322. or topic 228n. 82. 164. 202. 50. G. D. relative distinctions conveyed by ten 276 –7 time of event. A. with indefinite and defini conjugations 207–8 transparent contexts 170–2. J.and pronouns 138 and pronouns used as determiners 27–8 see also phrasal stress. 268 textual-situational ostension 161–3. 265 uniques. 247 with most 247 non-inclusive uses 247n post-nominal 81–2 surnames. 253 absolute 8 confusion with identifiability 15n. 214 gender in 137 proximate or obviative 135 stressed or unstressed 28–30 zero realization 317 Thompson. R. S. 318 Ultan. G. 111. 260 and inclusiveness 7–12. with the definite article 254 universal quantification 179. T. 290 substantive determiners. morphological markin of 232 topic prominent languages 230. 26 universal quantifiers 11. 285. B. 230 subjunctive mood. 285n. 295 Tritton. 275 Travis. 148. J. 233. 233 – 4 topic continuity 334 topic demonstratives 115 topic noun phrases. M. 221–3 traditional grammar 2. 265. 98. 320. 269 –70 Tura. 150 unaccusative verb hypothesis 89. 304 vagueness. 334. C. C. S. 88. 173 Traugott. 231–2 topicalization 222–3. 30. 91 truth-based theory of sentence meaning 255 – 6. and anaphoric demonstrative referen 114 –15 tone languages 64 “terraced-tone” languages 66n topic 190. 336 Thaeler. L. and definiteness 255 uniqueness implication 258–60. 321 Sudji!. O. E. hypothetical nature 277 substantive categories 43. 237. 180. 335 subject agreement. 334 structural position of 230. 252 incompatible with the indefinite article 9. P. A. 148. 230–1 traces 219. 82 synchronic studies 276 syntax xvi. 268 relative to a context 7–8 uniqueness clause. 108 Suñer. 335. 230. 233 – 4. E. 230. 230 subject-object asymmetries 154 subject-second languages 63. 282–307 synthetic structure. colloquial use of article with 122 Svane. 290 in the animacy hierarchy 213–14. 331–9 substantive occurrences of the definite article 164. shift to analytic structure 327 Szabó. A. 198. 45–6 transitive verbs. preference for a definite in 237 subject-initial languages 106. in verb agreement 212–13 subject position. 60. 332. word stress languages Suárez. 233 topic-introducing frames 230. M. 139 superlatives definiteness ecect 246 – 8. 241 Taraldsen. 276 ter Meulen. 235 thematic structure 227 theme-rheme opposition 227 theoretical survey xv theta-chain 242–3 theta-features (θ-features) 140 third person. K. M. 24 uniqueness 8 –11. 231 topic-comment opposition 227–8. 78. 67 subject. 95 T (Tense) 42. A. or referent identification 46 tone. 210 markers of 48n. 62 Thaeler. 301. 267– 8. 105 . 118. 230 –1. S. 335 or subject 228n. S. A. 103. definition of 315–18. 208 Szabolcsi.

113 wide scope. 201 Wackernagel. 299. 106. 164. 45n. I. 271. J. 163 – 4. D. L. 334. and Det position 64 Westermann. 65n VOS langua