r

THE NEW PSYCHOI10GY OF LANGUAGE

Cognitive and Functional Approaches to Language Structure

Volume 2

Edited by

Michael TOlnasello Max Planck Institute jiJr Evoluti()nm~v AnthlD()l

Leipzig Gennany

1m LAWRENCE ERLBAUM ASSOCIATES PliBLISI IERS 2003 Mahwah New Jersey

Introduction Some Surprises for Psychologists

Michael Tomasello Max Planck Institute

Linguistics can sometimes he a technical discipline with a rcalit and a v()shycabuuyall its OWII For this reason psychologists have oneil ailcd (or linshyguists to tell them what langllage is-that is give them a good according to the latest theory-so tbey can go on to study its sion processing and acquisition But mllch of the lllcolcti(d fralll(ork and vocabulary of modern linguistic theories relies on the calegories and terminology of traditional Wt~steru linguistics Traditional Wcsltrn lillgllisshytics arose historically in the Middle Ages (Iinl11 Greek and ROinan mainly for the teaching of Latin as a language of scholarship OUIlS mel verbs su~jects and objects predicate adjcClins and pr((II([I( lIominaL- arc manifestly not phenomena that were ueated by psychologists or ((11 linshyguists with a psychological bent with the goal of descrihing hm t Illlt peo

of the world speaking more than 5000 dilferenl comprehend and use a naturallangllage Many of thcm arc notR at all to many non-European languages (Croft in press

It may be that some of these categories arc indeed lIscful for the

tory purposes of psycholinguists But some lllay not be it is in each case all

empirical question And that is oIle of the revolutionary aspects of the IlCW

- wave oflinguistic theories that By under the banner of FUllctiona mdor Cognitive Linguistics Although they too IISC technical of it from the traditional vocabulary-in principle each

defined with respect to the function it serves in real proccsses of communication I n addition to this general fUllctional (Hicillal i( 1Il (( )[nishy

2 TOMASELlO

tive-Functional (Usage-Based) linguists also make the cognitive commitshyment to couch their definitions and explanations as much as possible in theon~tical constructs and terminolobY that are compatible with those of the other Cognitive Sciences (Lakoff 1990) This makes the work more acshycessible to psychologists and indeed it is even possible now that gists can share in the discussion and help to identify psychologically real linshyguistic entities involved in processes of linguistic communication

This is the reasoning behind the title The New PI)choloY Language which is descriptive of the chapters both in Tomasello (1998) and in the current volume Structural linguistics adopts many categories of traditional Western linguistics uncritically-indeed positing them as innate asoectB of a supposed universal grammar-and then goes OIl to create new categories based not on their cross-linguistic applicability or on their psyshychological plausibility but rather on their formal adequacy within the framework of a speciflc mathematical theory of language (Thus when a formal advance is made in the theory as in the new minimalism [Chomsky 199~1 it is automatically assumed to be a part of universal grammar with no empirical verification deemed necessary) Cognitive-Functional Lilli gllistics on the ottwr hand adopts the categories of traditional Western linshyguistics only tentatively and provisionally based on their correspondence to the actual patterns of lise of particular people llsing particular languages when it creates new categories itjustilles them on the basis of how people in a particular language or set of languages use them in acts of commnnication

In the introduction to the first volume I attempted to give an overview of CogJlitive-Functional Linguists for psychologists and psycholinguists in the hopes that this might provide them with some new perspectives for

basic processes of linguistic communication (Tomasello In the more modest introduction to this the second volume I simply wish to highlight and to briefly explore some of the discoveries-or in some cases rediscoveries with modern reformulations-of modern CognitiveshyFunctional (Usage-Based) Linguistics with special reference to those that seem to have most direct relevance for psychologists Many of these disshycOeries-or at least the new light in which they are cast in modern Usshyage-Based theories-will be surprising to psychologists and psycholil guists who have not kept up with recent research on such things as grammatical analyses of non-Indo-European languages grammaticali shyzatiol1 in language history the relation between written and spoken lanshyguage and the relation between language and human cognition and soshycial interaction In my opinion a serious consideration of these new facts about language could change fundamentally the way psychologists and psycholinguists go about their business

INTRODUCTION 1

Spoken Language Does Not Work Like Written Language

Everyone agrees that the primary foclis of Linguistics and (l

Psycholinguistics should he spoken language Spoken lallgllage was 1gt1ishymalY by many tens of thousands of years ill h llman history alld indccd llllshytil quite recently the majority of human beings on the a written language at all Today spoken language is still

years in individual ontogeny and the struggles 01 lllallY childl learning to read-as compared with the relative ease with which I(middotalll to speak-attests to the unnaturalness of written language

The problem is that learning to use a wriUell language-Ilot 10 llWlltioll metalinguistic skills for talking about it as ill Western grallllll1

schools-profoundly influences the way e think about language Olsou 1994 pp 258-2(5) argued this poin t forcefully in a series of

some of which are (a) Writing was responsible historically lill pecL~ of spoken language into consciolls awareness that is lor peeLgt of language into object of reflection analysis and design (Iraquo 10

writing system brings all aspects of what is said in spokcll language illlO awareness and those aspect of spoken language thal are nol rcprcsclllnl

written language are extremely difficult to bring into c()n~(i()usllcSS and Those aspects of spoken language represented by writtell language arc

felt by individuals erroneously to be a complete model of languagc llld once this model has been internalized it is extremely dificuh to llllthil1k il and look at spoken language naively

The way to deal with this problem of course is to focus lot Oil

matical sentences found introspectively-as is commou ill milch of Lillshyguistics-but rather to actually observe record and analyc spontallcOllS spoken speech Ford Fox amp Thompson this volullle) This is llot as easy as it sounds and indeed it is only with the inventioll of al]fmlahlc reshycording equipment (and resources for paying transcribers) that it has 1)(_ corne a possibility at all With the invention of cOIllputational lools for tag-

and searching transcripts of spoken language a whole lIew world 01 corpus linguistics is opening up that allows for Ihe analysis of d((ClIl-imiddotd

corpuses that represent what people aetnally do when tiln Iwak laquo-g Biber ct aI 1998 Sinclair 1991) Here is a partial list of SOllH of tillt filldshy

ings that emerge when one looks at spontaneous spoken speech (-ISS) ill comparisons with

bull There is very little in SSS that conesponds to a sentence as many people discovered when they first read transcripts of the inl(Hlllai COllvnsshy

tions of politicians as recorded on the infamous Wateq~atc tapes Pcople speak in intonation units which consist of prosodically and semantically

III

4 5 TOMASELLO

coherent stretches of language typically containing only one new piece of information (DuBois this volume) These in tonalion lin ils arc lypically

units of one sort or another NOlIll Phrases Adpositional Phrases Clauses) but only sometimes are they entile sentences on the model of written language

bull What are often thought of as prototypical utterances in a language acshytually are not For installce uttelances like the English ]ohn bought a moshy

in which there arc full HOUIlS (ie noun phrases) designating both of the main participants are extremely rare in SSS (but reasonably freshyquent in writing) In SSS what people prefer to do mostly is to introduce the main referent in one intonation unit and then predicate something about it in another (often using a pronominal reference to the just introshyduced entity) as in hey ylt1 know that guyJohn down at the poolhall he bought a Harley if you can believe that (Chafe 19941998)

bull What are thought of as the prototypical lISCS of ceIlain linguistic conshystructions often are not For example textbooks tell us that English relative clauses serve to restrict reference as in The motorcycle that he bought uses diesel fuel and they often do do this ill writing But it turilS out in

SSS people very seldom use a relative clause to restrict the refershyence of the primary participant (subject) which as noted previously is most often a pnmoun Also people seldom use the word thaI to introduce a relative clause in SSS This leads once again to more natural utterances like ya know that motorcycle he bought [it uses diesell (Fox amp Thompshyson I

bull Utterances high in transitivity (an agellt does something to cause a change of state in a patient) which are oftcn used as thc prototype of a senshytence in mallY languages arc not so frequent in SSS In one analysis Thompson and Hopper (in pless) fmllid that only about one quarter of the clausal intonation units ill SSS had two palticipan IS and many of these were low in transitivity (primary participant not very agentive or secondary parshyticipan t did not undergo change of state) There were also lIlany QiSF~n~l verbal predicates instead of lexical verbs (ef5 have a hard timp go to allthp trouble 0 V-ing wander around etc)

bull When one systematically compares stich things as noun phrases suborshydinate clauses of all types focus constructions of all types and many others one finds that SSS and written language are very different grammatically (Miller amp Weinert 1998) Many constructions occllr ollly or mainly in speech for example imperatives and interrogatives or only in writing for example some types of complex nominals (eg a rigorous and valid exshyamination of Applied Economics that consists of three papers) hut not in both

INTRODUlt nON

These are enough examples to make the point The reallhillg-spolltashyneow spoken speech-has properties of its own that are dilkrcllt in SOlIW

cases very dillclent from the intuitive model oflanguagc tilt literate Cdlshy

cated people carry around in their heads This internalized 1II0dd llIay of

course be used to generate hypotheses about the structure of SSS blll the bct is that SSS must he studied in its own right by the normal processes 01

scientific observation and experimentation however ditIindt and this may be

Grammar Arises Historically From Language Use

Although it is not well known in the Cognitive Science COl1l 111 II 11 it lite lact is that virtually all linguists who are involved in the detailed analysis of indishyvidual languages cross-linguistically-mostly known as lingnistic

bull now agree that there are vel) few if any specitic grammatical conslluniol1s or markers that are universally present in all lanf5uages There are mallY lallshyguages that simply do not have one or the other of n~lati( clauses sentential complements passive (oBstructions grammatical markers fell tense grammatical markers of evidentiality ditransitives topic markers ( wjJUla (to be) case malking of grammatical roles subjullctive llIood defi shynite and indefinite articles incorporated nouns plural markltrs and Oil

and on Typological research has also established beyond a rcsonahlc doubt that not only are specific grammatical constructions no ullinlsa but hasically none of the so-called minor word classes or English th help to constitute particular (onstructions (eg prepositiolls ltuxiian v(r)s

articles adverhs compiemcntizers LlIld the sal across languages either (Croft in press Dryer I

This does not mean that there are no language universals-there dcshymonstrably are-but only that we must look for those ulliv(Tsals ill besides particular linguistic items and constructions Olle place to look is human cognition and of course that is one of the central tCllets of ( live Linguistics Talmy (this volume) ontlines foUl concept structllrillg ~y_ terns that by hypothesis underlie all languages Thus all IIlllla1l conceptualize the world in terms of certain configurations or space ami time force dynamics and causality perspective and attentiol1al distribushytion and so languages as conventional symbolic systems design cd to C()lll shy

municate about this world obviously reflect these uH1ceptllaliztti()lIs as well Kemmer (this volume) analyzes how many difIcrellt iangmigcs COllshy

strue events and elaborate their participants proposing a llllivcrsal eH~l1t model that then different languages instantiate differently ill their various constructions Haspelmath (this volume) illustral(s graphicallv sornc oj i1lC

6 7 TOMASELLO

interesting and complex ways in which universal forms of conceptualization get symbolized into languages cross-linguistically with both some universal patterns and also a healthy dose of language-specific idiosyncrasies Anshyother place to look for universals is human communication in the sense of the communicative goals and needs of human beings-some of which are universal and some of which are particular to particular speech communishyties Comrie (this volume) outlines some possible linguistic universals due to the kinds of things that humans need to talk about most urgently and the ways they need to talk about them in order to avoid ambiguities and achieve their communicative goals

If grammatical items and constructions are not universally given to hushyman beings then where do they come from Beginning in the last century historical linguists have observed that many grammatical items in a lanshyguage seem to come from more contentfullexical items Some of the bestshyknown European examples are as follows

bull The main future tense marker in English comes from the full lexical verb will as in I will it to hapjJen At some point expressions arose of the form Itll happen (with the volitional component of will bleached out) Similarly the original use of go was for movement (Im going to the SlOTI) and this became Im gonna do it tomorrow (with the movement bleached out)

bull The English past perfective using have is very likely derived from senshytences such as I have afinger broken or I have the prisoners bound (in which have is a verb of possession) This evolved into something like I have broshyken afinger (in which the possession meaning of have is bleached out)

bull English phrases snch as on the top ofand in the side ofevolved into on top ofand inside ofand eventually into atop and inside In some languages relator words such as these spatial prepositions may also become atshytached to nouns as case markers (although not in English)-in this inshystance as possible locative case markers

bull In French the main negative is the expression ne pas as in Je ne sais ~ Currently in spoken French the ne is becoming less often used and jiaS is becoming the main negative marker But the word pas was at one point the word for step with the expression being something like the English not one bit or not one step further

In addition larger constructions themselves are producLs of grammatshyicalization processes albeit these processes may be somewhat different and so they have been called syntactitization (Givan 1979 1995) The basic idea is that instead of sequences of words becoming one word or a word changing

INTRODUCTION

from a more referential to a more grammatical fllnction or a word tllmillg into a grammatical morpheme in this case whole phrases takc ()Il a lICW kind of organization that is loose discourse sequences often acr()ss illtonlti()1l units become tighter syntactic constructions SOllie possible examples

bull Loose discourse sequences such as Hf jmlleri Ihe door find il oJIIIIIllIav become syntacticized into Hf tmlled tlte door oJITI (ltI resultativt COllstrucshytion)

bull Loose discourse sequences such as AI JoJiillld 1( JIXI Jilli) lie jJlays in a band may become My bOYrifllll filays Jim) ill 1 I(IId Or similarly My boyfriend He rids hOTSes HI lifts Oil tlifill Illay bccolllc My boyfrimd who riries hones bets on thllli

bull Similarly if someone expresses the belief that Mary will wcdjollll anshyother person might respond with an assent I bdilve Ihlll (lilowcil iJy a repetition of the expressed belief that A1a) will wed 111111 which iJcshycome syntacticized into the single statement I bdielll Ihlll Mill) lIIil1l1lid John

bull Complex sentences may also derive from discourse sequcllces or inishytially separate utterances as in I wanl it I buy il enllving into wllnl to bny it

Interestingly along with plenty of idiosyncratic gralllllIaticaliratioll paths in individual languages there would seem to be some lllliver~al ()r nearly universal grammaticalization and syntactitizatioll paths as ell Among the most widely attested arc such things as (a) main verb ~ auxilshyiary verb ~ tense-aspect-mood marker (eg a process begun hy Ellglih will [future] and have [perfective]) (b) demonstrative ~ definite article (eg English the from that) (c) the numeral one ~ indefinite article (Spalli~h unoa French un English a) and (el) demonstrative ~ cOllIplemcnti(T (eg in English I know that ~ I know that shl (()lIIing) These hqgtpen scpashyrately in separate languages presumably attesting to commoll processes of change based on universal principles of human cognitioll and lillgllistic communication (Croft 2(00)

Bybee (this volume) proposes some specific explanations for these (Olllshymon grammaticalization paths in terms of cognitive and connnllnicatie processes well known to psychologists such as automatization habituation decontextualization (emancipation) categorization pragmatic inflTtlJ(shying and others These processes occur as individuals use pieces of language in communication over time with speakers cOllstantly trying to say no lllorc than is necessary and listeners trying to make sure that speakers say enough that they can understand adequately the intended message Van Hoek (this

8 TOMASELLO

certain processes of reference and anaphora across clauses and intonation units operate the way they do in lan~llage Her exshy

focuses on the way people package their conceptualizations for purposes of interpersonal communication

The Units of Language Are Many and Various and Do Not Constitute A Grammar

In traditional Western linguistics we speak of The Grammar of a lanshyguage and Chomsky has followed in this tradition speaking of children

with A Grammar But languages as they are really spoken and of The Grammar of a lanshy

guage as a coherent entity many interesting strnctures must simply be igshynored For example it is well known that in mlditionalterms English is an SVO (Subject-Verb-O~ject) language su~jects typically precede the verb and agree with it in number Thus we say

She play~ the piano They piaL the piano

this way we say

There ~ my shoe Here is my shoe There are mv shoes Here are my shoes

In this case it is the element following the verb that agrees with it in numshyber and so is by that criterion its subject (Making matters even more comshyplicated the cry similar looking utterance It is m~ shoe does not also have the f(mn It art m1 Ihoes) It is also well known that many so-called ergative languages have ergative organization in for example first and second pershyson utterances but acclisative organization in third person ullerances

can also be split based on tense DeLancey 1981) is that different constructions in a language often have their

own (1IOsyncratic properties that do not lit neatly into the rules of The Grammar Fillmore Kay and OConner in their famous 1988 paper in Language (reprinted in abridged form in this volume) explore some of the many and various idiosyncratic constructions of English focusing especially on the construction exemplified in utterances such as She wouldn Zivpound in

New YOTh much less Boston vVhereas it was always known that all languages have some idioms metaphors proverbs and quirky constructions what this paper underlines is the fact that many constnlctions in a language are in fact mixtures of more rellular and more idiomatic subconstructions

INTRODliCTION 9

Subsequent studies on various other odd cOllstruction haq tllrlwd up many other similar examples most famously

the nominal extraposltlOn construction (Michaeli~ II 1ulIlmmiddotcht 19~j6) as in Its amazinf the JeotJuJ wmltIIl lint

the WXDY construction as in 11111 1111111 dlJshy

as in fie smikd ii1 Wll) illo I(

constructioll (JackendofL 19~JI) as ill f II caref awa_1

bull the -el construction as in Thpound Tilher lhl) IIrl thl nirll IIII rill

the incredulity construction as in Him bl a doror

These constrtlctions are notjnst totally weird idioms but rarhn tlin represhysent complex mixtures of regular and idiomatic componen Is and so in t 11shy

ditional Linguistics it is difficult to know what to do with th(1I1 The theoretical move in traditional as well as CllOl1lskian lillguistic has

always been to simply designate some items and constrlluiollS or a bll shyare then to tlie lexicoll

has been most clearly instantiated in CholllskYs ( 19HO) disshytinction between the Core and the Periphery in The Grall1l1lal or a lallshyguage More recently it is also evident in the vVords and Rliks approaclI of Pinker (1999) and Clahsen (999) in which all irregular asp(ch or a language are in the lexicon-and so must be learned by rote-wlicr(ls all the regular aspects of a language are a part of its grammaJ lIId ~() Edl ulIshy

der a rule that then generates its structural description The plOhlt-lII again is that this tidy distinction is very diflindt to maintain ill tlw bee or mixed constructions such as those listed in which it is al1llost lO segregate the regular and idiomatic aspens To look 1l1OIC

one example the incredulity construction (Alv rnalhn rid II

can generale lew ex-In some ways it is like other English (OIlitlllctioIlS

(eg it has SVO ordering the NPs are regular) but of cour( til( S is marked as an o~ject pronoun (accusative case) and the verb is Ilollfillite (not marked for agreemell t) And so the question is Is tli is a IIIc-b~lscd construction or an idiom If it is an idiom it mllst be called I prodllcti( idiom The problem is that there are thousands and thousands ofproltillcshytive idioms in a language that are regular and idiomatic in myriad fliifnshyent ways-so that they merge into more ngular constructiOlS with 110

clear break (Nunberg amp Wasow The discovery-perhaps best credited to Bolinger (I but dlle

to the work of Fillmore Kay and colleagues-is that there is no deal disshytinction between the core and the DeIiDhclv of a bmrIlHT and this llllshy

Introduction Some Surprises for Psychologists

Michael Tomasello Max Planck Institute

Linguistics can sometimes he a technical discipline with a rcalit and a v()shycabuuyall its OWII For this reason psychologists have oneil ailcd (or linshyguists to tell them what langllage is-that is give them a good according to the latest theory-so tbey can go on to study its sion processing and acquisition But mllch of the lllcolcti(d fralll(ork and vocabulary of modern linguistic theories relies on the calegories and terminology of traditional Wt~steru linguistics Traditional Wcsltrn lillgllisshytics arose historically in the Middle Ages (Iinl11 Greek and ROinan mainly for the teaching of Latin as a language of scholarship OUIlS mel verbs su~jects and objects predicate adjcClins and pr((II([I( lIominaL- arc manifestly not phenomena that were ueated by psychologists or ((11 linshyguists with a psychological bent with the goal of descrihing hm t Illlt peo

of the world speaking more than 5000 dilferenl comprehend and use a naturallangllage Many of thcm arc notR at all to many non-European languages (Croft in press

It may be that some of these categories arc indeed lIscful for the

tory purposes of psycholinguists But some lllay not be it is in each case all

empirical question And that is oIle of the revolutionary aspects of the IlCW

- wave oflinguistic theories that By under the banner of FUllctiona mdor Cognitive Linguistics Although they too IISC technical of it from the traditional vocabulary-in principle each

defined with respect to the function it serves in real proccsses of communication I n addition to this general fUllctional (Hicillal i( 1Il (( )[nishy

2 TOMASELlO

tive-Functional (Usage-Based) linguists also make the cognitive commitshyment to couch their definitions and explanations as much as possible in theon~tical constructs and terminolobY that are compatible with those of the other Cognitive Sciences (Lakoff 1990) This makes the work more acshycessible to psychologists and indeed it is even possible now that gists can share in the discussion and help to identify psychologically real linshyguistic entities involved in processes of linguistic communication

This is the reasoning behind the title The New PI)choloY Language which is descriptive of the chapters both in Tomasello (1998) and in the current volume Structural linguistics adopts many categories of traditional Western linguistics uncritically-indeed positing them as innate asoectB of a supposed universal grammar-and then goes OIl to create new categories based not on their cross-linguistic applicability or on their psyshychological plausibility but rather on their formal adequacy within the framework of a speciflc mathematical theory of language (Thus when a formal advance is made in the theory as in the new minimalism [Chomsky 199~1 it is automatically assumed to be a part of universal grammar with no empirical verification deemed necessary) Cognitive-Functional Lilli gllistics on the ottwr hand adopts the categories of traditional Western linshyguistics only tentatively and provisionally based on their correspondence to the actual patterns of lise of particular people llsing particular languages when it creates new categories itjustilles them on the basis of how people in a particular language or set of languages use them in acts of commnnication

In the introduction to the first volume I attempted to give an overview of CogJlitive-Functional Linguists for psychologists and psycholinguists in the hopes that this might provide them with some new perspectives for

basic processes of linguistic communication (Tomasello In the more modest introduction to this the second volume I simply wish to highlight and to briefly explore some of the discoveries-or in some cases rediscoveries with modern reformulations-of modern CognitiveshyFunctional (Usage-Based) Linguistics with special reference to those that seem to have most direct relevance for psychologists Many of these disshycOeries-or at least the new light in which they are cast in modern Usshyage-Based theories-will be surprising to psychologists and psycholil guists who have not kept up with recent research on such things as grammatical analyses of non-Indo-European languages grammaticali shyzatiol1 in language history the relation between written and spoken lanshyguage and the relation between language and human cognition and soshycial interaction In my opinion a serious consideration of these new facts about language could change fundamentally the way psychologists and psycholinguists go about their business

INTRODUCTION 1

Spoken Language Does Not Work Like Written Language

Everyone agrees that the primary foclis of Linguistics and (l

Psycholinguistics should he spoken language Spoken lallgllage was 1gt1ishymalY by many tens of thousands of years ill h llman history alld indccd llllshytil quite recently the majority of human beings on the a written language at all Today spoken language is still

years in individual ontogeny and the struggles 01 lllallY childl learning to read-as compared with the relative ease with which I(middotalll to speak-attests to the unnaturalness of written language

The problem is that learning to use a wriUell language-Ilot 10 llWlltioll metalinguistic skills for talking about it as ill Western grallllll1

schools-profoundly influences the way e think about language Olsou 1994 pp 258-2(5) argued this poin t forcefully in a series of

some of which are (a) Writing was responsible historically lill pecL~ of spoken language into consciolls awareness that is lor peeLgt of language into object of reflection analysis and design (Iraquo 10

writing system brings all aspects of what is said in spokcll language illlO awareness and those aspect of spoken language thal are nol rcprcsclllnl

written language are extremely difficult to bring into c()n~(i()usllcSS and Those aspects of spoken language represented by writtell language arc

felt by individuals erroneously to be a complete model of languagc llld once this model has been internalized it is extremely dificuh to llllthil1k il and look at spoken language naively

The way to deal with this problem of course is to focus lot Oil

matical sentences found introspectively-as is commou ill milch of Lillshyguistics-but rather to actually observe record and analyc spontallcOllS spoken speech Ford Fox amp Thompson this volullle) This is llot as easy as it sounds and indeed it is only with the inventioll of al]fmlahlc reshycording equipment (and resources for paying transcribers) that it has 1)(_ corne a possibility at all With the invention of cOIllputational lools for tag-

and searching transcripts of spoken language a whole lIew world 01 corpus linguistics is opening up that allows for Ihe analysis of d((ClIl-imiddotd

corpuses that represent what people aetnally do when tiln Iwak laquo-g Biber ct aI 1998 Sinclair 1991) Here is a partial list of SOllH of tillt filldshy

ings that emerge when one looks at spontaneous spoken speech (-ISS) ill comparisons with

bull There is very little in SSS that conesponds to a sentence as many people discovered when they first read transcripts of the inl(Hlllai COllvnsshy

tions of politicians as recorded on the infamous Wateq~atc tapes Pcople speak in intonation units which consist of prosodically and semantically

III

4 5 TOMASELLO

coherent stretches of language typically containing only one new piece of information (DuBois this volume) These in tonalion lin ils arc lypically

units of one sort or another NOlIll Phrases Adpositional Phrases Clauses) but only sometimes are they entile sentences on the model of written language

bull What are often thought of as prototypical utterances in a language acshytually are not For installce uttelances like the English ]ohn bought a moshy

in which there arc full HOUIlS (ie noun phrases) designating both of the main participants are extremely rare in SSS (but reasonably freshyquent in writing) In SSS what people prefer to do mostly is to introduce the main referent in one intonation unit and then predicate something about it in another (often using a pronominal reference to the just introshyduced entity) as in hey ylt1 know that guyJohn down at the poolhall he bought a Harley if you can believe that (Chafe 19941998)

bull What are thought of as the prototypical lISCS of ceIlain linguistic conshystructions often are not For example textbooks tell us that English relative clauses serve to restrict reference as in The motorcycle that he bought uses diesel fuel and they often do do this ill writing But it turilS out in

SSS people very seldom use a relative clause to restrict the refershyence of the primary participant (subject) which as noted previously is most often a pnmoun Also people seldom use the word thaI to introduce a relative clause in SSS This leads once again to more natural utterances like ya know that motorcycle he bought [it uses diesell (Fox amp Thompshyson I

bull Utterances high in transitivity (an agellt does something to cause a change of state in a patient) which are oftcn used as thc prototype of a senshytence in mallY languages arc not so frequent in SSS In one analysis Thompson and Hopper (in pless) fmllid that only about one quarter of the clausal intonation units ill SSS had two palticipan IS and many of these were low in transitivity (primary participant not very agentive or secondary parshyticipan t did not undergo change of state) There were also lIlany QiSF~n~l verbal predicates instead of lexical verbs (ef5 have a hard timp go to allthp trouble 0 V-ing wander around etc)

bull When one systematically compares stich things as noun phrases suborshydinate clauses of all types focus constructions of all types and many others one finds that SSS and written language are very different grammatically (Miller amp Weinert 1998) Many constructions occllr ollly or mainly in speech for example imperatives and interrogatives or only in writing for example some types of complex nominals (eg a rigorous and valid exshyamination of Applied Economics that consists of three papers) hut not in both

INTRODUlt nON

These are enough examples to make the point The reallhillg-spolltashyneow spoken speech-has properties of its own that are dilkrcllt in SOlIW

cases very dillclent from the intuitive model oflanguagc tilt literate Cdlshy

cated people carry around in their heads This internalized 1II0dd llIay of

course be used to generate hypotheses about the structure of SSS blll the bct is that SSS must he studied in its own right by the normal processes 01

scientific observation and experimentation however ditIindt and this may be

Grammar Arises Historically From Language Use

Although it is not well known in the Cognitive Science COl1l 111 II 11 it lite lact is that virtually all linguists who are involved in the detailed analysis of indishyvidual languages cross-linguistically-mostly known as lingnistic

bull now agree that there are vel) few if any specitic grammatical conslluniol1s or markers that are universally present in all lanf5uages There are mallY lallshyguages that simply do not have one or the other of n~lati( clauses sentential complements passive (oBstructions grammatical markers fell tense grammatical markers of evidentiality ditransitives topic markers ( wjJUla (to be) case malking of grammatical roles subjullctive llIood defi shynite and indefinite articles incorporated nouns plural markltrs and Oil

and on Typological research has also established beyond a rcsonahlc doubt that not only are specific grammatical constructions no ullinlsa but hasically none of the so-called minor word classes or English th help to constitute particular (onstructions (eg prepositiolls ltuxiian v(r)s

articles adverhs compiemcntizers LlIld the sal across languages either (Croft in press Dryer I

This does not mean that there are no language universals-there dcshymonstrably are-but only that we must look for those ulliv(Tsals ill besides particular linguistic items and constructions Olle place to look is human cognition and of course that is one of the central tCllets of ( live Linguistics Talmy (this volume) ontlines foUl concept structllrillg ~y_ terns that by hypothesis underlie all languages Thus all IIlllla1l conceptualize the world in terms of certain configurations or space ami time force dynamics and causality perspective and attentiol1al distribushytion and so languages as conventional symbolic systems design cd to C()lll shy

municate about this world obviously reflect these uH1ceptllaliztti()lIs as well Kemmer (this volume) analyzes how many difIcrellt iangmigcs COllshy

strue events and elaborate their participants proposing a llllivcrsal eH~l1t model that then different languages instantiate differently ill their various constructions Haspelmath (this volume) illustral(s graphicallv sornc oj i1lC

6 7 TOMASELLO

interesting and complex ways in which universal forms of conceptualization get symbolized into languages cross-linguistically with both some universal patterns and also a healthy dose of language-specific idiosyncrasies Anshyother place to look for universals is human communication in the sense of the communicative goals and needs of human beings-some of which are universal and some of which are particular to particular speech communishyties Comrie (this volume) outlines some possible linguistic universals due to the kinds of things that humans need to talk about most urgently and the ways they need to talk about them in order to avoid ambiguities and achieve their communicative goals

If grammatical items and constructions are not universally given to hushyman beings then where do they come from Beginning in the last century historical linguists have observed that many grammatical items in a lanshyguage seem to come from more contentfullexical items Some of the bestshyknown European examples are as follows

bull The main future tense marker in English comes from the full lexical verb will as in I will it to hapjJen At some point expressions arose of the form Itll happen (with the volitional component of will bleached out) Similarly the original use of go was for movement (Im going to the SlOTI) and this became Im gonna do it tomorrow (with the movement bleached out)

bull The English past perfective using have is very likely derived from senshytences such as I have afinger broken or I have the prisoners bound (in which have is a verb of possession) This evolved into something like I have broshyken afinger (in which the possession meaning of have is bleached out)

bull English phrases snch as on the top ofand in the side ofevolved into on top ofand inside ofand eventually into atop and inside In some languages relator words such as these spatial prepositions may also become atshytached to nouns as case markers (although not in English)-in this inshystance as possible locative case markers

bull In French the main negative is the expression ne pas as in Je ne sais ~ Currently in spoken French the ne is becoming less often used and jiaS is becoming the main negative marker But the word pas was at one point the word for step with the expression being something like the English not one bit or not one step further

In addition larger constructions themselves are producLs of grammatshyicalization processes albeit these processes may be somewhat different and so they have been called syntactitization (Givan 1979 1995) The basic idea is that instead of sequences of words becoming one word or a word changing

INTRODUCTION

from a more referential to a more grammatical fllnction or a word tllmillg into a grammatical morpheme in this case whole phrases takc ()Il a lICW kind of organization that is loose discourse sequences often acr()ss illtonlti()1l units become tighter syntactic constructions SOllie possible examples

bull Loose discourse sequences such as Hf jmlleri Ihe door find il oJIIIIIllIav become syntacticized into Hf tmlled tlte door oJITI (ltI resultativt COllstrucshytion)

bull Loose discourse sequences such as AI JoJiillld 1( JIXI Jilli) lie jJlays in a band may become My bOYrifllll filays Jim) ill 1 I(IId Or similarly My boyfriend He rids hOTSes HI lifts Oil tlifill Illay bccolllc My boyfrimd who riries hones bets on thllli

bull Similarly if someone expresses the belief that Mary will wcdjollll anshyother person might respond with an assent I bdilve Ihlll (lilowcil iJy a repetition of the expressed belief that A1a) will wed 111111 which iJcshycome syntacticized into the single statement I bdielll Ihlll Mill) lIIil1l1lid John

bull Complex sentences may also derive from discourse sequcllces or inishytially separate utterances as in I wanl it I buy il enllving into wllnl to bny it

Interestingly along with plenty of idiosyncratic gralllllIaticaliratioll paths in individual languages there would seem to be some lllliver~al ()r nearly universal grammaticalization and syntactitizatioll paths as ell Among the most widely attested arc such things as (a) main verb ~ auxilshyiary verb ~ tense-aspect-mood marker (eg a process begun hy Ellglih will [future] and have [perfective]) (b) demonstrative ~ definite article (eg English the from that) (c) the numeral one ~ indefinite article (Spalli~h unoa French un English a) and (el) demonstrative ~ cOllIplemcnti(T (eg in English I know that ~ I know that shl (()lIIing) These hqgtpen scpashyrately in separate languages presumably attesting to commoll processes of change based on universal principles of human cognitioll and lillgllistic communication (Croft 2(00)

Bybee (this volume) proposes some specific explanations for these (Olllshymon grammaticalization paths in terms of cognitive and connnllnicatie processes well known to psychologists such as automatization habituation decontextualization (emancipation) categorization pragmatic inflTtlJ(shying and others These processes occur as individuals use pieces of language in communication over time with speakers cOllstantly trying to say no lllorc than is necessary and listeners trying to make sure that speakers say enough that they can understand adequately the intended message Van Hoek (this

8 TOMASELLO

certain processes of reference and anaphora across clauses and intonation units operate the way they do in lan~llage Her exshy

focuses on the way people package their conceptualizations for purposes of interpersonal communication

The Units of Language Are Many and Various and Do Not Constitute A Grammar

In traditional Western linguistics we speak of The Grammar of a lanshyguage and Chomsky has followed in this tradition speaking of children

with A Grammar But languages as they are really spoken and of The Grammar of a lanshy

guage as a coherent entity many interesting strnctures must simply be igshynored For example it is well known that in mlditionalterms English is an SVO (Subject-Verb-O~ject) language su~jects typically precede the verb and agree with it in number Thus we say

She play~ the piano They piaL the piano

this way we say

There ~ my shoe Here is my shoe There are mv shoes Here are my shoes

In this case it is the element following the verb that agrees with it in numshyber and so is by that criterion its subject (Making matters even more comshyplicated the cry similar looking utterance It is m~ shoe does not also have the f(mn It art m1 Ihoes) It is also well known that many so-called ergative languages have ergative organization in for example first and second pershyson utterances but acclisative organization in third person ullerances

can also be split based on tense DeLancey 1981) is that different constructions in a language often have their

own (1IOsyncratic properties that do not lit neatly into the rules of The Grammar Fillmore Kay and OConner in their famous 1988 paper in Language (reprinted in abridged form in this volume) explore some of the many and various idiosyncratic constructions of English focusing especially on the construction exemplified in utterances such as She wouldn Zivpound in

New YOTh much less Boston vVhereas it was always known that all languages have some idioms metaphors proverbs and quirky constructions what this paper underlines is the fact that many constnlctions in a language are in fact mixtures of more rellular and more idiomatic subconstructions

INTRODliCTION 9

Subsequent studies on various other odd cOllstruction haq tllrlwd up many other similar examples most famously

the nominal extraposltlOn construction (Michaeli~ II 1ulIlmmiddotcht 19~j6) as in Its amazinf the JeotJuJ wmltIIl lint

the WXDY construction as in 11111 1111111 dlJshy

as in fie smikd ii1 Wll) illo I(

constructioll (JackendofL 19~JI) as ill f II caref awa_1

bull the -el construction as in Thpound Tilher lhl) IIrl thl nirll IIII rill

the incredulity construction as in Him bl a doror

These constrtlctions are notjnst totally weird idioms but rarhn tlin represhysent complex mixtures of regular and idiomatic componen Is and so in t 11shy

ditional Linguistics it is difficult to know what to do with th(1I1 The theoretical move in traditional as well as CllOl1lskian lillguistic has

always been to simply designate some items and constrlluiollS or a bll shyare then to tlie lexicoll

has been most clearly instantiated in CholllskYs ( 19HO) disshytinction between the Core and the Periphery in The Grall1l1lal or a lallshyguage More recently it is also evident in the vVords and Rliks approaclI of Pinker (1999) and Clahsen (999) in which all irregular asp(ch or a language are in the lexicon-and so must be learned by rote-wlicr(ls all the regular aspects of a language are a part of its grammaJ lIId ~() Edl ulIshy

der a rule that then generates its structural description The plOhlt-lII again is that this tidy distinction is very diflindt to maintain ill tlw bee or mixed constructions such as those listed in which it is al1llost lO segregate the regular and idiomatic aspens To look 1l1OIC

one example the incredulity construction (Alv rnalhn rid II

can generale lew ex-In some ways it is like other English (OIlitlllctioIlS

(eg it has SVO ordering the NPs are regular) but of cour( til( S is marked as an o~ject pronoun (accusative case) and the verb is Ilollfillite (not marked for agreemell t) And so the question is Is tli is a IIIc-b~lscd construction or an idiom If it is an idiom it mllst be called I prodllcti( idiom The problem is that there are thousands and thousands ofproltillcshytive idioms in a language that are regular and idiomatic in myriad fliifnshyent ways-so that they merge into more ngular constructiOlS with 110

clear break (Nunberg amp Wasow The discovery-perhaps best credited to Bolinger (I but dlle

to the work of Fillmore Kay and colleagues-is that there is no deal disshytinction between the core and the DeIiDhclv of a bmrIlHT and this llllshy

2 TOMASELlO

tive-Functional (Usage-Based) linguists also make the cognitive commitshyment to couch their definitions and explanations as much as possible in theon~tical constructs and terminolobY that are compatible with those of the other Cognitive Sciences (Lakoff 1990) This makes the work more acshycessible to psychologists and indeed it is even possible now that gists can share in the discussion and help to identify psychologically real linshyguistic entities involved in processes of linguistic communication

This is the reasoning behind the title The New PI)choloY Language which is descriptive of the chapters both in Tomasello (1998) and in the current volume Structural linguistics adopts many categories of traditional Western linguistics uncritically-indeed positing them as innate asoectB of a supposed universal grammar-and then goes OIl to create new categories based not on their cross-linguistic applicability or on their psyshychological plausibility but rather on their formal adequacy within the framework of a speciflc mathematical theory of language (Thus when a formal advance is made in the theory as in the new minimalism [Chomsky 199~1 it is automatically assumed to be a part of universal grammar with no empirical verification deemed necessary) Cognitive-Functional Lilli gllistics on the ottwr hand adopts the categories of traditional Western linshyguistics only tentatively and provisionally based on their correspondence to the actual patterns of lise of particular people llsing particular languages when it creates new categories itjustilles them on the basis of how people in a particular language or set of languages use them in acts of commnnication

In the introduction to the first volume I attempted to give an overview of CogJlitive-Functional Linguists for psychologists and psycholinguists in the hopes that this might provide them with some new perspectives for

basic processes of linguistic communication (Tomasello In the more modest introduction to this the second volume I simply wish to highlight and to briefly explore some of the discoveries-or in some cases rediscoveries with modern reformulations-of modern CognitiveshyFunctional (Usage-Based) Linguistics with special reference to those that seem to have most direct relevance for psychologists Many of these disshycOeries-or at least the new light in which they are cast in modern Usshyage-Based theories-will be surprising to psychologists and psycholil guists who have not kept up with recent research on such things as grammatical analyses of non-Indo-European languages grammaticali shyzatiol1 in language history the relation between written and spoken lanshyguage and the relation between language and human cognition and soshycial interaction In my opinion a serious consideration of these new facts about language could change fundamentally the way psychologists and psycholinguists go about their business

INTRODUCTION 1

Spoken Language Does Not Work Like Written Language

Everyone agrees that the primary foclis of Linguistics and (l

Psycholinguistics should he spoken language Spoken lallgllage was 1gt1ishymalY by many tens of thousands of years ill h llman history alld indccd llllshytil quite recently the majority of human beings on the a written language at all Today spoken language is still

years in individual ontogeny and the struggles 01 lllallY childl learning to read-as compared with the relative ease with which I(middotalll to speak-attests to the unnaturalness of written language

The problem is that learning to use a wriUell language-Ilot 10 llWlltioll metalinguistic skills for talking about it as ill Western grallllll1

schools-profoundly influences the way e think about language Olsou 1994 pp 258-2(5) argued this poin t forcefully in a series of

some of which are (a) Writing was responsible historically lill pecL~ of spoken language into consciolls awareness that is lor peeLgt of language into object of reflection analysis and design (Iraquo 10

writing system brings all aspects of what is said in spokcll language illlO awareness and those aspect of spoken language thal are nol rcprcsclllnl

written language are extremely difficult to bring into c()n~(i()usllcSS and Those aspects of spoken language represented by writtell language arc

felt by individuals erroneously to be a complete model of languagc llld once this model has been internalized it is extremely dificuh to llllthil1k il and look at spoken language naively

The way to deal with this problem of course is to focus lot Oil

matical sentences found introspectively-as is commou ill milch of Lillshyguistics-but rather to actually observe record and analyc spontallcOllS spoken speech Ford Fox amp Thompson this volullle) This is llot as easy as it sounds and indeed it is only with the inventioll of al]fmlahlc reshycording equipment (and resources for paying transcribers) that it has 1)(_ corne a possibility at all With the invention of cOIllputational lools for tag-

and searching transcripts of spoken language a whole lIew world 01 corpus linguistics is opening up that allows for Ihe analysis of d((ClIl-imiddotd

corpuses that represent what people aetnally do when tiln Iwak laquo-g Biber ct aI 1998 Sinclair 1991) Here is a partial list of SOllH of tillt filldshy

ings that emerge when one looks at spontaneous spoken speech (-ISS) ill comparisons with

bull There is very little in SSS that conesponds to a sentence as many people discovered when they first read transcripts of the inl(Hlllai COllvnsshy

tions of politicians as recorded on the infamous Wateq~atc tapes Pcople speak in intonation units which consist of prosodically and semantically

III

4 5 TOMASELLO

coherent stretches of language typically containing only one new piece of information (DuBois this volume) These in tonalion lin ils arc lypically

units of one sort or another NOlIll Phrases Adpositional Phrases Clauses) but only sometimes are they entile sentences on the model of written language

bull What are often thought of as prototypical utterances in a language acshytually are not For installce uttelances like the English ]ohn bought a moshy

in which there arc full HOUIlS (ie noun phrases) designating both of the main participants are extremely rare in SSS (but reasonably freshyquent in writing) In SSS what people prefer to do mostly is to introduce the main referent in one intonation unit and then predicate something about it in another (often using a pronominal reference to the just introshyduced entity) as in hey ylt1 know that guyJohn down at the poolhall he bought a Harley if you can believe that (Chafe 19941998)

bull What are thought of as the prototypical lISCS of ceIlain linguistic conshystructions often are not For example textbooks tell us that English relative clauses serve to restrict reference as in The motorcycle that he bought uses diesel fuel and they often do do this ill writing But it turilS out in

SSS people very seldom use a relative clause to restrict the refershyence of the primary participant (subject) which as noted previously is most often a pnmoun Also people seldom use the word thaI to introduce a relative clause in SSS This leads once again to more natural utterances like ya know that motorcycle he bought [it uses diesell (Fox amp Thompshyson I

bull Utterances high in transitivity (an agellt does something to cause a change of state in a patient) which are oftcn used as thc prototype of a senshytence in mallY languages arc not so frequent in SSS In one analysis Thompson and Hopper (in pless) fmllid that only about one quarter of the clausal intonation units ill SSS had two palticipan IS and many of these were low in transitivity (primary participant not very agentive or secondary parshyticipan t did not undergo change of state) There were also lIlany QiSF~n~l verbal predicates instead of lexical verbs (ef5 have a hard timp go to allthp trouble 0 V-ing wander around etc)

bull When one systematically compares stich things as noun phrases suborshydinate clauses of all types focus constructions of all types and many others one finds that SSS and written language are very different grammatically (Miller amp Weinert 1998) Many constructions occllr ollly or mainly in speech for example imperatives and interrogatives or only in writing for example some types of complex nominals (eg a rigorous and valid exshyamination of Applied Economics that consists of three papers) hut not in both

INTRODUlt nON

These are enough examples to make the point The reallhillg-spolltashyneow spoken speech-has properties of its own that are dilkrcllt in SOlIW

cases very dillclent from the intuitive model oflanguagc tilt literate Cdlshy

cated people carry around in their heads This internalized 1II0dd llIay of

course be used to generate hypotheses about the structure of SSS blll the bct is that SSS must he studied in its own right by the normal processes 01

scientific observation and experimentation however ditIindt and this may be

Grammar Arises Historically From Language Use

Although it is not well known in the Cognitive Science COl1l 111 II 11 it lite lact is that virtually all linguists who are involved in the detailed analysis of indishyvidual languages cross-linguistically-mostly known as lingnistic

bull now agree that there are vel) few if any specitic grammatical conslluniol1s or markers that are universally present in all lanf5uages There are mallY lallshyguages that simply do not have one or the other of n~lati( clauses sentential complements passive (oBstructions grammatical markers fell tense grammatical markers of evidentiality ditransitives topic markers ( wjJUla (to be) case malking of grammatical roles subjullctive llIood defi shynite and indefinite articles incorporated nouns plural markltrs and Oil

and on Typological research has also established beyond a rcsonahlc doubt that not only are specific grammatical constructions no ullinlsa but hasically none of the so-called minor word classes or English th help to constitute particular (onstructions (eg prepositiolls ltuxiian v(r)s

articles adverhs compiemcntizers LlIld the sal across languages either (Croft in press Dryer I

This does not mean that there are no language universals-there dcshymonstrably are-but only that we must look for those ulliv(Tsals ill besides particular linguistic items and constructions Olle place to look is human cognition and of course that is one of the central tCllets of ( live Linguistics Talmy (this volume) ontlines foUl concept structllrillg ~y_ terns that by hypothesis underlie all languages Thus all IIlllla1l conceptualize the world in terms of certain configurations or space ami time force dynamics and causality perspective and attentiol1al distribushytion and so languages as conventional symbolic systems design cd to C()lll shy

municate about this world obviously reflect these uH1ceptllaliztti()lIs as well Kemmer (this volume) analyzes how many difIcrellt iangmigcs COllshy

strue events and elaborate their participants proposing a llllivcrsal eH~l1t model that then different languages instantiate differently ill their various constructions Haspelmath (this volume) illustral(s graphicallv sornc oj i1lC

6 7 TOMASELLO

interesting and complex ways in which universal forms of conceptualization get symbolized into languages cross-linguistically with both some universal patterns and also a healthy dose of language-specific idiosyncrasies Anshyother place to look for universals is human communication in the sense of the communicative goals and needs of human beings-some of which are universal and some of which are particular to particular speech communishyties Comrie (this volume) outlines some possible linguistic universals due to the kinds of things that humans need to talk about most urgently and the ways they need to talk about them in order to avoid ambiguities and achieve their communicative goals

If grammatical items and constructions are not universally given to hushyman beings then where do they come from Beginning in the last century historical linguists have observed that many grammatical items in a lanshyguage seem to come from more contentfullexical items Some of the bestshyknown European examples are as follows

bull The main future tense marker in English comes from the full lexical verb will as in I will it to hapjJen At some point expressions arose of the form Itll happen (with the volitional component of will bleached out) Similarly the original use of go was for movement (Im going to the SlOTI) and this became Im gonna do it tomorrow (with the movement bleached out)

bull The English past perfective using have is very likely derived from senshytences such as I have afinger broken or I have the prisoners bound (in which have is a verb of possession) This evolved into something like I have broshyken afinger (in which the possession meaning of have is bleached out)

bull English phrases snch as on the top ofand in the side ofevolved into on top ofand inside ofand eventually into atop and inside In some languages relator words such as these spatial prepositions may also become atshytached to nouns as case markers (although not in English)-in this inshystance as possible locative case markers

bull In French the main negative is the expression ne pas as in Je ne sais ~ Currently in spoken French the ne is becoming less often used and jiaS is becoming the main negative marker But the word pas was at one point the word for step with the expression being something like the English not one bit or not one step further

In addition larger constructions themselves are producLs of grammatshyicalization processes albeit these processes may be somewhat different and so they have been called syntactitization (Givan 1979 1995) The basic idea is that instead of sequences of words becoming one word or a word changing

INTRODUCTION

from a more referential to a more grammatical fllnction or a word tllmillg into a grammatical morpheme in this case whole phrases takc ()Il a lICW kind of organization that is loose discourse sequences often acr()ss illtonlti()1l units become tighter syntactic constructions SOllie possible examples

bull Loose discourse sequences such as Hf jmlleri Ihe door find il oJIIIIIllIav become syntacticized into Hf tmlled tlte door oJITI (ltI resultativt COllstrucshytion)

bull Loose discourse sequences such as AI JoJiillld 1( JIXI Jilli) lie jJlays in a band may become My bOYrifllll filays Jim) ill 1 I(IId Or similarly My boyfriend He rids hOTSes HI lifts Oil tlifill Illay bccolllc My boyfrimd who riries hones bets on thllli

bull Similarly if someone expresses the belief that Mary will wcdjollll anshyother person might respond with an assent I bdilve Ihlll (lilowcil iJy a repetition of the expressed belief that A1a) will wed 111111 which iJcshycome syntacticized into the single statement I bdielll Ihlll Mill) lIIil1l1lid John

bull Complex sentences may also derive from discourse sequcllces or inishytially separate utterances as in I wanl it I buy il enllving into wllnl to bny it

Interestingly along with plenty of idiosyncratic gralllllIaticaliratioll paths in individual languages there would seem to be some lllliver~al ()r nearly universal grammaticalization and syntactitizatioll paths as ell Among the most widely attested arc such things as (a) main verb ~ auxilshyiary verb ~ tense-aspect-mood marker (eg a process begun hy Ellglih will [future] and have [perfective]) (b) demonstrative ~ definite article (eg English the from that) (c) the numeral one ~ indefinite article (Spalli~h unoa French un English a) and (el) demonstrative ~ cOllIplemcnti(T (eg in English I know that ~ I know that shl (()lIIing) These hqgtpen scpashyrately in separate languages presumably attesting to commoll processes of change based on universal principles of human cognitioll and lillgllistic communication (Croft 2(00)

Bybee (this volume) proposes some specific explanations for these (Olllshymon grammaticalization paths in terms of cognitive and connnllnicatie processes well known to psychologists such as automatization habituation decontextualization (emancipation) categorization pragmatic inflTtlJ(shying and others These processes occur as individuals use pieces of language in communication over time with speakers cOllstantly trying to say no lllorc than is necessary and listeners trying to make sure that speakers say enough that they can understand adequately the intended message Van Hoek (this

8 TOMASELLO

certain processes of reference and anaphora across clauses and intonation units operate the way they do in lan~llage Her exshy

focuses on the way people package their conceptualizations for purposes of interpersonal communication

The Units of Language Are Many and Various and Do Not Constitute A Grammar

In traditional Western linguistics we speak of The Grammar of a lanshyguage and Chomsky has followed in this tradition speaking of children

with A Grammar But languages as they are really spoken and of The Grammar of a lanshy

guage as a coherent entity many interesting strnctures must simply be igshynored For example it is well known that in mlditionalterms English is an SVO (Subject-Verb-O~ject) language su~jects typically precede the verb and agree with it in number Thus we say

She play~ the piano They piaL the piano

this way we say

There ~ my shoe Here is my shoe There are mv shoes Here are my shoes

In this case it is the element following the verb that agrees with it in numshyber and so is by that criterion its subject (Making matters even more comshyplicated the cry similar looking utterance It is m~ shoe does not also have the f(mn It art m1 Ihoes) It is also well known that many so-called ergative languages have ergative organization in for example first and second pershyson utterances but acclisative organization in third person ullerances

can also be split based on tense DeLancey 1981) is that different constructions in a language often have their

own (1IOsyncratic properties that do not lit neatly into the rules of The Grammar Fillmore Kay and OConner in their famous 1988 paper in Language (reprinted in abridged form in this volume) explore some of the many and various idiosyncratic constructions of English focusing especially on the construction exemplified in utterances such as She wouldn Zivpound in

New YOTh much less Boston vVhereas it was always known that all languages have some idioms metaphors proverbs and quirky constructions what this paper underlines is the fact that many constnlctions in a language are in fact mixtures of more rellular and more idiomatic subconstructions

INTRODliCTION 9

Subsequent studies on various other odd cOllstruction haq tllrlwd up many other similar examples most famously

the nominal extraposltlOn construction (Michaeli~ II 1ulIlmmiddotcht 19~j6) as in Its amazinf the JeotJuJ wmltIIl lint

the WXDY construction as in 11111 1111111 dlJshy

as in fie smikd ii1 Wll) illo I(

constructioll (JackendofL 19~JI) as ill f II caref awa_1

bull the -el construction as in Thpound Tilher lhl) IIrl thl nirll IIII rill

the incredulity construction as in Him bl a doror

These constrtlctions are notjnst totally weird idioms but rarhn tlin represhysent complex mixtures of regular and idiomatic componen Is and so in t 11shy

ditional Linguistics it is difficult to know what to do with th(1I1 The theoretical move in traditional as well as CllOl1lskian lillguistic has

always been to simply designate some items and constrlluiollS or a bll shyare then to tlie lexicoll

has been most clearly instantiated in CholllskYs ( 19HO) disshytinction between the Core and the Periphery in The Grall1l1lal or a lallshyguage More recently it is also evident in the vVords and Rliks approaclI of Pinker (1999) and Clahsen (999) in which all irregular asp(ch or a language are in the lexicon-and so must be learned by rote-wlicr(ls all the regular aspects of a language are a part of its grammaJ lIId ~() Edl ulIshy

der a rule that then generates its structural description The plOhlt-lII again is that this tidy distinction is very diflindt to maintain ill tlw bee or mixed constructions such as those listed in which it is al1llost lO segregate the regular and idiomatic aspens To look 1l1OIC

one example the incredulity construction (Alv rnalhn rid II

can generale lew ex-In some ways it is like other English (OIlitlllctioIlS

(eg it has SVO ordering the NPs are regular) but of cour( til( S is marked as an o~ject pronoun (accusative case) and the verb is Ilollfillite (not marked for agreemell t) And so the question is Is tli is a IIIc-b~lscd construction or an idiom If it is an idiom it mllst be called I prodllcti( idiom The problem is that there are thousands and thousands ofproltillcshytive idioms in a language that are regular and idiomatic in myriad fliifnshyent ways-so that they merge into more ngular constructiOlS with 110

clear break (Nunberg amp Wasow The discovery-perhaps best credited to Bolinger (I but dlle

to the work of Fillmore Kay and colleagues-is that there is no deal disshytinction between the core and the DeIiDhclv of a bmrIlHT and this llllshy

4 5 TOMASELLO

coherent stretches of language typically containing only one new piece of information (DuBois this volume) These in tonalion lin ils arc lypically

units of one sort or another NOlIll Phrases Adpositional Phrases Clauses) but only sometimes are they entile sentences on the model of written language

bull What are often thought of as prototypical utterances in a language acshytually are not For installce uttelances like the English ]ohn bought a moshy

in which there arc full HOUIlS (ie noun phrases) designating both of the main participants are extremely rare in SSS (but reasonably freshyquent in writing) In SSS what people prefer to do mostly is to introduce the main referent in one intonation unit and then predicate something about it in another (often using a pronominal reference to the just introshyduced entity) as in hey ylt1 know that guyJohn down at the poolhall he bought a Harley if you can believe that (Chafe 19941998)

bull What are thought of as the prototypical lISCS of ceIlain linguistic conshystructions often are not For example textbooks tell us that English relative clauses serve to restrict reference as in The motorcycle that he bought uses diesel fuel and they often do do this ill writing But it turilS out in

SSS people very seldom use a relative clause to restrict the refershyence of the primary participant (subject) which as noted previously is most often a pnmoun Also people seldom use the word thaI to introduce a relative clause in SSS This leads once again to more natural utterances like ya know that motorcycle he bought [it uses diesell (Fox amp Thompshyson I

bull Utterances high in transitivity (an agellt does something to cause a change of state in a patient) which are oftcn used as thc prototype of a senshytence in mallY languages arc not so frequent in SSS In one analysis Thompson and Hopper (in pless) fmllid that only about one quarter of the clausal intonation units ill SSS had two palticipan IS and many of these were low in transitivity (primary participant not very agentive or secondary parshyticipan t did not undergo change of state) There were also lIlany QiSF~n~l verbal predicates instead of lexical verbs (ef5 have a hard timp go to allthp trouble 0 V-ing wander around etc)

bull When one systematically compares stich things as noun phrases suborshydinate clauses of all types focus constructions of all types and many others one finds that SSS and written language are very different grammatically (Miller amp Weinert 1998) Many constructions occllr ollly or mainly in speech for example imperatives and interrogatives or only in writing for example some types of complex nominals (eg a rigorous and valid exshyamination of Applied Economics that consists of three papers) hut not in both

INTRODUlt nON

These are enough examples to make the point The reallhillg-spolltashyneow spoken speech-has properties of its own that are dilkrcllt in SOlIW

cases very dillclent from the intuitive model oflanguagc tilt literate Cdlshy

cated people carry around in their heads This internalized 1II0dd llIay of

course be used to generate hypotheses about the structure of SSS blll the bct is that SSS must he studied in its own right by the normal processes 01

scientific observation and experimentation however ditIindt and this may be

Grammar Arises Historically From Language Use

Although it is not well known in the Cognitive Science COl1l 111 II 11 it lite lact is that virtually all linguists who are involved in the detailed analysis of indishyvidual languages cross-linguistically-mostly known as lingnistic

bull now agree that there are vel) few if any specitic grammatical conslluniol1s or markers that are universally present in all lanf5uages There are mallY lallshyguages that simply do not have one or the other of n~lati( clauses sentential complements passive (oBstructions grammatical markers fell tense grammatical markers of evidentiality ditransitives topic markers ( wjJUla (to be) case malking of grammatical roles subjullctive llIood defi shynite and indefinite articles incorporated nouns plural markltrs and Oil

and on Typological research has also established beyond a rcsonahlc doubt that not only are specific grammatical constructions no ullinlsa but hasically none of the so-called minor word classes or English th help to constitute particular (onstructions (eg prepositiolls ltuxiian v(r)s

articles adverhs compiemcntizers LlIld the sal across languages either (Croft in press Dryer I

This does not mean that there are no language universals-there dcshymonstrably are-but only that we must look for those ulliv(Tsals ill besides particular linguistic items and constructions Olle place to look is human cognition and of course that is one of the central tCllets of ( live Linguistics Talmy (this volume) ontlines foUl concept structllrillg ~y_ terns that by hypothesis underlie all languages Thus all IIlllla1l conceptualize the world in terms of certain configurations or space ami time force dynamics and causality perspective and attentiol1al distribushytion and so languages as conventional symbolic systems design cd to C()lll shy

municate about this world obviously reflect these uH1ceptllaliztti()lIs as well Kemmer (this volume) analyzes how many difIcrellt iangmigcs COllshy

strue events and elaborate their participants proposing a llllivcrsal eH~l1t model that then different languages instantiate differently ill their various constructions Haspelmath (this volume) illustral(s graphicallv sornc oj i1lC

6 7 TOMASELLO

interesting and complex ways in which universal forms of conceptualization get symbolized into languages cross-linguistically with both some universal patterns and also a healthy dose of language-specific idiosyncrasies Anshyother place to look for universals is human communication in the sense of the communicative goals and needs of human beings-some of which are universal and some of which are particular to particular speech communishyties Comrie (this volume) outlines some possible linguistic universals due to the kinds of things that humans need to talk about most urgently and the ways they need to talk about them in order to avoid ambiguities and achieve their communicative goals

If grammatical items and constructions are not universally given to hushyman beings then where do they come from Beginning in the last century historical linguists have observed that many grammatical items in a lanshyguage seem to come from more contentfullexical items Some of the bestshyknown European examples are as follows

bull The main future tense marker in English comes from the full lexical verb will as in I will it to hapjJen At some point expressions arose of the form Itll happen (with the volitional component of will bleached out) Similarly the original use of go was for movement (Im going to the SlOTI) and this became Im gonna do it tomorrow (with the movement bleached out)

bull The English past perfective using have is very likely derived from senshytences such as I have afinger broken or I have the prisoners bound (in which have is a verb of possession) This evolved into something like I have broshyken afinger (in which the possession meaning of have is bleached out)

bull English phrases snch as on the top ofand in the side ofevolved into on top ofand inside ofand eventually into atop and inside In some languages relator words such as these spatial prepositions may also become atshytached to nouns as case markers (although not in English)-in this inshystance as possible locative case markers

bull In French the main negative is the expression ne pas as in Je ne sais ~ Currently in spoken French the ne is becoming less often used and jiaS is becoming the main negative marker But the word pas was at one point the word for step with the expression being something like the English not one bit or not one step further

In addition larger constructions themselves are producLs of grammatshyicalization processes albeit these processes may be somewhat different and so they have been called syntactitization (Givan 1979 1995) The basic idea is that instead of sequences of words becoming one word or a word changing

INTRODUCTION

from a more referential to a more grammatical fllnction or a word tllmillg into a grammatical morpheme in this case whole phrases takc ()Il a lICW kind of organization that is loose discourse sequences often acr()ss illtonlti()1l units become tighter syntactic constructions SOllie possible examples

bull Loose discourse sequences such as Hf jmlleri Ihe door find il oJIIIIIllIav become syntacticized into Hf tmlled tlte door oJITI (ltI resultativt COllstrucshytion)

bull Loose discourse sequences such as AI JoJiillld 1( JIXI Jilli) lie jJlays in a band may become My bOYrifllll filays Jim) ill 1 I(IId Or similarly My boyfriend He rids hOTSes HI lifts Oil tlifill Illay bccolllc My boyfrimd who riries hones bets on thllli

bull Similarly if someone expresses the belief that Mary will wcdjollll anshyother person might respond with an assent I bdilve Ihlll (lilowcil iJy a repetition of the expressed belief that A1a) will wed 111111 which iJcshycome syntacticized into the single statement I bdielll Ihlll Mill) lIIil1l1lid John

bull Complex sentences may also derive from discourse sequcllces or inishytially separate utterances as in I wanl it I buy il enllving into wllnl to bny it

Interestingly along with plenty of idiosyncratic gralllllIaticaliratioll paths in individual languages there would seem to be some lllliver~al ()r nearly universal grammaticalization and syntactitizatioll paths as ell Among the most widely attested arc such things as (a) main verb ~ auxilshyiary verb ~ tense-aspect-mood marker (eg a process begun hy Ellglih will [future] and have [perfective]) (b) demonstrative ~ definite article (eg English the from that) (c) the numeral one ~ indefinite article (Spalli~h unoa French un English a) and (el) demonstrative ~ cOllIplemcnti(T (eg in English I know that ~ I know that shl (()lIIing) These hqgtpen scpashyrately in separate languages presumably attesting to commoll processes of change based on universal principles of human cognitioll and lillgllistic communication (Croft 2(00)

Bybee (this volume) proposes some specific explanations for these (Olllshymon grammaticalization paths in terms of cognitive and connnllnicatie processes well known to psychologists such as automatization habituation decontextualization (emancipation) categorization pragmatic inflTtlJ(shying and others These processes occur as individuals use pieces of language in communication over time with speakers cOllstantly trying to say no lllorc than is necessary and listeners trying to make sure that speakers say enough that they can understand adequately the intended message Van Hoek (this

8 TOMASELLO

certain processes of reference and anaphora across clauses and intonation units operate the way they do in lan~llage Her exshy

focuses on the way people package their conceptualizations for purposes of interpersonal communication

The Units of Language Are Many and Various and Do Not Constitute A Grammar

In traditional Western linguistics we speak of The Grammar of a lanshyguage and Chomsky has followed in this tradition speaking of children

with A Grammar But languages as they are really spoken and of The Grammar of a lanshy

guage as a coherent entity many interesting strnctures must simply be igshynored For example it is well known that in mlditionalterms English is an SVO (Subject-Verb-O~ject) language su~jects typically precede the verb and agree with it in number Thus we say

She play~ the piano They piaL the piano

this way we say

There ~ my shoe Here is my shoe There are mv shoes Here are my shoes

In this case it is the element following the verb that agrees with it in numshyber and so is by that criterion its subject (Making matters even more comshyplicated the cry similar looking utterance It is m~ shoe does not also have the f(mn It art m1 Ihoes) It is also well known that many so-called ergative languages have ergative organization in for example first and second pershyson utterances but acclisative organization in third person ullerances

can also be split based on tense DeLancey 1981) is that different constructions in a language often have their

own (1IOsyncratic properties that do not lit neatly into the rules of The Grammar Fillmore Kay and OConner in their famous 1988 paper in Language (reprinted in abridged form in this volume) explore some of the many and various idiosyncratic constructions of English focusing especially on the construction exemplified in utterances such as She wouldn Zivpound in

New YOTh much less Boston vVhereas it was always known that all languages have some idioms metaphors proverbs and quirky constructions what this paper underlines is the fact that many constnlctions in a language are in fact mixtures of more rellular and more idiomatic subconstructions

INTRODliCTION 9

Subsequent studies on various other odd cOllstruction haq tllrlwd up many other similar examples most famously

the nominal extraposltlOn construction (Michaeli~ II 1ulIlmmiddotcht 19~j6) as in Its amazinf the JeotJuJ wmltIIl lint

the WXDY construction as in 11111 1111111 dlJshy

as in fie smikd ii1 Wll) illo I(

constructioll (JackendofL 19~JI) as ill f II caref awa_1

bull the -el construction as in Thpound Tilher lhl) IIrl thl nirll IIII rill

the incredulity construction as in Him bl a doror

These constrtlctions are notjnst totally weird idioms but rarhn tlin represhysent complex mixtures of regular and idiomatic componen Is and so in t 11shy

ditional Linguistics it is difficult to know what to do with th(1I1 The theoretical move in traditional as well as CllOl1lskian lillguistic has

always been to simply designate some items and constrlluiollS or a bll shyare then to tlie lexicoll

has been most clearly instantiated in CholllskYs ( 19HO) disshytinction between the Core and the Periphery in The Grall1l1lal or a lallshyguage More recently it is also evident in the vVords and Rliks approaclI of Pinker (1999) and Clahsen (999) in which all irregular asp(ch or a language are in the lexicon-and so must be learned by rote-wlicr(ls all the regular aspects of a language are a part of its grammaJ lIId ~() Edl ulIshy

der a rule that then generates its structural description The plOhlt-lII again is that this tidy distinction is very diflindt to maintain ill tlw bee or mixed constructions such as those listed in which it is al1llost lO segregate the regular and idiomatic aspens To look 1l1OIC

one example the incredulity construction (Alv rnalhn rid II

can generale lew ex-In some ways it is like other English (OIlitlllctioIlS

(eg it has SVO ordering the NPs are regular) but of cour( til( S is marked as an o~ject pronoun (accusative case) and the verb is Ilollfillite (not marked for agreemell t) And so the question is Is tli is a IIIc-b~lscd construction or an idiom If it is an idiom it mllst be called I prodllcti( idiom The problem is that there are thousands and thousands ofproltillcshytive idioms in a language that are regular and idiomatic in myriad fliifnshyent ways-so that they merge into more ngular constructiOlS with 110

clear break (Nunberg amp Wasow The discovery-perhaps best credited to Bolinger (I but dlle

to the work of Fillmore Kay and colleagues-is that there is no deal disshytinction between the core and the DeIiDhclv of a bmrIlHT and this llllshy

6 7 TOMASELLO

interesting and complex ways in which universal forms of conceptualization get symbolized into languages cross-linguistically with both some universal patterns and also a healthy dose of language-specific idiosyncrasies Anshyother place to look for universals is human communication in the sense of the communicative goals and needs of human beings-some of which are universal and some of which are particular to particular speech communishyties Comrie (this volume) outlines some possible linguistic universals due to the kinds of things that humans need to talk about most urgently and the ways they need to talk about them in order to avoid ambiguities and achieve their communicative goals

If grammatical items and constructions are not universally given to hushyman beings then where do they come from Beginning in the last century historical linguists have observed that many grammatical items in a lanshyguage seem to come from more contentfullexical items Some of the bestshyknown European examples are as follows

bull The main future tense marker in English comes from the full lexical verb will as in I will it to hapjJen At some point expressions arose of the form Itll happen (with the volitional component of will bleached out) Similarly the original use of go was for movement (Im going to the SlOTI) and this became Im gonna do it tomorrow (with the movement bleached out)

bull The English past perfective using have is very likely derived from senshytences such as I have afinger broken or I have the prisoners bound (in which have is a verb of possession) This evolved into something like I have broshyken afinger (in which the possession meaning of have is bleached out)

bull English phrases snch as on the top ofand in the side ofevolved into on top ofand inside ofand eventually into atop and inside In some languages relator words such as these spatial prepositions may also become atshytached to nouns as case markers (although not in English)-in this inshystance as possible locative case markers

bull In French the main negative is the expression ne pas as in Je ne sais ~ Currently in spoken French the ne is becoming less often used and jiaS is becoming the main negative marker But the word pas was at one point the word for step with the expression being something like the English not one bit or not one step further

In addition larger constructions themselves are producLs of grammatshyicalization processes albeit these processes may be somewhat different and so they have been called syntactitization (Givan 1979 1995) The basic idea is that instead of sequences of words becoming one word or a word changing

INTRODUCTION

from a more referential to a more grammatical fllnction or a word tllmillg into a grammatical morpheme in this case whole phrases takc ()Il a lICW kind of organization that is loose discourse sequences often acr()ss illtonlti()1l units become tighter syntactic constructions SOllie possible examples

bull Loose discourse sequences such as Hf jmlleri Ihe door find il oJIIIIIllIav become syntacticized into Hf tmlled tlte door oJITI (ltI resultativt COllstrucshytion)

bull Loose discourse sequences such as AI JoJiillld 1( JIXI Jilli) lie jJlays in a band may become My bOYrifllll filays Jim) ill 1 I(IId Or similarly My boyfriend He rids hOTSes HI lifts Oil tlifill Illay bccolllc My boyfrimd who riries hones bets on thllli

bull Similarly if someone expresses the belief that Mary will wcdjollll anshyother person might respond with an assent I bdilve Ihlll (lilowcil iJy a repetition of the expressed belief that A1a) will wed 111111 which iJcshycome syntacticized into the single statement I bdielll Ihlll Mill) lIIil1l1lid John

bull Complex sentences may also derive from discourse sequcllces or inishytially separate utterances as in I wanl it I buy il enllving into wllnl to bny it

Interestingly along with plenty of idiosyncratic gralllllIaticaliratioll paths in individual languages there would seem to be some lllliver~al ()r nearly universal grammaticalization and syntactitizatioll paths as ell Among the most widely attested arc such things as (a) main verb ~ auxilshyiary verb ~ tense-aspect-mood marker (eg a process begun hy Ellglih will [future] and have [perfective]) (b) demonstrative ~ definite article (eg English the from that) (c) the numeral one ~ indefinite article (Spalli~h unoa French un English a) and (el) demonstrative ~ cOllIplemcnti(T (eg in English I know that ~ I know that shl (()lIIing) These hqgtpen scpashyrately in separate languages presumably attesting to commoll processes of change based on universal principles of human cognitioll and lillgllistic communication (Croft 2(00)

Bybee (this volume) proposes some specific explanations for these (Olllshymon grammaticalization paths in terms of cognitive and connnllnicatie processes well known to psychologists such as automatization habituation decontextualization (emancipation) categorization pragmatic inflTtlJ(shying and others These processes occur as individuals use pieces of language in communication over time with speakers cOllstantly trying to say no lllorc than is necessary and listeners trying to make sure that speakers say enough that they can understand adequately the intended message Van Hoek (this

8 TOMASELLO

certain processes of reference and anaphora across clauses and intonation units operate the way they do in lan~llage Her exshy

focuses on the way people package their conceptualizations for purposes of interpersonal communication

The Units of Language Are Many and Various and Do Not Constitute A Grammar

In traditional Western linguistics we speak of The Grammar of a lanshyguage and Chomsky has followed in this tradition speaking of children

with A Grammar But languages as they are really spoken and of The Grammar of a lanshy

guage as a coherent entity many interesting strnctures must simply be igshynored For example it is well known that in mlditionalterms English is an SVO (Subject-Verb-O~ject) language su~jects typically precede the verb and agree with it in number Thus we say

She play~ the piano They piaL the piano

this way we say

There ~ my shoe Here is my shoe There are mv shoes Here are my shoes

In this case it is the element following the verb that agrees with it in numshyber and so is by that criterion its subject (Making matters even more comshyplicated the cry similar looking utterance It is m~ shoe does not also have the f(mn It art m1 Ihoes) It is also well known that many so-called ergative languages have ergative organization in for example first and second pershyson utterances but acclisative organization in third person ullerances

can also be split based on tense DeLancey 1981) is that different constructions in a language often have their

own (1IOsyncratic properties that do not lit neatly into the rules of The Grammar Fillmore Kay and OConner in their famous 1988 paper in Language (reprinted in abridged form in this volume) explore some of the many and various idiosyncratic constructions of English focusing especially on the construction exemplified in utterances such as She wouldn Zivpound in

New YOTh much less Boston vVhereas it was always known that all languages have some idioms metaphors proverbs and quirky constructions what this paper underlines is the fact that many constnlctions in a language are in fact mixtures of more rellular and more idiomatic subconstructions

INTRODliCTION 9

Subsequent studies on various other odd cOllstruction haq tllrlwd up many other similar examples most famously

the nominal extraposltlOn construction (Michaeli~ II 1ulIlmmiddotcht 19~j6) as in Its amazinf the JeotJuJ wmltIIl lint

the WXDY construction as in 11111 1111111 dlJshy

as in fie smikd ii1 Wll) illo I(

constructioll (JackendofL 19~JI) as ill f II caref awa_1

bull the -el construction as in Thpound Tilher lhl) IIrl thl nirll IIII rill

the incredulity construction as in Him bl a doror

These constrtlctions are notjnst totally weird idioms but rarhn tlin represhysent complex mixtures of regular and idiomatic componen Is and so in t 11shy

ditional Linguistics it is difficult to know what to do with th(1I1 The theoretical move in traditional as well as CllOl1lskian lillguistic has

always been to simply designate some items and constrlluiollS or a bll shyare then to tlie lexicoll

has been most clearly instantiated in CholllskYs ( 19HO) disshytinction between the Core and the Periphery in The Grall1l1lal or a lallshyguage More recently it is also evident in the vVords and Rliks approaclI of Pinker (1999) and Clahsen (999) in which all irregular asp(ch or a language are in the lexicon-and so must be learned by rote-wlicr(ls all the regular aspects of a language are a part of its grammaJ lIId ~() Edl ulIshy

der a rule that then generates its structural description The plOhlt-lII again is that this tidy distinction is very diflindt to maintain ill tlw bee or mixed constructions such as those listed in which it is al1llost lO segregate the regular and idiomatic aspens To look 1l1OIC

one example the incredulity construction (Alv rnalhn rid II

can generale lew ex-In some ways it is like other English (OIlitlllctioIlS

(eg it has SVO ordering the NPs are regular) but of cour( til( S is marked as an o~ject pronoun (accusative case) and the verb is Ilollfillite (not marked for agreemell t) And so the question is Is tli is a IIIc-b~lscd construction or an idiom If it is an idiom it mllst be called I prodllcti( idiom The problem is that there are thousands and thousands ofproltillcshytive idioms in a language that are regular and idiomatic in myriad fliifnshyent ways-so that they merge into more ngular constructiOlS with 110

clear break (Nunberg amp Wasow The discovery-perhaps best credited to Bolinger (I but dlle

to the work of Fillmore Kay and colleagues-is that there is no deal disshytinction between the core and the DeIiDhclv of a bmrIlHT and this llllshy

8 TOMASELLO

certain processes of reference and anaphora across clauses and intonation units operate the way they do in lan~llage Her exshy

focuses on the way people package their conceptualizations for purposes of interpersonal communication

The Units of Language Are Many and Various and Do Not Constitute A Grammar

In traditional Western linguistics we speak of The Grammar of a lanshyguage and Chomsky has followed in this tradition speaking of children

with A Grammar But languages as they are really spoken and of The Grammar of a lanshy

guage as a coherent entity many interesting strnctures must simply be igshynored For example it is well known that in mlditionalterms English is an SVO (Subject-Verb-O~ject) language su~jects typically precede the verb and agree with it in number Thus we say

She play~ the piano They piaL the piano

this way we say

There ~ my shoe Here is my shoe There are mv shoes Here are my shoes

In this case it is the element following the verb that agrees with it in numshyber and so is by that criterion its subject (Making matters even more comshyplicated the cry similar looking utterance It is m~ shoe does not also have the f(mn It art m1 Ihoes) It is also well known that many so-called ergative languages have ergative organization in for example first and second pershyson utterances but acclisative organization in third person ullerances

can also be split based on tense DeLancey 1981) is that different constructions in a language often have their

own (1IOsyncratic properties that do not lit neatly into the rules of The Grammar Fillmore Kay and OConner in their famous 1988 paper in Language (reprinted in abridged form in this volume) explore some of the many and various idiosyncratic constructions of English focusing especially on the construction exemplified in utterances such as She wouldn Zivpound in

New YOTh much less Boston vVhereas it was always known that all languages have some idioms metaphors proverbs and quirky constructions what this paper underlines is the fact that many constnlctions in a language are in fact mixtures of more rellular and more idiomatic subconstructions

INTRODliCTION 9

Subsequent studies on various other odd cOllstruction haq tllrlwd up many other similar examples most famously

the nominal extraposltlOn construction (Michaeli~ II 1ulIlmmiddotcht 19~j6) as in Its amazinf the JeotJuJ wmltIIl lint

the WXDY construction as in 11111 1111111 dlJshy

as in fie smikd ii1 Wll) illo I(

constructioll (JackendofL 19~JI) as ill f II caref awa_1

bull the -el construction as in Thpound Tilher lhl) IIrl thl nirll IIII rill

the incredulity construction as in Him bl a doror

These constrtlctions are notjnst totally weird idioms but rarhn tlin represhysent complex mixtures of regular and idiomatic componen Is and so in t 11shy

ditional Linguistics it is difficult to know what to do with th(1I1 The theoretical move in traditional as well as CllOl1lskian lillguistic has

always been to simply designate some items and constrlluiollS or a bll shyare then to tlie lexicoll

has been most clearly instantiated in CholllskYs ( 19HO) disshytinction between the Core and the Periphery in The Grall1l1lal or a lallshyguage More recently it is also evident in the vVords and Rliks approaclI of Pinker (1999) and Clahsen (999) in which all irregular asp(ch or a language are in the lexicon-and so must be learned by rote-wlicr(ls all the regular aspects of a language are a part of its grammaJ lIId ~() Edl ulIshy

der a rule that then generates its structural description The plOhlt-lII again is that this tidy distinction is very diflindt to maintain ill tlw bee or mixed constructions such as those listed in which it is al1llost lO segregate the regular and idiomatic aspens To look 1l1OIC

one example the incredulity construction (Alv rnalhn rid II

can generale lew ex-In some ways it is like other English (OIlitlllctioIlS

(eg it has SVO ordering the NPs are regular) but of cour( til( S is marked as an o~ject pronoun (accusative case) and the verb is Ilollfillite (not marked for agreemell t) And so the question is Is tli is a IIIc-b~lscd construction or an idiom If it is an idiom it mllst be called I prodllcti( idiom The problem is that there are thousands and thousands ofproltillcshytive idioms in a language that are regular and idiomatic in myriad fliifnshyent ways-so that they merge into more ngular constructiOlS with 110

clear break (Nunberg amp Wasow The discovery-perhaps best credited to Bolinger (I but dlle

to the work of Fillmore Kay and colleagues-is that there is no deal disshytinction between the core and the DeIiDhclv of a bmrIlHT and this llllshy

10 11 TOMASEI J()

dermines the whole idea of The Grammar of a language as a clearly defined set of rules It is interesting and important that when linguists who have worked for years ill the Cholllskiall tradition look Gu-efully at particular grammatical items and constructions they find that many of them that were at one time considered members of the same categOlY (eg compleshymentizer) or construction (eg complement clause) turn Ollt to be very dif fcrent from one another in detail-and so not assimilable to the same rigid rule (Cullicovcr 1999 Jackendoft 1996)

The altcrmltive is to conceive of a language as a structured inventory of symbolic units each with its own structure and fllnction (Langacker 1987) _These units may vary in both their complexity and generality For exshyample the Olle word utterance Forf is a very simple and concrete construcshytion llsed for a specific fllnction in the game of golf Thank you and DOll

mention il are II1ultiword cOflstrunions used for relatively specific social functiolls Some other constructions are composed of specific words along with slots into which whole classes of items llIay fit for example Down with

and Hooray jiJr There arc also constructions that are extremely genshyeral and abstTltlct Thus the ditransitive construction in English protoshytypically indicates transfer of possession and is represented by utterances such as lie gave the doctor money abstractly described as NP+VP+NP+NP AbshystTact linguistic constrnctions such as this have their own meanings ill relashytive independence of the lexical items involved and indeed this is the source of much of the creativity of langlwge (Goldberg 1995) Abstract

~ constructions are thus an important part of the inventory of symbolic reshysources that language users control-and they do much of the work that would be done by core grammar in more traditional accounts-bUl they are best seen as just one form that linguistic constructions may take

In genera the breakdown or the distinction between linguistic core and linguistic periphery is a genuine scientific discovery about the way language works and sorting out its implications will playa key role in creatshying a new psychology oflangllage Vhen we conceive of linguistic construcshytions as cognitive schemas of the same type as we tind in other cognitive

lt skills that is as relatively automatized procedures for gettillg things done (in this case communicatively) it is quite natural that they should not be of only two kinds (regular and idiomatic) hut rather that they should valy from simple to complex and independently from concrete to abstract in many complex ways

Frequency Counts

Individuals do not hear abstract constructions they hear only individual IItshyterances To create abstract constructions they mllst find patterns in the language they bear around them Children begin with constructions based

INTRODU(TIC)N

on concrete items and phrases they then discover 1 variety or rliatlvdv loshycal constructional patterns and only laler do tlley discoHT ilIon gell(ral patterns among these local constructiollal patterns (Tomasello 1992 2000) But as children create mon~ general constructions they do lIoi throwaway their more item-based and local constructions The ide1 tl111 people operate always and only with the most ahstract structlllcs 111lt11 linshyguists can find is what Langacker (1987) called the rulf-lisjlllfllt) It rllkcts a very deep difference in the theoretical goals or hmKtlliltIuists alld IHOIC

psychologically oriented linguists In cognirivey and functionally oriented (usage-hased) approltlclJ(s plOshy

pie can possess abstTact cognitive strnctures that they use ill ctrtlill illshystances but they still operate on some occasions with the 1II0le COIHT(C

structures that instantiate the abstraction Asjust a handftd if llIlllY thoushysands or tens of thousands of relatively cOllcrete alld fixed expressiolls that native speakers of English control (which mayor lllltl) Hot i IIstlIlt IIt(middot

more abstract constructions) Im siml)ly amazed loom rtlPl)whflflOI iI rOil

kefP out of this That was a dose mlf Its a moltn Iipriorilils Jcinlll lillll lolilll(

Id do il all otW again Im surprisld to tnti thai Do wzal)III( ( lold 11

whalyou mean f thought youd Tlevf ask I fav 1(IIlf 1101 You ((11 11w IIII IVhere did you find it He busY right now rou mil t belil1l1 1 wllrd II( Inl a II d on and on (Pawley amp Syder 1983)

Bybee and Scheibman (1999) provided evidence that peopk SOllllllIlHS produce complex utterances-which they know at sume level have illtcnlal structure-as single processing unit~ They analyze in some depth nrio(s Ilses of the English word dont and find thaI in highly frequellt and 1(11shy

tively fixed expression like 1 don f know people lend 10 red lice the ation of don I in some cases so rmlCh thaI it IS harelv recoglliza)le ir Jjgt(II(d

to in isolation Thus the most common pronullciation of dOIl1 1111070 is acshytually something more like ldunno and in some cases the Cxpressioll is barely more than a characteristic intonation contour This salllc )edUClioll

or the word dont does not occur in other less frequent exprCiSiOllS and constructions Although most adults can allalyze this expressioll illto it- components-for ~xample if a questionel- persists they call say each or tht words slowly and emphatically I DONT KNOW-bOlll a proctshying point of view its great frequency has made it a productioll rowil1c Bybee (1995) argued thaI the token frequency of an expressioll MT( to entrench it in a speakers repertoire and make it a processing lllii frequency-repeated instantiations of the same pallern bll with diff(T(1I1

concrete items-entrenches the pattern but also at the same lilllt makes il more generally applicable to more items Thus young childrcll ill form and use only very concrete and local construclional islallds (based Oil

specific lexical items) but with high type freqnency in one or 11101( Ioli for example VVheres (hr X wanna X Mort X It (J X Im X-inJ ii Put X 1111(

12 13 T()MASEILO

Mommy ~~ it [ets X it Throw X X gone I X-ed it Sit on the X here Thenlts a X X broken (Braine 1976 Lieven Pine amp Baldwin see Tomasello 2000 for a review of the evidence)

Frequency also plavs a crucial role in grammaticalization and language Thus it is well known that the linguistic constructions that are most

resistant to change al(~ those that are most fn~qllent That is why most ilTegshyular verbs in a language are typically highly frequent (eg in English the verbs to he and to have) Bybee and Thompson (in press) analyzed the examshyple of the subjunctive mood in Canadian French which has basically been lost However in a few highly frequent fixed expressions it lives on (as it also does in frequent English expressions like If I were you At the same lime highly frequent expressions also in some contexts become grammaticalized and so changt their function sometimes retaining the old hmction in othtr contexts (as in the English main verbs have and goand their more recent instantiations as auxiliary verbs as well) In the context of language acquisition Brooks Tomasello Lewis and Dodson (1999) arshygued and presented evidence thai the entrenchmenl of particular verbs in particular constructions (in both comprehension and prodllction) is a rna-

CIctor preventing children from overgeneralizing their abstract conshystructions 10 inappropriate verbs This finding (in combination with that of Brooks amp Tomasello 1999 who demonstrated the importance of two other usage-based f~lctors) thus solves in large measure the puzzle of why children

grammatical rules indiscriminately with their enshytire be expected to if they possessed the abstract ruics that formal grammar writers often attribute to them Pinker 1984

and entrenchment raises the specter of Behaviorism which as is well known was exorcised from Linguistics once and for all by Chomsky (191)9) Butjust because frequency and entrenchment were imshyportant concepts for behaviorists-who knew little of the structure of lanshyguage-does not mean that they are useless in other more cognitively and

sophisticated approaches It turns out that both the type and token frequency with which particular constructions are used makes an enormous difference both in their historical f~lte and ill the way they are unshyderstood acquired cognitively represented and used by contemporary speakers of a languagt

CONCLUSION

Linguistics as a discipline hovers between the Humanities and the BehavshyioralCognitive Sciences For much of it history Linguistics consisted solely of the analysis of texts and the teaching of rules Many linQuists thus

INTROIllI( nON

did not consider it their concern to worry about psvcilOiogicall(lttlily or to acquire expertise with the kinds of rigorons methods of dat~l statistical analysis that are the fOllndation (lIthe Heitavioral( cllces But with the rise of Cognitive Science as an prise with the rise of new technolog-ies that make possihle the and analysis of real live linguistic communication and wil h 111lt liiC or ( nitive-Functional (Usage-Based) approaches to ling-uistic IlHmy tile balmiddot ance is beginning to tip toward the side ofsciencc In a ut()pi~ll1 llllllit linshyguists and psYtchologists will work togetlrer to investigate the alttu~tl psychological processes by means of which human beings (( produce and acquire a natLIrallangliage The chapters ill Ihis VOlllIlH-ltlS well as those in the first volume-represent theoretical approaches Ihal middotill help us to make progress IowaId thai goal

REFERENCES

Biber D (19HK) Variation ((nJS s]eth rind writing (anlhriftgC (ltllll bridgt 1111 tTiil P ( Biber n Conrad S amp R]lpen R (99S) COIflllgt 11flf(IIisin 1ilorlJlg 1(1IIWt 1m lilli IIIii

US Cambridge Cambridgt Ll n iversity Prcss _ Bolinger D (1977) Mfillling ([IiiIm New York Longman

Braine M (1976) Chidrells lirst word combinalions I(IIf~iIIlf1h oIIi oli(I lI(lIri ill

Child )fiolnnenl 41(No I)

HlOoks P amp Tomasello M (1991) Bow young childrell cOlltrain till a I ((Hlstluctiolls f_(Jngu(lg~ 75 720-7~R

bull Brooks P TomaseHn M Lewis 1 amp Dodsoll K (UJlY) 1 low children oir lt11 C(l lilHlll strtlcltllT errors The entrenchmenl hypothesis Child )nlloIII 11 1-117

Byhet J (IY9fraquo Regular morphoh)Y and the lexicon lfgllllgt (Ild COgIIIIIlI 111111( iii 4~c-4fI5

HybecJ amp ScheiblTlan bull J (llYl) The ltfreCl ollisage Oll ltIlt)res or cOllSlilnIH TIllt 1(lt1( tion of don in English jnguiII(I 37 7i-c)6

Bybee amp Thompson S (in press) Three frellltl(Y tImiddotos ofsnllax lnH(fdill~middot 01111 Jiit

ley IiUgllisit SOIilly

Chafe W (1994) )510(1 muoliIW fllld liml Chicago LuinTsitv of (hidO VIS

Chate t (I 9l)H) Language and lhe flow ofthollght In ~1 Tomasello (Ed) III lli PHI Iud

of lanfU(JW Cognilive and i(n(lional ajljlrJ)adws (PI 111-10) Mlhlt I Lm](il(C

Erihalllll Associates

Chomsky N (IlfI9) A review of B F Skinners VNh1 hehavior 1(IIllII~ Ji t--- Ch()m~ky N (19HO) Rules alld (peselltations iJehmllmJi IIlId Iimlll )rilIn 7 1-11 Chomsky N (1993) minimalist prog~lm () linfllisti tlHfHY I K Ible ( S Kest (lk)

A viewmn Building 20 (pp 1-13) Cambridge M MIT Inmiddotss

Clahs(n I L (I 99H) rxical entris and rules of languag lI1UllitiisciplillIll siudy (II (el Ill inllection B~hl1Pi()ral (lui 11m Srilll((s 22 98()-9~)~l

Croft VI (2()OO) HXjJlaini1lg language rwngl An lImlIliolill) aHIJar1 London Longmall Croft W (~002) Hadiwl mllsrurtilII grammar Ox()[ri Oxlord I ni(I I(ss

bull CnIlicovcr P (1999) SYlllaclir uls Oxford Oxf(ml Univ(rsin Ir(ss DeLancey S (I (lSI) An interpretHiDn of plit ergativity and rdalt1 1gt11 1(IS 1llIglllI~middot

62l67

12 13 T()MASEILO

Mommy ~~ it [ets X it Throw X X gone I X-ed it Sit on the X here Thenlts a X X broken (Braine 1976 Lieven Pine amp Baldwin see Tomasello 2000 for a review of the evidence)

Frequency also plavs a crucial role in grammaticalization and language Thus it is well known that the linguistic constructions that are most

resistant to change al(~ those that are most fn~qllent That is why most ilTegshyular verbs in a language are typically highly frequent (eg in English the verbs to he and to have) Bybee and Thompson (in press) analyzed the examshyple of the subjunctive mood in Canadian French which has basically been lost However in a few highly frequent fixed expressions it lives on (as it also does in frequent English expressions like If I were you At the same lime highly frequent expressions also in some contexts become grammaticalized and so changt their function sometimes retaining the old hmction in othtr contexts (as in the English main verbs have and goand their more recent instantiations as auxiliary verbs as well) In the context of language acquisition Brooks Tomasello Lewis and Dodson (1999) arshygued and presented evidence thai the entrenchmenl of particular verbs in particular constructions (in both comprehension and prodllction) is a rna-

CIctor preventing children from overgeneralizing their abstract conshystructions 10 inappropriate verbs This finding (in combination with that of Brooks amp Tomasello 1999 who demonstrated the importance of two other usage-based f~lctors) thus solves in large measure the puzzle of why children

grammatical rules indiscriminately with their enshytire be expected to if they possessed the abstract ruics that formal grammar writers often attribute to them Pinker 1984

and entrenchment raises the specter of Behaviorism which as is well known was exorcised from Linguistics once and for all by Chomsky (191)9) Butjust because frequency and entrenchment were imshyportant concepts for behaviorists-who knew little of the structure of lanshyguage-does not mean that they are useless in other more cognitively and

sophisticated approaches It turns out that both the type and token frequency with which particular constructions are used makes an enormous difference both in their historical f~lte and ill the way they are unshyderstood acquired cognitively represented and used by contemporary speakers of a languagt

CONCLUSION

Linguistics as a discipline hovers between the Humanities and the BehavshyioralCognitive Sciences For much of it history Linguistics consisted solely of the analysis of texts and the teaching of rules Many linQuists thus

INTROIllI( nON

did not consider it their concern to worry about psvcilOiogicall(lttlily or to acquire expertise with the kinds of rigorons methods of dat~l statistical analysis that are the fOllndation (lIthe Heitavioral( cllces But with the rise of Cognitive Science as an prise with the rise of new technolog-ies that make possihle the and analysis of real live linguistic communication and wil h 111lt liiC or ( nitive-Functional (Usage-Based) approaches to ling-uistic IlHmy tile balmiddot ance is beginning to tip toward the side ofsciencc In a ut()pi~ll1 llllllit linshyguists and psYtchologists will work togetlrer to investigate the alttu~tl psychological processes by means of which human beings (( produce and acquire a natLIrallangliage The chapters ill Ihis VOlllIlH-ltlS well as those in the first volume-represent theoretical approaches Ihal middotill help us to make progress IowaId thai goal

REFERENCES

Biber D (19HK) Variation ((nJS s]eth rind writing (anlhriftgC (ltllll bridgt 1111 tTiil P ( Biber n Conrad S amp R]lpen R (99S) COIflllgt 11flf(IIisin 1ilorlJlg 1(1IIWt 1m lilli IIIii

US Cambridge Cambridgt Ll n iversity Prcss _ Bolinger D (1977) Mfillling ([IiiIm New York Longman

Braine M (1976) Chidrells lirst word combinalions I(IIf~iIIlf1h oIIi oli(I lI(lIri ill

Child )fiolnnenl 41(No I)

HlOoks P amp Tomasello M (1991) Bow young childrell cOlltrain till a I ((Hlstluctiolls f_(Jngu(lg~ 75 720-7~R

bull Brooks P TomaseHn M Lewis 1 amp Dodsoll K (UJlY) 1 low children oir lt11 C(l lilHlll strtlcltllT errors The entrenchmenl hypothesis Child )nlloIII 11 1-117

Byhet J (IY9fraquo Regular morphoh)Y and the lexicon lfgllllgt (Ild COgIIIIIlI 111111( iii 4~c-4fI5

HybecJ amp ScheiblTlan bull J (llYl) The ltfreCl ollisage Oll ltIlt)res or cOllSlilnIH TIllt 1(lt1( tion of don in English jnguiII(I 37 7i-c)6

Bybee amp Thompson S (in press) Three frellltl(Y tImiddotos ofsnllax lnH(fdill~middot 01111 Jiit

ley IiUgllisit SOIilly

Chafe W (1994) )510(1 muoliIW fllld liml Chicago LuinTsitv of (hidO VIS

Chate t (I 9l)H) Language and lhe flow ofthollght In ~1 Tomasello (Ed) III lli PHI Iud

of lanfU(JW Cognilive and i(n(lional ajljlrJ)adws (PI 111-10) Mlhlt I Lm](il(C

Erihalllll Associates

Chomsky N (IlfI9) A review of B F Skinners VNh1 hehavior 1(IIllII~ Ji t--- Ch()m~ky N (19HO) Rules alld (peselltations iJehmllmJi IIlId Iimlll )rilIn 7 1-11 Chomsky N (1993) minimalist prog~lm () linfllisti tlHfHY I K Ible ( S Kest (lk)

A viewmn Building 20 (pp 1-13) Cambridge M MIT Inmiddotss

Clahs(n I L (I 99H) rxical entris and rules of languag lI1UllitiisciplillIll siudy (II (el Ill inllection B~hl1Pi()ral (lui 11m Srilll((s 22 98()-9~)~l

Croft VI (2()OO) HXjJlaini1lg language rwngl An lImlIliolill) aHIJar1 London Longmall Croft W (~002) Hadiwl mllsrurtilII grammar Ox()[ri Oxlord I ni(I I(ss

bull CnIlicovcr P (1999) SYlllaclir uls Oxford Oxf(ml Univ(rsin Ir(ss DeLancey S (I (lSI) An interpretHiDn of plit ergativity and rdalt1 1gt11 1(IS 1llIglllI~middot

62l67

TOMASEIH)14

Dryer M (1997) An grammatical r(latiom llnivcrsaP InJ Bybeej Haiman amp S Thompson (Eds) Essays on Itl1lgllage ldion and langI Iype (pp 115-144) Amsterdam Nethershyhlllds John I~el~ialllim

Fillmore C fuiye P amp OConnor M (1988) Regularity aud idiomaticit) in grammatical constructions The case of let alont rangnagl 64 01-538

Fox B amp Thompson S (1990) A discourse (xplanatioll of Thlt Grammar of relative clawfs in English convfrsaliol1 UmgU(lf1 66 217-31 n

CivI)Il T (1979) Oil undrslanding grammar New York Academic Prltss GiVtlll T (1995) FUllciollolism (lnd gml1l11ilr Amsterdam Netherlands John Be Ilja III ins

Goldherg A (1995) Conslmetions A ul])slnutioll grmmnatmiddot approach to atrwnml structure Chishy

cago University of Chicag Press ~ __ Jr()Clt gt ~ Jackendoff R (199b) TWlstlll the mght away lan~1wgp I 1 5~4-)lJ Kay P amp Fillmore C (1999) Crammalica) constructions and linguistic gtlwralizltltions 1(1Ilshy

fIUlW 75 1- [akoff C (1987) Womm fin (lnli danIlrous Ihings What mtpgones Je(lfai about tltt mind Chishy

cago U liversit of Chicago Press Lakoll G (1990) The Invariancc Hypothesis Is absllltlCt reason based on illlage schemas Cogshy

nitive Linguistits I 19-74 Langacker R (1987) Foundatiolls ofrollililW graltwiCIl (Vol I) Palo Alto CA Stanford Univershy

sity Press Litven E Pine J amp Baldwin G (1997) Lexically-based learning and early grammatical dcshy

veloprnentnurJwl of Child jmguagt 24 117-220 Michatlis L amp Lambncht K (1996) Toward a construction-based thenl of language funcshy

tion The case of nominal extrapositioll Umgllagt 72215-247 MillerJ amp Wcincrl R (ImI8) Spont(lnfOliS sjmken l(lnguage Oxf)rd Oxford University Press -J unberg C Sag I amp Wasow T (1994) Idioms Ianguage 70491-538 Olson D (1994) 71w world on papl Cambridge Cambridge University Press Pawley l amp Snyder F (19R~) Two pmzles for linguistic theory InJ Richards amp R Smith

(Eds) Lllng1wfI and mmmunilfltion (pp 185-2(9) Nltw York Longman Pinker S (1984) la1~laI Rarrability and iangtp dnlfloplllml Cambridge MA Harvard U nishy

versi Ly Press Pinker S (1989) Lmwhilily and (Offlliiotl The (I(q11isilion of 1Jerb-wgIITTlent 1Imiddot1Ire Camshy

bridge IA Hananl University Press Pinker S (1999) Hrd (lnd Iuu New York Morrow Press Sinclair (1991 CmlJ1L fOl(ordallee (wd roilo(tion ()xfold Oxl)rd University Press Thompson S amp Hopp P (in press) Trmsitivity clause strncture and argulllenl structure

Evidenn from conversation In J Bybee amp P Hoppcr (Eds) FrqllPII(Y and the emergenf( or linflsti( liruelure Amsterdam John Benjamins

Tomasello M (1992) First nils A rase study in f(rl~ Ilrammalim dnJelrrjmlrnl New York Camshybridge Universily Press

Tomasello M (1998) Introduction The cognitive-functional perspective Oil language stmcshylure III M Tomastllo (Ed) The 1WW fisycholofY oIlIfIUIgt Cognitive arid I1t111tiorw apshy

proaches tn lallguage Im(11I1 (pp 1-25) Mahwah NJ Lawrence EriballIll Associates Tomasello M (2000) Do young children have adnlt syntactic competence Cogrlilion 74

209-2gt3