encyclopedia of language & linguistics || history of linguistics: discipline of linguistics

Sprachwissenschaft.’ Beitrage zur Geschichte der Sprach-wissenschaft 13, 115–126.

Schmitter P (2003b). Historiographie und Narration: Meta-historische Aspekte der Wissenschaftsgeschichtsschrei-bung der Linguistik. Seoul: Sowadalmedia. Tubingen:Gunter Narr (in Komission).

Schreyer R (2000). ‘What’s wrong with the historiographyof linguistics?’ Beitrage zur Geschichte der Sprachwis-senschaft 10, 205–208.

Sebeok T A (ed.) (1975). Current issues in linguistics. Vol.13: Historiography of linguistics (2 vols). The Hague:Mouton.

Simone R (1975 [1973]). ‘Theorie et histoire de la linguis-tique.’ Historiographia Linguistica 2, 353–378.

Simone R (1995). ‘Purus historicus est asinus: Quattromodi di fare storia della linguistica.’ Lingua e Stile30(1), 117–126.

Streitberg W (ed.) (1916–1936). Geschichte der indogerma-nischen Sprachwissenschaft seit ihrer Begrundung durchFranz Bopp (6 vols). Strassburg: Karl J. Trubner/Berlin:Walter de Gruyter.

Swiggers P (1997). Histoire de la pensee linguistique. Ana-lyse du langage et reflexion linguistique dans la cultureoccidentale, de l’Antiquite au XIXe siecle. Paris: PressesUniversitaires de France.

Szemerenyi O. (1971). Richtungen der modernen Sprach-wissenschaft 1: Von Saussure bis Bloomfield 1916–1950.Hidelberg: Winter

Taylor D J (ed.) (1987). The history of linguistics inthe classical period. Amsterdam & Philadelphia: JohnBenjamins.

Tagliavini (1963). Storia di parole pagane e cristiane attra-versoi tempi. Brescia: Morcelliana

History of Linguistics: Discipline of Linguistics 341

History of Linguistics: Discipline
of Linguistics N Smith, University College London, London, UK
� 2006 Elsevier Ltd. All rights reserved.

Introduction

Language makes us human.Whatever we do, language is central to our lives,

and the use of language underpins the study of everyother discipline. Understanding language gives us in-sight into ourselves and a tool for the investigation ofthe rest of the universe. Martians and dolphins, bono-bos and bees, may be just as intelligent, cute, adept atsocial organization, and morally worthwhile, butthey don’t share our language, they don’t speak‘human’. Linguistics, the scientific study of language,seeks to describe and explain this human faculty. It isconcerned with three things: discovering preciselywhat it means to ‘know a language’; providing tech-niques for describing this knowledge; and explainingwhy our knowledge takes the form it does.

These concerns may seem too obvious to need dis-cussing, but the complexity of our knowledge oflanguage becomes strikingly apparent when we seesomeone lose their language after they have had astroke, or when we observe an infant who has yet toacquire the faculty that we deploy so easily. To under-stand these three concerns, we need a theory, and thatis what linguistics provides.

The Meaning of ‘Language’

That linguistics is ‘the scientific study of language’ hasbecome a cliche, but what it means to be ‘scientific’

may not always be obvious, and what people meanwhen they use the word ‘language’ variesfrom occasion to occasion. Consideration of what isinvolved in being scientific is deferred until later inthe essay, for now it suffices to observe that only a fewaspects of language have been illuminated by theo-retical (scientific) linguistics, so there are many areaswhere it has little, if anything, helpful to say. Thesituation is akin to that in biology, viewed as thescience of living things. Despite their importance tous, biology has nothing to say about the definition ofpets; similarly, despite their relevance to us, linguisticshas nothing to say about the definition of dialects.

In everyday usage, ‘language’ is used differently,depending on whether it is construed as a propertyof the individual, of society, of the species, or as anautonomous entity in the world. Linguists working inthe tradition of ‘generative’ grammar, the frameworkthat has dominated linguistics for the last 50 years,argue that an ‘individual’ approach to language islogically prior to any other, but in principle we havethe possible domains shown in (1), each suggestingdifferent kinds of questions:

(1)
Language and the Individual Language and the Brain Language and Society Language and the Species Language and Literature Language and the World
Looking at ‘Language and the Individual’, thecentral question raised is ‘what constitutes our‘‘knowledge of language’’?’ What properties or attri-butes does one have to have to be correctly described

342 History of Linguistics: Discipline of Linguistics

as a speaker of English, or Burmese, or any other‘natural language’ – the term linguists use to refer tolanguages naturally acquired and spoken by humans,as opposed to the ‘artificial’ languages of logic orcomputing? An extension of this question is howand where knowledge of language is representedin the brain, and what mechanisms need to be postu-lated to enable us to account for our use of thisknowledge. Neurolinguistics is an area of remark-able growth, supported by technological advances inimaging.

Under ‘Language and Society’, sociolinguists raisequestions such as ‘What are the social variants (class,age, gender, power) that determine, or correlate with,the use of particular forms of the language?’ Awomanmight use some pronunciations or grammatical con-structions with statistically significantly greater fre-quency than a man of the same age, or a female of adifferent generation. For the world’s multilingualmajority, social considerations may even determinewhich language is used in specific situations. A Swissfrom Graubunden might use Romansh at home, SwissGerman at work, and High German at a conference.

Looking at ‘Language and the Species’, we mightbe preoccupied with the puzzle that all human chil-dren learn their first language with seeming effortless-ness, while the young of other species, howeverintelligent, show minimal such ability. Linguists, andresearchers in related fields, investigate not onlywhether this claim to uniqueness is indeed true but,if it is, how the faculty evolved.

When we turn to the relation between Language andLiterature, we confront several issues: ‘What is literaryform?’; that is, what are the linguistic properties thatmake something a novel or a novella, a sonnet or anepic? How are literary effects achieved? What are thelinguistic characteristics of a particular style?

Looking at ‘Language and the World’ raises issuesof three different kinds. First, how does languagerelates to things outside the head? That the word‘London’ refers to the capital of the United Kingdomis innocuous enough as an informal claim, but itraises interesting, and vexed, philosophical questions.The debate revolves around the status of language asa property of an individual, rather than as an entitywith independent existence. This ‘external’ notion oflanguage is presupposed by those who write irateletters to the press, inveighing against split infinitives,and lamenting the fact that our language is becomingdegenerate, either because of the sloppiness of mod-ern youth, the pernicious influence of text messaging,or the role of multiculturalism. The third issue is inmany ways the most obvious and the most puzzling:Why are there so many languages? Why does ‘human’have so many dialects?

Knowledge of Language

The generativist claim that study of the individual’sknowledge of language must be the first or exclusivefocus of a scientific linguistics is controversial; that itis a possible, indeed necessary, focus is not seriouslyin doubt. This individualistic claim implies that lin-guistics is a branch of psychology, ultimately biology,rather than, say, of sociology. This is not to deny thatthere are interesting domains of knowledge that takethe social conditions of language use as their centralfocus; it is rather to claim that there is a psychologicalenterprise which looks at one branch of human cog-nition and which lends itself to rigorous investigationand, moreover, that it is logically prior to looking atthe exploitation of this knowledge in society. Thisfocus on knowledge is highlighted in the claim thatthe subject of linguistics is ‘I-language’, rather than‘E-language’, where the ‘I’ stand for internal to aparticular individual, and ‘E’ stands for external(to the mind of the individual). A corollary of thisorientation is that the descriptions that linguists de-vise are said to be ‘psychologically real’, where this isnot a claim about psychological experimentation orthe kind of evidence used in formulating particularlinguistic hypotheses, but is simply the claim that weare investigating the human mind and that currenttheory is the closest approximation to the truth thatwe have.

The mind is ultimately a product of the brain (andother systems), and evidence about the mental cansometimes be gleaned from studies of the neural. Ingeneral, however, linguists remain agnostic about thedetails of the relation between the mind and the brain(frequently referring simply to the ‘mind/brain’).That is, we devise theories of a sub-part of humanknowledge, but whether that knowledge is localizedin the temporal lobe of the left hemisphere, or isdistributed throughout the brain, or whatever, is lessimportant. This is not because of lack of interest,but simply because – at present – theories of neuralstructure are too embryonic to cast much light onlinguistic generalizations. Different languages allowdifferent word orders, so that Japanese puts the verbat the end of the sentence and English puts it in themiddle. Linguistic theory must provide the means fordescribing and ultimately explaining this fact, but atpresent we have no inkling of how the differencebetween a Japanese and an English speaker might beneurally implemented, so the neurological structureof (this bit of) the language faculty is still a closedbook.

What do you have to know to count as a ‘speaker’of a language? If you say you speak English, it impliesthat you understand English, too. The point may


seem trivial, but knowledge of language is neutralbetween speaking and hearing; both activities drawon the same fund of knowledge. There is no knownillness or accident which leaves you able to speak onlyEnglish and understand only Portuguese, for instance.This is not to deny that you may be better at talkingthan listening; or that you may suffer brain damagethat leaves you unable to speak while you remainperfectly able to understand. A particularly poignantexample of this is provided by Bauby’s autobiographi-cal account of ‘locked-in’ syndrome, where a strokeleft the author speechless, but with his languageand his ability to understand intact. In normal, non-pathological, cases, however, your ability to utter (2a):

(2a)
Giraffes have long necks (2b) Giraffes have necks long
involves the same ability that enables you to under-stand (2a), and also to judge that someone who mis-takenly says (2b) has got it wrong. The implication ofthis observation is that the primary focus of linguis-tics is on characterizing this neutral knowledge, rath-er than the mechanisms of speaking, hearing, andjudging that are parasitic on it. In other words,linguistics is (largely) about one form of cognition,and only secondarily about the deployment of thatcognitive ability. In the standard terminology, this isknown as the ‘competence-performance’ distinction.Your knowledge of language (your competence)underlies your ability to speak, to understand, andto give judgements of well- or ill-formedness (yourperformance). You may be knocked unconscious andbe temporarily unable to speak or understand, butyour knowledge typically remains intact – you havecompetence with no ability for performance. Theconverse situation, in which you could perform inthe absence of any competence, does not occur,though it may characterize the ‘linguistic’ capabilitiesof the average parrot, which may be able to utterentertaining sentences of what sounds like English,but presumably without the knowledge of Englishgrammar that underlies our abilities.

To count as a speaker of English, you need first toknow a large number of words: not just nouns, verbs,and adjectives – words such as cat and go and pretty,whose meaning is relatively transparent, but itemssuch as the, under, and however, whose meaningand use are less easy to specify. Of course, not every-one has the same vocabulary: I may know technicalterms in linguistics that you are ignorant of, and youmay be familiar with words pertaining to reggae orarachnology that I don’t know, but if either of us wereignorant of words such as mother or and, peoplemight be justifiably reluctant to classify us as speakersof English.

In addition to knowing the words of a language,you need to know what to do with those words – youneed to know the grammar. Our knowledge of lan-guage falls into two compartments – the vocabulary(or ‘lexicon’) and the ‘computations’ we can carry outusing that vocabulary. This computational system,comprising syntax and morphology, is surprisinglycomplex, and enables us to produce baroque exam-ples such as Chomsky’s (1995: 88) Who do you won-der whether John said solved the problem? Suchsentences are of marginal acceptability and citingthem may strain the tolerance of outsiders, but thismarginal status may itself provide crucial evidence foror against some theoretical claim concerning ourknowledge. Henceforth, I shall assume that you andI have the same I-language, abstracting away fromdifferences in vocabulary and grammar. Fortunately,it’s enough for the present purposes to look at themore basic, but nonetheless rich and surprising,knowledge we have of words as simple as be andthe. Consider the examples that follow, which illus-trate a wide range of things you know, even if youweren’t previously aware of knowing them. It’s self-evident that is and have mean different things, asshown in (3), but sometimes they seem to be usedinterchangeably as in (4):

(3a)
Tom is a problem (3b) Tom has a problem
(4a)
Tim is yet to win the Booker prize (4b) Tim has yet to win the Booker prize
How is it that something as basic as is can sometimesget the same interpretation as has and sometimes adifferent one? Or consider the so-called definitearticle (the), which is often said to mark the distinc-tion between entities which are already familiar andthose which are new, as in (5a) and (5b) respectively:

(5a)
My friend likes the penguins (5b) My friend likes penguins
But this characterization is not adequate to accountfor the rather macabre effects found in the newspaperreport in (6b) beside the relatively unexceptionable(6a):

(6a)
The woman had lived with the dead man fortwo years
(6b)
The woman had lived with a dead man for twoyears
Still less can it account for the fact that on occasionthe presence or absence of the seems to indicate thedifference between subject and object, as in (7):

(7a)
This man is in charge of my brother (7b) This man is in the charge of my brother


In (7a) this man has control of my brother; in (7b) mybrother has control of this man. So what does thereally mean? Does it even make sense to ask such aquestion?

Take a more complex example: the word last ismultiply ambiguous: apart from its use as a noun ora verb, it can function as an adjective meaning either‘final’ or ‘previous’, as illustrated in (8):

(8a)
This is your last chance (8b) Your last example surprised me
This ambiguity can result in dialogues which havestrikingly different interpretations, as in the alterna-tives in (9):

(9)
Q ‘ ‘What were you doing in Paris?’’ A1 ‘ ‘Oh, I was collecting material for my last
book’’
A2 ‘ ‘Oh, I’m collecting material for my last
book’’

Answer 1 is itself ambiguous, with either meaningpossible for last (though ‘previous’ is the more easilyaccessible); Answer 2 has only the interpretation thatthe book under discussion is planned to be the finalone I write. The difference must be attributable tothe contrast between the past and the present tense,as that is the only way the sentences differ, but it’s notobvious why sentences should be ambiguous or notdepending on the tense they contain.

Linguists thrive on such ambiguity, as it regularlyprovides evidence for structural differences that maynot be otherwise apparent. A simple example isprovided by the inscrutable notice outside our localschool, given in (10):

(10)
This school accepts girls and boys under six
Whether the school accepts girls of any age but onlysmall boys, or no children over six is indeterminatewithout more information. As we shall see under thesection ‘Describing Knowledge of Language,’ (10)has two quite different syntactic structurescorresponding to the two meanings. Similarly thefact that (11) has a number of different interpreta-tions can give us clues as to how to analyze thevarious possibilities:

(11)
My son has grown another foot
If my son has become taller, the example is parallel to(12a); if he is a freak or a remarkably successfulgardener, there are other possibilities, as shown in(12b) and (12c), suggesting that another foot in (11)may be correctly analyzed either as a measure phraseor as a direct object:

(12a)
He has grown by another foot (12b) He has grown a third foot
(12c)
Another foot has been grown (in thisflowerpot)
Such differences of interpretation make the complex-ity of our knowledge apparent, but unambiguousexamples can be just as illuminating and can simulta-neously provide evidence against the traditional phil-osophical claim that meaning can be adequatelytreated in terms of truth. Thus, we know that (13):

(13)
My first wife gave me this watch
suggests rather strongly that I have been marriedmore than once, but I can utter it truthfully despitehaving been married only once: my only wife ispresumably my first wife. The example is misleading,not false, and so implies that there is much more tomeaning than mere truth. As shown by Chomsky’s(1957) famous Colorless green ideas sleep furiously,structure and meaning (syntax and semantics) candissociate, so we also know that, despite being initi-ally plausible and syntactically unexceptionable, (14)is meaningless:

(14)
More people have visited Moscow than I have
All the preceding examples illustrate both our knowl-edge of vocabulary and how it interacts with (syntac-tic) structure. The responsibility of linguistics is todescribe the full range of such facts, not just forEnglish, but for all human languages. Then, in virtueof its scientific pretensions, it has to (attempt to)explain why these facts rather than any others arethe ones that occur – again both in English and inother languages. To do justice to the richness of whatwe know, it is necessary to distinguish not just thelexicon and the computational system, but to differ-entiate among syntax, semantics, morphology, pho-nology and phonetics, and to relate this knowledge topragmatics – how we interpret utterances in context.

Take our knowledge of morphology, the internalstructure of words. We know that thick, thicker,thickest, and thicken are all words of English, butthat there is no thinnen to accompany thin, thinner,thinnest. We know that thick relates to thicken andthat rich relates to enrich, whereas richen is slightlyodd, and enthick is impossible. This knowledge can’tjust be a result of our never having heard thinnen orenthick before, you may never have heard textedbefore, as in ‘‘I’ve just texted an urgent message toFred’’, but you know that that is possible. As linguists,we may also know that some languages, such asVietnamese, have almost no morphology: words inthis language have none of the internal structurecharacteristic of affix-rich items such as indecisive-ness or rearranged. On the other hand, some (poly-synthetic) languages, such as Inuktitut (Eskimo) or


Mohawk pile one affix on top of another so thatwords are often strikingly complex, and correspondto whole sentences in English. Baker (2001: 87) givesthe Mohawk example in (15) with the meaning ‘‘Hemade the thing that one puts on one’s body ugly forher’’:

(15)
Washakotya’tawitsherahetkvhta’se’
Our knowledge of phonology, the sound structure oflanguage, is equally rich. We know that past, spat,and stap are possible words of English, indeed they allexist; that stip and stup are also possible words, eventhough they happen not to exist; but that satp, ptasand tpas are not even possible words. Apart fromknowing the segmental make-up of words, we alsohave knowledge of ‘supra-segmentals’: that photo-graph is stressed on the first syllable, photographeron the second, and photographic on the third. Twopoints need to be made: first, we ‘know’ this in thesense that we produce the correct pronunciations ondemand, and we recognize that deviations from thesepronunciations are slips of the tongue or foreigners’mistakes; that is, knowledge of language need not beimmediately available to conscious introspection.Second, the characterization in terms of ‘first’, ‘sec-ond’, and ‘third’ syllable is actually not the correcttheoretical characterization of our knowledge. Aswe shall see below, rules of grammar (includingphonology) cannot count.

We know more. In an example such as (5a) above,My friend likes the penguins, we have to account forthe pronunciation of the before the initial ‘p’ of pen-guins: a pronunciation rather different from that ofthe same lexical item the when it occurs before avowel, as in My friend likes the otters. Knowledgeof this kind is supplemented by phonetic knowledgewhich is even harder to bring to consciousness: thatthe ‘t’ in photographer is aspirated, but the ‘t’ inphotograph is not; that the ‘r’ in grime is voiced, butthat in prime it is slightly devoiced; that the vowel islonger in wed than in wet. Such facts belong to thedomain of phonetics, the field that deals withthe sound properties of language in general, ratherthan the sound structure of a particular language.

Our phonological knowledge is not self-contained,but may interact in complex ways with our knowl-edge of the rest of the grammar. We know that (16a)has an alternative pronunciation of the kind given in(16b), where is is ‘contracted’ to’s, but that (17a)cannot be matched by the impossible (17b) (impossi-ble is indicated by the asterisk), despite the apparentsimilarity of the examples:

(16a)
The prime minister is a war criminal (16b) The prime minister’s a war criminal
(17a) The president is a war criminal and the prime
minister is too
(17b)
*The president is a war criminal and the primeminister’s too
An understanding of such asymmetries requires inves-tigation of the relation between syntactic and phono-logical processes, and relies on an analysis of emptycategories: entities that have syntactic and semanticproperties but are silent.

In addition to phonology and morphology, we needto account for the (semantic) fact that sentences havemeaning. The examples in (18) exploit most of thesame words but their meanings are radically different:

(18a)
My friend likes the penguins (18b) The penguins like my friend (18c) My friend doesn’t like the penguins
Moreover, the semantics is ‘compositional’ – exceptfor idioms, the meaning of a sentence is a function ofthe meaning of its parts, and their syntactic configu-ration. The meaning difference between (18a) and(18b) is dependent on which item is subject andwhich object, notions that can be defined syntacti-cally. In fact, life is a little more complicated thanthat, as the semantic interpretation of ‘subject’ is notuniform, and we need to advert to ‘thematic relations’involving ideas of agentivity and patienthood, asshown by the minimal pair in (19):

(19a)
John undertook the surgery reluctantly (19b) John underwent the surgery reluctantly
John is the subject in both sentences, but is the agent(the surgeon) in the former, and the patient (in bothsenses) in the latter. These relations are internal to asingle sentence, but we also need to relate (the mean-ings of) different sentences. There are two possi-bilities: relations which depend on the meaning ofindividual words, and relations which are purely sen-tential in that they are independent of such lexicalrelations. An example of the former is illustrated by(20):

(20a)
Mozart persuaded da Ponte to write a libretto (20b) Da Ponte intended to write something
where (20b) follows, by virtue of the meaning ofpersuade from (20a). An example of the latter isprovided by pairs such as (21a) and (21b), where thetruth of (21a) guarantees the truth of (21b):

(21a)
Torture is immoral and should be illegal (21b) Torture is immoral
In the next section, I will outline some of the descrip-tive mechanisms exploited by (generative) linguistics;then I will try to show how we can approach anexplanation for at least some phenomena, looking at


a range of examples from English and elsewhere, anduse this extension to substantiate the claim that lin-guistics is a science. Throughout, I shall concentrateon syntax. Phonology and phonetics, morphology,and semantics are rich disciplines in their own right,each with a massive literature, but the essence of theanalysis of sentences is their syntactic structure. Andlife is finite.

Describing Knowledge of Language

Sentences have structure of various kinds. Returningto the example, My friend likes the penguins, we needto describe it in different ways at several distinct‘levels of representation’: phonological, semantic,and syntactic. Thus, it can be pronounced in a varietyof ways – with stress on friend or on penguins, forinstance, with concomitant differences of interpreta-tion. Restricting attention to the syntax, it is intuitive-ly clear that my and friend, and the and penguins ‘gotogether’ in a way that friend and likes, and likes andthe do not. Each initial word of My friend and thepenguins enables us to pick out some individual orindividuals in the world, whereas friend likes andlikes the have no such function within them. Thisintuition is accounted for in terms of ‘constituency’represented by means of a simplified tree diagram ofthe kind in (22):

(22)

The top of the tree (IP) indicates that the wholesequence ‘‘My friend likes the penguins’’ is an ‘I(nflec-tion) P(hrase)’{it used to be called ‘Sentence’, but theterminology has changed to reflect changes in ourunderstanding}. The IP ‘branches’ into an NP anda VP, where ‘NP’ means ‘Noun Phrase’, that is asequence consisting of a Noun (N) and somethingelse, and ‘VP’ stands for ‘Verb Phrase’, that is a se-quence consisting of a Verb (V) and something else,here in this instance another Noun Phase. The verb isthe (present-tense) form likes, and the two NounPhrases each consist of a Noun (here the singularfriend and the plural penguins) preceded by a ‘Det(er-miner)’, respectively my and the. Each of ‘IP’, ‘NP’,‘VP’, ‘N’, etc., are referred to as ‘nodes’ in the tree;IP, NP, and VP, etc., are said to ‘dominate’ every-thing below them, and to ‘immediately dominate’

everything immediately below them. So VP domi-nates all of V, NP, Det, N, the, and penguins, butimmediately dominates only V and NP, which areknown as ‘sisters’. Once one has got used to thejargon, the advantages of such an analysis are many:it simultaneously shows the linear sequence of items –the order they come in – and the relationships amongthe component parts: so the and penguins are moreclosely related, by virtue of being NPs, than are likesand the which do not form a ‘constituent’ of any kind.A constituent is defined as any sequence of items thatcan be traced exhaustively to a single node in the tree:likes and the can be traced back to VP (and indeed IP),but these nodes dominate other material, too (pen-guins, for instance) so likes the, like friend likes, arenot constituents.

We now have an explicit way of characterizingthe example This school accepts girls and boysunder six. The two interpretations of the object,girls and boys under six, can be represented withdifferent constituent structure as in (23):

(23a)
[girls] and [boys under six] (23b) [girls and boys] [under six]
where the brackets mark the constituents, and indi-cate that the ‘scope’ of under six is respectively eitherjust boys (23a) or includes girls and boys (23b).

In addition to this syntactic constituent structure,there is morphological structure to deal with: the factthat penguins is plural is marked by the addition ofthe suffix –s to the base penguin, and the oppositeorder (with s- prefixed to penguin to give spenguin) isimpossible (in English). Investigating the full range ofsuch facts in the world’s languages is a matter ofintensive research, and addresses the same immediatetask of accounting for how it is that native speakerscan have the intuitions and make the judgements ofwell- and ill-formedness that they do.

This last point bears elaborating. One of thesurprising facts about our linguistic ability is that itextends to knowing what is impossible as well aswhat is possible: we have intuitions of ill-formednessor ‘negative knowledge’. I have already traded on thisfact in assuming that, even though you had probablynever heard either example before, you would agreethat Giraffes have necks long is wrong, whereas I’vejust texted an urgent message to Fred is acceptable.The point can be generalized: the fact that one canrecognize mistakes and distinguish them from newbut well-formed creations is evidence for the rule-governed nature of the language faculty. Mistakespresuppose norms, or rules. It is also noteworthythat there are ‘impossible’ mistakes: some logicallypossible errors just don’t happen, even though onemight expect them to. Consider an example from


language acquisition and the task of the child inworking out how questions and statements of thekind in (24) are related:

(24a)
The children are playing truant (24b) Are the children playing truant?
There are all sorts of hypotheses children might en-tertain: move the auxiliary (are), move the thirdword, permute the first and second constituents, andso on. The kinds of mistake that children do make,however, show that their hypotheses overlap withthese in interesting ways. First, they sometimesmake mistakes of a kind for which there is no obviouspattern in the input, even though they may betheoretically well motivated: examples such as the‘auxiliary copying’ in (25):

(25a)
Is the steam is hot? (25b) Are the children are playing truant?
Second, they never try out any hypothesis that wouldinvolve them in counting: their attempts always rangeover modifications of linguistic structure, never ofmathematical structure. It seems that all rules in alllanguages are what is called ‘structure-dependent’ –they depend on notions such as constituent, NounPhrase, and so on, but not ‘third word’. Moreover,children seem not to need to learn this fact – it is aprinciple that guides their language acquisition fromthe start: it is innate. Claims of innateness have beenunnecessarily controversial in modern linguistics. Noone doubts that humans are innately (genetically)different from cats, chimpanzees, and dolphins, andthat this difference underlies our ability to acquirelanguage. Equally, no one doubts that humans ac-quire different languages depending on the environ-ment they are brought up in: if children are broughtup in Turkey rather than Greece, they learn Turkishrather than Greek. It is obvious that both nature andnurture have a crucial role to play. Where controversyis justified, and where empirically different claims canbe tested, is in the detail of what needs to be ascribedto the child’s ‘initial state’, of what precisely is innateand what has to be acquired on the basis of experi-ence. Explaining structure-dependence is an areawhere innateness has been repeatedly (and controver-sially) defended with a form of argument based on the‘poverty of the stimulus’ – the idea that we end upknowing things that it is impossible, or at least im-plausible, to think that we could find in the input.Consider examples more complex than those above,such as (26):

(26a)
The children who were naughty are playingtruant
(26b)
Are the children who were naughty playingtruant?
If ‘moving the third word’ or ‘moving the (first) aux-iliary’ were really the correct way of characterizingthe relation in (24) one would expect to find examplemistakes like that in (27):

(27a)
Who the children were naughty are playingtruant?
(27b)
Were the children who naughty are playingtruant?
Such mistakes simply do not occur. Of course, it isalways (usefully) dangerous to say that somethingdoes not happen: it may happen in the next utteranceone comes across. But this means that the claim iseminently falsifiable (see below), and can anyway bechecked by looking for relevant counterexamples inthe literature. A nice example of this kind is providedby Neeleman and Weerman’s (1997) account ofacquisitional differences between Dutch and English.They predicted that Dutch children should, andEnglish children should not, produce sentences withan adverb intervening between a verb and its object,as in (28):

(28)
I will eat quickly the yoghourt
They ransacked the largest international corpus ofchild data in checking their predictions, and happilyfound no exceptions.

Formalizing our knowledge of language demands acomplex toolkit, only a tiny fraction of which hasbeen given here, but such formalization is a necessaryprerequisite to finding explanations, to assimilatinglinguistics to the scientific enterprise. Given the toolsdeveloped here, we can make general hypothesesabout the nature of language and begin to test themon a wider range of data from English and elsewhere.

Explanation in Language

Examples involving structure-dependence enableone to address the demand for explanation in addi-tion to description. Let’s pursue the issue by lookingat the occurrence of items such as any, ever, or any-thing in English (so-called ‘Negative Polarity Items’).At a descriptive level, it is sufficient simply to contrastpossible and impossible sentences of the sort seenin (29a) and (29b), where those in (29a) are fullyacceptable but those in (29b) are ungrammatical, orinfelicitous, or just wrong:

(29a)
John ate something/ some salad (29b) *John ate anything/ any salad
But why is there this contrast? The example in (30)shows that any(thing) can occur happily enough innegative statements, but it occurs unhappily in posi-tive statements:


(30) John didn’t eat anything/ any salad

Looking at such negative examples, the generaliza-tion seems to be that any(thing) needs to occur with(be ‘licensed by’) a negator. But such an account isinadequate in two different ways: first, (31) showsthat it is not just negators that are relevant, but avariety of elements behave in a similar fashion. Thisclass includes questions, conditionals, and other itemsthat there is no space to characterize:

(31a)
Did John eaten anything/ any salad? (31b) If John ate anything/ any salad, I’d be amazed (31c) Everyone who has any sense has left already (31d) John denied having eaten any of the cakes
Second, even if we restrict ourselves to negatives, itstill seems that life is more complicated than we mightwish – (32a) is unsurprisingly fine but, despite beingnegative, (32b) is unacceptable and none of (32c) to(32e) is acceptable:

(32a)
Something/ some spider bit him in the leg (32b) *Anything/ any spider didn’t bite him in the leg (32c) *Anything is annoying me (32d) *Anything isn’t annoying me (32e) *John denied any of the accusations
That is, our first approximation that any needs to belicensed by a negative fails in both directions – somesentences with negatives do not allow any; some sen-tences without a negative do allow any. The nextobvious assumption might be that any(thing) has tobe preceded by a negator of some kind (not or n’there), but (33) shows that this hypothesis is inade-quate: it works for (33a) and (33b) but not for (33c)or (33d) – where nothing is another negator:

(33a)
The fact that he has come won’t changeanything
(33b)
Nothing will change anything (33c) *The fact that he hasn’t come will change
anything
(33d) *That nothing has happened will change
anything

The examples in (33a) to (33d) suggest another pos-sibility: perhaps the negator has to be in the sameclause as the item (any) being licensed? In (33a), thenegator and anything are in the same clause (compare‘‘This won’t change anything’’), whereas in (33c) and(33d), the negator is in a different clause. We aregetting closer, but (34) shows that this is still inade-quate as an explanation, as here the negator andanything are blatantly in different clauses, but theresult is well-formed.

(34)
I don’t think he has eaten anything
The claim that the negative (or other item) must bein the same clause as any fails: some sentences have

the negative in a different clause and are nonethelessgrammatical; some have the negative in the sameclause and are ungrammatical. The correct expla-nation necessitates an appeal to the notion of‘c-command,’ a relation between ‘nodes’ in a tree.To make this comprehensible and plausible, we needto introduce a little more of the technical machineryof generative grammar.

The representation of sentence structure in termsof trees of the kind shown in (22) can obviously beextended to show the structure of (29a), as shownin (35), where the only novel feature is the uncontro-versial claim that some is a kind of Determiner:

(35)

More complex sentences require more complexconfigurations. Thus, the salient property of an ex-ample such as (33a) ‘The fact that he has come won’tchange anything’ is that the subject is not just a nounor noun phrase, but a noun phrase containing anothersentence ‘he has come’. To a first approximation itwould have the (simplified) form given in (36),and the ungrammatical example in (33c) *The factthat he hasn’t come will change anything would becharacterized by the tree given in (37):

(36)

Some of the details of the tree have been included forthe sake of those who are already familiar with syn-tax. So the Complementizer Phrase (CP), optionallyheaded by a Complementizer such as that, and the I’(a constituent intermediate in size between a sentence(IP) and an Inflection element like will) are there for


the cognoscenti. But two things in these trees areimportant for everyone: first, that they contain aconstituent Neg(ation), itself a subpart of a Neg-P(hrase); and second, that it makes sense to talk ofone item being higher in the tree than another. That is,in (36), the ‘Neg’ is higher in the tree than anything,whereas in (37) the ‘Neg’ is lower in the tree thananything.

(37)

To make this account rigorous, we need to defineexactly what is meant by ‘higher’ and ‘lower’, andthat is what is meant by ‘c-command’: a node A in atree c-commands another node B if and only if thefirst branching node dominating A also dominatesB. In (36), Neg c-commands anything because thefirst branching node above Neg (i.e., NegP) alsodominates the NP anything; in (37), Neg does notc-command the word anything because the firstbranching node above Neg (again NegP) does notdominate anything.

It may seem as if we are using a sledgehammer tocrack a nut, but the beauty of the analysis is that c-command is not just an arbitrary condition intro-duced to account for a narrow range of data inEnglish. Rather it extends in two directions: it is amajor and essential ingredient in the explanation firstof a range of other phenomena in English; and secondto a wide range of phenomena in other languages,indeed in all languages: c-command is universal.

Before illustrating other uses of c-command, notethat if it is universal, we would like an explanation forhow that is possible. The obvious answer is that it isinnate, part of the faculty of language that differenti-ates humans from other organisms and explains whyall kids but no kittens acquire language. If correct,certain implications follow immediately: c-commandis not a condition that children acquiring their first

language need to learn, rather (like structure-depen-dence) it acts as a constraint that determines the kindof hypotheses they can come up with in masteringtheir first language.

Let us look at one generalization of the usefulnessof c-command in English: its use in ‘binding theory’,the part of linguistics that deals with the distributionof pronouns and reflexives. It is a commonplace thatreflexive pronouns such as myself, yourself, himself,and so on, have to agree (or ‘be compatible’) withtheir antecedent – the entity they refer back to, sothe examples in (38) are fine, but those in (39) areungrammatical:

(38a)
I admire myself (38b) The judge admires himself (38c) The waitress might flatter herself
(39a)
*I admire yourself (39b) *He admires herself (39c) *The waitress flattered ourselves
There are all sorts of other interesting complicationswith reflexives: if there are two possible antecedents,the sentence is ambiguous, so in (40) herself can referto either the nurse or the woman:

(40)
The nurse showed the woman some documentsabout herself
but this is true only if the two potential ante-cedents are in the same clause as the reflexive: (41)is unambiguous, and herself can refer only to theprincess, because the queen is in a different clause:

(41)
The queen said the princess had disgracedherself
Neither of these extra considerations accounts forwhy (42a) is unambiguous and (42b) is simply un-grammatical:

(42a)
The mother of the princess has disgracedherself
(42b)
*The brother of the princess has disgracedherself
The question is why herself in (42) can’t refer back tothe princess, but only to the mother or the brother,resulting in the judgements indicated. The answer isthat the antecedent of the reflexive must not only becompatible and in the same clause, but must also c-command it. The structure of possessives such as theprincess’s mother or the mother of the princess is amatter of contention, but what is not in dispute is thatprincess is lower in the tree than mother or brotherand hence does not c-command the reflexive:compare the trees in (43a) and (43b) for (38c) and(42a):


(43a)

(43b)

In both trees, the underlined DP (The waitress in(43a), The brother/mother of the princess in (43b))c-commands herself, but the crossed-out DP Theprincess in (43b) does not c-command herself so can-not act as its antecedent.

C-command is pervasive in the syntax of English,not just in accounting for polarity items and reflex-ives. More strikingly, it is pervasive in the syntax ofevery other human language. Consider (Cantonese)Chinese. Cantonese has a rich selection of sentence-final particles with a wide range of meanings fromconveying a feeling of intimacy to indicating whichelement in the preceding sequence is the focus. InEnglish, we can indicate this focus by means of stress,giving rise to the kind of difference in (44a) and (44b):

(44a)
John only watches football (he doesn’t play it) (44b) John only watches football (not cricket)
It’s even possible, with suitable pause and stress, tohave (45b) with the same interpretation as (45a):

(45a)
Only John watches football (not Bill) (45b) John only, watches football (not Bill)
Just as in English, Cantonese uses stress to identify theintended focus from the set of possible foci, and theoperator zaa3 (only) then associates with this in-tended focus, as in (46), which can have the variousinterpretations shown in (47):

(46)
Billy tai zukkau zaa3 Billy watch football zaa3
(47a)
Only Billy watches football (not Peter) (47b) Billy only watches football (he doesn’t play it) (47c) Billy only watches football (not cricket)
There is good evidence (see Law, 2004) that zaa3occurs in some C position of the sentence, and hencec-commands everything preceding it in the example in(46): see the tree in (48), (C is arguably final inCantonese, not initial as it is in English):

(48)

But to talk simply in terms of linear precedence orword order is inadequate. Cantonese also has a pro-cess of topicalization whereby a constituent – e.g.,zukkau (‘football’) – can be moved to the front ofthe sentence, where it is attached even higher in thetree than zaa3, and marked with le1 (the 1 indicates ahigh level tone). This is shown in (49a), with a rangeof putative translations in (49b) to (49d). Crucially, asindicated by #, (49d) is not a possible interpretationof the Cantonese sentence.

(49a)
zukkau-le1, Billy tai t zaa3 (49b) Football, only Billy watches (49c) Football, Billy only watches (49d) #Only football does Billy watch
Why this should be so is indicated in the tree in (50),where zukkau is not c-commanded by zaa3: (The ‘t’,for ‘trace’ in (48a) and (49a) indicates where thetopicalized constituent zukkau moved from).

(50)

Because zaa3 does not c-command zukkau, theattempted interpretation in (49d) is simply impossi-ble. The examples are extremely simple, indeedextremely oversimplified, but the moral is clear:the same abstract syntactic condition (c-command)


operates in Chinese just as it does in English, and inevery other language.

It is worth emphasizing that successful explana-tions for one class of data are good to the extentthat they generalize to phenomena for which theywere not devised. C-command was not invented toaccount for Chinese, but the fact that it automaticallyaccommodates quite subtle data in that languagelends support to a theory that incorporates it. Thepoint can be illustrated more widely. Every time onedraws a tree of the kind illustrated above, one makespredictions about the well-formedness of a host ofother sentences. It is striking that the trees in (36)and (37) exhibit a defining property of human lan-guage – its recursive power. That is, the possibilityof including one sentence inside another sentence,potentially without limit, gives rise to the infiniteexpressive power of natural language syntax.

Linguistics as a ‘Science’

Making testable predictions of this kind is one of thehallmarks of science, and we can now elaborate on theclaim that linguistics is ‘scientific’. For any disciplineto be scientific it must satisfy (at least) the conditionsin (51):

(51a)
It must seek explanation (51b) It must pursue universals (51c) This will necessarily involve idealization which
may well give rise to a tension betweencommonsense and science

(51d)
Most crucially, it will make falsifiablepredictions
The scientific enterprise is a search for explanatorylaws or principles. That is, linguists – like physicistsor molecular biologist – seek not only data, but alsodata that can be used as evidence for some theoreticalclaim. Consider again the analysis of reflexives. Earlywork in linguistics of the sort best exemplified by thework of Bloomfield (1935) provided detailed exem-plification of a wide range of reflexive constructionsfrom a variety of languages, but stopped short oftrying to explain their distribution. One of theachievements of generative grammar has been pre-cisely to explain – in terms of ‘binding theory’ – whyreflexive pronouns have the distribution they do. Toelaborate a little on the discussion given alreadyunder the ‘Explanation in Language’ section, theappearance of a reflexive pronoun is determined byprinciple A of binding theory which says that a reflex-ive must be ‘bound’ in some domain. As we saw, thismeans that it must have an antecedent which alsomeets a number of other conditions. Principle A is incontrast with Principle B, which determines the

distribution of ‘ordinary’ pronouns. That is, betweenthem the principles account for the range of factsdiscussed above as well as for the contrast betweenJohn admires him and John admires himself; why onecan construe John and him as referring to the sameperson in (52b) but not in (52a), even though thelatter seems to include the former as a proper subpart,and a host of other facts:

(52a)
John expects to see him (52b) I wonder who John expects to see him
Evidence for – or against – the claims of bindingtheory, or any part of the theoretical edifice, canbe drawn from a wide variety of domains: the distri-bution of words in sentences; the acquisition of theirfirst language by children, and of second andsubsequent languages by both children and adults;the historical change of language over time; the pro-cessing of language – be it production or perception –in normal and abnormal circumstances; the problemsthat arise in pathology, as a result of language distur-bance caused by a stroke or a tumor, and so on. Inevery case, explanation calls for concentration onthose data that can provide evidence: the data them-selves are trivial until embedded in a theory that canproduce testable hypotheses.

A concomitant of this search for explanation is thatthe generalizations made must carry over in relevantways to all languages, not just to English or Latin orChinese. That is, the quest for laws entails that anyscience must pursue universals, even if that meansnarrowing the domain of inquiry. This position hastwo implications: first, that the same principlesshould apply to Dutch and Hindi and Chinese – so‘all languages’ is to be construed literally; but second,that the domain of application of these principles maynot be superficially obvious. To take the second ob-servation first, it is well-known that there are so-called ‘emphatic’ reflexives, as illustrated in (53a)and (53b), which raise difficulties for any simpleanalysis of reflexivization in general:

(53a)
John himself came (53b) John came himself
These ‘reflexives’, so labeled because they include themorpheme {self }, have somewhat different propertiesfrom ‘real’ reflexives: for instance, they don’t haveany thematic role, (came takes only one argument –you can’t ‘come somebody else’), but simply empha-size the importance of the one role mentioned. Onthe other hand, they clearly do obey some of the sameconstraints as ordinary reflexives, as witness thepeculiarity of the examples in (54):

(54a)
*The boy came herself


(54b)
*The boy’s mother himself came
This duality suggests that it might be necessary – as atemporary measure – to limit the domain of bindingtheory to arguments taking a thematic role, leavingthe emphatic examples to be accommodated laterafter further research. The situation is parallel tothe development of a scientific theory of motion.For Aristotle, all motion fell within the ambit of histheory of movement, even the movement of flowersgrowing. Galileo was able to provide a unified ac-count of terrestrial and heavenly motion by restrict-ing attention to mechanical motion and excludingbiological growth. This should not be viewed as aretreat to a position where whatever you say turnsout to be true, simply because you have excludedthose areas where what you say is false. Rather it isan attempt to define an area where we can beginto understand the complexity of the real world byfocusing on phenomena that are comprehensible.

This narrowing is of two kinds: first, one can sim-ply ignore data which fall outside the generalizationone is attempting to explain; second, there is scientificidealization – the pretence that things are simplerthan they really are. This is justified because suchsimplification enables one to approach an under-standing of the abstract principles which underliecomplex phenomena. Such idealization in linguisticswas first made explicit in Chomsky’s distinction be-tween competence and performance and his claimthat ‘‘linguistic theory is concerned primarily withan ideal speaker-listener, in a completely homoge-neous speech-community.’’ We all know that realspeech communities are not homogeneous, but theforce of the idealization is that the heterogeneitythat does exist is not a necessary component in anadequate characterization of our knowledge of lan-guage or how we come by that knowledge. Considerin this latter respect the simplifying assumption – thestriking idealization – that first language acquisitionis ‘instantaneous’. It is obvious that children take aconsiderable time to master the intricacies of theirfirst language. Given how complex the knowledgethey end up with is, it may still be justifiable to talkof the surprising speed with which they reachthis mastery, but it is not by any stretch of the imagi-nation ‘instantaneous’. So what is the force of theassumption? Consider the acquisition of negation.

Most, perhaps all, children go through a stage inwhich they produce negative sentences with the nega-tive marker (no or not in English) in peripheral posi-tion in the sentence – i.e., first or last – as in (55a) and(55b), heard from two different two-year-olds:

(55a)
No computer on (55b) Computer on no
The context made it clear in each case that the forceof the utterance was an order not to turn the comput-er on. Superficially, it looks as if the two children havedifferent grammatical systems (though they wereequally proficient at understanding adult instruc-tions, suggesting that their grammar was more so-phisticated than might appear). What is relevanthere, however, is the fact that – as far as is known –both children will end up with the same grammaticalknowledge of English negation. That is, the differentstages they go through in their acquisition of thedetails of the grammar has no effect on the knowledgethey end up with – their adult competence. This claimmay, of course, be false. It might turn out that adultswho uttered (55a) as children have different gram-mars from those who uttered (55b) as children. It’spossible, but there is no evidence to that effect, andthe idealization to instantaneity is accordingly justi-fied. If one of the things we wish to explain is howhumans can progress from a stage in which they areapparently language-less to a stage of adult knowl-edge, it is advantageous to be able to abstract awayfrom the different paths they may take in acquiringthat knowledge. The idealization also simplifies theaccount of the initial state of the language faculty:what needs to be attributed to the mental make-up ofhuman infants to explain the fact that they do, whileinfant chimps do not, acquire language.

Idealization of this kind is in turn likely to involve atension between commonsense and science. Theclaim of instantaneous language acquisition seemsblatantly silly until one considers more carefullywhat it means. Consider a second example, againfrom first language acquisition. Children regularlymispronounce the words they are learning, sometimeswith surprising results, as in the case of the puzzlepuzzle. When he was about two and a half, my son –like many children – used to pronounce puddle as‘puggle’ ([pVgel]). He was perfectly consistent, andused to pronounce words of a comparable kind withthe same kind of deformation: so bottle became‘bockle’, pedal became ‘peggle’, and so on. The obvi-ous explanation for this behavior was that, for rea-sons of motor control, he was unable to pronouncepuddle. But at the same time as he made this mispro-nunciation, he was also making ‘mistakes’ withwords such as zoo, pronounced as ‘do’, lazy, pro-nounced as ‘lady’, and so on. The result was striking:although he pronounced puddle as ‘puggle’, heconsistently pronounced puzzle as ‘puddle’ ([pVdel]),so the reason for the former ‘mistake’ could clearlynot be that he was incapable of the appropriate motorcontrol. He could pronounce ‘puddle’, but only as hisversion of puzzle not for puddle. So the commonsenseexplanation of the phenomenon was wrong. An


obvious alternative explanation was that he couldn’thear the difference, but that hypothesis wasn’t muchmore plausible either, as his pronunciations of the twowords were consistently different, indicating that hemust be able to perceive the contrast. So the second‘obvious’ commonsense explanation was equallyproblematic. The correct explanation was providedby Marcy Macken, who demonstrated that there wasa perceptual problem, but not between puzzle andpuddle, but rather between puddle and puggle. Ofcourse, puggle is not a word of English, so I had failedto observe relevant examples. Words like riddle andwriggle provide a straightforward minimal pair, butthey had not been in my son’s vocabulary. Fortunate-ly, Macken observed that other examples made thecase as well as the (missing) minimal pair did. Wordssuch as pickle were intermittently pronounced ‘pittle’([pit?l]) suggesting that there was indeed perceptualconfusion. The puzzle puzzle could only be solvedwhen the difference between a variety of other exam-ples was simultaneously taken into account.

I have gone on about this example at such lengthbecause it illustrates the beauty of being (potentially)wrong. The most crucial part of the scientific enter-prise is that it makes testable (or ‘refutable’ or‘falsifiable’) predictions. Because my son regularlydistinguished puddle and puzzle, and similar exam-ples, I had claimed explicitly that he had no perceptu-al problem. Macken showed that I was wrong and, onthe basis of my own data, showed how I was wrong,leading to an improvement in our general under-standing of language acquisition, and the languagefaculty more generally. Such falsifiability is pervasivein linguistics as in all the sciences, and suggests thatmany, perhaps all, our hypotheses and principles willbe in need of revision when we get a better under-standing of what is going on. It follows that bindingtheory, which I have appealed to above, is probablywrong, and will need replacing by some more sophis-ticated theory in due course. Again this is to be wel-comed, though we must simultaneously guard againstthe danger of ‘naive falsificationism’. There are al-ways going to be contrary data that one’s currenttheory cannot explain. This is not a reason for simplyjettisoning the theory and whatever insights it mayprovide, but a point of departure for refinement andextension. A clear example is provided by the theoryof parametric variation, and the striking revision ofhis earlier work in Chomsky’s current MinimalistProgram (1995).

I have suggested that, like all principles of gram-mar, binding theory should be universal. But there areproblems. Even though (virtually) all languages havereflexives, their distribution is subject to slightly dif-ferent conditions in different languages. Consider

again the contrast between (40), The nurse showedthe woman some documents about herself, and (41),The queen said the princess had disgraced herself,where the former is ambiguous but the latter is unam-biguous. The contrast was attributed to the fact that(in English) the antecedent of a reflexive must be inthe same clause. So far so good, but if one takesequivalent examples in Chinese, it turns out that theequivalent of (40) is unambiguous, and the equivalentof (41) is ambiguous. The theory would appear to havebeen refuted: a prediction was made, it was tested, andfound to be false. But simply giving up the theorywould be defeatist, and it would also mean giving upthe explanation for the data it does account for.

The solution is interesting: the universality of bind-ing theory (and likewise for other subtheories of thegrammar) is maintained, but some latitude is allowedin the definitions involved – they are ‘parametrized’,as the jargon has it. In this case, all reflexives have tohave an antecedent, but language learners have tochoose (on the basis of the data they are exposed to)among several other options: whether they arelearning a language in which that antecedent has toappear in the same clause or in some other well-defined domain; whether the antecedent has to be asubject or can bear other grammatical relations, andothers. In Chinese, the antecedent of a reflexive mustbe a subject, so (40) is unambiguous; on the otherhand, that antecedent does not have to be in the sameclause, so (41) is ambiguous. If you are worried thatthis is too simple a get-out, an analogy with incestmay be helpful: all cultures appear to have an incesttaboo forbidding sexual relations between relatives(for instance, fathers and their daughters). Thetaboo is universal. But how that taboo is instantiatedis culture-specific: for example, some groups allowcousins to marry, others do not. The situation withregard to language and language-learning is some-what more complex than the cultural example, be-cause there are many more choices to be made. Theacquisitional task is more complex than it would havebeen if all languages were exactly like English, but it isnot as severe as one might fear. The idea is that the fullrange of parametric choices in language is availableto the child prior to experience – they are in somesense innate – and the child’s task reduces to choosingfrom a set of language structures options it already‘knows.’

Beyond Language: Pragmatics and theLanguage of Thought

We have looked at a wide range of examples illustrat-ing some of our knowledge of phonology, morpholo-gy, semantics, and (mainly) syntax, but we also have


knowledge that goes beyond words and sentences.Consider (56a) and (56b): as a remark about Fred,(56a) is fine, with stress on bats as indicated by thebold print, but as a reply to the question in (56b) it isanomalous:

(56a)
Fred has written a book about bats
(56b)
Who has written a book about bats?
Such discoursal knowledge must be distinguishedboth from syntactic knowledge of the kind that tellsus that (57) is ungrammatical:

(57)
Fred has written about bats a book
and from real world knowledge of the kind thatprompts our scepticism about (58a) and (58b):

(58a)
Bananas have legs (58b) Your saucer is being aggressive again
Someone who utters (56a) in response to (56b) prob-ably needs remedial English lessons; someone whoutters either of the sentences (58a) or (58b) is eithera linguist or in need of psychiatric help.

This brings us into the realm of pragmatics, ourinterpretation of utterances in context, and to therelation of language to thought. The examples in(58a) and (58b) are felt to be odd not because ofour linguistic knowledge – you get the same effectwhatever language you translate them into – but be-cause we know that the world isn’t like that. It isour encyclopedic knowledge that tells us this, notknowledge of our language (English). However,when we interpret someone’s utterances in some con-text, we habitually use both our knowledge of English(or whatever other language we are using) and ourencyclopedic knowledge. Suppose you hear (3a) Tomis a problem. Your knowledge of English vocabularyand grammar provides you with a meaning for thesentence, but it doesn’t tell you enough to act. Is yourinterlocutor looking for sympathy, asking you to dosomething about it, hoping for a denial? Any or all ofthese may be what you decide is the case on a partic-ular occasion, but you carry out this construal on thebasis of your knowledge of the speaker, of Tom, ofyour past exchanges with both of them, and so on,indefinitely. The core notion involved is what is‘relevant’, an idea that has been made explicit inRelevance Theory, an important extension of linguis-tics. We are now beyond the language faculty and canhand over responsibility to other disciplines; butone final question needs to be addressed: What islanguage for?

There are two standard answers: for thought andfor communication. Both answers are true, but bothneed a little hedging. First, we can obviously commu-nicate without using language by means of coughs,

sniffs, gestures, and so on. But language is far moresubtle than any other system known: conveying spe-cific negative or conditional propositions by means ofgestures or sniffing is not obviously possible. Innu-merable other creatures have complex communica-tion systems, but none of them, as far as we know,have anything with the recursive power of humansyntax (see Sperber and Wilson, 1995; Hauser et al.,2002). Second, the system we use to think with musthave a great deal in common with the natural lan-guages we speak, but it is not identical to them. Thelanguage of thought can include elements that naturallanguages cannot – visual images, for instance; andnatural languages have properties that would be un-necessary, or even unhelpful, in the language ofthought – pronouns, for instance. If I tell you thatshe is beautiful, it’s of no use to you storing that inmemory as ‘she’ is beautiful; it has to be stored with aname or some other description replacing she. None-theless, language has a central role in each of thesedomains, linking perception and articulation on theone hand to thought processes on the other. Thismeans that the output of our language faculty mustbe ‘legible’ to these other systems. Language acts as acode linking representations of sound to representa-tions of meaning. These representations must then bein a form that makes it possible for the sensorimotorapparatus to convert them into pronunciations andpercepts, and for the conceptual system to use themfor thinking, especially inference.

So, linguistics provides an account of each ofsyntax, phonology, morphology, and semantics, andhow they relate to each other; pragmatics then tells ushow such purely linguistic representations relate tothe language of thought – the medium in which wethink and carry out inferencing. This relation under-lies our ability to interpret the world and the people init, but the linguistic component is only the first stepon the journey. We normally take someone who utters‘torture is immoral’ to believe that torture is immoral,and we expect to be able to predict (at least some of)their actions on the basis of this. But people may lie,and about that linguistics has nothing to say.

See also: Data and Evidence; Language of Thought; Lin-

guistic Anthropology; Phonetics: Overview; Phonology:

Overview; Pragmatics: Overview; Principles and Para-

meters Framework of Generative Grammar; Psycholin-

guistics: History; Syntax-Semantics Interface.

Bibliography

Baker M C (2001a). ‘The natures of nonconfigurationality.’In Baltin M & Collins C (eds.) The handbook of contem-porary syntactic theory. Oxford: Blackwell. 407–438.

History of Linguistics in Central and South America 355

Baker M C (2001b). The atoms of language: the mind’shidden rules of grammar. Oxford: Oxford UniversityPress.

Bauby J-D (1997). The diving-bell and the butterfly.London: Fourth Estate.

Bloomfield L (1935). Language. London: Allen & Unwin.Carston R (2002). Thoughts and utterances: the pragmatics

of explicit communication. Oxford: Blackwell.Chierchia G & McConnell-Ginet S (2000). Meaning and

grammar: an introduction to semantics2. Cambridge,MA: MIT Press.

Chomsky N (1957). Syntactic structures. The Hague:Mouton.

Chomsky N (1965). Aspects of the theory of syntax.Cambridge, MA: MIT Press.

Chomsky N (1994). ‘Chomsky, Noam.’ In Guttenplan S(ed.) A companion to the philosophy of mind. Oxford:Blackwell. 153–167.

Chomsky N (1995). The minimalist program. Cambridge,MA: MIT Press.

Fabb N (2002). Language and literary structure: the linguis-tic analysis of form in verse and narrative. Cambridge,UK: Cambridge University Press.

Fromkin V (ed.) (2000). Linguistics: an introduction tolinguistic theory. Oxford: Blackwell.

Fromkin V, Rodman R & Hyams N (2002). An introduc-tion to language.

Gussenhoven C (2002). Phonology: analysis and theory.Cambridge, UK: Cambridge University Press.

Hauser M, Chomsky N & TecumsehFitch W (2002). ‘Thefaculty of language: what is it, who has it, and how did itevolve?’ Science 298 (Nov. 22), 1569–1579.

Heim I & Kratzer A (1998). Semantics in generative gram-mar. Oxford: Blackwell.

History of Linguistics in Central anO Zwartjes, Universiteit van Amsterdam, Amsterdam,The Netherlands

� 2006 Elsevier Ltd. All rights reserved.

Spanish and Portuguese missionaries dedicated them-selves to the study of non-Indo-European languages,resulting in the publication of a great number of worksin or about the Amerindian languages. Although mis-sionaries based and structured their works in agree-ment with the traditional Greco-Latin grammaticalframework, they were forced by the linguistic factsto search for unconventional solutions in theirdescriptions, which never had been studied before inEurope. There is no doubt that their approaches werein many cases original and creative on all subdisci-plines of grammar: phonology; orthography (the ren-dering of sounds never heard or written before);

Huddleston R & Pullum G K (2004). The Cambridge gram-mar of English. Cambridge, UK: Cambridge UniversityPress.

Hudson R A (1990). English word grammar. Oxford, UK:Blackwell.

Hudson R A (1996). Sociolinguistics (2nd edn.).Cambridge, UK: Cambridge University Press.

Jackendoff R (2002). Foundations of language: brain,meaning, grammar, evolution. Oxford, UK: OxfordUniversity Press.

Law A (2004). ‘Sentence-final focus particles in Cantonese.’Ph.D. thesis, University College London.

Macken M (1980). ‘The child’s lexical representation: ThePuzzle-Puddle-Pickle evidence.’ Journal of Linguistics16, 1–17.

McGilvray J (1999). Chomsky: language, mind, andpolitics. Cambridge, UK: Polity Press.

Neeleman A & Weerman F (1997). ‘L1 and L2 word orderacquisition.’ Language acquisition 6, 125–170.

Radford A (2004). English syntax: an introduction.Cambridge, UK: Cambridge University Press.

Smith N V (1973). The acquisition of phonology: a casestudy. Cambridge, UK: Cambridge University Press.

Smith N V (1989). The Twitter Machine: reflections onlanguage. Oxford, UK: Blackwell.

Smith N V (1999/2004). Chomsky: ideas and ideals.Cambridge, UK: Cambridge University Press.

Smith N V (2002). Language, bananas and bonobos:linguistic problems, puzzles and polemics. Oxford:Blackwell.

Sperber D & Wilson D (1995). Relevance: communicationand cognition. Oxford: Blackwell.

d South Americamorphology–morphosyntax (their model predatedours, not yet elaborated, used to describe phenomenasuch as agglutination, incorporation, different wordstructure, processes of derivation, and compositionin typologically different languages); syntax (e.g.,topicalization, ergativity, different word order, thedichotomy active and nonactive, incorporation); andlexicography – another challenge for the missionariesand an important tool for the translation of religioustexts.

Missionary works are important sources for othersubdisciplines of linguistics, such as sociolinguistics(language variety, language change, standardization),pragmatics, translation theories, cultural anthropol-ogy; for reasons of space these will not be treated inthis article. Descriptive linguistics always requires anunderlying theoretical framework and reflects the lin-guistic thinking of the period in question. Interest in

encyclopedia of language & linguistics || history of linguistics: discipline of linguistics

Documents