Integrational linguistics and the structuralist legacy

Integrational linguistics and thestructuralist legacy

Roy Harris2 Paddox Close, Oxford OX2 7LR, UK

I. Introduction

Integrationism is a product of the structuralist legacy in modern linguistics. Thatlegacy in turn has in¯uenced current perceptions of integrationism, and is itselfcomplex. Structuralism in what is sometimes called the ``broad'' sense (Crystal,1992, p. 370) is usually seen as a European phenomenon developing from the workof Saussure. Structuralism in the ``narrow'' sense is seen as an American phenom-enon associated particularly with the work of Bloom®eld and his followers: it is saidto be ``characterized by a general behaviouristic attitude and a rather restrictiveconception of scienti®c method, inherited from neopositivism and based on thenotion of veri®ability'' (Lepschy, 1970, p.110). These two structuralisms are aboutas close to each other as Chicago is to Geneva. In spite of super®cial resemblances(in, for example, some areas of terminology or the analysis of particular examples)they have little but the name in common. As is evident from his review of Saussure'sCours (Bloom®eld, 1923), Bloom®eld never grasped either the theoretical basis orthe originality of Saussure's position. Bloom®eld's conversion to behaviourismserved only to widen the gulf separating his own from Saussure's view of language.There could be no clearer testimony to this than that provided by Bloom®eld'ssemantics, which appeals to laboratories and chemical science to establish themeaning of the word salt (Bloom®eld, 1935, p. 139). This is not structuralism in theEuropean sense. At the very least, a Saussurean would say, it confuses faits de languewith faits de parole. It seems fairly certain that Saussure would no more have dreamtof endorsing the Bloom®eldian account of linguistic meaning than of lecturing at theUniversity of Geneva in his nightshirt.Each structuralism left its own kind of legacy in the history of linguistics. Inte-

grationism (Wolf and Love, 1993; Harris, 1998; Harris and Wolf, 1998) is often seenas being an essentially neo-Saussurean enterprise. More surprisingly, however, it hasalso been seen as a reversion to Bloom®eldian behaviourism. This paper commentson some aspects of those perceptions. In so doing it takes up and supplements somebriefer observations already published in ``Saussure, generative grammar andintegrational linguistics'' (Harris, 1995).

LANGUAGE&

COMMUNICATION

Language & Communication 19 (1999) 45±68

0271-5309/98/$Ðsee frontmatter#1998DrRoyHarris. PublishedbyElsevierScienceLtd.All rights reserved.

PII: S0271-5309(98)00017-2

2. Integrationism and the history of linguistics

Since integrationists treat the importance of context as an article of faith, let us®rst try to contextualize the question `Ìs integrational linguistics neo-Saussurean?''It is a question most likely to arise, no doubt, in discussing the history of modernlinguistics; and here straight away a problem arises. It is unfortunately the case thatfor many students the history of modern linguistics reduces to a list of ``big names''of which Saussure's is undoubtedly among the ``biggest'') and an associated list of-isms (beginning with comparativism and including structuralism, functionalism, gen-erativism, and so on). On this basis. one can draw neat diagrams which supposedlydisplay the relationships and `ìn¯uences'' that have structured the narrative that theomniscient historian of linguistics tells. Thus if the answer to our question turns outto be an unquali®ed ``yes'', integrationism might be slotted into such a frameworkvia a genealogical line that descends from the belly of the name ``Saussure'' and thusmakes integrationists ®rst cousins of linguists engaged in otherÐquite di�erentÐneo-Saussurean enterprises.And this is odd. For it is di�cult to imagine an integrationist feeling any strong

temptation to engage in the kind of formal analysis that has preoccupied themajority of neo-Saussureans (and the majority of structuralists) for most of thetwentieth century. This is not because integrationists feel that this kind of analysishas already been ``done'' and therefore does not need doing again. Rather, it isbecause of a feeling that doing that kind of analysis at best distracted linguists frommore important questions, and at worst diverted academic linguistic inquiry intodead ends from which it has yet to be extricated.So my ®rst point is that this would be a deplorable way to approach the history of

linguistics (although it is an approach that is commonly encountered not only in theteaching of the history of linguistics but in the teaching of many other branches ofthe history of ideas). If the question `Ìs integrational linguistics neo-Saussurean ?'' israised in this kind of context, the only appropriate integrationist response seems tome to be to reject the question altogether.It is undeniable, however, that integrationists do often discuss Saussure, state their

position by reference to Saussure, and have even contributed in some measure to amodern revaluation of the work of Saussure (Harris, 1983, 1987; Komatsu andHarris, 1993; Komatsu and Wolf, 1996, 1997). Does not all that su�ce to qualifythem as paid-up neo-Saussureans? And is not the perception that they are neo-Saussureans con®rmed by what integrationists themselves have said? For example.on p.319 of volume 1 of the Proceedings of the XVth International Congress of Lin-guistics (Wolf and Love, 1993) we read that the ``goal of an integrational linguisticsis [...] one which Saussure failed to achieve''. The goal in question is identi®edby reference to the famous statement about linguistic realities on p. 128 of the Coursde linguistique geÂneÂrale (Saussure, 1922). `Ìn order to determine to what extentsomething is a reality, it is necessary and also su�cient to ®nd out to what extent itexists as far as the language users are concerned.'' In their Proceedings paper Wolfand Love contrast this position with the more recent generativist view that muchlinguistic knowledge is `ìnaccessible to consciousness''. Thus the integrationist

46 R. Harris/Language & Communication 19 (1999) 45±68

approach is overtly presented as ``reinstating'' (the term they use) the importance of``speakers views about the form in which they communicate''.Now such an account of the matter might appear to have quite clear historical

implications: viz. that integrationism is in some sense or in some respects a rever-sion to an earlier Saussurean view of linguistic realities. That interpretation, it seemsto me, would be quite mistaken for various reasons. In the ®rst place, because of theambivalence of the phrase ``speakers views''. There has been much debate in modernlinguistics about the status of linguistic `ìntuitions''Ðan unfortunate and mislead-ing term if ever there was oneÐand this debate cannot be continued here. But itshould be noted that Saussure nowhere advocates that the linguist should go roundasking informants for their viewsÐfor example, about what they think such-and-such a word means or how it should be used or pronounced. Whereas that is pre-cisely what integrationists are prepared to do and have done in particular instances(e.g. Wolf et al., 1996; Davis, 1997).In the second place, it is extremely doubtful whether what Saussure regarded as

linguistic realities are linguistic realities for the integrationist. The Saussurean state-ment about linguistic realities that Wolf and Love quote comes from a passage ofwhich the main purpose is to make clear the di�erence between synchronic anddiachronic linguistics. The linguistic realities in question are ``synchronic'' realities.Now integrationists do not accept the synchronic-diachronic dichotomy, which isone of the foundational Saussurean distinctions. So it is hard to see how they are inany theoretical position to endorse or reinstate the version of psychological realismthat Saussure was advocating.In the third place, Saussure's linguistic realities are, faits de langue, not faits de

parole. And again the langue±parole dichotomyÐat least, as drawn by SaussureÐisnot one that integrationists can accept.Thus as soon as we begin to probe the comparison between Saussure's focus on

the language user and the integrationists insistence on the lay-orientation of anyviable linguistics, the similarity quickly starts to vanish before our eyes. All we areleft with is a kind of disembodied analogy, a theoretical counterpart to the grin ofthe Cheshire cat.

3. Encompassing the language myth

A somewhat di�erent interpretation of the relationship between integrationism andstructuralism is captured in Nigel Love's pithy phrase ``transcending Saussure''(Love, 1989). Love wrote: ``The Saussurean language myth is not so much to berejected as encompassed and transcended'' (Love, 1989, p.817). A pedant might per-haps fasten on the hedging formulation ``not so much to be rejected as. . .'' and say:``This won't do: is to be rejected or not?'' And another question that comes to mind iswhether it is possible to transcend something and subvert it at the same time.Love, fortunately, has since given us a clear illustration of how he understands the

`èncompassing and transcending'' operations. This is in his paper ``The locus oflanguages in a rede®ned linguistics'' (Love, 1990). Although Love does not explicitly

R. Harris/Language & Communication 19 (1999) 45±68 47

say there that he is showing us how to encompass and transcend the language myth,I shall take it that this is in fact what he is about; and I shall draw from this examplesome conclusions about the structuralist legacy in general.What Love, it seems to me, accepts as a lay linguistic reality is that in many

communities people do regard themselves as speakers and/or writers of a particularlanguage, as witnessed e.g. by the fact that they have no hesitation in identifying thatlanguage by name (as e.g. `Ènglish'', ``French'', etc.). To this extent their behaviourmight be regarded as con®rming the Saussurean doctrine that the world is full oflanguages (langues in the plural). Furthermore, they mightÐor might notÐentertainabout the language (sc. the language they believe themselves to be users of) beliefswhich correspond to Saussure's version of the language myth: e.g. that it provides a®xed code which enables them to exchange their thoughts with other members of thesame linguistic community. If they are educated Europeans it is more than likelythat they will indeed entertain such beliefs, because that is roughly what their edu-cation will have conditioned them to believe. Perhaps they will be less likely toendorse a Saussurean view of the relationship between speech and writing: theymight take a thoroughly unSaussurean view and believe that the written form oftheir language is somehow superior to spoken forms, or at least provides a standardby reference to which ``correctness'' is to be measured. In other words, the lay viewof languages may not on all points run exactly parallel to canonical Saussureandoctrine, but at least there is no crucial con¯ict on issues concerning the existence of a®xed code and its communicational function. To the extent that this is so, for at leastsome linguistic communities, and to the extent that the integrationist recognizes it tobe so, the integrationist account may be said ``not so much to reject as to encompass''the language myth. How about transcending it ? We will come to that later.First of all, let us not overlook the important point that `èncompassing'' the lan-

guage myth in this sense does not constitute either an endorsement or a developmentof Saussure's theoretical position on langue and parole. The language myth is beingtaken on board not as a profound theoretical insight but as an approximate de factodescription of how lay language users see their own linguistic situation. And in thatsense Saussure is both right and irrelevant: that is to say, the only reason whySaussure enters into the picture at all is that he put forward a theory that, in certainbasic respects, captures what the integrationist recognizes as an important complexof lay linguistic beliefs. That is why SaussureÐand not e.g. Mao Tse-Tung or Mar-ilyn MonroeÐfeatures in the story.In the second place, my caveat ``to the extent that the integrationist recognizes''

this congruence between Saussure's language myth and a complex of lay beliefs wasmeant to allow for the consideration that in any case that congruence involves acertain oversimpli®cation. Saussure's language myth applies, strictly speaking, onlyat the theoretical level of the idiosynchronic system; whereas it is doubtful that manylay language-users think of their ``language'' as an idiosynchronic system at all, andif they do they are perhaps unlikely to have a name for it. (The kind of English wespeak and write at 2 Paddox Close, Oxford, may well be recognized by us and othersas idiosyncratic, but it is not recognized as a separate language called ``Harrisian'',either by the denizens of 2 Paddox Close or by anyone else).


Third, `èncompassing'' the language myth as a lay linguistic reality does not entailaccepting the position that some theorists have argued for, according to which alanguage is simply whatÐand whateverÐits users collectively think it is. (The casefor one version of this view is put forward with admirable clarity in Pateman 1983.but it seems to me a case that the integrationist must reject. See below.)Fourth, none of this commits the integrationist to the assumption that the lan-

guage myth is universal, i.e. the foundation of all lay linguistic beliefs in all com-munities. It so happens that Western linguistics is a discipline of which, historically,structuralism is an important phase. But history might have taken a di�erent path.Unless these four provisos are clearly borne in mind the notion that integration-

ism `èncompasses'' the Saussurean language myth seems open to all kinds ofpotential misconceptions of a more or less ethnocentric kind.

4. A lay-oriented linguistics

Before proceeding further it will be as well to clear up a point already mentionedabove concerning the language-user's own view of the language. This is relevant tothe question of the structuralist legacy for at least two reasons. One is that for theintegrationist a viable linguistics must be `èssentially lay-oriented'' (Harris, 1981,p. 80). The other is that it was Saussure who ®rst introduced into linguistic theorythe notion that linguistic description should attempt to capture the language-user'sperspective. This objective was later referred to as ``psychological reality'', althoughthat is not a term Saussure ever used.The issue of psychological reality became controversial when it was linked to the

question of evaluating what was called the `àdequacy'' of linguistic descriptions.Tests were made to try to determine whether certain ``rules'' postulated by (generative)linguists were actually used by language-users in their production or comprehensionof utterances. There was much argument about ``real time'' linguistic computationgoing on in speakers' brains, etc. Something akin to this is still a live issue in thecomputer-based investigation of natural language processing, where connectionistshave questioned whether language `ìs rule governed as well as rule described''(Cottrell, 1994). But the point of mentioning it here is to recall that Saussure'sposition was (as far as it goes) quite unequivocal. The Cours tells us explicitly that

the sum total of deliberate, systematic classi®cations set up by a grammarianstudying a given linguistic state a-historically must coincide with the sum totalof associations, conscious or unconscious, operative in speech (Saussure, 1922,p. 189).

This is the passage that has to be cited in order to rebut any claim that Saussure wasever a ``hocus-pocus'' linguist (Crystal, 1991, p. 166). And it has to be realized that inthis passage ``speech'' is a structuralist technical term (i.e. la parole). Here, it mightperhaps be said, we see Saussure coming as close as he ever did to adopting (avant lalettre) an integrationist position. But `às close as he ever did'' is not all that close.


Why not? Because for Saussure la langue is still a bi-planar structure existingindependently of the circumstances of its application. For Saussure, no individual(s),however eminent, control(s) what is meant by what is said. Meaning belongs to thesystem, not to the user.But there is also a more far-reaching reason which becomes apparent if we turn the

exegetical spotlight on the phrase ``. . .associations, conscious or unconscious. . .''. It isdi�cult to make sense of this formulation unless we assume that Saussure takes itfor granted that speakers, as agents of the speech act (la parole), do not alwaysunderstand what they are up to when they speak (just as hens do not grasp what isgoing on when they lay eggs). Perhaps their bodies or their brains ``know'' but``they'' don't (or may not).So this is rather like arguing about whether cyclists (or their brains) have some-

how `ìnternalized'' (enough of) the mechanics of equilibrium to enable them to rideon two wheels without falling o�. Well, in one (rather trivial) sense, they must havedone, or else they would not be able to balance on the saddle long enough to getthem from here to the corner of the street. And since linguists always assume theequivalent of cycling ability (i.e. that people can actually speak their own language)this is not, to say the least, a very compelling line of argument. Worse still, if (some)connectionists are right, it may well be the case that Saussure's requirement is not somuch di�cult to ful®l as meaningless.But it does not deserve on that account to be con¯ated with anotherÐbroaderÐ

issue, which Saussure, to give him credit where credit is due, wisely avoided. Itconcerns a substantive matter on which, for a variety of reasons, the pass has sincebeen sold by the academic guardians of such disciplines as linguistics, philosophy,psychology, etc. This is the whole question of construing a linguistic ability asknowledge. (Cf. ``Does a hen know how to lay eggs?'', ``Do cyclists know how tocycle?'', etc.) When we look again at the passage from Saussure cited above (Saus-sure, 1922, p. 189), we see that it carefully skirts the issue of what language-usershave to know i.e. in order to use their language (langue), to engage in parole. Andthere is a very good structuralist motive for this evasion. It hinges on the importanceof Saussure's proviso `òperative in speech'' (mise en jeu dans la parole). For at thelevel of speech, it is impossible to draw any clear line between items of knowledgethat are ``linguistic'' and items of knowledge that are ``non-linguistic''. (Is knowingthe name of the present British Prime MinisterÐand thus being able to refer to himby name, as in saying ``Tony Blair is going to reshu�e his cabinet''Ðan item oflinguistic or of non-linguistic knowledge? We cannot, other than quite arbitrarily,analyse particular utterances in terms of two discrete categories of knowledge theydisplay.) Saussure, showing far greater philosophical sophistication than many ofhis successors, seems to have realized this, and therefore took care not to talk aboutla langue as a matter of what speakers know.Later linguists fell into exactly this trap. They then had to ®nd escape routes from

it, of which perhaps the most notorious and implausible is the Chomskyan invention(Chomsky, 1986, pp. 265 �.) of a metalinguistic verb ``to cognize'' (= ``know in thespecial sense of knowing a language as construed by Chomsky''). This Byzantinemanoeuvre outstrips all competitors. (One wonders whether, had the more general


question arisen of translating abilities into forms of knowledge, we would haveheard a similar account of how the competent cyclist cognizes riding a bicycle.)Metalinguistic dodges such as these are part of the structuralist legacy in the sense

that they are clearly motivated by a desire to hitch structuralism to the cognitivebandwagon, i.e. to validate it as ``psychologistic structuralism'' (Love, 1984), just asBloom®eld had attempted to hitch his structuralism to the currently fashionable beha-viourist bandwagon. Whether Saussure would have approved of any of these laterattempts to make linguistics a psychologically ``realistic'' discipline is another matter.The problem with this aspect of the legacy from an integrationist point of view is that itis di�cult to see how it can be pursued very far on an empirical basis without turninglinguistics into what integrationists would regard as a study of the biomechanicalunderpinnings of language rather than a study of linguistic communication.

5. Languages as objects of belief

Another part of the structuralist legacy is what I will call the ``collectivist'' con-ception of languages. Its rationale, put very crudely, runs as follows. If Saussure wasright to insist that the criterion for accurate (synchronic) linguistic description isconformity to what the language-user recognizes, consciously or unconsciously, asconstituting structural linguistic relations, thenÐat the end of the day and on thebroadest social scale of interpretationÐa language just is what its users recognize assuch. So, to put it even more crudely, a language (say `Ènglish'') is de®ned by whatself-recognizing speakers of that language recognize it as being. They do not recog-nize themselves as speaking e.g. ``French''; they do not recognize e.g. ``chien'' as anEnglish word; they do not recognize nouns like ``knife'' as having a grammaticalgender in the sense of requiring a special form of the de®nite article; etc. But they dorecognize their language as being called `Ènglish''; they do recognize it as having aword ``dog'' for a certain kind of domesticated quadruped; they do recognize it ashaving a sibilant su�x for forming plurals, etc. In short, collectivist linguistics positsthat English is what the collectivity of its users recognize as English.Circular as it may sound when stated thus baldly, this is a linguistic doctrine

which in practice has close links to the notion that languages have specially privi-leged users called ``native speakers''. It also has close links to the notion that com-petent speakers of a given language may describe this language accurately on thebasis of their own introspections about what they do as speakers. These, I wouldsuggest, are notions that are quite fundamental to the development of structuralism.Fundamental in the sense that if they had been seriously questioned nothing like thediscipline now recognized in Western culture as modern linguistics could ever havebeen put in place.Pateman (1983) contrasts what I am calling the ``collectivist'' view with four other

possible answers to the question ``What is a language?''. One answer is that a lan-guage is a natural kind (``naturalism''). Another is that a language is an abstractobject (``Platonism''). A third is that a language is a name given to a set of linguisticobjects (``nominalism''). The fourth is that a language is a social fact, which is also a


(or even the only) linguistic fact (``sociologism''). Whereas the answer Patemanfavours says that although a language is a social fact, that social fact is not itself alinguistic fact. What this involves, spelt out more formally, is the following (whereEnglish is taken as a sample language):

1. S believes, of the language S speaks, that it is English; and2. S1 believes, of the language S1, speaks, that it is English; if and only if3. S believes S1, is a speaker of English,4. S1, believes S is a speaker of English,5. S believes S1, believes S is a speaker of English,6. S1 believes S believes S1 is a speaker of English.

Put less formally, in Pateman's words, ``when we take individuals collectively it isthey who de®ne what English is, and this they can do consequent upon believing ofthemselves and each other that they are speakers of English.'' Accordingly,

the underlying reality of the English language as a socio-political fact is itsappearance as the intentionally inexistent object of speakers' mutual beliefs andit is its place in a ``package'' of mutual beliefs which sustains any individualspeaker's belief in his or her speakerhood (Pateman, 1983, p. 120).

By an `ìntentionally inexistent object of belief'' Pateman says he means simplythat such an object may not exist as anything other than an object of belief (cf. wit-ches).Now although integrationists are no less cautious than Pateman about the `èxis-

tence'' of languages like English, this account of languages as ``socio-political facts''is not one an integrationist would rush to endorse. For it begs the question by pre-supposing the infallibility of the collectivity. This objection has nothing to do withlanguage as such: it is based on the unacceptability of taking the beliefs of any groupabout themselves as self-validating. To develop Pateman's own example, considerthe claims of a sect whose members are convinced that they are all witches. Thesincerity of the beliefs is not the issue. Many communities have doubtless in the pastentertained false beliefs about themselves, and perhaps many still do. The point isthat we usually require something more, against which beliefs can be measured.That is why examples of speech produced by those who claim to be speakers ofEnglish cannot be held to demonstrate evidentially the validity of that belief unlesssome independent criterion of what English is can be made available. Failing that,the argument ``There must be an English language because we speak it'' reducesupon examination to ``There must be an English language because that is what wecall what we speak''. And this sounds suspiciously like the position Pateman rejectsas ``nominalism''.An integrationist would also point out that the inexistent object of belief that

Pateman describes does not seem to match up at all with any ``socio-political'' factthat our everyday experience would tempt us to call `Ènglish''. For many self-pro-claimed speakers of English patently do not agree with one another about the cri-teria for speaking it, or about one another's views on the topic. Curiously, Pateman


claims that one of the advantages of his account is that it allows for disagreementbetween individual speakers as to what English is (what is grammatical in English,etc.). But it doesn't. For if what they count as English di�ers, it is incoherent tomaintain that they all agree that they all speak it. In other words, Pateman'scondition (2) is violated. (Cf. the witches. If it turns out that individual witches donot agree upon what a witch is, it is mere equivocation to say they all agree they arewitches.)

6. Restructuring Saussure

Saussure would certainly not have approved the theoretical basis of integration-ism. But nor would he have endorsed some positions which are more recognizablethan integrationism as developments of structuralism. In this connexion it is inter-esting to compare integrationism with glossematics. Hjelmslev's major contributionto European structuralism was to take Saussure's distinction between form andsubstance as seriously as Saussure had said it should be taken. (This was a distinc-tion Bloom®eld and his followers never grasped in the ®rst place. The nearestBloom®eldian approach to Saussurean ``form'' was a set of ``habits''. For some ofBloom®eld's followers it was no more than extrapolation from distributional rela-tions.) For Hjelmslev, however, the Saussurean dictum la langue est une forme et nonune substance held the key to a theory of the linguistic sign of which Saussure himselfhad never succeeded in giving a full or entirely successful explication. (The clearestdiscussionÐclearer than Hjelmslev'sÐis to be found in Fischer-Jùrgensen, 1966).Readers of the Cours are told that it is the failure to understand the formal nature

of la langue which is responsible for all or most of our misconceptions concerning it.

The importance of this truth cannot be overemphasized. For all our mistakes ofterminology, all our incorrect ways of designating things belonging to la langueoriginate in our unwittingly supposing that we are dealing with a substancewhen we deal with linguistic phenomena (Saussure, 1922, p. 169).

Hjelmslev seems to have taken this warning even more seriously than Saussurehimself did. For Saussure does not trouble to spell out with any care exactly how weare to understand the distinction between form and substance. What he does, as Ihave pointed out elsewhere, is to supply analogies which in fact complicate the issuebecause they are not mutually consistent (Harris, 1987, pp.118�.). Hjelmslevundoubtedly saw (and rightly) that this was quite unsatisfactory from a theoreticalpoint of view. and therefore undertook to explicate Saussure's distinction for him.Now this may be one way of ``transcending Saussure'', but it is not the way inwhich integrationism transcends Saussure. From an integrational perspective, whatHjelmslev does is not so much transcend Saussure as attempt to restructure him.This well-intentioned restructuring operation turns out to be something of a dis-

aster. Hjelmslev, obsessed with the idea of linguistic description as a ``calculus'', triesto assimilate the distinction between form and substance to the (mathematical)


logician's distinction between constant and variable. Accordingly, he contrasts``form'' as ``the constant in a manifestation'' with ``substance'' as ``the variable in amanifestation'' (Hjelmslev, 1961, p.134). But this is deeply problematic in at leasttwo respects. One is that the logical distinction between constant and variable isalready a distinction between symbols; and it is di�cult to see how, without invol-ving circularity or regress, such a distinction can be applied without further ado toidentifying linguistic fundamentals of the kind Saussure is concerned with. Hjelm-slev attempts to give a technical de®nition of the distinction by saying that a con-stant is a ``functive whose presence is a necessary condition for the presence of thefunctive to which it has function'', whereas a variable is one whose presence is not anecessary condition (Hjelmslev, 1961, p. 131). But this explanatory path peters outrather disappointingly when a ``functive'' is de®ned as an `òbject that has functionto other objects'' and a ``function'' as `à dependence that ful®ls the conditions foran analysis''. Which seems in the end to tell us little more than that constancy (likevariation) is something in the eye of the analyst. In glossematics such questions tendto be dodged by falling back on the metaphysical notion of `ìmmanence'', itself notfurther de®ned. We are left to ponder which of the various philosophical senses ofthe term is apposite. Hjelmslev himself opposes `ìmmanence'' to ``transcendence'',an opposition with Kantian overtones. But all he seems to mean by this is the dif-ference between studying language as a means to the pursuit of other knowledge andstudying language as an end in itself. Which takes us back to the conclusion ofSaussure's Cours and the claim that ``the only true object of study in linguistics is lalangue, considered in itself and for its own sake'' (Saussure, 1922, p. 317)Ða con-clusion for which, notoriously, Saussure himself was not responsible.Uldall explicates the distinction between form and substance with respect to the

phoneme as follows:

the in®nite variety of sounds used in actual speech are seen to belong togetherin groups in such a way that the members of one and the same group can beexchanged without any change of meaning, while the interchange of two soundsbelonging to di�erent groups may lead to a change of meaning. In other words,there is something which remains constant when you exchange within the samegroup, but which changes when you go from one group to another, and which,consequently, must be supposed to be common to all the sounds belonging toone group (Uldall, 1944 Italics added.).

But the logic underlying this explanation seems to be very questionable; i.e. it isnot at all obvious why the mutual substitutability of sounds salva signi®cationeshould lead anyone to the conclusion that all the sounds concerned must havesomething in common (other thanÐcircularlyÐthe property of substitutability salvasigni®catione). But then it seems that a ``constant'' has been conjured up by simplyreifying one arbitrarily selected property.None of the above supplies a very cogent case for the distinction between form and

substance; while in practice it seems to leave the way open for individual linguists torecognize ``samenesses'' of all kinds and in all kinds of mutually incompatible ways.


And avoiding this free-for-all is exactly what a ``science'' of language was suppo-sedly designed to ensure.So if glossematics is taken as one way of transcending Saussure, it might be con-

strued as recognizing that you may as well be hung for a sheep as a lamb. Saussure'sdictum about the point of view preceding the object gets translated by Hjelmslev,(1961, p. 76) into the `ìmpossibility'' of taking `à description of substance as thebasis for the description of a language''. (Which is exactly what structuralism of theBloom®eldian variety assumes to be a necessary way of proceeding; i.e. ®rst collectyour ``data''. When one camp asserts to be impossible what a rival camp asserts tobe necessary, it becomes futile to pretend that all ``structuralists'' are singing to thesame theoretical tune.)Putting the priorities beyond any shadow of doubt, Hjelmslev writes:

On the contrary, the description of substance depends on the description of thelinguistic form. The old dream of a universal phonetic system and a universalcontent system (system of concepts) cannot therefore he realized [. . .] (Hjelm-slev, 1961, pp. 76±77).

Having thus insisted on the priority of linguistic form (cf. the generativist insistencethat the study of ``competence'' takes precedence over the study of ``performance''),Hjelmslev goes on to take the theoretical step that Saussure never took and never wouldhave taken; that is, to treat form as structured independently of substance.

7. Writing and speech

One consequence of this move is to bring glossematics into accord with inte-grationism (and against Saussure) on a somewhat unexpected point; namely, asregards admitting the parity of writing with speech.For Hjelmslev this follows from admitting that linguistic units are formal entities

independent of their expression in speech, writing, or any other material.

Thus the system is independent of the speci®c substance in which it is expressed;a given system may be equally well expressed in any one of several substances,e.g. in writing as well as in sounds.[. . .] The fact that articulated sound is themost common means of expression is not a consequence of any particularityinherent in the system, but is due to the anatomic-physiological constitution ofman (Siertsema, 1965, pp. 111±112).

Similar views are voiced by other glossematicians. According to Uldall, forexample:

it is only through the concept of a di�erence between form and substance thatwe can explain the possibility of speech and writing existing at the same time asexpressions of one and the same language. If either of these substances, the


stream of air or the stream of ink, were an integral part of the language itself, itwould not be possible to go from one to the other without changing thelanguage.[. . .] The form, then, will remain the same even if we change the sub-stance, as long as we do not interfere with its function. When we write a pho-netic or a phonemic transcription, we substitute ink for air, but the formremains the same, because the functions of each component form have not beenchanged [. . .] (Uldall, 1944).

Uldall proceeds on this basis to set up a unit called the ``cenia'' which spansspeech and writing.

Units of expression belonging to di�erent systems unite into groups analogousto the groups of sounds which we call phonemes and de®ned by all the membersof one group being functions of the same unit of content. A group of this kind Ishall call a cenia. Thus the speech-chain kat and the written chain ``cat'' belongto the same cenia, because they can be exchanged without a change of meaning,being functions of the same unit of content. But to the same cenia will alsobelong, it will be seen, any other unit from any other system of expression, if itis a function of the same unit of content (Uldall, 1944).

Although Uldall presents all this as a development of the Saussurean distinctionbetween form and substance, it is blatantly in con¯ict with the position that Saus-sure himself took. Saussure made it quite clear that in his terminology anything thatquali®ed as a langue must be a system of articulated speech. We have a quite cate-gorical statement in the Cours:

A language and its written form constitute two separate systems of signs. Thesole reason for the existence of the latter is to represent the former (Saussure,1922, p. 45).

This cannot be reconciled with the glossematic view, according to which speechtakes no priority over writing at all. For Hjelmslev the language itself is an entirelyabstract system, which can be manifested in any appropriate material form. Thisagain is born out by Uldall, who claims:

The system of speech and the system of writing are thus only two realizations ofan in®nite number of possible systems, of which no one can be said to be morefundamental than any other (Uldall, 1944).

Mainstream twentieth-century linguistics, however, including American structur-alism, continued to maintain Saussure's view as against Hjelmslev's. The followingstatement by Robins is typical:

Every natural language is primarily a spoken medium of communication, andthe forms of written languages [. . .] are obviously controlled and are to be


understood by reference to the essential spoken nature of language (Robins,1989, p.368).

Others have gone further and maintained the irrelevance of writing to linguisticanalysis. Martinet, for example, claims that

The linguist in principle operates without regard for written forms. He takesthem into consideration only in so far as they may on occasion in¯uence theform of vocal signs (Martinet, 1964, p. 17).

The occasional cases that Martinet seems to have in mind are those of pro-nunciations in¯uenced by orthography. But even here spelling is not strictly relevantto an orthodox structuralist synchronic description of speech: for synchronicdescription is not concerned with explaining why the pronunciation is as it is.One objection that has been raised to the glossematic position is that in fact the

written language does not always correspond exactly to the spoken language: so itbecomes implausible to treat both as equivalent manifestations of exactly the sameunderlying abstract system. There are often, for instance, words that are spelledalike but pronounced di�erently. Hjelmslev acknowledges that ``not all ortho-graphies are phonetic'' (Hjelmslev, 1961, p. 104). But he claims that this is irrelevant:`ìt does not alter the general fact that a linguistic form is manifested in the givensubstance'' (Hjelmslev, 1961, p. 105). He goes on to add that

the task of the linguistic theoretician is not merely that of describing the actu-ally present expression system, but of calculating what expression systems ingeneral are possible as expression for a given content system, and vice versa.(Hjelmslev, 1961, p. 105).

From an integrational point of view, it seems that Hjelmslev was confusing twoissues. The one on which he was right (at least, within the constraints of his ownversion of structuralism) was whether, in principle, it is possible for a given abstractsystem to have materially di�erent but exactly equivalent manifestations, involving,for example, quite di�erent sensory modalities. There seems no good reason to denythis, given the glossematic theoretical framework. For instance, suppose we have awaiting room in which a light ¯ashes when it is time for the next person in the queueto go in to an inner room and be attended to. An exactly equivalent system couldreplace the light by a buzzer, thus substituting an auditory for a visual signal. Fur-thermore, there might be a signi®cant distinction between the light ¯ashing once andthe light ¯ashing twice. Perhaps it ¯ashes once for the next man in the queue andtwice for the next woman in the queue, because men and women are attended toseparately. Again, this could be replaced by a distinction between one and twobuzzes on the buzzer. And in general, for any number of such distinctions, the buz-zer will do as well as the light. This would be a paradigm case to illustrate the glos-sematic thesis that the same abstract system can be manifested in exactly equivalentways, although in substantially di�erent signals.


For an integrationist, on the other hand, there is no equivalence between the lightsystem and the buzzer system, because the activities integrated are not the same.They belong to biomechanically di�erent dimensions. A blind person would not beable to respond to the light signals; nor a deaf person to the buzzer. But even in thecase of individuals with normal sight and vision there is no question of integrationalparity between the two systems. In order that two systems should have integrationalparity it would be necessary that the activities belong to the same perceptual mod-alities in both cases and also that the same biomechanical abilities he exercisedwithin those modalities. These conditions are clearly not satis®ed in the case of thelight and the buzzer. In other words, the glossematic position is in this respect atquite the opposite end of the theoretical spectrum from the integrationist position.In e�ect the glossematician simply abstracts from the actual human activitiesinvolved; whereas the integrationist places them in the forefront of the analysis.Hjelmslev, on the other hand, is wrong, even in terms of his own theoretical fra-

mework, to deny the relevance of the fact that orthographies do not always capturedi�erences in pronunciation. He seems not to realize the force of the objection. Theobjection is that a single written unit may correspond to two di�erent spoken units,as in the case of the noun refuse and the verb refuse, but this orthographic identitydoes not automatically obliterate any di�erence of meaning. All that happens insuch cases is that those familiar with written English learnÐto put it in traditionaltermsÐthat there are quite di�erent usages, grammatical constructions and pro-nunciations associated with a single written form. However, according to Hjelmslev,the fact that some orthographies are not phonetic shows that ``di�erent systems ofexpression can correspond to one and the same system of content'' (Hjelmslev, 1961,p. 105). The problem is that if cases like refuse are admitted as actual examples ofthis state of a�airs, it becomes theoretically possible to imagine writing systemswhich economize on their inventory of orthographically distinct words by allowinghomographs to proliferate. The reductio ad absurdum would be a writing systemwhich had only one word.Even Hjelmslev, presumably, would have regarded a single polysemous homo-

graph as inadequate to manifest in ink the abstract linguistic system of English,whatever that might be. But he o�ers no rule to determine the limit beyond whichthe deployment of homographs corresponding to phonetically or semantically dis-tinct units will result in non-equivalence between the spoken language and the writ-ten language.The di�culty is compounded by Uldall, who claims:

If we keep the units of content constant, we shall have the same languagewhatever system is used to make up the corresponding units of expression. [. . .]a system of any internal structure will do, provided that a su�cient number ofunits can be made up from it to express the units of content (Uldall, 1944).

This seems to imply that at least as many distinct units of expression are needed asthere are units of content in the language. Which in turn means that any vocal orgraphic systems which allow cases of homophony or homography automatically


misrepresent the content system of the language. But that in turn leads to the para-doxical conclusion that neither English speech nor English writing properly expressEnglish. It is paradoxical because without English speech or English writing it isdi�cult to see what kind of existence the English language would have. Further-more, if it is possible in principle that speech and writing may misrepresent thestructure of a language, there seems to be no a priori reason to assume that we canwith any assurance detect which elements of phonic or graphic manifestation cor-rectly represent the structure of the language and which, on the other hand, donot.In brief, glossematics shows us the muddle that results in linguistics when the

structuralist doctrine of the ®xed code is idealized to the point where it is assumed toexist independently of any speci®c materialization whatsoever. It is clear, therefore,that whatever similarity there may at ®rst sight seem to be between the integrationistposition on writing and the glossematic position is entirely super®cial. Theoretically,the two approaches could hardly be further apart.

8. Fixed codes and structuralism

The above discussion assumes that the ®xed-code doctrine is a basic tenet ofstructuralismÐand in that sense part of the ``structuralist legacy''Ðin at least theEuropean (Saussurean) version. But this assumption may beÐand has beenÐchal-lenged.It is challenged indirectly by John Joseph (Joseph, 1997) who questions whether

Saussure ever subscribed to the traditional Western ``language myth'' (Harris, 1981)and even whether Saussure had a theory of communication at all. Focussing on thefamous account of the speech act presented in the Cours (Saussure, 1922, pp. 27±28),Joseph points out:

Saussure never says explicitly that A and B must be speaking the same lan-guage, that A wishes or intends to transmit ideas to B, that B is able to interpretA's utterance as having precisely the meanings which correspond to the originalideas in A's mind, or that any of this is being o�ered as a model of ``commu-nication'' (Joseph, 1997, p.26).

On the basis of these lacunae, Joseph apparently concludes that Saussure did notimplyÐor may not have been implyingÐany of these things. As an argument exsilentio this ba�es credibility. That is to say, we are asked to take seriously thenotion that Saussure, in presenting this account of the speech act and omitting tomake the relevant disclaimers, wished the following possibilities to be entertained: (i)that A was speaking a language B did not understand and vice versa, (ii) that A didnot wish to say anying to B anyway, nor B to A, (iii) that B failed to grasp what Ameant, and A what B meant (which would follow from (i)) and (iv) that the exchangebetween A and B is NOT to be interpreted as exemplifying a typical episode oflinguistic communication (which would indeed follow from (i)±(iii)).


All one can say is that if this is so, Saussure must go down as one of the mostperverse (not to say incompetent) teachers in the history of the subject. It is ofcourse true that if (i)±(iv) re¯ect Saussure's intentions, then there is no place in thestory for a ®xed code. But nor is there any place for la langue at all, which is whatthe exchange between A and B is usually taken to illustrate. For if we are going tonitpick about lacunae in the text of the Cours, it should be noted that nowhere doesit actually state that the sounds A eventually utters represent a word or words in anylanguage, or that B interprets them thus. Exactly the same associationist pattern,matching notions to noises, would in principle do for an account of sobs, gasps,gulps and cries of all kinds. In fact, given Saussure's extreme vagueness about what``concepts'' are (faits de conscience), it would seem di�cult to exclude such possibi-lities if we are intent on looking for di�culties (however improbable) in the account.A more sensible interpretation of the passage Joseph cites is readily available,

however. It involves assuming that A and B are speaking the same langue, that eachdoes understand the other in virtue of both speaking this langue, that this is a typicalepisode of linguistic communication, and both participants intentionally engage in itwith a view to conveying speci®c verbal messages to each other. There is certainlynothing in Saussure's text which rules out that interpretation. And if that inter-pretation is right, then pace Joseph, we are indeed dealing with a classic example ofthe traditional view that linguistic communication is possible only ifÐand insofarasÐparticipants share a common verbal code for exchanging their thoughts. HadSaussureÐas Joseph seems to implyÐwished to challenge or at least not commithimself to that view, it is rather extraordinary that he did not spell out his caveats, ashe could easily have done.A quite di�erent objection to the interpretation of Saussurean structuralism as

®xed-code theory comes from Hewson (Hewson, 1992), who claims that ``Saussurianlangue is not, and never was, a code''. This is shown, according to Hewson, by theway Saussure appeals to the analogy of chess.

The analogy that Saussure uses for langue in the Cours is the game of chess; it isself-evident that the game of chess is not a code, where one item replacesanother on a one-to-one basis. A code, in fact, a�ects only the Saussurean sig-ni®ant symbolized by the chess pieces; it does not a�ect the Saussurean signi®eÂ,symbolised by the moves that the pieces make. In fact Saussure unwittinglymakes the point that a code makes no di�erence to the game by commenting thatone can exchange the wooden pieces for ivory pieces (i.e. one for one) withoutchanging the game in any way [. . .]. Codes are totally irrelevant to Saussureanlangue, which is a form (analogously the chess moves), not a substance (analo-gously the chess pieces) [. . .] In the game of chess an ivory piece can replace awooden piece, or a spice bottle can replace a lost rook, but a rook cannot replacethe moves that a rook makes. There is a world of di�erence between a chess pieceand the moves it can make [. . .] (Hewson, 1992, pp. 379±380).

What this shows, in the ®rst instance, is that Hewson has only partly understoodSaussure's reasons for appealing to the chess analogy (an analogy which, it must be


added, is applied somewhat di�erently in di�erent passages in the Cours. For dis-cussion, see Harris, 1987, 1993.) The distinction between form and substance is onlyone aspect of the comparison. But Hewson gets even this wrong. A rook is notsubstance in Saussure's sense. The substance is the material object standing on theboard. When a spice bottle is substituted for the missing piece, the substitution issubstantial: but the scent bottle then becomes a chess piece (which it was not beforethe substitution), and it does so in virtue not of its substantial resemblance to aconventional rook, but in virtue of now being subject to the rules of chess. And thatis where form resides. What Hewson fails to see is that the identity of a chess piece isde®ned by the rules, and the rules include stipulations about how it can move. Thishas nothing to do with Saussure's distinction between signi®ant and signi®eÂ. Boththe signi®ant and the signi®eÂ are formal units. So it is nonsense to say that signi®antsare code items whereas signi®eÂs are not. Either both are, or neither.In the second place, Hewson forces his own interpretation of what a code is on

Saussure's analogy. I am not sure what Hewson's notion of a code is (he supplies noexamples) except that it seems to involve `òne-to-one'' replacement of some kind. Butif so, then I imagine Hewson would deny that the operation of tra�c lights involves acode, since is di�cult to see what the signals are `òne-to-one'' replacements of. Andthis appears to be Hewson's main reason for denying that chess is a code. But there is achess code, just as there is a tra�c-lights code; and if there were not, then the rationaleof Saussure's analogy between chess and la langue would collapse. In the case ofchess, the code is stated explicitly in the rules, and in the case of tra�c lights in theHighway Code. In short, for Saussure the whole point about chess is that it is playedaccording to ®xed rules, and if you do not follow those rules you are not playing chess.In the third place, although Hewson claims that codes are ``totally irrelevant'' to

langue, the fact is that the Cours itself speaks of la langue as a code (Saussure, 1922,p. 31). Furthermore, this is not an editorial interpolation, but goes back to themanuscript sources. As De Mauro notes, ``l'interpreÂ tation de la langue comme5code4 remonte donc aÁ Saussure' (De Mauro, 1972, p. 423).In spite of the doubts expressed by Joseph and Hewson, I can see no cogent rea-

son for doubting that the integrationist interpretation of Saussure as a ®xed-codetheorist is well founded. The question that remains to be addressed is whether I amalso right in assuming that this is an essential part of the (European) structuralistlegacy. Or, to put it another way, would Saussurean structuralism make sense with-out a ®xed-code concept of languages? Could it remain theoretically intact if thatparticular component were removed?It is di�cult to see how it could. For what we are dealing with is a holistic concept

of structure. Saussure goes out of his way to make the point that linguistic units arenot to be thought of as a collection of independent items, a pile of bricks given inadvance of the building. On the contrary, it is the whole edi®ce which has to be inplace before there is any question of analysing its constituent parts. And this, whichI take to be the radical innovation that Saussure introduced into Western linguisticthought (Harris and Taylor, 1997, pp. 209 �.), is incoherent unless it is presupposedthat the relations between linguistic signs, together with the sound-sense correlationsconstitutive of each linguistic sign, are determinate for any given system. Which


amounts to saying nothing more nor less than that a language system (in the tech-nical sense of Saussurean langue) is to be conceived as a ®xed code. Fixed both in thesense that it is one and the same for all who use it, and also in the sense that anychange in its constitution automatically replaces it with something di�erent. That isprecisely why Saussure makes the point that if we alter the number of pieces in thegame of chess we must automatically play according to a di�erent ``grammar''(Saussure, 1922, p. 43). In other words, we get a (structurally) di�erent game.

9. Integrationism and behaviourism

Hewson not only takes issue with the integrationist interpretation of Saussureanstructuralism but also treats integrationism, as a ``return to the Behaviourism of the30s and 40s'' (Hewson, 1992, p. 381). Which, in e�ect, is to see it as part of thelegacy of Bloom®eldian structuralism.The main basis for this condemnation is a passage in Harris (1990) where the

general point is made that complying non-verbally to a request is no less a linguisticact than producing the verbal utterance expressing the request. The integrationistpoint here is that, although biomechanically quite di�erent, the two are com-plementary: these are integrated and integrating activities, not just events that hap-pen to occur sequentially, and the recognition of this integration is what gives eachthe status it has in the communication situation. Hewson comments as follows:

Harris even resurrects a form of the old Behaviourist chestnut (see Fries, 1954)that the overt behaviour following a speech act such as Open the door is themeaning of the act. He claims that B's `òpening the door in this particularsituation is a response which quali®es as `linguistic' no less indubitably than A'svocalization'' [. . .] Pandora's box is now wide open: If A frequently o�ers Bsome haggis, and B gets queasy every time, this response must also qualify as``linguistic'' (Hewson, 1992, p. 379).

This is evidently meant to be a reductio ad absurdum. How could feeling queasycount as language? Isn't this absurdity what integrationist logic commits one to? Theanswer is that it isn't. But before dealing with that, are we dealing with an oldbehaviourist chestnut at all?When Wittgenstein was accused of behaviourism he replied laconically: `Ìf I do

speak of a ®ction [sc. mental processes and states], then it is of a grammatical ®c-tion.'' (Wittgenstein, 1958, x307) And it is tempting to reply to Hewson with anintegrationist reformulation of this. But that would be tacitly acquiescing in a mis-representation of the behaviourist version of structuralism. Bloom®eld, for one,never held that any response to a speech stimulus was its meaning. Bloom®eldactually goes to some trouble in the chapter on meaning in his book Language(Bloom®eld, 1935) to make it very clear that this is not his contention. He distin-guishes quite carefully between the non-distinctive features of a speech situation andthe distinctive features: the latter alone, for Bloom®eld, contribute to the linguistic


meaning of a form. So Hewson's fatuous counterexample entirely misses the beha-viourist target (in part because Hewson fails in any case to distinguish betweeno�ering a haggis and merely uttering the word haggis). There is no parallel betweenfeeling queasy and shutting the door when requested to do so. In assimilating thetwo, Hewson manifestly misrepresents the integrationist claim. Feeling queasy ispresumably an involuntary reaction: shutting the door is not.Even more misdirected is Hewson's reference to the well-known article by Fries

(Fries, 1954), widely regarded at the time as a de®nitive statement of the (American)structuralist position on semantic criteria in linguistic analysis. For there it is madequite plain that ``personal'' meaning is quite di�erent from linguistic meaning. SoB's idiosyncratic aversion to haggis could not by any stretch of the imaginationcount as linguistically relevant.As for Hewson's claim to have discovered in Harris (1990) the ``behaviourist''

contention that the overt behaviour following a speech act is the meaning of the act,the most direct refutation is simply to cite the relevant passage in full. It reads:

In linguistic communication, what people do not say is just as important aswhat they do say. If A utters the words `Òpen the door'' and B in response saysnothing but simply opens it, then B has exhibited a knowledge of English whichpro tanto not only matches A's but is no less clearly demonstrated than in A'sutterance. This is not to say that opening the door is the only way in which Bmight demonstrate a linguistic pro®ciency adequate to this particular commu-nication situation; nor that it is the ultimate criterion of B's relevant linguisticknowledge. Both these assumptions would lead to a linguistics de®ned on anarrowly behaviouristic basis.

Nevertheless, B's behaviour in this situation is communicationally relevant, andhere the theoretical baby must not be thrown out with the behaviouristic bath-water. For there is no reason to deny that B's opening the door in this particularsituation is a response which quali®es as ``linguistic'' no less indubitably thanA's vocalization. It is not to say that every time a door is opened linguisticcommunication is taking place; any more than it is to say that every time agiven sequence of sounds is produced linguistic communication is taking place.But to acknowledge that in opening the door B makes a linguistically appro-priate contextual response to A's utterance is to recognize that B's actions arealso signs, on an equal footing with the signs expressed vocally by A. By soacting, B in turn signals to A; and, speci®cally, signals an interpretation of A'sown utterance. It is this reciprocity which is constitutive of communication. Byfocusing exclusively on A's utterance, orthodox linguistics introduces anasymmetry which simply ignores the mutual dependency of A's and B'scommunicative acts.

The theoretical implications of this mutual dependency are far-reaching. BothA's verbal act and B's non-verbal act have to be seen as integrated constituentsin an interactive continuum of communication. Divorced from that continuum,


neither has any communicational signi®cance whatsoever. Any attempt togive a systematic analysis of what A does independently of what B does restson a misconception of what is going on. It is like trying to describe a game oftennis as if it were being played by only one person, and the player on the otherside of the net did not exist. Such a description, we may he sure, howeverexhaustively it appears to cover the actions of the single player selected forattention, cannot make any sense of the game being played (Harris, 1990, pp.43±44).

From this, I suggest, it is clear that there is no endorsement of any alleged``behaviourist'' misconception about linguistic meaning, but, on the contrary, afairly clear rejection of ``the behaviourism of the 30s and 40s''.

10. Transcending Saussure

Having disposed of sundry misconceptions and potential misunderstandings ofthe ``structuralist legacy'', I would now like to go back to the notion that inte-grationism involves ``transcending Saussure'. In Harris 1995 I pointed out that thereare radical and less radical ways of transcending Saussurean structuralism, andthat the problem for the would-be transcenders is deciding at what point they willpart company with Saussure's theoretical assumptions and exactly on what groundto do so.Some attempts to transcend Saussure lead straight into theoretical incoherence. A

good example is Jakobson's attempt to transcend the Saussurean dichotomybetween synchrony and diachrony by combining both into what Jakobson called``dynamic synchrony''. If this is to be counted as more than mere playing withwords, it has to be given some substantive interpretation, and Jakobson (who didnot have Hewson's problems with recognizing Saussure as a code theorist) attemptsto do this by appealing to the alleged co-existence of di�erent ``sub-codes'' withinthe same linguistic community at the same time. Some of these sub-codes will bemore `àdvanced'' than others (from a diachronic perspective): hence the dynamismin the synchrony.Now integrationists also question Saussure's distinction between synchrony and

diachrony, but not for Jakobson's reasons. The integrationist motivation for therejection has a clear theoretical basis in the principle of cotemporality (Harris, 1998,pp.81�.): all linguistic agents and all linguistic acts are time-bound. Once thatprinciple is accepted, there is no theoretical room for Saussure's a-historicaldimension of linguistic study. But it is unclear that Jakobson's rejection has anytheoretical basis at all. In fact its point of departure seems to be a misreading ofSaussure's position. If Jakobson's separate ``sub-codes'' are independent systems (asJakobson seems to claim: Jakobson, 1990, p. 148), then they do not belong to anyidiosynchronic whole, in spite of their contemporaneity. But if, on the other hand,they are not systems in the Saussurean sense anyway, then there is no questionof their synchronicity. You cannot have your cake and eat it. In other words,


Jakobson fails to see how the Saussurean distinction between synchrony and dia-chrony is logically related to Saussure's holistic concept of a linguistic system. Thenotion of transcending the Saussurean dichotomy by trying to combine its twodichotomous components produces, as one might expect, theoretical mish-mash.What I take Love to be sayingÐand showingÐin Love (1990) is that the integra-

tionist conceives of transcending Saussure in a quite di�erent way. The task com-prises two essential parts. The ®rst of these is aptly called ``demythologization''. Thisconsists in showing how and why certain linguistic assumptions deriving from theWestern ``language myth'', whether entertained by the laity or by professional lin-guists, fail to stand up to scrutiny. (Because these assumptions are unfounded, mis-leading, irrelevant, self-contradictory, or otherwise repugnant to intelligentacceptance.) The second and subsequent part of the task has no generally recognizedlabel, but I will call it for convenience ``reorientation''. (Perhaps someone will thinkof a more satisfactory term.)Reorientation, as I envisage it, involves explaining what are or were the sources of

the myth that has just been exposed, and recouping from its demythologizationwhatever fragments of sense it contained. And here I should like to repeat what Ihave said elsewhere; namely, that much of what is wrong with modern linguisticsÐand with modern thinking about language generallyÐis wrong not because its pro-positions are straightforwardly false, but because the limited truth they contain ispresented in a badly skewed perspective. (It would be much healthier for everybodyif they were demonstrably false; but that would make the life of an intellectualaltogether too comfortable for comfort.)The structuralist legacy includes a great deal that it would be foolish for integra-

tionists to reject outright. For instance, that recognition of linguistic `ùnits'' ofvarious kinds is a function of their perceived contrast with other units of the sameorder: this seems to me a Saussurean breakthrough in our analytic understanding ofhow language works. But this insight is badly skewed when it is made to supportbelief in a ®ctional entity called ``the phoneme'' (of which, as a former colleague ofmine used to insist at great length, there are no more and no less than 44Ðor was it42? (I forget)Ðin ``the'' language called `Ènglish'').Where we ``transcend'' Saussure is in pointing out that any contrastive identi®-

cation of signs has to be relativized to the circumstances of particular commu-nication situations if it is to make sense. There is no contrast in vacuo. Or if thereis, it is an illusion produced by engaging in ``hocus-pocus'' linguistics. It is this illu-sion of the decontextualized contrast which is the baneful part of the legacy ofstructuralism.Now in some cases the demythologization will leave nothing that is worth

recouping and reorienting at all. For instance, if you demythologize the notion thatlinguistic forms have ®xed meanings, one consequence of this is that certainsemantic problems that orthodox linguists have spent much time arguing about arerevealed as non-problems. To illustrate this, take the distinction between homonymyand polysemy. There is a revealing case history in Ullmann (1959, p. 115). Ullmannmay be not unfairly described as a Saussurean structuralist or fellow-traveller, andhe agonizes over the question of whether the use of the word capo in certain North


Italian dialects to mean ``wheel hub'' is an example of homonymy or polysemy. Forcapo also means ``head'' (in reference to the part of the body supported by the neck).So are there two words here or just one word with two related meanings?Now once an integrationist has demythologized the distinction between homo-

nymy and polysemy, there is, as far as I can see, nothing remaining about the capocase that either requires or invites a reorientation. The problem that Ullmann triesto resolve (by deploying, incidentally, some very specious arguments that I cannotdissect here) is purely an artifact of the structuralist framework he is using. It hasno reality at all for the language users. (Or at least, if it has, Ullmann gives us noreason to think so.) This is not to say that I regard it as of no conceivable interestwhatsoever to examine what a wheel hub is called in Northern Italy: simply that the``linguistic'' problem of whether the word for ``wheel hub'' is ``the same word'' asthat for ``head'' is a bogus problem.The case with Love's ``relocation'' of the concept of ``languages'' is altogether

di�erent. Here we do indeed have something to ``relocate'', something that has sur-vived the demythologization. And Love ``relocates'' (or ``reorients'') it by showinghow it arises as a second-order construct, supported by a metalanguage and theavailability of writing.Anyone who has understood Love's message in all this (and, admittedly, it is fairly

sophisticated and not the kind of message to win a general election) seems to me tohave grasped at least two propositions. (Whether they are propositions you agreewith is a quite di�erent matter.) One is that the structuralist legacy is not to beignored (however wrongheaded Saussure might have been) because it re¯ects, how-ever indirectly, certain important lay views concerning language. The other is thatthrough the demythologization of that intellectual legacy, you are in a position tosee how the ideas it brings into play relate to your own understanding of your ownlinguistic experience. And that, for me, is doing integrational linguistics.That may seem to some too cautious or too anodyne a conclusion. If so, they

underestimate what is going on in integrational analyses such as Love's. I amreminded of the remark of a very experienced political commentator who whenasked to agree that the scale of the Conservative defeat in the 1997 UK generalelection meant that the Conservative party theorists needed to ``go back to thedrawing board'', replied: ``No. That would be disastrous. What they need is a newdrawing board.''Something similar applies in the linguistic case. Indeed, if the integrationist

critique of structuralism is on target, a new drawing board was not only needed butis already being provided.

References

Bloom®eld, L., 1923. Review of F.de Saussure, Cours de linguistique geÂ neÂ rale. Modern Language Jour-

nal, 8, 317±319.

Bloom®eld, L., 1935. Language. Allen & Unwin, London.

Chomsky, N., 1986. Knowledge of Language. Praeger, New York.


Cottrell, G.W., 1994. Connectionist approaches to natural language processing. In: Asher, R.E. (Ed.), The

Encyclopedia of Language and Linguistics, vol.2, Pergamon, Oxford, pp. 698±706.

Crystal, D., 1991. A Dictionary of Linguistics and Phonetics, 3rd ed. Blackwell, Oxford.

Crystal, D., 1992. An Encyclopedic Dictionary of Language and Languages. Blackwell, Oxford.

Davis, H., 1997. Ordinary people's philosophy: comparing lay and professional metalinguistic knowledge.

Language Sciences 19(1), 33±46.

De Mauro, T., 1972. de Saussure, F., Cours de linguistique geÂ neÂ rale, Edition Critique. Payot, Paris.

Fischer-Jùrgensen, E., 1966. Form and substance in glossematics. Acta Linguistica Hafnienisa X (1), 1±33.

Fries, C.C., 1954. Meaning and linguistic analysis. Language 30 (1), 57±68.

Harris, R., 1981. The Language Myth. Duckworth, London.

Harris, R., 1983. de Saussure, F., Course in General Linguistics. Duckworth, London.

Harris, R., 1987. Reading Saussure. Duckworth, London.

Harris, R., 1990. On rede®ning linguistics. In: Davis, H.G., Taylor, T.J. (Eds.), Rede®ning Linguistics,

Routledge, London, pp. 18±52.

Harris, R., 1993. Saussure, Wittgenstein and la reÁgle du jeu. In: HarreÂ , R., Harris, R. (Eds.), Linguistics

and Philosophy. The Controversial Interface. Pergamon, Oxford, pp. 219±131.

Harris, R., 1995. Saussure, generative grammer and integrational linguistics. In: De Mauro, T., Sugeta, S.

(Eds.), Saussure and Linguistics Today. Bulzoni, Roma, pp. 203±213.

Harris, R., 1998. Introduction to Integrational Linguistics. Pergamon, Oxford.

Harris, R., Taylor, T.J., 1997. Landmarks in Linguistic Thought. The Western Tradition from Socrates to

Saussure, 2nd ed. Routledge, London.

Harris, R., Wolf, G., 1998. Integrational Linguistics: a First Reader. Pergamon, Oxford.

Hewson, J., 1992. Review of Davis, H.G. and Taylor, T.J. (Eds), Rede®ning Linguistics. Canadian Jour-

nal of Linguistics 37(3), 378±381.

Hjelmslev, L., 1961. Prolegomena to a Theory of Language, revised ed. University of Wisconsin Press,

Madison (F.J. Whit®eld, Trans.)

Jakobson, R., 1990. Typological studies and their contribution to historical comparative linguistics [1958].

In: Waugh, L.R., Monville-Burston, M. (Eds.), Roman Jakobson, On Language, Harvard University

Press, Cambridge, MA, pp. 143±163.

Joseph, J.E., 1997. The ``language myth''myth: or, Roy Harris's red herrings. In: Wolf, G., Love, N.,

(Eds.), Linguistics Inside out. Roy Harris and His Critics. Benjamins, Amsterdam, pp. 9±41.

Komatsu, E., Harris, R., 1993. F. de Saussure, TroisieÁ me Cours de Linguistique GeÂ neÂ rale (1910±1911).

Pergamon, Oxford.

Komatsu, E., Wolf, G., 1996. F. de Saussure, Premier Cours de Linguistique GeÂ neÂ rale (1907). Pergamon,

Oxford.

Komatsu, E., Wolf, G., 1997. F. de Saussure, DeuxieÁ me Cours de Linguistique GeÂ neÂ rale (1908±1909).

Pergamon, Oxford.

Lepschy, G.C., 1970. A Survey of Structural Linguistics. Faber, London.

Love, N., 1984. Psychologistic structuralism and the polylect. Language & Communication 4 (3), 225±240.

Love, N., 1989. Transcending Saussure. Poetics Today 10 (4), 793±818.

Love, N., 1990. The locus of languages in a rede®ned linguistics. In: Davis, H.G., Taylor, T.J. (Eds.),

Rede®ning Linguistics. Routledge, London, pp. 53±117.

Martinet, A., 1964. Elements of General Linguistics. Faber, London, (E. Palmer, Trans.).

Pateman, T., 1983. What is a language? Language & Communication 3 (2), 101±127.

Robins, R.H., 1989. General Linguistics. An Introductory Survey, 4th ed. Longman, London.

Saussure, F. de, 1922. Cours de Linguistique GeÂ neÂ rale, 2nd ed. Payot, Paris, (R. Harris, Trans., F. de

Saussure, Course in General Linguistics. Duckworth, London, 1983).

Siertsema, B., 1965. A Study of Glossematics, 2nd ed. Nijho�, The Hague.

Uldall, H.J., 1944. Speech and writing. Acta Linguistica 4, 11±16.

Ullmann, S., 1959. The Principles of Semantics, 2nd ed. Jackson, Glasgow/Blackwell, Oxford.

Wittgenstein, L., 1958. Philosophical Investigations, 2nd ed. Blackwell, Oxford (G.E.M. Anscombe,

Trans.).

Wolf, G., Bocquillon M., de la Houssaye, D., Kryzek, P., Meynard, C., Philip, L., 1996. Pronouncing

French names in New Orleans. Language in Society 25, 407±426.


Wolf, G., Love, N., 1993. Integrational linguistics: an introductory survey. In: CrochetieÁ re, A., Boulanger,

J-C., Ouellon, C. (Eds.), Proceedings of the XVth International Congress of Linguists. Presses de

l'UniversiteÂ Laval, Sainte-Foy, 1, pp. 313±320.


Date post:	16-Sep-2016
Category:	Documents
Upload:	roy-harris
View:	215 times
Download:	1 times

Integrational linguistics and the structuralist legacy

Documents