Göran Sonesson

Unité de recherches linguistiques 7,

Centre national de recherches scientifiques,

Paris

in Papers from the 5th Scandinavian Conference of Linguistics, del IN, ss. 283-294. Stockholm: Almqvist & Wiksell International 1979.

“Ein jeder, weil er spricht, glaubt über die Sprache sprechen zu können.”

Goethe, quoted by Karl Kraus (1957)

Most textbooks will describe the subject matter of linguistics as the mapping of sounds onto meanings. Supposing linguistics is an empirical science whose task it is to describe a really existing object called language, it will be necessary to make sense of this characterisation. We will have to furnish it with an interpretation. Here, I do not want the term interpretation to be taken in the sense of model theory, which seems to conceive of the model and the interpreted formula as two pre-existing objects only to be brought together by the theory. Instead it should be taken in the sense of hermeneutics as a complex undertaking aiming at the establishment of the existence and characteristics of the subject matter, where the model itself is contributing to this very “effort after meaning”, which, according to Bartlett (1932), is a continuing. element of every intellectual enterprise in the widest sense of the word.

In a given linguistic theory, as for example in generative grammar, the notion of mapping forms a perfectly clear part of the meta-language. But while we can easily see to which objects of the life world sound and meaning correspond (namely to sound and meaning, respectively), it is by no means clear to which life world phenomenon, if any (and then, to which degree), the mapping should be related. I would like to call this a hermeneutic problem. Paul Ricoeur (1965.29-36) has discussed two hermeneutic traditions, the Aristotelian one, according to which every meaningful utterance is open to interpretation, and the exegetical, which applies an interpretation to an authoritative but ambiguous text. He then proposes a third definition, according to which interpretations apply only to multiple meanings, as for example in the case of symbols. Pulling the third sense somewhat in the direction of the first one, I will talk about interpretation when meaning stands out as a problem, even in case the multiple meaning turns out to be only provisional. Coming from a more general epistemological horizon, we enter the hermeneutic field when we start to look for a mediation between the life world of the subject, which is always ambiguous, and the constructed. world of a science, which, as such, is undetermined.

We may postpone discussion of’ different hermeneutic traditions, if we can agree that, on every reasonable interpretation, our text book definition implies that linguistics should tell us how, given a signifier, the speaker-hearer is able to retrieve the signified, and vice-versa, at least in the sense that it should specify what is required for. the signifier to correspond to a certain signified rather than another. This suggests the sign of the structuralist tradition, and more precisely, the semiotic function as conceived by Hjelmslev (1943): exactly as in the commutation test, the replacement of certain features on one plane entails the integral substitution of the unit on the other plane, while other features may be changed with no further consequences, so, in the perception of the speaker-hearer, some features are more essential than others for the recognition of the sign-Gestalt. This means that the semiotic function can be seen as an integral part of the intentionality of the perceptual act as described by Husserl. Of course, to say that is hardly orthodox glossematics. But it is in accordance with psycholinguistic experiments as well as with the intuitive foundations of structuralist theory (cf. Sonesson 1978a:38-67).

It will suffice here to admit that linguistics should describe the conditions on the pairings of content and expression, as an essential part of linguistic competence. Now, clearly, if the speaker-hearer knows anything about his language, he knows how to get from a given signifier to the corresponding signified and vice-versa. Of course, linguistics may describe these pairings in a more or less direct way, but I hope you will admit that, everything else being equal, the better theory is the one which most closely describes the reality of’ the semiotic function. If this is so, generative grammar us well as pragmatic analysis will come in for serious criticism. As an example, we will consider what has become of the notion of a performative or illocutionary act as first conceived by Austin (1962) after its integration into different linguistic frameworks. In our first section, we will have a look at the so-called performative hypothesis as presented by Ross, Lakoff, and Sadock. Then we will go on to consider pragmatic analysis.

The difference making no difference here, we may summarise the performative analysis, in the words of Ross (1970:259) by saying that every declarative sentence is supposed to “derive from deep structures containing one and only one superordinate performative clause whose main verb is a verb of saying”; in a similar way, every question is subordinated in the deep structure to a performative of questioning, every promise to a performative of promising, and so on. This is reminiscent of the definition given by Austin (1962:103) according to which the illocutionary act is “conventional at least in the sense that it could be made explicit by the performative formula”. Ross is consequently proposing one possible interpretation of the “implicit performative” of Austin.

There are many ways of criticising this contention: for example, with Anderson (1971) and Frazer (1971), one may argue that, in explicit performatives, many illocutionary acts may be realised at the same time, and that the illocutionary device does not have to be the uppermost clause of the sentence. One may argue, as I have done elsewhere (Sonesson 1978b:120), that it is impossible to make sense of the notion of’ a performative lexical insertion, and that the argument, in the case of explicit performatives, will lead to an infinite regression. More simply, one may argue, with Allwood (1976.183), that there is no point in transforming clearly pragmatic constraints into syntactic structures. As, last but not least, one may argue, with Bach (1971), that the arguments of Ross would justify the postulation of all sorts of implicit elements. Here I will consider another line of criticism.

A trivial argument for performative analysis is of’ the following type: given a sentence, as for example, “I will be there tomorrow”, we observe that it may be a statement, a warning, a promise, a threat, and so on, and we conclude that it may derive from different deep structures with different superordinate verbs: “I warn you that. . . “, “I promise you that...”, etc. Now, this analysis seems to me to be quite inadequate, for two reasons. a) It leaves unanalysed the capacity that the speaker-hearer has to retrieve these different meanings or, if you prefer, deep structures, when he is faced with the multiply-ambiguous surface structure. There may be a “signs zéro" somewhere in the system but not corresponding to every one of the oppositions. The performative hypothesis, then, is as helpful as explaining why morphine makes you sleep by invoking its soporific quality. That is why generative grammar seems to me to be deep structure distributionalism: as was also the case with the old distributionalism, it does not explain anything, it only notes the existence of a relation which it leaves unqualified, the only difference being that it supplies from a postulated deep structure the elements which do not appear on the surface. At least, I submit, this is the case here.

b) Second, the argument supposes that a surface structure realisation of a superordinate performative clause always transforms the sentence into an explicit illocutionary act. This, I will argue, is simply not the case. We may ignore here those linguistic acts which only work if they are not recognised or not intended to be recognised as such (for example, we do not say, I lie to you that... “, “I jeer at you that...”, etc.), since this could perhaps be explained by the “lurk-argument of Ross. Instead, let us consider a verb which always figures as an obvious candidate on the lists of performative verbs, from Austin onwards, i.e. “answer”. It will be argued here that the explicit usage of a superordinate clause where “answer” or “reply” appears an the main verb is neither a necessary nor a sufficient condition on the production of an act of answering. This should be fairly obvious. We may first note that, if you were fooling around saying to everybody, for example, “I answer you that you are a fool”, you could hardly be said to produce anything like an answer. So the performative formula is not sufficient. And, of course, it is not necessary, since we do have answers where, according to the performative hypothesis, the performative verb appears only in the deep structure. But now consider some more interesting examples which serve to show that the so-called explicit performative formula is quite ambiguous..

(1) Begin is as great a friend of peace as Nobel was. – (I reply

that) he is.

(2) Everybody vote for Mr. Palme! – (I reply that) I will.

(3) This nice Mr. Giscard really is a great statesman. – (I answer you that) Mouna is a much greater one.

Clearly, with our without the superordinate clauses we will here have, respectively, an act of agreement, an act of acceptance and an act of objection (For a more full treatments see my French dissertation, Sonesson 1978b:220-255). Inversely, we may observe that the utilisation of other so-called performative formulas, in an appropriate context, will still give us an act of answering.

(4) What the hell are you doing? – I pronounce you man and wife.

(5) Would you like a coup of tea? – Thank you!

(6) Do you love me? – Come here and kiss me!

Now, it should be clear that it is sufficient for a sentence to follow another sentence which qualifies as an act of questioning for it to acquire the illocutionary force of an answer. The so-called performative formula is of no consequence. Of course, the content will also turn out to be relevant, if we want to distinguish the notion of a proper answer from the more general notion of an answer (see the work of Åqvist, Belnap, and other logicians).

Another case in point is the existence of verbs accomplishing another illocutionary act than the one corresponding to the descriptive meaning of the verb: for example, in French, ‘Je t’emmerde” is not an act of annoying somebody but rather a challenge. Furthermore, a so-called explicit question, as “I ask you what time it is”, rather than being an act of questioning, should be considered an act of insisting or of repeating a question. An argument in favour of this analysis may be the existence of other possible retorts. For example, after “I ask you what time it is?”, but not after “What time is it?”, you may say “I haven’t heard that” or “You didn’t say that”, and those will not be answers.

I have argued that the performative hypothesis has nothing to tell us about the way meaning and expression are brought together in the semiotic act. It should be observed that an orthodox chomskyan interpretation will not save this hypothesis from futility. We may of course think of the implicit performative clause as a part of the biologically founded constraints on possible grammars, in Chomsky’s sense, but this will never explain how we are able to differentiate between different illocutionary acts. An innate cognitive structure could only derive one type of performative clause from undifferentiated perceptual evidence, i.e. the sentence. So that the innateness interpretation does not affect my claim.

I would now like to take my argument a little further. As a second essay to integrate Austin’s observations on the linguistic act into a more comprehensive framework, I would like to consider speech act analysis, originating in Strawson’s amalgamation of the illocutionary act with Grice’s analysis of non-natural meaning, followed up by Searle, Schiffer, Lewis, Wetterström, Parrett, Allwood, and others. From the point of view of linguistics, speech act analysis, in this sense, is a more faithful approach to Austin’s original observations. Some linguists, as for example Heringer, Fraser, and Gordon & Lakoff, will simply apply the ideas of Searle. Others, like Wunderlich and Allwood, continue in their own way the same tradition.

The linguistic act, as it occurs in the life world, is a rather complex phenomenon. Thinkers like Husserl and Schütz have shown that the meaning of the sign in the language system should be distinguished from what the speaker means by using the sign here and now (Husserl 1900; Schütz 1932). Since the linguistic act takes place in the “extra-linguistic” world, it should be seen as a lager semiotic unity. When we leave linguistics for semiotics. the object changes (since a larger part of the life world phenomenon becomes relevant) but the point of view should stay the same, that is, we continue to look for the semiotic function. This does not mean that we loose anything: on the contrary, what the speaker meant now appears as still another sign, and this reminds us that there must be something in the linguistic act over and above the system signifier which permits us to identify this supplementary signified.

The remarkable thing about speech act analysis is that is seems to conceive of communication as taking place without the mediation of a sign structure, as a kind of communion of intentions. From Grice to Schiffer over Strawson and Searle, there has been a continuing multiplication of the levels of intentions, supposedly necessary to distinguish an act of communication from acts of manipulation which do not make any use of conscious influence. But even with the postulation of the hundredth intention to intend, we do not get any closer to know how the speaker-hearer is able to recognise and to distinguish these acts. On the other hand, the semiotic function seems quite essential for the definition of the linguistic act. Speech act analysts completely ignores the existence of an expression plane. Even Allwood, who actually knows better, defines the communicative act as “an occurrence of behaviour connected with a parcel of communicative intentions” (Allwood 1978.10; but cf. Allwood 1976). If the communicativeness of the intentions is the same as that of the act, telepathy it would seem, will turn out to be communication. If telepathy works, it surely is communication, but then there must be a vehicle, be it of a hitherto unknown kind. Or suppose there is no mediation but God is furnishing each monad with information about the intentions of the other and that the monads are aware of this, then it would be strange to talk about communication.

Consider Searle’s (1969) argument for the viability of the speech act analysis. It is the only argument of the kind I know of. According to Searle, if we found letter-resembling forms in the sand (produced, for example, by the wind), we would not consider them to mean anything, since they have not originated in a speech act. But now suppose these forms were perfectly alike to letters produced by human beings. Then this would normally be sufficient for us to conclude that they have been produced intentionally by human beings. And if’ we have actually assisted at the formation of the letters by the wind, we will be inclined to see in the workings of the wind a message from God or from the local spirits. This is exactly as when a parapsychologist, seeing, if we are to believe him, pencils writing without being manipulated by any visible hand, concludes that there is, literally speaking, a ghost writer behind it. That is to say, when we identify an act of meaning, we do tend to look for an intention behind it. But clearly, it is not the other way round as Searle would have it: that is, it is not the case that we identify an act of meaning by recognising the intentions of the speaker. Consider the notorious case of the psychiatric patients discussing their problems with a computer: they do understand what the computer says, and they take it very seriously. It is only when they !earn that they have been counselled by a computer, that they disqualify what it has said. But the linguistic acts of’ the computer do not loose their meaning, they only loose their intentionality. The linguistic act is one of the numerous life world phenomena which, once recognised and assigned to a category, are usually thought to be performed consciously and with purpose. That is, purpose is implied by the execution of the act unless cancelled by the context. But the unintentional act cannot be considered a degenerate case of the intentional one, since the recognition of the intentions is of no importance for the identification of the act. Recognition follows the identification, not vice-versa.

Let us now see what we can make of Schiffer’s rather complex model of the speech act. This model is interesting in two respects: it defines the notion of mutual knowledge, which permits us to avoid the infinite regression of intentions in Grice, Searle, and Strawson (cf. Schiffer 1972:34). It will not be possible to discuss this feature here. Second, it doe,. in a way, integrate the sign structure in the speech act, by postulating that the public’s belief that p, i.e. the signifier, is in a certain relation R to the belief that p, i.e. the signified, is at least a partial reason for the public’s belief, after the enunciation of E, that p (Schiffer 1972:63). This is interesting, because it seems to say, as I just did, that the intentions are identified by the act, and not vice-versa. But then, why should we need, over and above this, a complex system of intentions? And in what sense is this a definition of meaning, if meaning is already presupposed in the definition? The relation E of Schiffer is in fact the semiotic function, so meaning is actually not at all explained by his laborious definition. Let us admit for a moment that Schiffer’s definition only aims at characterising what s means by x, when we already know what is meant by x in a given semiotic system. Even then, the definition is not really helpful. It only tells us that S should have a certain intention in enunciating x. But, it does not tell us how we get to know that S has the intention that we should have good reasons to believe that p, that our reason for this should at least partially be our belief that x is related to p by R, and that S has the intention to realise E, i.e. his enunciation of x. So that we must wonder how this other relation R2 is constituted which permits us to relate the signified intentions to their signifiers.

Now let us see how Schiffer takes care of the relation R. S will pronounce x in the circumstances C, only if he thinks that the enunciation of x will constitute evidence that he wants to say p.. The more x contains the factor f which is characteristic of a well-known medium of communication, the more probable is the realisation of this state of affairs. Schiffer now tries to derive the relation R of the speech act from the imitation of natural signs. Suppose that S want to signify to P that he is furious, and that there is mutual knowledge between S and P, that the sound “grrr” emitted by a dog signifies that it is furious; then S may use an imitation of this sound to signify his anger, and, once he has had this idea, it is very probable that P, in turn, becoming another S, will. use it again, to signify his anger. This clearly resembles Tarde’s doctrine of imitation as the basis of society.

I will make a few comments (cf. Sonesson 1976.52-66): a) supposing there is a natural repertoire of sounds, then, why should man have to learn them from dogs? Rather more probably, he will start out with some sounds proper to the species. b) Here we have, then, a dog, which “signifies” his anger. Now, either this is to say that the dog “means” something, in the Gricean sense, and then it will have to realise all the required intentions, and it will have to know mutually with a dog public a sound which it, in its turn, imitates, and so on infinitely. Or else, the dog does not mean anything at all, and then the constitution of the relation between the sound and the sense is only due to the human being imitating it, i.e. the constitution of the semiotic function is still not explained. This will also be the case if the bark of the dog means something “naturally”, in the sense of Grice, i.e. as the clouds mean rain. In fact, the cloud is an aspect of the total event of raining, together with many other changes in the weather state, and only to a human being will it “mean” rain. Only inside our particular life world is this a “natural” sign. To Nils Bohr and his followers, an electron leaping from one orbit to another is, in exactly the same sense, a sign of the electron’s charge. Everybody can see the relation between a cloud and the advent of rain, but it takes certain instruments and a theory to see the relation between an electron leap and its charge. But then, if S is to understand the dog as expressing a certain mental content by barking, he must already understand the semiotic function, which, I trust, is impossible to grasp for a dog. c) But even if S could understand the meaning of’ the sound emitted by the dog, this would not be sufficient to permit him to use the same sign about himself, and, even so. this does not justify P’s using it the same way. To do this, S and P should be able to identify the sensation of anger in themselves and in the dog as being the same kind of event. They should be able to label it in the same way, irrespective of its variable causes and the changing context. But this already seems to presuppose the semiotic function, which may be necessary to categorise reality (cf. Sonesson 1978a:49). d) But this is still not enough. In order to use the bark of the dog as a sign, S would have to adapt its sound shape for human usage, as always in the case of onomatopes. He would have to fix the limits of relevance beyond which the sign for anger changes into another sign. Given the original sound substance produced by the dog, he world have to constitute another sound form. In order to be used by S and P, the signifier should also be objectified. And that, once again, presupposes the semiotic function.

Once again, it should be observed that what is at stake here is not the conventional or innate nature of the sign function. As far as the argument goes, the semiotic function may be a phylogenetic or a octogenetic acquisition.’ This alternative is perhaps too simple. On the other hard, if the semiotic function is of a more social than biological nature, the argument will permit me to side with Durkheim, rather than with Tarde: some structure must precede the interactions of the individuals. So far, the argument says nothing about the nature of this structure.2

Notes

1 ) I have left out a large part of the original text which contained a critique of Allwood, since our difference turned out to be of a much deeper nature than I thought at first. Unfortunately, the discussion of this question obscured the essential argument of the essay.

2) In the philosophy section of the present conference Døør and 0lsen offered some pieces of criticism of Popper’s notion of “objective knowledge”. Their point remained rather obscure. In this essay I oppose speech act analysis, which derives the interactions from the individuals, to another approach which supposes the existence of some kind of interaction structure. In this sense I side with Popper. But this does not seem to be the same opposition as that between objective knowledge in the sense of Popper and “collective knowledge” as favoured by Døør and Olsen, since they too oppose the derivation of social structure out of isolated individuals.

Allvood, Jens, 1976. Linguistic Communication as Action and Cooperation. Dept. of Linguistics, Gothenburg.

1978: “On the Analysis of Communcative Action”, in Papers in Theoretical Linguistics 38. Dept. of Linguistics, Gothenburg.

Anderson, S.R. 1971. On the Linguistic Status of the Performative/ Constative Distinction. Indiana University Linguistics Club.

Austin, J.L. 1962: How to Do Things with Words. The Clarendon Press, Oxford.

Bach, L. 1971. “Concrete Syntax”, in Studies out in the Left Field. Linguistic Research, Inc., Edmonton; PP 151-155.

Bartlett, F. 1932: Remembering. Republished 1967. Cambridge University Press, Cambridge.

Frazer, Bruce 1971: An Examination of the Performative Analysis. lndiana University Linguistics Club, Bloomington.

Hjelmslev, L. 1943. Omkring Sprogteoriens grundlaeggelse. Republished 1966. Akademisk Forlag, Køpenhavn.

Husserl, L. 1900. Logische Untersuchungen. Zweite, angearbeitete Auflage 1913-1921. Niemayer, Halle.

Kraus, Karl, 1937: Die Sprache. Verlag Die Fackel, Wien. A. Auflage, hrsg. von H. Fischer, Kösel-Verlag KG, Munchen 1954.

Ricœur, Paul 1965. De l’Interprétation. Seuil. Paris.

Ross, J.R. 1970: “On Declarative Sentences”. In Jacobs & Rosenbaum, eds., Readings in English Transformational Grammar, Ginn & Coy Waltham, Mass.

Schiffer, S. 1972: Meaning. The Clarendon Press, Oxford.

Schütz, Alfred 1952: Der sinnhafte Aufbau der sozialen Welt. Re published 1974. Suhrkamp, Frankfurt/ A.

Searle, John 1969. Speech acts. Cambridge University Press, Cambridge.

Sonesson, Göran 1976: Pour une théorie du parleur. Working Paper.

1978a: Tecken och Handling. Bokförlaget Doxa, Lund.

1978b: Une Machine à Parler. Ecole des Hautes Études en Sciences Sociales, Paris.

To be published: A Theory of Semiotic Acts. Toronto UniversitySemiotic Circle Toronto.