DISCUSSION* |
Is Language
|
| In his sharp critique of contemporary theoretical
linguistics, Pavel Tichý speaks about a scandal (The Scandal of Linguistics,
From the Logical Point of view 3/92, 70-80). As a matter of fact, I am not quite
unsympathetic with such a sharp criticism of linguistics; but the view of language and of
linguistic theory presented in Tichý's essay seem to me to be so misguiding, that I doubt
that his advice presented in the essay could really help linguistics "to get out off
ground". Tichý claims that language is a matter of "making one's thought accessible to his neighbours", which can be achieved due to the fact that expressions have meanings. According to him, language is a code, and doing theory of language means cracking the code, revealing what individual expressions encode by pairing them with their meaning. As there are infinitely many expressions, meaning-assignment must be, according to Tichý, underlain by rules which speakers "store in depths of our minds"; and theory of language cannot aim at listing all the eventual expression-meaning pairs, but rather at the rules, it is to deliver generator of expression-meaning pairs. Now the contemporary linguistic theory, Tichý argues, produces no expression-meaning pairs, let alone systems for generating such pairs, and hence is an utter failure. What linguists actually do is, according to Tichý, pairing expressions with other expressions (be they called 'deep structures', 'logical forms' or whatever) and not with their extralinguistic meanings. Moreover, linguists supported by philosophers cut off the branch on which they should sit by putting into doubt the very existence of |
such entities as meanings. To gain a deeper insight into the nature of Tichý's criticism, let us first think about the idea of presenting the expression-meaning pairs. We have an expression, say 'shaving' or 'Jane shaved Fred', and we want to supplement it with its meaning. What is the meaning of 'shaving'? Well, it is, of course, shaving. What is that of 'Jane shaved Fred'? It is, of course, that Jane shaved Fred. So we may consider presenting the pairs <'shaving', shaving>, <'Jane shaved Fred', (that) Jane shaved Fred>, and similarly for other expressions. But this is hardly what Tichý has in mind; this would downgrade linguistics to triviality. (The situation would, of course, be less trivial if we wanted to present a generator of all such pairs, but still it would be enough to be able to generate all the well-formed English expressions; and I think that Tichý would agree that there is more to a theory of language than that.) So if the output of linguistics is to be a list of expression-meaning pairs, and if linguistics is to be a nontrivial matter, we have to require that the meaning of an expression - at least in some cases - be specified via an expression other than the one in question. This means that he must have in mind that, e.g., the meaning of 'shaving' should be presented by an expression other than 'shaving'. The natural way seems to be to require the meaning being given "directly". However, it is clear that we cannot articulate the meaning without the help of its name; hence "directly" can at best mean something like "by means of its 'True' name". Now what may be the reason for believing that there are names for meanings of expressions better than the expressions themselves? What may be the reason of |
JAROSLAV PEREGRIN, IS LANGUAGE A CODE ? |
74 |
| believing that there is, e.g., a better way to express what
'shaving' means than to use the word 'shaving' itself? Well, we may be convinced that
there is some "True Language of Being", sanctioned by God or by the Nature
Itself, which allows us to address things in a more straightforward way than any other
language does. Without claiming that such a mystic idea cannot make sense we clearly see
that it can be of no help for linguistics; for linguistics is a science, based on
justification, and how one could justify his belief that the language he is using to
present meanings is the "True" one? But Tichý is no mystic and this is hardly what he has in mind; so let us examine another line of thought. A non-mystical alternative to the idea of "the True Language of Being" is the idea of "the True Language of Science", the idea that our science is able to decide how should the best language with which the world could be described look like. This is the idea put forward, e.g., by logical positivists during the first half of this century; but the positivists clearly failed to give it a substantiation. In fact, instead of pointing out the right language to describe the world, the positivists succeeded only in making it clear that there is no such single language; in particular that there is no objective ground to make the decision between the physicalistic and the phenomenalistic language. And it is hard to see the rest of this century as bringing any change to this - on the contrary, the well-known puzzles of subatomic physics indicate that the idea of a Single True Language is even much more far-fetched than it appeared to the logical positivists(1). However, I think that there is still a third line of argumentation according to which Tichý's proposal could be defended, a line I myself |
would be sympathetic with. We may claim that the language by
which we present meanings should be better not in that it would be "closer to the
world", but because it would be in some sense canonical. What I mean is that
if we find some kind of schematic rendering of our language which is considerably simpler
and more comprehensible than natural language itself and which nevertheless retains the
semantically relevant properties of natural language, we do a job that really helps us
understand language, because we facilitate comprehensibility of the structure of
language (understanding structure, of course, more inclusively than syntactic structure)
and of the way language functions. I am convinced that this is indeed the only meaningful
way to grasp semantic theory; and this is also a way we could make sense of Tichý's
proposal. However, I am afraid that this is not Tichý's point of view. In fact if we accepted this view, then Tichý's criticism of contemporary linguistics would loose much of its strength. Linguists clearly do use various canonical languages to account for semantics of natural language. Of course, there are better and worse ways to do semantic schematization, and I think that one could articulate criteria according to which we could show the usual linguistic ways to be less ample than, say, the way urged by Tichý; but this is not what Tichý does - linguists, according to him, only map natural language expressions onto expressions of another language, whereas real semantics means mapping expressions on extralinguistic objects(2) . Let us consider the sentence 'Jane |
JAROSLAV PEREGRIN, IS LANGUAGE A CODE ? |
75 |
| shaved Fred'. I think that the way to account for its semantics usual among contemporary linguists might be something like [shavePret [Jane]Ag [Fred]Pat](3); whereas the way which Tichý himself proposes (e.g. in his recent book) would be something like lwt.[0shavewt 0Jane 0Fred]. The question now is: what makes [shavePret [Jane]Ag [Fred]Pat] into a pseudosemantics and lwt.[0shavewt 0Jane 0Fred] into semantics? The easy answer would be: the former is an uninterpreted string of letters, whereas the latter denotes something, namely a certain construction. This is not exactly true: the linguists' notation is not a mere string of letters, because the linguists do in a sense understand it, and if they are pressed, they would say something to the effect that it expresses a relation with its two roles filled by two individuals. So the difference may be at most that the linguists' notation is interpreted only quite vaguely, whereas Tichý's one exactly. But then we have to ask what is it that makes Tichý's notation exactly interpreted, and we would have to answer that it is an explicit definition, stating that lwt.[0shavewt 0Jane 0Fred] denotes the construction of doing so-and-so. This shows that lwt.[0shavewt 0Jane 0Fred] is a mere dispensable notation, and that the resulting expression-meaning pair should be <'Jane shaved Fred', the construction of doing so-and-so>. However, the natural language expression 'the construction of doing so-and-so' is in no lesser need of semantic explication than 'Jane shave Fred'; so we would first have to present a pair <'the construction of doing so-and-so', ...>. Either we end up in a vicious circle, or we must simply | take one part of language as semantically transparent.
However, in the latter case our semantic theory will be nothing more than a theory of
translation, of relations between expressions: we equate expressions of one part of
language (the "nontransparent" one) with those of another part (the
"transparent" one). Tichý makes fun of a student who claimed that meaning is an expression; and it is indeed clear that to consider meaning an expression is absurd. However, it is no less clear that in order to present meaning, we must use an expression. If we are to articulate the expression-meaning pair, we need a way to express the meaning. So semantic theory is always in a sense a theory of translation of one language into another; and thus we can hardly disapprove of a particular theory on the basis of saying that it is based on translation. We would have to articulate criteria according to which we could assess individual ways of translation from the point of view of their capability of showing us really something about semantics of natural language; and only such criteria would justify our pronouncements to the effect that one theory is better than another. How could we learn the meaning of an expression? In fact there are principally two ways: we can be shown how to use the expression (in some particular cases what kinds of things this expressions applies to), or we can be provided with the translation of the expression into a language we are familiar with. It is important to realize that the first way is an essentially practical way; and it cannot find way into a theory. I can show someone what 'shaving' means, but if I want to put it down, I must somehow express what is it what I am showing and I end up with either the trivial "'shaving' means shaving", or a less |
JAROSLAV PEREGRIN, IS LANGUAGE A CODE ? |
76 |
| trivial "'shaving' means doing such-and-such things with
such-and-such instruments". What semantic analysis can bring to light are only relations between expressions (especially the relation of consequence between sentences) and consequently the structure of language, not meanings of individual expressions. Equating expression with a formula of a canonical notation is not to pick up the thing the expression stands for, but rather only to point out the place of the expression within the system of language the canonical notation materializes. Let us illustrate this using - for the sake of simplicity - the simple possible worlds semantics instead of Tichý's more sophisticated apparatus of constructions. Within such semantics, the meaning of 'Jane shaved Fred' would be the set of those possible worlds in which Jane shaved Fred, i.e. {w | Jane shaved Fred in w}. But to understand which set {w | Jane shaved Fred in w} is, one would have to know what 'Jane shaved Fred' means; so the theory would give us a meaning only if we knew the meaning in advance. Someone might object that the class of relevant possible worlds should be specified in some more elementary way, say as {w | Jane made such-and-such-things to Fred using such-and-such-instruments in w}, but what is relevant here from the point of view of the meaning of 'Jane shaved Fred' is the purported synonymy relation between 'Jane shaved Fred' and 'Jane made such-and-such-things to Fred using such-and-such-instruments', not the possible world stuff. Possible worlds do an interesting work only in the context of language as a whole (or a nontrivial part of it); again they help to |
bring to light certain structural properties of language, not
meanings of individual expressions. And what holds for possible worlds, holds for any more
sophisticated apparatus, including Tichý's one. Tichý considers language as a code and linguistics as code-cracking. But this would mean that linguistics should take an expression and put it into his theory side by side with what it encodes. But this is, as we have just seen, simply impossible - the only thing that can be the subject matter to linguistic theory are - ultimately - relations between expressions, not relations of expressions to things. I do not deny that it sometimes is useful to view language as a code; many aspects of language, relevant especially for the methodology of science, can be made intelligible in this way. But there are other aspects of language which escape such an approach, and if linguistics is to be a real, full-fledged theory of language, it may make use of the language-code picture, but it cannot be captive of it. It is sometimes useful to view language as a code; but sometimes it is much more useful to view it, as Wittgenstein did, rather as a kind of toolbox. If I view language simply as a code, how can I make sense of the fact that each morning, when meeting my friends and colleagues, I use the expression "Good morning"?(4) If we, on the other hand, view language as a peculiar way of human behaviour and linguistics as a theoretical reconstruction of this behaviour, i.e. as the enterprise of finding rules which would "account for" the behaviour (i.e. which would generate the same patterns of behaviour as those factually observable), then we reach a picture essentially different from that put |
JAROSLAV PEREGRIN, IS LANGUAGE A CODE ? |
77 |
| forward by Tichý(5). We do
not assume that language has its origin in baptizing things - an obscure assumption the
possibility of justification of which is quite unclear. Certainly, we might try to base
the rules of our theoretical reconstruction of language on relating expressions to some
entities, but there would be no way to consider this move as "the discovery of the
true nature of language"; we would have to consider it as only one approach to
linguistic theory, which can account for some aspects of linguistic behaviour, not for
others. This vantage point allows us to see many other points of Tichý's criticism as unwarranted. First, Tichý takes the fact that we are able to use and understand the basically unlimited stock of expressions of our language to mean that using language means applying rules. I think that such a view blurs the important difference between behaviour that consists in an explicit following of rules and behaviour that is merely "regular". Do we really "explain the fact that a sentence which has never entered anyone's head can nevertheless be meaningful" by saying that meanings are assigned to expressions by means of general rules? What can such an explanation mean? If this means that we speakers really consciously use rules to assign meanings to words, then it is clearly false; if it means that rules are somehow existent in the heads of speakers, e.g. that they can be found somewhere among the neurons of the brain, then it is unwarranted; and if it means neither of this, then it is only another way of saying what it purports to explain. |
Second, we can also hardly understand Tichý's point about
the relationship between syntax and semantics. Tichý questions the possibility and
reasonability of dividing syntax from semantics; but the study of these two aspects of
language clearly can be separated quite smoothly: syntax is about which
expressions speakers use, whereas semantics is about how they use them. This
division is, indeed, artificial, but I do not see in what sense it is problematic. Tichý
claims that "a theory of language cannot explain why a sentence like 'Jane
shaved Fred' takes the shapes it does without reference to the fact that it serves to
convey a message whereby a definite relationship, that of shaving obtained
between two individuals, Jane and Fred'". However, I do not think that a theory of
language should explain something like that, and I do not think that any linguist believes
that it should. In fact, it is not clear in what sense the shape a sentence takes is
capable of being explained. To say "'Jane shaved Fred' takes the shape it does
because it serves to convey a message whereby the relationship of shaving obtained between
two individuals, Jane and Fred" is nothing else than to say that "'Jane shaved
Fred' takes the shapes it does because it serves to convey a message that Jane shaved
Fred" and thus resembles the explanation - borrowing a metaphor from Richard Rorty -
that opium makes you sleepy by talking about its dormitive power. Third, Tichý's claim that it is absurd to deny, as some philosophers do, the objective existence of meanings, is also hardly warranted. Tichý claims that to speak about meaningful expressions and to say |
JAROSLAV PEREGRIN, IS LANGUAGE A CODE ? |
78 |
| there are no meanings is a nonsense similar to speaking about Fred being a car-owner and to declare that there are no cars. To speak about owning cars and to deny the existence of cars indeed is absurd, but this is not a metaphor adequate to the issue of meaning. I think that a one can speak about Fred having headache and to claim that there are no headaches, but only aching heads (or, using Quine's more emphatic example, to speak about Fred doing something for the sake of somebody and to deny that there are sakes). If we take the view of language we did, then we can speak about being meaningful without speaking about meaning - to be meaningful primarily is not to be linked to an object, but rather to be part of a system of tools used in a certain way. Speaking about meanings is then only a way of accounting for the primary situation, in the similar sense as speaking about headaches is a way of accounting for people's aching heads. It is, of course, natural and plausible to speak about meanings (as we do about headaches), and in this sense meanings can be said to exist - but one of the traditional businesses of philosophers is finding out the minimal stock of items which our worlds must be considered to consist of (viz Occam's razor), and if it is the | case that we could always translate any talk concerning
meanings into a talk about meaningful expressions, then it makes clear sense to say 'there
are no meanings over and above their embodiment' (in the same sense as it would make sense
to say 'there are no headaches over and above aching heads'). To conclude, I must once
more stress the basic point of my criticism: I think that a theory of language based on
the view of language as a code, a theory such as the one developed in detail in Tichý's
book and papers, is both respectable and relevant for linguistics. The trouble starts only
when Tichý begins to insist that this is the "only true way" to view language,
that language is a code, and that the theory can thus "explain"
language over and above accounting for some of its regularities. Majority of linguistic
utterances factually made by people can simply not be made sense of from the
code-perspective, and if we want to really understand language, we need a much more
general perspective. |
JAROSLAV PEREGRIN, IS LANGUAGE A CODE ? |
79 |
| NOTES 1. And to see this we need not buy such extreme sceptic views of science as those put forward by such philosophers as Feyerabend or Foucault. 2. In this point Tichý's criticism is in fact akin to the well-known criticism of J.Katz's approach to semantics due to David Lewis. 3. It is, of course, hard to speak for all linguists; but I think that the example I use is a fair one, because it is fostered not only within the Prague School traditon which I consider myself related to, but also (in a more or less similar shape) within several trends based on the Chomskian phrase structure approach. 4. This is, of course, an extreme example, and it might suggest the idea that expressions function "primarily" in the encoding way, and exceptionally also "secondarily" as "empty means" of social contact. However, and this is the crucial point of the late Wittgenstein, such an idea would mean overlooking the factual vast diversity of ways we use language; there is no way to tell the "primary" from the "secondary" except taking a limited stance implied by a limited purpose, such as the building of the methodology of science. |
5. The usual objection to this approach
is that it enmeshes semantics into pragmatics. But what is the boundary between semantics
and pragmatics? To use the usual answer that semantics is about the relationship between
expressions and things, whereas pragmatics is about the relationship between expressions
and speakers who use them, is to beg the question: for given the totality of linguistic
utterances, what could be the criterion of deciding which part of them, or which aspects
of them amount to relations to things and which not? We can, of course, sort out a realm
of language which can be plausibly accounted for by treating expressions as related to
things; but if we do this, then we could hardly restrict ourselves to this artificially
delimited realm if we want to understand the general nature of language.
REFERENCES Pavel Tichý: The Scandal of Linguistics, From the Logical Point of View 3/92,
pp. 70-80 Jaroslav Peregrin |