meaning

The concept of meaning is every bit as problematic as the concept of mind, and for related reasons. For it seems to be the case that it is only for a mind that some things (gestures, sounds, marks, or natural phenomena) can mean other things. The difficulties that we have understanding the place of mind in nature spill over into, and create problems for, our understanding of the phenomena of meaning. Anyone who conceives of science as objective, and of objectivity as requiring the study of phenomena (objects and relations between objects) that exist and have their character independently of human thought, will face a problem with the scientific study of meaning.

One might attempt to overcome this problem by finding a natural relation holding between things that have meanings (i.e. signs) and the things they mean (or signify). The idea that meaning is, or needs to be, founded upon a natural relation is responsible for the lure of causal, picture, and onomatopoeic theories. Cause and effect and similarity of shape are thought to be relations which hold independently of being recognized by a mind.

Plato (Cratylus, 432–5) argued against the idea that the meaning relation could be founded purely on a natural relation of similarity. No signs, he observed, can be exactly similar to the thing signified without duplicating that thing in all respects. What counts as sufficiently similar is governed by convention (nomos) and there must, therefore, be an element of convention in every use of a thing as sign.

But while it appears unavoidable that meaning should have some component of convention, it does not seem unreasonable to seek a theoretical account of the general conditions under which such conventions, established for a relatively small number of linguistic elements (a basic vocabulary), would fix the meanings of complex terms and sentences constructed out of these elements. It would be the work of that branch of the science of logic known as compositional (or recursive) semantics to determine these general conditions, and the result would constitute an objective account of how parts of a language can mean something. If one added assumptions about how it is proper to fix the meanings of the basic elements — say they had to be tied in some way to sense experience — one could test any piece of discourse for meaningfulness by applying the principles of logic to analyse the discourse until it became clear whether its elements had been properly constituted or legitimized by experience, and hence whether it was a meaningful piece of discourse.

Early efforts in this direction were made by the classical British empiricists (Hobbes and Hume), but the important advances in logical technique at the turn of the 19th century gave those attracted to this approach a new self-consciousness and self-confidence, and they emerged as 'logical empiricists' (or 'logical positivists'). These were the philosophers (A. J. Ayer, Rudolf Carnap, and Moritz Schlick) who challenged the 'cognitive meaningfulness' of metaphysics, theology, and ethics, and whose slogan was 'the meaning of a sentence is its method of verification', for the analysis of a sentence would, if carried out, also show how one could go about verifying the sentence. (See falsification.)

The logical empiricists, however, differed over the way basic elements should be tied to sense experience. One group (including Ayer, Schlick, Mach, and Bertrand Russell) saw basic elements as grounded in subjective experience; another (including Otto Neurath and Carl Hempel) did not regard it as proper to anchor scientific discourse in subjective experience, and urged instead publicly accessible outcomes of publicly accessible (e.g. laboratory) procedures as the proper foundation for cognitive meaningfulness. The approach of this second group underwent a development as a result of the realization that what constitutes a test of something in science is not independent of the constellation of theories which scientists accept. The result was the abandoning of the idea that some elements of a language can be treated as basic with regard to the verification of all other sentences of that language. Hempel proposed instead a 'translatability criterion' according to which a whole body of discourse would be regarded as cognitively (i.e. scientifically) meaningful if it could be translated into a logically regimented language whose primitive vocabulary consisted of either 'logical locutions' or 'observation predicates'. None of the sentences, however, was to be regarded as basic for the purposes of verifying the sentences of the body of discourse thus translated. Instead of individual sentences confronting experience, the body of discourse as a whole stood or fell in the face of experience.

W. V. Quine and Donald Davidson developed further the line of this second group of positivists, by shifting attention from the attempt to say what it is to be cognitively meaningful to what is involved in one person understanding another person's (or culture's) language. In each case, they maintained, the interpreter has to work as a scientist works. The interpreter has to collect data, which consists in observations of the conditions under which the people whose discourse is to be interpreted will assent to, and dissent from, certain sentences. The interpreter also has to select basic elements (by proposing 'analytic hypotheses') and has to construct a theory which assigns truth conditions to basic elements and, through this, truth conditions to all the sentences of the language in such a way as to conform to the data. (Since the development of the logical apparatus, the compositional semantics, thought to be needed for this theory building, is due to Alfred Tarski; his name is often mentioned in this context.) As in Hempel's account, which is supposed to reflect the actual practice of natural scientists, the theory constructed stands or falls by its ability to accommodate the data taken as a whole.

As Quine and Davidson acknowledged, indeed urged, this approach to interpretation in an important sense undermines the idea that sentences or fragments of sentences have a meaning, or that names have a reference. The body of data is never adequate to determine uniquely one system of interpreting a language. There is no sense to the question what a noun really refers to, or what a speaker really means. In effect we should, in order to treat language as a natural phenomenon, abandon the notions of meaning and reference, and make use only of the concept of truth.

It was in the context of this challenge to the conceptions of meaning and reference that the 'causal theory of reference' was advanced by Saul Kripke and Hilary Putnam. According to this theory, A refers, for example, to a liquid chemical composed of two parts hydrogen and one part oxygen as 'water', because A heard B, C, D, etc. refer to it by that name, and they did so because they heard others do so; and so on back to an 'initial baptism' of that stuff with the name 'water'.

Although the phrase 'initial baptism' suggests a conscious attempt to establish a conventional pattern of linguistic response (as in a genuine christening), the initial baptism need not be regarded in this way. It could be treated as an unspecified natural occurrence, which under the circumstances holding at that time established a pattern of response to all essentially similar features of the environment. The similarity of feature, which is the basis of the pattern of response, need not be understood by those conforming to the pattern, although understanding could be achieved later by scientific investigation. This account thus underwrites the possibility of saying that what Aristotle meant by 'to hudor' was H₂O, even though it is obvious that Aristotle did not possess our understanding of the composition of water. It is clear from this illustration how the notion of meaning can be conceived as a relation which obtains independently of what humans think.

To its critics, the causal theory failed to provide a historically plausible account of the mechanisms by which reference is secured and maintained. But to those sympathetic to it, it offered an alternative to abandoning the use of these notions and accepting the conclusion drawn by Quine and Davidson that meaning and reference are too indeterminate for scientific study. (A corollary of this conclusion, to the effect that mind is insufficiently determinate for proper scientific study, was subsequently drawn by Bernard Williams.)

One might also try to escape Quine's conclusions by calling into question the positivist conception of science which underpins them. Those prone to read the theoretical pronouncements of scientists literally, i.e. 'realists', brush aside the possibility, which is frequently stressed by Quine, that any body of scientific evidence, e.g. electrical phenomena, can be accounted for in a variety of incompatible ways, not just by postulating the existence of electrons. They insist that we are often justified in selecting one of the competing accounts as the best explanation, and this outlook can be applied to the phenomena of language to endorse, as perfectly respectable scientific reasoning, the idea that the best explanation of linguistic phenomena will postulate mental states as the causes of those phenomena.

Although the word 'idea' no longer figures prominently, this approach in effect returns to the position of John Locke, who held that 'words, in their primary or immediate signification, stand for nothing but the ideas in the mind of him that uses them' (Essay, iii. ii. 2). Instead of ideas, the internal states most commonly appealed to are beliefs and desires, but the approach would still fall under the strictures of those philosophers and logicians who, since Frege, have indicted such theories on a charge of 'psychologism', for such theories do indeed shift the attention of those studying meaning away from the relationship between signs and the things they signify to the relationship between the (minds of) sign users and the things they signify.

Recent developments along these lines begin with an analysis of speaker's meaning advanced by H. P. Grice. We do quite often use the word 'meaning' to label what a speaker intended or was trying to do with his words. Thus Spooner's audience recognized that he meant to refer affectionately to the monarch when he used the words 'our queer dean'. This complex intention was analysed by Grice in terms of an intention to modify the beliefs or behaviour of the audience via the audience's recognition of that intention; and the presence of such intentions, Grice held, was what distinguished the ('non-natural') meaning of linguistic phenomena from the ('natural') meaning of natural phenomena (such as in 'those spots mean measles').

Criticisms of Grice's analysis focused on the difficulty of bridging the gap between this notion of (speaker's) meaning and the notion of word (or linguistic) meaning, for which we still need an account. Spooner's words after all meant something quite at variance with his intention; his words referred slurringly to a college official. Jonathan Bennett attempted to bridge the gap by integrating Grice's analysis with a sophisticated account of convention, which was devised by David Lewis. Under Lewis's analysis, a convention need not be established by conscious agreement (something which would presuppose, and hence could not explain, the existence of language) but is nevertheless more than a mere regularity because, however it was established, it is maintained by the recognition on the part of those who conform to it that conformity to it solves a 'coordination problem', i.e. eliminates certain disadvantages which would occur unless activities were coordinated. Bennett's account in effect rested linguistic meanings on the hardening into convention of speakers' meanings.

Bennett's approach has been criticized for not providing a convincing account of how what are essentially unstructured and unrelated primitive signalling acts (acts that express communication intention without presupposing a language) come to have the structure evident in all languages, namely that by which each speech act is composed of elements which make a similar contribution to a variety of different speech acts. A second criticism called into question the sense of crediting non-language users with intentions sufficiently complex to count as (Gricean) communication intentions. The first of these criticisms rests on an important feature of meaning, its systematic nature. Different theoretical traditions give different accounts of this essential structure, but agree at least on the principle of systematic relatedness. (In the controversy over whether chimpanzees in acquiring the use of Ameslan signs had achieved linguistic mastery, a crucial point was whether they had the ability to form signs of sufficient complexity.) (See primate language.)

Compositional semantics offers an account of this structure based on the principle that the meanings of complex expression are functions of the meanings of elementary components. In fashioning the central concepts of this approach Frege had insisted that a sentential (or propositional) unit is primary to language, that singular terms and predicate expressions made radically dissimilar contributions to sentences, and that one must not ask for the meaning (Bedeutung) of a word outside the context of a sentence. Bennett's account seems in particular to lack resources to explain how a set of internally unstructured and logically homogeneous signalling acts could yield expressions with the different logical functions of subject and predicate. Frege's ideas were transmogrified in the hands of Quine, and, as the importance Frege attached to reference decayed, Quine replaced Frege's maxim about looking for the meaning outside the context of a sentence by a stricture against seeking the meaning of an expression outside the context of a whole language.

A radically different account derives from the work of the linguist Ferdinand de Saussure, who saw as relevant to meaning not only the relation between sign and thing signified, but also the 'value' of the sign (word). The value of a word is a function of the words that can in some contexts be exchanged for the word ('strike' and 'hit' are interchangeable only in some contexts) and the words that stand in contrast to it. 'Sheep' and 'mutton' are used for the animal and its meat in English, where the French use 'mouton' for both. Thus the latter could have the value of neither of the former words.

Saussure's work provided an important inspiration for the movement known as structuralism. In the hands of the structuralists the relation between signified and signifier was virtually discarded, leaving the meaning of a sign resting on the structure of values constituted by the language as a whole. Coming from a different direction, the structuralists arrived at a linguistic holism which bore a number of striking similarities to that advanced by Quine and Davidson. In neither case could one look to some definite aspect of the world outside of discourse for what an expression meant or signified. If it made any sense to ask for the meaning of an expression, one had to look for it in its connections to other expressions.

Were this a viable account of linguistic meaning, it would support the idea that linguistic mastery (the ability thought by many to be essential to the possession of a mind) can be modelled in computer programs. A program, if it has the power to generate the right syntactic structure and vocabulary, could represent the way expressions interact with each other to constrain the formation of expressions and the assent to or dissent from sentences. Computer modelling along these lines has tended to draw more on the tradition which Frege founded than on that which Saussure founded. Such computer modelling leads to a second application of the 'realist' strategy mentioned above. Instead of postulating beliefs and intentions (in the head) to account for linguistic acts, what is postulated is an internal configuration formally isomorphic to a computer program (one we have yet to write) which accounts for what is conceived of as human linguistic 'printout'. Many cognitive scientists see their task as the writing of such a program, or possibly working in conjunction with physiologists to determine which of several possible programs the human mind actually runs. (See also artificial intelligence.)

There are, however, several doubts about the aspirations of such cognitive scientists. If these aspirations are based on a thoroughgoing holism, it is difficult to see how language mastery can be acquired (as humans evidently acquired it) in stages, for the whole of language would have to be mastered in order to have mastered any part of it. It is difficult also to see how two adult human beings could communicate with one another, so long as their vocabularies did not precisely coincide. For it would be impossible to determine whether a difference of opinion was based on one person being misinformed or the two speakers at cross purposes because they in effect spoke different languages.

Computer modelling need not rest on a holistic approach, but if there is thought to be some component of the meaning of an expression which is independent of its relationships to other expressions in the language, this component would seem to have to be sought in the relation that the expression bears to things outside the language. Computer models can be extended by connecting the machinery and software to 'external sensors', so that the printout bears an appropriate relation to the world. But this expedient leaves open the possibility that different sensors responding to different external features (or even direct interference from the programmer) could produce the same internal states and the same printout; insofar as this is possible, the computer fails to model the meaning of the words on the printout. Those, such as Jerry A. Fodor, who are convinced of the adequacy of computer models have tried to make a virtue of this difficulty and have advanced 'methodological solipsism ... as a research strategy in cognitive psychology'. Others, such as John Searle, have turned these difficulties into an a priori argument against the possibility of computer modelling of human languages and against the claim that the human mind is in all essential respects a computer.

So far we have considered approaches to meaning which look either to the relation between language and the world or to the relation between language and mind. (The latter may seek to preserve the meaning relation as a fit subject for scientific study by treating mental structures or intentions as natural, possibly purely physical, phenomena.) A third approach would argue that the difficulties of the first two arise from not recognizing that meaning must be studied in the context of a (three-term) relation between mind and the world and language, where the last of these is conceived of as a social phenomenon. This was an important feature of Saussure's theory, for he insisted that 'language never exists apart from social fact', but it is also an important principle of the outlook of such diverse thinkers as Hegel, Marx, Dewey, Mead, Vygotsky, and Wittgenstein.

It is from this perspective that Bennett's programme would be criticized for assuming that non-language users could have, and recognize in others, intentions as complex as required for Grice's analysis of (even) a simple act of signalling. Thought, from this perspective, cannot exist without (or prior to) language.

According to the social behaviourists, George Herbert Mead for example, it would be acceptable to regard social animals as affecting the behaviour of one another in the following complex way: the beginning of a pattern of behaviour (a gesture) on the part of one animal elicits from another a pattern of behaviour which modifies the development of that pattern as it unfolds in the behaviour of the first animal. But to behave as a human does in such a social interaction, the second animal would have to respond not only to the gesture, but to the relationship between the gesture and the object that had acted as stimulus to the gesture, and to do so from the standpoint of the first animal. This is what is required for the second animal to respond to something as meaningful, the concept of meaning implied here being 'an individual reaction which an object may call out'. For the meaning of an object to be 'ours' we must have adapted ourselves to a comprehensive set of reactions toward it, which we can not only adopt ourselves but do so also in the role of others who can respond in that way. Meaning is not, on this account, primarily a property of objects, John Dewey observed, it must be primarily a property of behavioural responses and derivatively of the objects that enter into those patterns.

Another development of an essentially social approach to meaning will be found in the later work of Wittgenstein. Wittgenstein interpreted Frege's suggestion that a word has a meaning only as part of a sentence, as a way of saying that things cannot have names except in a 'language game' (Investigations, I. 49), a label which Wittgenstein applied to social contexts in which words are used in structured ways of interacting with the world. All three, language, language users, and the world, are bound up in this concept, and its implications were spelled out in the advice, 'Don't ask for the meaning, ask for the use.' Wittgenstein, moreover, mounted controversial and much debated arguments against the possibility of a person assigning a meaning to a word on the basis of an intention to use that word to apply to something experienced privately. The use of language to describe subjective experience is dependent on and derivative from linguistic practices in which several people coordinate their interactions with each other. (See also Wittgenstein's philosophy of language.)

Whether one treats this third approach as conforming to or repudiating the demand for a scientific approach to language depends on whether one conceives the study of human social interaction as (at least potentially) continuous with the scientific study of nature. A tradition going back to Giambattista Vico in the early 18th century regards the study of human institutions and artefacts as the province of a study which, if scientific, involves methods quite discontinuous with the natural sciences. More recent developments in this tradition insist upon the need for the imaginative re-creation of the experience of other human beings, of the need for empathy or sympathy in grasping the meaning of a text, and generally refer to their methodology under the rubric 'hermeneutics'. Not all those, however, who insist on the importance of the social in the study of meaning adopt this methodological dualism. Whether or not one does adopt a dualist approach is related in an important way to the conception one has of scientific knowledge. This in turn rests on the way one conceives the relationship between human beings (in particular their minds) and the rest of nature.

(Published 1987)

— J. E. Tiles

Bibliography
• Ayer, A. J. (ed.) (1959). Logical Positivism.
• Bennett, J. (1976). Linguistic Behaviour.
• Davidson, D. (1984). Inquiries into Truth and Interpretation.
• Fodor, J. A. (1981). Representations.
• Frege, G. (1952). Philosophical Writings. Eds. and trans. P. T. Geach and M. Black.
• Kripke, S. (1980). Naming and Necessity.
• Mead, G. H. (1964). Selected Writings. Ed. A. J. Reck.
• Quine, W. V. (1960). Word and Object.
• Saussure, F. de (1960). Course in General Linguistics. Trans. W. Baskin.
• Wittgenstein, L. (1958). Philosophical Investigations. Trans. G. E. M. Anscombe.

---