2014 : WHAT SCIENTIFIC IDEA IS READY FOR RETIREMENT?

dan_sperber's picture
Social and Cognitive Scientist; CEU Budapest and CNRS Paris; Co-author (with Deirdre Wilson), Meaning and Relevance; and (with Hugo Mercier), The Enigma of Reason
The Standard Approach To Meaning

What is meaning? There are dozens of theories. I suspect however that little would be lost if most of them were retired and the others quarantined until we have had a serious conversation as to why we need a theory of meaning in the first place. Today I am nominating for retirement just the standard approach to meaning found in the study of language and communication.

There, "meaning" is used to talk about (1) what linguistic items such as words and sentences mean, and (2) what speakers mean. Linguistic meanings and speakers' meanings are quite different things. To know a word is to know what its meaning or meanings (if it is ambiguous) are. You acquire this knowledge when you learn to speak a language. You also acquire the ability to construct the meaning of a sentence on the basis of the syntax. The meanings of words and the contribution of the syntax to the meaning of sentences are relatively stable linguistic properties that vary over historical time and across dialects.

A speaker's meaning on the other hand is a component of an individual intention to modify the beliefs or attitudes of other people through communication.

What justifies, or so it seems, using the same word 'meaning' for these two quite different kinds of phenomena—a linguistic-community-wide stable feature of a language vs. an aspect of a social interaction—is a simple and powerful dogma that purports to explain how a speaker manages to convey her meaning to her audience. She does so, we are told, by producing a sentence the linguistic meaning of which matches her speaker's meaning. The job of the addressee then is just to decode.

Alas, this simple and powerful account of how we use linguistic meanings to convey our speakers' meanings is not true. This much is actually obvious to all students of language. The issue is: how far is this from the truth?

Take an ordinary sentence, say, "She went". Your competence as an English speaker provides you with all the knowledge of that sentence meaning that you need to make use of it either in speaking or in comprehension. This however does not come near to telling you what a speaker who utters this sentence on a given occasion might mean. She might mean that Susan Jones had gone home, that the cat had one day left the house and had never returned, or that the RMS Queen Mary 2 had just left the harbor. She might mean that the neighbor carried out her threat to go to the police; or, ironically, that her interlocutor had been a fool to imagine that their neighbor would carry out that threat. She might mean metaphorically that Nancy Smith had, at some point, wholly ceased paying attention. And so forth. None of these meanings is fully encoded by the sentence; some are not even partially encoded. That much is true not just of "She went" but also of the vast majority of English sentences (arguably of all of them). Linguists and philosophers are aware of this general mismatch between linguistic and speaker's meaning, but most of them treat it as if it were a complication of limited relevance that can be idealized away or left to be investigated by pragmatics, a marginal subfield of linguistics.

The dogma, then, comes with an annotation: the basic coding-decoding mechanism that makes communication possible is quite cumbersome. Using it involves being wholly explicit. Luckily, there is a shortcut: you can avoid the verbosity of full explicitness and rely on your audience to infer rather than decode at least part of your meaning (or all of it if you use, for instance, a novel metaphor).

There are two problems with this dogma. The first is that the alleged basic mechanism is never used. You never fully encode your meaning. Often, you don't encode it at all. The second problem is that, if we are easily able to infer a speaker's meaning from an utterance that does not actually encode it, then why, in the first place, do we need the alleged basic encoding-decoding mechanism that is so unwieldy?

Imagine a tribe where people who want to go from their valley to the sea always follow a well-trodden path across a low mountain pass. According to the tribe's sages however, this path is just a shortcut and the real way (without which there couldn't even have been a shortcut) is a majestic road that goes straight up to the top of the mountain and then straight down to the sea. Nobody has ever seen that road, let alone travelled it, but it has been so much talked about that everybody can visualize it and marvel at the sages' wisdom. Linguistics and philosophy are the home of many such sages.

Most of the time, semanticists start from the dogma I have just criticized. They provide elaborate, often formal analyses of linguistic meanings that match the contents of our conscious thoughts. Are linguistic meanings really like this? Only a minority of researchers is exploring the idea that they might be a very different kind of mental objects. Unlike beliefs and intentions, linguistic 'meanings' may be just as inaccessible to untutored consciousness as are syntactic properties. They must, on the other hand, be the right kind of objects to serve as input to the unconscious inferences that achieve comprehension.

Pragmaticists and psycholinguists should, for their part, acknowledge that the meanings actually conveyed by our utterances may be not at all like individual sentences written in our minds in the 'language of thought', but rather like partly clear, partly vague reverberating changes in our cognitive environment.

The old dogma that linguistic meanings and speakers' meaning match denies or discounts a blatant gap. This gap is filled by intense cognitive activity of a specifically human kind. Let's retire the dogma and better explore the gap.