Pragmatics is the study of what speakers mean when they say something, and how hearers understand it. It is different from semantics, which is the study of the meaning inherent in a sentence. Part of the meaning of a sentence is constant, and comes from the words used in it and how they are arranged. But this is not enough to determine what a speaker means on a particular occasion of utterance. The hearer has to use the context, some common-sense assumptions, and cues about the speaker's intention in order to make inferences about their meaning.

A simple example is We went there on Tuesday. Semantically, the word we means the speaker plus one or more others, and there means a place at some distance from the speaker. Each time this sentence is uttered, the reference might change: a different group of people, a different place. The hearer probably knows who's speaking, but they also need to know or work out from context who the other people in we are: partner, family, school party? They also need to be able to get what place there is. In this sentence already these two tasks of reference assignment are different. The word there almost certainly means somewhere that's already been mentioned, or is being pointed out. But the speaker might not have mentioned any other people, and it might be up to the hearer to infer that the speaker usually means their family when they say 'we' went somewhere. Or it could depend on the place: if there is an amusement park then we might mean the whole family, but if it's a theatre it might mean only the adult partners.

More subtly, Tuesday doesn't name a single day. It is in its own way indexical, like today or there, it's just that not every day can be referred to as Tuesday. The hearer has to enrich the speaker's utterance and infer that they went there last Tuesday, since this is normally what Tuesday means in most contexts. But if the speaker was describing a visit to a country house last year, the point of their story might have been that it was closed on Tuesdays and they hadn't realized it. This information might already be available to the hearer: hearing We went there on Tuesday they can infer that We couldn't get in.

Inference and implication

Pragmatics is the study of how all these contextual and inferential factors enable us to make sense of utterances. I've been using the terms 'speaker' and 'hearer' but of course all these applies to writing and Sign language just as much as to speech. If you find a sticky note on your computer monitor that just says Three times!! you will probably understand exactly what it means, and what you're supposed to do about it. It wouldn't be there unless the person who wrote it intended you to understand it, and to derive the right conclusion about what they wanted you to do, or make you feel. Communication is usually a successful affair: we do pick up on most cues, even subtle ones, and do mostly make correct judgements about what other people are thinking.

Of course, there is no guarantee in any of this. But the difference between semantics and pragmatics is that the semantic part of language is a code that puts in quite definite meanings, as far as it goes. The pragmatic component of understanding is not deductive or encoded. It is based on non-demonstrative inference, and seeks the most likely interpretation; or, if the cues are right, a superficially less likely interpretation. We don't always say what we mean in the clearest way.

The border between semantics and pragmatics is debatable: you can distinguish overt from covert communication, explicit from implicit, and what is said from what is implied and what is communicated. The example I've been using so far is fairly explicit: after resolving any ambiguity, working out who and what is being referred to, and enriching vague words, you have a clear statement that We went there on Tuesday, and this might be enough for the hearer to understand exactly what the speaker meant.

Not just words but intonation contributes to meaning: the speaker might be denying that they went on Wednesday. So far everything that has been 'said' is grammatically or lexically or phonetically encoded: we can tell people how English works, and from this they can work out what is 'said'.

However, the speaker might have meant considerably more by it. Depending not only on how it's said (such as hesitantly, nervously, warmly, or with shock), but also on common knowledge or prior assumptions or previous speech between these individuals, it could be an accusation that the hearer failed to turn up: or a warning that the speaker knows the hearer is lying about having been there all week: or a recommendation that it was worth going to again: or an invitation to go there. There is no way you can say any of these are 'encoded' by conventional means in the words or even the intonation, but it is quite normal for speakers and hearers to understand these implications.

Suppose John says 'Would you like some coffee?' and Mary answers 'Coffee would keep me awake.' She hasn't replied yes or no, but it clearly means yes (if she's trying to study for an exam) or no (if she needs some sleep). We don't know which, but we do know that John should be able to work out which Mary meant: she wouldn't have answered him like that if she hadn't expected him to be able to infer her meaning. A very large part of ordinary language use works like this.

History

Pragmatics began as a concern of philosophy and is now incorporated into the mainstream of linguistics. The main contributors at first were philosophers: Wittgenstein, Austin, Searle, and Grice. In the early twentieth century Wittgenstein developed his atomistic doctrine of meaning: sentences were pictures or models of things in the world. This was taken up by the logical positivist school, who declared sentences meaningless if they didn't state some verifiable fact. In the 1930s pragmatics began as a reaction to these doctrines. Wittgenstein renounced his former position and showed that utterances could be used to do many different things, and for a large class of sentences the meaning was in the use.

J.L. Austin distinguished the sentence, the static grouping of words, from the utterance, the particular use of it on one occasion. He also showed that, apart from obvious non-statements such as questions and commands, there were many sentences that looked like statements but were actually doing quite different jobs: these performatives included promising, naming, judging, marrying, and many others. When you say I promise to be there you're not just describing yourself as promising, you're actually doing the promising. In a marriage ceremony or court verdict, saying certain words counts as performing a specific act. This speech act theory was developed by John Searle after Austin's early death.

H.P. Grice made the major contribution of showing that a lot of communication is not in the words. The words give us encoded information about what the speaker's intention was. The speaker intended to tell us something, and intended us to recognize that intention. The words themselves can fall far short of the full meaning. He said there was a Cooperative Principle in communication: give your hearers just the right information that is useful at that point in the conversation, for the current purpose of that conversation. This principle unfolds into Grice's maxims, of sticking to the truth, of saying not too much and not too little, of saying it in a clear and brief way, and of making what you say relevant.

A lot of what we say violates these maxims, and Grice said that when this happens we look for some ulterior intention the speaker must have had, in order to save the assumption that the Cooperative Principle was still being followed. He called this implicature. So irony and metaphor aren't literally true, evasive speech doesn't say as much as the speaker knows, and if something doesn't seem relevant we look for a connexion that could make it so. The hearer infers what the speaker intended by using that form of speech rather than the simplest and most direct.

Dan Sperber and Deirdre Wilson are an anthropologist and a linguist who have developed Relevance Theory, one of the main modern approaches to pragmatics. They postulate that humans have a general cognitive principle of seeking the most relevant information to enable them to update their beliefs and plan their actions. They call the quantity we seek to maximize 'relevance': it varies with the amount of new information we're given, and inversely with how difficult it is to process it or get access to existing assumptions it's relevant to.

They also say we have a specific communicative principle of relevance, which is that we can assume that overt communication, whether speech or some other act of attracting someone's attention to something, is intended to be relevant, and not just relevant but significantly relevant: when a speaker says something, it's intended to be worth the hearer's effort to process it. So a speaker could say something direct: that's the cheapest and simplest way of communicating. When instead they choose a roundabout or metaphorical or apparently unconnected thing to say, they must have intended something extra in it to make it worth the hearer's extra effort.