Speech Acts

Speech-act theory was first put forward by the Oxford philosopher J.L. Austin. Austin wasn’t a linguist, but he was interested in trying to understand philosophical ideas by analysing language (this was the ‘ordinary language’ school of philosophy, which flourished in Oxford in the years after World War Two). Austin’s branch of philosophy was ethics: the study of morals, goodness and how to lead a good life. So, for example, he wanted to understand what was involved in keeping and breaking promises. He quickly realised that giving the truth conditions for a sentence like I promise I’ll give you your money back doesn’t tell you much about the ethics of keeping a promise: what a promise really is, i.e. what the word promise really means.

Austin made a distinction between performative and constative sentences. Constatives are the kinds of sentences we’ve been looking at in the previous chapter: sentences that say something that might be true or false, like It’s raining, Clover ate a mouse and All mammals sleep, etc. As we’ve seen, truth-conditional semantics seems to work quite well here. Performatives, on the other hand, don’t so much say something as do something. Examples are I promise to pay you back tomorrow, I name this ship the USS Enterprise and I solemnly swear that I am telling the truth, etc. Giving the truth conditions for these sentences seems pointless. Instead, to understand what they really mean, you have to understand what they do, and how they do it.

To do this, Austin further distinguished what he called the locutionary act of saying a sentence (that is, actually uttering it) from the illocutionary act, this is what the sentence is used to do. The illocutionary act carries illocutionary force, and that’s what the utterance of the sentence does. The illocutionary force of a performative is what it’s meant to do. So, the illocutionary force of a promise is that, by uttering the sentence under the right conditions (these are important, as we’ll see right away), you actually make a promise to someone to do something. Performatives perform speech acts, and speech acts have illocutionary force. The illocutionary force of I hereby pronounce you husband and wife is to marry a couple, for example.

Alongside their illocutionary force, speech acts have felicity conditions, which specify what makes a speech act work, and perlocutionary effects, which are the result of the speech act. So the felicity conditions for promising are, for example, that I have to be capable of fulfilling the promise. There’s no point in saying to someone – however much you might like to – I promise you all the money in the world, immortal life and a planet of your own. The reason is that we can’t fulfil the promise, and so the basic felicity condition isn’t met. For the marriage ceremony, there’s no point in most of us saying to some random couple we might know I hereby pronounce you husband and wife. In order to carry out the speech act, you have to have the civil or religious power to marry people; obviously you also need the people, and they need to want to get married. All of these are felicity conditions.

The perlocutionary effect of a promise is that you are under the obligation it expresses; you’ve got to do what you promised to do or you’re a bad person, guilty of the misdemeanour of breaking a promise and, in principle, open to whatever punishment that might exist in your society or legal system. The perlocutionary effect of the marrying speech act is that the couple so married are stuck with each other until death or divorce do them part.

Here’s another example, just to get these ideas clear:

The speech act carried out here is the act of naming the starship. That’s the illocutionary force: by saying (1), given the right felicity conditions, you actually give the vessel its name.

The felicity conditions in this case might be a bit demanding. First, you need a nice, newly-built starship (not so easy to find). Second, you need to have the power to utter (1), perhaps invested in you by the Galactic Federation (or perhaps not). Third, you need to accompany the naming act with a conventionally appropriate gesture such as breaking a bottle of champagne on the warp-drive boosters. And so on. The perlocutionary effect of (1), if all the felicity conditions are fulfilled, is that the Enterprise is then a starship in good standing, able and ready to boldly go where no man has gone before.

So, if you walk into your living room one quiet evening and say (1) in front of your family and friends, you haven’t named a starship. At best, you’ve made a fool of yourself; at worst, they’ll call the people in white coats. But the truth conditions of (1) are exactly the same whether the speech act is successfully carried out or not.

There are lots of speech acts. In fact, the more you think about them, the more there are: persuading, asking and ordering are among the really obvious ones. So, let’s take asking questions. The felicity conditions for asking a question are, at least, that you don’t know the answer (hence Do bears shit in the woods? isn’t a real question, since in our culture everyone is assumed to know the answer), that the person you’re asking has a reasonable chance of knowing the answer (so What is the meaning of life? is not a reasonable question to put to a non-guru or a non-semanticist). The illocutionary force is that the question is successfully put: we could call this interrogativity. The perlocutionary effect, if all goes well, is that you’ll get the answer.

One important point here is that questions – or, more precisely, interrogatives – have a special syntax in many languages. In English, as we briefly saw in Chapter 4, they involve ‘inverting’ the auxiliary in front of the subject NP. But the illocutionary force of a question and interrogative syntax are distinct things, although they often go together. We can use a syntactic interrogative to give an order, as in Why don’t you shut up? Conversely, we can ask a question without interrogative syntax, as in I heard you’re getting married, which, in the right context, could naturally be taken as asking someone whether they’re getting married.

What’s important about speech acts is that they show how language can be used, in the right contexts, with the right people, to do more than its structural elements on their own seem to mean. Speech acts are a great example of how speaker meaning adds a further dimension to language, and greatly adds to its already formidable expressive power. What’s also important is that people have to cooperate to make speech acts work. Most of the felicity conditions involve some kind of mutual knowledge; if I ask you a question or make you a promise in such a way that the speech acts actually have a chance of working, I have to recognise your role in the felicity conditions, illocutionary force and perlocutionary effects, and you have to recognise mine. It’s impossible to perform speech acts in a vacuum. These ideas about implicit co-operation and mutual knowledge are even more central to the notion of implicature, which is the next topic.

Implicatures

The notion of implicature was put forward by another Oxford philosopher connected to the ordinary language school: H. Paul Grice. Implicatures (a term Grice invented) are forms of reasoning we all use all the time in normal linguistic interactions, but they are not logical deductions, inferences or entailments, in the sense that they are not captured – and should not be captured – by the rules of a logical system like propositional or predicate logic.

As a philosopher, Grice was interested in human rationality, and, in particular, how people co-operate reasonably in everyday life, particularly in exchanging information. His main idea was the Co-operative Principle, which divides into various conversational maxims. There are four of these, as follows:

A natural initial reaction to this is that either it is a pretty poor theory of how people communicate, or some sort of philosophical pie in the sky. It’s absolutely obvious that people aren’t truthful, they witter on, they go off the point and they don’t express themselves clearly. But, and this was Grice’s point, and this is the clever part: people flout the maxims all the time, but since everyone has an awareness that the maxims should be obeyed, if I recognise that you’re flouting a maxim, I put this down to a communicative intention on your part that I should recognise this, and so I do recognise this and this allows you to achieve a certain communicative effect.

A simple example comes from metaphors. If I say to you, ‘You’re the cat’s whiskers’, this is obviously not true. In straight truth-conditional terms, the sentence is false and that’s that. But why would I sit around uttering false sentences? You assume, first, that I’m not insane (ok . . . ). Second, you assume that I’m not a liar (of course people do tell lies, but communication cannot proceed at all unless a basic assumption of veracity is made; this is one of the reasons why lying is bad). So, if I’m sane and truthful, why am I talking nonsense? Why am I saying something which is patently untrue? This is where you recognise my intention to communicate something other than what I actually said, thanks to the Co-operative Principle. I’m obviously violating the Maxim of Quality by saying something false, and I’m violating the Maxim of Manner by saying something obscure about cat-parts. So you recognise my intention to do that (note that if you assume I’m lying or mad, you implicitly abandon the Co-operative Principle; it is a principle of rational interaction and rationality involves good faith), and you recognise my intention that you will recognise my intention. So you conclude that I must be trying to say something more, i.e. that I think you’re pretty special and nice (since these are general properties of cats’ whiskers, by a vague cultural convention).

The really interesting thing about all of this is that it relies on each of our abilities to recognise the other’s intentions as they violate a maxim. I have to recognise the contents of your mind – in a way, I’m reading your mind with the help of the Co-operative Principle, and you’re reading mine. This ability to think about – and recognise – someone else’s intentions, thoughts and beliefs, and to recognise that they might be different from your own, is known as Theory of Mind. Each of us actually has an intuitive theory of what other people’s minds are like: the sorts of thoughts, beliefs, desires and intentions others have. We readily – almost unthinkingly – attribute minds of this kind to other humans, and sometimes to pets and computers (‘this bloody thing doesn’t want to save my file!’, which is of course nonsense as computers no more have desires than tables or parsnips do). Babies, up to about age one, and some people with autism may have impaired Theory of Mind: they can’t recognise that the contents of other minds might be different from their own.

However, the kind of reasoning about others’ intentions that’s involved here goes like this: ‘My interlocutor said something they know to be false, and they expect me to recognise that, and that I will recognise that they intended me to recognise that and they expect me to conclude on this basis that they really meant . . . ’. All of this involves some syntactically complex reasoning, arguably involving structures (of thought, not necessarily of actual sentences) like those we saw in Chapter 4 using recursive application of PS-rules. There could be a link between Theory of Mind and the fact that our minds can employ recursive syntactic rules.

Let’s look at some more examples of implicatures in action. Grice distinguished three types of implicature: conventional implicatures, generalised conversational implicatures, and (particular) conversational implicatures. The last kind is what we’ve just seen: an implicature is computed ‘on the fly’, using the Co-operative Principle. We really do do this all of the time.

We can illustrate conventional implicature with but. Compare the following sentences:

In propositional logic, we can obviously write (2) as p & q, p being ‘Fred is kind’ and q being ‘Fred/he is handsome’ (let’s ignore the business of figuring out who the pronoun refers to). So clearly (2) is true iff both conjuncts are true. No problem there.

What about (3)? How would we render it in propositional logic? This too has to be p & q. (3) is also true iff both ‘Dinsdale is vicious’ and ‘Dinsdale is fair’ are true. But there’s an obvious difference between but and and (which becomes quite clear if you swap them around in (2) and (3)). The difference is that but comes with a conventional implicature that there is a contrast of some kind between the two conjuncts. We can roughly capture this by glossing (3) as ‘Dinsdale is vicious and, despite this, he is fair’. In (2), no such extra gloss is needed.

Generalised conversational implicatures arise in what are also known as indirect speech acts. These are speech acts whose illocutionary force is, in a way, disguised, and some reasoning via the Co-operative Principle is needed to figure it out. But the cases are so common that these are not done ‘on the fly’, but are quite general (at least in a given culture). A classic example is:

This sentence has the syntax of a question (the auxiliary can precedes the subject NP you). The auxiliary can means, roughly, ‘to be able’. But (4) is not normally taken as a question about an individual’s salt-passing capacities: compare Can you swim the Channel?, which normally would be taken as a question about someone’s abilities. In fact, if I answer (4) by saying ‘Yes thanks; I’m an able-bodied adult and as such quite capable of lifting and moving a salt cellar in your general direction’, and do nothing, you’d understandably be ticked off (and your dinner would remain inadequately seasoned). On the contrary, if I respond by saying nothing at all, or saying ‘Of course’ or ‘Here!’, and actually passing you the salt, I’m doing what you wanted.

How does this work? Under normal conditions, there’s not much point in asking someone if they are capable of carrying out physically undemanding actions that are obviously feasible for them (contrast the reaction to (4) if I’m recovering from a broken arm and you’re a physiotherapist). So in asking the question, I’m asking a question whose answer is obvious; in speech-act terms, the question doesn’t meet the felicity conditions for interrogative illocutionary force. I recognise this, and so realise that your intention was not a request for information but rather a (polite) request for me to do something. Since I obviously can pass the salt, you’re asking me if I will pass the salt. I recognise your intention for me to recognise this in virtue of your deliberate flouting of the Maxim of Quantity (no point asking a question whose answer is obvious), and, duly, pass the salt. This is a generalised conversational implicature since the sentence in (4) is commonly used in exactly this way: the implicature isn’t computed ‘on the fly’. Example (4) is an indirect speech act, whose illocutionary force as a request is computed as just described. The basic felicity condition is that it is trivially possible for me to pass the salt (hence the violation of the Maxim of Quantity), the illocutionary force is that of a request and the perlocutionary effect is that you get the salt.

A very important kind of implicature are scalar implicatures. These have to do with scales of various kinds, in particular, logical scales. A famous example has to do with numbers. In normal circumstances, in simple positive declarative sentences, using a number to modify a noun logically entails that the same sentence is true for all positive numbers lower than that number. For example, look at (5):

If (5) is true, then it’s also true that 74,999 people came to the match, 74,998 people came to the match and so on. Now, look at the following little dialogue:

1. 1. A: How many people came to the match today?

   2. B: Three.

B’s answer is true if 75,000 people came to the match, as we’ve just seen. But it’s obviously not co-operative! The Maxims of Quality and Quantity (sometimes it’s hard to tell which, but that needn’t concern us unduly here) require us to be as precise as possible. B’s answer is naturally interpreted by A as ‘exactly three’ (by A’s recognition that if there had been 75,000 people there, B would have said so, and hence that B must have intended to say that there were exactly three people and no more there). In simple positive declarative sentences, numbers mean ‘at least n’ logically, but ‘exactly n’ pragmatically. So I can truthfully, but not at all co-operatively, tell you I have one child, when in fact I have two. Again, if I say ‘I have one child’, the implicature is ‘exactly one’, given the Co-operative Principle.

Here’s another example of a scalar implicature. Going back to the quantifiers of predicate logic, it should be clear that universal quantification entails existential quantification. So, if all cats are wise, it follows that some cats are wise. Formally:

(Here ‘F’ just stands for any predicate). So I can truthfully say ‘Some mammals sleep’ even if I know that all mammals sleep, given (7). But the Co-operative Principle forces me to recognise that if I had meant ‘all mammals sleep’ I would have said ‘all mammals sleep’; saying ‘some mammals sleep’ causes the implicature (clearly not a logical entailment) that ‘not all mammals sleep’.

In general, what the number examples and the quantifier example show us is that the Co-operative Principle leads us to always make the strongest statement we can, where ‘strongest’ means ‘most likely to be false’. A universal statement is more likely to be false than an existential one; for the universal to be false, just one nonsleeping mammal, for example, is needed, while for the existential to be false, no mammal can sleep. It’s easier for an existential to be true than it is for a universal, and therefore harder for it to be false. Therefore, existentials are ‘weaker’ than universals, and so the Co-operative Principle leads us to state universals when we think they hold, rather than existentials, despite both being true in the same situation.

For the same reason, we naturally interpret one-way implications as biconditionals, i.e. the if of propositional logic is often interpreted as iff. We saw this in the lawn-mowing example in the last chapter (see ‘Better living through semantics (Part II)’). Your parents naturally interpret If I mow the lawn, then you give me £20 as If and only if I mow the lawn, you give me £20. This is because the iff-statement is stronger: it’s true in one less case than the if-statement (just the one where I don’t mow the lawn and you give me £20).

Similarly for the numerals. If there were roughly 75,000 people at the match (say, the 75,000-seat stadium was full) then I should say that, rather than the equally true ‘three people’. Of course, 75,000 can be an uncooperative number too: if I say ‘75,000 people live in London’ or ‘there are 75,000 humans on the planet’, I’m being just as uncooperative as when I say there were three people at the match, since in fact there are many more in both cases. Again, to say 75,000 is true, but an uncooperatively weak statement.

A final example of implicature. If a loyal husband pops out to the pub, saying to his wife ‘I’m not leaving you, dear’, the wife will probably get worried. Again, if hubby’s intention is indeed just to have an innocent pint, what he’s said is true: it’s not the case that he’s leaving his wife. But, by alluding to p (‘I’m leaving you’), he has implicated that p is at least within the bounds of possibility. The poor wife, intuitively calculating her husband’s intentions using the Co-operative Principle (applying her Theory of Mind), naturally gets worried. Under certain conditions, ¬p can implicate the possibility that p. Shakespeare’s famous line ‘Methinks the lady doth protest too much’ describes a case like this.

*

What we’ve seen here is, I hope, enough to give you an idea of what pragmatics can tell us about how language is used in order, very often, to convey much more than just ‘what is said’. Of course, this is just the sketchiest overview, but it captures the flavour of things.

It may be useful to end this chapter, and the chapters on the internal workings of language, by making a three-way distinction between utterance, sentence and proposition. The utterance is what is said in a given context, complete with implicatures, illocutionary force, etc. This is the real, everyday use of language that we all practice all the time. The sentence is the formal object, generated by PS-rules and systematically connected to a phonological and a semantic representation (we saw how to get from syntax to semantics, but not how to get from syntax to phonology). The proposition is a logical form, capable of bearing truth values. So, for Clover ate a mouse we have:

Sentence:

One implicature of this utterance is that Clover ate exactly one mouse; the logical form of the proposition doesn’t say this, it merely ‘translates’ the indefinite article a of the sentence into the existential quantifier. The existential quantifier, remember, means ‘at least one’, not ‘exactly one’. But if Clover had eaten more than one mouse, the utterance would most likely have been Clover ate some/two/fifty mice, in line with what we’ve seen here regarding scalar implicatures. There are other implicatures here: the past tense ate refers to a time before the present (i.e. the time of the utterance), but probably not to a prehistoric, Stone Age cat-eating-mouse event; it’s more likely to refer to a past time of interest and relevance to the interlocutors. I’ve left aside the whole complicated question of the syntax, semantics and pragmatics of tenses here.

There’s lots more, of course, but by now it should be apparent that our ordinary use of language in ordinary situations involves, in addition to all of the complexities of phonetics, phonology, morphology, syntax and semantics, the constant interplay of implicatures and speech acts, both based on our intuitive use of Theory of Mind in all of our interactions with each other (and sometimes with our pets and computers, too).