J.L. Austin, answered from the texts and cited to the page.
The question is best approached by first noticing what it is not. When I determine what locutionary act someone has performed — what words they used, with what sense and reference — I have not yet touched the question of what they were doing in saying those words.1 "It is going to charge" may be perfectly clear as a locution and yet leave entirely open whether I was warning you, informing you, predicting for my own amusement, or something else again.2
That further question — what act was performed in saying something, as opposed to the act of saying something — is the question of illocutionary force.3 The range of such acts is very wide. Asking or answering a question, pronouncing sentence, making an appointment, issuing a criticism, giving a description — these are all illocutionary acts, though I would not pretend this is a clearly defined class.4
What they share is that the force of the utterance is not settled by its meaning in the sense of sense and reference; force and meaning must be distinguished, just as sense and reference had to be distinguished within meaning itself.5 "Shut the door" has a perfectly determinate meaning, and yet whether it was uttered as an order, an entreaty, or a plea remains a further question entirely.6
It is worth pressing on why this matters philosophically. For too long, problems about utterances were treated as problems about locutionary usage alone — what the words meant — and what I called the descriptive fallacy commonly arose from mistaking a problem of illocutionary force for a problem of meaning.7 Once we see that stating something is performing an illocutionary act just as much as warning or ordering, the old picture of statements as simply true-or-false descriptions standing apart from "doing" things collapses.8
The constative/performative distinction I began with was a useful starting point, but the deeper analysis shows that every utterance has illocutionary force — stating included — and the distinction between doing and saying cannot be used to mark off a special class of performatives from the rest.9
it might be perfectly possible, with regard to an utterance, say 'It is going to charge', to make entirely plain 'what we were saying' in issuing the utterance, in all the senses so far distinguished, and yet not at all to have cleared up whether or not in issuing the utterance I was performing the act of warning or not.How to Do Things with Words, pp. 105–106
It may be perfectly clear what I mean by 'It is going to charge' or 'Shut the door', but not clear whether it is meant as a statement or warning, &c.How to Do Things with Words, pp. 105–106
I explained the performance of an act in this new and second sense as the performance of an 'illocutionary' act, i.e. performance of an act in saying something as opposed to performance of an act of saying something.How to Do Things with Words, pp. 106–108
asking or answering a question, pronouncing sentence, making an appointment or an appeal or a criticism, making an identification or giving a description, and the numerous like. (I am not suggesting that this is a clearly defined class by any means.)How to Do Things with Words, pp. 105–106
I want to distinguish force and meaning in the sense in which meaning is equivalent to sense and reference, just as it has become essential to distinguish sense and reference within meaning.How to Do Things with Words, pp. 106–108
We may be quite clear what 'Shut the door' means, but not yet at all clear on the further point as to whether as uttered at a certain time it was an order, an entreaty or whatnot.Philosophical Papers, pp. 255–256
for too long philosophers have neglected this study, treating all problems as problems of 'locutionary usage', and indeed that the 'descriptive fallacy' mentioned in Lecture I commonly arises through mistaking a problem of the former kind for a problem of the latter kind.How to Do Things with Words, pp. 106–108
Surely to state is every bit as much to perform an illocutionary act as, say, to warn or to pronounce.How to Do Things with Words, pp. 139–141
whenever I 'say' anything (except perhaps a mere exclamation like 'damn' or 'ouch') I shall be performing both locutionary and illocutionary acts, and these two kinds of acts seem to be the very things which we tried to use as a means of distinguishing, under the names of 'doing' and 'saying', performatives from constatives.How to Do Things with Words, pp. 139–141