From Encyclopedia of Contemporary Literary Theory, ed. Irena Makaryk (Toronto, 1993):

Speech-act theory, which took shape between 1939 and 1959 in the lectures and addresses of John L. Austin, a prominent Oxford philosopher of ordinary language, focuses on how speech utterances themselves perform deeds in particular contexts.

The speech act model of language offers insight into contextual use and meaning and situates speech behaviour admidst the social and institutional circumstances in which it arises. . . .

That words themselves can be deeds [is] the basic premise of Austin's theory.

From A Glossary of Literary Terms by M. H. Abrams, 7th edition (Harcourt, 1999):

Speech-Act Theory, developed by the philosopher John Austin, was described most fully in his posthumous book How to Do Things With Words (1962), and was explored and expanded by other "ordinary language philosophers," including John Searle. . . . Austin's theory is directed against traditional tendencies of philosophers (1) to analyze the meaning of isolated sentences, abstracted from the context of a discourse and from the attendant circumstances in which a sentence is uttered; and (2) to assume, in what Austin describes as a logical obsession, that the standard sentence—of which other types are merely variants—is a statement that describes a situation or asserts a fact and can be judged to be either true or false.

In [a speech act] that is not an assertion, the prime criterion . . . is not its truth or falsity, but whether or not the act has been performed successfully, or in Austin's term, "felicitously." A felicitous performance of a particular illocutionary act depends on its meeting "appropriateness conditions" which obtain for that type of act; these conditions are tacit linguistic and social (or institutional) conventions, or rules, that are shared by competent speakers and interpreters of a language. For example, the successful performance of an illocutionary act of promising, such as "I will come to see you tomorrow," depends on its meeting its special set of appropriateness conditions: the speaker must be capable of fulfilling his promise, must intend to do so, and must believe that the listener wants him to do so. . . .

. . . Austin established an initial distinction between two broad types of locutions: constatives (sentences that assert something about a fact or state of affairs and are adjudged to be true or false) and performatives (sentences that are actions which accomplish something, such as questioning, promising, praising, and so on). . . . Austin, however, drew special attention to the "explicit performative," which is a sentence whose utterance itself, when executed under appropriate institutional and other conditions, brings about the state of affairs that it signifies. Examples are "I name this ship the Queen Elizabeth"; "I apologize"; "I call this meeting to order": "Let spades be trumps," [or "I now pronounce you husband and wife."]

A number of deconstructive theorists have [called speech-act theory] a prime instance of the performative, in that it does not refer to a pre-existing state of affairs, but brings about, or brings into being, the characters, action, and world that it describes.