promise, or merely offering a prediction, depending on conventional and other social features of the situation. A crude test of illocutionary force is the ‘hereby’ criterion: one’s utterance has the force of, say, a warning, if it could fairly have been paraphrased by the corresponding ‘explicitly performative’ sentence beginning ‘I hereby warn you that . . .’.) Speech act theory interacts to some extent with semantics, especially in the case of explicit performatives, and it has some fairly dramatic syntactic effects as well. A third branch of pragmatics (not altogether separate from the second) is the theory of conversation or theory of implicature, founded in the 1960s by Grice. Grice noted that sentences, when uttered in particular contexts, often generate ‘implications’ that are not logical consequences of those sentences (‘Is Jones a good philosopher?’ – ‘He has very neat handwriting’). Such implications can usually be identified as what the speaker meant in uttering her sentence; thus (for that reason and others), what Grice calls utterer’s meaning can diverge sharply from sentence-meaning or ‘timeless’ meaning. To explain those non-logical implications, Grice offered a now widely accepted theory of conversational implicature. Conversational implicatures arise from the interaction of the sentence uttered with mutually shared background assumptions and certain principles of efficient and cooperative conversation. The philosophy of linguistics studies the academic discipline of linguistics, particularly theoretical linguistics considered as a science or purported science; it examines methodology and fundamental assumptions, and also tries to incorporate linguists’ findings into the rest of philosophy of language. Theoretical linguistics concentrates on syntax, and took its contemporary form in the 1950s under Zellig Harris and Chomsky: it seeks to describe each natural language in terms of a generative grammar for that language, i.e., a set of recursive rules for combining words that will generate all and only the ‘well-formed strings’ or grammatical sentences of that language. The set must be finite and the rules recursive because, while our informationprocessing resources for recognizing grammatical strings as such are necessarily finite (being subagencies of our brains), there is no limit in any natural language either to the length of a single grammatical sentence or to the number of grammatical sentences; a small device must have infinite generative and parsing capacity. Many grammars work by generating simple ‘deep structures’ (a kind of tree diagram), and then producing multiple ‘surface structures’ as variants of those deep structures, by means of rules that rearrange their parts. The surface structures are syntactic parsings of natural-language sentences, and the deep structures from which they derive encode both basic grammatical relations between the sentences’ major constituents and, on some theories, the sentences’ main semantic properties as well; thus, sentences that share a deep structure will share some fundamental grammatical properties and all or most of their semantics.
As Paul Ziff and Davidson saw in the 1960s, the foregoing syntactic problem and its solution had semantic analogues. From small resources, human speakers understand – compute the meanings of – arbitrarily long and novel sentences without limit, and almost instantaneously. This ability seems to require semantic compositionality, the thesis that the meaning of a sentence is a function of the meanings of its semantic primitives or smallest meaningful parts, built up by way of syntactic compounding. Compositionality also seems to be required by learnability, since a normal child can learn an infinitely complex dialect in at most two years, but must learn semantic primitives one at a time.
A grammar for a natural language is commonly taken to be a piece of psychology, part of an explanation of speakers’ verbal abilities and behavior. As such, however, it is a considerable idealization: it is a theory of speakers’ linguistic ‘competence’ rather than of their actual verbal performance. The latter distinction is required by the fact that speakers’ considered, reflective judgments of grammatical correctness do not line up very well with the class of expressions that actually are uttered and understood unreflectively by those same speakers. Some grammatical sentences are too hard for speakers to parse quickly; some are too long to finish parsing at all; speakers commonly utter what they know to be formally ungrammatical strings; and real speech is usually fragmentary, interspersed with vocalizations, false starts, and the like. Actual departures from formal grammaticality are ascribed by linguists to ‘performance limitations,’ i.e., psychological factors such as memory failure, weak computational capacity, or heedlessness; thus, actual verbal behavior is to be explained as resulting from the perturbation of competence by performance limitations. See also GRAMMAR , MEANING, SPEECH ACT THEORY , THEORY OF DESCRIPTIONS , TRUT. W.G.L.