The uses of reality: a reply to

Introduction Computerized language corpora have inspired some of the most important insights in recent linguistics. They have shown us, for example, that actual language use is less a matter of combining abstract grammar rules with individual lexical items, and more a matter of collocation; that there are grammatically possible utterances which do not occur, and others which occur with disproportionate frequency; that in systematic descriptions of occurrences, grammar and lexis cannot be as easily separated as they have been traditionally, either in pedagogy or in linguistics. Ronald Carter is right to find such insights 'exciting', and his own work with Michael McCarthy on the CANCODE corpus, has added to them. As his article illustrates very well, the grammatical constructions we find in actual conversations are not always accounted for in traditional grammars.