The paper begins by observing that the notion of what a language consists of is problematical, reflected in one scholar's comment that, for the interpreter of texts, “language appears to begin where analytical grammar leaves off”. Section 2 describes ‘speech act formulas’ as conventional bundles of eight or nine different features, including several that are not normally considered part of grammar or lexical items, such as discourse function, idiomaticity constraints, special ‘musical’ features such as voice quality and volume, and body language. Section 3, viewing the period before 1970, asks what place was given to speech formulas in analytic grammars, on the one hand, and in treatments of language by conventional lexicographers and other humanists (scholars chiefly interested in language as an expression of human affairs), on the other. Section 4 comments on the circumstances that led some linguists and lexicographers in the 1970s to treat speech formulas as playing a central rather than a peripheral role in linguistic competence. The final section asks whether in recent decades ‘usage-based’ or ‘discourse-based’ approaches, often drawing on large electronic corpora, have led linguists to modify their views of what languages are, or whether most are still working with a grammar-and-lexicon centric model. It seems that certain methodological and theoretical biases act as conservative forces, keeping linguists focused on grammatical and lexical form, while paying relatively little attention to the full complexity of speech formulas and their role in fluency, idiomaticity, coherence, appropriateness, wit and other highly-valued facets of ordinary language use.