This article synthesises progress made in the description of spoken (especially conversational) grammar over the 20 years since the authors published a paper in this journal arguing for a rethinking of grammatical description and pedagogy based on spoken corpus evidence. We begin with a glance back at the 16th century and the teaching of Latin grammar in England, with its emphasis on speaking the target language. Later grammars were dominated by written standards, a situation that persisted till the 20th century, when recording technology and spoken corpora enabled new insights into the grammar of everyday speaking. We highlight those insights which especially challenge grammars derived only or mainly from written sources. We evidence the view that conversational grammar is non-sentence-based, co-constructed and highly interactive, and that it poses questions concerning metalanguage. We briefly review debates concerning spoken grammar and ELT/ESL pedagogy. We then consider 21st-century Internet technologies and e-communication, and implications for the spoken/written grammar distinction, arguing that description and pedagogy may need to undergo further rethinking in light of the multi-modality which characterises e-language. SOME HISTORICAL NOTES Why this article? A quarter of a century ago, English language pedagogy had entered a new era, one in which the homogenous notion of 'proficiency' had been replaced by skills-based, communicative approaches that included the speaking skill. Amid this revolution, grammar remained largely unscathed, clinging to the models and descriptions beloved of the old era. Against this background, in 1995, we published two papers advocating a greater role for the grammar of everyday speaking in the teaching of English as a second or foreign language, including one in this journal (Carter and McCarthy 1995; McCarthy and Carter 1995). We offered evidence from spoken data that everyday conversations manifested common grammatical phenomena that were marginalised in description and neglected in pedagogy. Twenty years on, we now aim to synthesise interim achievements, and to ponder where the study of spoken grammar and its applications might take us in the future. Many of the arguments against the incorporation of spoken grammar into second language pedagogy have been