Despite all the work that has been done on first- and second-language acquisition, we know surprisingly little about how languages are learnt, and even less about how they can best be taught. Theories come and go, assertions are plentiful, facts are in short supply. This is nowhere more true than in the area of grammar. The trouble with teaching grammar is that we are never quite sure whether it works or not: its effects are uncertain and hard to assess. If we teach rules, sometimes students manage to apply them and sometimes they don’t. Practice may have some effect, but carry-over to spontaneous production is often disappointing. If students speak more correctly as time goes by, is this because of our teaching, or would they have got better anyway? Research on methodology is inconclusive, and has not shown detectable and lasting effects, for instance, for implicit versus explicit instruction, for inductive versus deductive learning, or for separated-out study of structure versus incidental focus on form during communicative activity. Understandably, teachers are unsure how much importance they should give to grammar, what grammar they should teach, and how they should teach it. Language-teaching fashions consequently oscillate from one extreme, where grammar is given star billing, to the other, where it is backgrounded or completely ignored.