My approach to the study of language is based on the assumption that knowledge of language can be properly characterized by means of a generative grammar, i.e. a system of rules and principles that assigns structural descriptions to linguistic expressions. On this view, the basic concepts are those of 'grammar' and 'knowledge of grammar'. The concepts of 'language' and 'knowledge of language' are derivative: they involve a higher level of abstraction from psychological mechanisms and raise additional (though not necessarily important) problems. Of central concern, from this point of view, will be to determine the biological endowment that makes it possible for a grammar of the required sort to develop in human beings provided that they are exposed to some appropriate body of experience. This biological endowment may be regarded as a function that maps a body of experience into a particular grammar. The function itself is commonly referred to as universal grammar (u.g.) and can be expressed, in part, as a system of principles that determine the class of accessible particular grammars and their properties. Recent work suggests that u.g. consists, on the one hand, of a theory of so-called core grammar and, on the other, of a theory of permissible extensions and modifications of core grammar. Given the intricate internal structure of u.g., it can account for the superficially highly diverse grammars and languages that do in fact exist. Thus, what appear to be quite different systems of knowledge may arise from relatively little experience. A number of subsystems of u.g. have now been explored, each with its distinctive properties and possibilities of variation. Some current proposals concerning these systems are sketched, and some consequences considered with regard to the nature and acquisition of cognitive systems (including systems of knowledge) more generally.