Language has traditionally been divided into grammar and vocabulary. Crudely, the former consisted of elements of the generative system of the language and the latter was the stock of fixed nongenerative ‘words’. Recently, this analysis has been challenged and shown to be seriously misguided from both strictly linguistic and pedagogical points of view (Lewis, 1993; Nattinger & DeCarrico, 1992; Willis, 1990).

In fact, language consists broadly of four different kinds of lexical items, the constituent ‘chunks’ of any language. Each chunk may be placed on a generative spectrum between poles ranging from absolutely fixed to very free. Although it is true that traditional vocabulary is usually close to the fixed pole, and grammar structures are frequently close to the free pole, this fact obscures the vastly more numerous and in many ways more interesting items that occur nearer the middle of the spectrum. These items may be ‘words’, or ‘structures’ in traditional language teaching terms, but, as we will see, most typically they are lexical items of types not recognised in most teaching material.

Lexical items are socially sanctioned independent units. These may be individual words, or full sentences – institutionalised utterances – that convey fixed social or pragmatic meaning within a given community. This definition clearly entails that lexical items are dependent on agreement within a particular social group; what is a lexical item in American English may not be so in British English.