In general, studies of fixed expressions—idioms, formulae such as proverbs and catchphrases, and anomalous or ill-formed collocations—concentrate on their typological and syntagmatic properties. Attention is given to such things as the degree of their lexical and syntactic frozenness, or their transformation potential; and even the primary characteristic of idioms, their non-compositionality as lexical units, may be seen as a matter of the interpretation of a syntagm. However, it is their paradigmatic properties which are of importance in relation to interaction. Fixed expressions represent meaningful choices on the part of the speaker/writer. They are single choices (see Sinclair 1987b: 321 and passim), and, as with other kinds of lexical item, their precise values and force should be considered in terms of the paradigm operating at each slot or choice. By taking into account paradigmatic as well as syntagmatic aspects, it is possible to assess the way in which fixed expressions contribute to the content, structure and development of a text.
Fixed expressions, especially highly colourful and metaphorical idioms and proverbs, are comparatively infrequent. They appear to be more frequent in spoken text than written, although to date there are few extensive studies of their actual distribution. Strässler assesses the frequency of idioms, excluding phrasal verbs, in spoken discourse as around one per 4.5 minutes of conversation (1982:81). A survey of 240 English proverbs (Arnaud and Moon, forthcoming) finds that there are around 33 instances of proverbs per million words of OHPC, 1 and that the average frequency of each of the proverbs is much less than one occurrence per million words: this list of proverbs consists of those best known to informants in a small survey, and it should be pointed out that the more frequent of these proverbs nearly always occur in exploited or truncated forms, not the canonical citation forms. So in setting out to evaluate the textual contribution of fixed expressions, it is in fact difficult to find a text where their density is sufficiently high to make valid observations. A densely populated text would be atypical; while a densely populated section of a text would be unrepresentative by being decontextualized. With these caveats, I want to consider an editorial from The Guardian as a basis for discussion. The choice of this text is governed by the fact that it is fairly short, and contains a sufficient number of fixed expressions for