On the persistence of grammar in discourse formulas: a variationist study of that

This article provides evidence that, just as lexical meaning is retained in grammaticization, grammatical conditioning persists in fixed discourse formulas. Despite their high frequency and formulaic status, such formulas are not completely autonomous from the productive constructions from which they emerge. This evidence comes from a variationist analysis of that and zero complementizer in a corpus of spoken Canadian English. Testing syntactic, semantic, and discourse-pragmatic factors proposed to account for the variation, we focus on claims that frequent collocations have developed as discourse formulas. Multivariate analysis shows that, although the variation is largely lexically constrained, that serves to demarcate the boundaries of two clauses with lexical content, while zero tends to occur when the clauses function like a single unit. Moreover, the linguistic conditioning of that in frequent collocations that behave like discourse formulas parallels its conditioning in the general construction. These findings suggest that the principle of semantic retention or persistence should be extended to grammar.