1 author: Mira Ariel Tel Aviv University 59

Discourse proposes, grammar disposes

The findings presented in section 2 show grammar to be freer than discourse use. The use patterns revealed in discourse are a rather small subset of the options allowed for by the grammar. We now turn to see that a similar selection process operates when we shift from discourse to grammar. But first, we have to explicate what it means ‘to shift from discourse to grammar’. It’s quite obvious, in fact trivial, to assume that discourse makes use of grammar. But can we say that grammar too makes use of discourse? Research in the last few decades has convincingly shown that the answer here is yes, very much so. The idea is that discourse patterns, such as those discussed in section 2, sometimes turn into grammatical conventions. Yesterday’s salient discourse patterns may become today’s grammatical conventions. I examine a few such cases in section 3. In fact, the examples selected for section 2 were chosen with a view to potential grammaticization processes. I now show how some of the discourse patterns noted above may have grammatical consequences. This is what I mean by grammar ‘selecting’ from discourse. Some, although not all, discourse pat- terns result in chunking into conventionalized form/function correlations. In other words, they turn grammatical. It is salient discourse patterns, for the most part frequently used ones, that have a propensity (but no more than that) to be selected for grammaticization convention.

1. 
   DISCOURSE PROPOSES, SYNTAX DISPOSES
   

The first discourse pattern we discussed in section 2 involved complement- taking verbs, such as I think . . . We noticed some salient discourse patterns in connection with such predicates. Regarding the matrix, present tense think is the leading verb, guess is the second, I is the leading subject, and the sentence tends to lack optional constituents (adverbs, direct objects). All this, we noted, adds up to the profile of a stance framing constituent. Regarding the com- plement we noted that it tends to revolve around the current discourse topic, and in general be the contextually relevant portion of the speaker’s utterance. It also tends not to be introduced with a that. Looked at functionally, then, it seems that the relationship between the complement and the matrix here very much parallels the relationship between some proposition and an epistemic stance adverbial, such as apparently or maybe. Indeed, see how I think and maybe are used interchangeably in the following:
19. a. SCOTT: ... Maybe we could –