1.Introduction This paper investigates the effect of changing syntactic categories on speaker choice, using the needs to be done ∼ needs doing alternation as the testing ground. The two alternants in this alternation have different syntactic properties, and so we hypothesize that the syntactic preferences of the alternation's environment influence alternant choice. Our hypothesis is that, all else being equal, speakers prefer to use an alternant with a syntactic category that is more prototypical given the rest of the sentence. We show an effect of structural bias that argues for this hypothesis.
2.Speaker choice and syntactic alternations For a given situation, there is a large, possibly infinite, set of sentences that could be said, all expressing the same core idea. Different sentences in the set may be more or less appropriate for the situation, and thus better or worse choices for the speaker. The relative appropriateness of a sentence is presumably a complicated calculation, yet speakers choose their sentences fluidly. So how do speakers choose among the various sentences they can use to express an idea? This is a big question, too big to be answered completely with our current state of knowledge. However, instead of looking at speaker choice in sentencelevel variation, we can simplify the problem by investigating how speakers choose between alternants in a syntactic alternation. A syntactic alternation is a situation in which there are multiple related phrases expressing the same semantic idea with different syntactic forms. Canonical examples of this are that-omission and the passive, dative, and genitive alternations. Looking at syntactic alternations rather than sentence-level variability simplifies the problem in three important ways: