Functional constraints, usage, and mental grammars: A study of speakers’ intuitions about questions with long-distance dependencies

This paper describes an experimental study which attempts to reconcile two usage based approaches to questions with long distance dependencies (LDDs): the Lexical Template Hypothesis (Dąbrowska 2004, 2008; Verhagen 2005, 2006) and Goldberg's BCI (“Backgrounded Constituents are Islands”) constraint (Goldberg 2006; Ambridge and Goldberg 2008). The study replicates Ambridge and Goldberg's (2008) results supporting the BCI constraint; but it also shows that (1) LDD questions with think and say, the verbs which are part of the hypothesised templates, are judged to be more acceptable than predicted by BCI and (2) BCI cannot explain complementizer effects (why LDD questions with that are judged less acceptable than questions without that). The results also suggest that there are considerable individual differences in speakers' sensitivity to the constraint. Thus, the two hypotheses are complementary: BCI explains why certain LDD questions are more acceptable than others, and hence accounts for differences in the frequency of prototypical and unprototypical LDD questions, while the lexical template hypothesis explains the effects of the frequency of use on speakers' mental grammars.