Pluralities of Questions

This paper proposes a novel semantic analysis of the quantificational variability data discovered by Berman (1991). We suggest that the adverb of quantification in those data quantifies over semantic questions. Its domain is a division of the original question into subquestions, where a legitimate division into subquestions is one in which each member contributes towards the answer to the original question, and together the answers to all subquestions provide the complete answer to the original question. Thus the question itself is associated with a part/whole structure, based on information. We show that there are quantificational variability effects in which the matrix verb is exclusively question-embedding. These data pose a problem for other theories of quantificational variability in questions (specifically Berman 1991 and Lahiri 2002) and motivate our analysis. There are other desirable consequences of our theory, including flexibility in what counts as a subquestion and flexibility in what counts as a complete answer. Beyond quantificational variability, associating questions with a part/whole structure receives independent motivation from questions that occur in collective and cumulative embedded contexts.