The pragmatics of (non-)exhaustivity in questions

Morgan Moyer

Rutgers University

Friday, December 13, 2019
2 p.m.

Lattimore 513


The status of the non-exhaustive reading of questions has sparked much debate among researchers studying the semantics of questions. On the one hand, it appears constrained by the linguistic form of the question, and yet it appears independently licensed by non-exhaustive discourse goals. In this talk, I present a series of studies that systematically investigate the factors giving rise to non-exhaustive readings of embedded questions.

In the first experiment, I explore the linguistic constraints on non-exhaustivity, fully-crossing several surface-level cues discussed in the literature to explore how necessary they are to the reading. In the second experiment, I pit linguistic form against discourse goals and find that while form matters, both exhaustive and non-exhaustive readings are also modulated by discourse goals.

Finally, I present the results of a corpus study using the British National Corpus, investigating all occurrences of root and embedded questions, to better understand the force of these cues. We coded questions according to the relevant set of factors identified in Experiment 1 and quantified the link between linguistic cues and interpretation by asking (given an assertion of a particular know-wh question form), what is the most likely/acceptable answer (exhaustive or non-exhaustive) that a hearer could have given? I found that while the corpus data might validate some of the intuitions in the literature about question form, the data suggest that participants were not actively recruiting the structure available from their experience.

Combined, the results of these studies suggest interpreting (non-)exhaustivity is not a matter of a coarse-grained distinction between semantics versus pragmatics, at least as traditionally construed. Rather, it appears that interpretation arises from a hearer calculating how best to resolve the speaker’s missing information given what they can infer about the speaker's goal, and how the speaker posed the question in a given context.