Two steps towards a pragmatic of questions

Besides a rich literature on biased questions, the pragmatics of matrix questions remains largely unexplored from both a theoretical and empirical perspective. In this talk I will present two studies addressing this gap in our knowledge. This first study investigates how we select between various question forms based on possible prior knowledge. For instance, if I know that the final of the 2018 world cup was between France and Croatia, many options are available to inquire about the winner, including wh-questions ("Who won?"), polar questions ("Did Croatia win?"), and alternative questions ("Did Croatia or France won?"). We show that the notion of entropy (expected informativity of the answer) van Rooij (2003) proposed as a measure of question utility captures most of participants behavior in our experiments. The second study looks at NPIs in questions. Various proposal have been proposed to explain the distribution of NPIs in questions, based on syntactic, semantic and pragmatic factors, but the empirical facts are not very clear to begin with. We tested 26 different types of questions and compared them to the predictions of various theories. Our results suggests that three types of questions can be distinguished: questions where NPIs are clearly acceptable, questions where they are clearly unacceptable, and questions where they are slightly degraded. Interestingly, these results do not align with the predictions of van Rooij's entropy account, which was primarily designed to capture the distribution of NPIs.