Published online by Cambridge University Press: 24 November 2020
Implicatures serve as an important testing ground for examining the process of integrating semantic and pragmatic information. Starting with Bott & Noveck (2004), several studies have found that implicature computation is costly. More recently, attention has shifted toward identifying contextual cues that modulate this processing cost. Specifically, it has been hypothesized that calculation rate and processing cost are a function of whether the Question Under Discussion (QUD) supports generating the implicature (Degen 2013; Degen & Tanenhaus 2015). In this paper, we present a novel elicitation task establishing what the relevant QUDs are for a given context (Experiment 1). In Experiment 2, a sentence-picture verification study, we extend earlier findings about the effect of QUDs on scalar inference to a different kind of pragmatic inference: it-cleft exhaustivity. For both inferences, we find that under QUDs that bias toward calculation, there is no increase in reaction times, but under QUDs that bias against calculating the inference we observe longer reaction times. These results are most compatible with a constraint-based account of implicature, where QUD is one of many cues. Additionally, we explore whether our findings can be informative in narrowing down precisely what aspect of the inferential process incurs a cost.
For their insightful comments and suggestions, we thank the three anonymous reviewers, Hannah Rohde and John Tomlinson, the audiences at AMLaP 24 and NELS 49, and especially Chris Kennedy and Michael Tabatowski. We are grateful to Bob van Tiel for sharing experimental materials with us and to Zsolt Veraszto for technical help. All mistakes and shortcomings are our own.