Proceedings of the Annual Meeting of the Cognitive Science Society

Studies addressing the question of whether scalar inferencesgenerally incur a processing cost have yielded conflicting re-sults. Constraint-based accounts, which seek to unify theseconflicting results, make a prediction which we test here: theprobability of an interpretation and the speed with which it isprocessed depends on the contextual support it receives. Wemanipulated contextual support for the scalar inference in twotruth-value judgment experiments by manipulating a lexicalfeature (presence of partitive “of the”) and a pragmatic fea-ture (the implicit Question Under Discussion). Participants’responder type – whether their majority response was prag-matic (reflecting the inference) or literal (reflecting its absence)– was the main predictor of response times: pragmatic re-sponses were faster than literal responses when generated bypragmatic responders; the reverse was true for literal respon-ders. We interpret this as further evidence against costly infer-ence accounts and in support of constraint-based accounts ofpragmatic processing.