Association with explanation-conveying constructions predicts verbs’ implicit causality biases

Abstract

Abstract Given a sentence such as Mary fascinated/admired Sue because she did great, the verb fascinated leads people to interpret she as referring to Mary, whereas admired leads people to interpret she as referring to Sue. This phenomenon is known as implicit causality (IC). Recent studies have shown that verbs’ causality biases closely correspond to the verbs’ semantic classes, as classified in VerbNet, a lexicon that groups verbs into classes on the basis of syntactic behavior. The current study further investigates the relationship between causality biases and semantic classes. Using corpus data we show that the collostruction strength between verbs and the syntactic constructions that VerbNet classes are based on can be a good predictor of causality bias. This result suggests that the relation between semantic class and causality bias is not a categorical matter; more typical members of the semantic class show a stronger causality bias than less typical members.