Proceedings of the Annual Meeting of the Cognitive Science Society

Evaluations along a positivenegative dimension can be measured either explicitly (via self-report) or implicitly (via re-sponse interference tasks). Whether implicit evaluations encode relational information (e.g., A causes B) or only co-occurrence information (AB) has been debated extensively. 1,082 participants observed a machine being activated bycausally responsible stimuli and dispensing rewards in the presence of merely associated, but not causal, stimuli. Eval-uations of causally responsible vs. associated stimuli were measured implicitly and explicitly. Explicit and implicitevaluations of causally responsible stimuli were more positive than those of associated stimuli, both in the presence (Study1) and absence (Study 2) of verbal instructions about the operation of the machine. Study 3 eliminated temporal primacyand overshadowing as explanations of the effect. Supporting propositional theories, these findings suggest that implicitevaluations are sensitive not only to co-occurrence but also to relational information, whether conveyed verbally or learnedsolely from experience.