ABSTRACT: Suggests that a theory of attitudes (i.e., evaluations or effects associated with a social object) should specify (1) how attitude objects are represented and (2) what kind of representation constitutes an attitude. Topics addressed include state-of-the-art attitude measurement and concerns about the construct validity of the attitude concept. Data are presented in support of the suggestion that strong effects of attitude on action can and do occur when the actor is not attentionally focusing on the attitude. Data suggest that attitudes may be measured effectively with techniques resembling those used to study implicit memory.