In contemporary attitudes-and-attraction research, attraction has been viewed as a multidimensional construct. Moreover, the effects of dissimilar and similar attitudes have been shown to vary with the facets of attraction measured. The hypotheses tested are that (1) only the proportion of similar attitudes relevant to the social context or interaction goals affects behavioral attraction (i.e. interpersonal distance between the participant and targets), and (2) the proportion of similar attitudes influences affective attraction (i.e. Byrne's attraction measure), regardless of attitude relevance. Two experiments were conducted with classroom activities (Experiment 1) and a writing workshop (Experiment 2) as the social contexts. The results of both experiments supported the hypotheses. Clearly, a solely affective measure of attraction seems inadequate for understanding the similarity–attraction relationship. Copyright © 2002 John Wiley & Sons, Ltd.