This research explored how basic semantic dimensions of evaluation, potency, and activity influence attributions of who causes an interpersonal event. Two experiments examined how the dimensions affected (a) college students' judgments about which particular social identities were better agents or patients for generic positive and negative actions and (b) their judgments about who caused specific actions described by sentences containing the same identities. The semantic characteristics of the nouns and verbs and "goodness of thematic role" accounted for 44% of the variance in attribution judgments. Experiment 3 examined the effects of evaluation and potency on state verbs. Attributions were again made to the thematic role whose traits matched the valence of the state's evaluation. More potent subjects and more potent verbs elicited stronger attributions. Thus, basic dimensions of affective meaning that underlie the words used to describe interpersonal interactions help account for observers' judgments of who is responsible for those events.