Despite their effect on attitude structure, actions that engender high commitment do not necessarily lead to attitude inference. The hypothesis that attitude inference follows behavior only when the individual is highly committed to the behavior, cannot avoid thinking about the attitudinal implications, and cannot reinterpret the behavior as attitudinally-irrelevant was tested with 50 female undergraduates. A 2 x 2 design (belief-relevant cue vs. belief-irrelevant cue x list issues for own topic vs. for confederate's topic) was used with a low commitment (belief-relevant cue, own topic) control condition. After agreeing to the request to tape record a consonant speech, subjects who overheard a confederate offer a belief-relevant reason for agreeing to tape record a speech on a different issue, produced more consonant arguments for their own topic and showed greater attitude polarization than did subjects in the control or the remaining three experimental conditions. Subjects in the belief-relevant-own-topic condition also generated the greatest number of consonant arguments, suggesting that attitude inference following one's own behavior may be mediated by the cognitive responses one generates. (Author/JAC)