Two modes of disconfirmation of national stereotypes were compared. It was hypo-thesized that stereotypes should be revised either by stereotype-inconsistent cases from the primary target (i.e., direct disconfirmation) or by instances which belong to an alternative target but possess features stereotypic of the primary target (indirect disconfirmation). It was further predicted that politically liberal individuals would show greater change in their national stereotypes than conservatives. Two experiments were conducted in which the salience of directly and indirectly disconfirming information regarding different nations was manipulated. The results indicated that indirect disconfirmation generally had as much impact as direct disconfirmation, but that the effects were particularly visible among liberal participants. Conservatives tended to respond in a reactive manner which preserved or exaggerated their original views. A new theoretical framework for the study of stereotype change and implications for its application were discussed.