The research reported in this article examined the conditions under which persuasive arguments are most effective in changing university students' attitudes and expressed behavior with respect to affirmative action (AA). The conceptual framework was a model that integrated the theory of reasoned action and the elaboration likelihood model of persuasion. Studies 1 and 2 established effective manipulations of positive-negative AA information, and peripheral-central routes of processing. Study 3 implemented these techniques, and a path analysis was carried out testing the differential effects of valence of information processed via different routes on AA evaluative beliefs, attitudes, intention, and expressed behavior. Results indicated that positive AA messages processed centrally (i.e., for meaning) resulted in significantly more positive evaluative beliefs. Modifications to the original model resulted in a final model with excellent fit to the data that supported the mediating role of intention in the AA attitude-behavior relationship, as predicted by the theory of reasoned action. The findings highlight potential benefits of interventions for improving support for AA policies, provided that positive information is processed at a central, evaluative level. (Contains 5 tables and 4 figures.)