A study examined the effects of training in argumentation and of initial candidate preference on perceptions and evaluations of presidential candidate debates. Subjects, 490 undergraduate speech communication students, were randomly assigned to three experimental conditions--control (no training), criteria only, and criteria plus explanations for their application--just prior to viewing the debate. Subjects also completed a questionnaire on demographic information and candidate preference (Ronald Reagan, Walter Mondale, Other, and Undecided). Subjects then viewed the televised debate. Subjects' post-debate perceptions of the candidates were measured using two 35-item semantic differential scales and a single scale representing a global assessment of performance in the debate. Finally each subject rank ordered ten performance-related criteria in terms of the importance of each to the subject's evaluations of the candidate. The most noticeable result centered on the uncommitted subjects' perceptions of the candidates. The effects of training on those who had an initial commitment to a candidate were more ambiguous. A statistically significant training/preference interaction effect was observed for the subject's perceptions of Reagan on both the "capability" and "personableness" dimensions. Uncommitted subjects who had viewed Reagan more negatively on the "capability" dimension as a result of training evaluated Mondale in just the opposite way. Those uncommitteds who were trained evaluated Reagan's performance less favorably than did those in the control group. While these findings might suggest that only a portion of those exposed to training are influenced by it, the effect of such training can be significant in some elections. (HTH)