This study, based on the two-wave questionnaire data collected from legislative candidates in Iowa, attempts to test the “congratulation-rationalization effect,” a highly provocative hypothesis that John Kingdon formulated regarding politicians' beliefs about voters. The hypothesis asserts that winning candidates tend to develop complimentary beliefs about voters while losing candidates tend to develop beliefs deprecating to voters. The results of analysis indicate, however, no significant difference between winners and losers in terms of the direction and magnitude of changes in their beliefs about voters, suggesting that the hypothesis is invalid. When the hypothesis is reformulated in terms of “dissonance states” rather than “election outcomes,” the evidence is strongly supportive. Among winners, those who perceive a high degree of dissonance more than those who perceive little dissonance tend to change their beliefs about voters in a favorable direction. Conversely, among losers, those who perceive a high degree of dissonance more than those who perceive little dissonance tend to change their beliefs in an unfavorable direction. Therefore, the “congratulation-rationalization” hypothesis can be sustained only if cast in direct dissonance terms.