Data from the National Election Studies were examined in an effort to isolate cognitive dissonance of two kinds: dissonance arising from a behavioral commitment in the form of voting, and dissonance arising from inconsistencies associated with having supported the losing candidate. Feeling thermometer ratings of the two principal presidential candidates obtained before and immediately after six elections (1972, 1980, 1984, 1988, 1992, and 1996) were analyzed. Regression estimates supported a dissonance reduction explanation of observed attitude changes. Voters, as compared to nonvoters, tended to increase the evaluative distance between candidates after an election, whereas supporters of the losing candidate were more likely than supporters of the winning candidate to decrease such evaluative distances. An additional examination of voters yielded results consistent with dissonance theory: After the election, respondents reporting favorable evaluations of both candidates (a difficult choice) tended to spread comparative candidate evaluations compared to respondents who were favorable toward only one candidate (an easy choice). The results both support and cast doubt on prior studies.