This study investigated the extent to which the third-person effect—the tendency of people to estimate greater impact of media messages on ‘other people’ than on themselves—might depend upon question-contrast effects (i.e. self-serving comparisons triggered by back-to-back questions dealing with effect on others and oneself), the order of questions, and respondents' levels of background political knowledge. Two hundred and eighty-seven subjects participated in two experimental studies involving questions about media coverage of President Clinton's possible role in the ‘Whitewater Affair’, his alleged frequent policy reversals, the O. J. Simpson murder trial, and child molestation charges against Michael Jackson. Both experiments resulted in significant third-person effects that did not depend upon having the same respondents answer both questions; means for single-question (no contrast) conditions did not differ significantly from comparable means in two-question (contrast) conditions. No significant main effects of question order were observed. In Experiment 1 a significant interaction between political knowledge and question order was found, such that a negative relationship between knowledge and perceived impact on oneself emerged when the ‘self’ question followed a question about perceived effects on others. Experiment 2 replicated the interaction for two of three news stimuli, and indicated that it was not a product of differences in the personal importance of issues. Implications of these results for understanding the third-person effect are discussed.