Communication scholars describe a pervasive ‘third person effect’ wherein people see mass media as more likely to affect other people than themselves. Two experiments are reported demonstrating that this effect is not a universal response to the issue of social influence, but occurs in specific social comparative contexts. In Experiment I respondents judged the impact on self and other of three types of media content-negative content, positive content, and public service campaigns. Comparison others varied on two dimensions, vagueness and closeness. A third person effect was found for both negative and positive content, but was more pronounced for negative content. The effect was also more pronounced in comparisons with vague and distant others. In contrast, respondents saw themselves as relatively vulnerable to influence from public service campaigns. Moreover, the direction of perceived self-other differences varied with respondents' perceptions of the desirability of the intended influence. In Experiment 2 respondents judged the impact on self and other of media violence and drink-driving campaigns. Results confirmed a perception of relative invulnerability to negative content and indicated that comparisons with vague others, and particularly with vague-distant others like ‘the average person’, facilitate such perceptions. Perceived self-other differences on the issue of drink-driving were less evident and varied with the perceived desirability of the intended influence. Results are discussed in terms of the ego-defensive and self-enhancing functions of social comparisons.