Response-effects research has shown that survey questions shape and channel public opinion. Our study examines the degree to which the variance in people's cognitions about and attitudes toward crime policy proposals is a methodological artifact and the degree to which it represents media effects. Four different forms of a survey were used to manipulate the order of the proposal (death penalty vs. rehabilitation) and criterion (cognition vs. attitude). Results show that experimental manipulations affected cognitions or attitudes only for the death penalty. Respondents who were asked the cognition measure first showed a high level of consistency between the net valence of their arguments and their attitudes. Respondents who were asked their attitudes first reported relatively more affective arguments, a phenomenon that we label affective priming. Exposure and attention to specific media content influenced cognitions and attitudes differentially.