We examine the effect of respondent of consequentiality on a split-sample binary-choice/multinomial-choice stated-preference survey. We fail to observe the knife-edge results predicted in the consequentiality literature in the binary-choice data but do observe them in the multinomial-choice data. In the multinomial-choice data, only respondents who believed the survey was at least somewhat likely to affect future policy behaved consistently with theoretical predictions. Overall, we conclude that failing to control for perceived consequentiality can lead to false conclusions regarding marginal effects and welfare estimates. This is true in both the binary- and multinomial-choice contexts.