I collect data on how subjects acquire information about risky choices in both real and hypothetical settings using process-tracing software called Mouselab. On average, there are no significant differences across settings in the amount of time subjects take to make a choice or the completeness of the information they acquire. Subjects also acquire information in sequences consistent with an integration model of decision-making, such as expected utility theory or prospect theory. I do not find significant differences in risk preferences across settings, on average, but I do find that subjects' risk preferences are related to the completeness of the information that they acquire and where they start their information acquisition. (JEL C91, D80, D83).