We offer a new account for the well-documented empirical fact that people display inconsistent risk preferences. Our approach is to apply a theory of focusing (Kőszegi and Szeidl, 2013) to choice under risk. In our model, the decision maker may focus more on (and thus overweight) the upside or the downside of each prospect, depending on which side permits the starker comparison, and consequently the decision maker displays menu-dependent risk attitudes. We show that this highly parsimonious model accounts for many of the observed anomalies in choice under risk, and we distinguish this theory from existing theories of risky choice.