This paper presents new evidence on the distribution of risk attitudes in the population, using a novel set of survey questions and a representative sample of roughly 22,000 individuals living in Germany. Using a question that asks about willingness to take risks on an 11-point scale, we find evidence of heterogeneity across individuals, and show that willingness to take risks is negatively related to age and being female, and positively related to parental education and height. We test the behavioral relevance of this survey measure by conducting a complementary field experiment, based on a representative sample of 450 subjects, and find that the measure is a good predictor of actual risk-taking behavior. We use a more standard lottery question to measure risk preference, and find similar results regarding heterogeneity and determinants of risk preferences. We also estimate the coefficient of relative risk aversion for each individual in the sample based on lottery responses. Using five questions about willingness to take risks in specific domains - car driving, financial matters, sports and leisure, career, and health - the paper studies the impact of context on risk attitudes, finding a strong but imperfect correlation across contexts. Using data on a collection of risky behaviors from different contexts, including traffic offenses, portfolio choice, smoking, occupational choice, participation in sports, and migration, the paper compares the predictive power of all of the risk measures. Strikingly, the general risk question predicts all behaviors whereas the standard lottery measure does not. The best overall predictor for any specific behavior is typically the corresponding context-specific measure. These findings call into the question the current preoccupation with lottery measures of risk preference, and point to variation in risk perceptions as an understudied determinant of risky behavior.