We develop an experimental method to elicit subjective beliefs about the ordering of mortality risk over different causes of death. The experimental procedure emphasizes incentive-compatibility, so that the individual has a positive financial incentive to respond truthfully. We also consider the extent to which individuals have subjective beliefs for sub-segments of the population that are more accurate than their beliefs about the risks for the population as a whole. We propose several hypotheses concerning the degree of familiarity of the risks, and find that the evidence supports those hypotheses. The evidence also suggests that there is no discernible difference between beliefs elicited using hypothetical or real financial rewards in the elicitation format we use. Our findings restore some confidence in the ability to elicit beliefs about mortality risks, and therefore to get reliable estimates of the monetary value of a statistical life.