A series of experiments is used to investigate the extent to which peoples valuation of a risky prospect is affected by the values from which a participant selects a response. Three variables were considered: a smaller risk-free amount, a larger risky amount, and the probability of winning the larger amount. There were three conditions: in each, two of the three variables were held constant, and participants chose the value for the third variable that made the risky and risk-free options worth the same to them. This was done first by a free-choice valuation, and then, with different participants, by choosing one of four options that were either all below or all above the population free-choice median. The options presented had a strong effect on valuation of the missing variable. This effect remained even when the free-choice and multiple-choice conditions were presented within subjects. This demonstrates that people showing rational and consistent risk evaluation strategies could have their risk aversion manipulated by context. Overall, the experiments suggest that peoples propensity for risk aversion is manipulable by context. This is problematic finding for traditional, context-independent, theories of decision under risk.