Conventional theories of tradeoffs between health risks and money suggest that willingness to pay (WTP) for a reduction in the probability of harm should be proportional to the probability increment. In contrast, studies using Contingent Valuation (CV) generally find that WTP is less than proportional to the probability increment. Because the departure from proportionality may reflect a failure to effectively communicate the small risks that are of interest rather than an inconsistency between survey respondents' preferences and standard theory, we test the effect of variations in instrument design on estimated sensitivity to risks associated with automobiles, food, pneumonia, and blood transfusion. Variations include using probability analogies for risk communication, eliciting indifference risk for a specified cost, and distinguishing survey respondents by self-reported confidence in their response to WTP questions. Overall, estimated WTP remains less than proportional to the risk increment. Supplementing numerical probabilities with analogies does not substantially influence estimated sensitivity to scope, although eliciting indifference risks for specified cost and excluding respondents who report limited confidence in their answers to WTP questions yield results that are more consistent with theoretical predictions. Ongoing work tests the effect of using risk ladders and other visual representations to aid in communicating risk in CV studies.