How do consumers combine multiple risk items when forming overall risk judgments? Drawing on fuzzy trace theory and categorical averaging, this research finds that adding a low-risk item to a high-risk item reduces overall risk perception because people reason categorically about risk. They impose categorical distinctions on quantitative risk information, and when combining categorical information, they tend to average across categories instead of adding. Eight studies in the context of medical and health product decisions (N = 5,152) provide convergent evidence showing that when items in different risk tiers are considered together, they are consistently evaluated in a subtractive manner, leading to a higher likelihood of taking the objectively riskier medication (studies 1A, 1B, 1C, and 2). This effect is eliminated when the probability of one side effect is exceedingly high (study 3) or when the task requires reliance on verbatim representation of probabilities (study 4). The effect also disappears when risk information is presented graphically in a way that emphasizes the additive property of multiple risks (studies 5A and 5B). The findings have important implications for the fields of risk perception, risk communication, and consumer health and medical decision making.