There is little information concerning what people think about when completing questionnaires that assess perceptions of risk, and even less for questionnaires assessing unrealistic optimism. The thoughts of 40 participants who displayed unrealistic optimism about risks of skin cancer were elicited using think aloud methods, when completing both direct and indirect ratings of unrealistic optimism. The most common thoughts overall concerned exposure to the sun, and features such as skin colouring. Thoughts concerning prevalence, reasons for risky behaviour and admissions of ignorance were more common for indirect measures of unrealistic optimism than for direct measures. The direct unrealistic optimism measures yielded more optimistic ratings for those participants who did not mention symptoms or signs of skin damage, and those who mentioned thoughts about prevalence. Participants seem to be drawing upon different sources of information when completing superficially similar direct and indirect measures of unrealistic optimism, which may explain why these measures are usually only modestly associated. People do not seem to think about numerical probabilities when estimating risk, but instead appear to focus on issues such as exposure to risk, and concrete bodily symptoms and signs. This may at least partially explain why attempts to influence behaviour by providing probabilistic information are generally unsuccessful.