In everyday life, people encounter smaller rewards with higher probability than larger rewards. Do people expect this reward--probability regularity to hold in experimental settings? To answer this question, we tested whether people's behavior in probability judgment tasks is affected by the correlation between reward size and reward probabilities. In Study 1, we asked people to judge reward probabilities under uncertainty. In line with the ecological reward--probability correlation, people assumed that larger rewards were less likely than smaller rewards. In Study 2, we tested the prediction that people's information search and integration depend on the representativeness of the environment. Participants performed an experience-based probability judgment task in which they sampled outcomes from unknown gambles until they felt confident to estimate the probabilities of the gambles' outcomes. We manipulated the reward--probability relationship of the gambles in 3 experimental groups. Rewards and reward probabilities were negatively correlated, positively correlated, or not correlated at all. A negative correlation mimics the ecological reward--probability relationship often present in real life. We analyzed people's search effort and whether they integrated sample-based uncertainty into their judgments. We found that people sampled fewer outcomes in the ecologically representative condition than in the other 2 conditions. However, people did not integrate sample-based uncertainty in their judgments: In all conditions people treated the observed outcomes as representative of the underlying outcome distribution. People's prior beliefs about regularities in environments provides a potential explanation of why people often rely on small sample sizes when making judgments and decisions from experience.