Several past studies have identified that dysphoric undergraduates (those with depressed mood) sometimes give estimates of their control over outcomes closer to the actual contingency than estimates made by the nondepressed. This "depressive realism" phenomenon is typically found in tasks with zero control and frequent outcomes. The present paper investigates this phenomenon with a more powerful design for comparing estimates. Experiment 1 manipulated motivation for accuracy to compare the prevalent self-serving motivational bias explanation to an information processing bias perspective. Depressive realism was found under the typical conditions, but under conditions of higher motivation for accuracy, all participants overestimated their control on the key task. Response patterns appeared to influence estimates via the proportion of exposure to different trial events. Experiment 2 confirmed an influence of response frequency on estimate bias. In Experiment 3 the depressive realism pattern was found for positively valent outcomes, but the reverse pattern occurred for negatively valent outcomes. Both those with and without depressed mood showed biased processing on some tasks. The implications for contingency processing in depressives and the general population are addressed.