Retrospective evaluations of aversive episodes were studied in the context of a general model of "judgment by prototype" that has been applied in other situations. Unpleasant sounds of variable loudness and duration were the stimuli. In Experiment 1, continuous reports of annoyance closely tracked variations of noise intensity. Hypotheses about the determinants of retrospective evaluation were examined in Experiment 2. Experiment 3 confirmed a prediction of judgment by prototype: The effects of sound duration and intensity are additive in multitrial experiments. Experiment 4 confirmed a robust preference for aversive episodes that are "improved" by adding a period of reduced aversiveness.