In two experiments, the influence of musical endings on duration judgments was examined. In Experiment 1, subjects were asked to compare the duration of paired folk tunes that could vary in their degree of resolution. Results showed that melodies ending on the leading tone interval were underestimated relative to songs ending on the conventional tonic. Experiment 2 further revealed that time estimates were influenced by a melody's preceding temporal accent structure. Melodies ending earlier than expected were judged significantly shorter, particularly when the tune ended on the leading-tone interval. Conversely, tonic endings were judged substantially longer when they appeared to occur "too late" in time. Results are discussed in terms of a contrast model which emphasizes the role of expectancies on duration judgments.