In two experiments, the performance of listeners with different amounts of musical training (high skill, low skill) was examined in a two-alternative forced choice time-detection task involving simple five-cycle acoustic sequences. In each of a series of trials, all listeners determined which of two pattern cycles contained a small time change. Sequence context was also varied (regular vs. irregular timing). In Experiment 1, in which context was manipulated as a between-subjects variable, high-skill listeners performed significantly better than low-skill listeners only with regular patterns. In Experiment 2, in which context was manipulated as a within-subjects variable, the only significant source of variance was pattern context: All listeners were better at detecting time changes in regular than in irregular patterns. The results are considered in light of several hypotheses, including the expectancy/contrast model (Jones & Boltz, 1989).