Four adult humans made repeated choices between two time-based schedules of points exchangeable for money: a fixed-interval schedule and a progressive-interval schedule that began at 0 s and increased in fixed increments following each point delivered by that schedule. Under reset conditions, selection of the fixed schedule not only produced a point but also reset the progressive interval to 0 s. Reset conditions alternated with no-reset conditions, in which the progressive-interval duration was independent of fixed-interval choices. Fixed-interval duration and progressive-interval step size were varied independently across conditions. Subjects were exposed to all step sizes in ascending order at a given fixed-interval value before the value was changed. Switching from the progressive-interval schedule to the fixed-interval schedule was systematically related to fixed-interval duration, particularly under no-reset conditions. Switching occurred more frequently and earlier in the progressive-schedule sequence under reset conditions than under no-reset conditions. Overall, the switching patterns conformed closely to predictions of an optimization account based upon maximization of overall reinforcement density, and did not appear to depend on schedule-controlled response patterns or on verbal descriptions of the contingencies.