Three pigeons were studied under a pair of equal fixed-ratio schedules and a pair of equal variable-ratio schedules. Each pair was arranged as independent, concurrent schedules and also in a non-independent relation where each peck in a schedule counted toward the response requirement of both schedules. The non-independent pair of variable-ratio schedules maintained much higher changeover rates than any of the other three arrangements. Thus, two factors seemed necessary for generating high changeover rates. Responding on a schedule had to count toward the response requirement of both schedules, and the component schedules had to be variable. These data are consistent with the hypothesis that changeovers are at least partly controlled by the probability of reinforcement following a changeover.