Previous research has produced conflicting results regarding the effects of component duration on interactions in multiple schedules. In Experiment 1, potential sources of this conflict were evaluated. Both the effects of absolute reinforcement rate and carry-over effects (hysteresis) from a preceding condition were isolated. When lO-sec components were used, the sensitivity of relative response rate to relative reinforcement rate was affected very little by hysteresis effects and absolute reinforcement rate, but it was systematically reduced as a function of the number of prior conditions. Sensitivity to relative reinforcement rate was also substantially higher with the 10-sec components than with 2-min components. In Experiment 2, this effect of component duration was decomposed into two separate effects. Contrast effects during presentation of a target component with a constant reinforcement rate were greater the shorter the target component was itself; but they were smaller the shorter the alternative component in which reinforcement rate was varied. The latter effect was smaller and more unreliable across subjects. The existence of these two separate effects demonstrates that the usual method of studying component duration-that is, holding all components equal in duration-systematically causes underestimation of the effects of the component duration, and obscures the different processes controlling the two effects.A major emphasis in free-operant research has been the development of quantitative descriptions of behavior. One of the most influential of such descriptions is the generalized matching law, given by Equation 1, which has been applied to schedules in which two components are available, either simultaneously or successively (B refers to the behavior in a component; R refers to the corresponding reinforcement rate; b refers to a bias independent of the reinforcement rates, and a to the sensitivity to relative reinforcement rate). The assumption underlying this application is that variation in the interactions between the components of the schedule can be captured adequately by the two parameters of the expression. For this assumption to be sustained, parameter variations that should have important effects on behavior should yield corresponding changes in either bias or sensitivity.(1)An important schedule parameter that seems to challenge the use of Equation 1 as a description of multiple schedule behavior is the duration of the schedule components. Given that contrast effects in one component of the schedule are an inverse function of the reinforcement rate during the alternative component (see Williams, 1983a, for a review), it seems intuitively plausible that the duration of the components should be an important determinant of the size of the contrast effects. When each component is long, responding at most points in the component is distant from the alternative reinforcement schedule; when each component is short, responding at any point is close to the alternative schedule. It seems plausible, ...