Rats (Experiment 1) and pigeons (Experiment 2) responded on several concurrent fixed interval variable interval schedules. The programmed rate of reinforcement varied from 15 to 240 reinforcers per hour across conditions for each component. The rate of, but not the time spent, responding on each component usually changed within sessions. The patterns of changes in response rates within the session were similar enough for the two components that the bias and sensitivity to reinforcement parameters of, and the percentage of the variance accounted for by, the generalized matching law did not change within the session. These results imply that within-session changes in responding do not cause problems for assessing the validity of the generalized matching law when subjects respond on concurrent fixed interval variable interval schedules. They may help to explain why the matching law provides a good description of the data.(1)Response rates often change systematically within experimental sessions (see, e.g., McSweeney, 1992). In the present experiments, the implications of these withinsession changes in responding for the generalized matching law (GML), the leading description of concurrentschedule performance (Baum, 1974), were examined. The GML appears in Equation 1. The rates ofresponding emitted on, the time spent responding on, and the rates of reinforcement obtained from, one component ofa concurrent schedule are symbolized by PI' T I , and R I , respectively. The same variables for the other component are symbolized by P z , T z , andR z . The a and b parameters are bias and sensitivity to reinforcement, respectively. Bias represents preference for a component that is not explained by differences in the rates ofreinforcement provided by the components. Sensitivity represents the degree to which preference changes with changes in the ratio of the rates of reinforcement.~= changes also produced within-session changes in the parameters and fit ofthe GML. In that case, the parameters and fit ofthe GML would depend on the time in the session at which they were measured. Because longer sessions would sample different values of the parameters and fit than shorter sessions (McSweeney, 1992;McSweeney, Roll, & Cannon, 1994), the parameters and fit would also vary with session length, an undesirable result.Within-session changes in the parameters and fit of the GML might occur ifthe within-session changes in response rates differed for the two components of a concurrent schedule. Suppose that responding during each component increased to a peak and then decreased within experimental sessions, as responding often does (see, e.g., McSweeney, 1992). Suppose also that the withinsession patterns differed for the two components (e.g., the peak was reached at different times). Then, the ratio of the rates of responding during the two components would change within the session. The fit and parameters of Equation I would also change, unless the ratio of the obtained rates ofreinforcement varied in the same way as the ratio ofres...