Four human subjects worked on all combinations of five variable‐interval schedules and five reinforcer magnitudes (¢/reinforcer) in each of two phases of the experiment. In one phase the force requirement on the operandum was low (1 or 11 N) and in the other it was high (25 or 146 N). Estimates of Herrnstein'skwere obtained at each reinforcer magnitude. The results were: (1) response rate was more sensitive to changes in reinforcement rate at the high than at the low force requirement, (2)kincreased from the beginning to the end of the magnitude range for all subjects at both force requirements, (3) the reciprocal ofkwas a linear function of the reciprocal of reinforcer magnitude for seven of the eight data sets, and (4) the rate of change ofkwas greater at the high than at the low force requirement by an order of magnitude or more. The second and third findings confirm predictions made by linear system theory, and replicate the results of an earlier experiment (McDowell&Wood, 1984). The fourth finding confirms a further prediction of the theory and supports the theory's interpretation of conflicting data on the constancy of Herrnstein'