The present study was aimed at identifying procedural parameters that affect the generalization gradient of human predictive learning. Three such parameters were identified in the present experiment: (1) prior discrimination learning on the generalization testing dimension, (2) steady-state testing, and (3) within-subjects testing of the entire range of generalization test stimuli. Each of these procedural parameters lowered/sharpened the gradient. These results are remarkably similar to
findings from animal conditioning experiments (Jenkins & Harrison, 1960; Kalish & Haber, 1963) and highlight important analogies between the two areas of research.