The Shape of Some Wavelength Generalization Gradients

An impressive series of studies, begun by Guttman and Kalish (1956), has brought modern operant techniques to bear on the classical problem of generalization. Most of these papers test hypotheses based on the theories of Hull and Spence. The data have favored some hypotheses, but not others. For example, transposition was successfully predicted from the generalization gradient (Honig, 1958), and discrimination training led to a pre- dicted "peak shift" (Hanson, 1959), although the shape of the shifted curve was not derived in detail. On the other hand, there was no evidence of a relation between discrim- inability (size ofjnd) and the shape of the gradient (Guttman & Kalish, 1956). Hull's ideas on gradient summation were not confirmed (Kalish & Guttman, 1959). The role of motiva- tion seemed more complex than had been hoped (Thomas & King, 1959). Gradients of ex- tinction did not look as expected.2