Experiment 2 also included a test for the generalization-of-extinction in group BETWEEN, by further extinguishing
the GS in each subgroup, before re-tests with the original S+ (see also Dubin & Levis, 1973; Vervliet et al., 2006). These
tests revealed a gradual recovery of outcome-expectancy to the S+ with progressively dissimilar GSs. Hence, the amount
of generalized extinction depended crucially on stimulus similarity, thereby establishing the gradient of generalizationof-
extinction. Interestingly, the orderly gradient manifested itself already on the first test trial in this case (contrary to
generalization of acquisition). Probably, the extinction learning phase induced discrimination learning over the generalization
test dimension. In addition, the two types of generalization were positively correlated: Those participants that showed
stronger generalization of acquisition also showed stronger generalization of extinction.