Experiment 2 also produced an orderly gradient of extinction generalization, based on stimulus similarity. This is in line
with comparable findings in an earlier predictive learning study by Vervliet et al. (2006). Importantly, this suggests that the
generalization of extinction depends on a mechanism similar to the one that produces generalization of acquisition. However,
Vervliet et al. (2006) obtained evidence in a separate experiment that the generalization of extinction is intrinsically weaker
than the generalization of acquisition: after acquisition and extinction of the same S+, a generalization stimulus evoked
a moderate return of outcome-expectancy at test. This suggests that the acquisition effect generalized more easily to the novel generalization stimulus than the extinction effect.