Experiment 2 was primarily designed to evaluate the effects of within-subjects versus between-subjects testing methods
on the shape of the generalization gradient. When focusing on the first test trial alone, amarked difference emerged between the two groups: Group WITHIN showed the expected orderly decrease of outcome-expectancy over progressively dissimilar stimuli, whereas group BETWEEN exhibited a nearly flat gradient. This shows that when each subject was tested with only one of the generalization stimuli (and in the absence of prior discrimination training on the test dimension), generalization was virtually complete to the current set of generalization stimuli. In other words, responding to the test stimulus was not a function of (dis)similarity with the S+ (along the chosen continuum). This is a highly surprising finding, because participants were probably able to perceptually discriminate between the circles of differing size (see results from Experiment 1).Nevertheless, an orderly gradient did emerge in group BETWEEN over the course of extinction trials. Extinction of outcome expectancy proceeded faster with progressively dissimilar generalization stimuli. Because resistance-to-existence is often viewed as a measure of the strength of earlier learning, these differential extinction effects can be seen as the manifestation of a generalization gradient of acquisition.