Experiment 1 was designed to evaluate the effects of prior discrimination training and steady-state testing on the shape
of a generalization gradient in human predictive learning. Group INTRA received prior discrimination training with the S+
on the generalization dimension, whereas group INTER received prior discrimination training with the S+ on an irrelevant
dimension. As expected, the results revealed a lower and sharper gradient in group INTRA as compared to group INTER. This
shows that discrimination training can influence the level of generalization to different-but-similar stimuli, but the effect
is dimension-dependent. This is in line with observations in animal conditioning preparations (e.g., Jenkins & Harrison,
1960; Spence, 1937). The observed differences between the two groups were most pronounced on the first test trial. A
comparison between the first test block and the average over all test blocks indicated that repeated testing in a steady-state
testing procedure (where the original S+ is occasionally reinforced, but the generalization test stimuli are not) lowered the
gradients in a stimulus-dependent manner. A steady-state testing procedure can be seen as a discrimination procedure
(between the S+ and the range of generalization stimuli); it is therefore not surprising that it sharpened the gradients
further. Taken together, Experiment 1 showed that prior discrimination learning on the generalization test dimension and
a steady-state procedure during testing can sharpen/lower the gradient of generalization in human predictive learning.