Convergent validation of information processing constructs with Pavlovian methodology.

These two experiments introduce a new nonverbal measure of stimulus structure by merging Garner's (1974) distinction of separable and integral stimulus properties with the field of Pavlovian conditioning. According to one rule, two sets of stimulus (separable vs. integral) were constructed. In each experiment a differential compound conditioning procedure was used with one group of subjects who were trained with a subset of the separable material and then were tested with an unfamiliar subset. The same procedure was used with a second group but with integral materials. Both experiments showed a reversal of conditioning effects, depending on the properties of the stimuli. Separable compounds showed results as predicted by elementaristic conditioning theories: The associative values of the elements summed up. With integral compounds, the associative values of the elements were irrelevant. Instead, integral compounds were processed and classified primarily on the basis of similarity.