The function of imagery as a mediator in relational learning

Abstract

Two general hypotheses were tested in a transposition experiment. One hypothesis was that individuals differ in the structure of their internal representations of stimulus information. The differences in internal representational structure were thought to have a function in determining whether or not a relationship among the stimuli would be the outcome of learning. The other hypothesis was that relational information would be stored in memory in an imagery format. On each of twenty-five or more trials, subjects were presented with three squares differing in size by a constant ratio. They were consistently reinforced for choosing the medium sized one. On the test trials following training, the stimulus configuration was shifted toward the larger end of the size dimension. Continuing to select the medium sized square indicated that the relationship had been learned. Three variables were manipulated in the experiment. The salience of the relational information was manipulated by exposing subjects to one or two sets of stimuli in training. Only relational responses were reinforced in the two problem condition; but, relational and absolute responses were reinforced simultaneously in the one problem condition. The other two variables were the size ratio of the squares and whether or not squares were displayed or named on the test trials. Names had been learned for each square prior to the experiment. A battery of tests that measured verbal and imagery processes was administered. Half of the subjects consistently selected the medium sized square. Choice reaction times were faster to stimuli with large size ratios than to stimuli with small size ratios when squares were displayed on test trials, but there was no difference when stimuli were named. The salience of the relational information following two problems in training affected the percentage of relational responses only in the stimulus display condition on the test trials. The inference to be drawn from these outcomes is that relational information can be encoded concurrently in an image and a verbal format. Differences in cognitive structure were interpreted as indicating that trans posers were capable of forming abstract images but non-trans posers were capable of only concrete images.