Acquisition and transfer of a higher-order conditional discrimination performance in the Japanese monkey

Abstract

Two Japanese monkeys were trained on a higher-order conditional discrimination task with two colors (red and purple) which consisted of a matching-to-sample and an oddity-from-sample discrimination. External stimuli (patterns on a key, etc.) served as conditional discriminative stimuli for matching and oddity. Both subjects mastered this complex task. Then, the stimulus control established in the training was examined with test trials inserted among baseline trials. Subjects correctly responded on the test trials in which the incorrect comparison stimulus specified by the sample was no longer presented. However, subjects could not respond correctly when both sample and comparison stimuli were replaced with two new colors (yellow and yellowgreen). Thus, no evidence that the identity-difference relation of stimuli controlled the subjects' responding was obtained even in the presence of the higher-order conditional discriminative stimuli controlling their matdsing and oddity responses. These results suggested that the specific relation between the sample and the correct comparison stimulus mainly controlled two-color matching and oddity behaviors of Japanese monkeys. Fujita (1982) analyzed the stimulus control in two-color matching-to-sample behaviors of Japanese monkeys by examining transfer performances to new colors, using "non-effective" trials in which no differential reinforcement was made. Although his monkeys failed to match unfamiliar test stimuli, they showed somewhat successful matching performances for familiar test stimuli for which the successive and the simultaneous discriminations between the stimuli were previously established. He concluded that the general relation of "identity-difference" between the sample and the comparison stimulus controlled, at least in part, two-color matching-to-sample behavior of Japanese monkeys. (In other words, subjects acquired , although only incompletely, the general rule to choose the same comparison stimulus as the sample.) As he suggested, the apparent weakness of such relational control can be interpreted in two ways. The first possible interpretation assumes that the general relation of stimuli (i.e. identity-difference) controlled the subject's behavior with much strength (which was labeled as the "general-relation control") only in the presence of baseline stimuli. In other words, the baseline stimuli themselves might be a conditional discriminative stimulus to control the subject's generalized matching behavior. In this case, the accuracy of performance on test trials might deteriorate simply because the conditional discriminative stimulus (i.e. the baseline stimuli) is absent on those trials. The second possible interpretation is that the general-relation controlled the subject's behavior with least strength, and instead, the specific relation between the sample and the comparison stimuli mainly controlled his behavior (the "specific-relation control": subjects acquire a set of sample-specific rules such as, for example, to choose red if the sample is red.). In this case, the deteriorated matching accuracy