Generalized identity matching-to-sample in Cebus apella

Our interest was whether or not conditions could be created under which capuchin monkeys (Cebus apella) might display generalized identity matching-to-sample (i.e., high accuracy scores on initial tests with stimuli that had not previously appeared on matching-to-sample trials). We employed procedures designed to minimize the effects of several variables that past research suggested might interfere with such emergent matching. These included unwanted stimulus control by location, failure to discriminate defining stimulus features (i.e., stimulus generalization), and stimulus novelty. In Experiment 1, two monkeys displayed accurate simple simultaneous discrimination with three sets of three stimuli each, presented in any of nine locations on a touchscreen-equipped computer monitor. In Experiment 2, the same monkeys were exposed to a 0-delay identity matching training procedure involving one of the stimulus sets from Experiment 1. After the identity-matching baseline was established, identity-matching tests were presented with other three-stimulus sets (two from Experiment 1 and three new sets). For certain sets, the critical initial identity-matching test was preceded by a history of simple discrimination reversals involving the stimuli that would appear on the test. Overall, identity-matching test scores were substantially higher than might be expected by "chance." Moreover, results suggested that a simple-discrimination reversal history might facilitate identity-matching accuracy. In Experiment 3, two additional new stimulus sets were introduced. Extensive discrimination reversal overtraining was conducted with one of the sets. With the other, reversal training was comparable to that provided in Experiment 2. With both stimulus sets, initial identity-matching test scores were substantially above chance. Notably, scores with the reversal overtraining set were perfect for one monkey and near-perfect for the other. In overview, our results suggest that capuchin monkeys may be capable of generalized identity matching when given training that supplies the behavioral prerequisites for the tests.