Adventitious differential reinforcement of emergent sample-specific behavior and common coding in pigeons' matching-to-sample

Abstract

Matching-to-sample performance is enhanced if responses to the sample stimuli are required, and further enhanced if different responses are required to the different samples. Even when not required, differentiated sample responding may emerge spontaneously, either overtly or as an hypothesized covert "coding" response. Manabe, Kawashima, and Staddon (1995) trained budgerigars on a matching-to-sample task with two samples mapped onto each com parison stimulus (many-to-one training), where sample-specific vocal responses were required for one set of samples. Sample-specific responding emerged for the other set of samples and was congruent with respect to the comparison mappings. They concluded that the emergent behavior reflected emergent relations and called the behavior "naming", but it is possible that unintended adventitious differential reinforcement produced the effect. Experiment 1 was designed to replicate these findings in pigeons and determine whether naming or adventitious differential reinforcement provides the best explanation. In Phase 1, 17 pigeons were trained to respond to different stimulus locations in the presence of red and green stimuli. In Phase 2 these stimuli became samples in a matching-to-sample task, with required sample-specific responding as in Phase 1. In Phase 3, a matching task with form samples was added so that 11 pigeons received many-to-one (MTO) training while 6 received one-to-one (OTO) training. Differential sample responding was not required for the new task, but it emerged for 7 MTO birds and 1 OTO bird. Adventitious differential reinforcement of the emergent behavior was clearly indicated for 5 of these MTO birds. The adventitious differential reinforcement hypothesis was generally supported, but little support was found for the naming hypothesis, which strongly predicted emergent sample-specific responding for OTO birds. In Experiment 2, the color samples were mapped onto new comparisons. Then the form samples were tested in a single session for transfer to the new comparisons. Those pigeons that had differentiated their responses to the form samples tended to show greater positive transfer, supporting the contention that the sample-specific behavior apparent in Experiment 1 is equivalent to the coding response that mediates positive transfer following many-to-one training.