Pigeons' short-term memories for surprising vs. expected reinforcement and nonreinforcement

Abstract

Pigeons performed a delayed matching-to-sample task in which they matched red and green disks as comparison stimuli to samples of food and no food. The birds were also taught a discrimination between two lines: vertical (S+) followed by food and horizontal (S-) followed by no food. The two kinds of trials were then chained in infrequent probes such that (a) S+ and S- preceded samples of food and no food, (b) a longer than usual delay occurred, and then, (c) the comparison stimuli were presented. Therefore, in probes when S+ preceded food and S- preceded no food, the samples were "expected. " But in probes when S+ signaled no food and S- signaled food, the samples were "surprising. " Matching to surprising samples was more accurate than matching to expected samples. This result completes a pattern of findings implying that surprising reinforcers enhance learning and also persist (are longer rehearsed) in short-term memory. © 1979 Psychonomic Society, Inc.