Two-alternative learning situations with partial reinforcement

The comparative effects of reward and nonreward on learning are considered in connection with a two-alternative learning situation. Conditions are more general in nature than those discussed in an earlier article. In the statistical model proposed, the question of whether reward and nonreward are equivalent in their effects on learning reduces to testing a composite hypothesis on a multivariate probability distribution. An asymptotic test of this hypothesis is described, and its use is illustrated with data from psychological experiments.This paper was prepared with the partial support of the Office of Naval Research, and may be reproduced in whole or in part for any purpose of the United States government.I wish to express my gratitude to Professors J. Neyman and E. L. Scott of the University of California, Department of Statistics, for their constant assistance and encouragement throughout the research that led to this paper and during its preparation, and to Professor F. W. Irwin of the University of Pennsylvania Department of Psychology for his many helpful suggestions and comments.     

Partial reinforcement and partial delay of reinforcement effects with 72-hour intertrial intervals and interpolated continuous reinforcement

Three groups of rats ran 108 trials in a straight runway, one trial every 3 days. On the first 44 trials, one group received continuous (and immediate) reinforcement (CRF), a second group 50 per cent partial reinforcement (PRF), and the third group a 50 per cent schedule of partial delay of reinforcement (PDR). All groups received CRF on the next 20 trials, and extinction on the last 44 trials. The PRF and PDR groups extinguished at approximately the same rate, and significantly more slowly than the CRF group.     

A partial reinforcement extinction effect despite equal rates of reinforcement during Pavlovian conditioning

In 4 experiments rats received appetitive Pavlovian conditioning followed by extinction. Food accompanied every trial with the conditioned stimulus (CS) for the continuously reinforced groups and only half of the trials for the partially reinforced groups. In contrast to previous experiments that have compared the effects of partial and continuous reinforcement, the rate at which food was delivered during the CS was the same for both groups. The strength of the conditioned response during extinction weakened more rapidly in the continuously than in the partially reinforced groups. The results demonstrate that the partial reinforcement extinction effect is a consequence of the nonreinforced trials with the CS, rather than the rate at which the unconditioned stimulus is delivered during the CS.     

Resistance to continuous delay of reinforcement or extinction following partial delay or partial reinforcement in acquisition: A direct comparison

Rats received either partial reward (PR) or partial delay (PD) in acquisition with one, two, or three delay or nonreward trials followed by an immediately rewarded trial or one delay or nonreward trial followed by an immediately rewarded trial. These four groups were then split in half and given either continuous delay or continuous nonreward (extinction) in a “response persistence” phase. In addition, two continuously reinforced groups, one experiencing continuous delay, and the other experiencing extinction were included. The results showed that response persistence was greater when PD groups were given continuous delay rather than extinction, but the opposite was true for PR groups. The “length” (1, 2, or 3 versus 1 nonreward or delay trial) also transferred to the response persistence phase with the length 1, 2, and 3 conditions being more persistent than the 1 length conditions. The results were discussed with respect to theoretical notions of response persistence.     

The free-operant partial reinforcement effect: A discrimination analysis

Twelve continuously reinforced rats were extinguished in a modified operant chamber, with an S^D for goal-approach after each bar-press. Three groups of twelve rats each trained under FR 10 were extinguished with the S^D for goal-approach after every 6, 10, or 14-bar-presses, respectively. Results showed: (1) number of bar-presses to extinction was a direct function of the bar-press to S^D ratio during extinction, (2) number of food or goal-approaches was independent of both training and extinction conditions, and (3) prior to the breakdown in discrimination a revised response-unit hypothesis, based on a discrimination analysis of the effect of intermittent reinforcement, accurately predicted the number of bar-presses for each FR group.     

A further study of partial reinforcement in the turtle

Two matched groups of mature painted turtles, Chrysemys picta picta, were trained in a simple runway, one with partial and the other with consistent reinforcement, following which both groups were extinguished. The partially reinforced animals ran more slowly in acquisition, but showed somewhat greater resistance to extinction. The results are compared with those obtained in analogous experiments with other animals.