A Partial Reinforcement Extinction Effect Despite Equal Rates of Reinforcement During Pavlovian Conditioning.

Abstract

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.For many years the partial reinforcement extinction effect (PREE) was a mainstay of research in the field of animal learning and behavior. Within the context of instrumental conditioning, the PREE can be defined as a resistance to the detrimental effects of extinction following trials in which responding is intermittently rather than consistently paired with reinforcement (for a review see Mackintosh, 1974). So widespread was the PREE in studies of instrumental conditioning-and such was the paucity of evidence for the effect within Pavlovian conditioning-that for some time the PREE was thought to differentiate these two forms of conditioning (Kimble, 1961). However, the possibility that the PREE contradicted the single-process rule of learning (Guthrie, 1935;Hull, 1943) has not stood the test of time. Many studies of Pavlovian conditioning have now demonstrated that a conditioned stimulus (CS) that is intermittently paired with an unconditioned stimulus (US) will show a resistance to extinction relative to a CS that is consistently paired with a

Pavlovian conditioning and cumulative reinforcement rate.

“…The Partial Reinforcement Extinction Effect (PREE) refers to the observation that a CS (or response) that has been reinforced on every trial will undergo more rapid extinction than a CS (or response) that has been reinforced on only a fraction of trials (Mackintosh, 1974). This difference in extinction has been observed even for CSs that have been matched for cumulative reinforcement rate during conditioning by, for example, using short presentations of the partial CS and long presentations of the continuous CS (Bouton, Woods, & Todd, 2014;Haselgrove, Aydin, & Pearce, 2004). While the experiments by Haselgrove et al and Bouton et al used fixed CS-US intervals, and are therefore susceptible to issues of response timing, it is not obvious how timing of US expectancy would produce the difference in extinction they reported.…”

Appetitively motivated instrumental learning in SynGAP heterozygous knockout mice.

“…This is because the impact of reinforcement schedules on extinction learning described above is preferentially seen in appetitive instrumental learning. It is still debatable whether equivalent manipulation in Pavlovian conditioning (e.g., by degrading the CS-US contingency) would reliably yield a PREE-like phenomenon (Gormezano & Coleman, 1975;Miller & Capaldi, 2006; but also see Haselgrove, Aydin, & Pearce, 2004;Haselgrove & Pearce, 2003;Pearce, Redhead, & Aydin, 1997), because interpretation would be severely confounded by expected changes in the acquired CS-US associative strength (Rescorla & Wagner, 1972). Hence, our failure to observe any changes in the extinction of a conditioned response acquired by means of Pavlovian conditioning is not too surprising.…”

Perceptual learning and representational learning in humans and animals

“…These results could not be captured either by the Pearce-Hall associative model or by a Bayesian model that did not develop a representation of the world by a generative world model. Finally, Courville et al (2006) used the partial reinforcement extinction effect (Gibbon, Farrell, Locurto, Duncan, & Terrace, 1980;Haselgrove, Aydin, & Pearce, 2004;Rescorla, 1999) to show that fixed-rate stochastic reinforcement might not induce faster learning. This is an interesting result, because the Pearce-Hall and the Bayesian models provide opposite predictions, and animal behavior follows the prediction of the Bayesian model.…”