A step function in pigeons’ temporal generalization in the peak shift task

Abstract

Pigeons were reinforced for pecking a key following one signal duration (S+) but not following another signal duration (S-). The S+ and S-were 2.52 and 5.67 sec, respectively, counterbalanced across birds. Subsequent generalization tests with a range of signal durations revealed a step function, with high response rates for all durations on the S+ side of the distribution, low response rates for all durations on the S-side, and an intermediate rate for the intermediate duration. A comparison group of pigeons trained with only the S+ duration showed a flat generalization function. For the discriminationtrained birds, the delay between signal termination and opportunity to respond was subsequently varied during generalization testing. A step function again appeared, and no evidence of subjective shortening over the delay was found. The overall pattern of results suggests that the birds categorized the temporal signal into two classes and retained a categorical code over the delay.Understanding how antecedent stimuli gain control over reinforced responding is an important part of the experimental analysis ofbehavior. Stimulus control is reflected by changes in responding as a function of changes in some aspect of the training stimulus. Generalization (similar responding despite variation in the stimulus) and discrimination (responding differently when the stimulus is changed) are opposite ends of a continuum of stimulus control.Guttman and Kalish (1956) conducted a classic study ofstimulus control. Pigeons were intermittently reinforced for pecking a key in the presence of a monochromatic light source. The spectral stimulus used in training varied across groups of pigeons. Subsequent generalization tests with a range of spectral wavelengths on either side of and including the training value showed orderly generalization gradients, with highest response rates at the training value and decreasing rates of responding at increasingly different test values.The nature ofstimulus generalization has since received a great deal of experimental attention. One general finding has been that generalization gradients become sharper after discrimination training, in which periods of reinforced responding in the presence of a positive training This research was supported by a Natural Sciences and Engineering Research Council ofCanada research grant held by the first author. The authors thank M. Johnston, D. Kelly, and S. Reid for assistance with the research. Correspondence should be addressed to M. L. Spetch, Department of Psychology, University of Alberta, Edmonton, Alberta, Canada, T6G 2E9 (e-mail: mspetch@psych.ualberta.ca).ortoK.Cheng. School of Behavioural Sciences, Macquarie University, Sydney, NSW 2109, Australia (e-mail: kcheng@bunyip.bhs.mq.edu.au).-Accepted by previous editor; Robert A. Rescorla stimulus (S+) are intermixed with periods during which responding is not reinforced in the presence of a negative stimulus (S-). Moreover, intradimensional discrimination training, in which the S+ and S-differ only along the dim...

Temporal generalization and peak shift in humans

“…Spetch and Cheng (1998), with pigeons, found that the generalization gradients that the birds produced resembled step functions best. The birds responded at low rates in the presence of the S and all stimuli that were more similar to the S than to the S and responded at high rates in the presence of the S and all stimuli that were more similar to the S than to the S .…”

“…Our generalization tests did not provide any feedback to participants, whereas Spetch and Cheng (1998) interspersed reinforced S trials among the test trials, although in one experiment generalization testing was conducted in extinction. Perhaps the continued presence of reinforcement and the extended training and testing encouraged the pigeons to treat the durations as belonging to one of two categories, either "S "-like or "S "-like.…”

“…In one study, Spetch and Cheng (1998) investigated peak shift with stimulus duration, yet this experiment did not produce peak shift. Pigeons' generalization gradients were best described as a step function following intradimensional discrimination training.…”

“…The pigeons responded at low rates in the presence of the S and all stimuli that were more similar to the S than to the S , and they responded at high rates in the presence of the S and all stimuli that were more similar to the S than to the S . Spetch and Cheng (1998) suggested that the pigeons had responded as if the stimuli were categorical.…”

Generalisation: mechanistic and functional explanations

“…Yet other gradients are step functions, low on the S-side, high on the S+ side (e.g. Spetch and Cheng 1998).…”