CONTRAST, GENERALIZATION, AND THE PROCESS OF DISCRIMINATION¹

Abstract

A combination of generalization gradients for reinforcement and extinction fails to predict the generalization gradient produced by training a discrimination between two stimuli (Honig, Thomas, & Guttman, 1959;Hanson, 1959). One reason for this failure is that the usual generalization gradient for extinction reflects only an inhibitory effect of extinction, whereas extinction has not only an inhibitory effect but also an excitatory one. The inhibitory effect is an over-all decrease in responding over the stimulus continuum.The excitatory effect appears during the formation of a discrimination as an increase in responding in the presence of the stimulus correlated with reinforcement. This increase is called behavioral contrast . Moreover, an adequate description of the process of discrimination must take into account both the inhibitory and the excitatory effects of extinction.The primary purpose of this research is to compare generalization gradients for extinction and reinforcement obtained by a procedure that allows both effects of extinction to be measured. A second purpose is to examine the contribution of contrast to postdiscrimination generalization gradients. In this method, a pigeon's key pecking is first reinforced in the presence of ten values of a stimulus continuum. To obtain the reinforcement gradient, responding is then reinforced in only two of the values and extinguished in the remaining eight; or to obtain the extinction gradient, responding is extinguished in two of the values and reinforced in eight. With this method of maintained generalization gradients (cf. Pierrel, 1958), the effect of extinction on reinforced responding can be measured as increases in the rate of responding over the base-line rate maintained by reinforcement in all ten stimuli. METHOD ApparatusA modified picnic icebox contained the experimental chamber (Ferster Sc Skinner, 1957). A transparent, but otherwise standard, response key, 0.75 inch in diameter and faced with cover glass to minimize scratching, was mounted on one wall, which was painted a uniform flat black. An effective force of about 15 grams operated the key. The pigeon occasionally had access to grain for 3 seconds (reinforcement) through a 2-by 2-inch opening beneath the key. 'The present research was supported by grant NSF-G8621 to Harvard University from the National Science Foundation.Two 6-watt white lamps mounted behind the panel, one on each side of the key, were the only illumination in the chamber. A white masking noise was always present.The stimulus continuum was the spatial orientation of the apex of a triangle. A black isosceles triangle, 0.5 inch high and 0.25 inch along the base, was mounted on a white background 0.5 inch behind the transparent key. This triangle and the background were attached to a synchronous motor that slowly but continually rotated clockwise at a speed of slightly less than 0.19 degree per second (about twice the speed of the minute hand of a clock). The rotation of the triangle was in a plane perpendicular to the pigeon's...

SOLUTION OF THE INTERMEDIATE SIZE PROBLEM BY PIGEONS1

“…These failures to obtain any evidence for the role of relationships supported Spence's (1937Spence's ( , 1942 contention that the specific stimuli of the training set control the stimulus preferences in a test. Pigeons were the first species given the intermediate size problem to behave consistently as predicted by Spence' (Reynolds, 1961;Catania and Gill, 1964). Such behavior, called behavioral contrast by Skinner (1938), also was investigated by Pavlov (1927) who called it positive induction.…”

Orthogonal stimulus variation and attention in dimensional contrast

“…visual flicker rate (e.g., D. S. Blough, 1975;Catania & Gill, 1964;Hinson, 1988;Hinson & Higa, 1989;Reynolds, 1961). In addition, dimensional contrast is typically a large effect that can easily represent a 50%-100% increase in response rate (e.g., Hinson & Tennison, 1997).…”

“…The model states that discrimination performance is based on the joint contribution of a function representing discriminability between stimuli and a function representing the allocation of attentional resources. The discriminability function is given by Previous work has focused on the phenomenon of dimensional contrast, a reliable finding from maintained generalization procedures (e.g., D. S. Blough, 1975;Hinson & Higa, 1989;Hinson & Malone, 1980;Reynolds, 1961). Dimensional contrast refers to an enhancement of discrimination performance for relatively similar positive (S+) and negative (S -) training stimuli, compared with more dissimilar stimuli.…”

Stimulus variation and dimensional contrast

“…One of the more prominent results from maintained generalization studies of dimensional stimulus control is dimensional contrast (see D. S. Blough, 1975;Catania & Gill, 1964;Malone & Staddon, 1973;Reynolds, 1961). Dimensional contrast refers to the relative enhancement of response differences in border regions of empirical stimulus control gradients.…”

“…Dimensional contrast effects have been observed in a variety of discrimination tasks, including ones involving spatial orientation stimuli (e.g., Catania & Gill, 1964;Malone, 1975;Reynolds, 1961), visual wavelength stimuli (D. S. Blough, 1975), auditory amplitudes (P. M. Blough, 1980), and visual flicker rates (Hinson, 1988). Furthermore, the effects are dependent on relative rates of reinforcement during S+ and S-, rather than on the presence or absence of reinforcement (D. S. Blough, 1975).…”