Steepened generalization gradients after massed extinction to the CS

Steepened generalization gradients after massed extinction to the CSl Eliot Hearst. NATIONAL INSTITUTE OF MENTAL IlEALTH2 Roger Popp4"n. S'DtNFORD UNIVERSITY Ab8traet Pigeons received intermitient reinforcement for pecking at a vertical-line stimulus. Ss were then tested for generalization to other line orientations under one of the following conditions: (a) 20 min. of massed extinction to the vertical line, followed by a standard generalization test; (b) 20 min. of massed extinction to a horizontal line, followed by the generalization test; or (c) generalization test without any prior extinction. Both pre-extinguished groups showed steeper gradients than the group that had not been extinguished. Problem A popular technique for obtaining stimulus generalization gradients involves the presentation of several different stimulus values to individual Ss during extinction of the conditioned response. The results of such tests are complicated by the fact that gradients of relative generalization become reliably steeper as extinction progresses (e. g., Thomas & Barker, 1964; Friedman & Guttman, in press). Therefore the steepness of gradients depends to some extent upon the duration of the generalization test. Many writers interpret this progressive steepening to mean that the slope of separate extinction curves for different test stimuli varies with the proximity of the stimulus to the CS; the rate of extinction is presumably more -rapid at stimuli far from CS than at values close to CS. There is, however, another possibility to be considered in analyzing these results. Perhaps progressive steepening is due to general effects of extinction and would occur even if a variety of different stimuli were not presented during extinction. This possibility could 32 Pre-extinction af 90 D 28 ~ z Pre-extinc tion 01 0° 24 ili 20 « ~ 16 z 12 ::; ::: 8 4 0 -90 -45 0 DEGREES OF TILT Fig. I. Gradients of relative generalization for the three experimental groups. Psychon. Sci., 1965, Vol. 2. be tested by extinguishing responses at one stimulus value for a prolonged period before administering the standard generalization test. If such a period of massed extinction prior to generalization testing produces gradients that are reliably steeper than those without prior extinction, it would suggest that the opportunity for Ss to compare stimuli during testing, and thus to extinguish at differential rates to the various test stimuli, is not critical in accounting for the steepeningin-extinction effect. Therefore, in this experiment, gradients for two groups of Ss, each given massed extinction immediately before standard generalization testing, were compared with gradients from a group that received no preextinction. Me,hod The Ss were 30 experimentally naive White King hen pigeons, maintained throughout the study at 75% of their free-feeding weights. Details of the key pecking apparatus, the method of varying line tilt, and the general procedure have been described in an earlier report (Hearst et al., 1964). Fourteen days of training on a I-min. variable interval (VI) schedule for grain reinforcement preceded generalization test days. Throughout these training periods, and all subsequent tests, 30-sec . stimulus-on periods alternated with 10-sec. stimulus-off (blackout) periods. During stimulus-on periods the CS (vertical line) was projected on the key and Vl reinforcement was possible. During stimulusoff periods the chamber was dark and no reinforcement was possible. All sessions lasted 50 min. The Ss were randomly assigned to three different experimental groups , each with N = 10, for generalization testing on Day 15. On that day each S was first given a 10-min. warmup period under the same reinforcement conditions as before. This warmup period was immediately followed by one of thre e experimental treatments, depending On the group to which each S had been assigned: For Group I, there was a 20-min.periodduringwhichthe vertical line continued to appear on the key during stimulus-on periods but no key pecks were reinforced. This period of massed extinction waS immediately followed by a standard generalization test. The treatment for Group 2 was exactly the same as for Group 1 except that during massed extinction a horizontal line was projected on the key. Group 3 received a standard generalization test without any previous period of massed extinction. Eight different orientations of the stimulus line were presented in a mixed order during generalization tests. There were 10 presentations of each line orientation for a total of 80 stimulus-on periods. Ss could not obtain any reinforcement during generalization tests. Results Gradients of relative generalization are shown in Fig. 1. On the abscissa the CS is deSignated as 00 and the horizontal line as +90 0 and -90 0 , as is conventional in studies employing the line-tilt dimension. The groups that had been extinguished at either the 00 or 90 0 stimulus both showed steeper generalization gradients than the group that had received no prior extinction. Analysis of variance of the individual values that entered into the computation of these relative 83 500 400 (J) w (J) z 0 :l; 300 w ..g .. a: ...J f- 200 z w ::E 100 "'- Pre-exline/ion 0 -90 -45 010· 0 + 45 + 90 OEGREES OF TILT Fig. 2. Gradients of absolute generalization for tbe three experimental groups. gradients revealed a significant interaction between "stimuli" and "extinction treatment" (F= 2.56; df= 14/189; p< .01), which supports the conclusion that generalization gradients are reliably steeper following a period of massed extinction. In an additional analysis, an index of gradient slope (Thomas & Barker, 1964) was obtained for each individual S in the experiment by calculating the percent of each S's total generalization test responses that occurred to the CS (i. e., Responses to the 00 Stimulus/ Total Responses to All Eight Stimuli). When these indices were analyzed between groups by appropriate MannWhitney U Tests, differences between the "no preextinction" group and either of the two extinguished groups were significant at beyond the .02 level (2-tail test). Slope differences between the two extinguished groups were not significant (U= 38). Figure 2 displays the absolute response frequencies from which the relative gradients of Fig. I were derived. The group that had been pre-extinguished at the CS emitted the fewest total responses during the test, the oiher extinguished group an intermediate number of responses, and the nonextinguished group the most responses . Differences in total test responses between the nonextinguished group and the group pre-extinguished at the CS were significant beyond the .02 level (MannWhitney Test), but the other group differences in total test responses did not attain an acceptable level of significance. As would be expected, Ss given massed extinction at the CS made many more responses (Mean =936.1 responses) (...truncated)