Structural features in multiple response repertoires: Effects of response restriction.

Abstract

Data from three experiments are reported in which college students (Exp. 1), preschool children (Exp. 2), and Mongolian gerbils (Exp. 3) were given access to an experimental environment in which we measured four different unconstrained activities. Once these multiple response baselines were established, we restricted access to one of the four activities and observed the changes which occurred in the three unrestricted alternatives. In all three experiments, we typically observed that only one of the alternatives was selected to substitute for the restricted activity. Some structural characteristics of the baseline response repertoire were examined to determine if they could be used to predict which of the alternative activities would increase during response restriction. The data indicate that the temporal pattern exhibited by a response during baseline sessions was effective in predicting response changes during the subsequent restriction phase in all three experimental preparations. Specifically, the alternative response which had temporal patterns most similar to the restricted response substituted for the latter during the restriction period.

The Effects of Response Cost and Response Restriction on a Multiple‐response Repertoire With Humans

“…It has been over 20 years since Dunham's first propositions were presented (1971); it is time for their current status to be examined. The hierarchical rule has not been supported by any experiment with adult humans (Crosbie, 1988(Crosbie, , 1990a(Crosbie, , 1990c(Crosbie, , 1991Dunham et al, 1986, Experiment 1; the present experiments), and in those experiments in which it has been supported (Dunham et al, 1986, Experiments 2 and 3), the hierarchical and temporal similarity rules made identical predictions. Given these failures to replicate and the uncertain theoretical status of the rule since the abandonment of the implicit avoidance explanation, the rule should be considered suspect and perhaps abandoned.…”

“…Although most responses were reasonably stable, greater confidence could be placed in the results if more sessions were used and all responses were stable before the conditions were changed. Furthermore, a reversal (ABA) design (i.e., baseline, restriction or response cost, then a return to baseline) would highlight any changes produced by the experimental conditions, and make the procedure more similar to that of Dunham et al (1986). For these reasons, an ABA design was used in the present experiment and conditions did not change until all responses were stable.…”

“…Response cost temporarily reduced the rate of all unpunished responses, but they eventually recovered to their baseline levels. Dunham's hierarchical rule was not supported, but there was some support for the greatest follower proposition; the data were collected before those of Dunham et al (1986) were published, so the temporal similarity rule was not assessed. In another study (Crosbie, 1991), all 10 responses (0 to 9) were reinforced on a single schedule (VI 15 s, + 5 cents), but only the target response "3" was punished (-25 cents for every response).…”

“…In summary, Dunham's work with gerbils had three major findings: (a) Punishment increased the most probable unpunished baseline response, (b) punishment decreased the most frequent follower of the punished response, and (c) these effects were not produced only by the aversive aspects of punishment, because response restriction also had similar consequences. In an assessment of the external validity of these findings, Dunham, Cornwall, and Hurshman (1986) performed a systematic replication of the gerbil studies in which human subjects engaged in games or leisure activities, and response restriction was employed instead of electric shock. In Experiment 1, university students played pinball, pool, darts, and a video game in a multiple-response repertoire.…”

“…To maintain conditions comparable to those in Dunham and Crosbie's studies, all subjects received 13 sessions during Phase 1 (similar in number to those in Grantmyre, 1982, andCrosbie, 1990c, and more sessions than in Dunham et al, 1986) and 12 sessions during Phase 2 (response cost or response restriction). During the response cost phase, baseline conditions continued, but Response 3 also produced a 10-point loss for every second that the cursor was in Square 3.…”