A transformation of self-discrimination response functions in accordance with the arbitrarily applicable relations of sameness, more than, and less than.

In Experiment 1, 2 experimental subjects were given pretraining of nonarbitrary relations that brought their responses under the control of four contextual stimuli; same, opposite, more than, and less than. One control subject was not exposed to this pretraining. The 2 pretrained subjects and the 3rd nonpretrained subject then received training in six arbitrary relations, the following four relations being the most critical: same/A1-B1, same/A1-C1, less than/A1-B2, more than/A1-C2. All 3 subjects were then tested for seven derived relations, the following three relations being the most important: same/B1-C1, more than/B1-C2, less than/B1-B2. The 2 pretrained subjects, but not the nonpretrained subject, showed the derived relations. One of the stimuli (B1) from the relational network and two novel stimuli (X1 and X2) were then used to train three different self-discrimination responses on three complex schedules of reinforcement. That is, all 3 subjects were trained to pick X1 if they had not emitted a response, to pick B1 if they had emitted one response only, and to pick X2 if they had emitted two responses only. The 2 pretrained subjects, but not the nonpretrained subject, showed the predicted transformation of self-discrimination response functions in accordance with the relations of sameness, more than, and less than (i.e., no response, pick B2; one response only, pick C1; and two responses only, pick C2). In Experiment 2, 2 new subjects were employed, and the arbitrary relational training and testing phases were modified to control for a procedural artifact that may have contributed to the results of the first experiment. Experiment 2 replicated the findings of Experiment 1. The pattern of results support the utility of a relational frames approach to understanding derived stimulus relations.