Investigating the effects of pretraining associations between compound-stimulus elements on stimulus overselectivity

When a compound stimulus comprising multiple elements signals the contingencies for responding, not all those elements necessarily come to control behavior. Previous research suggests that such stimulus overselectivity depends on the strength of the associations between the elements, with stronger within-compound associations attenuating overselectivity. We investigated whether establishing associations between stimulus elements in a separate pretraining phase would modulate overselectivity. During pretraining, 384 adult human participants completed a matching-to-sample task in which associations between elements were trained directly, indirectly via many-to-one training, or not at all. Then, they chose between two compound stimuli composed of pretrained elements, or of novel elements, in a simultaneous-discrimination task. Overselectivity was assessed by measuring choice of individual stimulus elements in an unreinforced test phase. There were no significant differences in overselectivity levels depending on pretraining history, perhaps because the effects of pretraining did not generalize or the type of association formed during pretraining differed from the within-compound associations that modulate overselectivity. Nevertheless, overselectivity did tend to be lower when elements had been associated together directly or indirectly, providing some preliminary evidence that pretraining associations may reduce overselectivity. Overall, the present study highlights the need for further work to establish the generality of our findings, and to identify the boundary conditions under which prior experience with individual stimulus elements does, or does not, modulate stimulus overselectivity.