Date created: | Last Updated:

Category: Project

Description: Influential models of causal learning and inference assume that learning about generative and preventive relationships are symmetrical to each other. This symmetry is most clearly captured in the Rescorla-Wagner (RW) model, where excitation and inhibition are represented by positive and negative associative strengths. In the laboratory, prevention learning is studied using a feature negative design (A+/AB- discrimination). Previous studies from our lab have shown that many participants do not infer a direct prevention causal structure when presented with a feature-negative discrimination with a unidirectional outcome (Lee & Lovibond, 2021; Lovibond & Lee, 2021). Melchers, Wolff and Lachnit (2006) suggested that direct prevention learning in humans might be encouraged by use of a bidirectional outcome that can either increase or decrease from baseline. Here we test the viability of this claim against an alternative possibility that a bidirectional outcome encourages encoding of an causal relationship in the opposite direction, where the negative feature (B) directly causes a reduction in the outcome. In Experiment 1, we found that a substantial proportion of participants reported such an “Opposite Causal” structure. These participants showed the lowest outcome predictions when B was combined with a novel cause in a summation test, and predicted a direct reduction in the outcome when B was presented alone. In Experiment 2, B blocked learning to a novel cue that was directly paired with a reduction in the outcome, and this effect was strongest among participants who endorsed an Opposite Causal structure. Together, these experiments provide the first demonstration of a causal mechanism of prevention learning that involves generative connections. We conclude that evidence of direct prevention learning in humans remain elusive, and suggest instead that a modulatory structure, akin to negative occasion-setting in animals, best accounts for inhibitory learning effects in humans.