Context-outcome associations underlie context-switch effects after partial reinforcement in human predictive learning

Abstract

Predictive value for continuously reinforced cues is affected by context changes when they are trained within a context in which a different cue undergoes partial reinforcement. An experiment was conducted with the goal of exploring the mechanisms underlying this context-switch effect. Human participants were trained in a predictive learning situation in which a cue received partial reinforcement while a target cue received continuous reinforcement in context A (C1) and another target cue was presented unreinforced in context B (U2). Participants in group Partial-One did not receive partial reinforcement in context B, while participants in group Partial-Both received the same training they received in context A, but with different cues. When target cues were tested in group Partial-One, greater responding in context A than in context B was found. Differences were smaller in cue U2 than in cue C1. No differences across contexts were found in group Partial-Both. These results are in agreement with the hypothesis that context-switch effects after partial reinforcement are mainly due to the formation of direct context-outcome associations, though the difference on the effect size on the reinforced and unreinforced cues suggests that a modulator mechanism may be also responsible for these context-switch effects.