Referential learning effects on behavioral value and behavior causation

According to goal-directed models of behavior causation, the degree to which an agent expects a response to lead to a valued outcome determines the value they attribute to it and their engagement in this response. Conversely, a response that is not expected to lead to a valued outcome should not be valued or enacted. However, it is conceivable that a response that used to lead to a positive outcome retains some degree of residual value from these past co-occurrences. This idea stems from findings that an evaluative conditioning effect - a change in the valence of a conditional stimulus (CS) after having been paired with an unconditional stimulus (US) - can subsist even if the CS no longer signals the occurrence of the US. We tested whether the same principles apply to response values, that is, whether pairing a response with an outcome changes the value of the response and whether this value remains when the response is no longer paired with the outcome. We furthermore assessed whether any potential residual value stems from (a) an intact referential relation between the response and its past outcome or (b) a change in the intrinsic value of the response, and (c) whether such a residual value would be sufficient for response engagement. Using a novel task, a learned contingency between a response and a valued outcome was removed and liking of and engagement in the response were compared to those of a response for which the same contingency was absent from the start. In line with goal-directed models, we did not observe residual liking nor responding for the response that previously led to the positive outcome relative to the response that never did.