Still at the choice-point: action selection and initiation in instrumental conditioning

Contrary to classic stimulus-response (S-R) theory, recent evidence suggests that, in instrumental conditioning, rats encode the relationship between their actions and the specific consequences that these actions produce. It has remained unclear, however, how encoding this relationship acts to control instrumental performance. Although S-R theories were able to give a clear account of how learning translates into performance, the argument that instrumental learning constitutes the acquisition of information of the form "response R leads to outcome O" does not directly imply a particular performance rule or policy; this information can be used both to perform R and to avoid performing R. Recognition of this problem has forced the development of accounts that allow the O and stimuli that predict the O (i.e., S-O) to play a role in the initiation of specific Rs. In recent experiments, we have used a variety of behavioral procedures in an attempt to isolate the processes that contribute to instrumental performance, including outcome devaluation, reinstatement, and Pavlovian-instrumental transfer. Our results, particularly from experiments assessing outcome-selective reinstatement, suggest that both "feed-forward" (O-R) and "feed-back" (R-O) associations are critical and that although the former appear to be important to response selection, the latter-together with processes that determine outcome value-mediate response initiation. We discuss a conceptual model that integrates these processes and its neural implementation.