Neuro-cognitive mechanisms of learning and implementing context-dependent action-effect associations as a prerequisite for goal-directed behavior
Final Report Abstract
The question how people manage to act in a goal-oriented manner is one of the core questions in psychology. A fundamental prerequisite of goal-oriented behavior is to correctly recognize contingencies between a certain action (e.g., pressing a green button) performed in a certain context (e.g., the coffee maker in the cafeteria) leading to a certain perceivable effect or outcome (e.g., coffee). What is remarkable is that people are extremely fast in learning such contingencies. By means of behavioural measures and functional magnetic resonance imaging (fMRI), one main goal of the proposed project was to elucidate the neurocognitive mechanisms that underlie the rapid learning and retrieval of such contingencies. By focusing on non-incentive action outcomes (e.g., different actions followed by different naturalistic sounds) our project contrasts with previous studies in the tradition of the instrumental conditioning framework which have typically emphasized the motivational role of incentive action effects (e.g., different actions followed by different primary rewards like chocolate or juice). Moreover, by focusing on rapid initial learning processes our project contrasts with previous studies in the tradition of the ideomotor learning framework which typically assessed behavioural and neural measures after longer practicing periods following the introduction of novel contingencies (emphasizing the automaticity of action-outcome integration processes). One of the most important insights from our studies is that despite the mentioned differences to previous approaches, similar brain regions were nevertheless engaged during the initial learning of novel non-incentive action-outcome contingencies. This included areas belonging to the classical ‘reward system’ such as the anterior striatum and orbitofrontal cortex regions which have been implicated in previous instrumental learning studies involving incentive outcomes. But this also included the supplementary motor area and certain parietal cortex areas previously implicated in studies examining more automatized outcome integration processes. On a more fine-grained level, we could show that the specific involvement of these areas dependent on several additional constraints. For instance, when the novel contingencies were explicitly instructed, orbitofrontal cortex and striatal areas were specifically engaged via functional couplings with the lateral prefrontal cortex. By contrast, when novel contingencies were incidentally learned, orbitofrontal cortex and striatal areas were engaged independently of the lateral prefrontal cortex. This suggests that the same action-outcome contingencies can be learned through quite different neural processes, yet always under involvement of similar striatal and orbitofrontal regions. Moreover, we found that the engagement of some of these areas in action outcome integration process was additionally sensitive to the general predictive context. For instance, the supplementary motor area was most strongly engaged when an additional antecedent stimulus predicted that a certain response would entail a certain outcome. Similar effects were present in striatal and orbitofrontal sub-regions, but these were found to be additionally depending on meta-learning processes generating insight into the existence of specific contingencies under incidental learning conditions. Finally, this all said, action-outcome learning not only involved classical reward system areas but it also involved mid-to-posterior sections of the putamen. This is an area that has previously been implicated in habitual behavior rather than goal-oriented behavior. Importantly, however, learning-related changes in the putamen were predictive of subsequent action-outcome strength only when action-outcome associations were triggered by the perception of the respective outcome but not when nominally the same associations were internally activated by the anticipation of the outcome of one’s own future action. Together this suggests that actionoutcome contingencies are learned by both the reward system and the habit system, but the conditions under which these two neural representations are retrieved differ and the way they differ is, in turn, consistent with previous distinctions between habitual and goal-directed behavior.
Publications
- (2012). Early markers of ongoing action-effect learning. Frontiers in Psychology, 3:522
Ruge, H., Krebs, R. M., & Wolfensteller, U.
- (2012). Frontostriatal mechanisms in instruction-based learning as a hallmark of flexible goal-directed behavior. Frontiers in Psychology, 3:192
Wolfensteller, U., & Ruge, H.
- (2013). Functional integration processes underlying the instruction-based learning of novel goal-directed behaviors. Neuroimage, 68, 162-172
Ruge, H., & Wolfensteller, U.
(See online at https://doi.org/10.1016/j.neuroimage.2012.12.003) - (2014). Response selection difficulty modulates the behavioral impact of rapidly learnt action effects. Front Psychol, 5, 1382
Wolfensteller, U., & Ruge, H.
(See online at https://doi.org/10.3389/fpsyg.2014.01382) - (2015). Distinct fronto-striatal couplings reveal the double-faced nature of response-outcome relations in instruction-based learning. Cognitive, Affective and Behavioral Neuroscience, 15, 349-364
Ruge, H., & Wolfensteller, U.
(See online at https://doi.org/10.3758/s13415-014-0325-4) - (2016). The neural basis of integrating pre- and postresponse information for goal-directed actions. Neuropsychologia, 80, 56-70
Frimmel, S., Wolfensteller, U., Mohr, H., & Ruge, H.
(See online at https://doi.org/10.1016/j.neuropsychologia.2015.10.035)
