The Processing of Immediate and Delayed Feedback – A Critical Role of the Learned Type of Association?

从反馈中学习使我们能够适应环境。最近的证据表明，反馈加工的神经机制依赖于反馈预测事件和反馈本身之间的延迟。虽然对于即时反馈来说，纹状体很重要，但延迟几秒钟的反馈会招募海马体。然而，到目前为止，即时反馈学习和延迟反馈学习在认知和神经机制上的差异还没有被清楚地阐明。我们想要检验的主要假设是，反馈延迟对反馈加工的影响可以通过即时反馈和延迟反馈的不同信用分配机制来解释。信用分配指的是，如果其他事件介入，如何在事件和反馈之间建立关联的问题。最近，人们发现了一种基于时间邻近性的学习机制。这种学习机制是由纹状体和皮质之间的运动回路调节的，这表明它更多地指的是动作与反馈的整合，而不是刺激-反馈联系。因此，我们假设FRN，一种与纹状体和前扣带皮质相连的事件相关电位(ERP)成分，在动作反馈方面比刺激反馈联系更明显，但仅限于即时反馈。对于延迟反馈，我们假设占优势的是一种不同的信用分配机制，即反馈锁定的大脑区域重新激活，代表相关事件的处理。因此，反馈锁定的大脑活动应该不同于动作反馈和刺激反馈的关联。因此，我们假设，在纹状体外可视区域产生的事件相关电位成分N170对(视觉)刺激反馈比动作反馈关联更明显，但仅对延迟反馈。最后，功能神经成像将被用来进一步支持反馈锁定重新激活在延迟反馈加工中起重要作用的假设。具体地说，我们假设不同类别的视觉刺激和反馈之间的关联导致了反馈锁定大脑活动的不同模式，这种模式由海马体介导，只对延迟的反馈，而不是立即的反馈。机器学习技术(多体素模式分析)将被用来从反馈锁定的大脑活动中预测相关刺激(人脸或场景)的类别。

Learning from feedback enables us to adapt to the environment. Recent evidence has shown that the neural mechanisms of feedback processing depend on the delay between the feedback-predicting event and the feedback itself. While for immediate feedback the striatum is important, feedback that is delayed by a few seconds recruits the hippocampus. To date, however, the differences in the cognitive and neural mechanisms between learning from immediate and delayed feedback have not clearly been elucidated. The main assumption we want to test in this project is that feedback delay effects on feedback processing can be explained by different mechanisms of credit assignment for immediate and delayed feedback. Credit assignment refers to the question, how an association between an event and feedback can be established, if, for example, other events intervene. Recently, a learning mechanism based on temporal proximity has been identified. This learning mechanism is mediated by a motor circuit between striatum and cortex, suggesting that it refers more to the integration of actions with feedback than to stimulus-feedback associations. Accordingly, we hypothesize that the FRN, an event-related potential (ERP) component linked to the striatum and anterior cingulate cortex, is more pronounced for action-feedback than stimulus-feedback associations, but only for immediate feedback. For delayed feedback, in turn, we assume that a different mechanism of credit assignment prevails, namely the feedback-locked reactivation of brain regions representing the processing of the associated event. Feedback-locked brain activity should thus differ between action-feedback and stimulus-feedback associations. Accordingly, we hypothesize that the N170, an ERP component generated in extrastriate visual areas, is more pronounced for (visual) stimulus-feedback than action-feedback associations, but only for delayed feedback. Finally, functional neuroimaging will be used to further support the assumption that feedback-locked reactivation plays an important role for delayed feedback processing. Specifically, we hypothesize that associations between different categories of visual stimuli and feedback lead to distinguishable patterns of feedback-locked brain activity, mediated by the hippocampus, only for delayed but not immediate feedback. Machine-learning techniques (multivoxel pattern analysis) will be used to predict the category of the associated stimulus (faces or scenes) from feedback-locked brain activity.