University of Amsterdam, Department of Psychology, Amsterdam, The Netherlands; Radboud University, Behavioural Science Institute, Nijmegen, The Netherlands.
University of Amsterdam, Department of Psychology, Amsterdam, The Netherlands; Vrije Universiteit, Department of Cognitive Psychology, Amsterdam, The Netherlands.
Neuroimage. 2018 Nov 1;181:170-181. doi: 10.1016/j.neuroimage.2018.07.014. Epub 2018 Jul 7.
Reinforcement learning (RL) in humans is subserved by a network of striatal and frontal brain areas. The electrophysiological signatures of feedback evaluation are increasingly well understood, but how those signatures relate to the use of feedback to guide subsequent behavioral adjustment remains unclear. One mechanism for post-feedback behavioral optimization is the modulation of sensory processing. We used source-reconstructed MEG to test whether feedback affects the interactions between sources of oscillatory activity in the learning network and task-relevant stimulus-processing areas. Participants performed a probabilistic RL task in which they learned associations between colored faces and response buttons using trial-and-error feedback. Delta-band (2-4 Hz) and theta-band (4-8 Hz) power in multiple frontal regions were sensitive to feedback valence. Low and high beta-band power (12-20 and 20-30 Hz) in occipital, parietal, and temporal regions differentiated between color and face information. Consistent with our hypothesis, single-trial power-power correlations between frontal and posterior-sensory areas were modulated by the interaction between feedback valence and the relevant stimulus characteristic (color versus identity). These results suggest that long-range oscillatory coupling supports post-feedback updating of stimulus processing.
人类的强化学习（RL）由纹状体和额前脑区网络支持。反馈评估的电生理特征越来越被了解，但这些特征如何与利用反馈来指导后续行为调整相关尚不清楚。反馈后行为优化的一种机制是对感觉处理的调制。我们使用源重建 MEG 来测试反馈是否会影响学习网络和与任务相关的刺激处理区域中振荡活动源之间的相互作用。参与者在概率 RL 任务中使用试错反馈学习彩色面孔和反应按钮之间的关联。多个额区的 delta 频带（2-4 Hz）和 theta 频带（4-8 Hz）功率对反馈效价敏感。枕部、顶叶和颞叶区域的低和高 beta 频带功率（12-20 和 20-30 Hz）区分了颜色和面孔信息。与我们的假设一致，额区和后感觉区之间的单试功率-功率相关性受反馈效价与相关刺激特征（颜色与身份）之间的相互作用调节。这些结果表明，长程振荡耦合支持反馈后刺激处理的更新。
