Institute of Psychology, University of Osnabrueck, Osnabrueck, Germany.
Psychophysiology. 2021 May;58(5):e13789. doi: 10.1111/psyp.13789. Epub 2021 Feb 28.
In reinforcement learning, adaptive behavior depends on the ability to predict future outcomes based on previous decisions. The Reward Positivity (RewP) is thought to encode reward prediction errors in the anterior midcingulate cortex (aMCC) whenever these predictions are violated. Although the RewP has been extensively studied in the context of simple binary (win vs. loss) reward processing, recent studies suggest that the RewP scales complex feedback in a fine graded fashion. The aim of this study was to replicate and extend previous findings that the RewP reflects the integrated sum of instantaneous and delayed consequences of a singular outcome by increasing the feedback information content by a third temporal dimension. We used a complex reinforcement-learning task where each option was associated with an immediate, intermediate and delayed monetary outcome and analyzed the RewP in the time domain as well as fronto-medial theta power in the time-frequency domain. To test if the RewP sensitivity to the three outcome dimensions reflect stable trait-like individual differences in reward processing, a retesting session took place 3 months later. The results confirm that the RewP reflects the integrated value of complex temporally extended consequences in a stable manner, albeit there was no relation to behavioral choice. Our findings indicate that the medial frontal cortex receives fine graded information about complex action outcomes that, however, may not necessarily translate to cognitive or behavioral control processes.
在强化学习中，自适应行为取决于根据先前决策预测未来结果的能力。奖励正性（RewP）被认为在中前扣带皮层（aMCC）中编码奖励预测误差，只要这些预测被违反。尽管 RewP 在简单的二元（赢 vs. 输）奖励处理背景下得到了广泛研究，但最近的研究表明，RewP 以精细的分级方式对复杂反馈进行了缩放。本研究旨在复制和扩展先前的发现，即 RewP 通过增加第三个时间维度的反馈信息量来反映单一结果的即时和延迟后果的综合总和。我们使用了一项复杂的强化学习任务，其中每个选项都与即时、中期和延迟的货币结果相关联，并在时域中分析 RewP 以及时间频率域中的额眶额中线θ功率。为了测试 RewP 对三个结果维度的敏感性是否反映了奖励处理的稳定的特质样个体差异，我们在 3 个月后进行了重测。结果证实，RewP 以稳定的方式反映了复杂时间扩展后果的综合价值，尽管它与行为选择没有关系。我们的研究结果表明，额眶额中线皮层接收关于复杂动作结果的精细分级信息，但这些信息可能不一定转化为认知或行为控制过程。
