Neural Mechanisms of Rule-Based Behavior

项目摘要/摘要 在我们的一生中，我们学会了成百上千条规则，这些规则定义了我们在特定情况下应该如何行动。例如, 在餐馆里，我们遵循一套规则，指导我们点餐、用餐和付款的方式。通过学习 使用规则，我们可以优化我们的行为，最大化社会和身体上的回报。扰乱某人的能力 学习和遵守规则可能是病态的。这种干扰与许多神经精神疾病和 神经退行性疾病，如精神分裂症和痴呆症，这些疾病具有很高的社会和经济价值 成本。为了开发针对这些疾病的新的、机械信息的治疗方法，我们必须首先开发一种 详细了解支持规则的神经机制。 在这里，我们的目标是了解大脑如何灵活地学习、遵循和在几种不同的规则之间切换。 将大规模、多区域电生理学与猴子新的行为模式相结合，我们将研究 基于规则的灵活行为的两个方面： 首先，人们必须能够发现在新的情况下应该遵循哪条规则。这需要集成信息 从一套已知的规则中决定哪一条规则适合这种情况。我们的首要目标是 将利用我们的大规模记录技术来区分关于前额叶相对作用的假说 大脑皮层、顶叶皮质和基底节在整合反馈和决定遵循哪条规则方面发挥作用。 其次，我们的目标是了解如何学习、表示和执行多个规则。具体来说，我们将 测试规则表示是结构化的假设：计算上相似的规则使用相似的神经 机械装置。这种结构的理论化是为了让我们在新的情况下快速学习新的规则。为此， 猴子将学习并执行与计算相关的多个规则。在我们的第二个目标中，我们将使用 慢性和急性电生理学的结合，通过追踪规则的神经表征 学习。这将区分关于规则的神经表示是如何构造的以及它是如何构造的假设 涉及其他类似的规则。同时，我们的第三个目标将使用相同的录音来理解规则 对刺激表示采取行动，将其转化为符合规则的反应。我们将检验三个理论 包括一种新的动态模型，该模型假设规则通过动态转换来起作用 神经活动子空间之间的神经表示。 虽然我们提出的研究在本质上是基础的，但我们相信这是机械论的重要的第一步。 了解包括精神分裂症在内的几种精神疾病的核心认知缺陷。我们相信 这一理解将通过导致新的认知诊断和治疗来改善心理健康 精神错乱。特别是，我们希望利用我们的结果来开发生理标记物，以改进检测， 允许早期干预，并指导有针对性的治疗。

PROJECT SUMMARY/ABSTRACT Over our lifetime, we learn hundreds of ‘rules’ that define how we should act in a given situation. For example, when in a restaurant, we follow a set of rules that guide the way we order, eat, and pay for a meal. By learning and using rules, we can optimize our behavior and maximize social and physical rewards. Disrupting one’s ability to learn and follow rules can be pathological. Such disruptions are associated with many neuropsychiatric and neurodegenerative disorders, such as schizophrenia and dementia, where they carry high social and economic costs. To develop novel, mechanistically-informed, treatments for these diseases, we must first develop a detailed understanding of the neural mechanisms that support rules. Here, we aim to understand how the brain flexibly learns, follows, and switches between several different rules. Combining large-scale, multi-region electrophysiology with novel behavioral paradigms in monkeys, we will study two aspects of flexible rule-based behavior: First, one must be able to discover which rule to follow in a new situation. This requires integrating information from the world to decide which rule, from a set of known rules, is the correct one for the situation. Our first aim will leverage our large-scale recording techniques to distinguish hypotheses about the relative role of prefrontal cortex, parietal cortex, and basal ganglia in integrating feedback and deciding which rule to follow. Second, we aim to understand how multiple rules are learned, represented, and executed. Specifically, we will test hypotheses that the representation of rules is structured: computationally similar rules use similar neural mechanisms. Such structure is theorized to allow us to rapidly learn new rules in new situations. To this end, monkeys will learn and perform multiple, computationally-related, rules. In our second aim, we will use a combination of chronic and acute electrophysiology to track the neural representation of a rule through learning. This will distinguish hypotheses about how the neural representation of a rule is structured, and how it relates to other, similar, rules. In parallel, our third aim will use the same recordings to understand how rules act on stimulus representations to transform them into rule-appropriate responses. We will test three theories of cognitive control, including a novel dynamic model that hypothesizes rules act by dynamically transforming neural representations between subspaces of neural activity. While our proposed research is basic in nature, we believe it is an important first step in a mechanistic understanding of the core cognitive deficits of several mental illnesses, including schizophrenia. We believe this understanding will improve mental health by leading to new diagnostics and treatments for cognitive disorders. In particular, we hope to use our results to develop physiological markers that will improve detection, allow for earlier intervention, and guide targeted treatments.