PriorCircuit:Circuit mechanisms for computing and exploiting statistical structures in sensory decision making

动物智力的一个决定性特征是能够发现、表示和更新感知环境中统计规律的知识，为适应行为服务。提取和利用统计规律通常是隐含的--没有指导，而且是自动的--是人类婴儿语言习得等复杂行为的基础。大脑是如何做到这一点的？我们的目标是通过研究听觉辨别任务中的决策过程来理解内隐统计学习背后的神经计算，其中统计数据是精细参数化的，并且随着时间的推移而变化。我们最近的工作表明，人类、大鼠和小鼠可以跟踪感官统计数据，从而相应地对它们的行为进行偏向。我们的电生理询问进一步表明，后顶叶和前扣带回皮质代表了从学习到的统计数据中产生的这些偏向。这些电路节点为研究统计结构的内隐学习提供了一个明确的切入点。基于这些发现，我们的目标是实现对不同大脑区域如何随时间整合感觉事件的机械性理解，以便构建关于环境的先前信念，从而反过来影响感觉感知和/或行动计划。我们将通过将人类、大鼠和小鼠的新颖和高度可量化的行为范例与大规模电生理记录和啮齿类动物的功能丧失方法相结合来实现这一点。这些实验将提供一套理论模型，用于解释观察到的现象，并提出进一步的操作，允许我们排除支持内隐统计学习的计算的不同假设。这个项目的结果将促进对大脑如何产生和编码关于统计规则的信念的理解，并计算出适当的决定来优化奖励。

A defining feature of animal intelligence is the ability to find, represent and update knowledge of statistical regularities in the sensory environment, in service of adaptive behavior. Extracting and exploiting statistical regularities often occurs implicitly-without instruction, and automatically-and is the bedrock for complex behaviors such as language acquisition in human babies. How does the brain achieve this? Our goal is to understand the neural computations behind implicit statistical learning by investigating decision making in auditory discrimination tasks where the statistics are finely parameterized and change over time. Our recent work has shown that humans, rats and mice can track sensory statistics to bias their behaviour accordingly. Our electrophysiological interrogation has further implicated the posterior parietal and anterior cingulate cortices in representing these biases that arise from the learnt statistics. These circuit nodes provide a clear entry point for investigating the implicit learning of statistical structures.Building on these findings, we aim to achieve a mechanistic understanding of how different brain areas integrate sensory events across time in order to construct 'prior' beliefs about the environment that can in turn influence sensory perception and/or action planning.We will achieve this by combining novel and highly quantifiable behavioural paradigms in humans, rats and mice with large-scale electrophysiological recordings and loss-of-function approaches in rodents. These experiments will inform a suite of theoretical models that will serve to explain the observed phenomena, and propose further manipulations that would allow us to rule out different hypotheses for the computations that support implicit statistical learning. The results from this project will advance understanding of how the brain generates and encodes beliefs about statistical regularities and computes appropriate decisions to optimize reward.