Cortical circuit dynamics underlying multisensory decision making

项目摘要 为了在复杂的3D环境中导航和引导运动，人类和动物必须做出无数的 判断它们的自我运动方向或航向。每一个都是一个多感官的知觉决定， 能够通过结合来自视觉、前庭和听觉的信号来实现更高的准确性和精确度。 动觉与此同时，大脑必须决定何时采取行动（例如，到 快速改变方向以避开障碍物），并预测成功的可能性， 行动上决策的这些特征--选择的准确性、响应时间（RT）和信心--已经被 心理学家研究了世纪，但主要是为了一个单一的模态，而不是更多的 整合多种感官证据来源的自然案例。此外，多感觉的神经基础 整合主要是在单个细胞或大脑区域的水平上进行研究，而几乎所有的感知和 认知功能取决于人口水平的计算和区域之间的通信。解决 这些差距，我们训练猴子执行视觉前庭方向辨别任务， 选择，RT和信心通过后决策赌注（PDW）。在执行任务期间，我们将记录 整体活动同时从两个关键节点的感觉皮层网络代表视觉和 前庭自我运动线索（MST和PIVC，分别），以及一个节点（外侧顶内区，LIP）， 下游决策网络将感官证据转化为运动计划。这些地区 单独与航向知觉有关，但很少有人知道它们的协调活动是如何进行的。 只有通过人口记录才能观察到的模式支持多感官决策。目标1， 根据这项提议，我们将量化感觉神经群体之间的协调活动，并测试 来自多感觉整合的感知改善取决于它们之间的耦合强度， 超出了独立考虑其活动所能解释的范围。当视觉和前庭线索 人为地陷入冲突，我们会问航向估计的相对精度是否以及如何 从这些感官群体中解码出来的信息预测了选择和信心， 多感官证据积累模型在目标2中，我们将扩展对区域间 互动的决策阶段，量化的强度和时间之间的功能耦合LIP和 两个感官区域的每一个。这种协调活动的相对定时可以指示前馈与 反馈过程，揭示了感知决策如何通过感觉和 对一个提议（或对一个行动计划的承诺）的信念程度。结果将产生新的见解， 感知证据的呈现和读出及其相关的（不）确定性程度，并将推进一个 在多感官任务中对决策计算的群体和电路级理解。

Project Summary To navigate and guide locomotion in a complex 3D environment, humans and animals must make countless judgments of their direction of self-motion, or heading. Each of these is a multisensory perceptual decision, able to achieve greater accuracy and precision by combining signals from the visual, vestibular, and kinesthetic senses. At the same time, the brain must decide when to commit to a course of action (e.g., to quickly change direction to avoid an obstacle), and make predictions of the likelihood of success in that action. These features of a decision—choice accuracy, response time (RT), and confidence—have been studied by psychologists for over a century, but primarily for only a single modality, instead of the more natural case of integrating multiple sources of sensory evidence. Moreover, the neural basis of multisensory integration is largely studied at the level of individual cells or brain regions, whereas nearly all perceptual and cognitive functions depend on population-level computations and communication between areas. To address these gaps, we trained monkeys to perform a visual-vestibular heading discrimination task which measures choice, RT, and confidence via a post-decision wager (PDW). During performance of the task, we will record ensemble activity simultaneously from two key nodes in the sensory cortical network representing visual and vestibular self-motion cues (MST and PIVC, respectively), as well as one node (lateral intraparietal area, LIP) in the downstream decision network that converts sensory evidence into a motor plan. These regions have individually been linked to heading perception, but little is known about how their coordinated activity patterns, observable only though population recordings, support multisensory decision making. In Aim 1 of the proposal, we will quantify the coordinated activity across sensory neural populations and test whether the perceptual improvement from multisensory integration depend on the strength of coupling between them, beyond what can be explained by their activity considered independently. When visual and vestibular cues are artificially placed in conflict, we will ask whether and how the relative precision of heading estimates decoded from these sensory populations predicts choice and confidence, guided by predictions of a multisensory evidence accumulation model. In Aim 2, we will extend our investigation of inter-areal interactions to the decision stage, quantifying the strength and timing of functional coupling between LIP and each of the two sensory areas. The relative timing of this coordinated activity can indicate feedforward versus feedback processes, revealing how perceptual decisions evolve via recurrent loops between sensation and degree of belief in a proposition (or commitment to a plan of action). The results will yield new insights into the representation and readout of sensory evidence and its associated degree of (un)certainty, and will advance a population- and circuit-level understanding of decision computations in a multisensory task.