The Cognitive Neuroscience of Human Category Learning

描述(申请人提供)：范畴化是人类最重要的认知技能之一。它让我们能够在危险的世界中航行，找到食物、住所和朋友。现在有压倒性的证据表明，人类有多个类别学习系统，这些系统在神经解剖学上基本上是分开的，通过质的不同规则学习，并适应了学习不同类型的类别结构。自然要研究的下一个问题是这些不同的系统是如何相互作用的。这是一个重要的问题，因为在日常生活中，我们必须经常在不同的分类系统之间切换(例如，显性的和程序性的)。例如，驾驶的许多组成部分都是程序性的，但同时也需要一些明确的决定。在一个明确的决定之后，如果不能迅速切换回程序策略，可能会极大地增加事故的风险。这项拟议的研究研究了外显和程序性类别学习是如何协调的，以及控制是如何在这些系统之间转移的。我们采取综合、跨学科、融合的操作方法，将行为学、神经心理学(帕金森氏病患者)、功能磁共振成像和经颅磁刺激研究结合起来，目标是建立和测试调节分类和系统切换行为的大脑回路的生物学详细计算模型。这项提议是为了继续一项计划，该计划(A)提供了许多现有证据，表明人类有多个类别学习系统，(B)制定了调节每个系统的神经网络，以及(C)发现了这些系统的许多独特特性。在前一阶段(2R01 MH3760)，我们在几个领域取得了重大进展。其一是理解不同系统中的学习是如何协调的。为此，我们报告了逐个试验在显性和程序性分类策略之间切换的证据 非常困难。这项拟议的研究继续我们对系统相互作用的研究，有三个目标。目标1是确定系统转换的认知成分。目标2是确定系统切换的神经基础，目标3是开发和测试系统切换的生物学详细计算模型。我们开发的模型应该能够提供来自目标1和目标2的所有数据的准确账户，以及来自各种已发表的单位记录研究的数据。此外，该模型将对药物、基因和局灶性病变如何影响行为做出具体预测，并对可能改善类别学习中的系统切换的行为和药物干预做出新的预测。

DESCRIPTION (provided by applicant): Categorization is among the most important cognitive skills that humans possess. It allows us to navigate in a dangerous world, and to find food, shelter, and friends. The evidence is now overwhelming that humans have multiple category-learning systems, which are largely neuroanatomically separate, learn by qualitatively different rules, and have adapted to learning different types of category structures. A natural next question to investigate is how these various systems interact. This is an important problem because during daily life we must often switch between different categorization systems (e.g., explicit and procedural). For example, many components of driving are procedural, but at the same time some explicit decisions are required. Following an explicit decision, a failure to quickly switch back to a procedural strategy could greatly increase the risk of an accident. The proposed research studies how explicit and procedural category learning are coordinated and how control is transferred between these systems. We take an integrative, cross-disciplinary, converging operations approach that combines behavioral, neuropsychological (with Parkinson's disease patients), functional magnetic resonance imaging, and transcranial magnetic stimulation studies with the goal of building and testing biologically detailed computational models of the brain circuits that mediate categorization and system-switching behavior. This proposal is to continue a program that (a) provided much of the existing evidence that humans have multiple category-learning systems, (b) mapped out the neural networks that mediate each system, and (c) discovered many unique properties of these systems. During the previous period (2R01 MH3760), we made significant progress in several areas. One was to understand how learning in the various systems is coordinated. Toward this end, we reported evidence that trial-by-trial switching between explicit and procedural categorization strategies is extremely difficult. The proposed research, which continues our investigations of system interactions, has three aims. Aim 1 is to identify the cognitive components of system switching. Aim 2 is to identify the neural basis of system switching, and Aim 3 is to develop and test a biologically detailed computational model of system switching. The model we develop should be able to provide accurate accounts of all data from Aims 1 and 2, as well as data from various published single-unit recording studies. In addition, the model will make specific predictions about how drugs, genes, and focal lesions should affect behavior and it will make novel predictions about behavioral and pharmacological interventions that might improve system switching in category learning.