类别学习中规则学习和规则转换的认知和神经机制

Effective and appropriate behavior in the human environment requires the ability to form categories of objects, people, and events, and to link these categories with appropriate responses. Category learning is one of the hottest topics in equcational, cognitive and neural science. There are three major theories or models of how categories are represented: Rule-based, Prototype-based, and Exemplar-based models. Rule-based models assume that category learning is a process of discovering an explicit rule to maximize accuracy. Prototype-based models assume that stimuli are categorized on the basis of their similarity to category prototypes stored in memory. A category prototype is generally defined as the average, or most typical, member of a category. Exemplar-based models assume that the categorization of a new exemplar is based on the similarity of the new exemplar to the representations of all previously encountered exemplars stored in memory. However,our world is rich and colorful. Category members differ along many dimensions and on many specific features. Most researches on category acquisition have focused on category of only two dimensions or two feature values. For example, the famous task of Information-Integration category learning is a two dimensions perceptual learning task. The task of "5-4 category structure" and "Linearly separable categories" adopted two values of feature. Neuroscience researches suggested that categorization is not dependent on any single neural system. Many studies indicated that basal ganglia play a significant role on category learning and even found adversely affect between the head of the caudate and prefrontal cortex during rule learning. These finding indicates that the head, body and tail of the caudate nucleus may play a main role which is easily be neglected because of its C-shape structure and deep location. By using many methods, such as Model Fitting, Eyetracking and fMRI technology, this project will explore the cognitive and neural mechanism of category learning. The cognitive mechanism includes the representation, attention and learning strategy, The neural mechanism includes the rule generation, feedback, and feature perception. This project is expected to provide a new account of "Dimensions Sort Theory (DST)" on cognition and a new hypothesis of "The Region Center of Caudate" on neural mechanism. The study is separated into three steps. First step is to test that DST modal provides a better account than the exemplar-based the prototype-based modal by a new category structure. Second step is to explore the neural mechanism of rule learning, especially the head, body and tail of the caudate. Third step is to explore the neural mechanism of rule shifting, especially the connective function between body of caudate with globus pallidus, putament, and nucleus accumbens.

类别学习是关于人类如何把纷繁复杂的世界知识分门别类地进行存储和学习，一直是教育心理学和认知科学的最热点的问题之一，为此，研究者提出了多种认知模型。然而，已有的研究忽略了学习材料的多维性和多特征性，并在脑机制的研究中未能注意到基底核内部呈现细长结构状的尾状核的头、中和尾部在类别学习中的协同作用。本项目采用新的材料范式，方法上结合行为、数学模型、眼动技术和功能核磁共振技术，内容上涵盖类别表征、注意和学习策略。试图在认知机制上，提出和检验新的模型- - "特征维度排序模型"（DST）；在神经机制上，探索"尾状核"的头、中和尾部在维度搜索和排序中的协同作用，试图建立以尾状核为中心的类别学习神经机制。研究分三部分进行，第一，探索类别学习中的规则学习，提出和检验DST认知模型；第二，探索规则学习的神经机制，研究尾状核的头、中和尾部的协同作用；第三，探索规则转换的神经机制，研究"尾状核中部"的连接作用。

类别学习是关于人类如何把纷繁复杂的世界知识分门别类地进行学习，一直是认知神经科学的最热点的问题之一，本项目通过采用新的材料范式，方法上结合行为、数学模型、眼动技术和功能核磁共振技术，探索了类别学习中的规则形成、特征识别和规则转换。400名大学生被试参加了本项目的系列实验，其中300名参加了核磁共振实验。研究发现：（1）特征独特性对类别学习产生了显著的影响，类别学习可能是一个维度权量（Weighting）和排序的过程，进而提出了特征维度排序模型（DST）这个新观点。DST主要观点是：类别学习后形成的类别表征可能是“维度序列”，而不是前人主张的“原型”（Prototype）或“样例”（Exemplar）。例如，人类对“猫”这个动物的类别表征或许是“头长什么样子，身体长什么样子，尾巴长什么样子”的维度序列表征。根据DST假设，类别学习的关键是形成维度序列，即维度的先后顺序。在类别学习中，被试可以通过尝试错误的学习方法来赋予维度权重，并依此对维度进行排序，形成维度序列表征，最终以规则序列形式保持在长时记忆里。在运用类别知识进行归类活动的时候，将搜索长时记忆里的规则序列，排在前列的维度所在的规则更容易被提取作为外显规则。如果使用该规则仍然不能解决问题，或者得到负反馈，人们会更替其它规则来解决问题。（2）通过探索规则学习中的特征识别、反馈和规则转换的神经网络，发现了参与规则学习不同过程的不同的神经网络系统，其中视觉皮层参与了特征识别，感觉运动网络和默认网络参与了规则的运用，突显网络协同边缘系统参与了规则的转换。（3）通过探索类别学习概括化机制（Generalization），发现了基底核，特别是尾状核对类别学习有重要的作用。