学习是人类成为万物之灵的重要原因。作为“学习的重要器官”——人脑，堪称世界上最复杂的物质系统。但受限于技术与研究方法，人类对学习的研究一直处在思辨或外显行为层面。直到20世纪70年代认知神经科学的兴起，学习研究才上升到了神经生理层面。人类对学习的研究从纯粹的教育、哲学研究转变为跨越神经科学、教育学、心理学等多个学科的研究。时至今日，教育与认知科学的结合已经成为国际上备受关注的新兴研究领域，教育神经科学成为新的学习科学。 学科融合给学习科学特别是学习障碍的研究带来了新的机遇。数学障碍是最受关注的学习问题之一，但与阅读障碍相比，数学障碍这一细分领域的研究发展相对滞后。国内外研究均公认工作记忆是数学障碍的核心缺陷，因此在工作记忆四成分模型提出以后，作为数学障碍发生的最主要原因，工作记忆与数学能力之间的关系应做更深入的探讨。 本研究采取了测验法、实验法、多维Rasch模型、聚类、等值等多种研究方法和技术。首先基于测验法，在比较多种差异模型的基础上，确定了标准分数模型与回归模型相结合的数学障碍儿童鉴别程序和排除因素，并以此为准鉴别出了数学障碍儿童。其次，以工作记忆四成分模型为指导编制了成套测验，数据分析结果表明，测验信度、效度均达到心理测量学要求。之后基于工作记忆测验结果，利用聚类分析法对鉴别出的数学障碍儿童进行了分类，并进一步分析了各类儿童工作记忆和数学能力的特点。最后，在分析了工作记忆与数学能力关系的基础上，对不同类别数学障碍儿童进行了综合干预，并对干预效果进行了评估。 本研究的主要结论如下： 1. 数学障碍的鉴别应将标准分数模型与回归模型相结合，既考虑数学测验成绩的稳定性，又兼顾测验信度，可以避免单一模型的缺陷。 2. 工作记忆四成分成套测验在四至六年级学生群体中施测，其信度、效度均达到心理测量学要求。 3. 数学障碍的流行率为8.1%，通过聚类分析得到了3种类型的数学障碍：视空间信息障碍型、弥漫型、信息存储与整合障碍型。 4. 工作记忆各成分与数学能力、各数学领域能力之间相关显著。工作记忆四成分在各数学领域学习中发挥的作用不同；中央执行系统的作用最大；情境缓冲器缺陷会导致数学成绩的全面落后；视空间模板和语音回路的作用存在此消彼长的关系。 5. 基于Rasch模型的学业诊断方法，提高了数学知识教学干预的针对性。结合数学知识教学和工作记忆训练的综合干预模式对数学障碍学生的两种能力提升均有帮助。 6. 数学障碍儿童对资源教室、干预师资有较高需求，特殊教育资源的配置应当以学生的认知特点、学科能力等为基础，以满足发展需求为中心。

Learning ability is the most important factor for human beings to become the lord of creation. As an important organ of "learning", brain is the most complex system in the world. However, restricted by technology and researching methodology, the learning research of human being has been in speculative or explicit behavior level. Up to the 1970s, along with the rise of cognitive neuroscience, learning research development has merely reached the physical level. The field of learning research has transformed into cross-cognitive science, neuroscience research, pedagogy, psychology and other disciplines. The combination of education and cognitive science has become the focus of attention. Educational neuroscience will evolve into a new discipline about learning. Multi-disciplinary merge has brought new opportunities to researches on learning disabilities. Mathematical disability is one of the most concerned problems in the field of learning disabilities. Compared with reading disabilities, however, the development of mathematical disability is lagging. As one of the causes of mathematical disability, the relationship between working memory and mathematical ability deserves a deeper exploration, especially after the proposal of episodic buffer. Firstly, based on the psychological tests and the comparison among a variety of the discrepancy models, the procedure of indentifying the children with mathematical disability was produced. It was created by combining SDS model and Regression model. In addition, the children with mathematical disability were successfully picked out. Secondly, with the guide of the multi-component model of working memory, a set of tests were made and the according data exploded that the test reliability and validity all meet the standards of psychological measurements. Afterwards, with the help of test outcomes, the children with mathematical disability were categorized through clustering methodology and the features of various children's working memory and mathematical disability were further dug out. Lastly, after analyzing the relationship between working memory and mathematical disability, the intervention training was carried out among children with different types of mathematical disability, and the according effects were also evaluated. This research has drawn the following key conclusions: 1. The Combination of SDS model and regression model should be put into account in the discrimination of mathematics learning disability. Considering the stabilization of test results and reliability, it will avoid the defect of single model. 2. The model of four components was tested among the groups from Grade 4 to 6, and the analyzing result of test data shows that reliability and validity of working memory’s test meet the request of psychometrics. 3. On the choice of cluster model, Ward’s method is the best one. Three subtypes of mathematical disabilities’ have been find, namely, visuospatial sketchpad defect, diffuse defect, information storage and integration defect. 4. The analysis shows that, there is significant relationship between working memory and the pupils’ mathematical ability. Central executive and episodic buffer play an important role in mathematical learning. Visuospatial sketchpad and phonological loop play different roles in the process of mathematical learning. 5. Rasch model improves the pertinence of teaching intervention. The result of data analysis shows that comprehensive intervention have been a great help to both working memory and mathematical ability of mathematical disabilities’ children. 6. Special education resource configuration shoud be based on students' cognitive characteristics, ability of the cognitive, to meet development demand as the center.