人脑是世界上最为复杂的智能系统，从发展的角度来看，个体出生后的儿童期是脑智发育最关键的阶段之一：大脑神经系统和认知行为经历着长期而复杂的协同发展过程，这一阶段的脑智发育直接或间接影响着后续阶段的发展。因此，对儿童脑智发育机理的研究已成为国际脑科学与心理学领域的重大前沿课题。发展认知神经科学（Developmental Cognitive Neuroscience）领域关注的核心科学问题之一，是理解大脑如何随年龄的增长而产生高度特异化且交互协同的功能模块，以便支持各种复杂而高级的认知与情感功能。而这些神经系统之间是如何相互协作来促进认知与情感功能的成熟，到目前为止仍然没有定论。早期的成熟理论认为，随着大脑特定皮层区域解剖结构的成熟，每个脑功能模块都会“各司其职”地支持特定认知功能的发展。而技能学习假说理论则认为，在新的认知功能发生阶段，儿童与成人在技能获得时关联的皮层相同，而脑区激活的权重和模式则不同，并且儿童行为的神经变化将反映出成年人获得更复杂技能后所观察到的变化。后来，以Mark Johnson为代表的研究者们提出的交互式特异化理论则认为，存在一种特殊的通用发育架构性质的脑功能模块，可以共激活于多神经生物模式之下，支持着不同认知能力的发展。最近，John Duncan等提出了多元需求系统模型，该模型认为额顶网络系统通过不同的神经激活模式支持着多重认知功能的执行，是认知灵活切换的基础，可能在儿童跨认知域发展中起到了组织整合不同层级的神经计算资源的作用。基于交互式特异化和多元需求系统模型的理论观点，我们提出了本论文的关键科学问题：在儿童的脑智发育过程中，多元需求额顶系统是否存在着一种通用性的神经发育机制，通过层级化的分布式神经表征组织方式，支持着儿童跨认知域的发展。

本论文通过融合发展心理学、无创功能性磁共振成像（fMRI）、跨认知域的多任务范式和创新性的层级化神经表征建模等跨学科技术方法，以注意网络、工作记忆和情绪感知等任务为例，来研究了适合于跨认知域的通用性神经表征架构及其发育规律，主要包括以下四个方面：1）儿童认知子域发展的特异化神经发育基础；2）儿童跨认知子域发展的分布式神经表征基础；3）儿童跨认知域发展的多元需求额顶系统基础；4）儿童跨认知域发展的多层级分布式神经表征基础。首先，从儿童最基本的核心认知能力——注意的不同认知子域出发，研究一通过使用注意网络范式，探索了警觉、定向和执行注意三种认知子域的特异化发育模式；在行为水平上，我们发现：警觉注意最早从7岁起逐渐达到类似成人的成熟水平，定向和执行注意则表现为贯穿儿童中晚期的持续发育模式。相对于18-22岁的成人被试，儿童表现出了额顶背侧注意系统较弱分离，表明了儿童认知子域各自独特的神经特异化进程尚未发育成熟。因此，研究一从认知子域的水平印证了成熟理论的神经特异化观点。然而，根据交互式特异化理论的观点，发育中不同神经系统的特异化，可能还存在相互协同作用以促进认知的发展。为此，在研究一的基础上，研究二进一步探索了儿童跨认知子域发展的分布式神经表征基础，考察了7-12岁儿童与成人之间不同注意认知子域相关的多变量（分布式）神经表征在共激活额顶背侧注意网络系统中的发育特征。我们发现：儿童在额顶背侧注意系统中表现出了更低的跨认知子域神经表征稳定性和更弱的分布式神经概化水平，表明了多元需求额顶系统通过分布式神经表征的组织方式来整合支持着注意加工中不同认知子域的发展。据此，我们推测该系统可能是儿童认知发展的通用性神经表征基础。研究一和二在多认知子域的层级水平上，一定程度上印证了儿童脑智发育的交互式特异化理论，即额顶背侧注意系统的多神经生物模式整合对跨认知子域发展的支持。基于以上发现以及多元需求系统在多认知任务的额顶共激活规律，我们进一步提出（研究三）：多元需求额顶系统中层级化的神经表征组织方式可能同样支持着儿童其他认知域的发展。为此，我们整合了多个经典的认知任务范式（如注意网络测试、N-Back工作记忆和情绪匹配任务），考察了跨认知域发展的多任务神经共激活模式，并构建7-12岁儿童年龄组不同认知域的脑功能激活图谱工具包，探索了多元需求额顶系统在儿童跨认知域发展中多任务状态下的脑功能共激活模式。基于研究二的推论和研究三的发现，我们提出：需要从不同层级的信息加工机制出发，依靠更为底层的基于神经元群体的共激活信息编码规则，考察多元需求额顶系统在儿童跨认知域发展中的分布式神经表征机制。为了回答该问题，研究四利用跨认知域的多层级分布式神经表征建模方法，系统考察了儿童跨认知域发展的神经表征模式的发育规律，我们发现了儿童在多元需求额顶系统中表现为更弱的跨认知域神经概化水平，即更低的层级化分布式神经表征。综上，我们推测在儿童脑智发育过程中，多元需求额顶系统可能作为一个潜在的通用性架构，通过组构性的信息编码（Compositional Coding）方式，来进行不同层级认知功能对应神经表征信息的分离和整合，从而支持着儿童跨认知域的发展。

本研究从儿童跨认知域发展的视角，结合最新发展认知神经科学和层级化分布式神经表征建模手段，在认知行为层面上，阐明了7-12岁儿童在注意加工中认知子域的异质性发育规律；在脑激活层面上，揭示了7-12岁儿童不同认知子域的神经特异化发育特征，验证了儿童脑认知中交互式特异化理论的构想；在分布式神经表征层面上，进一步揭示了多元需求额顶系统在儿童跨认知域发展中的通用构架特性及其层级化的分布式神经表征组织方式，扩展了当前儿童脑智发育理论前沿。我们研究发现为理解儿童认知与情绪发展的规律提供了新的启示，推动了儿童神经发育基础的实践创新，有助于启发基于认知发展的人工智能算法。为研发脑智异常发育（如自闭症、ADHD等）的客观生物标记和为我国儿童青少年基础教育与心理健康评估提供了影像学方面的实证性证据支持。

Human brain is the most complex intelligent system in the world. From the perspective of development, childhood is one of the most critical stage during brain development: our neural system and cognitive behavior undergo a long-term and complex process of coordinated development. The neurodevelopment of children in this period will directly or indirectly affects their subsequent stages. Therefore, research on the mechanism of children's brain development has become a major frontier topic in the field of international brain science and psychology. One of the core scientific issues of developmental cognitive neuroscience is to understand how our brain develop highly specialized yet interacting neural modules with growth in order to support a wide spectrum of cognitive and affective functions. It is still inconclusive how these neural systems interplay and work together to promote the maturity of cognitive and emotional functions. The early maturational perspective believed that as the anatomical structure of a specific cortical area matures, each brain function module will "perform their duties" to support the development of corresponding cognitive functions. The skill-learning hypothesis held view that children and adults have same linking cortex when acquiring skills at the stage when new cognitive functions generate, but the weights and patterns of brain activation are different, and the neurological changes in children’s behavior will reflect the changes observed in adults after acquiring more complex skills. Later, the interactive specialization theory proposed by Mark Johnson argued that there is a special brain function module with the properties of a general developmental architecture to support the development of different cognitive abilities, which can co-activate in multiple neurobiological models. Recently, John Duncan et al. proposed a multi-demand system model. The advancing model believes the frontal-parietal network system supports execution of multiple cognitive functions through different neural activation modes, which is the basis for flexible cognitive switching, and may play a role in coordinating and integrating different levels of neural computing resources in the development of children’s brain across cognitive domains. Based on the interactive specialization and multi-demand system model, we put forward the scientific questions of this paper: whether the multi-demand frontal-parietal system have a general neural representation pattern under different cognitive subdomain tasks, and how this pattern supports the development of children’s multiple cognitive domains through a hierarchical distributed neural representation organization.

Integrating traditional developmental psychology, non-invasive functional magnetic resonance imaging in emerging cognitive neuroscience, multiple task paradigm across cognitive domains, innovative hierarchical distributed neural representation modeling and interdisciplinary technologies, we aim to explore a general neural representation framework and its developmental rules suitable for multiple cognitive domains, mainly including the following four aspects: 1) The neural specialization basis for children’s multiple cognitive subdomains; 2) The distributed neural representation basis for children’s multiple cognitive subdomains’ development; 3) The basis of multiple demand frontal-parietal system in children's multiple cognitive domain development; 4) The hierarchical distributed neural representation pattern for children’s multiple cognitive domain development. First of all, starting from cognitive subdomains of the most basic core cognitive function (i.e., attention), we examined the functional specificization basis of neural system in regard to children's attention subdomain development. The first study explored the heterogeneous developmental patterns of three core attention abilities by using the attention network test paradigm combined with fMRI technology. At behavioral level, we found that alerting reached adult-like level maturity as early as 7 years old, orienting and executive attention continued to develop until the middle and late stages of childhood. Compared with adults, children exhibited immature neural specialization with weaker dissociation in frontal-parietal dorsal attention system related to attentional cognitive subdomains (that is, three attention processing components), which emphasized the unique neurobehavioral development patterns of children's cognitive subdomains. Based on this finding, the second study further examined the developmental characteristics of multivariate (distributed) neural representations related to different attentional cognitive subdomains between children aged 7-12 and adults. We found that children exhibited lower neural representation stability and weaker multi-voxel activation pattern of neural generalization level cross cognitive sub-domains in the frontal-parietal dorsal attention system. The results show that the neural generalization process in multiple demand fronto-parietal system based on the fine-tuned multivariate activation pattern is the developmental basis of multiple attention cognitive subdomains, and we speculate that it may serve as the general basis of generalizable representation to support children's cognitive development. Subsequently, by integrating multiple classic cognitive task paradigms (attention network test, numerical N-Back working memory and emotion matching tasks) and the latest functional magnetic resonance imaging technology and analysis methods, we explored multiple cognitive domains of neural activation pattern, and examined the multiple demand frontal-patietal system basis in the process of children’s cognitive development. Meanwhile, we also construct a set of high-quality brain function activation maps specific to cognitive domains and age. We believe that it is necessary to start from different levels of information processing mechanisms and rely on population-based co-activation information coding rules to examine the distributed neural representation mechanism of the multiple demand frontal-parietal system in the children cognitive development. Finally, based on the inferences of study 2 and the findings of study 3, study 4 used a hierarchical distributed neural representation modeling method across multiple cognitive domains to systematically investigate the developmental patterns of neural information representation in children and adults. We found that children exhibited weaker neural generalization across cognitive domains in multiple demand frontal-parietal system (network), with lower hierarchical distributed neural representation.