自心理学诞生以来，研究者便一直致力于探索人类认知能力的结构。认知能力通常被定义为人脑在问题解决过程中完成心理信息加工的能力。认知能力在个体和社会发展中起到至关重要的作用，然而其结构与神经机制一直尚不清晰。虽然基于传统的智力测验和因子分析，Cattell-Horn-Carroll（CHC）理论所提出的层级结构得到广泛认可。然而，随着认知神经科学的不断发展，基于认知范式的认知能力测量得到了越来越多的关注。尤其是在国际大规模队列研究中，研究者开始关注基于认知范式的一般认知能力及其神经基础。目前基于传统智力测验和认知范式的研究发现都存在一定的局限和问题。首先，传统智力测验的任务往往基于实践经验，其认知过程相对复杂，从而难以揭示一般认知能力的认知和神经机制；其次，虽然基于传统智力测验而提出的CHC理论得到了较高认可度，但认知能力的因子结构仍存在争议；第三，虽然基于认知神经科学的任务范式及其认知和神经机制更加清晰，但其主要任务是为了分离特定认知成份，而不是为了考察个体差异，因此其测量一般认知能力的信效度及其实践可行性仍需要证明；最后，在神经基础上，当前研究更多采用基于静息态的脑功能网络作为神经指标，但其信效度近年来受到了挑战。为回答上述问题，本研究拟采用认知神经科学和大样本个体差异的研究范式，通过三个系统研究考察一般认知能力的结构，并揭示其神经基础。

研究一旨在考察通过认知范式测量一般认知能力的可行性以及方法学的优化。本研究共招募了1605名大学生被试完成19个来自于注意，记忆，执行功能，反应时等领域的认知任务，其中682人同时完成了磁共振静息态和任务态（N-back）的数据采集。结果发现，19个任务所产生的20个行为指标具有较高的内部一致性信度（除一个任务指标为0.53外，其余均大于0.6），同时任务之间几乎全部（97%）表现出正相关，表明用认知范式作为一般认知能力测评的可行性。对任务进行随机重采样发现，随着任务指标数目的增加，一般认知能力的分半信度，与瑞文任务的相关，以及和脑影像的结果相关程度都显著提升。模型拟合发现，平均需要25个任务指标才能达到一般认知能力测量信度为0.8的水平，高于大部分传统智力测验中所含测验数目。进一步分析发现，相对于静息态，工作记忆任务状态下的大脑功能连接模式能够更好地预测个体的一般认知能力。这些结果为基于认知范式的一般认知能力测量提供了方法学的基础。

在研究一的基础上，研究二旨在发展一套基于认知范式的一般认知能力测评系统，并进一步揭示其认知成分和神经基础。本研究发展了77个不同的任务，覆盖了感知觉、加工速度、注意、学习与记忆、思维与推理、执行功能、语言与数学等领域。对超过200人的重测结果表明这些任务具有较高的重测信度。本研究进一步收集1148名大学生被试在所有任务的行为数据，其中173名被试同时完成了结构磁共振像和工作记忆，情景记忆、自然刺激（电影）和静息态等条件下的功能磁共振成像。结果发现，77个任务指标中除其中6个外均表现出较高的重测信度（> 0.5）。同时，几乎全部任务之间（93%）都表现出正相关。与研究一的预测一致，结果发现24个任务对一般认知能力测量的信度平均可以达到0.8以上。进一步分析发现，对一般认知能力贡献最高的任务包含N-back任务、复杂广度任务和图形推理任务，提示一般认识能力包含了工作记忆与流体智力成分。采用对一般认知能力上预测度最高的任务，我们发现，8个任务可以解释一般认知能力80%的变异，而18个任务可以解释90%的变异。脑成像结果支持了脑体积与脑功能连接模式对一般认知能力的预测作用，但不支持脑网络全局效率的预测作用；而不同磁共振条件下的功能连接模式对一般认知能力预测性揭示了“任务态>自然刺激>静息态”这一规律。

研究三采用研究二的数据，进一步考察认知能力的结构并探索各个成份的神经基础。研究对比了探索性因子分析、CHC理论和认知心理学理论这三种模型对数据的拟合优度，发现探索性因子分析给出了最佳的拟合结果。进一步对比探索性因子分析结果与CHC理论的结果后发现，探索性因子分析的结构里比CHC理论新增了一般认为属于执行功能成分的“抑制”和“切换”因子，而其余因子（加工速度-注意、情景记忆、空间短时记忆、推理-工作记忆和言语短时记忆）则均在CHC理论中有所对应。这一结果进一步发展了认知能力的结构模型。基于脑功能连接模式，研究发现，除了“抑制”和“工作记忆-推理”两个因子外，其余所有因子维度都能被至少一种模态下的脑功能连接模式预测。同特殊因子相比，一般认知能力的贡献边分布更为广泛，提示其需要多个脑网络的共同作用。另外，预测一般认知能力的边更多富集在额顶网络，这和额顶整合理论预测的基本一致。同时，各个特殊因子的核心贡献节点也和文献吻合。这些结果进一步发展了认知能力的结构模型，且为各因子维度提供了神经机制证据。

综上所述，本研究发展了基于认知范式的一般认知能力测验工具，并深入揭示了认知能力的结构及其神经基础。这些结果拓展了传统基于CHC的理论模型，并为发展新的一般认知能力测评工具提供了基础。

Since the inception of psychology, researchers have been dedicated to exploring the structure of human cognitive abilities. Cognitive abilities are typically defined as the brain's capacity to process mental information during problem-solving. These abilities play a crucial role in individual and societal development, yet their structure and neural mechanisms remain unclear. While the hierarchical structure proposed by the Cattell-Horn-Carroll (CHC) theory based on traditional intelligence tests and factor analysis has gained widespread acceptance, there is growing attention towards cognitive ability measurement based on cognitive paradigms in the field of cognitive neuroscience. Particularly in large-scale international studies, researchers have begun focusing on general cognitive abilities based on cognitive paradigms and their neural foundations. However, research findings based on traditional intelligence tests and cognitive paradigms both have limitations and issues. First, tasks in traditional intelligence tests are often based on practical experience, leading to relatively complex cognitive processes that make it challenging to reveal the cognitive and neural mechanisms of general cognitive abilities. Second, although the CHC theory proposed based on traditional intelligence tests has gained high acceptance, the factor structure of cognitive abilities remains controversial. Third, although cognitive paradigms based on cognitive neuroscience provide clearer cognitive and neural mechanisms, their primary goal is to separate specific cognitive components rather than examine individual differences, thus requiring further evidence for the reliability and feasibility of measuring general cognitive abilities. Lastly, current research on neural foundations predominantly uses resting-state brain functional networks as neural indicators, but their reliability has been challenged in recent years. To address these questions, this study aims to investigate the structure of general cognitive abilities and reveal their neural foundations through a cognitive neuroscience approach and large-sample individual difference studies across three systems.

Study One aims to examine the feasibility of measuring general cognitive abilities through cognitive paradigms and methodological optimization. This study recruited 1605 college students to complete 19 cognitive tasks from domains such as attention, memory, executive function, and reaction time, with 682 participants simultaneously undergoing data collection through resting-state and task-based (N-back) fMRI. The results indicated high internal consistency reliability (except one task index at 0.53) across the 20 behavioral indicators generated by the 19 tasks, with almost all tasks (97%) showing positive correlations, demonstrating the feasibility of using cognitive paradigms for general cognitive ability assessment. Random task resampling revealed significant improvements in split-half reliability, correlation with Raven's tasks, and correlation with brain imaging results as the number of task indicators increased. Model fitting revealed that an average of 25 task indicators is needed to achieve a reliability level of 0.8 for measuring general cognitive abilities, surpassing the number of tests contained in most traditional intelligence tests. Further analysis found that brain functional connectivity patterns during working memory tasks better predict individual general cognitive abilities compared to resting-state. These results lay the methodological foundation for measuring general cognitive abilities based on cognitive paradigms.

Building upon Study One, Study Two aims to develop a cognitive paradigm-based general cognitive ability assessment system and further elucidate its cognitive components and neural foundations. This study developed 77 different tasks covering domains such as perception, processing speed, attention, learning and memory, reasoning, executive function, language, and mathematics. Retesting results from over 200 individuals indicated high retest reliability for these tasks. Additionally, this study collected behavioral data from 1148 college student participants across all tasks, with 173 participants simultaneously undergoing structural MRI and functional MRI imaging under conditions including working memory, episodic memory, natural stimuli (movies), and resting state. Results showed high retest reliability (> 0.5) for all but 6 of the 77 task indicators, with almost all tasks (93%) exhibiting positive correlations. Consistent with Study One predictions, results indicated that an average of 24 tasks achieves a reliability level of 0.8 or higher for general cognitive ability measurement. Further analysis revealed that tasks contributing most to general cognitive ability include N-back tasks, complex breadth tasks, and graphic reasoning tasks, suggesting that general cognitive ability encompasses working memory and fluid intelligence components. Using tasks with the highest predictive validity for general cognitive ability, 8 tasks explained 80% of variability, while 18 tasks explained 90%. Brain imaging results supported the predictive role of brain volume and brain functional connectivity patterns in general cognitive ability but not global brain network efficiency. Moreover, functional connectivity patterns under different MRI conditions revealed a pattern of "task state > natural stimuli > resting state" in predicting general cognitive abilities.

Study Three utilized data from Study Two to further investigate the structure of cognitive abilities and explore the neural foundations of each component. The study compared the goodness of fit of exploratory factor analysis, CHC theory, and cognitive psychology theory models on the data, finding that exploratory factor analysis yielded the best fit. Further comparison of exploratory factor analysis results with CHC theory revealed that the exploratory factor analysis structure introduced "inhibition" and "switching" factors, commonly considered components of executive function, not present in the CHC theory, while other factors (processing speed-attention, episodic memory, spatial short-term memory, reasoning-working memory, and verbal short-term memory) corresponded to the CHC theory. This result further developed the structure model of cognitive abilities. Based on brain functional connectivity patterns, the study found that all factor dimensions except "inhibition" and "working memory-reasoning" could be predicted by at least one mode of brain functional connectivity patterns. Compared to specific factors, contributions to general cognitive abilities are more widely distributed across brain networks, indicating the need for the combined action of multiple brain networks. Additionally, edges predicting general cognitive abilities are more enriched in the frontoparietal network, consistent with predictions of the frontoparietal integration theory. Moreover, core contributing nodes of each specific factor align with literature findings. These results further develop the structure model of cognitive abilities and provide neural mechanism evidence for each factor dimension.

In summary, this study developed a cognitive paradigm-based general cognitive ability assessment tool and thoroughly elucidated the structure of cognitive abilities and their neural foundations. These findings expand upon traditional CHC-based theoretical models and lay the groundwork for developing new general cognitive ability assessment tools.