MRC Cognition and Brain Sciences Unit, University of Cambridge, UK.
Neuroimage. 2019 Oct 1;199:114-126. doi: 10.1016/j.neuroimage.2019.05.059. Epub 2019 May 24.
The neural architecture of semantic knowledge comprises two key structures: (i) A set of widely dispersed regions, located adjacent to the sensorimotor cortices, serve as spokes that represent various modality-specific and context-dependent contents. (ii) The anterior-temporal lobe (ATL) serves as a hub that computes the nonlinear mappings required to transform modality-specific information into pan-modality, multifaceted concepts. Little is understood regarding whether neural dynamics between the hub and spokes might flexibly alter depending on the nature of a concept and how it impinges upon behaviour. Using fMRI, we demonstrate for the first time that the ATL serves as a 'pivot' which dynamically forms flexible long-range networks with cortical modules specialised for different domains (in the present case, the knowledge about actions and places). In two experiments, we manipulated semantic congruity and asked participants to recognise visually presented items. In Experiment 1 (dual-object displays), the ATL increased its functional coupling with the bilateral frontoparietal action-sensitive system when the objects formed a pair that permitted semantically meaningful action. In Experiment 2 (objects embedded in a scene), the ATL augmented its coupling with the retrosplenial cortex of the place-sensitive system when the objects and scene formed a semantically coherent ensemble. Causative connectivity revealed that, while communication between the hub and spokes was bidirectional, the hub's directional impact on spokes dwarfed the strength of the inverse spoke-to-hub connectivity. Furthermore, the size of behavioural congruity effects co-varied with the strength of neural coupling between the ATL hub and action- / place-related spokes, evident both at the within-individual level (the behavioural fluctuation across scanning runs) and between-individual level (the behavioural variation of between participants). Together, these findings have important implications for understanding the machinery that links neural dynamics with semantic cognition.
（i）一组广泛分散的区域，位于感觉运动皮层附近，充当 spokes，代表各种模态特定和上下文相关的内容。（ii）前颞叶（ATL）作为一个枢纽，计算将模态特定信息转换为跨模态、多方面概念所需的非线性映射。关于hub 和 spokes 之间的神经动力学是否可能根据概念的性质及其对行为的影响而灵活改变，人们知之甚少。使用 fMRI，我们首次证明 ATL 作为一个“枢轴”，可以与专门用于不同领域的皮质模块（在本案例中，是关于动作和地点的知识）动态形成灵活的长程网络。在两项实验中，我们操纵语义一致性并要求参与者识别视觉呈现的项目。在实验 1（双对象显示）中，当对象形成允许语义上有意义的动作的对时，ATL 增加了与双侧额顶叶动作敏感系统的功能耦合。在实验 2（嵌入场景中的对象）中，当对象和场景形成语义连贯的整体时，ATL 增加了与位置敏感系统的后扣带皮层的耦合。因果连通性揭示，虽然hub 和 spokes 之间的通信是双向的，但 hub 对 spokes 的定向影响远远超过反向 spoke-to-hub 连通性的强度。此外，行为一致性效应的大小与 ATL 枢纽和动作/位置相关 spokes 之间的神经耦合强度相关，这在个体内水平（扫描过程中的行为波动）和个体间水平（参与者之间的行为变化）都很明显。这些发现对于理解将神经动力学与语义认知联系起来的机制具有重要意义。
