面孔是日常生活中最重要的一类视觉刺激。以往研究表明，不同于对一般客体的加工方式，人类优秀的面孔识别能力可能源于对面孔的整体加工。在传统的行为研究中，整体-部分效应 (whole-part effect, WPE) 和合成面孔效应 (composite-face, CFE) 是两个广泛使用的面孔整体加工能力的测量范式。但是，目前尚不清楚这两个效应反映了面孔整体加工相同或不同的认知机制。为此，在本论文中，我们借助功能磁共振成像 (fMRI) 和基于大被试量 (N = 200/188) 的个体差异法，从面孔选择性激活和面孔网络连接水平，考察了这两个经典的面孔整体加工任务的神经基础。首先，我们计算体素的面孔选择性，即对面孔和一般客体的反应差异，这一指标反映了体素对面孔偏好反应的程度。接下来，我们考察面孔选择性与WPE/CFE的关系，以定位两个效应在脑区激活水平的神经基础。我们发现，在梭状回面孔区 (fusiform face area, FFA)， WPE和CFE存在半球分离：WPE与左侧FFA的面孔选择性有显著正相关，而CFE与右侧FFA的面孔选择性有显著正相关。进一步的分析发现，WPE与左侧FFA面孔选择性的相关主要来自于左侧FFA对面孔反应的增强，而CFE与右侧FFA面孔选择性的相关主要来自于右侧FFA对一般客体反应的抑制。除此之外，我们在枕叶面孔区(occipital face area, OFA)、后颞上沟(posterior superior temporal sulcus, pSTS)以及全脑其他脑区均发现了两个任务的分离。接下来，我们从网络水平考察了两个任务在核心面孔网络内的静息态功能连接基础。我们采用了体素水平的全局脑功能连接的方法，计算每个体素与核心面孔网络内其它体素的平均连接强度。我们发现，WPE与右侧OFA和核心面孔网络内其他脑区连接的增强相关，特别是与右侧OFA和双侧pSTS之间的连接增强相关。而CFE与右侧FFA和核心面孔网络内其他脑区连接的减弱相关，特别与是右侧FFA和双侧pSTS之间的连接减弱相关。综上，本论文从面孔选择性激活和网络连接水平都发现两个经典的面孔整体加工效应有分离的神经基础，不仅为两个效应行为上的分离提供了神经学证据支持，也提示两个效应可能反应了面孔整体加工的不同认知机制。

Faces are one of the most important visual stimuli in our daily life. It has been long proposed that different from processing of general objects, our extraordinary face recognition ability stems from face specific holistic processing. Two widely-used behavioral hallmarks of holistic face processing are the whole-part effect (WPE) and composite-face effect (CFE). However, it remains unknown whether the two effects reflect similar or different aspects of holistic face processing. To answer these questions, in the present study, we used fMRI and individual differences approach in a large sample of participants (N = 200/188) to examine whether the WPE and CFE involved shared or distinct neural substrates from the perspectives of face-selective responses and functional connectivity (FC) within the core face network (CFN). First, we calculated face selectivity of each voxel as responses differences between faces and objects. Then we examined the relationship between the face selectivity and the WPE/CFE to locate their neural basis respectively. We found that the WPE and CFE showed hemispheric dissociation in the fusiform face area (FFA), that is, the WPE was correlated with face selectivity in the left FFA, while the CFE was correlated with face selectivity in the right FFA. Further, the correlation between the WPE and face selectivity was largely driven by the FFA response to faces, whereas the association between the CFE and face selectivity resulted from suppressed response to objects in the right FFA. In addition, we also observed dissociated correlation patterns of the WPE and CFE in other face-selective regions and across the whole brain. Next, we examined the neural basis of the two tasks at network level. With a voxel-wise global brain connectivity approach based on resting-state fMRI, we characterized global connectivity within the CFN by calculating the averaged FC of each CFN voxel to all the other CFN voxels, referred as within network connectivity (WNC). We found that a cluster in the right occipital face area (OFA) showed significant positive correlation between its WNC and the WPE, while a cluster in the right FFA showed significant negative correlation between its WNC and the CFE. Further, larger WPE was associated with stronger connection between the right OFA and bilateral posterior superior temporal sulcus (pSTS), while higher CFE was driven by weaker connection between the right FFA and bilateral pSTS. In sum, the present study indicated that the two behavioral hallmarks of holistic face processing have dissociated neural basis at both face-selective activation level and functional network level. Our findings provide neural evidence for the behavioral dissociations of the two effects in literature and suggest that these two effects reflect different mechanisms of holistic face processing.