面孔是我们在日常生活中最经常接触到的刺激之一。我们可以根据一个面孔获得关于该个体的身份信息以及各个社会范畴信息，如性别、种族以及职业等等。传统观点认为，对面孔的身份信息和各类社会范畴信息的加工是相互独立的过程。而近年来，越来越多的研究与模型认为身份信息和各类社会范畴信息之间的加工是相互影响的。前人已有一系列实验证明了对面孔的身份信息加工与面部表情加工的相互影响是不对称的，而面孔身份信息与各个社会范畴信息加工之间是否存在对称或不对称的相互影响关系，目前还没有系统且统一的结论。因此，本文选择了性别、职业与种族三个社会范畴，来探讨身份信息加工过程是否存在社会范畴信息加工的参与，以及这些社会范畴信息的加工过程是否存在身份信息加工的参与，从而根据不对称的相互影响关系建立对面孔各维度信息加工的层级交互模型。
本文通过三个研究，采用合成脸范式(composite paradigm)，将属于不同社会范畴或不同身份的面孔进行拼合，探究面孔身份信息加工与性别信息加工、种族信息加工以及职业信息加工的相互影响关系。
研究一通过两个实验探究性别信息加工与身份信息加工的相互影响。实验均采用合成脸范式。实验1选用不同性别的影视明星作为实验材料制作对齐脸与错开脸刺激，要求被试完成性别判断任务，发现上下半脸身份不相容性别相容与身份相容性别相容条件下性别判断的合成脸效应不存在显著差异，说明性别信息的加工过程中存在身份信息的参与。实验 2 要求被试完成对目标半脸的身份判断任务，发现身份不相容条件性别相容和身份不相容性别不相容条件下的身份判断的合成脸效应存在显著差异，说明身份信息的加工过程中存在性别信息加工的参与。
研究二通过两个实验探究职业信息加工与身份信息加工的相互影响。实验 3 选用影视明星与体育明星作为实验材料制作对齐脸与错开脸刺激，要求被试完成对目标半脸的职业判断任务，发现身份不相容职业相容与身份相容职业相容条件下职业判断的合成脸效应存在显著差异，说明职业信息的加工过程中存在身份信息加工的参与。实验 4 要求被试完成对目标半脸的身份判断任务，发现身份不相容职业相容条件和身份不相容职业不相容条件下身份判断的合成脸效应不存在显著差异，说明身份信息的加工过程中不存在职业信息加工的参与。
研究三通过两个实验探究对于本族或异族的面孔，种族信息加工与身份信息加工的相互影响是否存在差异。实验 5 选用高加索面孔与亚洲人面孔作为实验材料制作对齐脸与错开脸刺激，要求被试完成对目标半脸的种族判断任务，发现亚洲面孔相比于高加索面孔，身份不相容种族相容条件下种族判断的合成脸效应更大，说明不同种族信息的加工过程中，身份信息加工的参与存在差异。实验 6 要求被试完成对目标半脸的身份判断任务，发现目标面孔为亚洲面孔时，身份不相容种族相容条件和身份不相容种族不相容条件下对身份判断的合成脸效应不存在显著差异，而目标面孔为高加索面孔时，二者存在显著差异。说明对于本族面孔，身份信息的加工过程中不存在种族信息的参与；而对于异族面孔，身份信息的加工过程中存在种族信息的参与本文通过上述三个研究，发现面孔身份信息加工与性别信息加工、职业信息加工以及种族信息加工存在不对称的相互影响，并构建四者加工的层级交互模型。在层级交互模型中，上层信息的加工需要下一层信息的参与，而下层信息的加工则不一定需要上一层信息加工的参与。因此，该模型以身份信息为中心，性别信息为下层信息，职业信息为上层信息。而对于种族信息，不同种族的面孔在层级关系上出现了分离：对于本族面孔，种族信息为上层信息；对于异族面孔，种族信息为下层信息。以此建立了面孔身份信息加工与社会范畴信息加工的层级交互模型。

People can get a wealth of information from a person’s face, for example, the person’s identity and social categories, including gender, race, occupation and so on. Based on traditional views, different properties of human faces are processed independently. Recently, however, more and more theoretical models and empirical findings have shown that facial identity analysis and different kinds of social categorizations are interdependent. There are evidences that there is asymmetric influence between perception of face identity and expression. However, it remains elusive that whether the relationships between identity analysis and different kinds of social categorizations are reciprocal or asymmetrical. Therefore, in this research, the relationships between the processing of identity and the processing of  gender, race or occupation were discussed. We investigated whether face identification involves the processing of these three social categories, and vice versa, and proposed a  hierarchical  interactive model of face identification and social categorization  based on our findings.
In this research, using composite paradigm, faces from different social categories or different person were formed to make composite faces. In three studies, we investigated the relationships between face identification and the processing of gender, race and occupation.
In Study 1, we investigated the relationship between the processing of gender and face identification in 2 experiments. Experiment 1 was designed to test whether processing of gender involves face identification. We use faces of male and female movie stars to make aligned and misaligned composite faces. Participants were asked to judge the gender of a certain half of the composite faces. We  found that  there is  no  differences  between the composite effects of the faces with congruent identity, congruent gender and the faces with incongruent identity, congruent gender, which demonstrated  that the processing of gender doesn’t involve face identification. Experiment 2 were designed to test whether face identification involves processing of gender. Participants were asked to judge the identity of a certain half of the composite faces. We found there is a significant difference between the composite effects of faces with congruent identity, congruent gender and faces with congruent identity, incongruent gender, which demonstrated that face identification involves the processing of gender.
In Study 2, we investigated the relationship between the processing of occupation and face identification in 2 experiments. Experiment 3 was designed to test whether processing of occupation involves face identification. We use faces of movie stars and athletes to make aligned and misaligned composite faces. Participants were asked to judge the occupation of a certain half of the composite faces. We  found that  there is a significant difference between the composite effects of the faces with congruent identity, congruent occupation and the faces with incongruent identity, congruent occupation, which demonstrated that the processing of occupation involves face identification. Experiment 4 were designed to test whether face identification involves processing of occupation. Participants were asked to judge the identity of a certain half of the composite faces. We  found that  there is  no differences  between the composite effects of faces with congruent identity, congruent occupation and faces with congruent identity, incongruent occupation, which demonstrated face identification doesn’t involve the processing of occupation.
In Study 3, we investigated the relationship between the processing of race and face identification in 2 experiments. Experiment 5 was designed to test for own-race faces and other-race faces, whether processing of race involves face identification. We use Asian faces and Caucasian faces to make aligned and misaligned composite faces. Participants were asked to judge the race of a certain half of the composite faces. We found that the composite effects of Asian faces with incongruent identity, congruent race  was larger than that of Caucasian faces, which demonstrated that for own-race faces, while judging  race, people tend to pay more attention to  face identification  than other-race faces. Experiment 6 were designed to test whether face identification involves processing of race. Participants were asked to judge the identity of a certain half of the composite faces. We found that for Asian faces,  there is no differences between the composite effect of faces with congruent identity, congruent race and faces with congruent identity, incongruent race, which demonstrated that for own-race faces, face identification doesn’t involve the processing of race; however, for Caucasian faces, the difference is significant, which demonstrated that for other-race faces, face identification involves the processing of race.
In summary, in three studies, we found that there are asymmetric relationships between identity analysis and the processing of gender, occupation and race, and proposed  a hierarchical  interactive model of face identification and social categorization based on our findings.  In this hierarchical model, the processing of upper-level information involves the processing of bottom-level information, but not vice versa. Hence, identity is the center of the model, while gender is the bottom-level  information, and occupation is the upper-level information. There is a separation of own-race faces and other-race faces on the position of race: for own-race faces, race is the upper-lever information; for other-race faces, race is the bottom-level information.