The research on the ensemble coding of emotion in facial crowd originated from the discovery that the visual system can represent the low-level visual features averagely. Haberman and Whitney (2007) used the mean emotion discrimination task for the first time to reveal that the visual system can also make ensemble representation for high-level facial emotions in addition to the low-level visual stimuli. The experimental paradigms, methods of determining emotional units and emotional discrimination thresholds they used lay an important foundation for the future research. Since then, researchers have been trying to reveal the processing mechanism of ensemble coding of emotion in facial crowd by exploring the relationship with individual representation, the separation from the ensemble representation of low-level features and their neural activity characteristics. The results achieved are as follows: Although the ensemble coding has strong cross-domain stability, for different categories of group stimuli, the ensemble coding is domain specific. The process of ensemble coding may include not only automatic recognition of high-level facial stimuli and integration of emotional information, but also the calculation of low-level face features, which is regulated by attention resources. In addition, both ensemble representation and individual representation play a role in the process of extracting mean emotion, but the relationship between them changes dynamically between competition and coexistence. In terms of neural mechanism, Im et al. (2017) used fMRI to investigate whether there have been different brain mechanisms between ensemble coding of emotion in facial crowd and individual facial expression processing. The results showed that dorsal visual pathway is preferentially activated by the processing of crowd emotion. The activities of the intraparietal sulcus and superior frontal gyrus are more sensitive to crowd emotion perception, while the fusiform gyrus in the ventral visual pathway has more advantages in perceiving individual facial expression. However, because the experimental task is not a simple mean emotion discrimination task, which includes the social decision-making process, these results cannot exclude the influence of task goals and participants' intention. Therefore, it is necessary to accumulate a lot of research evidence to reveal the neural mechanism of the ensemble coding of emotion in facial crowd. In conclusion, facial expressions convey important information in social interaction. To understand the emotions of others or groups is an significant prerequisite for stimulating and adjusting social motivations as well as making decisions to pursue advantages and avoid disadvantages. If some faces appear in the form of a group, then the way we process them is different from that of an individual facial expression. When we identify emotion in facial crowd, the cognitive nervous system will automatically integrate the emotion information and extract the mean from it. This process is called ensemble coding. And it is influenced by some cognitive and social factors such as attentional resources, visual processing, social motivation, culture and gender. However, so far, no systematic model about its processing mechanism has been formed. In the future, eye movement technique, neuro-electrophysiology and brain imaging techniques should be used, combined with attention resources, visual working memory, facial local features, emotional valence and social cues to further expand the study of cognitive neural mechanisms and influencing factors. At the same time, people with cognitive and emotional disorders should also be concerned. In addition, the developmental trajectory of the ensemble encoding about facial expressions should be explored from a lifelong perspective, which will not only contribute to the enrichment and development of this basic research field, but also help to provide scientific basis for the diagnosis and treatment of cognitive and emotional disorders.