Our everyday experiences involve perceiving crowds of faces, whether it's in a public transport or speaking in front of people. Your brain tries to summarize this multidimensional environment into an average representation in a socially relevant information that compose the world surrounding us. This project seeks to better understand the mechanisms of facial expression averaging, by combining methods from cognitive psychology and computational modeling techniques. We created stimuli combining four distinct facial expressions (e.g anger/disgust; fearful/happy). After an online rating task, the stimuli were used to create sequences of faces varying in identity and intensity of emotional expression, which were presented in a temporal averaging task. Our results showed that people were significantly faster at detecting fearful faces than detecting happy faces, replicating the reaction patterns obtained in previous experiments. Complementary researches will need to be done in order to see if the asymmetry is consistent throughout other emotional pairings