Abstract
The current study tested whether the perception of angry faces is cross-culturally privileged over that of happy faces, by comparing perception of the offset of emotion in a dynamic flow of expressions. Thirty Chinese and 30 European-American participants saw movies that morphed an anger expression into a happy expression of the same stimulus person, or vice versa. Participants were asked to stop the movie at the point where they ceased seeing the initial emotion. As expected, participants cross-culturally continued to perceive anger longer than happiness. Moreover, anger was perceived longer in in-group than in out-group faces. The effects were driven by female rather than male targets. Results are discussed with reference to the important role of context in emotion perception.
Notes
1The main effect of target emotion was qualified by an interaction with participant's culture, F(1, 58)=5.64, p=.021, η p 2=.09 for attributed happiness and F(1, 58)=7.51, p=.008, η p 2=.12 for attributed anger. When the initial emotion was happiness, Chinese attributed more happiness to the face at which the movie was stopped (M=27.50, SD=20.61 vs. M=13.08, SD=15.03), F(1, 58)=13.69, p<.001, η p 2=.19, and less anger (M=61.48, SD=22.55 vs. M=76.45, SD=21.21), F(1, 58)=10.12, p=.002, η p 2=.15, than Americans. Moreover, the difference in happiness and anger ratings in the happy to anger and the anger to happy conditions respectively were larger for Americans, F(1, 58)=120.50, p<.001, η p 2=.68 for attributed happiness and F(1, 58)=134.01, p<.001, η p 2=.70 for attributed anger, than for Chinese, F(1, 58)=58.07, p<.001, η p 2=.50 for attributed happiness and F(1, 58)=59.26, p<.001, η p 2=.51 for attributed anger. This finding possibly reflects a larger tolerance for ambiguity among Chinese, and a tendency for categorisation in the American group (e.g., Peng & Nisbett, Citation1999). These cultural differences in intensity ratings did not affect the results described in the next section.
2Participants’ gender was initially included as a between-participants factor. Since gender differences were unsystematic and fewer than could have been expected by chance (p<.05), we will not report the participant gender results.