Registered report “Categorical perception of facial expressions of anger and disgust across cultures”

ABSTRACT

Previous research has demonstrated that individuals from Western cultures exhibit categorical perception (CP) in their judgments of emotional faces. However, the extent to which this phenomenon characterises the judgments of facial expressions among East Asians remains relatively unexplored. Building upon recent findings showing that East Asians are more likely than Westerners to see a mixture of emotions in facial expressions of anger and disgust, the present research aimed to investigate whether East Asians also display CP for angry and disgusted faces. To address this question, participants from Canada and China were recruited to discriminate pairs of faces along the anger-disgust continuum. The results revealed the presence of CP in both cultural groups, as participants consistently exhibited higher accuracy and faster response latencies when discriminating between-category pairs of expressions compared to within-category pairs. Moreover, the magnitude of CP did not vary significantly across cultures. These findings provide novel evidence supporting the existence of CP for facial expressions in both East Asian and Western cultures, suggesting that CP is a perceptual phenomenon that transcends cultural boundaries. This research contributes to the growing literature on cross-cultural perceptions of facial expressions by deepening our understanding of how facial expressions are perceived categorically across cultures.

KEYWORDS:

• Emotion
• facial expressions
• categorical perception
• culture

Acknowledgments

The authors would like to express their gratitude to Zhihe Pan and Jinpeng Fu for her valuable contributions in data collection and analysis from China, as well as to Hannah Tran for her valuable contributions in data collection from Canada.

Author contributions

1. X. Fang and D. Sauter developed the study concept and design. X. Fang collected data. X. Fang performed the data analysis and interpretation. X. Fang drafted the manuscript, and G. Kleef, K. Kawakami, D. Sauter provided critical revisions. All authors approved the final version of the manuscript for submission.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 It is noteworthy that conducting a power analysis for NHST with the same criteria (Cohen's d = 0.43 and α = .05) reveals that a sample size of only 45 participants from each cultural group would be sufficient to achieve 80% power. This sample size exceeds the majority of previous studies on categorical perception of facial expressions (e.g. Etcoff & Magee, 1992; Sauter et al., 2011; Young et al., 1997), or is comparable to them (Cong et al., 2019).

2 This pilot study was conducted together with another pilot study (see Fang et al., 2019), the purpose of which was to match the perceived intensity of various emotional expressions on Asian and White faces for use in future studies.

3 The intervals between the presented face pairs in the discrimination task of this study were carefully selected. When utilizing two-step intervals, a substantial difference in the degree of mouth opening between face A and face B was observed. Consequently, participants were able to accurately discern which face corresponded to the previously presented target face solely based on the degree of mouth opening. Conversely, when employing one-step intervals, the disparity in the degree of mouth opening between face A and face B was relatively small. As a result, participants were unable to determine which face matched the previously presented target face based solely on the degree of mouth opening. Therefore, we opted to utilize a one-step interval (10%) for our experiment to enable a more nuanced examination of the participants’ perceptual abilities compared to intervals involving two or more steps.

4 It is important to highlight that the exclusion of participant data in this research was not due to their failure to diligently complete the tasks as required. Rather, it was a consequence of the rigorous data selection criteria employed for the preregistered analysis. Nonetheless, even with this stringent approach, the valid sample size of participants in the present research still exceeds that of previous research on categorical perception of facial expressions. Furthermore, to mitigate data loss, we also conducted (generalized/) linear mixed-effects model analyses that were more tolerant of missing values. This analysis utilized data from 126 Canadian participants and 97 Chinese participants, yielding results consistent with those reported in the main text. For detailed information, please refer to the Supplementary Materials.

5 To better quantify the similarity between angry and disgusted expressions of each actor, we utilized computer vision techniques. By employing opencv-python for image grayscaling and dlib with a pre-trained model (Sagonas et al., 2013) for detecting facial landmarks, we generated 128-dimensional facial embeddings. These embeddings captured the essential features of each face. To measure the similarity between expressions, we calculated the Euclidean distance between their corresponding embeddings. The resulting distances were as follows: 0.37 (AF), 0.44 (AM), 0.45 (WF), and 0.53 (WM). These values indicate that Asian actors exhibited lower Euclidean distances compared to White actors, suggesting a higher degree of similarity within their angry and disgusted expressions.

Additional information

Funding

This work was supported by Social Sciences and Humanities Research Council of Canada [grant number 430-2020-00908].