Supporting a Racially Diverse Facial Dataset: Normative Valence and Arousal Ratings Across Race and Moderation by Race

The primary goal of this project was to collect normative emotional valence and arousal ratings using the RADIATE facial database. The RADIATE database is one of the few that is racially diverse, yet it is underutilized, due in part to a lack of normative valence and arousal ratings. A secondary goal was to explore whether the race of the rater moderated emotion ratings. As part of an ongoing study, 204 participants (Asian: 9, Black: 25, Latinx: 39, White: 131) were randomly assigned to one of 10 blocks of 36 faces. Each block included faces counterbalanced on race, gender, and emotion so that each participant rated an identical number of faces with respect to these categories. Participants viewed faces in Qualtrics and rated each on valence (from 1-9, unpleasant to pleasant) and arousal (from 1-9, low to high). A 4-way Race of Rater x Race of Face x Emotion x Gender repeated-measures ANOVA with repeated-measures on the last 3 factors was used for valence and arousal ratings. As expected, across racial face categories, happy faces were rated as more pleasant (M = 6.50) and sad faces as more unpleasant (M = 3.03). In addition, happy (M = 4.29) faces were rated more emotionally arousing than sad (M = 3.76) and neutral faces (M = 3.29). The race of the rater moderated valence but not arousal ratings. Black raters rated Asian females as happier than Asian males and Latinx raters rated Latinas as sadder than Latinos, with no other evident effects. Present results contribute to sparse valence and arousal data for the RADIATE dataset. Results further suggest that emotional faces are not rated in a universal manner as some emotion theories presume. Implications of the results and future research directions are discussed.