Valence in the Reading the Mind in the Eyes Task

The Reading the Mind in the Eyes task (RMET; Baron-Cohen, Wheelwright. Hill. Raste. & Plumb, 2001) is commonly used to assess theory of mind abilities in adults. In the task, participants pair one of four mental state descriptors with a picture of the eye region of a face. The items have varying emotional valence, and nearly 100 studies have examined whether performance on this task varies with item valence. However, efforts to address this question have been hampered by cross-study inconsistencies in how item valence is assessed. Thus, the goal of this study was to establish reference ratings for the valence of RMET items. In Study 1, we recorded valence ratings for each RMET item with a large sample of raters (n = 164). We illustrated how valence categories are essentially arbitrary and largely influenced by sample size. In addition, valence ratings were continuously distributed, further questioning the validity of imposing categorical distinctions. In Study 2, we used an archival dataset to demonstrate how the different categorization schemes resulted in conflicting conclusions about the association between item valence and RMET performance. However, when we examined the association between item valence and performance in a continuous manner, a clear U-shaped pattern emerged: Items that had more extreme valence ratings (negative or positive) were associated with better performance than items with more neutral ratings. We conclude that using the item valence ratings we report, and treating item valence as a continuous rather than categorical predictor, will help bring consistency to the study of the association between item valence and performance in the RMET.