Abstract
Most studies investigating emotion perception have used dichotomous response measures whereby each response is either correct or incorrect. We used a novel continuous response paradigm to investigate the precision of visual working memory for expressions of sadness, anger, and fear and to investigate whether biases in errors (e.g., incorrectly perceiving angry faces as fearful rather than sad) are evident early in visual processing. We created an "emotion wheel" by morphing three anchor expressions (anger/sad/fear) in 4% steps. On each trial (n = 750), a target face (an anchor or any randomly selected morph) appeared for 500ms. After a 900ms delay, participants (n=29) located the target face on the emotion wheel (comprised of 75 faces representing continuous variation in emotion). We measured the magnitude (degrees between target and response) and direction (e.g., towards or away from particular emotions) of response error. The magnitude of response error varied with proximity of the target to an anchor expression (smaller for unambiguous [target contained >75% of one emotion, m = 47°] vs. ambiguous [target contained < 75% of either emotion, m = 60°] targets, p < .001) and across expressions (smaller for unambiguous angry compared to sad or fearful expressions, ps < .001; smaller for ambiguous angry/fear and fear/sad blends than angry/sad blends, ps < .01). The direction of response biases for ambiguous targets favored threat-related expressions. Participants were biased towards anger and fear when viewing anger/sad and fear/sad blends, ps < .05; no bias was observed for angry/fearful blends. Collectively, our findings suggest prioritization of both direct (angry) and indirect (fearful) threat, as opposed to merely negative (sad) faces. Our results have important implications for emotion theory and for understanding threat-related biases in emotion processing.
Meeting abstract presented at VSS 2018