Biased judgments of emotion are resistant to changes in the prevalence of anger.

Emotional expressions are often nuanced and are thus evaluated with an unavoidable degree of uncertainty. We recently showed that perceivers are biased to interpret others' expressions negatively, especially when they are seen in crowds. We argued that the flexible nature of this bias is protective, and that it emerges because perceivers learn about the threatening nature of crowds over prolonged windows of time. Here we evaluated whether this negativity bias can recalibrate as perceivers learn about changes in the prevalence of threat in the short-term. Perceivers viewed single faces or crowds of four faces and indicated whether their expressions were happy or angry. We manipulated the prevalence of anger across three experiments, displaying angry faces on 75% or 25% of trials. We replicated our previous findings, again showing that observers were biased to evaluate others as angry, especially in crowds. Surprisingly, the strength of anger bias was not affected by the prevalence of anger, neither when perceivers learned about this implicitly nor when they were explicitly informed about it in advance. This suggests that anger bias is flexible, but more to contextual cues, like the number faces being evaluated, and less to statistical cues, like the prevalence of anger.