We investigated whether stereotype associations between specific emotional expressions and social categories underlie stereotypic emotion recognition biases. Across two studies, we replicated previously documented stereotype biases in emotion recognition using both dynamic (Study 1) and static (Study 2) expression displays. Stereotype consistent expressions were more quickly decoded than stereotype inconsistent expression on Moroccan and White male faces. Importantly, we found consistent and novel evidence that participants’ associations between ethnicities and emotions, as measured with a newly developed emotion Implicit Association Test (eIAT), predicted the strength of their ethnicity-based stereotype biases in expression recognition. In both studies, as perceivers’ level of Moroccan-anger and Dutch-sadness associations (compared with the opposite) increased, so did perceivers’ tendency to decode anger more readily on Moroccan faces and sadness on White faces. The observed stereotype effect seemed to be independent of implicit prejudice (Study 2), suggesting dissociable effects of prejudices and stereotypes in expression perception.