For social species, e.g., primates, the perceptual analysis of social interactions is an essential skill for survival, emerging already early during development. While real-life emotional behavior includes predominantly interactions between conspecifics, research on the perception of emotional body expressions has primarily focused on perception of single individuals. While previous studies using point-light or video stimuli of interacting people suggest an influence of social context on the perception and neural encoding of interacting bodies, it remains entirely unknown how emotions of multiple interacting agents are perceptually integrated. We studied this question using computer animation by creating scenes with two interacting avatars whose emotional style was independently controlled. While participants had to report the emotional style of a single agent, we found a systematic influence of the emotion expressed by the other, which was consistent with the social interaction context. The emotional styles of interacting individuals are thus jointly encoded.