Experimental evidence suggests an impairment in emotion perception in numerous psychiatric disorders. The results to date are primarily based on research using static displays of emotional facial expressions. However, our natural environment is dynamic and multimodal, comprising input from various communication channels such as facial expressions, emotional prosody, and emotional semantics, to name but a few. Thus, one critical open question is whether alterations in emotion perception in psychiatric populations are confirmed when testing patients in dynamic and multimodal naturalistic settings. Furthermore, the impact task demands may exert on results also needs to be reconsidered. Focusing on schizophrenia and depression, we review evidence on how emotions are perceived from faces and voices in these disorders and examine how experimental task demands, stimulus dynamics, and modality may affect study results.