Investigating the relative severity of emotion recognition deficit across different clinical and high-risk populations has potential implications not only for the prevention, diagnosis and treatment of these diseases, but also for our understanding of the neurobiological mechanisms of emotion perception itself. We reanalyzed data from 4 studies in which we examined facial expression and gender recognition using the same tasks and stimuli. We used a standardized and bias-corrected measure of effect size (Cohen’s D) to assess the extent of impairments in frontotemporal dementia (FTD), Parkinson’s disease treated by L-DOPA (PD-ON) or not (PD-OFF), amnestic Mild Cognitive Impairment (aMCI), Alzheimer’s disease at mild dementia stage (AD), major depressive disorder (MDD), remitted schizophrenia (SCZ-rem), first-episode schizophrenia before (SCZ-OFF) and after (SCZ-ON) medication, as well as unaffected siblings of partients with schizophrenia (SIB). Analyses revealed a pattern of differential impairment of emotion (but not gender) recognition, consistent with the extent of impairment of the fronto-temporal neural networks involved in the processing of faces and facial expressions. Our transnosographic approach combining clinical and high-risk populations with the impact of medication brings new information on the trajectory of impaired emotion perception in neuropsychiatric conditions, and on the neural networks and neurotransmitter systems subserving emotion perception.