This study investigates emotion recognition deficits as candidate neurocognitive endophenotypes for callous-unemotional (CU) behaviors. Using a twin design, we tested genetic correlations between child CU behaviors and poor processing of fearful and sad facial expressions. Participants were 504 twin pairs (209 MZ pairs; 295 DZ pairs) from the Quebec Newborn Twin Study, a longitudinal study of a population-based sample of twins. Teachers in kindergarten and first grade rated children’s CU behaviors and other behavior problems (attention deficit and hyperactivity symptoms, physical aggression, and depressive symptoms). In first grade (mean age 7 years), the children completed the visual subtest of the Diagnostic Analysis of Nonverbal Accuracy Scale 2 (DANVA-II) to assess emotion recognition from facial stimuli. Using structural equation modeling, we examined the genetic-environmental etiology of the association between fear/sadness recognition and child CU behaviors, controlling for other behavior problems and recognition of other emotions. We found a significant genetic correlation between poor fear recognition and CU behaviors that was independent of other behavior problems. Poor recognition of sadness was not significantly associated with CU behaviors after taking into account other behavior problems. Our results suggest that CU behaviors and fear recognition have a partly shared genetic aetiology. This provides support for poor fear recognition as a key neurocognitive endophenotype for CU behaviors. Future research should test a hypothesized causal chain from specific genes, through amygdala functioning and fear recognition, to CU behaviors, and identify specific environmental factors (including intervention) that may disrupt this chain.