The nature of facial affect recognition (FAR) deficits in subjects at ultra-high risk (UHR) for psychosis remains unclear. In schizophrenia, associations between FAR impairment and poor neurocognition have been demonstrated meta-analytically, but this potential link is understudied in the UHR population. Our study investigated a cross-sectional sample of UHR subjects (n. = 22) and healthy controls (n. = 50), with the Degraded Facial Affect Recognition (DFAR) Task and a neurocognitive test battery. Our primary aims were 1. to examine associations between FAR and neurocognition in UHR subjects and 2. to examine if associations differed between cases and controls. The secondary aim was to examine group differences in FAR and neurocognitive performance. In UHR subjects, FAR was significantly associated with working memory, a neurocognitive composite score and intelligence, and at trend level with most other assessed neurocognitive domains, with moderate to large effect sizes. There were no significant associations in controls. Associations between FAR and working memory and the neurocognitive composite score differed significantly between cases and controls. UHR subjects did not differ from controls on DFAR Task performance but showed significant deficits in three of six neurocognitive domains. Results may suggest that FAR is associated with working memory in UHR subjects, possibly reflecting a neurocognitive compensatory mechanism.