Background Knowing how impaired manual dexterity and finger proprioception affect upper limb activity capacity is important for delineating targeted post-stroke interventions for upper limb recovery. Objectives To investigate whether impaired manual dexterity and finger proprioception explain variance in post-stroke activity capacity, and whether they explain more variance than conventional clinical assessments of upper limb sensorimotor impairments. Methods Activity capacity and hand sensorimotor impairments were assessed using clinical measures in N = 42 late subacute/chronic hemiparetic stroke patients. Dexterity was evaluated using the Dextrain Manipulandum to quantify accuracy of visuomotor finger force-tracking (N = 36), timing of rhythmic tapping (N = 36), and finger individuation (N = 24), as well as proprioception (N = 27). Stepwise multivariate and hierarchical linear regression models were used to identify impairments best explaining activity capacity. Results Dexterity and proprioceptive components significantly increased the variance explained in activity capacity: (i) Box and Block Test was best explained by baseline tonic force during force-tracking and tapping frequency (adjusted R 2 = .51); (ii) Motor Activity Log was best explained by success rate in finger individuation (adjusted R 2 = .46); (iii) Action Research Arm Test was best explained by release of finger force and proprioceptive measures (improved reaction time related to use of proprioception; adjusted R 2 = .52); and (iv) Moberg Pick-Up test was best explained by proprioceptive function (adjusted R 2 = .18). Models excluding dexterity and proprioception variables explained up to 19% less variance. Conclusions Manual dexterity and finger proprioception explain unique variance in activity capacity not captured by conventional impairment measures and should be assessed when considering the underlying causes of post-stroke activity capacity limitations.