After stroke, patients must learn to use residual motor function correctly. Consistently, motor learning is crucial in stroke motor recovery. We assessed motor performance, practice-dependent on-line motor learning, and factors potentially affecting them in stroke patients. This is a cross-sectional observational study. Twenty-six patients with first brain stroke leading to upper limb motor deficit in the subacute or chronic timeframe were enrolled. They performed a Finger Tapping Task (FTT) with both the affected and unaffected limbs. We assessed how patients learn to perform motor tasks despite the motor deficit and the differences in performance between the unaffected and affected limbs. Furthermore, by randomizing the order, we evaluated the possible inter-limb transfer of motor learning (i.e. transfer of a motor skill learned in one limb to the opposite one). Moreover, sleep, attention, anxiety, and depression were assessed through specific tests and questionnaires. Improved FTT accuracy and completed sequences for the affected limb were observed, even if lower than for the unaffected one. Furthermore, when patients initially performed the FTT with the unaffected limb, they showed higher accuracy in subsequent task completion with the affected limb than subjects who started with the affected limb. Only anxiety and attentional abilities showed significant correlations with motor performance. This work provides relevant insights into motor learning in stroke. Practice-dependent on-line motor learning is preserved in stroke survivors, and an inter-limb transfer effect can be observed. Attentional abilities and anxiety can affect learning after stroke, even if the effect of other factors cannot be excluded.