This review addresses what is currently known about the time course of skill reacquisition after stroke. There is growing evidence that the natural logarithmic pattern of functional recovery can be modified by intensive task-oriented practice preferably initiated within 6 months after stroke. However, the impact of practice on the learning-dependent and intrinsic spontaneous mechanisms of neurological recovery is poorly understood. At least four probably interrelated mechanisms have been identified that drive motor and recovery after stroke: (1) salvation of penumbral tissue in the first days to weeks after stroke; (2) alleviation of diaschisis; (3) homeostatic and learning-dependent (Hebbian) neuroplasticity; (4) behavioral compensation strategies. These mechanisms underlying recovery are highly interactive, and operate in different, sometimes limited, time-windows after stroke onset. In line with these mechanisms of improvement after stroke, we present a hypothetical phenomenological model for understanding skill reacquisition after stroke. Translational research is important at this point to improve our knowledge about the neural correlates of what and how patients learn when they show functional improvement after stroke. This knowledge should serve as a basis to optimize the timing, focus and intensity of evidence-based rehabilitation interventions and to design innovative strategies to enhance motor recovery after stroke.