Stroke often leads to impairment of hand function. Over the following months a variable amount of recovery can be seen. The evidence from animal and human studies suggests that reorganization rather than repair is the key. Surviving neural networks are important for recovery of function and non-invasive techniques such as functional magnetic resonance imaging allow us to study them in humans. For example, initial attempts to move a paretic limb following stroke are associated with widespread activity within the distributed motor system in both cerebral hemispheres, more so in patients with greater impairment. Disruption of activity in premotor areas using transcranial magnetic stimulation prior to movement can impair motor performance in stroke patients but not in controls suggesting that these new patterns of brain activity can support recovered function. In other words, this reorganisation is functionally relevant. More recently, research has been directed at understanding how surviving brain regions influence one another during movement. This opens the way for functional brain imaging to become a clinically useful tool in rehabilitation. Understanding the dynamic process of systems level reorganization will allow greater understanding of the mechanisms of recovery and potentially improve our ability to deliver effective restorative therapy.