Over the last few decades, there have been considerable improvements in the outcome of stroke patients both as regards mortality and disability. At least some of these improvements can be attributed to better organization of services and improved rehabilitation. Many patients, however, remain severely disabled and we will need to develop new strategies in which the focus will be on reversing impairments rather than simply helping patients to adapt to unaltered impairments. For this to happen, neurological rehabilitation research will have to develop therapies that have a clearly defined rationale and are rooted in neurosciences, are clinically described, are addressed to a well-characterized target population and are evaluated using appropriate outcome measures. Few studies at present meet all these criteria. The recent revolution in our understanding of the nervous system as being soft-wired, of the potential for recovery through reorganization and of the central role of afferent information associated with normal activity is ground for optimism and indicates the direction in which future therapies should be sought. The paper considers some approaches to providing appropriate afferent information, including inputs such as that from electrotherapy, novel approaches to assisted activity and constraint-induced therapy. We are on the verge of a revolution in neurological rehabilitation. If we exploit the new understanding of the nervous system arising from basic neurosciences in developing and evaluating therapies we should be able to build on the achievements of the last few decades so that fewer of our patients have to carry the burden of severe disability. Copyright © 2002 Whurr Publishers Ltd.