A number of therapies in development for patients with central nervous system injury aim to reduce disability by improving function of surviving brain elements rather than by salvaging tissue. The current study tested the hypothesis that, after adjusting for a number of clinical assessments, a measure of brain function at baseline would improve prediction of behavioral gains after treatment.Twenty-four patients with chronic stroke underwent baseline clinical and functional MRI assessments, received 6 weeks of rehabilitation therapy with or without investigational motor cortex stimulation, and then had repeat assessments. Thirteen baseline clinical/radiological measures were evaluated for ability to predict subsequent trial-related gains.Across all patients, bivariate analyses found that greater trial-related functional gains were predicted by (1) smaller infarct volume, (2) greater baseline clinical status, and (3) lower degree of activation in stroke-affected motor cortex on baseline functional MRI. When these 3 variables were further assessed using multivariate linear regression modeling, only lower motor cortex activation and greater clinical status at baseline remained significant predictors. Note that lower baseline motor cortex activation was also associated with larger increases in motor cortex activation after treatment.Lower motor cortex activity at baseline predicted greater behavioral gains after therapy, even after controlling for a number of clinical assessments. The boosts in cortical activity that paralleled behavioral gains suggest that in some patients, low baseline cortical activity represents underuse of surviving cortical resources. A measure of brain function might be important for optimal clinical decision-making in the context of a restorative intervention.