Existing methods of assessing neurobehavioral responsiveness in severely brain-injured patients are limited by their inability to recognize subtle clinical changes over time. This study evaluates the Coma Recovery Scale (CRS), developed for use during acute rehabilitation. The CRS was designed to detect subtle changes in neurobehavioral status and to predict outcome in patients with sever alterations of consciousness. Acceptable levels of concurrent validity were established with the Disability Rating Scale ([DRS], r = -.93) and the Glasgow Coma Scale ([GCS], r = .90). The CRS was also found to have adequate interrater reliability (kappa = .83). Twenty-eight minimally responsive patients, unable to communicate or follow commands reliably, were evaluated on the CRS, the DRS, and the GCS. Although initial scores on all three scales were significantly correlated with outcome at discharge, change scores were more predictive of outcome. In addition, CRS change scores correlated higher with outcome (r = -.78, p less than .01) than DRS and GCS change scores. These findings indicate that the CRS is a sensitive measure of neurobehavioral responsiveness and suggest that rate of change during acute rehabilitation may be an important outcome predictor.