The Neurobehavioural Rating Scale (NRS) was developed to allow quantification of clinical observations of the behaviour of individuals following traumatic head injury. Initial validation of this instrument reported satisfactory interobserver reliability, as well as preliminary support for its validity in differentiating both severity and chronicity of head injury. In the present study, reliability and content validity of the NRS were replicated using a population of severe head injury patients undergoing inpatient rehabilitation. Concurrent, independent assessment using the NRS was conducted weekly by two members of an inpatient, traumatic head injury team. Acceptable levels of interobserver reliability were found, despite less control over observed behaviour in the naturalistic setting used. Content validity was also supported in this replication, although recommendations for 'tightening' the NRS included better definition of the seven-point severity rating for each item, as well as potential deletion of two items that contributed little to differential variance in the total NRS score. The NRS appears to be a promising clinical and research tool for assessing the neurobehavioural sequelae of traumatic head injury. Its strengths include efficiency of administration and the flexibility to administer through brief, structured interviews or observations in a naturalistic setting. Continued research on the NRS in different settings and with different populations is required to establish its validity.