This study describes inter-observer agreement between physicians and consistency of one observer on repeated scoring in the assessment of outcome 6 to 12 months after severe head injury. Observer agreement is expressed by Kappa, taking chance agreement into account. The study was conducted in "live" and "non-live" situations, using five- and eight-category outcome scales. Kappa values were considerably higher when the five-category scale was used. However, even with the five-category scale, disagreement was considerable, and the results indicate that accurate predictions of the quality of survival will be difficult to attain. Observer agreement should be taken into account in prognosis studies, and when different therapeutic regimens are evaluated on the basis of outcome.