Interobserver variability for six coma scales was assessed prospectively on a sample of 15 comatose children, by two physicians, in a double-blind fashion. The six scales were the Glasgow Coma Scale, the Simpson and Reilly Scale, the Children's Coma Score, the Children's Orthopedic Hospital and Medical Center Scale, the Jacobi Scale, and the 0 to IV Scale. Interobserver variability was measured by using disagreement rate and the kappa statistic. The results from both methods were generally concordant. The disagreement rate for the various items in the different scales ranged from a high of 0.20 to a low of 0.03. The disagreement rate was greater than 0.10 for verbal responses in the Children's Coma Score and Glasgow Coma Scale and for both items in the Children's Orthopedic Hospital and Medical Center Scale. The disagreement rate was 0.10 or less for the 0 to IV Scale and for all items in the Simpson and Reilly and Jacobi scales. The relatively high interobserver agreement for these scales makes them more suitable for the pediatric setting than the other three scales, since good agreement is essential for interpreting data reliably, both in clinical practice and for research.