An analysis was made of repeated bone-specific skeletal age assessments of 90 hand-wrist roentgenograms by 4 observers who used the Greulich-Pyle atlas. Replicability was changed only slightly by including the skeletal ages of the carpal bones when calculating mean skeletal ages for the hand-wrist. The exclusion of the carpals could be recommended on this basis. However, the carpals are very variable in age at onset of ossification and perhaps at age of achieving other stages of skeletal maturity. Consequently, they may provide valuable discriminant data. The replicability of assessments of skeletal age was not reduced in individuals whose hand-wrist skeletal ages were markedly divergent from their chronologic ages. When there was a wide range of skeletal maturity between the bones of a hand-wrist, replicability was lower for over-all assessments, but not for bone-specific assessments. Roentgenographic factors (exposure, positioning) did not affect replicability within the range of variation in these factors in the roentgenograms assessed in the present study. In the assessments of single roentgenograms and in estimates of increments in skeletal maturity, use of the means of assessments of paired observers rather than assessments by single observers significantly reduced the interobserver differences but not the intraobserver differences. The mean absolute interobserver differences for single roentgenograms (mean of all bones) was 0.43 years for single observers, but 0.23 years for paired observers. The corresponding mean differences for increments were 0.68 and 0.31 years. These findings show the desirability of using the means of assessments by paired observers.