Three observers rated 57 X-rays from normal healthy children in Project HeartBeat! twice each by CASAS, the computer-assisted version of the TW2 RUS bone age method. Differences between duplicates of individual bone ratings which reached or exceeded 1.0 unit (or 1 stage) were 5% within observer and 8% between observers for CASAS, and 17 and 33%, respectively, for the unassisted MANUAL method. In children followed longitudinally, CASAS scores increased much more steadily than MANUAL scores, largely because the bones were rated, in the former system, on a continuous rather than a discrete-integer scale. We conclude that CASAS is a more reliable and probably a more valid estimator of skeletal maturity than the MANUAL version of the TW2 RUS method.