Sequential stages in the development of the hand, wrist, and cervical vertebrae commonly are used to assess maturation and predict the timing of the adolescent growth spurt. This approach is predicated on the idea that forecasts based on skeletal age must, of necessity, be superior to those based on chronologic age. This study was undertaken to test this reasonable, albeit largely unproved, assumption in a large, longitudinal sample. Serial records of 100 children (50 girls, 50 boys) were chosen from the files of the Bolton-Brush Growth Study Center in Cleveland, Ohio. The 100 series were 6 to 11 years in length, a span that was designed to encompass the onset and the peak of the adolescent facial growth spurt in each subject. Five linear cephalometric measurements (S-Na, Na-Me, PNS-A, S-Go, Go-Pog) were summed to characterize general facial size; a sixth (Co-Gn) was used to assess mandibular length. In all, 864 cephalograms were traced and analyzed. For most years, chronologic age, height, and hand-wrist films were available, thereby permitting various alternative methods of maturational assessment and prediction to be tested. The hand-wrist and the cervical vertebrae films for each time point were staged. Yearly increments of growth for stature, face, and mandible were calculated and plotted against chronologic age. For each subject, the actual age at onset and peak for stature and facial and mandibular size served as the gold standards against which key ages inferred from other methods could be compared. On average, the onset of the pubertal growth spurts in height, facial size, and mandibular length occurred in girls at 9.3, 9.8, and 9.5 years, respectively. The difference in timing between height and facial size growth spurts was statistically significant. In boys, the onset for height, facial size, and mandibular length occurred more or less simultaneously at 11.9, 12.0, and 11.9 years, respectively. In girls, the peak of the growth spurt in height, facial size, and mandibular length occurred at 10.9, 11.5, and 11.5 years. Height peaked significantly earlier than both facial size and mandibular length. In boys, the peak in height occurred slightly (but statistically significantly) earlier than did the peaks in the face and mandible: 14.0, 14.4, and 14.3 years. Based on rankings, the hand-wrist stages provided the best indication (lowest root mean squared error) that maturation had advanced to the peak velocity stage. Chronologic age, however, was nearly as good, whereas the vertebral stages were consistently the worst. Errors from the use of statural onset to predict the peak of the pubertal growth spurt in height, facial size, and mandibular length were uniformly lower than for predictions based on the cervical vertebrae. Chronologic age, especially in boys, was a close second. The common assumption that onset and peak occur at ages 12 and 14 years in boys and 10 and 12 years in girls seems correct for boys, but it is 6 months to 1 year late for girls. As an index of maturation, hand-wrist skeletal ages appear to offer the best indication that peak growth velocity has been reached. Of the methods tested here for the prediction of the timing of peak velocity, statural onset had the lowest errors. Although mean chronologic ages were nearly as good, stature can be measured repeatedly and thus might lead to improved prediction of the timing of the adolescent growth spurt.