The accuracy of height predictions at various ages based on five different methods (Tanner-Whitehouse mark I; Tanner-Whitehouse mark II; index of potential height; Bayley-Pinneau; Roche-Wainer-Thissen) was compared at yearly intervals with final height achieved in 32 boys (78 predictions) and 100 girls (227 predictions) with constitutionally tall stature. The boys were initially seen at a mean (SD) chronological age of 12.5 (3) years whereas the mean chronological age in girls was 11.8 (2.1) years. In tall boys Tanner-Whitehouse mark II gives a good estimation of final height up to the bone age of 13 years with a mean overestimation of 1 cm. The overestimation of final height is higher in the bone age groups 13-14 years (2.7 cm) and 14-15 years (3.4 cm) mainly due to the tall boys with a height greater than 3 SD scores. Up to the bone age of 12 years the final height is massively overestimated by the Bayley-Pinneau method but this method give relatively accurate estimations thereafter. The estimated confidence limits are large (+/- 8 cm) for the two methods up to a bone age of 15 years. In tall girls the Tanner-Whitehouse mark II method was accurate from bone age nine to 12 years but overestimated final height in the bone age groups 12-13 years and 13-15 years by a mean of 1.8 and 1.4 cm respectively. The Bayley-Pinneau method overestimated final height in the bone age groups 12-14 years whereas the height predictions are accurate thereafter. Up to a bone age of 13 years the estimated confidence limits for the two methods are large, +/- cm, but tend to improve thereafter. It is concluded that there is no best or most accurate method for predicting adult height in tall children. There are methods of first choice differing with respect to sex and bone age. In addition, correcting factors may improve their accuracy and correct their tendency to overestimate or underestimate adult height.