The adult statures of a sample of 71 Canadian boys from the Saskatchewan Longitudinal Growth Study were predicted using the original TW Mark 1 and the new TW Mark 2 prediction equations. The subjects had a mean chronological age of 11.59 years (SD = 0.30), a mean RUS bone age of 11.62 'years' (SD = 1.18), a mean height of 145.0 cm (SD = 6.98) and a mean measured adult height of 177.2 cm (SD = 6.65). The Mark 2 equations improved the predictions over Mark 1 by an average of 0.2-0.6 cm and slightly reduced the range of errors. No improvement in the prediction of boys above the 75th centile of British standards was noted but 60-70% of boys below the 25th centile predicted better with the Mark 2 equations. This pattern may well be repeated in more extreme subjects. About 80% of individuals who predicted badly with the Mark 1 equations, i.e. with errors equal to or greater than 5 cm, improved their predictions when Mark 2 equations were used.