We performed a point-prevalence survey of 15,799 children, six to fourteen years old, who formed part of a prospective longitudinal study. Our purpose was to detect the prevalence of scoliosis and to investigate associated factors. On the basis of the initial screening, 934 children (5.9 per cent) were referred for additional clinical and radiographic examinations; 896 children returned for this second evaluation. A lateral spinal curve with a Cobb angle of more than 5 degrees was seen in 431 children (2.7 per cent of the 15,799 children). Only seventy-six children (0.5 per cent) had a curve that met our definition of idiopathic scoliosis (a curve of more than 10 degrees with concordant apical rotation). The point-prevalence rate was higher in girls, and it increased with age. The rate was 0.1 per cent (four of 5246) in the age-group of six to eight years, 0.3 per cent (sixteen of 5831) in the age-group of nine to eleven years, and 1.2 per cent (fifty-six of 4722) in the age-group of twelve to fourteen years old. With allowance for the fact that different definitions of idiopathic scoliosis have been used in earlier studies, our results suggest that the natural history of idiopathic scoliosis may be becoming more benign spontaneously.