In order to determine why topographic methods have shown a poor correlation with radiographically measured scoliosis in clinical studies, the accuracy of detection of the presence, side, apex, and magnitude of a scoliosis curve was determined topographically (by moiré fringe photography and by projected raster photography) in 104 patients attending a scoliosis clinic. The presence or absence of thoracic curves was correctly shown by the topograms in 77% of cases, and in the lower region (lumbar and thoracolumbar curves) in 79% of cases. For correctly identified curves, the greatest back surface rotation was, on average, 1.0 vertebral levels below the skeletal curve apex in the thoracic region and 0.5 levels below the apex in the lower region. The moiré fringe with the greatest asymmetry occurred on average at 1.5 and 1.8 vertebral levels above the spinal apex in upper and lower regions, respectively. The magnitude of the Cobb angle was determined to within +/- 5 degrees in 24% of cases by moiré measurements, and in 27% by the raster technique. The side of the scoliosis was incorrectly diagnosed by topography in ten patients with minimal or 'nonstandard' vertebral rotation. It was concluded that the presence, level, and side of a scoliosis curvature is well demonstrated by back surface topography in patients with 'standard' rotation, but the magnitude of the scoliosis cannot be determined from topograms sufficiently accurately for most clinical purposes.