Comparisons of back surface topography to monitor scoliosis are hampered by shape change due to other causes, primarily stance variations. The aim is to extract changes due only to the scoliosis from the observed changes. Surface (back) topography data were derived from 2 patients measured repeatedly (10x) at a single visit and from 45 patients measured at routine clinical visits. The measured back shapes were aligned to obtain the position of closest fit using a mathematical technique. The shape similarities were then quantified with a closeness-of-fit index. The similarity indices from the alignments were used to estimate noise factors due to postural variation, relative to typical patient change. Surface alignment has also been trialled using models with predictable topographic changes to determine whether the noise sources can be modelled. The values of the root-mean-square of the residual surface differences at all measured points after mathematical back shape alignment were used as a surface similarity index. The magnitude of this index for the repeatedly measured back shapes indicates the back shape variation due to stance, and has been compared with the variation found in normal patients. The similarity index has also been examined for cases both with and without modelled changes to estimate the effectiveness of stance modelling. The noise due to stance change was indicated by a surface similarity index which averaged 2.0 mm; the index for backs measured at routine visits averaged 3.5 mm. It was concluded that noise due to stance is significant and deserves to be recognised, in the comparison of back surfaces in three-dimensions. There is evidence that changes due to noise can be modelled.