The reliability of repeated measurements of thoracic kyphosis from rasterstereographic back shape imaging was examined in 10 volunteers across different age cohorts and in a static back phantom. Coefficients of variation (CV) for five repeated measurements of the 10 subjects ranged from 2.4% to 3.0% for the kyphosis parameters ‘Maximum’, ‘Vertebra prominens to thoracolumbar inflexion point’, and ‘Vertebra prominens to T12’. Mean standard error of measurement values (SEM) as a percentage of mean kyphotic angles were 5.3%, 4.4% and 4.8% for the same three kyphosis parameters, ‘Maximum’, ‘Vertebra prominens to thoracolumbar inflexion point’, and ‘Vertebra prominens to T12’ respectively. The inter-trial CV for repeated imaging of the back phantom ranged from 0.4% to 1.3%. Intraclass correlation coefficients (ICC) ranged from 0.98 to 0.99 for all measurements. Consistent results were demonstrated across the varying age cohorts. The reliability of rasterstereographic evaluation of thoracic kyphosis is largely influenced by postural variations, while minimal errors are attributed to internal system inaccuracies. These results suggest the clinical utility of rasterstereographic imaging in the detection of relatively small curve changes and hence in monitoring of kyphosis progression.