Video rasterstereography has been developed for optical back shape measurement and for biomechanical analysis of spinal and pelvic geometry. Analysis of one single measurement permits 3-dimensional reconstruction of the back surface and calculation of shape parameters including pelvis tilt and torsion. In addition, estimates of the lateral deviation of the spinal midline and of vertebral rotation are provided. Its extended analytic potential makes rasterstereography a very appropriate tool for functional examinations. The term "functional examinations" refers in this context to biomechanical analysis of functional movements of the spine and pelvis caused by quasi-continuous changes of posture, if these can be observed as changes in back shape. Two examples are given to illustrate the aim and performance of functional examinations. Shoe elevation is used to correct leg-length discrepancy and is therefore prescribed for prevention and correction of scoliosis produced by pelvic obliquity. In a previous study it was shown that simulating leg-length discrepancy by raising a foot causes the pelvis to perform a torsional movement about the transverse axis. In effect, this movement reduces to some extent the effect of shoe elevation; thus a larger elevation might give better results. 42 scoliotic patients underwent functional examination. Leg-length discrepancies were simulated in 7 steps, and the resulting back shape was analysed by rasterstereography. The measurements were corrected for pelvic torsion. This method provided satisfactory correspondence with radiographically recorded leg-length discrepancies, i.e. 0.7 mm +/- 11.2 mm. One specific advantage of this procedure is that it covers aspects relating to spinal lateral deviation and vertebral rotation. It is concluded, however, that the 7 measurements used are hardly sufficient for this application and that better results are therefore to be expected from extended series. Kyphosis and lordosis clearly depend on posture. This is confirmed in a functional examination where these angles are measured under voluntary changes of posture. In forward bending, the trunk straightens and kyphosis and lordosis angles decrease. The reverse case applies to backward bending. If this effect is taken into quantitative consideration, an improved accuracy of measurements is obtained by reference to a standardised, mathematically defined posture. The resulting rms-error of kyphosis/lordosis measurement is then reduced from 3.10 degrees/2.95 degrees to 1.65 degrees/1.40 degrees. These figures open up new applications in the follow-up of kyphotic and lordotic deformities.