The use of a camera for metrical purposes presupposes a knowledge of errors affecting the photograph. The aim of the present study was to test stereomodels of defined test objects with regard to inherent deformations. Three different test objects were used, a fine-lined grid and two test models resembling the teeth and the surrounding labial gingiva. The accuracy as given by the standard error of unit weight of absolute orientation was 0.018 mm (precision grid), 0.066 mm (convex surface) and 0.115-0.139 mm (jaw model). The precision under the test conditions as judged from the comparison of stereomodels from repeated exposures of the same object (jaw model) was between 0.009 and 0.032 mm. Since the stereomodel was deformed in a similar way irrespective of the test object used, it was concluded that part of the total error could be interpreted as a systematic deformation of the stereomodel. For furtherin vivoapplication of the method, with difficulties to identify reference points on the gingiva from repeated exposures, a simple way to estimate the accuracy of gingival measurements was derived by extrapolation from the standard error of the reference points on the teeth.