Radiographs of 50 men and 50 women from Taiwan were of high quality and were generally characterized by optimal contrast and no blurring of anatomical structures. Each film was traced and cephalometric landmarks were determined by an examiner who did not know the sex of the subject and was then checked by another examiner. Films were digitized into an x-y coordinate system twice on two separate occasions, and average generated values were taken as actual measurements. Superciliary ridges, frontal sinuses, external occipital protuberance, and mastoid processes were adopted as objects of lateral radiographic cephalometric measurements. With discriminant functions created from 18 established cephalometric variables, 100 cases were classified according to two sex groups with 100 percent accuracy. Mean male values for all angular measurements were smaller than mean female values, but mean male values for linear and proportional measurements were larger than mean female values. Angular and ratio measurements were used to depict patterns and to minimize effects of size increases or changes. The radiographic cephalometric method made it possible to describe three- dimensional characters of the skull on a roentgenogram that presented a two-dimensional image. 15 references, 4 tables, and 2 figures