The aims of this study were to reconstruct a three-dimensional (3D) model, to provide a 3D analysis using biplanar cephalograms [postero-anterior (PA) and lateral] with orthogonal projection, and to compare the differences between the 3D and two-dimensional (2D) analyses. The procedures were as follows: (1) to identify landmarks from a dry skull and to construct the norms for the spatial information of the skull from the 3D reconstruction using computerized tomography (CT); (2) to reconstruct a 3D model using biplanar cephalograms (PA and lateral); and (3) to compare the differences between the 2D and 3D analyses. Fifteen clearly visible landmarks identified on both films were used in this study. By comparing the data from the CT and the biplanar cephalograms, it was found that the accuracy for the 3D linear measurements from biplanar cephalograms was 98.9 per cent. However, the accuracy for the linear measurements from 2D and CT data was only 89.2 per cent. If the measurement of gonion (Go) to menton (Me) was excluded, the accuracy for the linear measurements from 2D and CT data was 95.1 per cent. When using a t-test to compare the linear distances of 2D-CT and 3D-CT data (Go to Me excluded), the difference was statistically significant (P < 0.05). The findings indicate that biplanar cephalograms with orthogonal projection are able to provide a 3D analysis that is more accurate than 2D analysis.