The goal of this study was to determine the ability of various horizontal and vertical reference lines to provide measurements of dentofacial asymmetries from posteroanterior (PA) cephalograms. Ten horizontal and 15 vertical reference lines, including best-fit lines and lines most commonly used in PA analysis, were tested. A model of a dry skull was devised to create 30 asymmetric positions of the maxillomandibular complex. The true transverse and vertical asymmetries were calculated based on measurements of changes in the position of 24 dental and skeletal landmarks. A PA cephalogram was obtained for each asymmetric position. The horizontal and vertical reference lines were constructed on each PA cephalogram, and measurements of transverse and vertical asymmetries were obtained relative to the individual reference lines. Linear regression analyses were used to compare the actual asymmetries with those measured cephalometrically, relative to the individual reference lines. The adjusted R(2) values for all 10 horizontal lines indicated excellent agreement between the true asymmetries and the measured vertical asymmetries. Ten vertical lines accurately represented transverse asymmetry. Vertical lines constructed between 2 midline points, with 1 point located on the lower part of the skull, were not valid. The best-fit line and all lines constructed as perpendiculars through midpoints between pairs of orbital landmarks showed excellent validity. Crista galli-anterior nasal spine and nasion-anterior nasal spine had the lowest validity and should not be used in cephalometric analysis of asymmetries. The position of anterior nasal spine will be altered in facial asymmetry involving the maxilla.