The purpose of the present study was to evaluate the reproducibility of the landmark identification in frontal cephalograms obtained with anteroposterior (AP) projection by comparing the errors in the AP cephalograms with those in the posteroanterior (PA) cephalograms. A couple of frontal cephalograms, one with AP projection and another with PA projection, were obtained in 20 adult individuals using a digital cephalometer. Five examiners identified 6 landmarks (Cg, ANS, Me, Lo, J, Ag) on the monitor-displayed radiographs. In order to assess the intra-examiner errors, the landmarks were identified twice with two weeks of time interval by one of the examiners. The intra- and inter-examiner landmark identification errors in AP cephalograms were evaluated, and compared with those in PA cephalograms. 1. Scattergrams illustrating the errors of landmark identification in frontal cephalograms obtained with AP projection showed characteristic noncircular envelope of error in both intra- and inter-examiner scattergrams. Crista galli and ANS showed vertical direction errors. Menton and Antegonion showed hirizontal and oblique direction errors respectively. 2. Inter-examiner landmark identification errors were significantly larger than intra-examiner error. 3. Comparing the errors of AP cephalograms with those of PA cephalograms, none of landmarks were shown statistically significant differences between the two image modalities. 4. Analyzing landmark identification errors into horizontal and vertical direction, all landmarks were presented no significant differences. The results of the present study propose that frontal cephalograms obtained with AP projection may be used clinically with the same landmark identification reproducibility as the conventional PA projections.