BACKGROUND: Studies about precision and reproducibility of cephalometric points, measurements and angles have shown a wide range of results, regarding the referential points of the drawings. OBJECTIVE: To analyze the reproducibility error in cephalometry when traces are done in a manual way or with a computerized method. METHODS: A descriptive study with an analytic observational design was done. Ten extraoral lateral radiographs, which presented high standards of clearness and resolution of the image, were taken from patients who were older than 15 years of age. Images excluded were those that presented distortion of size or shape, and also where no occlusal contact was evident. On every radiograph traces were done according to the Steiner’s cephalometric analysis. For the handmade drawing, 5 photocopies were taken, from each original trace, in order to determine the angular and linear measurements. The program used for the computerized cephalometric analysis was 6-Span; the computerized cephalometric analysis was done 5 times from every original trace. Results were registered in a table and the D’Ahlberg’s and the correlation coefficient tests were used in order to compare both methods. RESULTS: The computerized method showed higher error of reproduction in the majority of the measurements taken, with the exception of the angular and linear measure of I -I: NB. In terms of the correlation between both methods, nonsignificant difference regarding the errors was found in both groups.