The orthodontic documentation carried out in radiological clinics is essential for diagnosis, planning and control of the orthodontic treatment. Amongst the diagnosis elements used are the cephalometric measurements, and errors can be incorporated as they are obtained. The objective of this work was to evaluate the values of some cephalometric measurements obtained in three radiological clinics using 30 lateral cephalometric radiographs of the head randomly chosen from the archives of the Department of Child Dentistry (Discipline of Orthodontics), School of Dentistry of Piracicaba, UNICAMP. These values were compared with the measurements obtained manually (control group) using variance analysis and Tukeys and Friedmans tests. Of the fourteen cephalometric measurements evaluated, the results demonstrated that only four of them presented statistically significant differences: IMPA, 1-NA, FMA, and H-nose. We concluded that although the majority of the mean cephalometric values did not present statistically significant differences, a great variability in the results was found when all the radiographic values were compared. This variability may influence the interpretation of the cephalometric measurements. Hence, we suggest that when the practitioner receives orthodontic documentation, he/she should redo the cephalometric analysis and compare the measurement values found with those presented to him. In addition, the practitioner should be prepared to use other elements for diagnosis, planning and control of the orthodontic treatment.