The reproducibility of nine mandibular variables (linear dimensions and angles) assessed from panoramic radiographs with the Orthopantomograph 5 (Siemens) was investigated. Attention was given to the possible influence of recording the reference number of the head positioner with one or two radiographers. Two separate exposures of three groups of patients were made under different radiographic conditions, each group representing one method. Acceptable reproducibility was observed for the vertical and angular variables, the method variance being mostly within 3% of the total variance. Horizontal variables were clearly more unreliable. No statistically significant differences were observed between the reproducibility of the right and left sides. A negative correlation was found between the angular variables within two groups. For most variables, only small differences among the methods were found. The highest reliability was obtained when the same radiographer recorded the reference number of the head positioner and made both exposures. An accuracy study on five dried skulls showed an image magnification of approximately 18% to 21% for the vertical variables, whereas the gonial angle assessed from a panoramic film was almost identical to that measured on the dried mandible.