The aim of the present study was to evaluate and compare the reproducibility of cephalometric landmarks on (1) conventional films, and images acquired by storage phosphor digital radiography both on (2) hardcopy and (3) monitor-displayed versions. The material consisted of 19 cephalograms for each image modality. The phosphor plates were scanned in an image reader and the 10-bit normalized, raw data digital images were converted to 8-bit TIFF images for PC monitor-display. The digital hardcopies were produced in a laser printer. Six observers were asked to record 21 cephalometric landmarks on each conventional film, hardcopy, and monitor-displayed image. For the films and hardcopies, the landmark co-ordinates were recorded via a digitizing tablet. For the monitor-displayed images, the co-ordinates were recorded directly from the monitor using a dedicated Windows-based cephalometric program. Reproducibility was defined as an observer's deviation (in mm) from the mean between all observers. Differences between the image modalities and between the observers were tested by two-way analysis of variance for each landmark. There was a statistically significant difference between the reproducibility of film, hardcopy and monitor-displayed images in 11 of the 21 landmarks. There was no unequivocal trend that one modality was always the best. For a full cephalometric recording (the sum of all 21 landmarks), the monitor-displayed images (mean = 25.3 mm) had a lower precision than film (P < 0.005) and hard-copy (P < 0.02). There was no significant difference between film (mean = 21.8 mm) and hardcopy (mean = 22.8 mm). The lower reproducibility seen for the monitor-displayed images is most probably of little clinical significance.