In most of the studies on long-term radiographic evaluations of crestal bone levels adjacent to dental implants, no baseline radiographs taken immediately postsurgically had been obtained. The aim of this study was to test the reproducibility of a simple radiographic method for linear measurements of changes in bone levels and to evaluate changes in crestal bone levels adjacent to non-submerged ITI implants 1 year following the surgical procedure. From 128 patients enrolled in a clinical and radiographic longitudinal study 40 patients also had radiographs taken immediately postsurgically. They were, however, not obtained as "identical" images. The radiographs were mounted onto slides and projected on a screen. Mesially and distally from 57 implants triplicate linear measurements of the distance implant shoulder to bone crest were taken, using known dimensions of the implants as internal reference distances. The median difference of 213 (out of 228 possible) duplicate measurements was 0.00 mm (ranging from -1.72 mm to +1.47 mm when comparing the second to the third reading). Some 81% of the double measurements were within +/- 0.5 mm and the precision was 0.30 mm. In the immediate postoperative radiographs the median mesial bone level was located at 2.07 mm (distally 2.19 mm) from the implant shoulder. A statistically significant amount of bone loss in the first year was observed mesially (median = -0.78 mm) and distally (-0.85 mm) (Wilcoxon matched pairs signed rank test P < or = 0.001). No statistically significant influence of the implant location, the implant length, type of the implant (screw; cylinder) was observed (Kruskal-Wallis P > 0.05). The age of the patients was not correlated significantly to the amount of bone loss observed. In conclusion, methodological limitations existed when evaluating linear bone changes in non-identical radiographs using reference dimensions of the implants. The amount of postsurgical bone loss estimated in other studies was confirmed when using an immediate postoperative radiograph as a baseline.