The absence of a peri-implant radiolucency on radiographs is used as a criterion for implant success. The purpose of the present study was to evaluate the accuracy of diagnosing peri-implant radiolucencies using an experimental model. Astra Tech fixtures were inserted into 20 bony specimens. Later, the fixtures were removed and the prepared cavities were enlarged in 2 steps. The specimens were radiographed under strictly standardized circumstances in the 3 stages ("press-fit" = simulated osseointegration, and "0.1 mm space" and "0.175 mm space" = simulated connective tissue layer). All specimens were radiographed with 2 vertical angulations (0 degree and 15 degrees). Ten observers read the radiographs and evaluated bone-implant relationship on a 5-point scale. The interobserver agreement was low. Specificity was remarkably low, and sensitivity was moderate. A significant difference in diagnostic accuracy was found for 0.175 mm peri-implant-space compared to 0.1 mm space. The diagnostic accuracy was found to be highest for 0.175 mm peri-implant space specimens. It is concluded, that radiography seems to be an unreliable method for diagnosing peri-implant spaces, although accuracy improved at increasing space widths. Strict orthogonal projection angles did not improve diagnostic accuracy.