This study evaluated the diagnostic accuracy of periapical, tomographic, and cross-sectional occlusal radiographic techniques in the assessment of facial and lingual bone loss at implant obscured sites. An edentulous dry human mandible was prepared with implants in the incisor, premolar, and molar regions. Successive round bur sizes (No. 1 to 6) were used to create circumferential infrabony periodontal defects in the alveolar crest adjacent to the implants. Periapical radiographs were made with custom stents and imaged each site in a conventional parasagittal orientation. Linear tomograms imaged sites in a cross-sectional coronal orientation. Cross-sectional occlusal radiographs imaged anatomy in the axial plane. Radiographic anatomy mesial and distal to each implant was masked with opaque tape before evaluation by a panel of six observers. Images were displayed in pairs consisting of an image at baseline and a second image having an equal chance of displaying an increment of bone removal or no bone removal. Image pairs were displayed in random order A five-point scale was used to evaluate accuracy of detection of the presence or absence of changes in bone density as well as observer confidence in that assessment. ANOVA demonstrated significant variation due to image modality and increment of bone removal. Cross-sectional occlusal views provided significantly greater mean observer confidence scores than either periapical images or tomograms (P < .01). However, tomograms may provide greater utility in actual clinical practice. (Implant Dent 1995;4:13–18)