Objectives: This study assessed the influence of different voxel resolutions of two different CBCT units on the in vitro detection of periodontal defects. Methods: The study used 12 dry skulls with a maxilla and a mandible. Artificial defects (dehiscence, tunnel, fenestration) were separately created on the anterior, premolar and molar teeth using burrs. A total of 14 dehiscences, 13 fenestrations, 8 tunnels and 16 non-defect controls were used in the study. Images were obtained from two different CBCT units in six voxel sizes (voxel size: 0.080, 0.100, 0.125, 0.150, 0.160 and 0.200 mm 3 ). Kappa coefficients were calculated to assess both intra-and interobserver agreements for each image set. Results: Overall intraobserver kappa coefficients ranged between 0.978 and 0.973 for the 0.080-mm 3 images and between 0.751 and 0.737 for the 0.160-mm 3 images, suggesting notably high intraobserver agreement for detecting periodontal defects. CBCT performed significantly better at detecting fenestrations (p , 0.05) than tunnel and dehiscence defects. No statistically significant difference was found between the detection of dehiscence and tunnel defects (p . 0.05). Conclusions: A voxel size of 0.150 mm 3 was identified as the cut-off point for overall detection of periodontal defects. CBCT should be considered the most reliable imaging modality for the diagnosis of periodontal defects.