Periapical, panoramic, linear tomographic, and computerized tomographic radiographs were made of a partially dentate human mandible with four implants in place. Measurements taken from the radiographs and computer-generated images were compared to measurements made directly on the cross-sectioned test specimen. Periapical radiographs produced the most accurate measurements. Computerized and linear tomographic images produced the unique advantage of cross-sectional views of anatomic structures, but image blurring inherent to linear tomography and volume-averaging error inherent to computerized tomography affected the accuracy of measurements made from these images. The use of a dense dimensional reference object helped to compensate for radiographic distortion and is appropriate when projection radiographic techniques are used. The CT reference scale provided the most accurate method for interpreting measurements made from the computerized tomogram. The use of a dense dimensional reference object is inappropriate with CT because of its volume-averaging error.