The use of absolute CT numbers for in vivo tissue characterization is compromised by a number of technical and geometrical factors. A phantom simulating thoracic geometry and containing intrapulmonary "features" was scanned on three CT scanners allowing for assessment of CT-number variations with a wide number of scanning and geometric parameters. It was found that, with thoracic geometry, the absolute CT numbers of intrapulmonary features (e.g., solitary pulmonary nodules) can vary significantly due to a number of CT scanning parameters, such as geometry, CT scanner used, and/or time. Such variations must be taken into account when establishing criteria for characterizing tissue types using CT numbers. However, results show that variations in quantitative behavior with reasonable changes in patient geometry do not preclude meaningful characterization of solitary lung nodules (using CT-number averages) using CT scans for the three CT scanners studied, providing CT-number threshold data are derived from the same model scanner and operating conditions. The use of CT-number patterns within high-density intrapulmonary pathology (e.g., solitary pulmonary nodules) is compromised by the fact that high CT-number patterns were found to be a function of the reconstruction filter used and object size as well as being influenced by details of the surrounding medium.