To determine if computed tomography (CT) can accurately measure lung volume, we compared lung gas volume measured by helium dilution with the equivalent volume calculated from CT total lung volume and density in 13 supine dogs. CT lung gas volume underestimated helium volume by 34% (range: -63 to 0%). Studies of wooden lung phantoms varying in density from 0.082g/cc to 0.776g/cc showed that only 15% of this error could be mimicked by the phantoms. The rest of the discrepancy is attributed to the lung's irregular borders, and the sharp density gradients surrounding and within the lung that result in x-ray beam hardening, sampling limitations, and partial volume measurement errors. Serial biweekly measurements in three dogs for 14 weeks showed CT gas volume to be highly reproducible with less scatter than seen in the helium measurements. Density in the lungs of all dogs showed a uniform gradual decrease from approximately 0.60g/cc at the dependent surface to 0.20g/cc at the superior surface with relatively constant density at any horizontal level. These studies show that whereas CT underestimates gas volume in the lungs, serial measurements are highly reproducible in experimental studies and are a promising technique to monitor diseases or response to therapy. Density gradients in the lungs were sufficiently uniform so that disruption of the normal gradient may be an indicator of early lung disease.