The objective of this study was to develop a body composition method based on computerized tomography (CT) which would make it possible to divide the body into multiple compartments at the tissue and organ level. Eight healthy males (21-42 years old) with BMIs ranging from 18.6 to 25.3 kg/m2 were used for the methodological development. Areas of tissues, organs and air/gas were measured in 28 cross-sectional scans having defined and identical positions in all examined subjects. The area determinations were performed with the following attenuation intervals (given in Hounsfield units, HU): air, gas and lungs: -1001 to -191 HU; adipose tissue (AT): -190 to -30 HU; all other soft tissues and organs: -29 to +151 HU; skeleton: 152 to 2500 HU. Various tissue and organ areas in the -29 to +151 HU interval were obtained by means of cursor circumscriptions, while area determinations in other intervals were based on the number of pixels fulfilling given attenuation criteria. Volumes of tissues, organs and gas were obtained from corresponding areas and the distances between the scans. The body was divided into 12 main volumes of tissues, organs and gas that could be further subdivided by region. The main volumes observed (in litres; mean +/- s.d.) were: skeleton (subdivisible into dense skeleton, red and yellow bone marrow): 8.7 +/- 0.9; skeletal muscle: 31.9 +/- 5.1; visceral AT (subdivisible into intra- and retroperitoneal, cardiac, other thoracic AT): 3.0 +/- 1.7; intra- and retroperitoneal organs other than AT: 4.6 +/- 0.8; gastrointestinal gas: 0.25 +/- 0.09; heart: 0.61 +/- 0.12; lungs and bronchial air: 5.1 +/- 1.1; other thoracic organs: 0.32 +/- 0.08; mammary glands: 0.001 +/- 0.004; CNS (subdivisible into brain and contents of spinal channel): 1.6 +/- 0.15; air in sinuses and trachea: 0.19 +/- 0.05; subcutaneous AT: 11.6 +/- 2.8; skin: 2.4 +/- 0.39. Precision errors as determined from double analyses of different tissue volumes ranged from 0.01 to 0.3 litres. For validation purposes, CT-estimated organ weights were obtained by multiplying organ volumes by their assumed densities. The sums of all organ weights were then compared with the measured body weights. The error calculated from the individual differences between these weights was 0.6 kg (0.85%). The multicompartmentation technique described has a high validity and reproducibility and is applicable over a wide range of medical fields which require body composition measurements at the tissue and organ level.