The aim of the present study was to evaluate a model of body composition for assessing total body protein (TBP) mass using dual-energy X-ray absorptiometry (DXA), with either measured or assumed total body water (TBW); it was intended to provide a less complex or demanding alternative technique to, for example, the four-component model (4-CM). The following measurements were obtained in healthy adults (n 46) aged 18--62 years, and children (n 30) aged 8--12 years: body weight (BWt), body volume (BV; under-water weighing), TBW ((2)H-dilution space or predicted using an assumed hydration fraction of fat-free mass (HF(ffm))), bone mineral content (BMC; DXA) and fat-free soft tissue (FFST; DXA). TBP was calculated using the 4-CM (TBP = 3.05BWt -- 0.290TBW -- 2.734BMC -- 2.74BV) and the DXA model (TBP = FFST -- 0.2302BMC -- TBW). DXA measurements were obtained using the Lunar DPX (Lunar Radiation Corporation, Madison, WI, USA) or Hologic QDR 1000/W (Hologic, Waltham, MA, USA). Precision of the DXA model for TBP with measured TBW (4.6--6.8 % mean TBP) was slightly worse than the 4-CM (4.0--5.4 %), whereas that modelled with assumed HF(ffm) was more precise (2.4--5.2 %) because it obviated imprecision associated with measuring TBW. Agreement between the 4-CM and DXA model with measured TBW was also worse (e.g. bias, 15 % of the mean; 95 % limits of agreement up to +/-39 % for adults measured on the Lunar DPX) than when a constant for HF(ffm) was assumed (3.7 % and +/-21 % respectively). Most of the variability in agreement between these various models was due to interpretation of biological factors, rather than to measurement imprecision. Therefore, the DXA model, which is less complex and demanding than the 4-CM, is of value for assessing TBP in groups of healthy subjects, but is of less value for individuals in whom there may be substantial differences from reference 4-CM estimates.