OBJECTIVE: To compare 16 currently used total body fat methods to a six-compartment criterion model based on in vivo neutron activation analysis. DESIGN: Observational, inter-method comparison study. SUBJECTS: Twenty-three healthy subjects (17 male and 6 female). MEASUREMENTS: Total body water (TBW) was measured by tritium dilution; body volume by underwater weighing (UWW); total body fat and bone mineral by dual-energy X-ray absorptiometry (DXA), total body potassium (TBK) by whole-body 40 K counting; total body carbon, nitrogen, calcium, phosphorus, sodium and chlorine by in vivo neutron activation analysis; skinfolds/circumferences by anthropometry (Anth); and resistance by single-frequency bioimpedance analysis (BIA). RESULTS: The average of total body fat mass measurements by the six-compartment neutron activation model was 19.7 AE 10.2 kg (mean AE s.d.) and comparable estimates by other methods ranged from 17.4±24.3 kg. Although all 16 methods were highly correlated with the six-compartment criterion model, three groups emerged based on their comparative characteristics (technical error, coef®cient of reliability, Bland-Altman analysis) relative to criterion fat estimates, in decreasing order of agreement: 1. multi-compartment model methods of Baumgartner (19.5 AE 9.9 kg), Heyms®eld (19.6 AE 9.9 kg), Selinger (19.7 AE 10.2 kg) and Siri-3C (19.6 AE 9.9 kg); 2. DXA (20.0 AE 10.8 kg), Pace-TBW (18.8 AE 10.1 kg), Siri-2C (20.0 AE 9.9 kg), and Brozek-UWW (19.4 AE 9.2 kg) methods; and 3. Segal-BIA (17.4 AE 7.2 kg), Forbes-TBN (21.8 AE 10.5 kg), Durnin-Anth (22.1 AE 9.5 kg), Forbes-TBK (22.9 AE 11.9 kg), and Steinkamp-Anth (24.3 AE 9.5 kg) methods. CONCLUSION: Relative to criterion fat estimates, body composition methods can be organized into three groups based on inter-method comparisons including technical error, coef®cient of reliability and Bland-Altman analysis. These initial groupings may prove useful in establishing the clinical and research role of the many available fat estimation methods.