Existing models to estimate the metabolically active body cell mass (BCM) component in vivo remain incompletely developed. The classic Moore model is based on an assumed BCM potassium content of 120 mmol/kg. Our objectives were to develop an improved total body potassium (TBK)-independent BCM prediction model on the basis of an earlier model (Cohn SH, Vaswani AN, Yasumura S, Yuen K, and Ellis KJ. J Lab Clin Med 105: 305-311, 1985), to apply this improved model in subjects to explore the sex and age dependence of the TBK/BCM ratio, to develop a new TBK/BCM model on the basis of physiological associations between TBK and total body water (TBW) at the cellular level of body composition, and to fit this new model with available reference data. Subjects were 112 healthy adults who had the following components measured: TBW by 2H2O or 3H2O, extracellular water by NaBr, total body nitrogen by in vivo neutron activation, bone mineral by dual-energy X-ray absorptiometry, and TBK by whole body counting. Human reference data were collected from earlier published reports. The improved Cohn model-derived TBK/BCM ratio was (mean +/- SD) 109.0 +/- 10.9 mmol/kg and was not significantly related to sex and age. A simplified version of the new TBK-TBW model provided a TBK/BCM ratio almost identical (109.1 mmol/kg) to that derived by the improved Cohn model. The TBK-BCM prediction formula derived from the improved and new models [BCM (kg) = 1/109 x TBK (mmol); or BCM = 0.0092 x TBK] gives BCM estimates approximately 11% higher than the classic Moore model (BCM = 0.0083 x TBK) formulated on rough tissue composition estimates. The present analyses provide a physiologically based, improved, and validated TBK-BCM prediction formula that should prove useful in body composition and metabolism research.