The measurement of total-body electrical conductivity (TOBEC) has become one of the standard methods for the estimation of body composition in infants. We investigated, using non-human models, the effect on the accuracy of TOBEC-derived body composition estimates of alterations in physical and chemical characteristics of the fat-free mass (FFM). The effect of electrolyte type, concentration and volume on TOBEC was determined using 2, 3 and 51 solutions of six different chlorides and sodium bicarbonate. Equimolar concentrations yielded TOBEC values in accordance with known ion conductivities: H+ >> Ca2+ > Mg2+ > K+ > Na+ > Li+ and Cl- > HCO3-. The behaviour of these solutions was described very accurately over a wide range of concentrations (1-200 mM) by a simple exponential law. Dissolved egg-white protein, glycine and L-glutamine elicited no TOBEC signal. In vitro, using polyethylene bottles filled with physiologic saline, in the interval of 2-45 degrees C a linear relation was observed between temperature and TOBEC. Below the freezing point no TOBEC signal was elicited. The effect of tissue autolysis and body temperature on TOBEC was examined by repeated measurements of TOBEC and temperature in seven fresh infant minipig cadavers. Five minipigs were allowed to cool. Shortly after death TOBEC decreased by 2.5% per degrees C. Two animals were kept at constant temperature. The TOBEC signal showed a gradual increase of 9% after 7 h due to autolysis. We conclude that in vivo TOBEC measurements are affected by ion concentration (e.g. non-isotonic hydration changes), geometry (e.g. deviations in body shape), temperature (e.g. fever, skin cooling) and tissue autolysis (measurements after death). Proteins, molecules with strong dipole moments and ions trapped in crystalline structures do not significantly affect the TOBEC reading.