Total body electrical conductivity (TOBEC) has become an increasingly employed method of estimating lean and lipid mass in small birds and mammals. The TOBEC value of an animal is highly correlated with lean mass. Prediction of lean mass with a linear regression equation allows for estimation of lipid mass by subtracting lean mass from total body mass (2-stage model). The accuracy of 2-stage models for predicting lipid mass has been questioned by previous investigators, who suggested that lipid mass was predicted more accurately directly from body mass and TOBEC (direct model). We examined the accuracy of TOBEC for predicting lipid mass in the eastern woodrat (Neotoma floridana) and southern plains woodrat (Neotoma micropus). Four 2-stage models, varying in the use of TOBEC as the independent or dependent variable and in the use of a transformed TOBEC value, were tested. Three direct models were tested in which lipid mass was predicted from TOBEC value or transformed TOBEC value and body mass. The direct models performed much better than 2-stage models in predicting lipid mass, with a relative error in predicted lipid mass of 80% less than the best 2-stage model. There was little difference in predictive error between 2-stage models that used TOBEC as the independent or dependent variable. The TOBEC value transformed for body geometry did not greatly improve either the 2-stage or direct models. Total body electrical conductivity showed potential in the prediction of lipid mass in woodrats and was the most accurate non-invasive method of determining lipid mass available.