Body weight is one of the most important measurements in assessing nutritional status. Weight is also an important variable in equations that are needed to predict caloric expenditure and in indexes of body composition. Equations to predict the body weight of nonambulatory elderly persons with known errors should make use of body measurements that are indexes of the actual constituents of body composition and that can be collected regardless of an elderly person's level of mobility. The study samples consisted of a validation sample of 228 ambulatory white elderly persons, a cross-validation sample, and a clinical-validation sample of nonambulatory white elderly persons. Equations to predict body weight in the validation sample were derived from recumbent measures of arm and calf circumferences, triceps and subscapular skinfolds, and knee height. The accuracy, sensitivity, and validity of the equations were independently tested in the cross-validation sample; mean signed differences in actual and predicted weight ranged from 0.1 to 1.8 kg. In the clinical-validation sample, the mean differences were larger. Prediction equations are presented using two, three, and four body measurements to allow the selection of an equation based upon those measurements that are possible to collect from a nonambulatory person on an individual basis.