Previous methods used to collect human sweat for electrolyte analysis have been criticized because they involve only regional sampling or because of methodological problems associated with whole body-washdown techniques. An improved method for collection of whole body sweat from exercising subjects is described. It involved construction of a plastic frame that supports a large plastic bag within which the subject exercises. The subject and the equipment are washed with distilled, deionized water before exercise begins. After exercise is completed, the subject and equipment are again washed with water containing a marker not present in sweat (ammonium sulfate). Total sweat loss is calculated from the change in body mass, and the volume of sweat not evaporated is calculated from dilution of the added marker. Recovery of added water was 102 +/- 2% (SD) of the added volume, and recovery of added electrolytes was 99 +/- 2% for sodium, 98 +/- 9% for potassium, and 101 +/- 4% for chloride. Repeated trials (n = 4) on five subjects to establish the reproducibility of the method gave a coefficient of variation of 17 +/- 5% for sodium, 23 +/- 6% for potassium, and 15 +/- 6% for chloride. These values include the biological variability between trials as well as the error within the method. The biological variability thus appears to be far greater than the methodological error. Normal values for the composition of sweat induced by exercise in a hot, humid environment in healthy young men and women were (in mM) 50.8 +/- 16.5 sodium, 4.8 +/- 1.6 potassium, 1.3 +/- 0.9 calcium, 0.5 +/- 0.5 magnesium, and 46.6 +/- 13.1 chloride.