The effect of skinfold caliper, investigator technique, and various skinfold prediction equations on estimates of body fatness was investigated using college-age female basketball players. Four skinfold calipers, four investigators, and five prediction equations from the literature were selected for study. Skinfolds were obtained on 16 athletes for each of five sites by all four investigators using each caliper. Triceps and subscapular skinfolds showed less variation among investigators (mean differences of 1-4 mm) as compared to suprailiac, abdomen, and thigh skinfolds (mean differences of up to 5-6 mm). Differences among investigators were less using the Harpenden and Holtain than for the Lange and Adipometer calipers. All three sources of variation--caliper, investigator, and prediction equation--contributed significantly to the variability in estimates of mean fat content of this sample, which ranged from 14.1-28.1% depending on which of the 80 caliper-investigator-prediction equation combinations (four calipers, four investigators, and five equations) was used. Using any one prediction equation, the range in mean percent fat due to the 16 caliper-investigator combinations was 7%. The results illustrate the need for standardization of skinfold sites, caliper, and prediction equations for reliable estimation of body composition in a specific population.