Skinfold anthropometry is a widely practiced technique often with little appreciation of its limitations. The large residual error appearing in any regressions of body density versus sums of skinfolds is primarily due to biological causes, in particular the non-constancy of the ratio of subcutaneous to total body fat. Nevertheless this preliminary study shows that the residual error can be reduced by referencing to a gold standard other than body density. Using a difference technique involving in vivo neutron activation analysis and tritiated water dilution, this paper shows that at least in 20-29 year old normal subjects the residual error (expressed as a percentage of mean total body fat in the respective groups) can be reduced from 22% to 16% (p < 0.001) in males and from 17% to 9% in females (p < 0.0001). It is suggested that a large scale study could be initiated with this new gold standard to obtain accurate predictor relationships throughout the whole age range for both sexes.