1. Body-weight, body height, knee width, wrist width and skinfold measurements were made on males (n 139) and females (n 167) in three age-groups (20–22, 25–27 and 30–32 years). Percentage of body fal was calculated from skinfold thicknesses using regression equations according to Durnin & Womersley (1974), Three indices of relative weight were calculated: W/H2, W/Hp and W/Ŵ, where W is body-weight, H is body height, p is the exponent that made the index of relative weight independent of height and W is the weight, estimated from body height and frame size.
2. The standard error of the estimate of body-weight was only reduced by 5% in males and by 13% in females when, in addition to body height, knee width was taken into account.The addition of wrist width did not improve the accuracy of estimation of body-weight in either sex. Therefore in further analyses W was estimated from body height and knee width. In the present population the exponent p was 1·7 in males and 1·6 in females.
3. The correlations between the percentage of body fat and the indices, W/H2, W/HP, and W/Ŵ, were all very similar, being approximately 0·8 in both sexes.
4. A positive relationship was observed between percentage of body fat and knee width in females, which may be explained by an artifact of measurement.
5. In conclusion it can be stated that the accuracy of estimation of percentage of body fat was not improved when the index of relative weight was adjusted for knee width or wrist width in the present population. The W/H2was the most preferable of the three indices which were calculated.