BMI is an indicator of body composition (BC) and adiposity in particular. This status is the result of good correlations with indirect (e.g. predictive) two- and three-component models predicting body fat. Aim of the study is to measure the direct relation of BMI with total and segmental BC, e.g. of skin, muscle, bone, viscera and adipose tissue.BC constituents of 29 white elderly persons (17 females and 12 males, aged 78.1+/-6.9 years) were determined by direct dissection. Correlations and stepwise linear regression analysis with BMI (dependent variable) and all BC constituents (independent variables) were calculated.All tissues dissected indicated a too high unexplained variance except for bone that accounted for 84% of BMI variance in females and 61% in males (P<0.001). No other constituents improved the prediction.The BMI cannot be an appropriate adiposity index. The high proportion of unexplained variance between BMI and direct BC constituents limit its use as a whole body and as a segmental BC index. The BMI could be an index for Bone Mass instead.