The determination of sex and the estimation of stature from bones play an important role in identifying unknown bodies, parts of bodies or skeletal remains. In medico-legal practice statements on the probable sex of a decomposed body or part of a body are often expected even during autopsy. The present study was, therefore, restricted to few easily accessible dimensions from bones which were prepared only by mechanically removing soft tissues, tendons and ligaments. The specimens came from the Anatomical Institutes in Munich and Cologne from the years 1994-1998 including a total of 143 individuals (64 males and 79 females). The mean age was 79 years (46-108), the mean body height 161cm (134-189). The following measurements were taken: maximum humeral length (mean: 33.4cm in males; 30.7cm in females), vertical humeral head diameter (mean: 5.0cm in males, 4.4cm in females), humeral epicondylar width (mean: 6.6cm in males; 5.8cm in females), maximum ulnar length (mean: 26.5cm in males, 23.8cm in females), proximal ulnar width (mean: 3.4cm in males, 2.9cm in females), distal ulnar width (mean: 2.2cm in males; 1.8cm in females), maximum radial length (mean: 24.6cm in males; 22.0cm in females), radial head diameter (mean: 2.6cm in males, 2.2cm in females) and distal radial width (mean: 3.6cm in males; 3.2cm in females). The differences between the means in males and females were significant (P<0.0005). A discriminant analysis was carried out with good results. A percentage of 94.93% of cases were correctly classified when all measures of the radius were applied jointly, followed by humerus (93.15%) and ulna (90.58%). Applied singly, the humeral head diameter allowed the best distinction (90.41% correctly grouped cases), followed by the radial length (89.13%), the radial head diameter (88.57%) and the humeral epicondylar width (88.49%). The linear regression analysis for quantifying the correlation between the bone lengths and the stature led to unsatifactory results with large 95%-confidence intervals for the coefficients and high standard errors of estimate.