This article presents an analysis of intra- and inter-observer errors in pelvic measurements commonly used for sex diagnosis, as well as their impact on the results of discriminant unction analyses. This allows for an assessment of which variables are the most reliable. Moreover, it is suggested that the indirect technique of measuring of the sciatic notch (shadow image) be replaced by direct measurement (trigonometry). The concordance of sex diagnosis through discriminant functions, that is, the concordance between two observers, two sets of measurements and two techniques, ranges from 88 to 100 %. This variation is due to the proximity, for some bones, of the discriminant function analysis (DFA) result to the discriminant value itself.