Determinations of sex by subjective assessment of the skulls from a skeletal series of known sex were compared to fully independent assessments based on pelves of the same specimens. Within-sex correlations of cranial and pelvic morphologies measured on an android-gynecoid scale were smaller than expected. Subjective assessment by means of the skull compared favorably to that of the linear discriminant functions of Giles and Elliot; however, the direction of error was similar for both procedures. Of course, estimations based on the pelves were generally superior to both in terms of frequency and overall bias of error. The bias of sex estimation for palaeodemographic purposes is contingent upon completeness of skeletal remains.