A Method for Visual Determination of Sex, Using the Human Hip Bone Jaroslav Bruzek* U.M.R. 5809 du C.N.R.S., Laboratoire d’Anthropologie des Populations du Passe ´ Universite ´ Bordeaux I, 33405 Talence, France KEY WORDS human pelvis; sex determination; morphological traits; method ABSTRACT A new visual method for the determina- tion of sex using the human hip bone (os coxae) is pro- posed, based on a revision of several previous approaches which scored isolated characters of this bone. The efficacy of the methodology is tested on a sample of 402 adults of known sex and age of French and Portuguese origins. With the simultaneous use of five characters of the hip bone, it is possible to provide a correct sexual diagnosis in 95% of all cases, with an error of 2% and an inability to identify sex in only 3%. The advantage of this new method is a reduction in observer subjectivity, since the evalua- tion procedure cannot involve any anticipation of the re- sult. In addition, this method of sex determination in- creases the probability of a correct diagnosis with isolated fragments of the hip bone, provided that a combination of elements of one character is found to be typically male or female. Am J Phys Anthropol 117:157–168, 2002. © 2002 Wiley-Liss, Inc. Correct sex identification of the human skeleton is important in bioarcheological and forensic practice. Current opinion regards the hip bone (os coxae) as providing the highest accuracy levels for sex deter- mination. However, “simple” observations of the hip bone without any scoring of related traits should not be normally considered proper, despite the fact that the results may be surprisingly accurate. Only those procedures which define a precise and unambiguous methodology leading to an accurate diagnosis of sex should be employed. Three techniques for the visual evaluation of traits of the hip bone are: 1) the method of Phenice (1969), which uses three traits on the pubis, 2) the method of Iscan and Derrick (1984) using the poste- rior pelvis, and 3) the method of Ferembach et al. (1980) of sexing the entire pelvis through an evalu- ation of eleven traits. However, even if it is generally accepted that these methods provide satisfactory ac- curacy, only a few studies have tested their reliabil- ity in known-sex samples and, surprisingly, the re- sults are often ignored. The results from the method of Phenice (1969) have been inconsistent, as accuracy levels range from 59% (MacLaughlin and Bruce, 1990) to 96% (Schone, quoted in Sutherland and Suchey, 1991). Such inconsistency is due to the fact that sexual dimorphism of the whole hip bone should be consid- ered, and observations should not be restricted to the pubis (Novotny ´ , 1986; Bruzek, 1992). Following Lovell (1989), MacLaughlin and Bruce (1990), and Bruzek (1991), the reliability of the method of Phenice (1969) is probably ca. 80%. Furthermore, pubic preservation rarely exceeds 30% in archaeo- logical samples (Waldron, 1987). The method proposed by Iscan and Derrick (1984) provides an accuracy level of 90% (Iscan and Dun- lap, 1983), but it cannot be regarded as equivalent to the results found with methods using the entire hip bone. The accuracy of the method of Ferembach et al. (1980) has not been studied. Bruzek and Ferem- bach (1992) found 93% correct sex assignment, using a set of eight variables of the hip bone. However, this method necessitates highly trained observers with experience in morphological variability, because it is based on an ordinal scale of evaluation. When (Novotny ´ , 1981, 1988) studied the discrim- ination power of 14 sexually dimorphic traits of the hip bone frequently used in sex determination, he replaced a descriptive or ordinal evaluation of char- acters (e.g., small, wide, shallow) with binary scor- ing (yes or no) along an intermediate category mor- phology. He then selected the characters with the highest diagnostic value, reducing the analysis to three complex variables: the preauricular surface, the sciatic notch, and the inferior aspect of the hip bone. These characters reduce misclassification (Bruzek, 1991). In addition, Rogers and Saunders (1994) tested accuracy and reliability in a set of 17 individual traits on a small sample of identified pel- ves. Accuracy varied from 6.1–94.1% for each indi- vidual pelvic trait, and the greatest accuracy was obtained by combining the scoring of three traits. *Correspondence to: Jaroslav Bruzek, Laboratoire d’Anthropologie des Populations du Passe ´, UMR 5809 du CNRS, Universite ´ Bordeaux I, Avenue des Faculte ´s, 33 405 Talence, France. E-mail j.bruzek@anthropologie-u.bordeaux.fr Received 2 September 1997; accepted 21 August 2001. AMERICAN JOURNAL OF PHYSICAL ANTHROPOLOGY 117:157–168 (2002) © 2002 WILEY-LISS, INC. DOI 10.1002/ajpa.10012