AMERICAN JOURNAL OF PHYSICAL ANTHROPOLOGY 62205-208 (1983) Assessment of Race From the Pelvis M. YASAR ISCAN Department of Anthropology, Florida Atlantic Uniuersity, Boca Raton, Florida 33431 KEY WORDS Racial assessment, Discriminant function analysis, Pelves, American whites, American blacks ABSTRACT Racial assessment of human postcranial skeletal remains has been a major concern for forensic and skeletal anthropologists. Materials (N = 400) of the present study are from the Terry Collection and consist of 100 black and white American pelves of both sex with known age and race. Measurements were taken from the articulated pelves. Results of discriminant function analysis indicate classificatory accuracy may be a s high as 88%. Transverse pelvic breadth contributes more to the function than biiliac breadth and antero-posterior height. The females are more easily assessed racially than males. Although a highly reliable classification is produced, the results of the study should be employed with caution, a s samples were of questionable nutritional status and of low socioeconomic class. The femur is the most studied bone of the postcranial skeleton in the assessment of the biological affinity of a n individual (Stewart, 1962; Gilbert, 1976; Walensky, 1965).Obstetrically oriented radiographical investigations have shown that there are metrical differences, especially in the dimensions of the pelvic inlet, among human populations (Scheyer, 1934; Aiman, 1976; Torpin, 1951). However, a study by Todd (1929) on a cadavera1 population pointed out that the differences between American blacks and whites are small, admitting that this part of the skeleton has not been "entrenched" adequately. A later comparative osteological work reemphasized that pelves of the white population are larger than those of American Indians and blacks (Howells and Hotelling, 1936). The human pelvis has played a minor role in assessing the biological affiliation of skeletal remains found in archaeological and forensic sites (Bass, 1969,1979; Kerley, 1978; Krogman, 1962; Stewart, 1979). The present paper is a n attempt to develop a simple statistical technique to determine the racial difference between American blacks and whites, using pelvic measurements. MATERIALS AND METHODS Four hundred pelves (100 each sex-race group) of American blacks and whites were c,( 1983 ALAN R LISS. INC utilized as the main sample in this study. The age, sex, and race of all specimens in the sample were known. They are a part of the Terry Collection at the Smithsonian Institution of American skeletal material from the 19th and early 20th centuries. Before any measurements were made, the pelves were articulated with several rubber bands while leaving a small gap between the pubic bones, a procedure used by Howells and Hotelling (1936). Three measurements were selected by a stepwise discriminant function procedure from the original 12 dimensions and then incorporated into the analyses. The biiliac breadth was measured by a n osteometric board as the maximum distance between the iliac crests. The antere posterior height, also known as conjugate diameter, was taken from the sacral promontory to the pubic crests of both sides and the average of the two was recorded. The transverse breadth was obtained as the maximum distance between the arcuate lines of the pelvic inlet. A sliding caliper was used for these two dimensions. The main sample of the study was randomly divided into base (N = 75 for each sex and race) and test (N = 25 for each sex and race) samples. The base sample was used to Received September 28, 1981;accepted May 26, 1983 206 M.Y. CAN compute both descriptive and t-test statistics. The base samples of the same sex of both groups were utilized to develop several canonical discriminant function statistical models. The statistical analyses were carried out according to the latest version (version 8) of the Statistical Package for the Social Sciences subroutines (Nie et al., 1975; Hull and Nie, 1979). Discriminant scores of the test samples were calculated according to the unstandardized canonical discriminant function coefficients obtained from the base sample. In all these calculations the males of both races were treated separately from the females. blacks. The relative accuracy of a correct classification is shown in Table 3. In the base sample, the highest classification accuracy is 88%. In the function where age is not included this rate drops to 79% in males and 83% in females. Overall, females seem to show better predictive results than males. Classification results of the test sample correspond well with those of the base sample (Table 3). However, males of the test sample are predicted slightly higher than the females. DISCUSSION Race determination from the cranium has provided successful results. Howells (1970) RESULTS was able to classify a skull into a race with a As Table 1 illustrates, it is evident that in 90% accuracy. Giles and Elliot (1962) showed all variables whites of both sexes exceeded a slightly lower percentage in their analysis blacks of the corresponding sex, with a statis- of American blacks and whites. The present tically significant difference between the study reached a n accuracy of 83% in males means. The F-ratio statistic in the same table and 88% in females. Such a result, although signifies that in all dimensions the sample less successful than those based on cranial variances are the same. The age variance of measurements, is promising for various reasons. First, the pelvic area has not been used sample females is, however, heterogenous. The results of the canonical discriminant in race determination; second, the number of coefficients are presented in Table 2. The measurements involved were minimal; and standardized coefficient represents the rela- third, these measurements could be taken tive contribution of a variable to the func- with minimal interobserver errors. This study also showed that, of the three tion. It is clear that in multivariable functions the antero-posterior is the factor dimensions, transverse and biiliac breadths that contributes least while the transverse seem to be homogeneously distributed within breadth is the most powerful of the four var- a race. However, the antero-posterior height iables. The influence of age is undeniably is more varied. In previous studies such obhigh. The unstandardized coefficient is the servations were also noted. Research in obweighting coefficient to be used in calculat- stetrical disciplines has indicated that the ing the discriminant score for a given pelvis. antero-posterior height is a sensitive dimenThe positive values, i.e., larger than zero, sion of the human skeletal architecture and signify a n association with the white sample responds to environmental and socioeconompopulation. The reverse score is for the ical factors of the population. A major reTABLE I . Means, standard deuiations, and F and t statistics IN Variables' Male Age (in years) Biiliac breadth Transverse breadth Antero-posterior height Female Age (in years) Biiliac breadth Transverse breadth Antero-posterior height White Mean SD Black Mean SD 58.52 273.96 123.59 108.73 13.07 16.10 7.95 9.62 49.36 254.60 111.96 102.01 65.24 277.99 133.93 116.64 15.94 17.11 8.09 10.51 47.88 252.81 120.56 110.75 'All dimensions in millimeters. 'All t's in both races significant a t P < 0.001 ,'Significant at P < 0.026 (two-tailed test). = 75 each sex-race) F tL 15.40 15.27 7.39 8.54 1.39 1.11 1.16 1.27 3.93 7.55 9.28 4.53 20.70 15.95 6.82 9.26 1.6g3 1.15 1.40 1.29 5.75 9.32 10.95 3.64 207 ASSESSMENT OF RACE FROM THE PELVIS T A B L E 2. Canonical discriminant function coefficients Functions and variables Standardized coefficients 1)Biiliac breadth 1.000 Constant 2) Transverse breadth Constant 3) Biiliac breadth Age Constant 4) Transverse breadth Age Constant 5) Biiliac breadth Transverse breadth Constant 6) Biiliac breadth Transverse breadth Age Constant 7) Transverse breadth Antero-posterior height Age Constant 8) Biiliac breadth Transverse breadth Antero-posterior height Constant 9) Biiliac breadth Transverse breadth Antero-posterior height Age Constant Males Unstandardized' coefficients Females Unstandardized' coefficients .ooo 0.0637219 0.0604552 -16,04480 0.1336859 - 17.01109 0.0510116 0.0196656 - 14.65077 0.1181089 0.0244722 -16.41311 0.0175062 0.1036824 - 17.83939 0.0033984 0.1127945 0.0237207 - 16.59630 0.1 116498 0.0172727 0.0252114 - 17.59680 0.0169375 0.1004562 0.0117874 - 18.61806 0.0019737 0.1086451 0.0170556 0.0247657 -17.68839 1 - 16.84043 1.000 Standardized coefficients 0.1303004 1.000 - 15.34591 0.891 0.293 0.0567736 0.0205256 - 16.11140 0.1199950 0.0273070 -- 15.60514 0.0179339 0.1035484 -16.93482 0.921 0.390 0.281 0.795 0.844 0.363 0.883 0.452 0.290 0.776 0.122 0.839 0.361 0.056 0.844 0.438 - 16.27929 0.1003125 0.0425694 0.0337675 -18.12124 0.0147927 0.0975458 0.0253182 - 18.06558 - 0.00107 13 0.1016162 0.0429505 0.0341109 - 18.05033 0.770 0.387 0.483 0.232 0.745 0.230 -0.017 0.780 0.391 0.487 0.835 0.171 0.466 0.280 0.751 0.117 0.032 0.813 0.169 0.458 'Discriminant score less t h a n 0 would classify as black. T A B L E 3. Correct prediction percentage of base and test samples Function White Base Black 76.0 74.7 80.0 76.0 73.3 80.0 81.3 74.7 82.7 74.7 77.3 78.7 85.3 84.0 85.3 82.7 80.0 82.7 76.0 78.7 81.3 88.0 80.0 86.7 88.0 80.0 88.0 78.7 88.0 76.0 88.0 85.3 86.7 88.0 86.7 88.0 White Test Black 75.3 76.0 79.3 80.7 78.7 82.7 82.0 77.3 82.7 88.0 80.0 84.0 88.0 80.0 92.0 88.0 88.0 88.0 76.0 84.0 76.0 84.0 80.0 92.0 88.0 88.0 88.0 82.0 82.0 80.0 86.0 80.0 92.0 88.0 88.0 88.0 77.3 83.3 78.7 88.0 82.7 86.7 88.0 83.3 88.0 72.0 60.0 84.0 84.0 72.0 84.0 84.0 76.0 80.0 76.0 88.0 88.0 88.0 84.0 84.0 88.0 88.0 88.0 74.0 74.0 86.0 86.0 78.0 84.0 86.0 82.0 84.0 Average Average Males 1) 2) 3) 4) 5) 6) 7) 8) 9) Females 1) 2) 3) 4) 5) 6) 7) 8) 9) M.Y. ISCAN 208 sponse of this diameter is a reduction in height in individuals living under nutritionally and socioeconomical disadvantaged conditions (Nicholson, l945; Caldwell et al., 1934; Thoms, 1936). It has also been implied that such a decrease in antero-posterior height is due to reorganization of the sacroiliac joint resulting from the weight and child bearing functions of the pelvic girdle (Derry, 1923; Caldwell et al., 1934; Igcan, 1980). From this it can be concluded that transverse and biiliac diameters should be of value for determining race differences between populations since environmental impact is relatively less on these variables than on the antero-posterior diameter. There are the usual limitations in using discriminant function studies. These limitations include the representativeness of the sample, so that the result can be used on a racially similar population. This issue was raised by Birkby (1966) in regard to the study by Giles and Elliot (1962). A second issue is the change in a population through time. Angel (1976) pointed out that there is a secular change in pelvic dimensions since human prehistory. The occurrence of such a change is also supported by Igcan (1980) in his earlier study of the Terry Collection in which it was shown that, among other measurements, antero-posterior height in recent populations has become larger. One should then be cautious in applying the result of the present study to a prehistoric or a more recent population. The final issue deserving mention here is that the age variable plays an important role in the discriminant function, and exact age may be the most difficult factor to determine. In conclusion, this study is a n attempt to fulfill a need (Stewart, 1973) that is necessary for determining racial affinity from the postcranial skeletal remains. It should serve skeletal and forensic anthropologists in their analysis of remains found under commingled conditions as well as in a situation where cranially determined race identification may require support from the postcranial remains. Modern samples, however, should be tested before reaching a statistical decision based on the results of this study. ACKNOWLEDGMENTS This research was supported in part by the Smithsonian Institution (Ales HrdliEka Fund). The author is grateful to Dr. Walda Igcan for commenting on the manuscript. Many thanks to Mrs. Marjorie Wolf for her excellent typing. LITERATURE CITED Aiman, J (1976)X-ray pelvimetry of the pregnant adolescent: Pelvic size and the frequency of contraction. Am J. Obstet. Gynecol. 48t281-286. Angel, JL (1976) Colonial to modern skeletal change in the U.S.A. Am. J. Phys. Anthropol. 45r723-736. Bass, WM (1969) Recent developments in the identification of human skeletal material. Am. J. Phys. Anthropol. 30t459-461. Bass, WM (1979) Developments in the identification of human skeletal material (1968-1978). Am. J. Phys. Anthropol. 51:555-562. Birkby, WH (1966) An evaluation of race and sex identification from cranial measurements. Am. J. Phys. Anthropol. 24:21-27. Caldwell, WE, Moloy, HC, and D’Esopo, DA (1934) Further studies on the pelvic architecture. Am. J. Obstet. Gynecol. 28t482-497. Derry, DE (1923) On the sexual and racial characters of the human ilium. J. Anat. 58t71-83. Gilbert, MB (1976) Anterior femoral curvature: Its probable basis and utility as a criterion of racial assessment. Am. J. Phys. Anthropol. 45t601-604. Giles, E, and Elliot, 0 (1962) Negro-White identification from the skull, VIe Congres International des Sciences Anthropologiques e t Ethnologiques, Paris. 1t179-184. Howells, WW (1970) Multivariate analysis for the identification of race from crania. In TD Stewart (ed): Personal identification in mass disasters. Washington, D.C.: National Museum of Natural History, pp. 111121. Howells, WM, and Hotelling, H (1936) Measurements and correlations on pelves of Indians of the southwest. Am. J. Phys. Anthropol. 21t91-106. Hull, CH, and Nie, NH (1979) SPSS Update. New York: McGraw-Hill. iacan, MY (1980) Effect of age on pelvic dimensions. Florida Scientists 43(Suppl. 1):48(Abstr.). Kerley, ER (1978) Recent developments in forensic anthropology. Yrbk. Phys. Anthropol. 21t160-173. Krogman, WM (1962) The human skeleton in forensic medicine. Springfield, Illinois: Thomas. Nicholson, G (1945) The two main diameters at the brim of the female pelvis. J. Anat. 79(Part 3):131-135. Nie, NH, Hull, C H , Jenkins, JG, Steinbrenner, K, and Bent, DH (1975) SPSS. New York: McGraw-Hill. Scheyer, H E (1934) Anthropological and roentgenocological observations on the pelves of Chinese women. Chin. Med. J. 48t1228-1234. Stewart, TD (1962) Anterior femoral curvature: Its utility for race identification. Hum. Biol. 34:49-62. Stewart, TD (1973) Recent improvements in estimating stature, sex, age and race from skeletal remains. In AK Mant (ed): Modern Trends in Forensic Medicine, Vol. 3. London: Butterworth, pp. 193-211. Stewart, TD (1979) Essentials of forensic anthropology: Especially as developed in the United States. Springfield, Illinois: Thomas. Thoms, H (1936) Is the oval or female type pelvis a rachitic manifestation? Am. J. Obstet. Gynecol. 31t111115. Todd, TW (1929) Entrenched Negro physical features. Hum. Biol. 1t57-69. Torpin, TW (1951) Roentgenpelvimetric measurements of 3604 female pelves, White, Negro and Mexican compared with direct measurements of Todd anatomy collections. Am. J. Abstet. Gynecol. 62279-293. Walensky, NA (1965) A study of anterior femoral curvature in man. Anat. Rec. 151559-570.