Disproportionate Sexual Dimorphism in the Human Face W STUART HUNTER AND STANLEY M GARN T h e Department of Pnedia trzc Dentistry. tlze University of W e s t e r n Chitarto,London, Canada a n d Center for Htcrnan Groiutlz a n d Deuelopment, the Uritueiszty of Michigan, Ann Arbor, Michigan ABSTRACT Analysis of facial dimensions of 86 young adults and their 76 parents indicates that a disproportionate sexual dimorphism exists in the ramus of the mandible, demonstrating a regional difference in growth response. The male ramus is on the average 1 4 % longer than the female ramus, whereas other facial dimensions approximate an 8% sex difference. The findings have relevance to the analysis of skeletal remains and suggest the desirability of age specific discriminant function analysis for the sexing of adult mandibles. The sexing of human skulls has been a matter of interest for nearly 100 years as Giles and Elliot (‘63) have described. Since shape is obviously a n important factor in sex discrimination, the use of various cranial and facial dimensions for the sexing of crania has recently received much attention as a result of the development of discriminant function analysis. However, the general basis for discrimination for sex lies in the fact that male skulls are as a rule larger than female skulls. The purposes of this study were (1) to institute a pilot exploration of the nature of sexual dimorphisms i n facial dimensions of young adults and their parents, ( 2 ) to employ such dimorphisms as were found in the improvement of sex identification and (3) to study the effects of the dimorphisms on the shape of the face and on the vertical position of the dentition in the face insofar as these might relate to clinical orthodontic evaluations. METHODS AND MATERIALS The data consist of measurements taken from tracings of lateral head roentogenograms (all obtained under standard conditions) of 162 subjects from 38 nuclear families. I n these tracings, symmetry is assumed and the mid-point between left and right landmarks (Gonion, for example), was used. For a discussion of the technical details see Enlow (‘68) chapter 12. The average ages for the samples used are shown in table 1 . AM. J. PHYS. ANTHROP.,36: 13&138 The sample was gathered in Ann Arbor during 1966-67 and consists, with a few exceptions of established mid-western middle-class families of European origin. With two exceptions, all 162 subjects could establish occlusal contact with at least some posterior teeth of their own. The initial study utilized six measures of the periphery of the face and cranial base shown in figure 1 . Evaluation of these measures revealed disproportionate sexual dimorphism in the ramus and posterior cranial base. Therefore, a n additional analysis of the components contributing to the vertical dimensions of the posterior region of the face was undertaken to determine whether the mandible itself and the dentition are positioned differently according to sex. For this purpose, the PM vertical (Posterior Nasomaxilla) described by Enlow et al. (’69) was used as a reference line for three additional measures made parallel to the PM vertical as shown in figure 2. As a check on the apparently critical Articulare-Gonion measure, a n additional adjacent measure, Condylion-Gonion (parallel to the posterior border of the ramus) was also taken. All tracings and measures were repeated and discrepancies i n excess of 0.5 m m were corrected. Following the initial comparison of differences between means and the computation of sex differences as percentages of the female values, discriminate functions were obtained using a discriminant 133 134 W. STUART HUNTER AND STANLEY M. GARN TABLE 1 C o m p a r a t i v e a g e s of the p a r e n t s and t h e i r a d u l t oflspring N Parent Offspring 38 48 Parent Offspring 38 38 Mean Male 55.5 23.7 Female 52.8 23.8 \ S.D. Range 5.9 4.4 47-73 17-37 5.8 5.4 44-71 17-39 \ \ Neutral Occlusal Axis Fig. 2 Additional roentgenographic measurements obtained to examine the components of the posterior vertical dimensions of the face. The numbered measurements are: 7. CondylionGonion, 8. Sphenoethmoid-Condylion, 9. Condylion-Neutral Occlusal axis, 10. Neutral Occlusal Axis-Gonion. PTM is pterygomaxillary fissure. For further description of PM vertical, SE and NOA see text. Findings Fig. 1 The landmarks used for this roentgenographic study and the measures made between them. The numbers refer to measures as follows: 1. Articulare-Gonion, 2. Gonion-Menton, 3. Nasion-Menton, 4. Sella-Nasion, 5. Sella-Articulare, 6. Articulare-Gnathion. analysis program for two groups,' which permits the use of two or more variables. Although the computer analysis yields a cut off value for discrimination, i n some cases slight improvement could be obtained by empirically adjusting the cut off, which was done. Previous analysis of these data (Hunter and Garn, '69) had disclosed significant secular change in the face from parent to offspring generations. Hence, for the most part, the two generations were dealt with separately. As may be seen in table 2, analysis of the six measures shown in figure 1 disclosed that for ramus length (ArticulareGonion) the sex difference expressed as a percentage of the smaller (female) measure is in excess of 14% at both age levels. This is slightly higher than values which may be computed for Giles' ('64) data for a similar measurement. For four of the other five measures, the per cent differences range from 6.1% to 9.7% with a n average of 8.1 % . The remaining posterior vertical measure (Sella-Articulare) approximates a 13% difference. Although these measures are in no way exhaustive, they do sample the general structure of the face and suggest a more detailed examination of the posterior vertical dirnensions of the face. The analysis of some obvious cornponents of posterior vertical facial dimen1 Adapted by t h e University of Michigan Statistical Research Laboratory from UCLA program BMD04M. 135 DISPROPORTIONATE SEXUAL DIMORPHISM TABLE 2 Sex differences n n d percentnge sexztnl d i m o r p h i s m f o r S I X men n f n ci n l d i m e n s i o n s f o r p a l e i i t s tind f o r offspring Male Female Measurement N Parent Offspring 38 48 59.2 58.8 38 38 51.7 51.4 7.5 7.4 14.5 14.4 Parent Offspring 38 48 80.8 83.6 38 38 75.2 76.2 5.6 7.4 7.5 9.7 Parent Offspring 38 48 133.7 134.2 38 38 125.1 122.8 8.6 11.4 6.9 9.3 Parent Offspling 38 48 78.8 80.2 38 38 73.9 75.6 4.9 4.6 6.6 6.1 Parent Offspring 38 48 37.6 39.9 38 38 33.3 35.4 4.3 4.5 12.9 12.7 Parent Offspring 38 48 124.6 125.9 38 38 113.8 115.3 10.8 10.6 9.5 9.2 Mean N mm 1. Articulare- Gonion 2. Gonion- Menton 3. Nasion- Menton 4. Sella- Nasion 5. Sella- Articulare 6. Articulare- Gnathion Per cent 1 sexual dimorphism Group Mean Difference mm 1 Per cent sexual dimorphism x m -2f T X 100 xf TABLE 3 S e x dgferences a n d per c e n t sexual d i m o r p h i s m f o r f o u r posterior vertical height m e a n m e a s u r e m e n t s f o r parents a n d for offspring Male Female Measurement Group N Mean N mm 7. Condylion to Gonion 8. Spheno- ethmoid to Condylion 9. Condylion toNeutral Occlusal Axis 10. Neutral Occlusal Axis to Gonion 1 Mean Difference mm mm Per cent 1 sexual dimorphism Parent Offspring 38 48 70.8 71.3 38 38 62.0 62.9 8.8 8.4 14.2 13.4 Parent Offspring 38 48 31.7 32.0 38 38 30.5 30.7 1.2 1.3 3.9 4.2 Parent Offspring 38 48 44.2 43.5 38 38 40.3 38.5 3.9 5.0 9.7 13.0 Parent Offspring 38 48 25.0 26.2 38 38 20.5 22.6 4.5 3.6 22.0 15.9 Per cent sexual dimorphism xm-Ff %f x 100 sion (fig. 2) are shown in table 3 and indicate that (1) the disproportionate dimorphism is in the ramus rather than in its position relative to cranial base and (2) that although the dentition of males is positioned proportionately further from the cranial base than in females, it is not disproportionately so. The proportional excess for measures eight and nine combined (Sphenoethmoid to Neutral Occlusal Axis) is 7.2% for the parent sample and 9.1% for the adult offspring sample. Having established the magnitude and location of the sexual dimorphisms in these samples, attention was then given to the utilization of the dimorphisms in 136 W. STUART H U N T E R A N D STANLEY M. GARN TABLE 4 Results of various discriminant analyses for sex showing t h e best three combinations w h e n two, three, foiir,five, a i d szr variables are zised Per cent misclassified Variable Variable numbers names Parent sample 38 males 38 females Offspring sample 48 males 38 females Combined sample 86 males 76 females 9.2 13.2 10.5 11.6 8.1 11.6 13.0 11.1 14.2 Na-Me,S-Na,Ar-Gn S-Na,Ar-Gn,Co-Go S-Na,S-Ar,Ar-Gn 7.9 9.2 6.6 11.6 10.5 10.5 12.3 10.5 15.4 Go-Me,S-Na,S-Ar, Ar-Gn Go-Me,Na-Me,S-Na, CO-GO S-Na,S-Ar,Ar-Gn, CO-GO 6.6 8.1 15.4 7.9 7.3 9.9 6.6 8.1 11.7 Go-Me,Na-Me,S-Na, S-Ar,Co-Go Na-Me,S-Na,S-Ar Ar-Gn,Co-Go Go-Me,S-Na,S-Ar, Ar-Gn,Co-Go. 7.9 5.8 8.6 7.9 8.1 9.3 6.6 8.1 12.3 7.9 5.8 10.5 S-Na,Ar-Gn 1 Ar-Gn,Co-Go S-Ar,Ar-Gn All six variables 1 See text for description of abbreviations. the sexing of crania and mandibles, using discriminant analysis. From the ten measures obtained i n this investigation, six were selected for examination on the basis of their potential utility either for standardized roentgenographic or for craniometric studies. Thus, NasionMenton, Gonion-Menton, and Condylion-Gonion may be obtained either craniometric ally or roentgenographic ally. Sella-Nasion, Articulare-Gnathion (approximately equivalent to CondylionGnathion) and Sella-Articulare are essentially roentgenographic. All possible combinations of the above six measures were tested and the combinations of greatest discriminatory effectiveness when two, three, four and five variables are used were selected for inclusion in table 4. When three variables are used instead of two, the percentage misclassified as to sex is reduced from 9.2% for the best combination of two variables to 6.6% for the best combination of three variables for the parent sample. On the other hand, the use of three variables rather than two provides no improvement for the offspring sample. In general however, some reduction in misclassification is accomplished by using more variables, so that when all six are used, 7.9% of the parent sample is misclassified and only 5.8% of the offspring sample is misclassified. The sex differences for the parent generation and for the offspring generation were compared and found not statistically different. However, pooling of the data did not result in improved discriminatory effectiveness as may be seen i n the last column of table 4. In general, the results of the discriminant analyses for a contemporary population using a radiographic procedure compare favorably with those reported by Giles ('64) for the mandible alone and by Giles and Elliot ('63) and Kajanoja ('66) for crania using direct measurements on cadavers i n which misclassification accounts for from 13% to 20% of the samples. DISPROPORTIONATE SEXUAL DIMORPHISM DISCUSSION The amount of facial sexual dimorphism for the sample of adults studied appears to be on the order of 8% as i t is in fact for stature in this sample (8.3% for parents and 8.6% for adult offspring), except for measures associated with ramus height for which it is on the order of 14% . Disproportionate sexual dimorphism in the adult face then, is localized in the ramus of the mandible and seems to be especially large in the gonial area. The dimorphisms for two posterior face height measures (Articulare-Gonion and Sella-Articulare) are disproportionately larger in both age groups than the other perimeter measures, with percentage differences from 12.7% to 14.5%. The breakdown of the components of the posterior vertical dimensions (table 3) indicates that the disproportionate dimorphism lies in the ramus and not in the positioning of the mandible relative to the cranial base. With respect to sexual discrimination, either (or both) of the measures Articulare-Gnathion and CondylionGonion appears in all of the combinations shown in table 4 in which the most effective discriminant combinations are summarized. The large ramal dimorphism then proves to be both directly and indirectly useful in sex discrimination. It should be noted that the greater sexual dimorphism for the dimension in question exists both in the young adult group and in the parental group. Further, the fact that the sex differences for the ; 15.- - measures from Neutral Occlusal Axis to Gonion (fig. 2, table 3) are both absolutely and proportionately greater for the parent sample than for the offspring sample suggests that the localized dimorphism increases with age disproportionately in the gonial area. Hence, continuing mandibular bone growth during adulthood would appear to be selective with the gonial area the target site. Pooling the generations results in lessening of the discriminatory power as may be seen for the combined sample in figure 3 in which each bar represents the average percentage of subjects misclassified by the three best discriminant analyses as shown in table 4. Separation of the sample by generation reduces the classification errors on the average by approximately 3 % . An additional 3% reduction in misclassification is obtained by using four or more variables. Thus, separation of samples by age would a p pear to be as useful as the taking of additional measures in the sexing of unknown mandibles. It is of interest that the use of two mandibular measures, Articulare-Gnathion and Condylion-Gonion (or Articulare-Gonion) as shown in the second line of table 4, provides a rather good level of discrimination, especially for the parent sample of this study, thus providing a n effective means for sexing isolated mandibles. It should also be pointed out that quite good levels of sexual discrimination may be obtained for the human facial skeleton - +.- Y, -U lrl E 2 al 137 5 U L Number of variables used Fig. 3 The reduction in percentage of subjects misclassified by a discriminant analysis for sex when 2,3,4,5and 6 variables are used and when the sample is separated by age. For details see table 4. 138 W. STUART HUNTER AND STANLEY M. GARN generally, through concomitant use of tion of samples by age improves the ef(1) a measure of general facial size show- fectiveness of the discriminant function ing the usual amount of sexual dimor- and should be undertaken as the first phism such as Nasion-Menton or Articu- step in the sorting of commingled remains. lare-Gnathion and (2) either of the two Subsequent analysis might well follow the measures showing disproportionate sexual protocols outlined by Giles ('70b). This study then, points to the extent dimorphism - Sella-Articulare or Condylion-Gonion. Similarily , Giles ('70a) has of localized difference in the magnitude observed that for crania alone, the mea- of sexual dimorphism; it indicates that sure of maximum bi-zygomatic width pro- the relative magnitude of sexual dimorduces almost as good results (81.9% phism in the face continues to increase correct) as utilization of eight measures with age; it points to the discriminatory effectiveness of certain mandibular mea(83.3% to 86% correct). Disproportionate sexual dimorphism in sures and finally, to the extent of improvethe face is postpubertal as pointed out by ment in sex discrimination even i n adulthood when age is taken into account. Krogman ('62). A study by Boerman ('67) using mixed longitudinal-cross-sectional ACKNOWLEDGMENTS data from 5 to 17 years of age did not This study was supported in part by show disproportionate sexual dimorphism in the ramus by age 17. Since it clearly United States Public Health Service, grant exists in the data of the present study by HD02272 and by Medical Research Counthe beginning of the third decade of life, cil of Canada, grant MA 3547. The auit appears to be distinctly a feature of thors are indebted to the families who so adult growth, which in addition continues generously contributed their time to proto be enhanced at least into the sixth vide data for this study and to Richard L. decade of life. This offers at least a par- Miller for data reduction. tial explanation for Giles' ('7Oa) observaLITERATURE CITED tion that i t is the younger males and older females who tend to be inisclassified Birkby, W. H. 1966 A n evaluation of race and sex identification from cranial measurements, by discriminant analysis. Am. J. Phys. Anthrop., 24: 21-28. Consideration of shape differences Boerman, I. E. 1967 Vertical facial developalone reveals that the dimorphism under ment from ages five to seventeen. M.S. Thesis University of Michigan. discussion, while disproportionate for ramus height, is not manifested in abso- Enlow, D. H. 1968 The Human Face. Hoeber, New York. lute or metric size differences i n that di- Enlow, D. H., R. E. Moyers. W. S. Hunter and mension any more than it is in the manJ. A. McNamara 1969 A procedure for the analysis of intrinsic facial form and growth, dibular length dimension (see table 2). Am. J. Orthodont., 56: 6 2 3 . The use of percentages is therefore helpGiles, E. 1964 Sex determination by discrimful in clarifying relative amounts of inant function analysis of the mandible. Am. J. dimorphism. Phys. Anthrop., 21: 129-136. 1970a Sexing crania by discriminant Finally, the sexing of unidentified skelfunction analysis: effects of age and number of etal material using the sexual dimorphism variables. Proceedings, VIIIth International discussed here, assumes that a reasonable Congress of Anthropological and Ethnological number of subjects is available for study Sciences, Tokyo, 1 : 59-61. 1970b Discriminant function sexing of and that the sample includes approxithe h u m a n skeleton. In: Personal Identification mately equal numbers of males and i n Mass Disasters. T. D. Stewart, ed. National females. The data presented here are Museum of Natural History. Smithsonian In.from a contemporary population. Secular stitute, Washington. pp. 9 S 1 0 9 . change (Hunter and Garn '69) as well as Giles, E., and 0. Elliot 1963 Sex determination by discriminant function analysis of crania. Am. racial differences, probably preclude the J. Phys. Anthrop., 2 1 ; 53-68. use of the figures presented in tables 2 Hunter, W. S., and S. M. Garn 1969 Evidence and 3 for the sexing of earlier and rafor a secular trend i n face size. Angle Orthocially different populations, although dont., 39: 320-323. Birkby ('66) concluded that sex discrim- Kajanoja, P. 1966 Sex determination of Finnish crania by discriminate function analysis. ination in unknown populations can Am. J. Phys. Anthrop., 24: 2 S 3 3 . achieve levels of accuracy comparable to Krogman, W. M. 1962 The Human Skeleton in those found here. However, the separaForensic Medicine. C. C Thomas, Springfield.