AMERICAN JOURNAL OF PHYSICAL ANTHROPOLOGY 143:279–284 (2010) Determination of Sex From the Hyoid Bone Sarah C. Kindschuh,1 Tosha L. Dupras,2* and Libby W. Cowgill2 1 2 Department of Anthropology, Binghamton University, State University of New York, Binghamton, NY 13902-6000 Department of Anthropology, University of Central Florida, Orlando, FL 32816 KEY WORDS sexual dimorphism; discriminate function analysis; forensic anthropology ABSTRACT This article explores size differences related to sex in the hyoid bones from the Robert J. Terry Anatomical Collection. A series of measurements were taken from 398 hyoids, both fused and unfused. The inclusion of unfused hyoids in the study provides the opportunity to investigate previously unknown size differences between sexes as well as to determine their utility in determining sex. Two-way ANOVA was used to explore differences in hyoid size as related to ancestry and sex. Discriminant function analysis was employed to test the ability of the hyoids to be classiﬁed by sex. Six One of the most common uses of the hyoid in a forensic context is as an indicator of traumatic strangulation. Previous research has shown, however, that the fused hyoid is in fact a sexually dimorphic bone and therefore has the ability to aid the physical anthropologist in the determination of sex (Jelisiejew et al., 1968; Miller et al., 1998; Reesink et al., 1999; Kim et al., 2006). Male hyoids are generally larger than female hyoids in almost all dimensions, but particularly in total hyoid length and width (Jelisiejew et al., 1968; Miller et al., 1998). The hyoid body has also been found to be a sexually dimorphic component of the bone (Reesink et al., 1999), suggesting it may also be useful in the determination of sex. Three previous studies have used discriminant function analysis with varying degrees of success to develop a function that could be used to classify a hyoid as male or female (Miller et al., 1998; Reesink et al., 1999; Kim et al., 2006). Reesink et al. (1999) used three dimensions of the body to develop an equation that classiﬁed their sample with an overall accuracy of 76%, while Miller et al. (1998) obtained similar results with a function using ﬁve measurements, achieving an overall accuracy of approximately 72%. Most recently, Kim et al. (2006) used three measurements to produce a function that resulted in an overall accuracy of 88.2%. Unfortunately, the three previously mentioned studies, as well as the majority of previous hyoid research, focused solely on hyoids that had been recovered at the time of autopsy and whose preparation included leaving enough soft tissue in place between the hyoid body and greater cornua to maintain the overall shape of the bone. However, if a hyoid is found in an archaeological or forensic setting, unless the greater cornua are actually fused to the body, the likelihood of ﬁnding a intact hyoid is signiﬁcantly decreased due to tissue decomposition. In fact, of the 398 hyoids used in this study, all derived from the Robert J. Terry Anatomical Collection, 229 were unfused with either one or neither of the greater cornua fused to the body. On the basis of this random sample of the Terry Collection, the likeliC 2010 V WILEY-LISS, INC. discriminant function equations ranging in accuracy from 82% to 85% are provided, each of which is more accurate than many of the discriminant functions developed in past hyoid research, are simple to use, and can be used to estimate the sex of a hyoid regardless of its state of fusion. In addition to providing further information about the morphological form of the hyoid, these analyses provide a method that can be easily employed to assess sex of the individual from the hyoid bone. Am V 2010 Wiley-Liss, J Phys Anthropol 143:279–284, 2010. C Inc. hood that an unfused skeletonized hyoid will be recovered is greater than 50%, demonstrating the need for methods of sexing unfused as well as fused hyoids. The anatomical reasoning behind why some greater cornua fuse to the body and others do not is still somewhat unclear: while some studies have suggested that the incidence of fusion increases with age (O’Halloran and Lundy, 1987), many hyoids remain unfused through old age, and there is no signiﬁcant correlation between sex and bilateral fusion of the greater cornua to the body (Miller et al., 1998). The goals of this research are twofold. First, while trends in size sexual dimorphism have been documented in fused hyoids, it remains unclear if similar patterns of sex differences exist in the unfused elements. This study addresses this issue directly, by exploring patterns of hyoid sexual dimorphism in both fused and unfused hyoids. Second, given the high frequency with which unfused hyoid elements are found in skeletal samples, an equation for predicting sex from unfused hyoids may prove useful for bioarchaeological and forensic researchers. This research presents several discriminant function equations which are easy to use, practical, and can be implemented regardless of the state of fusion of the hyoid. MATERIALS AND METHODS Materials All hyoids utilized in this study are part of the Robert J. Terry Anatomical Collection, housed at the *Correspondence to: Dr. Tosha Dupras, Department of Anthropology, University of Central Florida, Orlando, FL 32816. E-mail: firstname.lastname@example.org Received 25 July 2009; accepted 8 March 2010 DOI 10.1002/ajpa.21315 Published online 27 May 2010 in Wiley Online Library (wileyonlinelibrary.com). 280 S.C. KINDSCHUH ET AL. TABLE 1. Breakdown of sample population Sex Condition Female Unfused Fused Number 103 95 Male Unfused Fused 126 74 Total 198 200 TABLE 2. Fused and unfused hyoid dimensions (also see Fig. 1) Dimension BL BH CWI CHI CL CWS CHS THL THW WCS Description Maximum length of body Maximum height of body Width of greater cornu at fusion point with body (inferior end) Height of greater cornu at fusion point with body (inferior end) Maximum length of greater cornu Greatest width of superior end of greater cornu Greatest height of superior end of greater cornu Total hyoid length – maximum length of hyoid, from anterior surface of the body to the superior ends of greater cornua (fused hyoids only) Total hyoid width – maximum distance between the widest points of the greater cornua (not pictured; fused hyoids only) Total width between the superior ends of the right and left greater cornua-taken from lateral edges (fused hyoids only) Smithsonian Institute’s National Museum of Natural History in Washington, D.C. For further information regarding the history and composition of the Terry Collection, refer to Hunt and Albanese (2005). A total of 398 hyoids, 200 males and 198 females, were selected for measurement (Table 1). Of these, 169 hyoids were fully fused with both greater cornua fused to the hyoid body, and 229 were unfused in which only one or neither of the greater cornua was fused to the body. In an attempt to develop a sampling population that was as uniform as possible, the collection was divided by age, sex, and ‘‘race.’’ Individuals between the ages of 20 and 79 were broken into groups delineated by 10 year increments, and within these age groups an attempt was made to obtain 20 hyoids from both black and white males and females. Unfortunately, some groups within the Terry Collection are underrepresented, namely white females below the age of 40, so that less than 20 hyoids were recovered from individuals within those ranges. Two sets of measurements were developed and applied depending on the state of fusion of each hyoid. Each measurement was chosen due to its use in previous hyoid research (Jelisiejew et al., 1968; Miller et al., 1998; Reesink et al., 1999; Lekı̂san et al., 2005; Shimizu et al., 2005; Kim et al., 2006). These measurements were also selected for their ability to provide an overall assessment of the size of each hyoid component. These measurements, outlined in Table 2 and demonstrated in Figure 1a,b, were taken to the nearest hundredth of a millimeter using standard digital calipers. To compensate for potential preservation biases and to maximize the amount of data available for analysis, both the right and left sides of each hyoid were measured. Although some fused hyoids did exhibit very faint lines at the point of fusion between the body and greater cornua, the majority of lines were either partially open or clearly American Journal of Physical Anthropology discernible. Hyoids in which no fusion point between the body and greater cornua could be detected were not included in the study. In addition, some hyoids displayed sporadic bony growths or spicules which could potentially interfere with the measurement being taken. When present, these growths were excluded from the measurement. In addition, a random subsample of hyoids was selected and remeasured in order to quantify intraobserver measurement error. Percent measurement error for 20 fused and 20 unfused hyoids (equal sex representation in each group) is shown in Table 3. For those measurements which were taken of both the right and left sides, the measurements of both sides were pooled as opposed to calculating the percent error of each individual side. The average amount of error for each measurement was below 5%, suggesting the measurements can be replicated with a low amount of error. Statistical methods A One-Sample Kolmogorov-Smirnov Test conﬁrmed that the data are distributed normally and therefore parametric statistical methods were used in this analysis. Box’s M tests were used to test for an equality of covariance among the samples used to develop the discriminant functions. Because Box’s M tests are extremely sensitive to violations of normality, an alpha level of 0.001 is frequently used (Garson, 2009) and was implemented here. The Box’s M tests did not attain signiﬁcance, suggesting the covariances are homogeneous, and that the assumptions of discriminant function analysis have not been violated. Several statistical procedures were used to examine size differences of the hyoids. First, paired samples ttests were used to test for differences in measurements between the right and left sides of the same hyoid. Sequentially reductive Bonferroni multiple comparison tests, which increase the power in detection of more than one false null hypothesis relative to the standard Bonferroni technique, were used to adjust the alpha level in these analyses (Holms, 1979; Rice, 1989). Second, two-way ANOVA (analysis of variance) was used to determine the main effects of ancestry and sex on hyoid form, as it is possible that individuals of European and African ancestry may display different patterns of sexual dimorphism. Therefore, potential interactions between ancestry and sex were explored in order to determine if separate ancestry-speciﬁc discriminant functions were necessary. Third, discriminant function analysis was used to develop equations utilizing a combination of measurements that have the ability to classify a hyoid as either male or female. Separate functions were developed using both fused and unfused hyoids. In addition, the likelihood that an unfused or fragmented hyoid may be recovered is high, and it is most likely that the body will survive due to its more dense composition. Therefore, a third function was developed using the two measurements taken of the body from unfused hyoids [maximum length of the body (BL) and maximum height of the body (BH)] in an attempt to test the usefulness of just the body in determining sex. Three sets of discriminant function analyses were performed based on these samples (fused hyoids, unfused hyoids, and the hyoid body alone). Each analysis was run twice, once including ancestry as a variable and once excluding ancestry as a variable, resulting in a DETERMINING SEX OF THE HYOID 281 Fig. 1. Fused (a) and unfused (b) hyoid measurements. Refer to Table 2 for a description of the measurements that correspond with each number (total hyoid width (THW) not shown). TABLE 3. Average percentage of error of fused and unfused hyoids Average error Dimensiona Fused Unfused BL BH CWI CHI CL CWS CHS THL THW WCS 4.72% 2.08% 4.52% 3.01% 2.19% 3.04% 4.44% 1.35% 0.85% 0.40% 0.91% 2.44% 3.58% 3.43% 0.47% 4.70% 4.67% a See Table 2 for abbreviations. total of six discriminant functions. Functions 1–3 include ancestry as a variable in order to acknowledge and compensate for the potential size differences between ancestries. If the ancestry of the hyoid in question is known, the coefﬁcient for the ancestry variable is provided and is simply input into the function. When using the function to estimate the sex of an individual of African ancestry a coefﬁcient of 1 is used while the coefﬁcient of 2 is used when estimating the sex of a European individual. In developing the discriminant functions, the signiﬁcance of each independent variable in the analysis was determined using stepwise statistics to ﬁnd the variables which set the Wilks’ lambda at a minimum. Once the stepwise analysis identiﬁed which variables were most valuable in determining sex, a second analysis was performed in which these variables were added independently in addition to ancestry. All statistical tests were run using SPSS 17.0. RESULTS Paired samples t-tests comparing right and left sides of fused and unfused males and females showed only one signiﬁcant difference (the height of the greater cornu at the fusion point with the body of fused female hyoids). Although the difference between the right and left sides at this point is statistically signiﬁcant (P 5 0.003), the difference between the average values of each side is still relatively small (0.2 mm). Given this, and the fact that the remainder of the analyses failed to attain signiﬁcance, it seems probable that hyoids only exhibit low levels of asymmetry. This suggests that although the data used to develop the functions presented here was recorded according to side, it is not necessary to discrim- inate between the right and left sides when using the functions. Therefore, averages of right and left sides are used in the following ANOVA and discriminant function analyses. This will allow the user to utilize the functions when a partial hyoid is recovered or when it is difﬁcult to distinguish one side from another. Measurement means and standard deviations for males and females of both ancestries are shown in Table 4. The results of the two-way ANOVAs are shown in Table 5. Sex is a signiﬁcant main effect on hyoid metrics for all measurements on fused hyoids, and all but one measurement on unfused hyoids. Ancestry is a significant main effect for four measurements on fused hyoids and three of the measurements on unfused hyoids. However, several of the two-way ANOVAs show a signiﬁcant interaction effect, indicating that the patterns of sexual dimorphism in the hyoid may differ between individuals of different ancestries. In all of the cases where a signiﬁcant interaction effect was detected, individuals of African descent show lower levels of sexual dimorphism in the hyoid than individuals of European descent. As signiﬁcant interaction effects between sex and ancestry were detected in the two-way ANOVA, it is useful to include ancestry as a variable in the discriminant function analyses. For each analysis that was performed, each step was statistically signiﬁcant (P \ 0.001) and the Wilks’ lambda was signiﬁcant for each function (P \ 0.001), suggesting the function is producing signiﬁcantly different discriminant scores for each group. High canonical correlations and eigenvalues close to 1.0 indicate that the distinction ability of the function is high. Functions 1–3 include ancestry as a variable, and the coefﬁcients of 1 and 2 correspond with individuals of African and European ancestry, respectively. Functions 4–6 were developed using pooled ancestries samples, and can be used if the ancestry of the individual is unknown. Table 6 shows the functions, the condition of the hyoid in which they are used, their cutoff points, and overall accuracies. Function 1, developed using fused hyoids, utilizes the hyoid’s total length, maximum length of the body, and height of the greater cornu at its fusion point with the body (inferior end) in addition to ancestry. Function 2 was developed using complete unfused hyoids and the variables needed are maximum length of the body, maximum height of the body, height of the greater cornu at its inferior end, and ancestry. Function 3 was developed using only those measurements taken of the hyoid body and therefore uses only the maximum length and height of the body in addition to ancestry. Functions 4–6 use the same variables as Functions 1–3, with the exception of ancestry, and have very similar accuracies. Accuracies range between 82% and 85% for all functions. American Journal of Physical Anthropology 282 S.C. KINDSCHUH ET AL. TABLE 4. Dimension means (mm) and standard deviations for males and females of both ancestries Fused Dimensiona BL Sex Mean Std. deviation N Mean Std. deviation N Black F M F M F M F M F M F M F M F M F M F M F M F M F M F M F M F M F M F M F M F M F M F M F M F M F M F M F M F M F M F M 20.29 23.93 22.17 24.55 20.86 24.12 10.79 12.13 10.56 12.11 10.72 12.13 4.02 4.32 4.33 5.01 4.12 4.54 6.73 7.49 6.58 8.35 6.68 7.75 27.23 30.34 26.46 31.27 27.05 30.65 3.21 3.24 2.98 3.55 3.16 3.34 4.20 4.37 4.07 4.90 4.17 4.55 33.51 37.47 31.63 37.87 33.02 37.60 37.21 41.48 41.09 43.97 38.33 42.25 37.84 42.45 42.45 44.51 39.11 43.15 2.47 2.16 2.21 2.69 2.54 2.34 1.31 1.16 1.07 1.21 1.24 1.17 0.69 0.47 0.70 1.05 0.70 0.77 0.99 0.70 0.92 1.17 0.97 0.95 3.24 3.16 3.14 3.30 3.21 3.20 0.56 0.57 0.44 0.74 0.54 0.64 0.85 0.76 0.52 0.91 0.79 0.85 3.58 3.75 3.45 3.40 3.62 3.62 4.06 3.68 4.75 5.28 4.60 4.36 5.42 4.80 6.14 5.88 5.96 5.23 65 51 28 23 93 74 66 51 29 23 95 74 66 51 28 23 94 74 66 51 28 23 94 74 47 33 14 17 61 50 49 35 13 17 62 52 49 35 13 18 62 53 63 48 22 23 85 71 64 49 26 22 90 71 47 35 18 18 65 53 20.30 23.40 21.04 25.11 20.55 24.17 10.87 12.63 10.46 12.25 10.73 12.46 3.78 4.06 3.65 4.59 3.74 4.29 5.95 7.05 5.71 7.18 5.87 7.11 27.55 31.27 28.10 31.37 27.68 31.31 3.15 3.15 3.03 3.21 3.12 3.18 4.19 4.45 4.14 4.46 4.18 4.45 1.90 2.44 1.72 2.02 1.87 2.41 1.09 1.24 1.03 1.12 1.08 1.20 0.58 0.65 0.47 0.74 0.55 0.74 0.76 0.85 0.58 0.81 0.71 0.83 3.01 2.58 2.68 3.17 2.93 2.84 0.46 0.63 0.42 0.79 0.45 0.70 0.66 0.87 0.57 0.90 0.64 0.88 N/A 68 68 35 56 103 124 68 69 35 56 103 125 62 68 30 52 92 120 62 67 30 52 92 119 55 46 17 37 72 83 55 48 18 38 73 86 56 48 18 37 74 85 White Total BH Black White Total CWI Black White Total CHI Black White Total CL Black White Total CWS Black White Total CHS Black White Total THL Black White Total THW Black White Total WCS Black White Total a Unfused Ancestry N/A N/A See Table 2 for abbreviations. DISCUSSION Results from the analysis of the two-way ANOVA reveal several interesting patterns regarding how hyoid metrics vary in relation to both ancestry and sex. First, ancestry was found to be a signiﬁcant main effect on four measureAmerican Journal of Physical Anthropology ments taken of fused hyoids and two measurements taken of unfused hyoids. Among these variables, hyoids of individuals of European ancestry are generally larger than those of African ancestries. Second, three of the analyses show signiﬁcant interactions between ancestry and sex. In general, hyoids of individuals of African descent display 283 DETERMINING SEX OF THE HYOID lower levels of sexual dimorphism than hyoids of European descent. This suggests that whenever possible, ancestry should be included as a variable when determining sex based on hyoid morphology. However, as is displayed in Table 6, the overall accuracy of the functions, which is the average of the accuracy at which males and females are correctly classiﬁed, changes very little, if at all, when ancestry is included as a variable. This may be due to the fact that only one measurement utilized in the functions (CHI) displays a signiﬁcant interaction between ancestry and sex and that this variable actually contributes very little to the sex estimation. When comparing the standarTABLE 5. Results of 2-way ANOVA showing the effects of ancestry and sex on hyoid measurements Fused Dimensiona BL BH CWI CHI CL CWS CHS THL THW WCS Sex Ancestry Sex-Ancestry Sex Ancestry Sex-Ancestry Sex Ancestry Sex-Ancestry Sex Ancestry Sex-Ancestry Sex Ancestry Sex-Ancestry Sex Ancestry Sex-Ancestry Sex Ancestry Sex-Ancestry Sex Ancestry Sex-Ancestry Sex Ancestry Sex-Ancestry Sex Ancestry Sex-Ancestry Unfused F P F 56.47 9.77 2.49 50.24 0.38 0.28 17.39 17.95 2.71 65.29 5.06 10.52 33.36 0.01 1.54 5.87 0.09 4.60 8.53 1.37 3.71 64.12 1.35 3.20 23.60 18.64 0.89 9.36 9.39 1.38 \0.001* 0.002* 0.117 \0.001* 0.540 0.597 \0.001* \0.001* 0.102 \0.001* 0.026* 0.001* \0.001* 0.909 0.217 0.017* 0.759 0.034* 0.004* 0.245 0.057 \0.001* 0.248 0.076 \0.001* \0.001* 0.346 0.003* 0.003* 0.243 156.32 18.39 2.90 128.78 6.50 0.02 44.27 4.80 13.49 129.99 0.20 2.69 46.32 0.40 0.19 0.77 0.08 0.74 4.43 0.02 0.05 a See Table 2 for abbreviations. * Statistically signiﬁcant results. P \0.001* \0.001* 0.090 \0.001* 0.011* 0.901 \0.001* 0.030* \0.001* \0.001* 0.658 0.103 \0.001* 0.529 0.664 0.383 0.780 0.390 0.037* 0.879 0.830 N/A N/A N/A dized canonical discriminant function coefﬁcients, which are used to assess each variable’s contribution to the discriminant function, only ancestry (when used) has a lower input than CHI. For example, in Function 1 CHI has a standardized coefﬁcient of 0.410 as compared to 0.480 for total hyoid length and 0.537 for body length. Similar differences are seen in the standardized coefﬁcients of Function 2 and when the ancestry variable is removed in Functions 4 and 5 the distance between CHI and the other variables increases. This also suggests that while knowing the ancestry of the individual may be helpful it is not particularly necessary in order to estimate sex. The discriminant functions also display an interesting pattern when the accuracies for each sex are compared. Overall the two functions developed for use on fused hyoids classify males with higher accuracy than females, while those functions developed for use on unfused hyoids classify females with higher accuracy than males. Both functions using the hyoid body classify females better than males. It is possible that in the case of fused hyoids it is easier to classify larger hyoids, which tend to be male. In contrast, the functions developed for unfused hyoids perform better on smaller elements, which tend to be female. One of the most important aspects of this research is that it presents four discriminant functions that were developed speciﬁcally to be used if an unfused hyoid is recovered. In the past, similar research has primarily used cadaveric hyoids which were held together with soft tissue, simulating a fused hyoid regardless of whether the greater cornua were actually fused to the hyoid body. While this would provide an idea of the size and shape of the hyoid while in anatomical position, a hyoid recovered in a bioarchaeological or forensic context may be skeletonized or lacking enough tissue to hold the greater cornua to the body. In this case, formulae developed from fused hyoids that use measurements such as total length or width of the bone cannot be used and the ideal method of sex estimation would be to use formulae developed speciﬁcally for use on unfused hyoids. In addition, each of the four functions to be used solely on unfused hyoids has an overall accuracy of greater than 80% which is signiﬁcantly higher than the majority of previously developed discriminant functions. While differing sample sizes and the use of only complete hyoids may have had some inﬂuence on the accuracies of the functions developed in the research previously mentioned in the introduction, it is also likely that the method of measurement played a role. For example, TABLE 6. Discriminant functions for determining sex from unfused and fused hyoids Function number Hyoid condition Function 1 Fused Function 2 Unfused Function 3 Body only Function 4 Fused Function 5 Unfused Function 6 Body only Discriminant functions 1–6 D 5 (0.133)(THL) 1 (0.219)(BL) 1 (0.444)(CHI) 1 (20.107)(A) – 12.598 D 5 (0.210)(BL) 1 (0.416)(BH) 1 (0.498)(CHI) 1 (0.118)(A) – 13.047 D 5 (0.311)(BL) 1 (0.498)(BH) 1 (0.027)(A) – 12.850 D 5 (0.153)(THL) 1 (0.220)(BL) 1 (0.395)(CHI) – 13.123 D 5 (0.240)(BL) 1 (0.413)(BH) 1 (0.430)(CHI) – 13.006 D 5 (0.313)(BL) 1 (0.495)(BH) – 12.827 Total function accuracy Accuracy of sexes 84.4% M: 87.3%; F: 81.9% 20.13 84.3% M: 80.5%; F: 89.1% 20.092 82.8% M: 80.6%; F: 85.4% 0.0640 84.4% M: 87.3%; F: 81.9% 20.0765 85.2% M: 82.2%; F: 89.1% 20.092 82.8% M: 80.6%; F: 85.4% Cutoff point 0.0715 See Table 2 and Figure 1 for dimensions. ‘‘A’’ refers to ancestry (African 5 1, European 5 2). Values above cutoff points are males, values below are females. American Journal of Physical Anthropology 284 S.C. KINDSCHUH ET AL. Miller et al. (1998) and Reesink et al. (1999) radiographed each hyoid and then took measurements from the images, while Kim and colleagues (2006) digitally photographed each bone and then took measurements from the photograph. Taking measurements from a radiograph or photograph can be problematic due to dimensional distortion resulting from the angle at which the photograph was taken, the presence of soft tissue, and the bone density of each hyoid (Lekı̂san et al., 2005). Because these studies were performed using cadaveric sample populations in which tissue was removed from the bone before the image was taken, there is the possibility that a small amount of remaining tissue could alter the measurement taken from the image. Similarly, differences in bone density may alter the measurement if some portions of the bone do not show up on the radiograph. In addition, the use of imagery to take measurements has led to the taking of measurements which are not readily available to one without the use of this technology. For example, the discriminant function developed by Kim and colleagues (2006) utilizes three measurements, one of which is the ‘‘distance from the narrowest segment of the greater horn to a point equidistant between the distal and proximal ends of the greater horn, measured through the central axis of the greater horn on the right’’ (981). Unfortunately this would be a very difﬁcult and time consuming measurement to take, particularly if the only equipment available was a set of calipers. Because a physical anthropologist attempting to determine sex from a hyoid will likely not have the resources or time, especially if they are in the ﬁeld, to radiograph or photograph the bone and take measurements from the image, it is important to use methods that can be easily replicable whether in a lab or ﬁeld setting. In light of these concerns, the simplest and likely most accurate method of data collection is that used in this study. One possible limitation of this study is the representa_can, 1983), and the hyoids used tiveness of the sample (Is to develop the discriminant functions in this analysis may not be comparable in size and shape to those in modern populations. This issue of representativeness is one that is common in metric studies based on the Terry Collection and as such the user of the discriminant functions should be aware of possible discrepancies between the size and shape of the individual in question and those remains in the Terry Collection. The functions presented here should be used in conjunction with other methods of sexing whenever possible. However, studies such as Albanese et al. (2008) have achieved success in developing metric methods of determining sex using the Terry Collection and testing those methods on independent samples outside the collection. Although the functions presented here have not been tested on an outside population so that their ability to classify hyoids outside the Terry Collection is still unknown, it is believed that the functions can be useful in both modern and archaeological contexts. While the exact frequency of hyoid recovery is unknown and is strongly dependent upon numerous taphonomic factors, should a hyoid be recovered it is believed that these functions can contribute to the sex estimation of the individual. CONCLUSION The analysis of hyoid morphology presented here supports previous ﬁndings of sexual dimorphism in fused hyoids. In addition, this study has established the presence of sexual dimorphism in unfused hyoids. Although American Journal of Physical Anthropology recovering a hyoid in multiple elements does not allow for an idea of the overall size of the bone, size differences between males and females of the individual elements are signiﬁcant enough to allow for differentiation between the two sexes. The development of four discriminant functions that can be used to sex an unfused hyoid, will allow for sex estimations to be performed on an unfused hyoid, even if the only available element is the body. Sex determination of an individual is often performed using a limited collection of skeletal elements, particularly within a bioarchaeological setting, prompting the development of methods to determine the sex of bones that may not normally be considered as useful in the process of developing a biological proﬁle. The hyoid is rarely if ever considered to be a bone that can aid in sex estimation. However, this study has determined the presence of considerable sexual dimorphism in both fused and unfused hyoids and has produced six discriminant functions that can be used within multiple scenarios and which utilize measurements that can be quickly and easily reproduced. This will provide the physical anthropologist with the opportunity to make use of every skeletal element present when assessing the sex of an individual. ACKNOWLEDGMENTS The authors thank Dr. David Hunt and the Smithsonian’s National Museum of Natural History for allowing access to the hyoids of the Terry Collection. Special thanks also to Dr. Chris Ruff and the anonymous reviewers who helped to make this a stronger manuscript. LITERATURE CITED Albanese J, Eklics G, Tuck A. 2008. A metric method for sex determination using the proximal femur and fragmentary hipbone. J Forensic Sci 53:1283–1288. Garson, DG. 2009. Multivariate GLM, MANOVA, and MANCOVA, from Statnotes: Topics in Multivariate Analysis. Retrieved on 10/09/2009 from http://faculty.chass.ncsu.edu/ garson/pa765/statnote.htm. Holms S. 1979. A simple sequentially reductive multiple test procedure. Scand J Stat 6:65–70. Hunt DR, Albanese J. 2005. History and demographic composition of the Robert J. Terry Anatomical Collection. Am J Phys Anthropol 127:406–417. _can Y. 1983. Assessment of race from the pelvis. Am J Phys Is Anthropol 62:205–208. Jelisiejew T, Jaroslaw S, Kuduk I. 1968. Morphologic studies on the hyoid bone in man. Folia Morphol 27:172–182. Kim D-I, Lee U-Y, Park D-K, Kim Y-S, Han K-H, Kim K-H, Han S-H. 2006. Morphometrics of the hyoid bone for human sex determination from digital photographs. J Forensic Sci 51:979–984. Lekı̂san I, Marcikić M, Nikolić V, Radić R, Selthofer R. 2005. Morphological classiﬁcation and sexual dimorphism of hyoid bone: new approach. Coll Antropol 29:237–242. Miller KWP, Walker PL, O’Halloran RL. 1998. Age and sexrelated variation in hyoid bone morphology. J Forensic Sci 43:1138–1143. O’Halloran RL, Lundy JK. 1987. Age and ossiﬁcation of the hyoid bone: forensic implications. J Forensic Sci 32:1655–1659. Reesink EM, Van Immerseel AAH, Brand R, Bruintjes TJD. 1999. Sexual dimorphism of the hyoid bone? Int J Osteoarchaeol 9:357–360. Rice WR. 1989. Analyzing tables of statistical tests. Evolution 43:223–225. Shimizu Y, Kanetaka H, Kim Y, Okayama K, Kano M, Kikuchi M. 2005. Age-related morphological changes in the human hyoid bone. Cells Tissues Organs 180:185–192.