A Comparison of Sexual Dimorphism Using Cranial Measurements of the Norse and Zalavar Populations in the William H. Howells Craniofacial Dataset

          Cranial measurements can indicate if the size of variables is indicative of a particular group of people. These population-specific features also vary between males and females within a population in an expression known as sexual dimorphism. It is hypothesized that the crania of the Norse and Zalavar samples are sexually dimorphic within respective populations and show similar patterns of sexual dimorphism when compared across populations. 12 craniometric variables of the Norse (n = 110) and Zalavar (n = 98) populations were acquired from the William W. Howells Craniometric Data Set. The six cranial measurements analyzed are the Maximum Cranial Breadth, Maximum Cranial Length, Mastoid Height, Parietal Chord, Occipital Chord, and Frontal Chord. The six facial measurements comprise the Nasal Height, Bizygomatic Breadth, Supraorbital Projection, Biorbital Breadth, Bimaxillary Breadth, and Orbital Height. Kruskal-Wallis, Dunn’s, and correlation tests were run to ascertain if and where statistical significance occurred. Kruskal-Wallis tests revealed statistical significance in all 12 of the variables, while Dunn’s tests clarified the differences consistently took place between males and females within populations, although Orbital Height and Occipital Chord were not statistically different. Maximum Cranial Length (Males <0.001, Females = 0.012) and Orbital Height (Males = 0.014, Females = 0.015) were the only variables statistically significant across populations. Males and females varied very little from their sex counterparts across populations. In contrast, the sexes within populations were highly sexually dimorphic, accepting the alternative hypothesis of consistent patterns between the Norse and Zalavar populations.

INTRODUCTION

          The studies of human variation and sexual dimorphism are prevalent in anthropological research, both for archaeological purposes and forensic identification techniques. Applying research to past populations can shed light on how these concepts presented themselves in the past and how a population has changed over time. The skull is an increasingly useful tool in identifying population-specific traits, as well as the variation that occurs between males and females within a population. Because people often have features that recognize them as belonging to a specific group, one may wonder how much proximity affects how similar or different cranial features of two populations are. The objective of this study is to answer the following question: “How do cranial and facial measurements compare between sexes across and within the Norse and Zalavar populations of the William W. Howells Craniometric Data Set?” The alternative hypothesis states that the crania of both populations will exhibit sexual dimorphism in six facial and six cranial measurements. Depending on the results of this first part, the sexual dimorphism levels will be comparable between the Norse and Zalavar populations and show little variation across both groups. Contrarily, the null hypothesis rejects the presence of sexual dimorphism in these twelve variables, and therefore, the only way sexual dimorphism can be compared across populations would be the absence of it.

          Human variation is the concept that certain features are indicative of a certain region or population. This can range from cultural to physical attributes, although for the purpose of this study, physical attributes will be emphasized. These traits are caused by both genetic and environmental factors and are manifested in several parts of the body, including the skeleton. Physical features are also affected by evolutionary development, natural selection, gene flow, genetic drift, and DNA mutations. For example, Kim et. Al (2021) found that East Asian maxillary sinuses are larger than those of European and African descent, contributing to the tendency for those of East Asian descent to contract sinusitis more often. Population-specific traits, such as the one described above, can affect the sizes of cranial and facial features, which can, in turn, be compared to other populations to analyze the similarities and differences between cultures. In general, cranial features tend to vary more within populations than across populations, especially concerning populations outside of Africa to coincide with the Out-of-Africa evolutionary model (Rivera, 2019). Although the diversity of cranial variation decreases the farther from Africa a population is, it is important to remember that modern humans share 99.9% of the same DNA, meaning that all human variation occurs on a very small level. This is a very relevant statement in anthropology because of the discipline’s history of weaponizing human variation to support racist ideology.

          Sexual dimorphism describes the phenotypic differences exhibited by males and females (del Bove et al., 2023). Compared to other primates, humans have a relatively low occurrence of sexual dimorphism concerning the skeleton, with the pelvis being its most accurate indicator. Because the cranium is often the most well-preserved bone in the archaeological record and contains some displays of sexual dimorphism, it is often used in forensic anthropology and bioarchaeology to estimate sex. In these nonmetric sex estimation techniques, five features of the cranium are focused on (Buikstra and Ubelaker, 1994; Walker, 2008). The nuchal crest, mastoid process, supra-orbital margin, supra-orbital ridge/glabella, and mental eminence are each scored on a scale from one to five, with one being very gracile and five being very robust. Estimation from these landmarks usually ranks from “probable female” to “probable male.” The use of facial features for these techniques suggests that they will be more indicative of sexual dimorphism than cranial features. Because these techniques are morphological and not morphometric, scoring can vary between anthropologists as the features are interpreted differently. Because there is more room for user error on scoring traits like these, quantitative measurements of cranial features are preferred by some anthropologists, especially when researching archaeological remains and making inferences on sexual dimorphism.

          Both the Norse and Zalavar populations are composed of medieval individuals found in Norway and Hungary. According to Howells (Evteev, 2021; Howells 1989), the Norse population is considerably homogenous, as 85% of the skeletal data were obtained from the same cemetery near St. Nicolaus Church in Oslo, Norway. Some ethnic variation may exist within the Norse population because of Swedish influence; however, both countries bordered each other and were ruled by the same monarch during the Medieval period, making the sample reliable in terms of population representation. The remains are said to date between 1220 and 1400, creating potential overlaps of this cemetery and the tail end of Viking activity, which winded down in the 11^th century. Little information is given on how Viking conquests may have contributed to influence from populations other than the Norwegian inhabitants, as they invaded much of Western and Eastern Europe, as well as parts of Central Asia and into the Baltics. Though they sold the people they captured into slavery, those buried at St. Nicolaus’s Church were likely in the same Christian community in Norway. Thus, the homogeneity of the sample will not be entirely affected by multiple ethnicities.

          The Zalavar population, meant to characterize those of Hungarian descent, is much more heterogeneous than the Norse population. Skeletal materials were analyzed from two different cemeteries thought to contain the remains of the diverse ethnic groups that inhabited the land. One of these groups, the Avars, originated from Mongolia and migrated to Central Europe in the sixth century C.E. (Gnecchi-Ruscone et al., 2022). Though they remained in power in Hungary for the next 200 years, genetic testing of skeletal data completed by Gnecchi-Ruscone et al. in association with the Max-Planck Institute concludes that this group retained many qualities of Central Asia. The Slavs are another group within the Zalavar population that have Eastern European and Northern Asian qualities that cannot simply be considered “Hungarian.” Germanic peoples from Northern Europe are also grouped into this Zalavar population. The only people from this sample considered truly Hungarian in modern times are the Magyars. However, since these cemeteries have been identified to be in use between the 9^th and 11^th centuries, the Magyars would have been the first to migrate to the area to begin establishing the societal frameworks that would eventually become Hungary. In actuality, the Magyars were associated with the Ural Mountains in modern-day Russia and, therefore, exhibit more Eurasian qualities (Perry and Perry, 1983). This mix of ethnicities reflects the fascinating diversity of medieval Hungaria. Unfortunately, this poses an issue when testing human variation within and outside the Zalavar population because potential sexual dimorphism within the Avars, Magyars, Germanic, and Slavic peoples is leveled and unrepresented in Howells’ collection of measurements. In this study, the individuals will be treated as one Zalavar sample.

MATERIALS AND METHODS

          Dr. Benjamin Auerbach, associated with the University of Tennessee, made the William W. Howells Craniometric Data Set public. Of the 2524 crania, 28 populations, and 82 cranial variables that Howells and his colleagues collected measurements from between 1965 and 1980 (Auerbach, 2014), six cranial and six facial measurements of the Norse and Zalavar populations will be utilized. These measurements include Maximum Cranial Length, Maximum Cranial Breadth, Bizygomatic Breadth, Nasal Height, Mastoid Height, Orbital Height, Bimaxillary Breadth, Biorbital Breadth, Supraorbital Projection, Frontal Chord, Parietal Chord, and Occipital Chord. Abbreviations and location classification of each variable are provided in Table 1, with all measurements in millimeters. The Norse Population (n = 110) consists of 55 males and 55 females, while the Zalavar Population (n = 98) contains 53 males and 45 females. All methods were accomplished via RStudio software, version 4.4.1 (2024), and there were no missing values present.

Statistical Analysis

            The original Howells Data Set was used for this analysis, but to target the research question pertaining to both population and sex, another identification column named ‘PopSex’ was added to organize four groups separated by population and sex. Norse Males and Norse Females were named “NM” and “NF,” and Zalavar Males and Zalavar Females were named “ZM” and “ZF”. Because cranial size variation between and across the two populations is the issue at hand, not shape variation, the data were not transformed using the geometric mean. All statistical analyses were performed on this new data set, beginning with descriptive statistics conducted by the RcmdrMisc package (Fox and Marquez, 2023). Table 2 encapsulates the mean, standard deviation, standard error, and interquartile range for side-by-side comparison. Hypothesis testing on each population-sex group is necessary to compare variation within and across both population samples, but first, the data set needs to be evaluated for parametric assumptions to decipher the correct method. Normality was assessed using Shapiro Wilk, Anderson-Darling (nortest package, Gross and Ligges, 2015), and D-Agostino Omnibus tests (fBasics package, Wuertz et al. 2024), of which the sample sizes fit the expectations for each test on all 12 variables. Then, Levene’s test was used to evaluate the homogeneity of variance (car package, Fox and Weisberg, 2019). The Supraorbital projection (SOS) variable was not normal for any of the four PopSex groups, while the Bizygomatic Breadth (ZYB) and Orbital Height (OBH) were not normal for two of the four groups. All 12 measurements meet homoscedasticity.

          Due to several variables not meeting normality, the Kruskal-Wallis test was run for all variables, as the data are considered nonparametric and continuous. This method of hypothesis testing returns an H statistic, degrees of freedom, and a p-value, the last of which indicates if variation exists at any point between the four population-sex groups. A p-value below 0.05 indicates statistical significance. Next, to keep on par with the nonparametric requirements, epsilon squared (ε²) tests (rcompanion package, Mangiafico, 2024) estimated the effect sizes of each variable per group to disclose the magnitudes of the significant differences (see Table 3). To pinpoint where these statistical significances occur, post hoc analyses via Dunn’s test were run using the dunn.test package (Dinno, 2024). The method in which Dunn’s tests were performed was Hochberg’s because of its ability to limit the false discovery rate. These are intended to demonstrate if these differences are between sexes among a singular population or across both populations. The last tests performed were correlations on each population-sex group to compare how the variables of each group correlate with each other. These were completed using Kendall’s Tau B method.

RESULTS

Descriptive Statistics

          This study found the mean, standard deviation, standard error, and IQR within and across the Norse and Zalavar populations. All of the mean values expressed males as larger than females within their own populations. The Biorbital Breadth (EKB) is the only variable that showcases females in both populations as having larger standard deviations. This was also the circumstance for the Zalavar Females in the Maximum Cranial Length (GOL), Maximum Cranial Breadth (XCB), Frontal Chord (FRC), and Parietal Chord (PAC), as well as for the Norse Females in the Occipital Chord (OCC). All other standard deviations show males as having equal or higher values. The two variables with a higher standard deviation for Norse Females (EKB and OCC) coincidentally have higher standard errors. The variables with a higher standard deviation than males for Zalavar Females (GOL, XCB, ZYB, EKB, FRC, and PAC) predictably display a higher standard error as well. In certain variables, the interquartile ranges appear as higher values for females than males (ZYB, MDH, EKB, PAC, and OCC for the Norse Population and ZMB, EKB, FRC, and PAC for the Zalavar Population), signifying that the central values are more widespread in females. The standard deviations, standard errors, and interquartile ranges suggest there are higher levels of variation within the female population-sex groups than in the male groups. Regardless of females on occasion having these higher values in the aforementioned descriptive statistics, the male mean values were larger, implying that each population is likely sexually dimorphic.

Statistical Analysis

          As stated earlier, these data do not meet all assumptions of normality and, therefore, underwent non-parametric testing to calculate analyses related to the hypothesis. Kruskal-Wallis tests confirm that there are statistical differences present between any of the four population-sex groups for all 12 variables. Effect size testing in the aftermath indicates if these differences are small, medium, or large. The epsilon squared column (ε²) of Table 3 reveals that most of the measurements range from medium to large sizes, although the OCC variable is unique in that it is the only small difference at an effect size of 0.059. GOL and ZYB hold the highest effect sizes, at 0.425 and 0.596, respectively. Other large effect sizes include GOL (0.425), XCB (0.265), MDH (0.276), and FRC (0.287), while the medium effect sizes are recognized in NLH (0.223), OBH (0.092), ZMB (0.211), EKB (0.234), SOS (0.221), and PAC (0.188).

          Although significant differences are found to exist somewhere among the four PopSex groups, the Kruskal-Wallis tests do not specify if the statistical significances occur within populations and/or across populations. Therefore, Dunn’s tests analyze the explicit pair of groups for each variable in which there are differences. The use of the Hochberg method within the test will lower the likelihood of making an error in which the null hypothesis was not rejected as it should have been. Ultimately, it will show if sexual dimorphism is present and if similar levels of sexual dimorphism exist between the Norse and Zalavar populations. Focusing first on the results within populations, all but the Orbital Height and Occipital Chord measurements are significantly different, indicating sexual dimorphism between those ten variables.  Looking at both aspects of the tests, the Maximum Cranial Length is the only variable that varies both within populations and across populations. The Orbital Height varies only across populations, according to these results. Because the Occipital Chord did not yield significant differences neither within or across populations, it can be assumed that the difference is more practical than statistical, as it was still found that males had larger measurements on average than females from both the Norse and Zalavar populations.

Correlations

          Correlations were run between all variables of each PopSex group and, in turn, compared to the other three groups, looking specifically for similar values of each pair of measurements. Kendall’s Tau B technique was used to compare two variables at a time, which was run with the complete observation method best for working with a full data set. From this analysis, little to no correlations were discovered between any of the four groups, as interpreted from the mixed concordant and discordant outputs. In fact, the largest correlation seen of any of the groups reaches a moderate 0.508 between EKB and ZYB for Norse Females, which is less than ideal if looking for strong correlations. Thus, there are minimal relationships between the measurements within the population-sex group.  From these scatterplots, the common array of slight or fair correlations for each group can be seen, as well as the complete lack of vibrant circles that would indicate strong correlations. Each group has similar levels of correlations, especially when focusing on the groups across populations, allowing the observer to suppose that they compare enough to indicate similar patterns of sexual dimorphism.

DISCUSSION

            This study endeavors to compare sexual dimorphism between two distant but European populations to learn if human variation affects levels of sexual dimorphism within populations. Recall that the null hypothesis rejects the occurrence of sexual dimorphism in these 12 cranial and facial variables, which would not allow examining sexual dimorphism patterns across the Norse and Zalavar populations. The alternative hypothesis predicts sexual dimorphism will be assessable within each population and exhibit similar patterns of sexual dimorphism across populations. Being able to identify how the cranium exhibits sex-specific traits allows for a more methodical technique in sex estimation. This is more important in bioarchaeology than the practical use of forensic anthropology in many cases because sex estimation in forensic anthropology is completed to build a biological profile of the individual. When using samples from the later Holocene, such as the two samples in this study, it is ideal to obtain exact measurements of specific cranial measurements to understand where sexual dimorphism is presented and by how much compared to other populations.

          Although females show more variation within their own population-sex group, as detailed by the standard deviations, standard errors, and interquartile ranges between some of the measurements, males still had larger means of all 12 variables that corroborate the hypothesis that sexual dimorphism is expressed through these cranial and facial measurements. The entire collection of variables chosen for this observational experiment returned as statistically significant. If they were found to differ only within populations and not at all across populations, it would be estimated that sexual dimorphism is the only explanation for the statistical significances, accepting the first part of the alternative hypothesis that these variables exhibit sexual dimorphism. If there are no statistical differences across populations, then the males of both populations and the females of both populations would be similar in these variable measurements and, therefore, display the same level of sexual dimorphism. However, the statistical analyses did not reveal clear-cut results like those described above. First, it cannot be confirmed that sexual dimorphism is conveyed in all 12 of these variables based on the lack of significant differences between Orbital Heights and Occipital Chords of the Norse and Zalavar populations. While the means and z-scores of OBH and OCC approximate that males are larger than females in both populations, it is not by a significant amount that cannot assume sexual dimorphism exists within the entire sample, let alone the population for each respective region. Thus, sexual dimorphism can only be projected in eight out of ten of the measurements.

          Next, a discussion of the post hoc results concerning GOL, OBH, and OCC is warranted. A Kruskal-Wallis test found there to be statistical significance in OCC, yet this difference is not expressed within nor across populations. One cannot overlook that there may be a practical difference between the sexes of this variable rather than a statistical difference and accept the null hypothesis in this case that there is no sexual dimorphism and, therefore, no level of comparison to make between the Norse and Zalavar samples. GOL shows a significant difference both within and across populations, showing that sexual dimorphism exists for each population but that the populations have different levels of sexual dimorphism. OBH not being significant within populations but across populations for both groups does reveal similar patterns of sexual dimorphism between the Norse and Zalavar, specifically because it does not manifest in either population. Therefore, the similar level in this case is that the populations correspondingly do not show sexual dimorphism in the exact same variables. As for the remaining nine measurements (XCB, ZYB, NLH, MDH, ZMB, EKB, SOS, FRC, and PAC), the only statistical significance found was between males and females within each population. Referring once again to the effect sizes in Table 3, the XCB, ZYB, MDH, and FRC variables show the largest amount of sexual dimorphism between the sexes. NLH, ZMB, EKB, SOS, and PAC show a medium amount of sexual dimorphism. GOL, OBH, and OCC effect sizes are unable to be obtained in regards to sexual dimorphism because it was either not proven by the post hoc results (p-value > 0.05) or because the effect size cannot be solely equated to the difference between sexes of one population. It can be inferred that the cranial measurements were more likely to display sexual dimorphism than facial measurements based on effect sizes.  These results evaluate sexual dimorphism in the same variables and a lack of sexual dimorphism in the same variables.

            The rejection of Orbital Height indicating statistical significance between sexes within populations is surprising due to its effectiveness in human sex estimation using the cranium, even if the populations of both the Norse and Zalavar have z-scores indicating males as larger than females. One of the factors attributing to this could be the variation of ethnicities within each of the samples, particularly in the Zalavar population. The two cemeteries containing Avars, early Magyars, Germanic, and Slavonic individuals (Evteev, 2021; Howells, 1989) may have enclosed information about sexual dimorphism between these central European, eastern European, and Central Asian ethnicities that was lost when data were collected under the umbrella term ‘Zalavar.’ Interestingly, the same results were presented for the Norse population, where the majority of the sample is from one cemetery, and the known ethnicities of the Norse population are from Norway and Sweden. Another potential cause of OBH showing little variation between sexes is that the differences are not in size but rather in shape. This may be a factor of why it is helpful in sex estimation, but it is not indicated in this study.

CONCLUSION

          Ultimately, the alternative hypothesis is partially accepted in that ten out of twelve variables show sexual dimorphism, except for the Orbital Height and the Occipital Chord, the first of which was only significantly different across populations, and the last was not significantly different within and across populations. All variables aside from the Maximal Cranial Length (and, of course, the Occipital Chord mentioned above) are significant either within populations or across populations. GOL is the only one significant in both, making it implausible to gain a confident understanding of the effect size of sexual dimorphism. In the cases of OCC and GOL, the first part of the hypothesis claiming sexual dimorphism cannot be applied, rejecting it in favor of the null hypothesis. However, for the second part, all able variables appear to show similar levels of sexual dimorphism, including the lack of sexual dimorphism. OBH is consistent across both populations in its indiscernibility to express variation between sexes, while OCC is consistently insignificant statistically across and within populations. GOL reveals the same amount of significance affecting the comparison of sexes both within populations and across populations. Because of this, the second part of the alternative hypothesis is accepted.

            This study can be taken in a future direction to investigate how the different ethnicities, particularly the Zalavar population, can be taken into account. After all, Howells’ data set is one that studies human variation, which includes migrants of a geographic location. The data could also be standardized or transformed in a follow-up study to expand on how shape differences may exhibit sexual dimorphism within the cranium (Milella et al., 2021). More ethically and reliably obtained skeletal data from African populations can be compared to European, Asian, and South American populations to measure how sexual dimorphism presents itself in the cranium in different geographic regions of the world. Making advancements in the analyses of human variation and sexual dimorphism are vital in understanding how humans have evolved and portray their environment through their bodies.