Introduction

Estimating sex on modern skulls is a frequently discussed topic in both forensic pathology and forensic anthropology. Recent studies have shown that the applicability of the established dimorphic traits is variable between different populations [1]. It has additionally been found that discriminant functions from modern American samples do not sex German skulls properly [2]. Secular change between Euro-Americans and Europeans appears to be variable in terms of stature, regardless of their common ancestry [3]. This trend might occur in skulls as well.

Sexual dimorphism can be seen in several features on the skull, either morphologically or by metric methods. The most reliable morphologic traits have been shown to be (in order of decreasing reliability) supraorbital ridge, glabella, decline of frontal bone, upper edges of orbits, and mastoid process [4]. Out of these, glabella and mastoid process can also be measured and analyzed quantitatively. Furthermore, these two traits are, according to Walker, the most dimorphic ones, either as single features or in combination [1]. The value of glabella projection and mastoid height has been confirmed metrically as well [5].

The computer software Fordisc® 3.1 performs discriminant function analyses on cranial and postcranial data in order to assess ancestry, sex, and stature of an unknown individual [6]. In contrast to other methods, it can use any possible combination of measurements for its computation. The Fordisc® data sample is based on the American Forensic Data Bank (FDB) and hence contains mostly data of skulls from different American populations. It also includes a small number of Asians and a few other populations, such as Europeans, South Americans, and Polynesians. Therefore, sex assessment cannot be reliably performed on skulls from populations other than the previously named [2, 7]. Furthermore, Howells’s collection of pre-twentieth century skulls from all over the world can be used in the program. However, Fordisc® still lacks a modern European sample to perform reliably on European forensic cases [2].

The purpose of the present study is to compare sexual dimorphism of Euro-American and German skulls and its respective development from the nineteenth to the twentieth century. This should lead to an explanation of Fordisc®’s unsatisfactory sex assessment on Germans and a better understanding of how to improve it. Therefore, a modern German cranial measurement sample has been used as an import dataset in Fordisc®.

Materials and methods

The two populations studied are referred to as “Euro-Americans” and “Germans.” The term Euro-Americans includes American citizens with known European ancestry. Since these former European ethnicities no longer play a significant role in mating behavior [8], there is relatively little structure in this group. However, Price et al. have been able to identify three subgroups: those with predominantly northwest or southeast European ancestry and Ashkenazi Jews [9]. The group named as Euro-American in this study consists of individuals with predominantly northwest European ancestry (known from surnames and ancestry information). It comes from three major sources: the Terry and Todd anatomical collections (nineteenth century), forensic cases, and the University of Tennessee donated collection (twentieth century). All are identified individuals with known birth and death years. German is meant as individuals with full German heritage. The information has been gathered from museum records (nineteenth century), cemetery data (nineteenth and twentieth centuries), and positively identified forensic cases (twentieth century). Skulls not fitting the criteria above have been removed from the dataset before analysis. Although, it cannot be excluded having a certain amount of unknown admixture within the groups since the data have been collected postmortem.

All cranial measurements have been collected with a MicroScribe G2X digitizer using the program 3Skull® [10] to compute the measurements from the coordinates. For the Fordisc® analyses, an import dataset has been created. It contains several groups. All Euro-Americans are obtained from the FDB. The German sample contains skulls from modern forensic cases in Mainz and Freiburg, from cemetery excavations near Würzburg and in Inden (North Rhine-Westphalia), and from the following collections: Berliner Gesellschaft für Anthropologie, Ethnologie und Urgeschichte (BGAEU), Anatomisches Institut der Charité Universitätsmedizin Berlin, the Smithsonian Institution, National Museum of Natural History, Washington DC, and the Morton Collection, University of Pennsylvania. One strong GF20 outlier has been excluded from analyses. Samples were broken into five groups as follows: Germans, (early and late) nineteenth and (early) twentieth centuries, Americans, late nineteenth century, and early and late twentieth century. The term “early” is used meaning “00 through 49” of a century, whereas “late” has been defined as “50 through 99” of a century. The distribution in different centuries follows the birth year of each individual. Sample sizes by group are given in Table 1. One should note that the German group does not include any individuals from the late twentieth century. Given the absence of large collections of contemporary (and probably also future) German skulls, much of this research will need to be continued using CT scans which are routinely made on autopsy cases in some German institutes for forensic pathology. The feasibility of such an approach has been demonstrated by Verhoff at al. [11], by Ramsthaler et al. [4], and in Australia by Franklin et al. [12].

Table 1 Sample sizes of all groups used for analysis
Full size table

For purposes of this paper, the following three highly sexually dimorphic measurements were chosen: glabella subtense (GLS), mastoid height (MDH), and bizygomatic breadth (ZYB). These measurements, computed from coordinates, correspond to Howells’s [13] definitions. The decision for GLS and MDH was made because of their morphologically dimorphic value mentioned above. Studies on metric sex discrimination have shown ZYB as the most dimorphic measurement [14].

Fordisc® has been used to show secular change within each group and differences among the groups. Therefore, the program has been run using the groups and measurements mentioned above. Fordisc® provides summary statistics, classification results for each group, Mahalanobis distances among groups, and a canonical analysis for graphical display of intergroup distances. In addition, SAS® was used for ANOVA analysis.

Various Fordisc® runs were made to explore relationships among the groups. The twentieth century German sample sizes are too small to be used to estimate reliable sex discriminant functions for Germans. Therefore, the relationships were then used to determine which of the samples might serve as a useful proxy for twentieth century German. Results from the most promising proxies were tested to ascertain their effectiveness in classifying modern Germans.

Results

Figure 1 shows the groups on the first two canonical axes, which together account for over 99 % of the intergroup variation. They summarize all differences between the studied groups, illustrating the presence of sexual dimorphism within the populations and the respective extent of secular change. All females are on the positive end and all males on the negative end of the x-axis. On the vertical axis, the Germans form a group with the late nineteenth century Americans for both sexes while the twentieth century Americans are shifted downward. This suggests that remarkable secular change has occurred in the Euro-Americans from the nineteenth to the twentieth century compared to the Germans. Table 2 shows the within-sex Mahalanobis distances between the twentieth century Germans and the other groups. In both sexes, the twentieth century Germans do not differ significantly from the nineteenth century Germans or the nineteenth century Americans, but they do differ significantly from both twentieth century American samples.

Fig. 1
figure 1

Graphic Fordisc® results showing group distributions along the first two canonical variates, which together display 99.3 % of intergroup variation

Full size image
Table 2 Mahalanobis distances of all groups from the twentieth century German sample
Full size table

Group means and standard deviations can be seen in Table 3. Very little secular change appears in the German group means. GLS increases in males and females, MDH and ZYB decrease in both sexes. In contrast to that, the values for Euro-Americans rise strongly for MDH, and ZYB falls much more than in their European counterpart. GLS shows fewer changes. Figure 2 illustrates this development graphically. Furthermore, it can be noted in terms of GLS and MDH that the amount of sexual dimorphism in Germans almost stays the same from the nineteenth to the twentieth century and is relatively small. The differences between American females and males are obviously larger and even increase over time. As in the Fordisc® graph, AL19 is very similar to the German groups.

Table 3 Group means of all three measurements with their respective standard deviation in brackets
Full size table
Fig. 2
figure 2

Plots of group means for each measurement illustrating the patterns of sex dimorphism

Full size image

Figure 2 suggests that the groups differ in sex dimorphism and the twentieth century Americans are more dimorphic than Germans. Table 4 shows the ANOVA test for group-sex interaction, which is the test for variation in sex dimorphism [15]. Group-sex interaction is not significant for ZYB, but both MDH and GLS show significant variation in sex dimorphism. This supports what could already be seen in Fig. 2: the amount of sexual dimorphism experiences secular change too for MDH and GLS.

Table 4 Analysis of variance test for group-sex interaction
Full size table

Sex classification for all groups has been performed using the AL19 and G19 discriminant functions. The percentage of correct classification was then computed and compared. Table 5 shows the results. In general, the Germans show lower classification rates (except for GF20) on both functions. Between the AL19 and the G19 function, the correct classification is almost equal. The lower correct classification for Germans may be explained through their lower grade of sexual dimorphism.

Table 5 Correct sex classifications on the AL19 and G19 discriminant functions in percent
Full size table

Discussion

Two major findings can be noted. First, German skulls from both centuries examined are similar to the nineteenth century Americans (see also Fig. 1). Second, there is a remarkable secular change in the American measurements between the nineteenth and the twentieth century, which is not seen in the Germans. There is also an apparent increase in the amount of American sexual dimorphism, predominantly in GLS and MDH, arising primarily because males have changed more than females.

It is not possible to state categorically that Germans are not experiencing secular change similar to Americans. Evidence from morphoscopic traits may shed some light on possible change. Walker [1] was only able to achieve correct sex classification in the upper 80 % range, which is similar to what Fordisc® achieves on the nineteenth century Euro-Americans. His European-derived samples were the nineteenth century Americans and the eighteenth century English. Walker observes that the nineteenth century Euro-Americans have higher glabella scores than the eighteenth century English, which may reflect secular change and suggests that Walker’s criteria may not work well on the twentieth century Americans.

Ramsthaler et al. [2] achieved about 94 % correct sex using morphoscopic criteria presented by Ascadi and Nemeskeri [16]. Their sample was German forensic cases said to date from 1968–2005. These are presumably death years, in which case their sample, like ours, contains birth years predominantly prior to 1950. In a later study, Ramsthaler et al. [4], using morphoscopic traits from Knussmann (see their Table 2), achieved 98 % sexing accuracy on 50 crania. These results were from CT scans performed in 2007 and 2008. This sample would contain individuals predominantly born after 1950. The improvement over the 2007 study may suggest greater sex dimorphism in post-1950 Germans. If so, it suggests that German secular change begins later than American, something that clearly requires further investigation.

Concerning secular change in Germany, it might be argued that relatively unrestricted migration in Europe provides a partial explanation for the phenomenon. Europe has a complex history of population movements, both into Europe and within it. Consequently, European populations are genetically relatively homogeneous compared to other continents [17]. Using genome-wide single-nucleotide polymorphism, Seldin et al. [18] were only able to subdivide Europe into a northern and southern group, with Germany falling into the northern. Therefore, we are confident that genetic heterogeneity has not unduly influenced our results.

Reasons for a stronger masculinization in Euro-American males are a matter of speculation. One possibility is a different amount of prenatal testosterone between the two nations. Whitehouse et al. [19] have shown recently that the level of umbilical cord blood testosterone at birth is positively correlated to facial morphology in the adult individual, meaning that more testosterone causes a more masculine-shaped face. The ratio of the second to fourth digit, regarded as a marker of prenatal testosterone, is also correlated with facial shape [20]. We have no direct evidence that Germans and Americans have different prenatal hormonal environments, but something of this nature is likely and should be investigated. With regard to glabella projection, Americans appear to have reversed the long-term trend toward gracilization [21].

There is no reason to believe that German secular change should proceed in the same manner as American, given significant environmental differences. Stature provides an example of American-European differences in secular trend. Komlos and Lauderdale [3] have shown that US whites, formerly the tallest people in the world, have been overtaken by the Dutch between 1930 and 1940. Since the 1960s, American heights have fallen below those of Germans, Danes, and Norwegians. What is also interesting is that the pace of secular change in American females has slowed more than males, resulting in an increase in sexual dimorphism in height and bone length [22], a change analogous to that seen in GLS and MDH.

Komlos and Baur [23] have argued that the American height deficit, compared to Europeans, may be due to the more comprehensive health care and social safety net afforded most Europeans. However, American secular change in cranial vault morphology has proceeded rapidly [24], wherein it is argued that there may be genetic components. Although the present paper presents evidence that sex dimorphism in Germans is less than in Americans, up to the mid-twentieth century, it does not address general cranial morphology. If sex dimorphism in the last half of the twentieth century increases, as may be suggested by Ramsthaler et al. [4], then it occurs later than in America. The next important step to be taken is to address secular changes in German cranial morphology more generally.

Conclusion

This paper underlines the suspected reasons for the current Fordisc® sample being less applicable on European skulls than on Americans. Sexual dimorphism is seen to be different between Euro-Americans and Germans. As American crania show greater sexual dimorphism, in general, it is not surprising to have clearer sex estimation for them.

However, it should be clear that a better understanding of European sex dimorphism is required. Efforts are now underway to expand European samples in Fordisc® which will improve their respective classification rates.