6

ATTRIBUTION OF SEX

Perhaps the greatest frustration for a number of scholars has been the fact that so much information about the Pompeian skeletal sample was lost through post-excavation processes. Though compromised, the collection of human remains that survive from Pompeii can still provide a wealth of information. Various methods can be applied to obtain a profile of the sample of victims in terms of sex ratio, ages-at-death, general health and population affinities.

Apart from providing us with insight into the lives and deaths of the people who became victims, this information can be used to test the assertion that the old, infirm, very young individuals and women made up the majority of the victims (Chapter 5).¹ Some authors have offered remarkable reasoning to account for such claims. Massa,² for example, stated that more women than men were discovered amongst the victims as ‘the wife and mother preferred to die than survive alone’. He also considered that women were more attached to their possessions than men and that a number died trying to save their jewellery and other valuables.

Associated artefacts were traditionally used to establish the sex of victims found during the course of excavations. For example, sex and age attributions were made for 194 of the thousand-plus victims documented in the excavation diaries. Of those that were identified as adults, 78 were said to be female on the basis of associated finds of earrings, necklaces and other jewellery and 35 were recorded as male.³ Physical anthropological techniques were not routinely applied for the determination of sex of in situ skeletal finds until the latter part of the twentieth century. Obviously, there was a need to employ the skeletal evidence to test the assumption that stereotypical associated finds provide an accurate indicator of the sex of an individual.

Juvenile sex determination

The reason that we are able to attribute sex to an individual skeleton is because sex-related differences can be observed on adult human bones. It is difficult to determine sex from juvenile bones because most of the variance between male and female skeletons only becomes apparent with the onset of puberty. Various methods have been proposed for juvenile sex determination but none of these are widely accepted.⁴

While male testosterone levels are generally very low before the onset of puberty, they do vary throughout development, which means that there are some age groups where it is theoretically possible to make a more reliable attribution of sex. Prior to puberty, male testosterone levels tend to be highest in the foetus from two months until birth. Nonetheless, the differences are too slight to be viewed with the naked eye and are only apparent from measurements on the pelvic bones of a number of individuals from a sample.

Juvenile sex identi fication has been attempted on the basis of statistical studies of sex-related differences, or dimorphism, in the foetal sciatic notch.⁵ Similarly, a study of the skeletons of children of known sex and age from the historic cemetery of Spitalfields, London, suggested that it was possible to establish the sex of juveniles between birth and five years of age in 70 to 90 per cent of cases on the basis of diagnostic morphological features of the pelvis and mandible.⁶

Most of these techniques, which are based on minor variations, are not appropriate for the majority of archaeological material as good skeletal survival is required for assessment. The available Pompeian sample, in particular, contained very few remains of individuals below five years of age.

It has been suggested that an estimate of sex can be made by comparing the degree of tooth development with the level of development of the post-cranial skeleton. This technique is based on the fact that there is a faster rate of post-cranial growth in female children, whilst teeth develop at about the same rate. This technique is not generally applicable to archaeological material as complete skeletons are required, along with large samples so that population norms can be established for dental and skeletal maturation rates.⁷

In theory, it is also possible to use teeth to establish the sex of older juveniles from some populations. It has been claimed that the teeth of males tend to be larger than those of females. As teeth do not increase in size once they are formed, it has been suggested that statistical studies of permanent tooth measurements could be used as an indicator of sex from the remains of children. In practice, this technique is problematic because male and female tooth size varies within and between populations. Further, environmental factors, such as maternal health during pregnancy and the nutrition of an individual in the period of tooth development, may also influence tooth size. In addition to these problems, the differences in tooth size are subtle and both intra and inter observer error can result in misidentification.⁸

Ultimately, there is no consensus that there is a reliable method for establishing the sex of juveniles from gross inspection and measurement of skeletal material. It appears that while there are morphological differences between pre-pubescent males and females, they are too subtle to enable accurate systematic detection.⁹Consequently, no attempt was made to distinguish male from female juvenile bones in the Pompeian sample, as no reliable method was available at the time of research.

In the long run, microbiology might provide more promising techniques for establishing the sex of juveniles from archaeological contexts. Examination of nuclear DNA should theoretically provide a most useful method for the determination of sex from juvenile skeletal remains. It has been claimed that nuclear DNA has been archaeological skeletal material Though this technique has definite potential, it can be quite difficult to obtain DNA that will yield readable sequences from archaeological skeletons where there is poor preservation of organic material.¹¹ It has been suggested that these problems will be minimized by the development of new DNA techniques as well as the application of new methods, like the analysis of the breakdown products of the organic components of tooth enamel.¹² To date, DNA analysis of samples from bones from Pompeii and Herculaneum has been very disappointing (see Chapter 9).
successfully extracted and analyzed from

to determine the sex of individuals.¹⁰

Attribution of sex from adult bones

The determination of sex from adult skeletal remains is based on the assumption that the bones of males are generally both more robust and larger than those of females.¹³ There are several factors which contribute to these differences. Hormonal changes at the onset of puberty are perhaps the most important. They result in an increase of androgens in males and oestrogens in females. Androgens help males develop bone and muscle more easily than females.¹⁴

Health, diet and lifestyle also determine the degree of muscle and bone development in individuals.¹⁵ Sexual dimorphism can, for example, be a reflection of the distribution of labour in a society and thus can vary considerably between populations. The bones of females can become extremely robust when they engage in heavy labour. In the absence of complete skeletons, bones from such individuals can be difficult to differentiate from those of males. For example, remarkably robust pre-Hispanic skeletons from Tlatilco were only identifiable as female on the basis of their unequivocally female pelves.¹⁶ The contribution of nutrition and health to the degree of sexual dimorphism in a population has been the subject of some controversy. It does appear that environmental stressors have a greater effect on males than females, which can lead to a decrease in skeletal sexual dimorphism, observable as a decrease in male height and robusticity.¹⁷

Genetic population differences may also account for differential sexual dimorphism.¹⁸ Some populations are relatively more robust than others, e.g. the skull of a female Indigenous Australian may be more robust than that of an Asian male. Conversely, the postcranial skeleton of an Indigenous Australian tends to be far more gracile than that of a European. However, due to the multifactorial nature of bone inheritance, the distinction between environmental and genetic factors is not straightforward.

To identify the sex of individuals with any degree of con fidence, it is necessary to know the parameters for sexual dimorphism in the specific population under investigation. When faced with an unknown population, as in the case of the Pompeian collection, it is essential to have a large sample to establish population norms.

Ideally, sexual attribution should be determined by an examination of the entire skeleton. This is impossible for the majority of the Pompeian material as a result of post-excavation disarticulation. Determination of sex for the different Pompeian bone samples was therefore based on both measurements and a standardized scoring system (see Table 6.1) from observations on samples of four different sets of bones, with emphasis on the pelvis (or innominate bone), followed by the humerus, the femur and the skull, including mandibles and teeth.¹⁹ Various statistical techniques were employed to demonstrate whether there was a clear separation of measurements into two groups and whether there was skewing towards either robust or gracile measurements or observations. Based on the assumption that males are more robust, this would provide an indication of the proportion of ‘males’ to ‘females’ in the sample.²⁰

Pelvis

The pelvis is generally considered the most reliable skeletal indicator of sex as a result of its biological function.²¹ The criteria that were used for sex determination were mostly visual. This enabled observations to be made quickly on the available sample.

The pubic area shows a marked degree of sexual dimorphism because increased hormonal activity at puberty results in a lengthening of the pubic

Table 6.1 Scoring system for the determination of sex from observations on bones Score Sex Attribution

1 hyperfemale
2 unequivocal female
3 more female than male
4 mid-range
5 more male than female
6 unequivocal male
7 hypermale

Notes: This scoring system was based on the five-point system devised by the ‘Workshop of European anthropologists’ in 1972 (Ferembach et al., 1980, 523). This scoring system is compatible with the 1994 standards (Buikstra et al. (eds), 1994, 19–21). I modified this system with the addition of two further scores to include equivocal cases for which a sexual attribution could be inferred.

bone in females to facilitate childbirth. The increase in length is greater at the symphyseal than the acetabular end of the bone.²² As a result, the pubic region is the most useful portion of the pelvis for objective visual assessment of sex. Sex determination based on observation of this area is considered to be as reliable as metric analysis.²³ Unfortunately, this region, which is relatively fragile, tends not to survive well in many archaeological contexts and is not as well represented as other areas of the pelves in the Pompeian sample.²⁴

Some pelvic features are considered to be exclusive to females. It was initially assumed that these resulted from changes to the pelvis due to stress from childbirth, or parturition. The two most commonly claimed to result from parturition are pitting on the dorsal surface of the pubic symphysis and the presence of a pre-auricular sulcus or groove. It has been suggested that these changes to the bone reflect ligament damage caused during childbirth. The pubic ligaments tend to become relaxed prior to delivery, possibly as a result of the production of hormones, like relaxin, during pregnancy. This enables some movement and separation of the pelvic bones during childbirth. Too much strain can result in a lesion, which could leave a scar in the form of pitting on the bone.²⁵

This interpretation of these observed changes has been challenged. It has been argued that they do not necessarily result from childbirth-related injury, which is significant as it means that it is possible for similar changes to be observed on male pelves. For example, two forms of pre-auricular sulcus have been identified. The more ‘female’ form that has traditionally been associated with parity is thought to result from changes to the pelvic joint ligaments during pregnancy and childbirth and has been called a groove of pregnancy (GP). The other form is related to the degree of ligament development at the sacro-iliac joint and is described as a GL.²⁶ It should be noted that some inter-population variation has been observed for the pre-auricular sulcus. It does not tend to be so apparent on the pelves of certain populations, such as the Bantu, where sexual dimorphism is generally less marked than in Europeans.²⁷

Though it has been argued that there is some degree of correlation between the presence of dorsal pitting of the pubic symphysis and parity,²⁸ there are other factors which contribute to its occurrence, as such changes have also been observed on the pelves of men and nulliparous women. The latter group includes women who have either not had a pregnancy come to term or who have not had a natural delivery, as in the case of a Caesarean section.²⁹ Furthermore, these lesions tend to be obliterated with advancing age.³⁰ A study of human and non-human mammals revealed that there was not a strong correlation between the degree of change at the pubic symphysis and the preauricular area.³¹

The uncertainty about the interpretation of these bony changes is a reflection of a major problem associated with skeletal identification. Many hypotheses about changes observed on bones are based on educated guesswork or observations on a limited number of cases. This is because access to skeletal material, of known individuals is limited by ethical considerations. Occasionally, it is possible to test hypotheses with well-documented archaeological material, as in the case of the large sample of eighteenth- and nineteenth-century skeletons with coffin plate and other documentary information that were excavated from the crypt of Christ Church, Spitalfields in London. Comparison between the skeletal evidence and biographic and genealogical data indicated that there was no correlation between childbearing and pitting in the areas where ligaments attach on the dorsal surface of the pubic symphysis and preauricular area. Instead, there was a significant correlation between pelvic size and pitting. It was concluded that these changes were more commonly observed on females than males because they tend to have a larger pelvic area.³²

In terms of the determination of sex for the Pompeian pelvic sample, the reasons for the appearance of these features is not as important as the fact that they tend to be correlated with females rather than males, though it is important to be mindful that presence of these features does not necessarily provide incontrovertible evidence for a female attribution.

Initially, both right and left pelvic bones were used for sexing as the results of the early stages of sorting suggested that there were a greater number of right than left bones. I decided that it would be useful to look at both sides to establish whether the sex ratios were the same. A quick visual assessment suggested the proportion of males to females was roughly equivalent on both sides. When the bones were finally sorted and broken pelves reconstructed, it was found that there was minimal difference between the numbers of innominates representing each side and I only recorded the left bones in detail.

Bones were originally de fined as adult only when epiphyseal fusion was complete (see Chapter 7). This definition, however, excluded certain bones where there was evidence of changes normally associated with female skeletons, such as a pre-auricular groove. The definition was cautiously altered to include bones where fusion was complete, except at the iliac crest and the tuberosity of the ischium. Fusion in these regions is often not completed until the third decade by which time the hormone changes to initiate sexual dimorphism have occurred and reproduction has been possible for some time.³³ It is notable that certain scholars³⁴ define innominates where fusion at the iliac crest has commenced as young adult rather than sub-adult. Sexual attributions based on innominate bones of individuals who had not yet attained complete maturity were noted separately.

Sex determination was based on a combination of ten observations and three measurements from a sample of 158 left adult and older adolescent innominate bones, mostly from the Sarno Baths.³⁵ This sample represents all the material that was available from the Pompeian collection.

Some of these features, such as the ventral arc, are known to be more useful discriminators than others. The range of features were chosen because they involved different parts of the bone, so that no matter how incomplete the specimen it would be possible to include it in the study. The more reliable indicators were employed as a baseline to enable population norms to be established for the pelves in the Pompeian sample.

As expected, the pelvis proved to be the most useful sex indicator of the samples of individual bones that were examined and should be used as a baseline for the interpretation of the other bones. It is notable, though not surprising, that the non-metric observations produced far better separation than the metric data. Since the morphology of the pelvis is based on biological function, the non-metric features were considered more reliable.

Both univariate and multivariate descriptive statistics produced similar results for the non-metric data (as typified by Figures. 6.1 and 6.2). Three main conclusions could be drawn with regard to the individual features. The first is that the best pelvic features for sex separation from the Pompeian material are the ventral arc, sub-pubic concavity and the sub-pubic angle, closely followed by the medial aspect of the ischio-pubic ramus and the obdurator foramen. The pre-auricular sulcus, sciatic notch and pubic tubercule are also good indicators but do not appear to separate the sample as well. The second is that the auricular area does not display any degree of bimodality (Figure 6.3) and, unlike the other features, is skewed towards the more female end of the range. It is clearly not a good sex separator but may perhaps be useful as a population descriptor. The third is that this research supported the assertion that dorsal pitting may not be specifically linked to parturition as it was also observed on pelves that were apparently male.

No matter how the data were treated in terms of statistical analysis, the pelvic observations, with the exception of the auricular area, consistently separated the sample into a higher proportion of males than females. This is at odds with the popular view that it was the women who were more likely to have become victims. The issue of sex determination from the pelvis, however, is complex and the results for individual non-pubic features do not necessarily reflect the actual ratio of males to females. Further consideration is required for the interpretation of the results.

It has been observed that female pelves tend to display more mid-range traits than males.³⁶ This certainly appeared to be the case for the Pompeian sample. Many of the pelves where the pubic region had not survived were rather androgynous in appearance and were difficult to classify. This phenomenon has been observed for other skeletal samples from central Southern Italy.³⁷ It is notable that Bisel considered the pelves in the Herculaneum skeletal sample to be highly dimorphic, though she probably had more complete bones in her sample (see below).³⁸ A number of pelves that were unequivocally female from examination of the os pubis, displayed either mid-range or male features, especially for the sciatic notch. In addition, it

Figure 6.1 Frequency histogram showing sex attribution based on the ventral arc

Figure 6.2 Frequency histogram of the factor scores for the first principal component of the pelvic non-metric traits used for the determination of sex

Figure 6.3 Frequency histogram of sex attribution based on the auricular area

has been suggested that older female pelves tend to exhibit more rugosity, which could lead to their being confused with those of males.³⁹

Another argument that could be put forward to explain the higher frequency of pelves with male attributes is that the sample may be skewed towards the more robust bones. It is difficult to assess this point, though the visualization of the ‘crumble factor’ on the skulls (Chapter 5) suggests that relative robusticity was not a major factor in the survival of adult material.

These factors probably partially explain the high male to female ratio. Nonetheless, there is no compelling evidence to suggest that the number of males to females should have been equivalent and it is extremely unlikely that the pelvic sample contained more females than males, even taking into account the tendency for mid-range female pelves to be misidentified.

Humerus

Seven measurements were made on just under 100 left humeri from the Forum Bath collection to determine the degree of sexual dimorphism in the humerus in the Pompeian sample.⁴⁰ The measurements were made with callipers, tape and an osteometric board according to standard definitions.⁴¹ Left bones were chosen to ensure that each bone represented a single individual. Previous studies of both left and right humeri have demonstrated that there are no apparent side related differences for these measurements.⁴² Dittrick and Suchey⁴³ found from discriminant function analyses of nine humeral measurements that multiple variables did not produce appreciably better results over the use of several of the best variables used singly.

In theory, the expected sex differences should be a re flection of the greater size and robusticity of males. Some of the measurements were employed because they have traditionally been considered good sex indicators.⁴⁴ The rest of the measurements were chosen on the basis of metric studies of sexual dimorphism from the humerus.⁴⁵ It has been claimed that, as for sex determination from the femur, several humeral measurements, analyzed singly, are more reliable sex indicators than cranial variables.⁴⁶

It has been found that proximal, or upper humerus measurements, generally provide more reliable sex prediction than distal, or lower humerus measurements.⁴⁷ Further, it has been observed that measurements of the humeral head are more diagnostic for sex determination than those of the femur.⁴⁸ It was, nonetheless, considered necessary to include distal measurements in this study, as not all the bones were complete. In addition, it has been found that for some populations, e.g. Indigenous Africans, distal measurements are more useful for sex determination.⁴⁹ France⁵⁰ found that diaphyseal or shaft diameters were much poorer sex indicators than either proximal or distal measurements. She suggested that the reason for this was that the diaphyseal shape displays a great deal of individual variation. The least circumference of the diaphysis was included in the data set as it enabled the inclusion of bones, which lacked articular processes.

The results for sex determination from the humerus suggested a more even distribution of males to females. Regardless of the technique used for dealing with the data, including exploratory tools, such as factor analysis, the results consistently separated the sample with a slightly higher frequency of males to females (Figure 6.4).⁵¹

Femur

Nine femoral measurements were made on a sample of over 160 bones from the Forum Bath collection.⁵² Only the left bone was used to avoid the possibility of the same individual being measured twice. Five standard measurements considered useful for sex determination were used. It has been claimed that sex determination from several femoral measurements, analyzed singly, is more reliable than that from cranial variables.⁵³ Additional measurements were included so that there would be sufficient measurements from incomplete bones to enable their inclusion in the study. Multivariate and other statistical methods were employed to determine whether two separate groups, presumably males and females, could be differentiated. The only metric data which appeared to be clearly bimodal were those for maximum length (Figure 6.5), though there was considerable overlap and some

Figure 6.4 Frequency histogram of the non-standardized factor scores of the humerus measurements

skewing towards the lower, presumably more female, measurements. In addition, the results of the multivariate analysis were consistent with those obtained from univariate visualization.

In contrast with the results of the non-metric pelvic observations, the results for the femur indicated that there was considerable skewing towards more gracile individuals. It is possible that this is a reflection of sample bias from the Sarno Bath ‘cottage industry’ of hinge manufacture from human femora for furniture restoration (Chapter 5).

Skulls

Sexual determination was based both on observations and metric examinations. Only unambiguously adult skulls were used for sex determination. Sixteen non-metric observations were used to distinguish sex from the Pompeian skulls.⁵⁴ The features that were chosen have traditionally been used for sex determination for archaeological populations.⁵⁵ These features employ characteristics that are considered to be good indicators of sex as they are based on relative robusticity and gracility. Like other non-metric traits, interpretation of the mid-range cases can be rather subjective. It has been noted that the usefulness of these indicators is dependent on prior knowledge

Figure 6.5 Frequency histogram of the maximum length of the femur

of the degree of sexual dimorphism of the population under investigation. Relative values were ascribed to different traits by the ‘Workshop of European Anthropologists’ so that these weights could be used to establish sex from a number of features.⁵⁶ However, the weightings they gave to nonmetric cranial traits are not necessarily applicable to an unknown population. Sex attribution was made in relation to the other skulls in the sample.⁵⁷

The non-metric skull data did not produce clearly de fined groups that could be interpreted as males and females. Different treatment of the data resulted in different relative proportions of males to females.

Examination of the individual traits showed that only a few features indicated possible bimodality.⁵⁸ Many of the cases for features were mid-range.⁵⁹ If the results for individual features displayed any evidence of skewing, it was generally towards the female range, though there were some features that showed an apparent bias towards males and one feature that separated into equal numbers of each sex.⁶⁰

A standardized but unweighted sex index was calculated, based on a linear combination of all the features, ranged around zero. If zero is used as a breakpoint, the ratio of males to females is 64 : 45 or 58.7 per cent : 41.3 per cent. The reliability of this result is limited, as it does not account for overlap. As mentioned above, a weighted index was also calculated, using the weightings recommended by the European Workshop of Anthropologists (WEA).⁶¹ This also produced a higher number of skulls with more female than male features, the ratio being 63 : 46 or 57.8 per cent : 42.2 per cent. The strong correlation between these two indices for sex separation is apparent when they are plotted against each other in a scattergram (Figure 6.6).

The scores for overall shape (SHA) were adjusted for comparison with the two indices. It is notable that the proportion of ‘females’ to ‘males’ was only comparable to those obtained from the weighted and unweighted indices if the mid-range scores were included with those identified as ‘female’. This produced a ratio of 64 : 45. Otherwise, the breakdown was about 26 (23.9 per cent) ‘females’, 38 (34.9 per cent) indeterminate and 45 (41.3 per cent) ‘males’. Principal components analysis did not produce a clear separation into two groups.⁶² If zero is used as a cut off point between ‘males’ and ‘females’, the frequency of ‘females’ to ‘males’ from the first principal component is 47.3 : 52.7. The second principal component was skewed to the more mid-range

Figure 6.6 Scattergram of the two sex indices for non-metric skull features

and ‘male’ range of scores with a frequency of 40.6 : 59.3 ‘females’ to ‘males’. The other three components did not separate the sample.

Even though the skull has been traditionally used as one of the major sex indicators, it was only possible to have limited confidence in the results for sex determination from the skull for the Pompeian sample. Part of the problem could perhaps be related to the sample, which was not very dimorphic. This was recognized as a problem by D’Amore et al.⁶³ who expressed concern about the accuracy of the results they obtained on almost the same sample as the one used for the current study. They attributed the much higher frequency of males to females they identified in the sample to an artefact of the misdiagnosis of robust females. It is worth noting that my analysis produced a completely different set of results with a higher incidence of skulls with female attributes. A number of explanations can be used to account for this. First, they included juveniles in their sample, though one would have expected this to skew the results to the female range. Second, they used different criteria, such as cranial capacity, for sex determination. Finally, the sample appears to reflect a population that is somewhat androgynous and difficult to sex with consistency. It is significant that D’Amore et al. had problems with the determination of sex from this sample (see below), especially as D’Amore has had considerable experience with material from the Campanian region.⁶⁴ This suggests that the Pompeian sample may differ from other Campanian samples. In contrast, Bisel⁶⁵ considered the male and female skulls from Herculaneum easy to differentiate by sex, though she had the advantage of complete skeletons for comparison (see below).

The problem of the determination of sex from visual criteria on skulls is not confined to the material from Pompeii. Howells⁶⁶ included a lengthy discussion of the problems he encountered in attempting to determine sex from the skulls of the various populations he used in his craniometric analyses in his publications. He documented the problem of lack of concordance between observers. He also noted a possible tendency to favour female attribution. This is consistent with the results from the Pompeian sample. It is notable that Howells relied on morphological appearance rather than measurements to assess sex from the skull.

Measurements

Choice of skull measurements for this study was determined by the constraints of preservation. The cranial vault tended to have much higher survival rates than the facial region for the majority of the sample and was able to provide the largest, most complete data set. Measurements were also chosen that could be used for comparisons with the work of other scholars. Twelve skull measurements were used to establish whether it would be possible to separate the adult skulls from the Forum Bath collection by sex based on a metric analysis. Standard definitions were used for these measurements.⁶⁷ Discriminant function analysis has traditionally been used for the determination of sex from skulls.⁶⁸ Discriminant function analyses are based on the extrapolation of parameters for sexual dimorphism from a known population onto unknown individuals to identify their sex. Interpopulation variation is significant enough to require the development of population-specific sets of equations.⁶⁹ As the Pompeian skeletal material derives from an unknown population, the use of a modern reference sample is likely to produce sex separations that do not reflect the actual sex ratio of the Pompeian victims. This analysis was not considered appropriate for the Pompeian sample.

The metric data proved to be more problematic for the determination of sex from the Pompeian skull sample than that based on inspection of nonmetric features.⁷⁰ No results that could be used with confidence were obtained from the metric study of sex separation. The variables were too weakly correlated to produce a successful principal components analysis. It could be argued that the lack of success in sex identification is a reflection of underlying heterogeneity of the Pompeian sample. This problem, however, is not restricted to heterogeneous material and can be illustrated by the following example.

A series of 25 head and facial measurements was made on a relatively homogeneous sample of 900 Swiss soldiers of the same age in order to design a few standard gas masks that would be suitable for a number of ‘typical’ faces. Principal components analysis was employed in an attempt to establish these ‘typical’ faces by reducing all the variables to a few principal components. The aim of the exercise was to rank all the individuals on the basis of the first principal component. The main proviso for the success of such an exercise was that the first principal component be well defined and account for a considerable portion of the total variance. The principal components analysis was ultimately based on ten of the measurements. The first principal component was found to account for only 43 per cent of the total variance and could not be used as an approximation of the sample. This was considered a surprising result given that the first principal component often accounts for more than 80 per cent of variance in morphometric studies. The conclusion drawn from this study was that there are numerous ways in which skulls and faces, in particular, can vary and it is not possible to represent them by a few dimensions without a significant loss of information.⁷¹

Mandibles

Four features were observed on all the available adult mandibles from the Forum and Sarno Bath collections.⁷² These features were chosen as they are based on relative robusticity and gracility and are generally accepted as good sex discriminators.⁷³ This appeared to be the case for the Pompeian sample. The first principal component accounted for 87 per cent of the variance and provided such good separation it could probably be used as a sex index (Figure 6.7). The split about the mean is very slightly in favour of males, with a frequency of females to males of 48 : 52.

Teeth

The teeth of males tend to be larger than those of females. It has been suggested that two crown diameters, the bucco-lingual, or cheek to tongue sides, and mesio-distal, taken at 90 degrees from the bucco-lingual measurement, especially when combined as an index, are good indicators of sexual dimorphism. Studies on living populations indicate that the canines are the most useful sex separators. Considerable variation has been documented between populations. Within population overlap is also significant, especially in females, and it has been suggested that teeth should not be relied on as the sole means of sex determination.⁷⁴

Two problems were encountered with the assessment of the Pompeian teeth for sexual dimorphism. The main problem was the lack of survival of teeth in the Pompeian sample, especially single-rooted anterior teeth, which had often dropped out of their sockets and been lost. The second problem

Figure 6.7 Frequency histogram of the factor scores of the first principal component identified from analysis of the observations from the mandible for the determination of sex

was that many of the surviving teeth displayed considerable attrition, which made it difficult to determine the original crown size.⁷⁵

All the available canines that did not have teeth on either side or were able to be removed from their sockets from the Forum and Sarno Bath collections were measured. Teeth that were not isolated in the mouth were not measured in situ because it was found to be extremely difficult to get the callipers between teeth, especially those that had been worn flat. Measurements were only made on teeth where the crown was substantially intact.

Unfortunately, not many teeth met these criteria and the sample size was limited to 9 maxillary canines and 19 mandibular canines. This sample is not large enough to draw conclusions about sexual dimorphism in the Pompeian canine sample.

Pompeii

Nicolucci⁷⁶ stated that the sample of one hundred Pompeian skulls he examined in his 1882 study was composed of 55 males and 45 females. Unfortunately, he neglected to document the criteria he used to make this assessment. In addition to the stated reasons for this study, the metric data collected by Nicolucci for his 1882 study were examined to determine whether sex attribution was based solely on visual inspection of if any metric features were used. If the latter were found to be the case, the degree of concordance between his sex attributions and those determined from the metric data would also be considered. Six of Nicolucci’s key measurements appeared to be comparable with the current craniometric study. An additional nine measurements were considered.⁷⁷

In addition, Nicolucci calculated several cranial indices from the measurements he made,⁷⁸ though the one he considered most important, especially in terms of population studies was the cephalic index (Chapter 3). This is defined, as are all indices, in terms of a ratio of one measurement to another, expressed as a percentage of the larger one.⁷⁹

The other measurement that Nicolucci and other nineteenth-century scholars considered most important was cranial capacity (Chapter 3). There are a number of methods for taking this measurement and Nicolucci was not clear about the technique that he employed. The most common way of making this measurement is to fill the cranial cavity with mustard seed or small shot. When it is filled to the foramen magnum, the shot is emptied into a cubic centimetre measuring glass to be read.⁸⁰ Nicolucci’s results for cranial capacity were investigated to determine whether it was used as a basis for sex separation.

The metric data collected by Nicolucci also did not prove to be useful for the determination of sex,⁸¹ which confirms the suggestion that this is a problem of the material and not the sample. It was not possible to establish the criteria that Nicolucci employed to separate his sample into males and females from the metric information. Perhaps the most interesting result from this exploration of Nicolucci’s data is revealed from an examination using cluster analysis. The fact that the data could not be forced into two groups suggests that the sample Nicolucci used was not comparable to that stored in the Forum Baths. The way the data split into specific groups perhaps implies that his sample was not random but was specially selected to include unusual specimens. It is notable that the alleged ‘negro’ skull he singled out formed a single cluster, implying that it really was significantly different from the other crania in his study.

The work of D ’Amore et al. on the skulls housed in the Forum baths produced slightly different results to those obtained in the current study.⁸² They poured millet into the skulls and then ordered them by increasing cranial capacity. The cranial index was also calculated so that the two features could be compared. These measurements were chosen for sex attribution because, according to Krogman,⁸³ female cranial capacity is generally about 200 cc less than that of males, and females have a relatively higher cranial index because their skulls are more rounded. These two elements formed the basis of their classification. They also employed other, perhaps more traditional, indicators of dimorphism, such as the presence of frontal bosses, sharp orbital margins, smaller zygomas and palates, and smaller muscle attachments to identify females.⁸⁴ They classified 43 skulls as female and 80 as male. These figures were calculated as percentages, namely 35 per cent and 65 per cent respectively as compared to Nicolucci’s breakdown of 45 per cent females to 55 per cent males.⁸⁵

The choice of cranial capacity and cranial index as the major sex indicators for this study requires some comment. These features are not commonly employed for sex segregation in current studies of archaeological skeletal material and are generally not recommended as criteria for sexing in physical anthropological texts.⁸⁶ Further, it has been asserted that the most useful sex indicators on the skull are to be found in the region of the face rather than in the area that houses the brain.⁸⁷ While it can be claimed that there is a slight difference between the average cranial indices of men and women in a population, the veracity of the assumption that males have greater cranial capacity than females has been questioned. Though there is some correlation between brain and body size, it is not certain what the extent of overlap is for the range of cranial capacity. It has been suggested that in the past, undue emphasis was placed on size differences for the determination of sex from skulls.⁸⁸ It has been further asserted that data may have been manipulated in the past, either consciously or unconsciously, to confirm preconceived ideas, namely that greater cranial capacity in males was a reflection of male superiority.⁸⁹

D ’Amore et al. were concerned about the results they obtained. Though it was not explicitly stated, this was possibly because they expected their evidence to support the notion that women were more likely to have been victims than men. They conceded that they did not know how many people were able to escape from the eruption or who they were. In their opinion, only a modest number of bodies had been recovered from Pompeii because the majority of skeletons were either lost or not collected during the course of excavation. They argued, therefore, that the higher male to female sex ratio they calculated was not necessarily an accurate reflection of the population of Pompeian victims. A comparison of their findings with the sex ratios obtained by other scholars, including Nicolucci, who worked on ancient and recent Italian skeletal material, led them to conclude that the number of males always exceeded the number of females in Italian samples. They explained this phenomenon as the result of a high incidence of robust females in these samples.⁹⁰

It is possible that this explanation is correct, though the argument presented by D’Amore et al. to support this claim can be criticized. It is worth noting that the source for all their comparative data was compiled in 1904.⁹¹ The methods for determining sex from skeletal material were improved considerably over the course of the twentieth century and it is possible that some of the sex attributions of these earlier works could be questioned. In addition, the documented tendency for a systematic bias towards male attributions from skeletal evidence⁹² is unlikely to have been recognized, let alone corrected for, by nineteenth- and early twentieth-century anthropologists. Another possibility that D’Amore et al. do not appear to have considered is that the scholars whose work they cited may not have based their studies on random samples. Nicolucci certainly did not state the criteria he used for selecting his sample. It is probably reasonable to assume that he chose the more complete skulls available to him for measurement, along with those that he found interesting, such as the supposedly ‘negroid’ skull. The former may well have led to a bias towards the more robust skulls, which may explain the slightly higher number of males than females in his sample.

D ’Amore et al. also explored the possibility of misdiagnosis as another reason for the higher frequency of males in their sample. They expressed reservations about the accuracy of their sex attributions as a result of the problems of sex determination based solely from the skull. Nonetheless, they apparently made no effort to compare their results with post-cranial bones from the Pompeian collection. They merely cited examples from the skeletal literature, which reinforced the view that sexual diagnosis from the skull was difficult and potentially inaccurate. The problems of sex determination from skulls were considered to result from the lack of unequivocal sex specific characteristics and variation between populations.⁹³

D ’Amore et al. concluded that there was probably considerable overlap between the sexes for the features that they chose for sex separation in their sample of Pompeian skulls. This diminished their confidence in the results that they obtained. They suggested that the Pompeian skeletal series was more robust than they expected and that skulls which they classified as males may well have belonged to females.⁹⁴

Recent re-examination of the skulls, mandibles and pelves of essentially the same material used in my study confirmed the results reported in this chapter. Sex attribution based on the pelvis and mandible yielded a higher incidence of males to females, while examination of the skulls suggested more females than males.⁹⁵

Herculaneum

Attribution of sex for the Herculaneum sample differed from that of Pompeii as individuals were represented by articulated, often complete, skeletons so there was no incentive to establish which bones were better indicators of dimorphism in the sample. As a result, comparisons of results for individual skeletal elements cannot easily be made between the two sites.

Of the 139 skeletons she studied, Bisel⁹⁶ determined 51 to be male and 49 to be female, the remaining 39 being juvenile and difficult to sex (see above). It is important to realize that these skeletons are included in all of the samples used by subsequent researchers.

Luigi Capasso established sex for 144 of the 162 skeletons available to him. He identified 83 as male and 61 as female. It is notable that he made sex attributions for juvenile skeletons, though he did acknowledge that there were problems with the technique he employed. He excluded a foetal skeleton that he sexed as female from these figures. He calculated the ratio of males : females as 1.38 : 1.⁹⁷

More recently, a sample of 215 Herculanean skeletons were studied by Petrone et al.⁹⁸ They did not attempt to establish sex from young juvenile bones but did make attributions for individuals aged from mid teen years. They also obtained slightly higher numbers of males than females, with 37.4 per cent of the total sample sexed as male and 31.8 per cent as female.

Sex and population affinities

It is important to recognize that features associated with sex can be population specific. The determination of the sex of an individual is inextricably linked with their population affinities.⁹⁹ A number of features that were considered to be possible sex indicators, like the auricular area, are probably more useful as descriptors of the population. This is true for other features for different populations, such as the shape of the chin, brow ridges and septal apertures of the humerus.¹⁰⁰ The fact that there are not many diagnostic features for the separation of the Pompeians into male and female categories can also be interpreted as a population feature.

Skulls, in particular, have long held a fascination for scholars (see Chapter 3) and have often been the primary bone used for analysis. This study indicates that, at least for the Pompeian sample, they are of limited value. In addition, the inability of the metric data to provide information on intrapopulation data in the form of sex separation does not augur well for their value as population discriminators for the Pompeian victims.

Conclusion

The assumption that the sample of Pompeian victims was skewed towards females is not supported by the skeletal evidence, which suggests that, if anything, the sample has more of a male bias. No explanation can be offered for such a bias, especially because it is difficult to interpret whether or not it is significant owing to the problem of overlap.

It is notable that different bones provided different sex ratios. The pelvic non-metric observations yielded a considerably higher incidence of males in the sample, whilst the humeral and skull measurements along with the mandible observations, suggested almost equal separation with a slightly higher frequency of males. In contrast, the results from the femur measurements implied a greater number of more gracile, presumably female, individuals in the sample. Similarly, the pelvic measurements and the non-metric skull observations suggest that the sample was composed of more females than males.

Interpretation of these results is dependent on the establishment of which bones and traits are the most useful indicators of sex for the Pompeian sample. The results from the femur can be questioned on the basis of sample bias as a number of the bones were removed from the Pompeian stores for secondary usage as hinges (see Chapter 5). From the results, it appears that the bones that are most useful features for sex separation for this sample are the non-metric pelvic traits, followed by the non-metric mandible traits and the humerus measurements.

The association between sex and population is well documented with sexual dimorphism varying between populations. It appears as if the Pompeian sample shows a tendency towards androgyny for certain features, especially in the pelvis and skull. This study indicates that the skull, and the craniometric data in particular, is not very useful as a sex indicator for the Pompeian sample and, by implication, cranial measurements are of limited value for the determination of population affinities for this sample. This is because they did not indicate any real separation into well-defined groups. This could be explained if the sample were heterogeneous, though other evidence that has been collected does not support this view. The results of this research suggest that these cranial measurements are not a good indicator for either sex or population affinities for the Pompeian sample.