SEX DETERMINATION OF ANCIENT HUMAN REMAINS

Transcription

1 1 SEX DETERMINATION OF ANCIENT HUMAN REMAINS A thesis presented by Peter Mayall to The Faculty of Arts University of Melbourne in fulfilment of the requirements for the degree of Master of Arts by Research in the field of Classics and Archaeology 2011 Produced on archival quality paper

2 2 TABLE OF CONTENTS ABSTRACT 6 LIST OF TABLES 7 LIST OF FIGURES 9 CHAPTER 1 10 INTRODUCTION 10 Research Aim 12 Problems 12 Thesis Organisation 12 CHAPTER 2 14 METHODOLOGY 14 Morphological Assessment 14 Population Variability 15 Reference Collections 17 Observer Error 18 Preservation and Recovery 18 Statistical Analysis of Ancient Remains.20 CHAPTER 3 25 REVIEW OF STUDIES INVOLVING DETERMINATION OF SEX 25 Cranium 25 Pelvis. 35 Post cranial...45 CHAPTER RESEARCH MATERIAL 74 Historical Setting..74 Equipment.77 Skeletal Material...77 CHAPTER RESULTS AND DISCUSSION...85 Analysis of pelvis.85

3 3 Analysis of cranium..87 Analysis of post cranial material.89 Summary of results..102 CONCLUSION 107 Key Findings 107 Contribution 108 Future Research 108 Summary 109 APPENDIX 109 Appendix 1. Data recording form Appendix 2. Study of techniques for determining sex at the elbow Appendix 3. A study to assess the morphological features of the femur 124 Appendix 4. Size of Sample Required for Sex Allocation..127 BIBLIOGRAPHY 130

4 4 SEX DETERMINATION OF ANCIENT HUMAN REMAINS A thesis presented by Peter Mayall to The Faculty of Arts in fulfilment of the requirements for the degree of Master of Arts in the field of Classics and Archaeology in the School of Historical and Philosophical Studies Classics and Archaeology The University of Melbourne Supervisors: Prof. Antonio Sagona Dr. Varsha Pilbrow 2011

5 5 Acknowledgements I am very grateful to my supervisors, Prof. Antonio Sagona, Classics and Archaeology, the University of Melbourne and Dr. Varsha Pilbrow, Department of Anatomy and Cell Biology, the University of Melbourne for allowing me to participate in the archaeological excavations at Samtavro in the Republic of Georgia in on which this study is based and also for their support, encouragement, expertise and patience in assisting me with the preparation of this thesis. I would also like to thank the curators of the osteology collection at the Museum of London who permitted me to access their database and skeletal specimens enabling me to research aspects of sex determination which were relevant to this thesis.

6 6 Abstract The determination of sex of ancient human remains is frequently difficult for several reasons. The skeletal material available for investigation is usually in poor condition being fragmented and fragile as well as incomplete and commingled. In addition, it often comes from an undocumented population and in numerically small samples, compounding the problem of obtaining a statistically significant outcome. This study aims to examine methods currently available for sex allocation focusing especially on those which can be usefully applied or modified to analyse ancient, fragmentary, incomplete and undocumented material. The available methods are then employed to determine the sex of human remains recovered from the ancient burial ground at Samtavro, Georgia. Sixty skeletal individuals were assessed using cranial and pelvic components as well as post cranial material. This analysis allowed the sex to be estimated for 87 per cent of the adults with a concordance between criteria of at least 80 per cent.

7 7 List of Tables Table 1. Ranked cranial traits according to their accuracy 29 Table 2. Cranid Measurements 31 Table 3. Accuracy levels for pelvic traits 44 Table 4. Humerus: accuracy of sex determination...48 Table 5. Accuracy of sex determination of distal humerus...51 Table 6. Femur metrics: accuracy in sex determination 57 Table 7. Tibial metrics: accuracy in sex determination...65 Table 8. Skeletal material recovered..79 Table 9. Relative occurrence of sex determinants from pelvic elements.80 Table 10. Occurrence of assessable cranial traits at Samtavro...81 Table 11. Humeri from Samtavro...82 Table 12. Specimens from Samtavro with femoral criteria...83 Table 13. Tibial criteria for sex determination at Samtavro...83 Table 14. Minor skeletal elements at Samtavro 84 Table 15. Sex allocation by pelvic criteria 86 Table 16. Sex determination of crania by morphological analysis.88 Table 17 Glenoid cavity length and sex allocation 90 Table 18. Measurements of humerus and sex allocation 92 Table 19. Femoral measurements to determine sex...94 Table 20. Section points for femur derived from other sources 95 Table 21. Comparison of sex determined from femoral and pelvic measurements 96 Table 22. Accuracy of individual femoral measurements to determine sex...97 Table 23. Determination of sex from tibia..98 Table 24 Allocation based on dental measurements..100 Table 25. Age and sex of subadults Table 26. Final sex allocation 104 Table 27. Analysis of individual tombs.106 Table 28. Elbow Criteria from Samtavro specimens..112

8 8 Table 29. Depth of Olecranon fossa..119 Table 30. Thickness of base of Olecranon fossa Table 31. Angulation of Medial Epicondyle.120 Table 32. Epicondylar breadth of humerus Table 33. Olecranon process Coronoid process distance..120 Table 34. Olecranon Coronoid Angle..121 Table 35. Maximum diameter of head of radius Table 36. Combined Results.121 Table 37. Measurements of angle of neck of femur.124 Table 38. Measurements of degrees of torsion of femoral neck..125 Table 39. Measurements of the bicondylar angle.125

9 9 List of Figures Fig.1. Graphical Representation of Section Point 21 Fig. 2. Scoring System for Cranial Features 27 Fig. 3. Gonial Angle 28 Fig. 4. Representation condylar region base of skull indicating measurements 33 Fig. 5. Method of mastoid measurement 34 Fig. 6. Sex differences in the pubic bone 37 Fig. 7. Sex differences in the greater sciatic notch 39 Fig. 8. Subpubic angle. 39 Fig. 9. Landmarks for determining sex of innominate 40 Fig. 10. Measurement of sciatic notch and acetabulum 41 Fig. 11. Auricular area and pre auricular sulcus 42 Fig. 12. Landmarks for measuring glenoid cavity length 46 Fig. 13. Measurements of the humerus 49 Fig.14. Carrying angle at elbow joint 52 Fig. 15. Trochlear morphology 52 Fig. 16. Olecranon morphology 53 Fig. 17. Olecranon coronoid angle 54 Fig. 18. Femoral measurements used for sex allocation 56 Fig. 19. Purkait s triangle...58 Fig. 20. Landmarks for measuring size and angle of femoral neck..60 Fig. 21. Male and Female femora...61 Fig. 22. Femora with different degrees of torsion..62 Fig. 23. Measurements of the tibia used for sex determination..64 Fig. 24. Scatter Plot for Coimbra Collection Data.72 Fig.25. Maps of Georgia 75 Fig. 26. Measuring Olecranon fossa depth.115 Fig. 27 Angulation of medial epicondyle of humerus.116 Fig. 28 Olecranon Coronoid Angle 117 Fig. 29. Chelsea Old Church sample.128 Fig. 30. Maximum femoral head diameter (Roman period)...129

10 10 CHAPTER ONE Introduction The determination of the sex of human skeletons is integral to the study of an ancient community. It enables the estimation its people s physical characteristics, such as stature, age and ethnicity as well as assisting in the interpretation of their social and cultural history. The validity of the assessment of these features is dependent on the accuracy of the sex allocation which requires a careful evaluation by the most appropriate methods of the sexual dimorphism of the skeletal elements under consideration. Sexual dimorphism in the context of physical anthropology refers to the skeletal differences which can be recognized between adult males and females based on visual and metric criteria. The dimensions of female skeletal parts as a percentage of male parts range from 90 to 96 per cent. 1 Sex determination from human skeletal remains is feasible by the recognition of these sexually dimorphic traits which mostly result from morphological adaptations associated with childbirth in females, particularly related to the pelvis and the greater skeletal robusticity in males which is pronounced in the cranium. The morphology of the pelvis and cranium show considerable sexual dimorphism and when assessed by visual and metric analysis provide the most accurate estimate for sex allocation. Metric analysis of other skeletal elements however can provide additional supporting evidence. The normal individual variation within populations which results in some females being more robust and some males being less robust produces an overlap of the sexes at the centre of the normal distribution where it is difficult to allocate sex based only on metric criteria. This physical variation also applies to individuals belonging to specific populations in which the population as a whole may be generally larger or smaller 1 Krogman and Iscan 1986, p. 190.

11 11 compared to other groups due to genetic variation, nutrition and physical activity. Determination of sex can be achieved with considerable accuracy in adults when the skeletal material is relatively complete, in good condition and comes from a known population. Krogman was able to allocate the sex in a documented population using the entire skeleton in with 100 per cent accuracy of cases, from the pelvis alone 98 per cent were correctly allocated, from the cranium some 92 per cent, from the pelvis and cranium combined in 98 per cent and from the long bones 80 per cent of the individuals were correctly allocated. 2 Unfortunately these circumstances do not often apply to material retrieved from archaeological excavations where the remains are in poor condition, disarticulated, commingled, fragmented and incomplete. In addition they are often from an unidentified population. The attribution of sex under these circumstances is much more difficult owing to the state of the skeletal remains and lack of parameters from comparable material and therefore less accurate. On the other hand it has also been noted that archaeological populations tend to be more sexually dimorphic and genetically homogenous than the ethnically and genetically mixed populations used in modern forensic studies. 3 Sexual dimorphism is evident in human foetuses particularly in the pelvis with the sciatic notch being wider in female foetuses and deeper in males and the subpubic angle larger in females. 4 This was substantiated by Boucher who found that there was a greater difference between the sexes of the foetal subpubic angle than in adults. 5 This is because in male infants the secretion of androgens beginning in prenatal life produces sexual dimorphism with a larger muscle mass, a higher birth weight and skeletal changes which are apparent in pelvic morphology. 6 These changes in the pelvis are further accentuated by the acceleration in growth which occurs at puberty. 7 This transformation is a complex and dynamic process affected by sex differences related to variations in the rates and directions of growth in specific areas of the pelvis as well as individual variation. 2 Krogman and Iscan 1986, p Meindl et al. 1985, p Boucher 1957, p Boucher 1957, pp Saunders 2008, pp Coleman 1969, p. 149.

12 12 Expansion occurs at the growth centers in the iliac crests, ischial tuberosities, acetabulum and margins of sacro-iliac joint but the enlargement is also due to selective resorption and deposition of bone within the individual bones of the pelvis. 8 Examples of these are the greater length of the pubic bone in females contributing to their wider pelvic inlet and the directional differences in growth of the inferior sections of the ischio-pubic ramus and ischial tuberosity with both growing in a more lateral direction in females resulting in a wider subpubic angle. 9 The problem In the archaeological context the material to be analysed is often fragmentary, fragile, incomplete, commingled and commonly from an undocumented population which makes the process of estimating sex with acceptable accuracy difficult. In addition the skeletal material obtained may be retrieved as isolated specimens or in such small numbers that they do not comprise a statistically significant sample on which to establish parameters for sex determination. The research aim The aim of this study is to review the techniques for sex allocation in commingled and fragmentary human remains and employ them to estimate the sex of specimens recovered from an ancient burial site at Samtavro, in the Republic of Georgia. Thesis organisation Chapter Two reviews the published literature on sex estimation from skeletal remains. The background to sexual dimorphism is discussed and the materials and methods utilised in sex determination for cranial and post-cranial remains examined. The results 8 Coleman 1969, p Coleman 1969, p. 148.

13 13 from previous research on sex determination from skeletal elements are then tabulated to assess and compare the accuracy achieved by the various methods. In the final section of the chapter new techniques are proposed for estimating sex from the dimorphic features of the elbow and the femur. These skeletal elements are often recovered from ancient burials due to their compact bone density, but had not been quantified in previous studies. New techniques are proposed and tested on skeletal individuals of documented sex from the Museum of London. Chapter Three focuses on the material from the burial ground at Samtavro. The historical context of the burial ground is documented, with particular reference to the cultural and demographic circumstances of the period in which these burials took place. The individual skeletal elements recovered in the field seasons carried out in 2008 and 2009 and were analysed for sex estimation are recorded in this chapter. The results of these investigations are discussed in Chapter Four. To assess the accuracy of the techniques the results obtained in this study are compared to those from previous investigations. In addition where possible the estimations of sex from the pelvic and cranial analyses are compared to those from other post-cranial elements. Finally a tentative sex is allocated to each individual by summing the assessments made for the major skeletal elements. The major findings of this study are discussed in the conclusion with possible contributions arising from this research. Areas in which future investigations may be helpful in sex determination particularly with reference to ancient skeletal material are also proposed.

14 14 CHAPTER TWO Methodology This chapter examines the main techniques involved in the evaluation of ancient human remains for sex allocation. At the outset some general principles in the methodology need to be reviewed before the specific approaches relevant to individual skeletal elements are analysed. These principles involve several issues: a) The assessment of sexual dimorphism which necessarily involves morphological evaluation utilising qualitative and quantitative assessment and the notation of observer errors inherent in these methods. b) The effect of population variability. c) The methods of statistical evaluation of the findings and the accuracy of the resultant estimations. d) The state of the skeletal material. Morphological Assessment In all human populations there is generally a discernible difference in the size and shape of male and female skeletal elements with males being on average about five per cent larger. There is an overlap however, with some males being small and gracile and some females being large and robust. The evaluation for sex allocation involves both quantitative assessment with the measurement of skeletal elements and qualitative assessment with a visual appraisal of sexually dimorphic features. Qualitative assessment is based on the degree of robusticity which is visually apparent in specific skeletal components. This visual assessment is standardized by relating it to a five point scale with females at one end of the scale and males at the other with an indeterminate result in the centre. An example is that applied to sex determination based on the skull (Fig. 1).

15 15 Metric standards for skeletal elements have been developed for various population groups which form the basis of quantitative assessment. Metric methods have the advantages of being objective and can be analysed statistically whereas visual assessments are subjective and relies on the experience of the observer. Data bases containing analyses of reference populations have been established and widely utilised for this purpose but due to population variability their applicability is limited when dealing with an undocumented population. Population variability The qualitative features of skeletal individuals, in particular those related to robusticity and childbirth are universal and not so dependent on population specificity in their assessment. By contrast the quantitative variation in skeletal morphology is population specific and cannot be generalized to other populations. This was demonstrated when cranial functions calculated from a 19 th century skeletal collection 10 provided inaccurate sex determinations when applied to a modern population derived from the Forensic Data Base in Tennessee. 11 It was revealed there had been significant changes in cranial measurements over the intervening time period which the authors presumed were due to different growth patterns associated with alterations in nutrition, physical activity and migration. The chronological variability which can exist in populations was demonstrated by Dittich and Suchey when investigating the remains of three Native American populations. Those from the early sites were shown to be much more robust than those from more recent sites confirming the necessity of establishing specific criteria for different populations. 12 In addition it has been found that whereas there is a good correlation for sex estimation between crania and pelves for females this is not the case for males who have greater variability in their cranial morphology which tends to underestimate the numbers of 10 Giles and Elliot 1963, pp Ousley and Jantz 1997, pp Dittrick and Suchey 1986, pp. 3.

16 16 males in a population. 13 Skull morphology is also affected by advancing age, Female skulls, for instance become larger and have a tendency to develop male features giving a bias towards male values. 14 This indicates that sex assessments must also take into account the age of the individuals on which the study is being performed. 15 Sexual dimorphism based on the morphological variation and size of virtually all postcranial elements has been examined in numerous studies There are many published examples of these, such as the comparative analyses of the humerus in Chinese, Japanese and Thai populations by Iscan et al., 16 from contemporary Crete by Kranioti and Michalodimitrakis 17 and from a Guatemalan rural population by Frutos. 18 Other long bones were studied in contemporary Japanese by Sakaue, 19 in South African whites by Steyn and Iscan, 20 in European and African Americans from the Terry collection by Iscan and Miller-Shaivitz 21 and in Munich and Cologne by Mall et al.. 22 In addition, the talus has been studied among prehistoric Polynesians by Murphy. 23 The significant degree of skeletal variation between populations necessitates that the population on which the sex determination is based needs to be quite specific, not only with regard to ancestry but also the time period in which it is placed. Environmental factors and cultural practices affecting levels of nutrition can influence sexual dimorphism. 24 Physical activity can also change the size and morphology of postcranial elements and result in an alteration in sexual dimorphism. 25 For example, sex differences in femoral shafts, with a greater cross-sectional diameter in males, were related to sexspecific cultural activity (long distance running) in the Pecos Pueblo sample investigated 13 Meindl et al. 1985, p Meindl et al. 1985, p Meindl et al. 1985, p Iscan et al. 1998, pp Kranioti and Michalodimitrakis 2009, pp Frutos 2005, pp Sakaue 2004, pp Steyn and Iscan 1997, pp Iscan and Miller-Shaivitz 1984, pp Mall et al. 2001, pp Murphy 2002, pp Gray and Wolfe 1980, pp Ruff and Hayes 1983, p. 394.

17 17 by Ruff and Hayes. 26 Societies with poor protein consumption have been shown to have a lower mean male height resulting in reduced sexual dimorphism, whereas societies with a secure food supply had a greater degree of sexual dimorphism. 27 Polygamous societies with greater competition between males had an increased mean male height. 28 Those in cold climates had lower mean male statures and reduced sexual dimorphism. 29 Where the research material is from a known population, from which the parameters determining sex have previously been established, a comparison with the material being analysed can accurately establish the sex but these criteria are not transferable to another population. Reference Collections Many of the techniques used for estimating sex were developed from museum skeletal collections of known sex. The most notable are the Terry collection at the Smithsonian Institution in Washington D.C. and the Hamann-Todd Collection in Cleveland, Ohio, in the USA. The Terry Collection predominantly came from university dissection specimens and comprised older individuals collected during the 20 th century. It is predominantly male and consists of lower class individuals, mainly from St. Louis, Missiouri. 30 The Hamann-Todd collection housed in the Cleveland Museum of Natural History came from local hospitals and medical schools and was collected between 1912 and The collection comprises 3,100 individuals who had birth dates from 1850 to Most were from the lowest social class, but some were from the middle classes. These collections have been extensively used to provide parameters for sex determination but as both represent individuals from the early 20 th century they may not be applicable to other populations from different time periods and different geographic areas. 32 Data bases of modern samples of skeletal remains have also been established, such as The Forensic 26 Ruff and Hayes 1983, pp Gray and Wolfe 1980, p Gray and Wolfe 1980, p Gray and Wolfe 1980, p Holland 1986, p Ousley and Jantz 1998, p Ubelaker 2000, p. 46.

18 18 Data Base in Tennessee. This has over 1,300 cases from 48 of the US states with a large proportion having documented sex and ancestry. 33 Observer Error Intra-observer errors (observations repeated over time by a single observer) and interobserver errors (observations repeated by multiple observers) can be significant and impact on the credibility of the final results. For example, when evaluating the distal humerus by visual assessment Rogers found the inter-observer error for sex determination to be an unacceptable 25%. 34 This was attributed to lack of observer experience and a failure to define precisely the scoring criteria. 35 These problems can be reduced by having adequately trained observers and where possible using standardised scores and visual templates. The precision of metric analysis is not as dependent on observer experience but is greatly improved by having well-defined measurements and accurately calibrated instruments. Preservation and Recovery The ability to recover human remains and determine the sex depends to a great extent depends on their state of preservation. Bone degradation in archaeological sites is determined by the amount of ground water and the ph of the soil with wet, acid conditions promoting bone destruction. Physical factors such as the depth and type of burial and other aspects such as secondary burials, looting and other human and animal activities also affect the condition and durability of the skeletal remains. The physical nature of particular bones will also determine which skeletal elements will survive. Willey et al. studied limb bones to assess their survival based on their mineral density. 36 They compared the bone mineral density of ancient skeletal material from the 33 Ousley and Jantz 2008, pp Rogers 2006, p Rogers 2006, p Willey et al. 1997, pp

19 19 14 th century massacre victims from Crow Creek, South Dakota, with contemporary remains from the Anthropology Department of the University of Tennessee. They found that the level of bone mineral density of skeletal elements from the modern sample was consistent with those from the pre historic site. They concluded that those bone elements with a higher mineral content were more likely to be recovered. The mid-shaft regions of the long bones were the most durable with a 90 per cent recovery rate, with the proximal segments recovered in 80 per cent of cases, but the distal segments were less durable with only the distal humerus reaching a recovery rate of 70 per cent. 37 The greater bone density of some smaller bones increases the likelihood of their recovery compared with many of the larger less dense skeletal elements. This particularly applies to the calcaneus and talus. When investigating a pre-historic American population Wilbur found that these were present in about 70 per cent of skeletal individuals in a suitable state to be measured and potentially used for sex determination. 38 The relative survival of elements of the human skeleton was also assessed by Waldron at the Romano-British site in London. 39 He calculated the percentage of bones present from those that would have been expected if the skeletons had been complete. The post cranial components which had at least one side recovered in at least 50 per cent of the individual skeletons included the glenoid cavity of the scapula, the proximal and distal segments of the humerus and radius and the proximal ulna, the proximal and distal femur, the distal tibia and the complete talus. 40 The survival of individual bones was quantified by Buikstra and Ubelaker who produced an inventory for commingled bones according to their completeness with cranial bones and small elements recorded separately, vertebrae by segments and long bones with epiphyses and each third of diaphysis recorded. The degree of completeness is then 37 Wlley et al. 1997, pp Wilbur 1998, p Waldron 1987, p Waldron 1987, Table 6.1.

20 20 classified as 1=>75%, 2=75-25% and 3=<25%, 41 which indicates the skeletal elements available for sex estimation. Statistical Analysis of Ancient Remains The type of statistical analysis which can be applied to ancient skeletal remains to allocate sex depends on whether the remains are from a documented or undocumented population, whether cranial or pelvic material is available for assessment and whether the sample consists mainly of post cranial material. Statistical analysis of an undocumented population The statistical analysis to determine the sex of individuals within a population for which the sex has been documented from post-mortem reports, coffin plates or some other reliable source, or is morphologically very close to a documented population, can be based on the parameters previously calculated for skeletal elements of that population. In the case of an undocumented population, where there are no comparable individuals of known sex, the problem is more complex as it must rely on other methods which can be applied to the various skeletal measurements which have been obtained in order to estimate the sex. This involves calculating the section point for each skeletal element, that is the numerical point which is the cut-off separating the majority of males and females. The method available to determine this figure depends on whether the pelvis is available to provide a reliable estimate of the sex from which other parameters can be derived, or whether the sex must be estimated only from other post-cranial elements. In either case the sex allocation can only be approximate as there will be an overlap between males and females in the middle of the standard curve for skeletal dimensions (Fig. 1). 41 Buikstra and Ubelaker 1994, p. 209.

21 21 Section point Fig. 1. Graphical representation of section point 42 Section point determined from mid-point between male and female means when sex pre-determined from pelvis Because sex can be assigned with considerable accuracy from the bones of the pelvis (see Table 5) in situations where a significant number of individuals have pelvic bones present the sex of these can be accurately determined. The long bones associated with the pelves can be used as parameters for the determining the sex of the remaining skeletons. Ideally to provide a statistically significant result there should be at least thirty males and thirty females present in the sample being analysed. 43 This method for determining the sex for poorly preserved, fragmentary, ancient remains was employed by Black who studied the femoral circumference. 44 Only a small proportion of the sample was able to be sexed by pelvic criteria but their utilization enabled population specific sexed standards for the circumference of the femora to be determined. Calculating the means for the male and female femora in which the sex had been provided from the pelvic criteria provided this. The section point as being the midpoint between the male and female means was then established (Fig. 27). Using this section point the rest of the femora were then to allocate the sex of the individuals where the pelves were not present. 42 Albanese et al. 2005, p Martinand Pierce 1994, p Black 1978, pp

22 22 This achieved sex allocations consistent with that previously obtained from the pelvic morphology, of 85 per cent. 45 This method was also used for three prehistoric Native American populations analysed by Dittrick and Suchey. 46 They initially estimated the sex using pelvic criteria which they had previously determined were 99 per cent accurate in an adult population. This enabled the establishment of standards to analyse the long bones of the three Native American populations. Section point of post-cranial remains determined from the mean of the whole sample Albanese and his colleagues have developed an alternative method for sex determination in the absence of firm sex determining criteria. They formulated a method that can be used to calculate the section points of post cranial remains in the absence of pelvic bones. 47 In his method the mean of the whole sample for each criterion was determined and this was then used as the section point. The method is valid if both sexes are present in roughly equal proportions with a bimodal distribution of the measurements. The mean of the whole sample provides the section point for determining the sex to an unknown individual. 48 The method was tested on samples taken from three skeletal collections. Two of these were from Portugal, the Skeletal Collection from the University of Coimbra and the Museum of Natural History in Lisbon and the third sample was from the St.Thomas Anglican Church, Belleville, Canada. All were adults of documented sex with no pathological conditions. Several measurements of the humerus and femur were taken for the study. Sub-samples of different sizes were assessed and it was found that those which contained more than forty individuals gave consistently more accurate results for sex determination than those of lesser numbers. The number of each sex in the sample also needed to be fairly even with a sex ratio of one sex being not more than one 45 Black 1978, p Dittrick and Suchey 1986, pp Albanese et al. 2005, pp Albanese et al. 2005, p. 145.

23 23 and a half times the other to obtain consistently accurate results excluding situations such as a battlefield where one sex would predominate. 49 The spread of data for each sex had to show a relatively normal distribution. Preferably too, the means of the measurements for each sex had to be significantly different so that the region of overlap was minimal. The accuracy for correct sex allocation under these circumstances when compared to the documented sex was found to be per cent accurate. 50 Overall Section Point derived by combining means Another approach used by Albanese et al. for calculating the section point to separate males and females involved calculating the means of multiple measurements on the same skeletal element. Then, from the sum of these, they produced an overall mean used as the overall section point. This gave allocation accuracies of per cent even in the circumstances where the sex ratio was more than 1.5:1. 51 Multivariate or univariate analysis Multivariate analysis to determine the sex of skeletal remains is applicable when the material is in good condition and complete but when dealing with ancient remains where the elements available for assessment are incomplete and in poor condition this is often not possible. Dittrick and Suchey applied discriminant function analyses using nine measurements from both the humerus and femur for sex determination. This necessitated the presence of a suitable number of intact skeletal elements. 52 Unfortunately all of the skeletal elements required to enable the measurements to be taken to perform a multivariate discriminant function analysis cannot always be obtained for many of the specimens being investigated, particularly from an archaeological site. To deal with this 49 Albanese et al. 2005, p Albanese et al. 2005, p Albanese et al. 2005, p Dittrick and Suchey 1986, p. 5.

24 24 situation, in addition to their analysis using multiple variables, Dittrich and Suchey compared an analysis using single variables. They found that this did not give significantly inferior results. This demonstrated that the use of several different measurements, analysed singly is reliable and is the usually the most appropriate method when many of the skeletal elements are deficient. The above methods allow for greater flexibility in the determination of sex of human remains. They provided considerable accuracy even when the available skeletal material is fragmented, incomplete, does not belong to a large sample and is from an undocumented population Albanese et al. 2005, p. 149.

25 25 CHAPTER THREE Review of studies involving the determination of sex from human skeletal material Most of the elements of the human skeleton have been assessed by visual and metric means for their effectiveness in the determination of the sex but not all have been found to be useful. In this chapter the results obtained by numerous investigators have been examined to evaluate which morphological traits are likely to give the best outcome for human remains from archaeological sites, the focus of this study.. Many previous studies have been performed on modern forensic skeletal remains which are not necessarily applicable to ancient material. For this project particular attention is given to skeletal elements that have been recovered from historic burials. Sex determination using cranial elements. The examination of cranial traits has shown significant sexual dimorphism owing to the increased robusticity occurring in males, with a larger skull and more obvious muscle markings. By contradistinction, many females retain many of the pre-pubertal features such as small size and lightness with some of the juvenile frontal and temporal bossing also present. These features have enabled sex to be allocated correctly by visual assessment in up to 96 per cent of documented skeletal individuals. 54 The relative difference in cranial dimensions between the sexes has also enabled metric analysis to be useful with correct allocation in up to 89 per cent of cases. 55 Owing to the subjectivity associated with the visual assessment of the cranium and the need to obtain more consistent outcomes with less observer errors, a scoring system for some cranial features has been established. This provides a more objective approach and 54 Williams and Rogers 2006, p Giles and Elliot 1963, pp

26 26 also enables comparative studies to be performed (Fig. 2). 56 Five criteria have been used which demonstrate the differences in size, shape and robusticity between the sexes. 57 The criteria are classified from one to five, from (definitely) female to (definitely) male as follows: 1 = female, 2 = probably female, 3 = neutral (indeterminate), 4 = probably male, 5 = male. The nuchal crest (occipital protruberance) projects from the occipital bone and its size is representational of the musculature of the skull. In a gracile female it is smooth and scores 1, in a robust male it forms a massive hook and scores 5. The mastoid process varies significantly in its size and is scored according to its volume rather than its length. Its size can be related to the external auditory meatus. If it is small it reaches only a small distance below the external auditor meatus (scoring 1), whereas if it is large it is several times longer and wider than the external auditory meatus (scoring 5). The supraorbital margin, the upper border of the orbit, may be a sharp ridge in females (scoring 1) or thick and round in males (scoring 5). The supraorbital ridge is a rounded elevation over the inner aspect of each orbit with a prominent eminence in the midline, the glabella. At the least they are smooth with little prominence of the frontal area (scoring 1), but in males there may be a massive ridge in the midline and prominent supraorbital ridges laterally (scoring 5). The mental eminence is rounded and comes to a point in females with little projection (scoring 1). In males it is square with a large projection (scoring 5). 56 Buikstra 1994, Fig. 4, derived from Ascadi and Nemeskeri 1970, Fig Buikstra 1994, p.19.

27 27 Fig. 2. Scoring system for cranial features 58 The gonial angle (Fig.3) of the mandible formed by the junction of the horizontal ramus (the body) and the ascending (vertical) ramus is a useful dimorphic feature particularly as it provides an objective measurement. The angle is generally larger in females than males. 59 It has been quantified to be greater than 125 degrees in females and less than 125 degrees in males. 60 The gonial angle is affected by changes to the muscles of 58 Buikstra 1994, Fig. 4, derived from Ascadi and Nemeskeri 1970, Fig Jensen and Palling 1954, p Williams and Rogers 2006, p. 731.

28 28 mastication which follows the loss of molars particularly if the loss is uneven or complete, so does not accurately indicate the sex in that situation. 61 Fig.3. Gonial angle 62 Accuracy of cranial traits in sex determination The accuracy of various cranial traits to estimate sex has been assessed by Williams and Rogers who examined 21 cranio facial features. 63 Rogers had identified that although these features had been widely used their credibility had not previously been established. 64 The study sample was randomly chosen from the skeletal collection of the University of Tennessee from Caucasians of European descent born during the 20 th century. Fifty specimens were selected with equal numbers of males and females. The females had a mean age of 62.6 years and the males 51.9 years. They ranked the cranial traits by comparing the sex obtained by blind assessment with the known sex and by evaluating the traits on two separate occasions by the same observer and comparing the results. At least 80 per cent accuracy for sex determination was chosen as the standard to be attained as this was regarded as the minimum accuracy which could be obtained by cranial measurement 65 and an intra observer error of not greater than 10 per cent was selected as this value was consistent with previous studies Oettle et al. 2009, pp Bass 1987, Fig Williams and Rogers 2006, p Rogers 2005, pp Giles and Elliot P Molto 1979, pp

29 29 Only six traits achieved the desired level of accuracy of at least 80 per cent for sex allocation and an intra-observer error of 10 per cent or less (Table 1). Table 1. Ranked cranial traits according to their accuracy. 67 Mastoid 92% Supraorbital ridge 88% Size and architecture of cranium 90% Zygomatic extension 83% Nasal aperture 83% Mandible-gonial angle 83% Williams and Rogers found the accuracy when based on a combination of these six traits was 94 per cent. They cautioned that as these criteria reflect robusticity, which is population specific, further investigations with specimens of different ancestry would be necessary to demonstrate the reliability of these criteria in assessing other populations. 68 Metric Analysis of Crania To avoid the subjectivity associated with visual assessment of cranial traits Giles and Elliot used ten cranial measurements in discriminant function analyses to assign the sex of individuals. 69 Three hundred specimens, comprising 75 male and female European Americans and African Americans of known sex and age were selected from the Terry collection in Washington and the Todd collection in Cleveland. The individuals were between 21 and 75 years of age and lived during the 19 th century and early 20 th centuries. 67 Williams and Rogers 2006, Table Williams and Rogers 2006, p Giles and Elliot 1963, pp

30 30 The following measurements were used: 70 Glabello-occipital length Maximum width Basion-bregma height Maximum diameter, bizygomatic Prosthion nasion height Basion-nasion Nasal breadth Palate-external breadth Opisthion-forehead length Mastoid length The discriminant function analysis achieved accuracy for sex determination of per cent. Giles and Elliott claimed was that while method gave an equivalent accuracy to visual examination, it had the added advantage that it did not require any specialised knowledge or experience of the cranial traits to achieve this result. A craniometric approach for sex determination has been used more recently by Wright who developed the CRANID software program. 71 The primary use of CRANID is to identify the geographical origin of unknown crania. Twenty nine cranial measurements (Table 3) are used to match an unknown cranium with a database of over 3000 crania of known sex from 74 worldwide geographic regions. Although a complete set of measurements is required for a full analysis, an estimate of sex can be performed with incomplete crania. A close match with a sex specific region provides an indirect assessment of sex. The cranial measurements used in the CRANID program are shown in Table Giles and Elliot 1963, pp Wright 2009, pp

31 31 Table 2. CRANID measurements 1. Max Cranial Length GOL 2. Nasio-occipital length NOL 3. Cranial Base length BNL 4. Basion-bregma height BBH 5. Max. Cranial Breadth XCB 6. Max. Frontal Breadth XFB 7. Biauricular breadth AUB 8. Biasterionic breadth ASB 9. Basion-prosthion length BPL 10. Upper Facial Height NPH 11. Nasal height. NLH 12.Orbital height OBH 13.Orbital breadth OBB 14. Bijugal breadth JUB 15. Nasal breadth NLB 16. Palate breadth MAB 17. Bimaxillary breadth ZMB 18. Zygomaxillary subtense SSS 19 Bifrontal breadth FMB 20. Nasion frontal subtense NAS 21. Biorbital breadth EKB 22. Interorbital breadth DKB 23. Cheek height WMH 24. Frontal chord FRC 25. Nasion-bregma subtense FRS 26. Parietal chord PAC 27. Bregma-lambda subtense PAS 28. Occipital chord OCC 29. Lambda-opisthion subtense OCS

32 32 Holland used metric analysis of the cranial base to develop a technique which could be applied to estimate the sex of fragmented crania (Fig. 4). 72 The sample was taken from the Terry collection and consisted of 50 males and 50 females with equal numbers of European Americans and African Americans between the ages of 20 and 50 years. 73 Nine measurements were taken from the skull as shown in Fig Length of occipital condyle (MLC) Width of occipital condyle (MWC) Minimum distance between condyles (MnD) Bicondylar breadth (BcB) Maximum interior distance between condyles (MxID) Length of foramen magnum (LFM) Width of foramen magnum (WFM) Length of basilar process (mid point of anterior margin of foramen magnum to basilar suture) Distance between post condyloid foramen (DF) 72 Holland 1986, pp Holland 1986, p Holland 1986, p. 204.

33 33 Fig. 4. Representation of the condylar region of the base of skull indicating the location of measurements. 75 This technique correctly allocated sex in per cent of cases, and, although it was less accurate than the method of Giles and Elliot, it may be of use in specimens where the cranium is fragmentary or deformed and only the base of the skull is intact Holland 1986, Fig Holland 1986, p. 205.

34 34 Mastoid Process Metric analysis of the mastoid process has been shown to be possibly superior to visual assessment in determination of sex. 77 Measurement of the mastoid height (Fig. 5), which is the length of the mastoid process below and perpendicular to the eye ear plane, was shown to be more accurate for sex allocation than visual methods. The exception is the visual assessment of the supramastoid crest (the zygomatic extension), which proved to be more accurate in females. 78 The supramastoid crest is a raised area of bone that forms the posterior root of the zygomatic process and was scored by Bernard on a scale from -2 (very faint - hyperfemale) to +2 (very prominent - hypermale). 79 The analysis of the mastoid is important because of its durability in situations where the available skeletal material is fragmented and incomplete. Fig. 5. Method of mastoid measurement 80 The cranium can provide useful information for sex allocation, but it does have limitations when applied to ancient remains from an undocumented population because 77 Bernard 2006, pp Bernard 2006, p Bernard 2006, p Giles and Elliot 1967, p. 58.

35 35 both the visual as well as the metric criteria being related to robusticity are population specific. The visual assessment of morphological traits is also subjective. Reliant largely on experience this issue has been overcome to some extent by the use of templates of morphological variations for the commonly assessed traits (Fig.1). The mastoid and the supraorbital margins have been ranked highly for accuracy by Rogers and both can be assessed using the template. They have the additional advantage that being composed of compact bone are frequently recovered from ancient burials. The mastoid process was recovered from 60 per cent of specimens from the Roman British site investigated by Waldron whereas intact crania were only present in 17 per cent. 81 The fragmented state of most crania limits the value of metric assessment although measurements of the mastoid process and base of the skull may be able to be used. Sex Determination with Post cranial Skeletal Material Pelvic sex determinants. The pelvic girdle provides many features which can accurately determine sex but its fragility means that it is not commonly recovered intact at ancient burial sites. The pubic bone in particular is vulnerable especially when it lies uppermost in a burial in the supine position. Breakages often occur in the superior and inferior pubic rami so many of the features of the anterior pelvis are often not available for assessment. The relative frequency of the survival of individual pelvic bones in material obtained from a Romano British site in London has been documented by Waldron. 82 The burial site was dated from the second to the fourth centuries AD and contained 112 burials with 88 adults. It was found that the acetabulum, the greater sciatic notch and the auricular surface were present in over 60% of adults in contrast to the pubic symphysis which was present in only 30% Waldron 1987, p Waldron 1987, p Waldron 1987, p. 60.

36 36 There are several morphological differences between male and female pelves related to childbirth which provide useful criteria for sex determination. Phenice studied these differences and described a method of sexing the pelvis. He used the ventral arc (a ridge on the anterior surface of the pubic bone mostly only found in females), the subpubic concavity (a lateral curve of the ischiopubic ramus which widens the pelvic outlet) and the medial edge of the ischio-pubic ramus at the outlet which is a sharp ridge below the pubic symphysis in females and a broad, rounded surface in males (Fig. 6). 84 Phenice claimed that his method was not subjective, was rapid and could be used accurately without requiring a great deal of observer experience. The three criteria were tested by Phenice on 275 adult individuals of known sex from the Terry Skeletal Collection. There were 72 African Americans (20 males, 52 females) and 203 European Americans (160 males and 43 females). It was found that 264 (96 per cent) had been correctly allocated. When the ethnic background and sex of the individuals was taken into account the accuracy for sex allocation was still at least 95 per cent. 85 Where at least one or two of the criteria definitely indicated male or female the sex determination was correct in at least 96 per cent of individuals. 86 Lovell found that when used with a series of older individuals the accuracy in determining sex was reduced to 83 per cent Phenice 1967, pp Phenice 1967, p Phenice 1967, p Lovell 1989, p. 119.

37 37 Fig. 6. Sex differences in the pubic bone. 88 Female Male Ventral arc (1 2) Subpubic concavity (3) Ischio-pubic ramus (4 5) 88 Phenice 1967, Fig. 1.

38 38 The greater sciatic notch in females is wide, usually forming an angle of about 60 degrees. In males the notch is narrower with an angle of about 30 degrees. 89 Walker developed a scoring system based on the morphology of the greater sciatic notch and provided drawings based on the range of variation between extreme males and extreme females (Fig. 7). 90 This aimed to reduce the subjectivity associated with purely visual assessment. A series of pelves from the Hamann-Todd Collection, the Terry Collection and the St. Bride s Church Collection in London were then evaluated. The first two collections represented relatively recent populations, whereas the St. Bride s sample belonged to a population from All were adults with documented sex and age at death. It would be expected that in a five point scale a score of three would indicate an indeterminate result. The results, however, showed that if the scoring system indicated individuals had a score of one, 90% were female and when the score was three 90 per cent were males. A score of two was indeterminate although it was closer to the female morphology. There were no significant differences between the African Americans and European Americans but the earlier English population from St.Bride s Church demonstrated females with scores of one in greater frequency and the males also had lower scores of one or two more often than the Americans. 92 Age differences were also noted with, both males and females of less than 50 years having a more feminine sciatic notch morphology but older males in particular tended to show more masculinisation with narrower greater sciatic notches. 93 Overall this method of assessing the greater sciatic notch accurately determined sex in 80 per cent of individuals Uberlaker 1978, p Walker 2005, Fig Walker 2005, p Walker 2005, p Walker 2005, p Walker 2005, p See also Bruzek 2002, pp

39 39 Fig. 7. Sex Differences in the Greater Sciatic Notch. Drawing by P.Walker. 95 Greater sciatic notch Female Male The subpubic angle (Fig. 8) is less than a right angle in males and greater than a right angle in females. 96 This has been ranked on a five point scale with a strongly obtuse rounded angle being definitely female, a right angle being indeterminate and a strongly acute A-form being definitely male. 97 Fig. 8. Subpubic angle 98 Male Female 95 Walker 2005, p Workshop of European anthropologists 1980, p Workshop of European Anthropologists 1980, p Bass 1987, Fig

40 40 The pubis is longer in females and this feature has been utilized by Washburn, combining it with the length of the ischium, to calculate the ischiopubic index. 99 He studied 300 adult skeletons from the Hamann Museum of Anatomy of the Western Reserve University which included 100 European-American males and females and 50 African American males and females. The index measures the relative lengths of the pubis (A C) and ischium (A B) from the point where they meet at the acetabulum where there is an irregularity or a notch (Fig. 9). 100 The index is calculated by Length of Pubis x 100/ Length of Ischium. It is an objective observation which Washburn found to be 15 per cent larger in females and could accurately determine the sex in over 90 per cent of individuals. 101 Fig. 9. Landmarks for determining sex of innominate 102 The pelvic inlet is broader and elliptical in shape in females and narrower and heartshaped in males. There is a considerable variation however, with a significant number of male and female pelves not conforming to their typical shape. 103 Young and Ince 99 Washburn 1948, p Bass 1987, p Washburn 1948, p Bass 1987, Fig Coleman 1969, p

41 41 performed radiological studies on the pelves of 500 females and 50 males and found 14 per cent of female pelves had male characteristics but also some male pelves had the female shape. 104 The true pelvis, defined as that area of the pelvis below the arcuate line, is deeper and narrower in males and shallower and wider in females. The acetabulum is relatively larger in males than females in order to articulate with the larger femoral head. This aspect has been adapted by Kelley to formulate the sciatic notch / acetabulum index which can be used for sex determination regardless of the pelvis being fragmented. 105 The index is calculated by Greater Sciatic Notch Width / Vertical Diameter of Acetabulum x 100 (Fig. 10). The cut off point was 87 for European Americans and Afro-Americans and 86 for Native Americans with the accuracy for sex determination of at least 90 per cent accuracy. Fig.10. Measurement of greater sciatic notch and acetabulum 106 A B greater sciatic notch width C D vertical diameter of acetabulum 104 Young and Ince 1940, p Kelley 1979, pp Kelley 1979

42 42 The muscle markings are more prominent in males, in keeping with their greater muscle mass. This is best assessed by comparison with skeletal material from the same population in which the sex has been reliably determined. It is not regarded as a reliable criterion but may be useful as a contributory feature (see Table 5). The sacral shape shows considerable sexual dimorphism and can be a useful indicator of sex. It is relatively longer and narrower in males and shorter and broader in females in keeping with their wider, shallower pelves. It is also more curved in males and flatter in females. 107 The auricular surface of the ilium, which articulates with the sacrum, is often elevated in females contributing to their wider pelvis. This is mainly evident at the posterosuperior margin of the articular surface. 108 This area is generally flat in males (Fig. 11). The pre-auricular sulcus is an elongated depression between the sciatic notch and the sacroiliac joint. It is most commonly present in females. It is rarely present in males and if so, is much shallower than in females (Fig. 11). 109 Fig. 11. Auricular area and pre-auricular sulcus Bass 1987, p Stewart 1979, p Ubelaker 1978, p Ubelaker 1978, p. 43.

43 43 The Obturator foramen shows significant dimorphism, being large and ovoid in males and small and triangular in females. 111 Brothwell regarded this feature to be only of minor value. 112 Bierry et al. however, using a quantitative radiological method to assess the shape of the obturator fossa, confirmed the differences in shape and were able to assign correctly sex using this trait in over 80 per cent of CT examinations from the University Hospital in Strasbourg. 113 Accuracy of pelvic criteria in sex determination Rogers and Saunders assessed 17 morphological pelvic traits from a sample of 49 adults excavated from a 19 th century cemetery in Belleville, Canada. The individuals which selected for evaluation had intact pelves and their sex had been documented from coffin plates. They ranked the traits according to the accuracy in sex determination and the degree of intraobserver error (precision) (Table 3). 114 The accuracy was evaluated by comparing the results for sex allocation with the known sex. Precision was determined by the percentage of cases which had their sex allocation reversed in a second assessment of a trial sample. This method does produce some contradictory results, notably that for sacral segments which had an accuracy of only 6.1 per cent (ranked 17 th for accuracy) but with a high precision for re-assessment of 3.2 per cent (ranked 5 th for precision) giving an overall ranking above pelvic inlet, ischio-pubic ramus and auricular surface. The first six traits had intra-observer errors below 5 per cent. Four traits, acetabulum size and shape, auricular surface height, pre-auricular sulcus and ischiopubic ramus shape had errors above 10 per cent which indicated that these features were difficult to evaluate. 115 Combinations of traits produced the most accurate results with obturator foramen shape combined with ventral arc or with true pelvis shape producing accuracy for sex allocation of 98 per cent. 111 Brothwell 1981, p Brothwell 1981, p Bierry et al. 2010, pp Rogers and Saunders 1994, Table Rogers and Saunders 1993, p

44 44 Table 3. Accuracy levels for pelvic traits Accuracy Precision Ranking Ventral arc Obturator foramen % 2 True pelvis Sacrum shape % 4 Subpubic concavity % 5 Pubis shape % 6 Muscle markings Dorsal pitting Acetabulum % 9 Preauricular sulcus % 10 Sacrum (posterior) % 11 Sciatic notch % 12 Ilium shape % 13 Sacral segments % 14 Pelvic inlet % 15 Ischiopubic ramus % 16 Auricular surface % 17

45 45. Because the pelvis directly reflects the functional differences between the sexes it is the most valuable skeletal element for sex allocation in undocumented specimens. 116 Unfortunately elements of the anterior pelvis are not often recovered in ancient specimens so many of the criteria from this area are frequently not available assessment. 117 This includes the traits described by Phenice as well as five of the six most accurate criteria noted by Rogers and Saunders (Table 3). Owing to this, criteria from other regions of the pelvis must be taken into account when allocating sex, particularly the greater sciatic notch, frequently recovered from historic burials and has been shown to be a reliable trait. Metric analysis of post-cranial elements Numerous workers have studied most post-cranial elements to determine their efficacy in sex allocation. The results of many of these have been tabulated to enable comparisons to be made regarding the reliability of metric analysis of the various morphological criteria in determining sex and in particular an assessment as to their suitability for allocating the sex of ancient remains. Special attention is given to criteria associated with joints because proximal as well as distal joint measurements are shown to be consistently more accurate in determining sex as compared to long bone lengths, shaft diameters and circumferences. 120 In addition, due to the compact nature of the bone in the region of joints these sections are more frequently recovered from ancient burials. 121 Scapula 116 Mays 1998, p Waldron 1987, p Waldron 1987, p Walker 2005, p Albanese et al. 2005, p Waldron 1987, p.59.

46 46 The glenoid cavity is a compact skeletal element often recovered from ancient burials. 122 It has been shown to be useful for sex allocation by Stewart who found that a glenoid cavity length of 36mm separated the sexes in a high percentage of cases (the exact figure was not given). 123 This finding was based on measurements taken from scapulae of the Terry Collection. The landmarks used for the measurement are shown in Fig. 12. Fig.12. Landmarks for measuring glenoid cavity length 124 Humerus The sexual dimorphism of the humerus has been assessed by various investigators. Many morphological features have been examined and the reliability of each criterion for sex allocation evaluated. These include: (a) Frutos 125 who examined forensic specimens from rural Guatemala, (b) Mall et al. 126 who looked at contemporary skeletal material from Cologne and Munich, (c) Sakaue 127 examined modern Japanese, 122 Waldron 1987, p Stewart 1979, pp Bass 1995, p Frutos 2004, p Mall 2001, p Sakaue 2004, p. 77.

47 47 (d) Steyn and Iscan 128 investigated skeletons from South Africa dating from 1863 to 1963, (e) Iscan et al. 129 focused on 20 th century samples from Thailand, China and Japan, (f) Dittrick and Suchey 130 studied the remains from Central California belonging to three cultural groups from 2500 BC, BC and after 500 AD, (g) Finally Kranioti 131 investigated contemporary specimens from Crete and France 132 European Americans, African Americans and Native Americans from the Tennessee Data Bank. These results have been tabulated in Table 4. They indicate that measurements of the humeral head, particularly the vertical diameter, minimum mid-shaft diameter and also the distal humerus, notably the epicondylar breadth, when used as single variables have accuracy for sex allocation of at least 80 per cent. These criteria have been used as a basis for discriminant function analysis which has produced an even greater accuracy. Kranioti found discriminant function analysis involving maximum length, head vertical diameter, minimum midshaft diameter and epicondylar breadth gave an accuracy of 91.1 per cent. 133 Similarly measurements of humeri from the Tennessee Data Bank involving bicondylar breadth, head diameter and midshaft minimum diameter were correct for sex estimation in 96 per cent of cases. 134 The most useful diameters for sex discrimination are shown in Fig. 13. The samples studied by Mall, Sakaue, Kranioti, Stein and Iscan were all of documented sex whereas those examined by Frutos had their sex estimated from the pelvis and cranium and those of Dittrich and Suchey from the pubis. These studies have confirmed that the most useful criteria for sex allocation of the humerus are those proximally, associated with the shoulder joint involving the head of the humerus and distally, associated with the elbow particularly the epicondylar breadth and distal articular breadth. These have been shown to allocate correctly sex in up to 90% of individuals. 128 Steyn and Iscan 1999, p Iscan et al. 1998, p Dittrick and Suchey 1986, p Kranioti 2009, p France 1997, pp Kranioti 2009, p France 1997, p. 169.

48 48 Table 4. Humerus: Accuracy of Sex Determination (%) Author Frutos Mall Sakaue Steyn Kranioti k k Dittrick and Suchey Iscan France white black Early Mid/Late AA EA NA* maximum length maximum head diameter vertical head diameter tranverse head diameter maximum midshaft diameter minimum midshaft diameter circumference midshaft epicondylar breadth distal articular breadth trochlear width 70 sagittal diameter trochlear. 89 med-lat diameter distal epiphysis 94 vertical diameter prox. epiph. 70 med-lat diameter prox. Epiph 88 deltoid circumference deltoid diameter bicondylar width least circumference * AA=Afro-Americans EA= Euro-Americans NA= Native Americans

49 49 Fig. 13. Measurements of Humerus. I I A-B Maximum length C-D Maximum Diameter of Head M-N Maximum diameter of midshaft S-T Minimum diameter of midshaft M-N Circumference of midshaft Z Least circumference of shaft I I Epicondylar Breadth

50 50 Radius The head of the radius has considerable sexual dimorphism shown by the measurements of the maximum and minimum diameters. Berrizbeitia demonstrated an accuracy of over 90 per cent using these parameters in specimens from the Terry collection. 135 The specimens were more likely to be female if the minimum diameter of the radial head was equal or less than 21mm and the maximum diameter 22mm. The individual is more likely male if the minimum diameter was 22mm or more and the maximum diameter 23mm or more. These results indicate that as the head of the radius is preserved in over 50% of historic burials 136 it is a useful sex discriminant. Dimorphic features associated with the elbow joint The carrying angle is the angle at the elbow joint between the upper and lower arm when the arm is fully extended and the hand supinated (Fig.14). This angle is about ten degrees greater in females than males and because of this variation, morphological differences are present in the distal humerus and proximal ulna. The elements of the elbow have been subjected to both quantitative and qualitative assessment for sex determination in previous research. Rogers has researched the distal humerus using visual criteria. 137 These studies involved modern adult skeletal material from the University of New Mexico collection (UNM) and the William Bass collection at the University of Tennessee (UTK), th century adults from the Christ Church, Spitalfields collection in London and th century adults from Portugal 139 as well as adolescent skeletons also from the Spitalfields collection. 140 Four visual criteria related to posterior aspect of the distal humerus including the morphology of the trochlear (Fig.15), the depth 135 Berrizbeitia 1989, 136 Waldron 1987, p Rogers 1999, pp ; 2006, pp ; 2009, pp Rogers 1999, pp Rogers 2006, pp Rogers 2009, pp

51 51 and shape of the olecranon fossa (Fig.16) and the angulation of the medial epicondyle. A summary of the results of the adult studies is shown in Table 5. Table 5. Accuracy of sex determination of distal humerus, modern 141 and historic 142 samples. Features 20th C 20th C 18 19th C 19 20th C adults adults adults adults (UNM) (UTK) (London) (Lisbon) Olecranon fossa (depth/shape) 91% 82% 63% 78% Angle of medial epicondyle 86% 86% 67% 66% Trochlear constriction 74% 88% 55% 64% Trochlear symmetry 74% 69% 67% 66% The individual traits were considerably less accurate for sex allocation in the historic samples compared to the modern collections. A combination of the criteria from the modern samples produced accuracy for sex determination of over 90 per cent, with the olecranon used as a guide in the 2% of indeterminate specimens. 143 The historic samples, on the other hand, had a combined result of about 80 per cent which included up to 20 per cent of indeterminate cases Rogers 1999, p Rogers 2006, p Rogers 1999, p Rogers 2006, p. 232.

52 52 Fig. 14. Carrying angle at the elbow joint. 145 Fig. 15. Trochlear morphology Purkait and Chandra 2004, Fig Rogers 2006, p. 233.

53 53 Fig.16. Olecranon morphology 147 Another feature which may be useful is perforation of the olecranon fossa. It is considered to be more often present in females than males but Krogman and Iscan have suggested that this is not a particularly reliable trait. 148 This feature has been studied in specimens from the Museum of London. The details of this study are presented in the appendix and the results confirmed that perforation of the olecranon fossa is twice as common in females. Purkait and Chandra also investigated dimorphic features associated with the carrying angle. Their investigation was carried out using specimens of known sex from the Bhopal Medico-legal Institute with male and female adult ulna from middle class individuals with a mean age of 45 years. 149 They found the best results were related to the proximal ulna, particularly the olecranon-coronoid angle which was accurate for assigning sex in 85 per cent of individuals (Fig. 17). 150 It was assessed by measuring the angle made by the line joining the tips of the olecranon and coronoid processes (A-B) extended to join the line of the posterior surface (D). The angle was greater in males (mean 22 degrees) than females (mean 14 degrees) Rogers 2006, p Krogman and Iscan 1986, p Purkait and Chandra 2004, pp Purkait and Chandra 2004, p Purkait and Chandra 2004, p. 926.

54 54 Fig. 17. Olecranon-coronoid angle 152 These studies have shown that skeletal elements related to the elbow joint can be useful sex determinants. Rogers demonstrated its usefulness using visual traits from the distal humerus with modern individuals but the method was not as conclusive with historic specimens. Purkait and Chandra used intact ulnas to show that the olecranon-coronoid angle had significant sexual dimorphism but as only the proximal ulna is often the only section recovered from ancient remains this is of limited value. As a result of this finding a study was carried out using only the proximal ulna. The angle was not found to be significantly sexually dimorphic but the distance between the tips of the olecranon and coronoid processes (A-B in Fig.17) was found to have appreciable dimorphism. The details of this study are presented in the appendix. Because components of the elbow joint are often recovered from ancient burials other possibly useful criteria have also been investigated as part of this current research and included in the appendix. 152 Purkait and Chandra 2004, Fig. 2.

55 55 Femur Various components of femora have been studied to determine sex and have been found to show considerable accuracy, ranking third behind the pelvis and cranium (Fig.18). The accuracy for sex allocation is enhanced by the availability of multiple femoral variables which can be measured. Femur length shows considerable dimorphism with female femora being 94 per cent the length of male femora, but because there is a considerable overlap between the sexes this measurement is not as useful for sex differentiation as other criteria. The diameters of the femoral head have been found to be more effective with the mean transverse and vertical diameters for males being significantly larger than females in both the white and native populations in South Africa. 153 Steyn and Iscan confirmed that the accuracy for sex discrimination using the femoral head was 86 per cent, and using the breadth of the distal femur 91per cent. 154 The accuracy of femoral elements in determining sex by various authors has been tabulated in Table 6. Different population groups have been included: Steyn and Iscan 155 studied white South Africans of known sex, Safont et al. 156 used Late Roman material from Spain with the sex estimated from pelvic and cranial criteria, Dittrich and Suchey 157 investigated pre-historic remains from Central California with the sex derived from the pubis, Sakaue, 158 modern Japanese material of known sex, France, 159 sexed specimens from the Tennessee Data Bank, Dibennardo and Taylor 160 studied European American individuals from New York of documented sex and MacLaughlin and Bruce 161 looked at a pre-historic Scottish population, with the sex estimated from the pelvis and cranium. Black 162 researched a poorly preserved burial site in Ohio where the sex of a portion of the specimens was also estimated from the pelvis from which parameters to assess the femurs of the remainder were derived. 153 Asala 2001, pp Steyn and Iscan 1997, pp Steyn and Iscan 1997, p Safont et al. 2000, p Dittrick and Suchey 1986, p Sakaue 2004, p France 1997, pp DiBennardo and Taylor 1979, p MacLaughlin and Bruce 1985, p Black 1978, p. 229.

56 56 Fig Femoral measurements used for sex allocation A-B Maximum femur length C-D Bicondylar length (both distal condyles in contact with base) G H Bicondylar width S-T Antero-posterior diameter of midshaft M-N Medio-lateral diameter of midshaft M-N Circumference of midshaft Y-Z Antero-posterior subtrochanteric diameter W-X Medio-lateral subtrochanteric diameter E-F Maximum diameter of femoral head G H 163 Bass 1987, Fig

57 57 Table 6. Femur metrics: Accuracy in Sex estimation Femur metrics Author Steyn Safont Dittrich Sakaue France Macl'lin Black DiBennardo early midlate AA* EA** maximum length bicondylar length bicondylar width max.head diameter a-p head diam 91 med-lat head diam 88 a-p subtrochanteric diam 69 med-lat. subtroch. diam subtroch. Circumference 88 ant-post midshaft diam med-lat. midshaft diam midshaft circumference a-p med condyle 81 a-p lateral condyle 91 transverse diam med con 70 transverse diam lat con 88 midshaft area(cross-sect) 86 max. a-p diam shaft 90 * African Americans **European Americans Purkait developed another method for determining sex from the proximal femur by taking measurements related to areas of muscle attachment 164. The study was performed on 280 documented adult femora (200 males, 80 females) from Central India. It takes advantage of the larger, more robust muscle attachments of males (at points B, greater trochanter and C, lesser trochanter) as well the sex specific morphology in females of the femoral head related to their pelvic anatomy (point A, the most lateral point of articular margin) (Fig.19). 165 Using the measurements singly the accuracy for sex determination varied from 62 to 84 per cent, the length between the greater and lesser trochanters (BC) being 164 Purkait 2005, pp Purkait 2005, pp

58 58 the most accurate (84 per cent). He found that a combination of two or three of the measurements gave only marginally better results. 166 Brown, using documented specimens from the Terry collection confirmed that the main determinant in the analysis of the proximal femur was the diameter from the most medial point on the greater trochanter to the highest point on the lesser trochanter (B-C, Fig 19). By using this alone an accuracy of 85.5 per cent was achieved but by combining it with the vertical diameter of the femoral head this was improved to 90 per cent. 167 The advantage of these methods is that they can be used on fragmented specimens Figure 19. Purkait s Triangle 168 Albanese has sought to overcome the problem of the population specificity of metric methods by evaluating an area on the proximal femur related to the size and angle of the femoral neck. This is determined by the length of the pubic bone and therefore sex specific but not dependent on the overall robusticity and size of the individual and therefore not population specific. 169 The pubic bone is relatively longer in females as the result of changes in growth occurring at puberty which widen the pelvis. At the same time the relationship of the femora to the pelvis has to be maintained to maximize bipedal 166 Purkait 2005, pp Brown 2007, pp Purkait 2005, Fig Albanese et al. 2008, pp

59 59 movement. This results in the angle between the neck and shaft of the femur demonstrating significant sexual dimorphism with females having a longer femoral neck and the larger angle. 170 Where the angle is 40 degrees or less it is likely to be male (83 per cent chance), but if it is 50 degrees or more, it is probably female (75 per cent chance). 171 This study was performed on over 300 individuals from the Terry Collection after first being tested on a documented sample from the Grant Collection in Toronto. A widely varied population was included, the individuals only being defined by date of birth. This indicates that the method would be useful to allocate sex in situations where specimens are recovered without any population data being available. 172 The measurements of the sides of a triangle involving the head of the femur and the greater and lesser trochanters as well as the angles produced by this triangle were used to develop equations which when combined with the maximum diameter of the femoral head correctly assigned sex in over 90 per cent of cases. 173 A spread sheet is available to enable the calculations to be easily performed. 174 The measurements taken were from the greater trochanter (GT) to the fovea capitus (FC), the greater trochanter to the lesser trochanter (LT) and the lesser trochanter to the fovea capitus (Fig.20). The advantage of this method of sex determination, as compared with that of Purkait, is that size variation does not affect the result so it is not population specific Albanese et al. 2008, p Krogman and Iscan 1986, p Albanese et al. 2008, p Albanese et al. 2008, p Albanese et al. 2008, p ( 175 Albanese et al p

60 60 Fig. 20. Landmarks for measurements of size and angle of femoral neck 176 Dimorphic features of the proximal and distal femur The adaptation of the female pelvis to childbirth and the resulting larger distance between the hip joints compared to the male pelvis produces several dimorphic features which may potentially be useful for sex determination. These include a reduced oblique length, which is the length of the femur measured on an osteometric board when both femoral condyles are aligned with the end of the board (Figs. 21). There is also a greater angle of inclination of the femoral shaft and a reduced angle of torsion. In the female the lateral condyle is larger than medial condyle, the upper part of shaft is bowed and there is no retorsion of head (inclining posteriorly) whereas in males the medial condyle is the larger, the shaft of the femur is straight and retorsion of the head is present. 177 These differences are illustrated in Fig. 21. The different degrees of femoral torsion are shown in Fig. 22. These differences in femoral morphology between sexes would appear to have some potential to be exploited in sex differentiation particularly in situations where there is no known population for comparison. They are also useful when the remains are not present in numbers large enough to provide a statistically significant sample to enable differentiation based on skeletal measurements. 176 Albanese et al. 2008, Fig Walmsley 1933, p. 288.

61 61 Fig.21. Female femur 178 Male femur Walmsley 1933, Fig Walmsley 1933, Fig. 4.

62 62 Torsion of proximal femur Torsion (anterior twist of the proximal femur) is more marked in males, being 18 degrees as compared to 6 degrees in females in the example given by Walmsley. 180 As an alternative to measuring the angle directly the degree of torsion can be assessed by the difference in height of the cervical tubercle and the femoral head when the distal femur is resting on the femoral condyles (Fig.22). 181 Fig. 22. Femora with different degrees of torsion (indicated by the difference in level between the head and cervical tubercle) 182 These criteria were tested in a preliminary trial on skeletal material from the Museum of London. The angle of the neck of the femur, the angle of torsion and the inclination of the femoral shaft were evaluated but in this small series no significant sexual dimorphism was detected. The details of the study are presented in the appendix. 180 Walmsley 1933, Figs Bass 1987, p Bass 1987, Fig

63 63 It can be seen that the best results were those obtained from the measurements of the components associated with the hip and the knee joints, the femoral head and distal femur (marked in bold in the table). The accuracy for sex determination for criteria associated with these was of the order of 90 per cent accuracy. The midshaft circumference was ranked next in accuracy with the femoral length being generally less effective. As the proximal and distal segments of the femur are not only accurate criteria, but also recovered from over 50 per cent of ancient burials 183 they are significant sex determinants. Tibia The tibia is a robust bone which demonstrates significant sexual dimorphism and has been shown to have criteria which are effective for sex determination. The measurements taken from the tibia are shown below (Fig.23). Steyn and Iscan investigating a South African Caucasian population of known sex found that the distal epiphyseal breadth was the most effective for sex discrimination followed by the proximal breadth, the antero-posterior diameter, the circumference and the transverse diameter. 184 The results obtained by other workers including Sakaue, 185 using contemporary documented Japanese material, Iscan et al. 186 looking at 20 th century Chinese, Japanese and Thai samples of known sex, Holland 187 documented specimens from the Hamann Todd Collection, Slaus and Tomicic 188 investigated tibia from mediaeval Croatian sites with the sex based on pelvic and cranial morphology and Safont 189 examined specimens from the Late Roman period in Spain in whom the sex also had been estimated from pelvic and cranial criteria. The following table shows the accuracy obtained by several researchers using various tibial measurements (those of Slaus and Safont are estimates). (Table 7). 183 Waldron 1987, p Steyn and Iscan 1997, Sakaue 2004, p Iscan and Miller-Shaivitz 1984, p Holland 1991, p Slaus and Tomicic 2005, p Safont et al 2000, p. 323.

64 64 Fig. 23. Measurements of Tibia used for sex determination A-B Length S-T Antero-posterior diameter at nutrient foramen M-N Medio-lateral diameter at nutrient foramen M-N Circumference at nutrient foramen

65 65 Table 7. Tibial Metrics: Accuracy in Sex Determination (%) Tibia metrics Author Sakaue Iscan Holland Steyn Slaus Safont Length a-p diam lat artic surface 89 a-p diam med artic surface 84 transverse diam lat a. s. 92 transverse diam med a. s. 86 prox epiph breadth a-p diam distal epiph 86 transverse diam dist epiph max distal artic breadth 83 min distal artic breadth 59 distal artic breadth 61 mid-shaft area 91 Circumference 80 max prox artic surface 95 med prox condyl artic width med prox condyl artic length 92 lat prox cond artic width 93 circumference nutrient for min circumference 81 These figures confirm the findings of Holland who determined that measurements of the proximal tibia were very useful in sex determination. He attributed this to the heavy level of stress affecting the knee joint and as the amount of stress was presumed to be greater in males it produced significant sexual dimorphism. 190 The nutrient foramen is also a suitable feature from which to take measurements as it is easily located, even on tibiae in poor condition because this portion of the shaft is frequently recovered. 190 Holland 1991, p. 221.

66 66 Talus The talus would be expected to display significant sexual dimorphism as it has a weightbearing function proportional to the weight and size differential between the sexes. 191 It has been investigated as an indicator of sex by Steele researching the Terry Collection. 192 Discriminant function analysis involving the width, length and height of the talus produced 88 per cent accuracy for sex allocation. Single measurements were not useful except for the maximum length of the Talus which was taken as the distance between the sulcus for the flexor hallicus longus posteriorly and the most anterior point on the articular surface for the navicular. 193 Measuring the maximum length of the talus alone gave accuracy for sex designation of 80 per cent. 194 It was further tested on two Native American populations in which the sex was correctly assigned in 85 and 80 per cent respectively. 195 Murphy studying a pre-historic Maori population with the sex estimated by pelvic criteria confirmed that the single best measurement was the talus length. 196 Wilbur, who assessed talus length and width and talus trochlear length and width, investigated Native American collections in whom the sex had also been estimated by pelvic and cranial morphology. 197 Any of these measurements taken alone had an estimated correct allocation of over 82 per cent. Barrett et al. 198 found that talus length, width and height measurements were estimated to be about 85 per cent correct but preferred the talus volume, combining all three measurements, which he found to be a simpler and more reliable measurement. 191 Barret et al. p Steele 1976, pp Steele 1976, p Steele 1976, pp Steele 1976, p Murphy 2002, pp Wilbur 1998, pp Barrett et al. 2001, pp

67 67 Calcaneus Wilbur, investigating a prehistoric Native American population, sexed previously on pelvic and cranial criteria, found that adequate measurements of the talus and calcaneus could be taken for about 70 per cent of the individuals in the study. They were used in discriminant function analysis and demonstrated an estimated accuracy for sex allocation of greater than 87 per cent. 199 Using measurements of calcaneus body height, load arm width or maximum length Wilbur estimated the sex to be correct in greater than 83 per cent of cases. 200 Steele found that sex estimation using the calcaneus alone was 79 per cent, proving it to be less reliable than the talus Patella The patella has been studied for its capacity to be used for sex discrimination and has been found to have considerable sexual dimorphism. Introna et al. 202 studied a documented contemporary Italian population and found that one function incorporating maximum height and thickness demonstrated an accuracy for sex allocation of 84 per cent. Mahfouz et al. 203 used CT scanning to distinguish variations in patella morphology and produced data which was subjected to statistical analysis and accurately identified the sex in over 90 per cent of individuals. Sex determination from dentition There is considerable sexual dimorphism in tooth size, an effective indicator in the estimation of sex. Garn et al. found the canines and second molars have greatest dimorphism with the buccolingual diameter being particularly effective in their analysis. 204 A multivariate approach using six measurements on the canine and second 199 Wilbur 1998, p Wilbur 1998, p Steele 1976, p Introna et al. 1998, pp Mahfouz et al. 2007, pp Garn et al 1966, p

68 68 molar from the mandible gave a correct result in 82 per cent of previously sexed individuals. The mandibular canine was the tooth which singly gave the best result (73 per cent). 205 Dental analysis has also been found to be effective for sex allocation in subadults. Using discriminant function analysis results in up to 97 per cent of individuals were consistent with those obtained from pelvic criteria. 206 The variation in tooth diameter between males and females is only mm so observer error may be a factor in the estimation. 207 Sex determination in Sub-adults Determining the sex in subadults is difficult and according to Bass any attempt is no better than a guess 208. Other workers would not completely agree with this assessment although the degree of sexual dimorphism is minimal. 209 The determination of sex in juveniles is important for demographic studies to determine biases which may be evident in the status and care of children. It also influences the estimation of age in subadults as females grow more rapidly and mature earlier than males. 210 Males are on average larger than females until puberty. The bone age and dental age of males approximately correspond but the earlier maturation of females produces a differential with their bone age being in advance of their dental age. These differences have led to three methods of sex estimation in juveniles. Gleiser and Hunt studied Xrays of the teeth and hands of living children and found that the individual is likely to be female if the bone age is in advance of the dental age but male if they correspond. 211 In females the bone age is advanced by approximately one year between 5 10 years, two years from years and one year from years Garn et al 1979, pp Rosing pp Hillson1996, p Bass 1985, p Weaver 1998, pp Saunders 2008, p Hunt and Gleiser 1955, p Krogman and Iscan1986, p. 95.

69 69 It was suggested that when dealing with skeletal remains the dentition combined with the knee would be the best sites to study. 213 Sex differences in the foetal pelvis have been noted by Boucher. 214 She studied stillborn fetuses, measuring the width and depth of the sciatic notch, the ischial and pubic lengths and the sub-pubic angle. Significant sex differences were found in the sciatic notch dimensions and the sub-pubic angle. Schutkowski looked at children s skeletons from birth to five years of age and found that males had greater sciatic notches that were deeper and narrower than girls. 215 These findings were not confirmed in a later study by Vlak et al. on juveniles from birth to fifteen years indicating that this is not a reliable indicator of sex in this age-group. 216 Non-elevation of the auricular surface has been noted in males in the age groups from prenatal to six months of age. 217 Weaver analysed 153 fetal and infant skeletons from the Smithsonian Institute. He found that the auricular surface was not elevated in over 90 per cent of males in the prenatal and six month age groups and in 73 per cent in the newborn group. In females it was not as accurate with 75 per cent correct in the prenatal group but only 54 and 43 per cent in the newborns and those aged six months. Sex differentiation in subadults is very often not possible but if suitable skeletal remains are recovered then this is achievable for some individuals. There is no established method, however, by which this can be consistently performed Hunt and Gleiser 1955, p Boucher 1957, p Schutkowski 1993, p, Vlak et al 2008, p Weaver 1980, p Rosing 1983, p. 149.

70 70 Discussion An examination the pelvic bones and to a lesser extent the cranium have been shown to provide the greatest accuracy for the sex determination of human skeletal remains. Applying a combination of the most accurate criteria for the pelvis produced a correct sex allocation in 98% of individuals with an intra-observer error of less than 5% (see Table 3). The level of accuracy however is variable and is dependent on the preservation of the skeletal material and the experience of the observer. In the case of the cranium visual assessment using the best criteria was accurate in 94% of subjects with an intra-observer error of 10% (see Table 1). Metric analysis of the cranium by Giles and Elliott produced results of 82 89%. The sex determined by the pelvis has been used as the benchmark in situations where the population being studied is undocumented with no other population specific skeletal material available to provide parameters for comparison. Many of the papers which dealt with undocumented skeletal material and referred to in this thesis, obtained their initial sex allocation from pelves and used the terms accuracy, correct allocation and sex determination in their analyses. (for example Dittrich and Suchey(1986), Murphy (2002), Safont et al. (2000), Slaus and Tomicic (2005)). This terminology is not strictly correct as the results should be qualified by stating their consistency with sex allocation obtained by pelvic assessment rather than true accuracy which can only be obtained from a documented source. It has been demonstrated that skeletal elements associated with joints have shown the most sexual dimorphism due to the greater stresses applied to them with physical activity, presumed to be greater in males, which accentuates the dimorphism associated with their normal greater robusticity. Criteria related to these have proven to be the most reliable

71 71 for sex allocation. 219 This has been confirmed from the published research material with measurements of the humerus involving the head diameter and epicondylar breadth, the femoral head and bicondylar breadth and the distal epiphyseal breadth of the tibia being prominent in their accuracy in sex allocation (Tables 4, 6, 7). Further investigations have been carried out for this study involving the proximal femur and components of the elbow joint to determine their value in determining sex. These elements were chosen as they are often recovered from ancient burials where other material is either not present or not suitable for analysis. The material examined was at the Museum of London, Osteology Department. The results for the femur involved the degree of torsion of the femoral head, the bicondylar angle and the angle of the femoral neck did not reveal significant dimorphism. By contrast, the results for the elbow components indicated that as well as the epicondylar breadth and the maximum diameter of the head of the radius, other criteria related to the elbow joint including the depth of the olecranon fossa and the distance between the tips of the olecranon and coronoid processes of the ulna showed significant dimorphism and could be potentially be usefully included in the assessment of ancient remains particularly in the absence of pelvic or cranial material. The details of these investigations are presented in the appendix. The skeletal remains recovered from an ancient archaeological site do not always provide the number of individual elements which would be normally regarded as statistically significant. It has been indicated (above) that forty specimens are necessary to obtain significant results although the scatter plot produced by Albanese where the accuracy for sex determination is assessed with respect to the sex ratio (with one sex up to double the other) and sample size (varying from less than 20 to more than 79) indicates that this is not necessarily correct (Fig. 24). The figures which were derived from data taken from the Coimbra Collection in Portugal show that even the samples with the lowest number of specimens (less than twenty, tabled on the right side of the graph) have accuracies at least approaching 80 per cent. 220 This would suggest that an acceptable result (accuracy 219 Albanese 2005, p Albanese 2005, 147.

72 72 for correct allocation of at least 80%) can be obtained in spite of the available samples being numerically relatively small when the skeletal criteria being assessed have significant sexual dimorphism. It should be highlighted that the accuracy of small samples is dependent on the criteria being assessed having significant sexual dimorphism otherwise spurious results may be obtained due to under sampling population diversity. Fig. 24. Scatter plot of accuracy by sex ratio and sample size for Coimbra Collection Data. 221 This proposition has been investigated as part of the current research by examining documented samples from the Museum of London. These confirmed that reliable results can be obtained with relatively numerically small samples when the criteria being examined exhibit significant sexual dimorphism. The details of this study is also presented in the appendix. The visual examination of morphology, being usually based on a scale of five gradations of characteristics, will show a middle grade in which the sex is indeterminate and the allocations on each side of the middle being possibly female or possibly male (e.g. for the cranium and sciatic notch). By contrast, using a metric analysis the numerical finite results make it much less likely to get an indeterminate result. This only occurs when the 221 Albanese et al. Fig. 2.

73 73 result exactly corresponds with the section point, which can usually be resolved by applying extra weighting to the most accurate criterion. In the case of metric analysis the overlap section of the bimodal curve (see Fig. 1) represents the indeterminate specimens (containing the larger females and smaller males). A degree of error in the sex estimation by metric analysis is inevitable owing to biological variation in human dimensions, even in the same population. The error in allocation is proportional to the area of the overlap in the bimodal curve and reduces as the difference between the male and female means increases.

74 74 Chapter Four Research Material Background Historical setting of the research population The focus of this study was to determine, as far as possible, the sex of ancient human remains recovered from the burial ground at Samtavro, in the Republic of Georgia (Fig. 25), during archaeological excavations carried out in 2008 and This was a part of an investigation into the physical anthropology and archaeology of the area carried out by a team from the University of Melbourne and the Georgian National Museum. Samtavro is on the periphery of Mtskheta, the former capital of Georgia. Samtavro was the main burial ground of Iberia, a kingdom that stretched across the east of Georgia from 300 BCE 580 CE. Iberia was part of the Graeco Roman culture with well established cities, sophisticated architecture and settled farm lands which have been described by Strabo. 222 He described two distinct populations, those on the plains were peaceful, practiced agriculture and had much in common with the Armenians to the south in contrast to the people in the surrounding mountains who were warriors and very similar to the Scythians in the north. In the region of Mtskheta there is archaeological evidence of a prosperous society during that period with dwellings with tiled roofs, bath-houses and a columned hall. Nearby Armaziskhevi was the site of elite burials with early burials in stone sarcophagi within a two-room mausoleum and associated with substantial funerary goods. Surrounding this site were significant residential buildings and a Roman bathhouse. 223 The main language of the elite was Greek as noted on inscriptions tombstones 222 Strabo Geog. III Braund 1994, p. 205.

75 75 and tiles produced locally. The Georgian language did not evolve until after the arrival of Christianity in the fourth century CE Braund 1994, pp

76 76 Fig. 25. Maps of Georgia, showing the location of Mtskheta, north of Tbilisi (a) and Samtavro (b). 225 During the third and fourth centuries BCE Roman influence in Iberia decreased and by the end of this period it was controlled by Persia. 226 Mtskheta was the principal city of Iberia, situated on the junction of major rivers from the north as well as from the east and west. The burials at Mtskheta attest to the rapidly expanding population from the first century CE. 227 Samtavro has over 3000 burials on a twenty hectare site dating from the second millennium BCE to the eighth century CE. 228 During this period Iberia was a stable and prosperous kingdom negotiating with the governments of Rome and Persia until it was absorbed into the Persian Empire. 229 It was a multicultural community with multiple ethnic origins particularly in evidence along the southern and northern borders, with numerous languages being spoken as noted by the Romans who required a large number of interpreters to assist them with their trading negotiations. 230 In spite of this evidence for a multi-ethnic population a morphological study of crania retrieved from previous excavations at Samtavro, which covered the period from the 5 th century BCE to the third century CE, show no marked difference in cranial features. 231 The crania are relatively long and narrow, of medium height with a narrow, fairly long face and a strongly projecting nose. This cranial type is described as being analogous to the Mediterranean branch of Europeans 232. It would appear that there has been at least to some degree, continuity in the population throughout the period from the Bronze and early Iron Age to the early medieval period. The tombs excavated during the present study date from the latter part of this era, from the first to the fifth centuries CE. Four different types of tombs have been recognized. Cist tombs lined by large stone slabs common in the fourth century, modest tile-lined tombs associated with the late 225 Sagona 2010, Fig Braund 1994, p Braund 1994, p Sagona et al., 2010, p Rapp 2003, pp Braund 1994, p Alekseev 1964, pp Alekseev 1964, p. 348.

77 77 Roman period, other modest tombs constructed of rectangular stone blocks and the poorest, comprising earthen pits with a stone or tile marker. 233 Examples of tile-lined tombs and stone cist tombs had radiocarbon readings of AD and AD respectively. 234 Evidence of secondary burials was present with remains being commingled and in addition, some ritual activity was in evidence with the purposeful re arrangement of long bones in some tombs. The secondary burial activity, as well as possible looting, accounted for the disarticulation, fragmentation and absence of many of the skeletal elements, which, together with the age of the remains and environmental deterioration, was responsible for the difficulties encountered in the investigation of this skeletal material. Equipment The basic equipment used for the assessment of the skeletal elements was a standard osteometric board to measure long bones, vernier calipers for assessing cranial diameters as well as a co-ordinate caliper to calculate subtenses (the depth of a curved surface of the skull). A spreading caliper was necessary to measure diameters over irregular cranial surfaces. Templates were constructed and used to assess various cranial and pelvic criteria as based on the diagrams shown in Fig. 2 and Fig. 7 respectively. In this study the subpubic angle was also assessed by comparing it to a template (Fig. 8) which depicted a ninety degree angle so even if only one half of the pelvis was present the subpubic angle was able to be estimated by being greater or less than forty-five degrees from the midline. Skeletal Material The skeletal material recovered during the excavations at Samtavro in was examined for this study. Twenty-four tombs yielded the remains of sixty individuals, forty-nine adults and eleven sub-adults. Adults were defined as having an estimated age 233 Sagona et al. 2010, p Sagona et al. 2010, p. 135.

78 78 of at least eighteen years. The age of the subadults was estimated by the dental eruption sequence 235 and the diaphyseal length of long bones. 236 The state of epiphyseal closure 237 distinguished subadults and young adults. The age of older adults was assessed by cranial suture closure 238. The basilar (spheno-occipital) suture is particularly reliable in the timing of its fusion (before 21years in males). 239 The assessment of dental attrition is also helpful 240 as well as the pubic symphysis 241 and auricular surface morphology. 242 Condition of Skeletal Remains Most of the skeletal remains were recovered from stone or tiled tombs with a small number from earthen pit graves. They were at the base of a slope and although appeared dry at the time of excavation no doubt would have been affected by ground water in the past. As noted above there was evidence of secondary human activity resulting in the fracture, displacement, commingling and loss of skeletal elements. The chart below (Table 8) shows the frequency of recovery of specific skeletal elements either in a complete or partial condition and containing at least some of the criteria which could be utilised for sex determination. These are detailed further when individual bones are recorded in Tables Ubelaker 1978, pp Ubelaker 1978, pp Brothwell 1981, p Buikstra and Ubelaker 1994, pp Steele and Bramblett 1988, p Smith 1984, p Katz and Suchey 1986, pp Lovejoy et al. 1985, pp

79 79 Table 8. Skeletal material recovered humerus 31 Radius 12 Ulna 17 Femur 22 Tibia 23 Fibula 6 talus 33 calcaneus 17 Patella 21 Pelvis 25 Cranium 24 Individual remains Total 60 Adults 49 Subadults 11

80 80 Pelvic elements. Twenty-seven (55 per cent) of the adult specimens had pelvic elements which displayed at least some of the traits most often used for sex allocation. The frequency that individual criteria were present and able to be assessed is shown in Table 9. Their relative importance, as determined by their accuracy in determining sex, is indicated in Table 3. Table 9. Relative frequency of occurrence of sex determinants in pelvic elements in adults from Samtavro. Ventral arc 16% (4) Subpubic concavity 24% (6) Ischial ridge 20% (5) Sciatic notch 76%(19) Subpubic angle 24% (6) Pubic length 8% (2) Pelvic inlet 20% (5) True pelvic size 20% (5) Acetabular size 40%(10) Muscle markings 52%(13) Ischial flaring 20% (5) Sacral shape 28% (7) Pre-auricular sulcus 72%(18) Auricular surface 52%(13)

81 81 Cranial elements The skeletal remains at Samtavro contained cranial elements suitable for sex allocation in 49 per cent (24) of individuals. The fragmented nature of the material made it impossible to assess many of the commonly used cranial traits although it was feasible to reconstruct some of the crania which allowed a limited assessment to take place. In Table 10 the particular cranial traits which were available and their frequency of occurrence are documented. Table 10. Occurrence of assessable cranial traits at Samtavro Cranium, size and shape 18% (9) supraorbital margins 24% (12) supraorbital ridges 24% (12) Glabella 14% (7) Mastoid 37% (18) nuchal crest 22% (11) mental eminence 35% (17) gonial angle 27% (13) There were no complete skulls present but the skulls which were partially reconstructed and had components present which could be used for sex allocation are included in this analysis.

82 82 Post-cranial skeletal remains This includes post-cranial material which is complete or incomplete but contains skeletal elements in which criteria can be assessed for sex allocation. Scapula: glenoid cavity There were 13 cases in which the scapula was at least partially present and the length of the glenoid cavity could be assessed. Humerus Humeri were present in 60 per cent of the individuals excavated but in most cases they were fragmented and incomplete as shown in Table 11. Table 11. Humeri from Samtavro. Number of skeletons (adults) 49 Number with humeri 31 (partial or complete) Number of complete humeri 9 Number of partial humeri 22 Femur There were 22 specimens in which some femoral elements were available for assessment. The number of individual skeletal criteria (variables) available for assessment is shown in Table 12.

83 83 Table 12. Specimens from Samtavro with femoral criteria. Variables Number Maximum length 11 Bicondylar length 8 Bicondylar width 15 Ant-post. midshaft 22 Med-lat. midshaft 22 Circumf. Midshaft 22 Subtroch. Ant-post 21 Subtroch med-lat 20 Max. head diameter 19 Tibia There were 23 specimens in which tibial criteria could be assessed. The frequency of individual skeletal elements available is indicated in Table 13. Table 13. Tibial criteria for sex determination Variables Numbers Maximum length 10 Antero-posterior diameter at nutrient foramen 23 Medio-lateral diameter at nutrient foramen 23 Circumference at nutrient foramen 23

84 84 Minor Elements The number of individuals in which at least one side of these minor elements was recovered is shown in Table 14. Table 14. Minor skeletal elements at Samtavro Number % Talus Calcaneus Patella 21 43

85 85 Chapter 5 Results and Discussion The skeletal remains for this study were excavated from the burial ground at Samtavro in Twenty-four tombs yielded the remains of sixty individuals, forty-nine adults and eleven sub-adults. Most of the remains were fragmented, incomplete and commingled. The initial sex allocation utilised the primary sex determinants, the pelvis and the cranium, where these were available, followed by an examination of the remaining postcranial remains. The methods of assessment of each of the skeletal components were undertaken as indicated in Chapter Two. The results of these analyses are tabulated below together with a discussion as to their significance. Analysis of Pelvis In twenty- seven specimens sex could be allocated using pelvic criteria. This was 55 per cent of the total sample of adult skeletons and comprised thirteen females and fourteen males. The individual specimens which were able to have their sex determined and the criteria utilised are shown in Table 15. It is noted that the sciatic notch and the preauricular sulcus were the most useful criteria, the sciatic notch was able to be assessed in 22 (80 per cent) of these cases and the pre-auricular sulcus in 18(66 per cent). Both have been shown to have accuracy for determining sex of over 85 per cent (see Table 5). Each pelvic criterion has been scored from one to three with a score of one indicating female, two indeterminate and three male. (The sciatic notch was originally scored from one to five but this has been reduced to one to three to align it with the other criteria). The final sex allocation is based on the mean of the scores available for each individual, with a score of two taken as the cut-off separating males and females. In three specimens the mean score corresponded with the cut-off so in allocating the sex in these

86 86 extra weighting was given to the score given by the sciatic notch as this has been demonstrated to be a particularly accurate feature (Table 3) and commonly could be assessed. Rogers used this method in which extra weighting was given to the most accurate criterion where there was an indeterminate result. 243 Table 15. Sex allocation by pelvic criteria. Pelvic ventral subpubic ischial sciatic subpubic aceta- muscle ischial sacral preauricular auricular mean sex Criteria Arc concavity ridge notch angle bulum markings flaring shape sulcus surface allocated spec. 1/SM1 5 (3) M 1/SM (3) M 1/SM7 2 (1) F 1/SM (1) F 2/SM2 3 (2) M 3/SM3 2 (1) F 3/SM4 3 (2) M 6/SM1 3 (2) 2 M 7/SM F 16/SM1 1 (1) 3 2 F 18/SM1 1 (1) 1 F 19/SM1 2 (1) F 22/SM M 22/SM M 22/SM3 1 (1) F 22/SM4 1 (1) 1 F 22/SM5 3 3 M 23/SM (2) M 27/SM1 2 (1) M 27/SM3 3 (2) M 28/SM1 5 (3) M 28/SM M 30/SM1 1 (1) F 32/SM (1) F Be c (3) M 34SM (1) F 498/SM1 1 (1) 3 2 F 243 Rogers 1999, p. 60.

87 87 There were 24 of the 27 specimens in whom pelvic elements were recovered which were able to be assessed by multiple pelvic criteria so these would be expected to be correctly assigned in over 80% of individuals based on the rate of correct allocation reported (Table 3). Two of the remainder were assessed by the sciatic notch and the other by the pre-auricular sulcus, both of which have been shown to have accuracy for correct allocation of over 80% (Table 5). The allocation of sex based on the pelvic criteria is therefore the benchmark by which the results determined from the analysis of other postcranial is appraised. Analysis of Cranium There were twenty-six adult specimens recovered with crania that had morphological traits present suitable for sex determination (Table 16). The traits were scored from one to five and the mean of the scores derived from those traits available from each individual calculated. A score of less than three indicated a female and greater than three, a male. A score of exactly three gave an indeterminate result. This occurred in two individuals (5/SM3 and 22/SM3) where only the mental eminence was available for assessment. On the basis of these results twenty-five individuals were able to have the sex allocated with two specimens indeterminate. The sex allocation derived from pelvic criteria is shown for comparison. It is noted than of the eleven specimens that had an associated pelvis the allocation using cranial traits was consistent in nine cases (82 per cent). Only four specimens could be satisfactorily reconstructed to obtain a significant number of the measurements required for CRANID analysis. Unfortunately many of these measurements were out of range for this program so no results for sex determination could be achieved. It is possible that these individuals were not included in the populations covered in CRANID but with over sixty populations in the database this would be very unlikely. It is more feasible that it was due to the incomplete, fragmented and damaged condition of the crania recovered made an accurate reconstruction very difficult. It is also possible that artificial deformation was also present. It is known to have occurred in the Caucasus and other crania have shown evidence of this. It may be

88 88 attributable to cultural practices or to some degree to birthing practices which encouraged this deformation to occur. 244 Table 16. Sex determination of crania by morphological analysis. size supraorbital supraorbital glabella mastoid nuchal mental gonial mean Allocated sex Specimen shape margins ridges Crest eminence angle Cranium Pelvis 1/SM F 1/SM M 1/SM F 2/SM F 2/SM F M 5/SM F 5/SM3 3 3 Indet 6/SM F M 7/SM F F 16/SM F 19/SM F 21/SM1 1 1 F 21/SM F 22/SM1 5 5 M M 22/SM M M 22/SM3 3 3 Indet F 22/SM7 1 1 F 27/SM M 28/SM1 4 4 M M 28/SM M M 29/SM F 30/SM F F 33/SM M M 33/SM F 34/SM F F 498/ F F 498/ F 244 Abdushelishvili 1968, p.181.

89 89 Analysis of Post-Cranial Skeletal Remains The investigation of the post-cranial material is based on the metric analysis of each skeletal criterion and then determining the section point separating males and females from the mean of the whole sample. The sample sizes for the individual criterion do not reach the minimum size for consistent allocation based on the work of Albanese 245 but it has been demonstrated that even with smaller samples accuracy correct allocation can be achieved in 80% of specimens (as discussed above). The final sex for each skeletal element is then estimated by a summation of the results for each of the individual criteria. Scapula The length of the glenoid cavity was measured as shown above in Fig. 11. There were twelve individuals in whom the sex could be estimated using this criterion (Table 12). The sex of individuals previously estimated from pelvic criteria, where this is available, is shown for comparison. Even with this relatively small number the method of sex allocation appears to be reliable with five out of the six specimens previously assessed by pelvic criteria having the same sex allocated. It is also noted that the section point derived from this examination (36 mm.) corresponded to that obtained by Stewart s analysis of material from the Terry collection Albanese 246 Stewart1979, p.98.

90 90 Table 17. Glenoid cavity length and sex allocation pelvic length and sex sex allocation allocation section point 36 (overall mean) specimens 1/SM1 39 M M 5/SM2 47 M M 6/SM1 38 M M 7/SM1 38 M F 16/SM1 36 indet. F 18/SM2 37 M 19/SM1 35 F F 21/SM2 33 F 21/SM3 36 indet. 22/SM1 38 M M 22/SM4 31 F 22/SM5 26 F 28/SM2 35 F

91 91 Humerus The section points for sex allocation based on humeral traits were taken as the mean of the whole sample. For comparison the section points obtained from the midpoint of the male and female means of the specimens previously sexed by pelvic criteria, as described above, are also shown. It is noted that there is very little difference in the results, 95 for the mean derived from the sample with associated pelvic/cranial material available as compared to 94 for the overall mean, (Table 18) indicating that the figures are acceptable in spite of the small sample sizes. The overall mean was used because it was derived from the larger sample. As noted previously Albanese found there was little difference in calculating the section point from the overall mean (when the sex ratio was not greater than 1:1.5) or the midpoint of the means derived from those specimens previously sexed from pelvic criteria. 247 Values, which exactly corresponded with the section point, were regarded as indeterminate. The final sex allocation for each specimen was determined by a summation of the sex estimations obtained for each of the available criteria. There were 17 skeletal individuals who had fairly complete humeri from which accurate measurements could be taken and also had been able to be assessed using pelvic and/or cranial traits. There was a good correlation (88 per cent) in the allocated sex between these assessments (Table 18). It is noted that the mid-shaft circumference correlated in 15 of 17 specimens and the least circumference of the shaft in 13 of 17 and the minimum mid-shaft diameter in 14 of 16 specimens. Humeral length correlated in only 6 of 9 specimens, this discrepancy was particularly apparent in specimen 27/3, with a comparatively small length, within the female range, whereas the other measurements were within the male range suggesting an error most likely due to mis-allocation of these commingled skeletal remains. 247 Albanese et al. 2005, p. 147.

92 92 Table 18. Measurements of humerus and sex allocation. length max. minimum Circumf. least sex pelvic midshaft midshaft midshaft circ. Sex section pt. (mid-pt. of male and female means) section pt. (overall mean) specimens 1/ SM7 331 M 30 M 22 M 85 M 81 M M F 1/SM8 307 F 20 F 16 F 61 F 57 F F F 2/SM2 302 F 20 F 16 F 73 M 59 F F M 3/SM4 307 M 25 M 19 M 76 M 75 M M M 6/SM1 19 F 17 F 56 F 51 F F M 7/SM1 21 F 16 F 62 F 55 F F F 16/SM1 297 F 17 F 17 F 59 F 54 F F F 18/SM1 18 F 16 F 55 F 53 F F F 19/SM1 296 F 21 F 14 F 59 F 54 F F F 22/SM3 302 F 19 F 16 F 58 F 58 F F F 23/SM2 326 M 22 indet. 19 M 67 M 65 M M M 27/SM3 247 F 24 M 19 M 68 M 64 M M M 28/ SM1 24 M 22 M 76 M 70 M M M 28/SM2 20 F 15 F 72 M 59 F F 28/SM3 22 indet. 18 indet. 67 M 64 M M M 32/SM1 18 F 15 F 58 F 55 F F F 33/SM1 26 M 20 M 74 M 67 M M M 34/SM1 312 M 21 F 16 F 65 F 61 indet. F F

93 93 Femur There were 27 femora available for assessment but only five were intact so in the remainder the analyses were incomplete. The mean for each criterion was taken as the section point, dividing males and females. (Table 19). The final sex allocation for each specimen was determined by the relative frequencies of each sex allocated for the individual criteria. Where the outcome was indeterminate the most accurate criteria as indicated in Table 6 was given extra weighting to obtain a result. Section points corresponding to those for the criteria in Table 19 which have been derived from other sources (Europeans) are shown in Table 20 for comparison. It is seen that substituting these figures into the calculations would not appreciably alter the results.

94 94 Table 19. Femoral measurements to determine sex max bicondylar ant-post med-lat circ ant-post med-lat Head sex length length Width midshaft midshaft midshaft subtroch subtroch diam. allocated (overall) Section pt specimen 1/SM1 441 M 446 M 35 M 31 M 101 M 37 M 39 M 51 M M 1/SM6 83 M 34 M 31 M 48 M M 1/SM7 30 M 27 M 89 M 27 indet. 31 F 43 F M 1/SM8 438 M 436 F 75 F 28 M 28 F 89 M 26 F 37 M 43 F F 2/SM1 64 F 26 M 25 F 81 indet. 25 F 30 F 44indet F 2/SM2 28 M 25 F 87 M M 3/SM2 79 M M 3/SM3 422 F 82 M 29 M 29 M 103 M 27 indet. 31 F M 6/SM1 385 F 23 F 24 F 75 F 28 M 32 F 42 F F 7/SM1 25 indet 25 F 77 F 27 indet. 30 F F 16/SM1 24 F 25 F 77 F 25 F 29 F 40 F F 18/SM1 422 F 418 F 72 F 25 indet 23 F 76 F 25 F 27 F 39 F F 18/SM2 74 F 20 F 20 F 67 F F 19/SM1 414 F 413 F 70 F 26 M 25 F 82 M 26 F 32 F 40 F F 21/SM1 425 F 420 F 84 M 26 M 27 M 87 M 25 F 31 F 46 M M 22/SM1 24 F 28 F 45 M F 23/SM1 31 M 36 M 47 M M 23/SM2 482 M 479 M 33 M 27 M 95 M 34 M 29 F 48 M M 23/SM3 80 M M 24/SM1 30 M 28 M 91 M 28 M 32 F 47 M M 27/SM1 74 F 26 M 22 F 80 F 41 F F 27/SM3 502 M 490 M 85 M 30 M 28 M 92 M 28 M 34 M M 28/SM1 23 F 25 F 82 M F 28/SM3 75 F 29 M 28 M 91 M 26 F 29 F M 30/SM1 25 F 30 F 41 F F 32/SM1 73 F 24 F 31 F 41 F F 34/SM1 438 M 433 F 75 F 25 indet 25 F 81 indet 30 M 31 M 42 F F 498/SM1 404 F 27 M 26 indet 89 M 24 F 35 M 48 M M

95 95 Table 20. Section points for femoral criteria derived from other sources. 248 (European Americans) Maximum length 436mm Circumference 86mm Antero-posterior diameter 28mm Medio-lateral diameter 26.5mm Bicondylar width 75 mm 249 The sex allocated from the femoral measurements was then also compared to the sex determined from the pelvic criteria (Table 36). They were found to correspond in 10 (77 per cent) of the 13 specimens which had an associated pelvis. The section point derived from the mid-point of the male and female means (the sexes predetermined from pelvic criteria) was nearly identical to the section point derived from the mean of the whole sample (137 as compared to 139). The estimations of sex derived from the limited number of femora available gave a result of 77 per cent consistency with that obtained from the pelvis. 248 DiBennardo and Taylor 1979, p Bass 1987, p.230.

96 96 Table 21. Comparison of sex determined from femoral and pelvic measurements section pt (overall mean) Allocated sex (from femurs) Sex from pelvis 1/SM1 M M 1/SM6 M 1/SM7 M 1/SM8 F F 2/SM1 F 2/SM2 M M 3/SM2 M 3/SM3 M 6/SM1 F 7/SM1 F 16/SM1 F F 18/SM1 F F 18/SM2 F 19/SM1 F F 21/SM1 M 22/SM1 F M 23/SM1 M 23/SM2 M M 23/SM3 M 24/SM1 M 27/SM1 F 27/SM3 M M 28/SM1 F M 28/SM3 F M 30/SM1 F 32/SM1 F F 34/SM1 F F 498/SM1 M

97 97 Table 22. Comparison of individual femoral measurements compared to pelvic criteria in determining sex max. bicond. bicond. ant-post med-lat circ ant-post med-lat head length length Width midshaft midshaft midshaft subtroch subtroch diam. 5/11 7/8 14/15 15/22 19/22 15/21 15/21 15/21 16/19 In the estimations of individual criteria the bicondylar length, the bicondylar width and the maximum head diameter contributed the most consistent results in this sample. (Table 22.) The maximum length of the femur gave the results which were the least consistent. This would indicate that consideration be given to applying greater weight to the more consistent criteria which would improve the results for the sex allocation. Tibia There were 24 specimens of the tibia from which measurements were able to be taken to allocate sex and of these 12 had associated pelves enabling a comparison of the results to assess the reliability of these estimations (Table 23). The section point for each of the tibial criteria examined was derived from the mean of the whole sample. Where the result was indeterminate extra weighting was given to the circumference at the nutrient foramen, as this has been shown to have high accuracy for sex determination (Table 7). There was a reasonable correlation between the measurements derived from the tibiae and those from the pelves with 9 of the 12 in which the sex had been previously determined by pelvic criteria (75 per cent) allocating the same sex. The most reliable criterion was the medio-lateral diameter of the nutrient foramen with an 83 per cent consistency.

98 98 Table 23. Determination of sex from tibia Antpost. Tibia length diam. at med-lat. Diam.at Circumference sex Pelvic sex nutrient foramen Nutrient foramen at nutrient f. allocated Allocation Section pt /SM1 20 F 26 M 80 indet Indet. F 1/SM7 376 M 42 M 27 M 110 M M 1/SM8 31 M 20 F 90 M M F 2/SM2 28 indet 24 M 90 M M M 3/SM2 32 M 24 M 97 M M 3/SM4 370 M 37 M 24 M 97 M M 6/SM1 24 F 22 F 75 F F 7/SM1 26 F 18 F 74 F F 16/SM1 317 F 23 F 18 F 66 F F F 18/SM1 26 F 19 F 68 F F F 19/SM1 347 F 24 F 22 F 85 M F F 21/SM1 29 M 20 F 88 M M 21/SM2 24 F 23 indet 74 F F 21/SM3 20 F 26 F 81 M F 22/SM1 35 M 24 M 105 M M M 23/SM2 364 M 35 M 24 M 100 M M M 23/SM3 35 M 23 indet 91 M M 27/SM1 352 M 20 F 26 M 81 F F 27/SM3 387 M 34 M 27 M 98 M M M 28/SM1 35 M 31 M 107 M M 28/SM3 324 F 29 M 22 F 81 M M M 32SM/1 29 M 19 F 79 F F F 34/SM1 347 F 30 M 21 F 83 M M F 498/SM2 325 F F

99 99 Dentition The teeth which were examined for sex determination were the upper and lower canines and the upper and lower second molars. The measurements taken were: Canine: labial ligual Molars: bucco lingual / mesial bucco lingual /distal The comparison of sex allocation based on dental criteria with that based on pelvic traits (Table 24) in this sample suggested that the dental measurements were not a reliable method of assessment, at least in this population, with only a 60 per cent correlation. This may in part be due to dental attrition making the measurements less than would be expected. There is evidence in Table 24 to suggest that this is the case with three out of the four incorrect allocations being males allocated as females when compared to the pelvic allocation. The range of the measurements for each criterion is only about 3 mm with the differences between male and female means only 1 mm making observer error a limiting factor in the allocation of sex by this method.

100 100 Table 24. Allocation based on Dental measurements UCLALI UM1BLM UM1BLD LCLALI LM1BLM LM1BLD Mean S.Pt Sex Pelvic sex 1/SM M 10.29M M 1/SM3 8.24M 8.24F F 2/SM2 6.64F 7.36F 8.48F 9.1F 8.76F F 5/SM2 9.02M 10.34F 11.27M 8.72M 9.62F 9.77F M 5/SM3 7.08F 7.43F 9.05F 8.07M F 6/SM1 7.41F 10.6M 9.65F 6.81F 9.96M 10.03M F 7/SM1 7.56F 10.26F 10.59M 7.1F 9.4F 10.52M M 16/SM1 7.17F 10.53M 8.97F 7.08F 10.13M 9.2F F 19SM1 7.47F 10.31F 9.9M 9.39F 9.2F F 21/SM1 8.38M 11.45M 10.29M 8.53M M 21/SM2 9.2M 11.61M 10.44M M 21/SM3 6.62F 10.36F 9F 7.5F 9.24F 8.18F F 22/SM1 8.51M 11.29M 10.81M 9.14M 10.62M 9.8F M M 22SM2 9.44F 8.37F F M 22/SM4 7.77M 8.17M M F 22SM5 6.2F F M 23SM1 9.88M 10.19M M 23SM3 7.6M 8.25M M 27SM1 7.03F 8.1M 8.42F F M 27/SM3 6.2F F 28/SM M 9.27F M M 28/SM2 8.09F 8.46F F 28/SM3 11.4M 9.89M 7.14F 9.67M 9.68F M M 29/SM1 8.04M 10.29F 9.72F M 30/SM1 7.44F 10.58M 8.97F 6.9F 9.97M 9.02F F F 32/SM1 5.98F F F 33SM1 7.92M 10.42M 10.64M 7.7M 10.57M 10.66M M M 34/SM1 7.76M 11.65M 11.07M 7.43F 10.26M 10.46M M 498/SM2 8.39M 10.61M 9.66F 7.65M M section point

101 101 Sex allocation of subadults There were eleven subadults identified and of these an indication of the sex was able to be estimated in three individuals. (Table 25). The sex of these three subadults was based on the relative age of the dental age as compared to their skeletal age. Comparing the pattern of dental development with a dental chart showing the sequence of dental eruption, enabled the dental age of individuals to be assessed. 250 Skeletal age was estimated by examining the post-cranial skeletal material to determine the stage of epiphyseal closure. These were then compared with published figures to determine the relevant age. 251 In two cases the skeletal maturity was greater than dental maturity suggestive of females and in one case the dental age corresponded to the skeletal age indicative of a male but in most individuals no estimate of the sex could be made. Table 25. Age and Sex of Subadults dental age skeletal age pelvis estimated sex approx 3/SM1 8-9y 12y F 3/SM5 Newborn? 4/SM1? 5SM1 5y? 16/SM2 1.5y? 18/SM y <13-15y M 22/SM6 2-3y? 27/SM2 6y? 28/SM4 <6y? 30/SM2 4y? 33/SM2 9-10y >13y F 491/SM1 <15y? 250 Ubelaker 1994, Fig Ubelaker 1994, Fig. 20.

102 102 An attempt was made to ascertain the sex of the subadult remains which were recovered but, as expected these results confirmed the difficulty in trying to allocate the sex of subadults with the three results which were obtained not being regarded as more than speculative. Summary of results There were a total of 60 individuals excavated from the burial ground at Samtavro. This included 12 subadults and 48 adults (Table 26). There were 44 adults (92 per cent) in whom the sex could be assessed by the methods detailed above. These included 22 females and 22 males. There still remained 5 adults in which the sex was not able to be estimated as the remains were too fragmented and incomplete to obtain useful criteria. The details of these cases are discussed below. Eleven individuals had both cranial and pelvic sex determinants present and in nine of these (82 per cent) the sex allocation derived from each of these was the same. Thirty-six adults (73 per cent) had pelvic and/or cranial elements recovered. It would be expected that the results obtained from the pelvic and cranial criteria would be more reliable not only because of their more obvious sexual dimorphism but they are also not dependent on having statistically significant numbers in the sample and are not population specific to the same degree as post-cranial elements. The two cases in which the cranial and pelvic results were inconsistent (specimens Tomb 2 /SM2 and Tomb 6 /SM1) had both been allocated as male on the pelvic criteria but female from the cranial criteria. It appeared that they were both young adult males, aged about 21 years based on dental eruption sequence. It has been shown that the crania of young adult males are most commonly misclassified as the morphological features on which they are classified are based on robusticity and take on an increasing masculine appearance with age. 252 A further specimen (Tomb 21/SM1) also had inconsistent 252 Meindl et al. 1985, p. 83.

103 103 results. In this case the bones were commingled with other individuals making the assignment of skeletal material difficult which may have led to an inaccurate allocation. Overall, in eighteen specimens (out of 34) in which there were multiple elements able to be assessed the allocated sex for all of the available skeletal elements were in complete agreement. Of the twenty nine adult specimens in which there were both pelvic and/or cranial elements present together with other post-cranial material, the allocated sex diagnosed by pelvic and/or cranial criteria corresponded to that diagnosed by post-cranial criteria in twenty five instances and in only four was there a discrepancy. There were five individuals with indeterminate sex. Three of these (Tomb 1/SM5, Tomb1/ SM9 and Tomb 1/SM11) had only scanty fragmented remains recovered, which appeared to be small and probably subadults with no features present to indicate the sex. The allocation from tomb1 was additionally complicated by the large amount of commingled skeletal material. The remains of Tomb 5/SM3 consisted of parts of the left mandible and maxilla and some loose teeth. The dental attrition indicated an age of approximately 35 years and the robusticity of the mental eminence suggested a male but this was not definite. The remaining specimen (Tomb 25 /SM1) had only feet bones present from which the sex could not be estimated.

104 104 Table 26. Final Sex Allocation. Sex Allocation Final Sex Specimen Allocation pelvic cranial postcranial glenoid humerus femur tibia 1/SM1 M M M indet M 1/SM2 F F 1/SM3 M M 1/SM4 F F 1/SM5 Indet 1/SM6 M M M 1/SM7 F M M M M 1/SM8 F F F M F 1/SM9 indet Indet 1/SM11 Indet Indet 2/SM1 F F F 2/SM2 M F F M M M 3/SM1 Subad F 3/SM2 M M M 3/SM3 F M M 3/SM4 M M M M 3/SM5 Subad? 4/SM1 Subad? 5/SM1 Subad? 5/SM2 F M M 5/SM3 indet Indet 6/SM1 M F M F F F F 7/SM1 F F M F F F F 16/SM1 F F Indet F F F F 16/SM2 Subad? 18/SM1 F F F F F

105 105 18/SM2 M F Subad M 19/SM1 F F F F F F F 21/SM1 F M M M 21/SM2 F F F F 21/SM3 Indet F F 22/SM1 M M M F M M 22/SM2 M M M* 22/SM3 F indet F F 22/SM4 F F F 22/SM5 M F M 22/SM6 Subad? 22/SM7 F F 23/SM1 M M 23/SM2 M M M M M 23/SM3 M M M 24/SM1 M M 25/SM1 Indet 27/SM1 M F F F 27/SM2 Subad? 27/SM3 M M M M M M 28/SM1 M M M F M M 28/SM2 F F F 28/SM3 M M M M M M 28/SM4 Indet Subadult 29/SM1 F F 30/SM1 F F F F 30/SM2 Subad? 32/SM1 F F F F F 33/SM1 M M M M 33/SM2 Subad F 34/SM1 F F F F Indet** F 491/SM1 Subad? 498/SM1 F F M F 498/SM2 F F F

106 106 Distribution of Sexes in Tombs The allocation of sex has enabled the distribution of the sexes within the tombs to be analysed (Table 27). There were ten tombs with only a single skeleton present, of these four were female, two were male, two were subadults and two were indeterminate. Six tombs contained two individuals; half of these had one female and one subadult present. The remaining tombs contained multiple skeletal individuals with no discernible pattern of sex distribution present. Table 27. Analysis of Individual Tombs Tomb No. Individuals Adults male female Indet. subadults sex Indet F M M F

107 107 Conclusion Key Findings This research has confirmed that data obtained by the methods discussed can identify the sex of skeletal individuals from an undocumented population, in which the material recovered from ancient burials is fragmented, fragile, incomplete and commingled in at least 80 per cent of specimens. These methods were applied to the human remains recovered from the burial ground at Samtavro. It was found that the sex was able to be estimated in 87 per cent of that sample with the acceptable level of reliability of at least 80 per cent based on consistency with results obtained from pelvic criteria and published results for specific skeletal elements. The assessment of pelvic and cranial criteria provided the most reliable results for sex allocation but other post-cranial remains made a significant contribution particularly where cranial and pelvic elements were not present. The importance of specific skeletal remains in sex determination depends not only on their sexual dimorphism but also on their physical structure which determines their possibility of recovery. Minor skeletal elements are often recovered and may be useful for sex estimation in the absence of other primary sex determinants. These include components of the elbow as well as bones of the feet particularly the talus. In the absence of a documented population, skeletal elements under investigation which have a significant margin between male and female means and minimal overlap in values can provide significant results even with a relatively small sample. In these situations the mean of the whole sample can be used as the section point dividing males and females as long as the number of each sex present is approximately equal.

108 108 The allocation of sex to sub-adults cannot be achieved easily or with any degree of accuracy by the morphological methods currently available. DNA analysis offers the only accurate assessment available at present. Contribution This research has demonstrated the methods which may be effectively employed to analyse ancient skeletal material to enable accurate results for sex determination for a significant proportion of individuals. This has involved not only criteria from major skeletal elements but components not commonly employed for sex allocation have been shown to be able to make a worthwhile contribution. Statistical approaches for allocating sex to skeletal individuals have been reviewed and modifications to the accepted statistical criteria have been shown to be feasible to accommodate the nature of the material under investigation without significantly affecting the accuracy of the end result. An objective evaluation of components of the elbow has been carried out with material at the Museum of London which indicated that the olecranon fossa depth and the distance between the olecranon and coronoid processes show significant dimorphism. When assessed individually and particularly when combined with the epicondylar breadth of the humerus and the head of the radius they should prove to be useful criteria for sex determination. Future Research Many of the methods for sex determination have been evaluated on only a limited population range. Further work is necessary to determine their viability when applied to other populations. This particularly applies to the minor skeletal elements which can provide significant supporting evidence.

109 109 Sex allocation for sub-adults using the morphological methods presently available is difficult and inaccurate. Future research using larger sub-adult samples are necessary to clarify this situation and provide additional techniques and more accurate assessments. Summary This study has investigated the methods can be utilised for the sex determination of ancient human skeletal material. Following this analysis, these methods, with some modifications to take into account the nature of the research material, have been applied to ancient human remains recovered from the burial ground at Samtavro. The sex was able to be estimated in over 80 per cent of individuals which was considered to provide a reliable allocation in over 80 per cent of skeletal remains studied.

110 110 Appendix One Data Recording Form: Sex Determination Burial Number Specimen Number Date of Excavation Pelvis (Ohio Data Collection Codebook scores) 1 / 2 / 3 ventral arc pos/ med/ neg sub-pubic concavity c'cav/ flat/ c'vex ischio-pubic ridge pos/ med/ neg Sciatic notch wide/ med / narrow Sub-pubic angle 90+ / 90 / 90- Pubic length L / R Ischial length L / R wide/ med /narrow True pelvis size shallow/ med / deep Acetab'm size/direct'n small,anterior/ med / big,lateral muscle markings smooth/ med / marked Ischial flaring present/ med/ absent Sacrum shape long/shallow/med/broad/deep Pre-auricular sulcus marked/ med / minimal Auricular surface raised /med / flat Cranium (Ohio Data Collection Codebook scores) size/shape small,smooth/med/large,rough (score 1 5 as per chart) supraorb. Margins supraorb.ridges glabella prominence small / med / large mastoid process nuchal crest mental eminence gonial angle 125+/ 125 / 125- Scapula glenoid length L / R

111 111 Humerus bicondylar width olecranon depth Radius head diameter: max. head diameter: min. Ulna coronoid olecranon Femur head diameter GT FC GT LT LT FC bicond.width max. length Patella max.height max.thickness L / R L / R L / R L / R L / R L / R L / R L / R L / R L / R L / R L / R L / R Talus max. length L /R Allocated Sex

112 112 Appendix Two Study of quantitative techniques for determining sex at the elbow. Reliable techniques for determining sex are limited when cranial and pelvic elements are scarce and the available skeletal remains are poorly preserved, fragmentary and commingled. This was the situation during the investigation of the late Roman early medieval burial ground excavated at Samtavro, Georgia. The post cranial material was to a large extent fragmented and incomplete making the assessment of sex difficult and unreliable. It was noted however that in 84% of fragmented specimens skeletal components from the elbow including the distal humerus, proximal ulna and head of radius were recovered in relatively good condition (Table 28). As a result of this finding a study was performed to test the likely benefit of using these criteria for sex allocation. Table 28. Elbow Criteria from Samtavro specimens Number of specimens able to be assessed 38 Specimens with assessable elbow criteria 32 Number with some upper limb bones intact 14 Completely fragmented but with assessable criteria 18 No assessable elbow criteria 6 In this poorly preserved skeletal material 50 per cent of specimens were completely fragmented and not able to be assessed by measurements involving intact bones whereas 84 per cent still had elbow components in a suitable condition to be investigated, each containing at least two potentially useful criteria. The elbow joint has skeletal elements which would appear to make them distinctly suitable for the determination of sex. These, comprising the distal humerus, proximal ulna and head of radius largely consist of dense compact bone so are particularly durable

113 113 and often recovered from ancient skeletal assemblages where either due to the age of the material or the nature of the environment, many other skeletal elements do not survive. In general, measurements associated with joints consistently demonstrate the greatest sexual dimorphism which is attributed to the differences in functional and occupational stresses. 253 The elbow also has the specific features associated with the larger carrying angle in females which would provide other opportunities for detecting discriminating traits. Some of these components of the elbow have been further investigated for this study to determine their applicability to the sex determination of ancient remains. The morphological features of the distal humerus were assessed by quantitative methods rather than the qualitative methods used by Rogers to determine if this provided more accurate and reproducible techniques for sex allocation as the visual techniques were not found to be as accurate when applied to historic collections. 254 They were also found to have a significant inter-observer error when assessed by an inexperienced observer. 255 The proximal ulna was also studied to assess whether this was of value to estimate sex in situations where the head of the ulna was the only surviving portion of that bone (in which case the technique used by Purkait which requires an intact ulna is not feasible). The head of the radius, which had been investigated previously by Berritbeitia 256 with a modern sample, was reassessed with historic specimens from the database of the Museum of London. The focus of this research was to determine whether a quantitative analysis of those components of the elbow that are often recovered from ancient burials could provide a reliable estimation of sex. Materials Skeletal material from the Chelsea Old Church population at the Museum of London was examined. The study involved forty-three randomly selected skeletal individuals from 253 Iscan et al. 1998, p. 26, see also Albanese et al. 2005, p Rogers 2006, p Rogers 2006, p Berrizbeitia 1989, pp

114 114 the period 1700 to 1850 on whom the sex had been accurately documented from coffin plates. Both left and right sides were measured. Elements from the left side were used with the right side being substituted when the left was not available. Not all individuals had complete elbow joint elements so specimen numbers for the individual skeletal features differed. In each analysis the proviso was that each sex had approximately equal representation. Methods Quantitative means were devised to study some aspects of the distal humerus previously assessed qualitatively by Rogers 257 and adapted the method of Purkait and Chandra 258 to measure features of the proximal ulna. The features of the distal humerus studied were the depth of the olecranon fossa, the thickness of the bone in the base of the olecranon fossa, the angulation of the medial epicondyle and the epicondylar breadth. The depth of the olecranon fossa was assessed using a contour gauge situated longitudinally over the olecranon fossa, the wires were depressed into the fossa and the deepest wire measured with digital calipers (Fig.26). The bone thickness in the base of the olecranon fossa was determined using spreading calipers. A more accurate measurement could have been taken using a needle micrometer but in trialing this method it was considered too likely to cause damage to fragile bone. If the fossa was found to be perforated or translucent when held up to light this was recorded and assumed to be less than 1mm in thickness. The angulation of the medial epicondyle was determined by measuring the angle between the flat surface of the trochlear and the anterior surface of the epicondyle (Fig. 27). Because of the curvature of the border the angle was measured as the flatter, medial 257 Rogers 1999, pp ; 2006, pp ; 2009, pp Purkait and Chandra 2004, pp

115 115 portion. Two hinged blades were adjusted visually to correspond to the size of the angle and then measured with a protractor. The epicondylar breadth was measured as indicated in Bass. 259 Fig. 26. Measuring Olecranon fossa depth 259 Bass 1987, p.158.

116 116 Fig. 27. Angulation of medial epicondyle of humerus Angle of medial epicondyle Proximal Ulna Olecranon Coronoid angle The angle between the posterior surface of the head of the ulna and the line joining the tips of the coronoid and olecranon processes was measured using a hinged blade with one arm held against the posterior surface of the head of the ulna and the other in contact with the tips of the olecranon and coronoid processes as indicated in Fig.28. Olecranon-Coronoid Distance The distance between the tips of the coronoid process and the olecranon over the midpoint of the trochlear notch was measured using a vernier caliper (A-B Fig.28).