Can Long Bone Structural Variability Detect Among-Population Relationships?

University of Massachusetts Amherst ScholarWorks@UMass Amherst

Doctoral Dissertations Dissertations and Theses

July 2017

Gina Agostini University of Massachusetts Amherst

Follow this and additional works at: https://scholarworks.umass.edu/dissertations_2

Part of the Biological and Physical Anthropology Commons

Recommended Citation Agostini, Gina, "Can Long Bone Structural Variability Detect Among-Population Relationships?" (2017). Doctoral Dissertations. 909. https://doi.org/10.7275/10008914.0 https://scholarworks.umass.edu/dissertations_2/909

This Open Access Dissertation is brought to you for free and open access by the Dissertations and Theses at ScholarWorks@UMass Amherst. It has been accepted for inclusion in Doctoral Dissertations by an authorized administrator of ScholarWorks@UMass Amherst. For more information, please contact [email protected].

CAN LONG BONE STRUCTURAL VARIABILITY DETECT AMONG- POPULATION RELATIONSHIPS?

A Dissertation Presented

GINA MARIE AGOSTINI

Submitted to the Graduate School of the University of Massachusetts Amherst in partial fulfillment of the requirements for the degree of

DOCTOR OF PHILOSOPHY

May 2017

Anthropology

CAN LONG BONE STRUCTURAL VARIABILITY DETECT AMONG-POPULATION RELATIONSHIPS?

A Dissertation Presented

GINA MARIE AGOSTINI

Approved as to style and content by:

______Brigitte M. Holt, Chair

______Laurie R. Godfrey, Member

______Joseph Hamill, Member

______John H. Relethford, Member

______Jacqueline Urla, Department Head Anthropology

DEDICATION

To Josephine, Mario, Vivian, and William

ACKNOWLEDGMENTS

I have many people to thank, starting with Blake, my partner in crime and person of infinite patience. I also thank my committee, Brigitte Holt, Laurie Godfrey, John

Relethford, and Joe Hamill, for their guidance and patient advice. I thank my parents,

Tracey and Michael, extended family, Shirley, Boomer, Katami, and “the Chinch.” I thank Sarah, Aurelie, Jill, Joe, Shwynn, Rachel, and Dan for their continued friendship even after I routinely abandoned them to collect data for 3-4 months at a time.

Many thanks also to the people who helped me during my travels. I am especially appreciative to Marius Loots and Erica L’Abbé for kindly offering a place to stay, good conversations around the braai, introducing me to biltong, showing me the Cradle of

Humanity, and otherwise facilitating my wonderful stay in South Africa. Thanks also to

Petra Maas for teaching me how to navigate Cape Town transit—congratulations on your many successes. Many thanks to both Gloria Cassanova and Luca Pagani for sharing their home, their social network, and for introducing me to the Bialetti—perhaps the most

important object I own. Thanks also to Ale Riga and Irene Dori for their friendship and

for providing a forum to “nerd out” about limb bone robusticity. And thanks also to

Margarida for showing me the in-and-outs of Lisbon, including the public versus private hospital systems!

I thank the many museum curators who made this (and other) work possible—

Diana Swales for accommodating my last-minute (somewhat panicked) request to visit

the collections upon learning my intended samples were unexpectedly unavailable (ah,

fieldwork…) and for hauling countless skeletons up from the basement for me to

measure. To Elsa Pacciani not only for facilitating access to the collections in Florence,

iv both for her lunchtime and travel companionship. To Marta Mirazon Lahr, Maria

Giovonna Belcastro, Hugo Cardoso, Mafalda Madureira, Linda Greyling, Brendan

Billings, and Alan Morris, all for their assistance accessing collections. And to my undergraduate research assistants: Amber Lopez, Jennifer Nadeau, Oshiomah Oyageshio, and Subhrangi Swain—I know you all have bright careers ahead of you.

I also thank the many people who offered advice over the years. Noreen von

Cramon-Taubadel, Jay Stock, and Kristian Carlson, all of whom offered thoughtful feedback on a grant proposal e-mailed out of the blue. I thank Jason Kamilar for serving as collaborator, sounding board, and for showing me the ropes of R. Krista Harper,

Lynnette Leidy-Sievert, Stephen King, and others who were always happy to meet, offer feedback, and otherwise serve as a support system. Many thanks to Michael Doube and

Richard Domander for providing much-needed macro help with ImageJ.

Finally, I thank the National Science Foundation (award #1411887), the Cultural

Heritage of European Societies and Spaces or “CHESS” program (award # OISE-

0968575), and University of Massachusetts Amherst, all of whom have provided funds to support this and related projects over the past several years.

ABSTRACT

CAN LONG BONE STRUCTURAL VARIABILITY DETECT AMONG- POPULATION RELATIONSHIPS?

MAY 2017

GINA MARIE AGOSTINI, B.A., UNIVERSITY OF ARKANSAS

M.A., NORTH CAROLINA STATE UNIVERSITY

Ph.D., UNIVERSITY OF MASSACHUSETTS AMHERST

Directed by: Professor Brigitte Holt

Phenotypic traits develop and are maintained by complex interactions between intrinsic

(molecular) and extrinsic (environmental) factors. While the influence of intrinsic factors on adult craniomandibular variation has been intensively studied, less is known about limb bones, in part because it is assumed that their plasticity obscures intrinsic signals, especially those fixed early in life. While diaphyseal regions are plastic in response to activity, the extent to which they also reflect (phylo)genetic autocorrelation has not been sufficiently addressed, particularly given the common practice of comparing long bones across populations unevenly dispersed in space and time.

Here I investigate the degree to which long bone lengths, joints, and diaphyses vary in their ability to detect intrinsic genetic patterns. I do this by calculating among- population genetic relationships via long bone variation in samples from England,

Southern Europe, and South Africa. I then test whether these predictions significantly match those generated via craniofacial variation and, further, whether they are supported by contextual (historical) information.

Given the innately plastic nature of diaphyseal regions, I further test whether differences in physical activity can obscure predicted genetic relationships. I do this by

vi adding a temporal component to genetic distance analyses via inclusion of Medieval samples and by partitioning several Southern European samples into “high intensity” and

“low intensity” subgroups based on recorded occupational data, using these to generate more genetic predictions.

Results show that all three long bone properties reflect among-population genetic structure, with length and joint dimensions doing so at levels comparable to those of the crania. Diaphyses, however, generate lower levels of among-population differentiation, presumably because their plasticity fuels more intrapopulation variation. Despite this, diaphyses still detect key components of population genetic structure, including genetic affinity shared among modern English and descendant South Africans, the close genetic relationships among modern Southern Europeans (even when subdivided by occupation), and the ancestral connection between Medieval and modern English samples. In total, these results suggest that all long bone properties can detect among-population genetic information, and further, that interpretations of behavior from limb bone variation can be strengthened if genetic relationships (or assumed relationships) are controlled for.

vii

TABLE OF CONTENTS Page

ACKNOWLEDGMENTS ...... iv ABSTRACT ...... vi LIST OF TABLES ...... x LIST OF FIGURES ...... xii LIST OF ABBREVIATIONS ...... xvii 1. PROJECT OVERVIEW ...... 1 1.1 Introduction ...... 1 1.2 Chapter 1 References ...... 7 2. BACKGROUND ...... 13 2.1 Chapter Overview ...... 13 2.2 Molecular Development ...... 13 2.3 Function Throughout Development ...... 24 2.4 Bone Functional Adaptation: Overview and Application ...... 30 2.5 “Applied” Bone Functional Adaptation ...... 38 2.6 Bone Functional Adaptation: Limitations and Concerns ...... 42 2.7 Chapter 2 References ...... 44 3. PROJECT JUSTIFICATION AND PREDICTIONS ...... 61 3.1 Project Justification ...... 61 3.2. Goals and Predictions ...... 61 3.3. Chapter 3 References ...... 64 4. MATERIALS AND METHODS ...... 65 4.1 Samples and Sample Organization ...... 65 4.2 Analytical Methods ...... 71 4.2.1 s and R-matrices ...... 71 4.2.2 Coefficient of Variation ...... 76 𝑸𝑸𝑸𝑸𝑸𝑸 4.3 Data Collection ...... 77 4.3.1 Long Bones ...... 77 4.3.2 Articular and Length Dimensions ...... 81 4.3.3 Crania ...... 82 4.4 Pre-Analysis Data Treatment ...... 83 4.4.1 Standardization ...... 83 4.4.2 Data Imputation ...... 85 4.5 Chapter 4 References ...... 88

viii

5. RESULTS ...... 96 5.1 Body Mass Regressions ...... 96 5.2 Justification of Sample Pooling ...... 100 5.3 Nature and Pattern of Variation ...... 100 5.4 Population Differentiation ...... 106 5.5 Interpopulation Relationships ...... 111 5.5.1 Relationship Matrices ...... 111 5.5.2 Distance Matrices and Multidimensional Scaling Plots ...... 118 5.5.2.1 Modern Populations ...... 118 5.5.2.2 Temporal Test (Modern and Medieval Samples) ...... 134 5.6 Mantel Randomization Tests ...... 150 5.7 Occupaton Test ...... 152 5.8 Chapter 5 References ...... 159 6. DISCUSSION ...... 160 6.1 Long Bones in Context ...... 160 6.1.1. Differentiation ...... 160 6.1.2. Patterning ...... 163 6.1.3. Population History ...... 171 6.1.3.1. South Africa ...... 171 6.1.3.2. South Europe ...... 186 6.2 Genetic Predictions by Funtional Traits ...... 197 6.2.1 Genetic Affinities of Plastic Traits ...... 197 6.2.2 How Can Plastic Traits Confer Genetic Structure? ...... 203 6.3 Project Limitations ...... 208 6.4 Chapter 6 References ...... 210 7. CONCLUSION ...... 222 7.1 Concluding Remarks ...... 222 7.2 Future Directions ...... 222 7.3 Chapter 7 References ...... 223 BIBLIOGRAPHY ...... 225

LIST OF TABLES

Table Page

Table 1. Summary of cross-sectional shape, rigidity, and strength properties ...... 41

Table 2. Sample context ...... 67

Table 3. Sample pool* for different anatomical regions...... 68

Table 4. Temporal and sample groupings, abbreviations used in this project ...... 68

Table 5. Software used for this research ...... 76

Table 6. Brief description of orientation protocols following Ruff (2002) ...... 78

Table 7. Joint measurements (in mm) from Ruff (2002) ...... 81

Table 8. Cranial measurements and landmarks ...... 83

Table 9. Standardization factors for cross-sectional properties ...... 85

Table 10. Body mass formulae ...... 85

Table 11. Results of OLS Regression for body mass, tibial and humeral length ...... 96

Table 12. Mean coefficients of variation for aggregate anatomical groupings (all populations)...... 100

Table 13. Mean coefficients of variation for skull, joint, cross-sectional metrics, and length metrics (all populations) ...... 103

Table 14. Mean coefficients of variation for cross-sectional metrics by bone ...... 103

Table 15. Mean coefficients of variation by cross-sectional property ...... 104

Table 16. Analysis of Variance on mean CV by time and region ...... 105

Table 17. Descriptions for anatomical CV ANOVA tests ...... 105

Table 18. Analysis of Variance on mean CV using cross-sectional properties ...... 106

Table 19. Unbiased QST values and standard errors (se) by broad anatomical groupings, modern populations ...... 106

Table 20. Unbiased QST values and standard errors (se) by broad anatomical groupings, modern and Medieval populations ...... 107

Table 21. Unbiased QST values and standard errors (se) for various long bone groupings, modern populations ...... 108

Table 22. Unbiased QST values and standard errors (se) for various long bone groupings, modern and Medieval populations ...... 109

Table 23. Unbiased QST values and standard errors (se) by bone, modern samples ..... 110

Table 24. Unbiased QST values and standard errors (se) by bone, modern and Medieval populations ...... 110

Table 25. R-matrices for craniofacial data, modern populations ...... 112

Table 26. R-matrices for joint data, modern populations ...... 112

Table 27. R-matrices for length data, modern populations ...... 113

Table 28. R-matrices for cross-sectional data, modern populations ...... 113

Table 29. R-matrices for craniofacial data, modern and Medieval populations ...... 116

Table 30. R-matrices for joint data, modern and Medieval populations ...... 116

Table 31. R-matrices for length data, modern and Medieval populations ...... 117

Table 32. R-matrices for cross-sectional data, modern and Medieval populations ...... 117

Table 33. Mantel test results for Modern populations based on 10,000 permutations ... 150

Table 34. Mantel test results for Modern and Medieval populations based on 10,000 permutations ...... 150

Table 35. Mean J and standard error (95% confidence) by population and occupational intensity ...... 156

Table 36. T-test results comparing heavy and light intensity occupations by sample .... 156

Table 37. T-test results comparing Bologna and Lisbon by occupational intensity ...... 157

Table 38. R-matrices for cross-sectional data, modern populations ...... 157

Table 39. Unbiased QST values reported in other studies ...... 162

LIST OF FIGURES

Figure Page

Figure 1. Femur midshaft casts of pre-Medieval Saxon (left), Medieval (middle), and modern (right) male Europeans. Note the increasing circularity due to decreasing anteroposterior elongation over time...... 3

Figure 2. A sample of diaphyseal casts generated for this project...... 5

Figure 3. A simplified diagram showing the hierarchical nature of embryonic long bone patterning based on an adult macaque skeleton. Purple: generalized expression domain of the PBX genes; Blue/Green gradient: collinear expression of HOXA/D genes along the proximodistal axis; Red: expression domain of SHH genes that regulate anteroposterior patterning of the zeugopod/autopod elements and limb bud outgrowth via reciprocal interaction with distally expressed FGF genes. Mutual inhibition of HAND2 (H2) genes and GLI3 establish anteroposterior polarity of the stylopod, which is patterned prior to SHH expression. Yellow boxes indicate later expression domains of GDF5 genes and WNT/B-Catenin, both implicated in joint interzone formation and subsequent maintenance of joint structures during development (along with noggin, IHH, and several other factors). Arrows reflect interactive relationships between genes or their relationship to specific bony regions. Solid arrows reflect broader/higher level genetic influences, while dotted arrows reflect more regional/localized patterns...... 22

Figure 4. Diagram of the double feedback loop of BFA. The plus and minus symbols reflect the loss or deposition of new bone in response to decreasing and increased strains, respectively. Image modified from Ruff et al., 2006...... 34

Figure 5. Model cross-section of a proximal tibia (G. gorilla) displaying bending axes of the anatomical and principal SMAs (I). All intersect at the neutral axis, or centroid, which runs longitudinally through the diaphysis...... 40

Figure 6. Locations of the samples included in this project ...... 66

Figure 7. Overview of occupations represented in the Lisbon and Bologna samples using categories presented in Cardoso and Henderson (2013) ...... 70

Figure 8. Green boxes indicate regions for which bone casts were generated. Image modified from that of Fischer-Dückelmann (1911) (public domain) ...... 80

Figure 9. Image showing the number (in parentheses) of measurements gathered for each arm (green) and leg (purple) joint. Image modified from that of Fischer-Dückelmann (1911) (public domain) ...... 82

Figure 10. An overimputation plot illustrating two anomalies (left) in the biorbital breadth measurement. These were both discovered to be entry errors and were corrected prior to analysis...... 86

xii

Figure 11. Plot of body mass on tibial length, pooled sex ...... 97

Figure 12. Plot of body mass on humeral length, pooled sex ...... 97

Figure 13. Plot of body mass on tibial length, male ...... 98

Figure 14. Plot of body mass on humeral length, male ...... 98

Figure 15. Plot of body mass on tibial length, female ...... 99

Figure 16. Plot of body mass on humeral length, female ...... 99

Figure 17. Bar plots of mean coefficient of variation (CV) for each broad anatomical region ...... 101

Figure 18. Figure showing the pattern of CV across skeletal metric traits ...... 104

Figure 19. Multidimensional scaling (MDS) plots generating using craniofacial distance matrices for modern populations ...... 120

Figure 20. Multidimensional scaling (MDS) plots generating using long bone length distance matrices for modern populations ...... 121

Figure 21. Multidimensional scaling (MDS) plots generating using long bone joint distance matrices for modern populations ...... 122

Figure 22. Multidimensional scaling (MDS) plots generating using long bone diaphyseal distance matrices for modern populations ...... 123

Figure 23. Multidimensional scaling (MDS) plots generating using proximal long bone joint distance matrices for modern populations ...... 124

Figure 24. Multidimensional scaling (MDS) plots generating using proximal long bone diaphyseal distance matrices for modern populations ...... 125

Figure 25. Multidimensional scaling (MDS) plots generating using distal long bone joint distance matrices for modern populations ...... 126

Figure 26. Multidimensional scaling (MDS) plots generating using distal long bone diaphyseal distance matrices for modern populations ...... 127

Figure 27. Multidimensional scaling (MDS) plots generating using right arm long bone joint distance matrices for modern populations ...... 128

Figure 28. Multidimensional scaling (MDS) plots generating using right arm long bone diaphyseal distance matrices for modern populations ...... 129

Figure 29. Multidimensional scaling (MDS) plots generating using left arm long bone joint distance matrices for modern populations ...... 130

xiii

Figure 30. Multidimensional scaling (MDS) plots generating using left arm long bone diaphyseal distance matrices for modern populations ...... 131

Figure 31. Multidimensional scaling (MDS) plots generating using leg long bone joint distance matrices for modern populations ...... 132

Figure 32. Multidimensional scaling (MDS) plots generating using leg long bone diaphyseal distance matrices for modern populations ...... 133

Figure 33. Multidimensional scaling (MDS) plots generating using craniofacial distance matrices for modern and Medieval populations...... 136

Figure 34. Multidimensional scaling (MDS) plots generating using long bone length distance matrices for modern and Medieval populations ...... 137

Figure 35. Multidimensional scaling (MDS) plots generating using long bone joint distance matrices for modern and Medieval populations ...... 138

Figure 36. Multidimensional scaling (MDS) plots generating using long bone diaphyseal distance matrices for modern and Medieval populations ...... 139

Figure 37. Multidimensional scaling (MDS) plots generating using proximal long bone joint distance matrices for modern and Medieval populations ...... 140

Figure 38. Multidimensional scaling (MDS) plots generating using proximal long bone diaphyseal distance matrices for modern and Medieval populations ...... 141

Figure 39. Multidimensional scaling (MDS) plots generating using distal long bone joint distance matrices for modern and Medieval populations ...... 142

Figure 40. Multidimensional scaling (MDS) plots generating using distal long bone diaphyseal distance matrices for modern and Medieval populations ...... 143

Figure 41. Multidimensional scaling (MDS) plots generating using right arm long bone joint distance matrices for modern and Medieval populations ...... 144

Figure 42. Multidimensional scaling (MDS) plots generating using right arm long bone diaphyseal distance matrices for modern and Medieval populations ...... 145

Figure 43. Multidimensional scaling (MDS) plots generating using left arm long bone joint distance matrices for modern and Medieval populations ...... 146

Figure 44. Multidimensional scaling (MDS) plots generating using left arm long bone diaphyseal distance matrices for modern and Medieval populations ...... 147

Figure 45. Multidimensional scaling (MDS) plots generating using leg long bone joint distance matrices for modern and Medieval populations ...... 148

xiv

Figure 46. Multidimensional scaling (MDS) plots generating using leg long bone diaphyseal distance matrices for modern and Medieval populations ...... 149

Figure 47. Rates of over- and under-prediction for Modern samples based on Mantel permutations. The “Equal” category refers to situations in which the randomized matrix correlation perfectly matched the correlation observed between the original matrices...151

Figure 48. Rates of over- and under-prediction for Modern and Medieval samples based on Mantel permutations ...... 151

Figure 49. A comparison of femoral and tibial J values (standardized) for heavy and light-industrial subpopulations from Lisbon and Bolognese samples, mean and 95% confidence level (error bars)...... 154

Figure 50. A comparison of humeral and radial J values (standardized) for heavy and light-industrial subpopulations from Lisbon and Bolognese samples, mean and 95% confidence level (error bars)...... 155

Figure 51. MDS plots based on distance matrices for diaphyseal properties (all long bones). Abbreviations: Lisbon light (Lis_Lht) and heavy (Lis_Heavy) intensity occupations, Bologna, light (Bo_Lht) and heavy (Bo_Heavy) intensity occupations .... 158

Figure 52. The four primary regions of South Africa in the early 1900s, two Afrikaner regions in the Transvaal (in purple) and two English regions (in green). Image modified from Wikimedia.org (open access)...... 176

Figure 53. Sample composition by birth and death decade in South Africa. Cape Town (top), Pretoria (middle), Johannesburg (Bottom). Purple band reflects Anglo-Boer War and period during which Boer concentration camps were in use. Red band reflects a period of government-sponsored female English emigrant initiatives and doubling of Transvaal popualtions (van Helton and Williams, 1983). Blue band reflects Apartheid. Note that British immigration into South Africa would steadily decrease up to Apartheid and would have some disruption during the recession and both World Wars...... 184

Figure 54. The Voortrekker Monument, situated just outside Pretoria, serves as a memorial to Afrikaner culture and the Great Trek ...... 185

Figure 55. Aerial view showing the flat profile of Bologna city center ...... 194

Figure 56. One of many hills upon which Lisbon is built ...... 195

Figure 57. People traverse a steep slope common to Lisbon ...... 196

Figure 58. A schematic showing mean midshaft cross-sectional shape of two hypothetical populations (A and B) before (left) and after (right) exposure to bending forces of identical loading direction and magnitude. While bone functional adaptation does generate a plastic response of similar magnitude in both populations, it does not erase the baseline differences that exist among these populations due to genetic or developmental regulation, meaning some genetic structure is still detectable...... 204

Figure 59. Similar to the image above, this figure is a schematic showing mean midshaft cross-sectional shape of two hypothetical populations (A and B) before (left) and after (right) exposure to bending forces of identical loading direction and magnitude. But in this scenario each population has a different, genetically-regulated sensitivity to mechanical stress (as indicated by the double-feedback diagrams—refer to Fig 2.1 of the Background). Under this model, Population B would require much higher levels of strain to elicit bone remodeling than does Population A, resulting in a cross-sectional shape that is little changed after exposure to bending force in Population B...... 205

xvi

LIST OF ABBREVIATIONS

BFA Bone functional adaptation

Bo Bologna

Bo_Heavy Bologna, heavy intensity occupations

Bo_Lht Bologna, light intensity occupations

CT_E Cape Town, South Africa, European ancestry

CSP Cross-sectional properties (diaphyseal properties)

CV Coefficient of variation

D-matrix Distance (genetic) matrix

Eng_Med_N Medieval England, northern regions

Eng_Med_S Medieval England, southern regions

IBD Isolation-by-distance

Ital_Med Medieval Italy

Jo_E Johannesburg, South Africa, European ancestry

Lis Lisbon

Lis_Heavy Lisbon, heavy intensity occupations

Lis_Lht Lisbon, light intensity occupations

MDS Multidimensional scaling

Pr_E Pretoria, South Africa, European ancestry

R-matrix Relationship matrix

Sass Sassari, Sardinia

SouSh South Shields, England

xvii

CHAPTER 1

PROJECT OVERVIEW

1.1 Introduction

Its ability to adapt to the mechanical demands of daily life, integration with other physiological systems, and fact it preserves well in fossil and archaeological contexts makes bone an invaluable means to understand trends across evolutionary time. The long bones of the arms and legs are one of few sources available to directly infer the lifestyle and behavior of relatives long extinct. The ways in which limb bones are interpreted depend largely upon subfield. Functional morphologists, for example, have long used the many entheses marring the surface of these bones as a “muscular roadmap,” one that can reveal many aspects of behavior, including locomotor repertoire, leaping affinity, grasping ability, and arboreal prowess in both extinct and extant primates (Fleagle and

Simons, 1982; MaClatchy et al., 2000; Jablonski et al., 2002). Bioarchaeologists use these same bumps and deformations to interpret activity and occupational stress in humans, and by broadening the scope to populations, entheseal scores in aggregate can

reveal a host of social phenomena, including patterns of sexual division, social

stratification, or subsistence/technological revolutions—though such interpretations are

made with more caution in recent years upon the discovery that entheses, in addition to

muscle use, are also susceptible to the effects of age, size, and sex (Lai and Lovell, 1992;

Chapman, 1997; Molnar, 2006; Mariotti et al., 2007; Villotte et al., 2010; Havelková et

al., 2011; Weiss et al., 2012; Lieverse et al., 2013; Villotte and Knüsel, 2013). The utility

of entheseal markers to understand people long past is joined by other aspects of long

bones, including joint dimensions, a useful means to interpret body size/mass (Auerbach

and Ruff, 2004; Ruff et al., 2012), and degenerative joint diseases, which can indicate

physical activity and workload (Ortner, 1968; Walker and Hollimon, 1989; Bridges,

1992).

Long bone diaphyses, long heralded for their exceptional plasticity, have also

seen widespread use as a template to interpret broad patterns of activity and have long

served as proxy for behavior in both small scale (e.g., terrain navigation or occupation)

and large scale (e.g., subsistence change over time, technological or behavioral

revolutions) analyses (Stock and Pfeiffer, 2001, 2004; Holt, 2003; Weiss, 2003; Ruff,

2005; Shackelford, 2007; Maggiano et al., 2008; Sparacello and Marchi, 2008; Sládek et

al., 2016) (Fig. 1). Diaphyeal (“cross-sectional”) properties have highlighted many important aspects of behavior, particularly locomotor behavior, across the hominin lineage. Examples include identifying arboreal prowess in Ardipithecus ramidus, the degree of bipedality across various Australopithecus or Homo species, and, because leg bone diaphyses have weakened considerably over time, a metric to identify when modern humans began to outsource the burdens of physical activity onto technology rather than their own bodies (Ruff et al., 1993, 1994, 1999, 2015; Trinkaus et al., 1994, 1998;

Trinkaus, 1997; Ruff, 2005, 2008, 2009). Diaphyseal properties have also been put

toward more specific ends, allowing bioarchaeologists to discern patterns of activity,

mobility, and even parse out handedness (Stock and Pfeiffer, 2001, 2004; Holt, 2003;

Weiss, 2003 a; Ruff, 2005; Shackelford, 2007; Maggiano et al., 2008; Sparacello and

Marchi, 2008; Sládek et al., 2016). They have also been used to infer broader

sociocultural trends, such as the sexual division of labor, instances of compassion toward

injured or disabled individuals, and changes to subsistence strategies over time (Ruff and

Hayes, 1983; Marchi et al., 2006; Shackelford, 2007; Wanner et al., 2007; Maggiano et

al., 2008; Sparacello and Marchi, 2008; Weiss, 2009).

In addition to function, the plastic nature of long bones also provides important clues about past stressors, including overall health and nutrition. Stunted growth reflects childhood malnutrition and stress (Jantz and Jantz, 1999; Kemkes-Grottenthaler, 2005;

Cardoso and Caninas, 2010) and declines to mineral content reflect diseases of senility

(Riggs et al., 1981, 1982; Hannan et al., 1992). Irregular contours and localized infections highlight trauma and speak to broader trends of violence and conflict (Lovell, 1997;

Walker, 2001). Long bone lengths and intralimb proportions reflect “ecogeographic” patterns of variation that arise due to environmental factors (e.g., temperature) (Holliday and Falsetti, 1995). And other features of long bones, such as femoral anteversion, are simple byproducts of development (Bonneau et al., 2011).

While the utility of long bones in the many varied contexts above is well established, one application of long bone morphology has been noticeably absent in an

increasingly DNA-focused world—whether they can be used to discern genetic

relationships among populations, an integral component in our understanding of

population history and our ancestral legacy. This point was recently articulated by von

Cramon-Taubadel and Weaver (2009:239, emphasis added):

“variation in the postcranium has been presumed primarily to reflect long-term adaptation across species and/or the actions of environmentally or behaviorally driven phenotypic plasticity. However, the extent to which aspects of postcranial variation are the result of neutral evolutionary forces has not been addressed formally”

The absence of long bones in such contexts is especially noticeable given accumulating evidence that many regions of the skeleton—including craniofacial, dental, and pelvic regions—readily reflect among-population relationships and expose the neutral evolutionary processes of population drift and ancient dispersals (Hanihara and Ishida,

2005; von Cramon-Taubadel, 2009; Betti et al., 2012).

Through this dissertation I investigate whether variation in long bone structure can detect the relationships among populations separated by space and time. The goal of the project is simple—to assess which aspects of long bone morphology better reflect genetic relationships (i.e., are under tighter genetic regulation) and which better reflect behavioral variation (i.e., are more plastic). Its execution is more cumbersome, and

includes analysis of craniofacial and long bone data that encompass both “traditionally

static” and “traditionally plastic” regions of long bones, including lengths, articular

dimensions, and diaphyseal properties in six bones drawn from the leg and both arms.

Diaphyseal metric, shape, rigidity, and strength data were gathered from fourteen

anatomical locations chosen to reflect the typical range of mechanical loadings

4 experienced by long bones in vivo. By its termination, this project generated 10,601 circumferential (cross-sectional) casts (figure 2) and nearly 150,000 individual data points for over 1,000 people across two continents.

Figure 2. A sample of diaphyseal casts generated for this project

The scope of this project is intentionally broad so that a holistic understanding of long bone variation and its utility in genetic reconstructions can be assessed. DNA, as all molecules, will degrade over time. While the past few decades have seen great advances in ancient DNA (aDNA) acquisition and analysis, there will always be fossil and modern skeletal remains for which genetic data cannot be extracted due to age, preservation, contamination, or an unwillingness to allow destructive sampling (for a thorough discussion on the relative impossibility of extracting aDNA in context of fossil hominins, please see Svante Pääbo’s 2014 book and references therein). As stated by Cheverud

(1988:958), “[g]enetic correlations and variances are difficult to measure even in the best

5 of circumstances…and are utterly impossible to measure in many instances, such as in paleontology.” Because of the patchy representation of fossil and bioarchaeological remains, it is necessary to extract as much information from a relatively limited a subset of skeletal material as possible—a subset governed largely by taphonomy and a lot of luck. Because of their bulky size and dense cortical structure, long bone fragments are relatively abundant in the fossil and bioarchaeological records, particularly the diaphyseal regions which are often shunned by scavenging carnivores who prefer the fattier epiphyses (Blumenschine, 1988). Because of the often contentious debate surrounding the classification of new hominin specimens (a trend so consistent it may as well be tradition), extracting genetic or phylogenetic information via morphological variation will always be an important complement to DNA-focused approaches.

The specific objectives of this dissertation are to clarify [1] whether long bones reflect among-population genetic structure to any degree, [2] which general parts of long bones do so best, [3] whether the variable and plastic nature of diaphyseal regions completely erases any meaningful genetic information, and [4] whether some control for among-population genetic structure should be incorporated into future studies of long bone variation—particularly those studies that rely on interpopulation (or interspecific) comparisons. This last point is particularly important, as in his 2004 paper on five

“common statistical blunders” in the anthropological literature, Sokal (2004:114) lists

“ignoring spatial, temporal, and phylogenetic autocorrelation” among traits as one of the most common offenses he encounters. He argues that such autocorrelation can amplify the significance of statistical methods commonly used to test for differences among samples—such as the ANOVAs, t-tests, and chi-squared tests that have featured

prominently in long bone analyses. If long bones are discovered to competently detect underlying among-population relationships, controlling for these will strengthen subsequent behavioral inferences.

1.2 Chapter 1 References

Auerbach B, and Ruff C. 2004. Human body mass estimation: A comparison of “morphometric” and “mechanical” methods. American Journal of Physical Anthropology 125:331–342.

Basso N, and Heersche J. 2006. Effects of hind limb unloading and reloading on nitric oxide synthase expression and apoptosis of osteocytes and chondrocytes. Bone 39:807–14.

Betti L, Cramon-Taubadel N, and Lycett S. 2012. Human Pelvis and Long Bones Reveal Differential Preservation of Ancient Population History and Migration Out of Africa. Human Biology 85:139–152.

Blumenschine R. 1988. An experimental model of the timing of hominid and carnivore influence on archaeological bone assemblages. Journal of Archaeological Science 15:483–502.

Bonneau N, Simonis C, Seringe R, and Tardieu C. 2011. Study of femoral torsion during prenatal growth: Interpretations associated with the effects of intrauterine pressure. American Journal of Physical Anthropology 145:438–445.

Bridges PS. 1992. Prehistoric arthritis in the Americas. Annual Review of Anthropology 21:67 –91.

Cardoso H, and Caninas M. 2010. Secular trends in social class differences of height, weight and BMI of boys from two schools in Lisbon, Portugal (1910–2000). Economics & Human Biology 8:111–120.

Chapman MN. 1997. Evidence for Spanish influence on activity induced musculoskeletal stress markers at Pecos Pueblo. International Journal of Osteoarchaeology 7:497– 506.

Cheverud J. 1988. A Comparison of Genetic and Phenotypic Correlations. Evolution 42:958.

Čuk T, Leben-Seljak P, and Stefancic M. 2001. Lateral asymmetry of human long bones. Variability and Evolution 9:19–32.

Fleagle JG, and Simons EL. 1982. The humerus of Aegyptopithecus zeuxis: A primitive anthropoid. American Journal of Physical Anthropology 59:175–193.

Hanihara T, and Ishida H. 2005. Metric dental variation of major human populations. American Journal of Physical Anthropology 128:287–298.

Hannan MT, Felson DT, and Anderson JJ. 1992. Bone mineral density in elderly men and women: Results from the framingham osteoporosis study. Journal of Bone and Mineral Research 7:547–553.

Havelková P, Villotte S, Velemínský P, Poláček L, and Dobisíková M. 2011. Enthesopathies and activity patterns in the Early Medieval Great Moravian population: Evidence of division of labour. International Journal of Osteoarchaeology 21:487–504.

Holliday TW, and Falsetti AB. 1995. Lower limb length of European early modern humans in relation to mobility and climate. Journal of Human Evolution 29:141– 153.

Holt BM. 2003. Mobility in Upper Paleolithic and Mesolithic Europe: Evidence from the lower limb. American Journal of Physical Anthropology 122:200–215.

Jablonski NG, Leakey MG, Kiarie C, and Antón M. 2002. A new skeleton of Theropithecus brumpti (Primates: cercopithecidae) from Lomekwi, west turkana, Kenya. Journal of Human Evolution 43:887–923.

Jantz L, and Jantz RL. 1999. Secular change in long bone length and proportion in the United States, 1800–1970. American Journal of Physical Anthropology 110:57– 67.

Kemkes‐Grottenthaler A. 2005. The short die young: The interrelationship between stature and longevity—evidence from skeletal remains. Am J Phys Anthropol 128:340–347.

Lai P, and Lovell NC. 1992. Skeletal markers of occupational stress in the Fur Trade: A case study from a Hudson’s Bay Company Fur Trade post. International Journal of Osteoarchaeology 2:221–234.

Lieverse AR, Bazaliiskii V, Goriunova O, and Weber AW. 2013. Lower limb activity in the Cis‐Baikal: Entheseal changes among middle holocene siberi an foragers. American Journal of Physical Anthropology 150:421–432.

Lovell NC. 1997. Trauma analysis in paleopathology. Yearbook of Physical Anthropology: 40:139-170.

MaClatchy L, Gebo D, Kityo R, and Pilbeam D. 2000. Postcranial functional morphology of Morotopithecus bishopi, with implications for the evolution of modern ape locomotion. Journal of Human Evolution 39:159–183.

Maggiano IS, Schultz M, Kierdorf H, Sosa T, Maggiano CM, and Blos V. 2008. Cross‐ sectional analysis of long bones, occupational activities and long‐distance trade of the Classic Maya from Xcambó—Archaeological and osteological evidence. American Journal of Physical Anthropology 136:470–477.

Marchi D, Sparacello VS, Holt BM, and Formicola V. 2006. Biomechanical approach to the reconstruction of activity patterns in Neolithic Western Liguria, Italy. American Journal of Physical Anthropology 131:447–455.

Mariotti V, Facchini F, and Belcastro M. 2007. The study of entheses: proposal of a standardised scoring method for twenty-three entheses of the postcranial skeleton. Collegium Antropologicum 31:291–313.

Molnar P. 2006. Tracing prehistoric activities: Musculoskeletal stress marker analysis of a stone‐age population on the Island of Gotland in the Baltic sea. American Journal of Physical Anthropology 129:12–23.

Ogilvie MD, and Hilton CE. 2010. Cross-sectional geometry in the humeri of foragers and farmers from the prehispanic American Southwest: exploring patterns in the sexual division of labor. American Journal of Physical Anthropology 144:11–21.

Ortner DJ. 1968. Description and classification of degenerative bone changes in the distal joint surfaces of the humerus. American Journal of Physical Anthropology 28:139–155.

Rhodes JA, and Knüsel CJ. 2005. Activity‐related skeletal change in medieval humeri: Cross‐sectional and architectural alterations. American Journal of Physical Anthropology 128:536–546.

Riggs BL, Wahner HW, Dunn WL, Mazess RB, Offord KP, and Melton LJ. 1981. Differential changes in bone mineral density of the appendicular and axial skeleton with aging: relationship to spinal osteoporosis. Journal of Clinical Investigation 67:328–335.

Riggs BL, Wahner HW, Seeman E, Offord KP, Dunn WL, Mazess RB, Johnson KA, and Melton LJ. 1982. Changes in Bone Mineral Density of the Proximal Femur and Spine with Aging. Journal of Clinical Investigation 70:716–723.

Ruff CB. 2005. Mechanical determinants of bone form: insights from skeletal remains. Journal of Musculoskeletal and Neuronal Interactions 5:202–12.

Ruff CB. 2008. Biomechanical Analyses of Archaeological Human Skeletons. In: M Anne, Saunders SR, editors. Second Edition. John Wiley and Songs, Inc. p 183– 206.

Ruff CB. 2009. Relative limb strength and locomotion in Homo habilis. American Journal of Physical Anthropology 138:90–100.

Ruff CB, and Hayes WC. 1983. Cross‐sectional geometry of Pecos Pueblo femora and tibiae—A biomechanical investigation: I. Method and general patterns of variation. American Journal of Physical Anthropology 60:359–381.

Ruff CB, Holt B, Niskanen M, Sladek V, Berner M, Garofalo E, Garvin HM, Hora M, Junno J-A, Schuplerova E, et al. 2015. Gradual decline in mobility with the adoption of food production in Europe. Proceedings of the National Academy of Sciences 112:7147–7152.

Ruff CB, Holt BM, Niskanen M, Sladék V, Berner M, Garofalo E, Garvin HM, Hora M, Maijanen H, Niinimäki S, et al. 2012. Stature and body mass estimation from skeletal remains in the European Holocene. American Journal of Physical Anthropology 148:601–617.

Ruff CB, McHenry HM, and Thackeray JF. 1999. Cross-sectional morphology of the SK 82 and 97 proximal femora. American Journal of Physical Anthropology 109:509–521.

Ruff CB, Trinkaus E, Walker A, and Larsen C. 1993. Postcranial robusticity in Homo. I: Temporal trends and mechanical interpretation. American Journal of Physical Anthropology 91:21–53.

Ruff CB, Walker A, and Trinkaus E. 1994. Postcranial robusticity in Homo. III: Ontogeny. American Journal of Physical Anthropology 93:35–54.

Shackelford LL. 2007. Regional variation in the postcranial robusticity of late upper paleolithic humans. American Journal of Physical Anthropology 133:655–668.

Sládek V, Ruff CB, Berner M, Holt B, Niskanen M, Schuplerová E, and Hora M. 2016. The impact of subsistence changes on humeral bilateral asymmetry in Terminal Pleistocene and Holocene Europe. Journal of Human Evolution 92:37–49.

Sokal RR. 2004. Raymond Pearl’s Legacy: The proper measure of man. American Journal of Human Biology 16:113–124.

Sparacello V, and Marchi D. 2008. Mobility and subsistence economy: A diachronic comparison between two groups settled in the same geographical area (Liguria, Italy). American Journal of Physical Anthropology 136:485–495.

Steele J, and Mays S. 1995. Handedness and directional asymmetry in the long bones of the human upper limb. International Journal of Osteoarchaeology 5:39–49.

Steele J. 2000. Handedness in past human populations: skeletal markers. Laterality: Asymmetries of Body, Brain and Cognition 5:193–220.

Stock J, and Pfeiffer S. 2001. Linking structural variability in long bone diaphyses to habitual behaviors: Foragers from the southern African Later Stone Age and the Andaman Islands. American Journal of Physical Anthropology 115:337–348.

Stock JT, and Pfeiffer SJ. 2004. Long bone robusticity and subsistence behaviour among Later Stone Age foragers of the forest and fynbos biomes of South Africa. Journal of Archaeological Science 31:999–1013.

Trinkaus E, Churchill SE, and Ruff CB. 1994. Postcranial robusticity in Homo. II: Humeral bilateral asymmetry and bone plasticity. American Journal of Physical Anthropology 93:1–34.

Trinkaus E, Ruff CB, Churchill SE, and Vandermeersch B. 1998. Locomotion and body proportions of the Saint-Césaire 1 Châtelperronian Neandertal. Proceedings of the National Academy of Sciences 95:5836–5840.

Trinkaus E. 1997. Appendicular robusticity and the paleobiology of modern human emergence. Proceedings of the National Academy of Sciences 94:13367– 13373.

Villotte S, Churchill SE, Dutour OJ, and Henry-Gambier D. 2010. Subsistence activities and the sexual division of labor in the European Upper Paleolithic and Mesolithic: Evidence from upper limb enthesopathies. Journal of Human Evolution 59:35–43.

Villotte S, and Knüsel CJ. 2013. Understanding entheseal changes: definition and life course changes. International Journal of Osteoarchaeology 23:135–146. von Cramon-Taubadel N. 2009. Revisiting the homoiology hypothesis: the impact of phenotypic plasticity on the reconstruction of human population history from craniometric data. J Hum Evol 57:179–190. von Cramon‐Taubadel N, and W eaver TD. 2009. Insights from a quantitative genetic approach to human morphological evolution. Evolutionary Anthropology 18:237– 240.

Walker PL. 2001. A bioarchaeological perspective on the history of violence. Annual Review of Anthropology 30:573-596.

Walker PL, and Hollimon SE. 1989. Changes in osteoarthritis associated with the development of a maritime economy among southern California Indians. International Journal of Anthropology 4:171–183.

Wanner IS, Sosa T, Alt KW, and Blos V. 2007. Lifestyle, occupation, and whole bone morphology of the pre‐Hispanic M aya coastal population from X cambó, Yucatan, Mexico. International Journal of Osteoarchaeology 17:253–268.

Weiss E, Corona L, and Schultz B. 2012. Sex differences in musculoskeletal stress markers: Problems with activity pattern reconstructions. International Journal of Osteoarchaeology 22:70–80.

Weiss E. 2003a. Effects of rowing on humeral strength. American Journal of Physical Anthropology 121:293–302.

Weiss E. 2003b. Understanding muscle markers: Aggregation and construct validity. American Journal of Physical Anthropology 121:230–240.

Weiss E. 2009. Sex differences in humeral bilateral asymmetry in two hunter‐gatherer populations: California Amerinds and British Columbian Amerinds. American Journal of Physical Anthropology 140:19–24.

CHAPTER 2

BACKGROUND

2.1 Chapter Overview

In this chapter, I present a multifaceted overview of the proximate mechanisms that collectively shape variation in adult skeletal form, beginning at the earliest developmental stages (molecular patterning) before describing the cellular mechanisms that facilitate bone functional adaptation (BFA) through adulthood. I conclude by illustrating how BFA is used to study long bones in a biomechanical context, and present the strengths and limitations of traditional approaches. Each subsection highlights variables that could affect variation in adult skeletal form, be they molecular, developmental, or functional.

2.2 Molecular Development

“Development is relevant to human evolution because the process by which genetic variation is “translated” into morphology is not a one-to-one correspondence. Instead, genetic variation impacts how a series of overlapping and hierarchical developmental events (e.g., patterning, segmentation, proliferation, etc.) unfold over an organism’s ontogeny. On a population level, each of these events adds to, subtracts, and/or alters the magnitude and direction of phenotypic variation in “morphospace.” The challenge of relating development to evolutionary change therefore lies in disentangling this complexity to discover both when and how during development variation is generated in complex phenotypes.” Young and Capellini (2016:127)

From the regression of dolphin hind limbs, to the extreme elongation of bat “fingers” to the complete absence of limbs in snakes, relatively simple changes in gene expression throughout development have profoundly influenced the evolution of vertebrate limb structures (Thewissen et al., 2006; Sakamoto et al., 2009; Sears et al., 2011). While there have been significant advances in our understanding of the machinery that drives limb

development, there remain many unknowns, and thus this is an area of intense (and

frequently contradictory) study. Understanding the molecular mechanisms that fuel

development is further stymied by the fact that genes operate within larger, hierarchical networks of inhibition, activation, and redundancy which makes it difficult to parse out exact function, and further by the fact that the same gene can have markedly different

functions at different developmental stages and in different contexts (Storm and

Kingsley, 1999; Carroll et al., 2005; Vokes et al., 2008; Zeller et al., 2009). As stated in a

review of molecular development of limb structures by Kronenberg (2003:336):

“[M]any important biological phenomena described here have no clear molecular explanation…The control of the profound and progressive slowing of bone growth in late fetal and then early post-natal life is not understood. The pathways used to determine the enormous variety of specific bone shapes and sizes that all arise using the common endochondral development process are unknown…The coordination of growth plate function with the development of joints, tendons and ligaments is under study, but our understanding remains preliminary.”

What is presented in this section is a general overview of human limb development as

best can be inferred at present, with an emphasis on those events whose effects are detectable in the final product: adult skeletal morphology.

The first morphological evidence of the limbs begins when four protrusions extend laterally from the embryonic body wall roughly four weeks postcoitus (Cohn and

Tickle, 1999; Rallis et al., 2005; Sears, 2008; Oberg et al., 2010). It appears that rostrocaudal positioning of the limbs is dictated largely by t-box (TBX) and paired like

homeodomain 1 (PITX1) expression in the flank. In mammals, TBX5 expresses near the

presumptive forelimb field, while TBX4 expresses in the presumptive hind-limb region

(Gibson-Brown et al., 1996; Minguillon et al., 2005). However, neither specifies limb

14 identity but rather triggers and maintains limb outbudding, perhaps by regulating fibroblast growth factor 10 (FGF10) (Minguillon et al., 2005; Fernandez-Teran and Ros,

2008). Experimental studies confirm that altering TBX or PITX1 expression can modify developing limb structures with long-lasting effects that persist into adulthood. For example, individuals who do not express TBX5 will be armless, those who do not express

TBX4 will possess only partially developed hind limbs, and PITX1 mutations are known to generate asymmetric hind-limb malformations across a wide range of taxa, including stickleback fishes and manatees (Shapiro et al., 2004; Minguillon et al., 2005; Gurnett et al., 2008).

As outpocketing advances, early proximodistal patterning is governed by homeobox “HOX” genes. It appears that the expression of HOX genes governing stylopodial elements originates in the lateral plate mesoderm prior to limb outgrowth, while subsequent patterning of zeugopodial and autopodial elements is contingent upon later expression of sonic hedgehog (SHH) genes at the posterior boundary of the limb bud. These overlapping expression domains may be one source of integration across neighboring skeletal elements (Chiang et al., 2001). There are four paralogous HOX gene complexes (A,B,C and D), the apparent result of multiple duplication events in the vertebrate lineage. Each is highly conserved, presumably because HOX genes are critical for regulating the body plan during development (Barham and Clarke, 2008). Only the A and D complex appear to be involved in limb development, however (Zákány and

Duboule, 1999; Carroll et al., 2005; Kmita et al., 2005). Because HOX genes are spatially collinear, meaning their physical location on the DNA corresponds to the physical location of the trait they induce, it is known that expression of HOX9-10 patterns

stylopodial structures while nested expression of HOX9-13 patterns the zeugopod. The

autopod is patterned largely by HOX11-13 and is distinct owing to “different molecular pathways during chondrogenesis and a distinctive third phase of HOXD gene expression”

(Davis et al., 2007; Hamrick, 2007:382). This unique developmental regime may explain the tendency for distal skeletal elements to be more variable than proximal ones, a trend referred to as the “proximal stabilization model” (Hinchliffe, 1991; Hallgrímsson et al.,

2002; Hamrick, 2007). HOX misexpression is associated with many limb anomalies depending on the timing of disruption and whether other regulatory elements are affected.

These anomalies include digit loss, polydactyly, changes to bone length, bipartitions, and malformations of the distal zeugopod and autopod (Favier et al., 1995; Davis and

Capecchi, 1996; Fromental-Ramain et al., 1996; Zákány and Duboule, 1999; Sears, 2008;

Sheth et al., 2012). In the most severe cases, the distal zeugopod and autopod are absent

and the adult limb truncated (Kmita et al., 2005).

Outbudding is associated with the appearance of the apical ectodermal ridge

(AER), an organizer that manifests as a length of rigid, ectodermal epithelial cells

spanning the distal-most tip of the limb bud. It exists for only a short time and will flatten

and wither as limb development nears the terminal phalanges (Fernandez-Teran and Ros,

2008; Doroba and Sears, 2010; Oberg et al., 2010). It must be continually maintained by signals from the underlying mesoderm or will decay to produce a truncated limb

(Fernandez-Teran and Ros, 2008; Doroba and Sears, 2010). The AER is involved in

proximodistal patterning, establishing the dorsoventral boundary, and also stimulates the proliferation of proximally situated mesenchymal cells in the “progress zone.” The AER is rich in the fibroblast growth factors (FGFs) that are critical for proper limb

development (Niswander et al., 1993; Fernandez-Teran and Ros, 2008; Doroba and

Sears, 2010; Oberg et al., 2010).

Cell proliferation, which fuels appositional growth, is governed by interplay between SHH and distal expression of FGF genes from the AER (Suzuki et al., 2004;

Carroll et al., 2005; Kmita et al., 2005; Zákány and Duboule, 1999). SHH is released by a

specialized aggregate of cells in the proximocaudal region of the limb bud shortly after

stylopod development (Chiang et al., 2001). This Zone of Polarizing Activity (ZPA) is another key limb organizer that provides positional information to mesenchymal cells, regulates HOX gene expression, and is especially important in patterning skeletal

elements across the anteroposterior (thumb-to-pinky) plane (Chiang et al., 2001; Carroll

et al., 2005; Tickle, 2006). SHH also stimulates mitosis, generating the supply stores (the

cells) to facilitate outbudding, and therefore plays a key role in establishing the identity

of primordial zeugopodal elements and the digit rays to follow (Johnson and Tabin, 1997;

Litingtung et al., 2002; Zhu et al., 2008; Bouldin et al., 2010). SHH upregulates FGF4

genes in the AER, and FGF4 upregulates SHH genes expressed in the ZPA, creating a

feedback loop between the two that ensures cell proliferation persists until the limb

template is complete (Fernandez-Teran and Ros, 2008; Bouldin et al., 2010). It is

possible that ectodermal hedgehog (HH) buffers the SHH/FGF4 feedback system,

reducing the size of the AER in situations where SHH is overexpressed, and increasing it

when SHH is underexpressed, thus ensuring uniform development across the limbs and

serving as a potential source of inter-limb integration (Bouldin et al., 2010). It appears

that simple shifts in SHH expression can explain major transformations of the limb

throughout evolution, such as regression of hind limbs in cetaceans and fin reduction in

fishes (Thewissen et al., 2006; Sakamoto et al., 2009).

The first evidence of the skeleton manifests as a continuous, longitudinal mass of condensed mesenchyme that begins in the proximal stylopod and extends distally, branching once at the junction of the zeugopod and later creating a mass that will ultimately form the autopod (Hartmann and Tabin, 2001). Joints arise with the onset of chondrification, and joint formation likely stems from the interaction between chondrocyte activators and inhibitors across sequential stages of development (Storm and

Kingsley, 1999; Koyama et al., 2008). One of the first molecular indicators of joint formation is the expression of growth differentiation factor 5 (GDF5) genes at consistent

intervals along each digit ray (Merino et al. 1999, Settle et al. 2003). However, GDF5

does not appear to induce joint formation itself, rather it induces and maintains the

growth of epiphyseal cartilages and adjacent joint structures, including soft tissue (e.g.,

ligaments) (Merino et al., 1999; Storm and Kingsley, 1999; Settle et al., 2003; Koyama et

al., 2008). GDF5 expression may therefore be a source of integration across neighboring

bones that ensures joint integrity and later functional competence. GDF5 expression is

also place-and-time-sensitive in that it cannot induce cartilage formation beyond specific

developmental windows (Storm and Kingsley 1999, Merino et al. 1999, Settle et al.

2003). See figure 3 for a simple schematic of the major genetic networks that regulate the

development of appendicular structures.

As limb development progresses, mesenchymal condensations are gradually

replaced with cartilage (Francis-West et al., 1999; Zeller et al., 2009). Just prior to the

onset of chondrification, cells within the GDF5 expression domains flatten and organize

into two thickened layers between which is a region of loosely packed cells—the site of subsequent cavitation events that will separate the contiguous mesenchymal structures

into neighboring bones (Francis-West et al., 1999; Hartmann and Tabin, 2001). Together,

these layers comprise the “interzone,” or first visible evidence the joint to follow (Storm

and Kingsley, 1999). While the initial positioning and formation of the interzone does not

appear sensitive to movement, subsequent joint cavitation does require mechanical

stimulus, and therefore even in the earliest stages of development, skeletal structures and

movement are connected (Mikic et al., 2000). The interzone (and perhaps GDF5 itself) is

induced by wingless type 14 (WNT14) which, when applied to nonjoint regions of the

digit, can induce a structure much like the interzone (Hartmann and Tabin, 2001; Settle

et al., 2003). Furthermore, WNT14 stimulates activation of the cluster of differentiation

44 (CD44) gene which limits the aggregation of cells in this region and also stimulates

production of hyaluronan, an extracellular lubricant and major constituent of synovial

fluid that will later fill the joint capsule (Francis-West et al., 1999; Hartmann and Tabin,

2001; Tamer, 2014). Hyaluronan is also found in extracellular fluid surrounding mature bone cells where it plays a role in both chondrocyte hypertrophy and later bone

remodeling (Burra et al., 2011; Mackie et al., 2011). Another critical component of joint

formation is the inhibition of bone morphogenetic protein (BMP) by antagonists noggin,

chordin, and members of the FGF family, all of which express in the developing joint at

various stages (Francis-West et al., 1999; Merino et al., 1999).

Locally in each rudimentary long bone, chondrification and ossification are

synchronized. During this process, resting chondrocytes are pulled in from epiphyseal

margins, organize themselves into parallel rows, actively proliferate, hypertrophy, and

19 finally undergo a form of unidentified death that some have termed “chondroptosis”

(Roach and Clarke, 2000; White and Wallis, 2001; Roach et al., 2004; Pineault et al.,

2015). It appears that parathyroid hormone like hormone (PTHLH) genes expressed in epiphyseal regions and indian hedgehog genes (IHH) expressed locally feedback on one another via transforming growth factor beta (TGFβ) to maintain chondrocyte proliferation and trigger osteoblast differentiation in the developing embryo (Ballock and OʼKeefe,

2003; Emons et al., 2011). As a chondrocyte passes from the epiphyseal (“resting”) region to the metaphyseal (“proliferation”) region it matures and secretes vascular endothelial growth factor (VEGF) and other factors to trigger the development of vascular structures (Nilsson et al., 2005). It is by controlling vessel development that the rate of ossification is kept roughly consistent with that of chondrogenesis (Mackie et al.,

2011). The chondrocyte swan song is seen with the secretion of extracellular matrix that generates a structure (“scaffold”) to be subsequently opened by osteoclasts to allow vessels and bone-forming osteoblasts to invade the voids left after chondroptosis and begin production of the bony matrix (Carter et al., 2004; Rolian et al., 2016). This area of mineralization forms the primary ossification center (or “bone collar”) in the middle of the diaphysis which will soon be joined by a variable number of secondary ossification centers situated in epiphyseal regions. This means that long bone appositional growth, both embryonic and postnatal, is largely contingent upon the degree of chondrocyte hypertrophy (Ballock and OʼKeefe, 2003), followed by rate of chondrocyte proliferation and synthesis of extracellular matrix, all of which are governed by both local and systemic factors (der Eerden et al., 2003).

This process is also affected by a number of systemic endocrine factors acting broadly across the skeleton to coordinate growth rate across the limbs (Ballock and OʼKeefe,

2003). Growth plate chondrocytes are therefore sensitive to those stressors that alter circulating hormone concentrations, such as illness or malnutrition (Nilsson et al., 2005).

Both systemic increases to growth hormone (GH) concentrations and its local application to epiphyseal regions in various developing mammalian species, for example, result in increased bone length, presumably by amplifying chondrocyte proliferation. The opposite is true of glucocorticoids (Nilsson et al., 2005). Insulin like growth factor 1

(IGF1), which is sensitive to GH, can increase the rate of proliferation (Ballock and

OʼKeefe, 2003). The effect of hormones on epiphyseal maturation persists until even the latest stages of long bone development: postnatal epiphyseal fusion. As a juvenile reaches maturity, the growth line separating diaphyseal and epiphyseal regions reduces and once

“the proliferative capacity of the growth plate chondrocyte is exhausted” fusion occurs and growth terminates at that epiphyseal junction (Nilsson et al., 2005:161; Emons et al.,

2011). Estrogen has an especially strong influence on the rate of growth plate development, and early-onset or delayed estrogen expose can accelerate or retard growth plate maturation and, as a result, impact final height attained (Nilsson et al., 2005; Emons et al., 2011:386).

Figure 3. A simplified diagram showing the hierarchical nature of embryonic long bone patterning overlaid on an adult macaque skeleton. Purple: generalized expression domain of the PBX genes; Blue/Green gradient: collinear expression of HOXA/D genes along the proximodistal axis; Red: expression domain of SHH genes that regulate anteroposterior patterning of the zeugopod/autopod elements and limb bud outgrowth via reciprocal interaction with distally expressed FGF genes. Mutual inhibition of HAND2 (H2) genes and GLI3 establish anteroposterior polarity of the stylopod, which is patterned prior to SHH expression. Yellow boxes indicate later expression domains of GDF5 genes and WNT/B-Catenin, both implicated in joint interzone formation and subsequent maintenance of joint structures during development (along with noggin, IHH, and several other factors). Arrows reflect interactive relationships between genes or their relationship to specific bony regions. Solid arrows reflect broader/higher level genetic influences, while dotted arrows reflect more regional/localized patterns.

Ossification is governed by many molecular substances, chief among them members of

the BMP superfamily. Despite their name, BMPs are important in the formation and

maintenance of cartilage, and while they are initially triggered by SHH, they express

autonomously by the time mesenchymal condensations appear (Drossopoulou et al.,

2000; Tsumaki et al., 2002). BMPs are also connected to chondrocyte hypertrophy

(Ballock and OʼKeefe, 2003). They likely govern many of the more “detailed” aspects of

skeletal development by regulating the rate of chondrification to affect things such as

bone length and identity (Dahn and Fallon, 2000). Mice with inhibited BMP expression

possess a severely diminished skeleton, likely because formation of the cartilaginous

precursor is impaired (Tsumaki et al., 2002). Conversely, individuals who overexpress

BMP possess longer, more robust limb bones, presumably because BMPs pull more

mesenchymal cells into the skeletal template (Brunet et al., 1998). This is seen in bats

(chiroptera) which possess the longest posterior digits relative to body size of any

mammal, an adaptation that keeps the interdigital webbing taught and facilitates powered

flight (Wang et al., 2010). Many genes in the developing bat forelimb have altered

expression patterns, particularly those of the BMP superfamily that regulate chondrocyte

activity. This is further supported by experimental studies that show the external

application of BMP2 increases digit length in developing mice, while application of BMP

antagonists stunt digit length (Sears et al., 2006).

While many of the examples discussed herein focused on “big scale”

developmental shifts (i.e., ones that could result in speciation events or wholesale loss of

limb structures), it is important to remember that genes are not an “all or nothing” proposition. While genetic knock-outs in nature are relatively uncommon, gene

expression is subject to ongoing modification throughout life. Any of the genetic

pathways described above can be tweaked during development to generate subtle changes

to baseline bone shapes prior to significant mechanical modification. Activity-based

interpretations of long bone morphology lies squarely in the degree to which the

subsequent plastic effects introduced by mechanical action can overcome these

intrinsically regulated differences in “baseline shape.”

2.3 Function Throughout Development

Like all skeletal elements long bones are regulated by intrinsic (molecular) factors, but they, perhaps more than any other skeletal region, also have a critical role in maintaining functional competency, and their susceptibility to mechanical stimulus is established at an early embryonic age. This means that long bones develop at the nexus of these intrinsic and extrinsic signals, which can make parsing out the relative effects of development and function a murky, but important, affair. In this section, I will highlight some of the

(frustratingly) difficult ways that these two different processes interact throughout limb bone development.

As briefly mentioned in section 2.2, mechanical stimulus is necessary to initiate and maintain joint formation and epiphyseal growth, and beginning around the seventh week of life, embryonic limbs will begin to involuntarily flex (Carter et al., 2004; Roddy et al., 2011 a; b). When this movement is blocked by pharamocological or genetic manipulation, articular surfaces become underdeveloped and uneven, and, in extreme cases, refusion of the joint can occur (Mikic et al., 2000; Dowthwaite et al., 2003;

Plochocki, 2004; Pitsillides and Ashhurst, 2008). The joint profiles of immobilized

embryos often take on a flattened, featureless appearance, and some aspects (e.g., the

joint-lining menisci) can regress entirely if not routinely exposed to movement (Mikic et

al., 2000). There is even some evidence that the placement of epiphyseal ossification

centers is driven largely as a response to movement pressures experienced by these

regions in utero (Carter and Wong, 1988). Mechanical stimulus is also important for

healthy bone growth. Movement triggers articular chondrocytes to express genes

favorable to epiphyseal growth—genes whose expression domains coincide with regions

enduring the highest levels of stress (Plochocki et al., 2006; Nowlan et al., 2008).

External manipulation of embryonic limbs can induce the expression of these

chondrification genes (Responte et al., 2012; Rolfe et al., 2013). The relationship

between biophysical stimulus and growth is important given recent modeling experiments

that show even modest changes to the number and proliferation rates of the embryonic

chondrocyte pool for a given bone can generate differences in adult long morphology as

great as those seen between the femora of humans and chimpanzees (Rolian, 2016). It is

worth noting that this pattern is not unique to the long bones, and similar results have

been reported for a number of other skeletal regions, including the temporomandibular

joint, cranial base, intervertebral joints, and articulation points of the pectoral and pelvic

girdles (Wang and Mao, 2002; Habib et al., 2005, 2007; Nowlan et al., 2010).

Even after birth physical activity continues to alter epiphyseal shape and

thickness. Finite elemental models show that the shear stresses of bipedalism in toddling children disproportionately impact the medial femoral condyle, stimulating an increase in its growth rate relative to the lateral and generating the varus (“knock-kneed”)

appearance characteristic of bipedal apes (Shefelbine et al., 2002). This presumably

requires concomitant change to the proximal tibia to maintain knee joint integrity.

Morphological studies confirm that a varus knee is not present at birth, but rather

emerges in tandem with toddling and erect posture. Wheelchair-bound children, therefore, will never acquire this feature (Tardieu and Trinkaus, 1994; Tardieu, 1999,

2010).

Separating function from development is made further difficult by the fact that many “functional traits” emerge in utero and therefore precede the behaviors they presumably reflect. This includes the anteroposterior elongation and asymmetry of the distal femoral epiphysis, both indicators of bipedalism, which are detectable at fetal

stages in humans but not in chimpanzees. Furthermore, the humeral deltoid tuberosity,

whose degree of robusticity is often used to infer the intensity of arm use in life (e.g.,

spear-throwing or rowing), is underdeveloped or absent in immobilized embryos

(Tardieu, 1999; Nowlan et al., 2010). Furthermore, directional asymmetries in arm bones are detectable at fetal stages, with the rates of left/right bias on par with those observed for adults (Pande and Singh, 1971). Whether these prenatal asymmetries arise from developmental programming or from arm movements in utero is unresolved (Blackburn,

2011). Interestingly, these trends remain detectable in adults as directional asymmetries

(DA), with robusticity of the dominant limb being a bit higher than that of the contralateral side. The degree of asymmetry intensifies around age one when handedness manifests and persists into adulthood, therefore serving as a useful means to infer handedness from skeletal remains (Blackburn, 2011). While diaphyseal shape is typically the most asymmetric in such studies, length and joint dimensions do show some

asymmetry with the side biases expected—a result typically attributed to the greater

plasticity/lower canalization of diaphyseal bone compared to length or articular

dimensions (Auerbach and Ruff, 2006; Blackburn, 2011).

There is even some evidence that developing bones can autonomously self-correct

when intended growth trajectories are pushed astray by environmental stresses. For

example, research shows that accelerated (i.e., “catch-up”) growth could ameliorate limb

asymmetries in developing rabbits when the growth of one tibia was artificially and

temporarily slowed. Interestingly, researchers discovered that the stimulus for catch-up

growth arose from within the growth plate itself, perhaps because of an a priori

established number of cell divisions necessary to generate the adult epiphysis that will be

accomplished even if growth rate is temporarily altered, a phenomenon called senescence

(van der Eerden et al, 2003:785 citing Baron et al., 1994 and Gafni and Baron, 2000,

Eamons et al. 2011). This is supported by transplantation exercises in which it was

discovered that “the growth rate of a transplanted growth plate depends on the age of the

donor and not on the age of the recipient” (Emons et al., 2011:386). If true, this could

explain one source of inter-limb integration reported in other studies, and provide a way

for internal developmental mechanisms to continually shape phenotypic variation through

adolescence, even in the face of significant external (environmental) pressure (Young and

Hallgrímsson, 2005). As early as the 1960s, the phenomenon of “catch-up” growth was

cited as evidence of developmental canalization (Prader et al., 1963). By following some intrinsic developmental plan, individual long bones work in concert to ensure functional competence across limbs is maintained throughout ontogeny.

While some influence of mechanical environment on early (embryonic) joint

structures is well established, there is conflicting evidence regarding its influence on the

joints, diaphyses, and lengths at later ages. One study found that adolescent and young

adult tennis players possessed ulnar and metacarpal elongation in their swing arm

compared to the contralateral arm, a pattern the authors connect to mechanically induced

growth acceleration (Krahl et al., 1994). Other researchers, however, conclude that

intense mechanical loadings actually slow growth and stunt length (Villemure and

Stokes, 2009). In a study of treadmill running sheep across three age categories,

(Lieberman et al., 2001) failed to find a consistent relationship between physical activity

and articular dimensions despite significant differences in diaphyseal robusticity. They

conclude that developmental canalization may temper the plastic response of articular

regions which, owing to their functional importance as joints, are developmentally

constrained. Lazenby et al. (2008:881) also suggest that the “necessity to maintain joint

integrity” places limits on the ability of articular morphology to adapt to functional loads

in vivo,” a point mirrored by Garofalo (2013:19) who states that “canalization is

important for the maintenance of joint congruence and, therefore, function.” In the

context of nonhuman primates, Lieberman et al. (2001) suggest that articular regions may

be more suitable to infer taxonomy, while the more plastic diaphyseal regions might be

more appropriate to discern behavior. Ruff (2002) and Auerbach and Ruff (2004) mirror

this sentiment, stating that articular regions are more useful than diaphyseal ones to estimate body mass because the former are less likely to vary in response to individual differences in activity or muscle use.

Whether functional integration occurs as the product of selection acting on development

(meaning it is the result of long term evolutionary pressures) or as an in vivo response to

mechanical stimulus throughout development (meaning it is an individual-level plastic response or homoiology), the end result is the same—those skeletal elements that operate in concert (e.g., mandibular ramus and temporal) tend to covary (Cheverud, 1984).

Interestingly, research shows that the ontogenetic timing of integration may vary by anatomical region and across taxa. For example, in a study of patterns of craniofacial integration throughout development, Ackermann (2005:189) discovered that some morphological integration patterns in chimpanzees, gorillas, and humans emerge early

(during infancy/childhood) and persist through to adulthood, while others emerge later.

Furthermore, the “overall level and patterning of morphological integration across adult apes and humans is remarkably similar” despite obvious differences in facial form.

Interestingly, human morphological variation was generally more flexible than that of the other great apes at comparable ontogenetic stages. However, it should be stated that most of the studies in the area of integration (developmental and/or functional) focus on individuals who are still growing. How strongly these properties reflect differences in adult-onset activity patterns is unclear.

Bones are not isolated entities, but rather elements that operate in concert and must therefore have some degree of integration to ensure proper development and, subsequently, efficient movement. Function is a key driving force across evolutionary

time, and many researchers have argued that selective pressure for functional efficiency has generated a strong degree of (developmental) integration among forelimb and hind- limb elements. For example, Leamy (1977) argued that the tendency for limb

homologues to covary is due, in large part, to function. He specifically argues that intense

selective pressures have acted across evolutionary time to maintain proper forelimb and hind-limb proportions among the tetrapods (Leamy, 1977). The implications of forelimb/hind limb integration persist for primates as well. It has also been suggested, for example, that the short, blunt human thumb was a byproduct of its integration with the homologous hallux—that as selective pressures for bipedalism acted on the latter, the former was a developmental tag-along (Rolian et al., 2009). Even diaphyseal properties may be subject to the strength of interlimb developmental integration. For example, Ruff

(2000) discovered that radial and tibial cross-sections of the Pecos Pueblo were both weaker than those of an East African population believed to engage in similar patterns of physical activity. This is result is somewhat surprising given the fact the radius and tibia are situated in entirely different loading environments, and would presumably have no biomechanical reason to vary in similar fashion. They are, however, developmental homologues—a connection that may explain this tendency to covary.

2.4 Bone Functional Adaptation: Overview and Application

In this section, I will present a different aspect of relationship between bone and function—the ongoing, dynamic relationship between mechanical stimulus and bone morphology. The idea that long bones adapt to their functional environment is hardly new. Wolff (1892), when speaking of the “arch-like” orientation of trabeculae in the proximal femur, argued that some aspects of long bone structure were developmentally designed to resist mechanical loadings. Although he did not suggest an ongoing connection between bone form and mechanical stimulus, he was the first to suggest a

30 relationship predictable enough to be summarized mathematically (Burr et al., 2002).

“Wolff’s law” grew to be a common fixture in the world of adaptive bone remodeling despite the subsequent debunking of his mathematical work and later protestations by skeptics who felt its “many flaws and exceptions…rendered the term somewhat useless”

(Pearson and Lieberman, 2004:65). Today, most modern practitioners agree that it doesn’t matter whether Wolff was absolutely right, but generally right. To distance themselves from the often tedious “strict versus generalized” debates, some researchers refrain from using the phrase “Wolff’s law” altogether, opting instead for more descriptive labels (e.g., “bone functional adaptation” or “BFA”) to describe the process by which bone converts mechanical stimulus into physiological response (Ruff et al.,

2006).

BFA occurs via a “double feedback” mechanism wherein a significant change to mechanical environment elicits a concomitant reorganization of bone tissue, thereby modifying morphology (Ruff et al., 2006; Fig. 4). BFA differs from earlier hypotheses in its flexibility—for example, the Mechanostat Hypothesis (Frost, 1987) held that the intensity of remodeling was contingent upon a gradient of strain-defined “stages.” Under

BFA, whenever long bones are exposed to biomechanical stimuli that exceed the norm

(probably somewhere around 0.35% over the average), cortical bone remodels in planes enduring the highest stress, refashioning the bone to withstand the new loads imposed upon it (Hung et al., 2013). Conversely, when the intensity of routine mechanical loads decreases, bone will wither and weaken. Cross-sectional properties (CSPs) are interpreted from diaphyseal geometry in a transverse (cross-sectional) plane. Maximum and minimum diameters, shape ratios, and total area therefore offer a means to estimate a

given bone’s ability to resist deformation (rigidity) and fracture (strength) in a given

plane of interest and, when the scope broadens, an indirect means to infer patterns of

physical activity and behavior (Ruff, 2008). This will be discussed in more detail in

section 2.5.

Microscopic inspection of adult diaphyseal cortical bone reveals a network of

interconnected bone cells locked within a dense, crystalline matrix. Perhaps the best- supported hypothesis about the cellular mechanisms behind BFA is the “fluid shear” or

“lacuno-canalicular” hypothesis, which holds that osteocytes—the bone cells that convert mechanical stimulus into physiological response—do so by registering changes in fluid movement through the bony matrix in which they are entombed (Vezeridis et al., 2006) .

Specifically, when a bone is bent, pressure increases at the region of highest compressive strain, forcing fluids to travel to areas of lower tensile strains via a series of interconnected lacunae and canaliculi (small and smaller tunnels) weaving through the bone. Osteocytes locked inside the lacunae register these fluid changes and respond by communicating with likewise entombed neighbors via long, tentacle-like protrusions

(“dendritic processes”). These osteocytic connections create a “cellular communication network (CCN)” that “[i]n effect…gives the bone a sort of nervous system” (Pearson and

Lieberman, 2004:69). The lacuno-canalicular network connects to large blood vessels, most of which originate in the medullary cavity. It is therefore probably no coincidence that mature osteoblasts (bone-building cells) are often found lining endosteal regions where they can hitch an easy ride deep into cortical bone when called upon to do so

(Kassem et al., 2008; Burra et al., 2011).

The fluid found within the cortical bone is a lifeline for the trapped osteocytes in that it

provides moisture, nourishment, and waste removal, and some researchers believe this is

primary impetus behind bone remodeling (Burr et al., 2002; Heino et al., 2002; Wang et

al., 2004). Bone is an environment, and the long-lived osteocytes have a vested interest in

not dying. Studies of unloaded bones show that small particles easily diffuse from the

blood vessel through the narrowest canaliculi, but that the spread of particles is greatly

enhanced with just a handful of mechanical loading cycles, particularly in regions under

tensile strain (Tate et al., 1998; Knothe Tate and Knothe, 2000). Furthermore, larger

particles, such as proteins and signaling molecules, need help moving through the tight

canalicular spaces. This help likely arises from mechanically induced changes to fluid

pressure (Wang et al., 2004). In this way, osteocyte survival is, to some degree,

contingent upon regular mechanical stimulus.

This coordinated activity among osteocytes, osteoblasts, and osteoclasts connects

bone form to function, and is therefore the cellular backbone of BFA. Research shows

that fluid shear “activates several cellular processes in osteocytes, including energy

metabolism, gene activation, growth factor production, and matrix synthesis,” and also

triggers the release of signaling molecules (Burger and Klein-Nulend, 1999:S103).

Among the products released by the osteocytes are nitric oxide (NO), prostaglandins

(PTG) and insulin like growth factor (IGF), all indicators of bone formation. NO and transforming growth factor beta (TGF-β) are continually released by osteocytes to inhibit osteoclast activity, and when disrupted, bone withers (Heino et al., 2002; Burger et al.,

2003). Prostaglandins (e.g., PGE2 and IGF) are released by osteocytes to stimulate

recruitment of pre-osteoblasts, prompting in bone enlargement (Burger and Klein-

Nulend, 1999). Interestingly, ecNOS is an enzyme similar to a vascular hormone whose expression is proportional to blood pressure and which presumably regulates blood vessel enlargement (Burger and Klein-Nulend, 1999; Chen et al., 1999; Ehrlich and Lanyon,

2002). Osteocytes also express nuclear factor kappa B (NFKB) to stimulate growth of blood vessels to nourish new bone (Jilka, 2013). Changes in fluid movement are apparently also detected by mechanosensitive osteoblasts, which express β1-integrin and

α-actinin under high shear stress to form cellular adhesion sites that anchor them to the surrounding bone matrix (Bass et al., 2002).

As osteoblasts produce bone, some become embedded within the new cortical layers and

transform into osteocytes, thereby expanding the mechanosensitive network (Franz‐

Odendaal et al., 2006; Burra et al., 2011). Bit by bit over time, the periosteal boundaries

of cortical bone shift in response to ever-changing biomechanical demands. It is clear that

these cells and their products (signaling molecules) are the primary BFA coordinators.

The aforementioned “generalized versus localized” debates surrounding BFA, therefore,

apparently hinge on how far these cellular products diffuse away from regions under

stress—a process difficult to model given the minute diameters and complex 3D

branching patterns of the lacuno-cannicular structures (Knothe Tate and Knothe, 2000;

Hung et al., 2013). Regardless of the specificity of the remodeling range, the cumulative effect of these localized cellular processes is periosteal expansion in regions under high loading, for example the swing arms of racquet-sport athletes (Heinonen et al., 1995;

Haapasalo et al., 2000; Daly et al., 2004; Ducher et al., 2005), and withering of bone when loading stresses are reduced, as is seen in people in low gravity environments or on prolonged bed rest (Rubin and Lanyon, 1984; Lanyon, 1992).

The relationship between long bone properties and mechanical stimulus is supported both by the manipulation of live animals and the study of people with unique lifestyles, such as athletes and obese individuals (Ruff and Jones, 1981; Rubin and

Lanyon, 1984; Heinonen et al., 1995; Bennell et al., 1997; Kontulainen et al., 2002;

Warner et al., 2002; Moore, 2008; Agostini, 2009; Agostini and Ross, 2011). Some generalizations can be summarized from these varied investigations: [1] long bones respond best to dynamic (rather than static) loadings, [2] relatively few loading cycles

(e.g., four daily wing flaps in roosters or ten daily jumps in mice) are sufficient to induce

bone activity, [3] higher frequency, higher intensity stimuli elicit a stronger bony

response (up to a point), and [4] static loads (e.g., sitting or standing) have little effect on

diaphyseal structure (Rubin and Lanyon, 1984 a; Lanyon, 1992). Each of these are best

explained by the fluid shear hypothesis, as higher intensity bending would cause more

shearing of the osteocyte membrane as fluid rushes away from regions of high pressure,

while more frequent (dynamic) bending cycles would generate repeated fluid movements that facilitate the rapid transport of both nourishment and waste products (Klein‐Nulend

et al., 1995; Bacabac et al., 2004; Pearson and Lieberman, 2004). Furthermore, it appears

that the degree of bony response to mechanical stress is to some degree metabolic—cells

perform better when given a rest between cycles (Burr et al., 2002).

Again, sex, age, and growth hormones have a significant here and can alter BFA

sensitivity. Bone cells possess sex-hormone receptors, and their behavior can be

moderated by hormone exposure, particularly to estrogen (Heino et al. 2002). Diaphyseal properties vary significantly by age and sex, in large part because of hormone-driven

differences in remodeling rates (plasticity) (Lindahl and Lindgren, 1967; Ruff and Hayes,

1983, 1984, 1988; Yano et al., 1984; Pearson and Lieberman, 2004; Riggs et al., 2004).

Research shows that throughout childhood, males and females enjoy similar levels of

bone remodeling, but this changes with the onset of adolescence when females begin to

experience estrogen-mediated “bone packing,” presumably so they can generate a

mineral store for use during lactation later in life (Järvinen et al., 2003). While this excess

mineral makes the bone stronger (relative to body size), it does so by rendering the

remodeling process “more sluggish,” meaning female bone is generally less responsive to

mechanical stimulus (Järvinen et al., 2003). Later in life as females reach menopause,

estrogen levels decline and bone becomes more sensitive. As such, women begin to lose

density in their load-bearing bones, especially in endosteal regions (Feik et al., 1996).

This process is particularly relevant for the study of diaphyseal properties because the body compensates for cortical bone loss by pushing the periosteal margin further outward from the neutral axis so that bone rigidity can be maintained in the face of declining material strength (Ruff and Hayes, 1982, 1983; Stein et al., 1998). However, these compensatory mechanisms are stimulated by physical activity which, being relatively low in modern populations, means the declines in mass and rigidity with age are not fully compensated for—a deficiency that disproportionately affects females (Ruff and Hayes,

1988; Feik et al., 1996). This phenomenon is one of the primary explanations for the high frequency of osteoporosis and hip-fracture in aging females (Barbour et al., 2014).

Research in sports medicine consistently shows that long bones are most responsive to activity when training starts prior to maturity, with the most intense remodeling responses observed during puberty (Haapasalo et al., 1994; Krahl et al., 1994;

Kontulainen et al., 2001). It appears that the benefits of this “hyper-reactivity” persist well beyond adolescence into middle-age, even if physical activity diminishes.

Furthermore, the gains to cortical bone achieved during adolescence are more exaggerated in females (Khan et al., 2000; Kontulainen et al., 2006). However, there is evidence that most activity-induced accumulation in adolescents is seen along the endosteal, not periosteal, bone surfaces (Bass et al., 2002).

2.5 “Applied” Bone Functional Adaptation

Activity-mediated changes to bone shape is the aggregate of these cellular processes.

Because long bones regularly modify themselves in response to new activity patterns,

subtle differences in diaphyseal shape should reflect the type and magnitude of internal

(e.g., muscle pull) and external (e.g., ground reaction forces) loadings imposed upon the

bone in life. Furthermore, rigidity (resistance to deformation) and strength (resistance to

fracture) estimates can be extracted from diaphyseal shape properties and used to gauge

the physical demands placed upon the bone in life (Ruff, 2008). Extracting biomechanical

data from bone design is accomplished by extending mechanical engineering principles

to long bone cross-sections (a “cross-section” is akin to a “slice” in CT scans). These methods will be described only briefly here (for a full review, see Ruff and Hayes, 1983;

Larsen, 1997).

Under bending, the diaphysis of a long bone experiences both compressive and

tensile stress—compression on the side of the bending, tension on the opposite. Think,

for example, of bending a green twig between your thumbs and pointer fingers—the

thumb-side of the twig will experience strong compressive stress, the finger-side, tensile.

This means that somewhere in between is a region of equilibrium in which the two

contrasting forces negate one another. This region, called the neutral axis, spans

diaphyseal length and is typically depicted in cross-sections as the centroid (see Fig. 5).

In beam-like structures, the margins furthest from the centroid opposite the plane of bending will experience the highest stress levels, with intensity decreasing toward the neutral axis. It should be noted, however, that the central placement of the neutral axis is approximate and in reality is true only when a bone is loaded under bending stress alone.

Once bending is combined with some other loading stress, such as torsion or axial

compression, there will be displacement of the neutral axis (Lieberman et al., 2004). This

displacement can affect the raw values of predicted cross-sectional properties (CSPs),

though the pattern of CSPs calculated through true (i.e., identified by use of strain gauges

on living specimens) and hypothetical neutral axes are consistent (Lieberman et al.,

2004).

All rigidity measures are calculated from a plane passing through the neutral axis.

Rigidity measures include second moments of area (SMA or I), which correlate to

distance from the neutral axis and reflect a bone’s ability to resist bending deformation in

a given plane of interest. Four SMAs are typically assessed in BFA studies, two that

correspond to the anatomical anteroposterior (AP) and mediolateral (ML) axes (I and I ,

X Y respectively), and two that correspond to the axes in which bone is strongest (I ) and

MAX weakest (I ), often called principal SMAs. Angle θ reflects the displacement (in

MIN degrees or radians) of the principal from anatomical axes and therefore provides an

estimate of the directionality of the strongest bending forces acting on that particular

section. Additional variables of interest are J, resistance to torsion (torsion is commonly

experienced by bone due to interacting effects of muscle antagonism) and total area,

which reflects the overall distribution of bone in a transverse plane and resistance to

compression. Strength (Z) properties are differentiated from rigidity (I/SMA) on the

premise that if a bone were to fail under extreme loading burdens, this failure would

happen in those regions under the greatest stress—namely the peripheral regions.

Strength measures standardize a given SMA against distance from the centroid/neutral

axis to account for the likelihood that a bone will fail (fracture) at these extreme margins.

The most common cross-sectional properties seen in BFA studies are summarized in

Table 1 and Figure 5.

Table 1. Summary of common cross-sectional shape, rigidity, and strength properties* Property Abbr. Measurement Criteria Reflects Longitudinal diaphyseal axis and Neutral Axis -- Centroid of the cross-section point of intersection of SMAs Distribution of bone throughout the Total or Cross- TA or Total area within the periosteal section, resistance to compressive Sectional Area CSA margins loadings Sum of squared distances from Bending resistance to I the ML axis (in mm4) anteroposterior-oriented loadings X Sum of squared distances from Bending resistance to mediolateral- I 4 Second the AP axis (in mm ) oriented loadings Moments of Y Sum of squared distances from Bending resistance to the strongest Area I the narrowest plane passing mechanical loadings (SMA) through the neutral axis (in mm4) MAX Sum of squared distances from Bending resistance to the weakest I the widest plane passing through mechanical loadings (perpendicular the neutral axis (in mm4) to I ) MIN The sums of any two Polar Second MAX perpendicular SMAs. For this Overall measure of robusticity, Moment of J project, J is calculated from resistance to torsional loadings Area I and I

MAX MIN Reflects orientation of principal SMAs relative to anatomical Angle of displacement of Imax Theta θ SMAs, and therefore planes of (relative to Ix) greatest and weakest bending rigidites I divided by distance from the Bending strength against Z centroid to the outermost fiber of X anteroposterior-oriented loadings the ML axis (in mm3) X I divided by distance from the Bending strength against Z centroid to the outermost fiber of Y mediolateral-oriented loadings the AP axis (in mm3) Y Section Moduli I divided by distance from Bending strength against strongest Z the centroid to the outermost MAX mechanical loadings fiber of the I plane (in mm3) MAX I dividedMAX by distance from Bending strength against weakest Z the centroid to the outermost mechanical loadings (perpendicular MIN fiber of the I plane (in mm3) to I ) MIN Polar Section The sums of any two Z MIN TorsionalMAX strength Modulus perpendicular section moduli P General shape of the section, general indicator of the orientation Ratio of any two perpendicular of bone rigidity (i.e., a ratio of 1 Shape ratios I /I bending rigidities. indicates the section is circular and MAX MIN that bone is equally resistant to bending deformation in all planes) *Information summarized from (Jungers and Minns, 1979; Ruff and Hayes, 1983; Larsen, 1997; Ruff, 2008)

2.6 Bone Functional Adaptation: Limitations and Concerns

There are limitations to the interpretability of cross-sectional properties, most centering on uncertainty about the scope and specificity of the remodeling process. Some researchers find fault in the theoretical foundations of this approach, arguing that key assumptions (specifically, that forces move through the centroidal axis) are without merit, and that interpretations based on cross-sectional geometry may be reliable in only a very general sense (Lieberman et al., 2004). Also problematic is the tendency for BFA-related research to focus primarily upon “the exceptions,” meaning individuals targeted for their unique activity patterns (e.g., athletes, obese individuals, prehistoric rowers, astronauts).

This would include an emphasis on preadults who are more likely to exhibit exaggerated remodeling responses and on animals subject to unnatural loading environments engineered to generate a strong bony response (Hert et al., 1969, 1971; Ruff and Jones,

1981; Rubin and Lanyon, 1984 a; Block et al., 1989; Bennell et al., 1997; Mosley et al.,

1997; Hagino et al., 2001; Faulkner et al., 2003; Lovejoy et al., 2003; de Souza et al.,

2005; Nikander et al., 2006; Fredericson et al., 2007; Moore, 2008; Agostini, 2009; Shaw and Stock, 2009; Agostini and Ross, 2011). Such research provides critical information about how bone responds to atypical loading environments but provides less information about its behavior in more natural contexts, such as majority-adult populations whose level of daily activity was neither “all” nor “nothing.” Because bone remodeling is triggered mainly by the extreme ranges of mechanical stimulus (overuse and underuse)

(Robling et al., 2006), the magnitude of remodeling response seen in these targeted studies may not be wholly applicable to those seen in populations drawn from more natural settings. As stated by Collard and Wood (2007:575), “while mechanical loading

has been shown to influence many skeletal characters, the applicability of some rather

extreme experimental studies (e.g., osteotomies) to natural variation is questionable.”

Another complicating factor for BFA is seen in studies that fail to connect

diaphyseal variation to known activity. Multiple studies of varying scope confirm that the relationship between diaphyseal shape and known behavior is not consistent in humans, non-human primates, or even across mammals, and that nonhuman primates, a group characterized by functional diversity far exceeding that seen across humans, often fail to show correspondence between observed biomechanical loadings and predictions based on diaphyseal properties (Demes et al., 1998, 1999, 2000, 2001; Polk et al., 2000; Lieberman et al., 2004; Weiss, 2005; Wescott, 2006; O’Neill and Dobson, 2008). For example, in a

survey of strepsirrhines, Demes and Jungers (1993) found and inconsistent relationship between humeral and femoral bending strengths and primary locomotor strategy, with several leaping primate species having unexpectedly low values while the slow-moving

Perodicticus potto had unexpectedly high values. Lovejoy et al. (2002, 2003) discovered structural properties of the proximal femur to be similar in humans and other apes despite the markedly different loading contexts from which each arises, a discovery that prompted them to support developmental shifts as the primary mechanism driving long

bone variation. In a comparison of wild and captive chimpanzees, Morimoto et al. (2011)

similarly concluded that ontogenetic programming explained the bulk of femoral

diaphyseal variability between wild and zoo chimpanzees, and that bone functional

adaptation had only a limited influence on morphology.

This problem presents for human samples as well. Wescott (2006), for example,

evaluated leg bone robusticity among several Native American groups who engaged in

43 different activity levels and who navigated different terrain types. He failed to find a consistent connection between mobility and midshaft femur shape. In a comparison of humeral shape across four groups that routinely rowed boats to three that did not, Weiss,

(2003) found that both males and females of ocean-rowing groups possessed significantly robust humeri even though this activity was associated largely with males. In a subsequent study, she found no significant difference in humeral robusticity between highly-worked 18th century Quebecois prisoners-of-war and more suburban

Southwesterners of a similar demographic profile (Weiss, 2005). To explain this discrepancy, she suggests that long bone robusticity may be influenced by a number of other factors that include age, sex, and ethnicity. Results such as these may indicate that activity-induced plasticity cannot always mask effects due to intrinsic causes, such as ontogenetic programming, different hormone concentrations, or genetic variation.

2.7 Chapter 2 References

Ackermann R. 2005. Ontogenetic integration of the hominoid face. Journal of Human Evolution 48:175–197.

Agostini GM. 2009. The relationship between body mass index and long bone morphology. MA Thesis: North Carolina State University, Raleigh

Agostini GM, and Ross AH. 2011. The effect of weight on the femur: A cross‐sectional analysis. Journal of Forensic Sciences 56:339–343.

Auerbach B, and Ruff C. 2004. Human body mass estimation: A comparison of “morphometric” and “mechanical” methods. American Journal of Physical Anthropology 125:331–342.

Auerbach BM, and Ruff CB. 2006. Limb bone bilateral asymmetry: variability and commonality among modern humans. Journal of Human Evolution 50:203–218.

Bacabac RG, Smit TH, Mullender MG, Dijcks SJ, Loon J, and Klein-Nulend J. 2004. Nitric oxide production by bone cells is fluid shear stress rate dependent. Biochemical and Biophysical Research Communications 315:823–829.

Ballock RT, and OʼKeefe RJ. 2003. The biology of the growth plate. The Journal of Bone and Joint Surgery (American) 85:715–726.

Barbour KE, Lui LY, Ensrud KE, Hillier TA, LeBlanc ES, Ing SW, Hochberg MC, Cauley JA, and Study of Osteoporotic Fractures (SOF) Research Group. 2014. Inflammatory markers and risk of hip fracture in older white women: the study of osteoporotic fractures. Journal of Bone and Mineral Research 29:2057–2064.

Barham G, and Clarke NM. 2008. Genetic regulation of embryological limb development with relation to congenital limb deformity in humans. Journal of Children’s Orthopaedics 2:1–9.

Baron J, Klein KO, Colli MJ, Yanovski JA, Novosad JA, Bacher JD, Cutler Jr GB. 1994. Catch-up growth after glucocorticoid excess: a mechanism intrinsic to the growth plate. Endocrinology 135:1367-1371.

Bass SL, Saxon L, Daly RM, Turner CH, Robling AG, Seeman E, and Stuckey S. 2002. The effect of mechanical loading on the size and shape of bone in pre‐, peri‐ , and postpubertal girls: A study in tennis players. Journal of Bone and Mineral Research 17:2274–2280.

Bennell KL, Malcolm SA, Khan KM, Thomas SA, Reid SJ, Brukner PD, Ebeling PR, and Wang JD. 1997. Bone mass and bone turnover in power athletes, endurance athletes, and controls: a 12-month longitudinal study. Bone 20:477–484.

Betti L, Balloux F, Amos W, Hanihara T, and Manica A. 2009. Distance from Africa, not climate, explains within-population phenotypic diversity in humans. Proceedings of the Royal Society B: Biological Sciences 276:809–814.

Blackburn A. 2011. Bilateral asymmetry of the humerus during growth and development. American Journal of Physical Anthropology 145:639–646.

Block JE, Friedlander AL, Brooks GA, Steiger P, Stubbs HA, and Genant HK. 1989. Determinants of bone density among athletes engaged in weight-bearing and nonweight-bearing activity. Journal fo Applied Physiology 67:1100–1105.

Bouldin CM, Gritli-Linde A, Ahn S, and Harfe BD. 2010. Shh pathway activation is present and required within the vertebrate limb bud apical ectodermal ridge for normal autopod patterning. Proceedings of the National Academy of Sciences 107:5489–5494.

Brunet LJ, McMahon JA, McMahon AP, and Harland RM. 1998. Noggin, cartilage morphogenesis, and joint formation in the mammalian skeleton. Science 280:1455–1457.

Burger EH, Klein-Nulend J, and Smit TH. 2003. Strain-derived canalicular fluid flow regulates osteoclast activity in a remodelling osteon—a proposal. Journal of Biomechanics 36:1453–1459.

Burger, and Klein-Nulend. 1999. Mechanotransduction in bone--role of the lacuno- canalicular network. FASEB journal : official publication of the Federation of American Societies for Experimental Biology 13 Suppl:S101–12.

Burr DB, Robling AG, and Turner CH. 2002. Effects of biomechanical stress on bones in animals. Bone 30:781–6.

Burra S, Nicolella DP, and Jiang JX. 2011. Dark horse in osteocyte biology: glycocalyx around the dendrites is critical for osteocyte mechanosensing. Communicative and Integrative Biology 4:48–50.

Carroll SB, Grenier JK, and Weatherbee SD. 2005. From DNA to diversity: molecular genetics and the evolution of animal design. 3rd ed. Malden: Blackwell Publishing Ltd.

Carter DR, Beaupré GS, Wong M, Smith LR, Andriacchi TP, and Schurman DJ. 2004. The mechanobiology of articular cartilage development and degeneration. Clinical Orthopaedics and Related Research 427S:S69–S77.

Carter DR, and Wong M. 1988. Mechanical stresses and endochondral ossification in the chondroepiphysis. Journal of Orthopaedic Research 6:148–154.

Chen KD, Li YS, Kim M, Li S, Yuan S, and Chien S. 1999. Mechanotransduction in response to shear stress. Roles of receptor tyrosine kinases, integrins, and Shc. Journal of Biological Chemistry 274:18393–18400.

Cheverud JM. 1984. Quantitative genetics and developmental constraints on evolution by selection. Journal of Theoretical Biology 110:155–171.

Chiang C, Litingtung Y, Harris MP, Simandl BK, Li Y, Beachy PA, and Fallon JF. 2001. Manifestation of the Limb Prepattern: Limb Development in the Absence of Sonic Hedgehog Function. Developmental Biology 236:421–435.

Cohn M, and Tickle C. 1999. Developmental basis of limblessness and axial patterning in snakes. Nature 399:474–479.

Collard M, and Wood B. 2007. Hominin homoiology: An assessment of the impact of phenotypic plasticity on phylogenetic analyses of humans and their fossil relatives. Journal of Human Evolution 52:573–584.

Dahn RD, and Fallon JF. 2000. Interdigital regulation of digit identity and homeotic transformation by modulated BMP signaling. Science 289:438–441.

Daly RM, Saxon L, Turner CH, Robling AG, and Bass SL. 2004. The relationship between muscle size and bone geometry during growth and in response to exercise. Bone 34:281–287.

Davis AP, and Capecchi MR. 1996. A mutational analysis of the 5’ HoxD genes: dissection of genetic interactions during limb development in the mouse. Development 122:1175–85.

Davis M, Dahn R, and Shubin N. 2007. An autopodial-like pattern of Hox expression in the fins of a basal actinopterygian fish. Nature 447:473–476.

Demes B, Fleagle JG, and Jungers WL. 1999. Takeoff and landing forces of leaping strepsirhine primates. Journal of human evolution 37:279–92.

Demes B, Jungers WL, and Walker C. 2000. Cortical bone distribution in the femoral neck of strepsirhine primates. Journal of Human Evolution 39:367–379.

Demes B, and Jungers WL. 1993. Long bone cross-sectional dimensions, locomotor adaptations and body size in prosimian primates. Journal of human Evolution 25:57–74.

Demes B, Stern JT, Hausman MR, Larson SG, McLeod KJ, and Rubin CT. 1998. Patterns of strain in the macaque ulna during functional activity. American Journal of Physical Anthropology 106:87–100.

Demes, Qin, Stern, and Larson. 2001. Patterns of strain in the macaque tibia during functional activity.

Doroba CK, and Sears KE. 2010. The Divergent Development of the Apical Ectodermal Ridge in the Marsupial Monodelphis domestica. The Anatomical Record 293:1325–1332.

Dowthwaite GP, Flannery CR, Flannelly J, Lewthwaite JC, Archer CW, and Pitsillides AA. 2003. A mechanism underlying the movement requirement for synovial joint cavitation. Matrix Biology 22:311–322.

Drossopoulou G, Lewis KE, Sanz-Ezquerro JJ, Nikbakht N, McMahon AP, Hofmann C, and Tickle C. 2000. A model for anteroposterior patterning of the vertebrate limb based on sequential long-and short-range Shh signalling and Bmp signalling. Development 127:1337–1348.

Ducher G, Courteix D, Même S, Magni C, Viala JF, and Benhamou CL. 2005. Bone geometry in response to long-term tennis playing and its relationship with muscle volume: A quantitative magnetic resonance imaging study in tennis players. Bone 37:457–466.

der Eerden BC, Karperien M, and Wit JM. 2003. Systemic and local regulation of the growth plate. Endocrine Reviews 24:782–801.

Ehrlich PJ, and Lanyon LE. 2002. Mechanical Strain and Bone Cell Function: A Review. Osteoporosis International 13:688–700.

Emons J, Chagin AS, Sävendahl L, Karperien M, and Wit JM. 2011. Mechanisms of Growth Plate Maturation and Epiphyseal Fusion. Hormone Research in Paediatrics 75:383–391.

Faulkner RA, Forwood MR, Beck TJ, Mafukidze JC, Russell K, and Wallace W. 2003. Strength indices of the proximal femur and shaft in prepubertal female gymnasts. Medicine and Science in Sports and Exercise 35:513–518.

Favier B, Meur LM, Chambon P, and Dollé P. 1995. Axial skeleton homeosis and forelimb malformations in Hoxd-11 mutant mice. Proceedings of the National Academy of Sciences 92:310–314.

Feik SA, Thomas CD, and Clement JG. 1996. Age trends in remodeling of the femoral midshaft differ between the sexes. Journal of Orthopaedic Research 14:590–7.

Fernandez-Teran M, and Ros M. 2008. The Apical Ectodermal Ridge: morphological aspects and signaling pathways. International Journal of Developmental Biology 52:857–871.

Francis-West PH, Parish J, Lee K, and Archer CW. 1999. BMP/GDF-signalling interactions during synovial joint development. Cell and Tissue Research 296:111–119.

Franz‐Odendaal TA, Hall BK, and W itten EP. 2006. Buried alive: How osteoblasts become osteocytes. Developmental Dynamics 235:176–190.

Fredericson M, Chew K, Ngo J, Cleek T, Kiratli J, and Cobb K. 2007. Regional bone mineral density in male athletes: a comparison of soccer players, runners and controls. British Journal of Sports Medicine 41:664–668.

Fromental-Ramain C, Warot X, Messadecq N, LeMeur M, Dollé P, and Chambon P. 1996. Hoxa-13 and Hoxd-13 play a crucial role in the patterning of the limb autopod. Development 122:2997–3011.

Frost HM. 1987. Bone “mass” and the “mechanostat”: a proposal. The Anatomical Record 219:1–9.

Gafni R, Baron J. 2000. Catch-up growth: possible mechanisms. Pediatric nephrology 14(7):616-619.

Garofalo EM. 2013. Environmental and genetic effects on growth of the human skeleton- --a bioarchaeological investigation. PhD Dissertation: Johns Hopkins University, Baltimore

Gibson-Brown JJ, Agulnik SI, Chapman DL, Alexiou M, Garvey N, Lee SM, and Papaioannou VE. 1996. Evidence of a role for T-box genes in the evolution of limb morphogenesis and the specification of forelimb/hindlimb identity. Mechanisms of Development 56:93–101.

Gurnett C, Alaee F, Kruse L, Desruisseau D, Hecht J, Wise C, Bowcock A, and Dobbs M. 2008. Asymmetric Lower-Limb Malformations in Individuals with Homeobox PITX1 Gene Mutation. The American Journal of Human Genetics 83:616–622.

Haapasalo H, Kannus P, Sievanen H, Heinonen A, Oja P, and Vuori I. 1994. Long-term unilateral loading and bone mineral density and content in female squash players. Calcified Tissue International 54:249–255.

Haapasalo H, Kontulainen S, Sievänen H, Kannus P, Järvinen M, and Vuori I. 2000. Exercise-induced bone gain is due to enlargement in bone size without a change in volumetric bone density: a peripheral quantitative computed tomography study of the upper arms of male tennis players. Bone 27:351–357.

Habib H, Hatta T, Rahman O, Yoshimura Y, and Otani H. 2007. Fetal jaw movement affects development of articular disk in the temporomandibular joint. Congenital Anomalies 47:53–57. Habib H, Hatta T, Udagawa J, Zhang L, Yoshimura Y, and Otani H. 2005. Fetal jaw movement affects condylar cartilage development. Journal of Dental Research 84:474–479.

Hagino H, Okano T, Akhter MP, Enokida M, and Teshima R. 2001. Effect of parathyroid hormone on cortical bone response to in vivo external loading of the rat tibia. Journal of bone and mineral metabolism 19:244–50.

Hallgrímsson B, Willmore K, and Hall BK. 2002. Canalization, developmental stability, and morphological integration in primate limbs. American Journal of Physical Anthropology S35:131–158.

Hamrick MW. 2007. Evolvability, Limb Morphology, and Primate Origins. In: Ravosa MJ, Dagosto M, editors. New York: Springer. p 381–401.

Hartmann C, and Tabin CJ. 2001. Wnt-14 plays a pivotal role in inducing synovial joint formation in the developing appendicular skeleton. Cell 103:341–351.

Heino TJ, Hentunen TA, and Väänänen KH. 2002. Osteocytes inhibit osteoclastic bone resorption through transforming growth factor‐β: Enhancement by estrogen*. Journal of Cellular Biochemistry 85:185–197.

Heinonen A, Oja P, Kannus P, Sievanen H, Haapasalo H, Mänttäri A, and Vuori I. 1995. Bone mineral density in female athletes representing sports with different loading characteristics of the skeleton. Bone 17:197–203.

Hěrt J, Lisková M, and Landa J. 1971. Reaction of bone to mechanical stimuli. 1. Continuous and intermittent loading of tibia in rabbit. Folia morphologica 19:290–300.

Hěrt J, Lisková M, and Landrgot B. 1969. Influence of the long-term, continuous bending on the bone. An experimental study on the tibia of the rabbit. Folia morphologica 17:389–99.

Hinchliffe JR. 1991. Developmental approaches to the problem of transformation of limb structure in evolution. In: J, Hurle JM, Summerbell D, editors. New York: Springer. p 313–323.

Hung BP, Hutton DL, and Grayson WL. 2013. Mechanical control of tissue-engineered bone. Stem Cell Research and Therapy 4:1–7.

Jilka RL. 2013. The Relevance of Mouse Models for Investigating Age-Related Bone Loss in Humans. The Journals of Gerontology Series A: Biological Sciences and Medical Sciences 68:1209–1217.

Johnson RL, and Tabin CJ. 1997. Molecular models for vertebrate limb development. Cell 90:979–990.

Jungers WL, and Minns RJ. 1979. Computed tomography and biomechanical analysis of fossil long bones. American Journal of Physical Anthropology 50:285–90.

Järvinen T, Kannus P, and Sievänen H. 2003. Estrogen and Bone—a Reproductive and Locomotive Perspective. Journal of Bone and Mineral Research 18:1921–1931.

Kassem M, Abdallah BM, and Saeed H. 2008. Osteoblastic cells: differentiation and trans-differentiation. Archives of Biochemistry and Biophysics 473:183–7.

Khan K, McKay HA, Haapasalo H, Bennell KL, Forwood MR, Kannus P, and Wark JD. 2000. Does childhood and adolescence provide a unique opportunity for exercise to strengthen the skeleton? Journal of Science and Medicine in Sport 3:150–164.

Klein‐Nulend J, Roelofsen J, Sterck JG, Semeins CM , and Burger EH. 1995. Mechanical loading stimulates the release of transforming growth factor‐β

activity by cultured mouse calvariae and periosteal cells. Journal of Cellular Physiology 163:115–119.

Kmita M, Tarchini B, Zàkàny J, Logan M, Tabin CJ, and Duboule D. 2005. Early developmental arrest of mammalian limbs lacking HoxA/HoxD gene function. Nature 435:1113–1116.

Knothe Tate ML, and Knothe U. 2000. An ex vivo model to study transport processes and fluid flow in loaded bone. Journal of Biomechanics 33:247–254.

Kontulainen S, Kannus P, Haapasalo H, Sievänen H, Pasanen M, Heinonen A, Oja P, and Vuori I. 2001. Good Maintenance of Exercise‐Induced Bone Gain with Decreased Training of Female Tennis and Squash Players: A Prospective 5‐Y ear Follow‐Up Study of Y oung and Old Starters and Controls. Journal of Bone a nd Mineral Research 16:195–201.

Kontulainen S, Macdonald H, and McKay H. 2006. Change in Cortical Bone Density and Its Distribution Differs between Boys and Girls during Puberty. The Journal of Clinical Endocrinology & Metabolism 91:2555–2561.

Kontulainen S, Sievänen H, Kannus P, Pasanen M, and Vuori I. 2002. Effect of Long‐ Term Impact‐Loading on M ass, Size, and Estimated Strength of Humerus and Radius of Female Racquet‐Sports Pl ayers: A Peri pheral Quanti tati ve Computed Tomography Study Between Young and Old Starters and Controls. Journal of Bone and Mineral Research 17:2281–2289.

Koyama E, Shibukawa Y, Nagayama M, Sugito H, Young B, Yuasa T, Okabe T, Ochiai T, Kamiya N, Rountree RB, et al. 2008. A distinct cohort of progenitor cells participates in synovial joint and articular cartilage formation during mouse limb skeletogenesis. Developmental Biology 316:62–73.

Krahl H, Michaelis U, Pieper HG, Quack G, and Montag M. 1994. Stimulation of bone growth through sports. A radiologic investigation of the upper extremities in professional tennis players. The American Journal of Sports Medicine 22:751–7.

Kronenberg HM. 2003. Developmental regulation of the growth plate. Nature 423:332– 336.

Lanyon LE. 1992. Control of bone architecture by functional load bearing. Journal of Bone and Mineral Research America 7:S369–S375.

Larsen CS. 1997. Bioarchaeology: interpreting behavior from the human skeleton. Cambridge: Cambridge University Press .

Lazenby R, Cooper D, Angus S, and Hallgrímsson B. 2008. Articular constraint, handedness, and directional asymmetry in the human second metacarpal. J Hum Evol 54:875–885.

Leamy L. 1977. Genetic and environmental correlations of morphometric traits in randombred house mice. Evolution [Internet]. Available from: http://www.jstor.org/stable/2407758

Lieberman D, Devlin M, and Pearson O. 2001. Articular area responses to mechanical loading: effects of exercise, age, and skeletal location. Am J Phys Anthropol 116:266–277.

Lieberman DE, Polk JD, and Demes B. 2004. Predicting long bone loading from cross‐ sectional geometry. American Journal of Physical Anthropology 123:156–171.

Lindahl O, and Lindgren AG. 1967. Cortical bone in man 1. Variation of the amount and density with age and sex. Acta Orthopaedica Scandinavica 38:133–140.

Litingtung Y, Dahn RD, Li Y, Fallon JF, and Chiang C. 2002. Shh and Gli3 are dispensable for limb skeleton formation but regulate digit number and identity. Nature 418:979–983.

Lovejoy CO, Meindl RS, Ohman JC, Heiple KG, White TD. 2002. The Maka femur and its bearing on the antiquity of human walking: Applying comtemporary concepts of morphogenesis to the human fossil record. American Journal of Physical Anthropology 119:97-133.

Lovejoy CO, McCollum MA, and Reno PL. 2003. Developmental biology and human evolution. Annual Review of Anthropology 32:85–109.

Mackie EJ, Tatarczuch L, and Mirams M. 2011. The skeleton: a multi-functional complex organ. The growth plate chondrocyte and endochondral ossification. Journal of Endocrinology 211:109–121.

Manica A, Amos W, Balloux F, and Hanihara T. 2007. The effect of ancient population bottlenecks on human phenotypic variation. Nature 448:346–348.

Merino R, Macias D, Ganan Y, Economides AN, Wang X, Q Q, N N, K T, Varona P, and Hurle JM. 1999. Expression and function of Gdf-5 during digit skeletogenesis in the embryonic chick leg bud. Developmental Biology 206:33–45.

Mikic B, Johnson TL, Chhabra AB, Schalet BJ, Wong M, and Hunziker EB. 2000. Differential effects of embryonic immobilization on the development of fibrocartilaginous skeletal elements. Journal of Rebabilitation Research and Development 37:127–134.

Minguillon C, Buono J, and Logan M. 2005. Tbx5 and Tbx4 Are Not Sufficient to Determine Limb-Specific Morphologies but Have Common Roles in Initiating Limb Outgrowth. Developmental Cell 8:75–84.

Moore MK. 2008. Body mass estimation from the human skeleton. PhD Dissertation: University of Tennessee, Knoxville.

Morimoto N, de León MS, and Zollikofer CP. 2011. Exploring femoral diaphyseal shape variation in wild and captive chimpanzees by means of morphometric mapping: a test of Wolff’s law. Anatomical Record 294:589–609.

Mosley JR, March BM, Lynch J, and Lanyon LE. 1997. Strain magnitude related changes in whole bone architecture in growing rats. Bone 20:191–198.

Nikander R, Sievänen H, Uusi-Rasi K, Heinonen A, and Kannus P. 2006. Loading modalities and bone structures at nonweight-bearing upper extremity and weight- bearing lower extremity: a pQCT study of adult female athletes. Bone 39:886–94.

Nilsson O, Marino R, Luca F, Phillip M, and Baron J. 2005. Endocrine regulation of the growth plate. Hormone Research in Paediatrics 64:157–165.

Niswander L, Tickle C, Vogel A, Booth I, and Martin GR. 1993. FGF-4 replaces the apical ectodermal ridge and directs outgrowth and patterning of the limb. Cell [Internet]. Available from: http://www.sciencedirect.com/science/article/pii/0092867493903913

Nowlan NC, Murphy P, and Prendergast PJ. 2008. A dynamic pattern of mechanical stimulation promotes ossification in avian embryonic long bones. Journal of Biomechanics 41:249–258.

Nowlan NC, Sharpe J, Roddy KA, Prendergast PJ, and Murphy P. 2010. Mechanobiology of embryonic skeletal development: Insights from animal models. Birth Defects Research C: Embryo Today: Reviews 90:203–213.

Oberg KC, Feenstra JM, Manske PR, and Tonkin MA. 2010. Developmental Biology and Classification of Congenital Anomalies of the Hand and Upper Extremity. The Journal of Hand Surgery 35:2066–2076. O’Neill MC, and Dobson SD. 2008. The degree and pattern of phylogenetic signal in primate long-bone structure. Journal of Human Evolution 54:309–322.

Pande BS, and Singh I. 1971. One-sided dominance in the upper limbs of human fetuses as evidenced by asymmetry in muscle and bone weight. Journal of Anatomy 109:457–459.

Pearson OM, and Lieberman DE. 2004. The aging of Wolff’s “law”: ontogeny and responses to mechanical loading in cortical bone. American Journal of Physical Anthropology 39:63–99.

Pineault KM, Swinehart IT, Garthus KN, Ho E, Yao Q, Schipani E, Kozloff KM, and Wellik DM. 2015. Hox11 genes regulate postnatal longitudinal bone growth and growth plate proliferation. Biology Open 4:1538–1548.

Pitsillides AA, and Ashhurst DE. 2008. A critical evaluation of specific aspects of joint development. Developmental Dynamics 237:2284–2294.

Plochocki JH, Riscigno CJ, and Garcia M. 2006. Functional adaptation of the femoral head to voluntary exercise. The Anatomical Record Part A 288:776–81.

Plochocki JH. 2004. Bilateral variation in limb articular surface dimensions. American Journal of Human Biology 16:328–33.

Polk J, Demes B, Jungers W, Biknevicius A, Heinrich R, and Runestad J. 2000. A comparison of primate, carnivoran and rodent limb bone cross-sectional properties: are primates really unique? Journal of Human Evolution 39:297–325.

Prader A, Tanner JM, von Harnack GA. 1963. Catch-up growth following illness or starvation: An example of developmental canalization in man. The Journal of Pediatrics 5(5):646-659.

Rallis C, Buono J, and Logan MP. 2005. Tbx3 can alter limb position along the rostrocaudal axis of the developing embryo. Development 132:1961–1970.

Relethford JH, and Jorde LB. 1999. Genetic evidence for larger African population size during recent human evolution. American Journal of Physical Anthropology 108:251–60.

Relethford JH. 1998. Genetics of modern human origins and diversity. Annual Review of Anthropology 27:1–23.

Responte DJ, Lee JK, Hu JC, and Athanasiou KA. 2012. Biomechanics-driven chondrogenesis: from embryo to adult. The FASEB Journal 26:3614–3624.

Riggs LB, Melton JL, Robb RA, Camp JJ, Atkinson EJ, Peterson JM, Rouleau PA, McCollough CH, Bouxsein ML, and Khosla S. 2004. Population‐based study of age and sex differences in bone volumetric density, size, geometry, and structure at different skeletal sites. Journal of Bone and Mineral Research 19:1945–1954.

Roach H, Aigner T, and Kouri J. 2004. Chondroptosis: a variant of apoptotic cell death in chondrocytes? Apoptosis 9:265–277.

Roach HI, and Clarke NM. 2000. Physiological cell death of chondrocytes in vivo is not confined to apoptosis. Journal of Bone and Jount Surgery (British) 82:601–613.

Robling AG, Castillo AB, and Turner CH. 2006. Biomechanical and molecular regulation of bone remodeling. Annual Review of Biomedical Engineering 8:455–498.

Roddy KA, Kelly GM, van Es MH, Murphy P, and Prendergast PJ. 2011a. Dynamic patterns of mechanical stimulation co-localise with growth and cell proliferation during morphogenesis in the avian embryonic knee joint. Journal of Biomechanics 44:143–149.

Roddy KA, Prendergast PJ, and Murphy P. 2011b. Mechanical influences on morphogenesis of the knee joint revealed through morphological, molecular and computational analysis of immobilised embryos. PloS one 6:e17526.

Rolfe R, Roddy K, and Murphy P. 2013. Mechanical Regulation of Skeletal Development. Current Osteoporosis Reports 11:107–116.

Rolian C, Kivell TL, Lemelin P, Richmond BG, and Schmitt D. 2016. The role of genes and development in the evolution of the primate hand. In: New York: Springer. p 101–130.

Rolian C, Lieberman DE, and Hallgrímsson B. 2009. The coevolution of human hands and feet. Evolution 64:1558–1568.

Rolian C. 2016. Tinkering with Growth Plates: A Developmental Simulation of Limb Bone Evolution in Hominoids. In: Boughner JC, Rolian C, editors. Hoboken: Wiley-Blackwell. p 139–166.

Roseman CC, and Weaver TD. 2007. Molecules versus morphology? Not for the human cranium. BioEssays 29:1185–1188.

Ross AH, Ubelaker DH, and Falsetti AB. 2002. Craniometric variation in the Americas. Human Biology 74:807–818.

Ross AH. 2004. Regional isolation in the Balkan region: An analysis of craniofacial variation. American Journal of Physical Anthropology 124:73–80.

Rubin CT, and Lanyon LE. 1984a. Regulation of bone formation by applied dynamic loads. Journal of Bone and Joint Surgery America 66:397–402.

Rubin CT, and Lanyon LE. 1984b. Regulation of bone formation by applied dynamic loads. Journal of Bone and Joint Surgery America 66:397–402.

Ruff CB, and Hayes WC. 1982. Subperiosteal expansion and cortical remodeling of the human femur and tibia with aging. Science 217:945–948.

Ruff CB, and Hayes WC. 1984. Bone-mineral content in the lower limb. Relationship to cross-sectional geometry. Journal of Bone and Joint Surgery America 66:1024– 1031.

Ruff CB, and Hayes WC. 1988. Sex differences in age‐related remodeling of the femur and tibia. Journal of orthopaedic research 6:886–896.

Ruff CB, and Jones HH. 1981. Bilateral asymmetry in cortical bone of the humerus and tibia—sex and age factors. Human Biology 53:69–86.

Ruff CB. 2000. Body size, body shape, and long bone strength in modern humans. Journal of Human Evolution 38:269–290.

Ruff CB. 2002. Long bone articular and diaphyseal structure in old world monkeys and apes. I: locomotor effects. American journal of physical anthropology 119:305– 42.

Ruff CB. 2008. Biomechanical Analyses of Archaeological Human Skeletons. In: M Anne, Saunders SR, editors. Second Edition. John Wiley and Songs, Inc. p 183– 206.

Ruff, Holt, and Trinkaus. 2006. Who’s afraid of the big bad Wolff?:“Wolff’s law” and bone functional adaptation.

Sakamoto K, Onimaru K, Munakata K, Suda N, Tamura M, Ochi H, and Tanaka M. 2009. Heterochronic Shift in Hox-Mediated Activation of Sonic hedgehog Leads to Morphological Changes during Fin Development. PLoS ONE 4:e5121.

Sears KE, Behringer RD, Rasweiler JJ, and Niswander LA. 2006. Development of bat flight: morphologic and molecular evolution of bat wing digits. Proceedings of the National Academy of Sciences 103:6581–6586.

Sears KE, Bormet AK, Rockwell A, Powers LE, Cooper L, and Wheeler MB. 2011. Developmental basis of mammalian digit reduction: a case study in pigs. Evolution & Development 13:533–541.

Sears KE. 2008. Molecular determinants of bat wing development. Cells Tissues Organs 187:6–12.

Settle SH, Rountree RB, Sinha A, Thacker A, Higgins K, and Kingsley DM. 2003. Multiple joint and skeletal patterning defects caused by single and double mutations in the mouse Gdf6 and Gdf5 genes. Developmental Biology 254:116– 130.

Shapiro MD, Marks ME, Peichel CL, Blackman BK, Nereng KS, Jónsson B, Schluter D, and Kingsley DM. 2004. Genetic and developmental basis of evolutionary pelvic reduction in threespine sticklebacks. Nature 428:717–723.

Shaw CN, and Stock JT. 2009. Habitual throwing and swimming correspond with upper limb diaphyseal strength and shape in modern human athletes. American Journal of Physical Anthropology 140:160–172.

Shefelbine SJ, Tardieu C, and Carter DR. 2002. Development of the femoral bicondylar angle in hominid bipedalism. Bone 30:765–770.

Sheth R, Marcon L, Bastida FM, Junco M, Quintana L, Dahn R, Kmita M, Sharpe J, and Ros MA. 2012. Hox genes regulate digit patterning by controlling the wavelength of a Turing-type mechanism. Science 338:1476–1480. de Souza S, Faintuch J, Valezi A, Anna A, Gama-Rodrigues J, de Fonseca I, Souza R, and Senhorini R. 2005. Gait cinematic analysis in morbidly obese patients. Obesity Surgery 15:1238–1242.

Sparks CS, and Jantz RL. 2002. A reassessment of human cranial plasticity: Boas revisited. Proceedings of the National Academy of Sciences 99:14636–14639.

Stein MS, Thomas CD, Feik SA, Wark JD, and Clement JG. 1998. Bone size and mechanics at the femoral diaphysis across age and sex. Journal of Biomechanics 31:1101–1110.

Storm E, and Kingsley D. 1999. GDF5 Coordinates Bone and Joint Formation during Digit Development. Dev Biol 209:11–27.

Suzuki T, Takeuchi J, Koshiba-Takeuchi K, and Ogura T. 2004. Tbx Genes Specify Posterior Digit Identity through Shh and BMP Signaling. Developmental Cell 6:43–53.

Tamer TM. 2014. Hyaluronan and synovial joint: function, distribution and healing. Interdisciplinary Toxicology 6:111–125.

Tardieu C, and Trinkaus E. 1994. Early ontogeny of the human femoral bicondylar angle. American Journal of Physical Anthropology 95:183–195.

Tardieu C. 1999. Ontogeny and phylogeny of femoro-tibial characters in humans and hominid fossils: functional influence and genetic determinism. American Journal of Physical Anthropology 110:365–377.

Tardieu C. 2010. Development of the human hind limb and its importance for the evolution of bipedalism. Evolutionary Anthropology: Issues, News, and Reviews 19:174–186.

Tate ML, Niederer P, and Knothe U. 1998. In vivo tracer transport through the lacunocanalicular system of rat bone in an environment devoid of mechanical loading. Bone 22:107–117.

Thewissen JG, Cohn MJ, Stevens LS, Bajpai S, Heyning J, and Horton WE. 2006. Developmental basis for hind-limb loss in dolphins and origin of the cetacean bodyplan. Proceedings of the National Academy of Sciences 103:8414–8418.

Tickle C. 2006. Making digit patterns in the vertebrate limb. Nature Reviews Molecular Cell Biology 7:45–53.

Tsumaki N, Nakase T, Miyaji T, Kakiuchi M, Kimura T, Ochi T, and Yoshikawa H. 2002. Bone Morphogenetic Protein Signals Are Required for Cartilage Formation and Differently Regulate Joint Development During Skeletogenesis. Journal of Bone and Mineral Research 17:898–906.

Umemura Y, Sogo N, and Honda A. 2002. Effects of intervals between jumps or bouts on osteogenic response to loading. Journal of Applied Physiology 93:1345–1348.

van der Eerden BCJ, Karperien M, Wit JM. 2003. Endocrine Reviews 24(6):782-801

Vezeridis PS, Semeins CM, Chen Q, and Klein-Nulend J. 2006. Osteocytes subjected to pulsating fluid flow regulate osteoblast proliferation and differentiation. Biochemical and Biophysical Research Communications 348:1082–1088.

Villemure I, and Stokes IA. 2009. Growth plate mechanics and mechanobiology. A survey of present understanding. Journal of biomechanics 42:1793–803.

Vokes SA, Ji H, Wong WH, and McMahon AP. 2008. A genome-scale analysis of the cis-regulatory circuitry underlying sonic hedgehog-mediated patterning of the mammalian limb. Genes & Development 22:2651–2663. von Cramon-Taubadel N, and Lycett SJ. 2008. Brief communication: human cranial variation fits iterative founder effect model with African origin. American journal of physical anthropology 136:108–13.

Wang L, Ciani C, Doty SB, and Fritton SP. 2004. Delineating bone’s interstitial fluid pathway in vivo. Bone 34:499–509.

Wang X, and Mao JJ. 2002. Chondrocyte proliferation of the cranial base cartilage upon in vivo mechanical stresses. Journal of Dental Research 81:701–705.

Wang Z, Dong D, Ru B, Young RL, Han N, Guo T, and Zhang S. 2010. Digital gene expression tag profiling of bat digits provides robust candidates contributing to wing formation. BMC Genomics 11:1–12.

Warner SE, Shaw JM, and Dalsky GP. 2002. Bone mineral density of competitive male mountain and road cyclists. Bone 30:281–286.

Weiss E. 2003. Effects of rowing on humeral strength. American Journal of Physical Anthropology 121:293–302.

Weiss E. 2005. Humeral cross‐sectional morphology from 18th century Quebec prisoners of war: Limits to activity reconstruction. American Journal of Physical Anthropology 126:311–317.

Wescott DJ. 2006. Effect of mobility on femur midshaft external shape and robusticity. American Journal of Physical Anthropology 130:201–213.

White A, and Wallis G. 2001. Endochondral ossification: a delicate balance between growth and mineralisation. Current biology 11:R589–91.

Willmore KE, Young NM, and Richtsmeier JT. 2007. Phenotypic variability: its components, measurement and underlying developmental processes. Evolutionary Biology 34:99–120.

Wolff J. 1892. Das Gesetz der Transformation der Knochen. DMW - Deutsche Medizinische Wochenschrift 19:1222–1224.

Yano K, Wasnich RD, and Vogel JM. 1984. Bone mineral measurements among middle- aged and elderly Japanese residents in Hawaii. American Journal of Epidemiology 119:751–764.

Young NM. 2003. A reassessment of living hominoid postcranial variability: implications for ape evolution. Journal of Human Evolution 45:441–464.

Young NM, Hallgrímsson B. 2005. Serial homology and the evolultion of mammalian limb covariation structure. Evolution 59(12):2691-2704.

Young NM, and Capellini TD. 2016. Out on a limb: development and the evolution of the human appendicular skeleton. In: Boughner JC, Rolian C, editors. New York: John Wiley and Sons Inc. p 101–137.

Zeller R, López-Ríos J, and Zuniga A. 2009. Vertebrate limb bud development: moving towards integrative analysis of organogenesis. Nature Reviews Genetics 10:845– 858.

Zhu J, Nakamura E, Nguyen M-T, Bao X, Akiyama H, and Mackem S. 2008. Uncoupling Sonic Hedgehog Control of Pattern and Expansion of the Developing Limb Bud. Developmental Cell 14:624–632.

Zákány J, and Duboule D. 1999. Hox genes in digit development and evolution. Cell and Tissue Research 296:19–25.

CHAPTER 3

PROJECT JUSTIFICATION AND PREDICTIONS

3.1 Project Justification

As seen in the Background, there are many different mechanisms by which intrinsic and

extrinsic processes incrementally sculpt skeletal variation. The paleoanthropological and

bioarcheological records are biased toward adults, so the question of whether adult long

bone properties reflect function to the exclusion of developmental, genetic, or endocrine

signals is not trivial. This project investigates patterns of long bone variation in naturally

occurring human populations engaging in relatively normal patterns of physical activity

(i.e., it does not rely on athletes, obese individuals, or lab-based manipulations) and

therefore reflects an assemblage of people likely to be encountered in the many

cemeteries, burial sites, and other taphonomic contexts for which skeletons studies are

traditionally studied. Another major strength of this research lies in its embrace of

methods widely used in craniofacial studies, meaning the relative success of long bone

genetic predictions can be compared to a large number of existing studies based on other

anatomical criteria and in other species. The need for deeper investigation into how

intraspecific/intrapopulation patterns of variation impact (phylo)genetic patterns is frequently invoked in the recent literature, and this project is designed to contribute to this end (Lycett and Collard 2005, Collard and Wood 2007).

3.2. Goals and Predictions

The goals of this project are three-fold: [1] to conduct a general assessment of whether

long bones reflect among-population genetic structure in mainland and immigrant

European populations, [2] to assess which general aspect of long bone morphology (e.g.,

joints, lengths, or diaphyses) best reflect among-population genetic relationships, and [3] to specifically test whether the plastic nature of diaphyseal bone is sufficiently strong to erase any meaningful evidence of genetic structure as is currently assumed in practice.

Based on information presented in the Background, it is possible to distill these goals into a number of targeted predictions:

1. If it is true that craniofacial variation better mirrors genetic

structure than do long bone elements, craniofacial variation will generate

more reliable among-population distance predictions than will any of the

long bone properties. Multidimensional scaling (MDS) plots based on

craniofacial distance matrices will show better regional clustering of the

samples than do those generated using various long bone properties.

2. If it is true that more canalized structures imply “a greater degree of

genetic constraint” (Lazenby et al. 2008:881) and that they are

therefore more buffered against environmental (behavioral)

perturbations, coefficients of variation (CV) for craniofacial and articular

traits will be lower than those reported for long bone lengths or diaphyseal

properties, indicating that variation in the former is more intrinsically

constrained. MDS plots for craniofacial and articular dimensions are

expected to be similar, and are expected to show geographic clustering to a

stronger degree than will diaphyseal properties which are likely more

susceptible to in vivo environmental effects (e.g., activity, malnutrition,

stress). Mantel tests are expected to show a stronger correlation between

craniofacial and articular distance matrices than either shows to distance

matrices generated by diaphyseal or length data. This is expected both

because of greater plasticity (in individuals) and higher variance (across populations) of the latter.

3a. If it is true that diaphyseal regions are far more plastic than other

skeletal regions, the introduction of Medieval populations to the analysis

(the “temporal test”) will have a stronger impact on genetic predictions

made by diaphyseal properties than those made by craniofacial or joint

properties. This is not to say an effect on the other skeletal regions is

unexpected, rather that the effect will be strongest in the diaphyseal data due to in vivo adaptation to different loading regimes (i.e., plasticity). It is expected that MDS plots for the temporal test will show clustering of populations by time (rather than geography) for diaphyseal data, while the other data types (crania, joints) will primarily show geographic organization.

Mantel correlations between craniofacial and diaphyseal distance matrices will weaken in the temporal test due to competing signals of genetics (crania) and function (diaphyses), and may not be significant.

3b. If it is true that diaphyseal regions are far more plastic than other

skeletal regions, the separation of South European samples into subsamples

based on “heavy intensity” and “light intensity” occupations will strongly

skew the genetic predictions made using diaphyseal properties due to

interference of functional signals, and may disrupt any genetic clustering

that was present for this region.

3.3. Chapter 3 References

Lazenby R, Cooper D, Angus S, and Hallgrímsson B. 2008. Articular constraint, handedness, and directional asymmetry in the human second metacarpal. J Hum Evol 54:875–885.

Lycett SJ, and Collard M. 2005. Do homoiologies impede phylogenetic analyses of the fossil hominids? An assessment based on extant papionin craniodental morphology. Journal of Human Evolution 49:618–642.

CHAPTER 4

MATERIALS AND METHODS

4.1 Samples and Sample Organization

This dissertation partitions the analyses into two broad categories. The first, called the

“modern test,” is conducted only on those samples that are modern, which is broadly defined as post-Medieval. In this study, most modern samples arise from the 19th and 20th

centuries, and are therefore Industrial or post-Industrial. The goal of the modern test is to

see if various long bones properties detect genetic relationships among geographically

dispersed popualtions for which biomechanical diversity should be limited. The second

test, called the “temporal test,” includes several Medieval samples from England and

Italy drawn from the same regions as the modern samples. The goal of the temporal test

is to amplify the potential effects of plasticity on genetic predictions by introducing

samples comprised of individuals from more physically demanding lifestyles (i.e., a time

period before the widespread use of public transit and other technological innovations

that would reduce the burden of manual labor). In this way, results generated from the

modern test can be compared to those from the temporal test to assess whether activity-

induced plasticity lead to any erasure of among-population genetic structure.

This project includes data on both males and females from thirteen samples,

arising from modern and Medieval medical and cemetery collections that serviced the

public. Each sample is presented with geographic and temporal context in table 2 and

figure 6. The total sample pool is presented in table 3. Sample size varies by anatomical

region (e.g., crania, femur) due to differences in preservation. Therefore, the sample sizes

for each analysis are presented in the associated tables (Results section). Because of

𝑄𝑄𝑆𝑆𝑆𝑆 65

insufficient sample sizes for the individual Medieval samples, these were pooled into

three, regionally defined categories (table 4). A multivariate analysis of variance was conducted for samples in each temporal grouping to ensure that there were no significant among-population differences based on craniofacial properties. While it would have been ideal to avoid data-pooling, it was necessary given lack of archaeological samples for which individuals possessed all skeletal elements necessary for this project.

Figure 6. Locations of the samples

Table 2. Sample context Sample Name Geographical Region Time Century Reference Comet Place England (South) Medieval 15th Smith et al. 1984, Roberts et al. 2007 Black Gate England (North) Medieval 8-12th Nolan 2010, Swales 2012 All Saint’s at England (North) Medieval 11-14th Diane Swales Fishergate (“York”) (collections curator), personal communication South Shields England (North) Post- 19th Diane Swales (“Church at St. Medieval/ (collections curator), Hilda”) Modern personal communication Roselle Italy Early 6-12th Elsa Pacciani (Tuscany) Medieval (collections curator), personal communication, Boccone et al. 2011 Sestino Italy Late 16th Elsa Pacciani (Tuscany) Medieval (estimated) (collections curator), personal communication, Boccone et al. 2011 Piazza della Signora Italy Medieval 14th Elsa Pacciani (Tuscany) (collections curator), personal communication, Boccone et al. 2011 Bologna Italy Modern 19th-20th Gualdi-Russo 2007, Radi (Tuscany) et al. 2013, Mariotti et al. 2015 Sassari Italy Modern 19th-20th Mariotti et al. 2004, (Sardinia) Belcastro et al. 2008 Lisbon Portugal Modern 19th-20th Cardoso 2006, Cardoso and Henderson 2010 Johannesburg South Africa (Interior) Modern 20th Dayal et al. 2009 Pretoria South Africa (Interior) Modern 20th L’Abbé et al. 2005 Cape Town/ South Africa (West. Cape) Modern 20th Ginter 2005 Stellenbosch

Table 3. Sample pool* for different anatomical regions Anatomical n n n Sample** Region Males Females Unknown South English Crania 8 14 5 Medieval Long Bones 20 29 9 North English Crania 29 21 4 Medieval Long Bones 82 65 10 Crania 16 18 5 England Long Bones 24 25 10 Crania 32 36 6 Italian Medieval Long Bones 70 62 10 Crania 21 23 -- Italy Long Bones 22 25 -- Crania 25 23 -- Sardinia Long Bones 25 25 -- Crania 60 27 -- Portugal Long Bones 64 29 -- Crania 32 25 -- Johannesburg Long Bones 31 27 -- Crania 36 34 -- Pretoria Long Bones 32 31 -- Crania 32 24 -- Cape Town Long Bones 40 30 -- *Sample sizes vary by analysis due to differences in preservation across skeletal elements (i.e., not all individuals were intact). Therefore, sample sizes are provided in the results tables for each Q analysis, which will also pertain to associated R- and D-matrices.

Table 4. Temporal and sample groupings, abbreviations used in this project Temporal Identifier/ Samples Sample Grouping Grouping Abbreviation Comet Plate Medieval South English Medieval Eng_Med_S Black Gate, York Medieval North English Medieval Eng_Med_N South Shields Modern England SouSh Roselle, Sestino, Piazza della Signora Medieval Italian Medieval Ital_Med Bologna Modern Italy Italy Sassari Modern Sardinia Sass Lisbon Modern Portugal Lis Cape Town Modern Cape Town CT_E Johannesburg Modern Johannesburg Jo_E Pretoria Modern Pretoria Pr_E

There are two samples in this study for which occupational data were available: Lisbon and Bologna. The most commonly recorded occupation for females in Lisbon was doméstica (83%), which could refer to either housewife or house keeper (Cardoso, 2005).

These are followed by student (7%), and finally servant, professor, shop worker and postal worker (3% each). Profession for males was varied, and included clerk, police agent, student, and a number of light and heavy industrial workers. These occupations are consistent with those found in other studies (Cardoso, 2006). Historical data indicate that individuals in Lisbon worked from a young age—at least by adolescence—onward

(Cardoso and Garcia, 2009; Cardoso and Caninas, 2010; Cardoso and Henderson, 2010,

2013).

Data provided by museum curators show that men from the Bologna sample were exclusively involved with occupations that required some degree of physical/manual labor, particularly brick-laying, which was in high demand after the destruction of the city walls prompted widespread construction efforts on the newly accessible land. There were also a number of light industrial and other skilled professions represented in this sample (e.g., student or tailor) Like Lisbon, however, occupations for Bolognese women were generally recorded as housewife, and therefore not informative as to the true scope of work that was truly conducted by females (Cardoso and Henderson, 2013; Mariotti et al., 2015). In fact, historical research shows that women in several Bolognese districts were preferred hires for factory work, primarily in textile factories (Kertzer and Hogan,

1989). The few women in this sample who have a secondary occupation are listed as either seamstresses or bracchianti (hired hand, most likely in the agricultural sector but who historically would take on many odd jobs). According to Mariotti et al. (2015:3),

“less than 5% of both men and women [in this sample] were employed in occupations requiring a certain degree of education (e.g., teacher, piano player).”

Because both samples vary broadly in the types of occupation represented and because

they are sufficiently large, they offer a unique opportunity to further probe the possible

interference of function on genetic predictions generated by diaphyseal properties.

Therefore, an additional round of analyses was conducted with the modern samples in

which the Lisbon and Bolognese samples were partitioned into two subsamples based on

occupational intensity as defined using Cardoso and Henderson’s (2013) classification

system: [1] a “light intensity,” category which combined Cardoso and Henderson’s “non-

manual” (e.g., typographer, clerk) and “light manual” (e.g., artisanal workers, tailors) designations, and [2] a “heavy intensity” category which corresponded to their “heavy manual” designation (comprised largely of laborers and industrial workers).

Figure 7. Overview of occupations represented in the Lisbon and Bologna samples using categories presented in Cardoso and Henderson (2013)

4.2 Analytical Methods

4.2.1 s and R-matrices

RMET𝑸𝑸 𝑺𝑺𝑺𝑺5.0, an independent software program created and made available by Dr. John

Relethford, was used to calculate and distance and relationship matrices (D-matrices

𝑆𝑆𝑆𝑆 and R-matrices, respectively) (Relethford𝑄𝑄 and Blangero, 1990). A is a value that

𝑆𝑆𝑆𝑆 reflects the percentage of total variation attributable to differences between𝑄𝑄 (rather than

within) populations and can be treated as a general indication of population

differentiation (Relethford, 2009). Note this author follows the conventions of Roseman

(2004) and von Cramon-Taubadel (2009) in use of the phrase “Q ” to specify

ST phenotypically derived population differentiation (in contrast to F , which is generated

ST from molecular data). This differs from previous articles in which “F ” was used in both

ST contexts interchangeably (Relethford and Blangero, 1990; Relethford, 1996). s were

𝑆𝑆𝑆𝑆 originally generated using allelic data that compared the deviation of subpopulation𝐹𝐹

heterozygosity to the mean regional heterozygosity across all populations under study.

However, Relethford and Blangero (1990) demonstrated that they could also be derived

using continuous phenotypic data. s range from zero (meaning no among-population

𝑆𝑆𝑆𝑆 differentiation, or “genetically identical𝑄𝑄 ”) to one (meaning complete differentiation, or no

genetic similarity). Because they are a relative measure, s are comparable across

𝑆𝑆𝑆𝑆 studies drawn from different types of data, including molecular𝑄𝑄 data or other phenotypic

traits (e.g., dental, fingerprint, or skin pigmentation) (Roseman, 2004).

A heritability estimate is necessary for the calculation of Q . This is a value

ST reflecting the proportion of within-population trait variation that is due to genetic

(intrinsic) variation rather than extrinsic environmental effects (Visscher et al., 2008). A

value for heritability nearer to one indicates the variance of that trait is almost wholly

explained by underlying genetic variation, while a value nearer to zero suggests the

opposite. While there are published heritability estimates for craniofacial data, those for

appendicular bones are scant and are typically derived from medical investigations of

heritable bone disorders, and are therefore over-represented by properties such as bone

mineral density, femoral neck length, and other regions susceptible to the effects of

osteoporosis. Furthermore, existing long bone heritabilities vary widely (0.2-to-0.8)

(Smith et al., 1973; Christian et al., 1989; Arden et al., 1996; Slemenda et al., 1996; Deng

et al., 2002; Hui et al., 2006; Tse et al., 2009). Because of the uncertainty regarding the

heritability of long bone properties targeted in this study (e.g., joints, lengths, and

diaphyses), each analysis was conducted multiple times using a range of heritability

values (0.55, 0.75, and 1). A value around 0.55 (or lower) is typical of craniofacial

studies, though some have reported values as low as 0.33 (Cheverud, 1982, 1989;

Relethford, 2001). A heritability of “1” carries the assumption that trait variation is

perfectly proportional to underlying genetic variation (Relethford, 1994; Strauss et al.,

2015). Because heritability and Q values are inversely related, this generates a

ST “minimum Q ” or minimum similarity score (Relethford, 2007). In reality, this value

ST could be higher than the minimum Q (indicating greater among-population

ST differentiation), but never lower (Relethford and Blangero, 1990). It is therefore the most

conservative estimate (Relethford, 1994) and is common in studies of anatomical features

for which heritabilities are unknown, such as the long bones, craniofacial units, or the

pelvis (von Cramon‐Taubadel, 2011; Betti et al., 2012, 2014) .

While Q values are useful devices to assess overall differentiation among populations,

ST they give no indication as to how those relationships are patterned. For this, a relationship

(R) matrix is a useful device. In classic genetic studies, R-matrices were generated using

allele frequencies. Multiple R-matrices would be derived—one for each allele under

investigation—and subsequently averaged to generate a final matrix that summarized the overall pattern of genetic relationships. An R-matrix is an n x n matrix, where n is the total number of populations under study (Relethford, 2003). Each element within the

matrix contains a value that reflects the relationship between two populations weighted

against study-wide mean allele frequency and by population size (von Cramon-Taubadel,

2011). Because these values are weighted, all elements of an R-matrix sum to zero. For

any two populations, a negative element indicates less similarity than the average while

positive values indicate greater similarity (and therefore closer genetic affinity) than the

average (Relethford, 2004; von Cramon-Taubadel, 2009, 2011). Furthermore, the

diagonal element of the R-matrix (i.e., that which appears to represent a population’s

relationship to itself) reflects the distance of that population from the centroid generated

for all populations under study. The average of these diagonal elements reflects the

unbiased Q value described above.

ST Relethford and Blangero (1990) showed that R-matrices could be derived from phenotypic variation under the assumption that each associated allele contributes to that trait equally and incrementally (“equal and additive effects” model). Another benefit for the R-matrix is that it is little affected by deviations from the assumption that all populations are equally influenced by gene flow from an imaginary, panmictic outside population, one of the core assumptions of the Harpending-Ward model from which the

Relethford-Blangero model is derived (Relethford and Blangero, 1990, Relethford,

1996). For detailed explanations about how R-matrices are generated using allelic and

quantitative data, refer to Relethford and Blangero (1990). It should be noted that while heritability values will affect the Q value and the magnitude of predicted relationships

ST in the R-matrix, it has no bearing on the pattern (relative values) of those relationships

(Relethford, 2004 a; von Cramon-Taubadel, 2009; von Cramon‐Taubadel, 2011) .

Therefore, only the R-matrices generated using a heritability of one are included in this

dissertation.

Due to the costs and destructive nature associated with such analyses, it was not realistic to gather genetic data on samples as large as those utilized in this study.

Therefore, craniofacial variation was used as a proxy for genetic data. This is justified by the many studies that confirm craniofacial variation accurately reflects interpopulation genetic relationships, that it does so as accurately as does DNA or protein variation, and that this appears true even for traits that arise from regions of high mechanical loadings

(e.g., associated with mastication) (Relethford and Jorde, 1999; Relethford, 2004 a; von

Cramon-Taubadel, 2009). Fourteen craniometric variables (see table 8) were used to predict the genetic relationships among these populations. To visualize the predicted interpopulation relationships, multidimensional scaling (MDS) was conducted on the distance matrix following Harvati et al. (2006). The distance matrix was generated by taking the square root of elements in the D matrix produced by RMET 5.0. 2 To test whether the interpopulation relationships predicted by craniofacial data were correlated to those generated by various long bone properties, a Mantel test was conducted between craniometric distance matrices and those generated by each form of

long bone data (e.g., lengths, joints, diaphyses). A Mantel test provides a means to assess

the strength of correlation among the elements of two matrices, and remains a robust test

despite the problems often associated with small samples (Cheverud, 1989). Furthermore,

such tests are “free of distributional assumptions, [as] the only assumption made is that

all permutations are equally likely under the null hypothesis” (Cheverud et al., 1989).

Because the matrices arise from the same samples, it is expected that they will lack

independence. This makes it difficult to tell whether reported correlations are truly

significant or are rather an artifact of this dependence (Smouse et al., 1986). Mantel tests

provide a way to gauge significance by assessing whether the correlation observed is

greater than that expected under random scenarios. It does so by holding one matrix at a

constant while randomizing (permuting) elements of the other before running a

predetermined number of correlation tests between the two (Ackermann and Cheverud,

2000). In this way, the correlation of the two original matrices (e.g., the craniofacial and

length distances matrices) can be compared to the distribution of results generated by the

10,000 randomized matrices. A one-tailed t-test was used to assess whether the Mantel

test was significant as a positive association is expected (Smouse et al., 1986). Mantel

tests were conducted using the package “ade4” for R as well as a free standalone software program PASSaGE, version 2, which also provides the distribution of permutations and information about over- and under-prediction rates (Dray and Dufour, 2007; Rosenberg and Anderson, 2011). Ten-thousand permutations were used for each Mantel test in this study—well over the recommended minimum of 500 (Cheverud et al., 1989; Ackermann and Cheverud, 2000).

Table 5. Software used for this research Software Purpose URL ImageJ Image analysis http://imagej.nih.gov/ij/ BoneJ Extract cross-sectional http://bonej.org/ properties MomentMacrov1.4 Extract cross-sectional http://www.hopkinsmedicine.org/fae/mmacro.html properties, calibrate sections PASSaGE, v2 Mantel tests http://www.passagesoftware.net/ RMET 5.0 Calculate FST and R- http://employees.oneonta.edu/relethjh/programs/ matrices R Statistical software https://www.r-project.org/ RStudio Interface for R https://www.rstudio.com/

4.2.2 Coefficient of Variation

To test whether variation is less constrained in “presumably plastic” regions (i.e.,

diaphyseal regions), a coefficient of variation (CV) was calculated for each diaphyseal

trait. The CV, reflecting trait dispersion about the mean, gives an idea of size-adjusted variation. It is calculated as a percentage by dividing the standard deviation by the mean and multiplying this value by 100 (Plavcan, 2012). Because it is a “size-independent” value, the CV can and has been used to look at differences between the sexes, among human populations, and even between different primate and hominin taxa (Wood and

Lieberman, 2001). Sex-specific and pooled-sex CVs were generated by population to assess whether there are any interpopulation or sex-specific differences in the patterning or magnitude of CVs. CVs were log-transformed prior to analysis following Collard and

Lycett (2007).

For this study, the CV analysis consists of two primary sections. For the first, the mean CV for each broad anatomical region (crania, joints, diaphyseal diameters, lengths)

was calculated to test for any significant difference among populations. For the second

analysis, a mean CV was generated for more specific anatomical regions (e.g., by specific joint, by cross-sectional region) to assess whether any intralimb pattern in CV was

76 detectable. The aggregation of variables by region and the use of mean CV by region are common practice (Wood and Lieberman, 2001; Lycett and Collard, 2005; Collard and

Lycett, 2007; Collard and Wood, 2007). Finally, a MANOVA was conducted to test whether there was any significant effect of population, time period, or an interaction on the “broad anatomical” CVs. Then, a separate MANOVA was conducted on the trait CVs for the total sample to test for an effect of anatomical region (either broad or specific).

4.3 Data Collection

4.3.1 Long Bones

Six long bones were included in this project: humerus and radius (bilateral), femur, and tibia. These represent a proximal and distal component of each limb, which can be useful to discriminate between systemic and localized differences in diaphyseal variation.

Furthermore, these bones endure very different loading environments, as the leg bones routinely support body mass and experience loads related to mobility, while the arm bones experience unique strains and stresses depending on the types of activities performed. Because limb dominance can impact long bone properties, presumably due to increased muscle mass, strength and flexibility of the dominant arm, both the right and left humerus and radius were analyzed when sample sizes were sufficiently large

(Schulter-Ellis, 1980; Steele and Mays, 1995; Steele, 2000; Čuk et al., 2001; Sládek et al., 2016). The right femur and tibia were prioritized, though it was necessary to substitute the left when the right bones were missing or damaged, particularly for the archaeological samples.

Because long bone diaphyseal shape is sensitive to errors during the orientation process, it is important to orient the bones in a way that is consistent from one individual

77 to the next (Ruff and Hayes, 1982). Standards described by Ruff (2002) were used for this project and are briefly summarized in table 6, though interested readers are directed to the original manuscript which provides more detailed descriptions of the methods used. An architect’s triangle was useful for ensuring each key landmark was leveled appropriately and plasticine clay was used to stabilize and level the bones.

Table 6. Brief description of orientation protocols following Ruff (2002) Bone Description Humerus A light pencil mark was made on the medial diaphyseal surface at the anteroposterior midpoint at a level consistent with the proximal-most aspect of the olecranon fossa. Another pencil mark was made at the anteroposterior midpoint of the diaphysis “just distal to the head and lesser tubercle.” (337). The bone was then oriented such that both pencil marks fell in the same plane and the distal epicondyle parallel to the resting surface. In specimens with marked humeral curvature it was often necessary to elevate both the proximal and distal ends to accomplish proper orientation. Radius The anteroposterior dimension of the radial head and distal facet were measured and the midpoint of each lightly marked in pencil. The bone was oriented such that both were in the same plane and the distal articular surface parallel to the resting surface. Femur The femur was oriented along longitudinal axis by leveling to the midpoint between the anteroposterior diaphyseal thicknesses “just distal to the lesser trochanter and just proximal to the condyles” (337) Tibia The anteroposterior dimension of the proximal and distal tibial facets were measured and the midpoint of each lightly marked in pencil. The bone was oriented such that all three midpoints lie were the same plane.

Fourteen periosteal molds were taken for all long bones at consistent percentages of biological length (figure 8). Lower percentages reflect distal regions and higher percentages reflect more proximal regions. These casting locations avoid the morphological complexities of the joint regions (i.e., a “joint effect”) and therefore provide a stronger test of diaphyseal shape variation. These regions were chosen largely by convention (e.g., the midshaft measurements and 35% humerus), however a few additional regions were added to bolster the potential to detect different within-bone locational effects. For example, previous research shows that the proximal (80%) femur may more strongly reflect body breadth than it does mechanical environment of the leg

78 due to a close association with the pelvis (Ruff, 1995). While computed tomography (CT) or radiographic data would have been ideal, the cost associated with such equipment precluded their inclusion in this research. Nevertheless, Stock and Shaw (2007) found a highly significant relationship between external dimensions gathered via molds and internal bone structures, making molds a reliable substitute for costly digital methods. A professional-grade polyvinylsiloxane dental putty and fixer were used to construct the molds as it dries quickly, is not destructive, and is demonstrated to be effective in previous studies (Stock and Shaw, 2007).

Following the casting process, each cast was traced onto white paper and scanned in grey scale at a resolution of 600 dpi. The image was then imported into Photoshop where individual cross-sections were cropped, rotated so that the ML axis was the horizontal, flipped horizontally (if necessary) to ensure proper mediolateral orientation, and converted into a duotone image. The files were then saved and processed by Bone J, a robust plugin for the open-source ImageJ photo analysis software (Abràmoff and

Magalhães, 2004; Doube et al., 2010; Schneider et al., 2012). Both software programs are freely available at the URLs found in Table 5. Biomechanical information was extracted using the Slice Geometry application within BoneJ. Prior to analysis and multiple times throughout image processing, a calibration image was used to ensure that values generated by the macro were appropriate. The calibration image and directions are available via the Johns Hopkins University website (see Table 5). To ensure that values reported by BoneJ were consistent with those reported by Moment Macro v 1.4 (an older macro used to extract biomechanical information through ImageJ and upon which BoneJ is based), both macros were used to process an array cross-sections of varying shape and

79 size and subsequent output compared to verify consistency. All variables listed on Table

1 were extracted for each diaphyseal cast, as well as basic metric data (e.g., principal and anatomical diameters, perimeter). Only a subset of these were included in the genetic distance analyses to limit the impact of data redundancy: total area, Z , shape ratio

P (maximum/minimum), anteroposterior and mediolateral diameter.

Figure 8. Green boxes indicate regions for which bone casts were generated. Image modified from that of Fischer-Dückelmann (1911) (public domain)

4.3.2 Articular and Length Dimensions

In addition to diaphyseal data, 47 measurements were taken across all long bones

(including both the left and right humerus and radius). These are presented along with brief descriptions in table 7 and figure 9. Again, standards presented in Ruff (2002) were followed, with the exception of humeral and femoral head depth, which were gathered via coordinate calipers oriented in a superoinferior direction.

Table 7. Joint measurements (in mm) from Ruff (2002) Bone Abbreviation Description FHAP Anteroposterior breadth of head FHSI Superoinferior breadth of head FHDPCor* Head depth (calipers oriented along superoinferior axis) MCML Mediolateral breadth of medial condyle LCML Mediolateral breadth of lateral condyle MCSI Superoinferior length of medial condyle Femur LCSI Superoinferior length of lateral condyle

LPML Mediolateral breadth of the lateral condyle LPAP Anteroposterior breadth of the lateral condyle MPML Mediolateral breadth of the medial condyle MPAP Anteroposterior breadth of the medial condyle TTAP Anteroposterior breadth of the talar facet TTML Mediolateral breadth of the talar facet RHHAP Anteroposterior breadth of head RHHSI Superoinferior breadth of head RHHDPCor* Head depth (calipers oriented along superoinferior axis) RTRML Mediolateral breadth of trochlea RCPML Mediolateral breadth of capitulum RTSIMed Superoinferior height of the medial trochlea Humerus RTSINarrow Superoinferior height of the trochlea at its narrowest margin (Bilateral) RTSILat Superoinferior height of the lateral trochlea RTRAPMed Anteroposterior breadth of the medial trochlea RTRAPNarrow Anteroposterior breadth of the trochlea at its narrowest margin RTRAPLat Anteroposterior breadth of the lateral trochlea RCPSI Superoinferior length of capitulum RCPAP Anteroposterior length of capitulum RRHAP Anteroposterior breadth of head Radius RRHML Mediolateral breadth of head (Bilateral) RRCAP Anteroposterior breadth of distal facet RRCML Mediolateral breadth of distal facet *Note that these variables were gathered via coordinate calipers and therefore do not follow Ruff (2002)

Figure 9. Image showing the number (in parentheses) of measurements gathered for each arm (green) and leg (purple) joint. Image modified from that of Fischer-Dückelmann (1911) (public domain)

4.3.3 Crania

Fourteen dimensions from Howells (1973) were used in this project (table 8). These have low levels of intra-observer error and include measurements frequently used to reconstruct interpopulation genetic distances (Relethford, 1994; Betti et al., 2009, 2010).

All measurements were taken with sliding and spreading calipers, and include breadth,

length, height and shape of various regions of the vault, face, and basicranium.

Table 8. Cranial measurements and landmarks Measurement Landmarks Abbreviation Cranial Vault Height Basion-Bregma CranVauHt Cranial Base Length Basion-Nasion CranBasLen Max. Cranial Breadth Euryon-Euryon MaxCranBth Max. Cranial Length Glabella-Opisthocranion MaxCranLen Biauricular Breadth Auriculare-Auriculare BiaurBth Min. Frontal Breadth Frontotemp-Frontotemp MinFronBth Nasal Height Nasion-Nasospinale NasalHt Nasal Breadth Alare-Alare NasalBth Orbital Breadth (left and right) Dacryon-Ectoconchion ROrbBth Orbital Height (left and right) None ROrbHt Interorbital Breadth Dacryon-Dacryon InOrbBth Foramen Magnum Length Basion-Opisthion ForMagLen

4.4 Pre-Analysis Data Treatment

Only individuals exhibiting epiphyseal fusion were included. Individuals displaying

pathology or malnutrition were excluded from analysis (New et al., 1997; Ortner, 2003;

Alexy et al., 2005).

4.4.1 Standardization

All data were standardized using sex-specific z-scores following Relethford (1994, 2002).

Prior to this, however, all cross-sectional properties were standardized against body mass and biological length measured to some power per standard protocol (Holt, personal communication). This is because long bones, in acting as support structures, reflect both the size (stature) and shape (body mass distribution) of the individual. This means that size and shape effects must be controlled for when comparing behavior among populations who naturally vary in height, mass, and proportion (Trinkaus, 1981).

Specifically, cross-sectional areas reflect the force of axial loadings (compression and

tension), while SMAs reflect bending and torsional forces, and are therefore affected by

both load force (e.g., mass) and moment arm (bone length) (Ruff, 2000). Therefore, the

former are standardized by body mass, while the latter are standardized by body mass and

bone length (see table 9). Because Ruff (2000) found that both arm and leg bones scale to

mass despite the clear differences in functional environment, we use the same

standardizing criteria for all bones.

Because these data arose from sources for which body mass was unknown, these

values were estimated using predictive formulae. While estimating body mass from the

same bones subject to analysis may seem circular, Ruff (2000:283) argues that there is

value to estimated body mass standardization in that “it is virtually certain that without

such control any behavioral inferences [derived from cross-sectional properties] will be confounded.” Body mass was estimated based on the superoinferior dimension of the femoral head as research shows that the relationship between femoral head diameter and mass is consistent across a “wide range of body types,” populations, and proportions, and is consistently shown to have a high correlation to body mass in human and nonhuman primates (Auerbach and Ruff, 2004:339-340). Body mass for European individuals was calculated using equations presented in Ruff et al., (2012) which were derived, in part, by some of the same individuals comprising this sample (table 10). Body mass for South

African samples was calculated using Ruff et al. (1997).

Femoral head data were unavailable for roughly 8% of the total sample—typically when the femur was absent or grossly damaged. Therefore, new predictive formulate were generated using an Ordinary Least Squares regression between with femoral-head-

84 derived body mass and either tibial or humeral maximum length (see table 11 in Results).

Because of the higher correlation in pooled sex as opposed to sex-specific samples, and because sex of individuals in the archaeological samples was often unknown, we used only the pooled-sex predictive formulae in this study. Regressions were carried out using the R package lmodel2 (Legendre, 2014).

Table 9. Standardization factors for cross-sectional properties Biomechanical property Standardization factor Areas (Area/Body Mass)*10 Second Moment of Area (SMA) (SMA/(Body Mass*Biological Length2 ))*10 Section moduli (Z ) (Z /(Body Mass*Biological Length))*2 10 5 4 P P

Table 10. Body mass formulae* Male Body Mass = 2.741 × FH 54.9 General Female Body Mass = 2.426 × FH 35.1 Ruff et al. 1997 kg mm − Pooled Body Mass = 2.268 × FH 36.5 kg mm − Male Body Masskg = 2.80 × FH mm 66.70 − European Female Body Mass = 2.18 × FH 35.81 Ruff et al. 2012 kg mm − Pooled Body Mass = 2.30 × FH 41.72 kg mm − kg mm −

4.4.2 Data Imputation

Missing data were a problem, in particular for the Medieval English and Italian samples.

For this reason, it was necessary to modify the project design and to use posthoc statistical imputation to create a sufficiently robust dataset. To increase archaeological sample sizes, the sexes were pooled and the analyses were performed using a single humerus and radius for the combined long bone tests, preferentially selecting the right and substituting the left when the right was unavailable. For all other tests (e.g., right arm, left arm, proximal elements, etc.) there were sufficiently large samples to allow for separate analysis of left and right arm elements. All missingness and imputation analyses

were conducted using R version 3.2.3. Data were first tested for normality and

missingness using the package “MissMech” (Jamshidian et al., 2014) and by visual inspection of histograms, residuals, and qq-plots generated with the base system and with the package “psych” (Revelle, 2015). Data were further scrutinized for anomalies, errors, and outliers using the overimputation function in the R package “Amelia II” on a trait-by- trait basis (Honaker et al., 2011). Overimputation tests are tests in which known observations are removed at random, imputed, and plotted against the “real” data with

90% confidence intervals. This is useful to identify those variables for which imputation is weaker so that they can be excluded from analysis, but also for identifying outliers.

Outliers were only removed in the event they were blatantly anomalous (e.g., had exceptionally high leverage on residual plots and/or fell well outside the data cluster on overimputation plots as seen in figure 10).

Figure 10. An overimputation plot illustrating two anomalies (left) in the biorbital breadth measurement. These were both discovered to be entry errors and were corrected prior to analysis.

Because RMET 5.0 is incapable of handing missing data, it was necessary to impute

missing values prior to conducting analysis. While it is common to delete incomplete

individuals (listwise deletion) or to use single imputation methods (e.g., means substitutions), research shows that both can produce biased results because there may be some underlying reason for the pattern of “missingness” (Wayman, 2003). Furthermore, means can generate artificially low error estimates and can also reduce variance (Adams et al., 2004; Nakagawa and Freckleton, 2011). Therefore, missing data were imputed using multiple imputation (MI) which accounts for both the relationships among the known and unknown variables (distinguishing it from regression methods, which focus only on the known) as well as the overall uncertainty with the imputation itself:

“Maintaining the original variability of the missing data is done by creating imputed values which are based on variables correlated with the missing data and causes of missingness. Uncertainty is accounted for by creating different versions of the missing data and observing the variability between imputed data sets.” (Wayman, 2003:4)

While initially developed for use in the social and medical sciences (in particular survey

data), MI is also used by ecologists and biologists. MI was conducted in R using Amelia

II. Five imputations were conducted with the data which were subsequently compared for

consistency before averaging (Berglund and Heeringa, 2014). Data were imputed by anatomical region rather than total dataset to avoid situations in which an entire bone would be imputed. Only individuals who were 60% complete were included for each type of analysis (e.g., craniofacial, diaphyseal analysis). A threshold of 60% was selected in order to maximize the archaeological sample sizes and is only slightly less than the 70%

87 threshold used in other studies that calculate genetic distance from phenotypic

(craniofacial) data (Pinhasi and von Cramon-Taubadel, 2009).

4.5 Chapter 4 References

Abràmoff MD, and Magalhães PJ. 2004. Image processing with ImageJ. Biophotonics International 11:36–42.

Ackermann RR, and Cheverud JM. 2000. Phenotypic covariance structure in tamarins(genus Saguinus): a comparison of variation patterns using matrix correlation and common principal component …. American Journal of Physical Anthropology 111:489–501.

Adams DC, Rohlf JF, and Slice DE. 2004. Geometric morphometrics: Ten years of progress following the “revolution.” Italian Journal of Zoology 71:5–16.

Alexy U, Remer T, Manz F, Neu CM, and Schoenau E. 2005. Long-term protein intake and dietary potential renal acid load are associated with bone modeling and remodeling at the proximal radius in healthy children. American Journal of Clinical Nutrition 82:1107–1114.

Arden N, Baker J, Hogg C, Baan K, and Spector TD. 1996. The heritability of bone mineral density, ultrasound of the calcaneus and hip axis length: a study of postmenopausal twins. Journal of Bone and Mineral Research 11:530–534.

Auerbach BM, and Ruff CB. 2004. Human body mass estimation: A comparison of “morphometric” and “mechanical” methods. American Journal of Physical Anthropology 125:331–342.

Belcastro MG, Rastelli E, Mariotti V. 2008. Variation of the degree of sacral vertebral body fusion in adulthood in two European modern skeletal collections. American Journal of Physical Anthropology 135:149-160.

Berglund P, and Heeringa SG. 2014. Multiple imputation of missing data using SAS. Cary, North Carolina: SAS Institute Inc.

Betti L, Balloux F, Hanihara T, and Manica A. 2010. The relative role of drift and selection in shaping the human skull. American Journal of Physical Anthropology 141:76–82.

Betti L, Cramon-Taubadel N, and Lycett SJ. 2012. Human pelvis and long bones reveal differential preservation of ancient population history and migration out of Africa. Human Biology 85:139–152.

Betti L, von Cramon-Taubadel N, Manica A, and Lycett SJ. 2014. The interaction of neutral evolutionary processes with climatically-driven adaptive changes in the 3D shape of the human os coxae. Journal of Human Evolution 73:64–74.

Boccone S, Chilleri F, Pacciani E, Moggi Cecchi J, Salvini M. 2011. The skeleton of a Medieval male with multiple traumatic fractures from Piazza della Signoria, Florence, Italy. International Journal of Osteoarchaeology 21:602-612.

Cardoso AF, and Henderson C. 2013. The Categorisation of Occupation in Identified Skeletal Collections: A Source of Bias? International Journal of Osteoarchaeology 23:186–196.

Cardoso AF, and Henderson CY. 2010. Enthesopathy formation in the humerus: Data from known age at death and known occupation skeletal collections. American Journal of Physical Anthropology 141:550–560.

Cardoso H, and Caninas M. 2010. Secular trends in social class differences of height, weight and BMI of boys from two schools in Lisbon, Portugal (1910–2000). Economics & Human Biology 8:111–120.

Cardoso H, and Garcia S. 2009. The Not‐so‐ Dark A ges: Ecology for human growth in medieval and early Twentieth Century Portugal as inferred from skeletal growth profiles. American Journal of Physical Anthropology 138:136–147.

Cardoso HF. 2005. Patterns of growth and development of the human skeleton and dentition in relation to environmental quality: A biocultural analysis of a sample of 20th century portuguese subadult documented skeleton. Ph.D. Dissertation:McMaster University, Ontario.

Cardoso HF. 2006. Brief communication: The collection of identified human skeletons housed at the Bocage Museum (National Museum of Natural History), Lisbon, Portugal. American Journal of Physical Anthropology 129:173–176.

Cheverud JM, Wagner GP, and Dow MM. 1989. Methods for the comparative analysis of variation patterns. Systematic Zoology 38:201–213.

Cheverud JM. 1982. Phenotypic, Genetic, and Environmental Morphological Integration in the Cranium. Evolution 36:499–516.

Cheverud JM. 1989. A comparative analysis of morphological variation patterns in the papionins. Evolution 43:1737–1747.

Christian JC, Yu PL, and Slemenda CW. 1989. Heritability of bone mass: a longitudinal study in aging male twins. American Journal of Human Genetics 44:429–433.

Collard M, and Lycett SJ. 2007. Does phenotypic plasticity confound attempts to identify hominin fossil species? Folia Primatologica 79:111–122.

Čuk T, Leben-Seljak P, and Stefancic M. 2001. Lateral asymmetry of human long bones. Variability and Evolution 9:19–32.

Dayal, MR, Kegley A, Strkalj G, Bidmos MA, Kuykendall KL. 2009. The history and composition of the Raymond A. Dart collection of human skeletons at the University of Witwatersrand, Johannesburg, South Africa. American Journal of Physical Anthropology 140(2):324-335.

Deng HW, Mahaney MC, Williams JT, and Li J. 2002. Relevance of the genes for bone mass variation to susceptibility to osteoporotic fractures and its implications to gene search for complex human diseases. Genetic Epidemiology 22:12–25.

Doube M, Kłosowski MM, Arganda-Carreras I, Cordelieres FP, Dougherty RP, Jackson JS, Schmid B, Hutchinson JR, and Shefelbine SJ. 2010. BoneJ: free and extensible bone image analysis in ImageJ. Bone 47:1076–1079.

Dray S, Dufour A. 2007. The ade4 package: Implementing the duality diagram for Ecologists. Journal of Statistical Software 22(4):1-20.

Ginter J. 2005. A test of the effectiveness of the revised maxillary suture obliteration method estimating adult age at death. Journal of Forensic Science 50(6):1303- 1309.

Graham JW. 2009. Missing data analysis: Making it work in the real world. Annual Review of Psychology 60:549–576.

Gualdi-Russo E. 2007. Sex determination from the talus and calcaneus measuremetns. Forensic Science International 171:151-156.

Harvati K, Weaver TD, and Hublin J. 2006. Reliability of cranial morphology in reconstructing Neanderthal phylogeny. In: K, Harrison T, editors. Netherlands: Springer. p 239–254.

Holt, and Benfer. 2000. Estimating missing data: an iterative regression approach. Journal of human evolution 39:289–96.

Honaker J, King G, and Blackwell M. 2011. Amelia II: A program for missing data. Journal of Statistical Software 45:1–47.

Hui SL, Koller DL, Foroud TM, Econs MJ, Johnston CC, and Peacock M. 2006. Heritability of changes in bone size and bone mass with age in premenopausal white sisters. Journal of Bone and Mineral Research 21:1121–1125.

Jamshidian M, Jalal SJ, and Jansen C. 2014. Missmech: an R package for testing homoscedasticity, multivariate normality, and missing completely at random (mcar). Journal of Statistical Software 56:1–31.

Kertzer DI, and Hogan DP. 1989. Family, political economy, and demographic change: The transformation of life in Casalecchio, Italy, 1861–1921 . Madison, Wisconsin: University of Wisconsin Press.

L'Abbé, EN, Loots M, Meiring JH. 2005. The Pretoria Bone Collection: a modern South African skeletal sample. HOMO--Journal of Comparative Human Biology 56:197-205.

Legendre P. 2014. Model II regression user’s guide, R package version 1.7-2. R Vignette.

Mariotti V, Zuppello M, Pedrosi M, Bettuzzi M, Brancaccio R, Peccenini E, Morigi M, and Belcastro M. 2015. Skeletal evidence of tuberculosis in a modern identified human skeletal collection (Certosa cemetery, Bologna, Italy). American Journal of Physical Anthropology 157:389–401.

Nakagawa S, and Freckleton RP. 2011. Model averaging, missing data and multiple imputation: a case study for behavioural ecology. Behavioral Ecology and Sociobiology 65:103–116.

New SA, Bolton-Smith C, Grubb DA, and Reid DM. 1997. Nutritional influences on bone mineral density: a cross-sectional study in premenopausal women. American Journal of Clinical Nutrition 65:1831–1839.

Nolan J. 2010. The early Medieval cemetery at the Castle, Newcastle-upon-Tyne. Archaeologia Aeliana 5(39):147-287.

Ortner DJ. 2003. Identification of pathological conditions in human skeletal remains. Second. San Diego: Elsevier Science.

Pinhasi R, and von Cramon-Taubadel N. 2009. Craniometric data supports demic diffusion model for the spread of agriculture into Europe. PloS one 4:e6747.

Plavcan JM. 2012. Body size, size variation, and sexual size dimorphism in early Homo. Current Anthropology 53(S6):S409-S423.

Radi N, Mariotti V, Riga, A, Zampetti S, Villa C, Belcastro MG. 2013. Variation of the anterior aspect of the femoral head-neck junction in a modern human identified skeletal collection. American Journal of Physical Anthropology 152:261-272.

Relethford JH. 1994. Craniometric variation among modern human populations. American Journal of Physical Anthropology 95:53–62.

Relethford JH. 1996. Genetic drift can obscure population history: problem and solution. Human Biology 68:29–44.

Relethford JH. 2001. Global Analysis of Regional Differences in Craniometric Diversity and Population Substructure. Human Biology 73:629–636.

Relethford JH. 2002. Apportionment of global human genetic diversity based on craniometrics and skin color. American journal of physical anthropology 118:393–8.

Relethford JH. 2003. Anthropometric data and population history. In: Herring DA, Swedlund AC, editors. Cambridge: Cambridge University Press. p 32–52.

Relethford JH. 2004a. Global Patterns of Isolation by Distance Based on Genetic and Morphological Data. Human Biology 76:499–513.

Relethford JH. 2004b. Boas and beyond: Migration and craniometric variation. American Journal of Human Biology 16:379–386.

Relethford JH. 2007. The use of quantitative traits in anthropological genetic studies of population structure and history. In: Crawford MH, editor. Cambridge: Cambridge University Press. p 187–209.

Relethford JH. 2009. Race and global patterns of phenotypic variation. American Journal of Physical Anthropology 139:16–22.

Revelle, W. 2015. psych: Procedures for Personality and Psychological Research, Northwestern University, Evanston, Illinois, USA, http://CRAN.R- project.org/package=psych Version = 1.5.8.

Relethford JH, and Blangero J. 1990. Detection of differential gene flow from patterns of quantitative variation. Human Biology 62:5–25.

Relethford JH, and Jorde LB. 1999. Genetic evidence for larger African population size during recent human evolution. American Journal of Physical Anthropology 108:251–60.

Roberts AM, Peters TJ, Robson Brown K. 2007. New light on old shoulders: palaeopathological patterns of arthropathy and enthesopathy in the shoulder complex. Journal of Anatomy 211:485-492.

Roseman CC. 2004. Detecting interregionally diversifying natural selection on modern human cranial form by using matched molecular and morphometric data. Proceedings of the National Academy of Sciences of the United States of America 101:12824–12829.

Rosenberg MS, and Anderson CD. 2011. PASSaGE: pattern analysis, spatial statistics and geographic exegesis. Version 2. Methods in Ecology and Evolution 2:229– 232.

Ruff CB. 1995. Biomechanics of the hip and birth in early Homo. American Journal of Physical Anthropology 98:527–574.

Ruff CB. 2000. Body size, body shape, and long bone strength in modern humans. Journal of Human Evolution 38:269–290.

Ruff CB. 2002. Long bone articular and diaphyseal structure in old world monkeys and apes. I: locomotor effects. American journal of physical anthropology 119:305– 42.

Ruff CB, and Hayes WC. 1982. Subperiosteal expansion and cortical remodeling of the human femur and tibia with aging. Science 217:945–948.

Ruff CB, Trinkaus E, and Holliday TW. 1997. Body mass and encephalization in Pleistocene Homo. Nature 387:173–176.

Schneider CA, Rasband WS, and Eliceiri KW. 2012. NIH Image to ImageJ: 25 years of image analysis. Nature Methods 9:671–675.

Schulter-Ellis FP. 1980. Evidence of handedness on documented skeletons. Journal of Forensic Science [Internet]. Available from: http://www.astm.org/DIGITAL_LIBRARY/JOURNALS/FORENSIC/PAGES/JF S11265J.htm

Slemenda CW, Turner CH, and Peacock M. 1996. The genetics of proximal femur geometry, distribution of bone mass and bone mineral density. Osteoporosis International 6:178–182.

Sládek V, Ruff CB, Berner M, Holt B, and Niskanen M. 2016. The impact of subsistence changes on humeral bilateral asymmetry in Terminal Pleistocene and Holocene Europe. Journal of Human Evolution 92:37–49.

Smith BH. 1984. Patterns fo molar wear in hunter-gatherers and agriculturalists. American Journal of Physical Anthropology 63:39-56.

Smith DM, Nance WE, and Kang KW. 1973. Genetic factors in determining bone mass. Journal of Clinical Investigation 52:2800–2808.

Smouse PE, Long JC, and Sokal RR. 1986. Multiple regression and correlation extensions of the Mantel test of matrix correspondence. Systematic Zoology 35:627–632.

Steele J. 2000. Handedness in past human populations: skeletal markers. Laterality: Asymmetries of Body, Brain and Cognition 5:193–220.

Steele J, and Mays S. 1995. Handedness and directional asymmetry in the long bones of the human upper limb. International Journal of Osteoarchaeology 5:39–49.

Stock JT, and Shaw CN. 2007. Which measures of diaphyseal robusticity are robust? A comparison of external methods of quantifying the strength of long bone diaphyses to cross‐secti onal geometric properties. A merican Journal of Physical Anthropology 134:412–423.

Strauss A, Hubbe M, Neves WA, Bernardo DV, and Atuí JV. 2015. The cranial morphology of the Botocudo Indians, Brazil. American Journal of Physical Anthropology 157:202–216.

Swales D. 2012. Life and stress: A bio-cultural investigation into the later Anglo-Saxon population of the Black Gate Cemetery, Newcastle-upon-Tyne. Ph.D. Dissertation:University of Sheffield

Trinkaus E. 1981. Neanderthal limb proportions and cold adaptation. In: Stringer CB, editor. London: Taylor and Francis. p 187–224.

Tse KY, Macias BR, Meyer RS, and Hargens AR. 2009. Heritability of bone density: regional and gender differences in monozygotic twins. Journal of Orthopedic Research 27:150–154.

Visscher PM, Hill WG, and Wray NR. 2008. Heritability in the genomics era — concepts and misconceptions. Nature Reviews Genetics 9:255–266. von Cramon-Taubadel N. 2009. Revisiting the homoiology hypothesis: the impact of phenotypic plasticity on the reconstruction of human population history from craniometric data. Journal of Human Evolution 57:179–190. von Cramon‐Taubadel N. 2011. The relative efficacy of functional and developmental cranial modules for reconstructing global human population history. American Journal of Physical Anthropology 146:83–93.

Wayman JC. 2003. Multiple imputation for missing data: What is it and how can I use it? In: Chicago, IL.

Wood B, and Lieberman DE. 2001. Craniodental variation in Paranthropus boisei: A developmental and functional perspective. American Journal of Physical Anthropology 116:13–25.

CHAPTER 5

RESULTS

5.1 Body Mass Regressions

Because it was necessary to estimate body mass for a small number of individuals lacking femoral head data, an ordinary least squares (OLS) regression was used to assess the correlation between femoral head-derived body mass and length of either the tibia or humerus. Regressions were run for each sex and a pooled sex sample. All analyses show a moderate-to-high correlation between bone length and mass, with the highest correlations reported for the pooled sex samples. Intercepts and slopes derived from the regression were used to estimate body mass for the roughly 8% of the sample lacking the

femoral head data. Results of the OLS are found in table 11, and plots of each bone on

femoral head-derived body mass in figures 11-16.

Table 11. Results of OLS Regression for body mass, tibial and humeral length Bone Sex n r R^2 Intercept Slope Tibia Max Pooled 604 0.6261274 0.3920356 -12.76381 0.2076528 Male 330 0.5112874 0.2614148 -0.2021667 0.1788331 Female 263 0.4635539 0.2148822 15.29244 0.1194199 Humerus Pooled 626 0.6927644 0.4799225 -29.51948 0.2943726 Male 336 0.5729956 0.3283239 -23.86245 0.2808636 Female 283 0.5670254 0.3215178 1.951030 0.1831761

Figure 11. Plot of body mass on tibial length, pooled sex

Figure 12. Plot of body mass on humeral length, pooled sex

Figure 13. Plot of body mass on tibial length, male

Figure 14. Plot of body mass on humeral length, male

Figure 15. Plot of body mass on tibial length, female

Figure 16. Plot of body mass on humeral length, female

5.2 Justification of Sample Pooling

A multivariate analysis of variance based on all standardized craniofacial variables conducted for the two Northern Medieval English samples (Blackgate and York) was not significant (F=1.09, p-value=0.415) nor was it significant for the three Italian Medieval samples (Piazza della Signora, Roselle, and Sestino) (F=1.09, p-value=0.222). Therefore, each subset was pooled into “North England Medieval,” “South England Medieval,” and

“Italian Medieval” groupings to boost sample sizes for the temporal analyses.

5.3 Nature and Pattern of Variation

The broad anatomical groupings show a consistent pattern across all samples and in both sexes (table 12, figure 17). Namely, of all four regions considered, the craniofacial traits and long bone lengths are the least variable, while the joint and cross-sectional dimensions are the most variable. For most samples, length is the least variable trait, while cross-sectional properties are the most variable, though there is little difference in the CVs generated for joint and cross-sectional dimensions nor between those generated for the crania and length traits.

Table 12. Mean coefficients of variation for aggregate anatomical groupings (all populations) Aggregate Sex Anatomical Region Female Male Pooled Crania 5.84 5.98 6.27 Joints 8.63 9.20 10.83 CSP Metrics 9.16 8.92 11.36 Length 5.95 5.99 7.18

100

Coefficient of Variation (Males) 12.00

10.00

8.00 Crania 6.00 Joints C.V. 4.00 Length Cross-Section 2.00

0.00 Med. Italy Italy Sardinia Portugal Med. England South England Africa

Coefficient of Variation (Females) 12.00

10.00

8.00 Crania 6.00 Joints C.V. 4.00 Length Cross-Section 2.00

0.00 Med. Italy Italy Sardinia Portugal Med. England South England Africa

Coefficient of Variation (Pooled Sex) 12.00

10.00

8.00 Crania 6.00 Joints C.V. 4.00 Length Cross-Section 2.00

0.00 Med. Italy Italy Sardinia Portugal Med. England South England Africa

Figure 17. Bar plots of mean coefficient of variation (CV) for each broad anatomical region

101

When long bone CVs are analyzed by anatomical region, a clear pattern emerges in

which proximal joints are less variable than the two distal joints across all three limbs and

in both sexes (table 13). The same is also true of the long bone length—the proximal

bones are less variable than the distal bones. There is, however, no consistent interlimb pattern for the metric cross-sectional data. For the arms, the midshaft dimensions tend to be more variable than are the distal dimensions, though there are exceptions (e.g., the

female left radius for which CV is comparable at both the 35% and 50% locations). For

the femur, the midshaft is less variable than either the distal or proximal locations

(particularly for women), while the CV values for the three tibial locations are generally

comparable to one another. Overall, the range of CV values for arm and leg CSP are

generally consistent, though leg CVs are slightly less than that for the arms, primarily due to the lower values reported by the tibia (table 14). See figure 18 for an image summarizing the overall pattern of CV across the skeleton based on metric data.

Table 15 reflects aggregrate CV each type of cross-sectional property averaged across all long bones. These results show that SMA circularity ratios for the anatomical axes (anteroposterior and mediolateral) generally have the lowest levels of variation, followed by total area (TA), and the circularity ratios for the principal (maximum and minimum) SMAs. These are followed by strength measures (Z and Z ). The SMA

P properties (I and J) were found to be the most variable overall. When considered with the

metric data presented above, these results suggest that metric and area data, followed by

rigidity indices are less variable than those properties which reflect strength and rigidity.

102

Table 13. Mean coefficients of variation for skull, joint, cross-sectional metrics, and length metrics (all populations) Sex Region Side Joint/Bone Female Male Pooled Skull -- -- 5.84 5.98 6.27 Right Shoulder 7.10 7.89 9.61 Right Elbow 8.88 9.34 11.07 Right Wrist 8.77 9.80 10.71 Long Left Shoulder 6.90 7.73 9.43 Bone Left Elbow 8.49 9.37 11.05 Joints Left Wrist 9.01 10.32 10.95 -- Hip 6.81 7.54 9.44 -- Knee 7.30 7.99 9.33 -- Ankle 7.41 7.72 9.22 Right Humerus 35% 8.86 7.82 11.05 Right Humerus 50% 9.75 8.73 11.61 Left Humerus 35% 8.32 8.11 10.99 Left Humerus 50% 9.35 9.48 11.56 Right Radius 35% 8.98 8.66 11.98 Right Radius 50% 9.43 8.84 11.86 Long Left Radius 35% 9.90 9.78 11.34 Bone Left Radius 50% 9.84 9.36 12.31 CSP -- Femur 20% 9.17 9.11 10.56 -- Femur 50% 7.97 8.25 9.78 -- Femur 80% 10.75 8.53 10.94 -- Tibia 35% 8.41 8.71 10.54 -- Tibia 50% 8.92 8.88 11.24 -- Tibia 65% 8.56 8.68 11.10 Right Humerus 5.85 5.46 6.75 Right Radius 6.26 5.88 7.55 Long Left Humerus 5.63 5.75 6.91 Bone Left Radius 6.02 6.03 7.43 Lengths -- Femur 5.72 5.93 6.84 -- Tibia 6.21 6.90 7.59

Table 14. Mean coefficients of variation for cross-sectional metrics by bone and by limb Sex Region Side Joint/Bone Female Male Pooled Right Humerus 9.31 8.28 11.33 Right Radius 9.21 8.75 11.92 Left Humerus 8.84 8.80 11.28 Left Radius 9.87 9.57 11.83 CSP -- Femur 9.30 8.63 10.43 Metrics -- Tibia 8.63 8.76 10.96 Right Arm 9.26 8.51 11.63 Left Arm 9.35 9.18 11.55 -- Leg 8.96 8.69 10.69

103

Table 15. Mean coefficients of variation by cross-sectional property Sex CSP Female Male Pooled TA 14.97 14.62 20.17 I 31.97 30.58 41.59 J 33.55 29.09 40.86 Z 24.77 24.15 31.84 24.63 24.15 31.79 / 11.23 10.44 10.90 𝐏𝐏 𝐙𝐙/ 19.37 19.54 19.71 𝐗𝐗 𝐘𝐘 𝐈𝐈 𝐈𝐈 𝐈𝐈𝐌𝐌𝐌𝐌𝐌𝐌 𝐈𝐈𝐌𝐌𝐌𝐌𝐌𝐌

Figure 18. Figure showing the pattern of CV across skeletal metric traits

104

The Analyses of Variance showed no significant effect of time period, geographical region, or an interaction on any broad CV grouping (see table 16). However, the effect of anatomical region is highly significant regardless of which long bone properties were included for analysis (see table 17 for descriptions of the different tests, and table 18 for the results). When considered with the mean CV presented above, these results suggest that the pattern of variability is relatively consistent across populations regardless of temporal or geographical origin. It is also consistent for both sexes. This suggests that there are perhaps mechanisms (internal or external) that limit variation in the crania and in long bone lengths, while these constraints are somewhat laxer in joint and diaphyseal properties.

Table 16. Analysis of Variance on mean CV by time and region Aggregrate Geographical Region Time Period Interaction Sex Grouping F p-value F p-value F p-value Crania 0.080 0.924 0.384 0.558 0.000 0.988 Joint 0.033 0.968 3.981 0.093 0.058 0.817 Female Length 0.501 0.629 1.446 0.274 1.665 0.244 CSP Metrics 0.855 0.471 2.196 0.189 0.225 0.652 Crania 1.577 0.282 0.198 0.672 0.048 0.833 Joint 1.968 0.220 0.182 0.685 0.925 0.373 Male Length 0.725 0.5222 4.397 0.0808 4.285 0.0839 CSP Metrics 0.764 0.506 0.529 0.494 0.070 0.800 Crania 0.695 0.530 0.047 0.834 0.308 0.596 Joint 0.607 0.5714 5.366 0.0537 0.293 0.6053 Pooled Length 0.470 0.643 0.922 0.369 2.006 0.200 CSP Metrics 3.572 0.0853 5.356 0.0539 1.017 0.3469

Table 17. Descriptions for anatomical CV ANOVA tests Test Variables Included Test 1 Broad anatomical groupings (craniometric traits, long bone lengths, joint metrics, and cross-section metrics) Test 2 Craniometric traits, long bone lengths, joint metrics, and cross-sectional shape ratios Test 3 Cross-sectional areas, SMAs, section moduli, shape ratios

105

Table 18. Analysis of Variance on mean CV using different cross-sectional properties Test 1* Test 2 Test 3 Sex DF F p-value DF F p-value DF F p-value Female 3 35.3 <0.001 3 21.37 <0.001 4 463 <0.001 Male 3 37.74 <0.001 3 25.21 <0.001 4 972.8 <0.001 Pooled 3 95.86 <0.001 3 39.5 <0.001 4 658.2 <0.001 See Table 17 for test description.

5.4 Population Differentiation

Results of the Q analysis for both the modern and temporal tests are shown in tables

ST 19-24. Note that lengths are included only in the “broad” anatomical tests, as subdivision into different anatomical regions (e.g., arms, legs) would result in too few variables for analysis. When looking at the modern Q results for the different long bone regions

ST (lengths, joints, or diaphyseal properties), we see that length and joint values are on par with those reported by the craniofacial properties, suggesting similar levels of among- population differentiation (table 19). With the inclusion of Medieval samples (table 20),

Q values increase moderately for the joint dimensions, suggesting higher levels of

ST interpopulation differentiation. However, there are slight decreases seen in the craniofacial and length dimensions which are again comparable to one another, and relatively little change to CSP values.

Table 19. Unbiased QST values and standard errors (se) by broad anatomical groupings, modern populations Test Variables n =0.55 se =0.75 se =1 se Crania -- 398 0.12𝟐𝟐 0.006 0.089𝟐𝟐 0.005 0.07𝟐𝟐 0.005 𝐡𝐡 𝐡𝐡 𝐡𝐡 Joints 310 0.114 0.004 0.083 0.004 0.061 0.004 All long CSP 282 0.056 0.004 0.038 0.003 0.025 0.003 bones Lengths 333 0.139 0.012 0.103 0.011 0.077 0.01 Abbreviations: Cross-sectional properties (CSP), heritability (h ), standard error (se). CSPs include Z , TA, feret shape, and AP and ML diameters, all2 standardized (see Materials and Methods). Refer to table 1 for descriptions of the various CSPs. P

106

Table 20. Unbiased QST values and standard errors (se) by broad anatomical groupings, modern and Medieval populations Test Variables n =0.55 se =0.75 se =1 se Crania -- 509 0.112𝟐𝟐 0.005 0.08𝟐𝟐 2 0.005 0.06𝟐𝟐 0.005 𝐡𝐡 𝐡𝐡 𝐡𝐡 Joints 445 0.143 0.004 0.107 0.004 0.08 0.004 All long CSP 338 0.059 0.003 0.039 0.003 0.025 0.003 bones Lengths 449 0.123 0.01 0.091 0.01 0.067 0.009 Refer to table 19 caption for legend of abbreviations found in this table

Q values derived from cross-sectional properties (CSPs) are smaller across-the-board,

ST suggesting less interpopulation differentiation than the other anatomical regions being analyzed. In both the modern and temporal tests, diaphyseal Q s are roughly half the

ST value of those reported for craniofacial properties. This trend is consistent at all levels

analysis—the broad anatomical groupings seen in tables 19-20, as well as the more

specific analyses that isolate arm from leg bones and proximal from distal bones (tables

21-22). At no point do CSPs show higher levels of differentiation than do length, joints,

or crania. While the inclusion of the Medieval samples did increase Q values for the

ST joint dimensions, particularly in the arm bones, there was little, if any, difference in

diaphyseal Q across the modern and temoporal tests. These uniformly low values for

ST the diaphyseal regions suggest that either within-population variation is especially high for cross-sectional properties or that variation in these properties is similar for all populations under study, resulting in a significant degree of among-population overlap and lower levels of differentiation, even with inclusion of the Medieval samples.

Results for the joint data across all levels of analysis, however, are quite different.

For the broad anatomical grouping in modern samples, Q values for the long bone joint

ST dimensions are within 2% of that reported for craniofacial variables (table 19) and these

values generally increase for the more targeted analyses, with the exception of the

107

proximal elements and the right arm in the modern test (table 21). Distal joints for

modern samples report higher Q values than do proximal ones, but there is little

ST consistency in Q strength by limb, with the leg reporting values intermediate to the left

ST (higher Q ) and right (lower Q ) arms (table 21). Once the Medieval samples are

ST ST added, however, joint Q s well exceed those of the crania, with the effect being

ST particularly strong for the arms (tables 20, 22, 24). The arm joints have much higher Q

ST values than do the leg joints, indicating less similarity across populations.

The patterning for diaphyseal properties is less clear, in part because all analyses

yield Q values that are comparably low in the modern test. Proximal Q values are

ST ST slightly higher than those for distal elements, leg Q values are slightly higher than those

ST of the arms, and there is no difference between the arms themselves (table 21). For the

temporal test, again proximal diaphyseal elements report slightly higher Q values than

ST do distal ones, legs slightly higher values than arms, and the left arm a slightly higher value than the right, though again these differences are small (table 22).

Table 21. Unbiased QST values and standard errors (se) for various long bone groupings, modern populations Test Variables n =0.55 se =0.75 se =1 se Proximal Joints 385 0.096𝟐𝟐 0.005 0.070𝟐𝟐 0.004 0.052𝟐𝟐 0.004 𝐡𝐡 𝐡𝐡 𝐡𝐡 Elements CSP 343 0.055 0.004 0.038 0.004 0.026 0.003 Distal Joints 324 0.129 0.008 0.096 0.007 0.071 0.006 Elements CSP 299 0.043 0.004 0.029 0.004 0.018 0.003 Right Joints 292 0.106 0.006 0.077 0.005 0.056 0.005 Arm CSP 278 0.034 0.005 0.021 0.004 0.012 0.004 Left Joints 295 0.157 0.006 0.118 0.006 0.088 0.006 Arm CSP 275 0.038 0.005 0.024 0.005 0.014 0.004 Joints 369 0.126 0.006 0.093 0.005 0.069 0.005 Legs CSP 321 0.055 0.004 0.037 0.004 0.025 0.003 Refer to table 19 caption for legend of abbreviations found in this table

108

Table 22. Unbiased QST values and standard errors (se) for various long bone groupings, modern and Medieval populations Test Variables n =0.55 se =0.75 se =1 se Proximal Joints 604 0.147𝟐𝟐 0.004 0.11𝟐𝟐 0.004 0.083𝟐𝟐 0.004 𝐡𝐡 𝐡𝐡 𝐡𝐡 Elements CSP 472 0.06 0.004 0.041 0.003 0.028 0.003 Distal Joints 477 0.125 0.006 0.093 0.006 0.069 0.005 Joints CSP 373 0.045 0.004 0.029 0.003 0.018 0.003 Right Joints 425 0.189 0.006 0.143 0.006 0.109 0.005 Arm CSP 341 0.04 0.005 0.025 0.004 0.014 0.004 Left Joints 409 0.177 0.006 0.134 0.005 0.101 0.005 Arm CSP 330 0.05 0.005 0.033 0.005 0.02 0.004 Joints 576 0.13 0.005 0.097 0.005 0.072 0.004 Legs CSP 424 0.056 0.003 0.038 0.003 0.025 0.003 Refer to table 19 caption for legend of abbreviations found in this table

The same general pattern is evident in bone-specific analyses (tables 23-24). For the

joints, among-population differentiation is lower for the proximal bones in the modern

test. The same is true of the temporal test with the exception of the right radius. The arm

joints have higher Q values than their leg homologous for both tests, again with the

ST exception of the right radius in the temporal test, which reports a Q value slightly under

ST that of the tibia (table 24).

This pattern differs for diaphyseal properties in that femoral Q is higher than all

ST other bones in both tests (tables 23 and 24). Leg bones generally report slightly higher

Q values than do the arms, though, again, differences are small, even in the temporal

ST test. Regarding interlimb differences, the femur has higher Q values compared to

ST homologues in the arm, while there is little difference among distal elements for both tests. The overall picture that emerges is that not only is among-population differentiation higher for the joints than the diaphyseal properties, but that it is lower for proximal joints than distal ones. The same is not true of among-population differentiation in diaphyseal

109 properties, where the femur consistently reported the highest Q , with little difference

ST reported for the other five bones.

Table 23. Unbiased QST values and standard errors (se) by bone, modern populations Bone Test n =0.55 se =0.75 se =1 se Joints 395 0.107𝟐𝟐 0.008 0.079𝟐𝟐 0.007 0.058𝟐𝟐 0.007 Femur 𝐡𝐡 𝐡𝐡 𝐡𝐡 CSP 370 0.07 0.006 0.049 0.005 0.035 0.005 Joints 377 0.162 0.009 0.122 0.009 0.093 0.008 Tibia CSP 334 0.051 0.005 0.035 0.005 0.024 0.004 Right Joints 392 0.113 0.006 0.084 0.005 0.062 0.005 Humerus CSP 347 0.051 0.006 0.035 0.006 0.024 0.005 Left Joints 379 0.138 0.006 0.103 0.006 0.078 0.005 Humerus CSP 337 0.048 0.007 0.033 0.006 0.022 0.005 Right Joints 308 0.165 0.012 0.125 0.012 0.095 0.011 Radius CSP 299 0.047 0.007 0.031 0.007 0.02 0.005 Left Joints 317 0.242 0.013 0.188 0.013 0.146 0.012 Radius CSP 298 0.049 0.008 0.033 0.007 0.021 0.006 Refer to table 19 caption for legend of abbreviations found in this table

Table 24. Unbiased QST values and standard errors (se) by bone, modern and Medieval populations Bone Test n =0.55 se =0.75 se =1 se Joints 629 0.119𝟐𝟐 0.007 0.088𝟐𝟐 0.006 0.066𝟐𝟐 0.006 Femur 𝐡𝐡 𝐡𝐡 𝐡𝐡 CSP 561 0.07 0.005 0.05 0.004 0.035 0.004 Joints 596 0.16 0.007 0.121 0.007 0.091 0.006 Tibia CSP 457 0.05 0.004 0.033 0.004 0.022 0.003 Right Joints 632 0.22 0.006 0.17 0.005 0.131 0.005 Humerus CSP 480 0.048 0.005 0.033 0.005 0.022 0.004 Left Joints 605 0.186 0.006 0.142 0.005 0.109 0.005 Humerus CSP 466 0.058 0.006 0.04 0.005 0.028 0.005 Right Joints 449 0.157 0.011 0.118 0.01 0.089 0.009 Radius CSP 388 0.051 0.007 0.034 0.006 0.022 0.005 Left Joints 452 0.216 0.011 0.166 0.011 0.128 0.009 Radius CSP 384 0.05 0.007 0.033 0.006 0.021 0.005 Refer to table 19 caption for legend of abbreviations found in this table .

110

5.5 Interpopulation Relationships

While the Q values can show the overall levels of population differentiation, they

ST cannot reveal the nature or pattern of among-population relationships. For this, relationship matrices (R-matrices) were generated to show the similarity among populations, while distance matrices were subjected to multidimensional scaling (MDS) so that visual presentations of population distance could be generated. Only the R- matrices and D-matrices corresponding to minimum Q are presented here for reasons

ST discussed in the Materials and Methods section.

5.5.1 Relationship Matrices

Tables 25-28 present the R-matrices for modern populations for each of the four broad anatomical groupings: craniofacial properties, long bone joints, long bone lengths, and diaphyseal (cross-sectional) properties. For each matrix, the main diagonal reflects that population’s distance from the centroid, with higher values reflecting “more distant” populations. The off-angle elements reflect the predicted relationship between the two populations corresponding to that element. Off-angle elements are color-coded to highlight similarity values such that green reflects populations that are more similar than the average (element value 0.01), red those which are more dissimilar than the average

(element value 0.01). ≥

Results ≤for the modern craniofacial R-matrices show geographic patterning.

Specifically, Southern European populations report closer relationships to one another than to the other samples under analysis, though this relationship is closer among the two

Italian samples. The the same is true of South African populations, which all share a positive relationship. England also shows a closer association to South African

111 populations than it does to Southern European populations (table 25). Each of the long bone R-matrices (joint, lengths, and diaphyseal properties) predicts the same general pattern of relationships among the Southern Europeans and among the South Africans as does the craniofacial R-matrix, though the predicted strength of these relationships is generally lower in the diaphyseal matrix, comparable for the joint matrix, and higher for the length matrix (tables 26-28). There are a few exceptions, particularly in the joint and

CSP R-matrices in which England shares no positive affinity to any of the South African samples, and in the length matrix in which England is not shown to have a positive affinity to Pretoria, but does have a positive affinity to Cape Town and Johannesburg.

Table 25. R-matrices for craniofacial data, modern populations SouSh Italy Sass Lis CT_E Jo_E Pr_E SouSh 0.037 Italy -0.034 0.121 Sass -0.029 0.013 0.122 Lis -0.007 -0.003 0.047 0.052 CT_E 0.001 -0.032 -0.062 -0.021 0.045 Jo_E 0.018 -0.044 -0.059 -0.032 0.030 0.042 Pr_E 0.001 -0.033 -0.042 -0.043 0.031 0.036 0.043 Abbreviations: SouSh (South Shields, England), Sass (Sassari, Sardinia), Lis (Lisbon, Portugal), CT_E (Cape Town, South Africa, European ancestry), Jo_E (Johannesburg, South Africa, European ancestry), Pr_E (Pretoria, South Africa, European ancestry)

Table 26. R-matrices for joint data, modern populations SouSh Italy Sass Lis CT_E Jo_E Pr_E SouSh 0.052 Italy -0.005 0.050 Sass -0.010 0.014 0.042 Lis -0.019 0.041 0.022 0.112 CT_E -0.005 -0.046 -0.034 -0.053 0.066 Jo_E -0.008 -0.032 -0.021 -0.047 0.039 0.025 Pr_E -0.026 -0.034 -0.023 -0.063 0.021 0.032 0.082 See table 25 for abbreviations

112

Table 27. R-matrices for length data, modern populations SouSh Italy Sass Lis CT_E Jo_E Pr_E SouSh 0.042 Italy -0.011 0.031 Sass -0.039 0.057 0.206 Lis -0.046 0.025 0.115 0.067 CT_E 0.022 -0.051 -0.118 -0.058 0.069 Jo_E 0.024 -0.031 -0.125 -0.067 0.066 0.061 Pr_E -0.013 -0.030 -0.107 -0.042 0.058 0.061 0.064 See table 25 for abbreviations

Table 28. R-matrices for cross-sectional data, modern populations SouSh Italy Sass Lis CT_E Jo_E Pr_E SouSh 0.043 Italy 0.001 0.021 Sass -0.017 0.007 0.027 Lis -0.014 0.005 0.016 0.026 CT_E -0.015 -0.016 -0.020 -0.011 0.015 Jo_E -0.011 -0.019 -0.026 -0.015 0.021 0.023 Pr_E -0.020 -0.010 -0.014 -0.014 0.014 0.014 0.019 See table 25 for abbreviations

Overall, there is less correspondance among the R-matrices for the temporal test than

those for the modern test. In particular, the inclusion of Medieval samples does reveal

some geographic patterning for the craniofacial data (table 29). The three modern South

African samples still show positive affinity to one another, though their predicted

relationship to South Shields, the modern English sample, has weakened and South

Shields now shows a positive relationship to the Northern and Southern Medieval English samples. The modern South European samples also retain a postiive affinity to one

another, but, interestingly, the Medieval Italian sample does not follow suit (discussed

below).

113

As seen with the modern test, the predicted strength of between-population relationships

(positive or negative) are generally lowest for the diaphyseal properties, highest for length dimensions, and intermediate for joint properties. That said, there are some interesting consistencies between the craniofacial and various long bone matrices (tables

30-32). For example, all four report some positive relationship among the English

Medieval and South Shields samples and among the three South African samples which consistently report a higher affinity to one another than to any other sample under analysis. All four R-matrices also detect a positive relationship among the three modern

South European samples which, again, show higher affinities to one another than to any other sample under analysis.

There are some differences as well, the most notable being the inconsistent nature of predicted relationships between the Medieval Italians and other samples from one test to the next. The cranifacial matrix, for example, shows a (very weakly) positive relationship between Medieval Italians and the two Medieval English samples and, interestingly, the three South African samples, while at the same time reporting a negative relationship to all three modern South European samples (table 29). The shared affinity among Medieval samples is maintained in the joint matrix, but any affinity to the

South African samples is lost here (table 30). Finally, the affinity among the Italian and

English Medieval samples does not present in the length or diaphyseal matrices (tables

31-32). In the length matrix, for example, the Medieval Italians now show the highest affinity to two modern South European samples: Sassari and Lisbon.

The long bone joint R-matrices show a weakly positive relationship between the

North English Medieval and South Shields samples, but the Southern Medieval sample

114 loses its positive affinity to the other English samples. The Medieval Italian sample still reports a positive affinity to the two English Medieval samples but now has no positive affinity to any other sample. Another key difference is that the diaphyseal R-matrix fails to report a positive relationship between any South African sample and the South Shields sample, while these relationships were, to some degree, detected by long bone lengths, joints, and craniofacial data (though the relationships between South Shields and individual South African samples was inconsistent from one test ot the next).

This variation in results suggests that while some coherent geographic (and presumably genetic) structure is present across all matrices, there are likely other mechanisms that can significantly shape variation in the Medieval samples, and in particular that variation in the different anatomical regions may reflect different environmental stimuli (functional or otherwise).

115

Table 29. R-matrices for craniofacial data, modern and Medieval populations Eng_ Eng_ South Ital_ Italy Sass Lis CT_E Jo_E Pr_E Med_N Med_S Sh Med Eng_ 0.034 Med_N Eng_ 0.039 0.075 Med_S South 0.011 0.013 0.023 Sh Italy -0.017 -0.009 -0.041 0.123

Sass -0.039 -0.087 -0.020 0.050 0.144 Ital_ 0.010 0.002 0.000 -0.013 -0.008 0.008 Med Lis -0.034 -0.044 -0.002 0.019 0.070 -0.022 0.073

CT_E -0.012 0.004 -0.005 -0.038 -0.046 0.002 -0.011 0.040

Jo_E 0.000 -0.010 0.011 -0.050 -0.046 0.001 -0.023 0.029 0.043

Pr_E -0.005 -0.011 -0.004 -0.037 -0.029 0.011 -0.033 0.028 0.035 0.039 Abbreviations: Eng_Med_N (Medieval England, northern region), Eng_Med_S (Medieval England, southern region), SouSh (South Shields, England), Sass (Sassari, Sardinia), Ital_Med (Medieval Italy), Lis (Lisbon, Portugal), CT_E (Cape Town, South Africa, European ancestry), Jo_E (Johannesburg, South Africa, European ancestry), Pr_E (Pretoria, South Africa, European ancestry). All samples are from modern time period unless otherwise specified.

Table 30. R-matrices for joint data, modern and Medieval populations Eng_ Eng_ South Ital_ Italy Sass Lis CT_E Jo_E Pr_E Med_N Med_S Sh Med Eng_ 0.051 Med_N Eng_ 0.070 0.146 Med_S South 0.001 -0.024 0.044 Sh Italy -0.043 -0.057 0.009 0.076

Sass -0.023 -0.042 -0.006 0.033 0.048 Ital_ 0.060 0.101 -0.017 -0.065 -0.027 0.084 Med Lis -0.039 -0.075 -0.010 0.071 0.038 -0.058 0.144

CT_E -0.018 -0.053 0.001 -0.021 -0.022 -0.027 -0.026 0.075

Jo_E -0.023 -0.045 -0.002 -0.008 -0.008 -0.029 -0.022 0.048 0.034

Pr_E -0.046 -0.042 -0.017 -0.008 -0.002 -0.032 -0.030 0.031 0.043 0.094 See table 29 for abbreviations

116

Table 31. R-matrices for length data, modern and Medieval populations Eng_ Eng_ Ital_ SouthSh Italy Sass Lis CT_E Jo_E Pr_E Med_N Med_S Med Eng_ 0.025 Med_N Eng_ 0.028 0.023 Med_S South 0.019 0.014 0.046 Sh Italy 0.015 0.004 -0.016 0.028

Sass 0.020 -0.048 -0.045 0.054 0.209

Ital_ -0.011 -0.009 -0.041 -0.010 0.027 0.037 Med Lis -0.012 -0.045 -0.048 0.026 0.121 0.028 0.077

CT_E -0.037 -0.014 0.033 -0.049 -0.114 -0.016 -0.050 0.089

Jo_E -0.020 0.017 0.028 -0.033 -0.127 -0.021 -0.065 0.077 0.066

Pr_E -0.037 0.003 -0.010 -0.031 -0.106 0.007 -0.038 0.068 0.066 0.068 See table 29 for abbreviations

Table 32. R-matrices for cross-sectional data, modern and Medieval populations Eng_ Eng_ South Ital_ Italy Sass Lis CT_E Jo_E Pr_E Med_N Med_S Sh Med Eng_ 0.007 Med_N Eng_ 0.025 0.032 Med_S South 0.006 0.015 0.032 Sh Italy -0.005 -0.006 -0.006 0.019

Sass -0.004 -0.008 -0.021 0.009 0.030 Ital_ -0.006 -0.009 -0.001 -0.002 -0.010 0.022 Med Lis -0.017 -0.017 -0.018 0.007 0.021 -0.006 0.033

CT_E -0.008 -0.021 -0.014 -0.011 -0.016 -0.005 -0.002 0.020

Jo_E -0.011 -0.032 -0.009 -0.012 -0.019 -0.001 -0.005 0.026 0.030

Pr_E -0.010 -0.024 -0.017 -0.005 -0.009 -0.003 -0.003 0.018 0.021 0.021 See table 29 for abbreviations

117

5.5.2 Distance Matrices and Multidimensional Scaling Plots

In addition to R-matrices, which reflect similarity among populations, distance matrices

were calculated to asses the structure of among-population relationships by subjecting them to multidimensional scaling (MDS) and plotting the first and second principal coordinates to allow for easy visualization of among-population relationships. The following figures show plots of the first and second principal coordinates organized by anatomical region for both modern and temporal tests. The data points on each plot are color-coded to correspond with the geographical region from which that population arose and a legend is provided with each figure. Results for modern samples are presented first, followed by results of the temporal test.

5.5.2.1 Modern Populations

Results of MDS tests for all four regions (craniofacial data, long bone joints, lengths and cross-sectional properties) show geographical patterning (figures 19-22). The first principal coordinate for all four anatomical regions seperates the Southern European populations from the English and South African populations. One consistent difference between the cranial and various long bone plots regards the position of Bologna. The cranial plot positions it more distant from the other two South European populations, while the three long bone plots show a much closer relationships among Italy, Portugal and Sardinia. The craniofacial plot also positions South Shields (modern England) as much closer to the three South African samples than do the three long bone plots where

South Shields is plotted as more distant, particularly in the joint and diaphyseal plots. In

118

general, the South African samples tighly cluster on all plots except for that

corresponding to the joints.

Separating the long bones into proximal and distal groupings has little effect on

the overall pattern, though the position of South Shields does shift on plots for the distal

diaphysis (figures 23-24). When the long bones are grouped by arm (figures 27-30) or leg

(figures 31-32), geographic patterning is still detectable in the right arm and in the leg,

particularly for the cross-sectional data. However, this organization weakens in plots of

the left arm, where joint plots now show clustering of South Shields to Johannesburg and

Cape Town, while Pretoria seperates along the y-axis. The diaphyseal plots for the left arm also show some loose geographic structure, but clustering among the three South

African populations is weaker here than was seen for either the right arm or leg diaphyseal plots. Interestingly, there tends to be distant placement of Pretoria relative to the other South African samples on joint and diaphyseal plots regardless of limb.

119

Figure 19. Multidimensional scaling (MDS) plots generating using craniofacial distance matrices for modern populations

120

Figure 20. Multidimensional scaling (MDS) plots generating using long bone length distance matrices for modern populations

121

Figure 21. Multidimensional scaling (MDS) plots generating using long bone joint distance matrices for modern populations

122

Figure 22. Multidimensional scaling (MDS) plots generating using long bone diaphyseal distance matrices for modern populations

123

Figure 23. Multidimensional scaling (MDS) plots generating using proximal long bone joint distance matrices for modern populations

124

Figure 24. Multidimensional scaling (MDS) plots generating using proximal long bone diaphyseal distance matrices for modern populations

125

Figure 25. Multidimensional scaling (MDS) plots generating using distal long bone joint distance matrices for modern populations

126

Figure 26. Multidimensional scaling (MDS) plots generating using distal long bone diaphyseal distance matrices for modern populations

127

Figure 27. Multidimensional scaling (MDS) plots generating using right arm long bone joint distance matrices for modern populations

128

Figure 28. Multidimensional scaling (MDS) plots generating using right arm long bone diaphyseal distance matrices for modern populations

129

Figure 29. Multidimensional scaling (MDS) plots generating using left arm long bone joint distance matrices for modern populations

130

Figure 30. Multidimensional scaling (MDS) plots generating using left arm long bone diaphyseal distance matrices for modern populations

131

Figure 31. Multidimensional scaling (MDS) plots generating using leg long bone joint distance matrices for modern populations

132

Figure 32. Multidimensional scaling (MDS) plots generating using leg long bone diaphyseal distance matrices for modern populations

133

5.5.2.2 Temporal Test (Modern and Medieval Samples)

Some noticable differences arise with the inclusion of the Medieval samples (figures 33-

36). The most apparent is the three Medieval samples plot near one another in both the cranial and joint plots, with South Shields (modern England) landing intermediate between the Medieval English and modern South African samples. The pattern for both long bone lengths and cross-sections differs in that modern England clusters with the

Medieval English samples, while Medieval Italy more closely aligns to the other South

European samples on the length plot and falls in intermediate space on the diaphyseal plots.

This pattern for the joints remains generally consistent when the long bones are separated into proximal and distal elements (figures 37 and 39), right arm (figure 41), or leg (figure 45) groupings. For the left arm joints, however, South Shields moves from a relatively intermediate space to form a clustur with the three South African samples form another (figure 43).

All cross-sectional plots show tight clustering of the three South African samples, a strong association between the Medieval English samples, and a strong assocaition among the modern South European samples. The relationship of South Shields to the

South African and Medieval English populations does vary by analysis. Plots for cross-

sectional properties of proximal elements show a close association among the three

English samples, the three South African samples, and somewhat looser clustering of the

modern South European samples with Medieval Italy falling intermediate to the English

and South European samples (figure 38). This differs from the distal elements which,

while also showing a cluster of the South African samples, now has all Medieval

134 samples, South Shields, and Bologna clustering together (figure 40). Predicted relationships also differ somewhat by limb, in that the right arm and leg plots show South

Shields to have a close affinity to the Medieval English samples, while the left arm shows it plotting intermediate between the South African and Medieval English samples (figures

42, 44, and 46). The same is true of the Medieval Italian sample which falls with the

Medieval samples for leg diaphseal properties (figure 46), in a weak cluster with the

Southern European samples for the right arm (figure 42), and a space intermediate for the left arm (figure 44). The somewhat shifting relationships of the South Shields sample and the Medieval Italian sample may well reflect the competing signals of genetics and function, in particular that some anatomical regions reflect one signal better than the other.

135

Figure 33. Multidimensional scaling (MDS) plots generating using craniofacial distance matrices for modern and Medieval populations.

136

Figure 34. Multidimensional scaling (MDS) plots generating using long bone length distance matrices for modern and Medieval populations

137

Figure 35. Multidimensional scaling (MDS) plots generating using long bone joint distance matrices for modern and Medieval populations

138

Figure 36. Multidimensional scaling (MDS) plots generating using long bone diaphyseal distance matrices for modern and Medieval populations

139

Figure 37. Multidimensional scaling (MDS) plots generating using proximal long bone joint distance matrices for modern and Medieval populations

140

Figure 38. Multidimensional scaling (MDS) plots generating using proximal long bone diaphyseal distance matrices for modern and Medieval populations

141

Figure 39. Multidimensional scaling (MDS) plots generating using distal long bone joint distance matrices for modern and Medieval populations

142

Figure 40. Multidimensional scaling (MDS) plots generating using distal long bone diaphyseal distance matrices for modern and Medieval populations

143

Figure 41. Multidimensional scaling (MDS) plots generating using right arm long bone joint distance matrices for modern and Medieval populations

144

Figure 42. Multidimensional scaling (MDS) plots generating using right arm long bone diaphyseal distance matrices for modern and Medieval populations

145

Figure 43. Multidimensional scaling (MDS) plots generating using left arm long bone joint distance matrices for modern and Medieval populations

146

Figure 44. Multidimensional scaling (MDS) plots generating using left arm long bone diaphyseal distance matrices for modern and Medieval populations

147

Figure 45. Multidimensional scaling (MDS) plots generating using leg long bone joint distance matrices for modern and Medieval populations

148

Figure 46. Multidimensional scaling (MDS) plots generating using leg long bone diaphyseal distance matrices for modern and Medieval populations

149

5.6 Mantel Randomization Tests

Results of the Mantel randomization tests for both the modern (table 33) and the temporal tests (table 34) report a significant correlation between the craniofacial distance matrix

and each of the long bone distance matrices. With regard to correlations to the craniofacial distance matrices (one of the primary project aims), the highest correlation is seen with the long bone length matrices, the weakest with the joint matrices, and diaphyseal matrices intermediate (table 33). Interestingly, the strength of correlation between craniofacial and various long bone matrices decreases in the temporal test, with the greatest decrease seen for the length matrix and relatively minor differences for the joint and diaphyseal matrices (table 34). Results generated by the Ade4 package and by

PASSaGE are consistent for each test.

Table 33. Mantel test results for Modern populations based on 10,000 permutations Ade4 PASSaGE r p-value Z r p-value Crania~Lengths 0.805 0.003 6.840 0.805 < 0.001 Crania~Joints 0.443 0.042 6.046 0.443 0.027 Crania~Diaphyses 0.591 0.016 3.987 0.591 0.004 Length~Joints 0.627 0.012 6.422 0.627 0.003 Length~Diaphyses 0.750 0.001 4.300 0.750 < 0.001 Joints~Diaphyses 0.645 0.012 3.895 0.645 0.002 Note that r is presented for both tests simply to ensure that treatment of the data by each statistical program was comparable.

Table 34. Mantel test results for Modern and Medieval populations based on 10,000 permutations Ade4 PASSaGE r p-value Z r p-value Crania~Lengths 0.637 < 0.001 12.205 0.637 < 0.001 Crania~Joints 0.408 0.010 13.222 0.408 0.008 Crania~Diaphyses 0.529 < 0.001 7.711 0.529 < 0.001 Length~Joints 0.297 0.039 13.397 0.297 0.029 Length~Diaphyses 0.656 < 0.001 8.109 0.656 < 0.001 Joints~Diaphyses 0.524 0.003 8.916 0.524 < 0.001

150

When investigating the results of the permutation tests generated by PASSaGE (figures

47-8), the rate of underprediction far exceeded the rate of overprediction for every test.

This is the result one expects to see should the relationships between matrices be stronger than that predicted by chance alone (Diniz-Filho et al., 2013).

Underprediction Overprediction Equal

100 90 80 70 60 50 40 30 20 10 0 Rates of Prediction of Rates

Underprediction Overprediction Equal

100 90 80 70 60 50 40 30 20 10 0 Rates of Prediction of Rates

Figure 48. Rates of over- and under-prediction for Modern and Medieval samples based on Mantel permutations

151

5.7 Occupaton Test

As a specific test of function, another round of population distance predictions were generated based on diaphyseal properties of the modern samples, but with the Lisbon and

Bologna samples separated into two subgroups: one reflecting “heavy intensity” occupations and one reflecting “light intensity” (light manual or non-manual) occupations. Prior to calculating genetic distance predictions, J (an overall indication of robusticity) was calculated for each diaphyseal location, standardized per protocols outlined in the Materials and Methods section, and compared across all four subsamples

(table 35, figures 49-50). These were then subject to a series of comparative t-tests

(unequal variances assumed, Welsh modification), the first testing for significant differences in J between heavy-intensity and low-intensity subgroupings from the same sample, and a second that tested for significant differences between Lisbon and

Bolognese subsamples within each occupational category (tables 36-37).

The first test confirmed significant differences in diaphyseal robusticity when comparing occupational categories within each sample, the results being significant for left and right arm diaphyses in both samples, and for those parts of the leg nearest the knee in the Lisbon sample (i.e., distal femur, midshaft and proximal tibia) (table 36). This result is expected if the differences in these two occupational designations are sufficiently strong to trigger bone functional adaptation in response to increased demands of physical activity in the heavy-intensity subgrouping, and it is interesting to note these effects seem more widespread in the Lisbon sample. The second test confirmed significant differences between the two samples when comparing subsamples of similar occupational demographics (table 37). The low-intensity subgrouping reported significant differences

152

in J for all fourteen diaphyseal sampling locations except for the midshaft left humerus

and distal femur. The heavy-intensity subgroupings reported significant differences for

all leg diaphyseal properties, but, interestingly, no significant difference between arm

diaphyseal properties except for the right radius (35%).

The minimum unbiased Q value was 0.026, comparable to that reported for

ST both the modern and temporal tests (tables 19 and 20). The R-matrix shows both the heavy-intensity and light-intensity Lisbon subgroupings have a positive relationship to most of the other Southern European populations, with the strongest affinity seen between the heavy-intensity subgroup and the Sardinian sample (table 38). The low-

intensity subgroup shows the strongest affinity to its heavy-intensity counterpart. For

Bologna, the highest affinity either the light-intensity or heavy-intensity subsamples has is to one another. While the heavy-intensity Bologna subsample does not share a positive affinity to either Lisbon subsample, the two light-intensity subsamples share a very weakly positive relationship. As expected, the pattern of other among-population relationships for the other samples are relatively unchanged. Figure 51 shows plots of the first and second principal coordinates based on MDS of the distance matrix. They show that all South European samples and subsamples fall together on the MDS plots, with the two heavy-intensity subsamples being most distant from one another. These results suggest that there may be some slight effect of occupational intensity on genetic predictions based on the more plastic diaphyseal regions, but that the overall pattern mirrors that reported for the original analysis (figure 22). In short, populations still cluster largely along geographic lines with some possible functional interference.

153

Tibia 65%

Tibia 50%

Tibia 35%

Femur 80%

Femur 50%

Femur 20%

0 200 400 600 800 1000 1200 1400 1600 Lis_Lht Lis_Heavy Bo_Lht Bo_Heavy

Figure 49. A comparison of femoral and tibial J values (standardized) for heavy and light-industrial subpopulations from Lisbon and Bolognese samples, mean and 95% confidence level (error bars).

154

Radius (Right) 50%

Radius (Right) 35%

Radius (Left) 50%

Radius (Left) 35%

Humerus (Right) 50%

Humerus (Right) 35%

Humerus (Left) 50%

Humerus (Left) 35%

0 50 100 150 200 250 300 350 Lis_Lht Lis_Heavy Bo_Lht Bo_Heavy

Figure 50. A comparison of humeral and radial J values (standardized) for heavy and light-industrial subpopulations from Lisbon and Bolognese samples, mean and 95% confidence level (error bars).

155

Table 35. Mean J and standard error (95% confidence) by population and occupational intensity Subpopulation Bone Region Lis_Heavy Lis_Lht Bo_Heavy Bo_Lht Humerus (R) 35% 236.3 ± 24.9 204.0 ± 12.6 219.5 ± 24.5 174.2 ± 21.1 Humerus (R) 50% 286 ± 30.7 246.1 ± 16.9 260.3 ± 28.2 208.4 ± 22.4 Humerus (L) 35% 235.1 ± 27.3 195.3 ± 13.0 203.5 ± 17.2 159.6 ± 18.1 Humerus (L) 50% 273.9 ± 33.9 222.5 ± 18.0 234.5 ± 22 193.6 ± 23.25 Radius (R) 35% 80.17 ± 9.7 61.4 ± 4 65.7 ± 6.6 49.2 ± 5.8 Radius (R) 50% 80.16 ± 7.9 64.20 ± 4 67.6 ± 8.3 51.7 ± 5.4 Radius (L) 35% 77.32 ± 8.3 57.63 ± 4.4 64.2 ± 5.9 47.8 ± 6.6 Radius (L) 50% 81.8 ± 9.1 60.1 ± 4.7 65.9 ± 6.2 47.4 ± 5.4 Femur 20% 1449.5 ± 123 1184.8 ± 70.1 1170.2 ± 130.1 1077.5 ± 115.3 Femur 50% 556.9 ± 33.9 505 ± 26.5 429.6 ± 31.1 384 ± 31.5 Femur 80% 693.2 ± 43.7 657.0 ± 35.8 611.6 ± 49.8 543.7 ± 45.8 Tibia 35% 366 ± 23.9 334.8 ± 21.0 261.2 ± 17.8 247.4 ± 24.1 Tibia 50% 507.4 ± 30.4 449 ± 28.1 364. ± 29.7 319 ± 34.1 Tibia 65% 809.6 ± 56.5 691 ± 44.4 557.6 ± 48.1 502 ± 62

Table 36. T-test results comparing heavy and light intensity occupations by sample Bologna Lisbon Bone Region F p-value F p-value Humerus (R) 35% 5.629 0.023 * 4.691 0.033 * Humerus (R) 50% 5.967 0.02 * 4.162 0.045 * Humerus (L) 35% 7.966 0.007 * 6.749 0.011 * Humerus (L) 50% 4.163 0.048 * 6.195 0.015 * Radius (R) 35% 11.57 0.001 * 14.16 < 0.001 * Radius (R) 50% 9.902 0.003 * 11.49 0.001 * Radius (L) 35% 8.552 0.006 * 14.55 < 0.001 * Radius (L) 50% 14.49 < 0.001 * 14.85 < 0.001 * Femur 20% 0.99 0.325 12.91 < 0.001 * Femur 50% 3.468 0.07 3.897 0.052 Femur 80% 3.445 0.07 1.05 0.308 Tibia 35% 0.607 0.44 2.207 0.141 Tibia 50% 3.04 0.088 4.553 0.036 * Tibia 65% 1.469 0.232 6.983 0.01 * Abbreviations: right (R), left (L). *indicates significant result.

156

Table 37. T-test results comparing Bologna and Lisbon by occupational intensity Heavy intensity occupations Light intensity occupations Bone Region F p-value F p-value Humerus (R) 35% 0.727 0.400 4.509 0.037 * Humerus (R) 50% 1.187 0.284 4.299 0.041 * Humerus (L) 35% 2.603 0.117 8.202 0.005 * Humerus (L) 50% 2.59 0.118 2.807 0.098 Radius (R) 35% 4.789 0.036* 10.12 0.002 * Radius (R) 50% 4.057 0.052 10.87 0.001 * Radius (L) 35% 4.536 0.042 * 5.181 0.025 * Radius (L) 50% 5.576 0.025 * 8.693 0.004 * Femur 20% 9.054 0.005 * 2.541 0.114 Femur 50% 27.74 < 0.001 * 27.14 < 0.001 * Femur 80% 5.724 0.022 * 12.61 < 0.001 * Tibia 35% 42.75 < 0.001 * 21.63 < 0.001 * Tibia 50% 41.83 < 0.001 * 26.84 < 0.001 * Tibia 65% 40.7 < 0.001 * 21.23 < 0.001 * Abbreviations: right (R), left (L). *indicates significant result.

Table 38. R-matrices for cross-sectional data, modern populations SouSh Bo_Heavy Bo_Lht Sass Lis_Lht Lis_Heavy CT_E Jo_E Pr_E SouSh 0.048 Bo_Heavy 0.001 0.027 Bo_Lht -0.004 0.011 0.011 Sass -0.016 0.000 0.002 0.020 Lis_Lht -0.011 -0.006 0.001 0.007 0.021 Lis_Heavy -0.027 -0.008 -0.001 0.017 0.011 0.031 CT_E -0.008 -0.021 -0.014 -0.020 -0.007 -0.019 0.026 Jo_E -0.003 -0.024 -0.015 -0.022 -0.010 -0.020 0.030 0.030 Pr_E -0.014 -0.014 -0.010 -0.014 -0.015 -0.012 0.021 0.022 0.024

157

158

5.8 Chapter 5 References

Diniz-Filho JF, Soares TN, Lima JS, Dobrovolski R, Landeiro V, de Telles M, Rangel TF, and Bini L. 2013. Mantel test in population genetics. Genetics and Molecular Biology 36:475–485.

159

CHAPTER 6

DISCUSSION

6.1 Long Bones in Context

Section Highlights:

1. Differentiation. Q s generated by lengths, joints, and crania are

ST comparable both to one another and to Q s published in other

ST studies. Q s generated by diaphyseal properties are lower, likely

ST due to higher levels of intrapopulation variation.

2. Patterning. Despite this, however, all skeletal properties detect

some elements of among-population genetic structure when

considered in aggregate. The relationships predicted by different

long bone properties show geographic clustering consistent with an

“isolation-by-distance” model.

3. Population History. Long bones, even diaphyseal properties,

accurately detect more nuanced aspects of documented population

history. Two specific examples are the ability to discern the English

component of modern white South Africans, and clustering of

Southern European samples irrespective of known differences in

physical activity and occupation and even when functional signals

are present.

6.1.1. Differentiation

The overall, the Q values generated by the craniofacial and long bone data support

ST prior studies that show the majority of variation is found within, rather than between,

160

populations. Depending on the scale of analysis (local, regional, global), s for

𝑆𝑆𝑆𝑆 craniofacial traits typically range from 0.03-0.14, indicating that 3-14% of𝑄𝑄 total variation

exists between populations, while the remaining 86-97% exists within (Relethford, 2009).

This is on par with s reported for molecular and mitochondrial DNA (Manica et al.,

𝑆𝑆𝑆𝑆 2007), and corresponds𝐹𝐹 to what is known about human genetic diversity—that it is very

low compared to other species—much lower than even for chimpanzees, which humans

outnumber 35,000 to 1 (Relethford, 1994; Bowden et al., 2012). The range of craniofacial

=0.55 Q s in this study (0.12 for the modern samples, 0.112 for the modern and 2 ST Medievalℎ samples) is consistent with others that show 11-14% of craniofacial or odontometric variation exists among populations at a global scale, with lower levels reported for more regionally focused analyses (Relethford, 2001, 2009; Hanihara and

Ishida, 2005). The minimum Q here are slightly lower those reported by von Cramon-

ST Taubadel (2009) for various craniofacial𝑠𝑠 modules, which ranged from 0.07-0.13.

However, this could be due to the fact that her analysis includes populations that are more globally dispersed, while this analysis includes only individuals of European ancestry

(which, by design, should limit phenotypic diversity and generate smaller Q values).

ST While the =0.55 Q values for diaphyseal properties were much lower 2 ST (ranging from 0.034ℎ -0.07), those for craniofacial, long bone length, and long bone joint

data range from 0.12-0.139 in the modern analysis and 0.112-0.143 for the temporal

analysis, which meets or exceeds most of the values produced in other studies based on

different anatomical criteria (table 39). They are also comparable to F values derived

ST from various molecular data sources, which are reported to range from 0.1-0.15 in global

studies (Relethford, 1998). They are, however, considerably lower than values reported

161 for skin color, as are all skeletal, anthropometric, or even genetic traits reported in this and other studies (Relethford, 2002). Because Mantel tests were significant for all long bone properties in both the modern and temporal tests (tables 33-34), it does appear that long bones are moderately-to-highly correlated to craniofacial variation and are therefore all capable of detecting among-population variation.

Table 39. Unbiased QST values reported in other studies Trait Unbiased Geographic Range Source (SE) of Samples 𝐒𝐒𝐒𝐒 Craniometric 0.09 (0.030)𝐐𝐐 Sub-Saharan Africa Relethford (2001) Craniometric 0.05 (0.0023) Europe Relethford (2001) Craniometric 0.04 (0.0022) East Asia Relethford (2001) Craniometric 0.08 (0.0028) Australasia Relethford (2001) Craniometric 0.12 (0.0032) Polynesia Relethford (2001) Craniometric 0.11 (0.0032) Americas Relethford (2001) Odontometric traits 0.11 (0.003) - Global Hanihara and Ishida 0.13 (0.004) (2005) Craniometric 0.11 (0.0018) Old World (excludes Relethford (2002) Australia) Craniometric 0.15 ( 0.0014) Global Relethford (2002) Skin color ( =0.66) 0.87 (0.0008) Old World (excludes Relethford (2002) 2 Australia) Skin color (ℎ =0.66) 0.88 (0.0007) Global Relethford (2002) Craniometric 2traits ( =1) 0.06 (0.002) – Old World (excludes Relethford (1994) ℎ 2 0.08 (0.002) Australia) Craniometric traits (ℎ =1) 0.09 (0.002) – Global Relethford (1994) 2 0.09 (0.002) Head measurements ℎ(living 0.0050 (0.0013) – Ireland (1861-1920) Relethford et al. individuals) ( =1) 0.0087 (0.0007) (1997) Body measurements2 (living 0.0045 (0.0009) – Ireland (1861-1920) Relethford et al. individuals) (ℎ =1) 0.0047 (0.0014) (1997) Craniometric (vault)2 0.12 Africa, Asia, von Cramon- ( =1) ℎ South/Central Taubadel (2009) 2 Europe Craniometricℎ (zygotemporal) 0.10 Africa, Asia, von Cramon- ( =1) South/Central Taubadel (2009) 2 Europe Craniometricℎ (base) 0.10 Africa, Asia, von Cramon- ( =1) South/Central Taubadel (2009) 2 Europe Craniometricℎ (upper face) 0.13 Africa, Asia, von Cramon- ( =1) South/Central Taubadel (2009) 2 Europe Craniometricℎ (palate/maxilla) 0.07 Africa, Asia, von Cramon- ( =1) South/Central Taubadel (2009) 2 Europe ℎ

162

Lymphoblast microsatellites 0.06 Africa, Asia, von Cramon- ( =1) South/Central Taubadel (2009) 2 Europe Craniometricℎ (PC reflecting nasal 0.24 Global (Howells) Roseman and projection and breadth) Weaver (2004) Craniometric (PC reflecting 0.33 Global (Howells) Roseman and “inverse relationship between Weaver (2004) upper nasal breadth and projection” 260) Craniometric (PC reflecting 0.08 Global (Howells) Roseman and zygomaxillary features, mastoid Weaver (2004) width) Craniometric (PC reflecting vault 0.19 Global (Howells) Roseman and chords, upper facial projection) Weaver (2004) Craniometric (PC reflecting 0.07 Global (Howells) Roseman and parietal, occipital contours) Weaver (2004) Craniometric (PC reflecting 0.04 Global (Howells) Roseman and mastoid properties, upper facial Weaver (2004) projection) * =0.55, unless otherwise specified. May arise from sex-specific or pooled-sex analyses.2 Standard errors are included in parentheses when available. “PC” means ℎ“principal component” in studies based on principal component analysis.

6.1.2. Patterning

Most of the current literature pertaining to among-population skeletal variation invariably focuses on craniofacial or dental traits. These studies show that there is a strong correspondence between genetic and phenotypic population structure, that the crania work as well as molecular data to elucidate interpopulation or interspecific relationships, and that distance or correlation matrices generated using morphological variation are highly correlated to those derived from molecular data (Cheverud, 1982, 1989;

Schneider and Blakeslee, 1990; Relethford et al., 1997; Relethford and Jorde, 1999;

Ackermann and Cheverud, 2000; Relethford, 2004; Hanihara and Ishida, 2005; Manica et al., 2007). The general consensus is that the craniofacial region is tightly integrated due to its critical role in housing the brain and sensory organs, unique mode of development, and key functional roles (e.g., masticatory apparatus), all of which limit the ability of selective pressure to significantly alter morphology (Lieberman et al. 2000; Cardini and

163

Elton, 2008).

Evaluation of among-population genetic distance typically reveals a strong correlation between geographic and genetic distances, indicating that populations close to one another in geographic space will also tend to cluster on multidimensional scaling

(MDS) plots. This “isolation by distance” (IBD) pattern is also seen in patterns of craniofacial variation, and likely stems from logistical factors in that closely situated populations engage in higher levels of gene flow than do distant ones (Relethford, 2004,

2009; von Cramon-Taubadel and Lycett, 2008; Smith, 2009; von Cramon-Taubadel,

2009). In this way, craniofacial variation can reflect key elements of population history and structure, including patterns of ancient migrations and the distance that descendant populations moved from their point-of-origin over time (Ross, 2004; González‐José et al., 2005). The patterns observed in the craniofacial and various long bone MDS plots for the modern samples are consistent with an IBD pattern in that they show distinct geographical clustering of South African and Southern European populations. The only deviation is the South Shields (English) sample which, despite being geographically disparate from South Africa, tends to be positioned near it on many of these plots.

However, this tendency is consistent with the known population history of South Africa during the colonial era and therefore does not violate IBD assumptions (to be discussed in section 6.1.3).

Relethford (1994) showed that overall human craniofacial variation was low, but that within-region variation closely mirrored DNA variation. This is supported by a number of other studies showing that when all variables are aggregated the whole cranium is selectively neutral, primarly reflecting gene flow and genetic drift

164

(Relethford, 2002, 2010; von Cramon‐Taubadel and W eaver, 2009) . This pattern

somewhat changes when the scope of analysis narrows to specific traits or regions (units)

within the crania, at which point some units reflect neutral evolutionary signals better than do others, perhaps due to functional or environmental interference on less canalized traits (Roseman, 2004; Roseman and Weaver, 2004; Harvati et al., 2006; Betti et al.,

2009, 2010; Martínez‐Abadías et al., 2009 ; Smith, 2009; von Cramon-Taubadel and

Smith, 2012). von Cramon-Taubadel (2011), for example, discovered that while most of the cranium conferred a genetic signal, any evidence of interpopulation genetic relationships was effectively erased when the scope narrowed to mandibular and alveolar traits, at which level function (mastication) appeared to be the predominate force shaping variation.

There is some evidence that a similar pattern may hold for the long bones as well.

When investigated in aggregate (i.e., all arm and leg bones together), the IBD pattern is stronger. However, once the analysis is partitioned into smaller units of inquiry (e.g., leg bones, arm bones), the geographical pattern remains detectable but the clustering on

MDS plots is at times weaker, particularly for diaphyseal regions. Furthermore, the

plotted relationships of some populations (e.g., Medieval Italy or South Shields) does

shift between these more focused analyses—for example, South Shields plots closer to the South African cluster on the left and right arm joint MDS plots (figures 41,43) but pulls closer to the Medieval English cluster on the leg joint MDS plot (figure 45). In neither does it tightly cluster with the other populations, but its location does shift depending on the anatomical region being analyzed. This could suggest that, similar to von Cramon-Taubadel (2011), certain skeletal regions are more susceptible to external

165

factors, including function, which weakens their ability to detect neutral processes.

Studies that investigate neutral evolutionary signals conveyed by long bones are limited, but they do exist. For example, Betti et al. (2012) investigated whether variation in pelvic and long bone dimensions in a large, global sample were consistent with the Out-of-

Africa (OoA) model of ancient human dispersal, and whether climate could explain any variation in pelvic and appendicular elements. They discovered that while pelvic variation was consistent with OoA expectations (and indeed matched cranial variation patterns), long bone variation was more affected by climatic variables. While they did not calculate Q or generate distance matrices, their results suggest that any utility of long

ST bones in genetic distance predictions would be effectively erased by climatic effects, at

least in these larger scale, global studies. In a similar study, von Cramon-Taubadel et al.

(2013) investigated whether among-population genetic affinities predicted using long

bone variation were affected by environmental variables across pre- and post-agricultural

populations. They similarly found that long bones (grouped by limb) were unsuccessful

at detecting genetic patterns, and in fact better reflected certain aspects of environment,

including temperature and latitude. Similar conclusions have been reached in studies of

global craniofacial variation, in which it was found that controlling for climatic variables

(e.g., temperature, vapor pressure) increased the strength of Mantel tests between genetic

distance matrices and those derived using various craniofacial units (Harvati and Weaver,

2008).

The differences between results from this dissertation and those of Betti et al.

(2012) and von Cramon Taubadel et al. (2013) may stem largely from methodological

differences, the most notable of which is that this study neither tests nor controls for

166

climatic differences. In fact, climatic effects are likely to be limited. According to the

Koppen Climatological Index, there is only a 14 point range between average

temperatures across the geographic regions under analysis. No region averages below

freezing or above 76 degrees Fahrenheit. This means the samples in this study encompass

a more restricted climatic range than those of both other studies, which also limits its

potential impact. Another primary difference is that the Betti et al. (2012) samples

represent a much broader geographical range, while the von Cramon-Taubadel samples

are drawn from a much broader temporal range. These are in addition to a number of

minor differences, including differences in standardization method (geometric mean as

opposed to standardizing by mass and physiological length), and a more extensive

emphasis on long bones, in particular diaphyseal properties, in this study. While both

Betti et al. (2012) and von Cramon-Taubadel (2013) include long bones in their analysis,

they were more focused on overall patterns of variation and therefore combined the

different long bone properties (lengths, articular dimensions, and diaphyseal diameter

data). They also restrict the diaphyseal data to the principal anatomical diameters,

generally from the midshaft (50%) region. As seen in the MDS plots and the strength of

Q values in this study, some aspects of bone morphology (e.g., lengths) report higher

ST levels of among-population and stronger correlations to craniofacial data than others.

Other researchers have investigated among-population differences in long bones as part of a broader “body proportion package” that also includes body breadth and mass dimensions. Much like the crania, researchers have parsed out ancient migration events by using similarity in body proportion as a proxy for among-population relationships. For example, Holliday (1997) discovered that European individuals from the Early Upper

167

Paleolithic had long-limbed proportions most akin to those of Sub-Saharan Africans, but

that these proportions changed quickly during the Last Glacial Maximum. He argues that

the similarity among early Europeans and Sub-Saharan Africans likely reflects some

genetic continuity between the two regions as early anatomically modern humans (AMH)

dispersed from Africa into Europe, but that subsequent selective pressures modified body proportion to that seen in more recent Europeans. These pressures were likely in response to either cold adaptation or related to an overall reduction in stature due to nutritional stress as game animals grew more scarce (Formicola and Giannecchini, 1999).

This may also help to explain why long bone length was far less variable than were articular dimensions, and, further, why length distance matrices consistently revealed the highest correlation to craniofacial matrices in the Mantel tests (tables 33-34).

This result is surprising given the large volume of research that illustrates the prevalence

of growth stunting due to early-life environmental stressors, such as illness, malnutrition,

and child labor (Steckel, 1995; Komlos, 1998). However, Holliday (1999:563) states that

limb lengths and proportion are under strong genetic regulation, an argument that would

imply canalization. This is supported by work by Silventoinen (2003) who concluded that

the majority (80%) of height variation in Western populations was due to genetic, rather

than environmental, factors, and further by the number of studies that show long bone

growth is, at least to some degree, regulated by the growth plate itself which can correct

for environmental insults (Emons et al., 2011).

Similar results are reported for studies of nonhuman primate as well. In an

investigation of the CV of cranial and postcranial elements across strepsirrhine and

tarsiiform taxa, researchers likewise found that CV values were lower for maximum long

168

bone lengths than for epiphyseal dimensions, though the authors posited no explanation

for this pattern other than the likelihood that smaller measurements are more subject to measurement error (Fulwood and Kramer, 2013). In a study of phylogenetic signal

conferred by long bones across nonhuman primates, O’Neill and Dobson (2008)

discovered that femoral cross-sectional properties did reflect phylogeny after being scaled

to a product of body mass and long bone length. However, articular dimensions did not

significantly detect phylogeny after they were standardized against body mass. The

authors conclude that joint dimensions are adaptive in nonhuman primates, being

optimally designed to support body mass as primates engage in different locomotor

modes, while the stronger phylogenetic signal detected by diaphyseal properties reflects

either the heritability of locomotor behaviors or that bone remodeling processes are under

some degree of genetic regulation (a point to be revisited in section 6.2).

There is also some evidence that long bone dimensions in both modern and

Medieval samples reflect some degree of canalization. In particular, a clear pattern

emerged for both sexes in which Coefficients of Variation (CVs) were lower for the

proximal arm and leg joints than for distal joints. This was mirrored by the length CVs,

which showed the humerus/femur to be less variable than the radius/tibia. This pattern is

consistent with expectations of what Hamrick (2001, 2007, referencing Hinchliffe

(1991)) called the “proximal stabilization” model which holds that the proximal skeletal

elements, which develop early embryonically, are generally less variable. This might also

explain why the proximal joint elements tended to generate more coherent clustering on

MDS plots, particularly for the temporal test—that these anatomical regions, being more

buffered against environmental perturbation, are more apt to reflect neutral evolutionary

169 signals and less likely to be influenced by functional/environmental stimuli. These results again mirror those of Holliday (1999) who discovered that length of the distal elements tended to vary more than did that of the proximal elements in a large sample distributed throughout Europe and Africa.

It is interesting, however, that patterns of CV among the diaphyseal variables do not adhere to the proximal stabilization model. In fact, the proximal-most point of the femur (80%) is the most variable region reported for females for all skeletal metric traits analyzed. For this, it does appear function is a likely explanation. In particular, the increased CV for femur 80% could stem from competing functional signals originating from the pelvis, or a “hip effect,” rather than reflecting axial loading patterns expected to predominate in the more distal regions of the leg. Ruff (2000) investigated overall bending and torsional strengths among geographically disparate populations with assumed similarity in physical activity (Pecos Puebloans and East Africans). He discovered J (scaled to mass and bone length) was comparable among the populations with one exception: femur 80%. He suggests that this region, being near to the hip, can introduce a “breadth effect” wherein the proximal femur is uniquely affected by body mass acting on a mediolateral moment arm (rather than axial). Indeed, the interpopulation difference he detected was erased when the femur 80% cross-sectional properties were scaled against bi-iliac breadth rather than physiological femur length.

Incidentally, this would also suggest that the more proximal regions of the femur—being more likely influenced by a mediolateral moment arm—would also be more susceptible to an effect of body mass, a trend that has been reported numerous times in the body

170

mass- and obesity-focused studies of diaphyseal cross-sectional properties (Moore, 2008;

Agostini, 2009; Agostini and Ross, 2011).

6.1.3. Population History

Another important outcome of this study was the prediction of among-population

relationships that can be supported by existing historical documentation. In particular,

two patterns consistently emerged both in the long bone and craniofacial analyses: [1] the

close association among the three modern South African samples and their affinity to the

modern English sample, and [2] the close association of the modern Southern European

samples despite clear differences in culture, activity, and occupation. The relevance of

these discoveries given historical context will be presented in turn.

6.1.3.1. South Africa

To understand why South African and English populations often plot near one another

despite thousands of miles of geographical separation, one must delve into colonial

history of South Africa. The emphasis will be largely on patterns of population

movement (migration and settlement) because genes follow people. The presentation of

South African history will begin in 1652, the year that a small number of Dutch migrants

arrived to the Cape region under the sponsorship of the Verenigde Oostindische

Compagnie (VOC). For many decades the VOC had dominated Asian trade routes, and

established the Cape settlement (along with several hundred others) to serve as a refreshment point for its ships (MacKinnon 2004, Schrire 2014). The first Cape migrants were overwhelmingly male (VOC sponsorship of female emigrants would come in the late 1700s) and primarily Dutch, though some French and Germanic influence was also

171

present and is reflected in Afrikaner (modern Dutch-descendant) genomes today

(MacKinnon, 2004; Greeff, 2007).

While the Dutch government intended for settlers to farm grain, a combination of

undervaluation and poor soil prompted the adoption stock farming to stave off starvation

and poverty (van der Merwe, 1994). Several years after the first immigrants arrived, the

VOC permitted several families to become independent farmers, and these “free burghs”

settled stock farming operations further inland where more farmable land was available

(Keegan, 1996). For several decades, the Dutch immigrants remained largely in the

southwest, establishing settlements nearby (e.g., Stellenbosch, Drakenstein). This lasted

until the early 18th century when, frustrated by the VOC’s continual price-fixing, a small group of settlers rebelled (Patterson, 1957). What followed can perhaps be seen as the emergence of a “cultural tendency” that would exemplify later Boer populations— namely, that “whenever… the colonial government, either Dutch or British, has encroached upon the interior provinces, the Boer population has followed the policy of receding before the aggressive foreign power, choosing independence rather than empire”

(Ridpath and Ellis, 1899:54). When it was felt that they were being too strongly regulated by the central governing powers, frontier settlers, or Trekboers, would frequently (with or without permission) push deeper inland to regions held by indigenous populations.

Small groups of Trekboers continued their eastward creep in the subsequent decades, creating their own farming settlements along the way. Because of the distance and the dangerous nature of the trails, the interior settlements were largely self-sufficient, and travel back to the more urban Cape region was restricted to essential supply runs, some trade, and civic obligations (Patterson, 1957; van der Merwe, 1994; Keegan, 1996).

172

It is likely that any gene flow between these interior settlements and the more populous and diverse Cape populations was limited. Hostile contact with the indigenous populations, in particular the Khoekhoen herders, was commonplace as Trekboers pushed further north and east onto Khoekhoen land (Heinrich and Schrire, 2014). Clashes and raids were routine until roughly 1713, when the arrival of smallpox coupled with a lengthy drought lead to the collapse of Khoe society (MacKinnon, 2004). The white population increased rapidly in the years following—a rate of nearly 3 percent annually through the 1700s (Greeff, 2007 citing (Gouws, 1981)). These increases were fueled almost exclusively by local reproduction than by the arrival of new immigrants

(Karayiorgou et al., 2004). As a result, there remains a high degree of consanguinity among living Afrikaners even today (Greeff, 2007), evidenced by low allelic diversity and inflated rates of heritable disorders traceable to original Cape colony founders

(Hayden et al., 1980; Brink et al., 1987; Goldman et al., 1996; Groenewald et al., 1998;

Tipping et al., 2001; et al., 2002; Karayiorgou et al., 2004). This is despite the fact that interracial marriage was fairly common through most of the 1700s, particularly between white males and admixed females in more urban regions where nearly 1 in 4 recorded marriages between 1688 and 1807 involved a ‘non-European’ spouse (Keegan, 1996).

In the late 1700s, the Cape region was temporarily overtaken by English forces and in 1805 it came under permanent English control. Shortly thereafter, thousands of

English immigrants arrived, most as part of a government-funded humanitarian initiative.

The new arrivals were given 100 acre plots with the intention that they would farm it, again providing a cheap source of goods for the mainland import market (MacKinnon,

2004). The English were more reluctant to wed indigenous peoples. Furthermore, they

173

enacted policies which, while not overtly racist in language, were clearly discriminatory in subtext (Keegan, 1996; MacKinnon, 2004). This holds some irony given that the

English government would emancipate all colony slaves in 1838 (Giliomee, 2003).

Boer frustrations intensified through the early 1800s, fueled largely by crowding

at the influx of British immigrants and frustratingly inadequate support by the English

government when it came to conflict with indigenous populations on the frontier lands

(Keegan, 1996). The frontier Trekboers continued their incremental expansion beyond

the margins of the Cape colony. They eventually moved beyond the Fish River onto a

swath of “neutral” land designated to buffer the colony from Xhosa lands to the east, a

legislative decision made in response to the frequent land disputes and cattle raids whose

recuperation commanded considerable military resources (Giliomee, 2003). While

settling homesteads here was illegal, this mattered little to the Boers who felt increasingly

disenfranchised. Piet Retief in 1837 published a manifesto in the Grahamstown Journal.

In it he listed a number of grievances as catalyst for an impending “great Boer

migration,” including the rushed emancipation of slaves, lack of protection for frontier

farms, religious differences, and a deep-felt sense of marginalization:

“Numerous reports having been circulated throughout ths colony, evidently with the intention of exciting in the minds of our countrymen a feeling of prejudice against those who have resolved to emigrate from a colony, where they have experienced for so many years past a series of the most vexatious and severe losses…We complain of the unjustifiable odium which has been cast upon us by interested and dishonest persons, under the cloak of religion, whose testimony is believed in England to the exclusion of all evidence in our favour; and we can foresee as the result of this prejudice nothing but the total ruin of the country...We quit this colony under the full assurance that the English government has nothing more to require of us, and will allow us to govern ourselves without its interference in future.”

174

This manifesto marked an official beginning of an Afrikaner Republic, or at least an organized pushed toward one. Starting in the mid-1830s, large numbers of Trekboers migrated further from the southwest Cape region than any group had before, pushing deeper into the northeastern interior of South Africa as part of a coordinated, mass migration later called the Great Trek (MacKinnon, 2004). Members of this migration were called Voortrekkers, and their goal was to settle lands outside of British influence where they could live autonomously (Patterson, 1957). Using ox-drawn wagons, groups

(sometimes numbering in the hundreds) would travel nearly 1000 inhospitable miles northward to the Transvaal region that today contains the major urban centers of Pretoria and Johannesburg, and to the Natal region to the east (Harrison, 1981) (see figure 52).

The Afrikaners of the interior would go on to have a long and acrimonious relationship with British rule and with the large and powerful Zulu populations who routinely killed and were killed by the Afrikaner migrants (Harrison, 1981). Bloody battles, massacres, and raids were a mainstay of life in the deep frontier.

In the following years, English interests remained primarily coastal, keeping hold of the expansive southwestern Cape Region (Cape Town) and acquiring the eastern Natal region which contained the strategic port city of Durban (figure 52). During this time, the

English government was largely content to leave the interior alone. Here, two Afrikaner

“Boer states” emerged as a result of the Great Trek: the South African Republic (or

Transvaal) to the North (encompassing modern Pretoria and Johannesburg), and the

Orange Free State immediately to the South (encompassing modern Bloemfontein).

These four districts would later become the first provinces after South African unification

175 in 1910, but by the late 1800s, the Boer states functioned more-or-less autonomously

(Muthien and Khosa, 1995).

Afrikaner autonomy was short-lived, however, and the English gaze would slowly shift inward upon the discovery that the Transvaal, inhospitable though it may be, was rich in gold. As stated by McClintock, (1991:106), “[o]nly upon the discovery of diamonds and gold were the Union Jack and the redcoats shipped out with any real sense of imperial mission.” It should be noted that the quantity of gold discovered here was not insignificant:

176

“[The most important gold strikes in the world] were unquestionably those of the Transvaal: the small-scale discoveries in the eastern Transvaal during the early 1880s were followed by the huge finds on the Witwatersrand beginning in 1885. Ten years later, the Transvaal had become the world’s top gold producer. Gold from the single region of Witwatersrand yielded roughly one-fifth of the globe’s total annual output…it derived solely from the reconstructed outcrop mines of the Reef*” Katz (1995:307- 308)

*The Reef describes a series of large, sloping stone outcrops that are rich in gold and which are disproportionately found in the Transvaal.

In fact, the economic boom generated by the mining industry has led some researchers to call this period “the Mineral Revolution” (Murray, 1992). In 1880 the British advanced on the Boer settlements in the Transvaal but were defeated, and for a time things were relatively uneventful. However, the Boer states had their share of English sympathizers, particularly those individuals with a capitalist bent. In a move that many Afrikaners today still herald as an intentioned push by wealthy mining CEOs vying for access to global markets, a failed “Jameson Raid” occurred in late 1895 as an attempt to incite local

English-sympathizers, particularly those in the mining industry, to revolt against the Boer

South African Republic. While the raid did not result in war officially, it is generally cited as being the primary catalyst, with some dispute among historians (Phimister,

1993).

In 1899, war was formally declared between the Boer states and the English government. While it was short in duration (three years), it resulted in a loss of nearly ten percent of the Afrikaner population (Grundlingh, 1999). Most deaths were of women and children who, after their homesteads were destroyed by English troops, were placed into several camps where they could be monitored (Krebs, 1992). The combination of

177

crowding, poor sanitation, and limited resources created an environment rife with disease,

particularly measles, and by the end of the war, the number of women and children who

perished (estimated to be just shy of 30,000) was twice that of men who died fighting on

both sides (McClintock, 1991, Krebs, 1992). The conditions in these camps were

apparently so deplorable that they sparked an ethical debate in homeland British media,

and also served as origin for the phrase “concentration camps,” infamously used by the

Germans in World War II with “clear reference to the British camps in South Africa”

(Heyningen, 2009:22).

Following the war in 1902, the country made a large-scale push towards industrial capitalism, a shift that pulled many Afrikaners from their struggling farming homesteads into the major cities throughout the 1920s and 1930s (Vincent, 2000).

However, the economic status of Afrikaners lagged significantly behind that of British descendants. Those who were willing to assimilate to English expectations found themselves relegated to lower class occupations associated with shop-keeping, low-level farming, or teaching (McClintock, 1991). Those who refused to assimilate were generally also refused work, as Afrikaners were overtly discriminated against by pro-English elites, particularly in the Transvaal (Vestergaard, 2001). However, this marginalization instigated a concerted effort among Afrikaners to unify their culture as distinct from the

English (McClintock, 1991). This included the elevation and cohesion of the “people’s

language” which had arisen organically through decades of Afrikaner interaction with the

many indigenous populations on the frontier lands (and which therefore contained many

different dialects). The language that emerged was the “Afrikaans” language still widely

spoken today and which was officially recognized in 1925 (Webb and Kriel, 2000). It

178

was also in the years following the Anglo-Boer war that that the first Afrikaans-only newspapers were published, several prominent Afrikaner political organizations were formed, and large-scale demonstrations grew more commonplace. What emerged in the first half of the 20th century was a renewed sense of Afrikaner culture that, despite

representing the defeated (both of the War and of the economic spoils) remained a thorn

in the side of the English government. In fact, by 1950 Afrikaners would gain the

political “upper hand” and instill conservative policies that many cite as leading to

Apartheid.

In addition to the threat of a burgeoning Afrikaner nationalism, the post-war

imperial government had other problems to contend with. By the early 20th century, the

combined effects of the Boer concentration camps and the male-dominated immigration

of earlier decades had resulted in a significant “female deficit.” This was particularly true

of Johannesburg, where the rate of unmarried men was roughly double that of unmarried

women (van‐Helten and W illiams, 1983) . The imbalance was cited by Witwatersrand

gold mining operators (economic powerhouses centered largely around Johannesburg) as

contributing to an unstable work force, and they began to lobby hard for female

immigration with the perception that marriage would anchor people to the general

vicinity of the mines (van-Helton and Williams, 1983). It was rather fortuitous, then, that

in England the women’s rights movement of the late 19th century had generated a half-

dozen or so female-operated “emigration societies” (Canot, 2013). These “matrimonial

colonization” organizations were developed to help independent women find work

abroad in the many English colonies now strewn throughout Canada, Australia, New

Zealand, and South Africa (Myers, 2001). While these societies existed prior to the

179

Anglo-Boer war, it was afterward when wealthy, pro-English South Africans saw in them an opportunity to force a cultural revolution, particularly in the recently acquired (and overwhelmingly Afrikaner) Transvaal regions. As stated rather bluntly in an anonymous letter written into and published by the newspaper The Saturday Review (Volume 94:364 emphasis added):

“Do we wish to prepare another Boer war by encouraging a predominately hostile population, or do we mean to Anglicise South Africa? Let there be no vain talk of ‘fusion of races’, at least before the twenty-first century. For the present century the new colonies will be either British or Boer according as one or other race numerically as well as politically predominates, and unless we wish to see the whole struggle repeated in the near future it is our bounden duty to people the Transvaal and Orange colony with English people of the right sort. We cannot assimilate the Boers, at present, if at all, though they may easily absorb our sparse minority in the colonies. If we cannot assimilate, we must swamp them. It is only cant to pretend that we mean anything else. The land must be so expropriated and divided that holdings can be supplied to as many industrious and capable Englishmen as can be induced to settle there. We will not rob the Boer farmers—though before the surrender terms were arranged we had ample right to confiscate their lands—but we will take care by buying up land that room is found for English farmers beside them.” Civis Britannicus, 1902

And while “Civis Britannicus” was probably unsupportive of of state-sponsored matchmaking, the reality was that homeland England had the opposite problem when it came to women in that colonial expansion had generated a dearth of single women far outnumbering both available domestic work and the eligible bachelors in need of wives

(Canot, 2013). Female emigration societies offered the government a unique opportunity to kill two birds with one stone, and they seized upon it.

180

In the decade following the war, thousands of women arrived to South Africa from

England under the auspice of finding independent employ in domestic service—this is in

addition to the regular stream of emigrants who arrived by private means (Bush, 1994).

The majority of these women were funneled directly to the Afrikaner-dominated

Transvaal region. While the emigration societies publicly touted this as a success for female independence in a patriarchal world, the hushed reality was that this sponsored emigration amounted to little more than glorified match-making within a larger social experiment designed to quell Afrikaner identity. As stated in a letter penned by Lady

Maud Shelbourne, head of one of these emigration societies, in 1906:

“I think you may take it for certain that the majority of the girls will drift into Johannesburg after a year or two’s service. From an imperial point of view I don’t think this matters in the least. The majority of men are in Johannesburg, and as one real object is to provide English wives for them, there is no harm in the girls coming up there” (Bush, 1994:390)

But it worked, and by the early 20th century, Johannesburg had “made the transition from the sprawling, overgrown boom town…into the ordered, class-bound European city of

the twentieth century” (van-Helton and Williams, 1983:37).

Some of the individuals in this study undoubtedly descended from these

intermarriages, and in fact nearly 50% of the Johannesburg sample and roughly 80% of

the Pretoria sample was born during the 10 years of most intensive female emigration or

during the following two decades (one or two additional generations) (see figure 53). The

close clustering of all South African samples makes sense in light of this broader political

context—specifically, that the intentional efforts to “Anglicize” the Transvaal region

through sponsored emigration, succeeded at increasing the representation of English

181

individuals in these two interior cities despite the early Afrikaner dominance in

establishing these populations.

However, it is also clear from the results that while the English contribution is detectable in modern South Africans, they are a distinct group. This is seen in the modern

MDS plots in which the English sample falls close to the South African samples along the x-axis but is distances along the y-axis. It is also supported by the temporal analysis which shows that when other English populations are included, the modern English sample tends to shift position, either falling near to the Medieval English samples or in a space intermediate to the Medieval English and modern South Africans. This would perhaps indicate that the Dutch genetic component and the contribution of indigenous peoples to the genome of modern South African whites “pushes” modern England into closer association with the other English samples in the temporal analysis.

There is, however, a tendency in the modern analysis for R-matrices and the length and joint MDS plots for the Pretoria sample to show the least affinity to South

Shields English of the three South African samples. This may, in part, stem from the fact that female English emigrants disproportionately settled in Cape Town (the primary port of arrival) or Johannesburg (Bush, 1994), but it may also a reflect stronger Afrikaner influence in Pretoria’s history. After another brief, and ultimately failed, Afrikaner revolt in 1914, a number of new Afrikaner political organizations would emerge. These would push hard for an “Afrikaner Nation” well into the 1970s and for “Afrikaner influence” into the present day. Even President Mandela entertained the idea of a “Volkstaat,” or independent Afrikaner region. Among the more conservative of these was the “Herstigte

(Reconstituted) Nationale Party (HNP), [whose aims include] a complete Afrikaner

182

hegemony and the relegation of English to a ‘second language’ status” (Ramutsindela,

1998:180). Many of these efforts were concentrated in Pretoria. Even today, about 82% of the faculty at the University of Pretoria speak Afrikaans (Webb, 2002), and the rate of

Afrikaans-speakers in Pretoria is roughly double that of nearby Johannesburg (2014 census). Pretoria is also home to a disproportionately larger share of white people than are Johannesburg or the Cape Town regions, and in fact contains more white people than any other city in the African continent, most of whom are Afrikaans-speaking.

Afrikaner sentiments predominate in Pretoria and the surrounding regions

(personal observations). Districts, streets, and even parks in Pretoria are often named for prominent figures in Afrikaner history—indeed, the name “Pretoria” come from Andries

Pretorius, an important figure in the post-Great Migration battle against the Zulu

(Harrison, 1981). The Voortrekker Monument (figure 54) is situated atop a small kop

(hill) just outside of Pretoria city center. This museum provides a historical retelling of the Great Trek and is the site of many important Afrikaner celebrations, music festivals, and reenactments. It was also the endpoint of the symbolic “Second Trek,” a 1938 re- enactment of the Great Trek (McClintock, 1991).

183

184

Figure 54. The Voortrekker Monument, situated just outside Pretoria, serves as a memorial to Afrikaner culture and the Great Trek

In nearly every analysis, regardless of which variables are used, both the crania and long bones are able to detect this critical component of South African population history—the contribution of the English to the modern genome of South African whites and the close genetic affinity of South African populations to one another, regardless of whether they are coastal or interior. At no point in the modern analyses do MDS plots place England with Southern European populations despite closer geographical proximity. This changes, however, for the temporal test, when the South Shields sample shows more affinity to the

Medieval English samples, particularly those drawn from the north of England which arise from a cemetery a short distance from Sheffield (where the South Shields sample originates).

185

6.1.3.2. South Europe

Lisbon is a port city situated on the Iberian Peninsula at the mouth of the Tagus River. It

is historically large and densely populated, and in fact remains one of the largest cities in the European Union today (Solsten, 1993). It has a unique and tumultuous history compared to other late 19th/early 20th century European cities. Just prior to 1930, after

nearly two decades of sociopolitical chaos and sixteen changes of government, a military

coup came into power and subsequently relinquished control to young professor,

economist and Catholic conservative António de Oliveira Salazar, who reluctantly took

the position only on the condition he receive complete budgetary autonomy (Baklanoff,

1979). He had his work cut out for him—at the beginning of his tenure, years of political

instability had resulted in a 36% reduction to average wages in tandem with a 3000%

increase to cost of living (Solsten, 1993; Cardoso, 2005). Salazar would keep his seat for

nearly 40 years, making his administration the longest-lived dictatorship in European

history, indeed one of the longest in history, outlasting the regimes of Mussolini, Tse-

Tung, and Stalin (Lewis, 1978). The regime would come to be known as the “Estado

Novo,” or “New State.”

It is not easy to get a sense of working class life in Lisbon as Salazar was very

private, and even today key resources from his 40-year rule remain hidden away (Ames,

1991). Efforts are further hampered by the widespread press censorship during the

Salazar years. Even the British Broadcasting Company, which maintained a presence in

Portugal during the Estado Novo, succumbed to governmental pressure by tempering

their reporting of “social instability or disturbances that took place…during the war

186

years" (Ribeiro, 2010:275). However, there is one type of data from the Estado Novo

regime that can be found in abundance: economic data. From government archives to

print news stories to historical records, data pertaining to the positive impact of Estado

Novo policies on various indicators of Portugal's economy are freely accessible and often

portray this as a time when the Portugal "crossed over the arbitrary threshold that divides

the 'less developed' from the 'more developed' countries" (Baklanoff, 1979:799). In fact,

between 1930-1960 the gross domestic product (GDP) showed modest increases and the

budget typically registered in the black (Baklanoff 1996). Furthermore, several newsprint

stories of the day praise Salazar for turning the deficit "into a surplus for the first time in

over fifteen years" while reducing unemployment and inflation rates (Teixera 2008:17-

18).

Yet these favorable gains in state finances and productive output appeared to have

done little to alleviate the poor living and working conditions of the lower classes, who

continued to struggle with poverty, over-work, and disease. This conundrum was noted in

the Winona Daily Newspaper (1968:25):

"[Salazar promised] then to bring the country out of financial chaos if given a free hand; and he did so by slashing expenses and balancing the budget. It has remained in the black ever since. ‘Only Salazar is right,’ is his philosophy. He rules a paradoxical land. While it has advanced in the past five years, Portugal still has the lowest standard of living and the highest illiteracy rate in Western Europe. Its currently is one of the world's most stable, and it has a favorable balance of payments situation and $1.2 billion in gold and foreign exchange reserves, but it is years behind neighbor Spain in economic growth.” (Flores 1968:25)

This is very different than what was seen in 20th century Bologna, where post-unification public education efforts resulted in very high literacy rates, even among the servant class

187

and in rural areas outside the city proper (Kertzer and Hogan, 1989). So while Salazar did

bring a black state ledger and working industry to Portugal, these did not naturally

translate into correspondent benefits for the working class. In the Estado Novo

government, the pro-corporate policies favored by Salazar meant that financial gains

went disproportionately to the pocketbooks of private owners, whose power was

legitimized by policy. There were no public unions, no labor organizations, not even

special interest clubs that were not explicitly approved by the Estado Novo, another

major departure from the decidedly “pro-labor” working-class of Bologna. The few

Portuguese labor unions that did exist had to adhere to criteria so restrictive they were of little benefit to most workers (de Oliveira Marques, 1972:182). The result was a system

that, despite leading to economic and productive growth, actually decreased worker's

rights, wages, and quality of life. Recent studies show that the 1930s/1940s boasted the

highest income concentration (a measure of income inequality) of any decade in the 20th

century, indicating that wealth in Portugal was unevenly concentrated in the upper class

(Alvaredo, 2008:10).

Income inequality intensified during World War II which, while impacting

income across the board, most strongly affected the already-struggling working classes

(Cardoso, 2005). The response of Salazar to the public cries for general wage increases at

this time was less than popular: "'I cannot indicate any other solution but to work and

produce more and to limit oneself to consume less," a response "translated into action by

a decree increasing hours of work and instituting a kind of family allowance (abono de

família) to be financed in part by workers' contributions" (Raby, 1988:71). This inequity

of pay did not decrease appreciably until the 1974 revolution and subsequent

188

democratization process, a time period long after that represented by the sample in this

dissertation (Alvaredo, 2008).

The late-blooming Industrial Revolution brought with it repercussions for the

working-class. Like many countries with a burgeoning industrial sector, the early 20th

century saw an influx of many people into Lisbon from the neighboring countryside. This

population boom was not met with a commensurate update to infrastructure.

Neighborhoods grew overcrowded and were comprised largely of the low-income housing necessary to support the many skilled workers living in Lisbon. Several freguesias in western and central Lisbon (districts represented by the sample in this dissertation) catered specifically to working class, port, and industrial workers. These sectors were characterized by overcrowding, poor working conditions, garbage buildup, and insufficient access to even basic utilities and sanitation (Cardoso, 2005, 2006).

The correlation between industrialization and skeletal variation is well studied.

One meta-analysis found that average stature decreased in multiple European populations

upon the arrival of industrialization. The authors conclude that "[t]he absolute gain in

income achieved by modern economic growth in the early industrial period was simply

too low to offset declines in health" (Komlos, 1998:794-5). In a recent comparison of growth velocities in Medieval and 20th century Portuguese populations, Cardoso and

Garcia (2009) found that industrial-era adolescents in Lisbon were actually shorter than

Medieval adolescents, and they point a causal finger toward child labor, which began around the age of twelve. This was a problem that was not fully addressed until after the

1950s (Cardoso and Garcia, 2009). Another study showed that boys from lower socioeconomic backgrounds in Lisbon throughout the 20th century were both lighter and

189

shorter than those from the middle and upper classes. Interestingly, this gap persisted despite the arrival of social health initiatives in the 1960s and 1970s, and were apparently little affected by the rapid economic growth seen post-1960 (Cardoso and Caninas,

2010). It is clear from multiple lines of evidence—skeletal, archival, and ethnographical—that life for the average working class person during the Estado Novo was unstable and strenuous with long working hours that began during early adolescence

(Cardoso and Garcia, 2009; Cardoso and Henderson, 2010).

The sample from Bologna is parallel to Lisbon in many ways. It is

contemporaneous (late 19th/early 20th centuries) and overlaps in economic context,

representing a time during which Bologna underwent increasing industrialization and

urbanization (though farming would remain more important here than in other regions of

Italy) (Kertzer, 1984). It also represents the working class, and the types of occupation

seen in Bologna were at least partially redundant to those seen in the Lisbon sample

(Mariotti et al., 2015). There are, however, some notable differences. While the sample

does eclipse a period of emerging industrialization, factory work was apparently less

prevalent in Bologna, where much of the occupation growth of the late 19th century was seen in the burgeoning public communications and transport sectors, particularly in the installation and maintenance of both trams and railways (Kertzer, 1984). There are also some cultural and political differences, in that the working-class Italian populations, particularly during the late 19th century, were far more apt to revolt against poor living

conditions and subpar wages and, unlike Lisbon, had the power of labor unions to levy better work conditions and opportunities to greater effect (Kertzer, 1984). In fact, some researchers argue that the roots of forthcoming socialism first spread out from the rural,

190 agrarian peoples who for years labored as braccianti and sharecroppers in the surrounding agricultural fields and who wielded considerable political sway by virtue of numbers and mobilization efforts alone (Cardoza, 1982).

Given that the Bologna and Lisbon samples reflect a wide array of occupation

(ranging from construction workers and heavy laborers to artisans, nurses and academics), it was not surprising to see a significant difference in arm robusticy when comparing the light- and heavy-intensity subsamples. However, it was somewhat surprising to find a significant difference between Lisbon and Bolognese subsamples within the light-intensity occupational category for these same arm values (table 37) and, further, to see that individuals in the Lisbon light-intensity designation had arm J values on par with those reported for the Bologna heavy-intensity workers (figure 50). At its surface, this seems to contradict expectations of the function hypothesis, which would presumably result in a closer predicted affinity between subsamples drawn from similar occupational strata. However, given what is known about the work demands under the

Salazar regime, it is likely that the increased arm robusticity seen in the Lisbon light- intensity subpopulation reflects the harsher work environment faced by the Lisbon working class in general.

Another interesting pattern to emerge is the fact that leg J values do not significantly differ for the heavy-intensity/light-intensity Bolognese subgroupings (table

36). However, in Lisbon significant differences are reported for diaphyseal regions about the knee (table 36, figure 49). Differences are even more extreme when comparing

Lisbon to Bologna, where tests show significant differences in femoral and tibial J values for every location measured except one: femur 20% for the light-intensity subsamples

191

(table 37). The fact that Lisbon individuals, heavy-intensity workers in particular, have

such robust leg bone structure compared to those of Bologna may reflect a different

aspect of functional adaptation that is not necessarily related to workload intensity but

rather to their respective urban environments. One key difference between the Lisbon and

Bologna city centers relates to geographical profile. Bologna city center is remarkably

flat (figure 55), in stark contrast to Lisbon, constructed upon seven hills, some of which

are quite steep (25-30% grade) (figure 56). Today there are several lifts and funiculars

intermittently dispersed throughout the city to help residents navigate these steep slopes

and stairways, however these were not installed until the late 1800s, and only in small

numbers with limited hours of operation. Even today it is common to see people

traversing up and down the funicular slopes on foot, preferring to walk rather than wait

for the slow lift to crawl uphill or who residing in one of the many side streets that

diverge from the main funicular track (the funicular does not make intermediate stops)

(figure 57). Another important difference between the two cities is that the construction

and operation of trams and the railway was more developed and appeared earlier in

Bologna, which saw widespread use of both in the mid-to-late 19th century. In fact, some employers even covered the 0.2 lira cost of the tram for its workers (Kertzer and Hogan,

1989). While Lisbon also had tram service, it came along later.

This point is important because terrain, particularly sloped terrain, alters the mechanical loading environment of the weight-bearing leg bones. In a large-scale study using a sample comprised of aggregate global populations, Whittey and Holt (2015) discovered that terrain slope significantly impacted midshaft shape and SMA properties of the leg bones, even after controlling for the confounding influences of subsistence

192

practice and latitude. Specifically, populations that routinely navigated steeper slopes

displayed marked anteroposterior expansion of the midshaft femur and tibia, and

increased I /I ratios. Similar results were reported by Marchi et al. (2006), who found

X Y that Neolithic and Late Upper Paleolithic populations from “hilly” regions also display

anteroposterior elongation and higher robusticity (I , I , J) values for the midshaft femur.

X Y And while arm and leg robusticity values do seem to show evidence of function-

induced plasticity, the MDS charts consistently show all four subsamples from Bologna

and Lisbon plotting with the modern Sassari (Sardinian) sample (figure 51). Furthermore,

there is no clear affinity among the two heavy-intensity subsamples as might be predicted

were function appreciably skewing genetic predictions, and R-matrix elements show the two Bologna subsamples as having the greatest affinity for one another despite arising from different occupational strata (results for the two Lisbon subsamples, while showing a positive affinity, are somewhat murkier in that the heavy_intensity subgrouping is more strongly related to Sassari) (table 38). These results suggest that while function can and does introduce plasticity to diaphyseal structures, these are not sufficient to completely mask among-popualtion genetic relationships. One possible explanation for this pattern is that the normal, day-to-day walking of sloped terrain is not sufficiently stressful to elicit a strong bone-remodeling response, meaning a default genetic pattern emerges. In one of few human strain gauge studies that incorporate terrain, Burr et al., (1996) found that the magnitude of strain acting on the tibia while walking uphill (even with a backpack) was only mildly higher than walking across level ground. Only when study participants jogged or sprinted up and downhill did strains drastically increase to a level 2-3 times than that observed in level walking. However, the sample size of their study was quite

193

small (n=2) and it was unclear what degree of incline was used (i.e., whether it was

comparable to the ~30% slope seen throughout Lisbon). Furthermore, as seen in table 35,

the leg diaphyseal rigidity scores (J) for both Lisbon subpopulations are much higher than

those reported the Bolognese samples. These patterns are consistent with expectations of terrain-induced alteration to bone rigidity, making it appear that the functional threshold was met. The fact may well be that both genetic and functional signals are detectible in long bone diaphyseal properties. Explanations as to how this may happen are presented in section 6.2

Figure 55. Aerial view showing the flat profile of Bologna city center

194

Figure 56. One of many hills upon which Lisbon is built

195

Figure 57. People traverse a steep slope common to Lisbon despite a waiting funicular

196

6.2 Genetic Predictions by Funtional Traits

Section Highlights:

1. Skeletal regions traditionally viewed as being highly plastic

(diaphyseal regions) were more variable than those believed to be

more canalized (long bone lengths and craniofacial traits).

2. While this did impact indicators of population differentiation (Q ),

ST diaphyseal properties could still detect some elements of among𝑠𝑠-

population relationships.

3. Two processes are explored that may explain how traits that are

more variable and/or more plastic could still confer genetic signals:

[1] ontogenetically programmed differences in baseline bone shape,

[2] genetically regulated differences in bone’s biomechanical

sensitivity.

6.2.1 Genetic Affinities of Plastic Traits

Overall, it appears that the low Q values and the high CV values for diaphyseal

ST variables are telling different parts of the same story: namely, that diaphyses are more

variable within populations, which generates lower predictions of among-population

differentiation observed in this analysis. This is consistent with the argument that these

regions are less canalized than are craniofacial or length elements, and/or that they are

more phenotypically plastic. However, as revealed by R-matrices, MDS plots, and

Mantel tests, these higher levels of variation do not fully obliterate the ability of diaphyseal properties to detect some among-population genetic structure. This is apparent in the clear geographic clustering on diaphyseal MDS plots not only for modern

197

populations, but also for the temporal test in which diaphyseal properties detected

regional patterning even in temporally disparate populations, particularly for the English

samples. While these two results seem contradictory, the overall message appears to be

that while diaphyseal regions are indeed more variable as a whole (resulting in low

overall differentiation), this variation is sufficiently different among-populations that

some genetic structure can be elucidated.

To understand what this means in broader context, much relevant work can be found in studies of homoiologies. Homoiologies are nonheritable traits acquired during the life of an individual, such as those that arise due to phenotypic plasticity (Lycett and

Collard, 2005; Collard and Wood, 2007, von Cramon-Taubadel, 2009). While homoiologies are often studied in the context of craniofacial variation (e.g., differences in the (phylo)genetic utility of masticatory versus nonmasticatory traits) the above descriptions apply well to the long bones, which possess highly “plastic,” mechanosensitive diaphyseal regions. Many relevant homoiology studies arise from interspecific comparisons of nonhuman primates. In a study of nonplastic (dentition) and low strain/high strain regions of the craniofacial skeleton, Wood and Lieberman (2001) discovered that the teeth—which are unique in that they do not remodel and are therefore not regarded as plastic—were no less variable than were features of the craniofacial skeleton. However, with regard to the skeletal traits, they did find that those drawn from high-strain regions were more variable than those from low-strain regions. Results from this dissertation are similar in that long bone lengths were much less variable than were diaphyseal dimensions, although prior predictions that joints would be more developmentally canalized (and therefore display constrained variation) was not

198

supported here, and in fact the joints had both higher CVs and lower Mantel test

correlations to craniofacial distance matrices than did lengths or diaphyseal traits.

Lieberman (1996) suggests that traits with higher heritabilities should be targeted for phylogenetic reconstructions because these are, presumably, more resistant to environmental perturbation. Results of this dissertation show that while it is true

diaphyseal properties were more variable than were craniofacial or length properties, they

were not appreciably more variable than were the joints (at most a percentage or two).

Furthermore, results of the Mantel tests show that diaphyseal distance matrices had a

higher correlation to craniofacial matrices than did joints on both the modern and

temporal tests, while MDS plots show that diaphyseal properties were quite capable of

detecting genetic substructure, albeit at lower magnitudes, in both R and D matrices. This

was even true with the apparent (significant) effect of function seen in diaphyseal

structure between high-intensity and low-intensity Portuguese and Bolognese

subsamples. However, it should be noted that there is an issue of evolutionary scale here

in that this dissertation focuses only relatively recent humans, while Wood and

Lieberman (2001) focus on a broader taxonomic scope that includes the extant apes,

colobus monkeys, and Australopithecus boisei. It may be that different evolutionary or

plastic processes are reflected at different taxonomic levels, in particular, some

researchers have argued that variation at higher taxonomic levels will primarily reflect

selective pressures, while variation at lower taxonomic levels (i.e., sub-specific) will

primarily reflect neutral evolutionary processes (Ackermann and Cheverud, 2000;

Ackermann, 2002).

199

However, these results are not unique. While the partitioning of traits into plastic/non-

plastic categories makes sense in light of evolutionary theory, there are a number of other

studies that likewise fail to support the hypothesis that only non-plastic/low-plastic

skeletal traits are useful to elucidate among-population/among-species relationships. In a

large (nearly 400 individual) sample of captive baboons, Roseman et al. (2010) found

that trait variance and genotype/phenotype association were relatively uniform across a

number craniofacial modules, including the masticatory module—a unit comprised of

“high strain” landmarks drawn from the alveolar and zygomaxillary regions. They conclude:

“that traits from one region of the cranium do not appear to be better for reconstructing phylogeny than any other. Because traits in one region are no more or less evolvable than any other, at least in this population, there is not any reason to expect the intrinsic variational properties of traits to affect their evolution in different ways.” (Roseman et al., 2010:11, emphasis added).

Collard and Wood (2001) found that masticatory and non-masticatory regions all performed comparably poorly at reflecting molecularly-derived phylogenetic relationships among the great apes and papionins. Lycett and Collard (2005) compared

CVs derived from various craniofacial traits in eight papionin species and found that traits from high strain (masticatory) regions were more variable than those from low-to- moderate strain regions, but that low-to-moderate traits were actually less variable than the dental traits. Furthermore, when looking at the congruence between character state matrices generated from molecularly derived phylogenies to those generated from craniofacial variables, they discovered that those traits drawn from high-strain regions

200

were more congruent than those from low- or moderate-strain regions. And similar to results from this dissertation, there were no sex-specific differences in CV patterning.

Collard and Wood (2007) extended this study to great apes and colobus monkeys.

Again they found significant differences in CV between higher strain, lower strain, and non-plastic traits, however variation in the non-plastic traits was not higher than that of

the low-plastic traits. They argue that function could explain this pattern if similarity in diet and chewing strategy varied along phylogenetic lines. In another study comparing higher-strain/lower-strain/dental traits among papionins, Collard and Lycett (2007) found that traits from higher-strain regions were less reliable at detecting interspecific

differentiation than were traits from the latter two regions, but only about 3% less

reliable. They conclude their paper with the recommendation that more traits, regardless of plasticity, are better for phylogenetic analyses.

Similar conclusions abound when the analytical scope narrows to recent human

samples. von Cramon-Taubadel (2009) compared genetic and craniofacial variation

(partitioned into masticatory and non-masticatory regions) among twelve human samples from throughout the Old World. Like results of the nonhuman primate studies referenced above, she discovered that masticatory regions were more variable, but again no less capable of detecting among-population relationships. In fact, results of the Mantel test comparing R-matrices generated from genetic data and the various craniofacial units revealed that one masticatory unit, the zygotemporal unit, was more strongly correlated to the genetic R-matrices than were two of the nonmasticatory regions. However, in a subsequent study comparing hunter-gatherer to agrarian craniomandibular morphology, von Cramon-Taubadel (2011) found that variation in mandibular and palate shape—those

201

regions likely under more intensive biomechanical loadings—reflected subsistence to the

exclusion of among-population relationships, but it was unclear whether this was due to

in vivo plastic responses or due to long-term selective pressures operating trans- generationally.

The CV patterns discovered here also mirror the results of a recent study assessing the degree of asymmetry between left and right long bones (Auerbach and Ruff,

2006). Based on correlation tests, Auerbach and Ruff (2006) find evidence that length is the most canalized, diaphyseal diameters the least, and that articular dimensions appear to fall somewhere intermediate (though, in this dissertation, articular dimensions were only slightly less variable than diaphyseal properties). However, their results differ in that they failed to find support for interdependence between length and diaphyseal properties within individuals. They therefore suggest each region is a developmental module with an independent growth trajectory and unique ability to respond to activity. While this dissertation did not assess within-individual correlations, the consistently high values reported for the length and diaphyseal Mantel tests in both the modern and temporal analyses suggest that these two skeletal regions are highly correlated for these samples—

perhaps even more so than either is to articular dimensions. While this is not an

ontogenetic study and therefore cannot discern whether this correlation is due to similar

ontogenetic growth trajectories, it does not appear that length and diaphyseal elements

are completely independent modules given their moderately-to-high correlation to

craniofacial distance matrices, to each other, and their ability to discern among-

population genetic relationships. However, given the higher diaphyseal CV and lower

QST values, it does appear that, as proposed by Auerbach and Ruff (2006), different long

202

bone properties (i.e., lengths, diaphyses) do have different susceptibilities to the effects of physical activity.

6.2.2 How Can Plastic Traits Confer Genetic Structure?

Based on what is known about ontogenetic and in vivo long bone biology, it seems likely that there are two possible mechanisms by which traits that are “highly plastic” and

“highly variable” could still detect some genetic information: [1] that morphological differences arise during ontogeny to generate “baseline” bone shapes that vary by population, and that subsequent in vivo plastic effects cannot override these defaults

(figure 58), or [2] that interpopulation differences in genetically or hormonally regulated remodeling thresholds cause some populations to be naturally “less plastic” than others.

Therefore, two populations enduring commensurate levels of mechanical stress might have a different magnitude of bone remolding response (figure 59).

Work in molecular genetics has identified several mutations that can alter diaphyseal robusticity. These include mutations in the FGF gene family, which leads to overgrowth of long bones (Colvin et al., 1996), and mutations to Abelson murine leukemia viral oncogene (ABL) alleles which generate significantly reduced cortical bone, reduced rates of bone remodeling, and increased likelihood for developing osteoporosis (Li et al., 2000). Mutations in the sclerostin (SOST) gene family generate thicker long bone cortices (Balemans et al., 2001), and myostatin (MSTN) mutations modify cortical bone structure, making long bones appear to have belonged to highly active individual when, in fact, the robust structure is simply a byproduct of myostatin deficiency.

203

204

205

Traditional diaphyseal studies have also detected some among-population structure in very young individuals who are not yet mobile, which may be evidence of a developmentally or genetically regulated “baseline” bone shape (Cowgill, 2008, 2010).

For example, Cowgill (2010) discovered significant interpopulation differences in

humeral and femoral midshaft diaphyseal properties across subadults at different ages in

temporally distinct Holocene and Pleistocene populations, with differences in cortical

area and J being detectible in infants under one year of age (likely prior to ambulation).

Similarly, Wescott (2005:6) discovered that despite high levels of within-population

variation of the proximal femur, plasticity was “not sufficient to erase the underlying population differences” that he observed, at least in broad population groupings. Like

Cowgill, subsequent research showed that these among-population differences in femoral

morphology were detectible by middle childhood, indicating they were established prior

to maturity (Wescott, 2006), a pattern that is consistent with some degree of intrinsic

regulation, perhaps canalization. This is further supported by emergent research which

confirms that many aspects of Neanderthal robusticity, including long bone robusticity,

are present at birth and are the more likely the result of different intrinsic (genetic or

developmental) regulatory mechanisms than the result of intense in vivo physical

activities in which Neanderthals undoubtedly engaged (Weaver et al., 2016).

There are also many examples of ways in which simple genetic mutations can

affect bone biomechanical sensitivity in vivo. For example, the same MSTN mutation

that impacts long bone robusticity in affected individuals also stimulates an exaggerated

remodeling response to fracture repair (Lorentzon et al., 2005; Hamrick et al., 2006;

Elkasrawy and Hamrick, 2010). Filvaroff et al. (1999) found that mutations to genes

206

expressing TGF-β significantly slowed bone resorption and generated long bones that

were stronger than would be predicted from their current loading environments. Lau et

al. (2013) discovered that silencing genes associated with IGF-1 expression can prevent

bone modeling altogether, even when individuals engage in physical activity.

The experimental studies referenced above are imperfect in that most over-rely upon mice and rats which have much higher metabolism, long bone development that extends beyond sexual maturity, retention of inactive growth plate, and which do not have osteonal remodeling (rather all remodeling is localized to the periphery of blood vasculature where osteoclasts and osteoblasts work in coordination) (Villemure and

Stokes, 2009; Jilka, 2013). However, they do show that genes are capable, at least to some degree, of tempering or intensifying the remodeling process and generating

morphological changes to diaphyseal properties at early ontogenetic stages.

Other researchers have hit upon the topic of “variation in plasticity” as well. In

particular, Roseman et al. (2010) presented several hypothetical scenarios in which morphologically derived phylogenetic reconstructions can be misleading due to natural, interpopulation differences in variation. Among these hypotheses, they include a plasticity-focused argument:

“If there is genetic variance in the extent to which some individuals form more bone than others in response to [environmental change] through genotype by environment interactions (keeping the variance of the factor the same), then both the mean and the variance of a trait changes as a result of a plastic response in the form of a shift in the mean and the unmasking of hitherto hidden genetic variation.” (Roseman et al. 2010:4)

They also point out that individuals may vary in response to environmental pressures,

207

thereby increasing trait variation without necessarily shifting the population mean

(assuming there are individuals at each extreme of the response—“overdevelopers” and

“underdevelopers” in the context of bone functional adaptation and of their example).

This is problematic because those individuals at the extremes may actually fall in the

“normal” range for another species/population, thereby muddling (phylo)genetic reconstructions.

Parsing out which model is more likely is not an easy task, and speaks to conundrums already encountered in previous studies. For example, the von Cramon-

Taubadel (2011) study on craniomandibular variation speaks largely to these two models—was the erasure of neutral evolutionary signal in the mandible the result of in vivo plasticity (i.e., the result of hyperplasticity) or did it stem from the alteration of genetic regulatory mechanisms via long-term selective pressures for reduced mandibular size (i.e., a change to baseline morphology)? Future research will need to address this ambiguity before the full scope of function-induced long bone variation can be fully appreciated. In particular, studies that specifically assess patterns of natural, among- population human variation in genes previous linked to BFA would be of great benefit in understanding long bone plasticity and the degree to which it might impact among- population genetic reconstructions.

6.3 Project Limitations

As mentioned throughout the Discussion, one possible limitation in the structure of this project is that the behavioral differences among the samples, particularly for the temporal test, were not great enough to properly test for bone functional adaptation. In a global

208

study investigating interpopulation craniofacial variation, Betti et al. (2010) found that

only in extreme (e.g., arctic) environments do selective pressures grow strong enough to

override default among-population genetic signals. While these samples are

geographically and temporally dispersed, the range of behavioral variation may have

been insufficient to stimulate BFA. A similar point is articulated by Roseman (2010) who

investigated whether high-strain (masticatory) regions showed increased variation in

skeletal traits. While he found no evidence for this, he suggested that his focus on captive

populations (all of which were fed the same diet) was too narrow, and that perhaps wild

populations, who consume more variable foodstuffs, would have provided a more robust

test as different levels of plastic responses across individuals would generate higher

variance.

While this project tried to account for this threshold issue with the inclusion of the

temporal test, the fact remains that “remodeling thresholds” are poorly understood, and

those data which do exist are drawn largely from unnatural settings (i.e., experimental

manipulations) and animal studies. If the behavioral differences among Medieval and

modern samples were insufficient to cross the remodeling threshold, it is possible that

genetic patterning observed here was simply because the Medieval biomechanical

environments were simply not “different enough” from those of the later industrial

populations. However, it is worth nothing that results of the occupation test suggest that

differences in biomechanical environment between heavy- and light-intensity occupational groups were sufficient to generate a significant functional signal in the diaphyses of contemporaneous, modern Southern European samples and that this did not significantly alter predicted relationships.

209

Another limitation is that this study does not explicitly test for differences that arise

ontogenetically. This is especially important given research showing that the skeletal

form of children, human children especially, consists of integrated units that are quite

flexible (Ackermann, 2005). Ackerman (2005) also found that adult humans and

chimpanzees have similar levels of integration, suggesting that adult form may be

achieve via number of different tracks:

“The fact that humans are most similar to chimpanzees in covariance structure during adulthood emphasizes that our understanding of adult variation—and particularly its link to evolutionary change—remains incomplete without a deeper understanding of variation throughout ontogeny.” (Ackermann 2005:195)

If it is also true that limb bone structure in human children is flexible, then it is possible

that the differences detected in adults here might stem more from unique aspects of

geographical location or from early-life cultural influences than from genetic or

developmental limitations. It should be noted, however, that the anthropological literature

has seen an increase in the number of ontogenetically-focused skeletal studies of late, so

this is likely to be an area of active and exciting research in the upcoming years (Ruff,

2003; Cowgill, 2008, 2010; Ruff et al., 2013).

6.4 Chapter 6 References

Ackermann R. 2002. Patterns of covariation in the hominoid craniofacial skeleton: implications for paleoanthropological models. Journal of Human Evolution 43:167–187.

Ackermann R. 2005. Ontogenetic integration of the hominoid face. Journal of Human Evolution 48:175–197.

210

Agostini GM, and Ross AH. 2011. The effect of weight on the femur: A cross‐sectional analysis. Journal of Forensic Sciences 56:339–343.

Agostini GM. 2009. The relationship between body mass index and long bone morphology. MA Thesis: North Carolina State University, Raleigh.

Alvaredo F. 2008. Top incomes and earnings in Portugal 1936-2004. Paris: Sciences de l’Homme et de la Societe.

Ames P. 1991 (May 1). Dictatorship's files keep secrets 17 years after revolution. Associated Press.

Auerbach BM, and Ruff CB. 2006. Limb bone bilateral asymmetry: variability and commonality among modern humans. Journal of Human Evolution 50:203–218.

Baklanoff EN. 1979. The political economy of Portugal’s old regime: Growth and change preceding the 1974 revolution. World Development 7:799–811.

Balemans W, Ebeling M, Patel N, Hul VE, Olson P, Dioszegi M, Lacza C, Wuyts W, Ende VJ, Willems P, et al. 2001. Increased bone density in sclerosteosis is due to the deficiency of a novel secreted protein (SOST). Human Molecular Genetics 10:537–43.

Betti L, Balloux F, Hanihara T, and Manica A. 2010. The relative role of drift and selection in shaping the human skull. American Journal of Physical Anthropology 141:76–82.

Betti L, Cramon-Taubadel N, and Lycett SJ. 2012. Human pelvis and long bones reveal differential preservation of ancient population history and migration out of Africa. Human Biology 85:139–152.

Bowden R, MacFie TS, Myers S, Hellenthal G, Nerrienet E, Bontrop RE, Freeman C, Donnelly P, and Mundy NI. 2012. Genomic tools for evolution and conservation in the chimpanzee: Pan troglodytes ellioti is a genetically distinct population. PLoS Genetics 8:e1002504.

211

Brink PA, Steyn LT, Coetzee GA, and van der Westhuyzen DR. 1987. Familial hypercholesterolemia in South African Afrikaners. PvuII and StuI DNA polymorphisms in the LDL-receptor gene consistent with a predominating founder gene effect. Human Genetics 77:32–35.

Brittanicus, C. 1902 (September 20). The need of women colonists in South Africa.The Saturday Review, Pp 364

Burr DB, Milgrom C, Fyhrie D, Forwood M, and Nyska M. 1996. In vivo measurement of human tibial strains during vigorous activity. Bone 18:405–410.

Bush J. 1994. “The right sort of woman”: female emigrators and emigration to the British Empire, 1890-1910. Women’s History Review 3:385–409.

Canot C. 2013. The undesirable spinster: the organized emigration of British single women, 1851-1914. Literature.

Cardini A, Elton S. 2008. Does the skull carry a phylogenetic signal? Evolution and modularity in the guenons. Biological Journal of the Linnean Society 93:813-834.

Cardoso HF, and Caninas M. 2010. Secular trends in social class differences of height, weight and BMI of boys from two schools in Lisbon, Portugal (1910–2000). Economics & Human Biology 8:111–120.

Cardoso HF, and Garcia S. 2009. The Not‐so‐ Dark A ges: Ecology for human growth in medieval and early Twentieth Century Portugal as inferred from skeletal growth profiles. American Journal of Physical Anthropology 138:136–147.

Cardoza AL. 1982. Agrarian elites and Italian fascism: The province of Bologna, 1901- 1026. New Haven: Princeton University Press.

Cheverud JM. 1982. Phenotypic, Genetic, and Environmental Morphological Integration in the Cranium. Evolution 36:499–516.

212

Cheverud JM. 1989. A comparative analysis of morphological variation patterns in the papionins. Evolution 43:1737–1747.

Collard M, and Lycett SJ. 2007. Does phenotypic plasticity confound attempts to identify hominin fossil species? Folia Primatologica 79:111–122.

Collard M, and Wood B. 2001. Homoplasy and the early hominid masticatory system: inferences from analyses of extant hominoids and papionins. Journal of human evolution 41:167–94.

Colvin JS, Bohne BA, Harding GW, McEwen DG, and Ornitz DM. 1996. Skeletal overgrowth and deafness in mice lacking fibroblast growth factor receptor 3. Nature Genetics 12:390–397.

Cowgill L. 2008. The ontogeny of recent and Late Pleistocene human postcranial robusticity. PhD. Dissertation:Washington University, St. Louis.

Cowgill LW. 2010. The ontogeny of Holocene and Late Pleistocene human postcranial strength. American Journal of Physical Anthropology 141:16–37.

Elkasrawy MN, and Hamrick MW. 2010. Myostatin (GDF-8) as a key factor linking muscle mass and bone structure. Journal of Musculoskeletal and Neuronal Interactions 10:56–63.

Emons J, Chagin AS, Sävendahl L, Karperien M, and Wit JM. 2011. Mechanisms of Growth Plate Maturation and Epiphyseal Fusion. Hormone Research in Paediatrics 75:383–391.

Filvaroff E, Erlebacher A, Ye J, Gitelman S, Lotz J, Heillman M, and Derynck R. 1999. Inhibition of TGF-beta receptor signaling in osteoblasts leads to decreased bone remodeling and increased trabecular bone mass. Development 126:4267–79.

Flores IM. 1968. Near 80, Salazar wants to continue to serve. Winona Daily News, Winoa, Minnesota. Pp 25.

Formicola V, and Giannecchini M. 1999. Evolutionary trends of stature in Upper Paleolithic and Mesolithic Europe. Journal of Human Evolution 36:319–333.

Fulwood EL, and Kramer A. 2013. Brief communication: Effect of size biases in the coefficient of variation on assessing intraspecific variability in the prosimian skeleton. American Journal of Physical Anthropology 152:151–155.

213

Giliomee H. 2003. The Afrikaners: Biography of a people. London: C. Hurst and Company Publishers Limited.

Goldman A, Krause A, Ramsay M, and Jenkins T. 1996. Founder effect and prevalence of myotonic dystrophy in South Africans: molecular studies. American journal of human genetics 59:445–52.

González‐José R, Neves W, Lahr MM, S, H, M H, and G. 2005. Late Pleistocene/Holocene craniofacial morphology in Mesoamerican Paleoindians: implications for the peopling of the New World. American Journal of Physical Anthropology 128:772–780.

Gouws NB. 1981. Bevolkingstendense van Blankes in Suid-Afrika voor 1820. Johannesburg: Randse Afrikaanse Univeriteit.

Greeff JM. 2007. Deconstructing Jaco: Genetic Heritage of an Afrikaner. Annual Review of Genomics and Human Genetics 71:674–688.

Groenewald JZ, Liebenberg J, I M, and Warnich L. 1998. Linkage disequilibrium analysis in a recently founded population: evaluation of the variegate porphyria founder in South African Afrikaners. American Journal of Human Genetics 62:1254–1258.

Grundlingh A. 1999. The bitter legacy of the Boer War. History Today 49:21–25.

Hamrick MW. 2001. Development and evolution of the mammalian limb: adaptive diversification of nails, hooves, and claws. Evolution and development 3:355– 363.

Hamrick MW. 2007. Evolvability, Limb Morphology, and Primate Origins. In: Ravosa MJ, Dagosto M, editors. New York: Springer. p 381–401.

Hamrick MW, Samaddar T, Pennington C, and McCormick J. 2006. Increased muscle mass with myostatin deficiency improves gains in bone strength with exercise. Journal of Bone and Mineral Research 21:477–483.

Hanihara T, and Ishida H. 2005. Metric dental variation of major human populations. American Journal of Physical Anthropology 128:287–298.

Harrison D. 1981. The white tribe of Africa: South Africa in perspective. Berkeley and Los Angeles: University of California Press.

Harvati K, Weaver TD, and Hublin J. 2006. Reliability of cranial morphology in reconstructing Neanderthal phylogeny. In: K, Harrison T, editors. Netherlands: Springer. p 239–254.

214

Hayden MR, MacGregor JM, and Beighton PH. 1980. The prevalence of Huntington’s chorea in South Africa. South African Medical Journal 58:193–6.

Heinrich AR, Schrire C, and C. 2014. Faunal analysis and the development of the meat industry at the VOC Cape in the 17th and 18th centuries. In: Walnut Creek: Left Coast Press. p 65–100.

Heyningen E. 2009. The Concentration Camps of the South African (Anglo‐Boer) W ar, 1900–1902. History Compass 7:22–43.

Hinchliffe JR. 1991. Developmental approaches to the problem of transformation of limb structure in evolution. In: J, Hurle JM, Summerbell D, editors. New York: Springer. p 313–323.

Holliday TW. 1997. Body proportions in Late Pleistocene Europe and modern human origins. Journal of Human Evolution 32:423–447.

Jilka RL. 2013. The Relevance of Mouse Models for Investigating Age-Related Bone Loss in Humans. The Journals of Gerontology Series A: Biological Sciences and Medical Sciences 68:1209–1217.

Karayiorgou M, Torrington M, Abecasis GR, Pretorius H, Robertson B, Kaliski S, Lay S, Sobin C, Möller N, Lundy LS, et al. 2004. Phenotypic characterization and genealogical tracing in an Afrikaner schizophrenia database. American Journal of Medical Genetics Part B: Neuropsychiatric Genetics 124B:20–28.

Katz EN. 1995. Outcrop and deep level mining in South Africa before the Anglo-Boer War: Re-examining the Blainey thesis. Economic History Review 48:304–328.

Keegan T. 1996. Colonial South Africa and the origins of the racial order. Charlottesville, Virginia: University Press of Virginia.

Kertzer D. 1984. Family life in Central Italy, 1880-1910: Sharecropping, wage labor, and coresidence. New Brunswick, New Jersey: Rutgers University Press.

Kertzer DI, and Hogan DP. 1989. Family, political economy, and demographic change: The transformation of life in Casalecchio, Italy, 1861–1921 . Madison, Wisconsin: University of Wisconsin Press.

Komlos J. 1998. Shrinking in a growing economy? The mystery of physical stature during the industrial revolution. Journal of Economic History 58:779–802.

Krebs PM. 1992. “The Last of the Gentlemen”s Wars’: Women in the Boer War concentration camp controversy. History Workshop Journal 33:38–54.

215

Lau K-H, Baylink DJ, Zhou X-D, Rodriguez D, Bonewald LF, Li Z, Ruffoni D, Müller R, Kesavan C, and Sheng M. 2013. Osteocyte-derived insulin-like growth factor I is essential for determining bone mechanosensitivity. American Journal of Physiology Endocrinology and Metabolism 305:E271–E281.

Lewis PH. 1978. Salazar’s ministerial elite, 1932-1968. Journal of Politics 40:622–647.

Li J, Sarosi I, Yan XQ, Morony S, C, Tan H, McCabe S, Elliott R, Scully S, Van G, et al. 2000. RANK is the intrinsic hematopoietic cell surface receptor that controls osteoclastogenesis and regulation of bone mass and calcium metabolism. Proceedings of the National Academy of Science 97:1566–1571.

Lieberman D. 1996. How and why humans grow thin skulls: experimental evidence for systemic cortical robusticity.

Lieberman D, Pearson O, Mowbray K. 2000. Basicranial influence on overall cranial shape. Journal of Human Evolution 38:291-315.

Lorentzon M, Swanson C, Eriksson A-L, Mellström D, and Ohlsson C. 2005. Polymorphisms in the aromatase gene predict areal BMD as a result of affected cortical bone size: the GOOD study. Journal of Bone and Mineral Research 21:332–9.

MacKinnon AS. 2004. The Making of South Africa: Culture and Politics. Upper Saddle River: Pearson Prentice Hall.

Manica A, Amos W, Balloux F, and Hanihara T. 2007. The effect of ancient population bottlenecks on human phenotypic variation. Nature 448:346–348.

Martínez‐A badías N , Esparza M , Sjøvold T, González‐ José R, Santos M , and Hernández M. 2009. Heritability of human cranial dimensions: comparing the evolvability of different cranial regions. Journal of Anatomy 214:19–35.

216

McClintock A. 1991. “No longer in a future heaven”: Women and nationalism in South Africa. Transition 51:104–123.

Moore MK. 2008. Body mass estimation from the human skeleton. PhD Dissertation: University of Tennessee, Knoxville.

Murray C. 1992. Black Mountain: Land, class, and power in the Eastern Orange Free State. Edinburg: Edinburg University Press.

Muthien YG, and Khosa MM. 1995. The kingdom, the Volkstaat and the New South Africa: Drawing South Africa’s new regional boundaries. Journal of Southern African Studies 21:303–322.

Myers JC. 2001. Performing the voyage out: Victorian female emigration and the class dynamics of displacement. Victorian Literature and Culture 29:129–146.

De Oliveira Marques AH. 1972. History of Portugal. Volume II: From empire to corporate state. New York: Columbia University Press.

O’Neill MC, and Dobson SD. 2008. The degree and pattern of phylogenetic signal in primate long-bone structure. Journal of Human Evolution 54:309–322.

Patterson S. 1957. The last trek: A study of the Boer people and the Afrikaner Nation. London: Routledge and Kegan Paul Ltd.

Phimister I. 1993. Unscrambling the scramble for Southern Africa: The Jameson Raid and the South African War revisited. South African Historical Journal 28:203– 220.

Powell JF, and Neves WA. 1999. Craniofacial morphology of the first Americans: pattern and process in the peopling of the New World. Yearbook of Physical Anthropology 42:153–188.

Raby DL. 1988. Fascism and resistance in Portugal: Communists, liberals and military dissidents in the opposition to Salazar, 1941-1974. Manchester: Manchester University Press.

Ramutsindela MF. 1998. Afrikaner nationalism, electioneering and the politics of a Volkstaat. Politics 18:179–188.

Relethford JH. 1994. Craniometric variation among modern human populations. American Journal of Physical Anthropology 95:53–62.

Relethford JH. 1998. Genetics of modern human origins and diversity. Annual Review of Anthropology 27:1–23.

217

Relethford JH. 2001. Global Analysis of Regional Differences in Craniometric Diversity and Population Substructure. Human Biology 73:629–636.

Relethford JH. 2002. Apportionment of global human genetic diversity based on craniometrics and skin color. American journal of physical anthropology 118:393–8.

Relethford JH. 2004. Global Patterns of Isolation by Distance Based on Genetic and Morphological Data. Human Biology 76:499–513.

Relethford JH. 2009. Race and global patterns of phenotypic variation. American Journal of Physical Anthropology 139:16–22.

Relethford JH. 2010. Population‐specific deviations of global human crani ometric variation from a neutral model. American Journal of Physical Anthropology 142:105–111.

Relethford JH, Crawford MH, and Blangero J. 1997. Genetic drift and gene flow in post- famine Ireland. Human Biology 69:443–465.

Relethford JH, and Jorde LB. 1999. Genetic evidence for larger African population size during recent human evolution. American Journal of Physical Anthropology 108:251–60.

Retief P. 1837 (February 2). Manifesto. The Grahamstown Journal.

Ribeiro N. 2010. Salazar’s influence in the BBC Portuguese service during World War II. Journalism Studies 11:257–269.

Ridpath JC, and Ellis ES. 1899. The story of South Africa: an account of the historical transformation of the Dark Continent by the European powers and the culminating context between Great Britain and the South African Republic in the Transvaal War. Chicago: C.W.Stanton Company.

Roseman CC, Willmore K, Rogers J, Hildebolt C, Sadler B, Richtsmeier J, and Cheverud J. 2010. Genetic and environmental contributions to variation in baboon cranial morphology. American Journal of Physical Anthropology 143:1–12.

Roseman CC, and Weaver TD. 2004. Multivariate apportionment of global human craniometric diversity. American Journal of Physical Anthropology 125:257–263.

218

Ross AH. 2004. Regional isolation in the Balkan region: An analysis of craniofacial variation. American Journal of Physical Anthropology 124:73–80.

Ruff CB. 2003. Ontogenetic adaptation to bipedalism: age changes in femoral to humeral length and strength proportions in humans, with a comparison to baboons. Journal of Human Evolution 45:317–349.

Ruff CB, Burgess LM, Bromage TG, Mudakikwa A, and McFarlin SC. 2013. Ontogenetic changes in limb bone structural proportions in mountain gorillas (Gorilla beringei beringei). Journal of Human Evolution 65:693–703.

Ruff CB. 2000. Body size, body shape, and long bone strength in modern humans. Journal of Human Evolution 38:269–290.

Saux O, Beck K, Sachsinger C, Treiber C, Göring H, Curry K, Johnson E, Bercovitch L, Marais A-S, Terry S, et al. 2002. Evidence for a founder effect for pseudoxanthoma elasticum in the Afrikaner population of South Africa. Human Genetics 111:331–338.

Schneider KN, and Blakeslee DJ. 1990. Evaluating residence patterns among prehistoric populations: clues from dental enamel composition. Human Biology 62:71–83.

Silventoinen K. 2003. Determinants of variation in adult body height. Journal of Biosocial Science 35:263–285.

Smith HF. 2009. Which cranial regions reflect molecular distances reliably in humans? Evidence from three-dimensional morphology. American Journal of Human Biology 21:36–47.

Solsten E. 1993. Portugal: A country study. Washington DC: GPO for Library of Congress.

Steckel R. 1995. Stature and the Standard of Living. Journal of Economic Literature 33:1903–1940.

Teixeira M. 2008. Caught on the periphery: Portuguese neutrality during World War II and Anglo-American negotiations with Salazar. Honors Thesis:University of Pennsylvania

Tipping AJ, Pearson T, Morgan NV, Gibson RA, Kuyt LP, Havenga C, Gluckman E, Joenje H, de Ravel T, Jansen S, et al. 2001. Molecular and genealogical evidence for a founder effect in Fanconi anemia families of the Afrikaner population of South Africa. Proceedings of the National Academy of Sciences 98:5734–5739.

van der Merwe PJ. 1994. The migrant farmer in the history of the Cape Colony, 1657- 1842. Athens: Ohio University Press.

219

van‐Helten JL, and W illiams K. 1983. “The crying need of South Africa”: the emigration of single British women to the Transvaal, 1901–10. Journal of Southern African Studies 10:17–38.

Vestergaard M. 2001. Who’s got the map? The negotiation of Afrikaner identities in post-apartheid South Africa. Daedulus 130:19–44.

Villemure I, and Stokes IA. 2009. Growth plate mechanics and mechanobiology. A survey of present understanding. Journal of biomechanics 42:1793–803.

Vincent L. 2000. Bread and honour: White working class women and Afrikaner nationalism in the 1930s. Journal of Southern African Studies 26:61–78. von Cramon-Taubadel N, and Lycett SJ. 2008. Brief communication: human cranial variation fits iterative founder effect model with African origin. American journal of physical anthropology 136:108–13.

von Cramon-Taubadel N, and Smith H. 2012. The relative congruence of cranial and genetic estimates of hominoid taxon relationships: implications for the reconstruction of hominin phylogeny. Journal of human evolution 62:640–53.

von Cramon-Taubadel N, Stock JT, and Pinhasi R. 2013. Skull and limb morphology differentially track population history and environmental factors in the transition to agriculture in Europe. Proceedings of the Royal Society of London B: Biological Sciences 280:20131337.

von Cramon-Taubadel N. 2009. Revisiting the homoiology hypothesis: the impact of phenotypic plasticity on the reconstruction of human population history from craniometric data. Journal of Human Evolution 57:179–190.

von Cramon-Taubadel N. 2011. Global human mandibular variation reflects differences in agricultural and hunter-gatherer subsistence strategies. Proceedings of the National Academy of Sciences 108:19546–19551. von Cramon‐Taubadel N, and W eaver TD. 2009. Insights from a quantitative genetic approach to human morphological evolution. Evolutionary Anthropology 18:237– 240.

Weaver TD, Coqueugniot H, Golovanova LV, Doronichev VB, Maureille B, and Hublin J-J. 2016. Neonatal postcrania from Mezmaiskaya, Russia, and Le Moustier, France, and the development of Neandertal body form. Proceedings of the National Academy of Sciences 113:6472–6477.

Webb V. 2002. English as a second language in Sotuh Africa's tertiary institutions: A case study at the University of Pretoria. World Englishes 21(1):49-61.

220

Webb V, and Kriel M. 2000. Afrikaans and Afrikaner nationalism. International Journal of the Sociology of Language 144:19–49.

Wescott DJ. 2005. Population variation in femur subtrochanteric shape. Journal of Forensic Science 50:1–7.

Wescott DJ. 2006. Effect of mobility on femur midshaft external shape and robusticity. American Journal of Physical Anthropology 130:201–213.

Whittey E, Holt B. 2015. Effects of terrain on reconstructions of mobility patterns. Paper presented at the annual meeting of the American Association of Physical Anthropologists, Atlanta, GA.

Wood B, and Lieberman DE. 2001. Craniodental variation in Paranthropus boisei: A developmental and functional perspective. American Journal of Physical Anthropology 116:13–25.

221

CHAPTER 7

CONCLUSION

7.1 Concluding Remarks

Perhaps one of the strongest messages to come out of this research is that even if among- population differentiation is low for diaphyseal regions, all three long bone properties

(length, articular dimensions, cross-sectional properties) showed a capability to detect among-population genetic structure in a relatively limited sample of modern and

Medieval populations of European ancesty. This was true even when functional signals were present. These results further support those of other studies which confirm that even

“plastic” skeletal traits, like diaphyseal properties, have some genetic utility. Future analyses that target long bone structural variation to interpret behavioral (functional) signals may be strengthened by controlling for these baseline genetic relationships, particularly if they employ an approach that relies upon broad, interpopulation or interspecific comparisons.

7.2 Future Directions

Future goals for this research include expanding the scope of the project to include

indigenous and admixed South African populations as a more robust test of the isolation-

by-distance model, additional samples from a new geographical region (the United

States), an additional time period for the English samples (8th-12th century Saxon), and an additional Italian Medieval sample from Northern Italy to try and parse out the somewhat erratic behavior of the Tuscan Medieval Italian samples used in this dissertation. It would

also be ideal to include a modern and 18th/19th century sample from the Netherlands in an

222

attempt to capture the Dutch influence in the modern South African samples, though such

collections have yet to be identified.

Other future goals include the division of larger samples into two or three

different time periods to allow for an investigation of “generational changes,” much like

Relethford et al. (1997) who looked at differences in unbiased Q values across three

ST subsequent generations of Irish populations during a period of mass migration, and

similar to classic work by Boas (1910) who looked at phenotypic plasticity in head

measurements in subsequent generations of mainland and immigrant populations.

Another goal is to delve more deeply into the anatomical patterning of both

Coefficients of Variation and population distance predictions to identify and formalize a

core set of long bone properties that are most reliable for elucidating among-population

genetic structure. Furthermore, the author intends to broaden the variables to include

among-population assessments of brachial and crural indices, as significant differences in proportion have been reported in studies of both adults and fetuses (Holliday and Ruff

2001).

Finally, extension of this work into assessments of interspecific variation among

nonhuman primate taxa, particularly the hominoids, is intended, as understanding within-

species variation patterns in long bone structural properties will be critical should they

ever be used to speak to the taxonomic affinities of existing and new fossil hominin

discoveries.

7.3 Chapter 7 References

Boas F. 1910. Report presented to the 61st congress on changes in bodily form of descendants of immigrants. Government Printing Office:Washington, DC.

223

Holliday TW, Ruff CB. 2001. Relative variation in human proximal and distal limb segment lengths. American Journal of Physical Anthropology 116(1):26-33

Relethford JH, Crawford MH, and Blangero J. 1997. Genetic drift and gene flow in post- famine Ireland. Human Biology 69:443–465.

Roseman CC, Weaver TD. 2007. Molecules versus morphology? Not for the human cranium. BioEssays 29:1185-1188.

224

BIBLIOGRAPHY

Abràmoff MD, and Magalhães PJ. 2004. Image processing with ImageJ. Biophotonics International 11:36–42.

Ackermann R. 2002. Patterns of covariation in the hominoid craniofacial skeleton: implications for paleoanthropological models. Journal of Human Evolution 43:167–187.

Ackermann R. 2005. Ontogenetic integration of the hominoid face. Journal of Human Evolution 48:175–197.

Ackermann RR, and Cheverud JM. 2000. Phenotypic covariance structure in tamarins(genus Saguinus): a comparison of variation patterns using matrix correlation and common principal component. American Journal of Physical Anthropology 111:489–501.

Adams DC, Rohlf JF, and Slice DE. 2004. Geometric morphometrics: Ten years of progress following the “revolution.” Italian Journal of Zoology 71:5–16.

Agostini GM, and Ross AH. 2011. The effect of weight on the femur: A cross‐sectional analysis. Journal of Forensic Sciences 56:339–343.

Agostini GM. 2009. The relationship between body mass index and long bone morphology. MA Thesis: North Carolina State University, Raleigh.

Alvaredo F. 2008. Top incomes and earnings in Portugal 1936-2004. Paris: Sciences de l’Homme et de la Societe.

Ames P. 1991 (May 1). Dictatorship's files keep secrets 17 years after revolution. Associated Press.

Auerbach BM, and Ruff CB. 2004. Human body mass estimation: A comparison of “morphometric” and “mechanical” methods. American Journal of Physical Anthropology 125:331–342.

225

Auerbach BM, and Ruff CB. 2006. Limb bone bilateral asymmetry: variability and commonality among modern humans. Journal of Human Evolution 50:203–218.

Baklanoff EN. 1979. The political economy of Portugal’s old regime: Growth and change preceding the 1974 revolution. World Development 7:799–811.

Ballock RT, and OʼKeefe RJ. 2003. The biology of the growth plate. The Journal of Bone and Joint Surgery (American) 85:715–726.

Barham G, and Clarke NM. 2008. Genetic regulation of embryological limb development with relation to congenital limb deformity in humans. Journal of Children’s Orthopaedics 2:1–9.

Basso N, and Heersche J. 2006. Effects of hind limb unloading and reloading on nitric oxide synthase expression and apoptosis of osteocytes and chondrocytes. Bone 39:807–14.

Berglund P, and Heeringa SG. 2014. Multiple imputation of missing data using SAS. Cary, North Carolina: SAS Institute Inc.

226

Betti L, Balloux F, Hanihara T, and Manica A. 2010. The relative role of drift and selection in shaping the human skull. American Journal of Physical Anthropology 141:76–82.

Betti L, Cramon-Taubadel N, and Lycett SJ. 2012. Human pelvis and long bones reveal differential preservation of ancient population history and migration out of Africa. Human Biology 85:139–152.

Blackburn A. 2011. Bilateral asymmetry of the humerus during growth and development. American Journal of Physical Anthropology 145:639–646.

Blumenschine R. 1988. An experimental model of the timing of hominid and carnivore influence on archaeological bone assemblages. Journal of Archaeological Science 15:483–502.

Bridges PS. 1992. Prehistoric arthritis in the Americas. Annual Review of Anthropology 21:67 –91.

227