1 THE EFFECTS OF GROWTH DISRUPTION ON ADULT SKELETAL MORPHOLOGY By ANNA ELIZABETH VICK A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2010
2 2010 Anna Elizabeth Vick
3 ACKNOWLEDGMENTS I am grateful to many colleagues, friends and family who have offered encouragement throughout this process. First, I would like to thank Dr. David Daegling for the time he has dedicated to my education. Dr. Daegling spent numerous hours reading this document and fielding questions with patience and thoroughness. I thank Dr. John Krigbaum and Dr. Mike Warren for serving on my committee and for the faith you both have in my abilities. Your kindnesses have served as great encouragement. I am grateful to Dr. Shawn Kneipp for serving as my external committee member and for p roviding a different perspective for this research. I am grateful to David Hunt at the Smithsonian and Lyman Jellema at the Cleveland Museum of Natural History for access to the collections used for this research and guidance while at those respective inst itutions. The time and expertise volunteered by Sarah Pelot and Janine Hinton were invaluable in gathering materials and operating the xray equipment at the Smithsonian. They are exceptional people. Dorothy Perelman and Carl Farona served as excellent hosts while collecting data in Cleveland and I appreciate that they provided me the opportunity to get to know the city while I was there. Next, I would like to express my gratitude to The University Womens Club and the Ruegamer Charitable Trust who prov ided financial assistance during my time at UF. I am thankful for Laura Regan for always being there for me, despite the geographical distance between us now. Laura, Calvin and Hobbes welcomed me into their home on numerous occasions and their hospitality aided this research. I am indebted to Chad Maxwell for teaching me many important life lessons regardless of whether or not I signed up for the course. While I have benefitted from Chads many skills (research, technical, ironing, etc.) over the last eight years, I am most grateful for his friendship. I owe a huge debt of gratitude to Jennifer Hotzman for showing me the ropes. Throughout the last few years, Jennifer has jumped in first and
4 enlightened me with the benefit of her experience. I owe muc h appreciation to Andy Thon for improving my quality of life over the last few years and I look forward to our future together I also would like to thank my family. My brothers, Brian and Stephen, have attended to the many needs of my family while I ha ve been away. They have great strength of character despite having such weak bones. My father, Gilbert Vick, has always had extreme faith in my abilities, even when fielding the simplest of math questions. My grandmother, Marguerite Hardee Greer, has alwa ys believed in the value of an education. I am just one of a number of students who owe their educational opportunities to her. My mother, Dr. Laura Greer Vick, was my first anthropology professor. Not only has she always provided the love, support and encouragement of a mother, but she also taught me everything I know about primates, that sometimes hitchhiking in a pig truck is necessary if you want to see the best archaeological sites and that bathing midday is the best way to avoid the crocodiles.
5 TABLE OF CONTENTS ACKNOWLEDGMENTS .................................................................................................................... 3 page LIST OF TABLES ................................................................................................................................ 7 LIST OF FIGURES ............................................................................................................................ 10 LIST OF ABBREVIATIONS ............................................................................................................ 13 ABSTRACT ........................................................................................................................................ 14 CHAPTER 1 INTRODUCTION ....................................................................................................................... 16 2 ADULT STATURE AND PROPORTIONS ............................................................................. 21 Human Growth ............................................................................................................................ 21 Adult Proportions ........................................................................................................................ 36 Secular Trends in Adult Height .................................................................................................. 38 Sexual Dimorphism ..................................................................................................................... 41 The Metabolic and Disease Implications of Growth Disruption.............................................. 46 3 SKELETAL INDICATORS OF GROWTH DISRUPTION ................................................... 49 Enamel Defects ............................................................................................................................ 49 Harris Lines ................................................................................................................................. 52 Other Skeletal Indic ators of Growth Disruption ....................................................................... 58 4 MATERIALS AND METHODS ............................................................................................... 62 Materials ...................................................................................................................................... 62 Methods ....................................................................................................................................... 66 Measurements ...................................................................................................................... 67 Linear Enamel Defects ........................................................................................................ 68 Harris Lin es .......................................................................................................................... 71 Statistical Analyses .............................................................................................................. 72 5 RESULTS: STATURE AND STRESS ..................................................................................... 80 St ature and Enamel Defects ........................................................................................................ 80 Stature and Harris Lines ............................................................................................................. 82 Stature and the Minimum Number of Stress Events ................................................................. 82 Stature and the Age at which Stress Events Occur ................................................................... 83 6 RESULTS: SEXUAL DIMORPHISM AND STRESS .......................................................... 101
6 7 RESULTS: PROPORTIONS AND STRESS ......................................................................... 106 Proportions and dental enamel defects .................................................................................... 106 Proportions and Harris lines ..................................................................................................... 107 Sc aling ........................................................................................................................................ 108 Scaling D ifferences Based on E namel D efects ................................................................ 108 Scaling D ifferences Based on Harris L ines ..................................................................... 110 Proportions and the Minimum Number of Stress Events ....................................................... 111 Proportions and the Age at which Stress Events Occur .......................................................... 112 8 RESULTS: TRADITIONAL AND NON TRADITIONAL MARKERS OF GROWTH DISRUPTION IN THE ADULT SKELETON ....................................................................... 158 Vertebral Neural Canals and Traditional Stress Markers ....................................................... 158 Vertebral Neural Canals and Stature ........................................................................................ 161 Head Circumference and Traditional Stress Markers ............................................................. 162 9 DISCUSSION ............................................................................................................................ 191 Stature ........................................................................................................................................ 191 Sexual Dimorphism ................................................................................................................... 195 Proportions ................................................................................................................................. 197 The Reliability of Stress Indicators .......................................................................................... 201 Study Limitations: Collection Effect ...................................................................................... 207 Study Limitations: Age Effect ................................................................................................. 207 10 CONCLUSIONS ....................................................................................................................... 212 LIST OF REFERENCES ................................................................................................................. 215 BIOGRAPHICAL SKETCH ........................................................................................................... 234
7 LIST OF TABLES Table page 3 1 Mean estimates for ages of enamel formation ...................................................................... 60 4 1 Description of sample population ......................................................................................... 77 4 2 Skeletal measurements recorded ........................................................................................... 77 4 3 Description of proportional indices ....................................................................................... 78 4 4 Coding dental enamel defects ................................................................................................ 78 5 1 Descriptive statistics and normality of osteometric variables ............................................. 86 5 2 ANOVA Stature estimates by dental enamel defects (CBTS) .......................................... 87 5 3 ANOVA Stature estimates by dental enamel defects (TS) ................................................ 88 5 4 ANOVA Proxies for stature by dental enamel defects (LEH) in females ....................... 89 5 5 ANOVA Stature estimates by the presence or absence of Harri s lines ............................ 90 5 6 Interaction effects of Harris lines (HL) and enamel defects (CBTS) on stature ................ 91 5 7 ANOVA Stature by minimum number of stress events (MNSE) .................................... 93 5 8 ANOVA Stature estimates by number of Harris lines ...................................................... 95 5 9 ANOVA Stature estimates by age of enamel defect ......................................................... 96 5 10 Results of resampling for male tibia length b y age of the enamel defect ........................... 97 5 11 ANOVA Stature estimates by infant enamel defect ......................................................... 98 5 12 ANOVA Stature estimates by age of Harris lines ............................................................. 99 6 1 Sexual dimorphism scores for unstressed individu als ....................................................... 104 6 2 Sexual dimorphism scores for stressed individuals ........................................................... 104 6 3 Differences in sexual dimorphism between unstressed and stressed populations a ......... 105 7 1 Pearson correlations of the components of ratios (N = 204) ............................................. 115 7 2 ANOVA Proportional indices by dental enamel defects (CBTS) .................................. 116 7 3 ANOVA Proportions by dental enamel defects (TS) ...................................................... 117
8 7 4 ANOVA Proportional indices by dental enamel defects (LEH) in females .................. 118 7 5 ANOVA Proportional indices by the presence or absence of Harris lines (HL) .......... 119 7 6 Interaction effects of Harris lines (HL) and enamel defects (CBTS) on proportions ...... 120 7 7 Scaling proportional indices to size (SKH) ........................................................................ 121 7 8 RMA regressions of log -transformed components of ratios divided by dental enamel defects (CBTS) ..................................................................................................................... 122 7 9 Comparison of RMA regressions (RMA2) divided by dental enamel defects (CBTS) .. 123 7 10 RMA regressions of log -transformed components of ratios divided by the presence or absence of Harris lines .................................................................................................... 124 7 11 Comparison of RMA regressions (RMA2) divided by Harris lines ................................. 125 7 12 ANOVA Proportional indices by minimum number of stress events (MNSE) ............ 126 7 13 ANOVA Proportional indices by number of Harris lines .............................................. 128 7 14 ANOVA Proportional indices by age of enamel defect ................................................. 129 7 15 ANOVA Proportional indices by infant enamel defect .................................................. 131 7 16 ANOVA Proportions by age of Harris line ..................................................................... 132 8 1 ANOVAs of vertebral neural canal measurements by dental enamel defects (CBTS) ... 165 8 2 Nonparametric tests of VNC measurements by enamel defects (CBTS) ......................... 168 8 3 ANOV As of vertebral neural canal measurements by Harris lines (HL) ......................... 169 8 4 Nonparametric tests of vertebral neural canal measurements by Harris lines (HL) ........ 172 8 5 ANOVAs of vertebral body height measurements by dental enamel defects (CBTS) .... 172 8 6 ANOVAs of vertebral body height measurements by Harris lines (HL) ......................... 173 8 7 Nonparametric tests of vertebral body height measurements by dental enamel defects (CBTS) .................................................................................................................................. 174 8 8 Nonparametric tests of vertebral body height measurements by Harris lines (HL) ......... 174 8 9 Comparison of RMA regressions (RMA2) divided by dental enamel defects (CBTS) .. 175 8 10 Comparison of RMA regressions (RMA2) divided by Harris lines (HL) ........................ 176
9 8 11 Correlations between vertebral neural canal measures and skeletal height (SKH) ......... 177 8 12 Correlations between head circumference and skeletal height .......................................... 177 8 13 ANOVAs of head circumference measurements by traditional stress markers ............... 178 9 1 Nonparametric correlations between age and height ......................................................... 210 9 2 ANOVA Stature estimates (age controlled) by dental enamel defects (CBTS) ............ 210
10 LIST OF FIGURES Figure page 3 1 Radiograph of a tibia with a Harris line ................................................................................ 61 4 1 Dental data collection form ................................................................................................... 78 4 2 Measuring hypoplastic teeth .................................................................................................. 79 4 3 Aging radiopaque transverse lines. T = total length of the tibia as defined by Byers (1991); S = the standard technique for measuring tibial length; D = the distance from the Harris line to the distal end of the bone. Age at formation is calculated by using the formula 1.15(T 2.33D) X 100/T. Tables in Byers (1991) allow for conversion of that percentage to an age range. ........................................................................................ 79 5 1 Boxplot of female tibia length divided by CBTS stress codes. Differences in tibia lengt h between stressed and unstressed groups were found to be significant ( p = 0.0271). This boxplot is used as an example of the distributional differences between stressed and unstressed groups. ............................................................................ 100 7 1 Scaling crural index to size .................................................................................................. 134 7 2 Scaling humerofemoral index to size .................................................................................. 134 7 3 Scaling radiohumeral index to size ..................................................................................... 135 7 4 Scaling intermembral index to size ..................................................................................... 135 7 5 Components of proportional ratios and long bone lengths scaled to skeletal height (male) .................................................................................................................................... 136 7 6 Components of proportional ratios and long bone lengths scaled to skeletal height (female) ................................................................................................................................. 137 7 7 RMA for crural indices divided by dental enamel defects (male) .................................... 138 7 10 RMA for sitting height indices divided by dental enamel defects (male) ........................ 141 7 11 RMA for intermembral indices divided by dental enamel defects (male) ........................ 142 7 12 RMA for crural indices divided by dental enamel defects (female) ................................. 143 7 13 RMA for humerofemoral indices divided by dental enamel defects (female) ................. 144 7 14 RMA for radiohumeral indices divided by dental enamel defects (female) ..................... 145 7 15 RMA for sitting height indices divided by dental enamel defects (female) ..................... 146
11 7 16 RMA for intermembral indices divided by dental enamel defects (female) .................... 147 7 17 RMA for crural indices divided by Harris lines (male) ..................................................... 148 7 18 RMA for humerofemoral indices divided by Harris lines (male) ..................................... 149 7 19 RMA for radiohumeral indices divided by Harris lines (male) ......................................... 150 7 20 RMA for sitting height indices divided by Harris lines (male) ......................................... 151 7 21 RMA for intermembral indices divided by Harris lines (male) ........................................ 152 7 22 RMA for crural indices divided by Harris lines (female) .................................................. 153 7 23 RMA for humerofemoral indices divided by Harris lines (female) .................................. 154 7 24 RMA for radiohumeral indices divided by Harris lines (female) ..................................... 155 7 25 RMA for sitting height indices divided by Harris lines (female) ...................................... 156 7 26 RMA for intermembral indices divided by Harris lines (female) ..................................... 157 8 1 Reduced major axis regressions of mediolateral vertebral neural canal diameters of the averaged thoracic vertebrae over vertebral body heights in males divided by enamel defects (CBTS). ....................................................................................................... 179 8 2 Reduced major axis regressions of anteroposterior vertebral neural canal diameters of the averaged lumbar vertebrae over vertebral body heights in males divided by enamel defects (CBTS). ....................................................................................................... 180 8 3 Reduced major axis regressions of anteroposterior vertebral neural canal diameters of L1 over vertebral body heights in females divided by enamel de fects (CBTS). ......... 181 8 4 Reduced major axis regressions of transverse vertebral neural canal diameters of the averaged lumbar vertebrae over vertebral body heights in females divided by enamel defects (CBTS). .................................................................................................................... 182 8 5 Reduced major axis regressions of mediolateral ver tebral neural canal diameters of the averaged thoracic vertebrae over vertebral body heights in males divided by Harris lines (HL). ................................................................................................................. 183 8 6 Reduced major axis regressions of mediolateral vertebral neural canal diameters of L1 over vertebral body heights in females divided by Harris lines (HL). ........................ 184 8 7 Reduced major axis regressions of transverse vertebral neural canal diameters of the averaged thoracic vertebrae over vertebral body heights in females divided by Harris lines (HL). ............................................................................................................................. 185
12 8 8 Reduced major axis regressions of mediolateral vertebral neural canal diameters of the averaged lumbar vertebrae over vertebral body heights in females divi ded by Harris lines (HL). ................................................................................................................. 186 8 9 Box plots of female skeletal height (SKH) divided by highest and lowest quartiles of vertebral neu ral canal anteroposterior diameters. .............................................................. 187 8 10 Box plots of male skeletal height (SKH) divided by highest and lowest quartiles of vertebral neural canal anteroposterior diameters. .............................................................. 188 8 11 Box plots of female femur length divided by highest and lowest quartiles of vertebral neural canal anteroposterior diameters size standardized using vertebral body heights. .................................................................................................................................. 189 8 12 Box plots of male maximum femur length divided by highest and lowest quartiles of vertebral neural canal anteroposterior diameters size standardized using vertebral body heights. ......................................................................................................................... 190 9 1 Scatter plot of STH by age in females ................................................................................ 211 9 2 Scatter plot of STH by age in males ................................................................................... 211
13 LIST OF ABBREVIATION S ANOVA Analysis of Variance AP Anteroposterior CBTS Conservative bilateral tooth stress. An individual is considered stressed using this criteria if they have bilateral defects on their anterior dentition. HL Harris lines or radiopaque transverse lines LEH Linear enamel hypoplasia. This coding scheme does not recognize hypoplastic pitting. Only linear defects are scored as hypoplasias. ML Mediolateral MNSE Minimum number of stress events. This score was an attempt to approximate the number of individual stress events responsible for a particular type of stress indicator. For Harris lines, this score was calculated using the number of Har ris lines present in a single bone. For teeth, the age ranges for multiple defects were considered. RMA Reduced major axis regression SKH Skeletal height STH Sitting height TS Tooth stress. This is the most lenient way of scoring enamel defects in this study. Any linear or pit defect is recorded as an indicator of stress, regardless of whether or not it is bilateral. VBH Vertebral body height VNC Vertebral neural canal
14 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy THE EFFECTS OF GROWTH DISRUPTION ON ADULT SKELETAL MORPHOLOGY By Anna Elizabeth Vick May 2010 Chair: David Daegling Major: An thropology Skeletal markers of growth disruption, enamel defects and radiopaque transverse lines, were compared to osteometric data to determine whether or not the stresses associated with the skeletal marker had an effect on adult stature, sexual dimorphism and proportions. Black males and females f ro m the Terry and Hamann Todd collections were used as the sample population. Previous r esearch on the process of catch up growth suggests that in all but the most extreme cases, the negative effects of growt h disruption are erased; h owever, anthropologists use trends in adult morphology to estimate the relative environmental stress on populations. Resampling statistics determined that stressed females were significantly shorter than their unstressed counterp arts when stress was determined by the presence or absence of enamel defects. Significant differences in stature were not found in males, nor were they found when Harris lines were used to determine stress. While significant differences in sexual dimorph ism were not found between stressed and unstressed groups, it is notable that stress had a visible e ffect in the female skeletal sample, but not the male. This finding is in stark contrast to m ost theories regarding sexually differentiated responses to st ress which presume that female growth is canalized while male morphology varies more in accordance with the environment. The crural index in males was the only proportional index investigated that was significantly related to
15 enamel defects. All other pr oportional differences observed were subtle. The data from this skeletal sample do not support the predicted models for how the environment affects human growth and morphology.
16 CHAPTER 1 INTRODUCTION Studies of historic and archaeological populations use stature and sexual dimorphism as indicators of health status in a population. The recurrent fluctuations in mean height that are found through the analysis of historical records can be correlated with economic o r environmental conditions during childhood (Komlos 1994; Steckel 1995, 1999; Steegmann 1985, 1991). Similarly, osteological reports on skeletal populations often include measures of stature or sexual dimorphism with paleo pathological data, providing i nformation on the health of a population (Cohen and Armelagos 1984; Goodman and Martin 2002; Steckel and Rose 2002). The theoretical basis for these studies rests on the idea that adult height is a product of both genetic potential and environmental inf luences: tall individuals may be closer to reaching their genetic potential in light of favorable environmental or economic circumstances. Changes in sexual dimorphism should also reflect physiological responses to environmental stress in that males and females do not respond equally to stressors (Stini 1985). It has been argued that growth stunting may be considered an adaptive advantage among the poor where individuals are able to maintain a normal weight for height because it takes fewer resources to feed a smaller body than a larger one (Nickens 1976; Seckler 1980, 1982). However, those who consider stunting to be a major public health concern point out that there is nothing adaptive about malnourishment and infection both known to contribute to stunting. Instead, stunting is used as an indicator for observing the more serious health effects of growth disruption such as impaired cognitive development and reduced fitness (Martorell 1989; Waterlow 1988a). This research investiga tes whether or not there is a correlation between skeletal indicators of growth disruption and adult stature and proportions. The human skeleton records episodes of
17 growth disruption in the form of developmental defects in dental enamel and radiopaque tr ansverse lines in long bones Indicators of growth arrest serve as an independent variable for comparison with skeletal measurements to determine whether or not the stress associated with observable markers has any effect on adult stature and skeletal pro portions. Research suggests that the stunting associated with growth disruption may be erased depending on the severity, timing and duration of the metabolic insult (Cameron, 2002; Tanner 1981). Growth disruptions are often only episodic and the human body experiences catch up growth once the conditions which caused the growth disruption are removed. Stature as an indicator of environmental stress can be seen in juveniles prior to the completion of growth. For example, in undernourished children, skel etal age may be retarded in comparison to chronological age. This is an indicator that a normal growth trajectory has been disrupted. However, once skeletal growth is complete, growth disruption is much harder to detect. Anthropologists use secular tren ds in adult stature to estimate the relative environmental stress on populations. If catchup is indeed complete after episodes of growth disruption, then fluctuations in stature should only occur under extreme circumstances such as chronic (as opposed t o episodic) undernourishment. The null hypothesis in this study is that no metric differences will be observed between the size and shape of individuals who have experienced growth disruption compared to those who have not, due to the corrective process of catch up growth. In addition, three alternative hypotheses will be tested. The first alternative hypothesis proposes that individuals with independent indicators of growth disruption or stress will have a smaller estimated stature than those individua ls without such indicators of stress. Studies documenting secular changes in stature claim that smaller
18 stature is an indicator of nutritional deprivation (Steckel 1995). Such studies would suggest that indicators of growth disruption should be positivel y associated with smaller stature relative to individuals without such stress markers. If support is found for this hypothesis, then evidence suggests that catch up growth following growth disruption is not always complete. The second alternative hypot hesis tested states that because female growth is believed to be canalized, or less affected by stress conditions than male growth, males and females with evidence of stress will exhibit less sexual dimorphism than those with no evidence of stress. Theoret ically, changes in sexual dimorphism should also reflect physiological responses to environmental stress. The work of Stini (1969, 1985) has shown that the longterm effects of nutritional deprivation and other environmental stressors have a greater effec t on males than females. Females, probably as an adaptation to the demands of lactation and gestation, store fat and nutrient reserves which are beneficial during periods of nutritional stress (Stini 1982). As a result, female body size is canalized in comparison to male body size. While there is always some degree of size dimorphism in human populations, dimorphism increases when males are able to reach their genetic potential and decreases during periods of nutritional stress. There are many ways to measure sexual dimorphism. Ruff (1987) states that changes in the degree of sexual dimorphism in long bone lengths may be more appropriate than cross sectional dimensions for addressing questions of how nutrition or general health influences human morphol ogy. Because the focus of this research is in how growth disruption, the product of ill health and/or malnutrition, affects sexual dimorphism, it is the dimorphism of long bone lengths and overall stature that will be analyzed here. The third alternati ve hypothesis tested states that proportional differences will be observed between those with independent evidence of growth disruption and those without.
19 Bogins (1999) studies of Japanese children have led to interesting conclusions about the relationshi p between genetics and environment. Bogin believes that genetic differences between Japanese and Europeans can explain differences in size, but that environmental factors may be powerful determinants of body proportion (Bogin, 1999:241). By attempting to control for genetic factors and using independent indicators of stress, this research provides a foundation for teasing out the effects of the environment on body size and proportion. Proportional differences are measured in terms of relative trunk to limb ratios and differences in proximal and distal limb lengths. Stini (1985) and Tanner (1963) both state that proportional differences in human form do not occur as a product of growth disruption, but a number of animal experiments suggest that testing an alternative hypothesis is worthwhile (Williams and Hughes, 1975; Wilson and Osbourn 1960). Studies of catch-up growth demonstrate that the process differentially affects body segments based on the stage of growth at which catch up occurs; at least unt il a normal growth trajectory is resumed (Fleagle et al. 1975; Tanner 1981). If proportional differences can be correlated with the timing of stress events, results may inform the understanding of mechanisms of catchup growth. While the process of c atch up growth has been well documented in animal models and in studies of human growth, very few of these studies follow individuals through to adulthood in any animal but laboratory rat s Assumptions are made about the environmental conditions of past populations based on stature estimations. Yet, there is no consensus on how catchup growth affects adult human morphology. The goal of this research is to bridge the empirical gap between the morphometric data collected by biological anthropologists and l ongitudinal studies of human growth. Osteological analyses assume a relationship between stature and stress. The importance of this research is that
20 it tests an assumption which has directed research in biological anthropology. This assumption has been used to measure human health and explain some aspects of human variation. For example, researchers investigating secular trends in stature are assuming an environmental effect. Meadows Jantz and Jantz (1999) state environmental forces, such as nutrition and disease, are the usual causes of secular change in overall size. In short, all avenues of research assume that the environment affects human biology; the ways in which it is affected are not as clearly understood. Chapters 2 and 3 provide backgrou nd information for understanding growth, growth disruption and indicators of such disruption that appear in skeletal materials. Chapter 4 outlines the analytical approaches used to address the hypotheses outlined and Chapters 5 through 8 report the results of those analyses. In C hapter 9 the implications of these analyses are discussed.
21 CHAPTER 2 ADULT STATURE AND PR OPORTIONS Human Growth Like all primates, humans have extended periods of growth in comparison to other mammals. From th e gestational stage through maturity, there is a common pattern of growth in normal, healthy human children (Bogin, 1999; Eveleth and Tanner 1990). Human growth can be divided into five basic stages: the prenatal period, infancy, childhood, the juvenile period, and adolescence. Although growth is continuous until maturity, growth rates are not constant. In humans, our most rapid rate of growth occurs prenatally, particularly during the period just before birth (Stinson 2000). Successful growth during the prenatal period is often measured by birth weight. Robson (1978) estimates that as much as 66% of the total variation in birth weight may be due to the prenatal environment as opposed to genetic factors althou gh others argue that it is less (Stinson 2000). Infancy is the period from birth to approximately three years of age. The end of the infant stage is most often associated with the age at weaning1Unlike the typical mammalian pattern, primates are among those animals which experience an extended growth period between weaning and puberty. Extended growth is believed to be cor related to an animals need to socialize prior to reaching reproductive age (Bogin 2003). N on -primate mammals exhibiting extended growth include animals such as wolves that live in complex social groups. In humans, this post -weaning period may be divided further into childhood and juvenile periods. During childhood ( three to seven years), the growth rate remains constant and it decelerates during the juvenile period ( seven to eleven years). (Bogin 1999; Stinson, 2000). The human growth rate slows throughout infancy. 1 Some sources consider infancy to end at age f ive or six based on the eruption of the first permanent tooth (Schultz 1969).
22 Adolescence is the period from around eleven years of age until the cessation of growth. It is during this period that a growth spurt occurs (Bogin, 1999). Although non human primates have a life cycle more similar to humans than that of non-primate mammals, humans represent an extreme of the primate spectrum. For example, there has been much discussion regarding whether or not the adolescent growth spur t seen in humans is unique to the human species or whether other primates experience a similar growth spurt (Bogin 1999; Leigh, 1996). Leigh (1996) demonstrated that humans are not unique among primates in experiencing an adolescent growth spurt in weigh t; nevertheless, the human growth spurt does occur at a later age than would be expected given the trajectory of other primates. Moreover, non-human primates do not appear to experience a spurt in skeletal growth of the magnitude or duration of growth spu rts seen in humans (Bogin 1999). Throughout growth there are considerable changes in body proportions as well as in growth rates. In early growth, the brain and head are proportionally larger in comparison to the rest of the body. For example, by ag e five, the brain is already 90% of its adult weight despite the fact that linear growth does not cease for at least another ten years (Tanner 1988). As an individual grows, relative limb length increases in comparison to the rest of the body (Bogin 1999) Bone formation begins in utero due to the ossification of preexisting connective tissues. The flat bones of the cranial vault, the mandible and the clavicle are referred to as intramembranous bone because ossification occurs within sheet like membranes. Most of the bones included in growth studies (including vertebrae and limbs) are referred to as endochondral bone because ossification begins in masses of hyaline cartilage, referred to as cartilage anlagen, which are in the location and approximate shape of the future bone.
23 Blood ve ssels and osteoblasts (bone forming cells) from the periosteum begin to invade the center of the cartilage model during the second or third intrauterine month (Steele and Bramblett 1988). This becomes known as the primary center of ossification and bones develop in the direction of their cartilaginous ends. In human long bones, secondary centers of ossification develop in the epiphyses. There are approximately 450 ossification centers at birth. As a general rule, most primary centers of ossification de velop before birth and most secondary centers develop after birth (White and Folkens 2005). The cartilaginous band that resides between the epiphysis and diaphysis is referred to as the epiphyseal plate. As new chondrocytes undergo cell division, older ones adjacent to the shaft ossify. Long bones lengthen through this process while the epiphyseal plates are active until m aximum length is reached when the diaphysis and epiphysis unite. The timing of epiphyseal closure is variable; however, in general, females mature one to two years earlier than males. The distal epiphysis of the humerus is the first epiphyseal plate to fuse in human long bones. This may begin at as early as nine years in females. The humeral head, one of the last plates to fuse, may not be complete until as late as 24 years in males (Buikstra and Ubelaker 1994). As a bone lengthens, the diameter of the shaft also increases as osteoblasts form compact bone beneath the periosteum. Osteoclasts dissolve bone adjacent to the marrow ca vity, increasing both the cross sectional diameter of the shaft and the size of the medullary cavity. Even after adolescence, this process slowly continues throughout the life of an individual. Growth Disruption Stature and proportions, such as sitting height, can be used as indicators of environmental stress prior to the completion of growth. For example, in undernourished children, skeletal age
24 may be retarded in comparison to chronological age, an indicator that a normal growth trajectory has been disrupted. A childs growth rate reflects, perhaps better than any other single index, his state of health and nutrition; and often indeed his psychological situation also. Similarly, the average values of childrens heights and weight reflect accuratel y the state of a nations public health and the average nutritional status of its citizens, when a ppropriate allowance is made for difference s, if any, in genetic potential. (Eveleth and Tanner 1990:1) Longitudinal and cross -sectional studies of growth ha ve been used to create standards or references for growth to help identify growth disruption (Tanner 1986). Ideally, these standards are population specific and are based on data from healthy individuals in a given population (Eveleth and Tanner 1990). Limiting standards to healthy individuals is an attempt to control for environmental effects, and population specific studies attempt to control for genetic effects. For example, comparisons of growth trajectories between children of African and European descent demonstrate differences in axial and appendicular growth; black children have shorter trunks, longer legs, and longer forearms than their white counterparts (Hamill et al. 1973; Nyati et al. 2006). For an individual child, mid -parental height (defined as the average height of an individuals mother and father) can be used to estimate a target range for growth as well as to predict a childs final height (Tanner 1986). Simple height -for age formulae are problematic due to genetic effects. For example, because of variations in average population heights, the World Health Organization ( WHO ) relies more heavily on weight -for -height data than height for age data (W HO, 1986). Because stunting is cumulative and catchup growth can erase growth stunting over time, it is difficult to make assumptions about growth disruption based simply on height -for age data. An example given in the Bulletin of the World Health Organ ization (WHO 1986) used the following example: if a four year old is found to have a greater height -for age disparity than a two year old, it is not necessarily because the four year old
25 is more malnourished. Statural growth is cumulative so the process of deprivation and subsequent retardation may have just had more time to affect the four year old. Microbial infections have been found to have a significant effect on human height (Beard and Blaser 2002). Not only are there are significant associations between bacterial infections and growth, but historical data find height reductions when microbial transmission rates are high. While growth is said to reflect health and nutrition (Eveleth and Tanner 1990), the two are inexorably intertwined. For exam ple, Helicobacter pylori infections, one of the most common infections in humans worldwide, have been found to have a negative association with growth when coupled with iron deficiency anemia (Choe et al. 2000). Deficiencies in iron, vitamin A, vitamin D, protein, thiamine, ascorbic acid, riboflavin, niacin, zinc, folic acid, and vitamin B12 are just a few of the nutrients that have been associated with inhibited antibody formation (Calder and Jackson 2000; Scrimshaw et al. 1968). Scrimshaw et al. (1968) found that not only were the consequences of infection more severe when malnutrition was an issue, but often infectious diseases made minor nutritional deficiencies much worse. Malina (1987) defines six classes of nutrients: carbohydrates, fats, protei ns, vitamins, minerals and water. Carbohydrates provide energy; proteins are primarily responsible for the growth, maintenance and repair of tissue; and vitamins and minerals serve regulatory functions. When caloric intake is too low, protein will be use d for basic energy needs. Once the basic needs are met, only then does the body devote resources to skeletal growth. There are numerous cases where nutritionally deprived children are of normal weight -for -height, but are stunted in height -for age compared to adequately nourished children (Eveleth and Tanner 1990). It is in this sense that height at a particular age reflects an individuals history of net nutrition (Steckel 1995: 1910).
26 Researchers have attempted to tease out the different dietary co mponents to see which have the greatest effect on growth. Observations of growth patterns in developing countries have led researchers to conclude that low protein has the most impact on growth of any nutritional deficiency (Martorell and Habicht 1986; S tini 1971). Nutritional supplementation projects in developing countries have served as natural experiments supporting th is claim (Silventoinen 2003). However, caloric as well as protein deficiencies have been found to negatively affect collagen produc tion, a protein which accounts for approximately 90% of a bones organic material. Vitamin D and calcium deficiencies have long been associated with poor mineralization of bone, but evidence suggests that deficiencies in a number of micronutrients may als o affect growth (i e zinc, phosphorus, magnesium, vitamin A, and iron) (Allen, 1994, 1995; Prentice and Bates 1994). An attempt to correlate dietary protein, as evidenced through stable isotope analysis, with femoral length in a South African bioarcha eological sample proved to be unsuccessful (Pfeiffer and Sealy 2006). Thus, in most human populations, dietary deficiencies probably involve multiple nutrients making it difficult to parse out the effects of various dietary components. For example, low protein diets are often low in energy and important micronutrients (Allen 1994). Stinis research on sex differences in response to nutritional stress led him to conclude that males suffered more as a consequence of protein deprivation, leading to a d ecrease in sexual dimorphism in protein deprived populations (Stini 1969, 1973, 1975a 1975b, 1982). This finding is believed to be the linked to the greater fat and nutrient reserves found in females. These fat and nutrient reserves are thought to be an adaptation for the increased metabolic demands of lactation and gestation in producing offspring. Energetic stress results in lower fecundity for women, but the same patterns have not been found in men (Bribiescas 2006;
27 Ellison 2003). Spermatogenesis requires a negligible energetic investment. Testosterone levels drop during periods of energetic stress, but this does not affect fecundity and testosterone levels quickly rebound when conditions improve. Testosterone affects muscle anabolism and male physiology favors reproductive function at the expense of building muscle during periods of en ergetic stress (Bribiescas 2001, 2006; Ellison 2003). As a result of these physiological differences, males experience a greater reduction of lean body mass during periods of nutritional inadequacy than do their female counterparts. When periods of sta rvation occur during growth, the reduction in body mass is accompanied by reduced skeletal growth (Stini 1975a). On the other hand, Stinson (1985) did not find irrefutable support for the hypothesis of sex differences in environmental sensitivity, likel y due to sex -biases in parental investment. Equality in stress levels of male and female children should not be assumed. Severe undernutrition leads to stunted adult size, but some researchers have questioned whether or not stunting matters (provided the brain escapes any lasting effect) (Eveleth and Tanner 1990:195). Furthermore, Eveleth and Tanner suggest an adaptive advantage to small size under adverse conditions, because smaller body size requires fewer resources (Mrquez and del ngel 1997; Nickens 1976; Seckler 1980; Stini 1975a). Assertions regarding the adaptive nature of stunting are controversial because of the inherent costs and questionable benefits to stunted individuals. Some researchers argue that small size as a result of depriv ation cannot have adaptive advantages due to the negative processes, such as infection and malnutrition, which cause growth stunting (Gopalan, 1988; Martorell 1989). A review of data on the subject found that smaller individuals were not able to perform more work relative to body size, nor did they experience reproductive advantages during times of scarcity (Stinson, 1992). Likewise, mortality profiles differentially favor taller individuals (Saunders and Hoppa
28 1993), and the conditions that lead to stunting also negatively affect cognitive development (Colombo et al. 1988; Martorell 1989). Catch up Growth Growth disruptions are often only episodic and the human body experiences a process called catch up growth once the adverse conditions which c aused the growth disruption are improved (Cameron, 2002; Tanner, 1981) Catch up may be achieved through an increase in the rate of growth or by postponement of growth completion, thereby lengthening the growth period. In a growing individual these proce sses can be recorded in the skeleton ; however, once skeletal growth is complete, it is more difficult to ascertain whether an individual experienced disrupted growth or subsequent catchup growth While catch up growth has been observed clinically in pediatric case studies and has been recreated in laboratory experiments on non -human animals, the mechanism behind catch up growth is poorly understood. Tanner (1963) theorized that catch up growth is a systemic, neuroendocrinologial process that recognizes a mismatch between the present and potential size and then adjusts the growth rate through a target seeking process. Tanner admitted that the hypothesis he presented was speculative, but thought that growth experiments of the time were hampered by the l ack of appropriate theoretical models under which to work (Tanner 1963). In contrast, Williams and Hughes (1975) concluded that Tanners hypothesis was too simple. Because catch up growth appeared to affect different body parts in different ways, these researchers felt that the resultant growth after rehabilitation was the product of an interaction between the catch up stimulus and the normal growth stimuli (Williams and Hughes 1975). Baron et al. (1994) suggested that the primary mechanism of catch up growth resided at the growth plate rather than in the central nervous system. Their hypothesis was based on an
29 experiment in which growth was artificially suppressed in a single growth plate. Rabbits were locally administered glucocorticoid unilaterally in the proximal tibia. The growth suppressant was removed prior to growth completion. The experimental tibia never attained the full size of the control; however, the growth rate in the experimental tibia surpassed the rate of growth in the control side. Left and right femoral lengths were not statistically different offering further support for the conclusion that catchup growth was localized to a single epiphyseal plate. While this experiment does not rule out the possibility that a regulating mechanism exists at the systemic level, it does suggest there may be a more localized response in individual bones. Due to the long human growth period and the difficulty in controlling for confounding variables in human populations, animal experiments have pr oved useful in providing models of catch up growth. Mammalian models of growth are often tested on mice or rats likely because of their relative availability, low cost and short life -span. However, the utility of applying nonprimate mammal models to humans is hampered most notably by their differences in life cycles. These differences are incredibly important when trying to apply the timing of nutritional insults in animal models to humans. First, it is likely that the early postnatal life of a mouse i s analogous to prenatal stages in a human. An excellent case in point is a study by Grove et al. (2005), which investigated the development of metabolic systems in relation to overfed and underfed animals by performing experiments on both rodents and prim ates ( Macaca fuscata). They found that the neurons for transmitting hormonal signals that modify feeding and energy expenditure develop in week three postnatally for rodents and in the third trimester for primates (Grove et al. 2005). In addition, rodents do not have a juvenile period comparable to humans (Bogin 1999). Because human infants are even more altricial at birth than non -human primates, growth periods in young rats and mice may be comparable to earlier prenatal growth in human
30 analogues. In addition, the presence of the adolescent growth spurt in humans cannot be matched in the rat; even compared to other primates, the absolute growth at puberty is greater in humans (Bogin 1999). Another criticism of using rat models for studying human growth is that skeletal growth in humans and rats may be too different to be compar ed. Tanner (1963) claims rats do not exp erience full epiphyseal closure. I n contrast, Martin et al. (2003) not only document ed the rates of epiphyse al closure in rats, but they suggest that the rates are subject to environmental pressures in a manner similar to those in humans. Rat bone does differ from humans, however, in that there is no secondary remodeling of cortical bone. The process of bone r emodeling removes and replaces older bone thereby repairing microscopic fatigue damage (Martin et al. 1998). Due to the short lifespan and small size of rodents, remodeling would be a worthless energetic expenditure. Martin et al. (2003) concluded that rat models are useful for studies of growth, but not for studies of aging. Genetically, the breeding populations of rodents and non-human primates used for laboratory research are fundamentally different. The breeding of rodents for experimental research is controlled in an attempt to create a genetically defined animal model, thereby eliminating a certain degree of variation between individuals. In captive primate populations, where selective breeding occurs, the attempt is generally to maintain genetic diversity within those populations (Ribeiro Andrade et al. 2004). While the ultimate effect of the relative genetic variability between primates and rodents may be small, it is important to recognize that the genetic effect on growth is likely to be a gr eater confounding variable in non-human primate studies than it is in rodent experiments.
31 The majority of catch up growth studies performed on rats or mice focus on the timing and duration of nutritional insults (Farnum et al. 2003; Widdowson and McCance 1963; Williams and Hughes 1975). While the results of these studies have varied, they have proven useful in establishing two facts. First, there is evidence that there are critical stages during growth at which the effect of growth disruption is greate r. Second, c atch up growth follows three possible trajectories: a rapid growth spurt when conditions improve, an extended period of growth, or, in extreme cases, both the rate and duration of growth may increase (Boersma and Wit 1997). The most importa nt limitation of catch up growth studies on non human primates is also one of the major hurdles in conducting longitudinal studies of human growth the time it takes for an individual to reach full adulthood. As such, not all growth studies of non-human primates have followed subjects through to growth completion. For example, Fleagle and colleagues (Fleagle et al. 1975; Fleagle and Samonds 1975) demonstrated that the response to growth arrest was not uniform throughout the body of Cebus albifrons Wh en resources are limited, they appear to be channeled from the most mature skeletal elements toward those with the most growing left to do. However, because the animals used in this study were not followed through adulthood, it could not be determined whe ther or not the effects of protein deficiency were permanent (Fleagle et al. 1975; Fleagle and Samonds 1975). While the findings of this study were very provocative and have been frequently cited in subsequent studies of catchup growth, data on those an imals are not available past two years of age. It was believed at the completion of the study that the animals had reached a normal growth trajectory, but Cebus albifrons does not reach physical maturity until around the age of five years. With the excep tion of rats, who are relatively short lived, catch up growth studies rarely follow individuals experiencing growth arrest to the age of skeletal maturity.
32 There is disagreement regarding the degree to which catch up growth occurs in humans. The results of catch up studies differ, likely due to variation in the duration, severity and timing of the insults, as well as in the triggers initiating catch up growth. Numerous examples exist which prove that catchup growth can and does occur in human populatio ns, although the degree of recovery varies (Delgado et al. 1987; Johnston and MacVean 1995; Li et al. 2004; Steckel 1987). For example, height data recorded for American slaves provide notable evidence for catch up growth. The average heights of Amer ican slave children were around the first centile compared to modern standards. However, young adult males were in the 25th centile and females were in the 30th centile based on the same standards (Steckel 1987). This pattern of growth is unusual and is thought to reflect improved nutrition during growth. Young children were weaned from breast milk relatively early and were fed a poor diet. Rates of infant and childhood mortality were high among American slaves, due in part to the fact that it was not until children entered the work force, by the ages of eight to twelve that they received a workers allocation of meat. It was at these ages that the process of catch up began and height records suggest that growth continued until 19 years in females and 21 years in males (Steckel 1987). Martorell et al. (1994) compiled evidence of other cases in which catch up occurred, but notes that evidence suggests catch up is more effective when the conditions improve at younger ages. Stinson (2000) points out tha t in studies of catch up growth at the population level, increased mortality of smaller individuals may contribute to the appearance of catch up growth. The surviving population may no longer include its shortest members. While the potential for catch u p growth is apparent, Martorell et al. (1994) found that retarded growth in early childhood resulted in shorter adults in most cases. Li et al. (2004) analyzed longitudinal data from the 1958 British birth cohort and confirmed this position.
33 Disadvantage d children experienced sizeable height deficits ( two to three centimeters ) at age seven years. Catch up growth erased some, but not all of this deficit, resulting in adults that were one centimeter shorter on average (Li et al. 2004). When growth is stunted, the growth trajectory changes, lengthening the period in which growth can occur (Golden 1994). However, unless an individuals environment changes and there is sufficient time left for growth, stunting will persist into adulthood. Age at menarche is used as a marker for determining maturation in human populations. Data on populations worl dwide compiled by Eveleth and Tanner (1990) found that menarche was delayed in chronically undernourished and high altitude populations; however, in most cases, Martorell et al. (1994) did not find that the delay was long enough to fully compensate for growth stunting. In fact, Martorell et al. (1994) find some evidence that for those whom experience accelerated growth late in the growth period; maturation may be triggered, leading to short adult stature. The conclusion of these researchers is that, altho ugh complete catch up is possible, for most stunted children the environment does not change, leading to shorter adults (Golden 1994; Martorell et al. 1994). Timing of Growth Disruptions Because the rate of growth is not constant from infancy through adulthood, it has been suggested that the relative sensitivity to episodes of growth disruption also varies. Eveleth and Tanner (1990) believe that growth is most sensitive to nutritional deprivation during times of peak growth velocity: infancy and adol escence. Experiments on animal models where test subjects were nutritionally deprived during various stages of growth do provide some evidence for the differential effects of growth disruption in relation to body segments which grow at a faster or slower rate. In rats, it has been found that the effects of undernutrit i on become progressively less important with age (Widdowson and McCance 1963). Tanners (1962)
34 synthesis of research on animal husbandry found that animal shape was affected by both the tim ing and intensity of insult; faster growing body segments appeared to suffer the most from nutritional insult. Wilson and Osbourn (1960) found that the rate of growth in late maturing regions of the body w as most affected by growth insults in mammals and birds. However, because these regions had greater time to recover, the growth disruption is not evident in the adult form. In the developing world, research suggests that poor environmental circumstances can cause a normal growth trajectory to falter by s ix months of age, if not earlier (Waterlow 1988b). In studies of secular trends in stature, Hauspie et al. (1996) and Stinson (2000) conclude that the vast majority of differences seen in adult height can be explained by growth disruptions prior to six y ears of age. Not only are growth disturbances greatest in the first two to three years of life, but nutritional stresses are also greatest during this time period (Beaton et al. 1990; Martorell and Habicht 1986; Martorell et al. 1995). Stinson (2000) lists the cultural norms surrounding breastfeeding as one of the most important culturally mediating factors associated with growth. Weaning is a particularly dangerous time of life because food is a source of pathogens and children no longer have the immun ities provided by mothers milk; rates of mortality tend to spike at this phase of life. To compound the problem, the nutritional needs are great during this phase of life because the growth velocity is high (Martorell et al. 1994). Bec ause growth appears to be more sensitive to environmental circumstances in early childhood (Hauspie et al. 1996; Martorell and Habicht 1986; Martorell et al. 1995; Stinson 2000; Wadsworth et al. 2002), and legs are the fastest growing region of the bo dy from birth until puberty (Dangour 2001; Floyd 2007; Gunnell 2002; Tanner 1962), it stands to reason that
35 individuals who experience growth disruption in childhood should have proportionally shorter lower limbs Numerous studies on human proportions have found support for this hypothesis, includ ing studies of secular change. Luo and Karlberg (2000) examined longitudinal data to see how different growth phases interacted, leading to adult shortness. The shortest adults experienced subnormal growth during multiple growth phases. When growth data are compared with midparental height, the relationship between genetics and height was not apparent until after infancy. Luo and Karlberg (2000) found that most cases of growth faltering during and after childhood were erased through catch up growth, but that the same is not true during the fetal and infant stages. In a subsequent study, Luo et al. (2001) found that adult shortness was associated with subnormal growth in any growth phase. They concluded that short adult stature in the developing world was primarily due to environmental factors rather than genetics because the stature data were adjusted by mid parental height, a proxy for the genetic ef fect. Despite being a period of rapid growth, adolescence is not considered to be a peak period of permanent growth stunting. Growth during adolescence correlates more closely with genetics than with environmental circumstances (Frisancho et al. 1980; Jo hnston et al. 1976; Stinson 2000). During adolescence, large strides can be made in growth under seemingly poor environmental circumstances. Adolescent growth patterns are more similar between individuals of similar genetic backgrounds than between individuals of similar socioeconomic classes (Frisancho et al. 1980; Johnston et al. 1976). However, the physical demands of this rapid growth phase are great and the adolescent growth spurt would mask growth rate changes associated with stunting during th is phase, making it harder to detect. One study found that although adolescent and early childhood growth are statistically independent, short adolescents
36 experience more stunting than their taller peers during periods of hardship (Hermanussen, 1997). Fo gel (1986) suggests that growth disruptions in adolescence lead to a delay in growth while it is the early disruptions which have a permanent effect. Studies on adolescent growth disruption indicate that growth is not immune to environmental privation dur ing adolescence, but the ultimate effects are not as easy to detect. Experimental research on the effect of temperature, physical activity and nutrition on bone elongation support s the theory that the early postnatal growth is the most critical phase (Serr at et al. 2009). When catchup growth occurs, it generally does so by extending the growth phase or increasing the rate of growth. In other words, growth disruptions change ontogenetic trajectories through heterochronic processes. Allometric differences are due to the fact that different body segments have different growth profiles (Vrba 1996). Adult Proportions Fleagle and Samonds (1975) did not suggest that the effects of growth disruption seen in Cebus albifrons were permanent, but their findings r aised the question of whether or not proportional differences can be seen in individuals who have experienced catch up growth as opposed to those who have never suffered growth disruption. Some, but not all, of the catchup growth studies using rats have found proportional differences in final form. For example, Williams and Hughes (1975) found that catchup in nose to rump length of the experimental rat group neared adult length in the control group. However, tail length in the experimental group never attained control levels. Early studies of growth disruption suggest that although adult body size can be affected by growth disruption, changes in proportions would be rare if they were to occur at all (Stini 1985; Tanner 1963). It is widely accepted that human proportions (i.e., limb to trunk ratios, etc.) appear to be under strong genetic control (Bailey et al. 2007; Bogin and Rios 2003; Eveleth and
37 Tanner 1990; Hamill et al. 1973; Holliday, 1997; Mueller, 1986; Turan et al. 2005; Warren et al. 2002), and that these generally conform to the geographic expectations of Bergmanns and Allens rules (Baker 1988; Hiernaux et al. 1975; Holliday and Falsetti 1995; Katzmarzyk and Leonard, 1998; Kurki et al. 2008; Roberts 1953; Roberts 1973; Ruff 1991; Ruff 1994). However, deviations in ecogeographic patterning also have been attributed to environmental effects such as health care and nutrition (Katzmarzyk and Leonard 1998; Kurki et al. 2008). Differences in proportions within populations suggest environmental pressures affect human proportions. Several studies have found that lower limb length is more sensitive to environmental change than trunk length, which in turn leads to differences in proportions. With an improved developmental environment, lower limb length increases at a greater rate than relative sitting height (Bogin and Rios 2003; Bogin et al. 2002; Dangour, 2001; Floyd 2007, 2008; Fredriks et al. 2005; Frisancho et al. 2001; Katzmarzyk and Leonard, 1998; Li et al. 2007; Malina et al. 2004; Stinson, 2000; Tanner 1992; Wadsworth et al. 2002). Other studies have found differences in relative lower limb length based on factors such as altitude or activity pattern (Bailey et al. 2007; Theintz et al. 1993). Heart disease and associated risk factors are inversely related to lower limb length while the relationship to trunk length is not as strong (DaveySmith et al. 2001). Fredriks et al. (2005) state that relative lower limb length may be a more sensitive indicat or of environmental circumstances than height. The relationship of proximal and distal limb elements is another proportional measure that has been associated with environmental effects. The length of the tibia has proven to be more variable than other limb segments, and distal elements tend to be more variable than proximal limb segments (Holliday and Ruff 2001; Smith and Buschang, 2004). Secular allometric trends
38 have shown that as the environment improves the relative length of the tibia increases ( Jantz and Jantz 1999; Jantz and Owsley 1984; Meadows and Jantz 1995). Longitudinal studies also demonstrate the relative sensitivity of the femur to environmental change (Smith and Buschang, 2004). Secular Trends in Adult Height Liu et al. (2004) estimat e the heritability of stature to be above 0.75 while recognizing that nutrition and disease play a role in determining adult height. Adult stature is a complicated indicator of growth performance because it is difficult to determine what an individuals ultimate height could be under ideal conditions. Estimates of stature can be made based on mid-parental height, but there is a great deal of variation between individuals. Therefore, secular trends in adult height are compiled at the population level wher e individual genetic effects are canceled out and differences in height can be attributed to the environment (Steckel 1995). Secular trends in stature refer to short term changes in growth and development rather than changes at the evolutionary level (Ru ff et al. 1984). Looking at intergenerational differences within family lines is another means by which individual variation is controlled in secular studies (Damon, 1969; Floyd 2007, 2008). Secular studies use historic data or skeletal samples to obs erve trends or fluctuations in human morphology over time. In particular, military, medical, prison and slave records have provided a wealth of data for observing these trends historically (Fogel 1986; Komlos 1994; Steegmann 1985; Steegmann and Haseley, 1988; Van Wieringen 1986). Availability of height data on the average population has become more available since the mid 18th century (Steckel 1995). The general trend over the last 100 years is that people are growing larger, faster and that physica l maturation occurs at a younger age (Eveleth and Tanner 1990; Lieberman 1982). This being said, it is important to realize that the average European American is only a few
39 centimeters taller than they were 100 years ago ; heights have fluctuated cyclica lly rather than steadily increasing (Fogel 1986; Steckel 1995). The fact that fluctuations in height correspond to the economic, nutritional, and social circumstances at the time of a specific birth cohort is of particular interest in secular studies of adult height. On the other hand, expectations of stunting were not met in one study of a mid 19th century poorhouse raising the question of the degree of severity in conditions needed for stunting to occur (Steegmann 1991). Presumably, secular increases in height can be attributed to the fact that today we are subjected to fewer growth inhibitors than in the past and are therefore more likely to reach our genetic potential (Lieberman 1982; Stini 1975a). Secular change is often attributed to changes in nutrition, but there are numerous other factors that have been associated with changes in height. Improvements in hygiene and sanitation have reduced exposure to infectious agents, allowing children to utilize nutrients for growth. Height has also been a ssociated with urban/rural residence patterns, reduced family sizes and occupation. It has been suggested that selection for taller mates has led to increased heights (Hauspie et al. 1996; Huss -Ashmore et al. 1982; Malina 1979; Steegmann 1985). As populations have migrated, heterosis has become a factor, but Malina (1979) suggests that the effect of heterosis is small at best in human populations. By correlating birth dates, height and historical data, researchers find that nutrition has the greates t impact on height from the prenatal period through early childhood (Fogel 1986; Hauspie et al. 1996; Hermanussen 1997; Steegmann 1985). More specifically, the greatest secular changes are seen in lower limb length rather than trunk length (Floyd, 2007; Himes 1979; Malina et al. 2004; Stinson 2000). In addition, it appears that the distal limb is affected more than the proximal limb segment (Floyd 2008; Meadows Jantz and Jantz 1999). Thus, the
40 results of secular studies follow the same trends fo und in studies correlating human proportions and environmental stress. Multiple studies have found that secular increases are greater in males than females (Engerman 1994; Himes 1979; Meadows Jantz and Jantz 1999; Kuh et al. 1991). As discussed prev iously, some researchers have observed differential response by males and females to negative environmental conditions, which may explain the findings in secular studies (Stini 1969, 1973, 1975a 1982). The theoretical basis for this finding will be disc ussed in further detail in the following section on sexual dimorphism. Within the last 50 years, researchers have questioned whether the secular trends observed from the mid 1800s through the mid 1900s were still occurring. Most evidence suggests that the secular trend has slowed if not stopped. Damon (1969) analyzed data on Harvard families from 18701965 and found that the intergenerational change in males was initially strong, but then declined. In fact, Damon suggested that the secular increase i n height has peaked. Other researchers have concluded that the secular trend has slowed or stopped in developed countries, but persists in the developing world (Hauspie et al. 1996; Roche 1979). If stature can be used to track environmental stress in hu man populations, and failure to reach genetic potential can be attributed to environmental stressors, the most reasonable mechanism for stunted growth must involve growth disruption and incomplete catch up growth. Therefore, it is desirable to investigat e how growth disruption and catchup growth affect adult morphology. Secular studies of adult form rely on understanding how growth disruption affects the end product, and unfortunately there are many assumptions that have not been tested. For example, many studies of growth disruption in humans do not follow individuals from birth through to growth completion (Checkley et al. 1998; Liu et al. 1998; Nabarro et al. 1988;
41 Smith and Buschang, 2004); thereby ignoring potential catchup. The longitudinal s tudies on growth that have been completed have concentrated on looking at patterns of growth rather on the effects of growth disruption (Garn and Rohmann 1966; Maresh 1955). As such, even when environmental effects can be assumed, a direct relationship between the environment, growth disruption and catch up growth cannot be systematically tested. Knowledge of the environment can only be gleaned through factors like social status (Kuh et al. 1991) or maternal smoking during pregnancy (Li et al. 2004). While this information is useful, it does not provide a full picture of the environmental circumstances, nor does it provide a record of chronic or acute stress events at the level of the individual. Sexual Dimorphism There are many ways to measure sexual dimorphism. Although in living populations stature is probably the most common measurement of size, simply because of the ease of measurement, the type of measurement considered is often determined by the questions being asked of the data. Ruff (1987) s uggests that cross -sectional studies of diaphyses are evidence of the mechanical forces applied to bone, reflecting activity patterns during life. On the other hand, Ruff states that changes in the degree of sexual dimorphism in long bone lengths may be m ore appropriate for addressing questions of how nutrition or general health influences human morphology. While size and shape are correlated in considerations of human form, Ruff (1987) makes an interesting proposal for how to best measure human sexual di morphism. It is also important to consider the scale on which to consider changes in sexual dimorphism. Studies of change in the degree of sexual dimorphism in anatomically modern human populations provide information on a secular rather than an evolution ary level. While some forces may act at the evolutionary and secular scale, others do not appear to be operative on both levels. Evolutionary forces change the genetic composition of a species and are used to
42 explain the differences observed between spec ies or over spans of time representing hundreds if not thousands of generations. The time depth of secular changes may be generational or even longer, but fluctuations in secular trends suggest that they are not genetic changes. The degree of sexual dim orphism in Homo sapiens has not been consistent throughout time and has been attributed to many factors including genetics, sexual selection, activity patterns and nutrition. The trend has been that sexual dimorphism declined from the Upper Paleolithic to the present (Borgognin i Tarli and Repetto 1997; Brace 1973; Brace and Ryan 1980; Frayer 1980; Frayer 1981; Frayer and Wolpoff 1985; Meiklejohn et al. 1984). These decreases have been associated with changes in subsistence or technology, most notably at the end of the Up per Paleolithic when large scale extinctions led to hunting of smaller game (Brace and Ryan, 1980; Frayer 1980, 1981). While the length of both male and female limb segments declined at that time, the decline was greater in males, therefore reducing the d egree of sexual dimorphism (Frayer, 1981). Researchers have also hypothesized that a further reduction in sexual dimorphism occurred as humans shifted from a hunting and gathering economy to agriculture (Armelagos and Van Gerven 1980; Boyd and Boyd, 1989; Frayer 1980; Frayer and Wolpoff 1985; Hinton and Carlson 1979; Kennedy et al. 1987; Lazenby 2002; Ruff 1987; Wolfe and Gray 1982). In contrast to data available for the Upper Paleolithic Mesolithic transition, which strongly support a decrease i n sexual dimorphism, data for a possible reduction in sexual dimorphism after the shift to agriculture are more equivocal, despite larger samples (Vick 2005). These changes have helped inform researchers investigating the processes controlling sexual dimorphism.
43 While recognizing the primacy of heritability in the determination of stature (estimated to be above 0.75), Liu et al. (2004) recognized that nutrition and disease also play a role in determining adult height; likewise, variation in sexual dimorphi sm leads researchers to conclude that genetic as well as environmental factors influence sexual dimorphism. Eveleth (1975) found that the degree of sexual dimorphism varied between populations; but did not meet the expectations of the nutritional hypothes is that the most dimorphic populations should have the highest quality diet. As such, Eveleth concluded that there must be a genetic factor controlling the level of sexual dimorphism in each population. Gustafson and Lindenfors (2004) indicates that bo th male and female stature is phylogen etically controlled ; however, there is no evidence that the degree of sexual dimorphism is significantly more or less pronounced in populations that are larger in stature. The results of this study indicate a genetic component to sexual dimorphism while challenging the notion that relative sexual dimorphism is a byproduct of overall size. Above the species level, Renschs rule states that sexual dimorphism increases with body size in taxa where males are the larger sex (Rensch 1959). This relationship may not be as strong in primates as in other taxa (Frayer and Wolpoff 1985); nevertheless, Leutenegger and Cheverud (1985) found that variation in body weight could explain 83% of the variance in weight dimorphism among primates suggesting body weight is the major factor contributing to sexual dimorphism in body size. Among hominoids orangutans and gorillas, the largest members, are the most sexually dimorphic; however, reports of relative dimorphism within chimpanzees and humans can differ depending on the method or element used to make the measurement (Lovejoy et al. 1989; McHenry 1991; Richmond and Jungers 1995). Male -male competition is a common explanation for how sexual dimorphism develops. Because there is g reater reproductive variance in males (Chagnon 1979; Trivers 1972), larger
44 body size may offer an advantage to males engaged in aggressive encounters, allowing them to pass their genes to subsequent generations more effectively. Male -male competition is greatest among primates where females assume the largest share of energy investment in the success of offspring (Trivers 1972). Based on the theory of sexual selection, sexual dimorphism should be greater in polygynous rather than monogamous societies b ecause there is more competition for access to females. Gray and Wolfe (1980) tested this hypothesis in humans and found no significant correlation between sexual dimorphism and mating pattern. Gaulin and Boster (1992) reanalyzed the data of Alexander et al. (1979) and found that sexual selection did not explain variation in sexual dimorphism when applied to cross -cultural studies in humans. However, they do not rule out the possibility that sexual selection has an effect in human populations, suggesting that perhaps human marriage practices have not been stable through time, thus obscuring any observable effects (Gaulin and Boster 1992). Ethnographic studies also show that the quality of parental investment varies depending on the gender of the offspr ing. Because the variance in reproductive success is generally greater for males, parents who can afford to invest in their offspring have a potentially greater return by investing in males (Hrdy 1990). Rivers (1982) studied survival rates under conditi ons of famine and disaster and found that while females may have a natural advantage under times of stress, males often receive cultural advantages that may more than make up for any differential survival rates. Culture thus may affect the degree of sexua l dimorphism manifest in a population. Holden and Mace (1999) found that sexual dimorphism in stature is negatively correlated with the amount of work women perform. Likewise, female juvenile mortality rates are higher than those for juvenile males in are as where females contribute less to subsistence. These patterns follow geographical patterns of sexual dimorphism and may suggest that more resources
45 are allocated toward female children when there are economic returns for such investments (Holden and Mac e 1999). In addition, biomechanical data suggest that reductions in the division of labor, as seen with the transition to agriculture, lessen the sexual differences found in the cross -sectional properties of bone (Ruff 1987). Holden and Mace (1999) compared populations in the Ethnographic Atlas (Murdock, 1967) with regard to marriage practices, subsistence and the division of labor. [They] concluded that in contemporary humans, neither hunting nor agriculture has an y effect on sexual dimorphism. [Instead] It is the amount of subsistence work done by men and women, rather than the type of subsistence practiced, which has an effect on sexual dimorphism in different societies. (p. 42) As women contribute more to the su bsistence economy, it appears that the degree of sexual dimorphism is reduced. Holden and Mace (1999) hypothesize that these females are taller due to improved nutrition during growth, but mild to moderate physical activity has also been found to enhance linear growth (Torun and Viteri 1994). Consequently, there may be biomechanical and nutritional explanations for this phenomenon. Holden and Mace (1999) used stature as their only measurement of dimorphism while Ruff (1987) used only the cross -sectiona l properties of bone. Biologists may also view sexual dimorphism as a product of optimal biomass distribution for the species. When conditions select for large males, it is advantageous for the female of the species to be as much smaller as possible whil e still being able to achieve reproductive success (Bramblett 1994). The optimum female size is large enough to bear the physical demands of labor, but small enough to reduce the metabolic demands associated with large size. By considering sexual dimorphi sm as an optimal biomass distribution, DeVore and Washburn (1963) propose that males and females may be better able to utilize their resources if they fill
46 different ecological niches. If niche divergence amplifies with increased sexual dimorphism, then th e selective pressures affecting males and females are progressively more different. Perhaps most importantly to the issue of growth disruption, a number of studies have shown that sexual dimorphism in stature can decrease when people are under nutritional stress and increase under conditions of optimal nutrition (Brauer 1982; Gray and Wolfe 1980; Lieberman 1982; Stini 1969, 1982; Wolaski and Kasprzak, 1976). The theoretical basis for this is found in the fact that males and females experience differen tial success in dealing with stressors like starvation and disease due to hormonal and metabolic differences (Ortner 2003; Stini 1969). The greater fat and nutrient reserves characteristic of human females are thought to be an adaptation for the increas ed metabolic demands of lactation and gestation in producing offspring. As a result of these physiological differences, males experience a greater reduction of lean body mass during periods of nutritional inadequacy than their female counterparts. When p eriods of starvation occur during growth, the reduction in body mass is accompanied by reduced skeletal growth (Stini 1975a). Studies of secular trends in adult height have found that male increases in stature are greater than for females as the environm ent improves (Himes 1979; Meadows Jantz and Jantz 1999; Kuh et al. 1991). The resulting change in sexual dimorphism conforms to the theoretical expectations based on the differential response to stressors of males and females. The Metabolic and Disease Implications of Growth Disruption In modern America, obesity is more of a health risk than malnutrition; however, studies suggest that obesity may actually be triggered by early metabolic insults. Studies of growth disruptions and subsequent catchup gro wth have been divided between those that focus on soft tissue or metabolic changes and those that measure bony responses to catch-up growth. These studies can be broadly divided into those that focus on length and those that focus on weight.
47 Catch up gro wth of skeletal tissue (length) attempts to return an individual to a normal growth trajectory; on the other hand, catch up in weight deficiencies appear to be controlled by a different mechanism that may overshoot the normal growth trajectory leading to o besity (Hindmarsh, 2004). Hales and Ozanne (2003) attribute the effects of early growth disruption on metabolic disorders to what they term the thrifty phenotype hypothesis. During periods of nutritional deficiency the body naturally diverts scant res ources to the areas of greatest need. Resources are diverted away from the pancreas to support the brain. When deprivation occurs in early development the stress causes alterations to the metabolic programming of an individual. The result is thrifty m etabolic functioning programmed to subsist on fewer resources. Thrifty functioning becomes a problem when nutritional resources become available. It has been found that individuals with a lower than normal birth weight (an indicator of a less than optima l fetal environment) can experience metabolic alterations that make them prone to conditions such as obesity and Type II diabetes (Cameron and Demerath, 2002). Recent research has made strides in coordinating growth statistics with evidence of metabolic c hanges in both animal models and humans, focusing on early growth disruption (particularly in the fetal or early postnatal periods) and subsequent problems of obesity. (Cameron and Demerath, 2002; Grove et al. 2005). The bulk of this c hapter has focuse d on studies of hard tissues because the applicability to skeletal biology and bioarchaeological populations; however, from a public health perspective, the most important reason to concentrate on skeletal catch up growth is the role it serves as an indica tor of more serious health problems in growing children. While the mechanism controlling catch up growth in length and weight seem to be entirely different, they are similar in that both are physiological responses to growth disruption. Recent research has investigated the
48 relationships between metabolic disorders, weight, and length. As mentioned earlier, studies have found that lower limb length appears to be the component on human height most closely associated with environmental stress th e better the environment, the longer the legs (Li et al. 2007). This relationship has been used as an explanation for why shortened relative lower limb length has been associated with a range of medical conditions including cancer (Gunnell et al. 1998b), coronary heart disease (Davey -Smith et al. 2001; Gunnell et al. 1998a; Han et al. 1997; Lawlor et al. 2004), diabetes (Davey-Smith et al. 2001; Lawlor et al. 2004), and other metabolic disorders (Davey-Smith et al. 2001; Han et al. 1997).
49 CHAPTER 3 SKELETAL INDICATORS OF GROWTH DISRUPTION Enamel Defects Enamel develops in increments beginning at the apex and laying down successive layers of enamel during mineralization. The pattern and timing of dental development appear to be under str ong genetic control, with less environmental effect than that seen in skeletal development (Garn and Rohmann 1966). Surface enamel on the anterior teeth forms between approximately 0.8 and 5.2 years of age, but continues on the third molars to roughly 11.3 years (Reid and Dean 2006). Age estimates for the formation of surface enamel are summarized in Table 3 1. During growth, any disruption caused by metabolic insults may stop enamel formation just as they may stop bone growth. However, unlike bone, e namel is acellular and does not continue to remodel in adulthood; therefore, enamel provides an excellent source for investigating growth disruptions from adult skeletal material. Enamel hypoplasias are a type of developmental defect characterized by a d eficiency in enamel thickness due to a disruption during amelogenesis, defined as the process of enamel formation (Goodman and Rose 1990). These are formed when ameloblasts stop secreting enamel matrix earlier than normal (Hillson 1996). Possible etiol ogical factors for enamel hypoplasias include heredity, local trauma, or systemic metabolic disturbance (Goodman and Rose 1990; Suckling 1989). The majority of hypoplastic defects are caused by growth disruptions such as illness and nutritional deficiencies (Hillson and Bond, 1997). Hereditary defects are considered to be rare and can often be identified due to the fact that they are generally severe and all of the teeth in a set are likely to be affected (Buikstra and Ubelaker 1994; Goodman and Rose 1 990). Conversely, local trauma will only affect a single or a few adjacent teeth. It is reasonable to assume dental defects are due to metabolic stress, rather than trauma, if
50 they are bilateral in nature (Floyd 2007; Goodman and Rose 1990; Hillson 1996). Responding to nutritional stress, disease, and other physiological insults enamel defects are considered to be sensitive albeit non-specific indicators of stress (Floyd 2007; Goodman et al. 1987; Goodman and Rose 1990; Hillson 1996; Littleton 2005; May et al. 1993). Enamel hypoplasias do not take long to form and may represent a single, short -term systemic insult rather than long -term chronic stress (Suckling 1989). Surface defects that are macroscopically visible may appear in the form of lines, pits, or the complete absence of enamel. The variation in appearance of enamel defects has been attributed to the severity of insult (Suckling 1989), the duration of the metabolic stress (Blakey and Armelagos 1985; Hutchinson and Larsen 1988), t he position of the defect on the tooth (Hillson and Bond 1997), or may be related to other unknown factors. When scoring hypoplasias, not all researchers choose to include hypoplastic pitting because of distinct differences in the way pits and lines are formed (Steckel et al. 2006). Linear hypoplasias form when matrix secretion ceases along perikyma grooves. Perikyma grooves, also referred to as imbricational lines, mark the successive layers of enamel formation by particular ameloblasts. King et al. (2002) define linear enamel hypoplasias as a greater than expected spacing between neighboring pairs of perikymata. Pit defects do not occur along an associated perikyma groove on the tooth surface and are formed when only small clusters of ameloblasts stop forming enamel (Hillson 1996; Hillson and Bond, 1997). Cross -sectional studies have found that the associated perikyma groove may not demonstrate any surface defects. While hypoplastic pits represent a cessation of enamel formation, they are poorly understood and cannot be aged by macroscopic means (Hillson and Bond, 1997).
51 The surface of any tooth only represents a fraction of the total developmental history recorded in the enamel. By sectioning a tooth and observing it microscopically, the recor d of growth disruption is more complete. Microscopically, the incremental lines of growth in enamel, referred to as striae of Retzius, can be observed. Accentuated striae of Retzius, also referred to as Wilson bands, are considered to be a more sensitive indicator of growth disruption than enamel hypoplasias and may take less time to form (Larsen 1997; Rose et al. 1985). Simpson (2001) has also found differences in the age at which Wilson bands and hypoplasias form. Wilson bands most commonly form bet ween 12 and 30 months, whereas hypoplasias are more common after 25 months of age. Most surface defects in enamel are associated with Wilson bands. Where surface defects exist independent of Wilson bands, it is suggested that the surface defect may not b e due to physiological stress (Goodman and Rose 1990). Microscopic examination is also important for observing the hidden appositional zone of enamel. Enamel deposition begins at the cusp and the first layers become covered by subsequent layers of enam el matrix. In molars, the hidden appositional zone may account for between 40 and 50% of the total enamel. In anterior dentition, the appositional zone may account for 15 20% of enamel (Hillson and Bond 1997). Studies indicate that the frequency of hypoplastic defects is greater in the anterior dentition (Condon and Rose 1992a; Goodman and Armelagos 1985). This fact, in combination with the percentage of hidden enamel in the posterior dentition, is the reason that many studies using enamel hypoplasi as concentrate on the anterior dentition (Martin et al. 2008; Reid and Dean 2000; Santos and Coimbra 1999). Goodman and Rose (1990) recommend that epidemiological studies focus on the maxillary central incisors and mandibular canines given that they ar e the teeth most sensitive to enamel defects; alternatively Wright (1997) recommends considering
52 each tooth independently and using the posterior dentition as a marker of stress severity. In consideration of these recommendations, individuals selected fo r inclusion in this study were not removed from the sample if posterior teeth were missing. When enamel defects on the posterior dentition were present, they were considered supporting evidence to the defects in the anterior dentition. The anterior denti tion provides a record of chronological growth disruption between birth and seven years of age, the period during which enamel forms (Goodman et al. 1980). Studies have found that the timing of enamel formation is more closely associated with chronologic al age than most other skeletal elements (White 2000). While there is population variation in enamel formation (Reid and Dean 2006), it is possible to age hypoplasias based on the position of the defect on the tooth crown with a fair degree of accuracy Amelogenesis begins at the occlusal tip of the tooth; however, growth does not occur in a simple linear fashion (Hillson and Bond 1997). Research into dental development has produced methods for estimating ages (i.e., aging) of hypoplasia formation (Co ndon and Rose 1992b; Goodman et al. 1980; Goodman and Rose 1990; Reid and Dean 2000). The ability to accurately age a growth disturbance provides an additional line of reference for investigating the effects of growth disruption on adult morphology. H arris Lines Harris lines, also known as radiopaque transverse lines, are areas where trabeculae are deposited in a transverse plane perpendicular to the diaphysis of a long bone as a result of disruption in normal metabolic function during growth. The term Harris line is a nod to one of the early researchers of radiopaque transverse line formation (Park 1964). Harris lines can be seen through radiographs (see Figure 31), and medical researchers have been able to document their occurrence in association with known episodes of stress such as disease, nutritional
53 deficiencies, and even psychological stress (Mays 1995). For the bio archaeologist, Harris lines are often used to document changes in health over time. An increased frequency of Harris lines may indicate inadequate nutrition or increased levels of disease or infection in a population (Larsen 2002). As an indicator of st ress, Harris lines are useful in that the presence of such lines in an adult indicate that some form of growth disruption occurred. However, individual variation in the formation of such lines and cortical remodeling mean that researchers need to be cauti ous when using these indicators to describe the health insults an individual has experienced. Understanding the formation of Harris lines is important for understanding the visual presentation of a disruption in skeletal growth. At the growth plate, the cartilage matrix is oriented along the longitudinal axis in which the bone is growing. The orientation of cartilage serves as the template for subsequent trabeculae formation (Martin et al. 1998) Park (1964) uses an analogy of a dam and pond to describ e how Harris lines are formed through growth disruption. When growth disruption occurs, the epiphyseal cartilage becomes a thin atrophic layer that serves as a dam behind which bo ne is laid down in along the transverse plane Once growth resumes, the os teoblastic activity is temporarily blocked by the dam It is the recovery that causes the transverse area to thicken creating a radiopaque line in the bone. Without growth disruption, trabeculae are oriented longitudinally. Based on what is known of thi s process, the relative thickness of a Harris line can be attributed more to the growth arrest recovery than to the growth arrest itself (Park 1964). The growth arrest event is only important in that it creates the dam. Periods of intense growth without a prior growth disruption do not produce such lines, but the line seen at epiphyseal fusion is equated to growth disruption without subsequent recovery (Park 1964).
54 What we know of the process of Harris line formation allows us to more correctly interpr et differences in what appears to be the relative severity of Harris lines. A thick, pronounced Harris line compared to a thin Harris line would tell you less about the disruption event and more about the recovery phase (and/or subsequent cortical remod eling). Through the process of resorption and cortical remodeling, thicker lines are more likely to persist than thin lines. The number or relative frequency of Harris lines in an individual may be better evidence that said individual experienced more gr owth disruption events, but interpretations need be cautious due to complicating factors, to be discussed below. Park (1964) found that growth disruption needed to be complete, or near complete, for line formation to occur. The severity of insult needed for line formation is best described by Garn et al. (1968), who were able to compare radiographs with longitudinal data from the Fels Research Institute. Because the participants in this research program were subject to regular health reviews, radiographi c data could be compared to health data to determine the cause of specific Harris lines and record the association between traumatic events and the formation of lines. Garn et al. (1968) found a statistically significant but low -order association between health insults and the formation of lines. In ten percent of cases where new lines formed, they could not be explained by any insult recorded in the health data. In other cases, lines were associated with illnesses such as whooping cough, chickenpox, min or surgeries, and, in one interesting case, with a routine small pox vaccination (Garn et al. 1968: 73), which would lead one to believe that even relatively minor insults have the ability to produce Harris lines. Therefore, it is hard to make assumptio ns about the types of stress which led to the formation of a Harris line, particularly in individuals without medical records.
55 There appears to be a range of individual variability with regard to the formation of Harris lines: some individuals seem to be more prone to forming lines than others. Garn et al. (1968) found more Harris lines in juvenile males than females, possibly a product of greater male susceptibility to the negative effects of physiological stress. But, female lines persisted into adult hood at a higher rate than those lines found in males (Hummert and Van Gerven, 1985). These points need to be considered when comparing differential effects of stress in males and females. Unlike teeth, bone continues to remodel over the course of an ind ividuals lifetime. As a result, Park stated that Harris lines were only occasionally encountered in the bones of adults (Park 1964). Moreover, Park ( 1964) believed that lines formed later in the growth phase would be more likely to persist into adult hood. Subsequent studies on the persistence of Harris lines have changed this view; Garn (1968) found that Harris lines formed in early childhood were more likely to persist than those created later in the growth phase. A more recent study found that Har ris lines were most commonly formed in the first year of life and in adolescence, both periods of rapid skeletal growth (Alfonso et al. 2005). As a product of cortical remodeling, older individuals are less likely than younger individuals to have visible Harris lines (Grolleau Raoux et al. 1997). Therefore the ability to determine any associated effects of Harris line formation are likely clouded as a product of advanced age. By understanding these limitations, researchers are better able to interpret the results of studies. Harris lines represent a metabolic disruption which affects the longitudinal growth of bone and the subsequent growth recovery. As such, researchers have attempted to correlate long bone length with the presence of Harris lines. The results of these studies have provided conflicting results. Goodman and Clarke (1981) compared the Harris lines in tibiae (based on
56 presence or absence) with tibial length. For females and a combined sex sample, tibiae with Harris lines were signific antly longer than those without. A more recent study found no significant difference in the length of long bones with and without Harris lines (Nowak and Piontek 2002). These conflicting results are no clearer in published studies of juvenile morphology and Harris lines. Blanco et al. (1974) found living children with Harris lines were significantly shorter than those without; however Mays (1995) found no relation between femur length and the presence of Harris lines. Obviously the results of these stud ies have not answered the question of the relationship between how growth disruptions affect bone length. Likely complicating the scenario is the fact that the density of a Harris line is not determined by the severity of insult so much as the quality of recovery. Because the thickest lines are likely to persist the longest in light of cortical remodeling, Harris lines present in adults may not represent those with the greatest insults, but instead those with the greatest recovery. Comparisons of Harri s lines with other indicators of growth disruption have also shed light on how Harris lines form. Linear enamel hypoplasias, like Harris lines, are a non -specific indicator of stress that can be aged in terms of when they formed in an individual. Compari sons of linear enamel hypoplasias and Harris lines do not suggest a one to -one relationship. Alfonso et al ( 2005) found that in the archaeological populations they studied the rates of linear enamel hypoplasias peaked between the ages of three to five On the other hand Harris lines we re most commonly formed in the first year and during adolescence. Weaning is known to be a period of stress and the enamel hypoplasias data reflect this stress. However, the ages of peak Harris line formation appear to be more strongly associated with phases of peak bone growth more so than with periods of stress as indicated by the linear enamel hypoplasias And, while Harris lines have been positively associated with stress events in some cases (Garn et al. 1968), t he data
57 presented by Alfonso et al. (2005), among others, suggest that Harris lines are not nearly as reliable an indicator of stress. Garn et al. (1968) made the important discovery that Harris lines maintained dimensional stability (Garn et al., 1968 : 72). Simply put, the process of bone remodeling did not cause the lines to migrate along the bone shaft. Where they were laid down, they remained. This was an important discovery, because it laid the groundwork for later studies which have allowed rese archers to age Harris lines (Byers 1991; Hunt and Hatch 1981). Now, with a degree of reliability, x rays allow us to determine when a person experienced the metabolic insult which produced a specific Harris line. Once researchers were able to accuratel y age Harris lines, they were then able to determine the periods during growth which lines were most likely to form. Park (1964), whose research focused on juveniles, assumed that the process of remodeling meant that the lines seen in adult individuals we re formed late in the growth phase and therefore had less time to undergo remodeling. Garn et al. (1968) reached a different conclusion. By measuring the distance of a line from the epiphysis, he concluded that the most persistent lines are formed in inf ancy or even in fetal life. There are obviously many variables that need to be considered when using Harris lines as an indicator of growth disruption. But, regardless of the variability in formation and obliteration due to resorption, they provide a li ne of evidence for growth disruption that is not otherwise available, and one that can be associated with certain phases of growth. Therefore, the Harris lines can be a useful tool in osteological analysis when used with an understanding of how they are f ormed and resorbed.
58 Other Skeletal Indicators of Growth Disruption Although enamel defects and Harris lines may be the most well studied indicators of growth disruption, they are certainly not the only evidence available in the skeleton. For example, in anthropometric studies of living children, head circumference is used as an indicator of substandard growth and development. Stoch and Smythe (1976) found significant differences in head circumference between a control group and those undernourished during infancy, the period during which head growth is most rapid. One longitudinal study on head circumference found that catch up occurred in the majority of cases, but that small head circumference persists into adulthood for many individuals (Brandt et al. 2003). Therefore, in an adult skeletal sample, the growth environment for those with average head circumferences cannot be assumed. Clark et al. (1986) suggested that vertebral neural canals could be used as an indicator of disrupted growth. Because ve rtebral neural canals cease growth in early childhood while the bodies of the vertebrae continue to grow, the relationship of these two measurements can indicate early childhood growth disruption. Small neural canals with normal vertebral body sizes indic ate growth disruption in early childhood followed by a period of catch-up growth. If both measurements are small, it would be an indicator of chronic stress (Clark 1988). Rewekant (2001) built on this study by observing differences in a vertebral canal i ndex (sagittal diameter/transverse diameter) in two archaeological populations of divergent socio-economic statuses. Smaller vertebral canal indices were more often associated with the more unhealthy population; however, statistically significant differe nces were only found in the male sample. Rewekant (2001) does not provide data that would inform whether smaller vertebral canal indices are tied to smaller sagittal vertebral canal diameters or larger transverse diameters in the group of lower socio -econ omic status. Clark (1988) suggests that the transverse diameter of vertebral neural canals complete growth later than the sagittal diameter and that the relationship
59 of these variables can thereby indicate whether growth disruption occurred early or late in growth, but he does not suggest that a particular vertebral neural canal index is indicative of anything in particular. At a historic Maya site where approximately half of the skeletal sample had active or remodeled porotic hyperostosis and/or cribra or bitalia, researchers suggest that anemia may have some effect on adult stature. Those individual who displayed the porous lesions associated with anemia were found to be approximately 0.5 c entimeters shorter on average than those who did not. However, t he results were not found to have statistical significance (Cohen et al. 1997). Although Garn et al. (1968) discovered that Harris lines could be associated with a recovery from chronic or acute anemia, in adults the anemia which caused the lesion could have occurred during a phase of life before or after the completion of growth, whether or not the lesion is active or healed. While the relationship between stature and anemia during the growth phase is an interesting one, the evidence available in an adu lt skeleton does not provide the information necessary for inclusion in this study. In addition, the number of individuals sampled with evidence of anemia was too small to be statistically significant. Vertebral neural canal size and head circumference ar e both, in part, a product of overall body size; as such, assumptions of independence are violated in any statistical interpretation of how these variables are related to overall body size. To test how meaningful these variables are as indicators of growt h disruption, they are compared to linear enamel hypoplasias and Harris lines in Chapter 8
60 Table 3 1. Mean estimates for ages of enamel formation Age in years a Upper dentition I1 1.1 4.2 I 2 1.8 4.8 C 1.7 4.8 P 3 2.0 6.0 P4 2.5 6.0 M 1 1.1 3.0 M 2 4.4 6.4 M 3 9.3 11.3 Lower Dentition I 1 0.8 3.4 I 2 1.0 3.8 C 1.4 5.2 P 3 1.0 6.0 P 4 2.0 7.0 M 1 1.0 3.1 M 2 4.2 6.2 M 3 9.3 11.3 a All estimates based on Reid and Dean (2006) except for the premolars which are based on Goodman et al. (1980).
61 Figure 3 1. Radiograph of a tibia with a Harris line
62 CHAPTER 4 MATERIALS AND METHOD S The purpose of this chapter is to introduce the materials and methods used to address the hypotheses of this study. The materials section describes the collections from which data was obtained as well as the demographic composition of the research sampl e. The methods section outlines the techniques used to obtain measurements and stress data on the skeletal material as well as the statistical procedures used to analyze this data. Materials To address the hypotheses outlined, it was necessary to obta in data from a skeletal series of adult males and females, both with and without skeletal indicators of growth disruption. Data was collected from the Terry Collection at the Smithsonian Institution and the HamannTodd collection at the Cleveland Museum of Natural History. The Terry and Hamann Todd collections were compiled from anatomical specimens in the St. Louis, MO and Cleveland, OH areas respectively. In both cases, the majority of the cadavers were unclaimed bodies in morgues or hospitals or those that were given to the state by family members (Hunt and Albanese 2005). Comparisons of the Terry and HamannTodd collections conclude that the two populations are osteometrically similar (Isan 1990). The nature of this collection provides several a dvantages for this study. First, the combined number of individuals found in these collections insures a large sample size for testing the hypotheses of this project. Second, morgue records allow for reasonable control over ancestry, age, secular trends and any other extenuating circumstances that the historical records may illuminate. For inclusion in this study, all individuals must be of like ancestry. This criterion attempts to control for the genetic component of human morphology. It is recognized t hat there are
63 differences in average stature, growth rates, sexual dimorphism, and proportions in different populations (Eveleth 1975, 1986; Eveleth and Tanner 1990). For example, Hamill et al. (1973) found that European Americans tend to have longer tr unks and African Americans tend to have longer lower limb s despite socioeconomic conditions. As such, ancestry cannot be ignored in any investigation of growth disruption. Both the Terry and the HamannTodd collections divide their skeletal collection s into white and black samples based on what they term ethnic origins or ethnicity respectively. Today, race is recognized as a cultural construct rather than a biological reality by anthropologists and it is important to recognize problems with makin g such population distinctions; most notably, that the standards for making such distinctions are not always uniform (AAPA, 1996; Cartmill, 1998) The white individuals in these collections are presumably Americans of European descent. While there is a great degree of variability within Europe, the situation becomes even more complicated when considering the ancestry of black Americans. The supposition is that these individuals are of African descent, but Africa has the highest degree of genetic dive rsity of any continent (Jorde et al, 2000) In addition, there is a high degree of genetic admixture in African -American populations. Historical records confirm that nearly all slaves brought to North America were from the coast and interior of western A frica although a few were from Mozambique or Madagascar. Unlike the Portuguese and Dutch, English slave traders did not have ties with specific ports and purchased slaves from Senegal to Angola. The Dutch traders concentrated on the ports in Angola and i t is estimated that a quarter of slaves in the colonies originated in this area of Africa (Wright 1990). Slave routes changed over the course of time and because slaves were treated as chattel rather than people, genealogical information is not available for early African -Americans. Recent studies of modern African -
64 American populations suggest an average of 20.9% European ancestry and 2.7% Native American ancestry, although these numbers vary geographically (Reiner et al. 2005). Furthermore, according to the rule of hypodescent (the one drop rule) which has been standard practice throughout most of American history, a child of a mixed union would be racially classified as belonging to the race of the parent that is considered to be socially inferior Therefore, an individual might be classified black even if the representation of African ancestry is minimal. Despite the inherent problems with any label of ancestry, these collections provide the best resource available for controlling genetic vari ation within my sample. To knowingly ignore such labels would be irresponsible given what is known about variations in human growth across populations (Eveleth and Tanner 1990; Silventoinen, 2003). Due to the paucity of white females in the two collecti ons, this research focused on those individuals classified as black, despite the diversity assumed with this label. The Terry collection houses 1,728 individuals and the Hamann Todd collection houses 3,100 human skeletons. To streamline the data collec tion procedure, the written records associated with each collection were used to establish ancestry thereby determining which individuals to pull for analysis. A second criterion for inclusion in this study is age at death. All individuals must be adults with united epiphyses to insure that maximum height was attained prior to death. However, it is important that there is minimal, if any, osteoporotic bone loss in the vertebrae. Bone mass peaks around the age of 18 years, and, shortly thereafter, the tra becular structure begins to diminish. The age -related changes become particularly pronounced after the age of 50 years although the relative severity of such changes is related to other factors such as sex and mechanical stress (Agarwal et al. 2004; Atki nson 1967; D'Ippolito et al. 1999; Larsen, 1997). Age was
65 determined through autopsy records. The youngest individual in this sample was a female who had reached the age of 19 year s (and full epiphyseal closure) before death and the oldest individual w as 35 years of age (see Table 4 1). By limiting the sample to a maximum age of thirty five I hope to avoid osteoporotic fracturing that may not be macroscopically visible. While the individuals in the Terry and Hamann Todd collections were indigents, li kely exposed to heavy labor, African -Americans are less susceptible to osteoporotic fracturing than their neighbors of European descent ( Mensforth and Latimerm 1989; Robling and Stout 2000). Unfortunately, vertebral body heights continue to grow slightly through early adulthood (Clark 1988). In a population exposed to manual labor, an individual is likely accumulating microfractures in their vertebrae before they finish growth. Ind ividuals with macroscopically visible pathologies that would affect height were excluded from the sample. The presence of dentition is a third consideration for inclusion in this study. Enamel defects are an important line of evidence for addressing the hypotheses outlined. Edentulous individuals were immediately excluded from analysis. Very few individuals retained all of their teeth, so it was necessary to determine, on a case by-case basis, whether the dentition provided enough data for analysis. The individual with the lowest total number of teeth (N=16) had one of each tooth type (e.g. central maxillar y incisor, lateral mandibular incisor) in the anterior dentition. The presence of posterior dentition was considered to be less important for inclusion in the study for two reasons: 1) the posterior dentition are not as susceptible to developing enamel de fects as the anterior dentition ( Condon and Rose 1992a; Goodman and Armelagos 1985), and 2) the amount of hidden appositional enamel means that they provide less information in macroscopic observations of surface enamel (Hillson and Bond 1997).
66 An addit ional advantage in using anatomical collections, like the Terry and Hamann Todd collections, is that it is possible to control for secular trends due to the availability of year -of birth and year -of -death data. The dates of birth for individuals in this s ample range from 1885 to 1934. Secular changes in morphology have been documented in these collections (Jantz and Meadows Jantz 2000; Meadows Jantz and Jantz 1999); therefore trends found in the results will be compared to the date of birth data to ma ke sure that the results are not potentially confounded by the secular changes (see Table 4 1). Tuberculosis was the leading cause of death among these individuals (42.6%). The use of antibiotics to effectively treat human diseases did not become a widesp read practice until the 1950s (Lederberg 2000), after the deaths of most individuals analyzed in this study. As a result, many of these individuals died from diseases that are treated with antibiotics today and likely suffered a higher infectious load d uring growth than the average child born today. Methods Once individuals were chosen for analysis, a quick survey of each skeleton was conducted to determine if the skeletal biology of the individual was consistent with the records and if all elements ap peared to be from the same individual. It was not necessary that each bone antimere was present, but one of each bone had to be available for measurement. For example, if a left and right tibia were not available, the individual was not excluded from ana lysis, but one tibia was necessary for inclusion. Individuals with pathologies that affected skeletal height were immediately excluded. Most of excluded individuals had pathological curvature of the spine or long bones, or exhibited severe arthritic lipp ing and/or compression of the vertebral bodies. Other pathologies were recorded for each individual. The teeth were examined to determine whether there was enough evidence available to make a statement regarding the presence or absence of enamel defects. This is the phase of analysis where the majority of potential
67 candidates were excluded for either lack of teeth or questionable association of the available teeth. Measurements Each individual was measured to obtain estimates of stature and proportional indices. Cranial measurements and vertebral neural canal measurements were also taken to see how they relate to stress. The measurements collected for each individual are summarized in Table 4 2 and include the maximum and bicondylar length of the femur, maximum tibial length, maximum humeral length, and maximum length of the radius measurements standard in any postcranial osetometric analysis. In addition, the height of the cranium (basion bregma), maximum height of the vertebral bodies (C2-L5), ant erior height of S1, and the articulated height of the calcaneus and talus were included so that skeletal height (SKH) and an estimated living stature could be calculated using the revised Fully technique (Raxter et al. 2006). While reliable regression fo rmulae are available for calculating estimated stature based on long bone measurements, they do not provide independence when individual bone lengths are compared to estimated stature as a means of determining proportional differences (see a list of propor tional indices in Table 4 3). In anthropomorphic studies, sitting height to standing height ratios are used as a measure of proportional differences. In this study, skeletal trunk height (STH), as defined by Holliday (1997), will be included to estimate relative trunk height to stature proportions. Skeletal trunk height is found by summing dorsal body heights of the thoracic vertebrae, lumbar vertebrae, and the ventral height of the sacrum. The femur, tibia and talus -calcaneus measurements were made usi ng a standard osteometric board. Head circumference from the Terry collection was obtained by using a tape measure placed at opisthocranion and held just above the brow ridge in a manner simulating the anthropometric techniques used in children (Zerran, 2 007). Head circumference measurements
68 from the Hamann Todd collection were acquired by referencing the autopsy records from the collections database. Cranial height, length and breadth were measured with spreading calipers and all remaining measurements were made with sliding calipers. Autopsies were performed on the individuals in both the Terry and HamannTodd collections; consequently, most individuals had their calotte removed or, in some cases, the skull was bisected sagittally. Skulls in the Hama nnTodd collection were wired into approximate anatomical alignment, but skulls in the Terry collection were articulated using wax prior to being measured. With the exception of head circumference and talus -calcaneus height, all measurements were collected using the landmarks and techniques described by Buikstra and Ubelaker (1994). For the sided elements, both sides were measured and an average was used for the calculation of stature. Linear Enamel Defects Surface defects in the enamel were recorded using the dental data collection form ( Figure 4 1 ). Enamel defects were observed macroscopically and recorded when the defect could be felt by rubbing a fingernail across the surface of the tooth. Because many of the teeth in these collections were glued into the alveolar sockets, a 10X hand loupe was used to insure that the enamel surface was not compromised. Because so much of the enamel in molars is unobservable without sectioning and the majority of hypopla sias are found on the incisors and canines, hypoplastic analyses in this study concentrate primarily on the anterior dentition (Condon and Rose 1992b; Goodman and Armelagos 1985; Hillson 1996; Steckel et al. 2006), although all defects were recorded wh ere they were encountered. Each tooth was assigned an LEH code based on the standards outlined in The Global History of Health Project: Data Collection Codebook (Steckel et al. 2006). By these guidelines, teeth are ranked from zero to three based on the relative number of hypoplastic lines observable. A description of how these codes are assigned can be found in Figure 4 1. Hypoplastic lesions
69 can be described as lines, bands or pits, and each defect was documented accordingly. Dental caries and opacit ies were also recorded for each tooth. For each developmental defect in tooth enamel, the distance from the cemento -enamel junction was recorded (Figure 4 2). For wide band defects, distance was measured from the near, middle and occlusal edge of the defe ct. These measurements were used to find the age at which the defect was formed using the methodology provided by Goodman et al. (19 8 0). This method was chosen over the Read and Dean (2000) method because the latter relies on crown heights for age estima tions. In these samples, numerous teeth were worn or broken so that accurate crown heights could not be obtained. Ritzman et al. (2008) and Martin et al. (2008) compared these techniques and found the largest difference to be in that the Goodman et al. (1980) method consistently underestimates age at formation for the earliest forming hypoplasias. While the difference is statistically significant, it is only a difference of 1 to 4 months a time frame that is not meaningful for the purposes of this ana lysis (Martin et al. 2008). For this analysis, hypoplasias were broadly divided into two categories: those that formed in infancy (during the first three years of life), and those that formed during childhood (approximately three to seven years). When a defect was estimated to have been formed at approximately th re e years of age, that defect was placed in the child rather than the infant category. Hypoplastic pits were not assigned an age estimate given that pit defects do not form along the imbrica tional line associated with their location on the crown surface; as such, they cannot be aged correctly based on macroscopic analysis (Hillson and Bond, 1997). Bands were aged based on the location of the occlusal edge of the defect using the standards o utlined in Buikstra and Ubelaker (1994). Dental defects were coded in a number of ways for data analysis. Tooth stress (TS) was recorded based on presence or absence (scored zero or one) of any enamel defect in an
70 individual. This coding scheme does not differentiate between lines, bands or pits. Nor does it take into consideration whether or not a dental defect is bilateral or on the anterior or posterior dentition. To further differentiate between individuals with different types of hypoplastic lesion s, the linear enamel hypoplasias (LEH) stress code was created. If an individual has hypoplastic pitting, in the absence of any linear enamel defect, they are removed from analysis using the LEH score. This criterion is used, in part, to satisfy those in dividuals who do not believe that pitting should be scored as a hypoplastic lesion. A more conservative scoring procedure (CBTS) was used in which an individual was considered stressed only when an individual had bilateral enamel defects on their anteri or dentition. Many of the individuals deemed stressed based on TS criteria and unstressed by CBTS criteria were removed from further analysis. However, eleven individuals were coded differently depending on the criteria used. Because the CBTS scorin g criterion is the most conservative scoring technique employed in this analysis, it is considered to be the standard. A summary of these coding schemes can be found in Table 4 4. In an attempt to distinguish between chronic and acute stress, an estimate of the minimum number of stress events (MNSE) was calculated for each individual. The MNSE score was based on the number of hypoplastic defects in a single tooth and then the anterior teeth were compared using the calculated age estimates for enamel defect formation. For example, if a single defect was present on all the maxillary, anterior teeth, but age ranges suggest they did not pertain to the same stress event, then the MNSE score would be equivalent to the number of age ranges represented. This min imum number was based on defects found on canines and incisors. If a defect was not bilateral, it was excluded from analysis due to the fact that trauma can cause lines on teeth that are not attributed to systemic stress (Hillson 1996). If the antimere was not
71 present, a tooth was tentatively retained for analysis and evidence from other teeth (including posterior dentition) determined whether or not it would be included in the analysis. Data were also collected summing the number of lines for each indi vidual based on CBTS criteria as well as the number of teeth affected in each individual. Harris Lines Radiographs were taken of the distal tibiae to score observable radiopaque transverse lines (also known as Harris lines). For inclusion, lines had to be visible to the naked eye and extend half -way across the bone (standards outlined by Garn et al. (1968)). The distal tibiae are used for this type of analysis because they have been shown to exhibit a relatively high frequency of Harris lines when compared to other skeletal elements (Garn et al. 1968; Park, 1964). Because of this, most Harris line research is performed on tibiae, including methods for estimati ng age at formation (Byers 1991; Garn et al. 1968; Goodman and Clark, 1981; Hummert and Van Ge rven, 1985). For consistency, the left tibia was x rayed unless it was unavailable or determined to be the inferior choice due to reactive bone formation or postmortem cracks in the bone. Individuals from the Terry collection were radiographed using 35 x 43 cm mammography film where approximately the distal three quarters of five tibiae could be placed on a single film. The x -ray equipment used while collecting data at the Cleveland Museum of Natural History was a Faxitron cabinet unit with internal dim ensions of approximately 16 x 18.25 inches. The unit has a central projection and uses 10 x 12 inch film. This meant that the medial malleolus was only visible on the film of the smallest individuals, although most tibiae could be oriented within the cab inet in such a way to film the entire length of the diaphysis. Byers (1991) technique for estimating the age at which a Harris line formed requires a measurement from the distal end of the bone. To correct for this problem, foil was taped to each bone at a measurement determined
72 to be the middle of bone length. Harris lines were then measured from the center point and the distance from the distal end was then calculated. Byers (1991) technique for age estimation was used because all individuals in this study are adult s. Unfortunately, Byers (1991) figures suggest that his calculations rely on a tibial length that includes the intercondylar tubercles. The standard osteometric technique for measuring the length of the tibia is to measure the distance from the articular surface of the lateral condy le to the medial malleolus a measurement that excludes the intercondylar eminences (Bass 1995; Buikstra and Ubelaker 1994). Figure 43 shows the difference between these measurements as well as the method for calculating the timing of Harris line form ation Different techniques for measuring the tibia have caused problem s in the past (see Jantz et al., 1994). Because tibial length was collected using the standard technique in this study, a correction of 2.46mm was added to the tibia length before Bye rs age calculations were performed. This correction was the difference found when measuring the tibia by including and excluding the intercondylar eminences of black individuals from the Terry collection (Waxenbaum et al. 2006). Due to equipment failure and collection error, radiographs were obtained for only 183 of the 204 individuals used in this study. Those 183 individuals were divided based on the presence or absence of Harris lines. An estimated age at formation was calculated to determine whethe r or not the age at which a stress event occurs has any effect on adult stature and proportions. Statistical Analyses Linear enamel hypoplasias and Harris lines are the two primary markers of growth disruption in this study; as such, tests of independenc e are necessary to evaluate the relationship between these variables. In this case, it was important to determine whether Harris lines and linear enamel hypoplasias are recording the same stress events. To address the question of
73 independence, a two tailed, two -by -two contingency table (a Fishers Exact test) was created using data on the presence or absence of linear enamel hypoplasias and Harris lines. Stress markers allow for the data to be grouped into two samples: those with and without a particular stress marker. Analysis of variance (ANOVA) was the principal analytical approach employed to test whether or not dimensional variables are statistically different between stressed and unstressed groups. ANOVAs were also used to compare osteometric data with meristic data on age at formation for both Harris lines and linear enamel hypoplasias. Subsequent analyses of age at formation correlate adult metrics and proportional indices with age data to examine whether the age of the inferred stressor has an effect on adult form. An ANOVA is appropriate only where assumptions of normality and homoscedasticity are met. Shapiro Wilks test for normality and Levines test of error variance were used prior to performing any ANOVA. Where the assumptions of norm ality and homoscedasticity were not met, the nonparametric Mann -Whitney U test was performed. All descriptive statistics, ANOVAs, and Mann -Whitney U -tests were performed using SPSS Statistics Grad Pack 17.0. The results of an ANOVA were considered to be significant at the p < 0.05 level. When results were found to be near significant (0.05 < p < 0.10), a resampling procedure, using Resampling Stats 5.0.2 software, was employed as an independent test. In these tests, the populations being compared were first pooled, and then resampled with replacement 10,000 times according to the original sample sizes. For each resampling, a mean difference was calculated. Probabilities were determined by how many times the resampling procedure yield ed a mean differe nce greater to or equal to the original difference. Resampling statistics provides a reliable, conservative method for obtaining probabilities irrespective of distributional assumptions (Simon, 2000).
74 Two -way factorial analyses of variance were used to co mpare osteometric data of the stressed and unstressed population broken down by males and females. It is understood that statistically significant differences in osteometric data are found between males and females, but with a two -way factorial analysis o f variance it is possible to observe any interactions between the variables of sex and stress. This makes it possible to determine if the effect of growth disruption affects males more than females. To address the question of whether or not proportiona l indices (listed in Table 4 3) expressed as ratios, differ in relation to stress, it was first necessary to apply Pearsons correlation to determine whether or not the components of these ratios are independent. These correlations were expected to be si gnificant because long bone length is related to overall body size; therefore, to qualify the results of ANOVAs between stressed and unstressed groups it was necessary to analyze differences in scaling. Reduced major axis (RMA) regressions were conducted on the logtransformed components of ratios in stressed, unstressed and a pooled sample using RMA: Software for Reduced Major Axis Regression (v. 1.17) (Bohonak, 2004). Reduced major axis regressions were chosen over least squares regressions due to the f act that error is assumed in both the X and Y variables. Using reduced major axis regressions to interpret shape has recently been supported by Smith (2009). Using the correlation coefficient from the regressions found using the Bohonak software, RMA2 software (Cole 1997) was then used to test whether or not the scaling differences between groups were statistically significant. Where RMA slopes were not different, intercept differences were tested by fitting a regression line through the pooled dataset and counting the number of stressed and unstressed individuals that fall above and below the line. This information was then placed in a 2x2 contingency table and Fishers exact test was used to test the difference (Tsutakawa and Hewett 1977).
75 The degree of sexual dimorphism in each group was found using the formula ln X Mln X F for each measurement and proportion, where X M is the male mean and X F is the female mean. The decision to use this method of calculating a score of dimorphism was based on Smith (1999), who demonstrated that compared to other methods for finding dimorphism ratios, this method is believed to have fewer problematic statistica l properties. A dimorphism score is a valuable tool for measuring the value of dimorphism between populations, but, given that the score offers no measure of dispersion, traditional statistical techniques cannot be used to determine whether or not there is a significant difference in sexual dimorphism between stressed and unstressed populations. To address that question, a bootstrap procedure was employed using Resampling Stats 5.0.2 software. The null hypothesis was that no difference in sexual dimorph ism exists between stressed and unstressed groups; the alternative (one -tailed) hypothesis is that sexual dimorphism is greater in unstressed groups. The stressed and unstressed, male and female populations were pooled and resampled with replacement using the original sample sizes. In each resampling, the log difference between male and female means was calculated for the stressed and unstressed groups. A difference in dimorphism was then found by subtracting the stressed measure of dimorphism from the u nstressed sample. Probabilities were determined by counting the number of times the resampling mean difference was greater or equal to the empirical measure of difference in sexual dimorphism which was calculated by subtracting the original dimorphism score in the stressed sample from the unstressed score. Each resampling procedure was run with 1,000 iterations except where probabilities approached significance. In that instance, the test was run with 10,000 iterations. Because the alternative hypothesi s was directional, the probability was calculated as 1 p where the empirical difference in dimorphism was negative. In these cases, the
76 original alternative hypothesis is not tested; instead, 1 p tests the alternative hypothesis that sexual dimorphism is greater in the stressed group.
77 Table 4 1. Description of sample population N Age at death Year of birth Mean Min. Max S.D. Mean Min. Max S.D. Male 107 28.52 20 35 4.342 1902.17 1885 1928 7.167 Female 97 27.31 19 35 4.251 1904.62 1888 1934 6.826 Total 204 27.95 19 35 4.331 1903.33 1885 1934 7.097 Table 4 2. Skeletal measurements recorded Skeletal element Measurements Cranium Cranial height (br ba) Maximum cranial length (gl op) Maximum cranial breadth (eu eu) Head circumference Cervical vertebrae (C2 C7) Maximum anterior body height Thoracic vertebrae (T1 T12) Maximum anterior body height Maximum posterior body height Mediolateral vertebral neural canal diameter Anteroposterior vertebral neural canal diameter Lumbar vertebrae (L1 L5) Maximum anterior body height Maximum posterior body height Mediolateral vertebral neural canal diameter Anteroposterior vertebral neural canal diameter Sacrum Maximum anterior height of S1 Femur Maximum length (right) Maximum length (left) Bicondylar length (right) Bicondylar length (left) Tibia Length (right) Length (left) Fibula Maximum length (right) Maximum length (left) Talus calcaneus Height (right) Height (left) Humerus Maximum length (right) Maximum length (left) Radius Maximum length (right) Maximum length (left) Ulna Maximum length (right) Maximum length (left) Physiological length (right) Physiological length (left) Clavicle Maximum length (right) Maximum length (left)
78 Table 4 3. Description of proportional indices Index Calculation a Crural (tibial leng th / max. femoral length) 100 Humerofemoral (max. humeral length / max. femoral length) 100 Radiohumeral (max. radi al length / max. humeral length) 100 Sitting Height (skeletal trunk height / skeletal height) 100 Intermembral ((max. humeral length + max. radial length) / ( max. femoral length + tibial length)) 100 a the average of left and right measurements were used when both were available. Table 4 4. Coding dental enamel defects Code Criteria CBTS a Individuals are scored as stressed (1) only when defects are on the anterior dentition and they are found bilaterally. Otherwise, individuals are scored as unstressed (0). TS Individuals are scored based on the presence (1) or absence (0) of any enamel defect. This is the most lenient of all the scoring procedures. LEH Individuals are scored based on the presence (1) or absence (0) of linear enamel defect(s). Hypoplastic pitting is not scored. a The CBTS scoring procedure is considered to be the standard for this analysis. Figure 4 1. Dental d ata c ollection f orm
79 Figure 4 2. Measuring hypoplastic teeth Figure 4 3. Aging radiopaque transverse lines. T = total length of the tibia as defined by Byers (1991); S = the standard technique for measuring tibial length; D = the distance from the Harris line to the distal end of the bone. Age at formation is calculated by using the formula 1.15(T 2.33D) X 100/T. Tables in Byers (1991) allow for conversion of that percentage to an age range.
80 CHAPTER 5 RESULTS: STATURE AND STRESS Of the 204 individuals analyzed in this study, radiographs were available for 183 of them. A G -test of independence was performed to determine whether or not Harris lines and hypoplastic defects should be considered independent variables. The G -test was chosen because it is consid ered to be a Model I design where individuals are grouped by common experiences that represent treatment effects (Sokal and Rohlf 1995). The results of this test allow us to accept the null hypothesis that Harris lines and hypoplastic defects are independent (G = 0.867, p = 0.352). The same result is found when the population if divided by sex (female G = 0.733, p = 0.392; male G = 0.206, p = 0.65). The osteometric variables were tested for normality and homoscedasticity to determine the appropriate statistical method for analysis. The osteometric variables in question proved to be normally distributed (see Table 5 1). However, the individual measurements of the summed variables (SKH and STH) were not normally distributed in every case. This was particularly true of the dorsal height measurements of the thoracic and lumbar vertebrae which were used to find skeletal trunk height. Stature and Enamel Defects The following analyses describe difference in stature, between those individuals with and without skeletal indicators of growth disruption. Stature in skeletal populations is estimated by various means; consequently, skeletal height (SKH), stature (as calculated using the revised Fully technique without age), sitting height (STH) and the long bone measurements most commonly used to estimate stature (maximum femur and tibia lengths) were all considered. For the purposes of this study, an individual is considered to be stressed if they exhibit bilateral enamel defects in the anterior dentition (CBTS). All stature estimates exhibited
81 homoscedasticity, as determined by Levines test of homogeneity. One -way analyses of variance (ANOVAs) were performed to determine whether there were significant differences in stature estimates between stressed and unstressed groups. No significant differences between groups were found in the male sample; however, females had near significant ( 0.05 < p < 0.10) differences for all measures (Table 5 2). Resampling statistics were used as an independent test of near significant results of ANOVAs. The results of these tests were significant for every case: SKH ( p = 0.0406), Fully ( p = 0.0414), ST H ( p = 0.0296), maximum femur ( p = 0.0466) and tibia (p = 0.0271). Unstressed groups were larger than stressed groups for all measurements among females. As an exemplar, b oxplots of the differences in tibia length can be seen in Figure 5 1. There are man y methods for coding dental enamel defects. For the purpose of comparison to other studies, supplementary coding techniques were employed and the data were analyzed accordingly. The coding criterion for tooth stress (TS) is based on the presence or absenc e of any hypoplastic lesion in an individual. Both hypoplastic pitting and lines have equal weight and a lesion need not be bilateral for inclusion. One -way analyses of variance (ANOVAs) were performed to determine whether there are significant differe nces in stature estimates between stressed and unstressed groups divided by TS criterion. No significant differences between groups were found in the male sample; however, females had near significant differences (0.05 < p < 0.10) for sitting height and tibia length (Table 5 3). Resampling statistics were used as an independent test of near significant results of ANOVAs and the results of these tests were significant in both cases: STH ( p = 0.0000) and Tibia length ( p = 0.0335). A p value of 0.0000, as in the case of STH, means that none of the resampling iterations matched the magnitude of the observed difference.
82 Because some researchers do not score pitting when scoring enamel hypoplasias, the linear enamel hypoplasia (LEH) code was created. Using t his criterion, individuals with hypoplastic pitting in the absence of linear hypoplasias were removed from further analysis. However, this was true of only one individual in the sample, a female. As such, the results for the male sample coded by TS and L EH criteria are the same. One -way analyses of variance (ANOVAs) were performed on the female sample to determine whether there are significant differences in stature estimates between stressed and unstressed groups divided by LEH criterion. Near signi ficant differences were found for the ANOVAs of sitting height and tibia length (Table 5 4). When resampling statistics were used, both sitting height and tibia length differences were found to be significantly different ( p = 0.0365 and p = 0.0299 respect ively). Stature and Harris Lines Individuals were divided into two groups based on the presence or absence of Harris lines and one -way ANOVAs were performed to determine whether there were any differences in stature estimates between these groups. The res ults of these ANOVAs (summarized in Table 5 5) show that there are no significant differences between groups, indicating no significant relationship between the existence of Harris lines and differences in stature estimates for the individuals examined. H owever, it is interesting to note that for all the measurements examined, those with Harris lines exhibit longer dimensions, on average, than those without. A series of two -way ANOVAs were performed to determine whether or not there was any interaction eff ect between Harris lines and dental enamel defects as influences on stature The ANOVA indicated no significant interaction (Table 5 6). Stature and the Minimum Number of Stress Events Stature estimates were compared to the minimum number of stress event s (MNSE) score calculated for each individual based on the distribution of enamel defects as described in the
83 methods section. Males in this sample had a MNSE score between zero and five; females had a score between zero and four. Levines test of homosc edasticity found that female skeletal height and revised Fully stature estimates had unequal error variances, violating the assumptions of the ANOVA; a Kruskal Wallis test was performed on these data to evaluate the differences between groups based on the MNSE score. Differences in female skeletal height were found to be 2 (4, 97) = 3.240, p = 0.518. The same was true of the female revised Fully stature 2 (4, 97) = 3.392, p = 0.494. One -way ANOVAs were conducted on the remaining cases (Table 5 7). In no instance do differences in the minimum number of stress events have a significant relationship to the stature estimates in question; however, a n apparent trend toward decreased stature with increased number of stress events is notable in the female sample. Stature estimates were also compared to the number of Harris lines in an individual (none, one or more than one). In the female sample, no individual had more than one H arris line, so this analysis was confined to males. One -way ANOVAs were conducted and in no instance do differences in the number of Harris lines have a significant relationship to the stature estimates in question. Although the differences are not signi ficant it is notable that average stature estimates are greatest in those with two or more Harris lines (Table 5 8). Stature and the Age at which Stress Events Occur Based on the relative location of a linear enamel hypoplasia, individuals in this study we re divided into broad groups based on age of occurrence of dental stress events: a) infant stress events (birth to three years), b) childhood stress events ( three to seven years), and c) those individuals who experienced stress during both infancy and ch ildhood. These ages were calculated using the method of Goodman et al. (19 8 0). The stature estimates for each of these groups were compared using ANOVAs to discover whether or not the age at which a stress event occurred had any effect on adult stature.
84 Due to unequal error variances in female measurements, as determined by Levines test of homoscedasticity, Kruskal Wallis tests were performed on this sample to evaluate differences 2 (3, 97) = 2.773, p = 0.428), revised Fully stature 2 (3, 97) = 2.863, p 2 (3, 97) = 5.274, p = 0.153) between groups based on the age of the stress event. These results indicate that there are no significant locational differences in these measurements based on th e age at which stress events occur. For all other indices, ANOVAs yield no significant results for comparisons of samples based on age of stress events; however, near significant (0.05 < p < 0.10) differences were found in male tibia lengths (Table 5 9). Two -way comparisons between samples based on the age of stress events were made using resampling statistics. For most of these comparisons, groups were not found to be significantly different; however, tibia length was found to be significantly different between those with only infant stress and those with only childhood stress (p = 0.0079) The results of the comparisons made using resampling statistics are summarized in Table 5 10. Although the results of these ANOVAs were insignificant, a pattern was observable in the female data. Mean lengths were highest among the unstressed and lowest when stress occurred in both infancy and childhood for all measurements. A second set of ANOVAs were performed comparing individuals with infant stress markers and t hose without any dental stress. Individuals who only formed enamel defects during childhood were excluded from analysis to isolate the effects of stress in infancy possibly the most critical period of growth disruption. This investigation yielded signi ficant results in tibia length among females (Table 5 11). Tibia length is significantly shorter in females who form enamel defects in infancy compared to an unstressed sample (p = 0.043) Using Byers technique for aging Harris lines, individuals in this study were divided into three broad groups based on age of occurrence of stress events: a) infancy (birth to three years),
85 b) childhood ( three to seven years), and c) those individuals who experienced stress during both infancy and childhood. No individual had visible Harris lines that formed during the juvenile or adolescent period. The stature estimates for each of these groups were compared using ANOVAs to discover whether or not the a ge at which a stress event occurred had any effect on adult stature. The ANOVAs yielded no significant or meaningful results given that the vast majority of Harris lines in this sample occurred during infancy. Only two females and four males exhibited Ha rris lines during childhood (Table 5 12).
86 Table 5 1. Descriptive statistics and normality of osteometric variables Variable Sex N Mean S.D. Min Max Statistic a p Humerus (mm) b F 97 312.3 15.24 280.00 349.00 0.989 0.615 M 107 339.9 18.57 296.50 382.50 0.991 0.668 Radius (mm) b F 97 237.6 13.63 202.50 276.00 0.989 0.613 M 107 266.2 14.03 234.00 301.00 0.994 0.942 Femur bic (mm) b F 97 437.2 25.15 380.00 490.50 0.987 0.462 M 107 475.1 25.68 410.50 530.50 0.980 0.099 Femur max (mm) b F 97 442.9 25.48 386.50 495.50 0.983 0.258 M 107 479.2 26.03 412.50 537.00 0.981 0.138 Tibia (mm) b F 97 366.8 22.50 319.00 420.00 0.986 0.400 M 107 402.3 25.01 344.00 472.00 0.994 0.902 SKH (cm) F 97 148.0 6.38 135.09 162.62 0.985 0.351 M 107 159.8 6.60 146.31 175.07 0.983 0.192 STH (cm ) F 97 45.0 2.34 39.70 50.10 0.989 0.607 M 107 48.0 2.12 41.44 54.53 0.988 0.431 a Normality tests performed using the ShapiroWilk statistic b Description of paired elements refers to the average
87 Table 5 2. ANOVA Stature estimates by dental enamel defects (CBTS) N Mean S.D. Levines F p Male Skeletal height (SKH) Unstressed 42 159.82 6.853 Stressed 65 159.74 6.486 0.405 0.004 0.948 Revised Fully stature Unstressed 42 172.16 6.893 Stressed 65 172.05 6.538 0.427 0.007 0.932 Sitting height (STH) Unstressed 42 47.96 2.260 Stressed 65 48.02 2.033 0.846 0.015 0.904 Max. femur length Unstressed 42 481.50 26.115 Stressed 65 478.75 25.884 0.993 0.285 0.594 Tibia length Unstressed 42 401.29 24.410 Stressed 65 403.15 25.337 0.758 0.143 0.706 Female Skeletal height (SKH) Unstressed 43 149.27 6.672 Stressed 54 147.04 6.018 0.711 2.982 0.087 a Revised Fully stature Unstressed 43 161.56 6.691 Stressed 54 159.29 6.061 0.752 3.054 0.084 a Sitting height (STH) Unstressed 43 45.50 2.285 Stressed 54 44.62 2.325 0.672 3.472 0.066 a Max. Femur length Unstressed 43 448.23 26.411 Stressed 54 439.46 24.702 0.886 2.837 0.095 a Tibia length Unstressed 43 371.86 22.271 Stressed 54 363.07 21.943 0.927 3.788 0.055 a a Near significant (0.05 < p < 0.10)
88 Table 5 3. ANOVA Stature estimates by dental enamel defects (TS) N Mean S.D. Levines F p Male Skeletal height (SKH) Unstressed 37 159.97 6.875 Stressed 70 159.66 6.499 0.478 0.053 0.819 Revised Fully stature Unstressed 37 172.29 6.914 Stressed 70 171.99 6.551 0.497 0.047 0.828 Sitting height (STH) Unstressed 37 48.03 1.950 Stressed 70 47.98 2.210 0.476 0.013 0.909 Max. femur length Unstressed 37 481.76 26.700 Stressed 70 478.81 25.583 0.838 0.311 0.578 Tibia length Unstressed 37 401.57 25.256 Stressed 70 402.87 24.848 0.969 0.066 0.798 Female Skeletal height (SKH) Unstressed 36 149.27 6.660 Stressed 61 147.29 6.147 0.753 2.209 0.141 Revised Fully stature Unstressed 36 161.57 6.666 Stressed 61 159.54 6.193 0.791 2.312 0.132 Sitting height (STH) Unstressed 36 45.56 2.301 Stressed 61 44.68 2.316 0.902 3.232 0.075 a Max. femur length Unstressed 36 447.86 26.784 Stressed 61 440.69 24.901 0.797 1.776 0.186 Tibia length Unstressed 36 372.42 22.606 Stressed 61 363.75 21.838 0.993 3.471 0.066 a a Near significant (0.05 < p < 0.10)
89 Table 5 4. ANOVA Proxies for stature by dental enamel defects (LEH) in females N Mean S.D. Levines F p Skeletal height (SKH) Unstressed 36 149.27 6.660 Stressed 60 147.29 6.199 0.845 2.182 0.143 Revised Fully stature Unstressed 36 161.57 6.666 Stressed 60 159.53 6.245 0.880 2.293 0.133 Sitting height (STH) Unstressed 36 45.56 2.301 Stressed 60 44.67 2.333 0.854 3.290 0.073 a Max. femur length Unstressed 36 447.86 26.784 Stressed 60 440.53 25.082 0.851 1.825 0.180 Tibia length Unstressed 36 372.42 22.606 Stressed 60 363.55 21.963 0.973 3.588 0.060 a a Near significant (0.05 < p< 0.10)
90 Table 5 5. ANOVA Stature estimates by the presence or absence of Harris lines N Mean S.D. Levines F p Male Skeletal height (SKH) Absent 56 159.62 6.926 Present 40 160.67 6.619 0.613 0.555 0.458 Revised Fully stature Absent 56 171.91 6.989 Present 40 173.01 6.641 0.637 0.599 0.441 Sitting height (STH) Absent 56 47.96 1.976 Present 40 48.20 2.340 0.411 0.290 0.591 Max. femur l e ngth Absent 56 480.32 27.403 Present 40 482.63 25.820 0.944 0.173 0.678 Tibia length Absent 56 401.71 25.796 Present 40 405.80 25.588 0.652 0.589 0.445 Female Skeletal height (SKH) Absent 56 147.82 6.631 Present 31 148.47 6.230 0.701 0.201 0.655 Revised Fully stature Absent 56 160.13 6.661 Present 31 160.71 6.269 0.718 0.157 0.693 Sitting height (STH) Absent 56 44.94 2.422 Present 31 45.27 2.237 0.385 0.377 0.541 Max. femur length Absent 56 441.86 26.551 Present 31 445.10 25.513 0.992 0.305 0.582 Tibia length Absent 56 365.18 22.820 Present 31 369.71 21.125 0.737 0.829 0.365
91 Table 5 6 Interaction effects of Harris lines (HL) and enamel defects (CBTS) on stature F p Male Skeletal Height (SKH) HL 0.478 0.491 CBTS 0.162 0.689 HL CBTS 0.013 0.910 Revised Fully Stature HL 0.518 0.473 CBTS 0.145 0.704 HL CBTS 0.013 0.909 Sitting Height (STH) HL 0.124 0.725 CBTS 0.233 0.631 HL CBTS 0.995 0.321 Max. Femur Length HL 0.194 0.661 CBTS 0.002 0.964 HL CBTS 0.045 0.833 Tibia Length HL 0.432 0.512 CBTS 0.717 0.399 HL CBTS 0.127 0.722 Female Skeletal Height (SKH) HL 0.309 0.580 CBTS 1.483 0.227 HL CBTS 0.000 0.997 Revised Fully Stature HL 0.255 0.615 CBTS 1.510 0.223 HL CBTS 0.000 0.993 Sitting Height (STH) HL 0.655 0.421 CBTS 2.914 0.092 HL CBTS 0.092 0.762 Max. Femur Length HL 0.366 0.547 CBTS 1.339 0.251 HL CBTS 0.132 0.717
92 Table 5 6. Continued F p Tibia Length HL 1.114 0.294 CBTS 2.389 0.126 HL CBTS 0.007 0.934
93 Table 5 7. ANOVA Stature by minimum number of stress events (MNSE) N Mean S.D. Levines F p Male Skeletal height (SKH) 0 42 159.82 6.853 1 19 159.36 6.305 2 27 160.99 7.036 3 15 158.53 5.428 4 3 160.32 7.214 5 1 149.47 0.437 0.788 0.561 Revised fully stature 0 42 172.16 6.893 1 19 171.69 6.452 2 27 173.32 7.037 3 15 170.79 5.465 4 3 172.56 7.264 5 1 161.85 0.460 0.785 0.563 Sitting height (STH) 0 42 47.96 2.260 1 19 48.06 1.842 2 27 48.20 1.907 3 15 47.66 2.359 4 3 49.03 2.440 5 1 44.39 0.889 0.850 0.517 Maximum femur length 0 42 481.50 26.115 1 19 475.21 26.749 2 27 484.89 28.133 3 15 474.27 21.382 4 3 477.67 18.583 5 1 451.00 0.592 0.744 0.592 Maximum tibia length 0 42 401.29 24.410 1 19 401.21 23.827 2 27 405.56 27.392 3 15 402.47 24.480 4 3 407.33 29.771 5 1 373.00 0.797 0.404 0.845 Female Skeletal height (SKH) 0 43 149.27 6.672 1 26 146.85 4.745 2 18 147.95 7.114 3 9 146.13 7.603 4 1 143.85 0.022 a, b
94 Ta b le 5 7. Continued N Mean S.D. Levines F p Revised fully stature 0 43 161.56 6.691 1 26 159.09 4.831 2 18 160.26 7.169 3 9 158.31 7.506 4 1 155.75 0.031 a, b Sitting height (STH) 0 43 45.50 2.285 1 26 44.46 1.949 2 18 44.81 2.662 3 9 44.51 2.884 4 1 46.43 0.210 1.063 0.379 Maximum femur length 0 43 448.23 26.411 1 26 441.08 21.583 2 18 443.00 29.100 3 9 431.00 23.696 4 1 410.00 0.330 1.402 0.240 Maximum tibia length 0 43 371.86 22.271 1 26 362.92 18.560 2 18 366.83 25.551 3 9 358.00 24.990 4 1 345.00 0.242 1.344 0.260 a Significant ( p < 0.05) b The assumptions of the F test are violated where conditions of homoscedasticity are not met.
95 Table 5 8. ANOVA Stature estimates by number of Harris lines N Mean S.D. Levines F p Male Skeletal height (SKH) No Harris lines 56 159.62 6.926 One Harris line 30 160.39 6.238 Two or more Harris lines 10 161.50 7.964 0.508 0.375 0.688 Revised Fully stature No Harris lines 56 171.91 6.989 One Harris line 30 172.73 6.276 Two or more Harris lines 10 173.84 7.947 0.560 0.395 0.675 Sitting height (STH) No Harris lines 56 47.96 1.976 One Harris line 30 47.94 2.280 Two or more Harris lines 10 48.97 2.466 0.852 1.031 0.361 Maximum femur length No Harris lines 56 480.32 27.403 One Harris line 30 482.33 22.934 Two or more Harris lines 10 483.50 34.539 0.197 0.093 0.912 Tibia length No Harris lines 56 401.71 25.796 One Harris line 30 405.53 24.895 Two or more Harris lines 10 406.60 28.968 0.802 0.298 0.743
96 Table 5 9. ANOVA Stature estimates by age of enamel defect N Mean S.D. Levines F p Male Skeletal height (SKH) No stress 42 159.82 6.853 Infant stress 10 163.30 6.875 Childhood stress 18 157.97 6.258 Infant & childhood stress 37 159.63 6.277 0.597 1.418 0.242 Revised Fully stature No stress 42 172.16 6.893 Infant stress 10 175.68 6.986 Childhood stress 18 170.31 6.340 Infant & childhood stress 37 171.92 6.293 0.571 1.432 0.238 Sitting height (STH) No stress 42 47.96 2.260 Infant stress 10 48.50 1.996 Childhood stress 18 48.07 2.060 Infant & childhood stress 37 47.86 2.063 0.986 0.247 0.863 Maximum femur length No stress 42 481.50 26.115 Infant stress 10 491.40 26.850 Childhood stress 18 468.89 27.685 Infant & childhood stress 37 480.14 23.535 0.921 1.840 0.144 Tibia length No stress 42 401.29 24.410 Infant stress 10 416.60 21.854 Childhood stress 18 392.89 24.726 Infant & childhood stress 37 404.51 25.067 0.993 2.148 0.099 b Female Skeletal height (SKH) No stress 43 149.27 6.672 Infant stress 13 148.04 4.231 Childhood stress 19 147.09 5.669 Infant & childhood stress 22 146.40 7.251 0.039 a, c Revised Fully stature No stress 43 161.56 6.691 Infant stress 13 160.27 4.314 Childhood stress 19 159.37 5.741 Infant & childhood stress 22 158.64 7.266 0.045 a, c
97 Table 5 9. Continued N Mean S.D. Levines F p Sitting height (STH) No stress 43 45.50 2.285 Infant stress 13 44.65 2.214 Childhood stress 19 44.65 2.280 Infant & childhood stress 22 44.57 2.527 0.684 1.138 0.338 Maximum femur length No stress 43 448.23 26.411 Infant stress 13 446.23 17.234 Childhood stress 19 441.26 24.299 Infant & childhood stress 22 433.91 28.294 0.082 1.633 0.187 Tibia length No stress 43 371.86 22.271 Infant stress 13 362.69 11.071 Childhood stress 19 367.11 23.664 Infant & childhood stress 22 359.82 25.284 0.027 a, c a Significant ( p < 0.05) b Near significant (0.05 < p < 0.10) c The assumptions of the F test are violated where conditions of homoscedasticity are not met. Table 5 10. Results of resampling for male tibia length by age of the enamel defect Age categories compared p No stress infant stress 0.9674 No stress childhood stress 0.1052 No stress infant and childhood stress 0.7180 Infant stress childhood stress 0.0079 a Infant stress infant and childhood stress 0.0856 Childhood stress infant and childhood stress 0.9500 a Significant ( p < 0.05)
98 Table 5 11. ANOVA Stature estimates by infant enamel defect N Mean S.D. Levines F p Male Skeletal height (SKH) No stress 42 159.82 6.853 Infant stress 47 160.41 6.510 0.474 0.173 0.679 Revised Fully stature No stress 42 172.16 6.893 Infant stress 47 172.72 6.555 0.491 0.151 0.698 Sitting height (STH) No stress 42 47.96 2.260 Infant stress 47 47.99 2.045 0.838 0.004 0.951 Maximum femur length No stress 42 481.50 26.115 Infant stress 47 482.53 24.418 0.672 0.037 0.848 Tibia length No stress 42 401.29 24.410 Infant stress 47 407.09 24.702 0.879 1.236 0.269 Female Skeletal height (SKH) No stress 43 149.27 6.672 Infant stress 34 147.25 6.211 0.760 1.845 0.178 Revised Fully stature No stress 43 161.56 6.691 Infant stress 34 159.48 6.246 0.789 1.937 0.168 Sitting height (STH) No stress 43 45.50 2.285 Infant stress 34 44.67 2.385 0.484 2.414 0.124 Maximum femur length No stress 43 448.23 26.411 Infant stress 34 439.76 24.414 0.703 2.085 0.153 Tibia length No stress 43 371.86 22.271 Infant stress 34 361.62 20.836 0.699 4.250 0.043 a a Significant ( p < 0.05)
99 Table 5 12. ANOVA Stature estimates by age of Harris lines N Mean S.D. Levines F p Male Skeletal height (SKH) No stress 56 159.62 6.926 Infant stress 36 161.47 6.384 Childhood stress 1 156. 70 Infant & childhood stress 3 152.33 4.149 0.304 2.044 0.113 Revised fully stature No stress 56 171.91 6.989 Infant stress 36 173.81 6.410 Childhood stress 1 169.06 Infant & childhood stress 3 164.71 4.233 0.318 2.018 0.117 Sitting height (STH) No stress 56 47.96 1.976 Infant stress 36 48.31 2.350 Childhood stress 1 49.06 Infant & childhood stress 3 46.53 2.310 0.590 0.798 0.498 Maximum femur length No stress 56 480.32 27.403 Infant stress 36 485.67 25.019 Childhood stress 1 461.00 Infant & childhood stress 3 453.33 18.930 0.382 1.678 0.177 Tibia length No stress 56 401.71 25.796 Infant stress 36 408.53 25.073 Childhood stress 1 384.00 Infant & childhood stress 3 380.33 20.526 0.472 1.588 0.198 Female Skeletal height (SKH) No stress 56 147.82 6.631 Infant stress 29 148.08 6.152 Childhood stress 2 154.18 5.917 0.806 0.933 0.398 Revised fully stature No stress 56 160.13 6.661 Infant stress 29 160.31 6.205 Childhood stress 2 166.41 5.639 0.783 0.901 0.410 Sitting height (STH) No stress 56 44.94 2.422 Infant stress 29 45.18 2.283 Childhood stress 2 46.56 0.733 0.286 0.511 0.602 Maximum femur length No stress 56 441.86 26.551 Infant stress 29 443.72 25.794 Childhood stress 2 465.00 7.071 0.291 0.772 0.465
100 Table 5 12. Continued N Mean S.D. Levines F p Tibia length No stress 56 365.18 22.820 Infant stress 29 368.62 21.333 Childhood stress 2 385.50 10.607 0.535 0.954 0.389 Figure 5 1. Boxplot of female tibia length divided by CBTS stress codes. Differences in tibia length between stressed and unstressed groups were found to be significant ( p = 0.0271). This boxplot is used as an example of the d istributional differences between stressed and unstressed groups
101 CHAPTER 6 RESULTS: SEXUAL DIMORPHISM AND STRESS The relative degree of sexual dimorphism between the two groups was found using the formula ln X Mln X F for stature estimates and long bone length, where X M is the male mean and X F is the female mean. The population was then divided based on the presence or absence of stress markers based on the CBTS and HL criteria described in the Methods section. The results of these calculations can be observed in Table 6 1 and Table 6 2. In t his study, the null hypothesis states that there are no metric differences between individuals who have experience d growth disruption compared to those who have not. The alternative hypothesis with regards to sexual dimorphism is that unstressed populatio ns will have higher rates of sexual dimorphism due to the fact that males are more likely to reach their genetic potential under favorable circumstances. A corollary with this alternative hypothesis is that female size should be similar in stressed and u nstressed populations due to fat and nutrient reserves which help females buffer the negative effects of nutritional stress (Stini 1969, 1982, 1985). In C hapter 5 the results indicate that w here significant differences in mean stature and long bone estim ates occur, these differences are attributed to the female portion of the sample, not the males. Expectations based on Stinis (1969, 1982, 1985) findings were that the opposite pattern would be observed. T he mean measures suggest that female measurement s vary more in accordance with stress than male measurements conducted in this study A comparison of sexual dimorphism scores across various measurements shows that for both stressed and unstressed populations, the greatest amount of sexual dimorphism is found in the radius and in tibia length in stressed individuals. The smallest dimorphism scores are found in sitting height estimates.
102 Differences in sexual dimorphism scores between stressed and unstressed groups vary depending on the type of stress ma rker used. Larger differences in sexual dimorphism scores are found by using CBTS markers than are found by using HL markers (Table 6 3). In addition, when the population is divided by CBTS criterion, sexual dimorphism is greater in the stressed group than in the unstressed group for every measure, the opposite of expectations based on Stinis (1969, 1982, 1985) findings. When the population is divided by HL markers the same was true only of stature estimates. It is important to note that the difference s in dimorphism scores are an order of magnitude smaller for the population divided by HL criterion than in the population divided by CBTS criterion and, as such, they are likely less meaningful. Because sexual dimorphism, as it is calculated here, is foun d by taking the difference of two means, there is no measure of dispersion associated with the sexual dimorphism score; consequently, traditional statistical analyses cannot be completed on the raw data that would allow us to reject the null hypothesis tha t claims no differences between stressed and unstressed groups. The (one tailed) alternative hypothesis is that sexual dimorphism is greater in unstressed groups. To address that question, a bootstrap procedure was employed using Resampling Stats 5.0.2 s oftware. The stressed and unstressed, male and female populations were pooled, and then resampled with replacement using the original sample sizes. In each resampling, the log difference between male and female means was calculated for the stressed and u nstressed groups. A difference in dimorphism was then found by subtracting the stressed measure of dimorphism from the unstressed sample. Probabilities were determined by counting the number of times the resampling mean difference was greater or equal to the empirical measure of difference in sexual dimorphism which was calculated by subtracting the original dimorphism score in the stressed sample from the unstressed score. Each resampling procedure was run with
103 1,000 iterations except where probabilitie s approached significance (in the case of radius length partitioned by CBTS ). In that instance, the test was run with 10,000 iterations. Because the alternative hypothesis was directional, the probability was calculated as 1 p where the empirical difference in dimorphism was negative. This procedure tests a second alternative hypothesis that sexual dimorphism is greater in the stressed sample. In no case did the statistical analysis support the first or second alternative h ypothesis (at the p < 0.05 significance level); i.e., the null hypothesis no metrically observable difference between the size and shape of individuals who have experienced growth disruption and those who h ave not could not be rejected The results o f this procedure are summarized in Table 6 3.
104 Table 6 1. Sexual dimorphism scores for unstressed individuals Stress Marker X M X F ln X M ln X F Sexual Dimorphism Skeletal height (SKH) CBTS 159.822 149.266 5.074 5.006 0.068 HL 159.616 147.821 5.073 4.996 0.077 Revised Fully Stature CBTS 172.162 161.556 5.148 5.085 0.064 HL 171.914 160.128 5.147 5.076 0.071 Sitting height (STH) CBTS 47.964 45.497 3.870 3.818 0.053 HL 47.961 44.940 3.870 3.805 0.065 Maximum femur length a CBTS 48.123 44.792 3.874 3.802 0.072 HL 47.978 44.143 3.871 3.787 0.083 Bicondylar femur length a CBTS 47.701 44.191 3.865 3.789 0.076 HL 47.571 43.563 3.862 3.774 0.088 Tibia length a CBTS 40.126 37.170 3.692 3.615 0.077 HL 40.171 36.486 3.693 3.597 0.096 Humerus length a CBTS 33.989 31.592 3.526 3.453 0.073 HL 34.098 31.124 3.529 3.438 0.091 Radius length a CBTS 26.637 24.101 3.282 3.182 0.100 HL 26.657 23.724 3.283 3.166 0.117 a This is the average of the left and right elements where both are available. Table 6 2. Sexual dimorphism scores for stressed individuals Stress Marker X M X F ln X M ln X F Sexual Dimorphism Skeletal height (SKH) CBTS 159.736 147.037 5.074 4.991 0.083 HL 160.665 148.472 5.079 5.000 0.079 Revised Fully Stature CBTS 172.049 159.289 5.148 5.071 0.077 HL 173.011 160.707 5.153 5.080 0.074 Sitting height (STH) CBTS 48.015 44.618 3.872 3.798 0.073 HL 48.199 45.265 3.875 3.813 0.063 Maximum femur length a CBTS 47.788 43.883 3.867 3.782 0.085 HL 48.185 44.452 3.875 3.794 0.081 Bicondylar femur length a CBTS 47.391 43.336 3.858 3.769 0.089 HL 47.776 43.889 3.867 3.782 0.085 Tibia length a CBTS 40.292 36.281 3.696 3.591 0.105 HL 40.546 36.982 3.702 3.610 0.092 Humerus length a CBTS 33.985 30.944 3.526 3.432 0.094 HL 34.088 31.419 3.529 3.447 0.082 Radius length a CBTS 26.612 23.482 3.281 3.156 0.125 HL 26.718 23.892 3.285 3.174 0.112 a This is the average of the left and right elements where both are available.
105 Table 6 3. Differences in sexual dimorphism between unstressed and stressed populations a Measurement CBTS markers p HL markers p Skeletal height (SKH) 0.0145 0.128 0.0022 0.472 Revised Fully Stature 0.0135 0.121 0.0027 0.464 Sitting height (STH) 0.0206 0.080 0.0022 0.407 Maximum femur length b 0.0135 0.214 0.0027 0.392 Bicondylar femur length b 0.0130 0.215 0.0031 0.411 Tibia length b 0.0283 0.064 0.0042 0.399 Humerus length b 0.0206 0.090 0.0098 0.247 Radius length b 0.0251 0.061 0.0048 0.368 a Difference measured as the unstressed score minus the stressed score. b This is the average of the left and right elements where both are available.
106 CHAPTER 7 RESULTS: PROPORTIONS AND STRESS The following analyses describe differences in proportional indices, expressed as ratios, between those individuals with and those w ithout skeletal indicators of growth disruption. The components of these ratios were significantly correlated for all proportions, whether the data were divided by gender or combined across genders (Table 7 1). This result means that although ratios are dimensionless, they are not independent of body size and allometric factors need to be considered when investigating differences between groups. Proportions and dental enamel defects Individuals were determined to be either stressed or unstressed ba sed on the presence or absence of bilateral dental enamel defects in the anterior dentition (CBTS) by the criteria described in the Methods section. A ll proportions, except for the crural index in males, exhibited homoscedasticity, as determined by Levine s test of homogeneity. For those proportions that were found to be homogenous, one -way analyses of variance (ANOVAs) were performed to determine whether there were significant differences in proportional indices between stressed and unstressed groups. I n no case were significant differences between groups observed (Table 7 2). In the case of the male crural index, the Mann-Whitney test found no significant differences in the crural index between the stressed and unstressed groups of males ( p = 0.080). Because researchers code enamel defects using different methods, different coding techniques were used to form a basis of comparison with other studies. The coding criterion for tooth stress (TS) is based on the presence or absence of any hypoplastic lesi on, regardless of the type or whether or not a lesion is found on its antimere. One -way ANOVAs were performed to determine whether there are significant differences in the distribution of proportional differences between stressed and unstressed groups div ided by TS criterion. No significant differences
107 between groups are found; however, males have near significant differences in crural indices (Table 7 3). Resampling statistics were used as an independent test of the near significant results of ANOVAs an d the differences were determined to be insignificant ( p = 0.9618) ; h owever, the resampling procedure is directional and tests the alternative hypothesis that the unstressed values are greater than the s tressed values. In this case, 1 p < 0.05, which me ans there is a significant difference between groups, but the difference is not in the direction proposed. In the case of the male crural index, values are significantly higher in the stressed group. Because some researchers do not score pitting when scor ing enamel hypoplasias, the linear enamel hypoplasia (LEH) code was created. Using this criterion, individuals with hypoplastic pitting in the absence of linear hypoplasias were removed from further analysis. However, this was only true of one female in the sample; as a result, the male sample coded by TS and LEH criteria are equivalent. One-way ANOVAs were performed on the female sample to determine whether there are significant differences in proportional indices between stressed and unstressed groups divided by LEH criterion. No significant or near significant differences were observed (Table 7 4). Proportions and Harris lines Individuals were divided into two groups based on the presence or absence of Harris lines (see Methods section) and one -way ANOVAs were performed to determine whether there were any differences in proportional indices between these groups. The results of these ANOVAs (summarized in Table 7 5) show that there are no significant differences between groups, indicating no relation ship between the existence of Harris lines and differences in proportional indices for the individuals examined. A series of two -way ANOVAs were performed to determine whether or not there was any interaction effect between Harris lines and dental enamel d efects in their potential influence on
108 proportions The results show that no interaction effect exists between these indicators of stress (Table 7 6). Scaling Correlations suggest that the components of ratios are not independent of body size (see Table 7 1). Because of this fact, a scaling analysis is necessary to qualify the results of the preceding ANOVAs. For example, if a stressed population is found to be larger than an unstressed one for any given measurement, there is a scaling effect that needs to be considered. A non -zero slope indicated that size influenced proportions in five of eight cases (see Figures 7 1 through 7 4). In three of eight regressions, the confidence interval for the slope included zero, signifying that the effect of skeletal height on proportions in those instances was negligible at best (Table 7 7). Based on the magnitude of the correlation coefficient it can be inferred that skeletal height is moderately related to the intermembral index in m ales and females and the humero femoral index in females. For the remaining proportional indices, less than nine percent (R2 were examined to see whether the numerator or denominator in each prop ortional ratio has greater effect on how proportional ratios are scaling (Figure 7 5 and Figure 76). The components of the ratios follow a pattern consistent with the negative slope exhibited when scaling ratios. A negative slope in Figures 71 through Figure 7 4 suggests that the variable in the denominator of a ratio is changing relative to size at a faster rate than the variable in the numerator. Scaling D ifferences Based on Enamel D efects Reduced major axis (RMA) regressions were used to analyze how the components of ratios scale to one another and to see whether or not scaling differs between groups designated
109 stressed or unstressed based on the presence or absence of enamel defects. Log transformed components of ratios were scaled to one another using the RMA software (Bohonak 2004). Differences in scaling exist between groups based on stress categories, but also vary by gender. The majority of RMAs find that the components of ratios scale isometrically to one another. For males, these ind ices include: the unstressed crural index, stressed and unstressed humerofemoral indices, stressed and unstressed radiohumeral indices, stressed and unstressed sitting height indices and the stressed intermembral index. Female isometric indices include: stressed and unstressed crural indices, the unstressed radiohumeral index, the unstressed sitting height index, and the stressed intermembral index. Three ratios are positively allometric, indicating that the numerator of the ratio scales at a faster rate than the denominator. These include the stressed male crural index, the stressed female radiohumeral index and the stressed female sitting -height index. The stressed and unstressed humerofemoral index in females and the unstressed male and female intermembral indices were found to be negatively allometric (Table 7 8). RMA2 software (Cole 1997) was used to test whether the scaling of skeletal dimensions between stressed and unstressed groups was significant ly differen t (Table 7 -9). Minor differences in slope and intercept values are noted in Table 7 8 and Table 7 -9; these differences reflect calculation differences between the Bohonak (2004) and Cole (1997) software and are likely due to rounding errors that are exaggerated in log transformed data. These differences are considered negligible and do not influence interpretation. RMA2 analyses find significant differences in slope between stressed and unstressed groups for the humerofemoral index and the intermembral i ndex in males. For these indices, the length of the humerus (or upper limb length) is scaling faster than femoral (or lower limb length) in the stressed group.
110 In cases where RMA slopes were not significantly different, intercept differences were tested by fitting a regression lined through the combined dataset and counting the number of individuals with and without enamel defects that fall above and below the line. This information was then placed in a 2x2 contingency table and Fishers exact test was used to test the difference (Tsutakawa and Hewett 1977). In no case were intercept differences significant. In other words, for any given value of X, the value of Y is not significantly different between samples. These RMA regressions are represented graphically in Figure 7 7 through Figure 7 16. The results of the ANOVAs between stressed and unstressed groups suggest that there are no significant differences in proportional indices between individuals grouped by the presence or absence of dental ename l defects. An investigation into the question of scaling suggests that allometric differences between such groups are subtle. Scaling D ifferences Based on Harris L ines Reduced major axis (RMA) regressions were also conducted to determine whether or not sc aling differs between those with and without Harris lines. Logtransformed components of ratios were scaled to one another using the RMA software of Bohonak (2004) in Table 710. Once again, differences in scaling exist between groups based on the prese nce or absence of Harris lines and also vary by gender. Male proportional ratios scale isometrically for those with and without Harris lines in all but the crural index. In this case males are positively allometric with tibial length increasing at a fas ter rate than femoral length. Females vary greatly. For females without Harris lines, the crural index, humerofemoral index, and the intermembral index scale isometrically. Females with Harris lines scale isometrically for the crural index, the radiohum eral index and the sitting -height index. Radiohumeral and sitting height indices scale positively in females without Harris lines. Humerofemoral and intermembral indices scale negatively in females with Harris lines, with femoral (or lower limb) length i ncreasing at a faster
111 rate than humerus (or upper limb) length. The length of the radius increases at a faster rate than the humerus and sitting height increases at a faster rate than skeletal height. RMA2 software (Cole 1997) was used to test whether th e allometric differences between those with and without Harris lines were significantly different (summarized in Table 7 11). The only significant differences in slope are for female humerofemoral indices and the intermembral indices. These are also the only instances where those individuals with Harris lines exhibit negative allometry. No intercept differences were found using Fishers exact test. RMA regressions representing the relationship between Harris lines and the scaling of proportional indices are depicted graphically in Figure 7 17 through Figure 726. ANOVAs suggest that there are no significant differences in proportional indices between those with and without Harris lines. RMA regressions suggest that there are some scaling differences bet ween groups, but that those differences are subtle. Large differences in slope as well as high r2 values for the regressions are necessary to interpret a scaling difference between groups as significant. R2 values for these regressions range from 0.29 to 0.84 with a median value of 0.735. Proportions and the Minimum Number of Stress Events Using the aging techniques for enamel hypoplasias described in the Methods section, a minimum number of stress events (MNSE) score was calculated for each individual. Levines test of homoscedasticity found that male crural indices had unequal error variances, violating the assumptions of the ANOVA; a Kruskal Wallis test was performed on this sample to evaluate the differences in crural indices between groups based on the MNSE score. The test was p = 0.54. One -way ANOVAs were conducted on the remaining cases (Table 7 12). In no instance do differences in the minimum number of stress events appear to affect the proportional indices in question.
112 Proportions were also compared to the number of Harris lines (none, one or more than one). In the female sample, no individual had more than one Harris line, so this analysis was confined to males. Due to unequal error variances in male crura l indices, as determined by Levines test of homoscedasticity, a Kruskal -Wallis test was performed on this sample to evaluate differences in crural indices between groups based on the age of the stress event. The p = 0.537. For all other indices, o ne -way ANOVAs were conducted and in no instance do differences in the number of Harris lines have a significant relationship to the proportions in question; however, near significant results were obtained for the inter membral index (Table 7 13). Resampling statistics were performed as an independent test for ANOVAs with near significant results. A significant difference in intermembral indices exists between those individuals with no Harris lines and those with one Ha rris line ( p = 0.0336). However, when those with no Harris lines and those with two or more lines were compared, the differences were insignificant ( p = 0.7798). Similarly, intermembral indices in those with one Harris line and those with two or more Har ris lines were not significantly different ( p = 0.986). Proportions and the Age at which Stress Events Occur Based on the relative location of a linear enamel hypoplasia, individuals in this study were divided into broad groups based on age of occurrence of dental stress events: a) infant stress events (birth to three years), b) childhood stress events ( three t o seven years), and c) those individuals who experienced stress during both infancy and childhood. The proportional indices of these groups were compared using ANOVAs to discover whether or not the age at which a stress event occurred had any effect on ad ult proportions. Due to unequal error variances in male crural indices, as determined by Levines test of homoscedasticity, a Kruskal Wallis test was performed on this sample to evaluate differences in crural indices between groups based on the age of the
113 (2, 65) = 1.37, p = 0.51. For all other indices, ANOVAs yield no significant results for comparisons of samples based on age of stress events (Table 7 14). A second set of ANOVAs were performed comparing indi viduals with infant stress markers and those without any dental stress (Table 7 15). This investigation did not find any significant results, but for the female radiohumeral index, a near significant score was obtained (p = 0.097). Resampling statistics were performed as an independent test for ANOVAs with near significant results and in this case the results were found to be significant ( p = 0.0447). For the male crural index, which did not pass the test for homogeneity of variance, the result of the Ma nnWhitney U test is near significant, 1.87, p = 0.062. Resampling statistics were used as an independent test of the near significant results of the MannWhitney U test. T he difference was determined to be insignificant ( p = 0.9 885); however, the r esampling procedure was directional and tests the alternative hypothesis that the unstressed values are greater than the stressed values. In this case, 1 p < 0.05, which means there is a significant difference between groups, but the difference is not i n the direction proposed. The male crural index is significantly higher in those with infant stress than in the unstressed group. Using Byers technique for aging Harris lines, individuals in this study were divided into three broad groups based on age of occurrence of stress events: a) infancy (birth to three years), b) childhood ( three to seven years), and c) those individuals who experienced stress during both infancy and childhood. No individual had visible Harris lines that formed during the juvenile or adolescent period. The stature estimates for each of these groups were compared using AN OVAs to discover whether or not the age at which a stress event occurred had any effect on proportions. The ANOVAs yielded no significant or meaningful results given that the vast
114 majority of Harris lines in this sample occurred during infancy. Only two females and four males exhibited Harris lines during childhood (Table 7 16).
115 Table 7 1. Pearson correlations of the components of ratios (N = 204) Radius Humerus Femur (max) Femur (bic) Tibia STH SKH Radius 1.000 0.915 a 0.868 a 0.873 a 0.916 a 0.683 a 0.909 a Humerus 0.915 a 1.000 0.906 a 0.907 a 0.909 a 0.694 a 0.923 a Femur (max) 0.868 a 0.906 a 1.000 0.997 a 0.928 a 0.651 a 0.949 a Femur (bic) 0.873 a 0.907 a 0.997 a 1.000 0.929 a 0.649 a 0.953 a Tibia 0.916 a 0.909 a 0.928 a 0.929 a 1.000 0.667 a 0.946 a STH 0.683 a 0.694 a 0.651 a 0.649 a 0.667 a 1.000 0.783 a SKH 0.909 a 0.923 a 0.949 a 0.953 a 0.946 a 0.783 a 1.000 a Correlation is significant at the 0.01 level (2 tailed).
116 Table 7 2. ANOVA Proportional indices by dental enamel defects (CBTS) N Mean S.D. Levines F p Male Crural index Unstressed 42 83.37 1.852 Stressed 65 84.31 2.405 0.048 a, b Humerofemoral index Unstressed 42 70.69 2.015 Stressed 65 71.13 2.136 0.574 1.156 0.285 Radiohumeral index Unstressed 42 78.39 2.333 Stressed 65 78.36 2.429 0.812 0.003 0.958 Sitting height index Unstressed 42 30.03 1.127 Stressed 65 30.08 1.110 0.716 0.053 0.818 Intermembral index Unstressed 42 68.76 1.869 Stressed 65 68.82 1.596 0.670 0.037 0.847 Female Crural index Unstressed 43 83.00 2.288 Stressed 54 82.70 2.745 0.378 0.349 0.556 Humerofemoral index Unstressed 43 70.60 2.114 Stressed 54 70.58 2.166 0.731 0.003 0.953 Radiohumeral index Unstressed 43 76.30 2.076 Stressed 54 75.87 2.228 0.937 0.953 0.331 Sitting height index Unstressed 43 30.50 1.247 Stressed 54 30.35 1.092 0.285 0.391 0.533 Intermembral index Unstressed 43 68.01 1.652 Stressed 54 67.93 1.675 0.886 0.050 0.823 a Significant ( p < 0.05) b The F test cannot be performed if conditions of homoscedasticity are not met.
117 Table 7 3. ANOVA Proportions by dental enamel defects (TS) N Mean S.D. Levines F p Male Crural index Unstressed 37 83.42 1.896 Stressed 70 84.21 2.373 0.116 3.067 0.083 a Humerofemoral index Unstressed 37 70.84 2.090 Stressed 70 71.02 2.104 0.741 0.165 0.685 Radiohumeral index Unstressed 37 78.37 2.401 Stressed 70 78.38 2.388 0.969 0.001 0.981 Sitting height index Unstressed 37 30.04 1.053 Stressed 70 30.06 1.149 0.618 0.008 0.929 Intermembral index Unstressed 37 68.88 1.943 Stressed 70 68.75 1.569 0.418 0.141 0.708 Female Crural index Unstressed 36 83.23 2.337 Stressed 61 82.60 2.650 0.631 1.378 0.243 Humerofemoral index Unstressed 36 70.84 2.046 Stressed 61 70.44 2.184 0.712 0.794 0.375 Radiohumeral index Unstressed 36 76.38 2.055 Stressed 61 75.87 2.218 0.628 1.259 0.265 Sitting height index Unstressed 36 30.54 1.328 Stressed 61 30.34 1.052 0.078 a 0.659 0.419 Intermembral index Unstressed 36 68.19 1.625 Stressed 61 67.83 1.674 0.838 1.016 0.316 a Near significant (0.05 < p< 0.10)
118 Table 7 4. ANOVA Proportional indices by dental enamel defects (LEH) in females N Mean S.D. Levines F p Crural index Unstressed 36 83.23 2.337 Stressed 60 82.58 2.667 0.612 1.461 0.230 Humerofemoral index Unstressed 36 70.84 2.046 Stressed 60 70.45 2.200 0.675 0.730 0.395 Radiohumeral index Unstressed 36 76.38 2.055 Stressed 60 75.95 2.152 0.742 0.938 0.335 Sitting height index Unstressed 36 30.54 1.328 Stressed 60 30.33 1.059 0.085 a 0.709 0.402 Intermembral index Unstressed 36 68.19 1.625 Stressed 60 67.88 1.641 0.943 0.760 0.386 a Near significant (0.05 < p< 0.10)
119 Table 7 5. ANOVA Proportional indices by the presence or absence of Harris lines (HL) N Mean S.D. Levines F p Male Crural index Absent 56 83.72 2.159 Present 40 84.13 2.297 0.956 0.797 0.374 Humerofemoral index Absent 56 71.11 2.224 Present 40 70.77 1.941 0.455 0.631 0.429 Radiohumeral index Absent 56 78.23 2.54 Present 40 78.41 2.282 0.592 0.124 0.725 Sitting height index Absent 56 30.07 1.013 Present 40 30.02 1.289 0.153 0.046 0.830 Intermembral index Absent 56 68.98 1.832 Present 40 68.56 1.488 0.306 1.423 0.236 Female Crural index Absent 56 82.67 2.534 Present 31 83.24 2.734 0.446 0.944 0.334 Humerofemoral index Absent 56 70.56 2.139 Present 31 70.78 2.283 0.597 0.197 0.658 Radiohumeral index Absent 56 76.22 2.194 Present 31 76.04 2.187 1.000 0.145 0.704 Sitting height index Absent 56 30.41 1.097 Present 31 30.50 1.141 0.722 0.132 0.718 Intermembral index Absent 56 68.06 1.64 Present 31 67.99 1.757 0.849 0.038 0.845
120 Table 7 6 Interaction effects of Harris lines (HL) and enamel defects (CBTS) on proportions F p Male Crural Index HL 0.396 0.531 CBTS 5.030 0.027 a HL CBTS 1.118 0.293 Humero femoral Index HL 0.901 0.345 CBTS 1.556 0.215 HL CBTS 0.505 0.479 Radio humeral Index HL 0.067 0.796 CBTS 0.074 0.786 HL CBTS 0.411 0.523 Sitting Height Index HL 0.141 0.709 CBTS 0.019 0.890 HL CBTS 1.054 0.307 Intermembral Index HL 1.668 0.200 CBTS 0.066 0.797 HL CBTS 0.540 0.464 Female Crural Index HL 1.258 0.265 CBTS 0618 0.434 HL CBTS 0.565 0.454 Humero femoral Index HL 0.261 0.611 CBTS 0.064 0.801 HL CBTS 0.179 0.673 Radio humeral Index HL 0.086 0.771 CBTS 1.111 0.295 HL CBTS 0.023 0.880 Sitting Height Index HL 0.252 0.617 CBTS 1.030 0.313 HL CBTS 0.154 0.696
121 Table 7 6. Continued F p Intermembral Index HL 0.021 0.884 CBTS 0.160 0.690 HL CBTS 0.003 0.959 a Significant ( p < 0.05) Table 7 7. Scaling proportional indices to size (SKH) Slope R 2 Crural index Male 0.089 0.069 Female 0.028 a 0.005 Humerofemoral index Male 0.042 a 0.017 Female 0.126 0.143 Radio humeral index Male 0.048 a 0.018 Female 0.081 0.058 Intermembral index Male 0.099 0.085 Female 0.101 0.150 a 95% C.I. includes zero
122 Table 7 8. RMA regressions of logtransformed components of ratios divided by dental enamel defects (CBTS) N RMA Intercept RMA Slope 95% CI R2 Scaling a Lower Upper Crural index Male (combined) 107 0.309 1.139 1.045 1.233 0.819 Positive Unstressed 42 0.274 1.116 0.989 1.243 0.873 Isometric Stressed 65 0.337 1.157 1.025 1.288 0.798 Positive Female (combined) 97 0.200 1.072 0.964 1.179 0.758 Isometric Unstressed 43 0.143 1.038 0.892 1.184 0.801 Isometric Stressed 54 0.236 1.093 0.929 1.257 0.710 Isometric Humero femoral index Male (combined) 107 0.145 0.997 0.898 1.097 0.734 Isometric Unstressed 42 0.081 0.862 1.001 0.747 Isometric Stressed 65 0.285 1.081 0.943 1.219 0.744 Isometric Female (combined) 97 0.104 0.845 0.754 0.935 0.725 Negative Unstressed 43 0.158 0.813 0.682 0.943 0.742 Negative Stressed 54 0.078 0.861 0.728 0.993 0.695 Negative Radio humeral index Male (combined) 107 0.053 0.965 0.864 1.066 0.706 Isometric Unstressed 42 0.231 1.082 0.886 1.277 0.681 Isometric Stressed 65 0.030 0.911 0.791 1.031 0.727 Isometric Female (combined) 97 0.396 1.185 1.066 1.304 0.759 Positive Unstressed 43 0.246 1.086 0.907 1.264 0.729 Isometric Stressed 54 0.492 1.249 1.081 1.417 0.767 Positive Sitting height index Male (combined) 107 0.658 1.061 0.902 1.221 0.398 Isometric Unstressed 42 0.738 1.098 0.834 1.361 0.435 Isometric Stressed 65 0.597 1.034 0.827 1.241 0.371 Isometric Female (combined) 97 0.977 1.212 1.035 1.388 0.489 Positive Unstressed 43 0.794 1.128 0.856 1.400 0.416 Isometric Stressed 54 1.125 1.280 1.036 1.524 0.532 Positive Intermembral index Male (combined) 107 0.016 0.908 0.832 0.984 0.815 Negative Unstressed 42 0.180 0.824 0.704 0.944 0.792 Negative Stressed 65 0.084 0.960 0.862 1.058 0.835 Isometric Female (combined) 97 0.054 0.884 0.808 0.960 0.822 Negative Unstressed 43 0.177 0.820 0.715 0.925 0.835 Negative Stressed 54 0.021 0.923 0.809 1.036 0.806 Isometric a Variables are considered to be isometric when the confidence interval includes a slope of 1.0.
123 Table 7 9. Comparison of RMA regressions (RMA2) divided by dental enamel defects (CBTS) N Intercept Slope 95% CI Scale t df p Intercept lower upper Diffs. Male Crural index Unstressed 42 0.274 1.116 0.996 1.250 Iso. Stressed 65 0.336 1.156 1.032 1.295 Pos. 0.454 53 0.326 0.326 Humerofemoral index Unstressed 42 0.079 0.864 0.736 1.014 Iso. Stressed 65 0.283 1.080 0.950 1.260 Iso. 2.241 49 0.015 a Radiohumeral index Unstressed 42 0.231 1.081 0.904 1.294 Iso. Stressed 65 0.032 0.910 0.798 1.038 Iso. 1.587 48 0.060 b 0.324 Sitting height index Unstressed 42 0.734 1.096 0.864 1.390 Iso. Stressed 65 0.597 1.034 0.848 1.261 Iso. 0.381 51 0.352 0.843 Intermembral index Unstressed 42 0.186 0.821 0.710 0.949 Neg. Stressed 65 0.084 0.960 0.866 1.063 Iso. 1.806 47 0.039 a Female Crural index Unstressed 43 0.146 1.040 0.903 1.196 Iso. Stressed 54 0.231 1.091 0.939 1.266 Iso. 0.481 50 0.317 0.543 Humerofemoral index Unstressed 43 0.154 0.815 0.695 0.956 Neg. Stressed 54 0.078 0.860 0.738 1.002 Iso. 0.499 50 0.310 0.837 Radiohumeral index Unstressed 43 0.241 1.082 0.919 1.274 Iso. Stressed 54 0.490 1.249 1.092 1.428 Pos. 1.391 48 0.085 b 1.000 Sitting height index Unstressed 43 0.784 1.123 0.885 1.426 Iso. Stressed 54 1.127 1.281 1.060 1.548 Pos. 0.881 48 0.191 1.000 Intermembral index Unstressed 43 0.182 0.818 0.720 0.929 Neg. Stressed 54 0.019 0.922 0.816 1.042 Iso. 1.395 50 0.085 b 1.000 a Significant ( p < 0.05) b Near significant (0.05 < p< 0.10)
124 Table 7 10. RMA regressions of logtransformed components of ratios divided by the presence or absence of Harris lines N RMA Intercept RMA Slope 95% CI R2 Scaling a Lower Upper Crural index Male (combined) 96 0.340 1.157 1.021 1.293 0.812 Positive Absent 56 0.286 1.125 1.001 1.250 0.836 Positive Present 40 0.423 1.207 1.020 1.394 0.778 Positive Female (combined) 87 0.134 1.031 0.915 1.148 0.724 Isometric Absent 56 0.177 1.057 0.912 1.202 0.749 Isometric Present 31 0.031 0.970 0.762 1.178 0.681 Isometric Humerofemoral index Male (combined) 96 0.139 0.994 0.890 1.098 0.738 Isometric Absent 56 0.139 0.995 0.853 1.136 0.730 Isometric Present 40 0.142 0.995 0.833 1.158 0.754 Isometric Female (combined) 87 0.117 0.837 0.742 0.932 0.724 Negative Absent 56 0.009 0.902 0.776 1.028 0.738 Isometric Present 31 0.364 0.688 0.552 0.824 0.728 Negative Radiohumeral index Male (combined) 96 0.077 0.981 0.864 1.098 0.661 Isometric Absent 56 0.037 0.955 0.794 1.115 0.620 Isometric Present 40 0.140 1.022 0.847 1.198 0.727 Isometric Female (combined) 87 0.396 1.186 1.055 1.318 0.736 Positive Absent 56 0.347 1.154 1.008 1.300 0.787 Positive Present 31 0.564 1.298 0.988 1.608 0.603 Isometric Sitting height index Male (combined) 96 0.566 1.020 0.853 1.186 0.365 Isometric Absent 56 0.382 0.936 0.744 1.129 0.437 Isometric Present 40 0.811 1.130 0.818 1.443 0.294 Isometric Female (combined) 87 0.984 1.215 1.034 1.396 0.526 Positive Absent 56 1.050 1.245 1.024 1.466 0.578 Positive Present 31 0.853 1.155 0.825 1.485 0.433 Isometric Intermembral index Male (combined) 96 0.037 0.897 0.818 0.976 0.815 Negative Absent 56 0.045 0.894 0.785 1.002 0.802 Isometric Present 40 0.017 0.906 0.787 1.025 0.841 Isometric Female (combined) 87 0.058 0.882 0.794 0.970 0.785 Negative Absent 56 0.043 0.935 0.818 1.052 0.791 Isometric Present 31 0.268 0.772 0.636 0.908 0.784 Negative a Variables are considered to be isometric when the confidence interval includes a slope of 1.0.
125 Table 7 11. Comparison of RMA regressions (RMA2) divided by Harris lines N Intercept Slope 95% C.I. Scale t d.f. p Intercept Diffs. Lower Upper Male Crural index Absent 56 0.269 1.114 0.998 1.244 Iso Present 40 0.376 1.179 1.011 1.374 Pos 0.611 44.213 0.272 0.214 Humerofemoral index Absent 56 0.095 0.969 0.841 1.116 Iso Present 40 0.172 1.013 0.862 1.190 Iso 0.426 47.140 0.336 0.301 Radiohumeral index Absent 56 0.026 0.947 0.801 1.120 Iso Present 40 0.132 1.017 0.858 1.205 Iso 0.608 48.656 0.273 0.308 Sitting height index Absent 56 0.383 0.937 0.764 1.148 Iso Present 40 0.917 1.179 0.899 1.546 Iso 1.380 46.127 0.087 b 0.682 Intermembral index Absent 56 0.076 0.878 0.767 1.004 Iso Present 40 0.031 0.931 0.818 1.061 Iso 0.648 49.192 0.260 0.210 Female Crural index Absent 56 0.192 1.066 0.930 1.222 Iso Present 31 0.086 1.004 0.811 1.242 Iso 0.496 35.239 0.311 0.507 Humerofemoral index Absent 56 0.015 0.898 0.782 1.033 Iso Present 31 0.369 0.685 0.563 0.834 Neg 2.337 36.615 0.013 a 0.500 Radiohumeral index Absent 56 0.339 1.148 1.012 1.302 Pos Present 31 0.541 1.282 1.011 1.624 Pos 0.856 33.114 0.199 1.000 Sitting height index Absent 56 0.976 1.211 1.015 1.445 Pos Present 31 0.894 1.174 0.885 1.557 Iso 0.195 35.917 0.423 0.180 Intermembral index Absent 56 0.040 0.933 0.824 1.057 Iso Present 31 0.279 0.766 0.643 0.914 Neg 1.902 36.506 0.033 a a significant ( p < 0.05). b near significant (0.05 < p< 0.10)
126 Table 7 12. ANOVA Proportional indices by minimum number of stress events (MNSE) N Mean S.D. Levines F p Male Crural index 0 42 83.37 1.852 1 19 84.51 1.650 2 27 83.78 2.241 3 15 84.93 3.244 4 3 85.21 3.452 5 1 82.59 0.001 a, b Humerofemoral index 0 42 70.69 2.015 1 19 71.11 1.766 2 27 70.66 2.331 3 15 72.24 2.174 4 3 70.30 0.508 5 1 69.96 0.385 1.551 0.181 Radiohumeral index 0 42 78.39 2.333 1 19 78.63 2.0151 2 27 78.36 2.925 3 15 77.86 2.183 4 3 78.38 1.460 5 1 80.82 0.256 0.384 0.859 Sitting height index 0 42 30.03 1.127 1 19 30.19 1.328 2 27 29.96 1.097 3 15 30.06 1.002 4 3 30.58 0.417 5 1 29.70 0.331 0.242 0.943 Intermembral index 0 42 68.76 1.869 1 19 68.84 1.435 2 27 68.56 1.719 3 15 69.47 1.532 4 3 67.72 1.439 5 1 69.28 0.767 0.836 0.527 Female Crural index 0 43 83.00 2.288 1 26 82.44 3.123 2 18 82.73 2.089 3 9 83.22 3.065 4 1 84.04 0.266 0.313 0.869
127 Table 7 12. Continued N Mean S.D. Levines F p Humerofemoral index 0 43 70.60 2.114 1 26 70.30 2.262 2 18 70.54 1.894 3 9 70.98 2.229 4 1 74.67 0.611 1.111 0.356 Radiohumeral Index 0 43 76.30 2.076 1 26 75.93 2.566 2 18 75.70 1.973 3 9 75.85 1.891 4 1 77.65 0.371 0.431 0.786 Sitting height index 0 43 30.50 1.247 1 26 30.28 1.198 2 18 30.28 1.001 3 9 30.45 0.887 4 1 32.28 0.525 0.837 0.505 Intermembral index 0 43 68.01 1.652 1 26 67.78 1.683 2 18 67.82 1.490 3 9 68.11 1.652 4 1 72.07 0.694 1.725 0.151 a significant ( p < 0.05). b The assumptions of the F test are violated where conditions of homoscedasticity are not met.
128 Table 7 13. ANOVA Proportional indices by number of Harris lines N Mean S.D. Levines F p Male Crural index no Harris lines 56 83.72 2.159 one Harris line 30 84.05 2.588 two or more Harris lines 10 84.39 1.095 0.031 a, c Humerofemoral index no Harris lines 56 71.11 2.224 one Harris line 30 70.47 1.953 two or more Harris lines 10 71.65 1.692 0.614 1.519 0.224 Radiohumeral index no Harris lines 56 78.23 2.540 one Harris line 30 78.31 2.379 two or more Harris lines 10 78.70 2.048 0.772 0.155 0.857 Sitting height index no Harris lines 56 30.07 1.013 one Harris line 30 29.90 1.161 two or more Harris lines 10 30.36 1.636 0.121 0.641 0.529 Intermembral index no Harris lines 56 68.98 1.832 one Harris line 30 68.26 1.385 two or more Harris lines 10 69.44 1.508 0.261 2.578 0.081 b a significant ( p < 0.05) b near significant (0.05 < p< 0.10) c The assumptions of the F test are violated where conditions of homoscedasticity are not met.
129 Table 7 14. ANOVA Proportional indices by age of enamel defect N Mean S.D. Levines F p Male Crural index No stress 42 83.37 1.852 Infant stress 10 84.93 1.377 Childhood stress 18 83.93 1.980 Infant & childhood stress 37 84.32 2.789 0.000 a, b Humero femoral index No stress 42 70.69 2.015 Infant stress 10 71.20 1.087 Childhood stress 18 71.05 1.925 Infant & childhood stress 37 71.15 2.462 0.113 0.392 0.759 Radiohumeral index No stress 42 78.39 2.333 Infant stress 10 78.91 1.972 Childhood stress 18 78.63 2.495 Infant & childhood stress 37 78.09 2.526 0.966 0.414 0.743 Sitting height index No stress 42 30.03 1.127 Infant stress 10 29.72 0.990 Childhood stress 18 30.46 1.410 Infant & childhood stress 37 29.99 0.944 0.289 1.162 0.328 Intermembral index No stress 42 68.76 1.869 Infant stress 10 68.87 0.770 Childhood stress 18 68.99 1.620 Infant & childhood stress 37 68.73 1.764 0.163 0.111 0.953 Female Crural index No stress 43 83.00 2.288 Infant stress 13 81.49 3.159 Childhood stress 19 83.22 2.607 Infant & childhood stress 22 82.95 2.504 0.605 1.45 0.233 Humerofemoral index No stress 43 70.60 2.114 Infant stress 13 70.03 2.655 Childhood stress 19 70.46 1.693 Infant & childhood stress 22 71.00 2.230 0.398 0.588 0.624 Radiohumeral index No stress 43 76.30 2.076 Infant stress 13 75.03 2.008 Childhood stress 19 76.52 2.492 Infant & childhood stress 22 75.81 2.021 0.791 1.58 0.200
130 Table 7 14. Continued N Mean S.D. Levines F p Sitting height index No stress 43 30.50 1.247 Infant stress 13 30.16 1.226 Childhood stress 19 30.36 1.205 Infant & childhood stress 22 30.45 0.935 0.586 0.296 0.828 Intermembral index No stress 43 68.01 1.652 Infant stress 13 67.52 1.813 Childhood stress 19 67.88 1.546 Infant & childhood stress 22 68.22 1.720 0.936 0.503 0.681 a significant ( p < 0.05) b The assumptions of the F test are violated where conditions of homoscedasticity are not met.
131 Table 7 15. ANOVA Proportional indices by infant enamel defect N Mean S.D. Levines F p Male Crural index No stress 42 83.37 1.852 Infant stress 47 84.45 2.554 0.016 a, c Humerofemoral index No stress 42 70.69 2.015 Infant stress 47 71.16 2.230 0.522 1.108 0.296 Radiohumeral index No stress 42 78.39 2.333 Infant stress 47 78.26 2.423 0.905 0.064 0.802 Sitting height index No stress 42 30.03 1.127 Infant stress 47 29.93 0.950 0.232 0.187 0.666 Intermembral index No stress 42 68.76 1.869 Infant stress 47 68.76 1.599 0.533 0.000 0.999 Female Crural index No stress 43 83.00 2.288 Infant stress 35 82.41 2.812 0.396 1.053 0.308 Humerofemoral index No stress 43 70.60 2.114 Infant stress 35 70.64 2.406 0.330 0.005 0.944 Radiohumeral index No stress 43 76.30 2.076 Infant stress 35 75.52 2.022 0.735 2.816 0.097 b Sitting height index No stress 43 30.50 1.247 Infant stress 35 30.34 1.044 0.248 0.351 0.555 Intermembral index No stress 43 68.01 1.652 Infant stress 35 67.96 1.762 0.711 0.017 0.896 a significant ( p < 0.05) b near significant (0.05 < p< 0.10) c The assumptions of the F test are violated where conditions of homoscedasticity are not met.
132 Table 7 16. ANOVA Proportions by age of Harris line N Mean S.D. Levines F p Male Crural index No stress 56 83.72 2.159 Infant stress 36 84.15 2.412 Childhood stress 1 83.39 Infant & childhood stress 3 84.20 0.920 0.200 0.299 0.826 Humerofemoral index No stress 56 71.11 2.224 Infant stress 36 70.55 1.880 Childhood stress 1 73.51 Infant & childhood stress 3 72.43 1.693 0.457 1.547 0.208 Radiohumeral index No stress 56 78.23 2.540 Infant stress 36 78.60 2.285 Childhood stress 1 75.63 Infant & childhood stress 3 77.03 1.653 0.478 0.876 0.456 Sitting height index No stress 56 30.07 1.013 Infant stress 36 29.94 1.236 Childhood stress 1 31.31 Infant & childhood stress 3 30.58 2.056 0.218 0.754 0.523 Intermembral index No stress 56 68.98 1.832 Infant stress 36 68.42 1.466 Childhood stress 1 70.40 Infant & childhood stress 3 69.60 1.391 0.365 1.330 0.269 Female Crural index No stress 56 82.67 2.534 Infant stress 29 83.26 2.821 Childhood stress 2 82.91 1.123 0.332 0.483 0.618 Humerofemoral index No stress 56 70.56 2.139 Infant stress 29 70.86 2.331 Childhood stress 2 69.62 1.177 0.540 0.399 0.672 Radiohumeral index No stress 56 76.22 2.194 Infant stress 29 75.97 2.213 Childhood stress 2 76.93 2.171 0.963 0.247 0.782 Sitting height index No stress 56 30.41 1.097 Infant stress 29 30.52 1.172 Childhood stress 2 30.22 0.684 0.693 0.134 0.875
133 Table 7 16. Continued N Mean S.D. Levines F p Intermembral index No stress 56 68.06 1.640 Infant stress 29 68.03 1.805 Childhood stress 2 67.33 0.724 0.435 0.179 0.836
134 Figure 7 1. Scaling crural index to size Figure 7 2. Scaling humerofemoral index to size Males y = 0.09x + 69.70 R = 0.069 Females y = 0.03x + 78.77 R = 0.005 74.00 76.00 78.00 80.00 82.00 84.00 86.00 88.00 90.00 92.00 130.00 140.00 150.00 160.00 170.00 180.00Crural IndexSkeletal Height (SKH) Males Females Males y = 0.04x + 77.63 R = 0.017 Females y = 0.13x + 89.29 R = 0.143 64.00 66.00 68.00 70.00 72.00 74.00 76.00 78.00 80.00 130.00 140.00 150.00 160.00 170.00 180.00Humerofemoral IndexSkeletal Height (SKH) Males Females
135 Figure 7 3. Scaling radiohumeral index to size Figure 7 4 Scaling intermembral index to size Males y = 0.05x + 86.02 R = 0.018 Females y = 0.08x + 64.01 R = 0.058 70.00 72.00 74.00 76.00 78.00 80.00 82.00 84.00 86.00 130.00 140.00 150.00 160.00 170.00 180.00Radiohumeral IndexSkeletal Height (SKH) Males Females Males y = 0.10x + 84.56 R = 0.085 Females y = 0.10x + 82.87 R = 0.150 60.00 62.00 64.00 66.00 68.00 70.00 72.00 74.00 76.00 78.00 80.00 130.00 140.00 150.00 160.00 170.00 180.00Intermembral IndexSkeletal Height (SKH) Males Females
136 Figure 7 5. Components of proportional ratios and long bone lengths scaled to skeletal height (male) y = 3.68x 109.44 R = 0.873 y = 2.43x 48.42 R = 0.747 y = 3.53x 161.20 R = 0.866 y = 1.74x 12.46 R = 0.673 y = 4.17x 60.88 R = 0.774 y = 7.21x 270.65 R = 0.913 200 300 400 500 600 700 800 900 1000 1100 140 145 150 155 160 165 170 175 180Average Length (mm)SKH (cm) Femur (max) Humerus Tibia Radius Radius + Humerus Tibia + Femur
137 Figure 7 6. Components of proportional ratios and long bone lengths scaled to skeletal height (female) y = 3.70x 104.73 R = 0.858 y = 2.06x + 7.34 R = 0.744 y = 3.18x 103.36 R = 0.811 y = 1.81x 30.86 R = 0.721 y = 3.87x 23.53 R = 0.784 y = 6.88x 208.09 R = 0.893 150 250 350 450 550 650 750 850 950 130 135 140 145 150 155 160 165Average Length (mm)SKH (cm) Femur (max) Humerus Tibia Radius Radius + Humerus Tibia + Femur
138 Figure 7 7. RMA for crural indices divided by dental enamel defects (male) Unstressed y = 1.12x 0.27 R = 0.873 Stressed y = 1.16x 0.34 R = 0.798 Pooled RMA y = 1.14x 0.31 R = 0.819 1.52 1.54 1.56 1.58 1.6 1.62 1.64 1.66 1.68 1.7 1.6 1.62 1.64 1.66 1.68 1.7 1.72 1.74Log TibiaLog Maximum Femur Unstressed Stressed Unstressed Stressed Pooled RMA
139 Figure 7 8. RMA for humerofemoral indices divided by dental enamel defects (male) Unstressed y = 0.86x + 0.08 R = 0.747 Stressed y = 1.08x 0.28 R = 0.744 Pooled RMA y = 0.10x 0.14 R = 0.734 1.46 1.48 1.5 1.52 1.54 1.56 1.58 1.6 1.6 1.62 1.64 1.66 1.68 1.7 1.72 1.74Log HumerusLog Maximum Femur Unstressed Stressed Unstressed Stressed Pooled RMA
140 Figure 7 9 RMA for radiohumeral indices divided by dental enamel defects (male) Unstressed y = 1.08x 0.23 R = 0.681 Stressed y = 0.91x + 0.03 R = 0.727 Pooled RMA y = 0.97x 0.057 R = 0.706 1.34 1.36 1.38 1.4 1.42 1.44 1.46 1.48 1.5 1.46 1.48 1.5 1.52 1.54 1.56 1.58 1.6Log RadiusLog Humerus Unstressed Stressed Unstressed Stressed Pooled RMA
141 Figure 7 10. RMA for sitting height indices divided by dental enamel defects (male) Unstressed y = 1.10x 0.74 R = 0.435 Stressed y = 1.03x 0.60 R = 0.371 Pooled RMA y = 1.06x 0.66 R = 0.398 1.6 1.62 1.64 1.66 1.68 1.7 1.72 1.74 1.76 2.16 2.17 2.18 2.19 2.2 2.21 2.22 2.23 2.24 2.25Log STHLog SKH Unstressed Stressed Unstressed Stressed Pooled RMA
142 Figure 7 11. RMA for intermembral indices divided by dental enamel defects (male) Unstressed y = 0.82x + 0.18 R = 0.792 Stressed y = 0.96x 0.08 R = 0.835 Pooled RMA y = 0.91x + 0.02 R = 0.806 1.7 1.72 1.74 1.76 1.78 1.8 1.82 1.84 1.86 1.86 1.88 1.9 1.92 1.94 1.96 1.98 2 2.02Log (Radius + Humerus)Log (Tibia + Max Femur) Unstressed Stressed Unstressed Stressed Pooled RMA
143 Figure 7 12. RMA for crural indices divided by dental ename l defects (female) Unstressed y = 1.04x 0.14 R = 0.801 Stressed y = 1.09x 0.24 R = 0.710 Pooled RMA y = 1.07x 0.20 R = 0.758 1.48 1.5 1.52 1.54 1.56 1.58 1.6 1.62 1.64 1.58 1.6 1.62 1.64 1.66 1.68 1.7 1.72Log TibiaLog Maximum Femur Length Unstressed Stressed Unstressed Stressed Pooled RMA
144 Figure 7 13. RMA for humerofemoral indices divided by dental enamel defects (female) Unstressed y = 0.81x + 0.16 R = 0.742 Stressed y = 0.86x + 0.08 R = 0.695 Pooled RMA y = 0.84x + 0.10 R = 0.725 1.42 1.44 1.46 1.48 1.5 1.52 1.54 1.56 1.58 1.6 1.62 1.64 1.66 1.68 1.7 1.72Log HumerusLog Maximum Femur Unstressed Stressed Unstressed Stressed Pooled RMA
145 Figure 7 14. RMA for radiohumeral indices divided by dental enamel defects (female) Unstressed y = 1.09x 0.25 R = 0.729 Stressed y = 1.25x 0.49 R = 0.767 Pooled RMA y = 1.19x 0.40 R = 0.759 1.28 1.3 1.32 1.34 1.36 1.38 1.4 1.42 1.44 1.46 1.44 1.46 1.48 1.5 1.52 1.54 1.56Log RadiusLog Humerus Unstressed Stressed Unstressed Stressed Pooled RMA
146 Figure 7 15. RMA for sitting height indices divided by dental enamel defects (female) Unstressed y = 1.13x 0.79 R = 0.416 Stressed y = 1.28x 1.13 R = 0.532 Pooled RMA y = 1.21x 0.98 R = 0.489 1.58 1.6 1.62 1.64 1.66 1.68 1.7 1.72 2.12 2.13 2.14 2.15 2.16 2.17 2.18 2.19 2.2 2.21 2.22Log STHLog SKH Unstressed Stressed Unstressed Stressed Pooled RMA
147 Figure 7 16. RMA for intermembral indices divided by dental enamel defects (female) Unstressed y = 0.82x + 0.18 R = 0.835 Stressed y = 0.92x 0.02 R = 0.806 Pooled RMA y = 0.88x + 0.05 R = 0.822 1.66 1.68 1.7 1.72 1.74 1.76 1.78 1.8 1.82 1.84 1.86 1.88 1.9 1.92 1.94 1.96 1.98Log (Radius + Humerus)Log (Tibia + Max Femur) Unstressed Stressed Unstressed Stressed Pooled RMA
148 Figure 7 17. RMA for crural indices divided by Harris lines (male) Absent y = 1.13x 0.29 R = 0.836 Present y = 1.21x 0.42 R = 0.778 Pooled RMA y = 1.16x 0.34 R = 0.812 1.48 1.50 1.52 1.54 1.56 1.58 1.60 1.62 1.64 1.66 1.68 1.70 1.58 1.60 1.62 1.64 1.66 1.68 1.70 1.72 1.74Log TibiaLog Maximum Femur Absent Present Absent Present Pooled RMA
149 Figure 7 18. R MA for humerofemoral indices divided by Harris lines (male) Absent y = 0.10x 0.14 R = 0.730 Present y = 0.10x 0.14 R = 0.754 Pooled RMA y = 0.99x 0.14 R = 0.738 1.42 1.44 1.46 1.48 1.50 1.52 1.54 1.56 1.58 1.60 1.58 1.60 1.62 1.64 1.66 1.68 1.70 1.72 1.74Log HumerusLog Maximum Femur Absent Present Absent Present Pooled RMA
150 Figure 7 19. RMA for radiohumeral indices divided by Harris lines (male) Absent y = 0.96x 0.04 R = 0.620 Present y = 1.02x 0.14 R = 0.727 Pooled RMA y = 0.98x 0.08 R = 0.661 1.28 1.30 1.32 1.34 1.36 1.38 1.40 1.42 1.44 1.46 1.48 1.50 1.44 1.46 1.48 1.50 1.52 1.54 1.56 1.58 1.60Log RadiusLog Humerus Absent Present Absent Present Pooled RMA
151 Figure 7 20. RMA for sitting height indices divided by Harris lines (male) Absent y = 0.94x 0.38 R = 0.437 Present y = 1.13x 0.81 R = 0.294 Pooled RMA y = 1.02x 0.57 R = 0.365 1.58 1.60 1.62 1.64 1.66 1.68 1.70 1.72 1.74 1.76 2.12 2.14 2.16 2.18 2.20 2.22 2.24 2.26Log STHLog SKH Absent Present Absent Present Pooled RMA
152 Figure 7 21. RMA for intermembral indices divided by Harris lines (male) Absent y = 0.89x + 0.05 R = 0.802 Present y = 0.91x + 0.02 R = 0.841 Pooled RMA y = 0.90x + 0.04 R = 0.815 1.66 1.68 1.70 1.72 1.74 1.76 1.78 1.80 1.82 1.84 1.86 1.84 1.86 1.88 1.90 1.92 1.94 1.96 1.98 2.00 2.02Log (Humerus + Radius)Log (Max Femur + Tibia) Absent Present Absent Present Pooled RMA
153 Figure 7 22. RMA for crural indices divided by Harris lines (female) Absent y = 1.06x 0.18 R = 0.749 Present y = 0.97x 0.03 R = 0.681 Pooled RMA y = 1.03x 0.13 R = 0.724 1.48 1.50 1.52 1.54 1.56 1.58 1.60 1.62 1.64 1.66 1.58 1.60 1.62 1.64 1.66 1.68 1.70 1.72 1.74Log TibiaLog Maximum Femur Absent Present Absent Present Pooled RMA
154 Figure 7 23. RMA for humerofemoral indices divided by Harris lines (female) Absent y = 0.90x + 0.01 R = 0.738 Present y = 0.69x + 0.36 R = 0.728 Pooled RMA y = 0.84x + 0.12 R = 0.724 1.42 1.44 1.46 1.48 1.50 1.52 1.54 1.56 1.58 1.58 1.60 1.62 1.64 1.66 1.68 1.70 1.72 1.74Log HumerusLog Femur Absent Present Absent Present Pooled RMA
155 Figure 7 24. RMA for radiohumeral indices divided by Harris lines (female) Absent y = 1.15x 0.35 R = 0.787 Present y = 1.30x 0.56 R = 0.603 Pooled RMA y = 1.19x 0.40 R = 0.736 1.32 1.34 1.36 1.38 1.40 1.42 1.44 1.46 1.48 1.50 1.44 1.46 1.48 1.50 1.52 1.54 1.56 1.58Log RadiusLog Humerus Absent Present Absent Present Pooled RMA
156 Figure 7 25. RMA for sitting heig ht indices divided by Harris lines (female) Absent y = 1.25x 1.05 R = 0.578 Present y = 1.16x 0.85 R = 0.433 Pooled RMA y = 1.22x 0.98 R = 0.526 1.58 1.60 1.62 1.64 1.66 1.68 1.70 1.72 1.74 2.12 2.14 2.16 2.18 2.20 2.22 2.24Log STHLog SKH Absent Present Absent Present Pooled RMA
157 Figure 7 26. RMA for intermembral indices divided by Harris lines (female) Absent y = 0.94x 0.04 R = 0.791 Present y = 0.77x + 0.27 R = 0.784 Pooled RMA y = 0.88x + 0.06 R = 0.785 1.68 1.70 1.72 1.74 1.76 1.78 1.80 1.82 1.84 1.84 1.86 1.88 1.90 1.92 1.94 1.96 1.98 2.00Log (Humerus + Radius)Log (Max Femur + Tibia) Absent Present Absent Present Pooled RMA
158 CHAPTER 8 RESULTS: TRADITIONA L AND NON TRADITIONAL MARKERS OF GROWTH DISRUPTION IN THE AD ULT SKELETON Linear enamel hyp oplasias and Harris lines are standard markers used to assess growth disruption in the adult skeleton. It has been suggested that the diameter of vertebral neural canals provide a similar line of evidence in that they are affected by stressors and cease g rowth early in life, thus preserving a marker of early stress events. Deficits in head circumference measurements, a common marker of growth disruption in anthropometric studies of living children, may also persist into adulthood in some individuals (Brandt et al. 2003). This chapter compares head circumference and vertebral neural canal shape to linear enamel hypoplasias and Harris lines to see if there is a significant relationship between these variables. Vertebral Neural Canals and Traditional Stress Markers The anteroposterior diameters of the lumbar vertebrae (L1 -L4) are 70% complete at birth (Jeffrey et al. 2003). The mediolateral ( transverse ) neural canal diameters and vertebral body heights continue to grow through childhood, and are thus more likely to experience catch up before completing growth in the case of growth disruption. Small anteroposterior vertebral neural canal diameters have been associated with low birth weight, maternal smoking and protein deficient childhood diets (Clark, 1988; Jeffrey et al. 2003). This investigation will determine whether vertebral neural canal (VNC) size is associated with other skeletal stress markers and whether or not VNC size is associated with adult stature and proportions. To determine the relations hip between vertebral neural canal size and other more traditional stress markers, individuals were divided into groups based on the presence or absence of enamel defects as well as Harris lines. These groups were tested for normality and homoscedasticity and one -way ANOVAs were performed on those groups meeting these assumptions. Where assumptions were not met, nonparametric Mann -Whitney U tests were
159 performed. ANOVAs were performed on anteroposterior (AP) and mediolateral (ML ) diameters and the vertebral canal index (AP/ ML ) on L1, averaged thoracic and averaged lumbar measurements. The ANOVAs were performed on a combined sex sample as well as on samples divided by sex. ANOVAs dividing the sample by enamel defects (CBTS) yielded significant res ults for the anteroposterior diameters of L1 and the averaged lumbar samples for the combined sex and male samples (Table 8 1). Mann -Whitney U tests yielded significant results for the averaged thoracic anteroposterior diameters in both the male and combi ned sex samples (Table 8 2). Significant results were found for mediolateral diameters in the case of the male averaged lumbar sample alone. The vertebral canal index measures were found to be significantly different for L1 (male and combined sex). In no case were female vertebral canal sizes in any dimension, significantly related to enamel defects (CBTS). Unlike the ANOVAs for enamel defects, ANOVAs dividing the sample by the presence or absence of Harris lines (HL) only yielded significant results for the female samples. In addition, the significant results ( p < 0.05) were for the mediolateral diameters of the averaged thoracic and lumbar measurements. These ANOVAs are summarized in Table 8 3 and the nonparametric tests are summarized in Table 8 4. Near significant (0.05 < p < 0.10) results were found for the vertebral canal index of L1 among females which prompted the use of r esampli ng statistics as an independent test. The resampling procedure tested the alternative hypothesis that the verteb ral neural canal index was greater in the unstressed group. No support was found for the alternative hypothesis in question (p = 0. 9774). However, the inverse relationship is significant (1 p = 0.0226). The vertebral canal index of L1 among females i s significantly greater in the stressed
160 sample. This means that the AP diameter of the L1 VNC is relatively greater than the ML diameter among stressed females. Because of the sex differences found in the one -way ANOVAs, two -way ANOVAs were performed to de termine whether or not there was any interaction effect between sex and the traditional stress markers. No significant results were found, but enamel defects (CBTS) and sex were found to have a near significant interaction effect for the averaged lumbar a nteroposterior diameters ( p = 0.079). Near significant results were also found for an interaction effect between Harris lines and sex for L1 and the averaged thoracic vertebral canal indices ( p = 0.094 and p = 0.057, respectively). Clark (1988) suggest s that the relationship between vertebral neural canals (VNC) and vertebral body heights (VBHs) is a critical one to examine. If VNC and VBH measurements are reduced, Clark considers this to be an indicator of chronic growth disruption. If VNC measuremen ts are reduced but VBH measurements are not, then catch up growth may have occurred. ANOVAs were performed on vertebral body heights divided by the presence or absence of enamel defects (CBTS) and Harris lines (HL). When assumptions of normality or homos cedasticity were not met, Mann -Whitney U tests were performed to test the difference between groups. The results of these tests, summarized in Tables 8 5 through 8 8, demonstrate that there are no significant differences in vertebral body height in sample s grouped by the presence or absence of enamel defects or Harris lines. Reduced major axis regressions (RMAs) of VNC measurements over VBH were performed to see if there were significant scaling differences between those with traditional indicators of st ress (enamel defects and Harris lines) and those without. A few significant differences were found in the slopes of RMAs in females. When the sample was divided by
161 enamel defects (CBTS), the significant differences were found in AP diameters of the lumba r vertebrae (L1 and averaged) (see Table 8 9). When the sample was divided by the presence or absence of Harris lines (HL), near significant differences (0.05 < p < 0.10) in slope were found in the transverse diameters of the lumbar vertebrae (averaged an d L1) (see Table 8 10). Significant ( p < 0.05) results suggest that VNC size increases at a faster rate than VBH in stressed groups. Intercept differences were also found for females in the averaged mediolateral measurement of the thoracic and lumbar vertebrae when the sample was divided by HL. In males, intercept difference s were found for the averaged mediolateral measurements of the thoracic vertebrae whether the sample was divided by CBTS or HL markers. Nea r significant intercept differences were found for the anteroposterior diameters of the averaged lumbar canals in males. The significant intercept differences suggest that for a given VBH, VNC size is significantly different. Where significant difference s in slope occur, intercept differences are meaningless and are therefore not calculated. Figures 8 1 through 8 8 graph the reduced major axis regressions in which significant or near significant differences in slopes or intercepts were found. It is impo rtant to note that in most of these regressions the R2 values are low, suggesting that allometry can only explain part of the observed variation. Vertebral Neural Canals and Stature Previous studies have determined that tall individuals tend to have wider vertebrae, and hence wider mediolateral measurements of their vertebral neural canals (Clark et al. 1985; Porter et al. 1987; Porter and Pavitt 1987); therefore independence should not be assumed between skeletal height and mediolateral vertebral neural canal diameters. In contrast, the same research found that anteroposterior measures of the vertebral neural canals were independent of tibial length or other anthropometrics (Clark et al. 1985; Porter et al. 1987).
162 Skeletal height (SKH) and the anteroposterior measurements of L1, the average lumbar and the average thoracic vertebral neural canals were all found to have normal distributions using the Shapiro Wilk test. Pearsons correlations and the non-parametric Kendalls tau correlations were perform ed comparing the anteroposterior vertebral canal of the averaged thoracic vertebrae and SKH, L1 and SKH, the averaged lumbar vertebrae and SKH. All correlations were found to be highly significant with the exception of the anteroposterior dimensions of the averaged lumbar canals among females (see Table 8 11). The correlation between skeletal height and anteroposterior vertebral neural canal diameters suggest that larger VNC diameters may be an effect of body size despite previously published results (Cla rk et al. 1985; Porter et al. 1987). As a heuristic exercise, the relationship between vertebral canal size and height was further illustrated by dividing the population into quartiles based on the anteroposterior diameter of L1 and comparing skeletal he ight. Box plots show that skeletal height is greater in those individuals in the highest quartile for L1 anteroposterior diameter than those in the lowest quartile ( Figures 8 9 and 8 10). However, because VNC diameters may be dependent on overall body si ze, anteroposterior diameters were size standardized by dividing the VNC measurements by vertebral body height. Size -standardized VNC measurements and SKH cannot be considered independent variables; therefore, the size-standardized box plots compare maxim um femur length between the highest and lowest quartiles in males and females (see Figures 8 1 1 and 8 1 2 ). Interpretation of these results is difficult because femur length is also correlated with body size. Head Circumference and Traditional Stress Markers Head circumference, an anthropometric indicator of stunted growth in children, is also a measurement typically included in autopsy reports. Two methods were used for acquiring head circumference measures in this study. All h ead circumference meas urements of individuals from
163 the Terry collection were collected by approximating the anthropometric technique used in growing children (Zerran, 2007). A flexible tape measure was placed around the head from opisthocranion to a point just above the brow r idge. This measurement was recorded in millimeters for each individual. Wax (approximating the width of the saw blade) was used to hold the skulls together for measurement in specimens where either the calotte was removed or the skull was hemisected. At the Hamann Todd collection, hemisected skulls are all reconstructed with wire and head circumference measures were difficult to obtain accurately. Due to this complication and the easy access to autopsy records in the HamannTodd database, I decided to o btain the Hamann Todd head circumference measurements from the collection s records. While I can vouch for the consistency of my own data collection techniques, a cursory inspection of the HamannTodd data caused me to question the reliability of head cir cumference measures provided in the autopsy records. Because both skeletal height and head circumference are not assumed to be dependent o n overall body size correlations were performed to examine this relationship in the study sample. Among the males and females measured in the Terry collection, a Pearsons correlation found that head circumference and skeletal height were significantly correlated (male p = 0.011; female p = 0.028). In the HamannTodd collection, a Pearsons correlation between head circumference and skeletal height was found to be insignificant ( p = 0.631) in the male sample. Because head circumference in the female sample is not normally distributed, a Kendalls tau correlation was performed and was not found to be significant ( p = 0.097). The results of these correlations are summarized in table 8 11. The fact that head circumference and skeletal height were not significantly correlated in the Hamann Todd sample cause me to question the integrity of
164 autopsy reports; accordingly they have been removed from all analyses comparing he a d circumference data to traditional stress markers. Data from the Terry collection was tested for normality and homoscedasticity. Wher e these assumptions were met, one -way ANOVA s were performed to de termine whether or not there was a significant difference in head circumference between groups with and wit hout traditional stress markers (enamel defects and Harris lines). Mann Whitney U tests were used on all non parametric data. No significant differ ence was found in the male sample when divided by dental enamel defects or Harris lines, or in the female sample when divided by enamel defects (CBTS) (Table 8 1 2 ). The female sample divided by Harris lines was not found to meet the assumptions of an ANOV A and so a Mann -Whitney U test was performed. The difference in head circumference when the population was divided by the presence or absence of Harris lines was found to be insignificant ( 1.166 p = 0.244).
165 Table 8 1. ANOVAs of vertebral neural ca nal measurements by dental enamel defects (CBTS) N Mean S.D. Shapiro Wilk Levine's F p Combined Sex L1 anteroposterior Unstressed 84 17.68 1.458 Stressed 119 17.17 1.369 Total 203 17.38 1.426 0.411 0.507 6.557 0.011 a L1 mediolateral Unstressed 85 21.00 1.945 Stressed 119 21.01 1.737 Total 204 21.00 1.822 0.505 0.104 0.001 0.971 L1 AP/ ML Unstressed 84 0.85 0.073 Stressed 119 0.82 0.072 Total 203 0.83 0.073 0.466 0.761 6.171 0.014 a Thoracic (average) anteroposterior Unstressed 85 15.23 1.205 Stressed 116 14.84 1.028 Total 201 15.00 1.120 0.044 a, c Thoracic (average) mediolateral Unstressed 85 16.76 1.413 Stressed 115 16.72 1.378 Total 200 16.74 1.390 0.120 0.544 0.050 0.823 Thoracic (average) AP/ ML Unstressed 85 0.92 0.073 Stressed 114 0.90 0.063 Total 199 0.91 0.068 0.561 0.337 4.050 0.046 a Lumbar (average) anteroposterior Unstressed 84 16.92 1.574 Stressed 119 16.46 1.584 Total 203 16.65 1.592 0.126 0.400 4.237 0.041 a Lumbar (average) mediolateral Unstressed 85 22.99 2.020 Stressed 119 22.59 1.729 Total 204 22.75 1.861 0.114 0.027 a, c Lumbar (average) AP/ ML Unstressed 85 0.74 0.061 Stressed 119 0.73 0.057 Total 204 0.74 0.059 0.432 0.624 1.252 0.264 Male L1 anteroposterior Unstressed 41 17.93 1.568 Stressed 65 17.18 1.286 Total 106 17.47 1.442 0.814 0.147 7.244 0.008 a
166 Table 8 1. Continued N Mean S.D. Shapiro Wilk Levine's F p L1 mediolateral Unstressed 42 21.89 1.570 Stressed 65 21.71 1.672 Total 107 21.78 1.628 0.418 0.915 0.316 0.575 L1 AP/ ML Unstressed 41 0.82 0.064 Stressed 65 0.79 0.060 Total 106 0.80 0.063 0.897 0.981 4.468 0.037 a Thoracic (average) anteroposterior Unstressed 42 15.45 1.319 Stressed 63 14.86 0.993 Total 105 15.10 1.164 0.106 0.046 a, c Thoracic (average) mediolateral Unstressed 42 17.44 1.276 Stressed 63 17.23 1.369 Total 105 17.31 1.330 0.174 0.881 0.644 0.424 Thoracic (average) AP/ ML Unstressed 42 0.90 0.074 Stressed 62 0.88 0.052 Total 104 0.88 0.062 0.251 0.019 a, c Lumbar (average) anteroposterior Unstressed 41 17.37 1.635 Stressed 65 16.50 1.618 Total 106 16.84 1.672 0.592 0.272 7.250 0.008 a Lumbar (average) mediolateral Unstressed 42 23.80 1.807 Stressed 65 23.12 1.674 Total 107 23.39 1.752 0.403 0.256 3.990 0.048 a Lumbar (average) AP/ ML Unstressed 42 0.73 0.054 Stressed 65 0.72 0.052 Total 107 0.72 0.053 0.726 0.901 1.869 0.174 Female L1 anteroposterior Unstressed 43 17.44 1.318 Stressed 54 17.16 1.476 Total 97 17.28 1.408 0.358 0.584 0.999 0.320 L1 mediolateral Unstressed 43 20.12 1.894 Stressed 54 20.16 1.416 Total 97 20.15 1.636 0.146 0.055 b 0.012 0.913
167 Table 8 1. Continued N Mean S.D. Shapiro Wilk Levine's F p L1 AP/ ML Unstressed 43 0.87 0.072 Stressed 54 0.85 0.071 Total 97 0.86 0.072 0.736 0.919 1.560 0.215 Thoracic (average) anteroposterior Unstressed 43 15.02 1.054 Stressed 53 14.81 1.078 Total 96 14.90 1.066 0.329 0.927 0.851 0.359 Thoracic (average) mediolateral Unstressed 43 16.10 1.223 Stressed 52 16.10 1.123 Total 95 16.10 1.163 0.341 0.590 0.000 0.999 Thoracic (average) AP/ ML Unstressed 43 0.94 0.064 Stressed 52 0.93 0.062 Total 95 0.94 0.063 0.599 0.804 0.995 0.321 Lumbar (average) anteroposterior Unstressed 43 16.48 1.396 Stressed 54 16.40 1.554 Total 97 16.44 1.479 0.417 0.433 0.078 0.781 Lumbar (average) mediolateral Unstressed 43 22.19 1.911 Stressed 54 21.95 1.583 Total 97 22.06 1.731 0.097 b 0.168 0.453 0.503 Lumbar (average) AP/ ML Unstressed 43 0.75 0.066 Stressed 54 0.75 0.057 Total 97 0.75 0.061 0.645 0.311 0.003 0.953 a Significant ( p < 0.05) b Near significant (0.05 < p< 0.10) c The assumptions of an ANOVA are violated where conditions of normality and homoscedasticity are not met.
168 Table 8 2. Nonparametric tests of VNC measurements by enamel defects (CBTS) N Mean Sum of Mann Wilcoxon Z p (2 tailed) Rank Ranks Whitney U W Combined Sex Thoracic (average) AP Unstressed 85 112.61 9571.5 Stressed 116 92.50 10729.5 3943.5 10729.5 2.421 0.015 a Lumbar (average) ML Unstressed 85 108.70 9239.5 Stressed 119 98.07 11670.5 4530.5 11670.5 1.268 0.205 Male Thoracic (average) AP Unstressed 42 61.32 2575.5 Stressed 63 47.45 2989.5 973.5 2989.5 2.286 0.022 Thoracic (average) AP/ ML Unstressed 42 57.89 2431.5 Stressed 62 48.85 3028.5 1075.5 3028.5 1.501 0.133 a Significant ( p < 0.05)
169 Table 8 3. ANOVAs of vertebral neural canal measurements by Harris lines (HL) N Mean S.D. ShapiroWilk Levine's F p Combined Sex L1 anteroposterior Unstressed 111 17.44 1.510 Stressed 71 17.41 1.304 Total 182 17.43 1.429 0.411 0.161 0.023 0.881 L1 mediolateral Unstressed 112 21.09 1.855 Stressed 71 20.82 1.765 Total 183 20.99 1.820 0.505 0.383 0.921 0.338 L1 AP/ ML Unstressed 111 0.83 0.072 Stressed 71 0.84 0.074 Total 182 0.83 0.073 0.466 0.993 0.765 0.383 Thoracic (average) anteroposterior Unstressed 109 15.07 1.120 Stressed 71 14.90 1.145 Total 180 15.00 1.129 0.044 a, c Thoracic (average) mediolateral Unstressed 110 16.78 1.384 Stressed 69 16.58 1.426 Total 179 16.70 1.400 0.120 0.828 0.874 0.351 Thoracic (average) AP/ ML Unstressed 109 0.91 0.066 Stressed 69 0.91 0.071 Total 178 0.91 0.068 0.561 0.434 0.007 0.935 Lumbar (average) anteroposterior Unstressed 111 16.79 1.643 Stressed 71 16.55 1.563 Total 182 16.70 1.612 0.126 0.535 0.964 0.327 Lumbar (average) mediolateral Unstressed 112 22.97 1.789 Stressed 71 22.57 1.951 Total 183 22.81 1.859 0.114 0.869 2.052 0.154 Lumbar (average) AP/ ML Unstressed 112 0.74 0.057 Stressed 71 0.74 0.058 Total 183 0.74 0.057 0.432 0.732 0.180 0.671
170 Table 8 3. Continued N Mean S.D. ShapiroWilk Levine's F p Male L1 anteroposterior Unstressed 55 17.61 1.462 Stressed 40 17.42 1.444 Total 95 17.53 1.450 0.814 0.916 0.395 0.531 L1 mediolateral Unstressed 56 21.82 1.773 Stressed 40 21.64 1.498 Total 96 21.75 1.658 0.418 0.105 0.261 0.611 L1 AP/ ML Unstressed 55 0.81 0.064 Stressed 40 0.81 0.060 Total 95 0.81 0.062 0.897 0.449 0.037 0.847 Thoracic (average) anteroposterior Unstressed 54 15.17 1.130 Stressed 40 15.03 1.251 Total 94 15.11 1.178 0.106 0.539 0.317 0.575 Thoracic (average) mediolateral Unstressed 55 17.27 1.420 Stressed 39 17.28 1.284 Total 94 17.27 1.358 0.174 0.471 0.002 0.964 Thoracic (average) AP/ ML Unstressed 54 0.89 0.063 Stressed 39 0.88 0.064 Total 93 0.89 0.064 0.251 0.925 1.194 0.277 Lumbar (average) anteroposterior Unstressed 55 17.01 1.652 Stressed 40 16.83 1.721 Total 95 16.94 1.675 0.592 0.773 0.282 0.597 Lumbar (average) mediolateral Unstressed 56 23.46 1.762 Stressed 40 23.36 1.775 Total 96 23.42 1.759 0.403 0.576 0.069 0.793 Lumbar (average) AP/ ML Unstressed 56 0.73 0.054 Stressed 40 0.72 0.048 Total 96 0.73 0.051 0.726 0.740 0.139 0.711
171 Table 8 3. Continued N Mean S.D. ShapiroWilk Levine's F p Female L1 anteroposterior Unstressed 56 17.27 1.550 Stressed 31 17.39 1.122 Total 87 17.31 1.407 0.358 0.036 a, c L1 mediolateral Unstressed 56 20.36 1.647 Stressed 31 19.77 1.516 Total 87 20.15 1.618 0.146 0.705 2.725 0.102 L1 AP/ ML Unstressed 56 0.85 0.074 Stressed 31 0.88 0.068 Total 87 0.86 0.073 0.736 0.985 3.913 0.051 b Thoracic (average) anteroposterior Unstressed 55 14.97 1.111 Stressed 31 14.74 0.989 Total 86 14.89 1.068 0.329 0.889 0.872 0.353 Thoracic (average) mediolateral Unstressed 55 16.30 1.173 Stressed 30 15.68 1.054 Total 85 16.08 1.165 0.341 0.257 5.892 0.017 a Thoracic (average) AP/ ML Unstressed 55 0.93 0.065 Stressed 30 0.95 0.059 Total 85 0.94 0.063 0.599 0.685 2.564 0.113 Lumbar (average) anteroposterior Unstressed 56 16.58 1.619 Stressed 31 16.20 1.272 Total 87 16.44 1.508 0.417 0.224 1.264 0.264 Lumbar (average) mediolateral Unstressed 56 22.48 1.693 Stressed 31 21.54 1.690 Total 87 22.14 1.742 0.097 b 0.503 6.161 0.015 a Lumbar (average) AP/ ML Unstressed 56 0.75 0.060 Stressed 31 0.76 0.063 Total 87 0.75 0.061 0.645 0.606 1.408 0.239 a Significant ( p < 0.05) b Near significant (0.05 < p< 0.10) c The assumptions of an ANOVA are violated where conditions of normality and homoscedasticity are not met.
172 Table 8 4. Nonparametric tests of vertebral neural canal measurements by Harris lines (HL) N Mean Sum of Mann Wilcoxon Z p (2 tailed) Rank Ranks Whitney U W Combined Sex Thoracic (average) AP Unstressed 109 92.92 10128.0 Stressed 71 86.79 6162.0 3606.0 6162.0 0.771 0.441 Female L1 AP Unstressed 56 42.920 2403.5 Stressed 31 45.950 1424.5 807.5 2403.5 0.536 0.592 Table 8 5. ANOVAs of vertebral body height measurements by dental enamel defects (CBTS) N Mean S.D. Shapiro Wilk Levine's F p Male L1 VBH Unstressed 42 26.36 1.387 Stressed 65 26.10 1.457 Total 107 26.21 1.429 0.324 0.817 0.846 0.360 Lumbar (average) VBH Unstressed 42 27.92 1.504 Stressed 65 27.84 1.504 Total 107 27.87 1.498 0.361 0.822 0.075 0.784 Thoracic (average) VBH Unstressed 42 20.30 0.883 Stressed 65 20.36 0.862 Total 107 20.34 0.867 0.055 b 0.741 0.124 0.726 Female L1 VBH Unstressed 43 25.41 1.678 Stressed 54 25.35 1.205 Total 97 25.38 1.426 0.071 b 0.027 a, b Lumbar (average) VBH Unstressed 43 27.17 1.553 Stressed 54 27.01 1.205 Total 97 27.08 1.365 0.318 0.092 b 0.344 0.559 Thoracic (average) VBH Unstressed 43 18.95 0.939 Stressed 54 18.69 0.787 Total 97 18.80 0.863 0.461 0.119 2.191 0.142 a Significant ( p < 0.05) b The assumptions of an ANOVA are violated where conditions of normality and homoscedasticity are not met.
173 Table 8 6. ANOVAs of vertebral body height measurements by Harris lines (HL) N Mean S.D. Shapiro Wilk Levine's F p Male L1 VBH Unstressed 56 26.08 1.512 Stressed 40 26.41 1.308 Total 96 26.22 1.433 0.324 0.456 1.244 0.268 Lumbar (average) VBH Unstressed 56 27.68 1.605 Stressed 40 28.18 1.348 Total 96 27.89 1.516 0.361 0.161 2.634 0.108 Thoracic (average) VBH Unstressed 56 20.31 0.887 Stressed 40 20.36 0.836 Total 96 20.33 0.862 0.055 b 0.624 0.062 0.803 Female L1 VBH Unstressed 56 25.40 1.555 Stressed 31 25.43 1.202 Total 87 25.41 1.432 0.071 a 0.222 0.006 0.939 Lumbar (average) VBH Unstressed 56 27.07 1.484 Stressed 31 27.18 1.131 Total 87 27.11 1.363 0.318 0.254 0.116 0.734 Thoracic (average) VBH Unstressed 56 18.85 0.855 Stressed 31 18.78 0.971 Total 87 18.83 0.893 0.461 0.678 0.152 0.697 a Near significant (0.05 < p< 0.10)
174 Table 8 7. Nonparametric tests of vertebral body height measurements by dental enamel defects (CBTS) N Mean Sum of Mann Wilcoxon Z p (2 tailed) Rank Ranks Whitney U W Male Thoracic (average) VBH Unstressed 42 52.74 2215.0 Stressed 65 54.82 3563.0 1312.0 2215.0 0.338 0.735 Female L1 VBH Unstressed 43 49.1 2111.5 Stressed 54 48.92 2641.5 1156.5 2641.5 0.033 0.974 Lumbar (average) VBH Unstressed 43 49.74 2139.0 Stressed 54 48.41 2614.0 1129.0 2614.0 0.232 0.816 Table 8 8. Nonparametric tests of vertebral body height measurements by Harris lines (HL) N Mean Sum of Mann Wilcoxon Z p (2 tailed) Rank Ranks Whitney U W Male Thoracic (average) VBH Unstressed 56 47.93 2684.0 Stressed 40 49.30 1972.0 1088.0 2684.0 0.238 0.812 Female L1 VBH Unstressed 56 43.89 2458.0 Stressed 31 44.19 1370.0 862.0 2458.0 0.053 0.958
175 Table 8 9. Comparison of RMA regressions (RMA2) divided by dental enamel defects (CBTS) N Intercept Slope 95% CI Scale t df p Intercept lower upper Diffs. Male L1 anteroposterior Unstressed 41 1.202 1.727 1.300 2.294 Pos. Stressed 65 0.713 1.375 1.076 1.756 Pos. 1.235 50 0.111 0.549 L1 mediolateral Unstressed 42 0.637 1.391 1.035 1.870 Pos. Stressed 65 0.612 1.375 1.074 1.760 Pos. 0.062 51 0.475 0.552 Avg. thoracic anteroposterior Unstressed 42 1.358 1.947 1.435 2.643 Pos. Stressed 63 0.825 1.526 1.191 1.955 Pos. 1.257 50 0.107 0.168 Avg. thoracic mediolateral Unstressed 42 0.961 1.684 1.240 2.288 Pos. Stressed 63 1.106 1.789 1.417 2.260 Pos. 0.319 50 0.375 0.047 a Avg. lumbar anteroposterior Unstressed 41 1.401 1.826 1.359 2.453 Pos. Stressed 65 1.421 1.826 1.431 2.330 Pos. 0.000 50 0.500 0.076 b Avg. lumbar mediolateral Unstressed 42 0.698 1.435 1.072 1.921 Pos. Stressed 65 0.583 1.348 1.051 1.729 Pos. 0.331 51 0.371 0.117 Female L1 anteroposterior Unstressed 43 0.358 1.138 0.858 1.509 Iso. Stressed 54 1.239 1.762 1.350 2.299 Pos. 2.290 50 0.013 a L1 mediolateral Unstressed 43 0.731 1.448 1.111 1.888 Pos. Stressed 54 0.701 1.429 1.088 1.876 Pos. 0.073 50 0.471 0.539 Avg. thoracic anteroposterior Unstressed 43 0.710 1.476 1.126 1.936 Pos. Stressed 53 1.019 1.722 1.305 2.273 Pos. 0.808 50 0.212 1.000 Avg. thoracic mediolateral Unstressed 43 0.801 1.571 1.182 2.089 Pos. Stressed 52 0.801 1.579 1.193 2.089 Pos. 0.024 49 0.490 0.680 Avg. lumbar anteroposterior Unstressed 43 0.905 1.480 1.103 1.986 Pos. Stressed 54 1.875 2.158 1.642 2.837 Pos. 1.906 49 0.031 a Avg. lumbar mediolateral Unstressed 43 0.776 1.480 1.099 1.993 Pos. Stressed 54 0.995 1.632 1.245 2.138 Pos. 0.492 49 0.312 0.306 a Significa nt ( p < 0.05) b Near significant (0.05 < p < 0.10)
176 Table 8 10. Comparison of RMA regressions (RMA2) divided by Harris lines (HL) N Intercept Slope 95% CI Scale t df p Intercept lower upper Diffs. Male L1 anteroposterior Unstressed 55 0.795 1.440 1.111 1.867 Pos. Stressed 40 1.196 1.714 1.283 2.291 Pos. 0.912 47 0.183 0.405 L1 mediolateral Unstressed 56 0.645 1.400 1.070 1.831 Pos. Stressed 40 0.696 1.429 1.035 1.971 Pos. 0.098 47 0.461 0.100 Avg. thoracic anteroposterior Unstressed 54 1.021 1.684 1.286 2.206 Pos. Stressed 40 1.441 2.000 1.465 2.730 Pos. 0.848 47 0.200 0.538 Avg. thoracic mediolateral Unstressed 55 1.172 1.842 1.427 2.378 Pos. Stressed 39 1.089 1.778 1.297 2.437 Pos. 0.178 46 0.430 0.093 b Avg. lumbar anteroposterior Unstressed 55 1.192 1.680 1.294 2.181 Pos. Stressed 40 1.812 2.095 1.548 2.836 Pos. 1.124 47 0.133 0.147 Avg. lumbar mediolateral Unstressed 56 0.533 1.320 1.015 1.717 Pos. Stressed 40 0.910 1.571 1.148 2.151 Pos. 0.865 47 0.196 0.147 Female L1 anteroposterior Unstressed 41 0.792 1.444 1.115 1.871 Pos. Stressed 65 0.728 1.400 1.014 1.933 Pos. 0.155 39 0.439 1.000 L1 mediolateral Unstressed 42 0.496 1.286 1.000 1.654 Iso. Stressed 65 1.094 1.700 1.273 2.270 Pos. 1.499 40 0.071 b 0.502 Avg. thoracic anteroposterior Unstressed 42 0.866 1.600 1.222 2.095 Pos. Stressed 63 0.568 1.364 0.987 1.885 Iso. 0.779 40 0.220 0.377 Avg. thoracic mediolateral Unstressed 42 0.765 1.550 1.185 2.028 Pos. Stressed 63 0.411 1.261 0.886 1.795 Iso. 0.956 38 0.173 0.041 a Avg. lumbar anteroposterior Unstressed 41 1.348 1.792 1.375 2.334 Pos. Stressed 65 1.421 1.833 1.284 2.617 Pos. 0.106 39 0.458 0.507 Avg. lumbar mediolateral Unstressed 42 0.558 1.333 1.024 1.736 Pos. Stressed 65 1.297 1.833 1.325 2.537 Pos. 1.563 39 0.063 b 0.042 a a Significa nt ( p < 0.05) b Near significant (0.05 < p < 0.10)
177 Table 8 11. Correlations between vertebral neural canal measures and skeletal height (SKH) Average thoracic AP L1 AP Average lumbar AP Combined Sex N 201 204 203 Pearson Correlation 0.291 0.300 0.287 Sig. (2 tailed) 0.000 a 0.000 a 0.000 a Kendalls tau 0.192 0.190 0.182 Sig. (2 tailed) 0.000 a 0.000 a 0.000 a Female N 96 97 97 Pearson Correlation 0.308 0.350 0.165 Sig. (2 tailed) 0.002 a 0.000 a 0.107 Kendalls tau 0.183 0.211 0.070 Sig. (2 tailed) 0.008 a 0.002 a 0.307 Male N 105 106 106 Pearson Correlation 0.323 0.349 0.361 Sig. (2 tailed) 0.001 a 0.000 a 0.000 a Kendalls tau 0.224 0.231 0.263 Sig. (2 tailed) 0.001 a 0.000 a 0.000 a a Significant ( p < 0.05) Table 8 1 2 Correlations between head circumference and skeletal height Collection Sex Correlation coefficient a p Terry Male 0.345 0.011 b Female 0.270 0.028 b Hamann Todd Male 0.082 0.631 Female 0.187 0.097 a Pearsons correlations were performed in all cases except the Hamann Todd females. Because head circumference measures among Hamann Todd females were nonparametric, Kendalls tau correlations were used. b Significant ( p < 0.05)
178 Table 8 13. ANOVAs of head circumference measurements by traditional stress markers N Mean S.D. Shapiro Wilk Levine's F p Divided by dental enamel defects (CBTS) Female Unstressed 29 510.66 14.659 Stressed 25 513.00 17.682 Total 54 511.74 16.016 0.587 0.705 0.284 0.596 Male Unstressed 28 526.14 17.016 Stressed 38 529.32 15.141 Total 66 527.97 15.914 0.053 0.997 0.637 0.428 Divided by Harris lines Female Unstressed 33 509.64 12.180 Stressed 20 514.55 20.985 Total 53 511.49 16.062 0.587 0.009 a, b Male Unstressed 37 529.49 14.521 Stressed 28 525.71 17.841 Total 65 527.86 16.014 0.053 0.203 0.883 0.351 a Significant ( p < 0.05) b The assumptions of an ANOVA are violated where conditions of normality and homoscedasticity are not met.
179 Figure 8 1. Reduced major axis regressions of mediolateral vertebral neural canal diameters of the averaged thoracic vertebrae over verte bral body heights in males divided by enamel defects (CBTS). Unstressed y = 1.69x 0.97 R = 0.052 Stressed y = 1.81x 1.13 R = 0.153 Pooled y = 1.77x 1.07 R = 0.105 1.14 1.16 1.18 1.2 1.22 1.24 1.26 1.28 1.3 1.32 1.34 1.26 1.27 1.28 1.29 1.3 1.31 1.32 1.33 1.34 1.35Average thoarcic mediolatral vertebral neural canal (ML VNC)Vertebral body height (VBH) Unstressed Stressed Unstressed RMA Stressed RMA Pooled RMA
180 Figure 8 2. Reduced major axis regressions of anteroposterior vertebral neural canal diameters of the averaged lumbar vertebrae over vertebral body heights in males divided by enamel defects ( CBTS). Unstressed y = 1.78x 1.33 R = 0.145 Stressed y = 1.79x 1.37 R = 0.044 Pooled y = 1.85x 1.44 R = 0.073 1.1 1.15 1.2 1.25 1.3 1.35 1.38 1.4 1.42 1.44 1.46 1.48 1.5 1.52Average lumbar anteroposterior vertebral neural canal (AP VNC)Vertebral body height (VBH) Unstressed Stressed Unstressed RMA Stressed RMA Pooled RMA
181 Figure 8 3 Reduced major axis regressions of anteroposterior vertebral neural canal diameters of L1 over vertebral body heights in females divided by enamel defects (CBTS). Unstressed y = 1.11x 0.32 R = 0.178 Stressed y = 1.80x 1.30 R = 0.064 Pooled y = 1.43x 0.77 R = 0.107 1.1 1.15 1.2 1.25 1.3 1.35 1.28 1.3 1.32 1.34 1.36 1.38 1.4 1.42 1.44 1.46 1.48L1 anteroposterior vertebral neural canal (AP VNC)Vertebral body height (VBH) Unstressed Stressed Unstressed RMA Stressed RMA Pooled RMA
182 Figure 8 4 Reduced major axis regressions of transverse vertebral neural canal diameters of the averaged lumbar vertebrae over vertebral body heights in females divided by enamel defects (CBTS). Unstressed y = 1.48x 0.90 R = 0.042 Stressed y = 2.128x 1.82 R = 0.010 Pooled y = 1.79x 1.34 R = 0.042 1.05 1.1 1.15 1.2 1.25 1.3 1.35 1.36 1.38 1.4 1.42 1.44 1.46 1.48 1.5Average lumbar anteroposterior vertebral neural canal (AP VNC)Vertebral body height (VBH) Unstressed Stressed Unstressed RMA Stressed RMA Pooled RMA
183 Figure 8 5 Reduced major axis regressions of mediolateral vertebral neural canal diameters of the averaged thoracic vertebrae over vertebral body heights in males divided by Harris lines (HL). Unstressed y = 1.84x 1.17 R = 0.121 Stressed y = 1.77x 1.08 R = 0.073 Pooled y = 1.81x 1.14 R = 0.102 1.14 1.16 1.18 1.2 1.22 1.24 1.26 1.28 1.3 1.32 1.34 1.26 1.27 1.28 1.29 1.3 1.31 1.32 1.33 1.34 1.35Average thoarcic mediolateral vertebral neural canal (MLVNC)Vertebral body height (VBH) Unstressed Stressed Unstressed RMA Stressed RMA Pooled RMA
184 Figure 8 6 Reduced major axis regressions of mediolateral vertebral neural canal diameters of L1 over vertebral body heights in females divided by Harris lines (HL). Unstressed y = 1.32x 0.55 R = 0.131 Stressed y = 1.64x 1.01 R = 0.405 Pooled y = 1.43x 0.70 R = 0.182 1.15 1.2 1.25 1.3 1.35 1.4 1.29 1.31 1.33 1.35 1.37 1.39 1.41 1.43 1.45 1.47L1 mediolateral vertebral neural canal (ML VNC)Vertebral body height (VBH) Unstressed Stressed Unstressed RMA Stressed RMA Pooled RMA
185 Figure 8 7 Reduced major axis regressions of transverse vertebral neural canal diameters of the averaged thoracic vertebrae over vertebral body heights in females divided by Harris lines (HL). Unstressed y = 1.58x 0.81 R = 0.026 Stressed y = 1.28x 0.43 R = 0.132 Pooled y = 1.51x 0.72 R = 0.056 1.1 1.12 1.14 1.16 1.18 1.2 1.22 1.24 1.26 1.28 1.3 1.2 1.22 1.24 1.26 1.28 1.3 1.32 1.34Average thoarcic mediolateral vertebral neural canal (ML VNC)Vertebral body height (VBH) Unstressed Stressed Unstressed RMA Stressed RMA Pooled RMA
186 Figure 8 8 Reduced major axis regressions of mediolateral vertebral neural canal diameters of the averaged lumbar vertebrae over vertebral body heights in females divided by Harris lines (HL). Unstressed y = 1.36x 0.60 R = 0.044 Stressed y = 1.85x 1.32 R = 0.243 Pooled y = 1.55x 0.87 R = 0.072 1.24 1.26 1.28 1.3 1.32 1.34 1.36 1.38 1.4 1.42 1.44 1.46 1.36 1.38 1.4 1.42 1.44 1.46 1.48 1.5Average lumbar mediolateral vertebral neural canal (ML VNC)Vertebral body height (VBH) Unstressed Stressed Unstressed RMA Stressed RMA Pooled RMA
187 Figure 8 9 Box plots of female skeletal height (SKH) divided by highest and lowest quartiles of vertebral neural canal anteroposterior diameters.
188 Figure 8 10. Box plots of male skeletal height (SKH) divided by highest and lowest quartiles of vertebral neural can al anteroposterior diameters.
189 Figure 8 1 1 Box plots of female femur length divided by highest and lowest quartiles of vertebral neural canal anteroposterior diameters size standardized using vertebral body heights.
190 Figure 8 1 2 Box plots of male maximum femur length divided by highest and lowest quartiles of vertebral neural canal anteroposterior diameters size standardized using vertebral body heights.
191 CHAPTER 9 DISCUSSION This study uses the presence or absence of indicators of growth arrest as an independent variable for comparison with skeletal measurements to determine whether or not the observable stress markers have any association with the skeletal morphology of adults. The null hypothesis is that there are n o metrically observable difference between the size and shape of individuals who have experienced growth disruption to those who have not, due to the corrective process of catch up growth. Three alternative hypotheses were tested with this research: H1: Individuals with independent indicators of growth disruption or stress have a smaller estimated stature than those individuals without such indicators of stress. H2: Because female growth is believed to be canalized, or less affected by stress conditio ns than male growth, males and females with evidence of stress have a lesser degree of sexual dimorphism than those without any evidence of stress. H3: Proportional differences are observable between those individuals with independent evidence of growth disruption and those without. Stature When the osteometric data associated with stature estimates were divided by the presence or absence of Harris lines to address the first alternative hypothesis, no significant difference was found. However, when dent al defects (CBTS) were used as the stress indicator, support was found for the first alternative hypothesis in the case of females, but not in the case of males. When dental enamel defects were coded by the TS criterion (any enamel defect) or LEH criterio n (any linear defect) the results were significant for females in sitting height (STH) and tibia length. Support for the first alternative hypothesis suggests that catchup growth following growth disruption is not always complete. We can therefore reje ct the null hypothesis as it applies to females, but not as it applies to the male sample. The fact that significant differences are found
192 with females but not with male s is of particular interest because it is the opposite of expectations based on how m ales and females are believed to respond to stress. Since females have greater fat and nutrient reserves, thought to be an adaptation for the increased demands of lactation and gestation, many researchers have hypothesized that they account for a canalize d growth trajectory in females. Due to this physiological difference between the sexes, males are thought to experience a greater reduction in lean body mass than females during periods of stress, and reduced body mass is accompanied by reduced skeletal g rowth during severe stress events (Stini 1975a). In considering the physiological differences between the sexes it is also important to note that Guatelli -Steinberg (1999) found that males and females are equally likely to form enamel hypoplasias. The o bserved frequency of Harris lines between the sexes vary among archaeological populations (e.g. Goodman and Clark, 1981; Hummert and Van Gerven, 1985; Martin, 1985; Rathbun, 1987). The relationship between stature and stress can be interpreted several ways The first explanation I will consider does not challenge prior notions of sex differentiated responses to growth disruption ; i nstead I hypothesize that a sex difference in parental investment may favor male children over female children, assisting the process of catch up growth in males. Culturally, males may be better attended. In situations where a ll may suffer, male children may recover sooner or more completely. In research on the demographics of natural disasters and famines, females appear to have some biological advantage over their male counterparts (Macintyre 2002); however, there are many famines in which females die at a higher rate during infancy and childhood (Agarwal 1990; Dyson 1991a 1991b; Kidane 1989; Mariam 1986). In famines, it has been suggested that the most feasible explanation for this trend is that female babies and children have less access to food than male children (Neumayer and Plmper 2007).
193 In other natural disasters, such as a flood, male success over females may be attributed to the survival skills that are more often taught to male children such as tree climbing, running and swimming (Neumayer and Plmper 2007). Whether or not preferential feeding of male children during times of stress characterized African -A mericans during the 1880s to 1930s is a question that could likely be addressed through a historical investigation, but there are also anthropological studies which may be informative. Ethnographic studies show that the quality of parental investment var ies depending on the gender of the offspring. The Trivers -Willard hypothesis states that parents should invest more in the offspring with the greatest reproductive variance. Because the variance in reproductive success is generally greater for human male s than females, parents who can afford to invest in their offspring have a potentially greater return by investing in males. On the other hand, a parent in poor condition may see a greater return (in terms of reproductive success in the next generation) b y investing in female offspring (Hrdy 1990; Trivers and Willard 1973). Because the skeletal remains used in this research were of Americans of African descent from the early 1900s most of whom were unclaimed by family members or given to the state at d eath (Hunt and Albanese 2005), it is fair to assume that the majority of these individuals were from a low socioeconomic background ; according to the Trivers Willard model, this scenario would favor female offspring. The effect suggested by the Trivers W illard hypothesis may, however, be most pronounced in nonindustrialized area of the world ; Keller et al. (2001) found no support for this hypothesis in a recent study of the modern United States T hey point out that any investment in male children may be beneficial in terms of reproductive success, thus counteracting the effect suggested by Trivers and Willard. CluttonBrock (1991) point s out that childrens work, rather than their direct reproductive success, may play the biggest role in the
194 inclusive f itness of parents. If males are contributing more to the subsistence economy, it may be more adaptive to invest in male offspring who will help both parents and siblings. Research shows that female juvenile mortality is higher in areas of the world where females contribute little to the subsistence economy (Arnold and Zhaoxiang 1986; Baigari 1986; Das Gupta 1987), and male juvenile mortality is higher than expected where the female contribution is high (Cronk 1989; Harpending and Pennington, 1991). A second explanation for why only female growth disruption is associated with reduced stature is that it actually may be an example of female physiological buffering at its finest. The only individuals sampled in this study were adults between the ages of 1 9 and 35 years at death. If juvenile mortality rates were higher for males than females the stressed females in this study may represent a wider range of stress severity than is observed in adult males Wood et al. (1992) refer to this type of problem as the osteological paradox. Differential mortality can affect the demography of a skeletal sample, and the subsequent interpretation of data from those samples. To address the osteological paradox in this study, it would be valuable to analyze hi storical documents to discern whether or not a sex difference in juvenile mortality existed during this time period. A third hypothesis for explaining the results of how stress and stature are related in males and females questions the core of the theory regarding physiological differences in stress response between the sexes. The typical explanation for the physiological stress buffering in females is always related to fat stores for the demands of lactation and gestation (Rivers 1982; Stini 1969) G reulichs (1976) study of secular changes among males and females remarked on the biological superiority of females, but also found that females were more variable than males in most morphological characteristics. While females do have the ability to re tain fat, this
195 fact seems less relevant during early growth than it is once an individual reaches puberty. Female children simply do not store fat in the way that female adults do. Owens (2002) found that in the United States, males are twice as likely t o die from parasitic infections as females. In some areas of the world they are nearly four times as vulnerable as females. However, this trend does not begin until men reach their midtwenties and may be hormonally related to the immunosuppressant natur e of testosterone or may be due simply to exposure risk. While male children are believed to be more vulnerable than their female counterparts, Owens (2002) reports that large differences between male and female mortality rates do not begin until after pu berty. Other research reports higher infant mortality in males as well as greater male vulnerability during infancy, while admitting that causal factors are poorly understood (Waldron 1983). When female mortality is higher, a cultural bias against femal e children is assumed to be the cause (Holden and Mace 1999; Neumayer and Plmper 2007; Waldron, 1983). The majority of stressed individuals used in this study developed stress markers in infancy (between birth and 3 years of age) a time period wher e lactation and gestation are hardly an issue. Research needs to focus on whether or not a sex differentiated vulnerability to disease and malnutrition exists during the early juvenile years independently of factors such as accidents and parental invest ment biases. If such differences exist during the growth period, what are the true causal factors? Physiological differences may only become critical in the later years of development as sex hormones begin to play a role. Sexual Dimorphism Resampling statistics were used to test the second alternative hypothesis that sexual dimorphism is greater in unstressed than stressed groups. No support was found for the alternative hypothesis. This hypothesis was formed on the premise that females have greate r resistance to environmental stress than their male counterparts; stressed males and females
196 would be expected to exhibit less sexual dimorphism due to the fact that female size is beli eved to be canalized while male size varies more with environmental conditions (Stini 1969, 1985). Not only were there no significant differences in sexual dimorphism scores between groups, but the differences based on enamel stress markers were in the opposite direction of expectations raw scores of dimorphism (calcula ted as ln X Mln X F) were greater in the stressed group. The greatest differences in dimorphism, although insignificant at p < 0.05, were for tibial (p = 0.064) and radial (p = 0.061) length. While it is interesting that both these measures are for distal limb elements (see discussion pertaining to proportions), the results are inconsistent For tibial length males have slightly longer tibiae among the stressed group and females have slightly longer tibiae in the unstressed group. For radial length, both males and females have slightly larger radii in the unstressed group. When Harris lines were used as the marker to differentiate stressed and unstressed groups, no compelling results were fou nd. The direction of dimorphism was not consistent and the p values found by resampling averaged 0.395 suggesting that the differences observed were not m eaningful. Harris line data as it relates to sexual dimorphism could be interpreted two ways: 1) Har ris lines are reliable markers of stress and the p values found suggest that sexual dimorphism is not different between stressed and unstressed groups, or 2) Harris lines are not a good indicator of stress given that they do not pick up any differences in dimorphism. Given other concerns with Harris lines (see discussion on the reliability of stress markers), it is reasonable to conclude that Harris line data does not contribute to understanding the relationship between sexual dimorphism and stress. A co rollary to the alternative hypothesis regarding dimorphism is that female size should be similar in stressed and unstressed groups while male size should be different. The results
197 from C hapter 5 found this to be untrue. When dental defects were used as t he stress marker, female measurements are si gnificant ly larger among the unstressed while male differences are not significant. Furthermore, because stressed females are shorter, dimorphism is actually greater in stressed populations, although not signifi cantly so. These findings challenge the hypothesis that sexual dimorphism is greater in unstressed groups as well as the theory that female growth is more canalized than in males. There are a few possible interpretations for why no support was found for either the null or alternative hypothe se s: 1) female growth is not more canalized than males and therefore dimorphism should not follow the pattern of the alternative hypothesis, 2) cultural buffering favors male children to a greater extent than biologica l buffering favors females, or 3) the alternative hypothesis is correct, but is not observed due to greater male mortality Each of these explanations are explored in the discussion of stature differences. Proportions The position that body proportions a re more heritable than body size (Mueller 1986; Stini 1975) has been challenged by numerous researchers over the past decade who have found t hat better environmental circumstances lead to longer relative l ower limb length (Bogin and Rios, 2003; Bogin et al., 2002; Dangour, 2001; Floyd, 2007, 2008; Fredriks et al. 2005; Frisancho et al. 2001; Li et al. 2007; Malina et al. 2004; Stinson, 2000; Wadsworth et al. 2002) and that t he relationship of proximal and dista l limb elements is associated with environmental effects. The length of the tibia has proven to be more variable than other limb segments, distal elements tend to be more variable than proximal limb segments (Holliday and Ruff 2001; Smith and Buschang 2004), and s ecular allometric trends have shown that as the environment improves the relative length of the tibia increases ( Meadows Jantz and Jantz 1999; Jantz and Owsley 1984; Meadows and Jantz 1995). Furthermore, because nutritional stress and growth disturbances are
198 greatest in the first two to three years of life, age of an insult may affect its influence (Beaton et al., 1990; Checkley et al., 1998; Martorell and Habicht, 1986; Martorell et al., 1995). The results from research on human proportions is what influenced the third alternative hypothesis, that proportional differences would be observable between those with and without indicators of growth disruption. Support for this alternative hypothesis is weak at best. To investigate how stress affe cted proportions, ANOVAs were performed on common proportional indices (crural index, humerofemoral index, radiohumeral index, sitting height index and intermembral index) to see whether or not these proportions differed based on the presence or absence of stress markers. None of these ANOVAs yielded significant results; however, near significant results were retested using a resampling procedure which yielded a significant result in the case of the male crural index divided by the presence or absence of e namel defects. Interestingly, the observed difference was in the opposite direction of that proposed by the alternative hypothesis. In this study, the crural index was found to be higher among the stressed group. Maximum tibia length was higher and the maximum femur length was lower among the stressed than among the unstressed. Because proportional indices are ratios and the components of the ratios are not independent, scaling procedures were conducted to qualify the results. Significant differences i n scaling were found in male humerofemoral and intermembral indices when the skeletal sample was divided according to presence or absence of enamel defects (CBTS). Figures 7 8 and 711 graph these scaling differences where femur (or lower limb ) length inc reases at a slower rate than humerus (or upper limb ) length in the stressed group. Significant differences in scal ing were found in female humero femoral indices and intermembral indices when the population was divided according to the presence, absence, and number of Harris lines. Figures 7 23 and 726
199 graph these scaling differences where femur (or lower limb ) length increases at a faster rate th an humerus (or upper limb ) length in the stressed group. T hese scaling differences are significant, but because the ANOVAs are not, the scaling effects are considered to be subtle. No significant scaling differences were observed for male crural indices where significant differences were found between stressed and unstressed groups. Proportional indices were also analyzed to see if the number of stress events could be associated with significant differences in proportions. The majority of these comparis ons were not significant. Although significant differences in intermembral indices were found between male individuals with no Harris lines and those with only one Harris line, when groups with no Harris lines and those with two Harris lines were compared there were no significant differences. Similarly, when intermembral indices in those males with one Harris line and those with more than one Harris line were compared, no significant difference was found. Observation of the mean values for these indice s do not allow for a simple explanation of these results and suggest that this may be a Type I statistical error (rejection of the null hypothesis when the null hypothesis is true). In most cases, proportional differences were not significantly associate d with the age at which a stress event occurred. The two exception s are for the radio humeral index among females and the crural index in males when the sample was divided into stress categories based on the presence or absence of enamel defects Within t he three age groupings, significant differences were found only between those individuals with CBTS enamel defects that occurred during infancy and those without stress during this time period (those individuals who only formed enamel defects during childh ood (3 7 years) were removed from this analysis to isolate the effects of infant stress). In the case of the radiohumeral index in females, the distal el ement
200 is relatively shorter. On the other hand, significant differences in the male crural index are due to longer distal limb segments among those stressed in infancy. To summarize, the proportional differences based on stress were subtle and did not follow the trends observed in previous research. N o proportional differences were found in sitting heig ht indices suggesting that leg -to -trunk proportions were not different between groups. There is some evidence that distal and proximal limb segments respond differently to stress; however, the direction of variation is not consistent. The expectation was that distal limb segments would be truncated in stressed groups, but in the case of male crural indices the opposite pattern was observed. A significant relationship between age and proportions only exists for one of the many comparisons made; however, t he significant difference does follow a pattern that has been suggested in previous studies. The pattern of these results creates more questions than it answer s The r2 values of the reduced major axis regressions have a median value of 0.735, but range from 0.29 to 0.84. This means that 16 71% of the variation observed can be attributed to an unmeasured variable. R2 values are particularly low when scaling SKH and STH. One lurking variable that should be considered is genetics Proportions are known to be strongly influenced by genetics (Mueller 1986; Stini 1975) and proportions are known to vary between populations (Eveleth, 1975, 1986; Eveleth and Tanner 1990; Hamill et al., 1973) Because population admixture among African Americans can be high (Reiner et al., 2005), it is possible that the genetic signature in this sample is creating too much noise to observe the effects of growth disruption on proportions. To truly observe environmental effects on proportions, it may be necessary to find a more genetically homogenous population with larger sample sizes.
201 The Reliability of Stress Indicators To interpret the results of the analyses used in this research, it is first necessar y to critique the stress markers used. Dental defects were coded in three ways for data analysis (see Table 4 4 for a summary of these coding criteria ). The TS coding criterion was the most lax in that any enamel defect (pit or line, bilateral or not) was recorded as stress, and the LEH coding criterion did not consider hypoplastic pits as enamel stress. The third, and most conservative coding criterion (CBTS) only considered and individual stressed if they had bilateral defects on their anterior denti tion. The concern with enamel defect coding criteria is that systemic stress is not the only cause of enamel defects. Using the TS criterion, a traumatic insult to the dentition that affects a single tooth is coded as systemic stress. The CBTS criterion eliminates individuals from analysis who have not retained the antimere of affected teeth. The conservative requirements of this coding criterion severely limited the number of individuals that could be included in this analysis. After selectively remov ing m any of the individuals deemed stressed based on TS criteria and unstressed by CBTS criteria were removed from further analysis eleven individuals in this study were coded differently depending on the criteria used. In C hapter 5, compari sons of stature estimates were performed using all three coding criteria (see Tables 5 2 through 5 4). The results of comparisons using TS and LEH coding criteria were nearly identical given that only one female in this sample had hypoplastic pitting in the absen ce of any linear enamel defect. The results of comparisons using LEH/TS coding criteria and the CBTS coding criterion were also very similar. Because the CBTS coding criterion is more conservative in recognizing defects as stress the fear is that stress ed individuals will be misrepresented as healthy. If unstressed individuals are expected to have longer measurements, this could artificially lower the unstressed mean. This was not found to be the case. In fact, more near significant differences were f ound using ANOVAs comparing groups
202 divided by the CBTS criteria. Because the CBTS scoring criterion is the most conservative scoring technique employed in this analysis, it was considered to be the standard in further analysis. A G -test performed in Chapt er 5 indicated that Harris lines and hypoplastic defects behaved as independent variables. Not only are Harris lines and enamel defects the product of different events during growth, individuals who are deemed stressed based on enamel defects are not neces sarily the same individuals who have visible Harris lines. It is for that reason that the two indicators of growth disruption need to be analyzed separately. For example, m easurements in the stressed group, as determined by enamel defects, were shorter t han in the unstressed group for eight of ten measurements, but for only five of these measurements were the differences determined to be significant (see Table 5 2) On the other hand, utilization of measurements grouped by Harris lines indicated that the bones of stressed individuals were longer in every measurement, though none of these differences were significant (see Table 5 5) It is these differences that cause me to question the reliability of one stress marker over another. The d ifferences in degree of reliability as stress indicators are likely due to the ways in which these stress markers form and their persistence in the adult skeleton Enamel defects are caused by a disruption in amelogenesis which, in the majority of cases, is due to ill ness or nutritional deficiency (Goodman and Rose 1990; Hillson and Bond, 1997). These markers form at the time of the insult and are indelible once formed. Harris lines also record stress, but a thick Harris line can be attributed more to a growth recov ery than to the growth arrest itself (Park 1964). A growth disruption must occur for a Harris line to form: however, growth disruption without adequate recovery will not leave as thick of a line as an individual who recovers fully. Bone remodeling may m ean that the most chronically stressed individuals are those without
203 observable Harris lines as adults. Martin et al. (1985) caution researchers that while children with poor nutrition are more likely to form Harris lines, there are fewer remnant scars among the nutritionally stressed. Blanco et al. (1974) conducted a study of Harris lines in Guatemalan children less than seven years of age. They found a trend for shorter height for age in children with Harris lines than those without. These differences were more pronounced in male children than among females. On the other hand, Goodman and Clark (1981) found very different results when analyzing Harris lines from the adult skeletal sample at Dickinson Mounds. Tibial length was significantly greater among adults with Harris lines than among those without, particularly in females. Goodman and Clark (1981) suggest that individuals genetically destined to be taller may have greater nutritional needs than shorter individuals and are therefore more likely to form Harris lines when such needs are not met. While taller individuals may be more likely to form Harris lines, there are other possible explanations for individuals with Harris lines being taller. Because thick Harris lines are linked to a positive recovery, it is possible the most severely stressed in any population present as unlined. It is also possible that taller individuals with Harris lines may have undergone the catchup growth process. Based on Baron et al.s (1994) experiments i n artificially suppressing growth plates, we know that catch up growth leads to an increase in growth rate (although full size was never attained in these studies). Recent research on catch up growth in weight has linked obesity with early metabolic insul ts which suggest the process of catch up may overshoot a normal growth trajectory in weight (Cameron and Demerath 2002; Grove et al. 2005; Hindmarsh 2004). While there is no evidence for the same catch up process in skeletal length, until researchers better understand the process of catchup growth, this possibility should not be ruled out.
204 In the investigation into the relationship of sexual dimorphism and stress, differences in dimorphism between stressed an d unstressed groups divided by the presenc e or absence of Harris lines were an order of magnitude smaller than those produced by enamel defects. The corresponding p values found that Harris line data was not meaningful. This comparison, in addition to the others listed above force s the conclusi on that 1) Harris lines and enamel defects are not recording the same stress events, 2) the relationship between Harris lines and stress is not well understood, and 3) enamel defects are a far more reliable marker of stress in the human skeleton than are H arris lines. Chapter 8 investigate d the utility of using vertebral neural canal (VNC) shape as a marker for growth disruption as suggested by Clark et al. ( 1985, 1986, 1988). ANOVAs suggest that there is a significant relationship between vertebral neur al canal dimensions and enamel defects (CBTS) in five out nine comparisons made in males. Of particular interest is the fact that differences in anteroposterior dimensions were significantly different in every case for males when the population was divided by enamel defects. In no case are vertebral neural canal sizes significantly related to enamel defects (CBTS) in females (see Table 8 1 and 82) Small anteroposterior vertebral neural canal dimensions cease growth earlier than mediolateral dimensions and have been associated with low birth weight, maternal smoking and a protein deficient diet ( Clark 1985; Jeffrey, 1988). Rewekant (2001) found smaller vertebral neural canal indices (anteroposterior diameter/ m ediolateral diameter) among a stressed archaeological population when compared to a population of higher socioeconomic status. Interestingly, Rewekant (2001) found that the significant differences were only for the male sample as is the case in this study. These findings raise the question of whether the selective advantage of maintaining a large vertebral canal is greater in females than males.
205 When vertebral neural canal sizes were compared using Harris lines as the fixed factor a different trend was observed. A mong females there signifi cant differences exist in the mediolateral diameter of the averaged thoracic and averaged lumbar vertebrae as well as in the vertebral canal index of L1 when the population was divided by the presence or absence of Harris lines (see Tables 8 3 and 8 4). I n the case of the vertebral canal index, the significant difference is due to larger mediolateral diameters among the unstressed segment of the population. Mediolateral dimensions of the vertebral neural canals continue to grow through childhood and are believed to be more likely to experience catch up growth than anteroposterior dimensions (Clark 1985, 1988; Jeffrey et al., 2003). Because Harris lines are more closely associated with growth recovery than growth disruption, the fact that there is a relati onship between Harris lines and mediolateral VNC dimensions may not be surprising. It is important to note that this response is only observable among females. The relationship between VNC dimensions and traditional markers of stress suggest there may be utility in using vertebral neural canals as a stress indicator; however, there are several caveats to that conclusion. First, the utility of using a vertebral canal index (anteroposterior diameter/mediolateral diameter) is presumably an attempt to get an understanding of the overall shape of the vertebral neural canal shape, but a small index may be due to a smaller anteroposterior diameter or a larger mediolateral diameter. Given that both measurements are significantly corre lated to overall body size, t his ratio will always need to be qualified. In this study, anteroposterior and mediolateral dimensions were far more informative than the vertebral neural canal index. While the relationship between vertebral canal dimensions and other indicators is an i mportant one, VNC shape did not prove useful for addressing the hypotheses of this study. Correlations comparing skeletal height and vertebral neural canal shape suggest that
206 height and VNC shape are not independent variables (see Table8 11) although VNC measurement s would not be useful for predicting stature Therefore, VNC dimensions cannot be used to infer stature differences between stressed and unstressed groups. Head circumference, a known indicator of stunted growth in children, was also compared to enamel defects and Harris lines. These investigations did not yield any significant relationships in the HamannTodd collection (where head circumference was determined from autopsy reports), or in the Terry collection (where head circumference was me asured). Regardless of the relationship with other stress markers, head circumference is significantly correlat ed with skeletal height in the Terry collection which means it cannot be used as a n independent stress marker in this study. The fact that head circumference and skeletal height are not significantly correlated in the Hamann Todd collection only serves to make me wary of the autopsy reports. In an adult skeletal sample, the available markers of growth disruption are limited, and understanding the se limitations is critical to the interpretation. No skeletal marker used in this study records the full picture of stress during growth. Harris lines and enamel hypoplasias form during acute insults to growth. T hey may represent a single negative episode in an otherwise healthy individual or t he y may be an acute mark in a chronic ally stressed individual ; it is not always possible to tell the difference skeletally. Numerous stress markers may certainly be suggestive of a chronically stressed indiv idual; on the other hand, it could be no more than a record of numerous acute events. Steckel (1995) suggests that stature provide s a record of chronic stress which can be observed by looking at the secular changes in a population T he nature of the grow th disruption (chronic vs. acute) could provide the explanation for why stress markers are not significantly related to stature in men.
207 Study Limitations: Collection Effect It is important to consider the collection effect that may be influencing the results of this study. The fact that the majority of these individuals were unclaimed by family suggests that they were from the lowest of socioeconomic classes. Therefore, the HamannTodd and Terry collections are not necessarily representative of the Af rican -American communities as a whole in Cleveland, OH and St. Louis, MO at the turn of the twentieth century (Hunt and Albanese 2005). Attempts to assess whether or not this sample is representative have proven inconclusive (Albanese 2003). The decis ion to combine materials from two osteological collections was made based on Isans (1990) conclusion that the two populations were osteometrically similar. To further ascertain the similarity of the two populations, a few simple variables were analyzed to determine whether or not the individuals I sampled from the two collections were similar: year of birth and skeletal height. When data for year of birth were compared, a significant difference was observed (p = 0.016). The mean year of birth in the H amann Todd collection was 1902.77 and the mean for the Terry collection was 1906.09. Because secular trends exhibit increasing height within these collections, skeletal height in the two collections were compared and no significant differences were observ ed (p = 0.138). In addition, there is no correlation between year of birth and skeletal height in the combined sample (p = 0.342). Since a secular trend cannot be observed in this sample and the mean difference between the collections is only 3.32 years, the difference in year of birth is therefore notable, but likely unimportant. Study Limitations: Age Effect In acquiring the sample population for this study I initially chose individuals between the ages of twenty and thirty-five years in order to anal yze the skeletal remains of individuals with fully fused long bones and who were not suffering from any age related changes that would
208 affect stature. Because the number of females in the Hamann Todd and Terry collections is much lower than the males, a f ew 19 year -old females with fully fused epiphyses were included to increase the final sample size. Age is not normally distributed within this sample, so nonparametric correlations (Kendalls tau b) were run comparing age to both skeletal height (SKH) and sitting height (STH) to determine whether or not an age effect was visible in the sample population. Age is not significantly related to SKH, but it is related to STH in males (see Table 9 1). A scatter plot of STH data against age suggests that some of the youngest individuals in the male sample did not attain full vertebral height before death ; however, the confidence interval for a linear regression line fit through the data included zero, signifying that the effect of age on STH is negligible (see Fi gures 9 1 and 92 ). Bass (1995) states that vertebral body heights almost at tain their full size by puberty; nevertheless, they do not fully fuse until 17 to 25 years of age. Including males under 25 in my initial analysis is likely the cause of the age e ffect observed in this sample. While collecting data for this study, individuals were excluded from analysis if the epiphyseal rings of the vertebrae were not present, but they were included if the epiphyseal rings were intact but still in the process of fusing. To see how this age effect influenced the results of this study, individuals under the age of twenty five were removed from the male sample. After removing these individuals, STH and B = 0.137, p = 0.079). ANOVAs were performed comparing SKH, STH, femur length and tibia length in stressed and unstressed groups (defined by the presence or absence of enamel defects (CBTS)) in this smaller sample (containing only 25 to 35 year old males). None of the se results were significant (see Table 9 2) which suggests that any age effect which exists in the original sample population did not significantly influence the
209 results of this study. This research therefore informs future methodology: in future studies where vertebral height is considered, conservative procedures should exclude males under the age of twenty -five years.
210 Table 9 1. Nonparametric correlations between age and height Skeletal Height (SKH) Sitting Height (STH) Correlation Significance (2 Correlation Significance (2 Females 0.092 0.195 0.133 0.060 Males 0.063 0.348 0.175 0.010 a a S ignificant at the 0.01 level (2 tailed) Table 9 2. ANOVA Stature estimates (age controlled) by dental enamel defects (CBTS) N Mean S.D. Levines F p Male Skeletal height (SKH) Unstressed 30 160.10 6.918 Stressed 53 159.74 6.620 0.530 0.055 0.816 Revised Fully Stature Unstressed 30 172.35 6.955 Stressed 53 171.99 6.666 0.527 0.054 0.816 Sitting height (STH) Unstressed 30 48.50 1.707 Stressed 53 48.07 2.0393 0.339 0.939 0.335 Max. femur length (mm) Unstressed 30 480.50 27.748 Stressed 53 477.86 25.167 0.387 0.196 0.659 Tibia length Unstressed 30 401.38 24.543 Stressed 53 403.50 25.495 0.829 0.136 0.714
211 Figure 9 1. Scatter plot of STH by age in females Figure 9 2. Scatter plot of STH by age in males R Linear = 0.039 39 41 43 45 47 49 51 53 55 18 23 28 33STHAge R Linear = 0.0549 39 41 43 45 47 49 51 53 55 18 23 28 33STHAge
212 CHAPTER 10 CONCLUSIONS The three alternative hypotheses tested in this study were based on previous research o n the relationship between growth and stress, and an interest in whether or not these relationships could be observed in a skeletal sample. Skeletal markers of growth disruption, enamel defe cts and radiopaque transverse lines were compared to osteometric data to determine whether or not the stresses associated with the skeletal marker had an effect on adult stature, sexual dimorphism and proportions. Previous r esearch on the process o f catch up growth suggests that the negative effects of growth disruption are often erased (e.g. Cameron 2002; Tanner 1981) ; h owever, anthropologists use trends in adult morphology to estimate the relative environmental stress on populations (Komlos, 1994; Steckel 1995, 1999; Steegman 1985, 1991) The premise behind such studies is that stressed populations are less likely to reach their genetic potential for height In addition, females are believed to be buffered biologically against stressors evident in that females tend to have greater fat and nutrient reserves (Stini 1969, 1973, 1975a, 1975b, 1982). In that females are buffered, male size is believed to very more in response to environmental circumstance, thereby decreasing the degree of sexual dimorphism found in stressed populations. Human proportions vary by population and some researchers have suggested they are under stronger genetic control than is overall body size (Mueller 1986; Stini 1975) However, numerous researchers have found that bett er environmental circumstances lead to longer relative lower limb length (Bogin and Rios, 2003; Bogin et al., 2002; Dangour, 2001; Floyd 2007, 2008; Fredriks et al. 2005; Frisancho et al. 2001; Li et al. 2007; Malina et al. 2004; Stinson 2000; Wadswo rth et al. 2002) and that t he relationship of proximal and distal limb elements is associated with environmental circumstances (Holliday and Ruff 2001; Meadows Jantz and
213 Jantz 1999; Jantz and Owsley 1984; Meadows and Jantz 1995; Smith and Buschang 2004). Furthermore, because nutritional stress and growth disturbances are greatest in the first two to three years of life, age of an insult may affect its influence (Beaton et al., 1990; Checkley et al., 1998; Martorell and Habicht, 1986; Martorell et al., 1995). An investigation into the relationship between stature and skeletal markers of stress found that females were significantly shorter than their unstressed counterparts when stress was determined by the presence or absence of enamel defects. Signif icant differences in stature were not found among males. When Harris lines were used to determine stress, no significant differences were found between stressed and unstressed groups. While significant differences in sexual dimorphism were not found betw een stressed and unstressed groups, it is notable that stress had a visible affect in the female skeletal sample, but not the male. Investigations into the relationship between proportions and stress yielded only one significant result. The crural index in males was significantly related to the presence or absence of enamel defects. Interestingly, crural indices were higher among the stressed group. The stressed group had longer tibia l lengths and shorter femoral lengths than their unstressed counterpa rts. Previous researchers found that the distal limb segment is more variable (Holliday and Ruff 2001; Smith and Buschang, 2004), and that the length of the tibia increases in secular studies where environmental conditions improve ( Meadows Jantz and Jant z 1999; Jantz and Owsley 1984; Meadows and Jantz 1995). All other proportional differences observed were subtle. Enamel defects and Harris lines were the pathological markers observed in the Terry and Hamann Todd collections with enough frequency to yield statistically significant results as markers for dividing a population into stressed and unstressed groups. Whil e other pathologies
214 were observed, they were immediately thrown out if they were believed to directly affect stature. Other pathologies could not be tied specifically to the growth period, so they were recorded but not considered in this analysis. Verteb ral neural canals were measured and compared to traditional stress markers to test their efficacy as independent markers of stress. A trend between small anteroposterior dimensions of the vertebral canal and enamel defects was evident in males, but ultima tely the fact that VNC dimensions were significantly correlated to stature ruled out the use of VNC dimensions as a stress marker of use in this study. A G -test comparing enamel defects and Harris lines concluded that these markers behaved as independent variables, and the results when dividing the population by these two markers were drastically different. The etiology of enamel defects is better understood than that of Harris lines and they are more widely accepted as an indicator of growth disruption ( Alfonso et al., 2005). Harris line data in this study can be used as an informative comparison between the two techniques, but the enamel defect data is more useful for comparing the actual relationship between skeletal morphology and stress. Skeletal ma rkers of stress represent acute insults and are an incomplete record of growth disruption. This fact needs to be considered when interpreting how the results of this study can be applied to research on human growth. In addition, it qualifies how effective ly knowledge regarding how growth disruption affects morphology can be applied to a bioarchaeological sample.
215 LIST OF REFERENCES AAPA (American Association of Physical Anthropologists). 1996. AAPA statement on biological aspects of race. Am J Phys Anth ropol 101:569570. Agarwal B. 1990. Social security and family: Coping with seasonality and calamity in rural India. In: Ahmad E, Drze J, Hills J, and Sen AK, editors. Social s ecurity in d eveloping c ountries. Oxford, U.K.: Oxford University Press. p 1712 44. Agarwal SC, Dumitriu M, Tomlinson GA, Grynpas MD. 2004. Medieval trabecular bone architecture: the influence of age, sex, lifestyle. Am J Phys Anthropol 124:3344. Albanese J. 2003. Identified skeletal reference collections a nd the study of human vari ation. Doctoral thesis. McMaster University Hamilton, Ontario Alexander RD, Hoogland JL, Howard RD, Noonan KM, Sherman PW. 1979. Sexual dimorphisms and breeding systems in pinnipeds, ungulates, primates, and humans. In: Chagnon NA, and Irons W, editors. Evolutionary b iology and h uman s ocial b ehavior. North Scituate, Mass.: Duxbury Press. p 402 435. Alfonso MP, Thompson JL, Standen VG. 2005. Reevaluating Harris lines a comparison between Harris lines and enamel hypoplasia. Collegium Antropol 29:393408. Allen LH. 1994. Nutritional influences on linear growth: a gener al review. Eur J Clin Nutr 48:S4589. Allen LH. 1995. Malnutrition and human function: a comparison of conclusions from the INCAP and nutrition CRSP studies. J Nutr 125:1119S 1126S. Armelagos GJ, Van Gerven DP. 1980. Sexual dimorphism and human evolution: an overview. J Hum Evol 9:437446. Arnold F, Zhaoxiang L. 1986. Sex preference, fertility and family planning in China. Popul Dev Rev 12:221246. Atkinson PJ. 1967. Variation in trabecular structure with age. Calcified Tissue Int 1(1):2432. Baigari R. 1986. Food crisis, nutrition, and female children in rural Bangladesh. Popul Dev Rev 12(2):307315. Bailey SM, Xu J, Feng JH, Hu X, Zhang C, Qui S. 2007. Tradeoffs between oxygen and energy in tibial growth at high altitude. Am J Hum Biol 19:662 668. Baker PT. 1988. Human adaptability. In: Harrison GA, Tanner JM, Pilbeam DR, and Baker PT, editors. Human b iology. 3rd ed. New York: Oxford University Press. p 439 -547.
216 Baron J, Klein KO, Colli MJ, Yanovski JA, Novosad JA, Bacher JD, Cutler GB. 1994. Catchup gr owth after glucocorticoid excess: a mechanism intrinsic to the growth plate. Endocrinology 135(4):13671371. Bass WM. 1995. Human o steology: a l aboratory and f ield m anual. Columbia, MO: Special Publication No. 2 of the Missouri Archaeological Society. Bear d AS, Blaser MJ. 2002. The ecology of height the effect of microbial transmission on human height. Perspect Biol Med 14(1):6173. Beaton G, Kelly A, Kevany J, Martorell R, Mason J. 1990. Appropriate uses of anthropometric indices in children. Geneva: ACC /SCN. Blakey ML, Armelagos GJ. 1985. Deciduous enamel defects in the prehistoric Americans from Dickson mounds: prenatal and postnatal stress. Am J Phys Anthropol 66:371380. Blanco RA, Acheson RM, Canosa C, Salomon JB. 1974. Height, weight, and lines of a rrested growth in young Guatemalan children. Am J Phys Anthropol 40:3948. Boersma B, Wit JM. 1997. Catchup Growth. Endocr Rev 18(5):646661. Bogin B. 1999. Patterns of Human Growth. Cambridge: Cambridge University Press. Bogin B. 2003. The human pattern of growth and development in paleontological perspective. In: Thompson JL, and Krovitz GE, editors. Patterns of growth and development in the genus Homo Cambridge: Cambridge University Press. p 15 44. Bogin B, Rios L. 2003. Rapid morphological change in l iving humans: implications for modern human origins. Comp Biochem Phys A 136:7184. Bogin B, Smith P, Orden AB, Varela Silva MI, Loucky J. 2002. Rapid change in height and body proportions of Maya American children. Am J Hum Biol 14:753761. Bohonak AJ. 2004. RMA: software for reduced major axis regression Version 1.17. San Diego, CA: San Diego State University. Available at: http://www.bio.sdsu.edu/pub/andy/RMA.html Borgognini Tarli SM, Repetto E. 199 7. Sex differences in human populations: change through time. In: Morbeck ME, Galloway A, and Zihlman AL, editors. Evolving female. Princeton, N.J: Princeton University Press. Boyd DC, Boyd CC. 1989. A comparison on Tennessee archaic and mississippian maxi mum femoral lengths and midshaft diameters: subsistence change and postcranial variability. Southeastern Archaeol 8(2):107116. Brace CL. 1973. Sexual dimorphism in human evolution. In: Brace CL, and Metress J, editors. Man in evolutionary perspective. Ne w York: John Wiley & Sons, Inc. p 238254.
217 Brace CL, Ryan AS. 1980. Sexual dimorphism and human tooth size differences. J Hum Evol 9:417435. Bramblett CA. 1994. Patterns of primate behavior. Prospect Heights, IL: Waveland Press. Brandt I, Sticker EJ, Lent ze MJ. 2003. Catch up growth of head circumference of very low birth weight, small for gestational age preter m infants and mental development in adulthood. J Pediatr 142(5):463468. Brauer GW. 1982. Size sexual dimorphism and secular trend: indicators of s ubclinal malnutrition? In: Hall RL, editor. Sexual dimorphism in Homo Sapiens : a question of size. New York: Praeger Publishers. p 245262. Bribiescas RG. 2001. Reproductive ecology and life history of the human male. Am J Phys Anthropol 44:148176. Bribie scas RG. 2006. Men: e volutionary and l ife h istory. Cambridge, MA: Harvard University Press. Buikstra JE, Ubelaker DH, editors. 1994. Standards for data collection from human skeletal remains. Fayetteville: Arkansas Archaeological Survey. Byers S. 1991. Tec hnical note: calculation of age at formation of radiopaque transverse lines. Am J Phys Anthropol 85:339343. Calder PC, Jackson AA. 2000. Undernutrition, infection and immune function. Nutr Res Rev 13:329. Cameron N. 2002. Human growth curve, canalization, and catch up growth. In: Cameron N, editor. Human g rowth and d evelopment. New York: Academic Press. p 1 20. Cameron N, Demerath EW. 2002. Critical periods in human growth and their relationship to diseases of aging. Yearb Phys Anthropol 45:159184. Cartmill M. 1998. The status of the race concept in physical anthropology. Am Anthropol 100:651660. Chagnon NA. 1979. Is reproductive success equal in egalitarian societies? In: Chagnon NA, and Irons W, editors. Evolutionary b iolog y and human social behavior: an anthropological perspective. North Scituate, MA: Duxbury Press. p 374401. Checkley W, Epstein LD, Gilman RH, Black RE, Cabrera L, Sterling CR. 1998. Effects of Cryptosporidium parvum infection in Peruvian children: growth faltering and subsequent catch up growth. Am J Epidemiol 148(5):497506. Choe YH, Kim SK, Hong YC. 2000. Helicobacter pylori infection with iron deficiency anaemia and subnormal growth at puberty. Arch Dis Child 82:136 140.
218 Clark GA. 1988. New method for assessing changes in growt h and sexual dimorphism in paleoepidemiology. Am J Phys Anthropol 77:105116. Clark GA, Hall NR, Armelagos GJ, Borkan GA, Panjabi MM, Wetzal FT. 1986. Poor growth prior to early childhood: decreased health and life -span in the adult. Am J Phys Anthropol 70:145160. Clark GA, Panjabi MM, Wetzal FT. 1985. Can infant malnutrition cause adult vertebral stenosis? Spine 10(2):165170. Clutton Brock T. 1991. The evolution of parental care. Princeton: Princeton University Press. Cohen MN, Armelagos GJ, editors. 1984. Paleopathology at the o rigins of a griculture. New York: Academic Press. Cohen MN, O'Connor K, Danforth ME, Jacobi KP, Armstrong C. 1997. Archaeology and osteology of the Tipu site. In: Whittington SL, and Reed DM, editors. Bones of the Maya. Washington: Smithsonian Institution Press. p 7886. Cole TM. 1997. RMA: software for reduced major axis regression. Baltimore, MD: The John Hopkins University, Department of Cell Biology and Anatomy. Colombo M, de Andraaca I, Lpez I. 1988. Mental development and stunting. In: Waterlow JC, editor. Linear growth retardation in less developed countries. New Y ork: Raven Press. p 201214. Condon K, Rose JC. 1992a. Intertooth and intratooth variability in the occurence of developmental enamel defects. J Paleopathol 2:6177. Condon K, Rose JC. 1992b. Intertooth and introtooth variability in the occurence of developmental enamel defects. In: Goodman AH, and Capasso LL, editors. Recent Contributions to the study of Enamel Developmental Defects Journal of Paleopathology, Monogrphic Prublications, No 2. Chieti, Italy: Edigrafital Teramo. p 6177. Cronk L. 1989. Low socioeconomic status and female -biased parental investment: the Mukogodo example. Am Anthropol 91:414429. D'Ippolito G, Schiller PC, Ricordi C, Roos BA, Howard GA. 1999. Age -related osteogenic potential of mesenchymal stromal cells from human vertebral bone marrow. J Bone Miner Res 14(7):11151122. Damon A. 1969. Secular trend in height and weight within old American families at Harvard, 18701965. Am J Phys Ant hropol 29:4550. Dangour AD. 2001. Growth of upper and lower -body segments in Patamona and Wapishana Amerindian children (cross -sectional data). Ann Hum Biol 28(6):649663.
219 Das Gupta M. 1987. Selective discrimination against female children in rural Punja b, India. Pop Dev Rev 13:77100. Davey -Smith G, Greenwood R, Gunnell D, Sweetnam P, Yarnell J, Elwood P. 2001. Leg length, insulin resistance, and coronary heart disease risk: The Caerphilly Study. J Epidemiol Commun H 55:867872. Delgado H, Sibirn R, Del gado ME. 1987. Evidence for catchup growth in adolescence. Lancet ii:1021. DeVore I, Washburn SL. 1963. Baboon ecology and human evolution. In: Howel FC, and Bouliere F, editors. African ecology and human evolution. New York: Viking Fund Press. p 335367. Dyson T. 1991a. On the demography of South Asian famines part I. Popul Stud 45(1):525. Dyson T. 1991b. On the demography on South Asian famines part II. Popul Stud 45(2):279297. Ellison PT. 2003. Energetics and reproductive effort. Am J Hum Biol 15:342351. Engerman S. 1994. Comment. In: Komlos J, editor. Stature, living standards, and economic development. Chicago: University of Chicago Press. p 205 209. Eveleth PB. 1975. Differences between ethnic groups in sex dimorphism of adult height. Ann Hum Biol 2(1):3539. Eveleth PB. 1986. Popualtion differences in growth: environmental and genetic factors. In: Falkner F, and Tanner JM, editors. Human growth: a comprehensive treatise. New York: Plenum Press. p 221 239. Eveleth PB, Tanner JM. 1990. Worldwide vari ation in human growth. Cambridge: Cambridge University Press. Farnum C, Lee A, O'Hara K, Wilsman N. 2003. Effect of short term fasting on bone elongation rates: an analysis of catch up growth in young male rats. Pediatr Res 53:33 -41. Fleagle JG, Sammonds KW, Hegsted DM. 1975. Physical growth of cebus monkeys, Cebus albifrons during protein or calorie deficiency. Am J Clin Nutr 28:246 253. Fleagle JG, Samonds KW. 1975. Physical growth of cebus monkeys ( Cebus albifrons ) during the first years of life. Growth 39:3552. Floyd B. 2007. Focused life history data and linear enamel hypoplasia to help explain intergenerational variation in relative leg length within Taiwanese families. Am J Hum Biol 19:358375. Floyd B. 2008. Intergenerational gains in relative knee height as compared to gains in relative leg length within Taiwanese families. Am J Hum Biol 20:462464.
220 Fogel RW. 1986. Physical growth as a measure of the economic well -being of populations: the eighteenth and nineteenth centuries. In: Falkner F, and Tanner JM, editors. Human growth: a comprehensive treatise. New York: Plenum Press. p 263281. Frayer DW. 1980. Sexual dimorphism and cultural evolution in the late Pleistocene and Holocene of Europe. J Hum Evol 9:399415. Frayer DW. 1981. Body size, weapon use, and natural selection in the European Upper Paleolithic and Mesolithic. Am Anthropol 83:5773. Frayer DW, Wolpoff MH. 1985. Sexual dimorphism. Ann Rev Anthropol 14:429473. Fredriks AM, van Buuren S, van Heel WJM, Dijkman Neerincx RHM, Verloove -Vanhorick SP, Wit JM. 2005. Nationwide age reference for sitting height, leg length, and sitting height/height ratio, and their diagnostic value for disproportionate growth disorders. Arch Dis Child 90:807812. Frisancho AR, Guilding N, Tanner S. 2001. Growth of leg is reflected in socioeconomic differences. Acta Med Auxol 33(1):4750. Frisancho AR, Guire K, Babler W, Borkan G, Way A. 1980. Nutritional influence on childhood development and genetic control of adolescent growth of Quechuas and Mestizos from the Peruvian lowlands. Am J Phys Anthropol 52:367375. Garn SM, Rohmann CG. 1966. Interaction of nutrition and genetics in the timing of growth and development. Pediatr Clin N Am 13:353379. Garn SM, Silverman MD, Hertzog KP, Rohmann CG. 1968. Lines and bands of increased density: their implication to growth and development. Med Radiogr Photogr 44(3):58 89. Gaulin SJC, Boster JS. 1992. Human marriage systems and sexual dimorphism in stature. Am J Phys Anthropol 89:467475. Golden MHN. 1994. I s complete catch up growth possible for stunted malnourished children? Eur J Clin Nutr 48 (suppl. 1):S58 S71. Goodman AH, Allen LH, Hernandez GP, Amador A, Arriola LV, Chavez A, Pelto GH. 1987. Prevalence and age at development of enamel hypoplasis in Mexi can Children. Am J Phys Anthropol 72:7 19. Goodman AH, Armelagos GJ. 1985. Factors affecting the distribution of enamel hypoplasias within the human permanent dentition. Am J Phys Anthropol 68:479493. Goodman AH, Armelagos GJ, Rose JC. 1980. Enamel hypopl asias as indicators of stress in three prehistoric populations from Illinois. Hum Biol 52:515 528.
221 Goodman AH, Clark GA. 1981. Harris lines as indicators of stress in prehistoric Illinois populations. In: Martin DL, and Bumstead MP, editors. Biocultural ad aptation: comprehensive approaches to skeletal analysis. Amherst, MA: University of Massachusetts. p 35 46. Goodman AH, Martin DL. 2002. Reconstructing health profiles from skeletal remains. In: Steckel RH, and Rose JC, editors. The backbone of history: he alth and nutrition in the Western hemisphere. New York: Cambridge University Press. p 11 60. Goodman AH, Rose JC. 1990. Assesment of systemic physiological perturbations from dental enamel hypoplasias and associated histological structures. Am J Phys Anthr opol 33:59110. Gopalan C. 1988. Stunting : significance and implications for public health policy. In: Waterlow JC, editor. Linear growth retardation in less developed countries. New York: Raven Press. p 265284. Gray JP, Wolfe LD. 1980. Height and sexual dimorphism of stature among human societies. Am J Phys Anthropol 53:441456. Grolleau Raoux JL, Crubezy E, Rouge D, Brugne JF, Saunders SR. 1997. Harris lines: a study of age associated bias in counting and interpretation. Am J Phys Anthropol 103:209217. Grove KL, Grayson BE, Glavas MM, Xiao XQ, Smith MS. 2005. Development of metabolic systems. Physiol Behav 86:646660. Greulich WW. 1976. Some secular changes in the growth of American-born and native Japanese children. Am J Phys Anthropol 45: 553568. Gun nell D. 2002. Commentary: can adult anthropometry be used as a 'biomarker' for prenatal and childhood exposures? Int J Epidemiol 31:390394. Gunnell DJ, Davey -Smith G, Frankel S, Nanchahal K, Braddon FEM, Pemberton J, Peters TJ. 1998a. Childhood leg length and adult mortality: follow up of the Carnegie (Boyd Orr) survey of diet and health in pre -war Britain. J Epidemiol Commun H 52(3):142152. Gunnell DJ, Davey -Smith G, Holly JMP, Frankel S. 1998b. Leg length and risk of cancer in the Boyd Orr cohort. Br Med J 317:13501351. Gustafsson A, Lindenfors P. 2004. Human size evolution: no evolutionary allometric relationship between male and female stature. J Hum Evol 47:253266. Hales CN, Ozanne SE. 2003. The dangerous road of catchup growth. J Physiol 547(1): 5 10. Hamill PVV, Johnston FE, Lemeshow S. 1973. Body weight, stature, and sitting height: white and negro youths 1217 years. Vital and Health Statistics, Series 11. Rockville, MD: Health Resources Administration.
222 Han TS, Hooper JP, Morrison CE, Lean MEJ. 1997. Skeletal proportions and metabolic disorders in adults. Eur J Clin Nutr 51:804809. Harpending H, Pennington R. 1991. Age structure and sex-biased mortality among Herero pastoralists. Hum Biol 63:329353. Hauspie RC, Vercauteren M, Susanne C. 1996. Secular changes in growth. Horm Res 45(suppl 2):8 17. Hermanussen M. 1997. Plasticity of adolescent growth in boys. Am J Hum Biol 9:469480. Hiernaux J, Rudan P, Brambati A. 1975. Climate and the weight/height relationship in subSaharan Africa. Ann Hum Bi ol 2:3 12. Hillson S. 1996. Dental Anthropology. Cambridge: Cambridge University Press. Hillson S, Bond S. 1997. Relation of enamel hypolplasia to the pattern of tooth crown growth: a discussion. Am J Phys Anthropol 104:89103. Himes JH. 1979. Secular chan ges in body proportions and composition. Monogr Soc Res Child Dev 44(3/4):2858. Hindmarsh PC. 2004. Commentary: Catchup growth in humans a comment on poverty, birthweight, and infant weight gain in Hertfordshire. Int J Epidemiol 33:12341235. Hinton RJ Carlson DS. 1979. Temporal change in human temporomandibular joint size and shape. Am J Phys Anthropol 50:325334. Holden C, Mace R. 1999. Sexual dimorphism in stature and women's work: a phylogenetic cross -cultural analysis. American Journal of Physical Anthropology 110:1: 27 (20 pages). Holliday TW. 1997. Body proportions in late Pleistocene Europe and modern human origins. J Hum Evol 32:423447. Holliday TW, and Falsetti AB. 1995. Lower limb length of European early modern humans in relation to mobility and climate. J Hum Evol 29:141153. Holliday TW, Ruff CB. 2001. Relative variation in human proximal and distal limb segment lengths. Am J Phys Anthropol 116:2633. Hrdy SB. 1990. Sex bias in nature and in history: a late 1980s reexamination of the "biol ogical origins" argument. Yearb Phys Anthropol 33:2537. Hummert JR, Van Gerven DR. 1985. Observations on the formation and persistence of radiopaque transverse lines. Am J Phys Anthropol 66:297306. Hunt DR, Albanese J. 2005. History and demographic composition of the Robert J. Terry anatomical collection. Am J Phys Anthropol 127:406417.
2 23 Hunt EE, Hatch JW. 1981. The estimation of age at death and ages of formation of transverse lines from measurements of human long bones. Am J Phys Anthropol 54:461469. H uss-Ashmore R, Goodman AH, Armelagos GJ. 1982. Nutritional Inference from Paleopathology. Adv Archaeol Method Theor 5:395474. Hutchinson DL, Larsen CS. 1988. Determination of stress episode duration from linear enamel hypoplasias: a case study from St. Ca therine's Island, GA. Hum Biol 60(1):93110. Isan MY. 1990. A comparison of techniques on the determination of race, sex, and stature from the Terry and Hamann Todd collections. In: Gill GW, and Rhine S, editors. Skeletal attribution of race. Albequerque, MN: Maxwell Museum of Anthropology. Jantz RL, Hunt DR, Meadows L. 1994. Maximum length of the tibia: how did Trotter measure it? Am J Phys Anthropol 93:525528. Jantz RL, Meadows Jantz L. 2000. Secular changes in craniofacial morphology. Am J Hum Biol 12: 327338. Jantz RL, Owsley DW. 1984. Temporal changes in limb proportionality among skeletal samples of Arikara Indians. Ann Hum Biol 11(2):157163. Jeffrey JE, Campb ell DM, Golden MHN, Smith FW, Porter RW. 2003. Antenatal factors in the development of the lumbar vertebral canal. Spine 28(13):14181423. Johnston FE, MacVean R. 1995. Growth faltering and catchup growth in relation to environmental change in children of a disadvantaged community from Guatemala city. Am J Hum Biol 7(6):731740. Johnston FE, Wainer H, Thissen D, MacVean R. 1976. Hereditary and environmental determinants of growth in height in a longitudinal sample of children and youth of Guatemalan and European ancestry. Am J Phys Anthropol 44:469476. Jorde L, Watkins W, Bamshad M, Di xon M, Ricker C, Seielstad MT, Batzer M. 2000. The distribution of human genetic diversity: a comparison of mtiochondrial, autosomal and Y chromosome data. Am J Hum Genet 66(3):979988. Katzmarzyk PT, Leonard WR. 1998. Climatic influences on human body size and p roportions: ecological adaptations and secular trends. Am J Phys Anthropol 106:483503. Keller MC, Nesse RM, Hofferth S. 2001. The Trivers Willard hypothesis of parental investment: No effect in the contemporary United States. Evol Hum Behav 22:343 360. Ke nnedy KAR, Deraniyagala SU, Roertgen JC, Chiment J, Disotell T. 1987. Upper Pleistocene fossil hominids form Sri Lanka. Am J Phys Anthropol 72:441461. Kidane A. 1989. Demographic consequences of the 19841985 Ethiopian famine. Demography 26(3):515522.
224 Ki ng T, Hillson S, Humphrey LT. 2002. A detailed study of enamel hypoplasia in a post medieval adolescent of known age and sex. Arch Oral Biol 47:2939. Komlos J, editor. 1994. Stature, living standards, and economic development. Chicago: The University of C hicago Press. Kuh DL, Power C, Rodgers B. 1991. Secular trends in social class and sex differences in adult height. Int J Epidemiol 20(4):10011009. Kurki HK, Ginter JK, Stock JT, Pfeiffer S. 2008. Adult proportionality in small -bodied foragers: a test of ecogeographic expectations. Am J Phys Anthropol 136(1):2838. Larsen CS. 1997. Bioarchaeology: Interpreting Human Behavior from the Human Skeleton. Cambridge: Cambridge University Press. Larsen CS. 2002. Bioarchaeology: the l ives and l ifestyles of p ast p eo ple. J Archaeol Res 10(2):119166. Lawlor DA, Taylor M, Davey Smith G, Gunnell DJ, Ebrahim S. 2004. Associations of components of adult height with coronary heart disease in postmenopausal women: the British women's heart and health study. Heart 90:745749. Lazenby RA. 2002. Population variation in second metacarpal sexual size dimorphism. Am J Phys Anthropol 118:378384. Lederberg J. 2000. Infectious history. Science 288(5464):287293. Leigh SR. 1996. Evolution of human growth spurts. Am J Phys Anthropol 1 01:455474. Leutenegger W, Cheverud JM. 1985. Sexual dimorphism in primates: the effects of size. In: Jungers WL, editor. Size and s caling in p rimate b iology. New York: Plenum Press. p 33 50. Li L, Dangour AD, Power C. 2007. Early life influences on adult leg and trunk length in the 1958 British birth cohort. Am J Hum Biol 19:836843. Li L, Manor O, Power C. 2004. Early environment and childto adult growth trajectories in the 1958 British birth cohort. Am J Clin Nutr 80(1):185192. Lieberman LS. 1982. Norm al and abnormal sexual dimorphic patterns of growth and development. In: Hall RL, editor. Sexual dimorphism in Homo sapiens : a question of size. New York: Praeger Publishers. p 263316. Littleton J. 2005. Invisible impacts but long-term consequences: hypop lasia and contact in central Australia. Am J Phys Anthropol 126:295304.
225 Liu Y, Xu F, Shen H, Liu Y, Zhao, LJ, Long J, Zhang Y, Xiao P, Xiong D, Dvornyk V, Li J Conway T, Davies KM, Recker RR, Deng H 2004. Genetic dissection of human stature in a large s ample of multiplex pedigrees. Ann Hum Genet 68:472488. Liu YX, Jalil F, Karlberg J. 1998. Growth stunting in early life in relation to the onset of the childhood component of growth. J Pediatr Endocr Met 11(2):247260. L ovejoy CO, Kern KF, Simpson SW, Mei ndl RS. 1989. A new method for estimation of skeletal dimorphism in fossil samples with an application to Australopithecus afarensis In: Giacobini G, editor. Hominidae. Milano: Jaka Book. p 103108. Luo ZC, Karlberg J. 2000. Critical growth phases for adu lt shortness. Am J Epidemiol 152(2):125131. Luo ZC, Low LCK, Karlberg J. 2001. Critical growth phases for adult shortness in Hong Kong Chinese. J Pediatr Endocr Met 14(6):757765. Macintyre K. 2002. Famine and the female mortality advantage. In: Dyson T, and Cormac G, editors. Famine demography: p erspectives from the past and present. Oxford, U.K.: Oxford University Press. p 240 259. Malina RM. 1979. Secular changes in size and maturity: causes and effects. Monogr Soc Res Child Dev 44(3/4):59102. Malina RM. 1987. Nutrition and growth. In: Johnston FE, editor. Nutritional anthropology. New York: Alan R. Liss, Inc. p 173196. Malina RM, Pea Reyes ME, Tan SK, Buschang PH, Little BB, Kozial S. 2004. Secular change in height, sitting height and leg length in rural Oaxaca, southern Mexico: 1968 2000. Ann Hum Biol 31(6):615633. Maresh MM. 1955. Linear growth of long bones of extremities from infancy through adolescence. Am J Dis Child 89:725742. Mariam MW. 1986. Rural vulnerability to famine in Ethiopia: 19581977. London: Intermediate Technology. Mrquez L, del ngel A. 1997. Height among prehispanic Maya of the Yucatn peninsula: a reconsideration. In: Whittington SL, and Reed DM, editors. Bones of the Maya: studies of ancient skeletons. Washington: Smithsoni an Institution Press. p 51 61. Martin DL, Goodman AH, Armelagos GJ. 1985. Skeletal pathologies as indicators of quality and quantity of diet. In: Gilbert RI, and Meikle JH, editors. The analysis of prehistoric diets. New York: Academic Press. p 227 279. Martin EA, Ritman EL, Turner RT. 2003. Time course of epiphyseal growth plate fusion in rat tibiae. Bone 32(3):261267.
226 Martin RB, Burr DB, Sharkey NA. 1998. Skeletal tissue mechanics. New York: Springer. Martin SA, GuatelliSteinberg D, Sciulli PW, Walker PL. 2008. Brief communication: comparison of methods for estimating chronological age at linear enamel formation on anterior dentition. Am J Phys Anthropol 135(3):362365. Martorell R. 1989. Body size, adaptation and function. Hum Organ 48:1520. Martore ll R, Habicht J -P. 1986. Growth in early childhood in developing countries. In: Falkner F, and Tanner JM, editors. Human growth: a comprehensive treatise. New York: Plenum Press. p 241 262. Martorell R, Kettel Khan L, Schroeder DG. 1994. Reversibility of s tunting: epidemiological findings in children from developing countries. Eur J Clin Nutr 48 (sup.1):S45S57. Martorell R, Schroeder DG, Rivera JA, Kaplowitz HJ. 1995. Patterns of linear growth in rural Guatemalan adolescents and children. J Nutr 125:1060S 1067S. May RL, Goodman AH, Meindl RS. 1993. Response of bone and enamel formation to nutritional supplementation and morbidity among malnourished Guatemalan children. Am J Phys Anthropol 92:3751. Mays SA. 1995. The relationship between Harris lines and ot her aspects of skeletal development in adults and juveniles. J Archaeol Sci 22:511520. McHenry HM. 1991. Sexual dimorphism in Australopithecus afarensis J Hum Evol 20:2132. Meadows Jantz L, Jantz RL. 1999. Secular changes in long bone length and proport ion in the United States, 1800 1970. Am J Hum Biol 110:5767. Meadows L, Jantz RL. 1995. Allometric secular change in the long bones from the 1800s to the present. J Forensic Sci 40(5):762767. Meiklejohn C, Schentag C, Venema A, Key P. 1984. Socioeconomic change and patterns of pathology and variation in the Mesolithic and Neolithic of w estern Europe: some suggestions. In: Cohen MN, and Armelagos GJ, editors. Paleopathology at the origins of agriculture. London: Academic Press. p 75100. Mensforth RP, Lati merm B.M. 1989. Hamann Todd collection aging studies: osteoporosis fracture syndrome. Am J Phys Anthropol 80:461479. Mueller WH. 1986. The genetics of size and shape in children and adults. In: Falkner F, and Tanner JM, editors. Human growth: a comprehens ive treaty (vol 3). New York: Plenum Press. p 145 168. Murdock GP. 1967. Ethnographic Atlas. Pittsburgh: University of Pittsburgh Press.
227 Nabarro D, Howard P, Cassels C, Pant M, Wijga A, Padfield N. 1988. The importance of infections and environmental fact ors as possible determinant of growth retardation in children. In: Waterlow JC, editor. Linear g rowth r etardation in l ess d eveloped c ountries. New York: Raven Press. p 165183. Neumayer E, Plmper T. 2007. The gendered nature of natural disasters: the impa ct of catastrophic events on the gender gap in life expectancy, 19812002. Ann Assoc Am Geogr 97(3):551566. Nickens PR. 1976. Stature reduction as an adaptive response to food production in Mesoamerica. J Archaeol Sci 3:31 41. Nowak O, Piontek J. 2002. Does the occurrence of Harris lines affect the morphology of human long bones. Homo 52(3):254276. Nyati LH, Norris SA, Cameron N, Pettifor JM. 2006. Effect of ethnicity and sex on the growth of the axial and appendicular skeleton of children living in a dev eloping country. Am J Phys Anthropol 130:135141. Ortner DJ. 2003. Identification of pathological conditions in human skeletal remains. New York: Academic Press. Owens IPF. 2002. Sex differences in mortality rates. Science 297:20082009. Park EA. 1964. The imprinting of nutritional disturbances on the growing bone. Pediatrics 33:815862. Pfeiffer S, Sealy JC. 2006. Body size among Holocene foragers of the Cape ecozone, Southern Africa. Am J Phys Anthropol 129:111. Porter RW, Drinkall JN, Porter DE, Thorp L 1987. The vertebral canal: II. H ealth and academic status, a clinical study. Spine 12(9):907911. Porter RW, Pavitt D. 1987. The vertebral canal: I. N utrition and development, an archaeological study. Spine 12(9):901906. Prentice A, Bates CJ. 1994. Adequacy of dietary mineral supply for human bone growth and mineralisation. Eur J Clin Nutr 48 (suppl. 1):S161 S177. Rathbun TA. 1987. Health and disease at a South Carolina plantation: 1840-1870. Am J Phys Anthropol 74:239253. Raxter MH, Auerbach BM, Ru ff CB. 2006. Revision of the Fully technique for estimating stature. Am J Phys Anthropol 130(3):374384. Reid DJ, Dean MC. 2000. Brief communication: the timing of linear hypoplasia on human anterior teeth. Am J Phys Anthropol 113:135139.
228 Reid DJ, Dean MC 2006. Variation in modern human enamel formation times. J Hum Evol 50:329336. Reiner AP, Ziv E, LInd DL, Nievergelt CM, Schork NJ, Cummings SR, Phong A, Burchard EG, Harris TB, Psaty BM Kwok P 2005. Population structure, admixture, and agingrelated p henotypes in African American adults: the cardiovascular health study. Am J Hum Gene t 76(3):463 477. Rensch B. 1959. Evolution above the species level. London: Methuen & Co. Rewekant A. 2001. Do environmental disturbances of an individual's growth and deve lopment influence the later bone involution processes? A study of two mediaeval populations. Int J Osteoarchaeol 11:433443. Ribeiro Andrade MC, Penedo MCT, Ward T, Silva VF, Bertolini LR, Roberts JA, Leite JPG, Cabello PH. 2004. Determination of genetic s tatus in a closed colony of rhesus monkeys (Macaca mulatta). Primates 45(3):183186. Richmond BG, Jungers WL. 1995. Size variation and sexual dimorphism in Australopithecus afarensis and living hominoids. J Hum Evol 29:229245. Ritzman TB, Baker BJ, Schwartz GT. 2008. A fine line: a comparison of methods for estimating ages of linear enamel hypoplasia formation. Am J Phys Anthropol 135:348-361. Rivers J. 1982. Women and children last: an essay on sex discrimination in disasters. Disasters 6:256267. Roberts DF. 1953. Body weight, race and climate. Am J Phys Anthropol 11:533558. Roberts DF. 1973. Climate and human variability. An AddisonWesley module in anthropology, vol 34. Reading, MA: AddisonWesley. Robling AG, Stout SD. 2000. Histomorphology of human c ortical bone: applications to age estimation. In: Katzenberg MA, and Saunders SR, editors. Biological a nthropology of the h uman s keleton. New York: Wiley Liss. p 187 205. Robson EB. 1978. The genetics of birth weight. In: Falkner F, and Tanner JM, editors. Human g rowth. New York: Plenum Press. p 285297. Roche AF. 1979. Secular trends in stature, weight, and maturation. Mon Soc Res in Child Dev 44(3/4):327. Rose JC, Condon K, Goodman AH. 1985. Diet and dentition: developmental disturbances. In: Gilbert RI, and Mielke J, editors. The analysis of prehistoric diets. New York: Academic Press. p 281 305. Ruff CB. 1987. Sexual dimorphism in human lower limb bone structure: relationship to subsistence strategy and sexual division of labor. J Hum Evol 16:391416.
229 Ruff CB. 1991. Climate and body shape in hominid evolution. J Hum Evol 21:81105. Ruff CB. 1994. Morphological adaptation to climate in modern and fossil hominids. Yearb Phys Anthropol 37:65107. Ruff CB, Larsen CS, Hayes WC. 1984. Structural changes in the femur with the transition to agriculture on the Georgia coast. Am J Phys Anthropol 64:125136. Santos RV, Coimbra CEA. 1999. Hardships of contact: enamel hypoplasias in Tup -Mond Amerindians from the Brazilian Amazon. Am J Phys Anthropol 109:111127. Saunders SR, Hoppa RD. 1993. Growth deficit in survivors and non-survivors: biological mortality bias in subadult skeletal samples. Yearb Phys Anthropol 36:127151. Schultz AH. 1969. The life of primates. New York: Universe Books. Scrimshaw NS, Taylor CE, Gordon JE. 1968. Interaction of Nutrition and Infection World Health Organ Monogr Ser 57. Seckler D. 1980. "Malnutrition": an intellectual odyssey. Western J Agr Econ 5(2):219227. Seckler D. 1982. "Small but healthy": a basic hypothesis in the theory, me asurement and policy of malnutrition. In: Sukhatme PV, editor. Newer concepts in nutrition and their implications for policy. Pune, India: Maharshtra Association for the Cultivation of Science Research Institute. p 127137. Serrat MA, Farnum CE, Williams R M, Lovejoy CO. 2009. Environment influences bone elongation during a critical period of postnatal growth. Am J Phys Anthropol S48:235. Silventoinen K. 2003. Determinants of variaiton in adult body height. J Biosoc Sci 35:263285. Simon JL. 2000. Resampling stats: the new statistics. Arlington, VA: Resampling Stats, Inc. Simpson SW. 2001. Patterns of growth disruption in La Florida. In: Larsen CS, editor. Bioarchaeology of Spanish Florida. Gainesville: University of Florida Press. p 146 180. Smith RJ. 2009. Use and misuse of the reduced major axis for line -fitting. Am J Phys Anthropol 140:476486. Smith SL, Buschang PH. 2004. Variation in l ongitudinal d iaphyseal l ong b one g rowth in c hildren t hree to t en y ears of a ge. Am J Hum Biol 16:648 657. Sokal RR, Rohlf FJ. 1995. Biometry. New York: W.H. Freeman and Company. Steckel RH. 1987. Growth depression and recovery: the remarkable case of American slaves. Ann Hum Biol 14(2):111132. Steckel RH. 1995. Stature and standard of living. J Econ Lit 33:19031940.
230 Steckel RH, Larsen CS, Sciulli PW, Walker PL. 2006. The g lobal h istory of h ealth p roject: d ata c ollection c odebook. Columbus, OH: The Ohio State University. Steckel RH, Rose JC. 2002. The backbone of history: health and nutrition in the western hemisphere. Cambri dge: Cambridge University Press. Steegmann AT. 1985. 18th century British military stature: growth cessation, selective recruiting, secular trends, nutrition at birth, cold and occupation. Hum Biol 57(1):7795. Steegmann AT. 1991. Stature in an early mid1 9th century poorhouse population: Highland Park, Rochester, New York. Am J Phys Anthropol 85:261268. Steegmann AT, Haseley PA. 1988. Stature variation in the British American Colonies: French and Indian War records, 1755 1763. Am J Phys Anthropol 75:413421. Steele DG, Bramblett CA. 1988. The anatomy and biology of the human skeleton. College Station, TX: Texas A&M University Press. Stini WA. 1969. Nutritional stress and growth: sex differences in adaptive response. Am J Phys Anthropol 31:417426. Stini W A. 1971. Evolutionary implications of changing nutritional patterns in human populations. Am Anthropol 73(5):10191030. Stini WA. 1973. Reduced sexual dimorphism in upper arm circumference associated with a protein deficit diet in a South American populati on. Am J Phys Anthropol 36:341352. Stini WA. 1975a. Adaptive strategies of human populations under nutritional stress. In: Watts ES, Johnston FE, and Lasker GW, editors. Biosocial i nterrelations in p opulation a daptation. Chicago: Aldine. Stini WA. 1975b. Ecology and human adaptation. Dubuque, IA: Wm C. Brown Company Publishers. Stini WA. 1982. Sexual dimorphism and nutrient reserves. In: Hall RL, editor. Sexual dimorphism in Homo sapiens : a question of size. New York: Praeger Publishers. p 391420. Stini W A. 1985. Growth rates and sexual dimorphism in evolutionary perspective. In: Gilbert RI, and Meikle JH, editors. The analysis of prehistoric diets. New York: Academic Press. p 191226. Stinson S. 1985. Sex differences in environmental sensitivity during gr owth and development. Yearb Phys Anthropol 28:123147. Stinson S. 1992. Nutritional adaptation. Ann Rev Anthropol 21:143170.
231 Stinson S. 2000. Growth variation: biological and cultural factors. In: Stinson S, Bogin B, Huss Ashmore R, and O'Rourke D, editors. Human biology: an evolutionary and biocultural perspective. New York: Wiley-Liss, Inc. p 425463. Stoch MB, Smythe PM. 1976. Fifteenyear developmental study on effects of severe undernutrition during infancy on subsequent physical growth and inte llectual functioning. Arch Dis Child 51:327336. Suckling GW. 1989. Developmental defects of enamel historical and present -day perspectives of the pathogenesis. Adv Dent Res 3(2):8794. Tanner JM. 1962. Growth at adolescence. Oxford: Blackwell Scientific Publications. Tanner JM. 1963. Regulation of growth in size in mammals. Nature 199(4896):845850. Tanner JM. 1981. Catchup Growth in Man. Br Med Bull 37(3):233238. Tanner JM. 1986. The use and abuse of growth standards. In: Falkner F, and Tanner JM, edi tors. Human Growth: a comprehensive treatise. New York: Plenum Press. p 95 109. Tanner JM. 1988. Human growth and constitution. In: Harrison GA, Tanner JM, Pilbeam DR, and Baker PT, editors. Human biology: an introduction to human evolution, variation, gro wth, and adaptability. New York: Oxford University Press. p 339435. Tanner JM. 1992. Growth as a measure of the nutritional and hygenic stature of a population. Hormone Research 38(suppl 1):106115. Theintz GE, Howald H, Weiss U, Sizonenko PC. 1993. Evide nce for reduction of growth potential in adolescent female gymnasts. J Pediatr 122:306 313. Torun B, Viteri FE. 1994. Influence of exercise on linear growth. Eur J Clin Nutr 48 (suppl. 1):S186S190. Trivers RL. 1972. Parental investment and sexual selectio n. In: Campbell B, editor. Sexual selection and the descent of man: 18711971. Chicago: Aldine Publishing Company. p 136179. Trivers RL, and Willard DE. 1973. Natural selection of parental ability to vary the sex ratio of offspring. Science 179(4068):909 2. Tsutakawa RK, Hewett JE. 1977. Quick test for comparing two populations with bivariate data. Biometrics 33:215219. Turan S, Bereket A, Omar A, Berber M, Ozen A, Bekiroglu N. 2005. Upper segment/lower segment ratio and armspan height difference in healt hy Turkish children. Acta Pdiatr 94:407413.
232 Van Wieringen JC. 1986. Secular growth changes. In: Falkner F, and Tanner JM, editors. Human growth: a comprehensive treatise. New York: Plenum Press. p 307331. Vick AE. 2005. Sexual dimorphism and the transit ion to agriculture: a meta -analysis. MA thesis. University of Florida, Gainesville. Vrba ES. 1996. Climate, heterochrony, and human evolution. J Anthropol Res 52(1):128. Wadsworth MEJ, Hardy RJ, Paul AA, Marshall SF, Cole TJ. 2002. Leg and trunk length at 43 years in relation to childhood health, diet and family circumstances; evidence from the 1946 national birth cohort. Int J Epidemiol 31:383390. Waldron I. 1983. Sex differences in human mortality: the role of genetic factors. Soc Sci Med 17(6):321333. Warren M, Holliday T, Cole T. 2002. Ecogeographical patterning in the human fetus. Am J Phys Anthropol S34:161162. Waterlow JC, editor. 1988a. Linear growth retardation in less developed countries. New York: Raven Press. Waterlow JC. 1988b. Observations of the natural history of stunting. In: Waterlow JC, editor. Linear growth retardation in less developed countries. New York: Raven Press. p 117. Waxenbaum EB, Hunt DR, Falsetti AB. 2006. To measure or not to measure: an analysis of maximum length of the tibia Proc Am Ac Forensic Sci: 305306. White TD. 2000. Human o steology. New York: Academic Press. White TD, and Folkens PA. 2005. The Human Bone Manual. New York: Elsevier. WHO. 1986. Use and interpretation of anthropometric indicators of nutritional sta tus. B World Health Organ 64(6):929941. Widdowson EM, McCance RA. 1963. The effect of finite periods of undernutrition at different ages on the composition and subsequent development of the rat. P R oy S oc L ond B B io 158(972):329342. Williams JPG, Hughes PCR. 1975. Catchup growth in rats undernourished for different periods during the suckling period. Growth 39:179193. Wilson PN, Osbourn DF. 1960. Compensatory growth after undernutrition in mammals and birds. Biol Rev Biol Proc Camb Phil Soc 35(3):324363. Wolaski N, Kasprzak E. 1976. Stature as a measure of environmental change. Curr Anthropol 17(4):548552. Wolfe LD, Gray JP. 1982. Subsistence practices and human sexual dimorphism of stature. J Hum Evol 11:575580.
233 Wood JW, Milner GR, Harpending HC, We iss KM. 1992. The o steological p aradox. Cur r Anthropol 33:343370. Wright DR. 1990. African Americans in the Colonial e ra. Franklin JH, and Eisenstadt AS, editors. Arlington Heights, IL: Harlan Davidson, Inc. Wright LE. 1997. Intertooth patterns of hypopla sia expression: implications for childhood health in the classic Maya collapse. Am J Phys Anthropol 102:233247. Zerran J. 2007. Gaining and Growing. Available at: h ttp://depts.washington.edu/growing/Assess/Anthro.htm#Head%20Circumference
234 BIOGRAPHICAL SKETCH Anna Elizabeth Vick was born in 1975, in Chapel Hill, North Carolina. She graduated from the University of North Carolina at Chapel Hill in 1998 with a Bachelor of Arts degree in anthropology. In 2005, Anna earned a Master of Arts degree in anthropology from the University of Florida.