Citation
Cranial size in relation to body mass and skeletal robusticity in modern humans

Material Information

Title:
Cranial size in relation to body mass and skeletal robusticity in modern humans
Creator:
Stubblefield, Phoebe Regina
Publication Date:
Language:
English
Physical Description:
xi, 84 leaves : ill. ; 29 cm.

Subjects

Subjects / Keywords:
Autopsies ( jstor )
Body size ( jstor )
Body weight ( jstor )
Clavicle ( jstor )
Correlation coefficients ( jstor )
Humans ( jstor )
Mathematical variables ( jstor )
Skull ( jstor )
Skull base ( jstor )
Waves ( jstor )
Anthropology thesis, Ph. D ( lcsh )
Dissertations, Academic -- Anthropology -- UF ( lcsh )
Genre:
bibliography ( marcgt )
theses ( marcgt )
non-fiction ( marcgt )

Notes

Thesis:
Thesis (Ph. D.)--University of Florida, 2002.
Bibliography:
Includes bibliographical references.
General Note:
Printout.
General Note:
Vita.
Statement of Responsibility:
by Phoebe Regina Stubblefield.

Record Information

Source Institution:
University of Florida
Holding Location:
University of Florida
Rights Management:
The University of Florida George A. Smathers Libraries respect the intellectual property rights of others and do not claim any copyright interest in this item. This item may be protected by copyright but is made available here under a claim of fair use (17 U.S.C. §107) for non-profit research and educational purposes. Users of this work have responsibility for determining copyright status prior to reusing, publishing or reproducing this item for purposes other than what is allowed by fair use or other copyright exemptions. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder. The Smathers Libraries would like to learn more about this item and invite individuals or organizations to contact the RDS coordinator (ufdissertations@uflib.ufl.edu) with any additional information they can provide.
Resource Identifier:
029665988 ( ALEPH )
51945780 ( OCLC )

Downloads

This item has the following downloads:


Full Text









CRANIAL SIZE IN RELATION TO BODY MASS AND SKELETAL ROBUSTICITY IN MODERN HUMANS











By

PHOEBE REGINA STUBBLEFIELD


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


2002



























Copyright 2002

by

Phoebe Regina Stubblefield














ACKNOWLEDGEMENTS

I first want to thank God for getting me through this doctoral process. I also want to thank the late Dr. William R. Maples and Dr. Susan Anton for their mentorship and instruction which led me to and through the University of Florida. I also thank Dr. Sue Boinski for her leadership and encouragement. My thanks go to Dr. Anthony Falsetti, Dr. Thomas Hollinger, and Dr. William Hamilton for their help during this process. My special thanks go to Margaret and Herb Gilliland for their gracious and generous support and concern. Many thanks go to the American Academy of Forensic Sciences' Lucas Grant and the National Academies Ford Foundation Fellowship, and the doctors and technicians of the medical examiners' offices in Florida and Michigan, without which and whom this work would not have been possible. Last of all I want to thank my parents for their support, Pamela for transcribing data, and my twin, Peggy, for being there when I needed her.















TABLE OF CONTENTS


ACKNOW LEDGEM ENTS............................................................................................... iii

LIST OF TABLES...................................................................................vi

LIST OF FIGURES................................................................................ ix

ABSTRACT..........................................................................................x

INTRODUCTION ............................................................................................................... 6

Body W eight Estimation in Hominids.............................................................................2
Research with Ectocranial M easurements........................................................................4
Research with Cranial Thickness M easurements .............................................................4

LITERATURE REVIEW ....................................................................................................9

Pensler and M cCarthy (1985)..........................................................................................9
Hartwig-Scherer and M artin (1992).................................................................................9
Nawrocki (1992).............................................................................................................10
Aiello and W ood (1994)................................................................................................12
Gauld (1996) ..................................................................................................................13
Discussion ...................................................................................................................... 14
The Project......................................................................................................................15

M ATERIALS AND M ETHODS......................................................................................18

Sample Selection.............................................................................................................18
Data Collection ...............................................................................................................19
Somatic Variables ..................................................................................................19
Cranial M easurements ...........................................................................................21
Postcranial M easurements .............................................................................................25
Estimation of Lean Body W eight...................................................................................26
Data Analysis..................................................................................................................27

RESULTS .......................................................................................................................... 31

The Combined Sample: Descriptive Statistics...............................................................31
Combined Sample: Relationships Between Variables ....................................................33
Combined Sample: Principal Component Analysis........................................................39
Combined Sample: Regression Analysis.......................................................................43
The Autopsy and Skeletal Samples: Somatic Descriptive Statistics................................44
Quality of Skinfolds......................................................................................................45
Autopsy and Skeletal Samples: Cranial Descriptives.....................................................50
Autopsy and Skeletal Samples: Relationships Between Variables..................................53









Autopsy and Skeletal Samples: Principal Component Analyses....................................58
A utopsy Sam ple: Regression A nalysis ..........................................................................63
Sum m ary ........................................................................................................................64

D ISCU SSIO N ...................................................................................................................67

REFEREN CES..................................................................................................................73

A PPEN D IX ........................................................................................................................80

BIO G RA PH ICA L SK ETCH ............................................................................................84














LIST OF TABLES


Tabk a

1 Aiello and Wood's results relative to body weight using a mixed hominoid
sam ple............................................................................. .... ....14
2 Cranial measurements....................................................................23

3 Clavicle measurements...................................................................25

4 Results of the one-sample t-test for sectioned and radiographic clavicle
measurements.............................................................................26

5 Descriptive statistics for somatic variables in the combined sample................31

6 Descriptive statistics for cranial variables of the combined sample..................32

7 Comparison of shared ectocranial measurements in Howell's "mixed race"
sample and the combined sample.......................................................32

8 Descriptive statistics for clavicle variables in the combined sample................. 33

9 Combined sample correlation coefficients between body weight and the
somatic variables (including clavicle)...................................................34

10 Combined sample correlation coefficients between body weight and the
cranial variables..................................................................... ...34

11 Combined sample correlation coefficients between the clavicle cortical
thickness and cranial thickness variables..............................................35

12 Combined sample correlation coefficients between the ectocranial and cranial
thickness variables.......................................................................36

13 Combined sample correlation coefficients between ectocranial and average
thickness variables....................................................................... 38

14 Combined sample correlation coefficients between the cranial index and
cranial thickness variables............................................................... 38

15 Combined sample correlation coefficients between the cranial index shape
categories and the cranial thickness variables..........................................39

16 Combined sample correlation coefficients between cranial index categories
and average vault thickness..............................................................39









17 Combined sample principal components using all variables (except age)..........41 18 Combined sample principal components of the cranial variables and body
w eight.....................................................................................42
19 Combined sample regression coefficients for body weight versus the cranial
variables....................................................................................44

20 Descriptive statitics for the somatic variables in the autopsy and skeletal
sam ples....................................................................................45
21 Autopsy sample fatness categories based on Frisancho (1990) table iv39 for
males using summed triceps and subscapular skinfold values......................46

22 Autopsy sample weight classifications based on weight (kg) and stature (cm)
and skinfolds..............................................................................47

23 Contingency table of weight classifications based on weight for stature and
combined triceps and subscapular skinfold thicknesses.............................48

24 Regression equations for estimating body density (d) from the logarithm of
triceps and subscapular skinfold measurements.....................................48

25 Autopsy sample descriptive statistics for fat weight estimates and the new
variable lean weight.......................................................................49

26 Weight classifications based on unadjusted weight for height versus lean
weight for height..........................................................................49
27 Autopsy sample descriptive statistics for the cranial variables.......................50

28 Skeletal sample descriptive statistics of the cranial variables........................51

29 Autopsy and skeletal samples descriptive statistics for clavicle variables...........52

30 Autopsy and skeletal samples two-tailed t-test for the cranial variables............52

31 Autopsy and skeletal samples correlation coefficients between body weight
and the somatic and clavicle measurements............................................53

32 Autopsy and skeletal samples correlation coefficients between body weight
(kg) and the ectocranial variables.......................................................54

33 Manipulated skeletal sample correlation coefficients for select
variables....................................................................................55

34 Skeletal sample correlation coefficients for body weight versus cranial
variables excluding low weight subject................................................57

35 Autopsy and skeletal samples correlation coefficients between body weight
(kg) and the cranial thickness variables................................................58
36 Skeletal sample correlation coefficients for additional ectocranial variables
versus vault thickness averages.........................................................58

vii










37 Autopsy sample principal components for the cranial variables..................... 60

38 Skeletal sample principal components using all variables (except age)..... .......61

39 Skeletal sample principal components using the cranial variables.................63

40 Autopsy sample regression coefficients for body weight versus cranial
variables....................................................................................64

41 Autopsy sample regression coefficients for lean weight versus cranial
variables....................................................................................65















LIST OF FIGURES



1 Ectocranial Measurements................................................................24

2 Locations of Cranial Thickness Measurements.........................................24

3 Skeletal Sample Weight for Height Class Distribution................................56














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

CRANIAL SIZE IN RELATION TO BODY MASS AND SKELETAL ROBUSTICITY IN MODERN HUMANS

By

Phoebe Regina Stubblefield

December 2002

Chair. Susan Ant6n
Cochair. Sue Boinski
Major Department: Anthropology

Body weight is critical to understanding primate biology. However, taphonomic

processes hamper acquiring this statistic from human skeletons or fossil hominids. Fossil hominid species are defined from crania that frequently lack associated postcrania. Thus body size information is frequently restricted to that which can be extracted from the skull. To address this issue, I designed a static allometric study using modem humans as the reference sample.
I tested two hypotheses regarding the relationship between cranial size and body size. The first hypothesis stated that measurements of external cranial dimensions (ectocranial) covary with body weight, and the second that cranial vault thickness measurements covary with body weight in relation to systemic skeletal robusticity. To test these hypotheses I sampled 147 adults of known body weight (100 from the Terry collection at the National Museum of Natural History and 47 from recent autopsies) for








sixteen ectocranial and seven cranial vault thickness measurements. To assess postcranial robusticity I also measured cortical thickness at four shaft locations on the clavicle. I analyzed the relationships in the data using correlation and regression analyses, and principal component analysis.

In contrast to the published literature that uses multi-species hominoid and

anthropoid data sets, cranial measurements were not well correlated with body weight. Vault thickness measurements were particularly poorly correlated. There was also no pattern of association between gross vault thickness and clavicular cortical bone thickness. The lack of relationship between body size and many cranial measurements in modern humans indicates that these measurements are largely unsuitable for body weight prediction.














INTRODUCTION


This study addresses the feasibility of predicting body weight in large-brained

hominids from cranial measurements. Body weight is critical to hominid biology because it is the best indicator of body size (Jungers 1984), it is essential to understanding dietary, locomotor, and reproductive behaviors in all primates, and because it is central to understanding brain development, or encephalization, in the hominids. An understanding of body size is critical to models of hominid expansion outside of Africa, as body size is directly related to home ranges, time spent foraging, ability to cope with seasonal food availability, predation risk, and food choice (McNab 2002). While there is abundant literature on postcranial techniques of body size estimation in hominids, sources using cranial measurements are limited. Therefore I proposed and tested several hypotheses involving the relationships between body size (as indicated by body weight) and cranial measurements.

My primary hypothesis is that if body components, such as the limbs or torso, are scaled in proportion to the whole body, then the cranium should follow. I test this proposal in two related hypotheses:
* HI: body weight in modern humans covaries with cranial measurements

* H2: estimates of lean body weight will have stronger covariance relationships

with cranial measurements than gross body weight will have with cranial

measurements

I employ both ectocranial measurements and measurements of vault thickness.

Cranial dimensions (ectocranial measurements) have been associated with size differences due to sex and to ancestry (Giles and Elliot 1962, 1963), so a link to body size might be








expected. However, the function of tri-layered cranial bone is still subject to discussion (Kennedy 1991), although Moss and Young (1961) attribute separate functions to each layer. Physiological functions of hemopoiesis and fat storage take place in the diploic bone of the vault; likewise cortical thickness in long bones has biomechanical determinants (Ruff 2000), and the same could be true for the cranial vault (Currey 1984, Demes 1985). I will test the relationship between vault thickness and body size with hypotheses 1 and 2. Hypothesis 2 derives from an assumption that cranial measurements and body size are physiologically linked (Lieberman 1996). Increasing obesity in Western adults (Flegal et al. 1998) makes an estimate of lean body weight critical in comparisons to the non-weightbearing cranium.
My secondary hypothesis is that thickness of the non-pathological neurocranium, or vault, in the hominids is primarily the product of general skeletal development. Therefore hominids with systemically thicker cortical bone in the postcrania will also have greater vault thickness, whereas hominids with slighter postcranial cortices will have lesser vault thickness. Because cranial contents directly influence the magnitude and directions of vault development, brain size might also affect vault thickness (e.g., Moss and Young 1961; Richards and Ant6n, 1991). I propose three additional hypotheses regarding vault thickness:
* H3: vault thickness is a function of general skeletal robusticity as indicated by

cortical thickness of long bones
* H4: an index of vault length and height covaries with vault thickness

* H5: weight of the brain significantly covaries with vault thickness


Body Weight Estimation in Hominids
Body weight is correlated with a variety of biological features including dietary behavior (Kay 1975, McHenry 1984), locomotion (Jungers 1984), population density (Martin 1981), and brain size (McHenry 1976, Holloway 1980, Rightmire 1986). Body mass, according to Lindstedt and Calder (1981:2), is the "primary determinant of ecological








opportunities, as well as of the physiological and morphological requirements of an animal." In order to understand the biology of fossil primates, body weight must be reconstructed from the recovered skeletal material.

Most techniques for estimating body weight in fossil primates use dental or postcranial elements. The dental studies (e.g., Conroy 1987, Shea 1983, Pirie 1978, Gingerich et al. 1982, Gingerich 1977, Dagosto and Terranova 1992) are mainly derived from nonhuman primate samples. The postcranial and some of the dental analyses tend to use proxies for body weight in order to cope with the lack of recorded live weights in the nonhuman parts of the sample (e.g., Wood and Stack 1980-cranial length, Steudel 1980-partial skeletal weight, Ruff 2000-bi-iliac breadth, Pirie 1978-cranial length, Corruccini and Henderson 1978-three skull measurements). Yet Hartwig-Scherer and Martin (1992) demonstrated that a strong relationship between a proxy variable and body weight does not guarantee a strong relationship to body weight for a third variable that has a strong relationship to the proxy.

McHenry (1976, 1991) and Hartwig-Scherer (1994) relied on known body weights for their samples (mixed primate in the former, human in the latter) for developing predictive equations from long bone measurements. In another study using known body weights, Baker and Newman (1957) related dry skeletal weight to living weight. Cranial predictors of body weight are needed because of the frequent lack of association between cranial and postcranial remains in the hominid fossil record. Reviews and atlases of the human fossil record (e.g., Larsen et al. 1998) shows a clear bias towards cranial specimens, and likewise recent definition of the new possible hominid species Sahelanthropus tchadensis (Brunet et al. 2002) was based on cranial elements alone. Essentially, fossil hominids are cranial species. Weidenreich (1946:40) comments on this from an earlier vantage in palaeoanthropology: "How have the manifold characters which determine a type to be weighted in the system of classification? Do qualities of the teeth count more than those of the bones of the jaws, or those of the skull more than those of the








limbs?" Weidenreich gave preference to cranial characters, believing that evolution leading to the modern human brain was the most salient theme of hominid evolution. Without cranial predictors giving access to body weight, the chance and rare finds of associated postcrania hamper further investigation of the biology of extinct hominids.

Research with Ectocranial Measurements

Ectocranial measurements have a long history in physical anthropology as

descriptive tools. Classic texts by Martin and Sailer (1957) and Howells (1973) define a variety of cranial measurements for the human skull, and the use of these measurements is well illustrated in descriptions of fossil hominids (e.g. Jacob 1972, Wolpoff 1980). The use of these measurements as predictive tools has had a strong role in forensic anthropology in estimating sex (e.g., Giles and Elliot 1962, Loth and Henneberg 1996) and ancestry (Giles and Elliot 1963) from the skull. Only a few publications link ectocranial measurements to body weight (Aiello and Wood 1994, Hartwig-Scherer and Martin 1992). Aiello and Wood demonstrated high correlations (r > .90) between some ectocranial measurements and body weight in a mixed primate sample. Hartwig-Scherer and Martin (1992) examined only two cranial variables and found no significant correlations in their static human sample.


Research with Cranial Thickness Measurements

The history of research surrounding cranial thickness in hominids has taken three primary approaches. The explanatory approach attempts to identify or propose the source or cause of normal adult cranial thickness. The descriptive approach is centered on a particular population or geographic area, and is concerned with describing the variation of cranial thickness due to age, sex or other biological variables. Finally, the pathological approach is concerned with the etiology of abnormal cranial thickness. This last aspect is well illustrated in the work of Ortner and Putschar (1981) and will only be discussed here when the pathological intersects with the other approaches.







5
The explanatory approach examines the source of cranial thickness and may present models explaining cranial vault structure. Proposed mechanisms for increased cranial vault thickness encompass behavioral, biomechanical (Washburn 1947, Demes 1985, Nawrocki 1992), to physiological (Ivanhoe 1979, Lieberman 1996). Behavioral explanations stress dangerous subsistence practices or interpersonal violence, both involving blunt trauma to the head (Weidenreich 1943). Brown (1994) proposed that a midline area of increased cranial thickness in the frontal and parietal bones of Australian aborigines was related to the practice of settling disputes by trading blows to the head with a wooden club or staff. As this behavior predominated in young adults he thought such behavior would easily select against thin vaults. He supported this idea with findings of a suite of injuries associated with the practice in late-Pleistocene aboriginal specimens dating to about 11,000 years BP. Others have suggested that the increased ectocranial thickness of H. erectus may be related to general skeletal hypertrophy in this species (sensu Kennedy, 1985 and Hublin, 1986).

While the behavioral hypotheses still want more support, such as a survey of cranial thickness in cultures that end disputes with head-bashing, the biomechanical and physiological models involve more discrete hypothesis testing. The emphasis on testing was stressed initially by Moss and Young (1961:281) who suggested a functional approach to craniology to provide "significance" to the rapidly accumulating sets of cranial measurements and morphological descriptions. They divided the human cranium into neural and facial functional components that were each subject to further functional subdivision. Demes (1985) proposed that the vault of the modern human be studied in terms of a thin-walled globe, biomechanically known as a "shell," rather than with beam theory. The shell morphology of the skull reduced transmission of bending stresses to the thinner walled portions of the anterior base by transmitting strain into the walls of the vault. Based on stress tests of models of the basicranium, she suggested that vault curvature influences wall thickness such that the more curved the vault bones, the less thick they are. However she also stated that more research was needed, since her study did not examine the







6
entire vault. Proposing that longer crania and larger faces would cause more bending stress in a cranium, Nawrocki (1992) tested Demes' theory on a sample of Pleistocene hominids and modern humans. He found cranial thickness to be associated with flatter vaults and larger faces in his sample, although the recent human subsample was not well correlated for vault thickness and wall curvature. In his sample of modern humans with known body weights (drawn from the Terry collection), Nawrocki did not find significant correlation between body weight and vault thickness.

The physiological hypotheses have dealt with external and internal forces on cranial thickness. Moss and Young (1961) described the inner and outer layers and diploe of the vault as independent functional units. The inner table follows the contours of the neural mass, including accommodating alterations caused by neural atrophy. Crests and ridges from muscular attachments only affect the outer layer (Washburn 1947). In addition to serving as a site for hemopoiesis the diploe also serves to lighten the total cranial bone mass, a conclusion supported by Currey (1984) as a biomechanical solution to minimize bone mass without compromising bone strength. Studies of artificial cranial deformation also suggest that internal and external tables may act independently (e.g., Ant6n, 1989) Other studies have related cranial thickness to geomagnetic field intensity (Ivanhoe 1979). Ivanhoe supported his proposal with a correlation coefficient between 0.6 and 0.7 for his Pleistocene hominid sample, as well as noting the robusticity of a few other mammalian genera in times of varying geomagnetic field intensity. Ivanhoe's study would be better supported by a more thorough review of Pleistocene fauna and consideration of modern human variation in cranial thickness.

Association between cranial thickness and activity level was suggested by Lieberman (1996), based on studies of human physiological response to exercise that showed an increase in growth hormone secretion with increased exercise either of long duration (Lassarre et al. 1974) or of high intensity (Felsing et al. 1992). His study of a small sample of pigs and armadillos demonstrated marked increase in both cranial and postcranial cortical








thickness in exercised versus unexercised siblings. In a survey of cranial thickness in Holocene hominids Lieberman showed that pre-industrial farmers had greater cranial thickness than their postindustrial counterparts, thus contradicting his null hypothesis that increased activity did not affect cranial thickness.

Descriptive studies are the most common product of cranial thickness research.

Todd (1924) produced the earliest rigorous descriptive study of modern humans, which has since been followed by several others (Twiesselmann 1941, Adeloye et al. 1975, Ohtsuki 1977, Kaito 1980, Pensler and McCarthy 1985, Brown 1987a, Ishida and Dodo 1990, Gauld 1992, Ross et al. 1998). While these studies all utilized direct bone measurement, several radiographic studies have also contributed to the literature of human cranial thickness (Roche 1953, Young 1957, Getz 1960, Israel 1973, 1977, Tallgren 1974, Smith et al. 1985, Brown et al. 1979, Brown 1987b). Part of this literature documented variation in cranial thickness in various human populations, such as Todd, Adeloye et al., and Pensler and McCarthy for American blacks and whites, Twiesselmann and Getz for various European populations, Kaito for Japanese, Brown for Northern Chinese and Australian aborigines, and Gauld for Native Americans. Many of these described a conflicting pattern of thickness change with increased adult age. Israel and Tallgren both reported from longitudinal radiographic studies, but the former found a pattern of cranial measurement increase with age, while the latter did not Smith and colleagues did not find significant difference in cranial thickness in a sample of Near Eastern crania, although they note that the older crania were thicker. In another diachronic study of modern and Neolithic Japanese, Ishida and Dodo found significantly greater cranial thickness in their Neolithic males.

Another focus of the descriptive study is the variation in cranial thickness in PlioPleistocene hominids (Hinton et al. 1938, Weidenreich 1943, Tobias 1967Jacob 1972, Wolpoff 1980, Singer and Wymer 1982, Trinkaus 1983, Wood 1984, Clarke 1985, Sonakia 1985, Kennedy 1985, 1991, Brown 1994, Ant6n and Franzen 1997). Particularly,








specimens of Homo erectus, Homo ergaster, and Homo neandertalensis appeared to have greater non-pathological cranial and postcranial bone thickness than seen in modern humans. Pathological cranial thickening has been clearly demonstrated in one H. erectus and two Neandertal specimens (Ant6n 1997), in the form of hyperostosis calvariae internma (HCI). Cranial thickness in Homo erectus had been considered a derived trait for the species (Andrews 1984, Stringer 1984), but Kennedy (1991) showed that the similar state in H. neandertalensis prevented cranial thickness from being an autapomorphy in H. erectus. Brown (1994) demonstrated that the vault thickness of Asian H. erectus falls within the range of variation seen in Australian aborigines. He concluded that the emphasis on cranial thickness in Pleistocene hominids, rather than illustrating a peculiar character of these large brained bipeds, actually displays a lack of familiarity with variation in modern human vault thickness. Lieberman (1996) supported Brown's work while reviewing additional Pleistocene hominids, and showed that the range of cranial thickness was not outside the variation in modem humans. However, while absolute cranial thickness may not separate H. erectus and H. sapiens, the relative contributions of cortical versus diploic bone to thickness may differ between the taxa (Hublin, 1986; Ant6n 1997). Increased thickness in H. erectus is generally achieved via cortical expansion, whereas diploic expansion accounts for most increases in thickness in modern humans (Ant6n 1997).

While the cranial thickness literature is not tidy, some themes are discernible.

Populations, at least in the United States, vary in cranial thickness, but only in location of greatest thickness, not in overall cranial thickness. Males and females do not differ in average thickness. Cranial thickness might increase with age, but even the longitudinal studies disagree. Modern human variation in cranial thickness encompasses that seen in fossil hominids particularly Homo erectus. Exercise has strong support for influencing cranial thickness (and skeletal robusticity in general), while behavioral, biomechanical, and







9
geomagnetic effects want more testing. Most relevant to this study, there is no clear consensus for a significant relationship between cranial thickness and body weight.













LITERATURE REVIEW


A limited body of literature exists on the relationships between cranial
measurements and body weight in modem humans, representing the work of five authors. Two authors focus on ectocranial measurements of the skull (Hartwig-Scherer and Martin 1992, Aiello and Wood 1994). Two others focus on vault thickness measurements (Pensler and McCarthy 1985 and Gauld 1996), while the fifth (Nawrocki 1992) uses a combination of measurements.

Pensler and McCarthy (1985)
The earliest work is Pensler and McCarthy's examination of the vault as a bone
donation site. They measured cranial thickness at two frontal and two parietal locations in 200 male and female autopsy subjects with recorded weights, race (American Blacks and Whites), and age at time of death. Correlation of the log transformed thickness and weight data yielded r-values ranging from 0.2 to 0.6. The authors did not support cranial thickness as a predictor of body weight since they found that a 75-pound difference in weight produced only a 1.5mm difference in vault thickness.

Hartwig-Scherer and Martin (1992)
These authors reported two cranial variables in a study of the effect of using proxies for body weight when predicting body size in hominids. The authors used a sample (n = 295) of subadult and adult hominoids (H. sapiens, Gorilla gorilla, Pan paniscus, Pan troglodytes, and Pongo pygmaeus) with recorded body weights. They subdivided the sample to provide results on static and ontogenetic allometry. Measurements of the skull, humerus, radius, femur and tibia were collected and major axis line fitting used on the log10 transformed data.
With one exception the human sample produced such low correlation coefficients 10







11
with body weight that no results were reported. The adult human sample (n = 19) yielded no reportable results for either cranial variable (skull length and basicranial length). The ontogenetic sample produced an r-value of 0.93 for basicranial length with body weight. Nonhuman hominoids had r-values ranging from 0.90 to 0.98 for basicranial length and body weight in their ontogenetic samples. Generally skull and basicranial length performed well for the nonhuman apes. Adult samples had r-values ranging from 0.90 to 0.96 for basicranial length, while ontogenetic values ranged from 0.90 to 0.98. The lowest correlation coefficient was 0.66 for skull length in adult G. gorilla. Hartwig-Scherer and Martin displayed an interesting similarity between subadult human and nonhuman ape basicrania, and also indicated a strong contrast between humans and other apes in the performance of cranial variables as predictors of body weight

Nawrocki (1992)
In a test of the effect of shape on cranial thickness Nawrocki took thirty-five

craniofacial and eighteen bilateral and sagittal vault thickness measurements on a sample of recent and archaic modern humans and extinct hominids. He combined some of the cranial dimension measurements into an index of sphericity that was intended to remove size. He later recognized that size was not removed as illustrated by the significant correlations between it and his cranial variables indicative of size (maximum length, maximum breadth, and basion-bregma height). The individual vault thickness measurements were combined into five variables of average thickness summarizing entire vault thickness with and without the measurement at asterion, thickness of the anterior and posterior halves of the vault, and posterior thickness without asterion.

Nawrocki examined the relationship between his cranial measurements and body size (as indicated by femur length, vertical femoral head diameter, body height and weight) in the modern humans drawn from the Terry collection. He found several significant correlations between the individual ectocranial measurements and body size, the strongest involving body height (body length and the femoral measurements), but did not provide







12
results for each variable. The mean thickness variables lacked significant correlation to any of his size variables. Individual thickness measurements yielded a few significant relationships, notably between the femoral measurements and asterion and the parietal bosses (r = 0.24 to 0.41 for the former, -0.21 to -0.25 for the latter). Correlation specifically to body weight was not described in detail, although Nawrocki did attribute a lack of similarity between his results and those of Pensler and McCarthy to low sample sizes (n < 30) and low body weights in the Terry sample.

Comparisons of the index of sphericity to entire vault thickness produced variable results throughout Nawrocki's sample. The modem humans (n = 122) did not have a significant r-value (p > .05), but the archaic hominids (n = 61; including H. neandertalensis, H. heidelbergensis, H. erectus, and Homo habilis) produced a coefficient of 0.487 (p <

0.0001). The combined sample (n = 183) yielded a coefficient of 0.553 (p < 0.0001). Based on the combined sample results Nawrocki concluded that the positive correlation between vault shape and vault thickness supported his hypothesis that long low crania will have thicker vaults than high globular crania. He also found support in the archaic sample for a moderate relationship between face size and vault thickness, with bizygomatic breadth and facial height length providing the best correlation coefficients (r = 0.55 and 0.44 respectively). The strongest associations in the recent subsample were between body weight and mandibular body thickness and basion-prosthion length (r = 0.34 and r = 0.30, respectively).
Nawrocki's test of Demes' biomechanical shell model for the human cranium was most strongly supported in the archaic portion of the sample. Longer and lower vaults with less wall curvature were significantly associated with thicker vaults, but only in the archaic sample. Larger faces were correlated with thicker vaults, most strongly in the archaic sample. Nawrocki's results suggest an alternate model may be needed to explain vault thickness in recent humans.








Aiello and Wood (1994)

Aiello and Wood studied cranial measurements in anthropoids in order to develop regression formulae for predicting body weight in fossil hominids. They collected fifteen ectocranial and fourteen postcranial variables on a sample of 250 anthropoids. These authors used six orbital variables (breadth, height, area, biorbital and interorbital breadths, and postorbital breadth), biporionic breadth, two palatal variables, and six basicranial variables (three on the foramen magnum and three on the occipital condyles). Twenty-four individuals in the sample were human subjects with known body weights obtained during autopsies. The nonhuman subjects had body weights drawn from the literature (Harvey et al. 1987). The sample was divided into anthropoid and hominoid subgroups. I am focusing on the hominoid subgroup because the authors did not report results for the human subjects alone.
The cranial variables showed uniformly high correlation coefficients with body

weight using logl0 transformed group means (Table 1), while weaker relationships existed among the postcranial variables. The best variables in the hominoid sample were orbital area (r = 0.98), biporionic breadth (r = 0.98), and orbital height (r = 0.98), based on having a high r-value, a low standard error of the estimate (SEE), and a low percentage prediction error. Aiello and Wood cautioned against using anthropoid-based formula for hominoids, because in their sample less than half of the hominoids received a predicted body weight within 20% of the known weight (their criterion for high percentage prediction error).

The postcranial variables were drawn from the femur, tibia and humerus. Three measurements of the humerus (maximum length, epicondylar breadth, and distal joint breadth) performed as well as the best cranial predictors in the hominoid sample. Poor performance of the leg variables was attributed to the difference in body proportion between humans and the other large-bodied apes. This point again illustrates an effect of using a mixed-taxa reference sample.









Table 1. AieHlo and Wood's Results relative to Body weight Using a Mixed Hominoid Sample

Variable N r Orbital breadth 12 0.96 Orbital height 12 0.98 Orbital area 12 0.98 Interorbital breadth 12 0.81 Biorbital breadth 12 0.95 Postorbital breadth 12 0.73 Biporionic breadth 12 0.98 Intercanine breadth 12 0.89 Palate length 12 0.89 Foramen magnum length 12 0.90 Foramen magnum breadth 12 0.92 Foramen magnum area 12 0.92 Occipital condyle length 12 0.93 Occipital condyle breadth 12 0.90 Occipital condyle area 12 0.94 N are from species/sex means (six hominoid genera, two sexes, n = 12) Gauld (1996)
Gauld examined the relationship between vault thickness and body weight in a large (n = 235) sample of extant anthropoids in order to predict body weight in extinct hominids. Although detailed information was not provided, Gauld indicated that some nonhuman body weights in the extant sample were obtained as averages from the literature (Harvey et al. 1987 and Jungers 1988). The modern humans (n = 27) were drawn from archaeological collections and had body weights predicted from McHenry's (1988) equations. Gauld measured cranial thickness at five locations on the vault (midfrontal, bregma, midparietal, midtemporal squama, and inion) and calculated correlation coefficients and best fit lines (least square, reduced major axis and major axis) to define the relationship to body weight. The results indicated generally strong relationships between vault thickness and body mass for the entire catarrhine sample and the hominoid subgroup. The correlation coefficients between these variables and body weight ranged from 0.75 to 0.94 for the hominoid portion of the sample. Gauld's examination of the residuals from the least squares regression showed that the hominids (Australopithecus africanus, H. sapiens, and H erectus) shared a







15
pattern of deviation from the catarrhine regression. The greater absolute vault thickness in H. erectus was demonstrated by the greater shift, than seen in the other hominids, of its residuals from the regression line.

In her discussion Gauld references the low correlation coefficients obtained by Pensler and McCarthy as "not unexpected, given the increased number of environmental factors contributing to individual phenotypic thickness variation" (p. 420). Yet Gauld's sample consisted of a small number of humans from coastal and inland California, Japan, and Fiji. Her sample must reflect as many if not more different environmental factors as a group of American Whites and Blacks. Also the large sample size used by Pensler and McCarthy reduced the possibility of creating a sample of divergent individuals. Gauld's sample size did not do this.

Discussion

These five sources have presented contrasting images of the relationship between body weight and cranial measurements in modern humans. Pensler and McCarthy, Nawrocki, and the more limited work of Hartwig-Scherer and Martin all indicated low to moderate correlation coefficients with body weight. In contrast are the works of Aiello and Wood and Gauld that demonstrated strong relationships. Two critical methodological differences separate these studies; the source of body weights and the type of samples used. The former group used human and/or nonhuman subjects with recorded body weights, and analyzed samples consisting solely of humans. The latter group relied on combinations of published mean body weights for primate species, measured, and predicted body weights and used mixed-primate samples for their analyses.
In light of empirical research, the relationships between cranial measurements and body weight, in any primate or organism, are best illustrated by methodologies using actual observations. The results of Pensler and McCarthy, Hartwig-Scherer and Martin, and Nawrocki have more relevance, as their results are based on actual observations of body weight. That these authors reported results from observations of modern humans makes the








results applicable to predicting body weight in modern humans and organisms most like them, that is fossil hominids.

Reliance on a mixed primate sample, as Conroy (1987) indicated, facilitates over and under prediction of errors. Aiello and Wood demonstrated awareness of this problem in noting that the greater emphasis on leg length in modern humans reduced the correlation coefficient between the femoral measurements and body weight in their hominoid sample. The adaptations of bipedality are a defining morphological trend of the hominids, and it is partnered with the equally defining trend of increase in cranial capacity. A collection of extant hominoids sampled for cranial measurements has the same potential for reduced covariance as hominoids sampled for leg measurements.


The Project
These five works indicate that further testing is needed of the utility of cranial measurements, whether ectocranial or of cranial thickness, for estimating body mass in large-brained hominids. Utilizing a large sample size composed exclusively of modern humans with recorded body weights, in combination with an array of ectocranial and vault thickness measurements, I tested the utility of cranial measurements as predictors of body weight. Unlike the authors in this review, I also used a postcranial indicator of robusticity to investigate the contribution of skeletal size to vault thickness.

I designed a static narrow allometry study (Shea 1983, Smith 1985) such that adult specimens of a single species were used as the study sample. A static study was acceptable because the research questions were not aimed at ontogenetic changes in the hominids. A narrow study avoids the uncertainty of predicting body weight from a multi-species plot, especially when the focal specimen may not be represented in the species used to generate the plot. Only modern humans were sampled in order to avoid prediction errors possible in mixed-primate samples (Conroy 1987). The results of this study are certainly applicable to modern humans and are also a resource for predictions for the large-brained extinct







17
hominids, since modem humans are the only available reference sample. Direct measurements of body weight were required in order to avoid the problems of intervening variables (Hartwig-Scherer and Martin 1992).














MATERIALS AND METHODS


Sample Selection
Several challenges were encountered in identifying satisfactory sources of human cranial and postcranial measurements. Requisite for this analysis were a generous number of intact crania as well as vital statistics such as body length and weight While determining that the Terry collection (National Museum of Natural History, NMNH) met this need, I also drew upon my rapport with local medical examiners in Florida and Michigan to collect data in their offices. This choice limited feasible research protocols for the entire study, as will be described below.

Several biases were present in the study sample due to use of the morgue as a data source:

1. A morgue population represents local demographics in terms of population

ancestry.

2. Primarily people of European descent populated the resource areas.

3. Better health maintenance causes fewer women than men to be subject to

autopsy (see Florida statute 406.11 and Michigan statute 52.202).

4. Many older females had to be excluded due to pathological thickening of the

cranial bones resembling hyperostosis frontalis internal (HFI; Ant6n 1997,

Hershkovitz et al. 1999).
In order to complete the study in a timely manner with these biases in effect, an exclusively male sample of predominately European descent was accepted.

Measuring points were also subject to sample bias, especially in the choice of a

postcranial measurement site. The postcranial site could not alter or interfere with autopsy







19
protocol and had to be accessible without causing unreasonable or unethical changes to the disposition of the deceased. Given these limitations the best site was the clavicle, which unlike other long bones is frequently cut as a part of the autopsy protocol. Cuts were made at midshaft for one (usually the left) of the clavicles for a direct measurement of cortical thickness. Terry clavicles could not be sectioned so these clavicles were radiographed and cortical measurements taken from radiographs.

The autopsy and skeletal samples differed in the availability of cranial

measurements. Access to the cranial base and face was severely limited by autopsy protocol, so only a few measurements were available from this area for the autopsy sample. This limitation did not exist in the skeletal collection, so an extended set of cranial measurements was taken. Forty-seven male subjects composed the autopsy sample, while 100 males composed the skeletal sample.


Data Collection
Somatic Variables

Certain descriptive data were recorded for both samples. Each specimen in the skeletal collection had documented sex, age, race, body weight and body length. In the morgue sex and age were determined by physical observation and written documentation, respectively. Ancestry was recorded according to documentation associated with the individual or based on the medical examiner's determination, which was largely based on external phenotype and surname. Supine body length was recorded in inches from a steel measuring tape or ruler. Body weight was recorded in pounds on a digital floor scale or more rarely extracted from hospital intake records. The latter source was used if the individual had been on fluid therapy shortly before death. The floor scale measurements followed the autopsy protocol, which meant that the weights included some amount of clothing and or drapes. Recorded weights were not adjusted for clothing since the few kilograms of difference are not of antemortem or postmortem medical concern (William Hamilton, M.D., pers. comm.). Care was taken however, to weigh any large medical








devices, such as head-movement restriction halos, but this occurrence was very rare. Age was recorded in order to examine the sample for variation in cranial thickness, since variation in thickness with increasing age is still disputed (Tallgren 1974, Israel 1977). Individuals were excluded from the autopsy sample when they had cranial fractures that interfered with the measurements and when the subject was emaciated.

Occasionally some specimens had missing data for reasons other than the

occasional missed measurement, such as the one subject in the skeletal sample lacking a recorded stature measurement. Other discrepancies were due to unexpected incompatibility of the research tools to human variation, such as when the spreading calipers accommodated most but not all morgue subjects for the measurement of cranial base length (basion to nasion) where basion had to be approached from the inside of the vault. Quite frequently contact with basion could not be made. Contact on the endocranial surface was also a problem for some of the thickness measurements, particularly when an elongated frontal crest interfered with the midfrontal measurement (avoidable in the morgue sample, but not in the skeletal sample). The grooves for the middle meningeal arteries on the parietal might have interfered with thickness measurements, but the broad point of contact on the caliper extension arms would not fit within these grooves. Another difficulty involved the radiolucent character of some of the clavicle cortices in the skeletal sample. Occasionally radiographic examination of the clavicle yielded no measurable cortical thickness of the posterior or inferior walls, so no measurement was entered. In order to preserve a large sample size for later statistical analyses, the missing data were handled with pairwise comparison, such that missing cases were excluded from the analysis, but the subject and all its attached variables with complete cases was not excluded.
While I excluded the emaciated in the autopsy sample, no such effort was feasible for the skeletal sample. The nature of this study, examining cranial predictors of body weight rather than load-bearing bones (e.g., Jungers 1988 on teeth), makes a minimum body weight seem nearly inconsequential, as long as the individual was alive while having that








weight. The practical reason for avoiding emaciated subjects is to include only mobile individuals. A real population will have a range of body weights including both extremes. A population of wild nonhuman primates is unlikely to ever have individuals with the upper extremes of weight and more likely to contain individuals of the lower extremes, especially if seasonality is a part of their ecology. Considering that a collection of nonhuman primates would not be discarded because all of the specimens were captured in the season of their lowest weights (if this season was even recognized), it is wasteful to discard emaciated humans from this skeletal sample. Weight data, which are intended for a species mean should not be based only on individuals from the middle and upper ranges of their weights. Since many Terry subjects were of very low weight (Jantz and Moore-Jansen 1988), I will examine the proportion of low weight individuals in the skeletal sample and the effects of their inclusion on the correlation analyses.

Cranial Measurements

A total of twenty-three cranial measurements were utilized in the entire study, comprising sixteen ectocranial measurements and seven vault thickness measurements (Table 2 and Figure 1). The ectocranial measurements were drawn from Howells (1973), Bass (1995), and Giles and Elliot (1963). In order to accommodate positioning restrictions in the autopsy sample, cranial height was taken from porion to bregma following Stewart's direction (see Bass, 1995). Both this measurement and height from basion to bregma were taken in the skeletal sample. Ectocranial measurements were taken with GPM sliding and Paleotech@ spreading calipers, and for the measurement porion-bregma, a Paleotech� radiometer.

The thickness measurements illustrated in Figure 2 were drawn chiefly from Gauld (1992) with two exceptions. Gauld's point at inion was replaced with a point at lambda due to inaccessibility of inion in the autopsy sample, as well as general interference of internal occipital structures at this location. Gauld's point at the posterior-inferior quadrant of the parietal was replaced with a point in the superior-inferior quadrant. Most of the thickness







22
measurements required points to be located on the skull before measurement could proceed. The point at midfrontal was located by taking the frontal chord with the sliding caliper and marking its midpoint on the skull. Parallel to this point on the temporal line was the point temporal crest. Midparietal was located at the intersection of two lines on the right parietal, one running anterior-posteriorly from the midpoint of the coronal suture to the lambdoidal suture, and the other from medial to lateral from the midpoint of the sagittal suture. The subjective middle of the two superior quadrants formed the other parietal points. In the skeletal sample the thickness measurements were taken with a Mitotuyo� digital caliper fitted with extension arms. The extension arms allowed measurements to be taken by inserting one arm into the foramen magnum, bringing the other arm into contact with the outer table of the cranium at the desired point, and sliding the internal arm gently into contact with the inner table.

Selection of cranial measurements was aimed at investigating various research goals. Many of this study's measurements are associated with estimation of ancestry in modern humans (Giles and Elliot 1962), and may be useful in examining relationships between shape of the cranium and vault thickness. The Cranial Index (maximum cranial breadth* 100/maximum cranial length), was used as an indicator of cranial shape and employed in correlation analyses with cranial thickness.









Table 2. Cranial Measurements.


Ectocranial Measurements Variable No. Variable Name and Description
1 Maximum length (glabella-occipitale)
2 Maximum breadth (euryon-euryon) 3 Cranial base length (basion-nasion)
4 Biauricular breadth *
5 Minimum frontal breadth (frontotemporale-frontotemporale)
6 Frontal chord (nasion-bregma) 7 Parietal chord (bregma-lambda)
8 Height from porion (porion-bregma)***
9 External Palate Breadth 10 Cranial Height (basion-bregma) 11 Bizygomatic breadth (zygion-zygion) 12 Facial Length (basion- prosthion) 13 Facial height 14 Nasal height (nasion-nasospinale) 15 Nasal breadth (alare-alare) 16 Mastoid length* Cranial Thickness Measurements 17 Midfrontal, at the midpoint of the nasion-bregma chord** 18 Temporal crest, parallel to midfrontal on the temporal crest** 19 Bregma**
20 Midparietal, at the intersection of horizontal and vertical lines that
divide the parietal into quadrants**
21 Midpoint of the anterior-superior quadrant of the parietal*** 22 Midpoint of the posterior-superior quadrant of the paietal*** 23 Lambda***
*Giles and Eliot 1963; **Gauld 1994; ***Bass 1995
































Figure 1. Ectocranial Measurements. Measurement numbers correspond to Table 1


Figure 2. Locations of Cranial Thickness Measurements. Numbers correspond to Table 1.








Posteranial Measurements

The clavicles in the autopsy sample were measured for length with sliding calipers, marked at midshaft and sectioned with a Stryker@ saw. Since the bone was still in situ some soft tissue was included in the length measurement. The amount of soft tissue was believed negligible since the blades of the calipers allowed for extreme compression of soft tissue. Cortical thickness was measured at the superior-, inferior-, anterior- and posteriormost aspects of the cortex at midshaft. Superior-inferior and anterior-posterior diameters were also taken in both samples, by placing the blades of the sliding caliper perpendicular to the curvature (if any) of the shaft at midpoint, in order to obtain the minimal distance. The clavicle measurements are summarized in Table 3. Table 3. Clavicle Measurements.

Variable No. Variable Name
1 Maximum length*
2 Anterior-posterior diameter at midshaft**
3 Superior-inferior diameter at midshaft**
Cortical thickness at midshaft from anatomical position**
4 Anterior Cortex
5 Posterior Cortex
6 Superior Cortex
7 Inferior Cortex
*Bass 1995, ** See Text

To obtain cortical thickness the Terry clavicles were placed directly on the

radiographic plate in order that refraction not alter the cortical measurements. To test this methodology five clavicles from biological specimens were radiographed and then sectioned at midshaft with a hacksaw. These test clavicles were filmed at 65 kvp for 135 minutes on unscreened cassettes, using Kodak EM-1 Mammography film. The results of a one-sample t-test (a=O.05, df=4) examining the difference in thickness measurements showed that the null hypothesis (Ho: -0) could not be rejected for 3 of the 4 cortical thickness measurements. These results, summarized in Table 4, were considered sufficiently satisfactory to proceed with the radiographic protocol. For the Terry clavicles the same film as noted above was used, at 60 KVP for 4 seconds on a screened cassette. The clavicles







26
were filmed in superior-inferior and anterior-posterior views so that cortical thickness could be measured on the same axes as in the autopsy sample. Measurements on the radiographs were taken with GPM sliding calipers. Maximum length of the clavicles was measured on the osteometric board.

Table 4. Results of the One-Sample t-test for Sectioned and Radiographic Clavicle Measurements

Variable t df Sig. (2-tailed) Mean Difference DIFFANT 1.812 4 0.144 03640 DIFFPOST -6.498 4 0.003 -0.1260 DIFFSUP -1.570 4 0.192 -0.2740 DIFFINF 0.066 4 0.950 0.01200 The variables shown are the difference between the respective anterior, posterior, superior, and inferior measurements on the cortical bone and their corresponding measurements on the radiograph.
Estimation of Lean Body Weight

The autopsy sample presented a special situation in that it was possible to take skinfold measurements that could be used to estimate fat mass, and by subtraction, lean body mass. Martin et al. (1985) questioned the reliability of skinfold measurements for estimating fat mass due to issues such as tissue compressibility, water content of subcutaneous fat, and the variable proportion of subcutaneous to deep body fat. They only support their use in conjunction with other anthropometrics like height and body weight. Shepherd (1991), while noting these and other potential sources of error, concludes that skinfold prediction of body fat can have less error than estimates from densitometry. Skinfold measurements are the most accessible means of estimating fat mass in the autopsy setting, since more conclusive techniques (e.g. densitometry) require excessive manipulation of the deceased. Skinfold measurements were taken in millimeters with a Lange skinfold caliper at the triceps and subscapular location for each individual. Three skinfold readings were taken at each site and the average of the three reported as the final skinfold measurement
With living subjects hydration, or the lack thereof, is a major influence on skinfold measurements. Although cadavers do not lose moisture through respiration and sweating as








in living subjects, they do lose moisture through traumatic and postmortem fluid loss, and from evaporation in a ventilated refrigerated environment. The effect of fluid loss on skinfold measurements in cadavers, at least at the triceps and subscapular sites, is a balance between dehydration from air movement and rehydration from lividity. (Lividity is the subtle (i.e., not an edema) pooling of tissue fluids due to the influence of gravity, such that uncompressed skin takes on a reddened or livid appearance.) On a supine body lividity is most apparent on posterior limb surfaces and the back itself. Lividity may more than compensate for any type of fluid loss including exsanguination, as remaining blood, lymph and cellular fluids stay in the posterior tissues. Also, cadavers are usually kept covered until autopsy, which would impede surface dehydration. Aiello and Wood (1994) allowed no more than 5 days for posthumous weighing in order to prevent effects of dehydration, but from personal observation this researcher noted that a draped body in a refrigerated environment lost lifelike skin texture in three days. Therefore subjects were avoided if they had been stored more than 3 days. In addition the skinfolds were compared to normative standards for Americans (Frisancho 1990) to determine if the skinfolds produced weight classifications similar to those predicted from height.

The autopsy and skeletal samples represent two data sets. The autopsy sample utilizes nine ectocranial, seven vault thickness, seven clavicle, and various organ measurements, in addition to the general statistics like weight, stature, and age. The skeletal sample is similar, but has seven additional ectocranial and no organ measurements. In the data analysis the intersection of the two samples is used as the Combined sample.

Data Analysis
All English measurements taken during data collection were automatically converted to metric equivalents within the data collection spreadsheet program Microsoft Excel 98. Subsequent statistical analyses involving descriptive statistics (Tables 4-6), correlation, linear regression, and principal component analysis (PCA) were conducted with SPSS 10.








Data analysis in all samples included production of sample descriptives and

calculation of Pearson's correlation coefficient between body weight and all applicable cranial and clavicle measurements. To investigate the role of skeletal robusticity in cranial thickness, the correlation between cranial thickness and clavicular cortical thickness was calculated in the three samples. The Cranial Index was employed in correlation analysis with the vault thickness variables for the skeletal sample only, since the cranial height measurement (porion-bregma) used in the autopsy sample was not applicable.

Regression analysis was reserved either for variables showing strong relationships to body weight, or when the regression coefficients were needed for comparison to those in the literature. I considered a strong relationship to be a correlation coefficient as high as the poorest performing postcranial measurement in modem humans (i.e. r = 0.69 for radius circumference, Hartwig-Scherer 1994).

Previous studies (Gauld 1992, Aiello and Wood 1994) have implied that body weight and cranial measurement data do not meet requirements of normality and homoscedasticity, and that logarithmic transformation of the variables is necessary before regression analysis. Transformation of the variables wasjustified by an examination of the plot of the residuals of the non-transformed data.

In studies of allometry there has been some debate over the correct type of linefitting model to use, with each model having its benefits. The traditional Model 1 (Sokal and Rohlf 1995) approach, Least Squares (LS) regression, assumes that the independent variable is sampled without error from its population. In reality this assumption cannot be met in most allometric studies where both independent and dependent variables are measured quantities. The Least Squares technique also has three methods of correction of bias that is introduced into data when log transformed variables are returned to their original states. Least Squares has been considered the better technique simply for predicting one variable from another (Smith 1994), which agrees with the goal of this study, to predict body weight from cranial variables. Model II (Sokal and Rohlf 1995) techniques, e.g.,







29
Reduced Major Axis and Major Axis, do not have the bias corrections, but they also do not assume that either of the variables are sampled without error. The Model II approach is also best when it is desirable to violate the restriction against extrapolating predictions beyond the range of the dependent variable. The LS technique, with its bias corrections and strength as a predictive tool, seems more appropriate for this study. Also, if the coefficients of determination are high, then the results of LS and RMA or MA analysis will be similar. LS will be the primary line fitting technique used, while RMA and MA will be conducted secondarily for comparison.

Three criteria will be used to determine if the 22 cranial variables are actually

suitable for predicting body weight A high correlation coefficient (above 0.69) is the first requirement. The second criterion is a low standard error of the estimate (SEE), which indicates the accuracy with which the regression equation depicts the strength of the relationship between dependent and independent variable. A low SEE also indicates that the confidence bands around the regression line will be relatively narrow. Finally I will examine the percentage prediction error (PE), which in low values indicates good predictive ability of the equation.

I explored a chronology issue generated by the sample design. A gap of several

generations exists between the Terry specimens and those from autopsy. Two-tailed t-tests were conducted to determine if there were significant differences in means between the two samples for the various anthropometric and skeletal measurements. Previous work (Warren et al. 2000) comparing recent forensic and early 9 century skeletal samples (Terry and Hamann-Todd collections) had suggested a secular change in cranial measurements, at least as concerned measurements used to estimate ancestry. As many of those same measurements (and possibly crania) are employed in this study, it was interesting to review differences in mean measurements and cranial indices between the two samples.

To produce a meaningful reduction of the many variables principal component
analysis (PCA) was conducted, for the three samples. PCA is a large sample size procedure








(at least 5 subjects per variable; Hatcher and Stepanski 1994), so to accommodate this restriction subsets of the variables were analyzed. In all PCA procedures non-log transformed data were used, and only the correlation matrix was analyzed. Retention of principal components was based on a combination of four criteria:

1. Eigenvalues greater than one for each component

2. Components should account for a reasonable (e.g. >10%) proportion of the variance

represented by all the components

3. Components should be relatively closely grouped on a scree plot

4. Components should be interpretable

A component was interpretable if it had at least three variables with significant loadings, and these variables measured the same construct (e.g. postbregmatic cranial thickness). Different components should have different highly loading variables measuring different constructs. In rotation the components should demonstrate "simple structure" (Hatcher and Stepanski 1994). Simple structure indicates that the rotated (varimax rotation, which produces uncorrelated components) factor pattern shows at least three high loading variables on a component while the remaining variables have near zero loadings. Also the individual rotated components must have variables with a dichotomous loading pattern (i.e. very high and very low). Strength of loading is subject to interpretation since in the social sciences a loading of 0.4 is sufficient (Hatcher and Stepanski 1994), but several of the variables in this study show loadings of 0.7 and higher. My previous experience with skeletal measurements in PCA causes me to consider a loading of 0.7 and greater to be high, 0.5 and 0.6 moderate, and values below 0.5 are low.














RESULTS


The Combined Sample: Descriptive Statistics In the interest of treating the sample as simply a collection of male modem humans the autopsy and skeletal data were combined into one large sample (n = 147). Only the variables in common to the two samples were used. The somatic variables (Table 5) are shown first, followed by the cranial and clavicular measurements for the combined sample (tables 6 and 7). Of these, 47 are autopsy subjects and 100 are skeletal. Table 5. Descriptive Statistics for Somatic Variables in the Combined Sample.

..Standard N Minimum Maximum Mean Deviation Age (years) 147 18 87 54.56 15.24 Weight (kg) 147 31.00 184.1 64.63 26384 Body length (cm) 146 152.40 199.40 172.83 9.08

The lightest individual (31 kg) belonged to the skeletal sample and the heaviest to the autopsy sample. Nearly half (42) of the skeletal sample were below 45 kg, while obesity was the most common extreme in the autopsy sample. I examine the low body weights in the skeletal sample in the correlation analyses for the skeletal sample, while fatness in the autopsy sample is addressed in the skinfold analysis. Descriptive statistics for the cranial and clavicular measurements in the combined sample follow.

The ectocranial measurements (Table 6) do not deviate from what is expected in a human sample. Combining four samples from Howells (1973) that would most likely resemble this American sample in ancestry created a comparison sample. The means of analogous measurements were averaged for three European (Norse, Zalavar, and Berg) and one West African (Dogon) samples (Table 7). Howells used a large variety of measurements in his study, several of which were also used in this study. He also provided 31







32
detailed descriptions of how the measurements were taken, so it was quite simple to verify
that the measurements he used were functionally identical to the ones in this study.
Table 6. Descriptive Statistics for Cranial Variables of the Combined Sample.

Variable Name N Minimum Maximum Mean Standard Deviation Maximum Length 146 166 207 186.40 7.59 Maximum Breadth 147 127 160 142.63 5.93 Minimum Frontal Breadth 144 86 120 98.99 5.76 Biauricular Breadth 147 109 136 123.29 533 Height from Porion 147 111 150 125.71 7.25 Cranial Base Length 136 91 121 101.83 5.36 Facial Height 129 57 96 71.05 7.15 Frontal Chord 147 98 126 113.58 5.29 Parietal Chord 147 99 136 116.19 7.08 Midfrontal 146 3 12 6.40 1.73 Temporal Crest 145 2 9 5.53 1.43 Bregma 147 2 10 6.57 1.42 Mdparietal 146 2 11 5.69 1.68 Anterior Parietal 147 3 11 6.05 1.52 Posterior Parietal 146 3 11 6.34 1.72 Lambda 147 3 16 7.71 2.17
Measurements in mm.
Table 7. Comparison of Shared Ectocranial Measurements in Howells' " Mixed
Race" Sample and the Combined Sample

Measurement Howell Combined Sample t Sig. 2-tailed Mean Mean Mean Difference Maximum Length 182.97 186.40 5.47 0.00 3.43 Maximum Breadth 142.04 142.63 1.20 0.23 .59 Biauricular Breadth 122.79 123.29 1.13 0.26 .50 Cranial Base Length 100.06 101.83 3.86 0.00 1.77 Facial Height 67.54 71.05 5.58 0.00 3.51 Frontal Chord 111.73 113.58 4.24 0.00 1.85 "arietal Chord 113.04 1 16.19 1 5.40 0.00 3.15


The combined sample is not broadly deviant from what one would expect for modern humans, although the one-sample t-test indicates several significant differences (Table 7). The hypothesis that a mean measurement from Howells such as maximum length is drawn from the same population as this American sample must be rejected. Table
7 shows that this hypothesis should be accepted for only two variables, maximum breadth and biauricular breadth. The lack of similarity between samples is most likely due to secular changes caused by the distance of years, gene flow, and nutritional status that







33
separates the two populations. The alternative is that Howells' collection of Europeans and West Africans bear little resemblance to this sample. This is unlikely, especially since a similar degree of difference is seen later between the Terry and autopsy samples (Table 30).

Comparative clavicle measurements are more difficult to find, but again the
combined sample measurements are not atypical for a sample of modem humans. The University of Tennessee Forensic Database (FORDISC 1996) contains mean values for clavicle length, AP and SI diameter. This database is composed of modern (late 20 century and later) forensic examinations in the United States. Table 8. Descriptive Statistics for Clavicle Variables in the Combined Sample N Minimum Maximum Mean Std. Deviation FORDISC Clavicle Length 147 119.0 184.0 154.25 1036 158.18 AP Diameter 147 8 110 13.31 8.29 12.70 SI Diameter 147 1 18 11.60 2.55 11.14 Anterior Cortical
Thickness 142 0.5 5.0 2.09 1.01 n/a Posterior Cortical
Thickness 146 1 6 2.57 0.81 n/a Superior Cortical
Thickness 147 1 5 236 0.85 n/a Inferior Cortical
Thickness 138 1 7 2.26 1.20 n/a Measurements are in mm. n/a: not applicable

Combined Sample: Relationships Between Variables
Tables 9 and 10 summarize the covariance relationships between body weight and the somatic and skeletal measurements. Table 9 reports stature and the clavicular measurements, while 10 reports the cranial measurements.

The clavicle measurements are strongly correlated with body weight, with anteriorposterior diameter of the clavicle being the only exception (Table 9). While highly significant, the r-values range from low to moderate (about 03 to 0.6), with the highest being 0.667 for inferior cortical thickness.

Many significant correlations exist between body weight and several of the
ectocranial variables in the combined sample (Table 10) with the variable height from porion having the highest r-value of 0.628. The next highest correlation coefficients (of about 0.4)







34
are for cranial base length, minimum frontal breadth, and maximum length. In contrast the
cranial thickness variables show few significant correlations and uniformly low r-values.
Temporal crest has the highest correlation to body weight with an r-value of 0.289.
Table 9. Combined Sample Correlation Coefficients Between Body Weight and the Somatic Variables (including Clavicle)

Measurement N r Significance Body Length cm 146 0.559** 0.000 Clavicle Length 147 .307** 0.000 AP Diameter 147 0.081 0331 SI Diameter 147 0.562** 0.000 Anterior Cortical Thickness 142 .508** 0.000 Posterior Cortical Thickness 146 0.348** 0.000 Superior Cortical Thickness 147 .581 ** 0.000 Inferior Cortical Thickness 138 0.667** 0.000
p < 0.05; ** p < 0.01
Table 10. Combined Sample Correlation Coefficients Between Body Weight and the Cranial Variables

Measurement N r Significance Maximum Length 146 0.418** 0.000 Maximum Breadth 147 .152 0.067 Minimum Frontal Breadth 144 0.430** 0.000 Biauricular Breadth 14 0.189* 0.022 Height from Porion 147 0.628** 0.000 Cranial Base Length 136 0.490** 0.000 Facial Height 129 0.381** 0.000 Frontal Chord 147 0.286** 0.000 Parietal Chord 147 0.057 0.492 Midfrontal 14 0.070 0.402 Temporal Crest 14 .289** 0.000 Bregma 14 .133 0.109 Midparietal 146 0.164* 0.048 Anterior Parietal 147 0.163* 0.049 Posterior Parietal 146 0.031 0.714 Lambda 147 -0.080 0338
* p < 0.05; ** p < 0.01
Correlation coefficients between the clavicular cortical thicknesses and the cranial
thickness variables (Table 11) do not support the hypothesis that cranial thickness is a
function of skeletal robusticity. The resulting trend is for a lack of significant correlation.
Table 11 shows only three significant r-values: Cranial thickness at lambda has a weakly
negative correlation to anterior cortical thickness of the clavicle, and the temporal crest








measurement shows weakly positive correlations to both superior and inferior cortical
thickness variables. These results support rejection of the robusticity hypothesis, as there is
no pattern of relatedness between cranial thickness and cortical thickness in the clavicle.
Table 11. Combined Sample Correlation Coefficients Between the Clavicle Cortical Thickness and Cranial Thickness Variables

Measurement Anterior Cortical Thickness Posterior Cortical Thickness N r Significance N r Significance Midfrontal 141 -0.020 0.813 145 0.005 0.954 Temporal Crest 140 0.101 0.237 144 0.033 0.698 Bregma 142 0.010 0.909 146 -0.021 0.802 Midparietal 141 0.045 0.595 145 0.106 0.205 Anterior Parietal 142 -0.036 0.671 146 -0.030 0.718 PosteriorParietal 141 -0.035 0.680 145 -0.006 0.944 Lambda 142 -0.190 0.024* 146 0.010 0.903

Measurement Superior Cortical Thickness Inferior Cortical Thickness N r Significance N r Significance Midfrontal 146 0.037 0.656 137 -0.010 0.904 Temporal Crest 145 0.208* 0.012 136 0.227** 0.008 Bregma 147 -0.024 0.772 138 0.069 0.421 Midparietal 146 0.102 0.220 137 0.120 0.164 Anterior Parietal 147 0.095 0.253 138 0.076 0377 Posterior Parietal 146 -0.030 0.723 137 0.044 0.612 Lambda 147 -0.158 0.056 138 -0.132 0.122
*p < 0.05; ** p < 0.01
Significant relationships between the individual ectocranial variables and the vault
thickness variables are more common (Table 12). Correlations coefficients are all positive
but lower than 0.400. Maximum length is the only variable to be significantly correlated to
all of the ectocranial measurements in the combined sample, although the relationship to
lambda is not as strong as with the other thickness measurements. Height from porion and
frontal chord are significantly correlated to all of the cranial thickness variables but lambda,
although the latter ectocranial variable is not grossly out of significance (p = 0.068). The
behavior of frontal chord suggests its correlation to the thickness variables is a reflection of
that for maximum length, as the two ectocranial variables both cover sagittal vectors.
Following this pattern it is notable that parietal chord is also only slightly out of significance








for three thickness variables. Cranial base length, also a sagittal measurement, follows
essentially the same pattern, although it greatly lacks significance to lambda.
Table 12. Combined Sample Correlation Coefficients Between the Ectocranial and Cranial Thickness Variables
Measurement Maximum Length Maximum Breadth N r Significance N r Significance Midfrontal 145 0.393** 0.000 146 0.157 0.059 Temporal Crest 144 0.354** 0.000 145 0.141 0.090 Bregma 146 0.355** 0.000 147 0.047 0.569 Midparietal 145 0.342** 0.000 146 0.226** 0.006 Anterior Parietal 146 0.372** 0.000 147 0.276** 0.001 Posterior Parietal 145 0.331 ** 0.000 146 0.169* 0.042 Lambda 146 0.185* 0.025 147 0.012 0.890

Measurement Minimum Frontal Breadth Biauricular Breadth N r Significance N r Significance Midfrontal 143 0.175* 0.037 146 0.222** 0.007 Temporal Crest 142 0.276** 0.001 145 0.281** 0.001 Bregma 144 0.144 0.086 147 0.166* 0.044 Midparietal 143 0.280** 0.001 146 0.193* 0.020 Anterior Parietal 144 0.247** 0.003 147 0.228** 0.006 PosteriorParietal 143 0.132 0.116 146 0.113 0.173 Lambda 144 -0.001 0.992 147 0.068 0.415

Measurement Height from Porion Cranial Base Length N r Significance N r Significance Midfrontal 146 0.201* 0.015 135 0.253** 0.003 Temporal Crest 145 0.262** 0.001 134 0.298** 0.000 Bregma 147 0.220** 0.007 136 0.249** 0.003 Midparietal 146 0.296** 0.000 135 0.209* 0.015 Anterior Parietal 147 0.303** 0.000 136 0.168 0.051 PosteriorParietal 146 0.181* 0.029 135 0.177* 0.040 Lambda 1 47 0.041 0.618 136 0.140 0.103

Measurement Facial Height Frontal Chord N r Significance N r Significance Midfrontal 129 0.233** 0.008 146 0.289** 0.000 Temporal Crest 129 0.348** 0.000 145 0.184* 0.026 Bregma 129 0.095 0.282 147 0.301** 0.000 Midparietal 128 0.181* 0.041 146 0.226** 0.006 Anterior Parietal 129 0.112 0.206 147 0.387** 0.000 Posterior Parietal 128 0.173 0.051 146 0.229** 0.005 Lambda 129 0.039 0.658 147 0.151 0.068








Table 12. Continued.

Measurement Parietal Chord
N r Significance
Midfrontal 146 0.182* 0.028 Temporal Crest 145 0.147 0.077 Bregma 147 0.299** 0.000 Midparietal 146 0.148 0.075 Anterior Parietal 147 0.161 0.052 Posterior Parietal 146 0.253** 0.002 Lambda 147 0.267** 0.001
* p < 0.05; ** p <0.01

The thickness variables are well correlated amongst themselves (data not shown), so in consideration of this average vault thickness variables were calculated. Anterior thickness, posterior thickness and average vault thickness were correlated to the ectocranial measurements (Table 13). The ectocranial variables are uniformly positively correlated with vault thickness. Maximum length is conspicuous for having the highest r-value in all thickness areas, indicating that increases in vault length are associated with moderate increases in vault thickness. In terms of average vault thickness, the highest correlation coefficients are for variables following cranial length, (maximum length, Frontal and parietal chords), followed by vault height measurements (height from porion, cranial base length). Cranial breadth measurements are the most poorly correlated, along with the single facial measurement facial height These data support Nawrocki's conclusion that larger vaults are associated with greater vault thickness.

To assess the effect of shape on vault thickness the cranial index and the vault thickness measurements were correlated (Table 14), but they showed few significant relationships, all of which are negative. Measurements at midfrontal and bregma indicate that as the cranial index increases thickness decreases. That is, as the skull approaches a spherical shape vault thickness at these midline points decreases.

The cranial breadth to length categories for the cranial index are correlated to the
vault thickness measurements in Table 15. The narrow, dolichocranic skulls lack significant correlation to any vault thickness measurements, but as the skulls become increasingly









Table 13. Combined Sample Correlation Coefficients between Ectocranial and Average Thickness Variables

Anterior Thickness Posterior Thickness Average Vault Measurement Thickness n r Sign. n r Sign. n r Sign. Maximum Length 143 0.482** 0.000 144 0.379** 0.000 146 0.443** 0.000 Maximum Breadth 144 0.138 0.100 145 0.195* 0.19 147 0.198* 0.016 Minimum Frontal 141 0.258** 0.002 142 0.188* 0.025 144 0.232** 0.005 Breadth
Biauricular Breadth 144 0.305** 0.000 145 0.175* 0.035 147 0.236** 0.004 Height from Porion 144 0.291** 0.000 145 0.240** 0.004 147 0.281** 0.001 Cranial Base Length 133 0.346** 0.000 134 0.211* 0.015 136 0.280** 0.001 Facial Height 129 0.288** 0.001 127 0.149 0.094 129 0.225* 0.011 Frontal Chord 144 0.339** 0.000 145 0.300** 0.000 147 0.339** 0.000 Parietal Chord 144 0.267** 0.001 145 0.268** 0.001 147 0.286** 0.000

Table 14. Combined Sample Correlation Coefficients Between the Cranial Index and Cranial Thickness Variables

Measurement Cranial Index
N r Significance
Midfrontal 145 -0.180* 0.030 Temporal Crest 144 -0.158 0.058 Bregma 146 -0.247** 0.003 Mdparietal 145 -0.089 0.289 Anterior Parietal 146 -0.080 0340 Posterior Parietal 145 -0.128 0.125 L.ambda 146 -0.145 0.081
* p < 0.05; ** p < 0.01

broad for their lengths significant correlations appear. The skull of average breadth for its

length (i.e., mesocranic) shows significant positive correlations to the anterior parietal

measurement The brachycranic skulls are significantly correlated to decreases at the

posterior parietal. Although consistent pattern is lacking in the location of significant
correlation, it is interesting that the rounder brachycranic skulls show mostly negative

correlation to vault thickness, since spherical shapes are better at resisting compression.

Nawrocki (1992) also found reduced vault thickness with reduced vault length. Correlation

between average vault thickness and the cranial index categories is non-significant, but

negative for the brachycranic skulls (Table 16).







39
Table 15. Combined Sample Correlation Coefficients Between the Cranial Index Shape Categories and the Cranial Thickness Variables

Category Dolichocrany Mesocrany
Measurement N r Significance N r Significance Midfrontal 44 0.104 0.500 73 0.157 0.186 Temporal Crest 43 0.234 0.132 73 0.112 0347 Bregma 44 0.032 0.835 74 0.083 0.482 Midparietal 43 -0.018 0.908 73 0.140 0.239 Anterior Parietal 44 -0.037 0.811 74 0.296* 0.011 Posterior Parietal 44 -0.032 0.834 73 0.181 0.125 Lambda 44 0.151 0.329 74 0.071 0.546

Category Brach crany
Measurement N r Significance Midfrontal 26 -0.041 0.842 Temporal Crest 25 0.210 0.210 Bregma 26 -0.071 0.732 Midparietal 25 -0.205 0326 Anterior Parietal 26 -0.270 0.182 Posterior Parietal 26 -0.391 * 0.048 Lambda 26 -0.039 0.498
* p < 0.05; ** p < 0.01

Table 16. Combined Sample Correlation Coefficients between Cranial Index Categories and Average Vault Thickness

Average Vault Thickness
N r Significance Dolichocrany 44 0.084 0.586 Mesochrany 74 0.194 0.097 Brachycrany 26 -0.269 0.186
*p < 0.05; ** p < 0.01

Combined Sample: Principal Component Analysis
The principal component analysis yielded more information on the relationships
between variables in the sample (Tables 17 and 18). Two analyses were performed for the
combined sample, the first using all of the variables except age (Table 17), and the second

only the cranial variables and body weight (Table 18).
Table 17 illustrates the number of components retained when using a large set of
variables and the "eigenvalue 1" rule. The seven components shown have unrotated

eigenvalues of 7.02, 3.81, 1.75, 1.47, 1.26, 1.07, and 1.03, respectively. Together these
components account for about 69% of the variance in the data set. The eighth component







40
has an eigenvalue close to 1.0, but it did not resemble any of the other criteria for retaining a component. In the combined sample some variables with moderate loadings also tended to load moderately on more than one component, which makes them good candidates for exclusion from future PCA with this data set (Hatcher and Stepanski 1994 p.477). Consider maximum length, which loads only moderately (about 0.5), but does so on components three and four. Variables with low loadings on more than one component also violate the "simple structure" rule. Other obviously discardable variables are minimum frontal breadth, height from porion, parietal chord and body length. Still others (see body weight) are suspicious for having low loadings on other components.

Of the seven components shown in Table 17, only the first two have high loading variables occuring in an interpretable pattern (shown in bold). Component 1 is a clavicle cortical thickness factor, with strong loadings from all four cortical thickness measurements. Body weight also has its strongest loading, a moderate 0.648, on this component (but note its loadings on the other components). This makes sense, as the PCA is derived from the correlation matrix, and body weight has some of its highest correlations (around 0.5) with the clavicle cortical measurements. Only one cranial measurements was strongly correlated with body weight (height from porion, r = 0.628), and some moderate loading can be seen from this variable on this component. Yet height from porion is another candidate for removal for loading on more than one component. Component 2 is clearly the "cranial thickness" component, with high loadings from the parietal thickness measurements and moderate loadings from the more anterior thickness measurements. This component is particularly well defined because even the moderately loading components have miniscule loadings on the other components. Components 3 through 7 are handicapped by either a lack of highly loading variables or insufficient number of variables. This limitation is due in part to the number and variety of cranial variables available to the combined sample.
In the second PCA, as in the first, only two components are interpretable (Table 18 in bold). The four components account for 63% of the total variance, having unrotated








Table 17. Combined Sample Principal Components Using All Variables (Except Age)

Component
1 345 7 Maximum 0.139 0.295 0.585 0.501 0.0504 0.17 0.15 Length
Maximum -0.0801 0.0965 0.10 0.151 0.876 -0.097 0.082 Breadth
Minimum 0.328 0.095 0.56 0.16 0.471 -0.0782 0.0627 Frontal
Breadth
Biauricular 0.00898 0.152 0.035 0.0556 0.817 0.21 -0. Breadth
Height from 0.502 0.147 0.291 0.57 0.191 0.0020 0.119 Mon
ranial Base 0.315 0.167 0.655 0.1 0.0469 0.24 -0.122 Length
Facial 0.32 0.125 0.743 -0.0882 0.0561 0.094( -0.077 Height
Frontal 0.11 0.228 -0.0378 0.733 0.234 0.14 0.186 Chord
Parietal -0.12 0.138 0.502 0.289 0.0330 0.143 0.49 Chord Midfrontal -0.12 0.56 0.226 0.325 0.0131 0.0131 -0.221 Temporal 0.126 0.598 0.291 0.0341 0.13 0.111 -031 Crest
Bregma -0.0981 0.63 0.10q 0.352 -0.099" 0.22E -0.018 Midparietal 0.176 0.813 0.073 0.000143 0.186 -0.122 0.0277 Anterior 0.055 0.802 -0.028 0.25 0.191 -0.032 0.023 Parietal 0.
Posterior 0.0281 0.85 0.093 -0.035 0.0584 -0.062 0.2 Parietal
Lambda -0.197 0.642 0.0573 -0.0912 -0.079q 0.30 0.211 Weight kg 0.648 0.0282 0.242 0.381 0.13 0.12 -0.19 Stature cm 0.327 0.0721 0.263 0.402 0.00281 0.591 -0.1 Clavicle 0.251 0.0391 0.152 0.0562 0.0792 0.81 -0.00503 Length
AP Diameter -0.0159 0.0554 -0.038 0.0714 -0.0491 -0.070 0.67
SI Diameter 0.515 0.0241 0.323 0.36 0.0121 -0.17 -0.344 Anterior 0.823 -0.0818 0.128 0.070 0.0097 0.064- 0.113 Cortical
Thickness
Posterior 0.739 0.0486 -0.00682 -0.187 -0.0786 0.24 0.0953 Cortical
Thickness
Superior 0.83 -0.00312 0.216 0.00187 0.0276 0.12! -0.126 Cortical
Thickness
Inferior 0.848 0.0213 0.13 0.204 -0.04 0.043 -0.095 Cortical
Thickness







42
eigenvalues of 5.62, 237, 1.50, and 1.22 respectively. Component 1 is easily interpretable
as a primarily postbregmatic cranial thickness component. The second component is not
easily defined. Cranial base length and facial height load highly (0.7) on the component,

but the relationship between the two variables is not clear. The only other variable with
exclusively high loading on this component is body weight. While it is a sound component
in terms of loadings, its interpretation is unclear. The high degree of inter-relatedness of
the variables in this study has made it difficult to arrive at components describing a simple
construct. Component 2 is the only evidence of body weight grouping with other variables.
Keep in mind that at the very least PCA is a technique to cause related variables to group
together when dealing with a large number of variables. Tables 17 and 18 show that when

the non-cranial variables are present body weight loads most highly with them, while
contributing moderate loadings on to components 3 and 4. With the clavicle measurements

removed, body weight groups with two cranial measurements (cranial base length and facial
Table 18. Combined Sample Principal Components of the Cranial Variables and Body Weight

Component
1 2 3 4
Maximum Length 0.29 0.52 0.62- 0.0271 Maximum Breadth 0.081 0.0513 0.157 0.861 Minimum Frontal Breadth 0.060 0.614 0.238 0.411 Biauricular Breadth 0.15 0.12 -0.00 0.80M Height from Porion 0.076 0.574 0.532 0.21 Cranial Base Length 0.17 0.777 0.112 -0.0252 Facial Height 0.13 0.774 -0.000429 -0.0431 Frontal Chord 0.18 0.12 0.662 0313 Parietal Chord 0.17 0.0784 0.75 -0.0564 Midfrontal 0.594 0.154 0.179 0.11q Temporal Crest 0.604 0.421 -0.106 0.135 Bregma 0.65M 0.0671 0.329 -0.043q Midparietal 0.761 0.18M -0.00871 0.204 Anterior Parietal 0.77 0.080 0.157 0.266 Posterior Parietal 0.841 0.041q 0.090q 0.043 Lambda 0.66 -0.122 0.194 -0.15 Weight kg -0.0386 0.74A 0.15 0.16







43
height), which were also "distracted" by the clavicle measurements. The small number of variables in the combined sample obscures the pattern of relationship, but a look at the separate parts of the sample provides more information.

Combined Sample: Regression Analysis Additional description of the covariance relationship between the cranial variables and body mass in this sample of modem humans required the use of line-fitting techniques to the scatter of data represented by the correlation coefficient. The r-values produced in this study, especially as seen in the log-transformed data, do not support the generation of predictive tools. As discussed in the protocol Least Squares (LS) regression is unfit for this type of data because neither the dependent nor independent variables were sampled without error. Aiello and Wood (1994) had found that if r-values are high any line-fitting technique will suffice, but those circumstances do not pertain here. This analysis is conducted in order to produce data more suitable for comparison to the literature.

The LS regression coefficients for the combined sample are shown in Table 19. The correlation coefficients differ from those shown in Table 10 because these are from logl0 transformed data. The highest r-value is for height from porion, which also happens to have the lowest standard error. The next highest r-values are for cranial base length and minimum frontal breadth. These results are atypically low for the literature, where log transformed cranial variables have r-values ranging from 0.7 to 0.9, with the majority being above 0.95 (Aiello and Wood 1994, Gauld 1996). However, the equivalent of only one measurement is shared by this study and the work of Aiello and Wood (1994; their biporionic breadth). Biporionic breadth is one of their three best performing variables for predicting body weight in a hominoid sample with an r-value of 0.98. In contrast, biauricular breadth in this study yielded an r-value of 0.18. These results are more similar to those of Nawrocki's (1992) human sample from the Terry collection. Similarly, the cranial thickness measurements have low r-values, the highest being 0.25 for temporal crest. Gauld (1996) reported r-values of 0.94 and higher for her hominoid sample at the







44
midfrontal, bregma, and midparietal measurement sites. The deviation of regression results
from the published literature will be addressed more fully in the discussion. A separate
examination of the two parts of the combined sample follows.
Table 19. Combined Sample Regression Coefficients For Body Weight Versus the Cranial Variables

Variable N r Slope Intercept SEEa F Sign. Maximum Length 146 0.39 3.56 -6.26 0.15 26.07 0.000 Maximum Breadth 147 0.14 1.20 -).80 0.16 2.72 0.102 Minimum Frontal Breadth 144 0.42 2.69 -3.59 0.14 30.80 0.000 Biauricular Breadth 147 0.18 1.50 -136 0.16 4.69 0.032 Height from Porion 147 0.63 4.06 -6.74 0.13 92.96 0.000 Cranial Base Length 136 0.48 330 -4.85 0.14 38.95 0.000 Facial Height 129 0.37 1.37 -0.78 0.15 19.97 0.000 Frontal Chord 147 0.29 2.31 -2.96 0.15 13.49 0.000 Parietal Chord 147 0.06 0.37 1.01 0.16 0.55 0.460 Midfrontal 146 0.07 0.09 1.70 0.16 0.79 0375 Temporal Crest 145 0.25 034 1.54 0.16 9.72 0.002 Bregma 147 0.14 0.23 1.60 0.16 3.02 0.084 Midparietal 146 0.15 0.18 1.65 0.16 3.27 0.073 Anterior Parietal 147 0.16 0.23 1.60 0.16 3.78 0.054 Posterior Parietal 146 0.02 0.02 1.76 0.16 0.06 0.806 Lambda 147 0.06 -0.08 1.85 0.16 0.57 0.450 aStandard error of the estimate.

The Autopsy and Skeletal Samples: Somatic Descriptive Statistics
The general descriptive statistics for the two samples are shown in Table 20. The
autopsy and skeletal samples differed distinctly in mean age, showing a 12-year difference.
This is appropriate to the origins of the two samples, the former being drawn from autopsies
conducted in the public interest, and the latter from subjects used in medical school anatomy

laboratory. Anatomical subjects tend toward higher ages, while subjects in the autopsy
come from both extremes of adult age. Also interesting is the difference in weight between
the two samples, since the autopsy sample is nearly twice as heavy. Again this reflects the
source of the two samples since elderly adults tend to be below average weight, as well as
the tendency towards obesity in recent American adults. Concern over obesity in the
autopsy sample warranted the sampling of the triceps and subscapular skinfold thicknesses








also represented in Table 20. These variables are discussed more thoroughly in the next section.
Table 20. Descriptive Statistics for the Somatic Variables in the Autopsy and Skeletal Samples.
SN Minimum Maximum Mean Standard I I Deviation Autopsy
age (years)47 18 82 46.83 15.19 weight (kg) 47 52.70 184.1 91.07 26.81 body length (cm) 47 152.40 199.40 176.94 10.01 Triceps Skinfold 47 3.0 39.3 15.53 9.78 Subscapular
Skinfold 47 6.0 433 2130 9.63 Skeletal
Age (years) 100 22 87 58.20 13.91 Weight (kg) 100 31.00 96.40 52.21 14.31 Body length (cm) 99 152.50 188.0 170.94 7.97

Quality of Skinfolds
To address the concern that skinfold measurements from the recently deceased would not resemble a living sample, the autopsy sample skinfolds were compared to the trends recorded in Frisancho (1990). Table 21 shows the distribution of fatness categories encompassed by the autopsy sample, based on the fatness categories established by Frisancho (1990: Table IV.39). These categories are based on fatness estimates from summed values of the triceps and subscapular skinfolds. There is no preponderance of specimens falling in the "Lean" or "Below Average" categories, as would be expected if dehydration were adversely affecting skinfold measurements. Conversely, the distribution places nearly half the sample in the above average category, which is in agreement with recent statistics (NHANES III) that show more than half of American adults are obese.
The quality of the skinfold measurements was assessed by comparing the weight classifications derived from the sum of the triceps and subscapular skinfolds to the weight classifications derived from weight for a given stature. The categories are shown below for each autopsy subject (Table 22). A chi-square analysis was conducted to determine if there was any significant difference in the classifications based on the two techniques (Table 23).








Table 21. Autopsy Sample Fatness Categories Based on Frisancho (1990) Table IV.39 for Males using Summed Triceps and Subscapular Skinfold Values

Age Below Above Rangea N Lean Average Average Averageb 18-24 4 2 2 25 2 1 1 30 4 1 1 2 35 3 3 40 9 1 4 4 45 5 1 2 2 50 6 4 2 55 3 1 2 60 6 3 3 65 2 2 70 1 1 75 1 1 80 1 1 Totals: 47 3 2 21 21
a Age ranges resemble Frisancho (1990) except for extending past 72 years. b Above Average has been combined with the category "Excess Fat" which appears in Table IV39.

The returned statistic, although not significant, was very close to 1, indicating that the two

techniques returned nearly identical classifications for the autopsy subjects. Therefore, the

skinfolds collected in the autopsy sample were not significantly altered by postmortem

effects, and were included in subsequent calculations of lean body weight.

Lean body mass was estimated by calculating a fat mass estimate according to the
technique of Frisancho (1990), involving the following steps. First body density was

estimated using equations adapted from Dumin and Womersley (1974) for each skinfold

and the sum of the two skinfolds (Table 24). Densities for the subjects from the autopsy

sample whose ages fell outside the groups of the density equations were calculated with the

nearest age group equation. The percent fat weight for each density was estimated using the

equation % fat weight = [(4.95/Density) - 4.50] *100 (Siri 1956). Each subject's

percentage of fat weight was applied to his body weight to arrive at an estimate of fat weight,
which was then subtracted from the body weight to arrive at the lean body weight. The








descriptive statistics for Percent Fat Weight and the new variable Lean Weight are found in
Table 25.
Table 22. Autopsy Sample Weight Classifications Based on Weight (kg) and Stature (em) and Skinfolds.

Age Weight Stature Weight Sum Triceps and Weight
(kg) (cm) Classification Subscapular Classification Skinfolds
19 68.1 171.5 Average 20.00 Average 24 93.9 179.1 Average (high) 59.00 Above average 28 71.7 177.8 Average 3133 Average 29 953 184.2 Average (low) 10.67 Below average (lean) 30 91.2 182.9 Average 3633 Average 32 154.3 190.5 Above average 40.00 Above average 32 183.8 199.4 Above average 81.33 Above average 36 114.8 1943 Above average 5833 Above average 37 88.5 170.2 Above average 44.67 Above average 37 86.2 172.7 Average high 46.67 Above average 40 75.8 177.8 Average 41.00 Above average 40 59.0 165.1 Below average 26.00 Average 40 87.6 190.5 Average 17.00 Below average 40 903 1753 Average high 47.00 Above average 41 71.7 168.9 Average 3633 Average 41 113.0 189.2 Above average 32.00 Average 43 109.3 175.3 Above average 78.67 Above average 44 74.9 1803 Average 22.00 Average 44 100.7 177.8 Above average 60.33 Above average 45 1293 190.5 Above average 4633 Above average 45 87.6 1803 Average 2733 Average 46 83.5 182.9 Average 11.00 Below average (lean) 46 54.4 162.6 Below average 15.33 Average low 47 77.6 168.9 Average 43.00 Above average 50 100.7 1803 Above average 3333 Average 52 104.8 186.7 Above average 37.67 Average high 54 52.6 160.0 Below average 2933 Average 54 118.4 193.0 Above average 65.67 Above average 54 83.5 177.8 Average 23.50 Average 54 82.6 157.5 Above average 50.33 Above average 55 953 182.9 Average high 3233 Average 56 1143 182.9 Above average 4533 Above average 59 1593 181.6 Above average 72.33 Above average 60 563 154.9 Average low 17.67 Average low 60 95.7 174.0 Above average 44.67 Above average 60 86.2 177.8 Average 50.00 Above average 62 73.5 177.8 Average 22.67 Average 63 93.5 172.7 Above average 50.33 Above average 63 96.2 182.9 Average high 2333 Average 65 79.9 177.8 Average 25.00 Average








Table 22. Continued.


68 83.5 174.0 Average 22.00 Average 74 57.6 162.6 Average low 12.00 Below average 79 80.8 165.1 Average high 49.67 Above average 82 53.5 172.7 Low 22.83 Average 32 79.4 177.8 Average 10.00 Below average (lean) 21 67.6 167.6 Average 21.00 Average 18 95.3 172.7 above average 38.67 Above average
Based on Frisancho (1990) Table IV.13

Table 23. Contingency Table of Weight Classifications based on Weight for Stature and Combined Triceps and Subscapular Skinfold Thicknesses

Below Average Average Above Average Totals

Weight for Staturea 5 24 18 47

Skinfold Thicknessb 6 19 22 47

Totals 11 43 40 94 Pearson Chi-Square 1.046, df= 2, p = 0.593 Based on Frisancho (1990) Figures IV.13a and IV39b respectively.
Table 24. Regression Equations for Estimating Body Density (D) from the Logarithm of Triceps and Subscapular Skinfold Measurements.

Age Groups Males D = a - (b*log triceps skinfold) 17-29 D = 1.1252 - (0.0625 * log T) 20-29 D = 1.1131 - (0.0530 * log T) 30-39 D = 1.0834 - (0.0361 * log T) 40-49 D = 1.1041 - (0.0609 * log T) 50-72 D = 1.1041 - (0.0662 * log T) D = a - (b*log subscapular skinfold) 17-29 D = 1.1312 - (0.0670 * log Sub) 20-29 D = 1.1360 - (0.0700 * log Sub) 30-39 D = 1.0978 - (0.0416 * log Sub) 40-49 D = 1.1246 - (0.0686 * log Sub) 50-72 D = 1.1334 - (0.0760 * log Sub) D = a - (b*log sum of triceps and subscapular skinfolds) 17-29 D = 1.1561 - (0.0711 * log Sum) 20-29 D = 1.1525- (0.0687 * log Sum) 30-39 D = 1.1165 - (0.0484 * log Sum) 40-49 D = 1.1519 - (0.0771 * log Sum) 50-72 D = 1.1527 - (0.0793 * log Sum) Adapted from Frisancho (1990) and Durnin and Womersley (1974).







49
Table 25. Autopsy Sample Descriptive Statistics for Fat Weight Estimates and the New Variable Lean Weight

N Minimum Maximum Mean Std. Deviation % Fat Weight Triceps 47 5.0 44.7 27.20 8.53 % Fat Weight Subscapular 47 10.9 40.0 26.35 7.15 Sum Triceps and Subscapular 47 10.00 8133 36.84 17.89 Skinfolds
% Fat Weight Sum 47 7.5 42.2 25.58 6.96 Average % Fat Weight 47 7.8 40.5 2638 7.03 Fat Weight (kg) 47 7.5 61.0 24.86 11.92 Lean Weight 47 44.77 130.65 67.48 15.97 Skinfold measurements are in mm. Percent fat weight for each skinfold was calculated according to the procedure in Frisancho (1990). % Fat Weight = [(4.95/Density) - 4.50]
*100. Average % Fat Weight = (% Fat Weight Triceps + % Fat Weight Subscapular + % Fat Weight Sum)/13.
An additional assessment was made to determine if the values returned as lean
weight actually reclassified the autopsy subjects into lower weight for height categories than that based on unadjusted weight. Table 26 shows the reclassification of the autopsy sample, with the large chi-square statistic indicating that the two treatments produced significantly different results. Individuals who were already of below average weight were not treated for lean weight, but were passed into the adjusted weight pool. Thirty-one subjects were reclassified as below average weight At least 6 of the 31 were drawn directly from the above average category. Skinfolds tend to underpredict fatness in the extremely obese, so the reclassification process was performing better than expected. The three individuals remaining in the above average category had unadjusted weights of greater than 150 kg. Table 26. Weight Classifications Based on Unadjusted Weight for Height Versus Lean Weight for Height


Below Average Average Above Average Total Unadjusted Weight vs. Stature 5 24 18 47 Lean Weight for Stature 35 9 3 47 Totals 40 33 21 94
Pearson Chi-Square 41.683 df=2 p= 0.000
*Based on Frisancho (1990) Figure IV.13.








Autopsy and Skeletal Samples: Cranial Descriptives

The descriptive statistics for the cranial and clavicular variables used in the two
samples are shown in Tables 27-29. The skeletal sample has a larger number (n =16) of ectocranial variables because these defleshed crania provided greater measurement access. Table 27. Autopsy Sample Descriptive Statistics for the Cranial Variables

Variable Name N Minimum Maximum Mean Standard Deviation Maximum Length 46 174 206 189.87 7.43 Maximum Breadth 47 128 160 142.62 5.67 Minimum Frontal Breadth 45 93 120 102.82 5.63 Biauricular Breadth 47 109 136 12432 6.05 Height from Porion 47 117 150 13230 6.71 Cranial Base Length 37 91 121 104.81 6.90 Facial Height 37 63 96 75.11 8.04 Frontal Chord 47 105 124 115.31 5.28 Parietal Chord 47 105 135 116.49 7.51 Midfrontal 46 3 9 6.60 1.66 Temporal Crest 46 2 9 6.11 1.57 Bregma 47 4 9 6.56 1.13 Midparietal 47 3 11 5.98 1.91 Anterior Parietal 47 3 10 6.20 1.57 Posterior Parietal 46 3 11 6.11 1.76 Lambda 47 3 11 6.84 1.60
Measurements in mm.
The trend of higher means in the autopsy sample is very obvious in these descriptive statistics. Secular changes between the population of the Terry collection and recent forensic cases have been documented (Jantz and Moore-Jansen 1988). Such change is hardly an issue when developing a sample for estimating body mass in fossil primates, especially since at least tens of thousands of years separate the fossil from its only available reference sample. Yet it is interesting to document significant difference between these two samples in order to understand the degree of difference encompassed by the combined sample. The results of the two-tailed t-tests for the cranial and clavicle variables are shown in Table 30, equal variance not assumed. Although the sample sizes are not equal, the t-test is robust to some violations, particularly when conducting a two-tailed test and when sample sizes are large (>30). An F test for equal variances was conducted, but Zar (1998) notes that this test is unreliable if the data deviate from normality. As a precaution the Mann-Whitney







51
U was also calculated because if the data are actually in compliance with the assumptions of
the t-test the Mann-Whitney U performs almost as well.
Table 28. Skeletal Sample Descriptive Statistics of the Cranial Variables

Variable Name N Minimum Maximum Mean Std. Deviation Maximum Length 100 166 207 184.81 7.15 Maximum Breadth 100 127 159 142.63 6.07 Minimum Frontal Breadth 99 86 109 97.24 4.94 Biauricular Breadth 100 109 133 122.80 4.91 Height from Porion 100 111 136 122.61 5.13 Cranial Base Length 99 91 113 100.72 4.17 Facial Height 92 57 86 69.42 6.07 Frontal Chord 100 98 126 112.77 5.12 Parietal Chord 100 99 136 116.05 6.90 Facial Length 100 79 121 93.66 7.84 Cranial Height 99 119 151 134.45 6.02 Bizygomatic Breadth 92 121 140 130.76 4.47 Nasal Breadth 100 20 32 25.00 2.50 Nasal Height 100 45 61 52.19 3.40 External Palate Breadth 100 44 75 58.58 6.03 Mastoid Length 98 20.0 36.0 28.796 2.932 Midfrontal 100 3 12 630 1.77 Temporal Crest 99 2 9 5.27 1.28 Bregma 100 2 10 6.58 1.54 Midparietal 99 2 10 5.55 1.56 Anterior Parietal 100 3 11 5.98 1.50 osterior Parietal 100 3 11 6.45 1.70 Lambda 100 4 16 8.12 2.28
Measurements in mm.

With one exception, the same pattern of significant difference is seen between the
autopsy and skeletal samples as existed between the combined sample and the Howells'
group. In this case the parietal chord is not significantly different between the samples, as it
was in the earlier treatment. These results support secular change as the source of variance
between the combined and Howells' samples, since the same type of difference is seen
within the temporally divided autopsy and skeletal samples. The cranial thickness variables
show significant difference for the temporal crest and lambda sites. Atypically, the skeletal
sample has the larger mean at lambda, deviating from the trend of larger means in the
autopsy sample.








Table 29. Autopsy and Skeletal Samples Descriptive Statistics for Clavicle Variables
Standard
N Minimum Maximum Mean Deviation Autopsy Sample
Clavicle Length 47 119 184 157.57 11.60 AP Diameter 47 8 24 11.93 2.53 SI Diameter 47 9 17 14.04 2.09 Anterior Cortical Thickness 47 1.5 5 3.12 0.86 Posterior Cortical Thickness 46 1 6 3.15 0.91 uperior Cortical Thickness 47 2 5 3.29 0.73 nferior Cortical Thickness 47 2 7 3.52 1.13 Skeletal Sample
lavicle Length 100 127.0 175.0 152.69 938 AP Diameter 100 9 110 13.97 9.85 SI Diameter 100 1 18 10.45 1.85 Anterior Cortical Thickness 95 .5 3.0 1.574 0.610 Posterior Cortical Thickness 100 1 4 2.29 0.59 Superior Cortical Thickness 100 1 4 1.93 0.48 Inferior Cortical Thickness 91 1 3 1.61 0.53 Measurements are in mm.
Table 30. Autopsy and Skeletal Samples Two-Tailed t-test for the Cranial Variables

Variable Name Autopsy Skeletal t Sig. 2- Mean Mann Z Sig. 2Mean Mean tailed Differem Whitney U tailed Maximum Length 189.87 184.81 -3.87 0.000 5.06 1360.50 -3.96 0.000 Maximum Breadth 142.62 142.63 0.013 0.990 0.01 2324.00 -0.11 0.914 Minimum Frontal 102.82 97.24 -5.73 0.000 5.58 1006.00 -5.28 0.000 Breadth
BiauricularBreadth 12432 122.80 -1.50 0.137 1.52 1970.00 -1.58 0.114 Height from Porion 132.30 122.61 -8.78 0.000 9.69 534.00 -7.55 0.000 Cranial Base Length 104.81 100.72 -3.38 0.001 4.09 1087.00 -3.65 0.000 Facial Height 75.11 69.42 -3.88 0.000 5.69 990.00 -3.72 0.000 Frontal Chord 115.31 112.77 -2.74 0.007 2.54 1704.00 -2.69 0.007 Parietal Chord 116.49 116.05 -034 0.735 0.44 2321.00 -0.12 0.904 Midfrontal 6.60 6.30 0.97 0.335 0.29 1967.50 -1.41 0.158 Temporal Crest 6.11 5.27 3.18 0.002 0.84 1468.50 -3.49 0.000 Bregma 6.56 6.58 -0.07 0.943 0.02 2304.00 -0.19 0.847 Midparietal 5.98 5.55 1.34 0.184 0.43 2022.50 -1.28 0.200 Anterior Parietal 6.20 5.98 0.81 0.420 0.22 2199.00 -0.64 0.526 Posterior Parietal 6.11 6.45 -1.08 0.281 0.34 2048.00 -1.07 0.284 Lambda 6.84 8.12 -3.93 0.000 1.28 1546.00 -336 0.001









Autopsy and Skeletal Samples: Relationships Between Variables

As for the combined sample, Pearson's correlation coefficient was calculated to
demonstrate the degree of covariance between body weight and the skeletal measurements

(Tables 31-33). These calculations are for non-transformed data.

Table 31. Autopsy and Skeletal Samples Correlation Coefficients Between Body Weight and the Somatic and Clavicle Measurements
Measurement N r Significance Autopsy Stature cm 47 .727** .000
Lean Weight kg 47 .917** .000 Olavicle Length 47 .199 .181 AP Diameter 47 0.076 .614 SI Diameter 47 0.340* 0.019 Anterior Cortical Thickness 47 0.019 .901 Posterior Cortical Thickness 46 -0.031 .836 Superior Cortical Thickness 47 .218 .141 Inferior Cortical Thickness 47 0.379** 0.009

Skeletal Stature cm 99 .275** .006
Clavicle Length 100 .253* .011 AP Diameter 100 -0.020 .845 SI Diameter 100 .071 .483 Anterior Cortical Thickness 95 0.080 .440 Posterior Cortical Thickness 100 .077 .446 Superior Cortical Thickness 100 .037 .717 Inferior Cortical Thickness 91 0.105 0321
* p < 0.05; ** p < 0.01

The two samples were similar in covariance between body weight and the noncranial variables (Table 31). In the autopsy sample body weight, lean body weight, and
stature were highly correlated. Pearson's r is especially high for the correlation between

body weight and lean weight, which is expected since lean weight is derived directly from
body weight. The clavicle measurements were largely disinterested in body weight, except
for SI diameter and inferior cortical thickness (a component of SI Diameter). While the
skeletal sample does share a significant correlation to stature, the covariance with the
clavicular measurements is only significant for clavicle length. None of the cortical
measurements covary significantly, probably due to the radiolucency of the cortices making
them appear thinner than in the autopsy sample.








There are no high correlations between body weight and the ectocranial
measurements (Table 32) in either sample, although several were significant. Maximum
length, maximum breadth, minimum frontal breadth, height from porion, and cranial base
length all had significant but low r-values in the autopsy sample. Several of the ectocranial
measurements were significantly correlated with body length, but still with low r-values. The
relationships in the skeletal sample between the ectocranial variables and body weight also
yielded low r-values, but fewer significant values than in the autopsy sample. height from
porion, frontal chord, cranial height, cranial base length, bizygomatic breadth, and mastoid

length were all significantly correlated.
Table 32. Autopsy and Skeletal Samples Correlation Coefficients Between Body Weight (kg) and the Ectocranial Variables
Measurement N r Sig. (2-tailed) Autopsy Maximum Length 46 0.490** 0.000
Maximum Breadth 47 0.277 0.060 Minimum Frontal Breadth 45 0.304* 0.043 Biauricular Breadth 47 0.124 0.407 Height from Porion 47 0.445** 0.002 Cranial Base Length 37 0.473** 0.003 Facial Height 37 0.260 0.110 Frontal Chord 47 0.181 0.176 Parietal Chord 47 0.031 0.727

Skeletal Maximum Length 100 0.147 0.145
Maximum Breadth 100 0.179 0.074 Minimum Frontal Breadth 99 0.084 0.406 Biauricular Breadth 100 0.154 0.126 Height from Porion 100 0.252* 0.011 Cranial Base Length 99 0.265** 0.008 Facial Height 92 0.126 0.232 Frontal Chord 100 0.206* 0.040 Parietal Chord 100 0.074 0.463 Facial Length 100 0.082 0.418 Cranial Height 99 0.244* 0.015 Bizygomatic Breadth 92 0.248* 0.017 Nasal Breadth 100 0.036 0.720 SNasal Height 100 0.139 0.169 External Palate Breadth 100 -0.007 0.947 Mastoid Length 98 0.227* 0.025








The correlation results for the skeletal sample were surprisingly different, so an attempt was made to determine the source of the difference by manipulating the sample. The first proposition was that the low body weight in some of the skeletal subjects was affecting the regression analysis. Following Nawrocki's (1992) methodology of setting a weight limit, the sample was sorted by weight and the lighter half removed, leaving 52 subjects with weight greater than or equal to 50 kg. Correlation analysis with the heavier subjects did not yield consistently greater r-values, as indicated by the five variables shown in Table 33. Maximum breadth increased greatly to a significant 031, but two variables lost their previous significance (height from porion and cranial base length). A second manipulation involved removing all the subjects of African descent, since only three (6%) of the 47 autopsy subjects were of African descent, versus 22% in the skeletal sample. This modification produced greater change (Table 33), bringing height from porion and cranial base length back into significant correlation with r-values closer to those seen in the autopsy sample. This phenotypically homogenous sample has higher correlation coefficients than the heavier skeletal sample, but its larger sample size must also be contributing to the improved results. Yet the heavier sample has greater size than the autopsy sample but has fewer significant results.
Table 33. Manipulated Skeletal Sample Correlation Coefficients for Select Variables

Heavier Skeletal Sample European Skeletal Sample Variable n r Sign. n r Sign. Maximum Length 52 0.122 0388 78 0.205 0.072 Maximum Breadth 51 0.307* 0.027 78 0.190 0.096 Minimum Frontal Breadth 51 0.117 0.410 77 0.023 0.845 Biauricular Breadth 51 0.248 0.076 78 0.268* 0.018 Height from Porion 51 0.219 0.119 78 0.302** 0.007 Cranial Base Length 50 0.275 0.051 77 0.318** 0.005
*p < 0.05; ** p < 0.01
Closer examination of the weight distribution of the skeletal sample (Figure 3) shows that most of the sample is of low weight for their height, based on Frisancho's (1990) standards for males. Low weight is the minimum weight category and would be








grounds for medical intervention in a living subject (Frisancho 1990), and in the skeletal sample includes males weighing from 30 to about 50 kg. With the low weight individuals excluded, the correlation coefficients (Table 34) more closely resemble those from the autopsy sample for unadjusted body weight (Table 31). Only cranial base length in Table 33 shows a significant correlation, however, which likely stems from the insufficient sample size (n = 29), a factor that also could have contributed to the correlation coefficients generally being lower here than in the autopsy sample. In contrast, the larger sample size of the combined sample, with its greater proportion of viable body weights, showed more significant correlations to the cranial variables (Table 10).


80




. M


to 4-J

30.


ibmv


aVWIh.


b~w low


Weight Class


Figure 3. Skeletal Sample Weight for Height Class Distribution









Table 34. Skeletal Sample Correlation Coefficients for Body Weight Versus Cranial Variables Excluding Low Weight Subjects

Variable N r Significance Maximum Length 29 0308 0.104 Maximum Breadth 29 0.254 0.184 Minimum Frontal Breadth 29 0.193 0315 Biauricular Breadth 29 0.231 0.227 Height from Porion 29 0.140 0.470 Cranial Base Length 29 0.496** 0.006 Facial Height 26 0.116 0.574 Frontal Chord 29 0.176 0.361 Parietal Chord 29 0.152 0.430 Facial Length 29 -0.084 0.663 Cranial Height 29 0.217 0.258 Bizygomatic Breadth 27 0.135 0.503 Nasal Breadth 29 0.009 0.964 Nasal Height 29 0.187 0.332 External Palate Breadth 29 -0314 0.098 Mastoid Length 28 0.367 0.055 Midfrontal 29 -0.044 0.822 Temporal Crest 29 -0.062 -.750 Bregma 29 0.128 0.509 Midparietal 29 -0.245 0.201 Anterior Parietal 29 -0.068 0.726 Posterior Parietal 29 0.031 0.873 Lambda 29 0.033 0.866
*p < 0.05; ** p < 0.01
In contrast to the ectocranial variables, none of the thickness variables were
significantly correlated to body weight in the autopsy sample, although in the skeletal
sample one of the cranial thickness measurements was significantly correlated to body
weight (Table 35). Lambda has a significant (p < 0.05) r-value of 0.216.
At this point I return to the three average vault thickness categories and their
relationships to the ectocranial variables. The correlation coefficients for the seven
additional variables used in the skeletal sample are shown in Table 36, and once again they
are all positive. Cranial height (basion-bregma distance) has the highest r-value to average
vault thickness (0.40), performing better than height from porion (r = 0.28) in the combined
sample (Table 13). As seen previously in the combined sample the facial measurements
produced only low significant correlation. Bizygomatic breadth is surprising for a breadth








measurement, since in the combined sample these performed poorly. The zygomatic arches bridge the vault near two major components of the masticatory musculature, so the higher rvalue may reflect the pathway of dissolution of masticatory forces. Table 35. Autopsy and Skeletal Samples Correlation Coefficients Between Body Weight (kg) and the Cranial Thickness Variables


Measurement N r Significance Autopsy Midfrontal 46 -0.029 0.847
Temporal Crest 46 0.184 0.221 Bregma 47 0.235 0.112 Midparietal 47 0.215 0.177 Anterior Parietal 47 0.201 0.176 Posterior Parietal 46 0.184 0.222 Lambda 47 0.117 0.435

Skeletal Midfrontal 100 0.066 0.517
Temporal Crest 99 0.105 0.299 Bregma 100 0.192 0.056 Midparietal 99 0.009 0.927 Anterior Parietal 100 0.136 0.177 Posterior Parietal 100 0.091 0365 Lambda 100 0.216* 0.031
* p < 0.05; ** p < 0.01


Table 36. Skeletal Sample Correlation Coefficients for Additional Ectocranial Variables versus Vault Thickness Averages

Anterior Thickness Posterior Thickness Average Vault Measurement Thickness n r Sign. n r Sign. n r Sign. Facial Length 100 0.238* 0.017 100 0.245* 0.014 100 0.268** 0.007 Cranial Height 99 0.352** 0.000 99 0.366** 0.000 99 0.402** 0.000 Bizygomatic Breadth 92 0.345** 0.001 92 0.222* 0.033 92 0.298** 0.004 NasalBreadth 100 0.048 0.633 100 0.137 0.175 100 0.114 0.258 Nasal Height 100 0.248* 0.013 100 0.034 0.740 100 0.130 0.196 External Palate Breadth 100 0.140 0.166 98 0.279** 0.005 100 0.247* 0.013 Mastoid Length 98 0.138 0.174 98 0.058 0.571 98 0.100 0.325


Autopsy and Skeletal Samples: Principal Component Analyses
The PCAs for the autopsy and skeletal sample were conducted with the same

protocol as the combined sample (pairwise handling of missing cases, Varimax rotation).







59
Only one analysis is provided for the autopsy sample (Table 37) because the small sample size (N=47) violates the minimal sample size rule (of thumb) of having three subjects per variable. Table 37 is shown to give an idea of how the variables are behaving although the data are insufficient. Five components are retained, but only two components are interpretable, as in the combined sample. The five components represent 71% of the variance in the data set and have unrotated eigenvalues of 5.45, 2.45, 1.78, 1.45, and 1.07 respectively. Component 1 is a cranial thickness component, with high loadings although not always simple loadings from the parietal sites. Temporal crest has a strong contribution as well. The second component is similar to the same component number in the second matrix for the combined sample (Table 17), as body weight loads with some cranial variables, but not in a simple fashion. Height from porion has high loadings on this factor which it did not in Table 17. If there were more variables this factor might be interpreted as a body size component, but currently it is undefinable.

The greater number of cranial variables in the skeletal sample created the possibility of more informative principal component analyses. The first PCA for the skeletal sample used all the cranial and somatic variables (Table 38) and retained eight components based on the eigenvalue "1" rule. Only the first six components are shown here since the last two are uninterpretable. The eight components represent 66% of the variance in the data set. The unrotated eigenvalues for the six components shown are 7.16, 3.76, 2.43, 2.10, 1.66, and 1.54 respectively.

Component 1 is the cranial thickness factor and is most strongly supported by high loadings from the parietal measurements. The second component is defined by facial measurements (facial length, nasal breadth, and external palate breadth), although the two other variables that, intuitively, should also load here (facial height and nasal height) also contribute to factor seven. It is quite typical for variables to group onto their own vector when they measure the same direction (e.g., the breadth measurements of factor 3), or lie along the same line, as in this case. Facial length could be removed from analysis because it








loads too highly on too many factors, but removing it will not cause nasal height to load onto Component 2. The strongly loading variables, as well as the moderate associations from the other facial measurements, defines Component 2 as a facial factor. Component Three is a cranial width factor and is supported by maximum breadth, biauricular breadth, bizygomatic breadth, and a small association with minimum frontal breadth. This latter variable has only moderate and low loadings throughout the component matrix, and is too well associated with the facial variables factor and the chord measurements of factor five.

Component 4 seems to be the closest approximation of a body size factor. Its
highest loading variable is stature, followed by clavicle length and body weight. These last two variables only have moderate loadings, and the interpretation of the factor is dependent on how little these variables load on to other components. Notice that cranial base length has its highest loading here too, an association we saw earlier in the PCA for the cranial variables in the combined sample (Table 17). As previously mentioned the fifth component Table 37. Autopsy Sample Principal Components for the Cranial Variables

Component

weight kg 0.10571 0.71081 03052, 0.02771 -0.106 Maximum Length 0.32301 0.63841 0.22 0.1103 0.3005 Maximum Breadth 0.18941 0.29871 0.1640 0.7527 0.0784 Minimum Frontal Breadth 0.00621 0.34181 0.59871 0.4852' 0.0379" Biauricular Breadth 0.08361 -0.0861 0.0833( 0.8243 0.0116 Height from Porion 0.03431 0.86471 -0.061( 0.0220 0.1465 Cranial Base Length 0.16071 032751 0.7484q 0.1564 -0.03.5 Facial Height 0.06224 0.00381 0.94164 0.0542 0.1100 Frontal Chord 0.2829 0.44411 -0.547q 0.4167 0.1119 Parietal Chord -0.005 0.1614 0.04744 0.03131 0.94731 Midfrontal 0.5167 -0.14q 0.0197 0.5299 -0.041 Temporal Crest 0.7392 -0.1054 0.2101 -0.046 0.1009! Bregma 0.6806 0.1412q -0.08 0.286r 0.1901 Midparietal 0.7186 0.2708 0.2010 0.1113E -0.2233 Anterior Parietal 0.6615 0.24241 -0.07- 0.5116 -0.2462 Inferior Parietal 0.88611 0.16531 0.0787 0.0770 0.039 Lambda 0.70761 0.10611 -0.11lt 0.11731 -0.0211











Table 38. Skeletal Sample Principal Components Using AH Variables (Except Age)

Component
1 2 3 4 5 Maximum Length 0.3465 0.2947 0.0625 0.35872 0.50521 0.00511 Maximum Breadth -0.0181 -0.2 0.78471 -0.1421 0.28161 -0.0652 Minimum Frontal Breadth 0.13251 0.33181 0.53553 0.00931 0.38133 -0.0743 Biauricular Breadth 0.14872 -0.0704 0.86954 0.07845 -0.09 -0.1545 Heightfrom Porion 0.32883 0.04141 0.25771 0.3224 0.64054 0.00681 CranialBaseLength 0.21831 0.29184 0.06851 0.52207 0.11497 0.13052 Facial Height 0.10897 0.52231 0.04264 0.0975- 0.1386 0.13621 Frontal Chord 0.16281 0.07521 0.16064 0.1571 0.6318 -0.0102 Parietal Chord 0.36611 0.1591 0.11471 0.24381 0.5815 -0.1222 Facial Length 0.16981 0.81251 -0.1162 0.1658 0.02171 0.16923 Cranial Height 0.30741 -0.3094 0.21014 0.35752 0.46224 -0.0007 BizygomaticBreadth 0.1977 0.2332t 0.81012 0.2244 -0.1131 -0.1383 Nasal Breadth 0.01941 0.74211 0.04471 0.228 0.05255 0.0052 Nasal Height 0.0269. -0.0841 0.1061 0.22184 -0.0031 -0.077 External Palate Breadth 0.15644 0.8564 0.10227 -0.066t 0.07385 0.066 MastoidLength -0.085S 0.32434 0.10139 0.41653 0.14821 0.0534 Midfrontal 0.5840E -0.0521 0.01171 0.23331 0.12644 -0.074 Temporal Crest 0.6815S 0.05204 0.234 0.15005 -0.1611 -0.119 Bregma 0.67951 0.04987 -0.0037 0.3755 0.11471 -0.115 Midparietal 0.81201 0.06671 0.1398- -0.2171 0.07497 0.0136 Anterior Parietal 0.7822 0.00141 0.03801 -0.011 0.21624 0.0551 PosteriorParieal 0.78501 0.11421 0.07901 -0.1411 0.20511 0.1896 Lambda 0.64871 0.2091 0.0407 0.2415I 0.07102 0.1384 Clavicle Length 0.00152 0.24181 -0.0282 0.67481 0.01421 0.0255 APDiameter -0.0231 -0.0097 -0.1521 -0.1111 0.53124 0.0625 SIDiameter 0.12293 0.02732 0.03321 0.08811 -0.0324 -0.178 Anterior Cortical Thickness 0.03274 -0.0177 0.00498 0.03514 0.15681 0.7480 PosteriorCortical Thickness -0.1261 0.41291 -0.3023 0.17047 -0.174 0.5227 Superior CorticalThickness -0.036( 0.09047 -0.054 0.11174 -0.065 0.8061 InferiorCortical Thickness 0.12947 0.0791SI -0.1758 0.06534 -0.019 0.7045 Weightkg 0.0563 -0.084 0.28141 0.5248 -0.004 0.1947 Stature 0.12351 0.03771 -0.0949 0.7259 0.165 0.1113








is associated with chord measurements of the vault in the sagittal plane. The loadings are only moderate but tend to be the highest loadings for these variables in this matrix, so the definition is clear. The last definable component has high loadings from three of the clavicle measurements and is therefore a clavicle cortical thickness factor.

In summary, this PCA using all variables from the skeletal sample yielded six

definable components. These components were retained based on the eigenvalue "1" rule, definability, and association with at least three variables of either high or simply structured loadings. The components are:

1. Cranial Thickness

2. Facial Variables

3. Cranial Breadth

4. Body Size

5. Sagittal Chords of the Vault

6. Clavicle Cortical Thickness

Components seven and eight are not defined because they both lack at least three high loading variables. An association between nasal height and facial height, two measurements which occur on the same line on the skull, is found in component seven. Lack of a third variable omits this factor from display. Factor eight has high loadings only from SI diameter of the clavicle.

The second PCA excluded the clavicle variables (Table 39). Five components were interpretable (six passed the eigenvalue test). Unrotated eigenvalues are 6.97, 2.79, 2.26, 1.99, 1.54, and 1.28, respectively. The cumulative variance in the six components is 67%. In response to the reduction in variables many loadings have increased but the associations and interpretations of the components have not changed. The defined components are:

1. Cranial Thickness
2. Facial Variables

3. Sagittal Chords of the Vault








4. Cranial Breadth
5. Body Size

Autopsy Sample: Regression Analysis
Least Square regression lines were fit for the relationships between the cranial measurement variables and body mass. Regressions for the autopsy sample were conducted for unadjusted and adjusted (lean) body weight All results are for logarithmically transformed
(base 10) variables so the r-values will not resemble those shown previously. Table 39. Skeletal Sample Principal Components Using the Cranial Variables

Component

Maximum Length 0.29001 0.29133 0.61703 1.1E-0. 0.2316 0.3804S Maximum Breadth -0.0414 -0.2364 0.3716 0.74401 -0.100 0.12441 Minimum Frontal Breadth 0.10411 0.2932 0.4701 0.50381 -0.0111 -0.087 Biauricular Breadth 0.1327- -0.090 0.0225 0.8957 0.1135 0.0607 Height from Porion 0.2918 0.0227 0.708 0.1620 0.3760q -0.0211 Cranial Base Length 0.215 0.28921 0.1735 0.04467 0.61961 0.2436 Facial Height 0.0743 0.5511 0.18573 0.01827 0.0287 0.6651 Frontal Chord 0.11571 0.04104 0.70165 0.08371 0.18939 0.1992 Parietal Chord 0.312q 0.16247 0.6602 0.0503 0.00681 -0.011q Facial Length 0.16701 0.83547 0.0320 -0.124 0.20731 0.1674q Cranial Height 0.29714 -0.3441 0.50361 0.1498 0.50 -0.09 Bizygomatic Breadth 0.17832 0.23963 -0.0234 0.8467 0.19101 0.04162 Nasal Breadth 0.0072 0.74457 0.09241 0.0342 0.2266q -030 Nasal Height -0.04 -0.0832 0.04721 0.11673 0.21185 0.82913 External Palate Breadth 0.14911 0.8543q 0.11155 0.10367 -0.096 0.0916 Mastoid Length -0.104q 0.30232 0.24443 0.0675 0.4830q 0.11285 Midfrontal 0.5491 -0.051 0.18731 0.0 0.1695 0.4499 Temporal Crest 0.6643 0.06204 -0.070 0.2711 0.1557 0.2993 Bregma 0.6631 0.0752. 0.10512 0.02163 0.3111 0.0170 Midparietal 0.8158 0.03014 0.15953 0.144 -0.183 0.0095 Anterior Parietal 0.7797 -0.0127 0.26892 0.01564 0.03382 0.0411 Posterior Parietal 0.7958 0.10904 0.22612 0.0478 -0.04 -0.0157 Lambda 0.637 0.24435 0.1169 -0.0031 0.2387 -0.1299 Weight kg 0.0597 -0.0313 -0.0031 0.21731 0.6126 -0.0066 Stature 0.10586 0.08268 0.21868 -0.1642 0.70218 0.2002








As previously seen in the combined sample, all r-values are relatively low. The

highest r for the autopsy sample (Table 38) is 0.55 for cranial base length versus body
weight maximum length is second with a value of 0.51. The regression results using lean

weight show weaker relationships (Table 39). The r-value for cranial base length drops to

0.48 and maximum length to 0.46. Height from porion has a large change from 0.49 to
032. Even with the log transformation to correct for lack of normality in the sample, the

correlations between cranial variables and body weight are not as high as indicated in the
literature (Aiello and Wood 1994; Gauld 1992). With such weak relationships between the
variables the use of the line fitting techniques is merely a formality.

Table 40. Autopsy Sample Regression coefficients for Body Weight Versus Cranial Variables

Variable N r Slope Intercept SEE F Sign. Maximum Length 46 0.51 3.48 -5.98 0.10 15.19 0.000 Maximum Breadth 47 0.26 1.81 -1.95 0.12 335 0.074 Minimum Frontal Breadth 45 0.35 1.79 -1.66 0.11 5.96 0.019 Biauricular Breadth 47 0.11 0.62 0.65 0.12 0.56 0.457 Height from Porion 47 0.49 2.69 -3.75 0.10 14.30 0.000 Cranial Base Length 37 0.55 2.30 -2.71 0.10 15.53 0.000 Facial Height 37 0.30 0.79 0.49 0.11 3.56 0.068 Frontal Chord 47 0.17 1.03 -0.17 0.12 1.39 0.244 Parietal Chord 47 0.06 0.24 1.45 0.12 0.14 0.713 Midfrontal 46 0.04 0.04 1.98 0.12 0.08 0.776 Temporal Crest 46 0.15 0.13 1.84 0.12 0.96 0.333 Bregma 47 0.19 0.29 1.71 0.12 1.68 0.201 Midparietal 47 0.21 0.18 1.81 0.12 2.10 0.155 Anterior Parietal 47 0.16 0.17 1.81 0.12 1.22 0.275 Posterior Parietal 46 0.18 0.17 1.82 0.12 1.42 0.240 Lambda 47 0.14 0.15 1.82 0.12 0.87 0357


Summary
To investigate the proposal that in modem humans cranial size scales in proportion to
body size, and that cranial vault thickness scales in proportion to systemic skeletal
robusticity, five hypotheses were tested:
* HI: body weight in modern humans covaries with cranial measurements








* H2: estimates of lean body weight will have stronger covariance relationships with

cranial measurements than gross body weight will have with cranial measurements
* H3: vault thickness is a function of general skeletal robusticity as indicated by cortical

thickness of long bones
* H4: an index of vault length and height covaries with vault thickness
* HS: weight of the brain significantly covaries with vault thickness

The results of this study support two and reject three of these hypotheses.
Table 41. Autopsy Sample Regression coefficients for Lean Weight Versus Cranial Variables

Variable N r Slope Intercept SEE F Sign. Maximum Length 46 0.46 2.46 -3.79 0.08 12.03 0.001 Maximum Breadth 47 0.20 1.06 -0.47 0.09 1.96 0.169 Minimum Frontal Breadth 45 0.32 1.26 -0.72 0.09 4.94 0.031 Biauricular Breadth 47 0.17 0.70 035 0.09 1.29 0.262 Height from Porion 47 032 1.33 -1.01 0.09 5.17 0.028 Cranial Base Length 37 0.48 1.52 -1.26 0.08 10.33 0.003 Facial Height 37 032 0.66 0.61 0.09 4.11 0.050 Frontal Chord 47 0.23 1.03 -0.30 0.09 2.47 0.123 Parietal Chord 47 0.04 -0.11 2.05 0.09 0.05 0.817 Midfrontal 46 0.07 0.05 1.78 0.09 0.22 0.642 Temporal Crest 46 0.26 0.18 1.68 0.09 3.20 0.081 Bregma 47 0.24 0.28 1.60 0.08 2.81 0.100 Midparietal 47 0.29 0.18 1.68 0.09 3.96 0.053 Anterior Parietal 47 0.19 0.16 1.70 0.09 1.70 0.199 Posterior Parietal 46 0.24 0.17 1.69 0.09 2.60 0.114 Lambda 47 0.25 0.21 1.65 0.09 3.02 0.089

Hypothesis One is certainly supported as indicated by the correlation analyses of
the combined sample and its subsets. Of the sixteen cranial measurements in the combined

sample, ten were significantly correlated with body weight More cranial variables showed
significant correlations to body weight in this sample than in the autopsy or skeletal,
because the larger sample size supported the weaker relationships. The strongest
relationships were for height from porion and cranial base length, the former having the
greatest r-value for this study, 0.628. This is still below my minimum desirable correlation
coefficient, of 0.69 for circumference of the radius.








Hypothesis Two is rejected, since the estimate of lean body weight was not better corrrelated to the cranial measurements. The skinfolds did produce meaningful reductions of the gross body weights, but the new variable had lower r-values, particularly for height from porion and cranial base length.

Although the clavicular measurements were well correlated to body weight in the combined sample, they were not also well correlated to vault thickness. These data do not support the idea that clavicular cortical thickness and vault thickness are both the product of the same systemic skeletal growth factors. Hypothesis Three is rejected.

The cranial index did covary significantly with cranial thickness, thus supporting
Hypothesis Four. Approaching equality between cranial breadth and length was associated with decrease in vault thickness at bregma and on the parietal.

The final hypothesis was not supported. Brain weights in the autopsy sample

showed only one significant correlation to vault thickness, at the anterior parietal location. No pattern of association is established.














DISCUSSION


At the inception of this study I proposed that the human cranium is scaled in proportion to the entire body. From this foundation I tested five working hypotheses. Those tests indicated that cranial size is, at best, only moderately related to body size, while vault thickness is, for practical purposes, completely unrelated to body weight. The estimate of lean body mass, although carefully executed and seemingly providing reasonable adjustments of subjects' weight classes to lower levels, was not better correlated with the cranial variables. Vault thickness was not well correlated to clavicular cortical thickness or brain weight. There were, however, interesting correlations between the Cranial Index and individual vault thickness measurements, although average vault thickness did not covary significantly with the Cranial Index or its categories.

These results initially appear to be very negative, when actually, at least for the first hypothesis, they affirm much of the previous literature. Pensler and McCarthy, Nawrocki, and Hartwig-Scherer and Martin all found low or insignificant correlations between cranial measurements and body weight in modern humans. The methodology in this study resembles those studies in the use of measured body weights, not weights drawn from the literature or predicted through other means. Another point of similarity is the use of modern humans as a separate (or sole) sample for analysis. These similarities are no surprise as I designed the methodology to have these elements, but my results confirm that many cranial measurements, especially vault thickness measurements, are poorly correlated to body weight. Critically, one could assume Nawrocki and Hartwig-Scherer had spurious results due to small sample sizes (n <30 and n = 19), respectively), and Gauld suggested








that Pensler and McCarthy's results were due to "increased environmental factors". However, my data indicate that neither is the cause of their results.

My results illustrate how small sample sizes obscure relationships. Sample size has had an obvious effect in this study on the appearance of significant correlations, even to the point that a sample nearing 50 is too small. The low weights in the skeletal sample obscured many significant correlations, but a sample excluding all very low weight individuals (as occurs in methodologies biased against the emaciated) had too few individuals to reveal the weaker relationships. Using a weight limit with no consideration of height allowed a larger sample which produced results resembling those found in the Autopsy sample, including additional correlations not seen in the smaller data set.

The primary divergence in methodology between my study and those with

conflicting results (Aiello and Wood, Gauld) is my focus on a strictly human sample, use of measured, not predicted or averaged, body weights, and avoidance of species means. The error of investigating relationships between variables when one lacks measurements of the variables is obvious from the position of empiricism-to observe the relationship, observations are needed. Yet when studying primates the only species in generous supply tends to be H. sapiens, so researchers must make do with what is available. Aiello and Wood were constrained for body weights for their nonhuman sample, which is understandable, but Gauld lacked measured body weights even for the humans in her study.

Departing from the data deficit issue, the multi-species data set for predicting body weight also presents a problem. The convenience of the multi-species sample is that it lowers the standard errors of the equations produced, but as Smith (1985) indicates, the primary concern of body weight prediction is not precision, it is accuracy. To improve accuracy, to arrive at an estimate that may actually have been the specimen's functional, living body weight, one must design a sample with specimens resembling the focal animal. Researchers Gauld (1996) and Aiello and Wood (1994) shared the goal of predicting body weight in fossil hominids. Although they used different cranial variables, their strategies







69
were the same and to some degree included the same data since they both drew nonhuman body weights from Harvey et al. (1987). Their samples may be quite effective for the smaller-brained hominids, such as the Australopithecines and possibly the habilines, but for the larger-brained species modern humans must be the best reference sample for accuracy. In their study of body weight prediction of extant and extinct Cercopithecidae, Delson et al. (2000) suggested that the results of Aiello and Wood were in part due to the extremes of body size in their primate taxa. Best fit lines for data sets with "mouse to elephant" weight ranges tend to be anchored by the extreme points rather than representing the variation in the middle. While this explanation may address Aiello and Wood's cranial results, it does not explain the overall difference in performance between their cranial and postcranial variables. Delson et al. (2000) found that postcranial variables performed better than dental or cranial variables in weight prediction in cercopithecids.
The multi-species data set is unacceptable when the measurements are focused on areas that are specialized for one, but not the majority of, species in the group. I suggest that their analysis may be valid for nonhuman apes, but the divergent evolution of humans in expansion of the vault and reduction of the face makes cranial measurements unsuitable predictive tools. How well can a collection dominated by small-brained apes represent the genetic and epigenetic forces resulting in the morphology of their large-brained relatives? It would be comparable to measuring five different kinds of car tires, adding Mack truck tires as the sixth group and predicting other Mack tires from the resulting formula. Aiello and Wood's postcranial measurements certainly reflect the effect of divergent evolution within their sample, as the femoral measurements produced weaker correlations than the humeral measurements because humans are specialized for bipedality, while changes in the humerus are not so derived.
Yet why are the same weak correlations not found in Aiello and Wood's (1994) cranial measurements? I propose two causes. The humans are outliers, just not as markedly so as for the leg measurements. More importantly, their nonhuman species







70
means in the postcranial analysis were based on even fewer individuals, frequently as few as three per mean. The nonhuman means represented even less diversity than they did in the cranial analysis, so the much greater number of humans (12 male/12 female) would have better differentiated the human species mean relative to the nonhumans. The authors did not publish the regression plot for their cranial or postcranial data (and Smith (2000) cautions against using regression formulae not showing such a plot), so we do not know where humans fall in relation to the other hominoids. I suspect, considering how my low r-value for logo0 transformed biauricular breadth contrasts with their biporionic breadth (r = 0.18 and r = 0.98, respectively) that humans did not converge with the other apes.

More support against strong associations between cranial measurements and body weight comes from the principal component analyses. Analysis of the Combined Sample and its subsets did not produce consistent associations between cranial variables and the components highly loaded for body weight or stature. The body size component in the Autopsy sample had high loadings from Body Weight and Height from porion, and a moderate contribution from Maximum length (Table 37). Independently the cranial variables show non-transformed r-values of 0.45 and 0.49, respectively to Body Weight. The Combined Sample shows different relationships, as Body Weight loads with Cranial base length and facial height on the second component (Table 18). Again, these cranial variables do not have high r-values (r = 0.49 and 038 respectively). Another interesting aspect of the PCA is that the vault thickness variables always loaded together on a single component, and never in association with body size.

Lean body weight was not better correlated with body weight. Either the quantity generated is unrelated to lean body mass due to errors in the prediction method, or the physiological relationship is not based on lean mass. Estimates of total body fat from skinfolds can be unreliable, particularly tending towards underestimation in very obese subjects (Shepherd 1991). An additional source of error is the use of equations for the body density estimate that were not developed for the population under study. However,







71
Durnin and Womersley's equations were developed from a European population. This type of error should not apply unless the level of specificity is very precise. The correlations in the skeletal sample indicate that leaner bodies produce lower significant correlation coefficients.

Pearson (2000) found weak correlations between robusticity in the clavicle and

robusticity in the humerus, but some of the commentators attributed this to size difference between the clavicle and the humerus. Pearson also used a robusticity index, while I compared gross thickness without an attempt at standardization. Dividing the thickness measurements by their respective lengths (chord lengths for the cranial thicknesses and clavicle length for the cortical measurements) might have produced more interesting results. Currently I also find only weak and mostly insignificant associations between the two types of bone thickness.

Moss and Young's (1961) work suggests the better methodology is to treat the
tables of the vault as independent functional units. They demonstrated that the inner table follows the contours of the brain and dura even after neural atrophy has reduced the cranial contents. With each layer serving specific functions in conjunction with their function as a unified structure, the best approach to analyzing vault thickness would be to measure the layers separately and together.

Regarding the brain weight results, I suspect brain weight is not the best indicatory of size indicator. I was proposing, based on work showing that brain and dura growth defines and restricts vault growth (Moss and Young 1961, Huggare and Ronning 1995), that vault thickness would be related to brain size. A space-filling size indicator, such as volume, may have been more appropriate, since it is expansion of the contents that directs the growth of the vault

My results support Nawrocki's conclusion that vault thickness increases with head size (Table 13). All of my significantly correlated ectocranial variables had positive relationships to vault thickness. The Cranial Index results, while informative, lack a large








sample of brachycranic individuals. There were insufficient significant correlations, but at least all the correlations for the brachycranic class were negative. The results suggest that the wider and shorter the vault gets, the thinner the vault gets in some locations.

Contrary to the literature, many cranial measurements are not suitable as predictors of body weight in modem humans, and by extension for large-brained hominids. Gross vault thickness measurements are particularly poor tools and should not be used. Cranial measurements may be effective for small-brained extant and extinct hominoids, as Aiello and Wood did have a large correlation coefficient for their primarily nonhuman ape sample. Body weight predictions for larger-brained fossil hominids based on cranial measurements must be considered suspect because, at best, the formula predicts morphology for a smallbrained hominid. It must also be considered that there are many other cranial measurements that were not used in this study, so a useful relationship may exist that is as yet undescribed. Clear associations do exist between vault length and height and vault thickness, and there is both theoretical and empirical support for rounder skulls having relatively thinner vaults. Although this study does not identify how skeletal robusticity contributes to vault thickness, it suggests a methodological approach that might be effective for future research.













REFERENCES


Adeloye A, Kattan KR and Silverman FN (1975) Thickness of the normal skull in the American blacks and whites. Am J Phys Anthropol 43: 23-30.

Aiello LC and BA Wood (1994) Cranial variables as predictors of hominine body mass. Am J Phys Anthropol 95:409-426.

Andrews P (1984) An alternative interpretation of the characters used to define Homo erectus. Cour Forsch Inst Senckenberg 69:167-175.

Ant6n SC (1989) Intentional cranial vault deformation and indirect changes of the cranial base and face. Am J Phys Anthropol 79:253-267.

Ant6n, SC (1997) Endocranial hyperostosis in Sangiran-2, Gibraltar-1, and Shanidar-5. Am J Phys Anthropol 102:111-122.

Ant6n SC and Franzen JL (1997) The occipital torus and developmental age of Sangiran-3. J Hum Evol 33:599-610.

Baker, PT and Newman, RW (1957) The use of bone weight for human identification. Am J Phys Anthropol 15:601-618.
Bass WM (1995) Human Osteology: A Laboratory and Field Manual. Columbia, MO:Missouri Archaeological Society.
Brown, P (1987a) Cranial vault thickness in Northern Chinese, European and Australian Aboriginal populations. English summary. Act Anthropol Sin 6:184-189.

Brown P (1987b) Pleistocene homogeneity and Holocene size reduction: the Australian human skeletal evidence. Archaeology in Oceania 22:41-71.

Brown P (1994) Cranial vault thickness in Asian Homo erectus and Homo sapiens. Cour Forsch Senckenberg 171:33-46.

Brown T, Pinkerton SK, and Lambert W (1979) Thickness of the cranial vault in Australian aboriginals. Archaeol Phys Anthropol Oceania 14:54-71.

Brunet M, Guy F, Pilbeam D, Mackaye H, Likius A, Ahounta D, Beauvilain A et al. (2002) A new hominid from the Upper Miocene of Chad, Central Africa. Nature 418:145-151.

Clarke R (1985) A new reconstruction of the Florisbad cranium with notes on the site. In E Delson (ed.): Ancestors: The Hard Evidence, pp 301-305. New York, NY: Academic Press.

Conroy, GC (1987) Problems of body-weight estimation in fossil primates. Int J Primatol 8(2):115-137.








Corruccini RS and Henderson AM (1978) Multivariate dental allometry in primates. Am. J Phys Anthropol 48:203-208.

Currey J (1984) The mechanical adaptations of bones. Princeton, NJ: Princeton University Press.
Dagosto M and Terranova CJ (1992) Estimating the body size of Eocene primates: a comparison of results from dental and postcranial variables. Int J Primatol 13(3):3(07-344.
Delson E, Terranova CJ, Jungers W1, Sargis EJ, Jablonski NG, and Dechow PC (2000). Body Mass in Cercopithecidae (Primates, Mammalia): Estimation and Scaling in Extinct and Extant Taxa. Anthropological Papers of the American Museum of Natural History No. 83.

Demes B (1985) Biomechanics of the primate skull base. F Beck, HW Hild, GW Kriz, HR Ortmann, KJ Pauly, TH Schiebler (eds.): Advances in Anatomy Embryology and Cell Biology, vol. 94. New York: Springer-Verlag, pp 1-52.
Durnin JV and Womersley J (1974) Body fat assessed from total body density and its estimation from skinfold thickness: Measurements on 481 men and women aged from 16 to 72 years. Br J Nutr 32:77-96.
Felsing NE, Brasel JA and Cooper DM (1992) Effect of low and high intensity exercise on circulating growth hormone in men. J Clin Endocrinol Metab 75: 157-162.

Flegal KM, Carroll MD, Kuczmarski RJ, and Johnson CL (1998) Overweight and obesity in the United States: Prevalence and trends, 1960-1994. Int J Obesity 22:39-47.
Frisancho AR (1990) Anthropometric Standards for the Assessment of Growth and Nutritional Status. Ann Arbor. University of Michigan Press.

Gauld SC (1992) An allometric study of anthropoid cranial bone thickness: Implications for body size estimation in early hominid species. PhD Dissertation, Department of Anthropology, University of California, Los Angeles.

Gauld SC (1996) Allometric patterns of cranial bone thickness in fossil hominids. Am J Phys Anthropol 100:411-426.

Getz B (1960) Skull thickness in the frontal and parietal regions. Acta Morphologica Neerlando-Scandinavica 3:221-228.
Giles E and Elliot O (1962) Race identification from cranial measurements. J For Sci 7: 147-167.
Giles E and Elliot O (1963) Sex determination by discriminant function analysis of crania. Am J Phys Anthropol 21:53-68.
Gingerich PD (1977) Correlation of tooth size and body size in living hominoid primates, with a note on relative brainsize in Aegyptopithecus and Proconsul. Am. J Phys Anthropol 47:395-398.
Gingerich PD, Smith BH and Rosenberg K (1982) Allometric scaling in the dentition of primates and prediction of body weight from tooth size fossils. Am J Phys Anthropol 58: 81-100.









Hartwig-Scherer S (1994) Body weight prediction in fossil Homo. Courier ForschungsInstitut Senckenberg 171:267-279.
Hartwig-Scherer S and Martin RD (1992) Allometry and prediction in hominoids: a solution to the problem of intervening variables. Am J Phys Anthropol 88:37-57.
Harvey PH, Martin RD and Clutton-Brock Th (1987) Life histories in comparative perspective. In BB Smuts, DL Cheney, RM Seyfarth, RW Wrangham and 'Tl Struhsaker (eds.): Primate Societies. Chicago: University of Chicago Press, pp 181-196.
Hatcher L and Stepanski EJ (1994) A Step-by-Step Approach to Using the SAS System for Univariate and Multivariate Statistics. Cary, NC: SAS Institute Inc.
Hershkovitz I, Greenwald C, Rothschild MM, Latimer B, Dutour O, Jellema LM, and WishBaratz S (1999) Hyperostosis frontalis interna: An anthropological perspective. Am J Phys Anthropol 109:303-325.
Hinton MAC, Oakley KP, Dines HG, King WBIR, Kennard AS, Hawkes CFC, Warren SH, Cotton MA, Le Gros Clark WE, and Morant GM (1938) Report on the Swanscombe skull. J Roy Anthropol Inst Great Britain Ireland 68:17-98.
Holloway RL (1980) Within-species brain-body weight variability: a reexamination of the Danish data and other primate species. Am J Phys Anthropol 53:109-121.
Howells WW (1973) Cranial Variation in Man. Papers of the Peabody Museum No. 67. Cambridge, MA: Harvard University.
Hublin J (1986) Some comments on the diagnostic features of Homo erectus. Anthropos (Brno) 23:175-187.
Huggare J and Ronning O (1995) Growth of the cranial vault: Influence of intracranial and extracranial pressures. Acta Odontol Scand 53:192-195.
Ishida H and Dodo Y (1990) Cranial thickness of modern and Neolithic populations in Japan. Hum Bio 62(3):389-401.
Israel H (1973) Age factor and the pattern of change in craniofacial structures. Am J Phys Anthropol 39:111-128.
Israel H (1977) The dichotomous pattern of cranialfacial expansion during aging. Am J Phys Anthropol 47: 47-52.
Ivanhoe F (1979) Direct correlation of human skull vault thickness with geomagnetic intensity in some northern hemisphere populations. J Hum Evol 8:433-44.
Jacob T (1972) Palaeoanthropological discoveries in Indonesia with special reference to the finds of the last two decades. J Hum Evol 2:473-485.
Jantz RL and Moore-Jansen PH (1988) A Data Base for Forensic Anthropology: Structure, Content and Analysis. Report of Investigations No. 47. Knoxville, TN: The University of Tennessee Department of Anthropology.







76
Jungers WL (1984) Aspects of size and scaling in primate biology with special reference to the locomotor skeleton. Yrbk Phys Anthropol 27:73-97. Jungers WL (1988) Relative joint size and hominoid locomotor adaptations with implications for the evolution of hominid bipedalism. J Hum Evol. 17:247-265. Kaito Y (1980) On the thickness of the Japanese skull. Tokyo Jikeikai Medical Journal 95:503-513.
Kay RF (1975) The functional adaptations of primate molar teeth. Am J Phys Anthropol 43:195-216.
Kennedy GE (1985) Bone thickness in Homo erectus. J Hum Evol 14:699-708. Kennedy GE (1991) On the autapomorphic traits of Homo erectus. J Hum Evol 20- 375412.
Larsen CS, Matter RM, and Gebo DL (1998). Human Origins: The Fossil Record. Prospect Heights, IL: Waveland Press

Lassarre C, Girad F, Durand J, and Raynaud J (1974). Kinetics of human growth hormone during submaximal exercise. J App Physiol 37(6):826-830. Lieberman DE (1996) How and why humans grow a thin skull: experimental evidence for systemic cortical robusticity. Am J Phys Anthropol 101: 217-236. Lindstedt SL and Calder WA (1981) Body size, physiological time, and longevity of homeothermic animals. Quar Rev Bio 56(1):1-16. Loth SR and Henneberg M (1996) Mandibular ramus flexure: A new morphologic indicator of sexual dimorphism in the human skeleton. Am J Phys Anthropol 99:473-485. Martin, AD, Ross WD, Drinkwater DT, and Clarys JP (1985) Prediction of Body Fat by Skinfold Caliper- Assumptions and Cadaver Evidence. Int J Obesity 9:31-39 Suppl 1. Martin, RA (1981) On extinct hominid population densities. J Hum Evol 10:.427-428. Martin R and Sailer K (1957) Lehrbuch der Anthropologie Vol. 1. Stuttgart: Gustav Fischer Verlag.
McHenry HM (1976) Early hominid body weight and encephalization. Am J Phys Anthropol 45: 77-84.
McHenry HM (1984) Relative cheek-tooth size in Australopithecus. Am J Phys Anthropol 45:297-306.
McHenry HM (1988) New estimates of body weight in early hominids and their significance to encephalization and megadontia. In FE Grine (ed.): Evolutionary History of the "Robust" Australopithecines. New York: Aldine de Gruyter, pp 133-148. McHenry HM (1991) Petite bodies of the "robust" Australopithecines. Am J Phys Anthropol 86: 445-454.








McNab BK (2002) The Physiological Ecology of Vertebrates. A View from Energetics. Ithaca: Cornell University Press.

Moss ML and Young RW (1961) A functional approach to craniology. Am J Phys Anthropol 18: 281-292.
Nawrocki SP (1992) A Biomechanical Model of Cranial Vault Thickness in Archaic Homo. Ann Arbor: University of Michigan Press.
Ohtsuki F (1977) Developmental changes in the cranial bone thickness in the human fetal period. Am. J Phys Anthropol 46:141-154.

Ortner DJ and Putschar WGJ (1981) Identification of Pathological Conditions in Human Skeletal Remains. Smithsonian Contributions to Anthropology 28. Washington, DC: Smithsonian Institution Press.

Pearson OM (2000) Activity, climate, and postcranial robusticity. Current Anthropol 41:569-607.

Pensler J and McCarthy JG (1985) The calvarial donor site: An anatomic study in cadavers. Plast Recon Surg 75: 648-651.
Pirie PL (1978) Allometric scaling in the postcanine dentition with reference to primate diets. Primates 19(3):583-591.

Richards G and Ant6n SC (1991) Craniofacial configuration and postcranial development of a hydrocephalic child (ca. 2500 BC to 500 AD): with a review of cases and comments on diagnostic criteria. Am J Phys Anthropol 85:185-200.
Rightmire GP (1986) Body size and encephalization in Homo erectus. Anthropos (Brno) 23:139-150.
Roche, AF (1953) Increase in cranial thickness during growth. Hum Bio 25: 81-92.

Ross AH, Jantz RL, and McCormick WF (1998) Cranial thickness in American females and males. J For Sci 43(2):26-272.
Ruff CB (2000) Body size, body shape, and long bone strength in modern humans. J Hum Evol 38:269-290.

Shea BT (1983) Size and diet in the evolution of African ape craniodental form. Folia Primatol 40:32-68.

Shepherd RJ (1991) Body Composition in Biological Anthropology. Cambridge, MA: Cambridge University Press.
Singer R and Wymer J (1982) Human skeletal remains. In R Singer and J Wymer (eds.): The Middle Stone Age at Klasies River Mouth in South Africa, pp 139-149. Chicago, IL: University of Chicago Press.
Siri, W.E. 1956. Body composition from fluid spaces and density: An analysis of methods. Berkeley, CA: University of California Radiation Laboratory Publication No 3349.






78
Smith RJ (1985) The present as a key to the past. Body weight of Miocene hominoids as a test of allometric methods for paleontological inference. In WL Jungers (ed.): Size and Scaling in Primate Biology. New York, NY: Plenum Press, pp 437-448.
Smith RJ (1994) Regression models for prediction equations. J Hum Evol 26:239-244.

Smith RJ (2002) Estimation of body mass in paleontology. J Hum Evol 43:271-287.
Smith P, Wax Y, Becker A and Einy S (1985) Diachronic variation in cranial thickness of near eastern populations. Am J Phys Anthropol 67: 127-133.
Sokal RR and Rohlf FJ (1995). Biometry: The Principles and Practice of Statistics in Biological Research. San Francisco, CA: WH Freeman and Co.
Sonakia A (1985) Early Homo from Narmada Valley, India. In E Delson (ed.): Ancestors: The Hard Evidence. New York, NY: Academic Press.

Steudel K (1980) New estimates in early hominid body size. Am. J Phys Anthropol 52:6370.

Stringer CB (1984) The definition of Homo erectus and the existence of the species in Africa and Europe. Cour Forsch Inst Senckenberg 69:131-143.

Tallgren A (1974) Neurocranial morphology and ageing-A longitudinal roentgen cephalometric study of adult Finnish women. Am J Phys Anthopol 41:285-294.
Tobias PV (1967) The cranium and maxillary dentition of Australopithecus (Zinjanthropus) boisei. In LSB Leakey, (ed.): Olduvai Gorge Vol. 2. Cambridge, Great Britain: University Press.
Todd TW (1924) Thickness of the male white cranium. Anat Rec 27: 245-256.
Trinkaus E (1983) The Shanidar Neandertals. New York, NY: Academic Press.
Twiesselmann F (1941) Methode pour l'evaluation de l'epasseur des parois craniennes. Bulletin du Musee Royal d'Histoire Naturelle de Belgique 17:1-33.
Warren MW, Walsh-Haney HA, Smith, KR, Stubblefield PR, Schultz J, and Falsetti AB (1999) Regional Differences in Cranial Morphology. Abstract. Proceedings of the Annual meeting of the American Association of Physical Anthropologists. Columbus, Ohio.

Washburn SL (1947) The relation of the temporal muscle to the form of the skull. Anat. Rec. 99:239-248.
Weidenreich F (1943) The skull of Sinanthropus pekinensis; A comparative study on a primitive hominid skull. Palaeontologica Sinica New Series D, No 10 Whole Series No 127:1-484.
Weidenreich F (1946) Generic, specific and subspecific characters in human evolution. In SL Washburn and D Wolffson (eds.): The Shorter Anthropological Papers of Franz Weidenreich Published in the Period 1939-1948. New York, NY: Viking Fund, pp 25-43. [Originally published Am J Phys Anthropol 4(4) n.s.]







79
Wolpoff MH (1980) Cranial remains of Middle Pleistocene European hominids. J Hum Evol 9:339-358.
Wood B (1984) The origin of Homo erectus. Cour Forsch Inst Senckenberg 69:99-111. Wood BA and Stack CG (1980) Does allometry explain the differences between "gracile" and "robust Australopithecines? Am J Phys Anthropol 52:55-62. Young, RW (1957) Postnatal growth of the frontal and parietal bones in white males. Am J Phys Anthropol 15:367-386.
Zar JH (1998) Biostatistical Analysis. Englewood Cliffs, NJ: Prentice Hall.













APPENDIX

Inventory of Male Subjects
Autopsy Sample N=47 This listing is sorted by age. Location is Michigan (mich) or Orlando (orl). Weight and body length were recorded in English measurements, unless otherwise noted, and converted to metric equivalents within the spreadsheet software.


Specimen Location Age Weight Body length (lbs) (in)
56 mich 18 210 68 58 mich 19 150 67.5 55 mich 21 149 66 24 orl 24 207 70.5 68 mich 28 158 70 62 mich 29 210 72.5 44 mich 30 201 72 39 mich 32 340 75 63 mich 32 405 78.5 46 mich 32 175 70 71 mich 36 253 76.5 34 mich 37 195 67 47 mich 37 190 68 20 orl 40 167 70 25 orl 40 130 65 32 mich 40 193 75 65 mich 40 199 69 23 orl 41 158 66.5 61 mich 41 249 74.5 54 mich 43 241 69 27 orl 44 165 71 38 mich 44 222 70 35 mich 45 285 75 48 mich 45 193 71 59 mich 46 184 72 73 mich 46 120 64 43 mich 47 171 66.5 40 mich 50 222 71 60 mich 52 231 73.5 18 orl 54 116 63








Autopsy Sample. Continued

Specimen Location Age Weight Body length (lbs) (in)
20 orl 54 261 76 26 orl 54 184 70 66 mich 54 182 21 orl 55 210 72 37 mich 56 252 72 57 mich 59 351 71.5 22 orl 60 124 61 64 mich 60 211 68.5 67 mich 60 190 70 52 mich 62 162 70 50 mich 63 206 68 70 mich 63 212 72 53 mich 65 176 70 42 mich 68 184 68.5 72 mich 74 127 64 69 mich 79 178 65 30 mich 82 118 68


Skeletal Sample N=100 Specimen numbers are as


in the NMNH Terry Collection.


Specimen Age Weight kg Stature cm
5 57 65.2 179.6 125 31 44.0 178.5 143 58 71.9 164.7 152 70 46.6 165.9 606 28 38.8 163 609 75 38.8 162 614 50 56.6 181.0 618 60 55.5 181 630 72 50.0 175.0 633 76 76.6 176.0 634 74 433 163.0 635 54 37.7 165.2 638 55 71.4 177.7 644 54 31.0 162.0 645 28 433 188.0 646 50 33.3 179 648 63 70.3 172 649 51 79.6 176 651 64 50.6 158.0 663 44 53.2 170 665 74 453 170.0 670 60 46.6 161.0








Skeletal Sample. Continued.


Specimen Age Weight kg Stature cm 671 65 62.7 158.0 674 58 59.9 171.0 675 27 42.0 176 701 55 52.2 165.0 705 46 38.9 171 707 26 42.2 187 708 68 34.4 167.0 709 50 41.1 155.0 711 61 63.3 187 712 47 60.0 166 713 69 51.0 176.0 714 71 43.3 169.0 717 29 51.1 160 718 22 55.5 180 719 25 51.1 170 720 54 72.1 156.0 721 60 43.0 184.0 724 42 58.8 170 725 57 31.8 163 730 71 82.1 167.0 731 63 61.0 166 743 53 533 178.0 746 71 41.1 161.0 747 45 62.4 172.0 748 85 50.7 182.0 750 80 39.8 174.0 756 47 49.7 165.0 762 59 38.8 171.0 763 46 54.8 179.0 764 58 40.6 168.0 772 56 63.4 186.0 787 50 52.8 175.0 802 36 35.1 168.0 803 64 45.2 161.0 805 87 57.0 168.0
806 78 60.1 168.0 812 54 39.9 172.0 813 60 47.4 175.0 814 55 36.1 176.0 842 73 81.0 165.0 843 60 54.1 161.0 852 66 38.7 169.0 853 62 54.0 176.5 855 67 38.7 158.0 861 65 52.5 174.0 865 69 41.5 169.5









Skeletal Sample. Continued.


Specimen Age Weight kg Stature cm 866 60 53.5 179.0 867 38 43.1 868 60 653 172.0 870 60 41.6 170.0 871 56 53.0 182.0 872 48 42.6 168.0 874 75 43.9 173.5 908 52 41.9 180.0 909 70 35.1 161.0 912 63 58.6 168.0 918 40 33.6 157.0 924 43 43.3 175.0 956 63 84.6 185.0 963 71 54.2 170.0 974 77 42.9 166.0 979 85 41.8 178.5 982 70 85.0 175.0 984 68 96.4 173.0 1087 62 88.3 182 1089 43 85.3 182.5 1110 68 48.8 175.5 1114 60 41.5 165.0 1125 69 52.2 164.0 1301 57 65.1 171.0 1384 65 62.5 167.5 1424 51 38.7 152.5 1430 70 37.2 160.0 1442 63 58.8 179.0 1492 64 37.6 167.5 136R 59 69.6 170.2 1783R 65 60.8 176.2 972R 65 37.9 172.2














BIOGRAPHICAL SKETCH


Phoebe R. Stubblefield received her Bachelor of Arts degree in anthropology at the University of California, Santa Barbara in 1990 and the Master of Arts degree at the University of Texas at Austin in 1993 with an emphasis in palaeoanthropology. She entered the University of Florida in 1995 under the direction of the late Dr. William R. Maples, and studied forensic anthropology with him until his death. Through her relationship with Dr. Maples, Ms. Stubblefield has performed numerous forensic analyses for medical examiners in Florida and New York, trained both federal and state law enforcement personnel, and acted as consult during the Valujet disaster. She has also assisted with historical grave recovery efforts in Oklahoma and Guatemala to identify victims of the Tulsa Riots and the Guatemalan civil war. While completing her doctoral research under the supervision and mentorship of Dr. Susan Ant6n, Ms. Stubblefield followed in Dr. Ant6n's footsteps and became a Ford Doctoral Fellow. Her research interests include skeletal biology, forensic anthropology, human evolution, and human variation.








I certify that I have read this study and that in my opinion it conforms to
acceptable standards of scholarly presentation and is fully adequate, rsc and quality, as a dissertation for the degree of Doctor of P ' y.



S- usan Ant6n, Chair
Assistant Professor of Anthropology

I certify that I have read this study and that in my opinion it conforms to
acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy.



Sue Boinski, Cochair
Associate Professor of Anthropology

I certify that I have read this study and that in my opinion it conforms to
acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy.



Anthony Falsetti\
Associate Professor of Anthropology

I certify that I have read this study and that in my opinion it conforms to
acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy.



Thomas Holli r
Associate Professor of
Anatomy and Cell Biology

This dissertation was submitted to the Graduate Faculty of the Department of
Anthropology in the College of Liberal Arts and Sciences and to the Graduate School and was accepted as partial fulfillment of the requirements for the degree of Doctor of Philosophy.

December 2002


Dean, Graduate School




Full Text

PAGE 1

CRANIAL SIZE IN RELATION TO BODY MASS AND SKELETAL ROBUSTICITY IN MODERN HUMANS By PHOEBE REGINA STUBBLEHELD A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULHLLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2002

PAGE 2

Copyright 2002 by Phoebe Regina Stubblefield

PAGE 3

ACKNOWLEDGEMENTS I first want to thank God for getting me through this doctoral process. I also want to thank the late Dr. William R. Maples and Dr. Susan Anton for their mentorship and instruction which led me to and through the University of Florida. I also thank Dr. Sue Boinski for her leadership and encouragement. My thanks go to Dr. Anthony Falsetti , Dr. Thomas Hollinger, and Dr. William Hamilton for their help during this process. My special thanks go to Margaret and Herb Gilliland for their gracious and generous support and concern. Many thanks go to the American Academy of Forensic Sciences' Lucas Grant and the National Academies Ford Foundation Fellowship, and the doctors and technicians of the medical examiners' offices in Florida and Michigan, without which and whom this work would not have been possible. Last of all I want to thank my parents for their support, Pamela for transcribing data, and my twin, Peggy, for being there when I needed her. iii

PAGE 4

TABLE OF CONTENTS ACKNOWLEDGEMENTS iii LIST OF TABLES vi LISTOFHGURES ix ABSTRACT x INTRODUCTION 6 Body Weight Estimation in Hominids 2 Research with Ectocranial Measurements 4 Research with Cranial Thickness Measurements 4 LITERATURE REVIEW 9 Pensler and McCarthy (1985) 9 Hartwig-Scherer and Martin (1992) 9 Nawrocki (1992) 10 Aiello and Wood ( 1994) 12 Gauld(19%) 13 Discussion 14 The Project 15 MATERIALS AND METHODS 18 Sample Selection 18 Data Collection 19 Somatic Variables 19 Cranial Measurements 21 Postcranial Measurements 25 Estimation of Lean Body Weight 26 Data Analysis 27 RESULTS 31 The Combined Sample: Descriptive Statistics 31 Combined Sample: Relationships Between Variables 33 Combined Sample: Principal Component Analysis 39 Combined Sample: Regression Analysis 43 The Autopsy and Skeletal Samples: Somatic Descriptive Statistics 44 Quality of Skinfolds 45 Autopsy and Skeletal Samples: Cranial Descriptives !!!!."."."!!."!!!!.50 Autopsy and Skeletal Samples: Relationships Between Variables ....53 iv

PAGE 5

Autopsy and Skeletal Samples: Principal Component Analyses 58 Autopsy Sample: Regression Analysis 63 Summary 64 DISCUSSION 67 REFERENCES 73 APPENDIX 80 BIOGRAPHICAL SKETCH 84 V

PAGE 6

LIST OF TABLES Table page 1 Aiello and Wood's results relative to body weight using a mixed hominoid sample 14 2 Cranial measurements 23 3 Clavicle measurements 25 4 Results of the one-sample t-test for sectioned and radiographic clavicle measurements 26 5 Descriptive statistics for somatic variables in the combined sample 31 6 Descriptive statistics for cranial variables of the combined sample 32 7 Comparison of shared ectocranial measurements in Howell's "mixed race" sample and the combined sample 32 8 Descriptive statistics for clavicle variables in the combined sample 33 9 Combined sample correlation coefficients between body weight and the somatic variables (including clavicle) 34 10 Combined sample correlation coefficients between body weight and the cranial variables 34 1 1 Combined sample correlation coefficients between the clavicle cortical thickness and cranial thickness variables 35 12 Combined sample correlation coefficients between the ectocranial and cranial thickness variables 36 13 Combined sample correlation coefficients between ectocranial and average thickness variables 33 14 Combined sample correlation coefficients between the cranial index and cranial thickness variables 33 15 Combined sample correlation coefficients between the cranial index shape categories and the cranial thickness variables 39 16 Combined sample correlation coefficients between cranial index categories and average vault thickness 39 vi

PAGE 7

17 Combined sample principal components using all variables (except age) 41 18 Combined sample principal components of the cranial variables and body weight 42 19 Combined sample regression coefficients for body weight versus the cranial variables 44 20 Descriptive statitics for the somatic variables in the autopsy and skeletal samples 45 21 Autopsy sample fatness categories based on Frisancho (1990) table iv39 for males using summed triceps and subscapular skinfold values 46 22 Autopsy sample weight classifications based on weight (kg) and stature (cm) and skinfolds 47 23 Contingency table of weight classifications based on weight for stature and combined triceps and subscapular skinfold thicknesses 48 24 Regression equations for estimating body density (d) from the logarithm of triceps and subscapular skinfold measurements 48 25 Autopsy sample descriptive statistics for fat weight estimates and the new variable lean weight 49 26 Weight classifications based on unadjusted weight for height versus lean weight for height 49 27 Autopsy sample descriptive statistics for the cranial variables 50 28 Skeletal sample descriptive statistics of the cranial variables 51 29 Autopsy and skeletal samples descriptive statistics for clavicle variables 52 30 Autopsy and skeletal samples two-tailed t-test for the cranial variables 52 3 1 Autopsy and skeletal samples correlation coefficients between body weight and the somatic and clavicle measurements 53 32 Autopsy and skeletal samples correlation coefficients between body weight (kg) and the ectocranial variables 54 33 Manipulated skeletal sample correlation coefficients for select variables 55 34 Skeletal sample correlation coefficients for body weight versus cranial variables excluding low weight subject 57 35 Autopsy and skeletal samples correlation coefficients between body weight (kg) and the cranial thickness variables 58 36 Skeletal sample correlation coefficients for additional ectocranial variables versus vault thickness averages 58 vii

PAGE 8

37 Autopsy sample principal components for the cranial variables 60 38 Skeletal sample principal components using £ill variables (except age) 61 39 Skeletal sample principal components using the cranial variables 63 40 Autopsy sample regression coefficients for body weight versus cranial variables 64 41 Autopsy sample regression coefficients for lean weight versus cranial variables 65 viii

PAGE 9

LIST OF nOURES Figure page 1 Ectocranial Measurements 24 2 Locations of Cranial Thickness Measurements 24 3 Skeletal Sample Weight for Height Class Distribution 56 ix

PAGE 10

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 CRANIAL SIZE IN RELATION TO BODY MASS AND SKELETAL ROBUSTIOTY IN MODERN HUMANS By Phoebe Regina Stubblefield December 2002 Chair Susan Ant6n Cochair: Sue Boinski Major Department: Anthropology Body weight is critical to understanding primate biology. However, taphonomic processes hamper acquiring this statistic from human skeletons or fossil hominids. Fossil hominid species are defined from crania that frequently lack associated postcrania. Thus body size information is frequently restricted to that which can be extracted from the skull. To address this issue, I designed a static allometric study using modem humans as the reference sample. I tested two hypotheses regarding the relationship between cranial size and body size. The first hypothesis stated that measurements of external cranial dimensions (ectocranial) covary with body weight, and the second that cranial vault thickness measurements covary with body weight in relation to systemic skeletal robusticity. To test these hypotheses I sampled 147 adults of known body weight (100 from the Terry collection at the National Museum of Natural History and 47 from recent autopsies) for X

PAGE 11

sixteen ectocranial and seven cranial vault thickness measurements. To assess postcranial robusticity I also measured cortical thickness at four shaft locations on the clavicle. I analyzed the relationships in the data using correlation and regression analyses, and principal component analysis. In contrast to the published literature that uses multi-species hominoid and anthropoid data sets, cranial measurements were not well correlated with body weight. Vault thickness measurements were particulariy poorly correlated. There was also no pattern of association between gross vault thickness and clavicular cortical bone thickness. The lack of relationship between body size and many cranial measurements in modem humans indicates that these measurements are largely unsuitable for body weight prediction. xi

PAGE 12

INTRODUCTION This study addresses the feasibility of predicting body weight in large-brained hominids from cranial measurements. Body weight is critical to hominid biology because it is the best indicator of body size (Jungers 1984), it is essential to understanding dietary, locomotor, and reproductive behaviors in all primates, and because it is central to understanding brain development, or encephahzation, in the hominids. An understanding of body size is critical to models of hominid expansion outside of Africa, as body size is directly related to home ranges, time spent foraging, ability to cope with seasonal food availability, predation risk, and food choice (McNab 2002). While there is abundant literature on postcranial techniques of body size estimation in hominids, sources using cranial measurements are limited. Therefore I proposed and tested several hypotheses involving the relationships between body size (as indicated by body weight) and cranial measurements. My primary hypothesis is that if body components, such as the limbs or torso, are scaled in proportion to the whole body, then the cranium should follow. I test this ptx)posal in two related hypotheses: • HI : body weight in modem humans covaries with cranial measurements • H2: estimates of lean body weight will have stronger covariance relationships with cranial measurements than gross body weight will have with cranial measurements I employ both ectocranial measurements and measurements of vault thickness. Cranial dimensions (ectocranial measurements) have been associated with size differences due to sex and to ancestry (Giles and Elliot 1962, 1963), so a link to body size might be 1

PAGE 13

expected. However, the function of tri-layered cranial bone is still subject to discussion (Kennedy 1991), although Moss and Young (1961) attribute separate functions to each layer. Physiological functions of hemopoiesis and fat storage take place in the diploic bone of the vault; likewise cortical thickness in long bones has biomechanical determinants (Ruff 2000), and the same could be true for the cranial vault (Currey 1984, Demes 1985). I will test the relationship between vault thickness and body size with hypotheses 1 and 2. Hypothesis 2 derives from an assumption that cranial measurements and body size are physiologically linked (Lieberman 1996). Increasing obesity in Western adults (Regal et al. 1998) makes an estimate of lean body weight critical in comparisons to the non-weightbearing cranium. My secondary hypothesis is that thickness of the non-pathological neurocranium, or vault, in the hominids is primarily the product of general skeletal development. Therefore hominids with systemically thicker cortical bone in the postcrania will also have greater vault thickness, whereas hominids with slighter postcranial cortices will have lesser vault thickness. Because cranial contents directly influence the magnitude and directions of vault development, brain size might also affect vault thickness (e.g.. Moss and Young 1961; Richards and Anton, 1991). I propose three additional hypotheses regarding vault thickness: • H3: vault thickness is a function of general skeletal robusticity as indicated by cortical thickness of long bones • H4: an index of vault length and height covaries with vault thickness • H5: weight of the brain significantly covaries with vault thickness Body Weight Estimation in Hominids Body weight is correlated with a variety of biological features including dietary behavior (Kay 1975, McHenry 1984), locomotion (Jungers 1984), population density (Martin 1981), and brain size (McHenry 1976, Holloway 1980, Rightmire 1986). Body mass, according to Lindstedt and Calder (1981:2), is the "primary determinant of ecological

PAGE 14

3 opportunities, as well as of the physiological and morphological requirements of an animal." In order to understand the biology of fossil primates, body weight must be reconstructed from the recovered skeletal material. Most techniques for estimating body weight in fossil primates use dental or postcranial elements. The dental studies (e.g., Conroy 1987, Shea 1983, Pirie 1978, Gingerich et al. 1982, Gingerich 1977, Dagosto and Terranova 1992) are mainly derived from nonhuman primate samples. The postcranial and some of the dental analyses tend to use proxies for body weight in order to cope with the lack of recorded live weights in the nonhuman parts of the sample (e.g.. Wood and Stack 1980— cranial length, Steudel 1980— partial skeletal weight. Ruff 2000— bi-iliac breadth, Pirie 1978— cranial length, Corruccini and Henderson 1978— three skull measurements). Yet Hartwig-Scherer and Martin (1992) demonstrated that a strong relationship between a proxy variable and body weight does not guarantee a strong relationship to body weight for a third variable that has a strong relationship to the proxy. McHenry (1976, 1991) and Hartwig-Scherer (1994) relied on known body weights for their samples (mixed primate in the former, human in the latter) for developing predictive equations from long bone measurements. In another study using known body weights, Baker and Newman (1957) related dry skeletal weight to living weight. Cranial predictors of body weight are needed because of the frequent lack of association between cranial and postcranial remains in the hominid fossil record. Reviews and atlases of the human fossil record (e.g., Larsen et al. 1998) shows a clear bias towards cranial specimens, and likewise recent definition of the new possible hominid species Sahelanthropus tchadensis (Brunet et al. 2002) was based on cranial elements alone. Essentially, fossil hominids are cranial species. Weidenreich ( 1946:40) comments on this from an earlier vantage in palaeoanthropology: "How have the manifold characters which determine a type to be weighted in the system of classification? Do qualities of the teeth count more than those of the bones of the jaws, or those of the skull more than those of the

PAGE 15

limbs?" Weidenreich gave preference to cranial characters, believing that evolution leading to the modem human brain was the most salient theme of hominid evolution. Without cranial predictors giving access to body weight, the chance and rare finds of associated postcrania hamper further investigation of the biology of extinct hominids. Research with Ectocranial Measurements Ectocranial measurements have a long history in physical anthropology as descriptive tools. Classic texts by Martin and Sailer (1957) and Howells (1973) define a variety of cranial measurements for the human skull, and the use of these measurements is well illustrated in descriptions of fossil hominids (e.g. Jacob 1972, Wolpoff 1980). The use of these measurements as predictive tools has had a strong role in forensic anthropology in estimating sex (e.g., Giles and Hliot 1%2, Loth and Henneberg 19%) and ancestry (Giles and Elliot 1963) from the skull. Only a few publications link ectocranial measurements to body weight (Aiello and Wood 1994, Hartwig-Scherer and Martin 1992). Aiello and Wood demonstrated high correlations (r > .90) between some ectocranial measurements and body weight in a mixed primate sample. Hartwig-Scherer and Martin (1992) examined only two cranial variables and found no significant correlations in their static human sample. Research with Cranial Thickness Measurements The history of research surrounding cranial thickness in hominids has taken three primary approaches. The explanatory approach attempts to identify or propose the source or cause of normal adult cranial thickness. The descriptive approach is centered on a particular population or geographic area, and is concerned with describing the variation of cranial thickness due to age, sex or other biological variables. Finally, the pathological approach is concerned with the etiology of abnormal cranial thickness. This last aspect is well illustrated in the work of Ortner and Putschar (1981) and will only be discussed here when the pathological intersects with the other approaches.

PAGE 16

5 The explanatory approach examines the source of cranial thickness and may present models explaining cranial vault structure. Proposed mechanisms for increased cranial vault thickness encompass behavioral, biomechanical (Washburn 1947, Demes 1985, Nawrocki 1992), to physiological (Ivanhoe 1979, Lieberman 1996). Behavioral explanations stress dangerous subsistence practices or interpersonal violence, both involving blunt trauma to the head (Weidenreich 1943). Brown (1994) proposed that a midline area of increased cranial thickness in the frontal and parietal bones of Australian aborigines was related to the practice of settling disputes by trading blows to the head with a wooden club or staff. As this behavior predominated in young adults he thought such behavior would easily select against thin vaults. He supported this idea with findings of a suite of injuries associated with the practice in late-Pleistocene aboriginal specimens dating to about 1 1,000 years BP. Others have suggested that the increased ectocranial thickness of H. erectus may be related to general skeletal hypertrophy in this species (sensu Kennedy, 1985 and Hublin, 1986). While the behavioral hypotheses still want more support, such as a survey of cranial thickness in cultures that end disputes with head-bashing, the biomechanical and physiological models involve more discrete hypothesis testing. The emphasis on testing was stressed initially by Moss and Young (1961:281) who suggested a functional approach to craniology to provide "significance" to the rapidly accumulating sets of cranial measurements and morphological descriptions. They divided the human cranium into neural and facial functional components that were each subject to further functional subdivision. Demes (1985) proposed that the vault of the modem human be studied in terms of a thin-walled globe, biomechanically known as a "shell," rather than with beam theory. The shell morphology of the skull reduced transmission of bending stresses to the thinner walled portions of the anterior base by transmitting strain into the walls of the vault. Based on stress tests of models of the basicranium, she suggested that vault curvature influences wall thickness such that the more curved the vault bones, the less thick they are. However she also stated that more research was needed, since her study did not examine the

PAGE 17

6 entire vault. Proposing that longer crania and larger faces would cause more bending stress in a cranium, Nawrocki (1992) tested Demes' theory on a sample of Pleistocene hominids and modem humans. He found cranial thickness to be associated with flatter vaults and larger faces in his sample, although the recent human subsample was not well correlated for vault thickness and wall curvature. In his sample of modem humans with known body weights (drawn from the Terry collection), Nawrocki did not find significant correlation between body weight and vault thickness. The physiological hypotheses have dealt with extemal and internal forces on cranial thickness. Moss and Young (1%1) described the inner and outer layers and diploe of the vault as independent functional units. The inner table follows the contours of the neural mass, including accommodating alterations caused by neural atrophy. Crests and ridges from muscular attachments only affect the outer layer (Washburn 1947). In addition to serving as a site for hemopoiesis the diploe also serves to lighten the total cranial bone mass, a conclusion supported by Currey (1984) as a biomechanical solution to minimize bone mass without compromising bone strength. Studies of artificial cranial deformation also suggest that intemal and external tables may act independently (e.g., Anton, 1989) Other studies have related cranial thickness to geomagnetic field intensity (Ivanhoe 1979). Ivanhoe supported his proposal with a correlation coefficient between 0.6 and 0.7 for his Pleistocene hominid sample, as well as noting the robusticity of a few other mammalian genera in times of varying geomagnetic field intensity. Ivanhoe' s study would be better supported by a more thorough review of Pleistocene fauna and consideration of modem human variation in cranial thickness. Association between cranial thickness and activity level was suggested by Lieberman (1996), based on studies of human physiological response to exercise that showed an increase in growth hormone secretion with increased exercise either of long duration (Lassarre et al. 1974) or of high intensity (Felsing et al. 1992). His study of a small sample of pigs and armadillos demonstrated marked increase in both cranial and postcranial cortical

PAGE 18

7 thickness in exercised versus unexercised siblings. In a survey of cranial thickness in Holocene hominids Lieberman showed that pre-industrial farmers had greater cranial thickness than their postindustrial counterparts, thus contradicting his null hypothesis that increased activity did not affect cranial thickness. Descriptive studies are the most common product of cranial thickness research. Todd (1924) produced the earliest rigorous descriptive study of modem humans, which has since been followed by several others (Twiesselmann 1941, Adeloye et al. 1975, Ohtsuki 1977, Kaito 1980, Pensler and McCarthy 1985, Brown 1987a, Ishida and Dodo 1990, Gauld 1992, Ross et al. 1998). While these studies all utilized direct bone measurement, several radiographic studies have also contributed to the literature of human cranial thickness (Roche 1953, Young 1957, Getz 1960, Israel 1973, 1977, Tallgren 1974, Smith et al. 1985, Brown et al. 1979, Brown 1987b). Part of this literature documented variation in cranial thickness in various human populations, such as Todd, Adeloye et al., and Pensler and McCarthy for American blacks and whites, Twiesselmann and Getz for various European populations, Kaito for Japanese, Brown for Northern Chinese and Australian aborigines, and Gauld for Native Americans. Many of these described a conflicting pattern of thickness change with increased adult age. Israel and Tallgren both reported from longitudinal radiographic studies, but the former found a pattern of cranial measurement increase with age, while the latter did not. Smith and colleagues did not find significant difference in cranial thickness in a sample of Near Eastern crania, although they note that the older crania were thicker. In another diachronic study of modem and Neolithic Japanese, Ishida and Dodo found significantly greater cranial thickness in their Neolithic males. Another focus of the descriptive study is the variation in cranial thickness in PlioHeistocene hominids (Hinton et al. 1938, Weidenreich 1943, Tobias 1967 Jacob 1972, Wolpoff 1980, Singer and Wymer 1982, Trinkaus 1983, Wood 1984, Clarke 1985, Sonakia 1985, Kennedy 1985, 1991, Brown 1994, Anton and Franzen 1997). Particularly,

PAGE 19

8 specimens of Homo erectus. Homo ergaster, and Homo neandertalemis appeared to have greater non-pathological cranial and postcranial bone thickness than seen in modem humans. Pathological cranial thickening has been clearly demonstrated in one H. erectus and two Neandertal specimens (Anton 1997), in the form of hyperostosis calvariae interna (HCl). Cranial thickness in Homo erectus had been considered a derived trait for the species (Andrews 1984, Stringer 1984), but Kennedy (1991) showed that the similar state in H. neandertalensis prevented cranial thickness from being an autapomorphy in H. erectus. Brown (1994) demonstrated that the vault thickness of Asian H. erectus falls within the range of variation seen in Australian aborigines. He concluded that the emphasis on cranial thickness in Pleistocene hominids, rather than illustrating a peculiar character of these large brained bipeds, actually displays a lack of familiarity with variation in modem human vault thickness. Lieberman (1996) supported Brown's work while reviewing additional Pleistocene hominids, and showed that the range of cranial thickness was not outside the variation in modem humans. However, while absolute cranial thickness may not separate H. erectus and H. sapiens, the relative contributions of cortical versus diploic bone to thickness may differ between the taxa (Hublin, 1986; Anton 1997). Increased thickness in H. erectus is generally achieved via cortical expansion, whereas diploic expansion accounts for most increases in thickness in modem humans (Anton 1997). While the cranial thickness literature is not tidy, some themes are discernible. Populations, at least in the United States, vary in cranial thickness, but only in location of greatest thickness, not in overall cranial thickness. Males and females do not differ in average thickness. Cranial thickness might increase with age, but even the longitudinal studies disagree. Modem human variation in cranial thickness encompasses that seen in fossil hominids particularly Homo erectus. Exercise has strong support for influencing cranial thickness (and skeletal robusticity in general), while behavioral, biomechanical, and

PAGE 20

geomagnetic effects want more testing. Most relevant to this study, there is no clear consensus for a significant relationship between cranial thickness and body weight.

PAGE 21

LITERATURE REVIEW A limited body of literature exists on the relationships between cranial measurements and body weight in modem humans, representing the work of five authors. Two authors focus on ectocranial measurements of the skull (Haitwig-Scherer and Martin 1992, Aiello and Wood 1994). Two others focus on vault thickness measurements (Pensler and McCarthy 1985 and Gauld 1996), while the fifth (Nawrocki 1992) uses a combination of measurements. Pensler and McCarthy (1985) The earliest work is Pensler and McCarthy's examination of the vault as a bone donation site. They measured cranial thickness at two frontal and two parietal locations in 200 male and female autopsy subjects with recorded weights, race (American Blacks and Whites), and age at time of death. Correlation of the log transformed thickness and weight data yielded r-values ranging from 0.2 to 0.6. The authors did not support cranial thickness as a predictor of body weight since they found that a 75-pound difference in weight produced only a 1.5mm difference in vault thickness. Hartwig-Scherer and Martin (1992) These authors reported two cranial variables in a study of the effect of using proxies for body weight when predicting body size in hominids. The authors used a sample (n = 295) of subadult and adult hominoids {H. sapiens. Gorilla gorilla. Pan paniscus. Pan troglodytes, and Pongo pygmaeus) with recorded body weights. They subdivided the sample to provide results on static and ontogenetic allometry. Measurements of the skull, humerus, radius, femur and tibia were collected and major axis line fitting used on the loglO transformed data. With one exception the human sample produced such low correlation coefficients 10

PAGE 22

11 with body weight that no results were reported. The adult human sample (n = 19) yielded no reportable results for either cranial variable (skull length and basicranial length). The ontogenetic sample produced an r-value of 0.93 for basicranial length with body weight. Nonhuman hominoids had r-values ranging from 0.90 to 0.98 for basicranial length and body weight in their ontogenetic samples. Generally skull and basicranial length performed well for the nonhuman apes. Adult samples had r-values ranging from 0.90 to 0.96 for basicranial length, while ontogenetic values ranged from 0.90 to 0.98. The lowest correlation coefficient was 0.66 for skull length in adult G. gorilla. Hartwig-Scherer and Martin displayed an interesting similarity between subadult human and nonhuman ape basicrania, and also indicated a strong contrast between humans and other apes in the performance of cranial variables as predictors of body weight Nawrocki (1992) In a test of the effect of shape on cranial thickness Nawrocki took thirty-five craniofacial and eighteen bilateral and sagittal vault thickness measurements on a sample of recent and archaic modem humans and extinct hominids. He combined some of the cranial dimension measurements into an index of sphericity that was intended to remove size. He later recognized that size was not removed as illustrated by the significant correlations between it and his cranial variables indicative of size (maximum length, maximum breadth, and basion-bregma height). The individual vault thickness measurements were combined into five variables of average thickness summarizing entire vault thickness with and without the measurement at asterion, thickness of the anterior and posterior halves of the vault, and posterior thickness without asterion. Nawrocki examined the relationship between his cranial measurements and body size (as indicated by femur length, vertical femoral head diameter, body height and weight) in the modem humans drawn from the Terry collection. He found several significant correlations between the individual ectocranial measurements and body size, the strongest involving body height (body length and the femoral measurements), but did not prx)vide

PAGE 23

12 results for each variable. The mean thickness variables lacked significant correlation to any of his size variables. Individual thickness measurements yielded a few significant relationships, notably between the femoral measurements and asterion and the parietal bosses (r = 0.24 to 0.41 for the former, -0.21 to -0.25 for the latter). Coirelation specifically to body weight was not described in detail, although Nawrocki did attribute a lack of similarity between his results and those of Pensler and McCarthy to low sample sizes (n < 30) and low body weights in the Terry sample. Comparisons of the index of sphericity to entire vault thickness produced variable results throughout Nawrocki's sample. The modem humans (n = 122) did not have a significant r-value (p > .05), but the archaic hominids (n = 61; including H. neandertalensis, H. heidelbergensis, H. erectus, and Homo habilis) produced a coefficient of 0.487 (p < 0.0001). The combined sample (n = 183) yielded a coefficient of 0.553 (p < 0.0001). Based on the combined sample results Nawrocki concluded that the positive correlation between vault shape and vault thickness supported his hypothesis that long low crania will have thicker vaults than high globular crania. He also found support in the archaic sample for a moderate relationship between face size and vault thickness, with bizygomatic breadth and facial height length providing the best correlation coefficients (r = 0.55 and 0.44 respectively). The strongest associations in the recent subsample were between body weight and mandibular body thickness and basion-prosthion length (r = 0.34 and r = 0.30, respectively). Nawrocki's test of Demes' biomechanical shell model for the human cranium was most strongly supported in the archaic portion of the sample. Longer and lower vaults with less wall curvature were significantly associated with thicker vaults, but only in the archaic sample; Larger faces were correlated with thicker vaults, most strongly in the archaic sample. Nawrocki's results suggest an alternate model may be needed to explain vault thickness in recent humans.

PAGE 24

13 Aiello and Wood (1994) Aiello and Wood studied cranial measurements in anthropoids in order to develop regression formulae for predicting body weight in fossil hominids. They collected fifteen ectocranial and fourteen postcranial variables on a sample of 250 anthropoids. These authors used six orbital variables (breadth, height, area, biorbital and interorbital breadths, and postorbital breadth), biporionic breadth, two palatal variables, and six basicranial variables (three on the foramen magnum and three on the occipital condyles). Twenty-four individuals in the sample were human subjects with known body weights obtained during autopsies. The nonhuman subjects had body weights drawn from the literature (Harvey et al. 1987). The sample was divided into anthropoid and hominoid subgroups. I am focusing on the hominoid subgroup because the authors did not report results for the human subjects alone. The cranial variables showed uniformly high correlation coefficients with body weight using loglO transformed group means (Table 1), while weaker relationships existed among the postcranial variables. The best variables in the hominoid sample were orbital area (r = 0.98), biporionic breadth (r = 0.98), and orbital height (r = 0.98), based on having a high r-value, a low standard error of the estimate (SEE), and a low percentage prediction error. Aiello and Wood cautioned against using anthropoid-based formula for hominoids, because in their sample less than half of the hominoids received a predicted body weight within 20% of the known weight (their criterion for high percentage prediction error). The postcranial variables were drawn from the femur, tibia and humerus. Three measurements of the humerus (maximum length, epicondylar breadth, and distal joint breadth) performed as well as the best cranial predictors in the hominoid sample. Poor performance of the leg variables was attributed to the difference in body proportion between humans and the other large-bodied apes. This point again illustrates an effect of using a mixed-taxa reference sample.

PAGE 25

14 Table 1. Aiello and Wood's Results relative to Body weight Using a Mixed Hominoid Sample Variable N r Orbital breadth 12 0.96 Orbital height 12 0.98 Orbital area 12 0.98 Interorbital breadth 12 0.81 Biorbital breadth 12 0.95 Postorbital breadth 12 0.73 Biporionic breadth 12 0.98 Intercanine breadth 12 0.89 Palate length 12 0.89 Foramen magnum length 12 0.90 Foramen magnum breadth 12 0.92 Foramen magnum area 12 0.92 Occipital condyle length 12 0.93 Occipital condyle breadth 12 0.90 Occipital condyle area 12 0.94 N are from species/sex means (six hominoid genera, two sexes, n = 12) Gauld (1996) Gauld examined the relationship between vault thickness and body weight in a large (n = 235) sample of extant anthropoids in order to predict body weight in extinct hominids. Although detailed information was not provided, Gauld indicated that some nonhuman body weights in the extant sample were obtained as averages from the literature (Harvey et al. 1987 and Jungers 1988). The modem humans (n = 27) were drawn from archaeological collections and had body weights predicted from McHenry's (1988) equations. Gauld measured cranial thickness at five locations on the vault (midfrontal, bregma, midparietal, midtemporal squama, and inion) and calculated correlation coefficients and best fit lines (least square, reduced major axis and major axis) to define the relationship to body weight The results indicated generally strong relationships between vault thickness and body mass for the entire catarrhine sample and the hominoid subgroup. The correlation coefficients between these variables and body weight ranged from 0.75 to 0.94 for the hominoid portion of the sample. Gauld's examination of the residuals from the least squares regression showed that the hominids (Australopithecus africanus, H. sapiens, and Herectus) shared a

PAGE 26

15 pattern of deviation from the catarrhine regression. The greater absolute vault thickness in H. erectus was demonstrated by the greater shift, than seen in the other hominids, of its residuals from the regression line. In her discussion Gauld references the low correlation coefficients obtained by Pensler and McCarthy as "not unexpected, given the increased number of environmental factors contributing to individual phenotypic thickness variation" (p. 420). Yet Gauld's sample consisted of a small number of humans from coastal and inland California, Japan, and Hji. Her sample must reflect as many if not more different environmental factors as a group of American Whites and Blacks. Also the large sample size used by Pensler and McCarthy reduced the possibility of creating a sample of divergent individuals. Gauld's sample size did not do this. Discussion These five sources have presented contrasting images of tiie relationship between body weight and cranial measurements in modem humans. Pensler and McCarthy, Nawrocki, and the more limited work of Hartwig-Scherer and Martin all indicated low to moderate correlation coefficients with body weight. In contrast are tiie works of Aiello and Wood and Gauld that demonstrated strong relationships. Two critical methodological differences separate these studies; tiie source of body weights and tiie type of samples used. The former group used human and/or nonhuman subjects witti recorded body weights, and analyzed samples consisting solely of humans. The latter group relied on combinations of published mean body weights for primate species, measured, and predicted body weights and used mixed-primate samples for their analyses. In light of empirical research, tiie relationships between cranial measurements and body weight, in any primate or organism, are best illustrated by metiiodologies using acttial observations. The results of Pensler and McCarthy, Hartwig-Scherer and Martin, and Nawrocki have more relevance, as tiieir results are based on actual observations of body weight That tiiese autiiors reported results from observations of modem humans makes tiie

PAGE 27

16 results applicable to predicting body weight in modem humans and organisms most like them, that is fossil hominids. Reliance on a mixed primate sample, as Conroy (1987) indicated, facilitates over and under prediction of errors. Aiello and Wood demonstrated awareness of this problem in noting that the greater emphasis on leg length in modem humans reduced the correlation coefficient between the femoral measurements and body weight in their hominoid sample. The adaptations of bipedality are a defining morphological trend of the hominids, and it is partnered with the equally defining trend of increase in cranial capacity. A collection of extant hominoids sampled for cranial measurements has the same potential for reduced covariance as hominoids sampled for leg measurements. The Project These five works indicate that further testing is needed of the utility of cranial measurements, whether ectocranial or of cranial thickness, for estimating body mass in large-brained hominids. Utilizing a large sample size composed exclusively of modem humans with recorded body weights, in combination with an array of ectocranial and vault thickness measurements, I tested the utility of cranial measurements as predictors of body weight. Unlike the authors in this review, 1 also used a postcranial indicator of robusticity to investigate the contribution of skeletal size to vault thickness. I designed a static nanx)w allometry study (Shea 1983, Smith 1985) such that adult specimens of a single species were used as the study sample. A static study was acceptable because the research questions were not aimed at ontogenetic changes in the hominids. A narrow study avoids the uncertainty of predicting body weight from a multi-species plot, especially when the focal specimen may not be represented in the species used to generate the plot. Only modem humans were sampled in order to avoid prediction errors possible in mixed-primate samples (Conroy 1987). The results of this study are certainly applicable to modem humans and are also a resource for predictions for the large-brained extinct

PAGE 28

17 hominids, since modem humans are the only available reference sample. Direct measurements of body weight were required in order to avoid the problems of intervening variables (Hartwig-Scherer and Martin 1992).

PAGE 29

MATERIALS AND METHODS Sample Selection Several challenges were encountered in identifying satisfactory sources of human cranial and postcranial measurements. Requisite for this analysis were a generous number of intact crania as well as vital statistics such as body length and weight. While determining that the Terry collection (National Museum of Natural History, NMNH) met this need, I also drew upon my rapport with local medical examiners in Florida and Michigan to collect data in their offices. This choice limited feasible research protocols for the entire study, as will be described below. Several biases were present in the study sample due to use of the morgue as a data source: 1 . A morgue population represents local demographics in terms of population ancestry. 2. Primarily people of European descent populated the resource areas. 3 . Better health maintenance causes fewer women than men to be subject to autopsy (see Florida statute 406. 1 1 and Michigan statute 52.202). 4. Many older females had to be excluded due to pathological thickening of the cranial bones resembling hyperostosis frontalis interna (HFI; Anton 1997, Hershkovitz et al. 1999). In order to complete the study in a timely manner with these biases in effect, an exclusively male sample of predominately European descent was accepted. Measuring points were also subject to sample bias, especially in the choice of a postcranial measurement site. The postcranial site could not alter or interf-ere with autopsy 18

PAGE 30

19 protocol and had to be accessible without causing unreasonable or unethical changes to the disposition of the deceased. Given these limitations the best site was the clavicle, which unlike other long bones is frequently cut as a part of the autopsy protocol. Cuts were made at midshaft for one (usually the left) of the clavicles for a direct measurement of cortical thickness. Terry clavicles could not be sectioned so these clavicles were radiographed and cortical measurements taken from radiographs. The autopsy and skeletal samples differed in the availability of cranial measurements. Access to the cranial base and face was severely limited by autopsy protocol, so only a few measurements were available from this area for the autopsy sample. This limitation did not exist in the skeletal collection, so an extended set of cranial measurements was taken. Forty-seven male subjects composed the autopsy sample, while 100 males composed the skeletal sample. Data Collection Somatic Variables Certain descriptive data were recorded for both samples. Each specimen in the skeletal collection had documented sex, age, race, body weight and body length. In the morgue sex and age were determined by physical observation and written documentation, respectively. Ancestry was recorded according to documentation associated with the individual or based on the medical examiner's determination, which was largely based on external phenotype and surname. Supine body length was recorded in inches from a steel measming tape or ruler. Body weight was recorded in pounds on a digital floor scale or more rarely extracted from hospital intake records. The latter source was used if the individual had been on fluid therapy shortly before death. The floor scale measurements followed the autopsy protocol, which meant that the weights included some amount of clothing and or drapes. Recorded weights were not adjusted for clothing since the few kilograms of difference are not of antemortem or postmortem medical concem (William Hamilton, M.D., pers. comm.). Care was taken however, to weigh any large medical

PAGE 31

20 devices, such as head-movement restriction halos, but this occurrence was very rare. Age was recorded in order to examine the sample for variation in cranial thickness, since variation in tiiickness witii increasing age is still disputed (Tallgren 1974, Israel 1977). Individuals were excluded from tiie autopsy sample when tiiey had cranial fractures that interfered with the measurements and when the subject was emaciated. Occasionally some specimens had missing data for reasons other tiian tiie occasional missed measurement, such as the one subject in die skeletal sample lacking a recorded stature measurement. Other discrepancies were due to unexpected incompatibility of die research tools to human variation, such as when die spreading calipers accommodated most but not all morgue subjects for the measurement of cranial base length (basion to nasion) where basion had to be approached from the inside of die vault Quite frequentiy contact with basion could not be made. Contact on the endocranial surface was also a problem for some of die thickness measurements, particularly when an elongated frontal crest interf"ered witii die midfrontal measurement (avoidable in die morgue sample, but not in die skeletal sample). The grooves for the middle meningeal arteries on die parietal might have interf"ered widi diickness measurements, but die broad point of contact on die caliper extension arms would not fit widiin these grooves. Anodier difficulty involved die radiolucent character of some of the clavicle cortices in die skeletal sample. Occasionally radiographic examination of die clavicle yielded no measurable cortical diickness of die posterior or inferior walls, so no measurement was entered. In order to preserve a large sample size for later statistical analyses, die missing data were handled widi pairwise comparison, such diat missing cases were excluded from die analysis, but die subject and all its attached variables widi complete cases was not excluded. While I excluded the emaciated in die autopsy sample, no such effort was feasible for the skeletal sample. The nature of this study, examining cranial predictors of body weight rather dian load-bearing bones (e.g., Jungers 1988 on teeth), makes a minimum body weight seem nearly inconsequential, as long as die individual was alive while having diat

PAGE 32

21 weight. The practical reason for avoiding emaciated subjects is to include only mobile individuals. A real population will have a range of body weights including both extremes. A population of wild nonhuman primates is unlikely to ever have individuals with the upper extremes of weight and more likely to contain individuals of the lower extremes, especially if seasonality is a part of their ecology. Considering that a collection of nonhuman primates would not be discarded because all of the specimens were captured in the season of their lowest weights (if this season was even recognized), it is wasteful to discard emaciated humans from this skeletal sample. Weight data, which are intended for a species mean should not be based only on individuals from the middle and upper ranges of their weights. Since many Terry subjects were of very low weight (Jantz and Moore-Jansen 1988), I will examine the proportion of low weight individuals in the skeletal sample and the effects of their inclusion on the correlation analyses. Cranial Measurements A total of twenty-three cranial measurements were utilized in the entire study, comprising sixteen ectocranial measurements and seven vault thickness measurements (Table 2 and Figure 1). The ectocranial measurements were drawn from Howells (1973), Bass (1995), and Giles and Elliot (1963). In order to acconmiodate positioning restrictions in the autopsy sample, cranial height was taken from porion to bregma following Stewart's direction (see Bass, 1995). Both this measurement and height from basion to bregma were taken in the skeletal sample. Ectocranial measurements were taken with GPM sliding and Paleotech® spreading calipers, and for the measurement porion-bregma, a Paleotech® radiometer. The thickness measurements illustrated in Figure 2 were drawn chiefly from Gauld (1992) with two exceptions. Gauld' s point at inion was replaced with a point at lambda due to inaccessibility of inion in the autopsy sample, as well as general interference of internal occipital structures at this location. Gauld' s point at the posterior-inferior quadrant of the parietal was replaced with a point in the superior-inferior quadrant. Most of the thickness

PAGE 33

22 measurements required points to be located on the skull before measurement could proceed. The point at midfrontal was located by taking the frontal chord with the sliding caliper and marking its midpoint on the skull. Parallel to this point on the temporal line was the point temporal crest. Midparietal was located at the intersection of two lines on the right parietal, one running anterior-posteriorly from the midpoint of the coronal suture to the lambdoidal suture, and the other from medial to lateral from the midpoint of the sagittal suture. The subjective middle of the two superior quadrants formed the other parietal points. In the skeletal sample the thickness measurements were taken with a Mitotuyo® digital caliper fitted with extension arms. The extension arms allowed measurements to be taken by inserting one arm into the foramen magnum, bringing the other arm into contact with the outer table of the cranium at the desired point, and sliding the internal arm gently into contact with the inner table. Selection of cranial measurements was aimed at investigating various research goals. Many of this study's measurements are associated with estimation of ancestry in modem humans (Giles and Elliot 1962), and may be useful in examining relationships between shape of the cranium and vault thickness. The Cranial Index (maximum cranial breadth* 100/maximum cranial length), was used as an indicator of cranial shape and employed in correlation analyses with cranial thickness.

PAGE 34

23 Table 2. Cranial Measurements. Ectocranial Measurements Variable No 1 L ividAiiuuiii lengui i^gidDeua-occipiiaicj 2 IV^aYiTniim rM^?i/Hl"ri loiit^/i^n <»ii»^7rt«i iviaAiiiiuiii uicauiii \^curyon-curyon_/ i^idiiicu uabc lengui (^Ddsion-nasionj 4 uXaUllC-Ulal urcauiii 5 iviiiuiiiuin ironuii ureaum (^ironioiemporaie-ironiotemporaiej 6 ri uiiuii ciiuiu v,n«^ioii-urcgma/ 7 raiicuti t^uuru i^uregnid-itunDua^ g ocigiii iruiii punon ^^ponon-Dregma^ 9 External Palate Breadth 10 Cranial Height (basion-bregma) 11 Bizygomatic breadth (zygion-zygion) 12 Facial Length (basionprosthion) 13 Facial height 14 Nasal height (nasion-nasospinale) 15 Nasal breadth (alare-alare) 16 Mastoid length* Cranial Thickness Measurements 17 Midfrontal, at the midpoint of the nasion-bregma chord** 18 Temporal crest, parallel to midfrontal on the temporal crest** 19 Bregma** 20 Midparietal, at the intersection of horizontal and vertical lines that divide the parietal into quadrants** 21 Midpoint of the anterior-superior quadrant of the parietal*** 22 Midpoint of the posterior-superior quadrant of the parietal*** 23 jk/^^i 1 r?i* Lambda*** *Giles and Hiot 1963; **Gauld 1994; ***Bass 1995

PAGE 35

Figure 1. Ectocranial Measurements. Measurement numbers correspond to Table 1

PAGE 36

Postcranial Measnrements The clavicles in the autopsy sample were measured for length with sliding calipers, marked at midshaft and sectioned with a Stryker® saw. Since the bone was still in situ some soft tissue was included in the length measurement. The amount of soft tissue was believed negligible since the blades of the calipers allowed for extreme compression of soft tissue. Cortical thickness was measured at the superior-, inferior-, anteriorand posteriormost aspects of the cortex at midshaft. Superior-inferior and anterior-posterior diameters were also taken in both samples, by placing the blades of the sliding caliper perpendicular to the curvature (if any) of the shaft at midpoint, in order to obtain the minimal distance. The clavicle measurements are sunmiarized in Table 3. Table 3. Clavicle Measnrements. Variable No. Variable Name 1 Maximum length* 2 Anterior-posterior diameter at midshaft** 3 Superior-inferior diameter at midshaft** Cortical thickness at midshaft from anatomical position** 4 Anterior Cortex 5 Posterior Cortex 6 Superior Cortex 7 Inferior Cortex *Bass 1995, ** See Text To obtain cortical thickness the Terry clavicles were placed directly on the radiographic plate in order that refraction not alter the cortical measurements. To test this methodology five clavicles from biological specimens were radiographed and then sectioned at midshaft with a hacksaw. These test clavicles were filmed at 65 kvp for 135 minutes on unscreened cassettes, using Kodak EM-1 Manmiography film. The results of a one-sample t-test (a=0.05, df=4) examining the difference in thickness measurements showed that the null hypothesis (Ho: \i =0) could not be rejected for 3 of the 4 cortical thickness measurements. These results, summarized in Table 4, were considered sufficiently satisfactory to proceed with the radiographic protocol. For the Teny clavicles the same fihn as noted above was used, at 60 KVP for 4 seconds on a screened cassette. The clavicles

PAGE 37

26 were filmed in superior-inferior and anterior-posterior views so that cortical thickness could be measured on the same axes as in the autopsy sample. Measurements on the radiographs were taken with GFM sliding calipers. Maximum length of the clavicles was measured on the osteometric board. Table 4. Results of the One-Sample t-test for Sectioned and Radiographic Clavicle Measurements Variable t df Sig. (2-tailed) Mean Difference DIFFANT 1.812 4 0.144 0.3640 DlFFPOSr -6.498 4 0.003 -0.1260 DIFFSUP -1.570 4 0.192 -0.2740 DIFFINF 0.066 4 0.950 0.01200 • ..i^ .^iwi^u^w >^v>.>TV^wi 1 «^^jyV/\^U T \^ OUIVIIUI, pUSl^^IlUl, aUp^llVJI, and inferior measurements on the cortical bone and their corresponding measurements on the radiograph. Estimation of Lean Body Weight The autopsy sample presented a special situation in that it was possible to take skinfold measurements that could be used to estimate fat mass, and by subtraction, lean body mass. Martin et al. (1985) questioned the reliability of skinfold measurements for estimating fat mass due to issues such as tissue compressibility, water content of subcutaneous fat, and the variable proportion of subcutaneous to deep body fat. They only support their use in conjunction with other anthropometrics like height and body weight. Shepherd (1991), while noting these and other potential sources of error, concludes that skinfold prediction of body fat can have less error than estimates from densitometry. Skinfold measurements are the most accessible means of estimating fat mass in the autopsy setting, since more conclusive techniques (e.g. densitometry) require excessive manipulation of the deceased. Skinfold measurements were taken in millimeters with a Lange skinfold caliper at the triceps and subscapular location for each individual. Three skinfold readings were taken at each site and the average of the three reported as the final skinfold measurement With living subjects hydration, or the lack thereof, is a major influence on skinfold measurements. Although cadavers do not lose moisture through respiration and sweating as

PAGE 38

27 in living subjects, they do lose moisture through traumatic and postmortem fluid loss, and from evaporation in a ventilated refrigerated environment. The effect of fluid loss on skinfold measurements in cadavers, at least at the triceps and subscapular sites, is a balance between dehydration from air movement and rehydration from lividity. (Lividity is the subtle (i.e., not an edema) pooling of tissue fluids due to the influence of gravity, such that uncompressed skin takes on a reddened or livid appearance.) On a supine body lividity is most apparent on posterior limb surfaces and the back itself. Lividity may more than compensate for any type of fluid loss including exsanguination, as remaining blood, lymph and cellular fluids stay in the posterior tissues. Also, cadavers are usually kept covered until autopsy, which would impede surface dehydration. Aiello and Wood (1994) allowed no more than 5 days for posthumous weighing in order to prevent effects of dehydration, but from personal observation this researcher noted that a draped body in a refrigerated environment lost lifelike skin texture in three days. Therefore subjects were avoided if they had been stored more than 3 days. In addition the skinfolds were compared to normative standards for Americans (Frisancho 1990) to determine if the skinfolds produced weight classifications similar to those predicted from height. The autopsy and skeletal samples represent two data sets. The autopsy sample utilizes nine ectocranial, seven vault thickness, seven clavicle, and various organ measurements, in addition to the general statistics like weight, stature, and age. The skeletal sample is similar, but has seven additional ectocranial and no organ measurements. In the data analysis the intersection of the two samples is used as the Combined sample. Data Analysis All English measurements taken during data collection were automatically converted to metric equivalents within the data collection spreadsheet program Microsoft Excel 98. Subsequent statistical analyses involving descriptive statistics (Tables 4-6), correlation, Unear regression, and principal component analysis (PCA) were conducted with SPSS 10.

PAGE 39

28 Data analysis in all samples included production of sample descriptives and calculation of Pearson's correlation coefficient between body weight and all applicable cranial and clavicle measurements. To investigate the role of skeletal robusticity in cranial thickness, the correlation between cranial thickness and clavicular cortical thickness was calculated in the three samples. The Cranial Index was employed in correlation analysis with the vault thickness variables for the skeletal sample only, since the cranial height measurement (porion-bregma) used in the autopsy sample was not applicable. Regression analysis was reserved either for variables showing strong relationships to body weight, or when the regression coefficients were needed for comparison to those in the literature. I considered a strong relationship to be a correlation coefficient as high as the poorest performing postcranial measurement in modem humans (i.e. r = 0.69 for radius circumference, Hartwig-Scherer 1994). Previous studies (Gauld 1992, Aiello and Wood 1994) have implied that body weight and cranial measurement data do not meet requirements of normality and homoscedasticity, and that logarithmic transformation of the variables is necessary before regression analysis. Transformation of the variables was justified by an examination of the plot of the residuals of the non-transformed data. In studies of allometry there has been some debate over the correct type of linefitting model to use, witii each model having its benefits. The tiaditional Model 1 (Sokal and Rohlf 1995) approach, Least Squares (LS) regression, assumes that the independent variable is sampled without error from its population. In reality this assumption cannot be met in most allometiic studies where botii independent and dependent variables are measured quantities. The Least Squares technique also has three methods of correction of bias that is intixxiuced into data when log transformed variables are rehimed to their original states. Least Squares has been considered tiie better technique simply for predicting one variable from anoUier (Smitii 1994), which agrees witii tiie goal of tiiis shidy, to predict body weight from cranial variables. Model II (Sokal and Rohlf 1995) techniques, e.g..

PAGE 40

29 Reduced Major Axis and Major Axis, do not have the bias corrections, but they also do not assume that either of the variables are sampled without error. The Model II approach is also best when it is desirable to violate the restriction against extrapolating predictions beyond the range of the dependent variable. The LS technique, with its bias corrections and strength as a predictive tool, seems more appropriate for this study. Also, if the coefficients of determination are high, then the results of LS and RMA or MA analysis will be similar. LS will be the primary line fitting technique used, while RMA and MA will be conducted secondarily for comparison. Three criteria will be used to determine if the 22 cranial variables are actually suitable for predicting body weight. A high correlation coefficient (above 0.69) is the first requirement. The second criterion is a low standard error of the estimate (SEE), which indicates the accuracy with which the regression equation depicts the strength of the relationship between dependent and independent variable. A low SEE also indicates that the confidence bands around the regression line will be relatively narrow. Finally I will examine the percentage prediction error (PE), which in low values indicates good predictive ability of the equation. I explored a chronology issue generated by the sample design. A gap of several generations exists between the Terry specimens and those from autopsy. Two-tailed t-tests were conducted to determine if there were significant differences in means between the two samples for the various anthropometric and skeletal measurements. Previous work (Warren et al. 2000) comparing recent forensic and early 19*^ century skeletal samples (Terry and Hamann-Todd collections) had suggested a secular change in cranial measurements, at least as concerned measurements used to estimate ancestry. As many of those same measurements (and possibly crania) are employed in this study, it was interesting to review differences in mean measurements and cranial indices between the two samples. To produce a meaningful reduction of the many variables principal component analysis (PCA) was conducted, for the three samples. PCA is a large sample size procedure

PAGE 41

30 (at least 5 subjects per variable; Hatcher and Stepanski 1994), so to accommodate this restriction subsets of the variables were analyzed. In all PCA procedures non-log transformed data were used, and only the correlation matrix was analyzed. Retention of principal components was based on a combination of four criteria: 1 . Egenvalues greater than one for each component 2. Components should account for a reasonable (e.g. > 10%) proportion of the variance represented by all the components 3. Components should be relatively closely grouped on a scree plot 4. Components should be interpretable A component was interpretable if it had at least three variables with significant loadings, and these variables measured the same construct (e.g. postbregmatic cranial thickness). Different components should have different highly loading variables measuring different constructs. In rotation the components should demonstrate "simple structure" (Hatcher and Stepanski 1994). Simple structure indicates that the rotated (varimax rotation, which produces uncorrelated components) factor pattem shows at least three high loading variables on a component while the remaining variables have near zero loadings. Also the individual rotated components must have variables with a dichotomous loading pattem (i.e. very high and very low). Strength of loading is subject to interpretation since in the social sciences a loading of 0.4 is sufficient (Hatcher and Stepanski 1994), but several of the variables in this study show loadings of 0.7 and higher. My previous experience with skeletal measurements in PCA causes me to consider a loading of 0.7 and greater to be high, 0.5 and 0.6 moderate, and values below 0.5 are low.

PAGE 42

RESULTS The Combined Sample: Descriptive Statistics In the interest of treating the sample as simply a collection of male modem humans the autopsy and skeletal data were combined into one large sample (n = 147). Only the variables in common to the two samples were used. The somatic variables (Table 5) are shown first, followed by the cranial and clavicular measurements for the combined sample (tables 6 and 7). Of these, 47 are autopsy subjects and 100 are skeletal. Table 5. Descriptive Statistics for Somatic Variables in the Combined Sample. N Minimum Maximum Mean Standard Deviation Age (years) 147 18 87 54.56 15.24 Weight (kg) 147 31.00 184.1 64.63 26.384 Body length (cm) 146 152.40 199.40 172.83 9.08 The lightest individual (31kg) belonged to the skeletal sample and the heaviest to the autopsy sample. Nearly half (42) of the skeletal sample were below 45 kg, while obesity was the most common extreme in the autopsy sample. I examine the low body weights in the skeletal sample in the correlation analyses for the skeletal sample, while fatness in the autopsy sample is addressed in the skinfold analysis. Descriptive statistics for the cranial and clavicular measurements in the combined sample follow. The ectocranial measurements (Table 6) do not deviate from what is expected in a human sample. Combining four samples from Howells (1973) that would most likely resemble this American sample in ancestry created a comparison sample. The means of analogous measurements were averaged for three European (Norse, Zalavar, and Berg) and one West African (Dogon) samples (Table 7). Howells used a large variety of measurements in his study, several of which were also used in this study. He also provided 31

PAGE 43

32 detailed descriptions of how the measurements were taken, so it was quite simple to verify that the measurements he used were functionally identical to the ones in this study. Table 6. Descriptive Statistics for Cranial Variables of the Combined Sample. Variable Name XT N Mimmum Maximum Mean Standard Deviation Maximum Length 146 166 207 186.40 7.59 Maximum Breadth 147 127 160 142.63 5.93 Minimum Frontal Breadth 144 86 120 98.99 5.76 Biauricular Breadth 147 109 136 123.29 5.33 Height from Porion 147 111 150 125.71 7.25 Cranial Base Length 136 91 121 ini c5 lUl.iiJ J.JO Facial Height 129 57 96 71.05 7.15 Frontal Chord 147 98 126 113.58 5.29 Parietal Chord 14/ 99 136 116.19 7.08 Midfrontal 146 3 12 6.40 1.73 Temporal Crest 145 2 9 5.53 1.43 Bregma 14/ 2 10 6.57 1.42 Midparietal 146 2 11 5.69 1.68 Anterior Parietal 147 3 11 6.05 1.52 Posterior Parietal 146 3 11 6.34 1.72 Lambda 147 3 16 7.71 2.17 Table 7. Comparison of Shared Ectocranial Measurements in Howells' "Mixed Race" Sample and the Combined Sample Measurement Howell Mean Combined Sample Mean t Sig. 2-tailed Mean Difference Maximum Length 182.97 186.40 5.47 0.00 3.43 Maximum Breadth 142.04 142.63 1.20 0.23 .59 Biauricular Breadth 122.79 123.29 1.13 0.26 .50 Cranial Base Length 100.06 101.83 3.86 0.00 1.77 Facial Height 67.54 71.05 5.58 0.00 3.51 Frontal Chord 111.73 113.58 4.24 0.00 1.85 Parietal Chord 113.04 116.19 5.40 0.00 3.15 The combined sample is not broadly deviant from what one would expect for modem humans, although the one-sample t-test indicates several significant differences (Table 7). The hypothesis that a mean measurement from Howells such as maximum length is drawn from the same population as this American sample must be rejected. Table 7 shows that this hypothesis should be accepted for only two variables, maximum breadth and biauricular breadth. The lack of similarity between samples is most likely due to secular changes caused by the distance of years, gene flow, and nutritional status tiiat

PAGE 44

33 separates the two populations. The alternative is that Howells' collection of Europeans and West Africans bear little resemblance to this sample. This is unlikely, especially since a similar degree of difference is seen later between the Terry and autopsy samples (Table 30). Comparative clavicle measurements are more difficult to find, but again the combined sample measurements are not atypical for a sample of modem humans. The University of Tennessee Forensic Database (FORDISC 1996) contains mean values for clavicle length, AP and SI diameter. This database is composed of modem (late 20* century and later) forensic examinations in the United States. Table 8. Descriptive Statistics for Clavicle Variables in the Combined Sample N Minimum Maximum Mean Std. Deviation FORDISC Qavicle Length 147 119.0 184.0 154.25 10.36 158.18 AP Diameter 147 8 110 13.31 8.29 12.70 SI Diameter 147 1 18 11.60 2.55 11.14 Anterior Cortical Thickness 142 0.5 5.0 2.09 1.01 n/a Posterior Cortical Thickness 146 1 6 2.57 0.81 n/a Superior Cortical Thickness 147 1 5 2.36 0.85 n/a Inferior Cortical Thickness 138 1 7 2.26 1.20 n/a Combined Sample: Relationships Between Variables Tables 9 and 10 summarize the covariance relationships between body weight and the somatic and skeletal measurements. Table 9 reports stature and the clavicular measurements, while 10 reports the cranial measurements. The clavicle measurements are strongly correlated with body weight, with anteriorposterior diameter of the clavicle being the only exception (Table 9). While highly significant, the r-values range from low to moderate (about 0.3 to 0.6), with the highest being 0.667 for inferior cortical thickness. Many significant conelations exist between body weight and several of the ectocranial variables in the combined sample (Table 10) with the variable height from porion having the highest r-value of 0.628. The next highest correlation coefficients (of about 0.4)

PAGE 45

34 are for cranial base length, minimum frontal breadth, and maximum length. In contrast the cranial thickness variables show few significant correlations and uniformly low r-values. Temporal crest has the highest correlation to body weight with an r-value of 0.289. Table 9. Combined Sample Correlation CoefHcients Between Body Weight and the Somatic Variables (including Clavicle) Measurement N r Significance Body Length cm 146 0.559** 0.000 Qavicle Length 147 0.307** 0.000 AP Diameter 147 -0.081 0.331 SI Diameter 147 0.562** 0.000 Anterior Cortical Thickness 142 0.508** 0.000 Posterior Cortical Thickness 146 0.348** 0.000 Superior Cortical Thickness 147 0.581** 0.000 Inferior Cortical Thickness 138 0.667** 0.000 p < 0.05; ** p < 0.01 Table 10. Combined Sample Correlation Coefficients Between Body Weisht and the Cranial Variables Measurement N r Significance Maximum Length 146 0.418** 0.000 Maximum Breadth 147 0.152 0.067 Minimum Frontal Breadth 144 0.430** 0.000 Biauricular Breadth 147 0.189* 0.022 Height from Porion 147 0.628** 0.000 Cranial Base Length 136 0.490** 0.000 Facial Height 129 0.381** 0.000 Frontal Chord 147 0.286** 0.000 Panetal Chord 147 0.057 0.492 Midfrontal 146 0.070 0.402 Temporal Crest 145 0.289** 0.000 Bregma 147 0.133 0.109 Midparietal 146 0.164* 0.048 Anterior Parietal 147 0.163* 0.049 Posterior Parietal 146 0.031 0.714 Lambda 147 -0.080 1 0.338 * p < 0.05; ** p < 0.01 Correlation coefficients between the clavicular cortical thicknesses and the cranial thickness variables (Table 1 1) do not support the hypothesis that cranial thickness is a function of skeletal robusticity. The resulting trend is for a lack of significant correlation. Table 1 1 shows only three significant r-values: Cranial thickness at lambda has a weakly negative correlation to anterior cortical thickness of the clavicle, and the temporal crest

PAGE 46

35 measurement shows weakly positive correlations to both superior and inferior cortical thickness variables. These results support rejection of the robusticity hypothesis, as there is no pattern of relatedness between cranial thickness and cortical thickness in the clavicle. Table 11. Combined Sample Correlation Coefficients Between the Clavicle Cortical Thickness and Cranial Thickness Variables Measurement Anterior Cortical Thickness Posterior Cortical Thickness N r Significance N r Significance Midfrontal 141 -0.020 0.813 145 0.005 0.954 Temporal Crest 140 0.101 0.237 144 0.033 0.698 Bregma 142 0.010 0.909 146 -0.021 0.802 Midparietal 141 0.045 0.595 145 0.106 0.205 Anterior Parietal 142 -0.036 0.671 146 -0.030 0.718 Posterior Parietal 141 -0.035 0.680 145 -0.006 0.944 Lambda 142 -0.190 0.024* 146 0.010 0.903 Measurement Supe rior Cortical Thickness Inferior Cortical Thickness N r Significance N r Significance Midfrontal 146 0.037 0.656 137 -0.010 0.904 Temporal Crest 145 0.208* 0.012 136 0.227** 0.008 Bregma 147 -0.024 0.772 138 0.069 0.421 Midparietal 146 0.102 0.220 137 0.120 0.164 Anterior Parietal 147 0.095 0.253 138 0.076 0.377 Posterior Parietal 146 -0.030 0.723 137 0.044 0.612 Lambda 147 -0.158 0.056 138 -0.132 0.122 Significant relationships between the individual ectocranial variables and the vault thickness variables are more common (Table 12). Correlations coefficients are all positive but lower than 0.400. Maximum length is the only variable to be significantly correlated to all of the ectocranial measurements in the combined sample, although the relationship to lambda is not as strong as with the other thickness measurements. Height from porion and frontal chord are significantly correlated to all of the cranial thickness variables but lambda, although the latter ectocranial variable is not grossly out of significance (p = 0.068). The behavior of frontal chord suggests its correlation to the thickness variables is a reflection of that for maximum length, as the two ectocranial variables both cover sagittal vectors. Following this pattem it is notable that parietal chord is also only slighdy out of significance

PAGE 47

36 for three thickness variables. Cranial base length, also a sagittal measurement, follows essentially the same pattern, although it greatly lacks significance to lambda. Table 12. Combined Sample Correlation Coefficients Between the Ectocranial and Cranial Thickness Variables Measurement Maximum Length Maximum Breadth N r Significance N r Significance Midfrontal 145 0.393*=' 0.000 146 0.157 0.059 Temporal Crest 144 0.354* =< 0.000 145 0.141 0.090 Bregma 146 0.355*=! 0.000 147 0.047 0.569 Midparietal 145 0.342*=* 0.000 146 0.226*=* 0.006 Anterior Parietal 146 0.372*=* 0.000 147 0.276*=* 0.001 Posterior Parietal 145 0.331*=* 0.000 146 0.169* 0.042 Lambda 146 0.185* 0.025 147 0.012 0.890 Measurement Minimum Frontal Breadth Biauricu ar Breadth N r Significance N r Significance Vlidfrontal 143 0.175* 0.037 146 0.222** 0.007 Temporal Crest 142 0.276** 0.001 145 0.281** 0.001 Bregma 144 0.144 0.086 147 0.166* 0.044 Vlidparietal 143 0.280** 0.001 146 0.193* 0.020 Anterior Parietal 144 0.247** 0.003 147 0.228** 0.006 'osterior Parietal 143 0.132 0.116 146 0.113 0.173 Lambda 144 -0.001 0.992 147 0.068 0.415 Measurement Height from Porion Cranial B ase Length N r Significance N r Significance Vlidfrontal 146 0.201* 0.015 135 0.253** 0.003 emporal Crest 145 0.262** 0.001 134 0.298** 0.000 Bregma 147 0.220** 0.007 136 0.249** 0.003 Vlidparietal 146 0.296** 0.000 135 0.209* 0.015 Anterior Parietal 147 0.303** 0.000 136 0.168 0.051 Posterior Parietal 146 0.181* 0.029 135 0.177* 0.040 Lambda 147 0.041 0.618 136 0.140 0.103 Measurement Facial Height Frontal Chord N r Significance N r Significance Midfrontal 129 0.233** 0.008 146 0.289** 0.000 Temporal Crest 129 0.348** 0.000 145 0.184* 0.026 Bregma 129 0.095 0.282 147 0.301** 0.000 Midparietal 128 0.181* 0.041 146 0.226** 0.006 Anterior Parietal 129 0.112 0.206 147 0.387** 0.000 Posterior Parietal 128 0.173 0.051 146 0.229** 0.005 Lambda 129 0.039 0.658 147 0.151 0.068

PAGE 48

37 Table 12. Continued, Measurement Parieta Chord N r Significance Midf rental 146 0.182* 0.028 Temporal Crest 145 0.147 0.077 Bregma 147 0.299** 0.000 Midparietal 146 0.148 0.075 Anterior Parietal 147 0.161 0.052 Posterior Parietal 146 0.253** 0.002 Lambda 147 0.267** 0.001 The thickness variables are well correlated amongst themselves (data not shown), so in consideration of this average vault thickness variables were calculated. Anterior thickness, posterior thickness and average vault thickness were correlated to the ectocranial measurements (Table 13). The ectocranial variables are uniformly positively correlated with vault thickness. Maximum length is conspicuous for having the highest r-value in all thickness areas, indicating that increases in vault length are associated with moderate increases in vault thickness. In terms of average vault thickness, the highest correlation coefficients are for variables following cranial length, (maximum length. Frontal and parietal chords), followed by vault height measurements (height from porion, cranial base length). Cranial breadth measurements are the most pooriy correlated, along with the single facial measurement facial height. These data support Nawrocki's conclusion that larger vaults are associated with greater vault thickness. To assess the effect of shape on vault thickness the cranial index and the vault thickness measurements were correlated (Table 14), but they showed few significant relationships, all of which are negative. Measurements at midfrontal and bregma indicate that as the cranial index increases thickness decreases. That is, as the skull approaches a spherical shape vault thickness at these midline points decreases. The cranial breadth to length categories for the cranial index are correlated to the vault thickness measurements in Table 15. The narrow, dolichocranic skulls lack significant correlation to any vault thickness measurements, but as the skulls become increasingly

PAGE 49

38 Table 13. Combined Sample Correlation Coefficients between Ectocranial and Average Thickness Variables Measurement Anterior Thickness Posterior Thickness Average Vault Thickness n r Sign. n r Sign. n r Sign. Maximum Length 0.000 144 0.379** 0.000 146 0.443** 0.000 Maximum Breadth 144 0.138 0.100 145 0.195* 0.19 147 0.198* 0.016 Minimum Frontal Breadth 141 0.258** 0.002 142 0.188* 0.025 144 0.232** 0.005 Biauricular Breadth 144 0.305** 0.000 145 0.175* 0.035 147 0.236** 0.004 Height from Porion 144 0.291** 0.000 145 0.240** 0.004 147 0.281** 0.001 Cranial Base Length 133 0.346** 0.000 134 0.211* 0.015 136 0.280** 0.001 Facial Height 129 0.288** 0.001 127 0.149 0.094 129 0.225* 0.011 Frontal Chord 144 0.339** 0.000 145 0.300** 0.000 147 0.339** 0.000 Panetal Chord 144 0.267** 0.001 145 0.268** 0.001 147 0.286** 0.000 Table 14. Combined Sample Correlation Coefficients Between the Cranial Index and Cranial Thickness Variables Measurement Cranial Index N r Significance Mi df rental 145 -0.180* 0.030 Temporal Crest 144 -0.158 0.058 Bregma 146 -0.247** 0.003 Midparietal 145 -0.089 0.289 Anterior Parietal 146 -0.080 0.340 Posterior Parietal 145 -0.128 0.125 Lambda 146 -0.145 0.081 broad for their lengths significant correlations appear. The skull of average breadth for its length (i.e., mesocranic) shows significant positive correlations to the anterior parietal measurement The brachycranic skulls are significanUy correlated to decreases at the posterior parietal. Although consistent pattern is lacking in the location of significant correlation, it is interesting that the rounder brachycranic skulls show mostly negative correlation to vault thickness, since spherical shapes are better at resisting compression. Nawrocki (1992) also found reduced vault thickness with reduced vault length. Coirclation between average vault thickness and the cranial index categories is non-significant, but negative for the brachycranic skulls (Table 16).

PAGE 50

39 Table 15. Combined Sample Correlation Coefficients Between the Cranial Index Shape Categories and the Cranial Thickness Variables Categorv Dolichocrany Mesocrany Measurement N r Significance N r Significance Midfrontal 44 0.104 0.500 73 0.157 0.186 Temporal Crest 43 0.234 0.132 73 0.112 0.347 Bregma 44 0.032 0.835 74 0.083 0.482 Midparietal 43 -0.018 0.908 73 0.140 0.239 Anterior Parietal 44 -0.037 0.811 74 0.296* 0.011 Posterior Parietal 44 -0.032 0.834 73 0.181 0.125 Lambda 44 0.151 0.329 74 0.071 0.546 Categor) Brach^ I'crany Measurement N r Significance Midfrontal 26 -0.041 0.842 Temporal Crest 25 0.210 0.210 Bregma 26 -0.071 0.732 Midparietal 25 -0.205 0.326 Anterior Parietal 26 -0.270 0.182 Posterior Parietal 26 -0.391* 0.048 Lambda 26 -0.039 0.498 * p < 0.05; ** p < 0.01 Table 16. Combined Sample Correlation Coefficients between Cranial Index Categories and Average Vault Thickness Average Vault Thickness N r Significance Dolichocrany 44 0.084 0.586 Mesochrany 74 0.194 0.097 Brachycrany 26 -0.269 0.186 * p < 0.05; ** p < 0.0 Combined Sample: Principal Component Analysis The principal component analysis yielded more information on the relationships between variables in the sample (Tables 17 and 18). Two analyses were performed for the combined sample, the first using all of the variables except age (Table 17), and the second only the cranial variables and body weight (Table 18). Table 17 illustrates the number of components retained when using a large set of variables and the "eigenvalue 1" rule. The seven components shown have unrotated eigenvalues of 7.02, 3.81, 1.75, 1.47, 1.26, 1.07, and 1.03, respectively. Together these components account for about 69% of the variance in the data set. The eighth component

PAGE 51

40 has an eigenvalue close to 1.0, but it did not resemble any of the other criteria for retaining a component. In the combined sample some variables with moderate loadings also tended to load moderately on more than one component, which makes them good candidates for exclusion from future PCA with this data set (Hatcher and Stepanski 1994 p.477). Consider maximum length, which loads only moderately (about 0.5), but does so on components three and four. Variables with low loadings on more than one component also violate the "simple structure" rule. Other obviously discardable variables are minimum frontal breadth, height from porion, parietal chord and body length. Still others (see body weight) are suspicious for having low loadings on other components. Of the seven components shown in Table 17, only the first two have high loading variables occuring in an interpretable pattern (shown in bold). Component 1 is a clavicle cortical thickness factor, with strong loadings from all four cortical thickness measurements. Body weight also has its strongest loading, a moderate 0.648, on this component (but note its loadings on the other components). This makes sense, as the PCA is derived from the correlation matrix, and body weight has some of its highest correlations (around 0.5) with Uie clavicle cortical measurements. Only one cranial measurements was strongly correlated with body weight (height from porion, r = 0.628), and some moderate loading can be seen from this variable on this component. Yet height from porion is another candidate for removal for loading on more than one component. Component 2 is clearly the "cranial thickness" component, with high loadings from tiie parietal thickness measurements and moderate loadings from the more anterior thickness measurements. This component is particularly well defined because even the moderately loading components have miniscule loadings on the other components. Components 3 Uirough 7 are handicapped by either a lack of highly loading variables or insufficient number of variables. This limitation is due part to the number and variety of cranial variables available to the combined sample. In the second PCA, as in tiie first, only two components are interpretable (Table 18 in bold). The four components account for 63% of the total variance, having unrotated in

PAGE 52

41 Table 17. Combined Sample Principal Components Using All Variables (Except Age) 1^ in Tv^ n ^ n t \w 1^1 1 1 Lnj 1 1 CI 1 1 ] I : I : 5 ( t 7 Maximum Length 0.13< ) 0.29i 5 0.58^ 5 0.50 0.050^ 0.17' 1 0.155 Maximum Breadth -0.0801 [ 0.096! 5 O.lOi 5 0.15 0.87( 5 -0.0970.0827 Minimum Frontal Breadth 0.321 i 0.095f 5 0.5< 5 O.K ) 0.47] -0.078: ! 0.0627 Biauricular Breadth 0.0089{ 0.152 0.035S 3 0.055< i 0.81' 0.2 -0.2 Height from porion 0.502 0.14*; 0.29] 0.57f 0.19] 0.0020: 0.119 Cramal Base Length 0.315 0.161 0.65f ) 0.1 0.046S 0.24^ -0.122 Facial Height 0.323 0.125 0.743 I -0.0882 0.056 0.094e -0.0779 Frontal Chord 0.114 0.228 -0.0378 0.733 0.234 0.14( 0.186 Parietal Chord -0.124 0.138 0.502 0.288 0.033C 0.143 0.492 Midfrontal -0.12 0.569 0.226 0.325 0.0137 0.0131 -0.222 Temporal Crest 0.126 0.598 0.291 0.0341 0.134 0.111 -0.316 Bregma -0.0981 0.63 0.106 0.352 -0.0997 0.228 -0.0180 Midpanetal 0.176 0.813 0.0736 0.000143 0.186 -0.123 0.0277 Anterior Parietal 0.0550 0.802 -0.0280 0.255 0.197 -0.03 2-: 0 023(] Posterior Parietal 0.0281 0.853 0.0938 -0.0358 0.0584 -0.062e 0.24 Lambda -0.197 0.642 0.0573 -0.0912 -0.0798 0.302 0.218 Weight kg 0.648 0.0282 0.242 0.389 0.137 0.12^ -0.197 Stature cm 0.327 0.0721 0.263 0.402 0.00285 0.59' -0.15 Clavicle Length 0.255 0.0391 0.152 0.0562 0.0792 0.81J -0.00503 AP Diameter -0.0159 0.0554 -0.0385 0.0714 -0.0495 -0.070J 0.676 SI Diameter 0.515 0.0241 0.323 0.368 0.0125 -0.177 -0.344 Anterior Cortical Thickness 0.823 -0.0818 0.128 0.0706 0.00907 0.064f 0.113 Posterior Cortical Thickness 0.739 0.0486 -0.00682 -0.187 -0.0786 0.24f 0.0953 Superior Cortical Thickness 0.83 -0.00312 0.216 0.00187 0.0276 0.12f -0.126 nferior Cortical Thickness 01848" 0.0213 0.135 0.204 -0.0408 0.043J -0.0959

PAGE 53

42 eigenvalues of 5.62, 237, 1.50, and 1.22 respectively. Component 1 is easily interpretable as a primarily postbregmatic cranial thickness component. The second component is not easily defined. Cranial base length and facial height load highly (0.7) on the component, but the relationship between the two variables is not clear. The only otiier variable witii exclusively high loading on tiiis component is body weight. While it is a sound component in terms of loadings, its interpretation is unclear. The high degree of inter-relatedness of the variables in this stiidy has made it difficult to arrive at components describing a simple constiuct. Component 2 is the only evidence of body weight grouping with other variables. Keep in mind tiiat at the very least PCA is a technique to cause related variables to group togetiier when dealing with a large number of variables. Tables 17 and 18 show that when tile non-cranial variables are present body weight loads most highly witii tiiem, while contributing moderate loadings on to components 3 and 4. With tiie clavicle measurements removed, body weight groups witii two cranial measurements (cranial base lengtii and facial Table 18. Combined Sample Principal Components of the Cranial Variables and Body Weight Component 1 : 2 4 Maximum Length 0.29£ 0.S21 0.62i omiA Maximum Breadth 0.0812 0.0513 0.15' 0.861 Minimum Frontal Breadth 0.060S 0.613 0.238 0.415 Biauricular Breadth 0.152 0.12 -0.00887 0.808 Height from Porion 0.0769 0.574 0.532 0.219 Cranial Base Lengtii 0.172 0.7771 0.112 -0.0252 Facial Height 0.133 0.774 -0.000429 -0.0435 Frontal Chord 0.189 0.121 0.662 0.313 Parietal Chord 0.174 0.078S 0.75 -0.0564 Midfrontal 0^94 0.155 0.179 0.116 Temporal Crest 0.605 0.421 -0.106 0.139 Bregma 0.65S 0.0671 0.329 -0.0436 Midparietal 0.769 0.189 -0.00871 0.204 Anterior Parietal 0.775 0.0802 0.157 0.266 Posterior Parietal 0.841 0.041S 0.0906 0.0436 Lambda 0.669 -0.122 0.194 -0.154 Weight kg -0.0386 0.745 0.155 0.165

PAGE 54

43 height), which were also "distracted" by the clavicle measurements. The small number of variables in the combined sample obscures the pattem of relationship, but a look at the separate parts of the sample provides more information. Combined Sample: Regression Analysis Additional description of the covariance relationship between the cranial variables and body mass in this sample of modem humans required the use of line-fitting techniques to the scatter of data represented by the correlation coefficient. The r-values produced in this study, especially as seen in the log-transformed data, do not support the generation of predictive tools. As discussed in the protocol Least Squares (LS) regression is unfit for this type of data because neither the dependent nor independent variables were sampled without error. Aiello and Wood (1994) had found that if r-values are high any line-fitting technique will suffice, but those circumstances do not pertain here. This analysis is conducted in order to produce data more suitable for comparison to the literature. The LS regression coefficients for the combined sample are shown in Table 19. The correlation coefficients differ from those shown in Table 10 because these are from logio transformed data. The highest r-value is for height from porion, which also happens to have the lowest standard eiror. The next highest r-values are for cranial base length and minimum frontal breadth. These results are atypically low for the literature, where log transformed cranial variables have r-values ranging from 0.7 to 0.9, with the majority being above 0.95 (Aiello and Wood 1994, Gauld 1996). However, the equivalent of only one measurement is shared by this study and the work of Aiello and Wood ( 1994; their biporionic breadth). Biporionic breadth is one of their three best performing variables for predicting body weight in a hominoid sample with an r-value of 0.98. In contrast, biauricular breadth in this study yielded an r-value of 0. 18. These results are more similar to those of Nawrocki's (1992) human sample from the Terry collection. Similariy, the cranial thickness measurements have low r-values, the highest being 0.25 for temporal crest. Gauld (1996) reported r-values of 0.94 and higher for her hominoid sample at the

PAGE 55

44 midfrontal, bregma, and midparietal measurement sites. The deviation of regression results from the published literature will be addressed more fully in the discussion. A separate examination of the two parts of the combined sample follows. Table 19. Combined Sample Regression Coefficients For Body Weight Versus the Cranial Variables Variable N r Slope Intercent F 1 Ni on IVlaAllUUIll L^Uj^in 1 AA J.JO -6.26 0.15 26.07 0.000 IViaAlIllUIll DrcaUin U. 14 1 irk l.ZU A OA 0.16 2.72 0.102 Minimum Frontal Breadth 144 0.42 2.69 -3.59 0.14 30 80 0000 Biauricular Breadth 147 0.18 1.50 -1.36 0.16 4.69 0.032 Height from Fori on 147 0.63 4.06 -6.74 0.13 92.96 0.000 Cranial Base Length 136 0.48 3.30 -4.85 0.14 38.95 0.000 Facial Height 129 0.37 1.37 -Q.78 0.15 19.97 0.000 Frontal Chord 147 0.29 2.31 -2.96 0.15 13.49 0.000 Parietal Chord 147 0.06 0.37 1.01 0.16 0.55 0.460 Midfrontal 146 0.07 0.09 1.70 0.16 0.79 0.375 Temporal Crest 145 0.25 0.34 1.54 0.16 9.72 0.002 Bregma 147 0.14 0.23 1.60 0.16 3.02 0.084 Midparietal 146 0.15 0.18 1.65 0.16 3.27 0.073 Anterior Parietal 147 0.16 0.23 1.60 0.16 3.78 0.054 Posterior Parietal 146 0.02 0.02 1.76 0.16 0.06 0.806 Lambda 147 0.06 -0.08 1.85 0.16 0.57 0.450 Standard error of the estimate. The Antopsy and Skeletal Samples: Somatic Descriptive Statistics The general descriptive statistics for the two samples are shown in Table 20. The autopsy and skeletal samples differed distinctly in mean age, showing a 12-year difference. This is appropriate to the origins of the two samples, the former being drawn from autopsies conducted in the public interest, and the latter from subjects used in medical school anatomy laboratory. Anatomical subjects tend toward higher ages, while subjects in the autopsy come from both extremes of adult age. Also interesting is the difference in weight between the two samples, since the autopsy sample is nearly twice as heavy. Again this reflects the source of the two samples since elderiy adults tend to be below average weight, as well as the tendency towards obesity in recent American adults. Concern over obesity in the autopsy sample warranted the sampling of the tiiceps and subscapular skinfold thicknesses

PAGE 56

45 also represented in Table 20. These variables are discussed more thoroughly in the next section. Table 20. Descriptive Statistics for the Somatic Variables in the Autopsy and Skeletal Samples. N Minimum Maximum Mean Standard Deviation Autopsy age (years) 47 18 82 46.83 15.19 weight (kg) 47 52.70 184.1 91.07 26.81 body length (cm) 47 152.40 199.40 176.94 10.01 Triceps Skinfold 47 3.0 39.3 15.53 9.78 Subscapular Skinfold 47 6.0 43.3 21.30 9.63 Skeletal Age (years) 100 22 87 58.20 13.91 Weight (kg) 100 31.00 96.40 52.21 14.31 Body length (cm) 99 152.50 188.0 170.94 7.97 QuaUty of Skinfolds To address the concern that skinfold measurements from the recently deceased would not resemble a living sample, the autopsy sample skinfolds were compared to the trends recorded in Frisancho (1990). Table 21 shows the distribution of fatness categories encompassed by the autopsy sample, based on the fatness categories established by Frisancho (1990: Table IV.39). These categories are based on fahiess estimates from summed values of the tiiceps and subscapular skinfolds. There is no preponderance of specimens falling in the "Lean" or "Below Average" categories, as would be expected if dehydration were adversely affecting skinfold measurements. Conversely, the distiibution places nearly half the sample in tiie above average category, which is in agreement witii recent statistics (NHANES III) tiiat show more tiian half of American adults are obese. The quality of the skinfold measurements was assessed by comparing the weight classifications derived from the sum of the triceps and subscapular skinfolds to tiie weight classifications derived from weight for a given stature. The categories are shown below for each autopsy subject (Table 22). A chi-square analysis was conducted to determine if tiiere was any significant difference in the classifications based on the two techniques (Table 23).

PAGE 57

46 Table 21. Autopsy Sample Fatness Categories Based on Frisancho (1990) Table IV.39 for Males using Summed Triceps and Subscapular Skinfold Values Age Below Above Range^ N Lean Average Average Average'' 18-24 4 2 2 25 2 1 1 30 4 1 1 2 35 3 3 40 9 1 4 4 45 5 1 2 2 yj 6 4 2 55 3 1 2 60 6 3 3 65 2 2 70 1 1 75 1 1 80 1 1 Totals: 47 3 2 21 21 ^ Age ranges resemble Frisancho (1990) except for extending past 72 years. Above Average has been combined with the category "Excess Fat" which apoears in Table IV.39. The returned statistic, although not significant, was very close to 1, indicating that the two techniques returned nearly identical classifications for the autopsy subjects. Therefore, the skinfolds collected in the autopsy sample were not significantly altered by postmortem effects, and were included in subsequent calculations of lean body weight. Lean body mass was estimated by calculating a fat mass estimate according to the technique of Frisancho (1990), involving the following steps. Firet body density was estimated using equations adapted from Dumin and Womersley (1974) for each skinfold and the sum of the two skinfolds (Table 24). Densities for the subjects from the autopsy sample whose ages fell outside the groups of the density equations were calculated with the nearest age group equation. The percent fat weight for each density was estimated using the equation % fat weight = [(4.95/Density) 4.50] *100 (Siri 1956). Each subject's percentage of fat weight was applied to his body weight to anive at an estimate of fat weight, which was then subtracted from the body weight to arrive at the lean body weight. The

PAGE 58

47 descriptive statistics for Percent Fat Weight and the new variable Lean Weight are found in Table 25. Table 22. Autopsy Sample Weight Classifications Based on Weight (kg) and Statnre (cm) and Skinfolds. Age Weignt vKg; Stature (cm) Weight Qassification Sum Triceps and Subscapular okiniolds Weight Classification 19 68 1 171 S /WCIagC ZU.UU Average 24 93 9 17Q 1 /vverage i,nign^ Above average 28 71 7 177 R 1 / / .o /werage Average 29 95 3 IRd "7 rwerage i^iow _> 1 f\ tin Below average (lean) 30 91 1 189 Q /werage Average 32 1543 100 ^ /\uove average 4U.UU Above average 32 183 8 1004 ADove average Above average 36 1 148 194 ^ t\\jo\c average CO A 1_ Above average 37 88.5 170 9 /\oove average /I/I fil Above average 37 86.2 177 7 /werage nign 40.0/ A IAbove average 40 75 8 177 R /werage A 1 /V^ 41. UU A 1 Above average 40 59 0 ijeiow average Zd.UO Average 40 87.6 /werage 1 /.UU Below average 40 90 3 17^ ^ /werage nign 4/.00 Above average 41 71 7 1^ Q /werage Average 41 113.0 189 ? rvDove average Average 43 109.3 175 ^ /\uove average Above average 44 74.9 180 ^ /werage OO AA ZZ.UU Average 44 1007 177 8 Above average 60.33 Above average 45 129.3 190 5 /\Dove average Above average 45 87.6 180 ^ /werage 27.33 Average 46 83 5 189 O Average 11.00 Below average (lean) 46 544 Below average 15.33 Average low 47 776 1^ Q Average 43.00 Above average 50 100 7 15?0 ^ Above average 33.33 Average 52 104 8 IR^i 7 loD. / Above average 37.67 Average high 54 52 6 i*^o O Below average 29.33 Average 54 1 1R4. 1 xo.*+ 1 o^ n Above average 65.67 Above average 54 83.5 177.8 Average 23.50 54 82.6 157.5 Above average 50.33 Above average 55 95.3 182.9 Average high 32.33 Average 56 114.3 182.9 Above average 45.33 Above average 59 159.3 181.6 Above average 72.33 Above average 60 56.3 154.9 Average low 17.67 Average low 60 95.7 174.0 Above average 44.67 Above average 60 86.2 177.8 Average 50.00 Above average 62 73.5 177.8 Average 22.67 Average 63 93.5 172.7 Above average 50.33 Above average 63 96.2 182.9 Average high 2333 Average 65 79.9 177.8 Average 25.00 Average

PAGE 59

Table 22. Continued. 68 83.5 174.0 Average 22.00 Average 74 57.6 162.6 Average low 12.00 Below average 79 80.8 165.1 Average high 49.67 Above average 82 53.5 172.7 Low 22.83 Average 32 79.4 177.8 Average 10.00 Below average (lean) 21 61.6 167.6 Average 21.00 Average 18 95.3 172.7 above average 38.67 Above average Based on Frisancho (1990) Table IV. 13 Table 23. Contingency Table of Weight Classifications based on Weight for Stature and Combined Triceps and Subscapular Skinfold Thicknesses Below Average Average Above Average Totals Weight for Stature^ 5 24 18 47 Skinfold Thickness'' 6 19 22 47 Totals 11 43 40 94 Pearson Chi-Square 1.046, df= 2, p = 0.593 Based on Frisancho (1990) Figures IV. 13^ and IV.39'' respectively. Table 24. Regression Equations for Estimating Body Density (D) from the Logarithm of Triceps and Subscapular Skinfold Measurements. Age (jroups | Males D = a (b*log triceps skinfold) 17-29 20-29 30-39 40-49 50-72 D = 1.1252 (0.0625 * log T) D= 1.1131 -(0.0530 * log T) D = 1.0834 (0.0361 * log T) D = 1.1041 (0.0609 * log T) D= 1.1041 -(0.0662* log T) D = a (b*log subscapular skinfold) 17-29 20-29 30-39 40-49 50-72 D= 1.1312 -(0.0670* log Sub) D = 1. 1360 (0.0700 * log Sub) D = 1.0978 (0.0416 * log Sub) D = 1.1246 -(0.0686* log Sub) D = 1.1334 (0.0760 * log Sub) D = a (b*log sum of triceps and subscapular skinfolds) 17-29 20-29 30-39 40-49 50-72 D= 1.1561 -(0.071 1 * log Sum) D = 1.1525 (0.0687 * log Sum) D= 1.1 165 -(0.0484* log Sum) D= 1.1519 -(0.0771 * log Sum) D = 1.1527 (0.0793 * log Sum) Adapted from Frisancho (1990) and Dumin and Womersley (1974).

PAGE 60

49 Table 25. Autopsy Sample Descriptive Statistics for Fat Weight Estimates and the New Variable Lean Weight N Minimum Maximum Mean Std. Deviation % Fat Weight Triceps 47 5.0 44.7 27.20 8.53 % Fat Weight Subscapular 47 10.9 40.0 26.35 7.15 Sum Triceps and Subscapular Skinfolds 47 10.00 81.33 36.84 17.89 % Fat Weight Sum 47 7.5 42.2 25.58 6.96 Average % Fat Weight 47 7.8 40.5 26.38 7.03 Fat Weight (kg) 47 7.5 61.0 24.86 11.92 Lean Weight 47 44.77 130.65 67.48 15.97 Skinfold measurements are in mm. Percent fat weight 'or each skinfold was calculated according to the procedure in Frisancho (1990). % Fat Weight = [(4.95/Density) 4.50] *100. Average % Fat Weight = (% Fat Weight Triceps + % Fat Weight Subscapular + % Fat Weight Sum)/3. An additional assessment was made to determine if the values returned as lean weight actually reclassified the autopsy subjects into lower weight for height categories than that based on unadjusted weight. Table 26 shows the reclassification of the autopsy sample, with the large chi-square statistic indicating that the two treatments produced significantly different results. Individuals who were already of below average weight were not treated for lean weight, but were passed into the adjusted weight pool. Thirty-one subjects were reclassified as below average weight At least 6 of the 3 1 were drawn directly from the above average category. Skinfolds tend to underpredict fatness in the extremely obese, so the reclassification process was performing better than expected. The three individuals remaining in the above average category had unadjusted weights of greater than 150 kg. Table 26. Weight Classifications Based on Unadjusted Weight for Heisht Versus Lean Weight for Height Below Average Average Above Average Total Unadjusted Weight vs. Stature 5 24 18 47 Lean Weight for Stature 35 9 3 47 Totals 40 33 21 94 Pearson Chi-Square 41.683 df =2 p= 0.000 Based on Frisancho (1990) Figure IV. 13.

PAGE 61

50 Autopsy and Skeletal Samples: Cranial Descriptives The descriptive statistics for the cranial and clavicular variables used in the two samples are shown in Tables 27-29. The skeletal sample has a larger number (n =16) of ectocranial variables because these defleshed crania provided greater measurement access. Table 27. Autopsy Sample Descriptive Statistics for the Cranial Variables Variable Name N Minimum Maximum Mean Standard Deviation Maximum Length 46 174 206 189.87 7.43 Maximum Breadth 47 128 160 142.62 5.67 Minimum Frontal Breadth 45 93 120 102.82 5.63 Biauricular Breadth 47 109 136 124.32 6.05 Height from Porion 47 117 150 132.30 6.71 Cranial Base Length 37 91 121 104.81 6.90 Facial Height 37 63 96 75.11 8.04 Frontal Chord 47 105 124 115.31 5.28 Parietal Chord 47 105 135 116.49 7.51 Midfrontal 46 3 9 6.60 1.66 Temporal Crest 46 2 9 6.11 1.57 Bregma 47 4 9 6.56 1.13 Midparietal 47 3 11 5.98 1.91 Anterior Parietal 47 3 10 6.20 1.57 Posterior Parietal 46 3 11 6.11 1.76 Lambda 47 3 1 11 6.84 1.60 The trend of higher means in the autopsy sample is very obvious in these descriptive statistics. Secular changes between the population of the Teny collection and recent forensic cases have been documented (Jantz and MooreJansen 1988). Such change is hardly an issue when developing a sample for estimating body mass in fossil primates, especially since at least tens of thousands of years separate the fossil from its only available reference sample. Yet it is interesting to document significant difference between these two samples in order to understand the degree of difference encompassed by the combined sample. The results of the two-tailed t-tests for the cranial and clavicle variables are shown in Table 30, equal variance not assumed. Although the sample sizes are not equal, the t-test is robust to some violations, particularly when conducting a two-tailed test and when sample sizes are large (>30). An F test for equal variances was conducted, but Zar (1998) notes that this test is unreliable if the data deviate from normality. As a precaution the Mann-Whitney

PAGE 62

51 U was also calculated because if the data are actually in compliance with the assumptions of the t-test the MannWhitney U performs almost as well. Table 28. Skeletal Sample Descriptive Statistics of the Cranial Variables M rvunimum Maximum Mean btd. Deviation JLTlOAJlllUlil l_^il£'tll LKAJ loo zU / 1 OA O 1 7.15 Maximum RrppHfh IVA/ 1 97 1
PAGE 63

52 Table 29. Aatopsy and Skeletal Samples Descriptive Statistics for Clavicle Variables N Minimum Maximum Mean Standard Deviation Au topsy Sample Qavicle Length 47 119 184 157.57 11.60 AP Diameter 47 8 24 11.93 2.53 SI Diameter 47 9 17 14.04 2.09 Anterior Cortical Thickness 47 1.5 5 3.12 0.86 Posterior Cortical Thickness 46 1 6 3.15 0.91 Superior Cortical Thickness 47 2 5 3.29 0.73 Inferior Cortical Thickness 47 2 7 3.52 1.13 Skeletal Sample Qavicle Length 100 127.0 175.0 152.69 9.38 AP Diameter 100 9 no 13.97 9.85 SI Diameter 100 1 18 10.45 1.85 Anterior Cortical Thickness 95 .5 3.0 1.574 0.610 Posterior Cortical Thickness 100 1 4 2.29 0.59 Superior Cortical Thickness 100 1 4 1.93 0.48 Inferior Cortical Thickness 91 1 3 1.61 0.53 Measurements are in mm. Table 30. Autopsy and Skeletal Samples Two-Tailed t-test for the Cranial Variables VariaWe Name Autopsy Mean Skeletal Mean t Sig. 2tailed Mean Difference Mann Whitney L Z Sig. 2tailed Maximum Length 189.87 184.81 -3.87 0.000 5.06 1360.50 -3.96 0.000 Maximum Breadth 142.62 142.63 0.013 0.990 0.01 2324.00 -0.11 0.914 Minimum Frontal Breadth 102.82 97.24 -5.73 0.000 5.58 1006.00 -5.28 0.000 Biauricular Breadth 124.32 122.80 -1.50 0.137 1.52 1970.00 -1.58 0.114 Height from Porion 132.30 122.61 -8.78 0.000 9.69 534.00 -7.55 0.000 Cranial Base Length 104.81 100.72 -3.38 0.001 4.09 1087.00 -3.65 0.000 Facial Height 75.11 69.42 -3.88 0.000 5.69 990.00 -3.72 0.000 Frontal Chord 115.31 112.77 -2.74 0.007 2.54 1704.00 -2.69 0.007 Parietal Choixl 116.49 116.05 -0.34 0.735 0.44 2321.00 -0.12 0.904 Midfrcmtal 6.60 6.30 0.97 0.335 0.29 1967.50 -1.41 0.158 Temporal Crest 6.11 5.27 3.18 0.002 0.84 1468.50 -3.49 0.000 Bregma 6.56 6.58 -0.07 0.943 0.02 2304.00 -0.19 0.847 Midparietal 5.98 5.55 1.34 0.184 0.43 2022.50 -1.28 0.200 Anterior Parietal 6.20 5.98 0.81 0.420 0.22 2199.00 -0.64 0.526 Posterior Parietal 6.11 6.45 -1.08 0.281 0.34 2048.00 -1.07 0.284 Lambda | 6.84 8.12 -3.93 0.000 1.28 1546.00 -3.36 0.001

PAGE 64

53 Autopsy and Skeletal Samples: Relationships Between Variables As for the combined sample, Pearson's correlation coefficient was calculated to demonstrate the degree of covariance between body weight and the skeletal measurements (Tables 31-33). These calculations are for non-transformed data. Table 31. Autopsy and Skeletal Samples Correlation Coefficients Between Body Weight and the Somatic and Clavicle Measurements Measurement N r Significance Autopsy Stature cm 47 0.727** 0.000 Lean Weight kg 47 0.917** 0.000 Qavicle Length 47 0.199 0.181 AP Diameter 47 0.076 0.614 SI Diameter 47 0.340* 0.019 Anterior Cortical Thickness 47 -0.019 0.901 Posterior Cortical Thickness 46 -0.031 0.836 Superior Cortical Thickness 47 0.218 0.141 Inferior Cortical Thickness 47 0.379** 0.009 Skeletal Stature cm 99 0.275** 0.006 Qavicle Length 100 0.253* 0.011 AP Diameter 100 -0.020 0.845 SI Diameter 100 0.071 0.483 Anterior Cortical Thickness 95 0.080 0.440 Posterior Cortical Thickness 100 0.077 0.446 Superior Cortical Thickness 100 0.037 0.717 Inferior Cortical Thickness 91 0.105 0.321 * p < 0.05; ** p < 0.01 The two samples were similar in covariance between body weight and tiie noncranial variables (Table 31). In the autopsy sample body weight, lean body weight, and statiire were highly correlated. Pearson's r is especially high for tiie correlation between body weight and lean weight, which is expected since lean weight is derived directiy from body weight The clavicle measurements were largely disinterested in body weight, except for SI diameter and inferior cortical tiiickness (a component of SI Diameter). While tiie skeletal sample does share a significant correlation to stature, tiie covariance witii tiie clavicular measurements is only significant for clavicle lengtfi. None of tiie cortical measurements covary significantiy, probably due to tiie radiolucency of tfie cortices making them appear tiiinner than in the autopsy sample.

PAGE 65

54 There are no high correlations between body weight and the ectocranial measurements (Table 32) in either sample, although several were significant. Maximum length, maximum breadth, minimum frontal breadth, height from porion, and cranial base length all had significant but low rvalues in the autopsy sample. Several of the ectocranial measurements were significantly correlated with body length, but still with low r-values. The relationships in the skeletal sample between the ectocranial variables and body weight also yielded low r-values, but fewer significant values than in the autopsy sample, height from porion, frontal chord, cranial height, cranial base length, bizygomatic breadth, and mastoid length were all significantly correlated. Table 32. Autopsy and Skeletal Samples Correlation Coefficients Between Body Weight (kg) and the Ectocranial Variables Measurement N r Sig. (2-tailed) Autopsy Maximum Length 46 0.490** 0.000 Maximum Breadth 47 0.277 0.060 Minimum Frontal Breadth 45 0.304* 0.043 Biauricular Breadth 47 0.124 0.407 Height from Porion 47 0.445** 0.002 Cranial Base Length 37 0.473** 0.003 Facial Height 37 0.260 0.110 Frontal Chord 47 0.181 0.176 Parietal Chord 47 0.031 0.727 Skeletal Maximum Length 100 0.147 0.145 Maximum Breadth 100 0.179 0.074 Minimum Frontal Breadth 99 0.084 0.406 Biauricular Breadth 100 0.154 0.126 Height from Porion 100 0.252* 0.011 Cranial Base Length 99 0.265** 0.008 Facial Height 92 0.126 0.232 Frontal Chord 100 0.206* 0.040 Parietal Chord 100 0.074 0.463 Facial Length 100 0.082 0.418 Cranial Height 99 0.244* 0.015 Bizygomatic Breadth 92 0.248* 0.017 Nasal Breadth 100 0.036 0.720 Nasal Height 100 0.139 0.169 External Palate Breadth 100 -0.007 0.947 Mastoid Length 98 0.227* 0.025

PAGE 66

55 The correlation results for the skeletal sample were surprisingly different, so an attempt was made to determine the source of the difference by manipulating Uie sample. The first proposition was that the low body weight in some of tiie skeletal subjects was affecting the regression analysis. Following Nawrocki's (1992) metiiodology of setting a weight limit, die sample was sorted by weight and the lighter half removed, leaving 52 subjects witii weight greater than or equal to 50 kg. Correlation analysis with the heavier subjects did not yield consistentiy greater r-values, as indicated by tiie five variables shown in Table 33. Maximum breadth increased greatiy to a significant 03 1, but two variables lost their previous significance (height from porion and cranial base lengtii). A second manipulation involved removing all tiie subjects of African descent, since only three (6 %) of tiie 47 autopsy subjects were of African descent, versus 22% in the skeletal sample. This modification produced greater change (Table 33), bringing height from porion and cranial base length back into significant correlation with r-values closer to tiiose seen in the autopsy sample. This phenotypically homogenous sample has higher correlation coefficients tiian Uie heavier skeletal sample, but its larger sample size must also be conhibuting to the improved results. Yet tiie heavier sample has greater size tiian tiie autopsy sample but has fewer significant results. Table 33. Manipulated Skeletal Sample Correlation Coefficients for Select Variables Heavier Skeletal Sample European Skeletal Sample Variable n r Sign. n r Sign. Maximum Length 52 0.122 0.388 78 0.205 0.072 Maximum Breadth 51 0.307* 0.027 78 0.190 0.096 Minimum Frontal Breadth 51 0.117 0.410 77 0.023 0.845 Biauricular Breadth 51 0.248 0.076 78 0.268* 0.018 Height from Porion 51 0.219 0.119 78 0.302** 0.007 Cranial Base Length 50 0.275 0.051 77 0.318** 0.005 * p < 0.05; ** p < 0.01 Qoser examination of tiie weight distribution of the skeletal sample (Figure 3) shows tiiat most of the sample is of low weight for tiieir height, based on Frisancho's (1990) standards for males. Low weight is the minimum weight category and would be

PAGE 67

56 grounds for medical intervention in a living subject (Frisancho 1990), and in the skeletal sample includes males weighing from 30 to about 50 kg. With die low weight individuals excluded, the correlation coefficients (Table 34) more closely resemble those from the autopsy sample for unadjusted body weight (Table 31). Only cranial base lengtii in Table 33 shows a significant correlation, however, which likely stems from tiie insufficient sample size (n = 29), a factor that also could have contributed to the correlation coefficients generally being lower here than in the autopsy sample. In contrast, tiie larger sample size of tiie combined sample, witii its greater proportion of viable body weights, showed more significant correlations to the cranial variables (Table 10). 80 T 60 40 S c ibow iwrage below low Weight Class Figure 3. Skeletal Sample Weight for Height Class Distribution

PAGE 68

57 Table 34. Skeletal Sample Correlation Coefficients for Body Weight Versus Cranial Variables Excluding Low Weight Subjects Vanable N r Significance Maximum Length 29 0.308 0.104 Maximum Breadth 29 0.254 0.184 Minimum Frontal Breadth 29 0.193 0.315 Biauncular Breadth 29 0.231 0.227 Height irom Ponon 29 0.140 0.470 Cranial Base Length 29 0.496** 0.006 racial Height 26 0.116 0.574 rrontal Chord 29 0.176 0.361 ranetal Chord 29 0.152 0.430 Facial Length 29 -0.084 0.663 Cranial Height 29 0.217 0.258 Bizygomatic Breadth 27 0.135 0.503 Nasal Breadth 29 0.009 0.964 XT ITT * I, i Nasal Height 29 0.187 0.332 cAicnitu rdiaie OFeaoui oo zy -0.3 14 0.098 Mastoid Length 28 0.367 0.055 Midfrontal 29 -0.044 0.822 Temporal Crest 29 -0.062 -.750 Bregma 29 0.128 0.509 Midparietal 29 -0.245 0.201 Anterior Parietal 29 -0.068 0.726 Posterior Parietal 29 0.031 0.873 Lambda 29 0.033 0.866 * p < 0.05; ** p < 0.01 In contrast to the ectocranial variables, none of the thickness variables were significantly correlated to body weight in the autopsy sample, although in the skeletal sample one of the cranial thickness measurements was significantly correlated to body weight (Table 35). Lambda has a significant (p < 0.05) r-value of 0.216. At this point I retum to the three average vault thickness categories and their relationships to the ectocranial variables. The correlation coefficients for the seven additional variables used in the skeletal sample are shown in Table 36, and once again they are all positive. Cranial height (basion-bregma distance) has the highest r-value to average vault thickness (0.40), performing better than height from porion (r = 0.28) in the combined sample (Table 13). As seen previously in the combined sample the facial measurements produced only low significant correlation. Bizygomatic breadth is suiprising for a breadth

PAGE 69

58 measurement, since in the combined sample these performed poorly. The zygomatic arches bridge the vault near two major components of the masticatory musculature, so the higher rvalue may reflect the pathway of dissolution of masticatory forces. Table 35. Autopsy and Skeletal Samples Correlation Coefficients Between Body Weight (kg) and the Cranial Thickness Variables Measurement N r Significance Autopsy Mi df rental 46 -0.029 0.847 Temporal Crest 46 0.184 0.221 Bregma 47 0.235 0.112 Midparietal 47 0.215 0.177 Anterior Parietal 47 0.201 0.176 Posterior Parietal 46 0.184 0.222 Lambda 47 0.117 0.435 Skeletal Midfrontal 100 0.066 0.517 Temporal Crest 99 0.105 0.299 Bregma 100 0.192 0.056 Midparietal 99 0.009 0.927 Anterior Parietal 100 0.136 0.177 Posterior Parietal 100 0.091 0.365 Lambda 100 0.216* 0.031 * p < 0.05; ** p < 0.01 Table 36. Skeletal Sample Correlation Coefficients for Additional Ectocranial Variables versus Vault Thickness Averages Measurement Anterior Thickness Posterior Thickness Average Vault Thickness n r Sign. n r Sign. n r Sign. Facial Length 100 0.238* 0.017 100 0.245* 0.014 100 0.268** 0.007 Cranial Height 99 0.352** 0.000 99 0.366** 0.000 99 0.402** 0.000 Bizygomatic Breadth 92 0.345** 0.001 92 0.222* 0.033 92 0.298** 0.004 Nasal Breadth 100 0.048 0.633 100 0.137 0.175 100 0.114 0.258 Nasal Height 100 0.248* 0.013 100 0.034 0.740 100 0.130 0.196 External Palate Breadth 100 0.140 0.166 98 0.279** 0.005 100 0.247* 0.013 Mastoid Length 98 0.138 0.174 98 0.058 0.571 98 0.100 0.325 Autopsy and Skeletal Samples: Principal Component Analyses The PCAs for the autopsy and skeletal sample were conducted with the same protocol as the combined sample (painvise handling of missing cases, Varimax rotation).

PAGE 70

59 Only one analysis is provided for the autopsy sample (Table 37) because the small sample size (N=47) violates the minimal sample size rule (of thumb) of having three subjects per variable. Table 37 is shown to give an idea of how the variables are behaving although the data are insufficient. Five components are retained, but only two components are interpretable, as in the combined sample. The five components represent 71% of the variance in the data set and have unrotated eigenvalues of 5.45, 2.45, 1 .78, 1.45, and 1 .07 respectively. Component 1 is a cranial thickness component, with high loadings although not always simple loadings from the parietal sites. Temporal ci^st has a strong contribution as well. The second component is similar to the same component number in the second matrix for the combined sample (Table 17), as body weight loads with some cranial variables, but not in a simple fashion. Height from porion has high loadings on this factor which it did not in Table 17. If there were more variables this factor might be interpreted as a body size component, but currently it is undefinable. The greater number of cranial variables in the skeletal sample created the possibility of more informative principal component analyses. The first PCA for the skeletal sample used all the cranial and somatic variables (Table 38) and retained eight components based on the eigenvalue "1" rule. Only the first six components are shown here since the last two are uninteipretable. The eight components represent 66% of the variance in the data set The unrotated eigenvalues for the six components shown are 7.16, 3.76, 2.43, 2.10, 1.66, and 1.54 respectively. Component 1 is the cranial thickness factor and is most strongly supported by high loadings from the parietal measurements. The second component is defined by facial measurements (facial length, nasal breadth, and external palate breadth), although the two other variables that, intuitively, should also load here (facial height and nasal height) also contribute to factor seven. It is quite typical for variables to group onto their own vector when they measure the same direction (e.g., the breadth measurements of factor 3), or lie along the same line, as in this case. Facial length could be removed from analysis because it

PAGE 71

loads too highly on too many factors, but removing it will not cause nasal height to load onto Component 2. The strongly loading variables, as well as the moderate associations from the other facial measurements, defines Component 2 as a facial factor. Component Three is a cranial width factor and is supported by maximum breadth, biauricular breadth, bizygomatic breadth, and a small association with minimum frontal breadth. This latter variable has only moderate and low loadings throughout the component matrix, and is too well associated with the facial variables factor and the chord measurements of factor five. Component 4 seems to be the closest approximation of a body size factor. Its highest loading variable is stature, followed by clavicle length and body weight. These last two variables only have moderate loadings, and the interpretation of the factor is dependent on how little these variables load on to other components. Notice that cranial base length has its highest loading here too, an association we saw earlier in the PCA for the cranial variables in the combined sample (Table 17). As previously mentioned the fifth component Table 37. Aatopsy Sample Principal Components for the Cranial Variables Component < weight kg 0.1057^ 0.7108* 0.3052' o.om: -0.1 06e Maximum Length 0.3230] 0.6384: 0.2204( 0.1103: 0.3005e Maximum Breadth 0.1894: 0.29874 0.1640^ 0.752T 0.0784S Minimum Frontal Breadth 0.0062: 0.34181 0.5987] 0.4852' 0.0379^ Biauricular Breadth 0.08364 -0.0862 0.0833e 0.8243^ 0.01 162 Height from Porion 0.0343: 0.8647S -0.06ie 0.02201 0.1465^ Cramal Base Length 0.16071 0.32751 0.7484S 0.15644 -0.035S Facial Height 0.06221 0.0038] 0.9416: 0.0542^ 0.11001 Frontal Chord 0.28294 0.4441^ -0.547S 0.4167f 0.1119e Parietal Chord -0.0052 0.1610' 0.0474: 0.0313: 0.94731 Midfrontal 0.5167£ -0.14{ 0.01971 0.5299: -0.041 Temporal Crest 0.7392! -0.105( 0.21011 -0.046' 0.1009i Bregma 0.6806^ 0.1412< -0.081 0.286: 0.190ie Midparietal 0.7186S 0.2708S 0.2010^ 0.1113S -0.223f Anterior Parietal 0.66152 0.2424S -0.0773 0.51 16i -0.246: Inferior Parietal 0.88612 0.16531 0.0787' 0.07701 0.039S Lambda 0.7076S o.io6is| -0.1 lie 0.11731 -0.021i

PAGE 72

61 Table 38. Skeletal Sample Principal Components Using All Variables (Except Age) Component 1 2 3 4 f 6 Maximum Length 0.3465i 0.294' 0.06251 0.35872 0.50521 0.00518 Maximum Breadth -0.018S -0.: 0.78471 -0.142( 0.2816< -0.0652 Minimum Frontal Breadth 0.1325J 0.3318S 0.53553 0.00931 0.38133 -0.0743 Biauricular Breadth 0.14872 -0.070« 0.8695( 0.0784S -O.OS -0.1543 Height from Porion 0.32883 0.04141 0.2577J 0.3224 0.6405
PAGE 73

62 is associated with chord measurements of the vault in the sagittal plane. The loadings are only moderate but tend to be the highest loadings for these variables in this matrix, so the definition is clear. The last definable component has high loadings from three of the clavicle measiwements and is therefore a clavicle cortical thickness factor. In summary, this PCA using all variables from the skeletal sample yielded six definable components. These components were retained based on the eigenvalue "1" rule, definability, and association with at least three variables of either high or simply structured loadings. The components are: 1. Cranial Thickness 2. Facial Variables 3. Cranial Breadth 4. Body Size 5. Sagittal Chords of the Vault 6. Qavicle Cortical Thickness Components seven and eight are not defined because they both lack at least three high loading variables. An association between nasal height and facial height, two measurements which occur on the same line on the skull, is found in component seven. Lack of a third variable omits this factor from display. Factor eight has high loadings only from SI diameter of the clavicle. The second PCA excluded the clavicle variables (Table 39). Five components were interpretable (six passed the eigenvalue test). Unrotated eigenvalues are 6.97, 2.79, 2.26, 1.99, 1.54, and 1.28, respectively. The cumulative variance in the six components is 67%. In response to the reduction in variables many loadings have increased but the associations and interpretations of the components have not changed. The defined components are: 1. Cranial Thickness 2. Facial Variables 3. SagittalChordsof the Vault

PAGE 74

4. Cranial Breadth 5. Body Size Autopsy Sample: Regression Analysis Least Square regression lines were fit for the relationships between the cranial measurement variables and body mass. Regressions for the autopsy sample were conducted for unadjusted and adjusted (lean) body weight. All results are for logarithmically transformed (base 10) variables so the r-values will not resemble those shown previously. Table 39. Skeletal Sample Principal Components Using the Cranial Variables Component 1 2 3 A £ 6 Maximum Length 0.29001 0.29133 0.61703 l.lE-0* 0.23168 0.38049 Maximum Breadth -0.041£ -0.2m 0.3716< 0.74401 -0.1001 0.12441 Minimum Frontal Breadth 0.10415 0.29321 0.47012 0.50381 -0.0112 -0.0875 Biauricular Breadth 0.13278 -0.0905 0.0225<) 0.8957 0.11353 0.06077 Height from Porion 0.29188 0.0227S 0.70858 0.1620<) 0.37605 -0.0216 Cranial Base Length 0.2156 0.28922 0.17351 0.04467 0.61961 0.24369 Facial Height 0.07438 0.55119 0.18573 0.01827 0.0287 0.66512 Frontal Chord 0.11573 0.04108 0.70169 0.08371 0.18939 0.19929 Parietal Chord 0.3125 0.16247 0.66027 0.05038 0.00681 -0.0116 Facial Length 0.16703 0.83547 0.03209 -0.1244 0.20739 0.16746 Cranial Height 0.29714 -0.3441 0.50361 0.14989 0.507 -0.0983 Bizygomatic Breadth 0.17833 0.23963 -0.0234 0.84673 0.19101 0.04162 Nasal Breadth 0.00726 0.74457 0.09241 0.03422 0.22665 -0.304 Nasal Height -0.004 -0.0833 0.04726 0.11673 0.21189 0.82913 External Palate Breadth 0.14916 0.85435 0.11159 0.10367 -0.0963 0.09162 Mastoid Length -0.1045 0.30232 0.24443 0.06756 0.48305 0.11289 Midfrontal 0.54919 -0.051 0.18733 0.02865 0.16958 0.44997 Temporal Crest 0.66431 0.06204 -0.0709 0.27113 0.1557 0.29933 Bregma 0.66316 0.07523 0.10512 0.02163 0.3119 0.01709 Midparietal 0.81587 0.03016 0.15953 0.1445 -0.1836 0.00959 Anterior Parietal 0.77977 -0.0127 0.26892 0.01564 0.03382 0.0411 Posterior Parietal 0.79587 0.10904 0.22612 0.04786 -0.0407 -0.0157 Lambda 0.6379 0.24435 0.11696 -0.0031 0.23872 -0.1299 Weight kg 0.05972 -0.0313 -0.0031 0.21739 0.61264 -0.0066 Stature 0.10586 0.08268 0.21868 -0.1642 0.70218| 0.20025

PAGE 75

64 As previously seen in the combined sample, all r-values are relatively low. The highest r for the autopsy sample (Table 38) is 0.55 for cranial base length versus body weight maximum length is second with a value of 0.51. The regression results using lean weight show weaker relationships (Table 39). The r-value for cranial base length drops to 0.48 and maximum length to 0.46. Height from porion has a large change from 0.49 to 032. Even with the log transformation to correct for lack of normality in the sample, the correlations between cranial variables and body weight are not as high as indicated in the literature (Aiello and Wood 1994; Gauld 1992). With such weak relationships between the variables the use of the line fitting techniques is merely a formality. Table 40. Autopsy Sample Regression coefficients for Body Weight Versus Cranial Variables Variable N r Slope Intercept SEE F Sign. Maximum Length 46 0.51 3.48 -5.98 0.10 15.19 0.000 Maximum Breadth 47 0.26 1.81 -1.95 0.12 3.35 0.074 Minimum Frontal Breadth 45 0.35 1.79 -1.66 0.11 5.96 0.019 Biauricular Breadth 47 0.11 0.62 0.65 0.12 0.56 0.457 Height from Porion 47 0.49 2.69 -3.75 0.10 14.30 0.000 Cranial Base Length 37 0.55 2.30 -2.71 0.10 15.53 0.000 Facial Height 37 0.30 0.79 0.49 0.11 3.56 0.068 Frontal Chord 47 0.17 1.03 -0.17 0.12 1.39 0.244 Parietal Chord 47 0.06 0.24 1.45 0.12 0.14 0.713 Midfrontal 46 0.04 0.04 1.98 0.12 0.08 0.776 Temporal Crest 46 0.15 0.13 1.84 0.12 0.96 0.333 Bregma 47 0.19 0.29 1.71 0.12 1.68 0.201 Midparietal 47 0.21 0.18 1.81 0.12 2.10 0.155 Anterior Parietal 47 0.16 0.17 1.81 0.12 1.22 0.275 Posterior Parietal 46 0.18 0.17 1.82 0.12 1.42 0.240 Lambda 47 0.14 0.15 1.82 0.12 0.87 0.357 Summary To investigate tiie proposal that in modem humans cranial size scales in proportion to body size, and that cranial vault thickness scales in proportion to systemic skeletal robusticity, five hypotheses were tested: • HI : body weight in modem humans covaries with cranial measurements

PAGE 76

65 • H2: estimates of lean body weight will have stronger covariance relationships with cranial measurements than gross body weight will have with cranial measurements • H3: vault thickness is a function of general skeletal robusticity as indicated by cortical thickness of long bones • H4: an index of vault length and height covaries with vault thickness • H5: weight of the brain significantly covaries with vault thickness The results of this study support two and reject three of these hypotheses. Table 41. Autopsy Sample Regression coefficients for Lean Weight Versus Cranial Variables Variable N r Slope Intercept SEE F Sign. Maximum Length 46 0.46 2.46 -3.79 0.08 12.03 0.001 Maximum Breadth 47 0.20 1.06 -0.47 0.09 1.96 0.169 Minimum Frontal Breadth 45 0.32 1.26 -0.72 0.09 4.94 0.031 Biauricular Breadth 47 0.17 0.70 0.35 0.09 1.29 0.262 Height from Porion 47 0.32 1.33 -1.01 0.09 5.17 0.028 Cranial Base Length 37 0.48 1.52 -1.26 0.08 10.33 0.003 Facial Height 37 0.32 0.66 0.61 0.09 4.11 0.050 Frontal Chord 47 0.23 1.03 -0.30 0.09 2.47 0.123 Parietal Chord 47 0.04 -0.11 2.05 0.09 0.05 0.817 Midfrontal 46 0.07 0.05 1.78 0.09 0.22 0.642 Temporal Crest 46 0.26 0.18 1.68 0.09 3.20 0.081 Bregma 47 0.24 0.28 1.60 0.08 2.81 0.100 Midparietal 47 0.29 0.18 1.68 0.09 3.96 0.053 Anterior Parietal 47 0.19 0.16 1.70 0.09 1.70 0.199 Posterior Parietal 46 0.24 0.17 1.69 0.09 2.60 0.114 Lambda 47 0.25 0.21 1.65 0.09 3.02 0.089 Hypothesis One is certainly supported as indicated by the correlation analyses of the combined sample and its subsets. Of the sixteen cranial measurements in the combined sample, ten were significantly correlated with body weight More cranial variables showed significant correlations to body weight in this sample than in the autopsy or skeletal, because the larger sample size supported the weaker relationships. The strongest relationships were for height from porion and cranial base length, the former having the greatest r-value for this study, 0.628. This is still below my minimum desirable coirelation coefficient, of 0.69 for circumference of the radius.

PAGE 77

66 Hypothesis Two is rejected, since the estimate of lean body weight was not better conrelated to the cranial measurements. The skinfolds did produce meaningful reductions of the gross body weights, but the new variable had lower rvalues, particularly for height from porion and cranial base length. Although the clavicular measurements were well correlated to body weight in the combined sample, they were not also well correlated to vault thickness. These data do not support the idea that clavicular cortical thickness and vault thickness are both the product of the same systemic skeletal growth factors. Hypothesis Three is rejected. The cranial index did covary significantly with cranial thickness, thus supporting Hypothesis Four. Approaching equality between cranial breadth and length was associated with decrease in vault thickness at bregma and on the parietal. The final hypothesis was not supported. Brain weights in the autopsy sample showed only one significant correlation to vault thickness, at the anterior parietal location. No pattern of association is established.

PAGE 78

DISCUSSION At the inception of this study I proposed that the human cranium is scaled in proportion to the entire body. From this foundation I tested five working hypotheses. Those tests indicated that cranial size is, at best, only moderately related to body size, while vault thickness is, for practical purposes, completely unrelated to body weight. The estimate of lean body mass, although carefully executed and seemingly providing reasonable adjustments of subjects' weight classes to lower levels, was not better correlated with the cranial variables. Vault thickness was not well correlated to clavicular cortical thickness or brain weight. There were, however, interesting correlations between the Cranial Index and individual vault thickness measurements, although average vault thickness did not covary significantly with the Cranial Index or its categories. These results initially appear to be very negative, when actually, at least for the first hypothesis, they affirai much of the previous literature. Pensler and McCarthy, Nawrocki, and Hartwig-Scherer and Martin all found low or insignificant correlations between cranial measurements and body weight in modem humans. The methodology in this study resembles those studies in the use of measured body weights, not weights drawn from the literature or predicted through other means. Another point of similarity is the use of modem humans as a separate (or sole) sample for analysis. These similarities are no surprise as I designed the methodology to have these elements, but my results confirm that many cranial measurements, especially vault thickness measurements, are poorly correlated to body weight. Critically, one could assume Nawrocki and Hartwig-Scherer had spurious results due to small sample sizes (n < 30 and n = 19), respectively), and Gauld suggested 67

PAGE 79

68 that Pensler and McCarthy's results were due to "increased environmental factors". However, my data indicate that neither is the cause of their results. My results illustrate how small sample sizes obscure relationships. Sample size has had an obvious effect in this study on the appearance of significant correlations, even to the point that a sample nearing 50 is too small. The low weights in the skeletal sample obscured many significant correlations, but a sample excluding all very low weight individuals (as occurs in methodologies biased against the emaciated) had too few individuals to reveal the weaker relationships. Using a weight limit with no consideration of height allowed a larger sample which produced results resembling those found in the Autopsy sample, including additional correlations not seen in the smaller data set. The primary divergence in methodology between my study and those with conflicting results (Aiello and Wood, Gauld) is my focus on a strictly human sample, use of measured, not predicted or averaged, body weights, and avoidance of species means. The error of investigating relationships between variables when one lacks measurements of the variables is obvious from the position of empiricism— to observe the relationship, observations are needed. Yet when studying primates the only species in generous supply tends to be H. sapiens, so researchers must make do with what is available. Aiello and Wood were constrained for body weights for their nonhuman sample, which is understandable, but Gauld lacked measured body weights even for the humans in her study. Departing from the data deficit issue, the multi-species data set for predicting body weight also presents a problem. The convenience of the multi-species sample is that it lowers the standard errors of the equations produced, but as Smith (1985) indicates, the primary concern of body weight prediction is not precision, it is accuracy. To improve accuracy, to arrive at an estimate that may actually have been the specimen's functional, living body weight, one must design a sample with specimens resembling the focal animal. Researchers Gauld (1996) and Aiello and Wood (1994) shared the goal of predicting body weight in fossil hominids. Altiiough they used different cranial variables, tiieir strategies

PAGE 80

69 were the same and to some degree included the same data since they both drew nonhuman body weights from Harvey et al. (1987). Their samples may be quite effective for the smaller-brained hominids, such as the Australopithecines and possibly the habilines, but for the larger-brained species modem humans must be the best reference sample for accuracy. In their study of body weight prediction of extant and extinct Cercopithecidae, Delson et al. (2000) suggested that the results of Aiello and Wood were in part due to the extremes of body size in their primate taxa. Best fit lines for data sets with "mouse to elephant" weight ranges tend to be anchored by the extreme points rather than representing the variation in the middle. While this explanation may address Aiello and Wood's cranial results, it does not explain the overall difference in performance between their cranial and postcranial variables. Delson et al. (2000) found that postcranial variables performed better than dental or cranial variables in weight prediction in cercopithecids. The multi-species data set is unacceptable when the measurements are focused on areas that are specialized for one, but not the majority of, species in the group, I suggest that their analysis may be valid for nonhuman apes, but the divergent evolution of humans in expansion of the vault and reduction of the face makes cranial measurements unsuitable predictive tools. How well can a collection dominated by small-brained apes represent the genetic and epigenetic forces resulting in the morphology of their large-brained relatives? It would be comparable to measuring five different kinds of car tires, adding Mack truck tires as the sixth group and predicting other Mack tires from the resulting formula. Aiello and Wood's postcranial measurements certainly reflect the effect of divergent evolution within their sample, as the femoral measurements produced weaker correlations than the humeral measurements because humans are specialized for bipedality, while changes in the humerus are not so derived. Yet why are the same weak correlations not found in Aiello and Wood's ( 1994) cranial measurements? I propose two causes. The humans are ouUiers,just not as markedly so as for the leg measurements. More importantly, their nonhuman species

PAGE 81

70 means in the postcranial analysis were based on even fewer individuals, frequently as few as three per mean. The nonhuman means represented even less diversity than they did in the cranial analysis, so the much greater number of humans (12 male/12 female) would have better differentiated the human species mean relative to the nonhumans. The authors did not publish the regression plot for their cranial or postcranial data (and Smith (2000) cautions against using regression formulae not showing such a plot), so we do not know where humans fall in relation to the other hominoids. I suspect, considering how my low r-value for logio transformed biauricular breadth contrasts with their biporionic breadth (r = 0.18 and r = 0.98, respectively) that humans did not converge with the other apes. More support against strong associations between cranial measurements and body weight comes from the principal component analyses. Analysis of the Combined Sample and its subsets did not produce consistent associations between cranial variables and the components highly loaded for body weight or stature. The body size component in the Autopsy sample had high loadings from Body Weight and Height from porion, and a moderate contribution from Maximum length (Table 37). Independently the cranial variables show non-transformed r-values of 0.45 and 0.49, respectively to Body Weight. The Combined Sample shows different relationships, as Body Weight loads with Cranial base length and facial height on the second component (Table 18). Again, these cranial variables do not have high r-values (r = 0.49 and 038 respectively). Another interesting aspect of tiie PCA is Uiat tiie vault thickness variables always loaded togetiier on a single component, and never in association with body size. Lean body weight was not better correlated witii body weight. Ether tiie quantity generated is unrelated to lean body mass due to errors in the prediction metiiod, or tiie physiological relationship is not based on lean mass. Estimates of total body fat from skinfolds can be unreliable, particularly tending towards underestimation in very obese subjects (Shepherd 1991). An additional source of error is the use of equations for the body density estimate tiiat were not developed for tiie population under sftidy. However,

PAGE 82

71 Dumin and Womersley's equations were developed from a European population. This type of error should not apply unless the level of specificity is very precise. The correlations in the skeletal sample indicate that leaner bodies produce lower significant correlation coefficients. Pearson (2000) found weak correlations between robusticity in the clavicle and robusticity in the humerus, but some of the commentators attributed this to size difference between the clavicle and the humerus. Pearson also used a robusticity index, while I compared gross thickness without an attempt at standardization. Dividing the thickness measurements by their respective lengths (chord lengths for the cranial thicknesses and clavicle length for the cortical measurements) might have produced more interesting results. Currently I also find only weak and mostly insignificant associations between the two types of bone thickness. Moss and Young's (1961) work suggests the better methodology is to treat the tables of the vault as independent functional units. They demonstrated that the inner table follows the contours of the brain and dura even after neural atrophy has reduced the cranial contents. With each layer serving specific functions in conjunction with their function as a unified structure, the best approach to analyzing vault thickness would be to measure the layers separately and together. Regarding the brain weight results, I suspect brain weight is not the best indicatory of size indicator. I was proposing, based on work showing that brain and dura growth defines and restricts vault growth (Moss and Young 1961, Huggare and Ronning 1995), that vault thickness would be related to brain size. A space-filling size indicator, such as volume, may have been more appropriate, since it is expansion of the contents that directs the growth of the vault My results support Nawrocki's conclusion that vault thickness increases with head size (Table 13). All of my significantly coirelated ectocranial variables had positive relationships to vault thickness. The Cranial Index results, while informative, lack a large

PAGE 83

sample of brachycranic individuals. There were insufficient significant correlations, but at least all the correlations for the brachycranic class were negative. The results suggest that the wider and shorter the vault gets, the thinner the vault gets in some locations. Contrary to the literature, many cranial measurements are not suitable as predictors of body weight in modem humans, and by extension for large-brained hominids. Gross vault thickness measurements are particularly poor tools and should not be used. Cranial measurements may be effective for small-brained extant and extinct hominoids, as Aiello and Wood did have a large correlation coefficient for their primarily nonhuman ape sample, Body weight predictions for larger-brained fossil hominids based on cranial measurements must be considered suspect because, at best, the formula predicts morphology for a smallbrained hominid. It must also be considered that there are many other cranial measurements that were not used in this study, so a useful relationship may exist that is as yet undescribed. Qear associations do exist between vault length and height and vault thickness, and there is both theoretical and empirical support for rounder skulls having relatively thinner vaults. Although this study does not identify how skeletal robusticity contributes to vault thickness, it suggests a methodological approach that might be effective for future research.

PAGE 84

REFERENCES Adeloye A, Kattan KR and Silverman FN (1975) Thickness of the normal skull in the American blacks and whites. Am J Phys Anthropol 43: 23-30. Aiello LC and BA Wood (1994) Cranial variables as predictors of hominine body mass. Am J Phys Anthropol 95:409-426. Andrews P (1984) An alternative interpretation of the characters used to define Homo erectus. Cour Forsch Inst Senckenberg 69: 167-175. Anton SC (1989) Intentional cranial vault deformation and indirect changes of the cranial base and face. Am J Phys Anthropol 79:253-267. Ant6n, SC (1997) Endocranial hyperostosis in Sangiran-2, Gibraltar-l, and Shanidar-5 Am J Phys Anthropol 102: 1 1 1-122. Ant6n SC and Franzen JL (1997) The occipital torus and developmental age of Sangiran-3 J Hum Evol 35:599-610. Baker PT and Newman, RW (1957) The use of bone weight for human identification. Am J Phys Anthropol 75:601-618. Bass WM (1995) Human Osteology: A Laboratory and Field Manual. Columbia, MO:Missoun Archaeological Society. Brown, P (1987a) Cranial vault thickness in Northern Chinese, European and Australian Abonginal populations. English summary. Act Anthropol Sin 6:184-189. Brown P (1987b) Pleistocene homogeneity and Holocene size reduction: the Australian human skeletal evidence. Archaeology in Oceania 22:41-71. Brown P (1994) Cranial vault thickness in Asian Homo erectus and Homo sapiens. Cour Forsch Senckenberg 777:33-46. Brown T Pinkerton SK^d Lambert W (1979) Thickness of the cranial vault in Australian abongmals. Archaeol Phys Anthropol Oceania 74:54-71. Brunet M, Guy F, Pilbeam D, Mackaye H, Ukius A, Ahounta D, Beauvilain A et al. (2002) A new homimd from the Upper Miocene of Chad, Central Africa. Nature 4 18: 145151 . Clarke R (1985) A new reconstruction of the Rorisbad cranium with notes on the site. In E Delson (ed.): Ancestors: The Hard Evidence, pp 301-305. New York, NY: AcadeiSc Press: S(2M 15-^7 Problems of body-weight estimation in fossil primates. Int J Primatol 73

PAGE 85

74 Corruccini RS and Henderson AM (1978) Multivariate dental allometry in primates. Am J Phys AnthropoU«:203-208. Currey J (1984) The mechanical adaptations of bones. Princeton, NJ: Princeton University Dagosto M and Terranova Q (1992) Estimating the body size of Eocene primates: a comparison of results from dental and postcranial variables. Int J Primatol 1 i(3):307-344. Delson E, Terranova CJ, Jungers Wl, Sargis EJ, Jablonski NG, and Dechow PC (2000) Body Mass m Cercopithecidae (Primates, Mammalia): Estimation and Scaling in Extinct gd Extant Taxa. Anthropological Papers of the American Museum of Natural History No. Demes B (1985) Biomechanics of the primate skull base. F Beck, HW Hild GW Kriz HR Ortmann, KJ Pauly, TH Schiebler (eds.): Advances in Anatomy Embryology and Cell ' Biology, vol. 94. New York: Springer-Verlag,pp 1-52. Dumin JV and Womersley J (1974) Body fat assessed from total body density and its 7?^Sre Br J NuS?2-'^^ Measurements on 481 men and women aged from 16 to Felsing NE, Brasel JA and Cooper DM (1992) Effect of low and high intensity exercise on circulating growth hoimone in men. J Clin Endocrinol Metab 75: 157-162. ^!^^T^'f f'^" *^"czmarski RJ, and Johnson CL (1998) Overweight and obesity in the Umted States: Prevalence and trends, 1960-1994. Int J Obesity 22:39-47. Frisancho AR (1990) Anthropometric Standards for the Assessment of Growth and Mutntional Status. Ann Arbor. University of Michigan Press. Gauld SC (1992) An allometric study of anthropoid cranial bone thickness: Implications tor body size esdmation m early hominid species. PhD Dissertation, Department of Anthropology, Umversity of California, Los Angeles. S:lrl^ffi<^ltt2^^^ Giles E and EUiot O (1962) Race identification from cranial measurements. J For Sci 2JrmTs'So^V2T5^'4'^^ S'il^n iTT^ Coirelatio'^ of tooth size and body size in living hominoid primates, 47^391398 '° ^'Syptopithecus and Proconsul. Am. J Phys Anthro^l Sis? T' BH and Rosenberg K (1982) Allometric scaling in the dentition of pnrnates and prediction of body weight from tooth size fossils. Am J Phys Anthrol^l 58:

PAGE 86

75 Hartwig-Scherer S (1994) Body weight prediction in fossil Homo. Courier ForschungsInstitut Senckenberg 777:267-279. Hartwig-Scherer S and Martin RD (1992) AUometiy and prediction in hominoids: a solution to the problem of intervening variables. Am J Phys Anthropol 88:31-51. Harvey PH, Martin RD and Qutton-Brock Th (1987) Life histories in comparative perspective. In BB Smuts, DL Cheney, RM Seyfarth, RW Wrangham and TT Struhsaker (eds.): Primate Societies. Chicago: University of Chicago Press, pp 181-196. Hatcher L and Stepanski EJ (1994) A Step-by-Step Approach to Using the SAS System for Univariate and Multivariate Statistics. Cary, NC: SAS Institute Inc. Hershkovitz I, Greenwald C, Rotivschild MM, Latimer B, Dutour O, Jellema LM, and WishBaratz S (1999) Hyperostosis frontalis interna: An anthropological perspective. Am J Phys AnUiropol 709:303-325. Hinton MAC, Oakley KP, Dines HG, King WBR, Kennard AS, Hawkes CFC, Warren SH, Cotton MA, Le Gros Clark WE, and Morant GM (1938) Report on the Swanscombe skull. J Roy Antiiropol Inst Great Britain Ireland 68: 17-98. Holloway RL (1980) Within-species brain-body weight variability: a reexamination of the Danish data and other primate species. Am J Phys Antiu-opol 53: 109-121. Howells WW (1973) Cranial Variation in Man. Papers of the Peabody Museum No. 67. Cambridge, MA: Harvard University. Hublin J (1986) Some comments on the diagnostic features of Homo erectus. Anthrooos (Brno) 25:175-187. ^ Huggare J and Ronning O (1995) Growth of tfie cranial vault: Influence of intiacranial and extracranial pressures. ActaOdontol Scand 53 .192-195. Ishida H and Dodo Y ( 1990) Cranial Uiickness of modem and Neolithic populations in Japan. Hum Bio 62(3):389-401. Israel H (1973) Age factor and die pattern of change in craniofacial structures. Am J Phys AnUiropol 39:111-128. Israel H (1977) The dichotomous pattern of cranialfacial expansion during asdne Am J Phys Antiu-opol 47: 47-52. b & b Ivanhoe F (1979) Direct correlation of human skull vault fliickness witii geomagnetic mtensity m some northern hemisphere populations. J Hum Evol 8:433-44. Jacob T (1972) Palaeoantiiropological discoveries in Indonesia witii special reference to tiie finds of the last two decades. J Hum Evol 2:473-485. Jantz RL and Moore-Jansen PH (1988) A Data Base for Forensic Anthropology: Stiiicture Content and Analysis. Report of Investigations No. 47. Knoxville, TN: The University of ' 1 ennessee Department of Anthropology.

PAGE 87

76 Jungers WL (1984) Aspects of size and scaling in primate biology with special reference to the locomotor skeleton. Yrbk Phys Anthropol 27:73-97. Jungers WL (1988) Relative joint size and hominoid locomotor adaptations with implications for the evolution of hominid bipedalism. J Hum Evd. 77:247-265. KaitoY(1980) On the thickness of the Japanese skull. Tokyo Jikeikai Medical Journal 95:5(B-513. Kay RF ( 1975) The functional adaptations of primate molar teeth. Am J Phys Anthropol 43:195-216. ^ f Kennedy GE (1985) Bone thickness in Homo erectus. J Hum Evol 14:699-708. Kennedy GE (1991) On the autapomorphic traits of Homo erectus. J Hum Evol 20: 375412. Larsen CS, Matter RM, and Gebo DL (1998). Human Origins: The Fossil Record. Prospect Heights, IL: Waveland Press Lassarre C, Girad F, Durand J, and Raynaud J (1974). Kinetics of human growth hormone dunng submaximal exercise. J App Physiol 57(6):826-830. lieberman DE (1996) How and why humans grow a thin skull: experimental evidence for systemic cortical robusticity. Am J Phys Anthropol 101: 217-236. Undstedt SL and Calder WA (1981) Body size, physiological time, and longevity of homeothermic animals. C^ar Rev Bio 5(^1): 1-16. Loth SR and Henneberg M (1996) Mandibular ramus flexure: A new morphologic mdicator of sexual dimorphism in the human skeleton. Am J Phys Anthropol 99:473-485. Martin, AD, Ross WD, Drinkwater DT, and Clarys JP ( 1985) Prediction of Body Fat by Skinfold Caliper Assumptions and Cadaver Evidence. Int J Obesity 9:31-39 Suppl 1. Martin, RA (1981) On extinct hominid population densities. J Hum Evol i0:427-428. Martin R and Sailer K (1957) Lehrbuch der Anthropologie Vol. 1. StuttgartGustav Fischer Veriag. McHenry HM (1976) Early hominid body weight and encephalization. Am J Phys Anthropol 45: 77-84. 45^25^306^ ^^^^ ^^'^^'''^ cheek-tooth size in Australopithecus. Am J Phys Anthropol McHenry HM (1988) New estimates of body weight in early hominids and their significance to encephalization and megadontia. In FE Grine (ed.): Evolutionary History of the Robust Australopitheanes. New York: Aldine de Gruyter, pp 133-148. So^lSfiiSJiT'"'' °' A"^t-»«Pithecines. Am J Phys

PAGE 88

77 McNab BK (2002) The Physiological Ecology of Vertebrates. A View from Energetics. Ithaca: Cornell University Press. Moss ML and Young RW (1961) A functional approach to craniology. Am J Phvs Anthropol 18: 281-292. Nawrocki SP (1992) A Biomechanical Model of Cranial Vault Thickness in Archaic Homo. Ann Arbor: University of Michigan Press. Ohtsuki F (1977) Developmental changes in the cranial bone thickness in the human fetal period. Am. J Phys Anthropol 46.141-154. Ortner DJ and Putschar WGJ (1981) Identification of Pathological Conditions in Human Skeletal Remains. Smithsonian Contributions to Anthropology 28. Washington, DC: Smithsonian Institution Press. Pearson OM (2000) Activity, climate, and postcranial robusticity. Current Anthropol 47:569-607. Pensler J and McCarthy JG (1985) The calvarial donor site: An anatomic study in cadavere. PlastReconSurg 75:648-651. Pirie PL (1978) Allometric scaling in the postcanine dentition with reference to orimate diets. Primates 79(3):583-591. Wchards G and Anton SC (1991) Craniofacial configuration and postcranial development of a hydrocephalic child (ca. 2500 BC to 500 AD): with a review of cases and comments on diagnostic cnteria. Am J Phys Anthropol 85: 185-200. (1986) Body size and encephalization in Homo erectus. Anthropos (Brno) Roche, AF (1953) Increase in cranial thickness during growth. Hum Bio 25: 81-92. Ross AH, Jante RL, and McCormick WF ( 1998) Cranial thickness in American females and males. J For Sci 45f2):26-272. Ewl 5^2^^^°'*^ ^"""^ ^""^ '° "^"^^"^ J Hum ftiSatoUO^J""^ ^'^^ '° *^ evolution of African ape craniodental form. Folia SX^Sity ftL?"'""^^^^ " Anthropology. Cambridge, MA: rif ^Nl-H^ri^^T' ^ ^ ^^2^ ""'"^^ ^^^^^^^ ^^n^ai'^In R Singer and J Wymer (eds )• SS^fcWca^^^^ Sl^cl^rnS^

PAGE 89

78 Smith RJ ( 1985) The present as a key to the past. Body weight of Miocene hominoids as a test of allometric methods for paleontological inference. In WL Jungers (ed.): Size and Scaling in Primate Biology. New York, NY: Plenum Press, pp 437-448. Smith RJ (1994) Regression models for prediction equations. J Hum Evol 26:239-244. Smith RJ (2002) Estimation of body mass in paleontology. J Hum Evol 45:271-287. Smith P, Wax Y, Becker A and Einy S (1985) Diachronic variation in cranial thickness of near eastern populations. Am J Phys Anthropol 67: 127-133. Sokal RR and Rohlf FJ (1995). Biometry: The Principles and Practice of Statistics in Biological Research. San Francisco, CA: WH Freeman and Co. Sonakia A (1985) Early Homo from Narmada Valley, India. In E Delson (ed.): Ancestors: The Hard Evidence. New Yoric, NY: Academic Press. Steudel K (1980) New estimates in early hominid body size. Am. J Phys Anthropol 52:63Stringer CB (1984) The definition of Homo erectus and the existence of the species in Africa and Europe. Cour Forsch Inst Senckenberg 69: 131-143. Tallgren A (1974) Neurocranial morphology and ageing— A longitudinal roentgen cephalometric study of adult Finnish women. Am J Phys Anthopol 47:285-294. Tobias PV (1967) The cranium and maxillary dentition of Australopithecus (Zinjanthropus) boisei. In LSB Leakey, (ed.): Olduvai Gorge Vol. 2. Cambridge, Great Britain: University Press. Todd TW (1924) Thickness of the male white cranium. Anat Rec 27: 245-256. Trinkaus E (1983) The Shanidar Neandertals. New York, NY: Academic Press. Twiesselmaim F (1941) Methode pour revaluation de I'epasseur des parois craniennes. BulleUn du Musee Royal d'Histoire Naturelle de Belgique 77:1-33. '^a^t^^^y HA, Smith, KR, Stubblefield PR, Schultz J, and Falsetti AB (1999) Regional Differences in Cranial Morphology. Abstract. Proceedings of the Annual meeting of the Amencan Association of Physical Anthropologists. Columbus, Ohio. R^' 99"S9-248^^ temporal muscle to the form of the skull. Anat. Weidenreich F (1943) The skull of Sinanthropus pe/dnensis; A comparative study on a 72™^°™°" Palaeontologica Sinica New Series D, No 10 Whole Series No ^^w.Tk'^'' ^^^^^191?.^"'^' fPf ""'^ subspecific characters in human evolution. In ^ Washburn and D Wolffson (eds.): The Shorter Anthropological Papers of Franz Weidenreich Published m the Period 1939-1948. New Yoric NYVikine Fund dd25^^ [Originally published Am J Phys Anthropol 4(4) n.s ] ' ^ ^ ' ^''^"^ PP

PAGE 90

79 Wolpoff MH ( 1980) Cranial remains of Middle Pleistocene European hominids. J Hum Evol 9:339-358. Wood B (1984) The origin of Homo erectus. Cour Forsch Inst Senckenberg 69:99-1 11. Wood BA and Stack CG (1980) Does allometry explain the differences between "gracile" and "robust Australopithecines? Am J Phys Anthropol 52:55-62. Young, RW (1957) Postnatal growth of the frontal and parietal bones in white males. Am J Phys Anthropol 75:367-386. Zar JH ( 1998) Biostatistical Analysis. Englewood Cliffs, NJ: Prentice Hall.

PAGE 91

APPENDIX Inventory of Male Subjects Autopsy Sample N=47 This listing is sorted by age. Location is Michigan (mich) or Orlando (orl). Weight and body length were recorded in English measurements, unless otherwise noted, and converted to metnc equivalents within the spreadsheet software. c * specimen Location Age Weight (lbs) Body length (in) 56 lo o 1 ri 21U 68 58 1111 1
PAGE 92

Autopsy Sample. Continued Specimen Location Age Weight (lbs) Body length (in) uri ZOi /o rtrl 1 $24 illiCll 21 nrl «:< '?1 o ZiU "79 /Z xx\\r\\ llllCll JO ZZ7Z ^9 /Z 57 Tt\\c\\ 1111 L-Xl /I, J 22 nrl 1 94 IZMDl 64 lilicil Zl 1 OO.J 67 llllCll 1 cin /u 52 xv\\c\\ lllicil 1^i9 lOZ /u 50 mich 63 206 68 70 mich 63 212 72 53 mich 65 176 70 42 mich 68 184 68.5 72 mich 74 127 64 69 mich 79 178 65 30 mich 82 118 68 Skeletal Sample N= 100 Specimen numbers are as in the NMNH Terry Collection. Specimen Age Weight kg Stature cm 5 57 65.2 179.6 125 31 44.0 178.5 143 58 71.9 164.7 152 70 46.6 165.9 606 28 38.8 163 609 75 38.8 162 614 50 56.6 181.0 618 60 55.5 181 630 72 50.0 175.0 633 76 76.6 176.0 634 74 43.3 163.0 635 54 37.7 165.2 638 55 71.4 177.7 644 54 31.0 162.0 645 28 43.3 188.0 646 50 33.3 179 648 63 70.3 172 649 51 79.6 176 651 64 50.6 158.0 663 44 53.2 170 665 74 453 170.0 670 60 46.6 161.0

PAGE 93

Skeletal Sample. Continued. Specimen Age Weight k« Stature cm 671 65 62.7 158.0 674 58 59.9 171.0 675 27 42.0 176 701 55 52.2 165.0 705 46 38.9 171 707 26 42.2 187 708 68 34.4 167.0 709 50 41.1 155.0 711 61 63.3 187 712 47 60.0 166 713 69 51.0 176.0 714 71 43.3 169.0 717 29 51.1 160 718 22 55.5 180 719 25 51.1 170 720 54 72.1 156.0 721 60 43.0 184.0 724 42 58.8 170 725 57 31.8 163 730 71 82.1 167.0 731 63 61.0 166 743 53 53.3 178.0 746 71 41.1 161.0 747 45 62.4 172.0 748 85 50.7 182.0 750 80 39.8 174.0 756 47 49.7 165.0 762 59 38.8 171.0 763 46 54.8 179.0 764 58 40.6 168.0 772 56 63.4 186.0 787 50 52.8 175.0 802 36 35.1 168.0 803 64 45.2 161.0 805 87 57.0 168.0 806 78 60.1 168.0 812 54 39.9 172.0 813 60 47.4 175.0 814 55 36.1 176.0 o42 73 81.0 165.0 843 60 54.1 161.0 852 66 38.7 169.0 853 62 54.0 176.5 855 67 ; 38.7 158.0 861 ( 55 52.5 174.0 865 ( 59 n.5 169.5

PAGE 94

Skeletal Sample. Continued. Specimen Age Weight kg Stature cm 866 60 53.5 179.0 867 38 43.1 868 60 65.3 172.0 870 60 41.6 170.0 871 56 53.0 182.0 872 48 42.6 168.0 874 75 43.9 173.5 908 52 41.9 180.0 909 70 35.1 161.0 912 63 58.6 168.0 918 40 33.6 157.0 924 43 43.3 175.0 956 63 84.6 185.0 963 71 54.2 170.0 974 77 42.9 166.0 979 85 41.8 178.5 982 70 85.0 175.0 984 68 96.4 173.0 1087 62 88.3 182 1089 43 85.3 182.5 1110 68 48.8 175.5 1114 60 41.5 165.0 1125 69 52.2 164.0 1301 57 65.1 171.0 OZ.J 167.5 1424 51 38.7 152.5 1430 70 37.2 160.0 1442 63 58.8 179.0 1492 64 37.6 167.5 136R 59 69.6 170.2 783R 65 60.8 176.2 972R 65 37.9 172.2

PAGE 95

BIOGRAPHICAL SKETCH Phoebe R. Stubblefield received her Bachelor of Arts degree in anthropology at the University of California, Santa Barbara in 1990 and the Master of Arts degree at the University of Texas at Austin in 1993 with an emphasis in palaeoanthropology. She entered the University of Rorida in 1995 under the direction of the late Dr. William R. Maples, and studied forensic anthropology with him until his death. Through her relationship with Dr. Maples, Ms. Stubblefield has performed numerous forensic analyses for medical examiners in Rorida and New York, trained both federal and state law enforcement personnel, and acted as consult during the Valujet disaster. She has also assisted with historical grave recovery efforts in Oklahoma and Guatemala to identify victims of the Tulsa Riots and the Guatemalan civil war. While completing her doctoral research under the supervision and mentorship of Dr. Susan Anton, Ms. Stubblefield followed in Dr. Anton's footsteps and became a Ford Doctoral Fellow. Her research interests include skeletal biology, forensic anthropology, human evolution, and human variation. 84

PAGE 96

I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, ia-sc^ and quality, as a dissertation for the degree of Doctor of Phil^S^^Sy. >usan Ant6n, Chair Assistant Professor of Anthropology I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. Sue Boinski, Cochair Associate Professor of Anthropology I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. Anthony Faisetti Associate Professor of Anthropology I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. Thomas Holli»^ Associate Professor of Anatomy and Cell Biology This dissertation was submitted to the Graduate Faculty of the Department of Anthropology in the College of Liberal Arts and Sciences and to the Graduate School and was accepted as partial fulfillment of the requirements for the degree of Doctor of Philosophy. December 2002 Dean, Graduate School