affiumn: fig.

4
5......1 .. fl
a

5b.. 3 a...

.22.... ‘ 3a

a)! .1 :ltalr 1'
.... (tilt...- :

I! .‘5’I7‘

: .3.

I;

. as .93; t

unite};

3...... Evie.
. .“4 a? flifl. .
13.51: hr... ain.“.
51.3.}:

v...

5.3%...

3.25.1...
3.3.5....

é... .. y

.x . . x9 .

21.5.! «1.1. ..
3:1,}..1

-.. ~12... ..

.1

$.41 «:3. 1..
3:1, 2! .
: .1}L3..13,.s 1...
7.

1
0:3 .5.
3.2.. II.
.I

itll
9"}?
1..

I.

 

 

 

!V...

LIBRARY
Michigan State

 

’I.

Vii:

This is to certify that the
thesis entitled

A COMPARISON OF TWO METHODS OF AGE DETERMINATION

USING HISTOMORPHOLOGY:

PERIOSTEAL VERSUS ENDOSTEAL SURFACE EQUATIONS

presented by

Andrea L. Clowes

has been accepted towards fulfillment
of the requirements for the

MS. degree in Forensic Science

 

 

 

mam

Major F’rofessor’s $gnature

MM 4,. MM

Date

MSU is an affirmative-action, equal-opportunity employer

LIBRARY
Michigan State
University

 

 

 

PLACE IN RETURN BOX to remove this checkout from your record.
To AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

6/07 p:/CIRC/Dale0ue.indd-p.1

A COMPARISON OF Two METHODS OF AGE DETERMINATION USING
HISTOMORPHOLOGY:
PERIOSTEAL VERSUS ENDOSTEAL SURFACE EQUATIONS
By

Andrea L. Clowes

A THESIS

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

MASTER OF SCIENCE
Forensic Science

2008

ABSTRACT

COMPARISON OF TWO METHODS OF AGE DETERMINATION USING
HISTOMORPHOLOGY:
PERIOSTEAL VERSUS ENDOSTEAL SURFACE EQUATIONS
By

Andrea L. Clowes

Establishing the age at death in a human remains case is a critical component in
positive identification. Forensic techniques of bone histology, the microstructural
analysis of tissue, are known to be accurate age estimators. This thesis is meant to
explore the accuracy of the Hauser et a]. method (1980) and compare it with the Kerley-
Ubelaker method (1978). By directly comparing the two methods, the applicability of the
equations of Hauser et al. are showcased next to those of Kerley-Ubelaker.

A sample of 30 femoral midshaft thin section slides (9 to 82 years of age at death)
from the Dominican Republic was borrowed from Dr. Douglas Ubelaker at the National
History Museum of Natural History. Age estimations were determined using the Kerley-
Ubelaker method (KUA) (Kerley and Ubelaker, 1978) and the subperiosteal (HPA),
endosteal (HEA), and average of endosteal/subperiosteal (HAA) equations of the Hauser
method (1980). Paired t-tests were used to determine if there was a significant
difference from the actual age at death for KUA, HPA, HEA, and HAA. Results revealed
that the Hauser subperiosteal age estimate (p=.05) was the only accurate equation for the
full sample. When analyzing the femoral thin sections of individuals older than 21 years
at death, KUA, HPA, and HAA were all found to be accurate. Both sample sets indicated
that Hauser endosteal age is an inaccurate age estimator with the Dominican Republic

sample.

ACKNOWLEDGMENTS

I would like to first acknowledge and thank my thesis committee, Dr. Todd
Fenton, Dr. Norm Sauer, and Dr. Christina DeJong, for your helpful advice in the
analysis and writing of this thesis. You gave me clarity in times of confusion. I would
also like to thank Dr. Douglas Ubelaker at the Smithsonian Institution for allowing me to
collect and borrow the slide sample utilized in this research study. I would also like to
thank Dr. MariaTeresa Tersigni; you were my greatest teacher in the field of histology.
You were patient and willing to go the extra mile. My friends and compatriots, the
Master’s and PhD Students of physical anthropology of Michigan State University, were
always ready to lend an ear to my ideas. I thank you for that. Specifically, I would like
to note the assistance of Lindsey Jenny and Wendy Lackey. You were generous with
advice and helped me organize my ideas. My roommate, Sara, should also be
acknowledged as she has been the main sounding board for all of my ideas. Thank you
for your unending patience. Finally, I would like to thank my mother, who has supported
me in everything I’ve done throughout my life and this thesis was no different. You have

given me my work ethic and desire to learn. I wouldn’t be here without you.

iii

TABLE OF CONTENTS

LIST OF TABLES ................................................................. ~ ..................... vii
LIST OF FIGURES ..................................................................................... x
CHAPTER 1
INTRODUCTION ..................................................................................... 1
CHAPTER 2
HISTOLOGY .......................................................................................... 6
Bone Composition and Cell Types ............................................................. 6
Bone Modeling and Remodeling ............................................................... 7
Primary and Secondary Bone .............................................................. 7
Modeling and Remodeling ................................................................. 8
Thin Sections ...................................................................................... 11
CHAPTER 3
AGE ASSESSMENT ............................................................................... 13
Specific Examples of Methods ................................................................. 13
Possible Factors Affecting Age Estimation .................................................. 17
Bone Utilized in Analysis ................................................................ 17
Physical Activity ........................................................................... 18
Location on Bone for Analysis ............................................................ 20
Sex Variation ................................................................................ 20
Population Variation ...................................................................... 21
Destruction of Bone ....................................................................... 23
Intra—Observer Error ....................................................................... 23
CHAPTER 4
ADDITIONAL HISTOMORPHOMETRIC APPLICATIONS .............................. 25
Bioarchaeological Applications ............................................................. 25
Human/Nonhuman Identification ......................................................... 29
Disease Pathologies ........................................................................... 31
CHAPTER 5
QUESTIONS AND HYPOTHESES ............................................................. 33
Research Question 1 .......................................................................... 33
Research Question 2 .......................................................................... 34
Question 2a ................................................................................. 34
Question 2a: Hypothesis ........................................................ 34
Question 2a: Expected Outcome ............................................... 35
Question 2b ................................................................................ 35
Research Question 3 ........................................................................... 36

iv

TABLE OF CONTENTS CONTINUED

Question 3a ................................................................................. 36
Question 3b ................................................................................. 36
Question 3b: Hypothesis ....................................................... 36
Question 3b: Expected Outcome ......... _. .................................... 37
Research Question 4 .............................................................................. 37
CHAPTER 6
MATERIALS AND METHODS ................................................................. 38
Sample ........................................................................................... 38
Methods .......................................................................................... 39
Kerley-Ubelaker .......................................................................... 4O
Hauser et a1 ................................................................................ 42
SubPeriosteal Surface ........................................................... 42
Endosteal Surface ............................................................... 43
Intra-observer error ....................................................................... 44
Statistical Analyses ....................................................................... 44
CHAPTER 7
RESULTS ........................................................................................... 47
Sample Results of Question 1 ................................................................. 47
Sample Results of Question 2 ................................................................. 48
Sample Results of Question 3 ................................................................ 53
Sample Results of Question 4 ................................................................. 57
CHAPTER 8
DISCUSSION ........................................................................................ 58
CHAPTER 9
CONCLUSION ...................................................................................... 64
APPENDICES
Appendix A ........................................................................................... 68
Appendix B ........................................................................................... 77
BIBLIOGRAPHY ................................................................................ 101

LIST OF TABLES

Table 1. Regression equations for Kerley-Ubelaker (1978) age at death estimation
(Y = Age in years; X = Osteon or Osteon fragment count) .................................. 39

Table 2. Regression equations for Hauser et a1. (1980) age at death estimation
(Y = Age in years; X = Haversian canal count) ................................................ 41

Table 3 — Pearson’s r Correlation Between Real Age (RA) and Kerley-Ubelaker Age
(KUA) ................................................................................................ 76

Table 4 — Pearson’s r Correlation Between Real Age (RA) and Hauser-P Age
(HPA) ................................................................................................ 77

Table 5 - Pearson’s r Correlation Between Real Age (RA) and Hauser-E Age
(HEA) ................................................................................................ 78

Table 6 — Pearson’s r Correlation Between Real Age (RA) and Hauser-Avg Age
(HAA). . . . ., .......................................................................................... 79

Table 7 - Paired Sample Statistics for Real Age (RA) and Kerley-Ubelaker Age (KUA),
N=3O ................................................................................................. 80

Table 8 — Paired Sample t-Test Results and Average Error (years) for Real Age (RA) and
Kerley-Ubelaker Age (KUA), N=30 ............................................................ 80

Table 9 — Paired Sample Statistics for Real Age (RA) and Kerley-Ubelaker Age (KUA),
n=25, RA>21 ........................................................................................ 81

Table 10 — Paired Sample t-Test Results for Real Age (RA) and Kerley-Ubelaker Age
(KUA), n=25, RA>21 .............................................................................. 81

Table 11 — Paired Sample Statistics for Real Age (RA) and Hauser-P Age (HPA),
N=30 ................................................................................................. 82

Table 12 — Paired Sample t-Test Results and Average Error (years) for Real Age (RA)
and Hauser-P Age (HPA), N=30 ................................................................. 82

Table 13 - Paired Sample Statistics for Real Age (RA) and Hauser-P Age (HPA), n=25,
RA>21 ............................................................................................... 83

Table 14 — Paired Sample t-Test Results for Real Age (RA) and Hauser-P Age (HPA),
n=25, RA>21 ....................................................................................... 83

vi

LIST OF TABLES CONTINUED

Table 15 — Paired Sample Statistics for Real Age (RA) and Hauser-E Age (HEA),
N=30 ................................................................................................. 84

Table 16 — Paired Sample t-Test Results and Average Error (years) for Real Age (RA)
and Hauser-E Age (HEA), N=30 ................................................................ 84

Table 17 — Paired Sample Statistics for Real Age (RA) and Hauser-E Age (HEA), n=25,
RA>21 ............................................................................................... 85

Table 18— Paired Sample t Test Results for Real Age (RA) and Hauser-E Age (HEA),
n=25, RA>21 ........................................................................................ 85

Table 19 —- Paired Sample Statistics for Real Age (RA) and Hauser-Avg Age (HAA),
N=30 ................................................................................................. 86

Table 20 — Paired Sample t-Test Results and Average Error (years) for Real Age (RA)
and Hauser-Avg Age (HAA), N=30 ............................................................. 86

Table 21 -— Paired Sample Statistics for Real Age (RA) and Hauser-Avg Age (HAA),
n=25, RA>21 ........................................................................................ 87

Table 22 - Paired Sample t-Test Results for Real Age (RA) and Hauser-Avg Age (HAA),
n=25, RA>21 ........................................................................................ 87

Table 23 —Percentages within 5 and 10 years; 95th percentile for Kerley-Ubelaker Age
(KUA), Hauser-P Age (HPA), Hauser-E Age (HEA), and Hauser-Avg Age (HAA)
estimation methods/equations, N=3O ............................................................. 88

Table 24 -Percentages within 5 and 10 years; 95th percentile for Kerley-Ubelaker Age
(KUA), Hauser-P Age (HPA), Hauser-E Age (HEA), and Hauser-Avg Age (HAA)
estimation methods/equations, n = 25, RA >21 ................................................. 88

Table 25- Paired Differences of the Mean for Kerley-Ubelaker Age (KUA), Hauser-P
Age (HPA), Hauser-E Age (HEA), and Hauser-Avg Age (HAA), N=30 .................. 89

Table 26- Paired Differences of the Mean for Kerley-Ubelaker Age (KUA), Hauser-P
Age (HPA), Hauser-E Age (HEA), and Hauser—Avg Age (HAA), n = 30, RA>21 ....... 89

Table 27 — Pearson’s r Correlation Among Real Age (RA), Kerley-Ubelaker Age (KUA),
Hauser-P Age (HPA), Hauser-E Age (HEA), and Hauser-Avg Age (HAA) ............... 90

Table 28 - Paired Sample Statistics for Kerley-Ubelaker Age (KUA) and Hauser-P Age
(HPA) ................................................................................................. 91

vii

LIST OF TABLES CONTINUED

Table 29 — Paired Sample t-Test Results for Kerley-Ubelaker Age (KUA) and Hauser-P
Age (HPA) ........................................................................................... 91

Table 30— Paired Sample Statistics for Kerley-Ubelaker Age (KUA) and Hauser-E Age
(HEA) ................... g .............................................................................. 92

Table 31 — Paired Sample t—Test Results for Kerley-Ubelaker Age (KUA) and Hauser-E
Age (HEA) ........................................................................................... 92

Table 32 — Paired Sample Statistics for Kerley-Ubelaker Age (KUA) and Hauser-Avg
Age (HAA) ........................................................................................... 93

Table 33 — Paired Sample t-Test Results for Kerley-Ubelaker Age (KUA) and Hauser-
Avg Age (HAA) ..................................................................................... 93

Table 34 — Paired Sample Statistics for Hauser-P Age (HPA) and Hauser-E Age
(HEA) ................................................................................................ 94

Table 35 — Paired Sample t-Test Results for Hauser-P Age (HPA) and Hauser-E Age
(HEA) ................................................................................................ 94

Table 36 - Paired Sample Statistics for Hauser-P Age (HPA) and Hauser-Avg Age
(HAA) ................................................................................................ 95

Table 37 — Paired Sample t-Test Results for Hauser-P Age (HPA) and Hauser-Avg Age
(HAA) .......................... ‘ ...................................................................... 95

Table 38 — Paired Sample Statistics for Hauser-E Age (HEA) and Hauser—Avg Age
(HAA) ................................................................................................ 96

Table 39— Paired Sample t Test Results for Hauser-E Age (HEA) and Hauser-Avg Age
(HAA) ................................................................................................ 96

Table 40- Pearson’s r Correlation Between Kerley-Ubelaker Age 1 (KUAl) and Kerley-
Ubelaker Age 2 (KUA2) ........................................................................... 97

Table 41 — Pearson’s r Correlation Between Hauser-P Age 1 (HPAl) and Hauser-P Age 2
(HPA2) ............................................................................................... 97

Table 42 — Pearson’s r Correlation Between Hauser—E Age 1 (HEAD and Hauser-E Age
2 (HEAZ) ............................................................................................. 98

Table 43 - Pearson’s r Correlation Between Hauser-A Age 1 (HAAl) and Hauser—A Age
2 (HAA2) ............................................................................................ 98

viii

LIST OF FIGURES

Figure l — Shaft of Long Bone. Displays Examples of Haversian Systems and
Volkmann’s Canals (Above). The Lower Image on the Right Exhibits a Magnified View
of an Osteon. Adapted from J unqueira and Carneiro (2005) ................................. 67

Figure 2 — The Longitudinal Aspect of a Basic Multicellular Unit (BMU) in Action, From

Right to Left, Representing the Formation of an Osteon. Adapted from Robling and
Stout, 2000) .......................................................................................... 68

Figure 3 — From Left to Right, Dell Dimensions 4550 Computer, Sony Camera Adapter
CMA-D2 on Top of a Sony Trinitron, and a Leica M2 12 Stereomicroscope Attached to a
Computer. Sigma Scan Pro 5.0 Program is on Screen of Computer. Michigan State
University, Forensic Anthropology Laboratory ............. ‘ .................................... 69

Figure 4 - Example of Anterior, Medial, Lateral, and Posterior Areas of Bone Used in
Analysis of Femoral Midshaft in Kerley-Ubelaker Method (1978) ......................... 70

Figure 5 — Example Of Fragmentary Osteon (Outlined in Red) on a Digital Image at 200x
in Association with an Intact Osteon (Indicated by Arrow). Adapted from Tersigni
(2005) ................................................................................................. 7 1

Figure 6 — Example of Kerley-Ubelaker (1978) Analysis. Circles are Identified
Haversian Canals (10 present) and Lines Indicate Osteon Fragments (5 present). Image
Not Used in Analysis ............................................................................... 72

Figure 7 — Periosteal and Endosteal Surfaces of Human Bone. Macroscopic view (Left)
of Femoral Midshaft Showing the Periosteal and Endosteal Bone Surfaces. Microscopic
View (Right) of Human Rib (100 X) Showing the Periosteal and Endosteal Surfaces
(right). Adapted from Tersigni (2005) .......................................................... 73

Figure 8 — Example of Areas of Bone Used in Analysis of Femoral Midshaft in Hauser et
a1. Method (1980). P: Periosteal Equations (Three Averaged Areas Each of Anterior,
Medial and Lateral) and E: Endosteal Equations (Three Averaged Areas Each of
Anterior, Media] and Lateral) ..................................................................... 73

Figure 9 — Microscopic (400x) View of Haversian Canal (Arrow) of a Human Tibia.
Adapted from Tersigni (2005) .................................................................... 74

Figure 10 — Example of Hauser et al. (1980) Analysis. Circles are Identified Haversian
Canals (29 present) and Lines Indicate l/2 Haversian Canals (4 present). Image Not Used
in Analysis ........................................................................................... 74

Figure 11 - Graph of Correlation Between Real Age (RA) and Kerley-Ubelaker Age
(KUA) ................................................................................................ 76

ix

LIST OF FIGURES CONTINUED

Figure 12 — Graph of Correlation Between Real Age (RA) and Hauser-P Age
(HPA) ................................................................................................. 77

Figure 13 - Graph of Correlation Between Real Age (RA) and Hauser-E Age
(HEA) ................................................................................................. 78

Figure 14 — Graph of Correlation Between Real Age (RA) and Hauser-Avg Age
(HAA) ................................................................................................. 79

Millie!!!

Forensic anthropology is the application of the knowledge of skeletal‘biology to a
medicolegal setting. In 1977, forensic anthropology achieved legal status with the
establishment of the American Board of Forensic Anthropology sponsored by the Physical
Anthropology Section of the American Academy of Forensic Sciences (AAFS) (Kerley,
1978). The goal of the forensic anthropologist is to examine human remains in order to
extract as much information as possible leading to a positive identification or an explanation
as to circumstances surrounding that individual’s death. Beyond the identification of
unknown human remains, research in forensic anthropology has been ongoing since the late
18005. Research has focused on skeletal growth and development as it relates to age and is
applicable to the reconstruction of the biological nature of prehistoric populations as well as
the legal solution of forensic cases (Kerley, 1978).

In forensic science applications as well as those bioarchaeologiCal, the purpose of

the human osteologist is to help answer the following questions (White, 2000; Byers,

2004):
1. Is bone the material analyzed?
2. Are the remains human or non—human?
3. Are the remains modern?
4. What is the minimum number of individuals represented in the sample?
5. What was the age of the individual at death?
6. Are the remains of an adult? If so, what is the sex?
7. What was the height of the individual in life?
8. What is the ancestral affiliation of the individual?
9. Are any pathologies present on the remains?

The methods for the analysis of the biological profile (estimations of sex, age, stature,
ancestry), as well as the rest of the questions posed above, have long been studied and

described in a number of texts. Some of the most notable include those by Krogman

(1962), Stewart (1979), Iscan (1988), Buikstra et a1. (1994), Bass (1995), Larsen (1997),
and White (2000).

The estimation of age at death, an essential element in the creation of the
biological profile, is the subject of the research contained herein. This is important
because establishing the age at death in a human remains case can be a critical
component in positive identification. By narrowing the age range, identification becomes
more likely. Forensic Anthropology typically relies on standard morphological
techniques for analyzing the age at death for both adults and subadults utilizing a
multiplicity of methods that focus on the pubic symphysis (Todd, 1920; McKern and
Stewart, 1957; Suchey and Katz, 1986; Brooks and Suchey, 1986), stemal rib ends (Iscan
et al., 1984; Iscan et al., 1985), the auricular surface of the os coxa (Lovejoy et al., 1985),
cranial suture closure (Meindl and Lovejoy, 1985, Mann et al., 1987) epiphyseal closure
(Buikstra and Ubelaker, 1994), long bone length (Hoffman, 1979; Stewart, 1979), dental
eruption and formation (Schour and Massler, 1941; Moores et al., 1963; Harris and
McKee, 1990; Ubelaker, 1999), and dental wear (Murphy, 1959; Scott, 1979) among
others.

Less pervasive throughout physical anthropology, forensic techniques of bone
histology can also provide methodologies that estimate age at death (Martin et al, 1998;
Robling and Stout, 2000). Histology is the microscopic study of plant and animal tissue
(Ham, 1974). Not commonly used in forensics, histology has nevertheless been the topic
of many innovative techniques of bone analyses for much of the 20‘h and 21St century.
Bone histology can be used to determine if an unknown piece of material is bone, if it is

human, or to identify the presence of pathologies. The estimation of age at death,

however, has been the focus of most histomorphological interest from the anthropological
community of recent years, and multiple methods have been created and employed
(Robling and Stout, 2000). In fact, the first published report of bone histology being used
to estimate the age at death of a individual dates back almost 100 years to 1911
(Balthazard and Lebrun, 1911).

In bone histomorphometry, morphological components on a cellular level, such
as bone cells and wall thickness, are measured (Eriksen, 1994; Kerley, 1965).
Histomorphological research has led to prediction equations on the age at death of human
individuals. These analyses typically depend on the measurements and/or tallies of
variables such as osteons, osteon fragments and Haversian canals.

Generally, histological age estimation techniques utilize either the outer cortex or
the entire bone cross-section. However, while histology can be an accurate method of
age estimation, problems can often arise with individuals whose remains are not fully
present, are fragmented, and/or are otherwise damaged. This results in an inability to
perform the analyses. Additionally, some methods of histomorphology can be
complicated. For example, specific analyses that focus on multiple characteristics require
a higher level of training and experience to achieve success (Ahlqvist and Damsten,

1969; Stout and Gehlert, 1979).

In 1980, an article by Hauser et al. entitled “Identification using
histomorphometry of the femur and tibia” (English translation) was published in the
French periodical, Acta Medicinae Legalis et Socialis. This article focused on a new
method of aging individuals through a histomorphological examination of the midshaft of

the femur. Since then, the only mention of the Hauser method in published material has

been in the 2000 article by Robling and Stout which presented a basic summary of
histological age estimation methods.

The Hauser method (1980) is made notable in part by its reported ease of use, as
only one variable, the Haversian canal, is tallied. Additionally, and probably more
important, the method utilizes two independent equations, one from the subperiosteal
(outer cortex) and one from the endosteal (inner cortex) surface of bone. While it is
always a plus to contribute an additional aging technique to histomorphology, the
exciting possibility of the Hauser method lies in its endosteal equation. According to
Hauser et al., the endosteal equation of bone is actually-more accurate than that of the
subperiosteum. As this author was unable to find any other histological aging methods
that utilize the inner cortex of bone, the article by Hauser et a1. provides a promising
alternative. The ability to analyze only the endosteal surface of bone allows for age
estimation when the subperiosteal surface of bone is damaged. If the endocortical
method developed by Hauser et a1. proves to be an accurate method, femora lacking an
outer cortex will be able to be analyzed for both forensic and archaeological purposes.

The objective of this study is to analyze the Hauser method (1980) of age
estimation in order to determine error rates, reproducibility, and comparability with a
known histological method. The subperiosteal and endosteal equations will be treated as
independent methods along with the average of the two age estimations provided by the
equations. The better—known Kerley-Ubelaker method also utilizes the midshaft of the
femur and will be included in this study for comparative purposes. Its variables focus
only on the outer cortex of the bone cross-section. Since its original presentation in 1965,

the Kerley technique has become a standard histological method for the determination of

age in the field of forensic anthropology (Kerley, 1965; Ahlqvist and Damsten, 1969;
Stout and Gehlert, 1980; Robling and Stout, 2000). In 1978, Kerley and Ubelaker offered
revisions of the microscopic field size for the method of determining age at death in
human cortical bone.

In order to accomplish my goals, I will examine midshaft femoral cross-sections
of 30 individuals of known ages on loan from Dr. Douglas Ubelaker at the National
Museum of Natural History. Predictions of age from each of the Hauser et al. (1980)
equations, as well as the average of the two estimations, will be compared to the actual
ages at death of the individuals. I will also evaluate the Kerley—Ubelaker method (Kerley,
1965; Kerley and Ubelaker, 1978) with this sample. This will allow for a comparison not
only among the Hauser equations, but also in contrast with that of the known Kerley-
Ubelaker. The Hauser et al. method provides an additional option when severe
fragmentation eliminates other histological methods, such as Kerley—Ubelaker. If the
Hauser method, specifically the equation which utilizes the endocortical surface, proves
to be an accurate method for estimating age, this opens the door for future research on the

medullary cavity of bone.

Chapter 2: Histology

Bone Composition and Cell Types

As a type of connective tissue, each of the 206 bones in the human body can be
studied microscopically. Bones are composed of three materials: water, mineral, and an
organic matrix (Martin et al., 1998). While some water in calcified bone is free (flowing
through pore space, not bound to minerals, capillaries, etc), a portion is bound to other
molecules. Sixty-five percent of bone is made up of its mineral aspect. For example,
Martin et al. (1998; p. 40) reports that “the mineralization of an osteoid (the organic
portion of extracellular bone) displaces part of its water” resulting in a change of water
content as new bone mineralizes. The mineral aspect of bone is mostly hydroxyapatite
crystals [Ca10(PO4)6(OH2)]. This substance affects the solubility of bone mineral. The
organic matrix of bone is largely composed of collagen, a structural protein. Collagen
can “spontaneously organize itself into strong fibers” and gives bone its flexibility and
tensile strength (Martin et al., 1998; p. 40).

Bone is a dynamic connective tissue that is in a constant state of change in
reaction to its mechanical and physiological environment. There are four types of bone
cells falling into the categories of those that resorb bone and those that form(ed) bone:
osteoblasts (bone forming cells), osteoclasts (bone-resorbing cells), osteocytes (bone
maintenance) and bone lining cells. Martin et al. (1998; p. 44) reports that osteoclasts
resorb bone by eroding around the border of a cell “at a rate of tens of micrometers per
day by first demineralizing the adjacent bone with acids and then dissolving its collagen
with enzymes”. Osteoblasts produce the organic portion of the bone matrix (osteoid). As

an osteoid calcifies, an osteoblast may become surrounded by the intercellular organic

substance and is transformed into an osteocyte (Ham, 1998). Osteocytes are housed in
lacunae (cavities) and communicate among themselves and osteoblasts through small
tunnels (canaliculi). Similar to osteocytes, bone lining cells are dormant osteoblasts that
“escaped being buried in newly formed bone and remained on the surface when bone
formation ceased” (Martin et al., 1998; p. 48). These cells maintain a route of
communication between themselves and osteocytes and are located on the endosteal and
periosteal surfaces of bone. Bone lining cells, also converted osteoblasts, begin the
remodeling of bone in response to mechanical stress, an extreme example of which is

bone fracturing.

Bone Modeling and Remodeling
Primary and Secondary Bone

Compact bone can be divided into the two subgroups of primary (including
circumferential lamellar bone and plexiform bone) and secondary bone (Martin et al.,
1998). Primary bone is laid out onto the periosteal (outer) surface of bone during growth
and development. Circumferential lamellar bone runs parallel to the surface of the bone
and incorporates blood vessels “such that each is surrounded by several circular
larnellae”, which then form a primary osteon/Haversian system with a primary Haversian
canal (through which blood vessels, lymph fibers and nerves pass) at its center (Martin et
al., 1998; p. 37). To some extent, lamellar bone will be nonrandomly distributed through
a cross-section reflecting an element of variation in the biomechanical loading between

individuals (Robling, 1998).

Secondary bone occurs through remodeling, resulting in the resorption of existing
bone with the creation of new lamellar bone (Martin et al. 1998). The cylindrical
structures in secondary bone are labeled secondary osteons/Haversion systems containing
about 16 lamellae surrounding the Haversian canal. Volkmann’s canals, each conveying
periosteal vessels, connect the Haversian canals to each other or to the marrow cavity and
run transverse (perpendicular) to Haversian canals (Ham, 1998). As part of the aging
process in some species (including humans), the majority of compact bone is replaced by
secondary bone, which results in newly formed osteons as well as whole or partially
absorbed older osteons. Secondary osteons/Haversian systems are circumscribed by
discrete bundles of lamellar bone termed a cement line (Mulhern and Ubelaker, 2001 ).
Trabecular bone also becomes secondary, though osteons are rarely produced in this
environment since most often they do not fit inside of the individual spicules of
trabeculae. Silberberg and Silberberg (1961) report that bone resorption is more
dominant than deposition in bone older than 60 years at death. Examples of lamellar
bone, Haversian canals, and Volkmann’s canals can be seen in Appendix A, Figure 1.
Modeling and Remodeling

Modeling is the sculpting of bones during growth and development (Martin et al.,
1998). More specifically, Ericksen et al. (1994; p. 2) states that bone modeling is “the
process by which the overall shape of bone is changed in response to physiologic and/or
mechanical change”. This process resultsfrom the independent work of cells such as
osteoclasts for bone resorption as well as osteoblasts for the formation of bone. This

results in a change of the overall shape and/or size of a bone. Martin et al. reports that,

while the process of modeling is slow and prolonged, it becomes more quiescent as the
skeleton reaches maturity (1998).

Remodeling is the life-long process by which bone is constantly renewed through
the removal of older portions of bone (resorption) and replaced with that which is newly
formed (mineralization) (Eriksen, 1994; Martin et al., 1998). This is usually in response
. to mechanical stressors (appearing for example, as fatigue damage) in order to “fine tune”
the skeleton. That is to say, remodeling works to keep the skeleton in both a state of
optimum mechanical competence (strength) and Optimum physiologic competence
(repair) (Frost, 1973). According to Frost (1985), the following are factors known to
affect both osteon remodeling and the accumulated osteon populations:

Age, chronologic
Life span

Sex

Maturation, skeletal
Species

Hormones
Electrolyte disorders
Metabolic

Genetic disorders
Toxic agents
Radiation damage
Bone growth

Bone remodeling patterns

The remodeling of a bone, according to Frost, is episodic, occurs throughout life
(though there is a reduction as growth stops), and involves the sequential actions of
osteoclasts (resorb bone) and osteoblasts (form new bone) working in basic multicellular
units (BMUs) (1973). The endosteal surface of bone is characterized by higher activity

through bone resorption than formation (Martin, 1998). It is this deficit in bone that “is

the reason for normal, age-related bone loss that occurs in men and women over the age
of 30-35” (Martin, 1998; p. 66).

Each BMU is an instrument of bone remodeling that consists of about ten
osteoclasts and several hundred osteoblasts that are activated by mechanical or chemical
signals (Martin et al., 1998). In this process, bone structural units are created through the
remodeling of cortical or trabecular bone. In a histological analysis, the extent to which
the “bone remodeling unit . . . resorbs the preexisting cortical bone will determine the
location of the reversal line and hence the area of the osteon” (Pfeiffer, 1998; p. 220).
This can be viewed in Figure 2 in Appendix A. An accumulation of remodeling activity,
as is associated with an increase in age, reflects in an increase in porosity and a decrease
in cortical bone mass (Eriksen, 1994). Tersigni (2005; p. 18) summarizes the six stages
involved the life cycle of an osteon:

1. Activation

Differentiated cells are collected from a precursor cell population and are

moved to the area in which the new osteon is to be formed.
2. Resorption

Osteoclasts resorb bone by moving longitudinally through the bone at a rate of

40nmlday. The BMU has a diameter of 200nm and a resorption area of a few

hundred um long.
3. Reversal

Transition from osteoclast activity to osteoblast activity takes several days, and

the reversal area is the lag space between the resorptive and refilling regions. The

reversal line will eventually form the cement line. Process takes about 30 days.

4. Formation
Osteoblasts appear at the periphery of the newly formed tunnel. They close
approximately 1-2nm per day. They leave a central passage (Haversian canal) for
blood vessels to maintain the BMU, bone matrix and to carry calcium and
phosphorus back and forth. This phase takes 3 months.

10

5. Mineralization

In this phase the osteoid is mineralized. There is a lag time of 10 days between the

deposition of osteoid and the mineralization. Thus there is an osteoid front

between the osteoblasts and the mineralized bone. Primary mineralization occurs
within the first few days after mineralization starts. Secondary mineralization
occurs as more minerals are added over the next several months.

6. Quiescence

After the BMUs have completed their process of remodeling the osteoclasts

disappear and the osteoblasts become osteocytes and maintain the bone.
Thin Sections

When performing a basic histological analysis, bone is examined cross-
sectionally. Although bone can be difficult to slice due to its calcification, several
methods have been developed to create thin sections. Bone is routinely decalcified,
dehydrated and embedded in paraffin before sectioning (Ham, 1998). Decalcified bone
has had its original mineral composition dissolved away. This method is helpful in the
preservation and thus dating of archaeological bone (e. g. 1,000 years B.P.) (Ham, 1988).
However, the canaliculi are no longer observable, making it impossible to view bones’
structural tissue on a dynamically changing level (Martin et al., 1998).

Two additional methods of producing thin sections of bone are available for more
complex analyses. The first method of sectioning undecalcified bone involves the use of
“special hard embedding media and special kinds of microtomes adapted for heavy wor ”
(Ham, 1998; p. 382). Similar to paraffin embedding, the bones are immersed in liquid
solutions, then polymerized and later cut into thin (less than ten micrometers or pm) or

thick (100 micrometers or um) sections (Martin et al., 1998). Bone that is burned,

archaeologically relevant, or otherwise damaged can be dry, weathered and more likely to

11

crumble when sliced. By utilizing this thin sectioning technique, it is protected during
the cutting process. This allows for a detailed analysis on a cellular level.

The older, second method, involves cross-sections termed ground bone sections,
and has been popular since before the advent of plastics in the 19305 (Martin et al.,
1998). The creation of ground bone sections involves the use of a fine saw to cut thin
slices of calcified bone (Ham, 1998). This is followed by the grinding of each section
until thin enough to transmit light. The thickness (50 or 100 micrometers for example) of
the thin sections is varied to suit the needs of the specific research question. Again,
cellular analysis is difficult with this method, but the canaliculi stand out, and the layered
nature of bone and osteocytes are viewable as dark cavities. The latter method has been

utilized in the sample studied for the analysis herein.

12

W

The aging of individuals is most easily determined through morphological
methods. Once growth and development have ceased, however, estimating age at death
can become more difficult. Furthermore, traditional morphological methods for
determining the age can be useless in archaeological cases in which the “most useful
indicators of age are often missing or obliterated in fragmented, eroded, or incomplete
skeletons” (Kerley, 1965; p. 149). Unlike morphological methods of analysis, the
microscopic evaluation of bone generally necessitates destruction of the bone itself.
When determining the age of individuals older than 50 years at death, histological
methods are required (Thompson, 1979). Creating thin sections generally involves
cutting, grinding and polishing a section of bone followed by viewing it through
polarized light. It is argued by Rosing et al. (2007) that while histology is more time
consuming and destructive than morphological methods, it also yields more precise

results.

Specific Examples of Methods

The tally of osteons or their fragments is commonly utilized in determining
chronological age (Maat et al., 2006). Kerley (1965) developed age estimation equations
involving similar tallies by using the femur, the tibia, and the fibula, though the femur is
the only method that is explicitly explained. This was a landmark study for establishing a
method of determining age histologically (Kerley, 1965; Ahlqvist and Damsten, 1969;
Stout and Gehlert, 1980; Robling and Stout, 2000). Ahlqvist and Damsten (1969)

compared the Kerley technique with a modified method they had devised that was less

13

laborious and required less skill. The results showed that both methods were valid,
though Kerley’s technique was more accurate. Later, these methods were again
compared with similar results (Stout and Gehlert, 1980). It was noted that while Kerley’s
method was more accurate, there was far less interobserver error with that of Ahlqvist
and Damsten. A later comparison by Stout and Stanley (1991) examined the accuracy of
determining the percentage of a given area covered by osteons used by Ahlqvist and
Damsten with the total count of osteons in the same given field as identified by Kerley.
Results reported that osteon counts were a more accurate feature of analysis in
histological age assessment than the percentage of cross-section covered.

Samson and Branigan (1987) utilized a contemporary sample from 58 individuals
of European ancestry between the ages of 16 and 91 at death to develop an age estimation
formula. They examined mean cortical thickness, mean Haversian canal diameter, and
the number of Haversian canals per unit area of bone. Sexual dimorphism was found in
these estimations. Female femoral cross-sections were not found to be accurate when
analyzed. This formula was then applied to the femora of Iron Age and Saxon remains to
estimate age at death. The authors reported very little variation when analyses were
replicated. Similar to the Hauser method (1980), Haversian canals were a key aspect of
the research and the method can be used on highly weathered or abraded bone.

Another aging technique involves taking a drilled core sample of bone four
millimeters in diameter in order to minimize destruction (Thompson, 1979). The bone
can then be left intact for subsequent analyses as well as to preserve the remains except
for the hole drilled into it. This method includes the stereological quantification of

secondary osteons and Haversian canals; the data are then applied to regression equations

14

that determine the age of the individual. In Thompson’s analysis, the lower (femur and
tibia) and upper (humerus) extremities were examined and yielded positive results,
although estimations of the lower extremities were identified as marginally more
accurate. Thompson (1981) continued the original study by core sampling a larger
sample (54 individuals) with known ages. The humeral method was determined to be too
inaccurate for identification applications. The femoral and tibial results, although they
exhibited general accuracy, indicated that the analysis was slightly more accurate for
females than males. Additionally, this method was compared among “white”, “black”
and “Eskimo” individuals and indicated a greater accuracy in the determination of age for
skeletons identified as “white” (Thompson, 1981).

Stout et al. (1994), in contrast to earlier methods, developed the use of the stemal
end of the fourth rib in place of long bone analysis. The authors disputed Thompson’s
(1979; 1981) and others’ methods of extracting small portions of bone, such as a core, by
stating that these are not complete representatives of bone. Stout et al. (1994) examined
intact osteon density, fragmentary osteon density, and total osteon population density.
Given the small sample size of the bone, this method is especially intriguing since a cross
section of the entire bone is used. The use of a complete cross-section results in a more
comprehensive assessment of the histology of the bone than with methods such as Kerley
(1965) or Thompson (1979; 1981). Additionally, use of the fourth rib leaves the long
bones intact for other osteometric analyses. The stemal end of the rib is also easily
accessible anatomically, and the method requires only a minimal amount of bone. This
allows a morphological analysis of the stemal end of the rib to still be possible in the

same sample.

15

Analyzed bones do not need to be of recent origin to be useful for age
determination. Stout (1986) utilized a prediction equation of age at death using a rib
sample to identify unknown skeletal remains over 400 years old. The age predicted was
consistent with ages reported through historical documentation. A study by Ericksen
(1997) focused specifically on the use of histology in bioarchaeological settings. The
sample included 122 Preceramic adult skeletons excavated from a cemetery in Chile
dating to approximately 3300 to 3000 BR Previous research on the remains had
developed and assigned morphological ages at death. The anterior cortex from the
midshaft of the femur was then extracted for histological analysis. Results suggested that
the use of histological analyses for the determination of historical age has tended to over-
age individuals (Ericksen, 1997). Ericksen hypothesized that this could be due to
different distributions of lifeways (such as differential environments, diets and/or
physical activities) between past and contemporary populations. Another factor in the
over-aging of individuals could be the application of equations that were initially
determined based on samples of older (e. g. mean = 503) individuals and then applied to
groups that died at younger ages (e. g. mean = 305). Additionally, caution is emphasized
in the analysis of skeletal microstructures as the “exfoliation of peripheral unremodeled
bone lamellae can cause over-aging as deeper, more remodeled areas come to lie nearer
the periosteal surface” (Ericksen, 1997; p. 65).

Overall, there is a variety of histological methods to determine age at death
utilizing bones from throughout the human skeleton. The most accurate methods have
incorporated either the rib or the femur as the location for analysis. Accurate methods of

age estimation for the microscopic analysis of bone have been available for over 40

16

years, but its utility in the forensic world had been surprisingly small (Maat et al., 2006).
With a more complete comparative understanding of aging techniques on past and
present populations, modifications of the methods of age assessment can more accurately
compensate for differences in bone structure. Given the difficulty of the age assessment
however, it is ultimately preferable to consider several indicators of age simultaneously

(Rosing et al., 2007).

Factors Affecting Age Estimation

Bones Utilized in Analysis
When assessing age by analyzing cortical remodeling, the sampling location is

key (Pfeiffer et al., 1995). Methods for the aging of human remains have involved a
variety of bones including the occipital (Cool et al., 1995), second metacarpal (Kimura,
1992), mandible (Singh and Gunberg, 1970; Drusini and Businaro, 1990), fibula (Kerley,
1965; Kerley and Ubelaker, 1978), ulna (Thompson, 1979), femur (Kerley, 1965;
Ahlqvist and Damsten, 1969; Kerley and Ubelaker, 1978; Thompson, 1979; Eriksen,

7 1991; Hauser et al., 1980; Samson and Branigan, 1987;), tibia (Balthazard and Lebrun,
1911; Kerley, 1965; Kerley and Ubelaker, 1978; Hauser et al., 1980), humerus (Iwamoto
et al., 1978; Thompson, 1979; Yoshino et al., 1994), sixth rib (Stout and Paine, 1992),
fourth rib (Stout et al., 1996), and clavicle (Stout et al., 1996). The femoral midshaft,
however, remains the most utilized area of the body (Chan et al., 2007).

Ribs are considered by some to be the ideal site for histological analysis because
an entire cross-section, rather than only a portion of the bone, can be analyzed, as in the

case of long bones (Crowder and Rosella, 2007). Additionally, ribs are noted to reflect

17

the minimal presence of biomechanical variation. Robling (1998) suggested the creation
of an age-at-death formula that takes mechanical loading into account. The applications
of these formulae have limitations, however, because ribs, having thin cortices, are less
resilient to taphonomic processes than long bones. Additionally, various aging
techniques rely on the specific areas (such as the middle third) of a particular rib (such as
the sixth rib). When skeletal remains are discovered with damaged or incomplete rib
sections, the positive identification of the specific ribs can prove impossible. Crowder
and Rosella (2007) have documented intracostal variation in rib histomorphometry. They
reported a high level of variable osteon counts among ribs in the same individual.
Therefore, the use of the rib for age determination is only a good choice when the
specific rib is identifiable and the cross-section is whole.

Other histological age estimation methods include analyses based on human teeth
and infants. Fitzgerald and Rose (2000) interpreted markers of internal microstructural
growth of teeth to estimate age. These markers were noted to be a valid area for age
estimation for situations in which the teeth are complete and other age estimation
methods are unable to be used. There have also been studies assessing the long bones of
infants and children for histological subadult aging (Jowsey, 1960; Kerley, 1965).
Unfortunately, subadult skeletons are thought to be too complex for analysis due to the
stacking of bone modeling and remodeling that makes analyses of cellular based
structures inconclusive (Stout, 1992).

Physical Activity
Pfeiffer (2000, p. 298) reports that “histological patterns of human cortical bone

appear to be somewhat distinctive among mammals, and they appear to vary temporally

18

and spatially”. The estimation of an individual’s age can be biased when the level of
physical activity of the general population analyzed is unknown. Indeed, when the
population upon which a particular method is based is unknown, the level of error
between the groups can be extremely high. If an age estimation method was based on a
highly active group of individuals, and this technique was then applied to one sedentary
individual, the results could be significantly biased. One clear example of this was
provided by Ubelaker ( 1974) who examined the ages at death of individuals from two
Late Woodland ossuaries in Maryland. He compared the gross aspects of the pubis with
the histological aging method developed by Ahlqvist and Damsten (1969). This method
was based on a modern Finnish population. When it was used to estimate the age of
Ubelaker’s sample, the results deviated highly from those obtained by pubic analysis
(1974). Ubelaker offers reasoning that this discrepancy may be due to unequal activity
levels.

Physical activity can significantly increase the error rate of an age estimation. An
animal study by Lieberman (1997) demonstrated that an increased level of mechanical
loading corresponds to an amplification of intracortical bone remodeling. Chan et al.
(2007) caution against the use of the posterior, lateral, and medial locales due to
significant circumferential variation caused by biomechanical loading and unloading.
Hauser et a1. (1980) specifically suggest that the posterior aspect of the periosteal surface
of bone not be included in analyses because of the presence of the linea aspera, a strong

muscle attachment site on the femur.

19

Location on Bone for Analysis

The midshaft is the most common area for analysis in long bones such as the
femur. Chan et al. (2007, p. 81) reports that “cross-sections evaluated from the proximal
region of the femoral cortex will show a negative age estimation bias” and “those taken
from the distal diaphysis will show a positive bias” in comparison with the standard
anterior midshaft age estimates. Tersigni (2005) reports significant inter- and intra-
sectional variation as related to age estimates by the Kerley (1965) method. Tersigni
(2005) determined that the middle two inches (approximately) of midshaft femoral bone
are acceptable for use in cross-sectional age analysis as this area is indicative of the
lowest error rates along its length.
Sex Variation

The idea of variation between the sexes has been present at least since the original
presentation of the Kerley (1965) method of age estimation. Kerley (1965), Stout and ‘
Paine (1992), and Stout et al. (1994) have all reported no difference in age estimation
based on the sex of the individual. Others have cited Kerley’s (1965) conclusion as a
reason for disregarding the sex of individuals altogether in their development of age
estimation methods (Ahlqvist and Damsten, 1969). On the other hand, Singh and
Gunberg (1970), Samson and Branigan (1987), and Yoshino et al. (1994) all only
examined male specimens, leaving the question of variation between the sexes
unaddressed.

Thompson (1979; 1981) and Ericksen (1991) decided to face this question head-
on by creating separate equations for males and females. Thompson (1981) reported

more accurate results in female estimations than in males. Thompson did not, however,

20

check accuracy of the sex equations when the results were pooled. Ericksen (1991) also
found that separate sex equations were more accurate, noting that females accumulate
intact osteons into their sixties while males can accumulate osteons into the one
hundredth year of life. An uneven distribution in the size of osteons between males and
females is an additional consideration relating to sexual variation. In a recent study by
Pfeiffer (1998), however, no sex differentiation in osteon size of the femoral midshaft
and stemal 1/3 of the sixth rib were noted among samples from an English population of
the 18th century, Canadians of the 19th century and South Africans of the 20‘h century.

Although differences in the accuracy of age assessments between sexes have been
reported to be negligible, females have been found to lose cortical thickness at a higher
rate than males (Walker, 1982). A negative correlation was also identified between
cortical thickness and age. Research by Walker (1982) recorded the greatest loss of
cortex in the long bones of females after the fifth decade. However, Robling and Stout
(1998) reported a rapid increase in the osteon density of females immediately after the
onset of menopause. These factors may affect the estimation of age of women older than
50 years at death. However, following Ahlqvist and Damstens’ study (1969), I will not
take sex into account in these analyses. Additionally, neither Kerley-Ubelaker (1978) nor
Hauser et al. (1980) separated their method equations by sex.
Population Variation

Ancestral variation is a notable factor in age determination. Thompson and
Gunness-Heyy (1981) asserted that Eskimos appear to have a higher rate of bone
turnover than European Americans. These researchers applied Thompson’s (1979)

histological method to 19th century Eskimo femora and obtained poor age estimates.

21

Unspecified human cadavers were the original population utilized in order to create the
technique achieved by Thompson (1979). Nothing is mentioned as to ancestral
derivations and variations among these specimens. Thompson and Gunness-Heyy (1981)
concluded that population-specific equations are necessary to account for genetic
differences in the remodeling of bone. Similarly, Ericksen and Stix (1991) used a 19'h
century American sample deriving from Philadelphia to test a technique created by
Ericksen based on a sample from the Dominican Republic. The method was determined
to be successful. However, the age of those falling outside the bounds of the method’s
standard error were generally over-estimated.

Kerley’s (1965) method, and its later revision by Kerley and Ubelaker (1978),
was originally based on a sample of 126 American military personnel said to be
predominantly white and male. Both the original Kerley method and its revision have
also been applied to samples of different ancestries. Fangwu (1983) tested a sample of
modern Chinese femora (n=35) using the original Kerley (1965) method. It was found
that approximately 42% of the estimated ages were within 1 5 years and 67% were within
:1: 10 years of actual age at death. Thirty-three percent of age estimations had error rates
of greater than :1: 10 years.

Additionally,Ubelaker (1981) tested the revised Kerley-Ubelaker (1978) method
on a sample of 114 femora from the Dominican Republic. A sub-sample of these
specimens was analyzed in the research for the present thesis. Ubelaker’s results
indicated that 42% of the age estimations were within :1: 5 years of the actual age at death,
while 62% were within :1: 10 years. Although the rate of error is higher than that

established by the Kerley method (based mainly on American soldiers), it is not high

22

enough to preclude the applicability of the method to a demographic reconstruction. The
large size of the sample analyzed tends to normalize the data.
Destruction of Bone

Another problem in age assessment utilizing an entire cross—section of the mid-
diaphyseal femur is that the act of cutting the cross-section is destructive. This
destruction can preclude analyses of the biological profile (i.e. stature, ancestry) as well
as biomechanical stresses. As Robling (1998; p.163) suggests, “histologic estimates of
age at death that rely on the microstructure of dynamically loaded bones (e. g., femur,
tibia) are affected by the mechanical usage of that limb”. Alternatives to an entire cross-
section of the femur being utilized include some comparably accurate age estimation
methods (Thompson, 1979; Iwaniec et al., 1998; Stout, 1994) that attempt to preserve the
integrity of the bone and/or rely on areas less subject to biomechanical stressors. These
methods accomplish this by drilling a core of bone (Thompson, 1979), disturbing only a
section of the bone (anterior femur) (Iwaniec et al., 1998), or the use of an entire cross-
section (rib) (Stout, 1994).
Intra-Observer Error

Variation in intra- and inter-observer analysis for histological techniques through
the comparison of differing identifying features (secondary osteons, Haversian canals,
osteon fragments) was studied by Lynnerup et al. (1998). The results revealed that the
only feature not resulting in large inter- or intra—observer error rates in the identification
of features was secondary osteons. Some difficulty was noted in positive identification of
Haversian canals through its possible overlapping (through the wide limits of

identification) with osteon fragments. The authors suggested that osteon fragments

23

should not be included in any analysis. Though unstated, it is assumed that Haversian
canals should also be excluded.

There is a key discrepancy in the research of Lynnerup et al. (1998). It did not
analyze the identification of Haversian canals as the sole feature of analysis. It is
possible that when a researcher is identifying only Haversian canals, the lines between
the canals and the osteon fragments are less blurred. In the independent identification of
Haversian canals, it is the thought of this researcher that, through intra-observer analysis,

the Haversian canals will not suffer high rates of error.

24

Chapter 4: Additional Histomorphometric Applications
Forensic anthropology and bioarchaeology have naturally overlapping goals in
histological analyses. This is true beyond histomorphological age estimations. When
bones of an individual are discovered, the same questions are asked whether they died
2 four weeks or four hundred years ago (Who were they? How did they end up here?). As

such, many techniques are applicable to both forensic and archaeological applications.

Bioarchaeological Applications

Bone is often the only type of human tissue to be preserved in archaeological
settings. As such, morphological and metric analyses of the skeleton are often used to
infer the lifeways of past populations. Both Robling (1998) and Pfeiffer (2000) reported
that histology can be used to deduce activity levels in human skeletal remains. Problems
arise when poor bone preservation or extreme fragmentation render these methods
insufficient. For example, when skeletons are excavated in archaeological situations,
bones can often suffer from post-depositional crushing (Boddington, 1987).

Bone histology has developed as a legitimate alternative to morphological and
metric methods. This is due to a recent increase in research in the microscopic analysis
of past populations. For example, Pfeiffer (1998) has used histology to compare Late
Pleistocene human samples with those of nineteenth century cadavers in order to analyze
the diversity between the populations through the histomorphometric identifications of
sex, age, and population/environment. She found that while these patterns appear to be

somewhat distinctive among mammals, they also vary temporally and spatially.

25

The most common argument against the use of histology in the analysis of
archaeological bone has been that alterations over time result in the obliteration of the
microstructure of bone (J ackes et al., 2001). These modifications, examined
histologically, can be separated into four categories according to Cipollaro et al. (1999; p.
3):

1. Good Preservation

The microstructure of bone is “completely preserved and the different structural

elements (osteons, osteon fragments, and interstitial lamellar systems) were

clearly distinguishable.”

2. Intermediate Preservation
There is partial alteration of the bone microstructure.

3. Poor Preservation
The majority of the overall bone microstructure is altered.

4. Very Poor Preservation

The overall microstructure is altered.

Differential bone degradation seems to be caused, at least in part, by the original
deposition of bones into humid and/or variable (non—stable) environments resulting in
microbial breakdown (J ackes et al., 2001). Additional decomposition factors suggested
by Child (1995) include burial temperature, alkaline and acidic soils, microbial
interactions, non-microbial inhibitory substances, extremes of temperature, and enclosed
environments. Interestingly, however, the degradation of bone does not “show a linear
relationship with time” (Jans et al., 2002; p. 349). A study by Bell et al. (1996), who
analyzed the histological cross—sections of eleven individuals, reports that early
postmortem alteration to skeletal tissues can appear as soon as three months after death.
In contrast, a specimen one year after death can remain in excellent microscopic cellular

condition.

26

J ackes (2000) suggests caution in employing histological analysis when analyzing
archaeological bone. All individuals should not be utilized as the process is inherently
destructive. Instead, it is suggested to give an overall consideration to what could be
done to produce the quickest, most accurate and most economical result. While
macroscopically bone may appear healthy, the outer portion of bone can possibly be
unusable for histological analysis on a microscopic level (J ackes, 1992).

Garland (1987) discussed the relationship between burial environment and
histological appearance in an article researching the histological structure of fossil bone.
It was found that bones which have had their histological structure destroyed or altered
may not be used for the purpose of biological aging by several of the current histological
methods. However, Haversian canals were reported as easier to identify than osteons and
osteon fragments in older disinterred bone. Specimens examined by Garland reflect the
possible positive use of Haversian canals for a burial dated to the first century AD.

While these factors can be an impediment in the analysis of archaeological
samples, histology can still be useful and, in some cases, completely applicable. For
example, Cipollaro et al. (1999) used histology alongside DNA amplification in the study
of remains almost 2000 years of age. While some of the bones were completely
degenerated, others in the analysis exhibited microstructures that were indistinguishable
from fresh bone, possessing osteons, osteon fragments, and interstitial lamellar systems.

Within archaeological contexts, bone histology is a useful path for the analysis of
human versus nonhuman remains and the age assessments of individuals. This is

especially applicable after the cessation of growth and development. This field

27

additionally contributes to the study of diet, disease and bone degeneration in present and
past populations.

Histology can be a useful tool in the determination of the preservation of bone
microstructure in heritage management (J ans et al., 2002). Excavation is a destructive
process no matter how careful the archaeologist. When a site is disrupted, it can never be
restored to its previous state, and material remains can be destroyed. It is because of
these factors that many archaeologists in heritage management currently prefer to
preserve remains in the environment in which they had survived since deposition. In
order to make in situ preservation a reasonable alternative to excavation, however, "more
knowledge is required about degradation mechanisms of archaeological materials" and
their influencing factors (J ans et al., 2002; p. 344). Histology can contribute significantly
to these preservation issues. The destruction of bone over time, touted as a weakness in
the use of histology in bioarchaeological research, has been made a strength in this
section of analysis.

The microscopic study of bone diagenesis can contribute more to interpretations
of the state of bone preservation than simple macroscopic evaluations. Thin sections of
bone must meet several criteria for histology to be a useful technique for archaeological
management research. These conditions include an overall “impression of the
preservation status of thebone” and demonstrate relatively small changes therein (J ans et
al., 2002; p. 350). With histology, degenerative changes can “be localized within the
bone, giving a more detailed picture of diagenesis within the bone” (J ans et al., 2002; p.

350). It is important to produce long-lasting, high quality thin sections when analyzing

28

bone for this purpose (Garland, 1989). These sections can provide avenues for research

in the future as new techniques become available.

Human/Nonhuman Identification

The most basic question asked upon discovering bone in archaeological contexts
is, “Is it human?”. This can generally be answered by looking at the morphology of bone
but can become far more difficult in situations of severe fragmentation. Microscopically,
animal bone can be readily differentiated from that of human, as they possess different
cortical and cellular patterning (Gray, 1941). A number of characteristics can be used to
help in this differentiation. Gray (1941) used the number of canaliculi that arise from a
lacuna in femora and parietals to differentiate human from nonhuman bone.

Plexiform bone is a main method of differentiation between human and
nonhuman bone. It is rarely found in humans and is characterized by a mixture of
lamellar and woven bone which often appears as a “brick wall” as it contains rectilinear
spaces that are residuals of past vascularity (Martin et al., 1998; Mulhem and Ubelaker,
2001). Plexiform bone is typical in large animals that have rapid growth rates, such as
cows (Martin et al., 1998).

While plexiform bone is a common distinguishing feature between human and
nonhuman bone, other methods are available. A study by Mulhem and Ubelaker looked
at the differences in osteon banding, defined as at least 5 distinctive rows of primary
and/or secondary osteons (2001). They determined that osteon banding not only occurs
more often in nonhuman bone, but that the bands that appear in human bone are far less

organized. For example, small animals such as rats have fine-fibered bone in their

29

cortices which presents as randomly arranged collagen fibers that are smaller and more
closely placed than woven bone (Martin et al., 1998).

Cattaneo et al. (1999) looked at possibilities surrounding the differentiation of
human from nonhuman bone in the case of badly burned bone fragments. They
conducted a study comparing histological, immunological, and DNA techniques on burnt
bone. This was meant to be an analysis using the “worst case scenario” with
temperatures between 800-1200°C in order to simulate the environment of a house or car
fire. Histologically, bones were analyzed through both the morphology of microscopic
thin sections and a quantitative, metric analysis. For the metric analysis an “algorithm for
histological species identification from fragments of the type of bones commonly found
among burnt remains, as for example in the present forensic cases” was developed (p.
188). Though simulated in severe temperature, the cellular matrix of the bone remained
relatively unchanged and was readily identifiable. This indicates that histological
analysis can be comparable to immunology of the DNA analysis of bone (at least in
reference to human versus nonhuman bone). It is also indicative that histology is
successful, at least to a certain degree, in analyzing burnt bone.

Taking into account osteon shrinkage that occurs with burning, the analysis
reported by Cattaneo et al. (1999) was based on both metric and morphological thin-
section histology. The histological reference data for burnt bone included three variables:
the diameter, perimeter and area of the present osteons and Haversian canals. This
information was then utilized to create an equation for the metric determination between
human and nonhuman bone. Conversely, the biomolecular analysis involved a test for

the presence of human albumin and DNA amplification. The authors concluded that the

30

quantitative histology conducted was far more successful (79% correct) than
biomolecular techniques (9.5% correct) in identifying the human from nonhuman origin

(Cattaneo at al., 1999).

Skeletal Pathologies

Archaeological histology samples can present information on the health of past
populations. Some skeletal pathologies are specific enough to differentiate from others
microscopically (Lovell, 2000). Paine and Brenton’s 2004 study focused on the analysis
and diagnosis of paleopathology through histology. Specifically, pellagra, a niacin
deficiency disease, was examined. The sample included 27 individuals from the
Raymond Dart collection in South Africa who were known to have died either from
pellagra or general malnutrition. The authors examined the relationships among
Haversian canals, lacunae, and secondary osteons of the rib samples. Distinct differences
were found between the two groups. Pellagra was shown to exhibitextremely thin
cortical bone and large Haversian canals in all samples. Other common characteristics
for this disease process include type H and double zonal secondary osteons, and
Howship’s (resorptive) lacunae. This study adds to the “understanding of the diet and
health of maize-dependant populations of both past and present” (Paine and Brenton,
2004, p. 156).

A similar study, presented by Cook et al., focused on the “histomorphometric ,
study of thin femoral head sections of a skeletal sample from the Dakhleh Oasis, Egypt,
dated from circa 36 BC. to 400 AD.” (1988, p. 23). The parameters measured include

the mean wall thickness, trabecular bone volume, surface volume, mean trabecular

31

diameter, and total resorption surface. A high total resorption surface (12.31%) was
noted in relation to hyperparathyroidism. The results highlight histology’s use as an
investigative tool in the detection of disease, specifically hyperparathyroidism in

archaeological samples dating at least 1,600 to 2,000 years B.P. (Cook et al., 1988).

32

Cha ter 5: uestions and H otheses

The main goal of this study is to test the Hauser et a1. method of
histomorphological age at death estimation. The establishment of error rates,
reproducibility and the overall accuracy of the Hauser et al. technique are the main
objectives in this project. Further, it is the intention of this study to examine the
endocortical equation of Hauser et al. specifically as an age estimator. A direct
comparison of the Hauser et al. method to that of Kerley-Ubelaker will showcase the
applicability of this method to one already well established. From this point on, the age
estimated by the average of the Kerley-Ubelaker osteon and osteon fragment ages will be
KUA; that of the subperiosteal Hauser et al. equation will be HPA; that of the endosteal
Hauser et al. equation will be HEA, and the estimated average of both Hauser et a1. age
equations (subperiosteal and endosteal) will be HAA. This study seeks to answer a

multiplicity of questions:

Research Question 1

What are the correlations between real ages from the Ubelaker slide sample and
the predicted ages for each histological method (KUA, HPA, HEA, and HAA)? Does
there appear to be an overall difference?

If the estimations of age based on each method are perfect, the resulting
correlation will be 1.0. The overall correlation between the age estimation of the Kerley-
Ubelaker method (KUA) and the chronological age should be close to 1.0; i.e. a
significant overall difference between them will not be indicated. Also, there will not be

an overall difference in the accuracy of the age estimation of the Hauser et a1. periosteal

33

equation (HPA), the endosteal equation (HEA), and the average of the periosteal and

endosteal Hauser et al. equations (HAA) when compared with the Ubelaker slide sample.

Research uestion 2
How does the real age (RA) from the Ubelaker slide sample compare with each
method’s (KUA, HPA, HEA, and HAA) estimated age? If present, what is the

directionality of these techniques?

Question 2a: Will there be a statistically significant difference in the overall age

estimations by: KUA, HPA, HEA, and HAA?

Question 2a: Hypotheses

Hora: No statistical difference exists between the mean numbers of years estimated from
real for KUA.

Hara: There is a statistical difference in the mean numbers of years estimated from real
for KUA.

Hoza: No statistical difference exists between the mean numbers of years estimated from
real for HPA.

Haza: There is a statistical difference in the mean numbers of years estimated from real
for HPA.

H03a2 No statistical difference exists between the mean numbers of years estimated from

real for HEA.

34

Ha3a: There is a statistical difference in the mean numbers of years estimated from real
for HEA.

H04a: No statistical difference exists between the mean numbers of years estimated from
real for HAA.

Ha4a: There is a statistical difference in the mean numbers of years estimated from real

for HAA.

Ouestion 2a: Expected Outcome
It is unknown if there will be a statistically significant difference in histological
age estimation generated in the overall usage of the Kerley-Ubelaker as well as the

Hauser-P, Hauser—E, and Hauser-Avg (HPA, HEA, HAA) Hauser et al. equations.

Question 2b: Is there a directionality of each method of age estimation (KUA, HPA,
HEA, and HAA) from the actual ages of individuals at death? If there is a directionality,
is it significant? Do the individual methods underestimate or overestimate the actual age
at death of individuals?

The expected outcome of the overall presence of directionalin is unknown for
each of the techniques analyzed (KUA, HPA, HEA, HAA). According to Ubelaker
(1981), the Kerley-Ubelaker method should prove to both underestimate and
overestimate age. The Hauser-P should prove to overestimate age and Hauser-E should
underestimate age based on preliminary research by the author. Similarly, the averaged
results of Hauser-P and Hauser-E (HAA) should prove to generally cancel out

underestimations and overestimations.

35

Research uestion 3
How do the histological aging techniques (KUA, HPA, HEA, and HAA) compare

among each other?

Question 3a: What are the correlations among KUA, HPA, HEA, and HAA? Does there
appear to be a difference?

The intercorrelation rates among the four methods when analyzed with the
Ubelaker slide sample should be close to 1.0 as all methods are expected to have

similarly high correlation rates when compared to the actual ages of individuals.

Question 3b: Is there a statistically significant difference in the estimations of age among

KUA, HPA, HEA, and HAA?

Research Question 3b: Hypotheses

Hor: There will be no statistical difference in the mean difference between KUA and
HPA.

Hal: There will be a statistical difference in the mean difference between KUA and
HPA.

H02: There will be no statistical difference in the mean difference between KUA and
HEA.

Ha2: There will be a statistical difference in the mean difference between KUA and HEA.

H03: There will be no statistical difference in the mean difference between KUA and

HAA.

36

Has: There will be a statistical difference in the mean difference between KUA and HAA.

H04: There will be no statistical difference in the mean difference between HPA and
HEA.

Ha4: There will be a statistical difference in the mean difference between HPA and HEA.

H05: There will be no statistical difference in the mean difference between HPA and
HAA.

Has: There will be a statistical difference in the mean difference between HPA and HAA.

H06: There will be no statistical difference in the mean difference between HEA and
HAA.

Has: There will be a statistical difference in the mean difference between HEA and HAA.

Question 3b: Expected Outcomes
The statistical difference among each of the histological aging techniques (KUA,

HPA, HEA, and HAA) is unknown.

Research Question 4

What will the intra—observer error correlation rate be in the analysis of the age
estimation equations (KUA, HPA, HEA, and HAA)?

If no intra-observer error is found, then the r-value of the correlation will be 1.0.
However, if the rate of intra-observer is high, the correlation will be closer to 0.0. The

correlation rate of intra-observer error in this study is unknown.

37

SEW
Sample

The sample in this analysis consists of slides on loan from Dr. Douglas Ubelaker
at the National Museum of Natural History in Washington, DC. The sample was
originally collected in the Dominican Republic in 1974. It encompassed the known age
and sex of a total of 158 skeletons representing underprivileged individuals “removed
from local cemeteries as a solutionto space problems” and were “socially classified as
Black or mulatto” (Ubelaker, 1981; p. 183). These individuals were reported to be
mainly laborers.

Thin sectioned in 1975, ground sections of 114 individuals were successfully
created from the midshaft cross section of the femur. Forty-four of the sectioned samples
were rejected based on:

“(1) displayed loss of bone from the periosteal surface due to erosion in the soil or
fragmentation during the preparation process, (2) displayed excessive infiltration of soil
or organic debris within the periosteal margin of the cortex, (3) were from cross sections

which displayed evidence of pathological disturbance, and (4) were from skeletons of
questionable age at death” (Ubelaker, 1981; p. 184).

Unidentified skeletal remains include a wide spectrum of individuals (sex, ancestry,
height). This diversity also carries over into age at death and, as such, the sample for this
study was chosen to reflect this. From the Ubelaker sample, 30 femoral cross sections
have been chosen (N=30), representing a broad age range of nine to 82 years of age at
death. As the glue utilized for some of the original mountings had begun to deteriorate,
the decision of which samples to choose from the original 114 was based on the age of

individuals at death and the overall quality of the Dominican Republic slides. The sex of

38

each individual was recorded, although it is not expected to be a significant variable

when analyzing age histomorphologically (Ubelaker, 1981).

Methods

This research was conducted in the Forensic Anthropology Laboratory in the
Department of Anthropology at Michigan State University. The slides were placed under
a Leica M2 12 stereomicroscope attached to a Dell Dimensions 4550 computer
(Appendix A, Figure 3). A stereomicroscope is the compilation of two microscopes
focusing on one point from different angles allowing for a three-dimensional and
correctly upright and lateral view of an image (neither upside down nor backwards). The
computer program, Sigma Scan Pro 5.0, was then utilized in acquiring the image of each
slide. Each image was captured at 10x with a correction factor multiplied by each of the
tally totals for Kerley-Ubelaker (osteons, osteon fragments, non-Haversian canals, and
percentage present lamellar bone) as well as Hauser et a1. (endosteal, periosteal).
Described below, the program also allowed for a tallying of specific points of interest for
each equation.

Captured images were used to analyze the Kerley-Ubelaker method (KUA), the
subperiosteal equation of the Hauser method (HPA), the endosteal equation of the Hauser
method (HEA), and the average of the Hauser equations (HAA). Each image was then
saved on the computer as the plain image (no field size) by way of a Sony camera adapter
CMA-D2 and viewed on a Sony Trinitron (Appendix A, Figure 3). Each captured image
was rectangular in shape (1.30 x 1.00 mm). As the fields for both Kerley and Hauser are

circular, a correction factor was then multiplied by the final tally for each individual in

39

each feature analyzed. My analysis was accomplished through the use of both the
captured computer image and the microscope. Each method was performed twice for the I
30 individual samples. These actions were performed in order to help ensure minimal
intra-observer error as well as provide easy reference of the analyses.

As Stout and Gehlert have stated, “the histomorphometrics of bones from
different individuals can be quite similar, especially if the individuals are similar in age”
(1979; p. 803). However, through the combination of age analysis due to cortical
remodeling, the osteon density of bone, reflections of biomechanical stressors due to
physical activity, and cortical rift patterns from altered biomechanical forces can result in
the individualization and positive identification of bone. The above authors accede,
however, that bone histomorphology requires a great deal of experience (Stout and
Gehlert, 1979; p. 804). As such, it should be noted that this author does not possess

extensive experience in histological age analysis and is partially self-taught.

Kerley- Ubelaker

The 1978 Kerley-Ubelaker revision of the original Kerley method (1965) was
utilized in this study. The sample included in this study is a subset (n=30) of the one in
which Ubelaker identified the error rate in examining the Kerley-Ubelaker method
(N=114 femora) averaging 10.43 years (1981). This analysis, however, did not specify
the variable used for the age estimations (see possibilities below). The author will take
into account past analyses and note it in the discussion section.

Each image was captured with Sigma Scan Pro 5.0 through the Leica M2 12

stereomicroscope at 100x. The anterior, posterior, medial, and lateral aspects of each

40

slide of midshaft femoral bone were assessed. Each area was examined within the
constraints of the field size for osteons (at least 80% complete), osteon fragments, and
lamellar bone and primary osteons. Osteons, osteon fragments, and primary osteons were
counted (number present tallied) and the percentage of present lamellar bone estimated
and placed into separate equations. According to a 1982 study by Stout and Gehlert
based on individuals aged 13-51 years, non-Haversian canals significantly underestimate
the actual age of individuals while osteon and osteon fragments will hold greater weight
in the final analysis. The number of present non-Haversian canals was collected but did
not receive weight in the overall age analyses. This was also true for the percentage
present of lamellar bone consistent with analyses by Stout and Gehlert (1980; 1982). It
was decided that the percentage would not be included due to the variability inherent in
collecting the data (estimating percentages). Additionally, Kerley-Ubelaker identify both
non-Haversian canals and percentage of present lamellar bone as possessing the highest
level of error (1978). As such, the regression equations utilized for statistical analysis of

the Kerley-Ubelaker method only used those of osteons and osteon fragments.

Table 1. Regression equations for Kerley-Ubelaker (1978) age at death estimation
(Y = Age in years; X = Osteon or Osteon fragment count)

 

 

 

 

 

Variable Formula

Osteon Age (0A) Y = 2.278 + 0.187X + 0.00226X2

Fragments Age (FA) Y = 5.241 + 0.509X + 0.017X2 - 0.00015X3
Osteon/Osteon 2 s 2 3
Fragment Average Y— (2.278 + 0.187X + 0.00226X )+ (52.241 + 0.- 09x + 0.017x - 0.000190
(KUA) '-

 

 

 

 

41

The field size directed by Kerley-Ubelaker (1978) is circular with an area of 2.06
mm2 (diameter of 1.62 mm). The area of the stereomicroscope field was rectangular with
an area of 1.30 mm2 (length by width of 1.30 mm x 1.00 mm). The correction factor was
then determined to be 2.06 mm2 -:- 1.30 mm2 = 1.585. The total tally number of each
feature was then multiplied by 1.585 and entered into the specific regression equation
(see Table 1). The age estimations for each regression equation were then averaged to

obtain a single age for the individual.

Hauser et al.

For the 1980 Hauser et al. method, the area of the circular field was decreased to
1.056mm2 (diameter 1.16 mm). The rectangular area of the stereomicroscope field of
1.30 mm2 (length by width of 1.30 mm x 1.00 mm) was then utilized with the Hauser
field size to determine the correction factor (CF). The equation involved was 1.056 mm2
-:- 1.30 mm2 = .812. The total tally for each of the two equations provided by Hauser et
al. was then multiplied by the CF. The resulting corrected tally was entered into each 'of
the regression equations (see Table 2). Two areas of each bone (subperiosteal and
endosteal) were utilized for these separate equations. As such, they were treated as
separate methods and are detailed below. The Hauser et al. technique of histological
aging utilizes three aspects for each of the equations of the femoral thin section with an
additional three sub areas for each of the aspects.

SubPeriosteal Surface
Hauser et al. (1980) decided not to use the posterior aspect of the femur for their

subperiosteal equation. This, they state, is due to the large variations in the size and

42

number of osteons for the posterior region as the linea aspera is a major muscle
attachment site. As such, the three areas of interest for the subperiosteal equation are
anterior, medial, and lateral. As with the Kerley-Ubelaker method, the field size was
placed onto a calibrated image captured by Sigma Scan Pro 5.0 through the Leica DM
2500 light microscope. Three areas (circular field size) of each aspect (anterior, medial,
and lateral) were then analyzed. The number of present Haversian canals were counted
within the given field size. If a canal was only half present, it was counted as one half in
the tally. If the Haversian canal was irregularly shaped, practically destroyed, or the
diameter was greater than 200 microns (ym), it was eliminated from the counts total.

The three areas of each aspect were then averaged, multiplied by the CF, and entered into

the periosteal surface equation (HPA).

Table 2. Regression equations for Hauser et al. (1980) age at death estimation
(Y = Age in years; X = Haversian canal count)

 

 

 

 

 

 

 

 

 

Variable Formula
Hauser-P Age (HPA) Y = 4.86X — 22.22
Hauser-E Age (HEA) Y = 7.96X — 45.49
Hauser-Average f4.86X — 22.22) + (7.96X — 45.49)
(HAA) Y = 2

Endosteal Surface

The posterior, medial, and lateral aspects of the endosteal surface of the femoral
midshaft were analyzed. Three areas (circular field size) of each aspect were then

examined through the quantification of present Haversian canals. This was followed by

43

the average of counts of each aspect analyzed. Additionally, the same constraints for the
periosteal equation remained (i.e. one half of a canal present was counted as one half and
irregularly shaped Haversian canals, those practically destroyed, and those with a
diameter greater than 200 microns (ym) were eliminated). The average of the aspects
themselves (i.e. posterior, medial, lateral) were then determined, the CF applied, and

entered into the endosteal equation (HEA).

Inna-observer error

For the purposes of this research, intra-observer error was also tested. Lynnerup
suggested that the tally of osteons was likely to be consistently identified (1998). Past
research suggests that high rates of intra-observer error with osteon fragments and
Haversian canals are more likely than low rates. As noted earlier in this thesis, however,
the error rates for the identification of Haversian canals as the sole feature of interest
were not analyzed (Lynnerup et al., 1998). As such, the estimate of likely intra-observer
error cannot be stated at this point for the identification of Haversian canals. As each
slide was analyzed twice, it was decided that, because 0f the additional experience
provided by the first estimation, the second set of estimations would be the one involved

in the statistical analyses (except for analysis of intra—observer error).

Statistical Analyses
A database was created on SPSS 14.0 containing burial number, sex,
chronological age at death, corrected Kerley-Ubelaker osteon age, corrected Kerley-

Ubelaker osteon fragment age, Kerley-Ubelaker non-Haversian canal age, Kerley-

Ubelaker percentage of lamellar bone age, Kerley-Ubelaker osteon/osteon fragment
average age (KUA), Hauser et al. periosteal surface age (HPA), Hauser et a1. endosteal
surface age (HEA), and Hauser et al. endosteal/periosteal surface average age (HAA). In
order to better understand the relationship between the chronological age of individuals
and the estimates created by each histological method (KUA, HPA, and HEA), Pearson’s
r correlations were run through SPSS. Additionally, in order to determine intra-observer
error, the Pearson’s r correlations were run comparing the first attempt of each technique
with the second. Following the accepted rate of error reported by Weinberg et al. (2005),
intraobserver error measurements greater than p = .05 (5%) are considered imprecise for
the purposes of this analysis.

The correlation coefficient “expresses the ratio of the covariance in x and y to the
product of the variation in x and the variation in y” (Bachman and Paternoster, 2004; p.
462). In other words, it indicates the magnitude and direction (positive or negative)
between the variables compared. The more similar the actual age at death is with each
estimate, the more similar the variation of x and of y. Bachman and Paternoster (2004; p.
462) report that “a correlation has a lower limit of -1.0 and an upper limit of +1.0.” If the
real age and estimated age are equal, the Pearson’s r correlation will be 1.

Paired sample t-tests were also run through SPSS 14.0. The intent was to
determine if the means of the real age from the Ubelaker slide sample significantly
differed with each method’s (KUA, HPA, HEA, and HAA) estimated age. Paired t-tests
were also run to determine if a statistically significant difference is present among KUA,
HPA, HEA, and HAA as well as to examine intra-observer reliability. Paired sample t-

tests can only be used with matched pairs (Bachman and Paternoster, 1997). Matched

45

groups or pairs t-tests are based on dependent samples in which the “cases in one sample
are deliberately selected so that they are comparable to the cases in another sample”
(Bachman and Paternoster, 1997; p.667). A paired t—test was used as the variables
compared were the estimated and real ages of the individuals from the Ubelaker sample.
As such, they were both dependent on the growth and development and/or the rates of
degeneration for each of the individuals analyzed in this research study. The t-test of the
matched group then tested the difference between the means for each pair (x = mean; x2 -
x1).

The pairs analyzed were contrasted and examined in order to understand the
statistical significance between the real ages with each of the methods for age estimation
(Kerley-Ubelaker, Hauser-P, Hauser-E, and Hauser-Av g). Upon the completion of the
initial age estimations, it was noted by the author that the youth of some individuals (9-
21) reflected, in some methods, highly inaccurate age estimations. As such, t-tests were
performed on the complete sample population with 29 degrees of freedom (N=30) and a
subset of the sample with 24 degrees of freedom (n = 25). This subset included
individuals older than 21 and was analyzed in order to more accurately estimate the effect

of extreme values of on the aforementioned methods.

46

Chapter 7: Results

Sample Results of Question 1

Research Question 1 asks whether the level of correlation between the reported
actual ages of the individuals at death included in the Ubelaker slide sample analyzed by
the various age estimation methods is equal to 1.0. The Pearson’s r correlation was run
through SPSS 14.0 between real ages at death of the Ubelaker slide sample (RA) and the
ages estimated by the Kerley-Ubelaker methods, the equations for Hauser—P (HPA) and
Hauser-E (HEA) as well as the average estimation created by averaging Hauser-P and
Hauser-E (HAA). When analyzing Pearson’s r, a value of -1.0 is equal to a perfect
negative relationship, a 0.0 value indicates no relationship between the two continuous
variables, and a correlation of 1.0 r value is a perfect positive relationship (Bachman and
Paternoster, 2004). The correlation itself provides an assessment of the strength of the
linear relationship between two variables.

The correlation between RA and KUA is .652, which is positive (Appendix B,
Table 3). This means that the higher the RA of the individual, the higher the reported
KUA. The correlation is significant at the .05 level. The correlation between RA and
age estimation of Hauser—P (HPA) is equal to .690 (p < .05), a slightly stronger positive
correlation with real age than that of Kerley-Ubelaker (Appendix B, Table 4). The results
of the RA and HEA correlation is slightly lower than that of both KUA and HPA with r
equal to .585 (Appendix B, Table 5), but is still significant at the .05 level. The
correlation between RA and the average of HPA and HEA (HAA) is .679, also significant
at the .05 level (Appendix B, Table 6). While the significance level for correlations were

all significant at the p=.05 level, all methods are also strong enough to have significant

47

_ correlations related to p=.01. These results indicate that a strong and positive relationship

exists between the actual age at death and each of the estimates.

Sample Results of Question 2

Statistical significance is the topic of Research Question 2. Paired sample t-tests
were conducted to indicate whether a statistically significant difference could be found in
the overall age estimations for each of the equations and methods utilized: Kerley-
Ubelaker (KUA), Hauser—P (HPA), Hauser-E (HEA), and the total average age
determined by both Hauser equations (HAA). For this research question, the significance
level is the probability of rejecting the null hypothesis that no mean difference exists
between the real ages and the age estimations from each method. All t-tests performed
were two-tailed with p-values of .05. Each p-value measures the strength of evidence
against the null hypothesis, representing the probability that a value will reach or exceed
the given data point under the null hypothesis (Johnson and Bhattacharyya, 2001). The
smaller the p-value identified, the stronger the evidence for and the higher the likelihood
that the results have a statistically significant difference.

Also of interest is the overall directionality of each method. The null hypothesis
states that there would be no statistical difference between the mean numbers of years
estimated from real (RA) by the Kerley-Ubelaker method (KUA). The alternative
hypothesis suggests a difference in the mean number of years for each variable,
indicating a significant over- or under-aging. If a statistically significant difference is
found, the method would not be an accurate age estimator of the Ubelaker slide sample

and would indicate that the result is unlikely to have occurred by chance (Johnson and

48

Bhattacharyya, 2001). However, a significant difference does not necessarily imply that
the difference between the means is large (Bachman and Paternoster, 2004). If no
significant difference is found, we would conclude that the Kerley-Ubelaker method is an
accurate estimation of age when examining the Dominican Republic Ubelaker slide
sample.

When testing the difference between the Kerley-Ubelaker method and real age, the
paired sample t-test indicated a mean age of 42.10 years for RA and 49.13 years for KUA
(Appendix B, Table 7). This is a 7.03 year difference between the two means,-indicating that
the estimate derived from Kerley-Ubelaker is (on average) consistently larger than the real
age. Of the ages estimated by KUA, 23 percent of the estimated ages were within 5 years of
the actual age at death, 46 percent were within 10 years, and 95 percent were within 30 years
(Appendix B, Table 23). The average from RA is 13.69 years (Appendix B, Table 8). This
means that when looking at the difference in absolute values between the actual age at death
and the estimated age for the sample of 30 individuals, the difference is 13.69 years. As is
reflected by Table 8 in Appendix B, the value oft at 29 degrees of freedom was -2.46. The t-
value was then entered into the t distributibn in order to determine the level of significance.
Since the two-tailed significance level was .02, which is less than .05, there is a statistical
difference between the means, and the null hypothesis was rejected at the p-value of .05.
Therefore, Kerley-Ubelaker is not an accurate method of estimating age for the Ubelaker
slide sample.

In order to determine whether the age of the sample affected the overall age
estimation and statistical significance of the Ubelaker slide sample, a paired sample t-test was

conducted including those only older than 21 years of age at death. The null and alternative

49

hypotheses were otherwise identical to those mentioned previously. In Appendix B, Table 9,
the mean ages were adjusted to 47.04 years for RA and 52.09 years for KUA. The mean
difference is noticeably smaller than in the first analysis, when individuals younger than 21
years at death were included. The difference between RA and KUA decreased from a 7.03 to
5.05 years, and was not statistically significant (p=. 12) (Appendix B, Table 10). The
percentage distance from the chronological age at death remained similar (Appendix B,
Table 24) with 24 percent, 44 percent, and 95 percent within 5, 10, and 30 years from real
age, respectively. As such, by excluding those 21 years of age at death or younger, the
Kerley—Ubelaker method as an age estimator is accurate.

Utilizing the same null and alternative hypotheses reflected above, the author found
that the mean age using HPA for n=30 was 46.73 and was also higher than the RA mean of
42.10 (4.63 years) (Appendix B, Table l 1). It was also revealed that 26 percent of the age
estimations by HPA are within 5 years of the reported ages at death of the Ubelaker sample
as is displayed in Table 23, Appendix B. Further, 43 percent were within 10 years and 95
percent were within 30 years of the actual age. The average error from RA for HPA was
13.81 years. The null hypothesis was not rejected because the mean difference in years was
not significant as p=. 16 (Appendix B, Table 12). Therefore, when analyzing the Ubelaker
sample of slides from the Dominican Republic, the Hauser-P equation is an accurate method
for age estimation.

When examined with a smaller subset of individuals, and once again with the
same null and alternative hypotheses stated above, the HPA mean age was approximately
4.82 years older than mean age of RA (Appendix B, Table 13). The percentage of

difference from the actual age at death by the HPA estimate was similar to the previous

50

results: 28 percent within 5 years, 25 percent within 10 years, and 95 percent within 30
years of the reported age at death of the Ubelaker slide sample (Appendix B, Table 24).
This difference is very similar to that of RA and HPA with the complete, 30 individual
sample. In Table 14 of Appendix B, a significance level of .24 is reflected. It is again
greater than the .05, and, therefore, an accurate method for age estimation of the Ubelaker
sample.

The mean difference in years for the endosteal surface of bone was 52.03, 9.92
years above that of RA, 42.10 (Appendix B, Table 15). When examining the accuracy of
each estimate that 13 percent of the slide samples were within 5 years of the reported age .
at death, 30 percent were within 10 years, and 95 percent were within fifty years.
Appendix B, Table 23, reports this data. Additionally, the average error was found to be
high at 19.60 years (Appendix B, Table 16). As p=.02, a significant difference was found
between RA and HEA. This indicates that HEA is not an accurate method for estimating
age for the complete sample of the Dominican Republic Ubelaker slides.

When excluding individuals aged 21 and younger and reflecting the previously
stated hypothesis, the mean of HEA was 57.35, still approximately 10.31 years greater
than that of RA, 47.04 (Appendix B, Table 17). Unique to this method, with the smaller
sample excluding the youngest five individuals from the Dominican Republic slide
sample, the accuracy decreased. Only four percent (i.e. n=1) of the sample was estimated.
to be within 5 years of its actual age, and 24 percent were within 10 years (Appendix B,
Table 24). On the other hand, 95 percent of the age estimations remained within 50 years
of the reported ages at death. A significant difference was found (p=.04) (Appendix B,

Table 18). Consequently, while age does not seem to be a factor in the overall estimation

51

of HEA, it'is still not an accurate method in determining age at death for the Ubelaker
slide sample.

A paired sample t-test was also performed for the average age estimations of both
Hauser equations (HAA). Again, the null hypothesis states that there will be no
significant difference in the mean number of years estimated from real for the Hauser-
Avg. The difference between the means of HAA and RA was 7.28 (Appendix B, Table
19). Distances from the actual ages at death of the Dominican Republic slide sample are
reported in Appendix B, Table 23, were 23 percent within 5 years, 43 percent within 10
years, and 95 percent within 40 years of the actual age at death. The t-value was -2.23
with 29 degrees of freedom. The average error rate was found to be 14.06 years
(Appendix B, Tables 20). A statistically significant difference was noted (p=.04) and the
null hypothesis was rejected. The average Hauser method was not found to be an
accurate estimator of age at death for the complete Dominican Republic Ubelaker slide
sample.

When examining HAA in relation to individuals older than 21 years at death, the
Hauser mean was 54.34, 7.30 years greater than 47.04 (Appendix B, Table 21). This is
slightly larger than the mean difference when compared with that of the complete sample.
The percentages of the ages estimated within 5, 10, and 40 years from the reported ages
given in the Ubelaker slide sample, were 20, 40, and 95 percent, respectively (Appendix
B, Table 24). The t-value was reported, as is shown in Appendix B, Table 22, to be -2.00
and the level of significance supported the null hypothesis at .06. As such, the Hauser-
Avg method, when taking into account individuals older than 21 years of age at death in

the Ubelaker sample, is an accurate age estimator.

52

The second part of Research Question 2 was focused on the overall directionality of
each method. It was found that all of the methods used to estimate age (KUA, HPA, HEA,
HAA) reflected negative mean differences, indicating a consistent overestimation of the age
of the sample individuals as a whole (Appendix B, Table 25). As such, there was a clear,
positive directionality to all of the methods. The differences ranged from 4.63 (HPA) to 9.93
(HEA) years difference.

When examining the mean differences for the sample subset of 25 individuals older
than 21 years, the overestimation persists. The age estimation methods of the highest and
lowest mean differences from real remain the same as with the complete sample of 30. The
differences range from 4.28 (HPA) to 10.31 (HEA) years (Appendix B, Table 25, Table 26).
KUA reflects a smaller mean difference from 7.03 down to 5.05, a difference of almost two
years. The Kerley-Ubelaker method (1965, 1978) seemed to clearly indicate a vulnerability
to individuals 21 years of age or younger at death. The equations developed by Hauser et al.
(1980) were not affected as extremely, with HEA gaining a mean difference, in years of .39.
HPA, in difference, grows closer to the RA (as with KUA) by .35 years. As HEA reflected a
negative movement from the complete sample analysis to the smaller age-affected subset and
HPA displayed a positive movement of roughly the same distance, the difference between the
two samples reflected an adjustment of only .02 years for HAA. For the purposes of this

analysis, all Hauser equations are less affected by youth than those of Kerley—Ubelaker.
Sample Results of Question 3

The first part of Research Question 3 examines the correlations between the

methods of age estimation themselves. The table containing the results of each of these

53

analyses can by found in Appendix B, Table 27. When comparing KUA and HPA, the r
value is high at .910. This indicates a very strong positive correlation between the
estimations of age by KUA to HPA. The correlation between KUA and HPA is almost
1.0, indicating a nearly perfect positive correlation with little difference indicated. The r
value for KUA and HEA, however, is much lower, although still strong at .688. The
Pearson’s r correlation between KUA and the total Hauser equations of HAA is, at .847,
also strongly positive. When examining the correlations of the methods with HPA, the r
value with HEA was found to have a istrong positive correlation at .740 and a very strong
positive correlation with HAA at .922. Pearson’s r value is the highest and strongest
when comparing HEA and HAA at .943. As HAA is the average of HPA and HAA, this
correlation and lack of differentiation is expected. All intercorrelations were found to be
significant at p = .05. These correlations are also strong enough to be significant at p =
.01.

In order to measure the level of significant differences among KUA, HPA, HEA, and
HAA for the Ubelaker’s Dominican Republic sample, paired t-tests were run through SPSS
14.0. The null hypothesis for each held that no statistically significant difference would be
found between the mean values of each method. The alternative hypothesis would reflect a
statistically significant difference between the values of the methods. These hypotheses were
found to be valid for all analyses under Research Question 3. If the level of significance was
found to be lower than the p-value of .05, the null hypothesis would be rejected and would
note a statistically significant difference between the variables (methods).

The differences between the age estimated by Kerley-Ubelaker (KUA) and all

possibilities of those estimated by Hauser (HPA, HEA, HAA), were of primary interest. Are

54

the two methods significantly different from one another? KUA and HPA reflected a mean
positive difference of 2.40 years, generally estimating age as slightly higher than Kerley-
Ubelaker (Appendix B, Table 28). The t-value was 1.3 with 29 degrees of freedom. The
significance level of this two-tailed test was .22, larger than the pre—stated p-value of .05, so
the null hypothesis was not rejected (Appendix B, Table 29). As reflected in Appendix B on
Table 30, the t-test was again run comparing KUA with HEA, indicating a mean negative
difference of -2.90 years. Resultantly, the endosteal Hauser equation will be likely, with the
Ubelaker slide sample, to underestimate those of Kerley-Ubelaker by almost three years.
The significance level was .44, well over the p-value and, once again, the null hypothesis
fails to be rejected (Appendix B, Table 31).

The mean difference is noticeably affected when comparing KUA with HAA (the
average of HPA and HEA). As the KUA - HPA mean difference was slightly positive, the
KUA - HEA mean difference was slightly negative, and the distance for the differences were
very similar, the mean difference between KUA and HAA was only -.25 years (Appendix B,
Table 32). The t-value was -.11 for the 29 degrees of freedom and, as displayed on Table 33
in Appendix B, the significance level was high at .92. No statistically significant difference
was found between the age estimations of Kerley-Ubelaker and Hauser for the Ubelaker
sample. Also reflected in the significance levels is the fact that the average of the two Hauser
equations is most similar to those of Kerley-Ubelaker.

Paired sample t-tests were then run to identify statistically significant differences
within the Hauser method. HPA and HEA had a negative mean difference of -5 .30 years
(Appendix B, Table 34). The t-value was -1.53 at 29 degrees of freedom (Appendix B,

Table 35); The p-value remained .05 and was again lower than the significance level of

55

.14. The null hypothesis was not rejected and no significant difference between the
endosteal and periosteal age estimations found. As HPA and HEA are averaged together,
the significance levels for each when compared to HAA are equal at .14, again greater
that .05 (Appendix B, Tables 37 and 39). The t-values were inversely equal at

-1.53 (HPA/HAA) and 1.53 (HEA/HAA) (Appendix B, Tables 36 and 38). All Hauser
methods have the same significance level of .14. As expected, there were no statistically
significant differences found between Hauser-P and Hauser-E.

Correlations examine the linear relationships between two variables. On the other
hand, statistical significance endeavors to address the specific idea proposed by each
hypothesis. The significance level is the probability of rejecting the null hypothesis when
it is true. The null hypothesis in this scenario holds that no statistically significant
difference will be found among KUA, HPA, HEA, and HAA. While correlations can
give one an idea of the overall relationship between two variables, it cannot state the level
of significance between them. This is reflected by the analyses for Research Question 3.
For example, the correlation between KUA and HPA was .910, and the t-test reported a
significance level was .217. Alternatively, KUA and HEA was .847 while the paired t-
test had a significance level of .917. The differences between correlation value and
significance level indicate that while the 1 to l correlation between KUA and HPA may
be higher than that of KUA and HEA, so, too, is the likelihood of incorrectly rejecting the
null hypothesis. This author believes that these differences are due to the small sample

size included in the study.

56

Sample Results Question 4

Pearson’s r correlations were run to determine the intra—observer error reliability of
the author. Each age estimation (KUA, HPA, and HEA) was performed twice. The second
estimation was utilized for the previous analyses. Each estimate will be termed 1 or 2
respectively in accordance with the first and second analyses [Kerley-Ubelaker Age 1
(KUAl) and Kerley-Ubelaker Age 2 (KUA2), Hauser-P Age 1 (HPAl) and Hauser-P Age 2
(HPA2), Hauser-E Age 1 (HEAD and Hauser-E Age 2 (HEA2), Hauser-Avg Age 1 (HAAl)
and Hauser-Av g Age 2 (HAA2)]. The correlations were determined in the same manner as
Research Question 1. The Pearson’s r value between Kerley-Ubelaker Age 1 (KUAl) and
Kerley-Ubelaker Age 2 (KUA2) was determined to be .907 (Appendix B, Table 40). This is
an especially strong positive correspondence. The correlations between HPA] and HPA2
was also very strong with an r value of .926 (Appendix B, Table 41). The r value for HEAl
and I-IEA2 is slightly lower though still a very strong positive correlation of .864 (Appendix
B, Table 42). HAA reflects an additional very high correlation between HAAl and HAA2
with an r value of .913 (Appendix B, Table 43). All correlations are significant at the .01

level, well within the acceptable level of p =.05 discussed by Weinberg et al. (2005).

57

SEW

All methods of age estimation examined in this study reflected significantly
positive correlations with the actual age of the individuals. The correlations show the
relationship of variables to one another but are not useful for indicating the overall
accuracy of each method. The correlations are all positive. This indicates that as the real
age (RA) of an individual increases, then, generally, so do each of the age estimation
methods (KUA, HPA, HEA, and HAA). The rank correlations of the methods from
lowest to highest are HEA, KUA, HAA, and HPA. HPA has the highest r value for the
Pearson’s r correlation at .690. The lowest value results from the HEA equation at .585.
HAA is close to the r value of HPA at .679.

The results of the paired sample t-test varied. When examining each method for
the complete sample of 30 individuals, the Kerley-Ubelaker method, the Hauser-E, and
the Hauser-Avg equations were not found to be accurate with statistically significant
differences present in the mean values for each. The Hauser method, when utilizing the
equation of the periosteal surface of the bone, exhibited no statistically significant
difference in the mean number of years from real. When compared to the RA,
directionality was present as well as both strong and in a positive direction for all of the
equations involved in the statistical analyses (KUA, HPA, HEA, and HAA). This means
that for the sample analyzed, it is more likely than not that the age estimation for each of
the methods will be older than that of the actual individual.

By eliminating the youngest individuals from the age estimations (5 individuals
aged 9-21 years at death), results generally tended to be more accurate. There is no

statistically significant difference for the Kerley-Ubelaker method when utilizing the

58

Ubelaker slide sample, and is thus an accurate method for estimating age. The Hauser-P
and Hauser-Avg equations were likewise accurate. However, the Hauser equation based
on the endosteal surface of bone had a statistically significant difference between the
mean ages of real and HEA. This resulted in an inaccurate age estimation for the given
Dominican Republic sample. This makes sense as the endosteal surface of bone is
mainly a site of bone resorption after the ages of 30-35 years (Eriksen, 1994; Martin,
1998). As a result, it would be expected that Hauser-E would be most accurate when the
individuals analyzed are 35 years of age or younger. This is in direct contrast to the
results found with the Kerley-Ubelaker method, which reflected a vulnerability to bias
with a younger sample set.

Discussed in the previous section, the mean difference in years is the average for
the total of the respective age estimations minus the average for the total of the actual
ages at death. The average error in years is the absolute value of the distance for each
estimated age from the real age at death of each individual. For example, an age
estimation of 42 for a 50 year old individual has an absolute distance of 8 years. Thus,
while the mean difference is focused on the total group average, the average error is
concentrated on the individual.

In 1981, Ubelaker estimated the ages of individuals to have an average error of
10.43 years using Kerley-Ubelaker for the complete sample analyzed (N=114) (1981).
This researcher has found the average error of the subset sample (n=30) for KUA to be
13.69 years (Appendix A, Table 8). Reasons for this difference could very well lie in the
decreased sample size. It is also possible that the use of osteons and osteons fragments as

the variables to quantify affected the estimations. As Ubelaker did not mention which

59

variables (osteons, osteon fragments, non-Haversian canals, and percentage of present
lamellar bone) were used in the original Kerley-Ubelaker analysis, it is not known
whether this is a valid factor. The Hauser-P is similar to that of Kerley-Ubelaker with an
average error of 13.81 years (Appendix A, Table 12). Hauser-Avg had a similar average
error at 14.06 years as is reflected in Table 16 of Appendix A. The highest average error
came from Hauser-E at 19.60 years (Appendix A, Table 20).

Statistically, the Hauser-P method was found overall to be most accurate with the
Ubelaker slide sample, as it is not different in a manner that is statistically significant
from the mean difference of the actual age of individuals. This holds true for both the
full sample (n=30) as well as the sample subset (n=25) of individuals older than 21 years.
Both the Kerley-Ubelaker method and the Hauser-Avg equations were not found to be
accurate with the initial sample of 30 but became more accurate when younger
individuals were taken out of the pool for analyses. There is a key difference to note,
however, as the p-value of KUA moves from .02 to .12. This is a large jump in
significance and indicates that those included in the sample that were 21 years of age or
younger had a significant effect on the accuracy of age estimations. Conversely, HAA’s
p-value is much smaller, moving from .03 to .06. This indicates that while youth may
have affected the overall estimation of age, this effect was relatively small. The Hauser-
E equation was found not to be accurate at both sample levels (N=30 and n=25). What is
interesting about the statistical results for this equation is that, unlike all of the others, the
full sample reflects a smaller mean difference from the subset of ages for individuals.
The overall difference between the significance levels for the two sample sizes, however,

was the smallest of all methods analyzed at .02 (.02 for N=30 and .04 for n=25).

60

The comparison among methods was performed in order to gain insight into the
similarities between the Kerley-Ubelaker (1975; 1978) and Hauser (1980) methods
themselves. Also of interest was the level of significant difference between the two
Hauser equations. Pearson’s r correlation was initially run, comparing KUA, HPA, HEA,
and HAA among one another, which exhibited interesting results. Although the initial
correlations for KUA and HPA with RA were .65 and .69 respectively, when correlated
with each other, the r value was incredibly strong at .91. The lowest correlations among
the equations, while still strong, were both related to that of Hauser-E (HEA/KUA, .69;
HEA/HPA, .74). When the paired sample t-test was utilized to examine the similarity
amongst Kerley-Ubelaker, Hauser-P, Hauser-E, and Hauser-Avg, the value of
significance in mean difference was low, keeping with hypotheses stating no significant
difference in the mean values of each age estimation existed. An interesting significance
level occurred between the two methods themselves (i.e. the endosteal and periosteal
equations of Hauser were averaged) as the significance level was .92.

The correlation rates were also utilized in order to determine the intra-observer
error rates. A p-value of .05 was chosen as a statistical limit measuring intra-observer
error (Weinberg et al., 2005). The correlation between the first and second analysis is
strong, with each of the methods at a value of .01, more accurate than that limit given.
The highest correlation is with the HPA method at .93 while the lowest is with HEA at
.86. Since both HPA and HEA consist of counts of present Haversion canals, the
consistency of their identification on the Ubelaker slides appears to be variable. As HPA

is the most consistent correlation method, it is indicative that the identification of

61

Haversion canals for the Hauser method on the subperiosteal surface of bone is
straightforward and simple.

What are some of the possible reasons responsible for differentiation among the
results, and what factors may have affected the estimations of age for each of the
methods? As was stated and addressed in the results section of the statistical analyses
and at the beginning of this discussion, subadult aging seems to have affected the overall
accuracy for the Kerley-Ubelaker method. Indeed, past research has suggested that a
subadult skeleton has been thought to be too complex for analysis due to the stacking of
bone modeling and remodeling for cellular based structures (Stout, 1992).

Another possible affecter of age estimations is sample size. Although the size of
the sample utilized for this study is larger that some, such as Ahlqvist and Damsten
(1969) with a sample of 20, Stout and Gehlert (1980) with 13, and Paine and Brenton
(2006) with 26 individuals, these smaller samples were present mostly for review of
methods rather than a complete analysis. A greater sample size would tend to
“normalize” the population and reduce the effects. of bias and outliers on the overall
estimations themselves due to a multitude of variables, some of which are listed below.

The level and rate of mechanical loads of the individuals analyzed in the Ubelaker
sample from the Dominican Republic is thought to be high as the population was
generally labeled as laborers (1981). As such, the rate of physical activity for these
individuals would be expected to be high. The rates of activity for the contemporary
American sample upon which the Kerley methods is based and the contemporary French
sample upon which Hauser et al. is based, whether sedentary or active, possibly played a

role in the age estimation for each of the methods. Chan et al. (2007) cautions against

62

use of the posterior area of the femur in determining age due to muscle connections at the
linea aspera. The periosteal surface of the Hauser method (1980) does not include the
posterior aspect of the femoral midshaft specifically in order to avoid a skewed age
estimation result. As Ubelaker (1974) mentioned, a discrepancy is possible due to
unequal activity levels among populations, specifically that of the Dominican Republic
sample and those from which the Kerley-Ubelaker (1965; 1978) and Hauser et al. (1980)
methods are based.

Finally, variations among the ancestral populations themselves could also be a
key to the differences in age estimation for the Kerley-Ubelaker (1965; 1978) and Hauser
et al. methods (1980). While Kerley did not mention the specific sample in the original
creation of the equations, it was reported that there was an overall success in applying his
method to skeletal remains from varying populations. These included testing
archaeological samples from Native American skeletons from Kentucky, Florida, and
Virginia, as well as the Aleutian and Philippine Islands with dated ages at death from 500
to 5,000 years before the time of analysis (1965). Holding with the results found in this
study, Ubelaker contested this success, reporting that the method produced high standard
errors with an average of 11 years from the femora of individuals from the Dominican
Republic who were classified as “Black or mulatto” (Ubelaker, 1978). The sample for
the Hauser et al. method was also based on contemporary cadavers, specifically French

(no further ancestral details given) (1980).

63

Chapter 9: Conclusion

The goal of this study was to test the overall Hauser et al. method (1980) as a
usable estimator of age at death. A sample of 30 individuals was collected to better
understand both the relationship between the Hauser (1980) and the Kerley-Ubelaker
(1965 ; 1978) methods of age estimation as well as to examine the accuracy and
similarities between the subequations for the endosteal and periosteal surfaces. This
study examined how Hauser et al. compares with a known and generally accepted age
estimator, Kerley-Ubelaker (Kerley, 1965; Kerley-Ubelaker 1978; Ubelaker, 1981;
Ahlqvist and Damsten, 1969; Robling and Stout, 2000). The Hauser method itself is a
compilation of two separate equations based on the endosteal and subperiosteal surfaces
of bone that the original authors suggest are successful for independent use.
Additionally, the accuracy of each of the methods or equations [Kerley-Ubelaker (KUA),
Hauser-P (HPA), Hauser-E (HEA), and the average of both Hauser equations (HAA)] at
estimating age, was also examined. Pearson’s r correlations and paired sample t-tests
were run through SPSS 14.0, comparing each method or equation (KUA, HPA, HEA, and
HAA). This research was performed in order to more closely examine how Kerley—
Ubelaker (1978) and Hauser et al. (1980) compare as methods, regardless of the accuracy
of the age estimation, with the Ubelaker sample.

When the above methods were examined for accuracy, preliminary results
suggested skewed or inaccurate age estimations of younger individuals (RAS 21 years at
death). As such, a second set of paired sample t-tests were run on a smaller sample that
excluded individuals from the Ubelaker slide sample who were equal to or younger than

21. years at death. This reduced the slide sample to n=25.

64

It was expected that the Kerley-Ubelaker method would both overestimate and
underestimate the age of the slide sample utilized from Dominican Republic (Ubelaker,
1981). This was found to be true; however, KUA as a whole tended to overestimate the
ages of individuals. It was unknown whether the Kerley-Ubelaker method would have
statistically significant mean differences. The expected outcomes were also unknoWn for
each of the t-test analyses when examining the accuracy of each of the Hauser (1980)
estimations (HPA, HEA, and HAA). When observing the intercorrelation of each
method or equation, no expectation was present on the similarities of each method. Intra-
observer error had no preconceived expectations as this author has no past history of
recording intra-observer error.

Hauser et al. (1980) introduced a method that can estimate age at death based on
counting the Haversian canals in two separate locations: subperiosteal and endosteal.
When estimating the age of individuals in the Ubelaker slide sample, the subperiosteal
equation reflected both the highest rate of accuracy of all methods analyzed and the
highest correlations (of KUA, HPA, HEA, and HAA) to real age at death. The
subperiosteal equation for Hauser was determined to be accurate for the Dominican
Republic sample; however, the intra-observer error rates were estimated to be less
accurate. The higher presence of both Haversion canals and osteon fragments could .
explain this and is consistent with results found by Lynnerup et al. (1998).

Conversely, the endosteal surface of the Hauser method was the least accurate and
displayed the lowest r value when analyzed against the actual age at death. These results
are significantly different from those reported by Hauser et al. (1980) who stated that the

equation of endosteal surface was more accurate than that of the subperiosteal. A strong

65

advantage of the Hauser et al. method could lie in its endosteal equation. When the
exterior of bone is damaged and unable to be analyzed, there are few methods of age
estimation that utilize the endosteal envelope of bone. Forensic anthropology and
bioarchaeology could both benefit from continued research in age estimation methods
utilizing the endosteal surface of bone. The sample used in this study represents
underprivileged individuals identified in records as "black" or "mullatto" (Ubelaker,
1981). Although a statistically significant difference was found between the actual age of
the individuals at death and the endosteal Hauser equation, this author hypothesizes that
the differences in physical activity and/or diet is largely responsible for the significant
mean differences.

In this thesis, Hauser et al. has been found to be a valid age estimator and more
accurate than that of Kerley-Ubelaker when applying the periosteal surface equation
(HPA) and the overall average of the Hauser equations (HAA) on the Dominican
Republic slide sample. Unfortunately, the endosteal equation of Hauser (HEA) is not an
accurate method of age estimation with the sample provided by Dr. Ubelaker.
Differences in these variables are factors that affect the rates of bone resorption and
formation along the endosteal surface of bone. As such, the endosteal surface equation of
Hauser et al. should be examined using a similar population as was initially utilized in its
development (1980).

I would not necessarily discount the use of the endosteal equation, however, as
the correlations between the actual age of individuals and age estimation provided by the
endosteal equations have a strong positive relationship. Instead, I would simply caution

its use when other methods of age estimation are available. The factors considered most

66

responsible for differences, according to this author, are small sample size, non—similarity
between sample populations for the Dominican Republic and that upon which the Hauser
method is based, as well as the age of the individuals at death. Future research should
examine the endosteal surface of bone utilizing a larger, European-based contemporary
sample of individuals 35 years of age and younger.

When comparing the Kerley-Ubelaker method to that of Hauser-Av g age, the
correlations were very high, and the mean difference between the methods was less than
a quarter of a year. The methods’ results were nearly identical and indicate a great
similarity between the two. As a next step in comparing the two methods, this author
suggests the Hauser method be directly compared to that of Kerley-Ubelaker using a
contemporary population that is either American (original Kerley sample) or French (the
Hauser sample) (Kerley, 1965; Hauser et al., 1980).

Future research should also specifically examine histological differences between
sexes. While this study did not test for differences between the sexes citing past research
(Kerley, 1965; Stout and Paine, 1992; and Stout et al. 1994), some studies have reported
differences in accuracy rates of estimating age between the sexes (Thompson 1981;
Ericksen 1991). Though not included in this thesis, the original Hauser et al. (1980)
article also included a method of age estimation based on the tibia. Although it was
reported to have a lower level of accuracy than the femoral Hauser method, the tibia
could offer a useful alternative in instances when the tibia is the only bone present for age

estimation.

67

Appendix A

68

Figure 1 — Shaft of Long Bone. Displays Examples of Haversian Systems and
Volkman’s Canals (Above). The Lower Image on the Right Exhibits a Magnified View
of an Osteon. Adapted from J unqueira and Carneiro (2005).

 

Periosteum

 

Endosteum Haversian Canal .
Lamellae Lacuna Cement

   
  

Haversian Canal

 

Osteocytes \ Cellular Processes

69

Figure 2 — The Longitudinal Aspect of a Basic Multicellular Unit (BMU) in Action, From
Right to Left, Representing the Formation of an Osteon. Adapted from Robling and
Stout, 2000).

Cutting cone

Forming
osteon Complete
Osteoclasts osteon
Osteoblasts

. 5* a

{ Resorptive

 

70

Figure 3 — From Left to Right, Dell Dimensions 4550 Computer, Sony Camera Adapter
CMA-D2 on Top of a Sony Trinitron, and a Leica MZ 12 Stereomicroscope Attached to a
Computer. Sigma Scan Pro 5.0 Program is on Screen of Computer. Michigan State
University, Forensic Anthropology Laboratory.

 

71

Figure 4 — Example of Anterior, Medial, Lateral, and Posterior Areas of Bone Used in
Analysis of Femoral Midshaft in Kerley-Ubelaker Method (1978).

 

72

Figure 5 — Example of Fragmentary Osteon (Outlined in Red) on a Digital Image at 200x
in Association with an Intact Osteon (Indicated by Arrow). Adapted from Tersigni
(2005).

 

73

Figure 6 — Example of Kerley-Ubelaker (1978) Analysis. Circles are
Identified Haversian Canals (10 present) and Lines Indicate Osteon Fragments
(5 present). Image Not Used in Analysis.

 

74

Figure 7 — Periosteal and Endosteal Surfaces of Human Bone. Macroscopic view (Left)
of Femoral Midshaft Showing the Periosteal and Endosteal Bone Surfaces. Microscopic
View (Right) of Human Rib (100 X) Showing the Periosteal and Endosteal Surfaces
(right). Adapted from Tersigni (2005).

Endosteal Periosteal

 

Endosteal
Periosteal /

Figure 8 — Example of Areas of Bone Used in Analysis of Femoral Midshaft in Hauser et
al. Method (1980). P: Periosteal Equations (Three Averaged Areas Each of Anterior,
Medial and Lateral) and E: Endosteal Equations (Three Averaged Areas Each of
Anterior, Medial and Lateral).

 

Figure 9 - Microscopic (400x) View of Haversian Canal (Arrow) of a Human Tibia.
Adapted from Tersigni (2005).

 

Figure 10 - Example of Hauser et al. (1980) Analysis. Circles are Identified
Haversian Canals (29 present) and Arrows Indicate V2 Haversian Canals (4
present). Image Not Used in Analysis.

 

76

Appendix B

77

Table 3 — Pearson’s r Correlation Between Real Age (RA) and Kerley—Ubelaker Age
(KUA)

 

 

 

 

 

 

 

Correlations
RA KUA

RA Pearson Correlation 1 .652**

Sig. (2-tailed) .000

N 30 30
KUA Pearson Correlation .652** l

Sig. (2-tailed) .000

N 30 30

**Correlation is significant at the .05 level (2-tailed)

 

 

 

Figure 11 — Graph of Correlation Between Real Age (RA) and Kerley-Ubelaker Age
(KUA)

 

 

 

 

O
0
0 ° 0
60.00 _ °
0
O
5 ° 0 o
x O o O
40.“) _ o o o
O
O
0 0
20.00 ..
I l I I I I
0 00 20.00 40 00 RA 60.00 80.00 100.00

78

Table 4 - Pearson’s r Correlation Between Real Age (RA) and Hauser-P Age (HPA)

 

 

 

 

 

 

Correlations
RA HPA

RA Pearson Correlation 1 .690**

Sig. (2-tailed) _ .000

N 30 30
HPA Pearson Correlation .690** 1

Sig. (2-tailed) .000

N 30 30

 

 

**Correlation is significant at the .05 level (2-tailed)

 

 

Figure 12 — Graph of Correlation Between Real Age (RA) and Hauser-P Age (HPA)

 

 

 

 

100.00 ‘
o
0
80.00 - 0 °
0
O 0

60.00 1 o
5 ° ° ‘
I o O o

0
40.00 ‘ oo o
o 0
20.00 4
o
o o
0.00 .
20.00 40.00 60.00 80.00 100.00

0.00

79

Table 5 - Pearson’s r Correlation Between Real Age (RA) and Hauser-E Age (HEA)

 

 

 

 

 

Correlations
RA HEA
RA Pearson Correlation l .585**
Sig. (2-tailed) .000
N 30 30
HEA Pearson Correlation .585** 1
Sig. (2—tailed) .000
N 30 30

 

 

 

**Correlation is significant at the .05 level (2-tailed)

 

 

Figure 13 - Graph of Correlation Between Real Age (RA) and Hauser-E Age (HEA)

100.00 ,

80.00 .

HEA

40.00 .

20.00 '

0.00 ‘

 

 

 

0 o
o
0
o
o
o

o

0 o
o
o
00
o
o
0
o
I I I I I
0 00 20.00 40.00 80.00 100.00

80

 

Table 6 — Pearson’s r Correlation Between Real Age (RA) and Hauser—Avg Age (HAA)

 

 

 

 

 

Correlations
RA HEA
RA Pearson Correlation l .679**
Sig. (2-tailed) .000
N 30 30
HAA . Pearson Correlation .679** l
Sig. (2-tailed) .000
N 30 30

 

 

 

**Correlation is significant at the .05 level (2-tailed)

 

 

Figure 14 — Graph of Correlation Between Real Age (RA) and Hauser-Avg Age (HAA)

100.00 ,

80.00 .

60.00 i

40.00 .

20.00 -

0.00

 

 

 

o
0
° 0
o
o
o o
o o
o
I I I f 1
0.00 20.00 40.00 80.00 100.00

81

 

Table 7 — Paired Sample Statistics for Real Age (RA) and Kerley-Ubelaker Age (KUA),

 

 

 

 

=30
Paired Sample Statistics
Std. Std. Error
Mean N Deviation Mean
RA 42.1000 30 18.88322 3.44759
KUA 49.1306 30 18.57120 3.39062

 

 

 

 

 

 

Table 8 — Paired Sample t-Test Results and Average Error (years) for Real Age (RA) and

Kerley—Ubelaker Age (KUA), N=30

 

Paired Sample t-Test

 

df

Sig. (2-tailed)

Avg. Error

 

RA—KUA

29

 

 

-2.465

 

.020

 

13.6886

 

 

At .05 Significance level p S .05

 

82

 

 

Table 9 — Paired Sample Statistics for Real Age (RA) and Kerley-Ubelaker Age (KUA),

 

 

 

 

=25, RA>21
Paired Sample Statistics
Std. Std. Error
Mean . N Deviation Mean
RA 47.0400 25 16.56170 3.31234
KUA 52.0885 25 18.01433 3.60287

 

 

 

 

 

 

Table 10 — Paired Sample t-Test Results for Real Age (RA) and Kerley-Ubelaker Age

(KUA), n=25, RA>21

 

Paired Sample t-Test

 

df

Sig. (2-tailed)

 

RA—KUA

 

24

 

-1.618

 

.119

 

 

At .05 Significance level p S .05

 

83

 

 

Table 11 —— Paired Sample Statistics for Real Age (RA) and Hauser-P Age (HPA), N=30

 

Paired Sample Statistics

 

 

 

 

 

 

 

 

Std. Std. Error
Mean N Deviation Mean
RA 42.1000 30 18.88322 3.44759
HPA 46.7310 30 23.89787 4.36313

 

Table 12 — Paired Sample t-Test Results and Average Error (years) for Real Age (RA)

and Hauser-P Age (HPA), N=30

 

Paired Sample t-Test

 

df

Sig. (2-tai1ed)

Avg. Error

 

RA—HPA

29

 

 

-1.453

 

.157

 

13.8116

 

 

At .05 Significance level p S .05

 

84

 

 

Table 13 — Paired Sample Statistics for Real Age (RA) and Hauser-P Age (HPA), n=25,

 

 

 

 

RA>21
Paired Sample Statistics
Std. Std. Error
Mean N Deviation Mean
RA 47.0400 25 16.56170 3.31234
HPA 51.3219 25 22.40213 4.48043

 

 

 

 

 

 

Table 14 — Paired Sample t-Test Results for Real Age (RA) and Hauser-P Age (HPA),

 

 

 

n=25, RA>21 .
Paired Sample t-Test
(if t Sig. (2-tailed)
RA — HPA 24 -1.193 .244

 

 

 

 

 

At .05 Significance level p S .05

 

85

 

 

Table 15 - Paired Sample Statistics for Real Age (RA) and Hauser-E Age (HEA), N =30

 

Paired Sample Statistics

 

 

 

 

 

 

 

 

Std. Std. Error
Mean ' N Deviation Mean
RA 42.1000 30 18.88322 3.44759
HEA 52.0254 30 27.76908 5.06992

 

Table 16 — Paired Sample t-Test Results and Average Error (years) for Real Age (RA)

and Hauser-E Age (HEA), N=30

 

Paired Sample t-Test

 

df

Sig. (2-tailed)

Avg. Error

 

RA-HEA

29

 

 

-2.397

 

.023

 

19.6013

 

 

At .05 Significance level p S .05

 

86

 

 

Table 17 — Paired Sample Statistics for Real Age (RA) and Hauser-E Age (HEA), n=25,

 

 

 

 

RA>21
Paired Sample Statistics
Std. Std. Error
Mean N Deviation Mean
RA 47.0400 25 16.56170 3.31234
HEA 57.3518 25 26.58337 5.31667

 

 

 

 

 

 

Table 18 — Paired Sample t-Test Results for Real Age (RA) and Hauser-E Age (HEA),

 

 

 

n=25, RA>21
Paired Sample t-Test
df t Sig. (2-tai1ed)
RA — HEA 24 -2.133 .043

 

 

 

 

 

At .05 Significance level p S .05

 

87

 

 

Table 19 — Paired Sample Statistics for Real Age (RA) and Hauser-Avg Age (HAA),

 

 

 

 

N=30
Paired Sample Statistics
Std. Std. Error
Mean N Deviation Mean
RA 42.1000 30 18.88322 3.44759
HAA 49.3782 30 24.10868 4.40162

 

 

 

 

 

 

Table 20 — Paired Sample t-Test Results and Average Error (years) for Real Age (RA)

and Hauser-Avg Age (HAA), N=30

 

Paired Sample t-Test

 

df

Sig. (2-tailed)

Avg. Error

 

RA—HAA

29

 

 

-2.230

 

.034

 

14.0593

 

 

At .05 Significance level p S .05

 

88

 

 

Table 21 — Paired Sample Statistics for Real Age (RA) and Hauser-Avg Age (HAA),

 

 

 

 

n=25, RA>21
Paired Sample Statistics
Std. Std. Error
Mean N Deviation Mean
RA 47.0400 25 16.56170 3.31234
HAA 54.3368 25 22.32866 4.46573

 

 

 

 

 

 

Table 22 — Paired Sample t-Test Results for Real Age (RA) and Hauser-Avg Age (HAA),

 

 

 

n=25, RA>21
Paired Sample t-Test
df t Sig. (2-tailed)
RA — HAA 24 - l .958 .062

 

 

 

 

 

At .05 Significance level p S .05

 

89

 

 

Table 23 — Percentages within 5 and 10 years; 95th percentile for Kerley-Ubelaker Age

(KUA), Hauser-P Age (HPA), Hauser-E Age (HEA), and Hauser-Avg Age (HAA)
estimation methods/equations, N=30

 

 

 

 

 

 

 

 

 

Method Percent Within Percent Within 95 Percent
5 years 10 years Within

Kerley-Ubelaker Age 23.33 46 30 years

(KUA)

Hauser-P Age (HPA) 26 43 30 years

Hauser-E Age (HEA) 13.33 30 50 years

Hauser-Average (HAA) 23.33 43 40 years

 

Table 24— Percentages within 5 and 10 years; 95th percentile for Kerley-Ubelaker Age

(KUA), Hauser-P Age (HPA), Hauser-E Age (HEA), and Hauser-Avg Age (HAA)
estimation methods/equations, n= 25, RA >21

 

 

 

 

 

 

 

 

 

Method Percent Within Percent Within 95 Percent
5 years 10 years Within

Kerley-Ubelaker Age 24 44 30 years

(KUA)

Hauser-P Age (HPA) 28 40 30 years

Hauser-E Age (HEA) 4 24 50 years

Hauser-Average (HAA) 20 40 40 years

 

90

 

 

Table 25 — Paired Differences of the Mean for Kerley-Ubelaker Age (KUA), Hauser-P
Age (HPA), Hauser-E Age (HEA), and Hauser-Avg Age (HAA), N=30

 

 

 

 

 

 

Paired Differences of the Mean
Mean
RA — KHA -7.0306
RA — HPA -4.6309'
RA — HEA -9.9253
RA — HAA -7.2781

 

 

 

 

Table 26 — Paired Differences of the Mean for Kerley-Ubelaker Age (KUA), Hauser-P
Age (HPA), Hauser-E Age (HEA), and Hauser-Avg Age (HAA), n = 30, RA>21

 

 

 

 

 

 

Paired Differences of the Mean
Mean
RA — KHA . -5.0485
RA — HPA -4.2818
RA—HEA -10.3ll7
RA - HAA -7.2968

 

 

 

 

91

Table 27 — Pearson’s r Correlation Among Real Age (RA), Kerley-Ubelaker Age (KUA),
Hauser-P Age (HPA), Hauser-E Age (HEA), and Hauser-Avg Age (HAA)

 

 

 

 

 

 

 

 

 

 

 

 

Correlations
RA KUA HPA HEA HAA
RA Pearson Correl. 1 .652** .690** .585** .679**
Sig. (2-tailed) .000 .000 .000 .000
N 30 30 30 30 30
KUA Pearson Correl. .652** 1 .910** .688** .847**
Sig. (2-tailed)' .000 .000 .000 .000
N 30 30 30 30 30
HPA Pearson Correl. .690** .910** 1 .740** .922**
Sig. (2-tailed) .000 .000 .000 .000
N 30 30 30 30 30
HEA Pearson Correl. .585** .688** .740** 1 .943**
Sig. (2-tai1ed) .000 .000 .000 .000
N 30 30 30 30 30
HAA Pearson Correl. .679** .847** .922** .943** 1
Sig. (2-tailed) .000 .000 .000 .000
N 30 30 30 30 30

 

 

**C0rrelation is significant at the .05 level (2-tailed)

 

92

 

Table 28 - Paired Sample Statistics for Kerley-Ubelaker Age (KUA) and Hauser-P Age

 

 

 

 

(HPA)
Paired Sample Statistics
- Std. Std. Error
Mean N Deviation Mean '
KUA 49. 1306 30 18.57120 3.39062
HPA 46.7310 30 23.89787 4.36313

 

 

 

 

 

 

 

Table 29 — Paired Sample t-Test Results for Kerley-Ubelaker Age (KUA) and Hauser-P
Age (HPA) '

 

Paired Sample t-Test

 

df t Sig. (2-tailed)

 

KUA — HPA 29 1.261 .217

 

 

 

 

At .05 Significance level p S .05

 

 

 

93

Table 30 — Paired Sample Statistics for Kerley-Ubelaker Age (KUA) and Hauser-E Age
(HEA)

 

 

 

 

Paired Sample Statistics
' Std. Std. Error
Mean N Deviation Mean
KUA 49.1306 30 18.57120 3.39062
HEA 52.0254 30 ’ 27.76908 5.06992

 

 

 

 

 

 

 

Table 31 — Paired Sample t-Test Results for Kerley-Ubelaker Age (KUA) and Hauser-E
Age (HEA)

 

Paired Sample t-Test

 

df t Sig. (2-tailed)

 

KUA - HEA 29 -.787 .438

 

 

 

 

At .05 Significance level p S .05

 

 

 

94

Table 32 — Paired Sample Statistics for Kerley-Ubelaker Age (KUA) and Hauser-Avg

 

 

 

 

Age (HAA)
Paired Sample Statistics
Std. Std. Error
Mean N Deviation Mean
KUA 49.1306 30 18.57120 3.39062
HAA 49.3782 30 24.10868 4.40162

 

 

 

 

 

 

Table 33 — Paired Sample t-Test Results for Kerley—Ubelaker Age (KUA) and Hauser-

 

 

 

Avg Age (HAA)
Paired Sample t-Test
df t Sig. (2-tailed)
KUA — HAA 29 -. 105 .917

 

 

 

 

 

At .05 Significance level p S .05

 

95

 

 

Table 34 - Paired Sample Statistics for Hauser-P Age (HPA) and Hauser-E Age (HEA)

 

Paired Sample Statistics

 

 

 

 

Std. Std. Error
Mean N Deviation Mean
HPA 46.7310 30 23.89787 4.36313
HEA 52.0254 30 27.76908 5.06992

 

 

 

 

 

Table 35 — Paired Sample t-Test Results for Hauser-P Age (HPA) and Hauser-E Age

 

 

 

 

 

 

(HEA)
Paired Sample t-Test
df t Sig. (2-tailed)
HPA -- HEA 29 -1.529 .137

 

 

At .05 Significance level p S .05

 

96

 

 

Table 36 — Paired Sample Statistics for Hauser-P Age (HPA) and Hauser-Avg Age
(HAA)

 

 

 

 

Paired Sample Statistics
Std. Std. Error
Mean N Deviation Mean
HPA 46.7310 30 _ 23.89787 4.36313
HAA 49.3782 30 24.10868 4.40162

 

 

 

 

 

 

 

Table 37 - Paired Sample t-Test Results for Hauser-P Age (HPA) and Hauser-Avg Age
(HAA)

 

Paired Sample t-Test

 

df t Sig. (2-tailed)

 

HPA — HAA 29 -1.529 .137

 

 

 

 

At .05 Significance level p S .05

 

 

 

97

Table 38 — Paired Sample Statistics for Hauser-E Age (HEA) and Hauser-Avg Age

 

 

 

 

 

 

 

 

(HAA)
Paired Sample Statistics
Std. Std. Error
Mean N Deviation Mean
HEA 52.0254 30 27.76908 5.06992
HAA 49.3782 30 24.10868 4.40162

 

 

Table 39 — Paired Sample t-Test Results for Hauser-E Age (HEA) and Hauser-Avg Age

 

 

 

(HAA)
Paired Sample t-Test
df t Sig. (2-tailed)
HEA — HAA 29 1.529 .137

 

 

 

 

 

At .05 Significance level p S .05

 

98

 

 

Table 40 - Pearson’s r Correlation Between Kerley-Ubelaker Age 1 (KUAl)
and Kerley-Ubelaker Age 2 (KUA2)

 

 

 

 

 

 

Correlations
KUA] KUA2
KUAl Pearson Correlation 1 .907**
Sig. (2-tailed) .000
N 30 30
KUA2 Pearson Correlation . .907** .l
Sig. (2-tailed) .000
N 30 30

 

 

**Corre1ation is Significant at the .01 level (2-tailed)

 

Table 41 — Pearson’s r Correlation Between Hauser-P Age 1 (HPAl)

 

 

 

 

and Hauser-P Age 2 (HPA2)
Correlations
HPAl HPA2
HPAl Pearson Correlation 1 .926**
Sig. (2-tailed) .000
N 30 30
HPA2 Pearson Correlation .926** 1
Sig. (2-tailed) .000
N 30 30

 

 

 

 

“Correlation is significant at the .01 level (2-tailed)

 

99

 

 

Table 42 — Pearson’s r Correlation Between Hauser-E Age 1 (HEAl)

 

 

 

 

and Hauser-E Age 2 (HEA2)
Correlations
HEAl HEA2
HEAl Pearson Correlation l .864**
Sig. (2-tailed) .000
N 30 30
HEA2 Pearson Correlation .864** l
Sig. (2-tailed) .000
N 30 30

 

 

 

 

**Correlation is Significant at the .01 level (2-tailed)

 

Table 43 - Pearson’s r Correlation Between Hauser-A Age 1 (HAAl)

 

 

 

 

and Hauser—A Age 2 (HAA2)
Correlations
HAAl HAA2
HAAl Pearson Correlation 1 .913**
Sig. (2-tailed) .000
N 30 30
HAA2 Pearson Correlation .913** 1
Sig. (2-tailed) .000
N 30 30

 

 

 

 

**Correlation is Significant at the .01 level (2-tailed)

 

100

 

 

BIBLIOGRAPHY

Ahlqvist, J, O Damsten, 1969. “A modification of Kerley’s method for the microscopic
determination of age in human bone.” Journal of Forensic Sciences. 14:205-212.

Bachman, R, R Paternoster, 1997. Statistical Methods for Criminology and Criminal
Justice, 2“(1 ed., New York, NY, McGraw Hill Co.

Bachman, R, R Paternoster, 2004. Statistics for Criminology and Criminal Justice, 2rld
ed., New York, NY, McGraw Hill Co.

Balthazard, Lebrun, 1911. “Les canaux de Havers de l’os humain aux differents ages.”
Anrfles d’hvgie’ne publique et de me’decine le’gfi, 15: 144-152.

Bass, WM. 1995. Human Osteology: A Laboratory and Field Manual, 4th edition,
Columbia, MO, Missouri Archaeological Society.

Bell, LS, MF Skinner, and SJ Jones, 1996. “The speed of postmortem change to the

human skeleton and its taphonomic significance.” Forensic Science International,
82: 129-140.

Boddington, A, 1987. “Chaos, disturbance and decay in an Anglo-Saxon cemetery.” In:
A Boddington, AN Garland, and RC J anaway (edS.) Death. Decay and
Reconstruction: Approaches to archaeology and f0ren_sic science. Manchester,
Manchester University Press, pp. 27-42.

 

Brooks, ST, J M Suchey, 1990. “Skeletal age determination based on the os pubis: a
comparison of the Ascaadi-Nemeskéri and Suchey-Brooks methods.” Human
Evolution, 5: 227-238.

Buikstra, JE, DH Ubelaker, 1994. Standards for Data Collection from Hum_an Skeletal
Remains, Arkansas Archaeological Survey Research Series 44, Fayetteville, AR.

Cattaneo, C, S DiMartino, S Scali, OE Craig, M Grandi, and RJ Sokol, 1999.
“Determining the human origin of fragments of burnt bone: a comparative study
of histological, immunological and DNA techniques.” Forensic Science
International, 102: 181-191.

Cera F, A Drusini, 1985. “Analisi critica e sperimentale dei metodi di determinazione
dell’eta attraverso le microsctrutture ossee. Quademi di Anatomia Pratica
XLI:105-121.

Chan, AHW, CM Crowder, and TL Rogers, 2007. “Variation in cortical bone histology

within the human femur and its impact on estimating age at death.” American
Journal of Physical Anthropology, 132280-88.

101

Child, AM, 1995. “Microbial Taphonomy of Archaeological Bone.” Studies in
Copservation, 40: 19-30.

Cipollaro, M, G Di Bernardo, A Forte, G Galano, L De Masi, U Galderisi, FM Guarino, F
Angelini, and A Cascino, 1999. Croatian Medical Journal, 40. Retrieved
3/13/2006 from http://www.cmj.hr/ 1999/40/3/ 1041 1967.htm.

Cook, M, E Molto, and C Anderson, 1988. “Possible case if Hyperparathyroidism in a
Roman period Skeleton from the Dakhleh Oasis, Egypt, diagnosed using bone

histomorphometry.” American Journal of Physical Anthropology, 75:23-30.

Cool, SM, JK Hendrikz, WB Wood, 1995. “Microscopic age changes in the human
occipital bone.” Journal of Forensic Science, 40:789-796.

Crowder, C, L Rosella, 2007. “Assessment of intra- and intercostals variation in rib
histomorphometry: its impact of evidentiary examination.” J ournagf Forensic
Science, 52:271-276.

Drusini, A, F, Businaro, 1990. “Skeletal age determination by mandibular
histomorphometry.” International Journal of Anthrcmplogy, 5: 235-243.

Ericksen, MF, 1991. “Histologic estimation of age at death using the anterior cortex of
the femur.” American Journal of Physical Anthropology, 84: 171-179.

Ericksen, MF, 1997. “Comparison of two methods of estimating age at death in a
Chilean preceramic population.” International J oum_al of Osteoarchaeology, 7:65-
70.

Ericksen, MF, AI Stix, 1991. “Histologic examination of age of the First African Baptist
Church adults.” American Journal of Physical Anthropology. 85:247-252.

Fangwu, Z, 1983. “Preliminary study on determination of bone age by microscopic
method.” Acta Anthropologica Silica, 2: 142-151.

Fitzgerald, CM, J C Rose, 2000. “Reading between the lines: dental development and
subadult age assessment using the microstructural growth markers of teeth.” In:

MA Katzenberg, SR Saunders (eds). Biological Anthromlogy of the Human
Skeleton, New York, NY, Wiley-Liss, Inc., pp. 187-213.

Frost, HM, 1973. Bone Remodeling and its Relationship to Metabolic Bone Diseases, IL,
Charles C Thomas.

Frost, HM, 1987. “Secondary osteon population densities: An algorithm for estimating
missing osteons.” American Journal of Physical Anthropology. 20:239-254.

102

Garland, AN, 1987. “A histological study of archaeological bone decomposition.” In: A
Boddington, AN Garland, and RC Janaway (edS.) Death, Decay and
Reconstruction: Approaches to archaeology and forensic science, Manchester,
UK, Manchester University Press, pp. 109-126.

Garland, AN, 1989. “Microscopical analysis of fossil bone.” Applied Geochemistry,
4:215-229. '

Gray, DJ, 1941. “Length of lacunae and number of canaliculi in bones of several
mammals.” The Anatomical Record. 81:163-169.

Ham, AW, 1974. Histology, PA, J B Lippincott Company.

Harris, EF, JH McKee, 1990. “Tooth mineralization standards for Blacks and Whites
from the middle southern United States.” Journal of Forensic Sciences, 34: 859-
872.

- Hauser, R, D Barres, M Durigon, and L Derobert, 1980. “Identification using

histomorphometry of the femur and tibia.” Acta Medicinae Legalis et Socialis
30:91-7.

Hoffman, JM, 1979. “Age estimations from diaphyseal lengths: two months to twelve
years.” Journal of Forensic Sciences, 24: 461-469.

Iscan, MY, SR Loth, and RK Wright, 1984. “Age estimation from the rib by phase
analysis: white males. Journal of Forensic Sciences, 29: 1094-1104.

Iscan, MY, SR Loth, and RK Wright, 1985. “Age estimation from the rib by phase
analysis: white females. Journal of Forensic Sciences, 30: 853-863.

Iscan MY, 1988. “Rise of Forensic Anthropology.” Yearboola of Physical Anthropology,
3 1 2203-230.

Iwamoto, S, E Oonuki, and M Konishi, 1978. “Study in the age-related changes of the
compact bone and age estimation: 2. one the humerus.” Acta Medica Kink;
University, 3:203-208.

Iwaniec, UT, TC Crenshaw, MJ Schoeninger, SD Stout, and MF Eriksen, 1998.
“Methods for improving the efficiency if estimating total osteon density in the

human anterior mid-diaphyseal femur.” American Journal of Physg
Anthropology. 107:13-24.

J ackes, M, 1992. “Paleodemography: problems and techniques.” In: SR Saunders, MA
Katzenburg (eds.) Skeletal Biology of Past Pefioples: Research Methods, NY,
Wiley-Liss., pp. 189-224.

103

J ackes, M, 2000. “Building the basis for paleodemographic analysis: adult age
determination.” In: MA Katzenburg, SR Saunders (eds.) Biological
Anthropology of the Human Skeleton. pp. 417—466.

J ackes, M, R Sherbume, D Lubell, C Barker, and M Wayman, 2001. “Destruction of
microstructure in archaeological one: a case study from Portugal.” International
Journal of Osteoarchaeology, 11:415-432.

J ans, MME, H Kars, CM Nielson-Marsh, CI Smith, AG Nord, P Arthur, and N Earl,
2002. “In situ preservation of archaeological bone: a histological Study within a
multidisciplinary approach.” Archaeometay. 44:343-352.

Johnson, RA and GK Bhattacharyya, 2001. Statistics: Principles and Methods. New
York, NY, John Wiley & Sons, Inc..

Junqueira, LC and J Carneiro, 2005. B_asic Histology, 11th edition, New York, NY,
McGraw-Hill, Co.

Kerley, ER, 1965. “The microscopic determination of age in human bone.” American
Journal of Physical Anthropology. 23: 149-164.

Kerley, ER, 1978. “Recent developments in forensic anthropology.” Yearbook of
Physml Anthropology. pp. 160-173.

Kerley, ER and DH Ubelaker, 1978. “Revisions in the microscopic method of estimating
age at death in human cortical bone.” Americzm Journal of Physical
Anthropology, 49:545-546.

Kimura, K, 1992. “Estimation of age at death from second metacarpals.” Zeitschrift ftir
Morphologie und AnthropologLe, 79: 169-181.

Krogman, WM, 1962. The Human Skeleton in Forensic Medicine, 3rd edition,
Springfield, IL, Charles C Thomas Publisher.

Larsen, CS, 2003. Bioarchaeology: Interpreting Behavior from the Human Skeleton,
Cambridge, MA, Cambridge University Press

Lovejoy, CO, RS Meindl, R Mensforth, TJ Barton, 1985. “Chronological metamorphosis
of the auricular surface of the ilium: a new method for the determination of adult '
Skeletal age at death.” American Journal of Physical Anthropology. 68: 15-28.

Lovell, NC, 2000. “Paleopathological description and diagnosis.” In: MA Katzenburg,

SR Saunders (eds.) Biological Anthropology of the Human Skeleton. New York,
NY, Wiley-Liss, Inc., pp. 217-248.

104

Lynnerup, N, B Frohlich, JL Thomsen, 2006. “Assessment of age at death by
microscopy: unbiased quantification of secondary osteons in femoral cross
sections.” Forensic Science International, 159S:SlOO-S103.

Lynnerup, N, JL Thomsen, and B Frohlich, 1998. “Intra- and inter-observer variation in
histological criteria used in age at death determination based on femoral cortical
bone.” Forensic Science Intermitional, 91:219-230.

Maat, GJR, A Mares, MJ Aarents, and NJD Nagelkerke, 2006. “Histological age
prediction from the femur in a contemporary Dutch sample: the decrease of

nonremodeled bone in the anterior cortex.” Joumal of Forensic Science. 51:230-
237.

Mann, RW, SA Symes, WM Bass, 1987. “Maxillary suture obliteration: aging the human
Skeleton based on intact or fragmentary maxilla.” J ournial of Forensic Science.
32: 148-157.

Maples, WR, 1978. “An Improved Technique Using Dental Histology for Estimation of
Adult Age.” Journal of Forensic Sciences. 23: 764-70.

Martin, RB, N Sharkey, and D Burr, 1998. _S_keletal Tissue Mechanics, New York, NY,
Springer-Verlag.

McKem, T, TD Stewart, 1957. “Skeletal age changes in young American males,
analyzed from the standpoints of identification. Technical report EP-45.
Headquarters, Quartermaster Research and Development Command, Natick, MA.

Meindl, RS, CO Lovejoy, 1985. “Ectocranial suture closure: a revised method for the
determination of skeletal age at death based on the lateral-anterior sutures.”
American Journal of Physical Anthropology, 68:29-45.

Moores, CFA, EA Farming, and EE Hunt, 1963. “Age variation of formation stages for
ten permanent teeth.” American Journal of Physical Anthropology. 21:205-213.

Mulhem, DM, DH Ubelaker, 2001. “Differences in osteon banding between human and
nonhuman bone.” Journal of Forensic Sciences, 46:220-222.

Murphy, TR, 1959. “Gradients of dentine exposure in human tooth attrition.” American
Journal of Physical Anthropology, 17: 179-185.

Narasaki, S, 1990. “Estimation of age at death by femoral osteon remodeling: application
of Thompson’s core technique to modern Japanese.” Journal Of the
Anthropological Society Of Nippon. 98:29-38.

105

Paine, R, B Brenton, 2004. “Histological analysis of ribs from a 20‘h century Black
South African population: Differentiating a microstructural pattern for pellagra
and general malnutrition.” American Journal of Physical Anthropology
Supplement, 38: 156. Abstract.

Pfeiffer, S, 1998. “Variability of osteon size in recent human populations.” American
Journal of Physical Anthropology, 106:219-227.

Pfeiffer, S, 2000. “Paleohistory health and disease.” In: MA Katzenburg, SR Saunders
(eds.) Biological Anthropology of the Human Ske_leton. NY, Wiley-Liss, Inc., pp.
287-302.

Pfeiffer, S, R Lazenby, J Chiang, 1995. “Cortical remodeling data are affected by
sampling location.” American Journal of Physical Anthropology. 106:219-227.

Robling, AG, 1998. “Histomorphometric assessment of mechanical loading history from
human Skeletal remains: the relation between micromorphology and
macromorphology at the femoral midshaft.” Dissertation. Columbia, MO,
University of Missouri.

Robling, AG and SD Stout, 2000. “Histomorphometry of human cortical bone:
applications to age estimation.” In: MA Katzenberg, SR Saunders (eds). ‘
Biological Anthropology of the Human Skeleton, New York, NY, Wiley-Liss,
Inc., pp. 187-213.

R0sing, FW, M Graw, B Marré, S Ritz-Timme, MA Rothschild, K Rotzscher, A
Schmeling, I Schroder, and G. Geserick, 2007. “Recommendations for the

forensic diagnosis of sex and age from skeletons.” Journal of Comparative
Human Biology, 58:75-89.

Rother, VP, G Kriiger, J Mechlitt, H Hunger, 1978. “Histomorphometrische sowie
regressions- und factor-analytische Untersuchungen von Alternsveréinderungen
des Humerus. Anatomischer Anzeiger, 144:346-365.

Samson, C, K, Branigan, 1987. “A new method of estimating age at death from
fragmentary and weathered bone” In: A Boddington, AN Garland, and RC
J anaway (eds.) Death, Decay zmd Reconstruction: Approaches to archaeology and
forensic science, Manchester, UK, Manchester University Press, pp. 101-108.

Schour, I, M Massler, 1941. “The development of the human dentition.” Journal of the
American Dental Association, 28: 1153-1160.

Scott, EC, 1979. “Dental wear scoring technique.” American Journal of Physical
Anthropology. 51:213—218.

106

Silberberg, M, R Silberberg, 1961. “Aging changes in cartilage and bone.” In: GH
Boume (ed.),
Structural Aspacts of Aging, New York, NY, Hafner.

Singh, 1, DL Gunberg, 1970. “Estimation of age at death in human males from
quantitative histology of bone fragments.” American Journal of Physig
Anthropology. 33:373-382.

Stewart, TD, 1979. Essentials of Forensic Anthropology: Especially as Developed in the
United States. Springfield, IL, Charles C Thomas.

Stoker, NG and BN Epker, 1971. “Age changes in endosteal bone remodeling and
balance in the rabbit.” Journal of Dent Research, 50, 1570-1574.

Stout, SD, 1986. “The use of bone histomorphometry in skeletal identification: the case
of Francisco Pizarro.” Journal of Forensic Sciences. 31:296-300.

Stout, SD, 1989. “Histomorphometric analysis of human skeletal remains.” In: K
Kennedy, and MY Iscan (eds). Reconstruction of Life from the Skeleton. New
York, NY Alan R. Liss, pp. 41-52.

Stout, SD, 1992. “Methods of determining age at death using bone microstructure.” In:
SR Saunders, MA Katzenburg (eds.) Skeletal Biology of Past Peoples: Research
Methods, New York, NY, Wiley-Liss, pp.21-36.

Stout, SD, WH Dietze, MY Iscan, and SR Loth, 1994.. “Estimation of age at death using
cortical histomorphometry of the stemal end of the fourth rib.” Journal of
Forensic Sciences, 39:778-784.

Stout, SD, SJ Gehlert, 1979. “Histomorphological identification of individuals among
mixed Skeletons.” Current Anthropology. 20:803-805.

Stout, SD, SJ Gehlert, 1980. “The relative accuracy and reliability of histological aging
methods.” Forensic Science International, 15:181-190.

Stout SD, and SJ Gehlert, 1982. “Effects of field size when using Kerley's histological
method for determination of age at death. “ American Journal of Physical
Anthropology. 58: 123-125.

Stout, SD, R Paine, 1992. “Brief communication: histological age estimation using rib
and clavicle.” American Journal of Physical Anthropology. 87:111-115.

Stout, SD, SL Simmons, 1979. “Histomorphometric determination of formation rates in
archaeological bone.” Calcified Tissue Research, 21:163-169.

107

Stout, SD, SC Stanley, 1991. “Percent osteonal bone versus osteon counts: the variable
of choice for estimating age at death” American Journal of Physical
Anthropology, 85:515-519.

Suchey, JM, D Katz, 1986. “Skeletal age standards derived from an extensive multiracial
sample of modern Americans.” American J oumal of Physical Anthropology.
69:269.

Tersigni, MA, 2005. “Serial long bone histology: inter- and intra-bone age estimation.”
Dissertation. Knoxville, TN, University of Tennessee.

Thompson, DD, 1979. “The core technique in the determination of age at death of
skeletons.” Journal of Forensic Sciences. 24:902-915.

Thompson, DD, 1981. “Microscopic determination of age at death in an autopsy series.”
Journal of Forensic Sciences, 26:470-475.

Thompson, DD, M Gunness-Hay, 1981. “Bone mineral-osteon analysis of Yupik-
Inupiaq Skeletons. American Journal of Physical Anthropology. 55:1-7.

Thompson, DD, CA Galvin, 1983. “Estimation of age at death by tibial osteon
remodeling in an autopsy series.” Forensic Science Intemationg, 22:203-211.

Todd, TW, 1920. “Age changes in the pubic bone: I,.the male white pubis.” American
Journal of Physical Anthropology. 3:285-334.

Ubelaker, DH, 1974. “Reconstruction of demographic profiles from ossuary skeletal
remains: A case study from the tidewater Patomic.” Smithsonian Contributions to
Antropology No. 18, Washington, DC, Smithsonian Institution Press.

Ubelaker, DH, 1981.”Approached to demographic problems in the northeast.”
Foundations of Northeast Archaeology, NY, Academic Press.

Ubelaker, DH, 1989. Human Skeletal Remains, second edition, Washington, DC,
Taraxacum Press.

Ubelaker, DH, 1998. “The evolving role of the microscope in forensic anthropology.”
In: K] Reichs (ed.), Forensic Osteology Advances in the Identification of Humfl
Remains, second edition, Springfield, IL, Charles C, Thomas.

Ubelaker, DH, JL Rife, 2006. “Approaches to commingling issues in archaeological
samples: A case study from Roman period tombs in Greece.” In: B. Adams and J.
Byrd, (eds). Recovery, Analysis and Identification of Commingled Human
Remains, Totowa, N .J .: Humana, forthcoming.

108

Walker, RA, 1982. “Assessment of human cortical bone dynamics and Skeletal age at
death from femoral cortical histology.” Thesis. Kent, Ohio, Kent State
University.

Weinberg, SM, NM Scott, K Neiswanger, and ML Marazita, 2005. “Intraobserver error
associated with measurements of the hand.” American Journal of Humfl
Biology, 17:368-371.

White, TD, 2000. Human Osteology, 2"d edition, San Diego, CA, Academic Press.

Yoshino, M, K Irnaizumi, S Miyasaka, and S Seta, 1994. “Histological estimation of age

at death using microradiographs of humeral compact bone.” Forensic Science
International, 64: 191-198.

109

 

 

I!

4881

1293 02956

3

20.6.1: .. . s