THREE-DIMENSIONAL GEOMETRIC MORPHOMETRIC SEX DETERMINATION AND MODELED FRAGMENTARY ANALYSIS OF THE HUMAN PUBIC BONE.

Chapter One: Introduction

Sexual dimorphism in the human pelvis is primarily based on the fact that females are

able to give birth, while males are not (1,2). This causes shape differences between the male and

female pelvis that forensic anthropologists are able to use to differentiate between the two sexes.

Traditionally, the analysis methods used to estimate a male or female origin of an unknown bone

are based on a visual analysis of specific parts of the pelvis. These visual analysis methods are

consistently 90-95% accurate when employed by an experienced forensic anthropologist (1,3).

Very few metric methods exist, and those that do are either based on these visual analysis

techniques or utilize interlandmark distances for analysis. The goal of the present research is to

refine a new method of three-dimensional geometric morphometric analysis in order to predict

an unknown bone as male or female, using only the pubic bone.

The pubic bone is a small, rectangular shaped portion of bone at the anterior aspect of the

pelvis. Everyone has a right and left pubic bone, meeting in the middle to form the pubic

symphysis. Visual analyses to estimate sex using the pubic bone include an analysis of the

ischiopubic ramus, the ventral arc, and the general shape of the bone (2). It is known that sexual

dimorphism exists in the pubic bone, so it follows that a metric analysis of these shape

differences should result in accurate sex predictions. Geometric morphometrics is the metric

analysis of shape and can be completed in a two- or three-dimensional technique (4). By utilizing

a three-dimensional geometric morphometric shape analysis, it should be possible to gain a much

more accurate and complete understanding of the shape differences between male and female

pubic bones. The overarching hypothesis of this research is that a geometric morphometric shape

analysis of the pubic bone will result in statistically accurate sex determinations on both whole

and fragmented human pubic bones and that certain landmarks will be more effective than others

in establishing a sex determination.

This research uses a sample from the University of New Mexico Maxwell Museum’s

Documented Skeletal Collection (5). This collection offers more than 300 nearly complete

skeletal individuals, all of which include known demographic information, such as sex, age-at-

death, ancestry affiliation, and cause of death. Many also include known information on parity

and occupation. All individuals in the collection have died within the last 50 years, making this

one of the most modern collections in the US. 213 adult individuals from this sample were

utilized for this research, which yielded 378 individual pubic bones (not all individuals had

complete right and left pubic bones). Each bone was placed into a rubber vice which held the

bone still and in place while data collection took place. A Microscribe digitizer was used to

collect eight landmark points from each pubic bone. After the initial data collection on all 378

bones was complete, a random group of 50 individuals (n=100) was re-digitized in order to test

the replicability of the method.

The statistical analyses used in this research included a Generalized Procrustes Analysis

in order to remove size and orientation from the data; a Principal Components Analysis to create

groupings based on similarities in variance within the data; and a series of Discriminant Function

Analyses to determine the predictive power of the sample to classify unknown bones as male or

female. In order to test the applicability of this method to more realistic fragmented pubic bones,

a modeled fragmentary analysis was performed by running Discriminant Functions Analyses on

all possible landmark combinations of three landmarks or more. This provided 218 possible

landmark combinations that represent fragmented bones missing landmarks. The results of these

analyses are presented and discussed in the following papers.

Also conducted as a smaller research project, was an overview of trophy skulls found in

Montana over the past few years. Trophy skulls and other retained body parts are generally

associated with times of war and summon imaginings of ancient civilizations, such as the Aztec

(6). While it is no longer common to keep the skull of one’s enemy, archaeological crania are

still found, kept, and often displayed (7). The question becomes, how does the forensic

anthropologist approach these crania, and where should their final resting place be? This research

presents six case reports of unknown individuals whose skull had been found and displayed in

some manner. It addresses the oddly high occurrence of trophy skulls in a state of relatively

small population (Montana), and analyzes the possible origination of each, determining that

modern trophy skulls may not be relics of war, but rather stumbled-upon keepsakes. It is also

argued that a change in terminology from trophy skull to souvenir skull be considered in order to

better reflect the background and provenience of these crania. Further analysis and discussion of

this research can be found in the third of the following articles.

References

1. White TD, Folkens PA. The Human Bone Manual. 1 edition. Burlington: MA: Academic

Press, 2005.

2. Kelley MA. Phenice’s Visual Sexing Technique for the Os Pubis: A Critique. American

Journal of Physical Anthropology 1978;48(1):121–2.

3. Bytheway JA, Ross AH. A Geometric Morphometric Approach to Sex Determination of the

Human Adult Os Coxa: A GEOMETRIC MORPHOMETRIC APPROACH TO SEX

DETERMINATION. Journal of Forensic Sciences 2010;55(4):859–64.

4. Zelditch ML, Swiderski DL, Sheets HD, Fink WL. Geometric Morphometrics for

Biologists. First. San Diego, California: Elsevier Academic Press, 2004.

5. UNM Maxwell Museum of Anthropology | Documented Collection. osteolab. 2018.

http://osteolab5.wixsite.com/osteolab/documented-collection (accessed July 9, 2018).

6. Andrushko VA, Latham KAS, Grady DL, Pastron AG, Walker PL. Bioarchaeological

evidence for trophy-taking in prehistoric central California. American Journal of Physical

Anthropology 2005;127(4):375–84.

7. Yucha JM, Pokines JT, Bartelink EJ. A Comparative Taphonomic Analysis of 24 Trophy

Skulls from Modern Forensic Cases. Journal of Forensic Sciences 2017;62(5):1266–78.

Chapter Two: Background Literature Review

The use of geometric morphometrics has become increasingly popular in forensic

anthropological research (1–3). Geometric morphometrics is, at its most basic level, the metric

analysis of shape (4). This method has been used in many biological studies for at least the past

three decades, but is relatively new in the field of Forensic Anthropology (4). Geometric

morphometrics differ from traditional methods because morphoscopic analysis does not

metrically assess bone, and interlandmark distance measurements do not fully capture the shape

of a bone. 3D geometric morphometric analysis is still a new technique in Forensic

Anthropology, and because of this, there are relatively few studies utilizing the method and there

is little standardization among the existing research.

Much of the existing geometric morphometric work focuses on sex and ancestry

assessment of the cranium (1,5,6). It has been shown, however, that metric analysis of the post-

cranial skeleton can be just as accurate or more accurate in sex estimation as the cranium and it is

accepted that the pelvis is the absolute best indicator of sex (7). Further research using 3D

techniques on the post-cranial skeleton is needed to help increase accuracy of sex estimation in

forensic and archaeologic contexts. Research focused on the pelvis most commonly investigates

changes in age, and few projects attempt to metrically determine sex (8,9). It is especially

important today to develop accurate metric analyses due to the need for reliable and objective

results in the court room (10). Improving the specificity of sex determination in the pelvis has

the potential to improve both modern forensic anthropology as well as analyses within

bioarchaeology.

The goal of this dissertation project will be to use a 3D Microscribe digitizer to assess sex

from the pubic bone, as well as to compile a list of standard landmarks that can be used in future

geometric morphometric studies of the pubic bone and pelvis. This project will also assess

patterns in morphological change related to age as an aid in determining whether or not the

differences between males and females change as individuals age. If available, ancestry will also

be recorded and tested to determine whether or not ancestry plays a significant role in shape

differences. The information gained from this research should be applicable to modern forensic

cases as well as other various scenarios, such as mass or commingled graves. The initial

hypothesis of this project, based on a review of the existing literature and experience is that this

3D geometric morphometric analysis will accurately determine the sex of unknown skeletons

based on the shape of the pubic bone.

The Human Pelvis

The primary function of the human pelvis is to transfer the weight of the upper body into

the legs during standing, walking, and running (11). The pelvis consists of two large bones,

known as the hip bones, or os coxae. These two bones connect anteriorly at the pubic

symphysis, and posteriorly with the sacrum at the sacroiliac joints (11). These bones form a very

stable structure, known as the pelvic ring, which allows close to no mobility, transfers weight

into the legs, and helps to protect and hold the abdominal organs (11). Each os coxae is actually

formed from the fusing of three different bones; the ilium, the ischium, and the pubis. These

three bones eventually fuse together at the acetabulum to form the large and complex structure of

the pelvis. The three bones of the os coxae grow throughout the juvenile years of each

individual, and sometimes do not fully fuse into one bone until the age of 25 (11). Such a long

growth period through childhood, puberty, and adulthood is part of what contributes to making

the pelvis the absolute best indicator of adult sex in the human skeleton; and speaks as to why

sub-adult sex estimation can be so difficult.

Although the pelvis has a unique morphology when compared to the rest of the human

skeleton, it develops much the same way as the other bones throughout the body, beginning with

an intra-uterine mesenchymal template which transforms into cartilage, and later ossifies to form

bone (11). The first indication of pelvic development in-utero occurs on day 28 when the lower

limb buds begin to develop (11). Following this, between days 34 and 36, the obturator, femoral,

and sciatic nerves rapidly extend into the developing limb bud, establishing their location before

any larger structures are formed (11). Chondrification, the process of transforming the

mesenchymal templates into cartilage, begins to occur between intra-uterine weeks six and seven

(11). Chondrification centers are well established in the ilium, ischium, and pubis by eight

weeks, and by the third intra-uterine month, the cartilaginous pelvis is well developed (11).

Ossification of the pelvis occurs through the process of endochondral ossification, during

which, chondrocyte cells begin the stimulation of mineralization and blood vessel growth, both

of which allow bone cells to be laid down within the cartilaginous template (11,12). Each of the

three bones in the os coxae begin ossifying at primary ossification centers which follow the

timing and location of the prior chondrification process (11). An ossified ilium is recognizable at

around four to five intra-uterine months, the ischium is recognizable after six months, and the

pubis begins ossification last, and is recognizable around the seventh month (11,13). All three

bones are easily identifiable by birth, and experience rapid growth during the first three months

of life, which then slows until approximately three years of age, after which, the growth slows

even more (11,13). During these first few years of postnatal life, multiple secondary ossification

centers form in each of the three bones (11). The secondary ossification centers help to expand

bone, much like the epiphyseal plates in long bones (11,13). They also help to form the

epiphyses of the ilium, the ischium, and the pubis, which will fuse later in life. A slow rate of

growth is maintained until puberty, during which a growth spurt causes rapid size and shape

changes. The ilium epiphyses are generally fully fused by about 20 years of age, the ischium is

generally completely fused by about 20-23 years of age, and the pubis is completely fused by 25

years of age (11,13). It should be noted however that shape changes continue to occur in the

pelvis over an individual’s lifespan. These changes can be due to age, giving birth, trauma,

disease and more.

Since the nineteenth century, osteologists have followed a simple rule, coined Wolff’s

Law after the man who developed the concept, that bone will adapt to the load placed upon it

(14). This adaptation affects the morphology of bone throughout the body and can be easily seen

in various contexts, one classic example being that of a professional weight lifter’s skeleton

exhibiting much larger muscle attachment sites on their bones than any non-weight lifter

counterpart. A lot of research has been done using this principle in an attempt to determine the

handedness of a skeletal individual due to the simple concept that one using their right hand

more often to complete physical tasks would develop larger muscles, and therefore larger

attachments throughout the bones of their right hand and arm (15–17). Unfortunately, no

concrete method was ever validated to determine the handedness of an unknown individual;

however a look into the symmetry and asymmetry of the human skeleton can offer some useful

insights (16).

Although asymmetry between the upper limbs cannot reliably predict handedness, it is

observed quite often (16). This observation occurs most frequently in archaeological samples

however, and does not seem to exhibit anything close to a pattern in modern populations, likely

due to the shift away from all-day physical labor in contemporary developed nations (16,18).

Asymmetry between the lower limbs is usually much smaller than asymmetry between the upper

limbs, and sometimes manifests as crossed symmetry (16). Crossed symmetry is when the bones

of the right arm are larger as well as the bones of the left leg, or vice versa (16). This seems to

occur due to the natural act of bracing oneself bilaterally, such as when a ball is thrown with the

right hand, and the individual steps forward with their left to compensate (16). However, crossed

symmetry is also not often seen in modern populations and likely does not affect the pelvis to

much degree. Locomotor requirements tend to produce much more symmetrical limb sets, which

explains why significant asymmetry between the lower limbs of humans is relatively rare (16). It

is unlikely that any significant asymmetry exists between the right and left pubic bones of

individuals seeing as the pubic bones are not directly affected by any common one-sided

movement of the body. Any asymmetry that does exist would likely be due to shape changes

related to age and may be easier to observe in older individuals.

This research focuses specifically on the adult pubic bone, which is one of the more

sexually dimorphic and age-revealing bones of the pelvis (12,19). The pubic bone is located on

the anterior portion of the os coxae and forms an articulation at the pubic symphysis. This joint

consists of a fibrocartilaginous joint which sits between the articular surfaces of the pubic bones

(20). This joint allows for approximately 2mm of movement in adults, and otherwise resists

tearing, rotating, shearing, and any further movement (20). The shape of the adult pubic bone

offers many clues as to the biological sex of the individual and can be used to quite accurately

assign a designation of male or female (12,21). The articular surfaces of the pubic bone

continually change shape and appearance as an individual ages (12,20,21). These changes can

be categorized into stages and thus used to estimate the age-at-death for an unknown individual

(12,21). The ability to accurately estimate sex and age from the pubic bone make it one of the

most important bones to recover when an unknown skeletonized individual is found, and

therefore accurate estimation methods are also necessary.

Traditional Sex Estimation

It is well known that the pelvis is the absolute best indicator of sex within the human

skeleton, although the cranium and other post-cranial elements can also be used (7,12,21). The

pelves of males and females are functionally different due to the basic fact that females give birth

(22). The pelvis of a female must be shorter and wider to offer space for childbirth while the

pelvis of a male tends to be taller and narrower (22). Other morphological traits have developed

between the two biological sexes as well due to the difference in functionality required in the

pelvis. These traits can only be used to determine the sex of an adult skeleton though, because

children do not develop these distinct traits until after puberty has occurred and their bones are

fully fused (12,21). Traditional sex estimation uses a visual analysis of these traits to determine

the sex of an unknown individual. In contrast, metric methods, such as geometric

morphometrics, use measurements and statistics to calculate a sex determination.

There are morphological indicators of sex on all three bones of each os coxa, and when

used in conjunction, a visual analysis performed by an experienced forensic anthropologist can

yield a 98% accuracy rate (7). The greater sciatic notch and the preauricular sulcus on the ilium;

the ischiopubic ramus ridge on the ischium; and the ventral arch, subpubic angle, and the body

shape on the pubic bone are some of the most reliable traits visible on the pelvis (12,21). These

traits are generally visually scored on a 1-5 scale, 1 being very female and 5 being very male, to

aid in the overall determination of sex (12,21). Each trait can be used to estimate sex alone, for

example, if the pelvis is fractured and only fragmentary pieces are recovered. It is possible for

female individuals to exhibit male traits however, and vice versa, meaning that a sex estimation

determined from just one trait may not be accurate (22). This is also true when utilizing traits of

the cranium to determine sex, and simply means that whenever possible, multiple traits should be

used in conjunction in order to determine the most accurate sex estimation possible.

There are surprisingly few metric methods to determine sex using the pelvis. A popular

method which uses both morphoscopic methods and metric analysis was published in 2008 (23).

This method requires the user to conduct a morphoscopic analysis of the subpubic concavity, the

ischiopubic ramus ridge, and the ventral arch. The user then enters these scores into a

spreadsheet which uses a logistic regression calculation to determine the overall probability of

the specimen being male or female (23). This method is popular because it offers a percentage

of likeliness, however it is still based on subjective morphoscopic analysis. A study in 2010 used

geometric morphometrics and a Microscribe digitizer to locate landmarks on the entire pelvis

(19). The research was conducted in much the same way as the current project, the largest

difference being that in 2010, the authors focused on the entire pelvis, rather than just the pubic

bone(19). This study resulted in a 100% correct classification rate for males and 98% for

females, an incredibly promising result for future metric analysis of the pelvis (19). This

geometric morphometric analysis is better than past morphoscopic analyses because it offers just

as high a rate of correct classification as an experienced anthropologist and it is metrically based,

which will inevitably stand up better in court if necessary (10,19). The main concern with this

particular method is that the entire pelvis is needed, which often is not available. To improve

upon this method, more specific research should be done using geometric morphometrics on

smaller portions of the pelvis, such as the pubic bone.

Age Estimation

The age of an unknown human skeleton can be estimated based on numerous parts of the

body, including the teeth, epiphyseal closure sites, the cranium, sternal rib ends, the pubic

symphysis and the auricular surface (21,22). Aging methods for sub adults are based on bone

growth and development, while aging methods for adults are based on bone degeneration (22).

Teeth are best used to determine the age of sub adults because human teeth erupt at very regular

intervals which can offer an accurate age estimation within a few months (24). Epiphyseal

closures are also best used for sub adults because the bones tend to develop at regular rates, and

are mostly to completely fused by ages 25-30 (25). The sutures on the cranium can also be used

to estimate age based on how open or closed they are, however, this method has been shown to

be inaccurate and is no longer widely used among forensic anthropologists unless it is the only

method available (26). The sternal rib ends, however, have proven to be quite useful in age

estimation. By analyzing the morphology and condition of the sternal end of the fourth rib, an

experienced forensic anthropologist can estimate a range for age at death of the individual (27).

The auricular surface and the pubic symphysis are the best age indicators in the body

(although the sternal rib ends have been shown to be just as accurate as the pubic symphysis)

(22,27). The auricular surface is the portion of ilium which articulates with the sacrum, together,

forming a tight joint which offers stability and little to no movement in the pelvis as a whole.

This surface has been shown to degenerate at a relatively regular rate, which can offer an age

range for unknown individuals (28). Similarly, the articulating surface of the pubic symphysis

can also be used to estimate the age of an unknown set of remains (29). Both methods are based

on descriptions and photographs of the bone surfaces at different ages. As with sex estimation,

the most accurate age estimations are determined when multiple analyses are used together.

Effects of Parturition

Forensic scientists have studied the effects of parturition on the human skeleton for many

years. Knowing the parturition status of an unidentified female skeleton could aid greatly in the

investigation to identify her, and is of great interest to forensic anthropologists (30,31). While it

has been shown that parturition and related events can affect the pelvis and leave changes and

marks, it is not the only event which can create these same marks and alterations (30,31). The

alterations found on the pelves of parous women can also be found on nulliparous women and

men (30). The amount of alterations or intensity of them also do not seem to correlate with

multiple pregnancies, and are sometimes entirely absent on parous and multiparous women (30).

This means that it is known that pregnancy can affect the shape and appearance of the pelvis, but

it cannot be concluded that pregnancy occurred when these alterations are viewed on the pelvis

of an unknown individual.

The main alterations which have been noted over time as possibly indicating a past

pregnancy include various alterations and manifestations of the preauricular sulcus, pubic pitting,

trabecular bone loss, and an extension of the pubic tubercle (30). Most of these alterations are

associated with changes in hormones and physical forces which are a product of the ligaments

and muscles stretching and relaxing around the pelvis in anticipation of childbirth (30). It has

also been shown that pregnant women can experience a decrease in bone density during their

pregnancy due to the transfer of calcium to the fetus and their breast milk (32–34).

Unfortunately, these alterations have also been found to be caused by general age changes,

urinary tract infections, minor and/or major traumas, surgery, joint and pelvic instability,

anomalies in the lumbosacral bones and joints, occupation-related activity, congenital anomalies,

degenerative conditions, as well as general pelvis and body size (30). The decrease in bone

density can also be caused by some of these other factors and is completely reversible in women

once the pregnancy has ended (32). Most of these other occurrences can happen to both men and

women, regardless of their parity history, making a forensic determination of pregnancy

impossible.

Considering how closely this research is looking at the shape of the pubic bone, and that

pregnancy can create alterations in the shape of the pelvis, it is important to work with a sample

that consists of females with available pregnancy records. A 2016 study showed that parturition

could affect the accuracy of age estimation when using the Suchey-Brooks pubic symphysis

method (35). The researchers showed that when their sample was separated into groups of men,

nulliparous women, and parous women, the parous women were inaccurately aged most often

(35). When they grouped all the females together however, there was no significant error in the

age estimations, meaning that the Suchey-Brooks method (created using a mixed parous and

multiparous sample) already accounts for the changes observed in some parous female samples

(35). This research shows, however, that parous females could present pubic bodies which are

quite different from nulliparous females, and which, could potentially bias this geometric

morphometric research. To control for this possibility, the method will be created using a

sample of both known parous and nulliparous females in an effort to account for these

alterations. It will also be noted though whether or not there is a difference in the results of

parous and nulliparous women when grouped separately in the statistics of this research.

Geometric Morphometrics

As stated earlier, geometric morphometrics is most simply a 3D analysis of shape (4).

The method has been used in biological fields for many years, although has only recently

become widespread in anthropology. Research using geometric morphometrics varies in many

ways, from focusing on age or sex estimation, to which elements of the skeleton are used. One

of the largest variances between studies, however, is the method of data collection. There are

two primary methods for gaining digitized data on shape variation. The first is to use a camera

to take photos of the element or bone surface, and to then use a computer to measure and record

landmark measurements and sliding semi-landmark measurements (2,36,37). While these

studies offer great potential and demonstrate the usefulness of 3D research, they are limited by

the fact that they use a 2D technique to analyze 3D shapes (36,37). A study on the use of sliding

semi-landmarks showed that when morphological variation is small, such as that between

modern humans, the differences between a landmark and a sliding semi-landmark study using

the same sample can be altered, thus raising questions about any results obtained by using sliding

semi-landmarks (36). The second method is to use a digitizer arm or 3D scanner, like a

computed tomography (CT) machine, to record 3D measurements in space. These methods tend

to be more precise, however they require more expensive tools as well as more experience to use

them effectively (1,38,39).

A study by San-Millan et al. used a camera to take images of the acetabulum of 327

males and 355 females of known sex and age from various collections in Europe to study shape

changes within the acetabulum as individuals age (37). Using the digitized images, the

researchers placed one landmark at the apex of the anterior acetabular horn of the lunate surface

and one on the apex of the posterior horn, as well as at 32 sliding semi-landmarks along the

outline of the lunate surface (37). The authors analyzed these points using various statistical

tests to determine shape changes related to sex and age of the individuals (37). The results found

significant shape differences in both sex and age between individuals; however, as females aged,

their acetabular shape began to look much more similar to males and it became harder to

distinguish between sexes of older ages (37). This age confluence could introduce errors when

attempting to assess the sex of individuals based on joint morphology. For this reason, this

dissertation project will also look for patterns in morphology changes related to age in the body

of the pubic bone. The authors of this study also point out that digitizing a 3D shape (the

acetabulum) using a 2D technology (camera photography) is limited and that further research

using 3D methods should be employed to expand on their results concerning acetabular

morphology between the sexes (37).

Within the second main area of data collection, using 3D digitizers or CT machines, there

are also various methods to collect data. The first method is collecting 3D points to measure

traditional interlandmark distances, such as maximum length of a bone (1,2). The second

method consists of recording multiple fixed landmarks to conduct a shape analysis of the element

in question (38,40,41). This method would record landmarks such as the medial intercondylar

tubercle and the maximum anterior point on the medial condyle of the proximal tibia (41). The

third method uses a mixture of fixed landmarks, similar to the second method, while adding

points evenly spaced across the bone surface (3). For example, Franklin et al. used fixed

landmarks on mandibles as well as three evenly-spaced points along the posterior ramus and

another three evenly-spaced points along the mandibular body in an attempt to gain a better

representation of the shape of the mandible (3). A project which needs traditional distance

measurements should use the first method, while projects researching the shape morphology of

skeletal elements should use either the second or third method. Currently, the research does not

show many advantages or disadvantages to adding evenly-spaced points to existing landmarks.

One study on intraobserver error in geometric morphometric methods, however, did show that

the largest amount of error comes from locating landmarks (39). Due to this error, using evenly-

spaced points might be more easily replicated, but future research is needed to substantiate this.

A study by Katherine Spradley and Richard Jantz focused on estimating ancestry from

the cranium used interlandmark distances measured by digitizing landmarks using a Microscribe

digitizer on craniums in the Forensic Anthropology Data Bank (FDB) collection housed at the

University of Tennessee (1). The authors used the digitizer to capture 31 landmarks from each

skull in their study, and were then able to calculate all possible interlandmark distances between

the 31 points, which resulted in 465 interlandmark distances per skull (1). Using these

landmarks, the authors also tested geometric morphometric shape analysis to compare to the use

of interlandmark distance measurements. The results of their study showed that the

interlandmark measurements considered nonstandard (or measurements that have not been

described before in research protocol books like Standards (42)), were more accurate in ancestry

estimation than were the geometric morphometric analyses (1). While geometric morphometrics

was not the most accurate method to determine ancestry in this study, the use of a Microscribe

digitizer proved accurate and useful in data collection (1).

The authors determined that the shape information provided by geometric morphometric

analysis is still important and that further research should be conducted in this area (1). The

authors also pointed out that using the digitizer prompts the observer to focus on one landmark at

a time, which reduces error, and then the landmark is automatically recorded through a computer

program, like Excel, which reduces recording error and immediately saves data (1). An earlier

project using geometric morphometric analysis of the cranium set out to develop a database

similar to FORDISC which could categorize unknown skulls’ ancestry and sex (43). The authors

collected 75 craniofacial landmarks on over 1,000 skulls of European, African, and Hispanic

ancestry using a Microscribe digitizer (43). While this study did not use interlandmark distances,

it did find significant shape differences between ancestries and the authors’ called for future

research in geometric morphometrics concerning both ancestry and sex determination (43). The

database created from this research is an ongoing project as more research contributes to the field

of geometric morphometrics. Although this dissertation will not be focusing on cranial

measurements, these studies provide support for continued research in the field of geometric

morphometrics and emphasizes the advantages of using a Microscribe digitizer (1,43).

Most studies using a digitizer focus on analyzing the shape of a bone or bone surface,

rather than measuring interlandmark distances. One of the earliest studies using a Microscribe

digitizer was a dissertation written by Frederick John Snow at the University of Tennessee in

2004. Snow’s goal was to assess sex from the scapulae of a sample that consisted of 241

individuals from the William M. Bass Donated Skeletal Collection, all of either White or Black

ancestry (44). The author collected ten landmarks on both right and left scapulae (when

possible) of each individual (44). The main analysis of this study showed that the mean centroid

size could significantly discriminate between males and females, but could not discriminate

between White and Black ancestries (44). The author observed that the canonical discriminant

function and principal component analysis were also able to significantly distinguish males from

females, and were able to observe shape differences between distinct ancestral backgrounds (44).

Snow concluded that this was the first study to use geometric morphometrics to exhibit sexual

dimorphism between male and female scapulae, and reported that his analysis indicates that

shape differences in ancestry are present (44). This early research clearly displayed the potential

of digitized geometric morphometrics and indicated the need for further study of the scapula.

A thesis project from the Boston University School of Medicine set out to test sex

assessment through metric, non-metric, and geometric morphometric analysis of the distal

humerus (38). This study used a Microscribe digitizer to collect 28 points on the distal ends of

227 humeri from the William M. Bass Donated Skeletal Collection at the University of

Tennessee (38). The analysis of this study was particularly interesting because the author ran

statistical analyses on all 28 landmarks together, only the posterior landmarks, and only the

anterior landmarks (38). This type of analysis could help in future work when presented with

fragmentary remains. The results of these analyses were not much different from each other;

males and females were classified correctly anywhere from 50-70% of the time, which are

relatively low classification rates (38). Overall, the author determined that the geometric

morphometric analysis was not more accurate than the metric and non-metric methods; although

it did show some shape differences and has potential for future research (38). The author stated

that her inexperience in using geometric morphometric techniques most likely contributed to the

overall low results of the geometric morphometric analysis and declared that more research was

needed to determine which landmarks are most useful in shape analysis of the distal humerus

(38). This is a significant preliminary study in the use of geometric morphometrics and

emphasizes the need for further research and experience in the data collection process.

A study by Bytheway & Ross in 2010 looked to determine how well the pelvis, when

digitized, could determine sex (19). The authors used 200 left os coxae from the Terry

Collection housed at the Smithsonian, and collected 36 landmark points per os coxa (19). The

authors found that the landmarks they wished to collect did not fit into the traditional

classifications of type I, II, or III (an explanation of these landmark types can be found in the

Materials and Methods section); so they revised classifications and descriptions to create

categories such as “constructed” or “traditional fuzzy” landmarks (19). The “fuzzy” landmarks

are considered to be areas of the bone, rather than one point, and are found and collected by

following the author’s description of the point (19). Overall, the authors found that by using

geometric morphometric analyses on the pelvis, they could determine sex 98-100% of the time

(19). This is an incredibly high rate of accuracy and is consistent or better than traditional

methods (generally 90-95% accuracy (45)) to determine sex using the pelvis. The authors

concluded that the pelvis is, as expected, a highly accurate element to use to determine sex; they

also determined that geometric morphometrics is a reliable and accurate method to apply (19).

The geometric morphometric analysis was able to display the direction of shape change between

males and females in different areas of the bone, information that is not available via traditional

analytical methods, and has great implications for identification in fragmentary contexts (19).

The authors also point out that the greatest amount of shape variation was found in the pubic and

ischium bones, as well as the connecting areas of the illium (19). This study did not perform an

evaluation of intraobserver error on data collection, however, so it is unknown how the “fuzzy”

landmark points would perform in further use of this method and the authors state their intention

of further research concerning this subject (19).

Most other studies using the pelvis to determine sex focus on morphoscopic and visual

index methods. One of the earliest metric analyses of the pubic bone was a thesis project by

Tiffany Burch in 2003. Burch measured the pubic width on the dorsal side of the pubic bone,

measuring the width between the medial edge of the pubic symphysis to the margin of the

obturator foramen (46). The author’s samples consisted of 110 males and 50 females which had

been collected at autopsy from the Department of Coroner, Los Angeles (46). This collection

was originally curated by Dr. J. Suchey and used to develop the widely used Suchey-Brooks

method to age skeletal individuals based on the pubic symphysis. Using samples collected from

autopsy offers the advantage of using a completely modern sample of known age, sex, and

ancestry. The author only used individuals 17-30 years in age, and later further narrowed this

field to 21-30 years old because younger females tended to classify incorrectly (46). The author

averaged the width of both right and left pubic bones for each individual, and then found

sectioning points to distinguish between male and female (46). The highest correct percentage of

classified males and females, aged 21-30 was 91.4%(46).

This is a high classification rate and is at least on par with traditional morphoscopic

methods. While this project did not complete a 3D analysis, a shape analysis, or use more than

one measurement or landmark, it is one of the only metric analyses of the pubic bone in an

attempt to determine sex. Burch is able to successfully show that the pubic bone is metrically

sexually dimorphic and that its size and shape should be further studied (46). The author also

utilized a sample collected from autopsy, although she utilized this sample almost 30 years after

it was originally collected (46). This gap in time may influence the results applicability because

the sample is not completely contemporary, although this would need to be tested. To improve

on this method, this dissertation research will utilize a contemporary sample taken from autopsy

to improve on the forensic applicability of the results.

The third method of data collection discussed above consists of using landmark points in

conjunction with points that are placed after measuring evenly-spaced distances across a bone.

The study by Franklin et al. set out to use geometric morphometrics to determine sex of

subadults based on the mandible (3). The study consisted of 96 subadult mandibles of known

age, sex and population, pulled from the Hamann-Todd Osteological Collection, the Raymond A.

Dart Collection of Human Skeletons, and the Natural History Museum of London (3). The

authors collected 38 bilateral points with a Microscribe digitizer (3). Most of the landmark

points collected were traditional, however the authors did add evenly-spaced points along the

posterior ramus and the mandibular body in an effort to more accurately capture the shape of the

entire mandible (3). While the authors were not able to significantly determine differences

between males and females, they did observe some shape differences which are promising for

future research to expand on. They also reported finding more significant differences between

populations, which indicate that subadult mandibles might be able to be classified by ancestry,

which could be quite useful in forensic contexts (3). This study also emphasizes the usefulness

of a digitizer and the resulting shape analysis it can provide (3).

Another method to utilize landmark data is to study landmark measurements through 3D

Computed Tomography (CT) scans. One project used 100 randomly selected CT scans of in situ

pelves to test the reliability of landmark distance measurements and the traditional Phenice traits

based on CT scans (47). The advantages of using modern CT scans are that the sample consists

of modern, living people, the researcher does not need access to the actual bone, and a computer

program is used to identify and complete all measurements meaning there is less subjective error.

The disadvantages of using CT scans usually involve obtaining access to the scans themselves.

It can be very difficult to obtain access to current medical records of living people due to the

many privacy laws currently in use. A researcher must also have access to the computer

software necessary to manipulate CT scans and gain accurate results. The authors of this study

however were able to reach results of 100% accuracy in male and female determination (47).

This is astonishingly high and further supports the fact that using the pelvis is the best way to

determine sex and supports a 3D metric analysis of the pelvis. While not all future researchers

will have access to modern medical records, this research clearly indicates need for further

research into 3D shape analysis, specifically of the pelvis because of the incredibly accurate

results achieved (47).

Many research projects focused on the pubic bone study how to improve age estimations

of skeletal remains (8,9). The pubic symphysis is one of the best age indicators on the human

skeleton, so understandably, most research focuses here. An example of this which uses a

somewhat different method of 3D shape analysis is a thesis by Gray (2011). Here, the author

uses 3D laser scanning which is similar to using a CT scan to perform a shape analysis, however

different measurements are obtained from a laser scan (48). The author attempted to determine

age from the symphyseal face of the pubic symphysis on a sample of 40 male pubic bones (48).

Gray used 3D laser scans of the symphyseal face to quantify and measure seven morphoscopic

traits from the Suchey-Brooks method of aging the pubic symphysis (48). The results showed

that rim completeness, billowing height and area, and depth of symphyseal face depression were

most closely correlated with correct age (48). While projects like this are continually improving

the precision of aging human skeletons based on a small portion of the pelvis, there is a gap

created which ignores how to determine sex from the same small portion of bone. Continued 3D

studies show that this is an accurate and reliable field to further study, and there is a clear need to

improve aspects of sex determination within this field.

Conclusion

The human pelvis is a widely studied element within forensic anthropology and has been

used as the best sex indicator for many decades (7,21,22). It is also known that the pelvis can

predict age as well as sex, although it is always important to use other methods in tandem when

possible (29,49). Although the current literature can conclude no true correlation between

pregnancy and alterations of the pelvis, it is known that some pregnancies can cause changes,

and future research may be able to fully establish what a pelvis could reveal about parturition

(30,35). Most traditional methods used on the pelvis to conduct these analyses consist of visual

observation and analysis however, and it has become more necessary over time to create

potentially more accurate metric methods to estimate this information for unidentified skeletal

remains (10). Due to this need, geometric morphometrics has become increasingly more popular

in the field of forensic anthropology, but there is still a gap visible in the existing metric studies

of the pelvis.

It is clear based on these previous studies that more research in the field of geometric

morphometrics is necessary. Researchers are still exploring the best methods to use within this

field and many of these studies indicate that geometric morphometrics has great potential (50). It

is also clear, however, that most of the larger projects have focused on sex and ancestry based on

the cranium (1,43). The projects which do focus on the pubic bone mainly investigate changes in

morphology due to age (48,51,52). This means that more research on the postcranial skeleton is

definitely needed, and that more research and literature on shape differences in the pelves of

males and females is needed (7,50). This dissertation will draw from aspects of past research,

including data collection methods, certain landmarks, and statistical analyses in an effort to

improve upon existing research, as well as to improve standardization throughout geometric

morphometric studies.

References Cited

1. Spradley K, Jantz RL. Ancestry Estimation in Forensic Anthropology: Geometric

Morphometric versus Standard and Nonstandard Interlandmark Distances. J Forensic Sci

2016;61(4):892–7. https://doi.org/10.1111/1556-4029.13081.

2. Lynch JJ, Cross P, Heaton V. Sexual Dimorphism of the First Rib: A Comparative

Approach Using Metric and Geometric Morphometric Analyses. J Forensic Sci

2017;62(5):1251–8. https://doi.org/10.1111/1556-4029.13421.

3. Franklin D, Oxnard CE, O’Higgins P, Dadour I. Sexual Dimorphism in the Subadult

Mandible: Quantification Using Geometric Morphometrics. J Forensic Sci 2007;52(1):6–10.

https://doi.org/10.1111/j.1556-4029.2006.00311.x.

4. Zelditch ML, Swiderski DL, Sheets HD, Fink WL. Geometric Morphometrics for

Biologists. First. San Diego, California: Elsevier Academic Press, 2004.

5. Abdel Fatah EE, Shirley NR, Jantz RL, Mahfouz MR. Improving Sex Estimation from

Crania Using a Novel Three-dimensional Quantitative Method , ,. J Forensic Sci

2014;59(3):590–600. https://doi.org/10.1111/1556-4029.12379.

6. Jung H, Woo EJ. Evaluation of Mastoid Process as Sex Indicator in Modern White

Americans using Geometric Morphometrics. J Forensic Sci 2016;61(4):1029–33.

https://doi.org/10.1111/1556-4029.13079.

7. Spradley K, Jantz RL. Sex Estimation in Forensic Anthropology: Skull Versus Postcranial

Elements: SEX ESTIMATION IN FORENSIC ANTHROPOLOGY. J Forensic Sci

2011;56(2):289–96. https://doi.org/10.1111/j.1556-4029.2010.01635.x.

8. Berg GE. Pubic Bone Age Estimation in Adult Women*. J Forensic Sci 2008;53(3):569–77.

https://doi.org/10.1111/j.1556-4029.2008.00712.x.

9. Wink AE. Pubic Symphyseal Age Estimation from Three-Dimensional Reconstructions of

Pelvic CT Scans of Live Individuals. J Forensic Sci 2014;59(3):696–702.

https://doi.org/10.1111/1556-4029.12369.

10. Lesciotto KM. The Impact of Daubert on the Admissibility of Forensic Anthropology

Expert Testimony. J Forensic Sci 2015;60(3):549–55. https://doi.org/10.1111/1556-

4029.12740.

11. Verbruggen SW, Nowlan NC. Ontogeny of the Human Pelvis. Anat Rec 300(4):643–52.

https://doi.org/10.1002/ar.23541.

12. White TD, Folkens PA. The Human Bone Manual. 1 edition. Burlington: MA: Academic

Press, 2005.

13. Schaefer M, Black S, Louise S. Juvenile Osteology: A Laboratory and Field Manual. San

Diego, California: Elsevier Academic Press, 2009.

14. Frost HM. A 2003 Update of Bone Physiology and Wolff’s Law for Clinicians. Angle

Orthod 2004;74(1):3–15. https://doi.org/10.1043/0003-

3219(2004)074<0003:AUOBPA>2.0.CO;2.

15. Wilczak CA. Consideration of sexual dimorphism, age, and asymmetry in quantitative

measurements of muscle insertion sites. Int J Osteoarchaeol 1998;8(5):311–25.

https://doi.org/10.1002/(SICI)1099-1212(1998090)8:5<311::AID-OA443>3.0.CO;2-E.

16. Auerbach BM, Ruff CB. Limb bone bilateral asymmetry: variability and commonality

among modern humans. J Hum Evol 2006;50(2):203–18.

https://doi.org/10.1016/j.jhevol.2005.09.004.

17. Ruff CB, Jones HH. Bilateral asymmetry in cortical bone of the humerus and tibia—sex and

age factors. Hum Biol 1981;53(1):69–86.

18. Danforth ME, Thompson A. An Evaluation of Determination of Handedness Using Standard

Osteological Measurements*. J Forensic Sci 2008;53(4):777–81.

https://doi.org/10.1111/j.1556-4029.2008.00741.x.

19. Bytheway JA, Ross AH. A Geometric Morphometric Approach to Sex Determination of the

Human Adult Os Coxa: A GEOMETRIC MORPHOMETRIC APPROACH TO SEX

DETERMINATION. J Forensic Sci 2010;55(4):859–64. https://doi.org/10.1111/j.1556-

4029.2010.01374.x.

20. Becker I, Woodley SJ, Stringer MD. The adult human pubic symphysis: a systematic

review. J Anat 2010;217(5):475–87. https://doi.org/10.1111/j.1469-7580.2010.01300.x.

21. Bass WM. Human Osteology: A Laboratory and Field Manual. 5th ed. edition. Columbia,

Mo: Missouri Archaeological Society, 2005.

22. White T, Black M, Folkens P. Human Osteology. 3rd ed. Burlington: MA: Elsevier

Academic Press, 2012.

23. Walker PL. Sexing skulls using discriminant function analysis of visually assessed traits.

Am J Phys Anthropol 2008;136(1):39–50. https://doi.org/10.1002/ajpa.20776.

24. Ubelaker D. Human Skeletal Remains: Excavation, Analysis Interpretation. 6th ed. Chicago,

IL: Aldine Publisher, 1978.

25. McKern T, Stewart TD. Skeletal Age Changes in Young American Males. Natick, MA:

Quarter Master Research and Development Command, 1957.

26. Garvin HM, Passalacqua NV. Current Practices by Forensic Anthropologists in Adult

Skeletal Age Estimation*: AGE ESTIMATION PRACTICES. J Forensic Sci

2012;57(2):427–33. https://doi.org/10.1111/j.1556-4029.2011.01979.x.

27. Hartnett KM. Analysis of Age-at-Death Estimation Using Data from a New, Modern

Autopsy Sample-Part II: Sternal End of the Fourth Rib*,†: AGE-AT-DEATH

ESTIMATION USING THE FOURTH RIB. J Forensic Sci 2010;55(5):1152–6.

https://doi.org/10.1111/j.1556-4029.2010.01415.x.

28. Lovejoy CO, Meindl RS, Pryzbeck TR, Mensforth RP. Chronological metamorphosis of the

auricular surface of the ilium: a new method for the determination of adult skeletal age at

death. Am J Phys Anthropol 1985;68(1):15–28.

29. Hartnett KM. Analysis of Age-at-Death Estimation Using Data from a New, Modern

Autopsy Sample-Part I: Pubic Bone*,†: J Forensic Sci 2010;55(5):1145–51.

30. Ubelaker DH, Paz JSDL. Skeletal Indicators of Pregnancy and Parturition: A Historical

Review. J Forensic Sci 57(4):866–72. https://doi.org/10.1111/j.1556-4029.2012.02102.x.

31. McFadden C, Oxenham MF. Sex, Parity, and Scars: A Meta-analytic Review. J Forensic Sci

2018;63(1):201–6. https://doi.org/10.1111/1556-4029.13478.

32. Reid IR. The skeleton in pregnancy and lactation. Intern Med J 32(9–10):433–4.

https://doi.org/10.1046/j.1445-5994.2002.00281.x.

33. Kovacs CS, Kronenberg HM. Pregnancy and Lactation. Primer on the Metabolic Bone

Diseases and Disorders of Mineral Metabolism. Wiley-Blackwell, 2013;156–64.

34. Kirby BJ, Ma Y, Martin HM, Favaro KLB, Karaplis AC, Kovacs CS. Upregulation of

calcitriol during pregnancy and skeletal recovery after lactation do not require parathyroid

hormone. J Bone Miner Res 28(9):1987–2000. https://doi.org/10.1002/jbmr.1925.

35. Bongiovanni R. Effects of Parturition on Pelvic Age Indicators. J Forensic Sci

2016;61(4):1034–40. https://doi.org/10.1111/1556-4029.13085.

36. Perez SI, Bernal V, Gonzalez PN. Differences between sliding semi-landmark methods in

geometric morphometrics, with an application to human craniofacial and dental variation. J

Anat 2006;208(6):769–784.

37. San-Millán M, Rissech C, Turbón D. Shape variability of the adult human acetabulum and

acetabular fossa related to sex and age by geometric morphometrics. Implications for adult

age estimation. Forensic Sci Int 2017;272:50–63.

https://doi.org/10.1016/j.forsciint.2017.01.005.

38. Berthelot CM. Metric, nonmetric, and geometric morphometric methods of sex estimation

using the distal humerus. 2014.

39. Menéndez LP. Comparing Methods to Assess Intraobserver Measurement Error of 3D

Craniofacial Landmarks Using Geometric Morphometrics Through a Digitizer Arm. J

Forensic Sci 2016. https://doi.org/10.1111/1556-4029.13301.

40. Brzobohatá H, Krajíček V, Horák Z, Velemínská J. Sex Classification Using the Three-

Dimensional Tibia Form or Shape Including Population Specificity Approach. J Forensic Sci

2015;60(1):29–40. https://doi.org/10.1111/1556-4029.12641.

41. Toon C. Sexual dimorphism at the proximal tibia: a geometric morphometric analysis. 2014.

42. Buikstra JE, Ubelaker DH, editors. Standards: for Data Collection from Human Skeletal

Remains. Arkansas Archeological Survey, 1994.

43. Ross AH, Slice DE, Williams SE. Geometric Morphometric Tools for the Classification of

Human Skulls. US Dep Justice 2010.

44. Snow FJ. Geometric morphometry analysis of the scapula: Implications for the

determination of sex and ancestry. 2004.

45. Klales AR, Ousley SD, Vollner JM. A revised method of sexing the human innominate

using Phenice’s nonmetric traits and statistical methods. Am J Phys Anthropol

2012;149(1):104–14. https://doi.org/10.1002/ajpa.22102.

46. Burch TC. Metric sex determination using minimum width of the pubic bone. 2003.

47. Decker SJ, Davy-Jow SL, Ford JM, Hilbelink DR. Virtual Determination of Sex: Metric and

Nonmetric Traits of the Adult Pelvis from 3D Computed Tomography Models*,†. J Forensic

Sci 2011;56(5):1107–14. https://doi.org/10.1111/j.1556-4029.2011.01803.x.

48. Gray A. Applicability of Three Dimensional Surface Scanning to Age-at-Death Estimations

based on the Human Pubic Symphysis. 2011.

49. Sutherland LD, Suchey JM. Use of the Ventral Arc in Pubic Sex Determination. J Forensic

Sci 1991;36(2):501–11.

50. Gonzalez PN, Bernal V, Perez SI. Geometric morphometric approach to sex estimation of

human pelvis. Forensic Sci Int 2009;189(1–3):68–74.

https://doi.org/10.1016/j.forsciint.2009.04.012.

51. Slice DE, Algee-Hewitt BFB. Modeling Bone Surface Morphology: A Fully Quantitative

Method for Age-at-Death Estimation Using the Pubic Symphysis. J Forensic Sci

2015;60(4):835–43. https://doi.org/10.1111/1556-4029.12778.

52. Stoyanova D. A computational method for age-at-death estimation based on the pubic

symphysis. 2012.

Chapter Three: Materials and Methods

The methods of this research derive from those within the field of geometric

morphometrics. At its most basic definition, geometric morphometrics is the analysis of shape

(1). The basic process of this consists of recording multiple homologous shapes (multiple pubic

bones), rotating them all onto the same plane and removing size as a variable (Generalized

Procrustes Analysis), and finally distinguishing differences between group clusters (Principal

Components Analysis, etc.) (1). This research will use a Microscribe Digitizer to collect 3D

points on multiple collections of pubic bones. This data will then be analyzed using the

computer software MorphoJ to identify differences in shape between male and female

specimens.

Hypothesis: by using 3D geometric morphometric analysis, differences in the shape of

female and male pubic bones can be identified and used to classify unknown individuals as either

male or female.

A pilot project was conducted on a small test sample of pubic bones and preliminary

statistics were run on the data gathered. The initial results are promising for this research overall

and indicate that the method and hypothesis will be successful when tested on larger sample

sizes.

It is well known within anthropology that when developing methods which are to be

applied to modern humans, the samples used should also consist of modern humans. In order for

this research to be forensically relevant, it must be based on a modern skeletal collection (2,3).

There are a series of modern human skeletal collections in the United States, the most popular

being the William M. Bass Donated Skeletal Collection at the University of Tennessee, the

Hamann-Todd Osteological Collection at the Cleveland Museum of Natural History, and the

Terry Collection at the Smithsonian. These collections consist of modern humans of known sex,

ancestry, age-at-death, and sometimes profession. Most studies within the existing literature use

samples from one or more of these collections (4–9). This research will use the University of

New Mexico Maxwell Museum Documented Collection, as well as the University of Montana

Skeletal Pubic Symphysis Collection, both of which consist of modern, documented individuals.

Materials

The Samples: The University of New Mexico Maxwell Documented Skeletal Collection

The primary sample will consist of individuals from the University of New Mexico

Maxwell Museum’s Documented Skeletal Collection. This collection was established in 1984

and now has over 300 individuals of both sexes, varying ages, and many population groups (10).

The known information for most individuals within this collection include sex, age-at-death,

population affinity, and cause of death (10). All donations after 1995 have also been asked to

provide past health information as well as occupation information so that more research using the

collection can be conducted (10). This is one of the best modern collections in the western

United States and is particularly important to forensic anthropology because it consists entirely

of individuals who have passed away within the last 50 years (10). The sample as a whole is

made up of 60% males, most of the males and females are aged 51-75, and 80% of the total

sample identified as White (10).

The Maxwell Documented Skeletal Collection has been chosen as the primary sample for

this research for multiple reasons. First, it is one of the most modern skeletal collections in the

United States. It is important to develop new methods using a contemporary collection so that

the method(s) is more likely to be accurate when used on modern forensic cases (11). This helps

to avoid biases that may be introduced by secular change in a population. Along with this, it is

also important to develop new methods using a collection that derives from the ancestry

affiliation of which it will be used on in the future (11). This is not always possible, especially

when modern admixture and self-identification are taken into account, but it is always important

to record ancestry when available and to notice any trends or correlations that may arise. The

Maxwell Documented Collection is identified as 80% White, which will correlate well with any

sample recovered in Montana.

“White” can mean many things, and a White sample from New Mexico will likely be

quite different from a White sample in Montana, but it is at least a similar overall characteristic

that can be used to combine and contrast the two samples. It is important to remember that the

ancestry classifications are self-reported, which means they represent which group of people

each individual identified with, and not necessarily where their genetic ancestors came from. In

the U.S., as with many other places, “White” has become an umbrella term for people with light

skin, who usually reside in a middle to high socioeconomic class, and who might recognize

having ancestors from Europe. In reality, white is a color and not an ancestry, meaning that the

genetic ancestry of people who identify as “White” could vary drastically. It has also been

shown that those who are multiracial, or identify with multiple ancestries or ethnicities, often

change their identification over time, and usually choose to list a single ancestry rather than

multiple ones (12). Self-reported information on ancestry may not be the most reliable, however

it does create a group basing that can still offer important information when compared and

contrasted.

Lastly, the Maxwell Collection offers a good sample for this research because most of the

individuals within the collection are 51-75 years old (10). Only individuals 21 and older will be

considered for this research to avoid morphologically indeterminate juveniles. Determining the

sex of a subadult human skeleton is nearly impossible, and it is generally considered bad practice

to do so within forensic anthropology. An overall older sample will provide more individuals

and hence more data for the current research.

The Samples: The University of Montana Skeletal Pubic Symphysis Collection

The second sample for this research will be curated throughout the duration of the

project. The Medical Examiners at the State Crime Lab in Missoula, Montana have agreed to

procure the pubic bodies from recently deceased individuals who receive autopsies. Families of

the deceased are asked to fill out a consent form to allow for the donation of the pubic bone

portions to the University of Montana. Through the use of this form, the family has the chance

to deny or allow donation, offer demographic information about the deceased, and offer access to

the deceased’s medical records. The collection process is simple; during autopsy, the medical

examiner makes four cuts using the bone saw, one on each ramus of the right and left pubic

bone, to easily remove the two pubic bodies around the pubic symphysis. The bones are then

soaked in hot water, bleach, and soap, and left to dry. Once dry, the procured samples are

transported to the University of Montana Anthropology Department, where they are curated into

the existing skeletal collection. Samples will be collected from all possible individuals aged 21

and older. Younger individuals will not be included in this sample to avoid the ambiguities

between male and female juvenile skeletons.

Samples will ideally be collected for two years, after which, sample collection will stop

so that analyses can proceed. The author hopes to collect data points from the samples

throughout this time frame, so that analyses can occur immediately after the collection process

ends. The overall size of this sample depends on the availability of remains, which in turn

depends on the autopsy rate in Missoula, as well as the willingness of families to donate tissue of

the deceased. In a perfect world, after two years, there would be at minimum 200 individuals

included in this sample; whether or not that number will be reached remains to be seen.

Whatever the size of the sample at the end of this project, the individuals curated will still

contribute to this research by determining whether or not they fit within the new method based

on the Maxwell Museum sample. The author will also consider this collection process as an

exercise in learning how to curate a new collection and how to work with multiple agencies in

the process of sample procurement. It is the hope of the author that regardless of the size of the

collection at the end of this project, the curation can continue through other individuals at the

University of Montana in order to continue building an irreplaceable collection which will result

in further research.

Equipment

The most important piece of equipment needed for this research is the Microscribe

Digitizer. This consists of a stylus at the end of a moveable arm, which is attached to a heavy

base which keeps the digitizer in place. The digitizer also comes equipped with a foot pedal to

ease the recording of data, and all necessary cords to connect to the research computer. The use

of the digitizer is simple, one must simply place the point of the stylus at the point which is to be

recorded, and simultaneously click the foot pedal to record the data point. This data point is

automatically entered into an Excel Spreadsheet. Once data collection is finished, the data can

be imported into other software, such as Past or MorphoJ. This software is necessary to perform

further statistical tests, such as a Generalized Procrustes Analysis, a Principal Components

Analysis, and a Canonical Variates Analysis.

Other necessary equipment includes a vice to hold each individual sample in place while

data recording occurs. In order to achieve the necessary stillness, a small, rubber clamp was set

into a larger, steel vice. The steel vice holds the clamp quite still, while still allowing the clamp

to be easily opened and closed to hold each bone. A crockpot, bleach, soap, and water are also

needed to clean each bone sample once it has been removed during autopsy. These materials are

readily available in the Medical Examiner’s office, and once the samples are clean and dry, they

are placed into individual bags and boxes to be transported to the University of Montana. Once

at the university, each individual is assigned a number to ensure anonymity during data

collection and analysis.

Methods

Background: Geometric Morphometrics

Geometric morphometrics has been used in a variety of fields for some time now and is a

relatively new method within anthropology. Biology, on the other hand, has utilized geometric

morphometrics for many years and quite a few interesting studies have been published using

variations of the methods behind geometric morphometrics. As in anthropology, biological

studies have also used photography, outlines and semi-sliding landmarks, as well as Microscribe

Digitizers to collect 3D data (13–17). Geometric morphometric data can also be used in simple

or complex ways, depending on how many traits are being observed and what kind of

phylogenetic or environmental traits they are being compared to (17). This section will outline a

sample of the various types of research within biology that have used the methods of geometric

morphometrics to inform on larger evolutionary questions.

Neosauropod dinosaurs lived during the Jurassic period, were herbivorous, and exhibited

the extreme end of gigantism (18). A 2007 study on these giant dinosaurs used geometric

morphometrics as one of their methods to research long bone scaling patterns (18). The authors

set out to test the hypothesis that neosauropod long bones scaled isometrically as well as to

investigate any paleobiological implications of these trends (18). Previous studies have indicated

that as neosauropods age, their long bones grow isometrically, meaning that no significant shape

change occurs through growth (18). This is somewhat surprising, considering the enormous

weight and size gain each individual experienced aging from juvenile to adult (18). Supporting

previous studies with more accurate geometric morphometric data would indicate that adult

neosauropods were quite limited in their locomotive movements and speed (18). The authors

also looked for any indication of allometric growth, which would mean that different body parts

grow at different rates. The samples used consisted of femurs and humeri and represented six

neosauropod taxa; both geometric morphometric and distance-measured data were collected on

all samples (18). The geometric morphometric data was collected through photography and

computer software which identified landmarks on each fossil (18).

After many multivariate statistical tests, the authors found that their data did support the

hypothesis that neosauropod long bones grew isometrically and did not exhibit a pattern of

allometry (18). Based on this data, as well as previous studies, and other observations (such as

the fact that neosauropod dinosaur long bones did not have a medullary cavity), the authors

hypothesize that with increasing mass, it might have been necessary for neosauropods to shift

more weight support to their limb bones and away from their limb muscles (18). This, along

with possible behavioral adaptations which could have included avoiding strenuous physical

activity, could have resulted in the lack of muscle “sculpting” exhibited in the long bones and the

observed isometric growth (18). It is likely that neosauropods were simply highly efficient

walkers, and that they did not change their locomotive patterns as they grew from juvenile to

adult (18). This research is a great example of how geometric morphometric analysis can help to

increase accuracy and further inform on existing research.

A later study on theropod dinosaurs, in 2012, used geometric morphometrics to help

answer remaining questions concerning macroevolutionary patterns of their cranial morphology

and how these may or may not relate to their feeding patterns (13). Theropod dinosaurs existed

for over 160 million years, and included the famed Tyrannosaurus rex. A long-standing question

in all evolutionary fields is whether or not phylogenetic constraint or functional adaptation is

more important in shaping species morphology. The authors used landmark-based two-

dimensional geometric morphometrics to complete their analysis (13). The authors attempted to

identify 24 type one and type two landmarks on each of the 51 specimens (one specimen per

species) available for study, although, because of incomplete fossils and missing data, they were

forced to create two data sets (13). The first included 26 species on which all landmarks could

be recorded, and the second included 36 species for which a reduced set of 13 landmarks could

be recorded (13). The second set maximized the number of landmarks while still including

representatives from all major subclades of nonavian theropod dinosaurs (13). The landmarks

were recorded through two-dimensional photography and computer software, including MorphoJ

(13). The authors then ran a series of multivariate statistics, including a Generalized Procrustes

Analysis, and multiple Principal Components Analyses, to compare the landmark data to bite

force and speed data, as well as phylogenetic data (13).

Through their analyses, the authors found that theropod crania differ primarily in relative

antero-posterior length and snout depth, and that “oviraptorosaurs deviate most strongly from the

‘typical’ and ancestral theropod morphologies” (13). They also found that large-bodied

carnivores independently converged on the same region of morphospace, while noncarnivorous

taxa generally fell out in distinct regions of morphospace and exhibited greater overall disparity

(13). Morphospace is considered the region in which landmarks from the crania are plotted into,

along with information about diet and environment, and which can then offer information about

broad patterns based on crania shape (13). Overall, the distribution of taxa in morphospace was

strongly correlated with phylogeny but only weakly correlated with biting and feeding behavior

(13). The authors conclude that phylogeny and not adaptive feeding behavior was the major

determinant of broad patterns in the skull shape of theropod dinosaurs (13). This research is a

good example of how geometric morphometrics can inform on larger evolutionary trends in

morphology and help to unravel remaining mysteries. The authors indicate that more research

will continue as more data, including possibly 3D techniques, become available (13).

A 2007 study on East African cichlids also set out to answer the question of whether or

not phylogeny or adaption held a stronger sway on shape change over time (14). East African

cichlids are famous for their intense variation and explosive speciation and offer great

opportunities for studies on evolution, diversification, and adaptative radiation (14). The authors

used over 1,000 specimens, which represented 45 species and 14 of the 17 major fish tribes in

Lake Tanganyika (14). Digital pictures were taken of the profile of each fish, and 17 landmarks

were then digitally recorded from each photo (14). The authors also coded each fish species for

its preferences of habitat, such as preferred water column depths, preferred substrate (mud, sand,

rock, etc.), feeding preferences and type of prey, the type of parental care given within the

species, the mating system (monogamy, polygamy, etc.), and it’s breeding type (mouthbreeders

or substrate guarders) (14). The phylogenetic analysis of each fish species consisted of

mitochondrial DNA sequencing and comparison (14). All of these types of data were then

compared to each other and run through statistical analyses to determine whether the adaptive

characteristics or the phylogenetic information was more influential to the shape changes among

the cichlids (14).

The authors found that the influence of phylogeny on the evolution of shape in the Lake

Tanganyika cichlids is small and that shape is much more influenced by ecology and habitat

adaptation (14). The authors were able to conclude that ecology plays a very large role in

generating morphological diversity (14). These results are different from the study on Theropod

dinosaurs which concluded that phylogeny was the more important factor (13). These varying

results are probably due to the fact that the research on Therapods consisted of animals which

spanned millions of years, while the research on cichlids consisted of fish that have relatively

recently speciated (13,14). Both studies display the informative power of geometric

morphometrics and how the method can help to further understandings of evolution, shape

change, phylogeny, and habitat adaptation.

Attempting to determine sex through the use of geometric morphometrics is not unique to

research on Homo sapiens; a recent study on neon flying squid (Ommastrephes bartramii) also

set out to find shape differences between males and females, as well as to attempt to differentiate

between stocks of the squid within the northwestern Pacific Ocean (15). This species of squid is

fished annually as a good source of protein for human consumption, and management of the

species and the two stocks is important to ensure the continued survival of the species for the

ocean ecosystem as well as a source of food for people (15). The researchers investigated the

shape of the squids’ upper and lower beaks, as well as the pigmentation levels of the beaks to see

if these characteristics could be used to determine sex and/or which stock (eastern or western)

they originated from (15). The upper and lower beaks of 216 random O. bartramii were digitally

photographed, and then 8 landmarks on the upper beak and 10 landmarks on the lower beak were

digitized (15). Semi-sliding landmarks were then added to both upper and lower beaks of all

specimens (15). The authors used a series a multivariate statistics to analyze the shape change

among the beaks, including a Generalized Procrustes Analysis, a Principal Components

Analysis, a MANOVA test, and thin-plate spline grids (15).

The results showed that the upper and lower beak shapes differed significantly between

the western and eastern stocks of squid, and could be used to differentiate between the two (15).

The authors also found that the degree of pigmentation of the upper beak changed between the

stocks, adding a second indicator to which regional group they originated from (15). The only

shape differences between sexes the authors could identify occurred in the western stock of

squid, and could not be used to differentiate males and females on random squid from both

stocks (15). The authors hypothesize that these results occur due to the variation of feeding

habits between the two stocks of squid but the sharing of a similar habitat by both sexes (15).

Shape differences between the sexes could not be identified in this study, however, the authors

indicate that further research will be done in an effort to determine the best way to differentiate

between the sexes using other parts of the squid (15). This research is a great example of how

geometric morphometrics can be used across multiple species in an effort to learn more about

shape differences in males and females as well as different geographical groups.

A very early study (2000) used a Microscribe Digitizer to determine how high in

taxonomic orders geometric morphometrics could effectively be used (16). The authors explain

that most geometric morphometric research focuses on one species, or perhaps one genus or

family (16). To test how much variability could be effectively studied through geometric

morphometrics, the authors digitized 23 skulls all representing different ordinal mammalian

groups (16). The authors chose 35 landmarks to collect on each skull, the identification of which

proved to be quite difficult considering the vast variability among mammalian crania (16).

Rather than using the word “homologous” to describe all of the landmarks between each species,

the authors defined their landmarks as “equivalent” (16). TPSSMALL and Morpheus computer

software was then used to analyze the Procrustes distances and the tangent space distances of the

data (16).

The authors found that the tangent space analyses were decent representations of the

Procrustes distances, however when the data was used to create a phylogeny, it displayed almost

no consensus with current (2000) morphological phylogenies (16). The authors conclude that

this study included too much variation for geometric morphometrics to be completely useful

(16). The essential shape analysis was correct; however, the data was unable to create any useful

comparison of phylogeny (16). The authors also explain that part of this is likely due to the

difficulty of choosing and identifying landmarks on such variable crania (16). This study is a

good example of researchers testing the boundaries of geometric morphometrics in biology as a

whole to show that the method is valid but must be applied to the correct type of specimens with

less variation among them.

It is also quite common for geometric morphometrics to be used in tandem with other

methods, such as earlier when pigmentation in squid was analyzed alongside their beak shapes

(15). Another recent study used geometric morphometrics to study the evolutionary patterns and

trait characteristics of ground squirrels (Marmotini) (17). The authors compared data sets of

body size, and cranial, mandibular, and molariform tooth shape to analyze variation, covariation,

and disparity patterns in a phylogenetic framework (17). A mitochondrial DNA dataset was used

for the phylogenetic analysis and comparison, while geometric morphometric data was used for

the rest of the analyses (17). The authors used a Microscribe Digitizer to collect 50 3D

landmarks from 136 specimens, representing 65 marmotine species (17). After these landmarks

were digitized, computer software was used to add semi-sliding landmarks along the midline of

each cranium (17). The authors also used a previously collected dataset of 2D landmarks to

gather information on molariform tooth shape and mandibular shape (17). This dataset consisted

of 58 marmotine species and also included regular landmarks and semi-sliding landmarks (17).

Multivariate statistics, such as GPA, PCA and MANOVA were used to compare and analyze all

of this data together (17).

The authors found strong correlations between body size and cranial traits, and that

evolutionary modes were concordant across traits. They also found “divergent dynamics on the

macroevolutionary landscape, with phenotypic disparity being differentially shaped by

convergence and conservatism” (17). These findings reiterate the mosaic nature of

morphological evolution and indicate that the evolution of ground squirrels is poorly captured by

single process descriptors (17). This research shows that when researching evolutionary

patterns, morphological traits should be studied in groups, against multiple phylogenetic and

environmental factors, rather than a single-trait approach. It also demonstrates how different

types of geometric morphometric data can be used together to help inform on many evolutionary

patterns and can be used for many types of research, no matter how simple or complex. The

success of these research projects using geometric morphometrics, and the ability to show shape

change within and between species, indicates that this type of analysis should yield useful

information when applied to male and female human pelves.

Landmarks

For this research, landmarks are defined as a specific point on bone that can be located on

each sample and subsequently recorded in 3D space (19). A preliminary list of landmarks was

used in the pilot project stage of this research and can be found in Appendix A. These

preliminary landmarks were chosen based on previous research, traditional landmark location,

and some new landmarks were chosen by the author (6,20,21). New landmarks were chosen in

an effort to improve existing landmarks and overall shape analysis. When collecting data on

landmark points, whether it is collected with a digitizer or through the use of a camera, it is

important to distinguish which type of landmark it is. Landmarks can be categorized as type I,

II, or III (1,2). Type I landmarks are considered the easiest to find and consist of one single point

on the bone where tissue transitions, such as an intersection of sutures (2,6,22). Type II

landmarks are considered the points of maximum curvature or greatest muscle attachment, an

example of this would be the euryon on the cranium (2,22). Type III landmarks are the most

extreme points of a structure overall, sometimes labelled as the “posteriormost” or

“anteriormost” points (2,6). A category that has since been added to this list by subsequent

studies is constructed landmarks, which is defined as “points corresponding to locations that are

defined using a combination of traditional landmarks and geometric information” (4). For

example, calculating and using the midpoint along a line of maximum width as a landmark.

Most studies using geometric morphometrics distinguish which types of landmarks are used to

help determine which types of points are most useful or which points introduce the most error

(2,4,22). Sliding semi-landmarks were not used in this study because when morphological

variation is small, such as that between modern humans, the differences between a landmark-

based study and a sliding semi-landmark based study using the same sample can be altered or

skewed, based on differences in initial alignment; thus raising questions about any results

obtained by using sliding semi-landmarks (23). Throughout the pilot project, the author tested

the preliminary landmarks to determine if they were easily located and if they accurately

captured the shape of the pubic bone (see Appendix A for landmark locations). These

preliminary tests also helped the author determine which type of landmark is most effective, and

if types II, III, and constructed landmarks could be easily replicated. The results of these tests,

and the explanation of the final set of landmarks can be found in a later section.

Statistics

Once all data is recorded, a series of statistical analyses will be run in order to identify

what, if any, differences exist between the shape of male and female pubic bones. For the

purpose of geometric morphometrics, shape has been define as “all the geometrical information

that remains when location, scale and rotational effects are filtered out from an object” (19,24).

The first statistical step of every geometric morphometric project is a Generalized Procrustes

Analysis, or GPA. Before any analysis is completed, the landmark coordinates include

information on size, shape, position and orientation (25,26). In order to analyze just shape

however, all of this other information must be excluded (26). A GPA eliminates the non-shape

variation, and rescales and rotates each sample so that they are relative to each other in the same

plane (1,26). By doing this, the GPA translates all landmark configurations to the same centroid,

scales all configurations to the same centroid size, and iteratively rotates all configurations until

the summed squared distances between the landmarks and their corresponding sample average is

a minimum (26). After the GPA is run, all configurations are super-imposed on each other, and

the resulting coordinates are called Procrustes shape coordinates and only include information

about the shape of the configurations (26). This statistical step alone does not offer a lot of

useful information, but rather, it prepares the data to be analyzed further in ways that will offer

more information about the shapes that are present.

The next step is to run a Principal Components Analysis, or PCA. A PCA is a way to

represent the variation that is present within the sample, and the goal of a PCA is to determine

which variable introduces the most amount of variation (19,27). When the PCA is run, it creates

a covariance matrix which shows the principal components of the data, and from this, the eigen

vectors of the present variation (19). The principal components provide a means of comparing

the relative importance of each dimension of the data (19). If the first two eigen vectors

represent 50% of the variance, or more, then the test can be considered successful. The PCA

also offers a scatter plot, which shows groupings or clusters and outliers (28). The PCA scatter

plot is very useful in determining whether or not the variation in the data is being explained by

the variable in question. To do this, the researcher can see on the plot which specimens are

grouping together and can then go to those specimens to see what they have in common. Vice

versa, the researcher can identify the specimens on the plot which graph far apart and can then

identify the differences between the physical specimens.

Further statistical tests will be explored once data collection at the Maxwell Museum is

completed. These future tests include Discriminant Function, MANOVA, and independent

group t-tests of centroid sizes. At this point, it will be determined which statistical analyses will

yield the most useful information concerning the differences between male and female pelves.

The Pilot Project

The goals of the pilot project stage of this research were to both familiarize the author

with geometric morphometric data and the Microscibe Digitizer equipment, as well as to indicate

whether or not this specific approach to sexing a skeleton would work. It was also used to refine

the methods and landmarks used in order to lessen error when the larger datasets are collected.

The project used specimens from the University of Montana Forensic Anthropology Teaching

Collection and consisted of 14 individuals, 9 males and 5 females, with a total of 24 pubic bones

available. There is no known provenience information for these individuals. 13 data points were

collected on each pubic bone using the Microscribe Digitizer. These points in space were

recorded in an Excel Spreadsheet, and then transferred to MorphoJ, a software program designed

to analyze 3D data.

Limited analysis was conducted on this data because of the nature of a pilot project, in

that it should only be used as an early indicator and test for what should follow when larger data

sets are analyzed. Generalized Procrustes Analysis and Principal Components Analysis were the

main statistical tests run on this preliminary data and showed promising results. The PCA Eigen

values clearly showed more than 50% of the variance within the first two principal components,

which indicates a successful test (see Appendix B for graph). The scatter plot of the PCA

displays a tight cluster in the middle, which consists of all male specimens, except for one

female. The rest of the female specimens are scattered throughout the rest of the graph (see

Appendix B for graph).

These results are promising because the male specimens are clearly clustering together,

which indicates that sex is explaining most of the variance within the sample. The largest

question is why the female specimens did not form a separate cluster. This may be answered by

the fact that females tend to undergo an intense form of trauma which alters the pelvis in

unknown ways, pregnancy and child birth (29). The lasting changes and effects of pregnancy on

the female pelvis is still poorly understood and it seems as though pregnancy affects different

women in different ways (29,30). There is no way to know if any of the female individuals

included in this preliminary research had ever been pregnant or given birth, but if some of them

had while others hadn’t, it may explain why they do not cluster together. It is also possible that

age may be explaining some of the variance in the sample. It is known that age affects shape

change in the pelvis, and it is hard to tease that out of this preliminary data, considering how

small the sample size is (21,31,32).

Intraobserver error was also tested for this data to determine how consistently the author

was recording landmarks. Unfortunately, the average standard error was 0.79, and overall error

ranged from 0.5-1.1, all of which are relatively high error rates. Some error can be explained by

inexperience of the author, which should improve with time. To help lower error further

however, the author determined which landmarks had the lowest and highest error rates.

Landmarks 1, 2, 4, and 13 (see Appendix A for landmark locations) displayed the lowest error

rates. Landmark 1 is the pubic tubercle, and 2, 4, and 13 are all landmarks which are found by

identifying maximum lengths, which means these are all quite easy to find. The landmarks with

the highest error rates (landmarks 5, 6, 7 and 8) were constructed landmarks, and proved quite

difficult to locate. In order to lessen overall error rates, the landmarks which proved to be the

most difficult to locate have been removed and the definitions of the easier, remaining landmarks

have been refined (see Appendix C for final list of landmarks).

It is also important to discuss the sample biases present in this pilot project and address

how they may be controlled for in future research. The largest bias is the small sample size,

which will easily be remedied by simply gaining access to larger collections, such as the

Maxwell Museum. This sample also included almost twice as many males as females. While

this is hard to control considering researchers generally want to use as many specimens as is

possible, it may be prudent to exclude some in some analyses to determine if evening out males

and females affects the data in any way. More than likely though, once the sample size reaches a

respectable number, the number of males and females will be quite close to each other and

should not present further issues. The individuals within this pilot research are also likely to

have varying ancestries, which can introduce variation in unknown ways. Knowing the ancestry

of individuals is one of the large advantages when using a documented collection. The question

of ancestry will be easily addressed through the curation of known individuals within Montana

and the use of the documented collection at the University of New Mexico. Once the ancestries

are known, individuals can be grouped according to ancestry to determine whether or not

ancestry affects the overall sex determination process. It will also be known whether or not the

female specimens were ever pregnant, which may help to explain the wide variation that is

already being observed between female individuals. Using samples that include known

provenience and health backgrounds will vastly improve the information offered by future

research based on this preliminary study.

Overall, the pilot project met the original goals of the researcher. Experience and

familiarity using the equipment and software was gained and will hopefully be evident in lower

error rates as future research occurs. The author was also able to refine the process of gathering

data and narrow down the landmarks which should be most useful in attempting to determine sex

of unknown individuals. Most importantly, the preliminary data offered promising results that at

least male specimens may cluster together and lead to a sex differentiation among larger sample

sizes. The pilot project acted as a very important learning experience and test for the larger

dissertation research that will follow and should prove to have been one of the most important

steps in creating a new method to metrically sex unknown human pelves.

Appendix A: Preliminary Landmark Descriptions and Types

Number

Landmark

Description

Type

Pubic Tubercle

Most prominent point of the pubic tubercle

Traditional

(II)

Superior Pubic

Symphysis

The most superior point of the pubic symphysis

Extremal

(III)

Inferior Pubic

Symphysis

The most inferior point of the pubic symphysis

Extremal

(III)

Lateral Border

Point on the lateral border of the pubic body which

would create the maximum breadth of the obturator

foramen

Extremal

(III)

5, 6

Symphysis

Width

The anterior (5) and posterior (6) points which

create the maximum width of the pubic

symphyseal surface

Constructed

Pubic

Symphysis

Midpoint of the pubic symphysis; found by

determining both the maximum length and width,

point is at their intersection

Constructed

Surface

Midpoint of the anterior surface; found by

determining both the maximum length and width,

point is at their intersection

Constructed

Posterior

Surface

Midpoint of the posterior surface; found by

determining both the maximum length and width,

point is at their intersection

Constructed

10, 11

Pubic Body

Height

The superior (10) and inferior (11) points which

create the maximum height of the pubic body

Constructed

12, 13

Pubic Body

Width

The medial (12) and lateral (13) points which

create the maximum width of the pubic body

Constructed

Anterior Surface,

Right Pubic Bone

Sympheseal Surface,

Right Pubic Bone

Posterior Surface,

Right Pubic Bone

Appendix B: Graphed Results of Pilot Project (Principal Components Analysis)

PCA Eigen Vectors

PCA Scatter Plot (arbitrary circle placed to show general clustering)

Appendix C: Final List of Landmarks, Descriptions and Types

Number

Landmark

Description

Type

Pubic Tubercle

Most prominent point of the pubic tubercle

Traditional

(II)

Superior Pubic

Symphysis

The most superior point of the pubic symphysis

Extremal

(III)

Inferior Pubic

Symphysis

The most inferior point of the pubic symphysis

Extremal

(III)

Lateral Border

Point on the lateral border of the pubic body which

would create the maximum breadth of the obturator

foramen

Extremal

(III)

5, 6

Pubic Body

Height

The superior (5) and inferior (6) points which

create the maximum height of the pubic body

Extremal

(III)

7, 8

Pubic Body

Width

The medial (7) and lateral (8) points which create

the maximum width of the pubic body

Extremal

(III)

References

1. Zelditch ML, Swiderski DL, Sheets HD, Fink WL. Geometric Morphometrics for

Biologists. First. San Diego, California: Elsevier Academic Press, 2004.

2. Toon C. Sexual dimorphism at the proximal tibia: a geometric morphometric analysis. 2014.

3. Lesciotto KM. The Impact of Daubert on the Admissibility of Forensic Anthropology

Expert Testimony. J Forensic Sci 2015;60(3):549–55. https://doi.org/10.1111/1556-

4029.12740.

4. Bytheway JA, Ross AH. A Geometric Morphometric Approach to Sex Determination of the

Human Adult Os Coxa: A GEOMETRIC MORPHOMETRIC APPROACH TO SEX

DETERMINATION. J Forensic Sci 2010;55(4):859–64. https://doi.org/10.1111/j.1556-

4029.2010.01374.x.

5. Berthelot CM. Metric, nonmetric, and geometric morphometric methods of sex estimation

using the distal humerus. 2014.

6. Costello A. An analysis of sexual dimorphism using geometric morphometrics of the femur

and tibia: The use of GM in assessing sex of fragmented remains. 2015.

7. Franklin D, Oxnard CE, O’Higgins P, Dadour I. Sexual Dimorphism in the Subadult

Mandible: Quantification Using Geometric Morphometrics. J Forensic Sci 2007;52(1):6–10.

https://doi.org/10.1111/j.1556-4029.2006.00311.x.

8. Jung H, Woo EJ. Evaluation of Mastoid Process as Sex Indicator in Modern White

Americans using Geometric Morphometrics. J Forensic Sci 2016;61(4):1029–33.

https://doi.org/10.1111/1556-4029.13079.

9. Lynch JJ, Cross P, Heaton V. Sexual Dimorphism of the First Rib: A Comparative

Approach Using Metric and Geometric Morphometric Analyses. J Forensic Sci

2017;62(5):1251–8. https://doi.org/10.1111/1556-4029.13421.

10. UNM Maxwell Museum of Anthropology | Documented Collection. osteolab. 2018.

http://osteolab5.wixsite.com/osteolab/documented-collection (accessed July 9, 2018).

11. Komar DA, Grivas C. Manufactured populations: What do contemporary reference skeletal

collections represent? A comparative study using the Maxwell Museum documented

collection. Am J Phys Anthropol 137(2):224–33. https://doi.org/10.1002/ajpa.20858.

12. Doyle JM, Kao G. Are Racial Identities of Multiracials Stable? Changing Self-Identification

Among Single and Multiple Race Individuals. Soc Psychol Q 2007;70(4):405–23.

13. Brusatte SL, Sakamoto M, Montanari S, Harcourt Smith WEH. The evolution of cranial

form and function in theropod dinosaurs: insights from geometric morphometrics: Geometric

morphometrics of theropod dinosaurs. J Evol Biol 2012;25(2):365–77.

https://doi.org/10.1111/j.1420-9101.2011.02427.x.

14. Clabaut C, Bunje PME, Salzburger W, Meyer A. GEOMETRIC MORPHOMETRIC

ANALYSES PROVIDE EVIDENCE FOR THE ADAPTIVE CHARACTER OF THE

TANGANYIKAN CICHLID FISH RADIATIONS. Evolution 2007;61(3):560–78.

https://doi.org/10.1111/j.1558-5646.2007.00045.x.

15. Fang Z, Chen X, Su H, Thompson K, Chen Y. Evaluation of stock variation and sexual

dimorphism of beak shape of neon flying squid, Ommastrephes bartramii, based on

geometric morphometrics. Hydrobiologia 2017;784(1):367–80.

https://doi.org/10.1007/s10750-016-2898-0.

16. Marcus L, Hingst-Zaher E, Zaher H. Application of landmark morphometrics to skulls

representing the orders of living mammals. Hystrix Ital J Mammal 2000;11(1).

17. McLean BS, Helgen KM, Goodwin HT, Cook JA. Trait-specific processes of convergence

and conservatism shape ecomorphological evolution in ground-dwelling squirrels: TRAIT-

SPECIFIC EVOLUTIONARY PROCESSES. Evolution 2018;72(3):473–89.

https://doi.org/10.1111/evo.13422.

18. Bonnan MF. Linear and Geometric Morphometric Analysis of Long Bone Scaling Patterns

in Jurassic Neosauropod Dinosaurs: Their Functional and Paleobiological Implications. Anat

Rec Adv Integr Anat Evol Biol 2007;290(9):1089–111. https://doi.org/10.1002/ar.20578.

19. Stegmann MB, Gomez DD. A brief introduction to statistical shape analysis. Inform Math

Model Tech Univ Den DTU 2002;15(11).

20. San-Millán M, Rissech C, Turbón D. Shape variability of the adult human acetabulum and

acetabular fossa related to sex and age by geometric morphometrics. Implications for adult

age estimation. Forensic Sci Int 2017;272:50–63.

https://doi.org/10.1016/j.forsciint.2017.01.005.

21. White TD, Folkens PA. The Human Bone Manual. 1 edition. Burlington: MA: Academic

Press, 2005.

22. Perlaza NA. Sex Determination from the Frontal Bone: A Geometric Morphometric Study. J

Forensic Sci 2014;59(5):1330–2. https://doi.org/10.1111/1556-4029.12467.

23. Perez SI, Bernal V, Gonzalez PN. Differences between sliding semi-landmark methods in

geometric morphometrics, with an application to human craniofacial and dental variation. J

Anat 2006;208(6):769–784.

24. Kendall DG. A survey of the statistical theory of shape. Stat Sci 1989;:87–99.

25. Bookstein FL. Landmark Methods for Forms without Landmarks: Morphometrics of Group

Differences in Outline Shape. Med Image Anal 1996;1(3):225–43.

26. Mitteroecker P, Gunz P, Windhager S, Schaefer K. A brief review of shape, form, and

allometry in geometric morphometrics, with applications to human facial morphology.

Hystrix Ital J Mammal 2013;24(1):59–66.

27. Shlens J. A tutorial on principal component analysis. 2014.

28. Wold S, Esbensen K, Geladi P. Principal component analysis. Chemom Intell Lab Syst

1987;2(1–3):37–52.

29. McFadden C, Oxenham MF. Sex, Parity, and Scars: A Meta-analytic Review. J Forensic Sci

2018;63(1):201–6. https://doi.org/10.1111/1556-4029.13478.

30. Ubelaker DH, Paz JSDL. Skeletal Indicators of Pregnancy and Parturition: A Historical

Review. J Forensic Sci 57(4):866–72. https://doi.org/10.1111/j.1556-4029.2012.02102.x.

31. Bass WM. Human Osteology: A Laboratory and Field Manual. 5th ed. edition. Columbia,

Mo: Missouri Archaeological Society, 2005.

32. White T, Black M, Folkens P. Human Osteology. 3rd ed. Burlington: MA: Elsevier

Academic Press, 2012.

Chapter Four: Implications and Relevance

The available research and literature in geometric morphometrics has been growing,

although, there are still gaps that need to be filled (1). It is also important to repeat existing

studies in order to replicate results, determine usefulness, and propose revisions to improve upon

past research. This project will draw on preexisting research and integrate new methods to study

the shape of the pubic bone in an effort to bolster the available information and further the

applicability of geometric morphometrics in forensic and bioarchaeological contexts. This

method has the potential to aid in the sex determination of both whole and fragmentary skeletal

remains in modern forensic anthropological casework and bioarchaelogical sites, mass graves,

and commingled remains. 3D methods are also becoming more and more popular in the research

setting, meaning it is important to study their reliability and test the boundaries of their uses in

each field. There is still a gap in the field of forensic anthropology when it comes to 3D

techniques, although research similar to this is quickly starting to bridge existing knowledge.

This research is also important because it discusses the difficulties and technical aspects of

curating a new and modern collection of human skeletal remains, a goal which is often

unattainable for many university programs. All of these aspects contribute to the importance and

contemporary relevance of the current research and indicate that this new method has the

potential to contribute greatly to the field of forensic anthropology.

It is rare in forensic and bioarchaeological contexts to recover entire skeletons (2–4).

Oftentimes, long periods of time will pass before a body is discovered, which means that in a

clandestine burial, there are plenty of chances for weather events, like rain, and outside forces,

like grazing cattle, to disturb, break, and move bones. In a forensic case, the body may have been

deliberately separated from itself, in order to more effectively hide it, and only parts may be

found. It is also very difficult to ensure that all elements for each individual are paired and

sorted correctly when in the context of a mass grave or commingled burial. For these reasons, it

is important to continue improving and revising more accurate ways to determine sex from

fragmentary remains of all parts of the body (2–4). This research focuses on the body of the

pubic bone, which is a very dense piece of bone that may stand better chances of surviving

through time to later be recovered than, other, less-dense parts of the skeleton. This type of

technique could not only improve modern forensic casework, but also help in the identification

of victims of mass or commingled graves.

A forensic anthropologist would count themself lucky if given an entire human skeleton

to analyze for a case. More often than not, case work consists of fragmentary remains, which

may or may not represent most elements of the body. Sometimes, only a cranium is recovered,

while other times, broken pieces of many bones are recovered. The anthropologist is still

expected to offer their expert opinion on as many aspects of the biological profile as possible.

This includes sex, age, stature, ancestry, and an analysis of trauma (5). Multiple methods exist to

estimate all of these characteristics; however, it is generally agreed that multiple methods should

be used in conjunction to make determinations. This can be quite difficult to do when presented

with an incomplete skeleton. Adding into the complicated nature of fragmentary remains is the

knowledge that metric analyses are more trusted by court-systems and juries than are subjective

morphoscopic analyses (6). There are few widely accepted metric assessments of sex and age

based on fragmentary pieces of the skeleton in the existing anthropological literature (7). The

current research will create a starting point for a new method which could greatly aid in the

metric sex assessment of fragmentary remains.

Research in metric sex assessment is growing and quickly being added to, however,

assessments in fragmentary portions are still lacking. A recent study looked to determine how

well the pelvis, when digitized, could determine sex (7). The authors used 200 left os coxae and

collected 36 landmark points per os coxa (7). Overall, the authors found that by using geometric

morphometric analyses on the pelvis, the could determine sex 98-100% of the time (7). This is an

incredibly high rate of accuracy and is consistent or better than traditional methods (generally

90-95% accuracy) to determine sex using the pelvis (8). The authors concluded that the pelvis is,

as expected, a highly accurate element to use to determine sex; they also determined that

geometric morphometrics is a reliable and accurate method to apply (7). One issue, making the

applicability of this research an issue, is that an entire os coxa is needed to make a sex

estimation. The current research will improve on this study further by focusing the method onto

one small portion of the pelvis, which could potentially be applied when the pelvis is recovered

in a fragmentary state.

Commingled remains and mass graves present unique problems in forensic and

bioarchaeological casework and sites. Usually, determining a minimum number of individuals

(MNI) is the first main goal, after which, assigning elements to individuals is attempted. Pair

matching elements, for example, both of these humeri belonged to the same individual, or, this

pelvis belongs with this cranium, can be very difficult (9). Many times, using a sex estimation of

different elements can aid in the individualization of bones. It is likely that in a commingled

setting, not all elements would be complete, meaning that accurate metric methods on small

portions of bone would offer an advantage to the anthropologist. Confidently sexing pubic bones

could potentially aid in the overall analysis of a mass and/or commingled grave, at the very least,

offering information about the male to female ratio of the recovered individuals. In some

situations, it may also help to associate a pelvis with a certain cranium or other elements in order

to attempt to complete more individuals. In a forensic mass grave context, it could also help to

identify unknown individuals by confidently assigning them to one sex or the other, especially if

the rest of the pelvis and/or cranium are missing or too fragmented to use.

3D technologies are becoming more and more common in many areas of research and

science (1,10). As 3D techniques become more common in physical anthropology laboratories, it

is vital to determine what type of data collection is most useful and to create standard methods of

collection that can be used in various scenarios. The ultimate goal of forensic anthropology is to

help identify victims and unknown individuals, so it is important that as a science, the results

stand up to other disciplines and the court systems in which we, as forensic anthropologists,

work. It is also important to continue working to determine which type of techniques can

accomplish which types of analysis. 3D technologies are becoming more common, unfortunately

though, many are still very expensive and not available to all researchers. The goal of this project

is to use a less expensive piece of equipment (Microscribe Digitizer) to create an accurate metric

method. The results of this project will determine whether or not the Digitizer should be used for

data collection similar to this and whether or not the proposed method is accurate.

It is also important in a modern forensic context to improve quantitative proof of results

because a subjective morphoscopic review is less likely to be accepted in a court of law. Metric

results are much harder to contest than a morphoscopic assessment and are more useful to the

expert witness and the victim because of the subjectivity of a morphoscopic assessment. Since

the 1993 Daubert decision, scientific standards for court cases have been tested and questioned

for all fields of inquiry, however, it is particularly important in forensic anthropology that we be

able to display the validity of our determinations considering how much of the field is based on

experience (6,11,12). Most of the standard methods in forensic anthropology consist of

morphoscopic analysis, such as the classification of a phase on the auricular surface to determine

age, and these methods are inherently subjective. It is still imperative to ensure consistency and

reliability in the application of forensic anthropological methods; because of this, many

researchers are moving towards more metric and 3D technologies which are not subjective and

can still offer accurate and reliable results and determinations (11). As with all new technologies,

it is important to standardize the way in which they are used to obtain results (11). It has been

shown that a digitized pelvis can accurately determine sex (7), so testing and refining this

method is one step closer to being able to standardize the use of a Microscribe Digitizer in the

determination of sex in the human pelvis.

Most literature concerning the pelvis focuses on how to improve age-at-death estimations

(13,14). Based on a review of related literature, metric sex determination of the pelvis falls in a

clear gap. This dissertation will begin to fill this gap by providing a full 3D shape analysis of the

pubic body. The existing literature which does focus on sex determination uses morphoscopic

techniques on the pelvis as a whole. While these techniques have proven trustworthy by an

experienced observer, they are subjective and individuals new to the technique will often make

mistakes (15,16). It is important to have methods that are capable of determining sex on

fragmentary remains when the entire pelvis is not available, and to have methods that are easily

replicable. A review of the existing literature on applications of geometric morphometrics to the

human pelvis is included in the background chapter of this dissertation.

It is well known and accepted within anthropology that when developing a method which

is intended to be used on modern humans, it must also be developed and based on modern

humans (17,18). A part of this project consists of attempting to curate a new sample of modern

pubic symphyses from autopsy. This process, while rewarding, has been quite slow and has

presented many problems and learning opportunities for the author. The process included gaining

approval from the Medical Examiner’s Office as well as the University’s Internal Review Board

to begin a new collection consisting of those deceased individuals who have received an autopsy.

It was also necessary to gain the support and cooperation of the local Sheriff’s Office and

Coroners; because Montana does not employ Medicolegal Death Investigators, the Coroners are

in charge of determining whether or not an individual needs an autopsy. Ideally, any time an

individual was sent for an autopsy, the involved Coroner would email the autopsy request form

to the author of this project, as well as the State Crime Lab, to allow the author quick access to

the demographic information of the deceased individual. The author would then determine if the

individual seemed like a good donor based on the age of the individual, and whether or not their

driver’s license denoted them as an organ donor. If the author was also able to obtain the next of

kin information, she could then contact them to ask for permission for the donation. The next of

kin would need to sign a consent form and return it to the author before the autopsy was

completed so that the Medical Examiners could view consent and subsequently remove the

sample from the body.

Communication and timing proved to be the most problematic aspects of this process. As

it was not their main duty, Coroners would often forget to email the request to the author as well

as the Crime Lab, resulting in many individuals not being considered for donation. When the

author was aware of a new individual for consideration, it was often impossible to obtain next of

kin consent before the autopsy was completed and the deceased had been moved to a funeral

home. The first issue of communication is something that could be improved on through

continuation of a large project such as this. The longer the Coroners are asked to send on

information, the more often it would hopefully occur. The second issue is more difficult to

improve on considering how difficult it is to ask a family for consent of a donation. If the next of

kin simply does not want to discuss it, or move too slowly, there is nothing more to be done

about it. These occurrences are to be expected however, not all who are asked will consent to a

donation of part of their loved one. It is the hope of the author that graduate students at the

University of Montana will continue this curation in an attempt to create a much larger and more

useful contemporary sample of pubic symphyses, and it is the hope that with time, the process

will become more regular and faster in order to improve on the number of samples obtained.

The importance of developing new methods based on contemporary, modern, human

skeletal collections is to ensure that new methods can be applicable to modern forensic cases

(19,20). Contemporary collections can offer an abundance of information and research material

to forensic anthropologists and bioarchaeologists. A contemporary collection can be used to

study sex and sexual dimorphism, age, growth and development, morphological variation,

nutrition, disease, and can support the creation of new and improved methods concerning

skeletal analysis (20–23). Anatomically modern Homo sapiens have existed for approximately

200,000 years and the modern human skeleton has changed throughout this time, a concept

referred to as secular change (24,25). Due to this morphological change over time, it is important

to develop new methods and research on contemporary collections of skeletal remains (20).

A recent study showed that secular change can happen relatively quickly, and can affect

traits on the pelvis that are regularly used in forensic anthropology (24). Samples from the

Hamaan-Todd Osteological Collection and the William M. Bass Donated Skeletal Collection

were compared to denote any changes in the ischiopubic ramus, the subpubic angle, and the

ventral arc of the pelvis (24). The Hamaan-Todd Collection includes individuals that were born

between the mid-nineteenth century and the early twentieth century, while the William M. Bass

Collection includes individuals born since 1940 (24). The results showed significant shape

differences in all three traits for females and the ventral arc for males (24). The sex

determination methods which utilize these traits still offer high accuracy determinations for both

populations, but with such significant shape changes occurring over time, it is unknown how

long those changes will not have an effect on these methods (24). This study demonstrates the

need for continued contemporary skeletal collections.

It is also important to develop new collections world-wide because regional

morphological variations can also affect the accuracy of new and established methods (20).

Cross-comparison of collections throughout the world could help further understandings of how

humans develop and age, what kinds of sexual dimorphism is present, and how different disease

processes and cultural differences may or may not be expressed in the human skeleton (20).

Developing methods that are population-specific can help to increase accuracy for that specific

region, and may aid further in forensic investigations of unknown identity and missing persons.

By developing a new method based on the University of New Mexico Maxwell Documented

Skeletal Collection (consisting of only individuals who have passed away in the last 50 years),

this research will develop a new method to assess the sex of a modern human skeleton based on

the pubic bone (26). This method will then be tested on the new contemporary sample of pubic

symphyses at the University of Montana (consisting of individuals who have passed away in the

last year) in order to determine if the method seems to be regionally specific, or, if like most sex

determination methods, it is accurate across regions. This method will be usable for modern

forensic cases because it was developed based on very recent contemporary populations, rather

than an older collection, such as the Hamaan-Todd Collection.

This dissertation project is a small, but necessary step towards a better understanding of

the bones which make up the human pelvis, specifically, the shape of the pubic bone. Based on

the existing knowledge of morphoscopic sex differences in the pubic bone, it is expected that a

thorough metric analysis will yield significant and useful results. Creating a method which will

allow future forensic anthropologists to metrically determine the sex of an unknown pubic bone

fragment can potentially aid both future research in forensics and bioarchaeology, as well as

expert witnesses in court. As 3D technology becomes more common, affordable, and understood,

it will become the norm for research and analysis and so it is important to continue filling in the

existing gaps with new hypotheses and methods. Ultimately, the goal of forensic anthropology is

to recreate the life history of unknown people based on their skeletal remains. Improving the

methods in which we do this is necessary and vitally important to ensure our accurate portrayal

of those individuals we aim to help. This project also explores the successes and failures of

attempting the curation of a new collection and may hopefully lead to better methods of

communication and curation in Montana in the future.

References

1. Zelditch ML, Swiderski DL, Sheets HD, Fink WL. Geometric Morphometrics for

Biologists. First. San Diego, California: Elsevier Academic Press, 2004.

2. Bongiovanni R, LeGarde CB. A Univariate Approach to Sex Estimation for the

Fragmentary Upper Limb. Journal of Forensic Sciences 2018;63(2):356–60.

https://doi.org/10.1111/1556-4029.13530.

3. Macaluso PJ. Testing the Effectiveness of Two Cranial Base Foramina for Metric Sex

Assessment of Fragmentary Remains: EFFECTIVENESS OF TWO CRANIAL BASE

FORAMINA. Journal of Forensic Sciences 2012;57(4):1017–21.

https://doi.org/10.1111/j.1556-4029.2012.02066.x.

4. Dupras TL, Schultz JJ, Wheeler SM, Williams LJ. Forensic Recovery of Human Remains:

Archaeological Approaches, Second Edition. CRC Press, 2016.

5. Buikstra JE, Ubelaker DH, editors. Standards: for Data Collection from Human Skeletal

Remains. Arkansas Archeological Survey, 1994.

6. Lesciotto KM. The Impact of Daubert on the Admissibility of Forensic Anthropology

Expert Testimony. Journal of Forensic Sciences 2015;60(3):549–55.

https://doi.org/10.1111/1556-4029.12740.

7. Bytheway JA, Ross AH. A Geometric Morphometric Approach to Sex Determination of the

Human Adult Os Coxa: A GEOMETRIC MORPHOMETRIC APPROACH TO SEX

DETERMINATION. Journal of Forensic Sciences 2010;55(4):859–64.

https://doi.org/10.1111/j.1556-4029.2010.01374.x.

8. Klales AR, Ousley SD, Vollner JM. A revised method of sexing the human innominate

using Phenice’s nonmetric traits and statistical methods. Am J Phys Anthropol

2012;149(1):104–14. https://doi.org/10.1002/ajpa.22102.

9. Best KC, Garvin HM, Cabo LL. An Investigation into the Relationship between Human

Cranial and Pelvic Sexual Dimorphism. Journal of Forensic Sciences 0(0).

https://doi.org/10.1111/1556-4029.13669.

10. Sholts SB, Walker PL, Kuzminsky SC, Miller KWP, Wärmländer SKTS. Identification of

Group Affinity from Cross-sectional Contours of the Human Midfacial Skeleton Using

Digital Morphometrics and 3D Laser Scanning Technology* ,‡: HUMAN MIDFACIAL

CONTOUR ANALYSIS. Journal of Forensic Sciences 2011;56(2):333–8.

https://doi.org/10.1111/j.1556-4029.2011.01701.x.

11. Christensen AM, Crowder CM. Evidentiary Standards for Forensic Anthropology. Journal

of Forensic Sciences 2009;54(6):1211–6. https://doi.org/10.1111/j.1556-4029.2009.01176.x.

12. Grivas CR, Komar DA. Kumho , Daubert , and the Nature of Scientific Inquiry:

Implications for Forensic Anthropology. Journal of Forensic Sciences 2008;53(4):771–6.

https://doi.org/10.1111/j.1556-4029.2008.00771.x.

13. Berg GE. Pubic Bone Age Estimation in Adult Women*. Journal of Forensic Sciences

2008;53(3):569–77. https://doi.org/10.1111/j.1556-4029.2008.00712.x.

14. Wink AE. Pubic Symphyseal Age Estimation from Three-Dimensional Reconstructions of

Pelvic CT Scans of Live Individuals. J Forensic Sci 2014;59(3):696–702.

https://doi.org/10.1111/1556-4029.12369.

15. Gray A. Applicability of Three Dimensional Surface Scanning to Age-at-Death Estimations

based on the Human Pubic Symphysis. 2011.

16. Burch TC. Metric sex determination using minimum width of the pubic bone. 2003.

17. Toon C. Sexual dimorphism at the proximal tibia: a geometric morphometric analysis. 2014.

18. Komar DA, Grivas C. Manufactured populations: What do contemporary reference skeletal

collections represent? A comparative study using the Maxwell Museum documented

collection. American Journal of Physical Anthropology 137(2):224–33.

https://doi.org/10.1002/ajpa.20858.

19. White TD, Folkens PA. The Human Bone Manual. 1 edition. Burlington: MA: Academic

Press, 2005.

20. Eliopoulos C, Lagia A, Manolis S. A modern, documented human skeletal collection from

Greece. HOMO - Journal of Comparative Human Biology 2007;58(3):221–8.

https://doi.org/10.1016/j.jchb.2006.10.003.

21. Hunt DR, Albanese J. History and demographic composition of the Robert J. Terry

anatomical collection. American Journal of Physical Anthropology 2005;127(4):406–17.

https://doi.org/10.1002/ajpa.20135.

22. Miller-Shaivitz P. The feasibility of assembling a contemporary human skeletal collection

for forensic anthropological education and research. 1996.

23. Tobias PV. On the scientific, medical, dental and educational value of collections of human

skeletons. Int J Anthropol 1991;6(3):277–80. https://doi.org/10.1007/BF02444955.

24. Klales AR. Secular Change in Morphological Pelvic Traits used for Sex Estimation. Journal

of Forensic Sciences 2016;61(2):295–301. https://doi.org/10.1111/1556-4029.13008.

25. Stringer C. Lone Survivors. New York: Henry Holt and Company, LLC., 2012.

26. UNM Maxwell Museum of Anthropology | Documented Collection. osteolab. 2018.

http://osteolab5.wixsite.com/osteolab/documented-collection (accessed July 9, 2018).

Chapter Five: Three-dimensional geometric morphometric sex determination of the human

pubic bone

3-D sex determination of the pubic bone

Katherine Baca1, Brandon Bridge2, Meradeth Snow3

Submitted to PLOSone with Revisions

1Department of Anthropology, University of Montana, Missoula, Montana, United States of

America

2Department of Economics, University of Montana, Missoula, Montana, United States of

America

3Department of Anthropology, University of Montana, Missoula, Montana, United States of

America

Abstract

Geometric morphometrics has become a more popular method in anthropology as three-

dimensional data and research become more widely recognized and accessible. This research

provides a refined method utilizing 3-D geometric morphometric analysis to determine sex from

the human pubic bone. The study used a sample of N=378 individual pubic bones from the

University of New Mexico Maxwell Documented Collection. Eight landmarks were digitized on

each individual bone using a Microscribe Digitizer. Results from the Principal Components

Analysis provide promising clustering between male and female groups, as well as indications

that the method may be ancestry-specific, and that parity may have an effect on the shape of

female pubic bones. The Discriminant Function analysis of the training data set resulted in

96.2% accuracy in predicting the correct sex, and the testing data set resulted in 95.5% accuracy,

P<0.0001. To test the ability to replicate this method, the author collected data a second time on

a random set of 50 individuals, N=100 pubic bones and reran the GPA, PCA, and discriminant

function analyses. This second test resulted in 96.5% accuracy of the training data set, and 93.8%

accuracy of the testing data set. To test interobserver error, the author collected all eight

landmarks from the same bone once a day, six days in a row. The PCA scatter plot of this test is

presented to exhibit the extremely low variance between each instance of measurement.

Introduction

The use of geometric morphometrics has become increasingly popular in forensic

anthropological research [1–3]. Geometric morphometrics is, at its most basic level, the metric

analysis of shape [4]. This method has been used in many biological studies for at least the past

three decades, but is not commonly used in the field of Forensic Anthropology [4]. Geometric

morphometrics differ from traditional methods because morphoscopic analysis does not

metrically assess bone, and interlandmark distance measurements do not fully capture the shape

of a bone. In contrast, geometric morphometrics metrically analyzes the shape of a bone in two

or three dimensions [4,5]. Metric analyses are more objective and generally require less training

and experience than visual analysis techniques [6]. Thus, it is important to continue improving

metric analyses to aid in the forensic anthropologists’ biological profile determinations.

Much of the existing geometric morphometric work focuses on sex and ancestry

assessment of the cranium [1,5,6]. It is accepted, however, that the pelvis is the best indicator of

sex in human skeletal remains [6,9,10]. Research has also shown that metric analysis of the post-

cranial skeleton can be just as accurate, if not more accurate, in sex estimation than the cranium

[10]. Further research using 3D techniques on the post-cranial skeleton is needed to help

increase accuracy of sex estimation in forensic and archaeological contexts. Research focused

on the pelvis most commonly investigates changes in age, and few projects attempt to metrically

determine sex [11,12]. The DSP method is one of the few large-scale metric methods of sex

determination using the pelvis and has been shown to be quite accurate [13]. This method utilizes

interlandmark distance measurements however, and is based on archaeological samples; both of

which are not ideal for use in modern forensic casework [13]. It is especially important today to

develop accurate metric analyses due to the need for reliable and objective results for instances

in which the remains might be evidence in a court room [14]. Improving the specificity of sex

determination in the pelvis has the potential to improve both modern forensic anthropology as

well as analyses within bioarchaeology.

It is important to develop new methods using a contemporary collection in order to

ensure the method’s accuracy when applied to modern forensic cases [15]. This helps to avoid

biases that may be introduced by secular change in a population. Anatomically modern Homo

sapiens have existed for approximately 200,000 years and the modern human skeleton has

changed throughout this time due to the environment and random genetic drift, a concept

referred to as secular change [16,17]. Due to this morphological change over time, it is important

to develop new methods and research on contemporary collections of skeletal remains [18]. A

recent study showed that secular change can happen relatively quickly, and can affect traits on

the pelvis that are regularly used in forensic anthropology [16]. Samples from the Hamaan-Todd

Osteological Collection and the William M. Bass Donated Skeletal Collection were compared to

denote any changes in the ischiopubic ramus, the subpubic angle, and the ventral arc of the pelvis

[16]. The Hamaan-Todd Collection includes individuals that were born between the mid-

nineteenth century and the early twentieth century, while the William M. Bass Collection

includes individuals born since 1940 [16]. The results showed significant shape differences in all

three traits for females, and the ventral arc for males [16]. The sex determination methods which

utilize these traits still offer high accuracy determinations for both populations, but with such

significant shape changes occurring over time, it is unknown how long changes will continue to

not affect the methods used [16]. This study demonstrates the need for continued contemporary

skeletal collections. Due to the locomotor demands of the lower limbs, it is unlikely that any

significant asymmetry exists between the right and left pubic bones; however, crossed symmetry

has been observed in the past, and it will be important to note whether any significant

asymmetrical patterns indicate that a right or left pubic bone would be more accurate in

determining male or female [17].

Along with this, it is also important to develop new methods using a collection that

derives from the ancestry affiliation for which it will be used in the future [15]. This is not

always possible, especially when modern admixture and self-identification are considered, but it

is always important to record ancestry when available and to notice any trends or correlations

that may arise in order to ensure that data derived from these methods is used most appropriately

and accurately in the future.

The goal of the present research is to provide a refined method using a 3D Microscribe

digitizer to assess sex from the pubic bone, as well as to compile a list of eight standard

landmarks that can be used in future geometric morphometric studies of the pubic bone and

pelvis. This research refines similar methods used in two previous studies which utilized a 3D

Microscribe Digitizer to estimate sex from landmarks located across the entire os coxa [6,20].

This research also noted patterns in morphological change related to age, ancestry, and parity in

females, all of which likely play a role in pelvis morphology, but to what extent is unclear. The

information gained from this research is applicable to modern forensic cases as well as other

various scenarios, such as mass or commingled graves, where it would be advantageous to

understand the demographics of the individuals involved, even if identifications may not be

possible.

The initial hypothesis of this research was that this method of 3D geometric

morphometric analysis will formulate a statistically accurate method of sex determination based

on the shape of the human pubic bone.

Materials and Methods

The main method of data collection for this research was done utilizing a Microscribe

Digitizer G2X. The sample bone is held in a small, rubber clamp, which is held by a heavy steel

vice to ensure the bone does not move during data collection. Due to the nature of recording

points in three-dimensional space, if the bone moves during data collection, all points must be re-

recorded. It is important to ensure the digitizer is calibrated and working correctly by pressing

the “home” button on the digitizer and testing the origin point by measuring a known distance in

millimeters or centimeters (setting determined by researcher) [21]. This data point is

automatically entered into an Excel Spreadsheet open on the computer. Once data collection is

finished, the data can be imported into other software, such as Past, Python, or MorphoJ [22, 23].

The subsequent analyses were performed using Python.

The University of New Mexico Maxwell Documented Skeletal

Collection

The larger sample for this research consists of individuals from the University of New

Mexico Maxwell Museum’s Documented Skeletal Collection. This collection was established in

1984 and now has over 300 individuals of both sexes, varying ages, and many population groups.

The known information for most individuals within this collection includes sex, age-at-death,

population affinity, and cause of death. All donations after 1995 have been asked to provide past

health information as well as occupation information so that more research using the collection

can be conducted. This is one of the best modern collections in the western United States and is

particularly important to forensic anthropology because it consists entirely of individuals who

have passed away within the last 50 years. The sample as a whole is made up of 60% males,

most of the males and females are aged 51-75, and 80% of the total sample identified as White

[24].

The Maxwell Documented Skeletal Collection has been chosen as the primary sample for

this research for multiple reasons. First, it is one of the most modern skeletal collections in the

United States. As discussed earlier, it is vital that methods which are to be applied to modern

human remains are developed from known modern collections [15]. Second, the Maxwell

Documented Collection is identified as 80% White, which allows for a large collection of similar

ancestry. This research utilized 213 of the 307 available individuals, 133 males and 80 females;

each individual had to be older than 18 years at time of death to avoid the indeterminate

morphology of juveniles; the oldest individual in the sample was 101. Each individual had to

offer at least one intact pubic bone. A total of 378 pubic bones were recorded, as not all

individuals offered intact right and left bones. Both the right and left bones were collected to

ensure the method was viable for use on both pubic bones, considering an anthropologist may

not have the luxury of choosing between the two when presented with a fragmented skeleton.

The right and left bones were not compared to each other (other than a brief check for significant

asymmetry) because orientation is one of the variables of information removed before a

geometric morphometric analysis can begin, so they can all be analyzed together.

Table 1. Demographic information

Number of

Individuals

Parous

Non-

Parous

Age

Group

18-35

Age

Group

36-50

Age

Group

51-60

Age

Group

60+

White

Ancestry

Hispanic

Ancestry

Black

Ancestry

Male

133

N/A

112

Female

Total

Number of

Individuals

213

133

190

Demographic information including the number of individuals, parous and non-parous females,

age groups, and self-reported ancestry of the 213 individuals included in the sample, note that

parity and ancestry information was not available for all individuals.

It is important to recognize that “White” is an ambiguous term when it comes to

determining ancestry or regional belonging; “White” can mean many things, and a White sample

from New Mexico will likely be quite different from a White sample elsewhere. Although, it is at

least a similar, overall characteristic that can be used to combine and contrast future population

data. Ancestry classifications are self-reported, which means they represent which group each

individual identified with, and not necessarily where their genetic ancestors came from. In

reality, white is a color and not an ancestry, meaning that the genetic ancestry of people who

identify as “White” could vary drastically. It has also been shown that those who are multiracial,

or identify with multiple ancestries or ethnicities, often change their identification over time, and

usually choose to list a single ancestry rather than multiple ones [18]. Self-reported information

on ancestry may not be the most reliable, however it does create a group basing that can still

offer important information when compared and contrasted, so long as the general issues with

the classification are taken into account.

Landmarks

For this research, landmarks are defined as a specific point on bone that can be located on

each sample and subsequently recorded in 3-D space [19]. The landmarks used in this study

were chosen based on previous research, traditional landmark location, and some new landmarks

were chosen by the author [27–29]. New landmarks were chosen in an effort to improve or

expound on existing landmarks and overall shape analysis.

When collecting data on landmark points, it is important to distinguish which type of

landmark it is. Landmarks can be categorized as type I, II, or III [4,30]. Type I landmarks are

considered the easiest to find and consist of one single point on the bone where tissue transitions,

such as an intersection of sutures [20,24,25]. Type II landmarks are considered the points of

maximum curvature or greatest muscle attachment, an example would be the euryon on the

cranium [24,25]. Type III landmarks are the most extreme points of a structure overall,

sometimes labelled as the “posteriormost” or “anteriormost” points [20,24]. A category that has

since been added to this list by subsequent studies is constructed landmarks, which is defined as

“points corresponding to locations that are defined using a combination of traditional landmarks

and geometric information” [26]. For example, calculating and using the midpoint along a line

of maximum width as a landmark. Constructed landmarks tend to be more difficult to find by

inexperienced users due to the need for a measurement of some kind to locate the landmark.

Landmark types II and III, which use extremes to locate, tend to be consistently easier to find.

Most studies using geometric morphometrics distinguish which types of landmarks are used to

help determine which types of points are most useful or which points introduce the most error

[24–26]. Sliding semi-landmarks were not used in this study because when morphological

variation is small, such as that between modern humans, the differences between a landmark-

based study and a sliding semi-landmark based study using the same sample can be altered or

skewed based on differences in initial alignment, thus raising questions about any results

obtained [32]. There are no visible suture intersections on an adult pubic bone, so only landmark

types II and III were utilized in this research. Type III were found to be reliably consistent when

they represented both extremes of a width or height. Table 2 lists the landmarks used and the

type of landmark they represent, figure 1 shows the approximate physical location of each

landmark.

Table 2. Written landmark descriptions and locations.

Number

Landmark

Description

Type

Pubic Tubercle

Most prominent point of the pubic tubercle

Traditional

(II)

Superior Pubic

Symphysis

The most superior point of the pubic symphysis

Extremal

(III)

Inferior Pubic

Symphysis

The most inferior point of the pubic symphysis

Extremal

(III)

Lateral Border

Point on the lateral border of the pubic body which

would create the maximum breadth of the obturator

foramen

Extremal

(III)

5, 6

Pubic Body

Height

The superior (5) and inferior (6) points which

create the maximum height of the pubic body

Extremal

(III)

7, 8

Pubic Body

Width

The medial (7) and lateral (8) points which create

the maximum width of the pubic body

Extremal

(III)

Fig 1. Location of landmarks. Anterior surface of the right pubic bone displaying approximate

locations of the eight landmarks recorded on each bone.

For the purpose of geometric morphometrics, shape has been defined as the geometrical

information that is left when location, scale and rotational effects are removed from an object

[19,28]. The first statistical step of this research was to perform a Generalized Procrustes

Analysis, or GPA. Before any analysis is completed, the landmark coordinates include

information on size, shape, position and orientation [34,35]. In order to analyze just shape,

however, all of this other information must be excluded [35]. A GPA eliminates the non-shape

variation and rescales and rotates each sample so that they are relative to each other in the same

plane [4,35]. By doing this, the GPA translates all landmark configurations to the same centroid,

scales all configurations to the same centroid size, and iteratively rotates all configurations until

the summed squared distances between the landmarks and their corresponding sample average is

at a minimum [35]. After the GPA is run, all configurations are superimposed on each other and

the resulting coordinates are called Procrustes shape coordinates and only include information

about the shape of the configurations [35]. This statistical step alone does not offer a lot of

useful information, but rather it prepares the data to be analyzed further in ways that will offer

more information about the shapes that are present.

The next step was a Principal Components Analysis, or PCA. A PCA is a way to

represent the variation that is present within the sample, and the goal of a PCA is to determine

which variable introduces the most amount of variation [19,31]. The Principal components

provide a means of comparing the relative importance of each dimension of the data [19]. If the

first two eigen vectors represent 50% or more of the variance then the test can be considered

competent at finding a linear classifier that effectively separates the classes (sex).

Lastly, a Discriminant Function analysis was run through Python to determine the

predictive power of the sample based on the two groupings of males and females. The first part

of a Discriminant Function uses multivariate F tests to determine whether or not there are any

significant differences between both groups (male and female) with regard to all variables [37].

This results in finding the statistically significant means across the groups, which can then be

used to classify all samples into which of the two groups they most likely to belong to [37]. The

analysis randomly breaks the data set into two sets, placing 70% of the data in the first and the

remaining 30% of the data into the second. The first set is used as a training set to teach the

software the difference between the two classifying groups (male and female). The second set is

used as a test to predict how well the machine learned to distinguish between the two groups.

Essentially, this will result in the predictive power of the method’s ability to determine an

unknown bone as either male or female.

In order to determine ease of reproducibility, the author resampled 50 individuals from

the previously sampled 213. The author then collected data for a second time on each of the 50

individuals, N=100 pubic bones. One of the advantages of geometric morphometrics is that the

data can be collected on any plane and rescaled to perform analysis; because of this, one of the

clearest ways to asses replicability was to simply run the entire statistical process again using

only the resampled data and compare these results to the results of the larger sample. This

analysis was performed exactly as described above, using Python software.

To test the presence of interobserver error, the author collected all eight landmarks from a

single bone, once a day, six days in a row, to determine whether or not the landmarks were

consistently recorded in the same location. This data collection occurred approximately one year

after the initial data collection, and the individual bone came from the University of Montana

Forensic Anthropology Laboratory Collection, which is not a collection of known provenience.

Due to this, the data collected from this individual was not added to the previous, large sample

from the University of New Mexico Maxwell Museum. The six instances of data collection were

treated as their own data set, and a GPA and PCA were run using MorphoJ. The resulting PCA

scatter plot, presented below, displays how much variance was present between each collection

instance.

Results

The first two eigenvectors of the PCA test represented 43% of the total variance, which

was close to, but not exceeding the ideal 50% for class separability. The resulting PCA scatter

plot displaying principal components one and two can be seen in Figure 2. The third principal

component represented 11.4% of the total variance, but was not mapped because it did not add to

the visualization of the data. The scatter plot was also coded to show males and females, age

groupings, parity of females, and ancestry groups (see Figures 2-4). While it does not in itself

offer definitive class separability, it is clear from Figure 2 that the males and females do show

some difference in how they cluster, and the male specimens cluster much more tightly than the

females. It should be noted that the only African American female individual in the sample is

represented by the two uppermost outliers of the scatter plot (right and left pubic bones). This

individual was not removed from the subsequent analysis because she provided two complete

pubic bones with valid measurements and was in line with the higher variance of females. While

a sample size of one is not nearly enough to confirm the necessity of an ancestry-specific

method, this does indicate that future research should use larger and more diverse samples to

determine whether or not ancestry is a defining variable.

Fig 2. PCA scatter plot. Principal Components Analysis scatter plot displaying male and female

groups.

Fig 3. PCA age scatter plot. Principal Components Analysis scatter plot displaying age groups

(males and females grouped together); group 1 (18-35), group 2 (36-50), group 3 (51-60), group

4 (61+).

Fig 4. PCA parity scatter plot. Principal Components Analysis scatter plot displaying available

data on parity of females (parous vs. non-parous).

By coloring the scatter plot based on age groupings, it is clear that as the sample groups

get older, the variance becomes greater (Figure 3). The younger age groups (groups 1 and 2)

cluster much tighter than groups 3 and 4. By coloring only the females with available data

concerning their parity, it can be seen that parous females exhibit a slightly greater variance than

non-parous females (Figure 4). Unfortunately, this data was not available for every female

individual; however, this data indicates the need for future research, especially considering how

little is known about the shape changes a female pelvis experiences after giving birth.

The Discriminant function analysis was run through Python software. The training

discriminant function analysis resulted in 96.2% accuracy between male and female predicted

group classifications. The testing set discriminate function analysis resulted in 95.5% accuracy

between male and female groups with a P-value of <0.0001. Figure 5 shows the bar graph of the

testing set discriminant function analysis.

Fig 5. Discriminant function bar graph. Results of the discriminant function test set. The x-

axis displays the degree of maximum separability, the y-axis displays the frequency.

This Discriminant Function analysis was validated through a second software program,

MorphoJ, by running the same tests (GPA, PCA, DFA). The testing set resulted in 96.3%

accuracy, and the test set resulted in 95.5% accuracy (P<0.0001).

The last analysis performed on this large data set, which consisted of running the GPA,

PCA, and discriminant function analysis on the second set of resampled data, N=100, resulted in

a training score of 96.5% accuracy, and a cross-validation score of 93.8% accuracy, P<0.0001.

The interobserver error test ran a GPA and PCA on a smaller data set which consisted of

six instances of landmark collection from the same bone. This resulted in a PCA scatter plot,

figure 6, which shows incredibly small amounts of variation between each instance of data

collection. All six instances of landmark collection graph on top of each other, clearly

representing that each data collection instance occurred on the same individual. The first two

principal components contained 78% of the variation, this indicates a successful PCA and does

not necessitate graphing of the remaining principal components due to their small amount of

represented variation.

Fig. 6 Results of the interobserver error PCA analysis. This chart shows the PC scores (X

axis displays PC 1; Y axis displays PC 2) of the six instances of data collection from the same

bone, by the same author, once a day, six days in a row.

No significant asymmetry was observed between the right and left pubic bones of

individuals. Most individuals’ right and left bones were relatively mirrored when displayed on

the PCA scatter plot. The mirroring is not exact; however, asymmetrical shape differences are

likely due to shape changes related to aging rather than sex differences. No further analysis was

conducted concerning asymmetry because it did not seem to affect the results of the research in

any significant way; this indicates that this method can be just as effective when either the right

or left pubic bone is used. The PCA scatter plot with each individual labeled can be viewed in

Appendix A.

Discussion

The main goal of this research was to test whether or not the method of digitizing

landmarks on the human pubic bone could be used to accurately distinguish between male and

female individuals. With a result of 95.5% accuracy and a p-value of <0.0001 from the

discriminant function, this method has been successful on this sample. Experienced forensic

anthropologists using morphoscopic methods can accurately determine sex approximately 90-

95% of the time, depending on the state of the bones and the experience of the user [6,20].

Similar studies, which collected landmarks across the entire os coxa reported similarly high

results, so it is promising to see that the method can be narrowed down to a smaller portion of

bone and still retain high accuracy [6,20]. These results indicate that this method could

potentially be used in forensic casework to aid in the identification of unknown human remains.

It could also be useful in instances of mass graves, commingled graves, or fragmentary remains

when portions of the pubic bones are recovered. This study helps to fill the existing gap of

metrically assessed sex determination methods in forensic anthropology and introduces an

applicable method of geometric morphometrics to the field.

Supporting forensic anthropological determinations in the court room with metric

analysis is becoming more important as more and more scientific expertise is used to support

criminal cases. With a 95.5% accuracy rate, the present method has great potential as an aid in

supporting a sex determination of an unknown individual. The validation of the discriminant

function analysis results using MorphoJ software indicates that this type of analysis is

consistently accurate when using various brands of software. This demonstrates that the data is

robust enough to exhibit the same pattern when analyzed in distinct software programs, which

also means it could be available to more future researchers. It is also quite promising that the

resampled data resulted in significant accuracy rates quite close to the larger sample percentages.

The resampled data pool was much smaller than the overall data set, which may explain some of

the difference. It is clear though that with experience using a Microscribe Digitizer, this method

should be replicable. The interobserver error test also shows that data collection can be

incredibly consistent and that using type II and type III landmarks did not create replication

issues. Future research should determine how long it takes for a user to become proficient at

collecting data using a digitizer. A forensic anthropologist should use multiple methods to

support their determinations and there are very few metric options available for sex

determination. This method offers a very accurate, metric determination which could be added to

the anthropologist’s traditional analysis. It is also vital that this method was developed using a

sample which only contains individuals that have died in the last 50 years because it is more

applicable on modern populations and there is less concern of secular changes biasing the results

when used on contemporary individuals.

Limitations of this research demographic sample size and applicability issues. While

large, the sample for this research includes many more males than females, which could be

improved in future research. The sample also consists almost entirely of White ancestry

individuals, so it is unknown whether ancestry differences introduce enough shape variation to

alter the effectiveness of this method. Within this sample, it is clear that the single African

American individual is an outlier, however, it is not known if this is due to her ancestry or some

unrelated reason. The same can be said for the effects of parity because this sample had

relatively few known instances of parity information. This method also requires that the pubic

bone be nearly complete for analysis, which is an unrealistic expectation in modern forensic

casework. It is suggested that future research test this method on larger samples of varying

ancestry, and on fragmented bones which may be missing landmarks, to improve the accuracy

and usefulness of this technique.

The data and code that support the findings of this study are available within the public

repository GitHub, and can be found at https://github.com/brandonbridge/GeoMorpho [36].

Acknowledgments

Thank you to the University of New Mexico Maxwell Museum for support and access to the

Documented Collection; and thank you to all the individuals and families of individuals who

donated their remains to science in order to make this research possible. Thank you to the

University of Montana Anthropology Department and the Office of Sponsored Programs for

publication fee funding.

References

1. Spradley K, Jantz RL. Ancestry Estimation in Forensic Anthropology: Geometric

Morphometric versus Standard and Nonstandard Interlandmark Distances. Journal of

Forensic Sciences. 2016 Jul;61(4):892–7.

2. Lynch JJ, Cross P, Heaton V. Sexual Dimorphism of the First Rib: A Comparative Approach

Using Metric and Geometric Morphometric Analyses. Journal of Forensic Sciences. 2017

Feb;62(5):1251–8.

3. Franklin D, Oxnard CE, O’Higgins P, Dadour I. Sexual Dimorphism in the Subadult

Mandible: Quantification Using Geometric Morphometrics. Journal of Forensic Sciences.

2007 Jan;52(1):6–10.

4. Zelditch ML, Swiderski DL, Sheets HD, Fink WL. Geometric Morphometrics for Biologists.

First. San Diego, California: Elsevier Academic Press; 2004.

5. Anastasiou E, Chamberlain AT. The Sexual Dimorphism of the Sacro-Iliac Joint: An

Investigation Using Geometric Morphometric Techniques. Journal of Forensic Sciences.

2013 Jan;58:S126–34.

6. Candelas González N, Rascón Pérez J, Chamero B, Cambra‐Moo O, González Martín A.

Geometric morphometrics reveals restrictions on the shape of the female os coxae. J Anat.

2017 Jan;230(1):66–74.

7. Abdel Fatah EE, Shirley NR, Jantz RL, Mahfouz MR. Improving Sex Estimation from

Crania Using a Novel Three-dimensional Quantitative Method , ,. Journal of Forensic

Sciences. 2014 May;59(3):590–600.

8. Jung H, Woo EJ. Evaluation of Mastoid Process as Sex Indicator in Modern White

Americans using Geometric Morphometrics. J Forensic Sci. 2016 Jul 1;61(4):1029–33.

9. Ammer S, Coelho J d’Oliveira, Cunha EM. Outline Shape Analysis on the Trochlear

Constriction and Olecranon Fossa of the Humerus: Insights for Sex Estimation and a New

Computational Tool,. Journal of Forensic Sciences. 2019;64(6):1788–95.

10. Spradley K, Jantz RL. Sex Estimation in Forensic Anthropology: Skull Versus Postcranial

Elements. Journal of Forensic Sciences. 2011 Mar;56(2):289–96.

11. Berg GE. Pubic Bone Age Estimation in Adult Women*. Journal of Forensic Sciences. 2008

May 1;53(3):569–77.

12. Wink AE. Pubic Symphyseal Age Estimation from Three-Dimensional Reconstructions of

Pelvic CT Scans of Live Individuals. J Forensic Sci. 2014 May 1;59(3):696–702.

13. Murail P, Bruzek J, Houët F, Cunha E. DSP: A tool for probabilistic sex diagnosis using

worldwide variability in hip-bone measurements. Bulletins et mémoires de la Société

d’Anthropologie de Paris. 2005 Dec 1;(17 (3-4)):167–76.

14. Lesciotto KM. The Impact of Daubert on the Admissibility of Forensic Anthropology Expert

Testimony. Journal of Forensic Sciences. 2015 May;60(3):549–55.

15. Komar DA, Grivas C. Manufactured populations: What do contemporary reference skeletal

collections represent? A comparative study using the Maxwell Museum documented

collection. American Journal of Physical Anthropology. 137(2):224–33.

16. Klales AR. Secular Change in Morphological Pelvic Traits used for Sex Estimation. Journal

of Forensic Sciences. 2016 Mar 1;61(2):295–301.

17. Auerbach BM, Ruff CB. Limb bone bilateral asymmetry: variability and commonality

among modern humans. Journal of Human Evolution. 2006 Feb 1;50(2):203–18.

18. Stringer C. Lone Survivors. New York: Henry Holt and Company, LLC.; 2012.

19. Eliopoulos C, Lagia A, Manolis S. A modern, documented human skeletal collection from

Greece. HOMO - Journal of Comparative Human Biology. 2007 Sep;58(3):221–8.

20. Bytheway JA, Ross AH. A Geometric Morphometric Approach to Sex Determination of the

Human Adult Os Coxa. Journal of Forensic Sciences. 2010 Mar 25;55(4):859–64.

21. Sanchez A. Microscribe and Craniometrics. Harris County Institute of Forensic Sciences.

2018;80.

22. Welcome to Python.org [Internet]. Python.org. [cited 2020 Feb 18]. Available from:

https://www.python.org/

23. Klingenberg CP. MorphoJ: an integrated software package for geometric morphometrics.

Molecular Ecology Resources. 2011;11:353–7.

24. UNM Maxwell Museum of Anthropology | Documented Collection [Internet]. osteolab.

2018 [cited 2018 Jul 9]. Available from: http://osteolab5.wixsite.com/osteolab/documented-

collection

25. Doyle JM, Kao G. Are Racial Identities of Multiracials Stable? Changing Self-Identification

Among Single and Multiple Race Individuals. Soc Psychol Q. 2007 Dec 1;70(4):405–23.

26. Stegmann MB, Gomez DD. A brief introduction to statistical shape analysis. Informatics and

mathematical modelling, Technical University of Denmark, DTU. 2002;15(11).

27. Costello A. An analysis of sexual dimorphism using geometric morphometrics of the femur

and tibia: The use of GM in assessing sex of fragmented remains [Internet]. Boston

University School of Medicine; 2015 [cited 2017 Mar 13]. Available from:

https://open.bu.edu/handle/2144/16027

28. San-Millán M, Rissech C, Turbón D. Shape variability of the adult human acetabulum and

acetabular fossa related to sex and age by geometric morphometrics. Implications for adult

age estimation. Forensic Science International. 2017 Mar;272:50–63.

29. White TD, Folkens PA. The Human Bone Manual. 1 edition. Burlington: MA: Academic

Press; 2005. 488.

30. Toon C. Sexual dimorphism at the proximal tibia: a geometric morphometric analysis

[Internet]. 2014 [cited 2017 Mar 13]. Available from: https://open.bu.edu/handle/2144/15329

31. Perlaza NA. Sex Determination from the Frontal Bone: A Geometric Morphometric Study.

Journal of Forensic Sciences. 2014 Sep;59(5):1330–2.

32. Perez SI, Bernal V, Gonzalez PN. Differences between sliding semi-landmark methods in

geometric morphometrics, with an application to human craniofacial and dental variation.

Journal of anatomy. 2006;208(6):769–784.

33. Kendall DG. A survey of the statistical theory of shape. Statistical Science. 1989;87–99.

34. Bookstein FL. Landmark Methods for Forms without Landmarks: Morphometrics of Group

Differences in Outline Shape. Medical Image Analysis. 1996;1(3):225–43.

35. Mitteroecker P, Gunz P, Windhager S, Schaefer K. A brief review of shape, form, and

allometry in geometric morphometrics, with applications to human facial morphology.

Hystrix, the Italian Journal of Mammalogy. 2013;24(1):59–66.

36. Shlens J. A tutorial on principal component analysis. 2014.

37. Poulsen J, French A. Discriminant function analysis. Retrieved from. 2008.

38. Bridge B. brandonbridge/GeoMorpho [Internet]. 2020 [cited 2020 Apr 1]. (GeoMorpho).

Available from: https://github.com/brandonbridge/GeoMor

16a

16b

17a

17b

19a

19b

20a

20b

28a

28b

31a

31b

32a

32b

33a

33b

36b

38a

38b

39a

39b

42a

42b

43a

43b

44a

44b

47a

47b

52a

52b

53a

53b

55a

55b

56a

56b

58a

58b

63a

63b

65a

65b

66a

66b

67a

67b

68a

68b

69b

72a 72b

80a

80b

81a

81b

83a

83b

84a

84b

85a

85b

87a

87b

89a

94a

94b

100a

100b

102a

102b

106b

108a

108b 112a

112b

113a

113b

118a

118b

119a

120a

120b

123a

123b 124a

126a

126b

128b

137a

137b

143a

143b

145a

145b

147a

151a

151b

153a

153b

154a

154b

157a

157b

162a

162b

163a

163b

164a

164b 170a

170b 174a

174b

176a

176b

177a

177b

178a

178b

179a

180a

180b

183a

183b

188a

188b

189a 189b 190a

190b

191a

191b

193a

193b

195a

195b

207a

207b

211a

211b

215a

215b

222a

222b

223a

224a

224b

226a

226b

227a 227b

228a

228b

229a 229b

232a

232b

234a

234b

237a

237b

238a

238b

240a

240b

242a 242b

243a

243b

244a

244b

245a

245b

246a

246b

247a

247b

252a 252b

254a

254b

256a

256b

257a

257b

260b

262a262b

265a

265b 266a

266b

268a

268b

274a

274b

279a

279b 280a

280b

281a 281b 282a

282b

287a 287b

290a

290b

292a

292b

293a

293b

294a

294b 296a

296b 298a

298b

301a

301b

302a

302b 303a

303b

306a

306b

307a

307b

21a

21b

37a

37b

46a

46b

48a

48b

49a

49b

51a

51b

62a

62b

74a

74b

77b

86a

93a

132b

134a

134b

140a

140b

142a

142b

156a

156b

158a

158b

159a

159b

160a

160b

168a

168b

172a

173b

175a

181b

184a

184b

185a

185b

186a

186b

187a

187b

192a

192b

194a

194b

196a

196b

198a

198b

201a

201b

202a

202b

204a

204b

208a

208b

210a

210b

212a

212b

213a

213b

218a

218b

220b

230a

230b

231a

241a

241b

248b

251a

251b

253a

253b

255a

255b

258a

258b

259a

259b

261a

261b

263a263b

264a

264b

267a

267b

269a

269b

270a

270b

271a

271b 272a

272b

273a

273b

275a

275b

277a

277b

278a

278b

283a

283b

284a

284b

285a

285b

286a

286b

288a

288b

291a

291b

295a

295b

297a

297b

299a

299b

300a

300b

304a

304b

305a

305b 308a

308b

309a

309b

Males

Females

Appendix A: PCA scatter plot displaying labeled individuals (a indicates the right bone, b indicates the left bone)

Chapter Six: Modeled Fragmentary Analysis using a Geometric Morphometric Sex

Determination Method of the Human Pubic Bone

Submitted to the Journal of Forensic Sciences

Katherine Baca, MA

University of Montana

Brandon Bridge, PhD

University of Montana

Meradeth Snow, PhD

University of Montana

Acknowledgements: Thank you to the University of New Mexico Maxwell Museum for their

support and access to their Documented Skeletal Collection; thank you to all of the individuals

and their families who have donated their remains to science to make research such as this

possible.

ABSTRACT

Sex determination of the human pelvis has traditionally been done through visual

analyses of morphoscopic traits and there are limited metric methods available to forensic

anthropologists to add metric credibility to these analyses. The goal of this research was to create

a new metric method using three dimensional geometric morphometrics to determine sex from a

fragmented pubic bone. The sample consisted of n=378 pubic bones from the University of New

Mexico’s Maxwell Museum Documented Skeletal Collection and eight landmarks were collected

from each bone. Statistical analyses and machine learning algorithms were used to mimic

fragmented remains that included tests run on each possible landmark combination of three or

more landmarks to simulate fragmented bones (218 combinations). The results of the modeled

fragmentary analysis consisted of 133 combinations which exhibit a 90% or higher accuracy in

sex prediction; and nine combinations which exhibit 95.5% accuracy in sex prediction. In

particular, three landmarks clustered around the ventral arc of the pubic bone performed the best,

indicating this is the most sexually dimorphic portion of the bone. These results indicate that

three-dimensional geometric morphometrics is a valid method to be applied to sex determination

in forensic anthropology.

KEYWORDS: Forensic Anthropology, Sex Determination, Geometric Morphometrics, Pubic

Bone, Fragmentary, Geomorph.

Sex determination using the pelvis has traditionally been done using a series of

morphoscopic traits, including the overall shape of the pelvic inlet, the ventral arc, the

ischiopubic ramus, subpubic concavity, and the greater sciatic notch (1). It is well accepted that

the pelvis is the best element to use to determine sex when available (2,3). An experienced

biological anthropologist can generally estimate sex from a visual analysis of the pelvis correctly

approximately 90-95% of the time (1–5). The pelvis exhibits a series of sexually dimorphic

differences as an individual matures, including a wider pelvic inlet and sub-pubic arch for

females, and a narrower greater sciatic notch and pubic bone for males, most of which are due to

the physiological ability of females to give birth (1,2,4–6). Very few accepted metric analyses of

these shape differences in the pelvis exist, but metric analysis is becoming more important as it

becomes more common for forensic anthropologists to testify as experts within the court system

(2,7).

Geometric morphometrics is an analysis of shape and can be performed in both two-

dimensional and three-dimensional planes (8). This research applies 3-D geometric

morphometrics to the problem of metrically determining sex from the human pelvis. The basic

process of this specific method consists of recording multiple homologous shapes (multiple

pubic bones), rotating them all onto the same plane, removing size as a variable (Generalized

Procrustes Analysis), distinguishing differences between group clusters (Principal Components

Analysis), and then performing a discriminant function analysis to determine the predictive

power of the method (8). This method is advantageous because it offers more metric information

about the specimen than traditional visual or interlandmark distance measurements do.

Traditional visual analyses rely on the expertise of the forensic anthropologist to judge the size

and shape of the bone by simply examining it. Whereas completing a statistical analysis on

multiple points in space (landmarks on bone) offers measurements between the x, y, and z

coordinates of each landmark and a much more complete and exact analysis of the bone as a

whole. Geometric morphometrics is commonly utilized in biological studies, however, is fairly

new to the field of anthropology (9–11). Existing research in biological anthropology mainly

utilizes the method for cranial ancestry estimation and age-at-death estimation; there are few

metric methods available for sex determination from the pelvis (12,13).

It is rare to recover complete sets of skeletal remains in most forensic and archaeological

scenarios (2). Often, remains have been buried or exposed to the elements for some time before

recovery, which results in broken, partially disintegrated, and/or incomplete bones available for

analysis. Due to this commonality, it is necessary to develop methods of sex estimation that can

be used on fragmentary remains. The hypothesis of this research is that a geometric

morphometric shape analysis of the pubic bone will result in highly statistically accurate sex

determinations on both whole and fragmented human pubic bones and that certain landmarks

will be more effective than others in establishing a sex determination.

Adding to the complicated nature of fragmentary remains is the knowledge that metric

analyses are more trusted by court-systems and juries than are subjective morphoscopic analyses

(7). An expert should use multiple methods to validate their determinations, and a precise metric

method to validate a sex determination could aid in establishing the credibility of the testimony.

This method could also be useful in other contexts, such as mass graves, commingled and

fragmentary remains, and bioarchaeological studies. In archaeological contexts when multiple

individuals are identified, one of the first demographics determined is how many males and

females are present; this method could easily offer a metric answer to this question.

Materials/Methods

In order to perform a three-dimensional geometric morphometric analysis of the pubic

bone, a Microscribe digitizer was utilized to collect a set of landmarks from each specimen. The

bone must be held stationary during data collection; if it were to move while collecting data

points, the collection for that specimen would have to be deleted and started again. The author

used a rubber clamp, bolted to the table, to hold the bone in place. The rubber clamp was able to

hold the bone still without damaging the bone. The bone could be placed into the clamp in any

manner as long as the landmarks were accessible to reach with the digitizer. Before data

collection, the digitizer was calibrated by pressing the “home” button and checking a known

measurement in millimeters (14). The stylus of the digitizer was placed at the point of the

landmark and a foot pedal was pressed to record that location; the x, y, z coordinate was then

entered into a spreadsheet which can be uploaded to any statistical software. Eight landmarks

were collected on each specimen when available, only specimens that were missing three or

fewer landmarks were utilized (missing landmarks were recorded as -9999 to differentiate them

from existing landmarks). See Table 1 for the description of the landmark locations; it was

necessary to always record landmarks in the same order (one through eight) so the computer

knew which landmark it was recording. The landmarks were chosen by the author based on

previous research and a pilot project which aided in narrowing down landmarks that were easily

identified and replicable (2). 213 individuals were utilized from the University of New Mexico’s

Maxwell Documented Skeletal Collection, which resulted in 378 total pubic bones. The Maxwell

Documented Collection consists only of donated individuals who have passed away in the past

50 years, making it one of the largest modern collections in the US. The collection houses over

300 individuals, 60% of which are male, the majority of adults are aged 51-75, and the sample as

a whole is self-identified as 80% White (15). All 213 individuals utilized were 18 or older to

avoid the indeterminate morphology present in juvenile pelves, and when available, both right

and left pubic bones were recorded.

Number

Landmark

Description

Pubic Tubercle

Most prominent point of the pubic tubercle

Superior Pubic

Symphysis

The most superior point of the pubic symphysis

Inferior Pubic

Symphysis

The most inferior point of the pubic symphysis

Lateral Border

Point on the lateral border of the pubic body which would create

the maximum breadth of the obturator foramen

5, 6

Pubic Body

Height

The superior (5) and inferior (6) points which create the maximum

height of the pubic body

7, 8

Pubic Body

Width

The medial (7) and lateral (8) points which create the maximum

width of the pubic body

Table 1: Description of the location and landmark type of the eight landmarks recorded on each

bone.

The statistical analysis began with transferring the raw data from the spreadsheets into

Python 3 (16). Before the data is manipulated in any way, it contains information on size,

position, and orientation; all of which must be removed in order to analyze shape alone (17,18).

A Generalized Procrustes Analysis was run to transform the data so that it is scaled and rotated

to the same plane, leaving only the shape information (8). Next, a Principal Components

Analysis was run to isolate which variables were providing the most variation within the sample

(19). The analysis first performs an eigen-decomposition on the covariance matrix of the

Procrustes shape coordinates and then sorts the eigenvalues and eigenvectors. The top two

eigenvectors exhibit the greatest variance in the data and are known as the first two principal

components (19). The component scores of each observation were then plotted on the first two

principal component axes to form a scatter plot to visualize the variation present within the data.

This visualization displays groupings, clusters, and outliers. This scatter plot indicated that it was

possible to distinguish between males and female groups based on all data points being present.

The bulk of the statistical analysis consisted of a series of discriminant function analyses

to determine the predictive power of the sample when classifying into two groups; males and

females. A discriminant function analysis in Python 3 is performed as a supervised machine

learning algorithm. “Supervised” in this case means the input data (the Procrustes shape

coordinates) are labeled male or female, and the program learns to predict the sex classification

from this input data. First, the data is randomly split into two sets, the training set and the testing

set. The training set in this study was 70 percent of the sample, with the remaining 30 percent

reserved as the testing set. The training set is used to train the prediction model while the testing

set is left out. The trained prediction model is then applied to the testing set to validate its

predictive accuracy. The training score reported in the model results indicates how well the

model predicts the classes of the samples in the training set, while the testing score reports the

model’s predictive accuracy on the data samples previously left out.

The discriminant function analysis is a dimensionality reduction technique similar to the

principal component analysis. The primary difference between the two techniques is that a

principal component analysis is an “unsupervised” attempt to project the samples onto a

subspace whose axes maximize the variance in the data, while the “supervised” discriminant

function analysis attempts to project the samples onto a subspace whose axes maximize the

separability between classes (20). The end result is the predictive power of the method’s ability

to determine whether an unknown bone originated from a male or a female individual. The first

discriminant function was run on the entire sample to establish a baseline for how well the

100

method worked when presented with whole, non-fragmented bones. To begin the simulated

fragmentary analyses, a discriminant function analysis was run without landmarks 1, 2, and 5,

which were the landmarks most commonly missing from the UNM sample. Next, in order to

determine which combinations of landmarks offer the best predictive power, every possible

combination of three landmarks or more was then run through a discriminant function test.

After the initial data collection, the author used a random number generator to create a

list of 50 individuals from the original data set, all of which provided both right and left pubic

bones. This resulted in a second sample of N=100 pubic bones, which the author collected data

from a second time, in order to test the replicability of this method. A GPA, PCA, and

discriminant function analysis was run on this second data set to determine whether or not the

method was replicated accurately by the researcher. It was also important to test interobserver

error to ensure the author could consistently identify the correct landmark locations.

Approximately one year after the initial data collection, the author collected the same eight

landmarks on one bone from the University of Montana Forensic Anthropological Laboratory

once a day, six days in a row. This resulted in a small data set of six instances of data collections

all representing the same bone. A GPA and PCA was run on this data set to determine how

consistent the author was when collecting landmarks on the same bone.

Results

The first discriminant function test, which included all available landmarks for all 378

specimens resulted in 96.2% accuracy based on the training data, and 95.5% accuracy based on

the testing data, P<.0001. These results are on par with other sex determination methods on the

pelvis, which as mentioned earlier range from approximately 90-95% (1–5). The second

discriminant function test removed landmarks 1, 2, and 5 from the samples and resulted in a

training set accuracy of 93.5% and a testing data set accuracy of 91.7%, P<.0001.

101

The larger discriminant function loop which tested all possible combinations of three

landmarks or more offered results for 218 different combinations. The least effective

combination consisted of landmarks 1, 4, and 8 and resulted in a training score of 86.9% and a

testing score of 77.6% accuracy. Nine different combinations all resulted in the highest testing

score of 95.5% accuracy, see table 2 for which landmarks were included in each of these

fragments. It should also be noted that 133 of the 218 (61%) possible combinations resulted in

testing scores of 90% or higher. Results for all 218 landmark combinations can be viewed in the

supplemental information.

Combination Number

Training Score

Testing Score

Landmarks Included

95%

95.5%

2, 3, 6

95.7%

95.5%

3, 6, 7

96.1%

95.5%

2, 3, 4, 6

95.4%

95.5%

2, 3, 5, 6

115

96.9%

95.5%

3, 4, 6, 7

121

95.7%

95.5%

3, 6, 7, 8

172

94.6%

95.5%

2, 4, 5, 6

177

96.5%

95.5%

3, 4, 5, 6, 7

181

95.4%

95.5%

3, 5, 6, 7, 8

Table 2: The nine landmark combinations which resulted in the highest testing scores, showing

their corresponding training scores and which landmarks were included in each combination.

The second data set which consisted of 50 individuals, n=100 pubic bones, to test

replicability of the method, resulted in a training set accuracy of 96.5%, and a testing set

accuracy of 93.8% accuracy. The last analysis based on the data set which consisted of six

102

instances of data collection from the same bone resulted in very low variance between instances

of landmark collection. Figure 1 displays a scatter plot graphing the PC1 and PC2 scores of each

of the six data collection instances and shows how close each instance of data collection is to

each other. This demonstrates a successful second use of the method on a separate, albeit

smaller, data set, as well as low intra-observer error when repeatedly tested on the same bone.

Figure 1: Results of the interobserver error PCA analysis. This chart shows the PC scores (X

axis displays PC 1; Y axis displays PC 2) of the six instances of data collection from the same

bone, by the same author, once a day, six days in a row.

Discussion

The results reported here clearly indicate that this three-dimensional geometric

morphometric method of sex determination has the potential to increase the accuracy and

credibility of sex estimations on both whole and fragmented human pubic bones. More than half

(61%) of all possible combinations of these landmarks result in an accuracy rate of 90% or

higher, which is in the same range as previously used visual analysis methods (1–5). This result

indicates that this method could be used on many differently sized and broken fragments of

pubic bone to gain an accurate estimation of sex. A single discriminant function was run without

103

landmarks one, two, and five, because those were the landmarks most commonly missing from

the UNM samples and easily replicated a real fragmentary context. A result of 91.7% is a

promising indication that this method worked well in an actual fragmentary context. Finally, nine

simulated fragments resulted in an accuracy result of 95.5%, which is just as high as the overall

accuracy of the method when all eight landmarks are included in the discriminant function

analysis.

When considering the nine combinations of landmarks that all resulted in 95.5%

accuracy, there are two specific results within that appear quite significant to the applicability of

this method. First, combination 44 utilized only landmarks three, six, and seven, which are the

inferior pubic symphysis, the inferior point which creates the maximum height of the pubic body,

and the medial point which creates the maximum width of the pubic body, relatively (see Figure

2). All three landmarks are clustered around the general area of the ventral arc. This indicates

that the ventral arc, which is already used in visual analysis sexing methods, is the most sexually

dimorphic area of the pubic bone (1,4). Secondly, landmarks three and six are utilized in eight of

the nine combinations with the highest accuracy results; the one combination that does not

include both three and six, does include six (combination 172). Figure 3 displays a heat map of

the landmarks’ predictive powers. To create this map, the average predictive power of each

combination containing landmarks one and two was calculated; the same for each combination

containing landmark one and three, one and four, and so on with every two-landmark duo. The

darker the color, the higher the average predictive power for combinations including that

landmark duo. It is clear that combinations including landmarks three and six have the highest

average predictive powers. The variance in the average of the predictive powers is not large

simply due to the high accuracy of the data as a whole. Landmarks three and six appear to be the

104

landmarks that offer the most shape information in relation to male or female individuals and are

most effective in predicting sex. This is quite promising to the real-world applicability of this

method because it indicates that very high results could still be obtained even if the forensic

anthropologist is only presented with a small piece of the inferior pubic bone. At this point, the

expert could perform both a visual analysis on the ventral arc, as well as a metric analysis using

this method to make their overall sex determination.

Figure 2: The locations of landmarks three, six, and seven; notice that they are clustered around

the ventral arc area of the pubic bone.

105

Figure 3: Heatmap displaying the average predictive powers of combinations including each

landmark due; note that landmarks three and six make consistently more powerful combinations

than other landmarks.

A limitation of this simulated fragmentary research is that not all of the high scoring

landmark combinations reflect real possible fragmented bones. For example, combination 77

which includes landmarks three, four, five, six, and seven, requires that essentially all sides of

the bone are present, which is highly unlikely if the bone is fragmented. This is the case with

most of the combinations which require more than four landmarks. Multiple combinations which

only require three or four landmarks to result in accuracy levels just as high as using all eight

landmarks are therefore an important finding. Combinations 24, 44, and 115 are particularly

applicable when it comes to simulating actual broken bone because each combination uses

landmarks that are relatively close together. A second limitation is that this research was

conducted using a sample of primarily white individuals, further research into ancestry specific

biases is needed. An early Principal Components Analysis exhibited the only African American

individual in the sample as an outlier; this could be due to her ancestry or an unknown variable,

more research is needed to determine the cause.

106

It is imperative to develop methods which can be used on fragmentary remains,

considering how often incomplete remains are recovered. This is true not only in forensic

contexts, but mass graves, commingled remains, and bioarchaeological contexts as well. It is

likely that in these scenarios not all elements would be complete, meaning that an accurate

metric method on a small portion of bone would offer an advantage to the anthropologist (2).

This method has been narrowed from the entire os coxa, to just the pubic bone, and then further

to fragmentary scenarios, ensuring that it is applicable and useful to actual recovered remains (2).

The method also appears to be easily replicable based on the second, repeated, data set, which

exhibits significant results very close to the accuracy exhibited by the larger sample and the

higher scoring simulated fragmentary samples. It is also clear that interobserver error is low for

this author. Future research should investigate how long it takes a new user to become

comfortable and adequate at locating and collecting landmark data.

Based on the promising results here, future research into this method is needed. The

method should be applied to larger samples to continue validating it as an established method so

that it can be used in legal contexts in the future. This research should continue to develop the

fragmentary application of the method as well; replication of this research is vital if it is to ever

be applied to actual forensic casework. Any ancestry specific biases should also be further

explored on larger samples to ensure the method can be used across populations. This method

may also have great potential in exploring the shape changes in female pelves related to

pregnancy and birth. Early analyses showed that the female specimens exhibited more variation

overall than did male specimens—this may be due to the traumatic event of giving birth,

however more research is needed to determine whether or not this is the case.

107

References

1. White TD, Folkens PA. The Human Bone Manual. 1 edition. Burlington: MA: Academic

Press, 2005.

2. Bytheway JA, Ross AH. A Geometric Morphometric Approach to Sex Determination of the

Human Adult Os Coxa: A GEOMETRIC MORPHOMETRIC APPROACH TO SEX

DETERMINATION. Journal of Forensic Sciences 2010;55(4):859–64.

https://doi.org/10.1111/j.1556-4029.2010.01374.x.

3. Spradley K, Jantz RL. Sex Estimation in Forensic Anthropology: Skull Versus Postcranial

Elements: SEX ESTIMATION IN FORENSIC ANTHROPOLOGY. Journal of Forensic

Sciences 2011;56(2):289–96. https://doi.org/10.1111/j.1556-4029.2010.01635.x.

4. Lovell NC. Test of Phenice’s Technique for Determining Sex from the Os Pubis. American

Journal of Physical Anthropology 1989;79:117–20.

5. Sutherland LD, Suchey JM. Use of the Ventral Arc in Pubic Sex Determination. Journal of

Forensic Sciences 1991;36(2):501–11.

6. McFadden C, Oxenham MF. Sex, Parity, and Scars: A Meta-analytic Review. Journal of

Forensic Sciences 2018;63(1):201–6. https://doi.org/10.1111/1556-4029.13478.

7. Lesciotto KM. The Impact of Daubert on the Admissibility of Forensic Anthropology

Expert Testimony. Journal of Forensic Sciences 2015;60(3):549–55.

https://doi.org/10.1111/1556-4029.12740.

8. Zelditch ML, Swiderski DL, Sheets HD, Fink WL. Geometric Morphometrics for

Biologists. First. San Diego, California: Elsevier Academic Press, 2004.

9. Fang Z, Chen X, Su H, Thompson K, Chen Y. Evaluation of stock variation and sexual

dimorphism of beak shape of neon flying squid, Ommastrephes bartramii, based on

geometric morphometrics. Hydrobiologia 2017;784(1):367–80.

https://doi.org/10.1007/s10750-016-2898-0.

10. Clabaut C, Bunje PME, Salzburger W, Meyer A. GEOMETRIC MORPHOMETRIC

ANALYSES PROVIDE EVIDENCE FOR THE ADAPTIVE CHARACTER OF THE

TANGANYIKAN CICHLID FISH RADIATIONS. Evolution 2007;61(3):560–78.

https://doi.org/10.1111/j.1558-5646.2007.00045.x.

11. Marcus L, Hingst-Zaher E, Zaher H. Application of landmark morphometrics to skulls

representing the orders of living mammals. Hystrix, the Italian Journal of Mammalogy

2000;11(1).

12. Spradley K, Jantz RL. Ancestry Estimation in Forensic Anthropology: Geometric

Morphometric versus Standard and Nonstandard Interlandmark Distances. Journal of

Forensic Sciences 2016;61(4):892–7. https://doi.org/10.1111/1556-4029.13081.

108

13. San-Millán M, Rissech C, Turbón D. Shape variability of the adult human acetabulum and

acetabular fossa related to sex and age by geometric morphometrics. Implications for adult

age estimation. Forensic Science International 2017;272:50–63.

https://doi.org/10.1016/j.forsciint.2017.01.005.

14. Sanchez A. Microscribe and Craniometrics. Harris County Institute of Forensic Sciences

2018;:80.

15. UNM Maxwell Museum of Anthropology | Documented Collection. osteolab. 2018.

http://osteolab5.wixsite.com/osteolab/documented-collection (accessed July 9, 2018).

16. Welcome to Python.org. Python.org. . https://www.python.org/ (accessed February 18,

2020).

17. Bookstein FL. Landmark Methods for Forms without Landmarks: Morphometrics of Group

Differences in Outline Shape. Medical Image Analysis 1996;1(3):225–43.

18. Mitteroecker P, Gunz P, Windhager S, Schaefer K. A brief review of shape, form, and

allometry in geometric morphometrics, with applications to human facial morphology.

Hystrix, the Italian Journal of Mammalogy 2013;24(1):59–66.

19. Stegmann MB, Gomez DD. A brief introduction to statistical shape analysis. Informatics

and mathematical modelling, Technical University of Denmark, DTU 2002;15(11).

20. Poulsen J, French A. Discriminant function analysis. Retrieved from 2008.

https://pdfs.semanticscholar.org/8e10/85b33d84d7015174e30e3e911272b6e1c2fe.pdf.

THREE-DIMENSIONAL GEOMETRIC MORPHOMETRIC SEX DETERMINATION AND MODELED FRAGMENTARY ANALYSIS OF THE HUMAN PUBIC BONE.

Students also viewed