You are on page 1of 7

Forensic Science International 244 (2014) 314.e1–314.

e7

Contents lists available at ScienceDirect

Forensic Science International


journal homepage: www.elsevier.com/locate/forsciint

Forensic Anthropology Population Data

Estimation of sex from the metric assessment of digital hand


radiographs in a Western Australian population
Rebecca DeSilva, Ambika Flavel, Daniel Franklin *
Centre for Forensic Science, The University of Western Australia, 35 Stirling Highway, Crawley, Western Australia, Australia

A R T I C L E I N F O A B S T R A C T

Article history: The forensic anthropologist is responsible for contributing to the identification of an unknown by
Received 26 May 2014 constructing a biological profile from their skeletal remains. Towards achieving this goal,
Received in revised form 15 July 2014 anthropologists can apply population and temporally specific standards with known error margins
Accepted 19 August 2014
to morphometric data collected from a decedent. Recent research relating to the formulation of sex
Available online 16 September 2014
estimation standards has focussed on the assessment of bones other than the traditionally favoured
pelvis and cranium, such as long bones of the appendicular skeleton. In particular, sex estimation
Keywords:
standards based on morphometric data from metacarpals and phalanges have reported classification
Sex discrimination
Hand bone measurements
accuracy rates of 80% (and above) based on a narrow range of populations. The purpose of this study is to
Population standards provide population-specific hand bone sex-estimation standards for a contemporary Western Australian
Sexual dimorphism population.
Forensic anthropology population data The present study examines digital right hand radiographs of 300 adults of known age, equally
represented by sex. A total of 40 measurements were taken in each hand (metacarpals and proximal
phalanges); the measurements were then analysed using univariate statistics and cross-validated direct
and stepwise discriminant function analysis. All hand bone measurements were significantly sexually
dimorphic, with a tendency for the width measurements to express a higher degree of dimorphism than
the length measurements. A maximum cross-validated classification accuracy of 91% was achieved with
a sex bias of -6%. The standards presented here can be used in future forensic investigations that require
sex estimation of hand bones in a Western Australian population.
ß 2014 Elsevier Ireland Ltd. All rights reserved.

1. Introduction includes estimating ancestry, sex, age, and stature through metric
and non-metric analyses of skeletal remains; this data can thus be
The use of forensic anthropology in medico-legal investigations used to narrow the pool of possible matching identities.
has become more common over time, with an increasing number Biological sex is a primary component of the osteobiography
of cases involving remains that are ‘problematic’ for a forensic and is generally one of the first to be assessed; standards for
pathologist [1,2]. Such cases are often referred to a forensic subsequent estimations (age and stature) are largely sex-specific
anthropologist and include those concerning skeletal, partially [1,6]. Sexual dimorphism is the biological foundation of sex
fleshed, charred or dismembered remains. Forensic anthropology estimation and is defined as the physical and behavioural
is the application of concepts derived from the theory and methods differences that occur between males and females [7,8]. Sex
of physical anthropology to a forensic investigative context differences in the shape, size and appearance of bones arise during
[2,3]. Analysis of human skeletal remains covers any aspect of development according to individual genetic markers and in
the biological profile (osteobiography) that aids towards establish- response to sex hormones during puberty; bone development
ing personal identity or manner of death [4,5]. A biological profile being dependent on a combination of genetic markers and
hormone exposure [8]. The age at which these sex-specific
morphological changes occur is subject to a number of population
specific genetic and environmental factors [8]. As the degree of
* Corresponding author at: Centre for Forensic Science, The University of Western
sexual dimorphism, and the age at which it occurs in males and
Australia, M420, 35 Stirling Highway, Crawley 6009, Western Australia, Australia.
Tel.: +61 8 6488 1232; fax: +61 8 6488 7285. females, varies between populations (who may differ geographi-
E-mail address: daniel.franklin@uwa.edu.au (D. Franklin). cally, temporally or by ancestry) sex estimation standards are

http://dx.doi.org/10.1016/j.forsciint.2014.08.019
0379-0738/ß 2014 Elsevier Ireland Ltd. All rights reserved.
314.e2 R. DeSilva et al. / Forensic Science International 244 (2014) 314.e1–314.e7

required to be population specific [9,10]. Standards used to analyse of Health (WA DoH) Picture Archiving and Communication
morphometric data are thus most accurate when applied to the Systems (PACS) database. The right hand of 150 adult males and
population from which they are derived [10,11]. 150 adult females are examined; the stated age range for males is
Although the pelvis is highly sexually dimorphic, and therefore 18.3–64.3 years (mean 41.9) and 18.5–68.4 years (mean 42.8) for
the preferred element for sex estimation, recent research has females. The medical images are acquired from various Western
worked towards quantifying the sex estimation potential of other Australian hospitals and relate to patient admissions for clinical
skeletal elements [12]. Previous research has demonstrated that evaluation of the hand-wrist (generally sprains and other soft
the sternum [9], femur [13], metatarsals [14] and metacarpals [15] tissue injuries) between 2011 and 2013. The standardised protocol
can be used to correctly classify sex with a high degree of expected for acquiring the hand-wrist X-rays is as follows: Focus Receptor
accuracy (above 80%). However, and with specific reference to the Distance (FRD): 100 cm; Focus Object Distance (FOD) 98 cm;
hand bones, Burrows et al. [11] demonstrated that the application Object Receptor Distance (ORD) 2 cm – resultant magnification is
of skeletal standards for metacarpals formulated for one popula- standardised at 2% (generally less than variation introduced by
tion can be less accurate when applied to another that is intra-observer measurement precision – see below). Only radio-
geographically removed; this can be attributed to population graphs that show little (or compete absence of) skeletal trauma
and temporal differences between the samples specific to each and/or anomalies in the metacarpals and proximal phalanges are
study. When standards are applied inappropriately, the errors included. In accordance with standard research ethics require-
highlighted by Burrows et al. [11] have a potentially significant ments and the inherent constraints of using medical data, the
impact on the value of anthropological evidence in legal radiographs are received anonymised, with only age and sex
proceedings. It is therefore crucial that population-specific information for each individual retained. In this study, ancestry of
standards are applied. the subjects is not known (nor is ethnicity recorded at hospital
The Western Australian (WA) population is deemed to be more admission as it is not deemed to be of medical relevance), however
ethnically diverse than other states and territories of Australia, ancestry is generally taken to approximate a representative sample
with Perth being one of the most diverse capital cities [16]. The of the contemporary Western Australian population (see above).
Western Australian population comprises 3.1% indigenous Aus-
tralians, which is higher than the Australian average (2.5%), but 2.2. Methods
considerably lower than the Northern Territory (26.8%) (Table 1)
[17]. Australian Bureau of Statistics [18] data indicate that 56.2% of 2.2.1. Measurements from digital radiographs
the WA population have one or more parent born overseas and 75% The 40 linear measurements used in this study follow
have an ancestry other than Australian (within two generations). previously published definitions [15,20] adapted for 2D images;
This compares with Australia as a whole where 46.2% of people four measurements are acquired for each metacarpal (MC) and
have one or more parent born overseas. In broad terms the proximal phalanx (PP) (Table 2 and Fig. 1). The OsiriX1 line-tool
population is predominantly Caucasian in all Australian states (but function is used to define the linear measurements (in milli-
not territories) [19]. As no Australia-wide anthropological studies metres).
have been undertaken, we treat the WA population as distinct,
albeit it is anticipated that the standards produced here are more 2.2.2. Statistical analyses
applicable to other Australian states than they would be to other A 6  6 precision study was performed to statistically quantify
non-Australian populations. the degree of intra-observer error and therefore determine the
The aim of the present study is to formulate sex-estimation quality of the subsequent data collected. Six randomly selected
standards based on metacarpal and phalange measurements of hand radiographs were measured a total of six times each, with a
digital hand radiographs, applicable to a Western Australian minimum of one day between repeats to minimise data recall. The
population. This study forms part of an ongoing project dedicated error associated with measurement repeats was calculated using
to the development of population specific anthropological the technical error of measurement (TEM), relative TEM (rTEM), and
standards for Australian forensic practitioners. coefficient of reliability (R) statistics (see [21–24] for detailed
explanations of these approaches).
2. Materials and methods Descriptive statistics (mean, standard deviation and range)
were calculated and assessed for outliers prior to subjecting
2.1. Materials the data to further statistical analyses. The mean hand bone
measurements of males and females were compared using
The sample comprises 300 posterior–anterior (PA) digital one-way ANOVA with sex as the dependent variable and
radiographs acquired from the Western Australian Department individual hand bone measurements as the independent variables.

Table 1
Australian Bureau of Statistics, 2011 Census data: Ancestry (primary response) by state.

NSW VIC QLD SA WA TAS NT ACT OTa Aust. Average

Australian Aboriginal 2.5% 0.7% 3.6% 1.9% 3.1% 4.0% 26.8% 1.4% 29.0% 2.5%
Non-indigenous Australian 20.5% 20.5% 23.2% 21.3% 19.2% 30.0% 17.7% 24.2% 13.2% 22.7%
North-West European 41.4% 42.0% 51.3% 50.6% 49.1% 49.7% 13.8% 45.1% 8.0% 45.3%
Southern and Eastern European 8.4% 12.7% 4.1% 9.9% 7.1% 2.3% 3.6% 7.4% 0.9% 8.4%
North African and Middle Eastern 3.8% 2.5% 0.5% 0.8% 0.9% 0.3% 0.3% 1.1% 0.2% 2.1%
South-East Asian 3.0% 3.0% 1.5% 2.0% 2.5% 0.5% 3.2% 2.8% 14.5% 2.5%
North-East Asian 6.1% 4.8% 2.5% 2.5% 3.5% 1.1% 2.1% 4.5% 13.7% 4.3%
Southern and Central Asian 3.6% 4.1% 1.5% 2.1% 2.4% 0.6% 1.7% 4.0% 0.8% 3.0%
Peoples of the Americas 0.8% 0.5% 0.5% 0.4% 0.5% 0.3% 0.4% 0.8% 0.1% 0.6%
Sub-Saharan African 0.5% 0.8% 0.6% 0.5% 1.5% 0.3% 0.6% 0.6% 0.3% 0.7%
Other 2.7% 1.9% 3.4% 2.4% 2.6% 5.0% 18.2% 2.8% 16.2% 1.1%
Not stated 6.9% 6.5% 7.3% 5.7% 7.6% 5.9% 11.6% 5.4% 3.2% 6.9%
a
Other Territories: Christmas Island, the Cocos (Keeling) Islands, and the Jervis Bay Territory.
R. DeSilva et al. / Forensic Science International 244 (2014) 314.e1–314.e7 314.e3

Table 2 of MC5 (F = 61.35; P < 0.001). Sex differences explain 17–55% of


Definitions of hand bone measurements.
sample variance.
Code Definition

ML The maximum linear distance between the most distal point of the 3.2. Discriminant function analyses
bone to the most proximal point of the bone.
LS The maximum linear distance between the most medial point and 3.2.1. Direct single variables
the most lateral point of the head of the bone. Demarking points were calculated for each of the 40 measure-
LB The maximum linear distance between the most medial point and
ments acquired from the hand to assess whether sex could be
the most lateral point of the base of the bone.
MB The maximum linear distance between the most medial and most accurately estimated using a single measurement. The most
lateral point of the mid-shaft region. accurate function for each bone is reported in Table 4 and
associated predicted probabilities in Fig. 2. Cross-validated
accuracies for the direct single variable functions ranged from
Discriminant function analysis (DFA) prediction models were 76.70% (Function 4; WMMC4) to 85.70% (Function 7; WBPP2). Eight
developed, allowing the sex of an individual to be estimated from of the ten functions are derived from base width measurements.
hand bone measurements [25,26]. The discriminant scores Function 10 (BWPP5) is the most accurate variable for estimating
produced are classified based on their relationship to a sectioning sex in this population with a combined cross-validated accuracy of
point (a value halfway between the male and female mean scores 85.30% and a sex bias of 4.00%.
[26]). Because male hands are significantly larger than females in
this population, if the discriminant score is lower than the 3.2.2. Multiple variable functions
sectioning point, it is likely to represent a female individual [27]. A A series of direct multiple variable discriminant analyses was
series of stepwise and direct multiple DFA’s were performed; performed for each metacarpal using combined measurements
functions are assessed according to both the cross-validated (ML, WB, WH and WM). Function 11, based on the measurement of
accuracy (percentage of correctly classified individuals) and the metacarpal two (Table 5) yielded a cross-validated classification
sex bias (the difference between the classification accuracy of each accuracy of 84.00% (sex bias 4.00%) and is the only function to
sex). A negative sex bias indicates that more females are correctly improve on the accuracy of the single measurement demarking
assigned than males, and vice versa. point for that digit.
A stepwise discriminant function analysis was performed for
3. Results the complete hand (Function 12; Table 5); eight variables were
selected and achieved a cross-validated classification accuracy of
3.1. Preliminary analysis 91.00% with a sex bias of 6.00%. Additionally, stepwise DFA of
measurements from each of the five individual digits were
Results of the precision test yielded rTEM values less than 5% calculated (potentially 8 measurements per digit), however only
(range 0.24–3.34%) and R-values approaching 1 (range 0.97–1.00). three functions (Function 13–15; Table 6) improved on the accuracy
Overall, the mean male measurement values were larger than of the single measurement demarking points. The fifth digit is the
female values (Table 3). The maximum length of MC3 had the most dimorphic (Function 15) with a cross-validated classification
largest mean sex difference (5.77 mm) and maximum mid-shaft accuracy of 87.70% and sex bias of 2.00%.
width of MC3 had the smallest mean sex difference (1.07 mm). In
general the maximum length and base widths had larger mean 4. Discussion
differences between the sexes than the head and mid-shaft widths.
ANOVA results were all statistically significant; the most sexually Sex is often the first component of the biological profile
dimorphic measurement was the base width of PP2 (F = 363.88; estimated and the remaining elements (age and stature) are
P < 0.001); the least sexually dimorphic was the maximum length generally assessed using sex-specific methods [1,28]. Misclassifi-
[(Fig._1)TD$IG] cation may have a flow on effect that reduces the accuracy of the
estimation of other biological attributes and ultimately compro-
mise the identification of an individual. Population specific sex
estimation standards afford the most accurate possible classifica-
tion [11,21,28]. Within a forensic context it is particularly
important to subject results to statistical interrogation and supply
the end-user with a measure of sexual dimorphism and a
quantification of the accuracy of the sex estimation standards
developed. The latter facilitates informed decisions as to which of
the available standards is most applicable in a particular
circumstance.
In the present study mean male measurements taken from
digital hand radiographs were significantly larger than mean
female measurements. In general the base, head, and mid-shaft
width measurements of all ten bones were more dimorphic
(greater between-sex variance) compared to the maximum length
measurements. The latter is congruent with previous research; the
mediolateral base width of MC2 was reported by Scheuer and
Elkington (1993) [20] as producing their highest index of
separation, and in Barrio et al. (2006) [29] it resulted in the
highest sex classification accuracy (91.00%). Manolis et al. (2009)
[30] also suggest that epiphyseal widths are the most accurate
Fig. 1. Hand bone measurements defined in Table 1. Measurements illustrated on indicators of sex as they generally exhibit morphometric sexual
metacarpal two (radiograph of an adult female right hand). dimorphism to a greater extent than length measurements.
314.e4 R. DeSilva et al. / Forensic Science International 244 (2014) 314.e1–314.e7

Table 3
Descriptive statistics of hand bone measurements (in mm).
#
Measurement Male (n = 150) Female (n = 150) Mean diff F R2 P-value

Mean SD Range Mean SD Range

Digit 1
***
MLMC1 48.86 3.21 40.50–57.81 44.60 2.88 36.72–51.90 4.26 146.13 0.33
***
WHMC1 17.03 1.41 12.97–20.73 14.92 1.11 11.81–17.64 2.11 207.59 0.32
***
WBMC1 16.61 1.60 13.76–21.77 14.70 1.20 12.69–18.14 1.91 137.31 0.41
***
WMMC1 9.91 1.00 7.38–13.03 8.78 0.76 6.83–10.81 1.13 121.60 0.29
***
MLPP1 34.38 2.31 29.53–39.86 31.55 2.14 26.80–37.24 2.83 120.85 0.29
***
WHPP1 15.41 1.23 8.56–15.73 13.66 0.97 7.19–13.84 1.75 186.55 0.30
***
WBPP1 12.27 1.25 11.90–19.09 10.79 0.99 10.01–15.62 1.48 129.26 0.38
***
WMPP1 8.27 0.91 5.98–10.75 7.15 0.81 4.98–8.92 1.12 128.01 0.30
Digit 2
***
MLMC2 74.75 4.55 62.66–88.17 68.98 3.96 60.42–78.40 5.77 137.62 0.32
***
WHMC2 21.09 1.58 14.11–20.23 18.55 1.27 11.93–17.98 2.54 234.42 0.33
***
WBMC2 17.12 1.35 17.56–25.36 15.32 1.20 15.09–22.03 1.80 147.63 0.44
***
WMMC2 9.45 0.82 7.68–11.89 8.23 0.70 6.64–10.46 1.22 191.69 0.39
***
MLPP2 43.30 2.47 37.42–49.53 40.52 2.49 34.64–47.65 2.78 93.84 0.24
***
WHPP2 17.89 1.02 9.68–15.44 15.80 0.87 9.62–13.05 2.09 363.88 0.39
***
WBPP2 12.46 0.94 15.41–20.96 11.15 0.68 13.70–18.01 1.31 191.18 0.55
***
WMPP2 10.37 0.78 7.95–12.42 9.08 0.68 7.27–11.11 1.29 233.75 0.44
Digit 3
***
MLMC3 69.16 4.40 57.26–82.90 63.83 3.65 55.37–74.86 5.33 130.23 0.30
***
WHMC3 15.83 1.35 14.38–21.27 13.98 1.18 11.20–17.11 1.85 160.04 0.39
***
WBMC3 17.38 1.32 13.52–20.06 15.40 1.17 12.39–18.56 1.98 189.27 0.35
***
WMMC3 9.05 0.71 7.41–10.88 8.04 0.64 6.36–10.00 1.01 169.59 0.36
***
MLPP3 48.24 2.99 40.68–56.09 44.85 2.61 38.96–52.13 3.39 109.60 0.27
***
WHPP3 17.55 1.11 10.67–16.10 15.40 0.86 10.22–13.62 2.15 353.43 0.41
***
WBPP3 13.23 0.97 15.15–20.16 11.78 0.75 13.34–17.85 1.45 210.08 0.54
***
WMPP3 10.61 0.86 8.56–12.57 9.10 0.73 7.11–10.71 1.51 268.07 0.33
Digit 4
***
MLMC4 61.56 4.39 50.46–75.21 56.75 3.37 49.45–64.91 4.81 113.25 0.28
***
WHMC4 14.25 1.25 12.06–19.24 12.63 1.02 10.69–15.55 1.62 151.46 0.35
***
WBMC4 14.92 1.30 11.41–17.61 13.19 1.04 9.92–15.19 1.73 159.90 0.34
***
WMMC4 7.53 0.73 5.49–9.28 6.50 0.57 5.24–8.21 1.03 183.10 0.38
***
MLPP4 45.36 2.85 37.91–52.40 41.68 2.44 35.68–48.19 3.68 158.22 0.33
***
WHPP4 16.23 1.08 10.19–15.37 14.32 0.93 9.35–12.88 1.91 268.68 0.42
***
WBPP4 12.35 0.88 13.87–19.58 10.98 0.72 12.19–16.43 1.37 214.88 0.47
***
WMPP4 9.94 0.86 7.58–12.22 8.36 0.72 6.21–10.13 1.58 297.82 0.50
Digit 5
***
MLMC5 56.23 5.68 46.28–68.80 51.95 3.53 35.77–60.70 4.28 61.35 0.17
***
WHMC5 15.30 1.28 11.79–18.37 13.42 1.05 10.13–15.23 1.88 194.78 0.38
***
WBMC5 14.83 1.30 12.26–18.51 12.99 1.05 11.07–17.05 1.84 181.03 0.40
***
WMMC5 8.74 0.93 6.97–11.20 7.45 0.73 5.71–9.45 1.29 179.56 0.38
***
MLPP5 36.05 2.16 29.04–42.20 32.82 2.16 27.89–39.77 3.23 158.22 0.35
***
WHPP5 15.31 0.99 8.15–12.70 13.45 0.82 7.54–11.52 1.86 314.58 0.29
***
WBPP5 10.38 0.84 13.27–18.88 9.35 0.75 11.50–16.13 1.03 124.17 0.51
***
WMPP5 8.53 0.84 6.41–11.07 7.16 0.74 5.16–8.97 1.37 223.08 0.43
#
Key: Measurement definitions in Table 2.
***
P < 0.001.

Previous studies examining the fleshed hand also affirm that measurements (both skeletal and fleshed) is related to functional
width and breadth measurements are more sexually dimorphic loading in response to mechanical stimulus [33–36]. Trinkaus et al.
than length measurements [31,32]. It has been suggested (1994) [33] and Wilczak (1998) [37] posit that longitudinal
that greater expression of sexual dimorphism in hand width measurements (such as maximum length) are less likely to be
affected by any physical stresses resulting from habitual activity.
Table 4
They concluded that variance in diaphyseal measurements were
Demarking points for single hand bone measurements (in mm).
heavily influenced by mechanical loading, as evidenced by the high
#
Measurement Demarking Combined Sex bias level of bilateral asymmetry exhibited by professional tennis
points cross-validated
players (up to 57% difference in diaphyseal dimensions), as
(M > F) accuracy
compared to four recent human samples (up to 14%). Maximum
Metacarpal bone length and articular dimensions were found to exhibit
Function 1. WBMC1 15.98 81.70% 2.00%
Function 2. WBMC2 19.82 81.30% 4.00%
minimal levels of asymmetry.
Function 3. WHMC3 16.22 81.30% 6.70% Sex estimation discriminant functions with high classification
Function 4. WMMC4 7.02 76.70% 6.70% accuracies are not necessarily reliable if they misclassify either
Function 5. WBMC5 14.36 80.30% 7.30% males or females at disproportionate levels. To this end, functions
Proximal Phalanx
with a sex bias 5% should be applied with caution, despite having a
Function 6. WBPP1 14.53 78.30% 0.70%
Function 7. WBPP2 16.84 85.70% 6.00% potentially high degree of expected cross-validated accuracy.
Function 8. WBPP3 16.48 87.00% 7.40% Generally, the functions that included multiple measurements
Function 9. WBPP4 15.27 81.30% 4.00% achieved a higher accuracy; this observation is also consistent with
Function 10. WBPP5 14.38 85.30% 4.00% previous research [15,20,29]. Posterior probabilities were calculated
Key: #Measurement definitions in Table 2. for all functions; these are used as a statistical indication the
[(Fig._2)TD$IG] R. DeSilva et al. / Forensic Science International 244 (2014) 314.e1–314.e7 314.e5

Fig. 2. Predicted probabilities of the sample for the most accurate measurements (in mm) for individual metacarpals 1–5. Where 0 is male extreme and 1 is female extreme;
demarking point is 0.5.

classification confidence of each individual. Higher posterior proba- for the radius [12]; 76% to 97% for the femur [12,46,47]; and 91% for
bility values are indicative of a discriminant score that is well above the tibia [47]. The range of classification accuracies presented in
(or below) the sectioning point [25]. No individuals within the sample the present study are comparable to those previously reported for
were classified with lower than 40% certainty. This provides some other skeletal elements; further confirming that hand bones from a
degree of statistical confidence that functions derived from the Western Australian population are sexually dimorphic and can be
Western Australian sample should be applicable to the broader used to accurately estimate sex.
population in a forensic context. Functions based on a sole metacarpal The classification models produced in this study are designed
variable had the lowest percentage of individuals (ranging from 55.80% for forensic application in Western Australia to assign sex in the
to 64.82%) classified at 80% certainty or above. This would suggest that complete hand or its individual bones. In an ideal situation, data
those functions are not as robust as the direct multiple variable and would be acquired from the complete hand in order to classify sex
stepwise discriminant functions, albeit reliable for application in using the stepwise model (Function 12) as it has the highest
instances where only an isolated bone is available for analysis. classification accuracy and a reasonable sex bias value. However,
In the present study, the highest cross-validated classification the likelihood of recovering a complete hand when a human body
accuracy (91.00%) was for Function 12, requiring a total of eight has already progressed to the skeletal stage of decomposition is
hand measurements (Table 6) from five proximal phalanges and potentially lower compared to other larger and more robust
the first metacarpal. This function also had the highest percentage skeletal elements. Waldron (1987) [48] calculated percentage
of individuals classified at 80% certainty and above (90.59%) recovery rates for different archaeological skeletal elements and
suggesting it is unlikely those individuals were classified by found a relatively strong correlation between the survival of
chance. Function 12 is therefore highly recommended for use on skeletal elements and their size. The phalanges were among the
complete hands. When considering multivariate functions for least well-represented skeletal elements, likely due to their
individual digits Function 11 for MC2 outperforms any single smaller size. Metacarpals were considered to be relatively resistant
variable for that digit, and uses all four measurements. Metacarpal to post-mortem damage, with recovery rates ranging from 39.80%
two is a highly recognisable digit and sexually dimorphic with a for the right fifth metacarpal to 63.60% for the left third metacarpal.
high classification accuracy [84.0%] and low sex bias ( 4.00%). Depending on the post-mortem interval, forensic cases may
In considering the accuracy of skeletal sex assessments in exhibit percentage recovery rates similar to those found by
general, the pelvis is highly dimorphic [38] as it has a specific Waldron (1987) [48], or perhaps even lower depending on the
morphology related to biological function (e.g. accommodating manner of death (e.g. post-blast remains). For this reason, sex
childbirth in females). For the pelvis alone, sex classification estimation standards were formulated based on a single hand
accuracy rates have been reported between 90% and 95% [39– bone; because these models do not require a complete hand, they
41]. High classification accuracies have also been reported for the can be applied to digital radiographs in situations where the hand
skull, ranging from 75% to 90% [42,43]. A high degree of expected bones have been disarticulated, fractured, or in cases of
sex classification accuracy has also been reported for the long commingled amputated limbs such as what may occur in disaster
bones: between 83% to 96% for the humerus [12,44,45]; 85% to 94% victim identification (DVI) situations.

Table 5
Direct multiple variable discriminant analysis of metacarpal two.
#
Equation: unstandardised coefficient, variable and constant Group centroids Sectioning point Correctly assigned Sex bias

Function 11. Metacarpal Two

(0.41 * MLMC2) + (0.315 * WBMC2) + (0.181 * WHMC2) + < 1.024 0.00 < 123; , 129 [84.0%] 4.00%
(0.567 * WMMC2) 17.108 , 1.024
Key: #Measurement definitions in Table 2; < = Male, , = Female.
314.e6 R. DeSilva et al. / Forensic Science International 244 (2014) 314.e1–314.e7

Table 6
Stepwise discriminant function analysis of the Western Australian adult sample.
#
Step Variables Unstandardised Standardised Wilk’s Structure Group Sectioning Correctly Sex bias
coefficient coefficient lambda point centroids point assigned

Complete Hand
Function 12. 1 WBPP2 0.325 0.308 0.450 0.803 < 1.371 0.00 91.00% 6.00%
2 WMPP4 0.610 0.486 0.411 0.727 , 1.371
3 MLPP5 0.257 0.572 0.395 0.530
4 WHPP5 0.590 0.470 0.378 0.469
5 WBPP1 0.190 0.210 0.369 0.575
6 MLPP2 0.199 0.495 0.360 0.408
7 WBPP3 0.327 0.325 0.352 0.792
8 MLMC1 0.090 0.275 0.346 0.509
Constant 18.107
Individual digits
Function 13. 1 WBPP3 0.576 0.573 0.457 0.923 < 1.175 0.00 86.00% 5.40%
2 WMPP3 0.530 0.423 0.429 0.804 , 1.175
3 MLMC3 0.058 0.233 0.418 0.561
Constant 18.56
Function 14. 1 WMPP4 0.759 0.604 0.500 0.864 < 1.154 0.00 85.70% 6.00%
2 MLPP4 0.141 0.373 0.441 0.601 , 1.154
3 WBPP4 0.306 0.310 0.427 0.820
Constant 17.74
Function 15. 1 WBPP5 0.580 0.527 0.486 0.872 < 1.175 0.00 87.70% 2.00%
2 WMMC5 0.295 0.246 0.454 0.659 , 1.175
3 WBMC5 0.226 0.265 0.446 0.686
4 WMPP5 0.391 0.309 0.437 0.734
5 WHPP5 0.425 0.338 0.429 0.548
6 MLPP5 0.113 0.251 0.419 0.618
Constant 16.75

Key: #Measurement definitions in Table 2; < = Male, , = Female.

5. Conclusions [5] SWAGANTH. Sex Assessment The Scientific Working Group for Forensic
Anthropology, 2010.
[6] D. Franklin, Forensic age estimation in human skeletal remains: current concepts
In the forensic anthropological assessment of skeletal remains, and future directions, Leg. Med. (Tokyo) 12 (2010) 1–7.
accurate sex estimation based on population-specific standards is a [7] A. Glücksmann, Sexual Dimorphism in the Human and Mammalian Biology and
Pathology, Academic Press, London, 1981.
valuable tool in narrowing the pool of possible matches. The [8] D.W. Frayer, M.H. Wolpoff, Sexual dimorphism, Annu. Rev. Anthropol. 14 (1985)
present study has outlined a series population specific sexing 429–473.
standards for Western Australia; such data did not previously exist. [9] D. Franklin, A. Flavel, A. Kuliukas, A. Cardini, M.K. Marks, C. Oxnard, et al.,
Estimation of sex from sternal measurements in a Western Australian population,
This represents a valuable addition to the database of human Forensic Sci. Int. 217 (2012) 230e1–230e5.
identification protocols currently being developed as part of on- [10] R.A. Lazenby, Population variation in second metacarpal sexual size dimorphism,
going research projects at the Centre for Forensic Science at the Am. J. Phys. Anthropol. 118 (2002) 378–384.
[11] A.M. Burrows, V.P. Zanella, T.M. Brown, Testing the validity of metacarpal
University of Western Australia. Importantly, these standards are
use in sex assessment of human skeletal remains, J. Forensic Sci. 48 (2003)
based on data that is representative of the contemporary Western 17–20.
Australian population. [12] M.K. Spradley, R.L. Jantz, Sex estimation in forensic anthropology: skull versus
postcranial elements, J. Forensic Sci. 56 (2011) 289–296.
[13] S.A. Asala, M.A. Bidmos, M.R. Dayal, Discriminant function sexing of fragmentary
Ethical approval femur of South African blacks, Forensic Sci. Int. 145 (2004) 25–29.
[14] A.G. Robling, D.H. Ubelaker, Sex estimation from the metatarsals, J. Forensic Sci.
42 (1997) 1062–1069.
This study was granted ethics approval by the Human Research
[15] D.T. Case, A.H. Ross, Sex determination from hand and foot bone lengths, J.
Ethics Committee of the University of Western Australia; Project Forensic Sci. 52 (2007) 264–270.
Number RA/4/1/4362. [16] OMI, Cultural Diversity in Western Australia: A Demographic Profile, Office of
Multicultural Interests, Government of Western Australia, 2013 http://www.
omi.wa.gov.au/resources/publications/info_sheets/Cultural_Diversity_2013.pdf
Acknowledgements (accessed 3.07.14).
[17] ABS, Census of Population and Housing – Counts of Aboriginal and Torres Strait
Islander Australians, 2011, Australian Bureau of Statistics, 2014 http://www.abs.
The authors thank Adj. Assoc. Prof. Rob Hart, Frontier Medical gov.au/ausstats/abs@.nsf/Lookup/2075.0main+features32011 (accessed 3.07.14).
Imaging International (Western Australia) for assistance with [18] ABS, 2011 Census Data: QuickStats WA Demographics, 2011.
obtaining and anonymising the digital radiographs used in this [19] ABS, Australian Bureau of Statistics [Table Builder Basic: online database],
Australian Bureau of Statistics, 2011 http://www.abs.gov.au/.
study. We also thank the valuable comments of two anonymous
[20] J.L. Scheuer, N.M. Elkington, Sex determination from metacarpals and the first
referees that helped to improve the final version of this manuscript. proximal phalanx, J. Forensic Sci. 38 (1993) 769–778.
[21] D. Franklin, A. Cardini, A. Flavel, A. Kuliukas, Estimation of sex from cranial
References measurements in a Western Australian population, Forensic Sci. Int. 229
(2013) 158e1–158e8.
[22] S.J. Ulijaszek, D.A. Kerr, Anthropometric measurement error and the assessment
[1] V. Braz, Anthropological estimation of sex, in: S. Blau, D. Ubelaker (Eds.), Hand-
of nutritional status, Br. J. Nutr. 82 (1999) 165–177.
book of Forensic Anthropology and Archaeology, Left Coast Press, California, 2009,
[23] R. Goto, C.G. Mascie-Taylor, Precision of measurement as a component of human
pp. 201–207.
variation, J. Physiol. Anthropol. 26 (2007) 253–256.
[2] D.C. Dirkmaat, L.L. Cabo, S.D. Ousley, S.A. Symes, New perspectives in forensic
[24] G.C. Marks, J.P. Habicht, W.H. Mueller, Reliability, dependability, and precision
anthropology, Am. J. Phys. Anthropol. 137 (2008) 33–52.
of anthropometric measurements. The Second National Health and Nutrition
[3] C. Cattaneo, Forensic anthropology: developments of a classical discipline in the
Examination Survey 1976–1980, Am. J. Epidemiol. 130 (1989) 578–587.
new millennium, Forensic Sci. Int. 165 (2007) 185–193.
[25] M.L. Patriquin, M. Steyn, S.R. Loth, Metric analysis of sex differences in South
[4] L. Scheuer, Application of osteology to forensic medicine, Clin. Anat. 15 (2002)
African black and white pelves, Forensic Sci. Int. 147 (2005) 119–127.
297–312.
R. DeSilva et al. / Forensic Science International 244 (2014) 314.e1–314.e7 314.e7

[26] M. Slaus, Z. Tomicic, Discriminant function sexing of fragmentary and complete [38] D. Franklin, A. Cardini, A. Flavel, M.K. Marks, Morphometric analysis of pelvic
tibiae from medieval Croatian sites, Forensic Sci. Int. 147 (2005) 147–152. sexual dimorphism in a contemporary Western Australian population, Int. J. Legal
[27] A. Agresti, Categorical Data Analysis, 2nd ed., John Wiley and Sons, USA, 2002. Med. (2014).
[28] D. Franklin, A. Cardini, A. Flavel, A. Kuliukas, The application of traditional and [39] T.W. Phenice, A newly developed visual method of sexing the os pubis, Am. J. Phys.
geometric morphometric analyses for forensic quantification of sexual dimor- Anthropol. 30 (1969) 297–301.
phism: preliminary investigations in a Western Australian population, Int. J. Legal [40] N.C. Lovell, Test of Phenice’s technique for determining sex from the os pubis, Am.
Med. 126 (2012) 549–558. J. Phys. Anthropol. 79 (1989) 117–120.
[29] P.A. Barrio, G.J. Trancho, J.A. Sanchez, Metacarpal sexual determination in a [41] D.H. Ubelaker, C.G. Volk, A test of the phenice method for the estimation of sex, J.
Spanish population, J. Forensic Sci. 51 (2006) 990–995. Forensic Sci. 47 (2002) 19–24.
[30] S.K. Manolis, C. Eliopoulos, C.G. Koilias, S.C. Fox, Sex determination using meta- [42] E. Giles, O. Elliot, Sex determination by discriminant function analysis of crania,
carpal biometric data from the Athens Collection, Forensic Sci. Int. 193 (2009) Am. J. Phys. Anthropol. 21 (1963) 53–68.
130e1–130e6. [43] D. Franklin, L. Freedman, N. Milne, Sexual dimorphism and discriminant function
[31] N.I. Ishak, N. Hemy, D. Franklin, Estimation of sex from hand and handprint sexing in indigenous South African crania, Homo 55 (2005) 213–228.
dimensions in a Western Australian population, Forensic Sci. Int. 221 (2012) [44] M.S. Robinson, M.A. Bidmos, The skull and humerus in the determination of
154e1–154e6. sex: reliability of discriminant function equations, Forensic Sci. Int. 186 (2009)
[32] K.E. Aboul-Hagag, S.A. Mohamed, M.A. Hilal, E.A. Mohamed, Determination of sex 86e1–86e5.
from hand dimensions and index/ring finger length ratio in Upper Egyptians, [45] J. Albanese, H.F.V. Cardoso, S.R. Saunders, Universal methodology for developing
Egypt. J. Forensic Sci. 1 (2011) 80–86. univariate sample-specific sex determination methods: an example using the
[33] E. Trinkaus, S.E. Churchill, C.B. Ruff, Postcranial robusticity in Homo. II: humeral epicondylar breadth of the humerus, J. Archaeol. Sci. 32 (2005) 143–152.
bilateral asymmetry and bone plasticity, Am. J. Phys. Anthropol. 93 (1994) 1–34. [46] J. Albanese, G. Eklics, A. Tuck, A metric method for sex determination using
[34] R. DiBennardo, J.V. Taylor, Sex assessment of the femur: a test of a new method, the proximal femur and fragmentary hipbone, J. Forensic Sci. 53 (2008) 1283–
Am. J. Phys. Anthropol. 50 (1979) 635–637. 1288.
[35] D.R. Carter, M.C. Van Der Meulen, G.S. Beaupre, Mechanical factors in bone growth [47] M.S. Robinson, M.A. Bidmos, An assessment of the accuracy of discriminant
and development, Bone 18 (1996) 5S–10S. function equations for sex determination of the femur and tibia from a South
[36] M. Moss, The design of bones, in: J. Goodfellow, P. Bullough (Eds.), Scientific African population, Forensic Sci. Int. 206 (2011) 212e1–212e5.
Foundations of Orthopaedics and Traumatology, William Heinemann Medical [48] T. Waldron, The relative survival of the human skeleton: implications for palaeo-
Books, London, 1980, pp. 59–66. pathology, in: A. Boddington, A. Garland, R. Janaway (Eds.), Death, Decay and
[37] C.A. Wilczak, Consideration of sexual dimorphism, age, and asymmetry in quan- Reconstruction; Approaches to Archaeology and Forensic Science., Manchester
titative measurements of muscle insertion sites, Int. J. Osteoarchaeol. 8 (1998) University Press, United Kingdom, 1987, pp. 55–64.
311–325.

You might also like