Journal of Forensic and Legal Medicine 37 (2016) 8e14

Contents lists available at ScienceDirect

Journal of Forensic and Legal Medicine

j o u r n a l h o m e p a g e : w w w . e l s e v i e r . c o m / l o c a t e / j fl m

Original communication

Assessment of craniometric traits in South Indian dry skulls for sex

determination
Balakrishnan Ramamoorthy, Mangala M. Pai*, Latha V. Prabhu, B.V. Muralimanju,
Rajalakshmi Rai
Department of Anatomy, Kasturba Medical College, Mangalore, Manipal University, Karnataka 575001, India

a r t i c l e i n f o

a b s t r a c t

Article history:
Received 5 June 2015
Received in revised form
24 August 2015
Accepted 3 October 2015
Available online 22 October 2015

The skeleton plays an important role in sex determination in forensic anthropology. The skull bone is
considered as the second best after the pelvic bone in sex determination due to its better retention of
morphological features. Different populations have varying skeletal characteristics, making population
specic analysis for sex determination essential. Hence the objective of this investigation is to obtain the
accuracy of sex determination using cranial parameters of adult skulls to the highest percentage in South
Indian population and to provide a baseline data for sex determination in South India. Seventy adult
preserved human skulls were taken and based on the morphological traits were classied into 43 male
skulls and 27 female skulls. A total of 26 craniometric parameters were studied. The data were analyzed
by using the SPSS discriminant function. The analysis of stepwise, multivariate, and univariate
discriminant function gave an accuracy of 77.1%, 85.7%, and 72.9% respectively. Multivariate direct
discriminant function analysis classied skull bones into male and female with highest levels of accuracy.
Using stepwise discriminant function analysis, the most dimorphic variable to determine sex of the skull,
was biauricular breadth followed by weight. Subjecting the best dimorphic variables to univariate
discriminant analysis, high levels of accuracy of sexual dimorphism was obtained. Percentage classication of high accuracies were obtained in this study indicating high level of sexual dimorphism in the
crania, setting specic discriminant equations for the gender determination in South Indian people.
2015 Elsevier Ltd and Faculty of Forensic and Legal Medicine. All rights reserved.

Keywords:
Skulls
Sex determination
Craniometric parameters
Discriminant function
Forensic anthropology

1. Introduction
The skeleton plays an important role in sex determination in the
eld of forensic anthropology. Though, entire skeleton is essential
for accurate results, it is rarely available completely and in good
condition. Therefore, it is important to apply appropriate methods
for gender determination from skeletal remnants.1,2
The pelvis, femur, tibia, humerus, radius, mandible and cranium
are useful in sex determination.3 Among these, pelvic traits which
are highly dimorphic are referred as the best gender indicator. But
the skull due to its better preservation is believed to be the second
among the best, after the pelvis in gender determination in

* Corresponding author. Department of Anatomy, Kasturba Medical College, Bejai

Campus, Mangalore, Manipal University, Karnataka 575 004, India. Tel.: 91 0824
2211767, 91 9845564685 (mobile).
E-mail addresses: drbalkris83@gmail.com (B. Ramamoorthy), mangala.pai@
manipal.edu (M.M. Pai), latha.prabhu@manipal.edu (L.V. Prabhu), murali.manju@
manipal.edu (B.V. Muralimanju), rajalakshmi.rai@manipal.edu (R. Rai).

archaeological contexts as it can ascertain individual sex with high

accuracy.1e3
The better stability of the dimorphic features of skull in comparison to other bony remnants is due to its durability to changing
environmental conditions. The various methods used for sex
determination are classied into molecular and morphological
methods.4 Robusticityand differences in the size are the most signicant factors contributing to the dimorphic features of the skull.5
Adult skull is preferred to sub-adult skull in sex determination
due to the hormonal control on the growth and development of
bones. The gender differences in skull occur due to the variability in
the bone development and its forces. Bone development in females
is earlier than in males and there is a drastic modication in the
appearance of female skull at the time of puberty. Thus the
morphological differences among the genders manifest more
evidently in the adult skeletons.6,7
The functional expression of modeling of the face and skull,
which starts acquiring the characteristic phenotypical features, is
considered as sexual dimorphism. Over expressed bone elevations

http://dx.doi.org/10.1016/j.jm.2015.10.001
1752-928X/ 2015 Elsevier Ltd and Faculty of Forensic and Legal Medicine. All rights reserved.

B. Ramamoorthy et al. / Journal of Forensic and Legal Medicine 37 (2016) 8e14

in this process in the form of tubercles, ridges, processes etc.,

characterize male skulls and differentiate them from females.7
The morphological features which indicate the sexual dimorphism help in accurate classication in more than 95% of the cases,
when skeletal remnants are available in good condition. A set of 14
indicators were introduced by Krogman in 1955 followed by
introduction of additional four characteristics by Krogman and
Iscan in 1986,such as the shape and size of the nasal aperture, size
of the nasal bone, the length of the zygomatic arch and chin shape
which allowed morphological gender estimation of the human
skull.4 Sex determination based on only morphological features as
done in traditional studies are variable because the features depend
on various environmental conditions and individual perceptions.8
The scientic measurement of the dimensions of the face and
skull bases is considered as craniometry, which provides information related to the race, gender and body type.3,9 Subsequently
there has been a change in trend from morphological methods to
morphometric analysis and statistical comparison like univariate
analysis and the use of indices. These methods also never allow
100% accurate sex determination due to overlap between male and
female values in many parameters.8
To overcome this, advanced analytical methods, such as multivariate linear discriminant functional analysis of metrical data are
used recently.8 Discriminant function analysis is an objective statistical technique selecting least number of traits and maximum
discriminatory effectiveness.10
Different populations have varying skeletal characteristics,
making population specic analysis for sex determination essential.11 In the present century of intermixing population, the population specic standards are of signicant practical relevance.
With all these implications, the objective of our study was to
analyze the accuracy of craniometric parameters for sexual
dimorphism in adult preserved skulls to the highest possible percentage in South Indian population by stepwise, multivariate and
univariate function analysis and to develop discriminant equations.
2. Materials and methods
This retrospective study was done in 70 adult preserved human
skulls of one of the major medical colleges conforming to the inclusion and exclusion criteria from October 2012 to May 2014, after
obtaining the Institute ethics committee clearance. The study
included well preserved fully ossied adult skulls without any
craniofacial injury. Skulls having obliterated anatomical landmarks,
distorted morphological features and with obvious bone pathology
were excluded from the study. Those skulls which could not be
accurately sexed based on morphological features were also
excluded. Mandible was excluded from this study. This was a
limiting factor as majority of skulls didn't have the corresponding
mandibles.
Electronic weighing machine, sliding vernier caliper, spreading
caliper, non-stretchable measuring tape, measuring scale, thread,
chalk and marker were used for measuring the parameters.
2.1. Methods
The sex of the adult preserved skull was initially determined by
standard morphological traits.12 Based on the morphological traits
70 skulls of unknown sex were classied into 43 male skulls and 27
female skulls. A total of 26 craniometric parameters were included
in the study as adapted from Moore- Jansen et al.,13 Buikstra and
Ubelaker13 and Deshmuk AG and Devershi DB.8
The parameters studied were weight, maximum cranial
circumference, maximum cranial length (glabella-opisthocranion),
maximum cranial breadth (euryon-euryon), bizygomatic breadth

(zygion-zygion), basion-bregma height (basion-bregma), cranial

base length (basion-nasion), basion-prosthion length (basionprosthion), maxillo-alveolar length (prosthion-alveolon), maxilloalveolar breadth (ectomolare-ectomolare), biauricular breadth
(auriculare-auriculare), upper facial height (nasion-prosthion),
upper facial breadth (frontomalare temporale-frontomalare temporale), minimum frontal breadth (frontotemporale-frontotemporale), nasal height (nasion-nasospinale), nasal breadth
(alare-alare), orbital height (distance between superior and inferior
orbital margins taken perpendicular to orbital breadth), orbital
breadth (dacryon-ectoconchion), biorbital breadth (ectoconchionectoconchion), interorbital breadth (dacryon-dacryon), frontal
chord (nasion-bregma), parietal chord (bregma-lambda), occipital
chord (lambda-opisthion), foramen magnum length (basion-opisthion), foramen magnum breadth (distance between lateral margins of foramen magnum at the points of greatest lateral curvature)
and mastoid length (vertical projection of mastoid process below
and perpendicular to eye-ear plane).8,13
All the measurements were taken to the nearest millimeter and
were converted to centimeter and weight was measured in grams.
Left side of each skull was chosen. If left side could not be used,
right side was used. After cleaning the preserved skulls, each skull
was serially numbered to minimize operator bias.
2.2. Statistical analysis
The data was tabulated in Microsoft excel worksheet and the
statistical analysis was done using the statistical package for social
sciences software IBM SPSS 20.0 (Chicago, IL, USA) for Windows.
The general descriptive statistics was done for all the craniometric
measurements providing mean and standard deviation (S.D)
separately for both male and female skulls. Student's t test was
performed to check if any signicant differences exist (p  0.05)
among genders.
Direct discriminant function analysis was done by stepwise,
multivariate and univariate methods and sectioning points were
calculated. Multivariate analysis is done separately for all the 26
parameters and the 6 parameters with p value of 0.000 at
descriptive statistical analysis, stepwise analysis was done to
determine the best dimorphic variables and univariate analysis was
done to gure out the contribution of each variable. The classication results of the above were cross-validated using leave one
out classication method.
Discriminant functional scores of male (Zm) and female (Zf)
skulls were obtained by applying the mean data of male and female
skulls respectively using the discriminant functional score
(Z) b0 b1X1 b2 X2 - - - - - - - - b26 X26 (b0 - constant, b1
e b26 are coefcients, X1 e X26 are variable of parameters).
Sectioning point (Z0) was obtained using the average mean value of
male and female skull variables. This program has its sectioning
point at zero. When the values of two group centroids are different,
the sectioning point is obtained by taking the average of the two
means. If the score is higher than the sectioning point, the individual is considered male; lesser than that, the individual is
considered as female.11
3. Results
3.1. Descriptive statistics
Out of the 70 skulls, 43 (61.4%) were male and 27 (38.6%) were
female. The mean value reveals statistical signicant difference
(p  0.05, Independent sample student's t test) across gender for
most of the craniometric parameters as shown in Table 1. Male

10

B. Ramamoorthy et al. / Journal of Forensic and Legal Medicine 37 (2016) 8e14

Table 1
Means, Standard deviations and p values of craniofacial parameters by sex in South Indian population.
Parameters

Mean

Weight
Maximum cranial circumference
Maximum cranial length
Maximum cranial breadth
Bizygomatic breadth
Basion-Bregma height
Cranial base length
Basion-Prosthion length
Maxillo alveolar length
Maxillo alveolar breadth
Biauricular breadth
Upper facial height
Upper facial breadth
Minimum frontal breadth
Nasal height
Nasal breadth
Orbital height
Orbital breadth
Biorbital breadth
Interorbital breadth
Frontal chord
Parietal chord
Occipital chord
Foramen magnum length
Foramen magnum breadth
Mastoid length
a

S.D

P value (Student's t test)

Male

Female

Male

Female

606.66
50.90
17.83
13.30
11.36
13.57
10.20
9.44
5.28
5.89
11.83
6.34
10.34
9.64
4.99
2.70
3.41
4.51
9.70
1.34
11.38
11.20
9.47
3.66
3.13
3.19

488.35
49.23
17.05
12.80
10.88
13.02
9.80
9.18
5.02
5.67
11.29
6.06
10.06
9.42
4.81
2.63
3.46
4.38
9.46
1.29
10.87
10.90
9.18
3.65
3.07
2.92

132.502
1.939
0.813
0.622
0.610
0.614
0.517
0.558
0.321
0.517
0.542
0.491
0.448
0.465
0.283
0.218
0.242
0.490
0.421
0.311
0.524
0.696
0.539
0.316
0.292
0.373

92.153
1.410
0.684
0.615
0.573
0.469
0.536
0.481
0.290
0.560
0.444
0.445
0.438
0.353
0.294
0.171
0.169
0.425
0.424
0.307
0.552
0.461
0.591
0.243
0.300
0.391

0.000a
0.000a
0.000a
0.002a
0.002a
0.000a
0.003a
0.053
0.001a
0.103
0.000a
0.017a
0.011a
0.034a
0.013a
0.141
0.365
0.283
0.025a
0.499
0.000a
0.057
0.039a
0.935
0.422
0.005a

Signicant at 5% level of signicance.

skulls have higher mean values than female skulls in all parameters
except orbital height.
The comparison of standard deviation suggests that male skulls
exhibit more variability than female skulls in all measurements
except cranial base length, maxillo alveolar breadth, nasal height,
biorbital breadth, frontal chord, occipital chord, foramen magnum
breadth and mastoid length as shown in Table 1.
3.2. Discriminant function analysis
3.2.1. Stepwise discriminant function analysis
Using the stepwise discriminant function analysis (Function 1),
two variables were accepted as the best discriminator of sexes
(Table 2). The best predictors of sex were biauricular breadth
followed by weight. The discriminant function was obtained and
sectioning point was estimated (Table 3). The discriminant function equation for sex determination was calculated as shown
below.

Discriminant score Z 18:534 constant

1:347  Biauricular breadth
0:005  Weight
Stepwise analysis method classied 77.1% of the skulls (72.1%
males and 85.2%females). Cross validation using Leave one out
method proved that the model was highly reliable with 77.1% of
skulls being correctly classied as shown in Table 4.

3.2.2. Multivariate direct discriminant function analysis

The direct discriminant function analysis (Function 2) of all 26
variables discriminating male and female skulls was developed and
sectioning point was obtained (Table 5). By multivariate direct
discriminant analysis method sex was determined in 85.7% of the
skulls of which 83.7% were males and 88.9% were females. The
model was proved to be fairly reliable by cross validation using
Leave one out method which classies 64.3% of skulls correctly as
shown in Table 6.
Direct discriminant function analysis (Function 3) of 6 variables
with p value of 0.000 at descriptive statistical analysis was developed and sectioning point was obtained (Table 7). Total classication accuracy of 80% of the skulls was determined of which 79.1%
were males and 81.5% were females. Cross validation of the results
gave an accuracy of 78.6% (Table 8).
3.3. Univariate direct discriminant function analysis
The best dimorphic variables, biauricular breadth and weight
were subjected to univariate direct discriminant function analysis.
Single variable direct discriminant function analysis of biauricular
breadth (Function 4) and weight (Function 5) discriminating male
skulls from female skulls was developed and sectioning point was
obtained as shown in Table 9.
Univariate direct discriminant analysis method for biauricular
breadth and weight correctly classied 72.9% of the skulls (72.1%
males and 74.1% females) individually. Cross validation using Leave

Table 2
Stepwise discriminant function analysis (Function 1) of craniofacial parameters for sex determination in South Indian population.
Step

Variables entered

Wilks' Lambda

Equivalent F-ratio

Degree of freedom

1
2

Biauricular breadth
Weight

0.779
0.695

19.269
14.718

1,68
2,67

B. Ramamoorthy et al. / Journal of Forensic and Legal Medicine 37 (2016) 8e14

11

Table 3
Stepwise discriminant function analysis (Function 1): Canonical discriminant function coefcients and sectioning points in South Indian population.
Variables

Unstandardized coefcients

Standardized coefcients

Structure point

Male and female centroids

Sectioning point

Biauricular breadth
Weight
(Constant)

1.347
0.005
18.534

0.683
0.608

0.803
0.743

M 0.518
F 0.824

0.153

Wilks' Lambda: 0.695, Eigen value: 0.439, Canonical correlation: 0.552.

Table 4
Classication accuracy of craniofacial parameters in South Indian population by
stepwise discriminant function analysis.
Craniofacial parameters and functions Malea n (%)
Function 1: Stepwise analysis
Original
Cross- validated
a

Femalea n (%)

31/43 72.1 23/27

31/43 72.1 23/27

85.2
85.2

Total (%)
77.1
77.1

Predicted group membership.

one out method proved that the model was highly reliable with
same results as shown in Table 10.

4. Discussion
The rst and foremost step in identication of an adult skeleton
is sex determination as this paves the way for further sex dependent techniques such as age and stature determination. The availability of the skeletal remains in complete form and the extent of
sexual dimorphism that runs in the population determine the
reliability of sex determination. It is widely established that the two
most sexually dimorphic skeletal remnants are the skull and the
pelvis.14
Sexual dimorphism exhibits itself in two varieties: male bones
are larger and strongly built than female bones and the male pelvis
is structured for bipedal striding, while the female pelvis displays

differences in size and shape that is adapted both for locomotion

and child birth.14
The important aspect of sexual dimorphism lies in the overlap of
the size of the male and female bones, yet the accuracy depends on
factors causing variation in gender.15 Sexual dimorphism seen in
the male skull is due to the pubertal changes that exhibit increased
muscular attachments whereas the female skull retains the juvenile features.14 In case of any unidentied skull it is essential to
determine the source of population and apply data accordingly
from that particular population or from a similar population.
Sex determination techniques depend on the cranial descriptive
features or on their measurements and ratios. Adequate training
and experience is essential for observational technique and can be
done with accuracy by the experts and not by a common person.
Hence in this study we chose to take measurements of the skull,
which represent the objective method of sex determination as
suggested by Steyn and Iscan.11
The accuracy of sex determination, apart from techniques
employed, also depends on the statistical method used and thus,
the need to apply appropriate statistics is essential for reliable
outcome. Hence, in the present study, we used stepwise, multivariate and univariate direct discriminant function analysis which
is the commonly used statistical technique for sex determination
from skeletal remains. We also compared the results of stepwise,
multivariate univariate direct discriminant function analysis as it
helps in establishing the appropriate cranial parameter which gives

Table 5
Multivariate direct discriminant function analysis (Function 2: All 26 variables): Canonical discriminant function coefcients & sectioning points in South Indian population.
All 26 Variables

Unstandardized coefcients

Standardized coefcients

Structure point

Centroids

Sectioning point

Weight
Maximum cranial circumference
Maximum cranial length
Maximum cranial breadth
Bizygomatic breadth
Basion-Bregma height
Cranial base length
Basion-Prosthion length
Maxillo alveolar length
Maxillo alveolar breadth
Biauricular breadth
Upper facial height
Upper facial breadth
Minimum frontal breadth
Nasal height
Nasal breadth
Orbital height
Orbital breadth
Biorbital breadth
Interorbital breadth
Frontal chord
Parietal chord
Occipital chord
Foramina magnum length
Foramina magnum breadth
Mastoid length
(Constant)

0.001
0.024
0.865
0.117
0.602
0.734
0.401
0.758
2.080
0.975
0.584
1.022
0.014
0.902
0.220
0.239
2.578
0.236
0.077
0.245
0.330
0.013
0.063
0.254
0.104
0.281
23.564

0.160
0.042
0.663
0.072
0.359
0.413
0.210
0.402
0.644
0.521
0.296
0.485
0.006
0.384
0.063
0.048
0.559
0.110
0.032
0.076
0.177
0.008
0.035
0.074
0.031
0.107

0.452
0.432
0.462
0.368
0.367
0.446
0.346
0.219
0.380
0.184
0.489
0.272
0.291
0.241
0.283
0.166
0.102
0.120
0.256
0.076
0.431
0.216
0.235
0.009
0.090
0.324

M 0.850
F 1.354

0.252

Wilks' Lambda: 0.458, Eigen value: 1.185, Canonical correlation: 0.736.

12

B. Ramamoorthy et al. / Journal of Forensic and Legal Medicine 37 (2016) 8e14

Table 6
Classication accuracy of craniofacial parameters in South Indian population by
multivariate discriminant function analysis (Function2: All 26 variables).
Craniofacial parameters and functions Malea n (%)

Femalea n (%)

Function 2: Multivariate analysis (All 26 variables)

Original
36/43 83.7 24/27
Cross- validated
28/43 65.1 17/27
a

88.9
63.0

Table 8
Classication accuracy of craniofacial parameters in South Indian population by
multivariate discriminant function analysis (Function3: 6 variables).

Total (%)

Craniofacial parameters and functions Malea n (%)

85.7
64.3

Function 3: Multivariate analysis (6 variables)

Original
34/43 79.1 22/27
Cross- validated
33/43 76.7 22/27
a

Predicted group membership.

the maximum idea about the gender differences in skulls in South

Indian population.
To the best of our knowledge there have been only a few studies
on the evaluation of dimorphic features of the skull in South Indians, with only a few parameters taken into consideration. Hence
we studied 26 craniofacial parameters on 70 adult preserved human skulls to determine the accuracy of craniometric variables in
sex determination and provide data for determination of sexual
dimorphism in South Indian population from skull measurements.
Studies determining the sexual dimorphism are available with
both childhood16 and adult1,2 skeleton remains. We studied
determination of sexual dimorphism in adult population as denitive features become more pronounced after puberty when hormone controlled bony changes, which helps in determining sexual
dimorphism occurs. Craniometric parameters were used to determine the dimorphism in our study as they are hard structures
which are unyielding and provide differentiation between genders
and population.
In our study mean value of all the craniometric parameters were
higher in males than females with the exception of orbital height.
Similar results were obtained by different studies on various population groups on craniofacial parameters.1,3,8e10,17-25 A comparative table showing most signicant craniometric parameters in
different population groups including our study population have
been shown in Table 11.
In this study, we developed a new set of sex discriminant
functions for the skull by using various parameters of South Indian
skulls. Stepwise discriminant function analysis helps in determining the best dimorphic variable automatically thereby aiding in
obtaining higher classication accuracies with less number of variables. Wilks' lambda determines the relative order in which the
variables enter the function. The program default criterion determines the F values to include or exclude a variable. Wilks
lambda value ranges between 1 and 0. The discriminating power of
the model is increased if the value is closer to zero. Based on this
the two variables selected are biauricular breadth and weight.
Applying other variables with these two variables does not increase
the classication accuracy.
On this aspect, we noted that the most sex-discriminating variable in our study was biauricular breadth, followed by weight.
Stepwise discriminant analysis with the best dimorphic variables
gave an accuracy of 77.1% (72.1% male skulls, 85.2% female skulls)

Femalea n (%)
81.5
81.5

Total (%)
80.0
78.6

Predicted group membership.

which was more than that reported by Osvaldo et al.20 (72%) but
less than that by Steyn et al.11 (85.7%), Saini et al.26 (85.5%) and
Marinescu et al.27 (88%).
Stepwise discriminant function analysis was found to be an
efcient method in our study which improved the sex classication
accuracy with only few dimorphic variables. Cross validation of the
results using leave one out method does not affect the classication accuracy. The stepwise analysis model used in different population groups by different researchers showed a classication
accuracy of the best predictors in each population group ranging
from 72% to 90%.1,11,15,19e21,26,27
Multivariate analysis is done with many variables, some of
which may be signicant by itself and some not. The signicant
variables potentiate the other insignicant ones thereby increasing
the percentage of correct sexing. A single variable may not accurately classify sex in 100% cases. Few percentages of cases still
remain unclassied. More the number of variables used to determine sex by discriminant function analysis, more is the percentage
of correct sex classication. The classication accuracy of sex
determination using multivariate analysis in the present study was
85.7% (83.7% male skulls, 88.9% female skulls) employing 26 variables. Cross validation of the results using leave one out method
reduced the classication accuracy to 64.3%.
There was a fall in the percentage of accuracy on applying the
cross validation method to the results of multivariate analysis in
our study but yielding the most reliable classication percentages.
The reason being, this method cross-checks the distribution of
subjects allocated to the groups to ensure their validity. Here each
case is classied using a discriminant function based on the
remaining sample. The accuracy rate of original sample and the
sample created by cross validation is demonstrated by this method.
In the present study, multivariate analysis of 6 variables with p
value of 0.000 gave a classication accuracy of 80% (79.1% males,
81.5% females). Using the multivariate discriminant function analysis model, different researchers in different population groups had
a total classication accuracy ranging from 70% to 90%.8,10,19,21e25,28
Studies done by using multivariate function analysis on preserved skulls in Northern India24,10 yielded an accuracy percentage
of 70% and 76.7% which was less than that reported in the present
study whereas in South Indians by Gupta et al.23 (90%) it was
higher. Another study in Nanded, India8 had an accuracy of 87.84%
which was comparable to our study.

Table 7
Multivariate direct discriminant function analysis (Function 3: 6 variables): Canonical discriminant function coefcients & sectioning points in South Indian population.
Variables

Unstandardized coefcients

Standardized coefcients

Structure point

Centroids

Sectioning point

Weight
Maximum cranial circumference
Maximum cranial length
Basion-Bregma height
Biauricular breadth
Frontal chord
(Constant)

0.003
0.075
0.513
0.318
0.956
0.182
24.597

0.413
0.131
0.393
0.179
0.485
0.097

0.674
0.643
0.689
0.665
0.729
0.642

M 0.570
F 0.908

0.169

Wilks' Lambda: 0.652, Eigen value: 0.533, Canonical correlation: 0.590.

B. Ramamoorthy et al. / Journal of Forensic and Legal Medicine 37 (2016) 8e14

13

Table 9
Univariate direct discriminant function analysis (function 4 and 5): Canonical discriminant function coefcients and sectioning points in South Indian population.
Variable

Unstandardized coefcient

Wilks' lambda

Eigen value

Canonical correlation

Male and female centroids

Sectioning point

Function 4: Biauricular breadtha

1.972
22.925 (constant)
0.008
4.726 (constant)

0.779

0.283

0.470

0.123

0.805

0.242

0.442

M 0.416
F 0.662
M 0.384
F 0.612

Function 5: Weighta
a

Standardized coefcient 1, Structure point 1.

Table 10
Classication accuracy of craniofacial parameters in South Indian population by
univariate discriminant function analysis.
Craniofacial parameters and functions Malea n (%)
Function 4: Univariate analysis-Biauricular breadth
Original
31/43 72.1
Cross-validated
31/43 72.1
Function 5: Univariate analysis-Weight
Original
31/43 72.1
Cross-validated
31/43 72.1
a

0.114

Femalea n (%)

Total (%)

20/27
20/27

74.1
74.1

72.9
72.9

20/27
20/27

74.1
74.1

72.9
72.9

Predicted group membership.

In univariate direct discriminant function analysis, single variables rather than multiple variables are assessed for their accuracy
in correct sexing. Thereby the method helps in determining the
classication accuracy of each variable individually. Biauricular
breadth and weight individually gave the same accuracy of 72.9%.
Single variable analysis clearly demonstrates that the predictive
value of sexual dimorphism depends mainly upon the sex
discrimination power of the individual variables rather than the
number of variables included in the study. Cross validation of the
results by using leave one out method yielded the same classication accuracy.
Biauricular breadth yielded an accuracy of 73.7% in Western
Australians21 similar to our study. Weight was assessed as a single
variable discriminant for sex determination in our study which has
not been reported in any study as far as our knowledge.
The success rate of sex determination in South Indian crania
using multivariate analysis was better than stepwise and single
variable analysis. The classication accuracy of female skulls was
higher than male skulls by discriminant analysis. It should be noted
that biauricular breadth and weight were found to be the highly
dimorphic variables specic for South Indian population and
should be always considered for sex determination in South
Indians.

The study emphasizes the need for multivariate analysis in sex

determination of skull and shows that neither single variable
analysis nor combination of the most dimorphic variables as in
stepwise analysis provides classication accuracy of high
percentages.
The measurements which are highly dimorphic in a particular
population may not be signicant in other population groups. The
results clearly indicate that if we choose measurements which are
highly dimorphic to a particular population, reliable results with
high degree of accuracy can be obtained.
Depending on two or three measurements which are not actually dimorphic to a particular population may miss valuable information and will generally yield low classication results. The study
gives valuable information on the most dimorphic variables that
gives high level of contribution in the determination of sex from
skulls in south Indian population.
Based on the study, we can conclude that the South Indian
population has a pronounced sexual dimorphism. It is imperative to
use the discriminant functions only to the cases which are known
to be coming from the similar population. In other words, the
population groups with similar sexual dimorphism can use this
formula.
Hence in case of unidentied skull the rst step is to identify the
source population and to apply a method based results from that
population group. Since the derived discriminate function equation
are known to be population specic and gender specic, there is a
requirement of similar equations to be derived for other
population.

5. Conclusion
Percentage classication of high accuracies were obtained from
our study population indicating high level of sexual dimorphism in
the crania, setting specic discriminant functions for the gender
determination in South Indian population.

Table 11
Literature references and the list of most signicant craniometric parameters in different population groups.
Literature references
Study population (South Indians)
Deshmuk AG and Devershi DB8 (Nanded- India)

Steyn M and Iscan MY11 (South African whites)

Kranioti EF, Iscan MY and Michalodimitrakis M15(Modern Cretan)

Osvaldo, Oliveira FD, Rachel, Tinoco LR, Eduardo, Daruge et al.20 (Brazilians)
Ogawa Y, Imaizumi K, Miyasaka S, Yoshino M25 (Modern Japanese)

List of most signicant craniometric parameters

Weight, maximum cranial circumference, maximum cranial length, basion-bregma
height, biauricular breadth and frontal chord (p < 0.001)
Maximum cranial circumference, maximum cranium length, maximum bizygomatic
diameter, basion-nasion length, maximum cranium breadth, cranium height, bregmalambda length, biasterionic breadth, mastoid length and palatal breadth (p  0.05).
Cranial length, basion-bregma, maximum frontal breadth, minimum frontal breadth,
bizygomatic breadth, nasal height, nasal breadth, basion-nasion, basion-prosthion,
nasion-prosthion, mastoid height, biasterionic breadth (p < 0.001)
Maximum cranium length, basion-bregma height, maximum vault breadth, maximum
frontal breadth, bizygomatic breadth, minimum frontal breadth, foramen magnum
length, foramen magnum breadth, basion-nasion length, nasion-prosthion height,
basion-prosthion length, mastoid height, biorbital breadth and nose height (p < 0.001).
Basion-bregma, nasion-lambda, nasion-basion, glabella-lambda, glabella-inion,
glabella-bregma, basion-lambda, zygion-zygion and mastoid-mastoid (p  0.05).
Maximum cranium length, basion-bregmatic height, cranium base length, maximum
cranium breadth, maximum frontal breadth, upper facial breadth and bizygomatic
breadth (p < 0.001).

14

B. Ramamoorthy et al. / Journal of Forensic and Legal Medicine 37 (2016) 8e14

Multivariate direct discriminant analysis gave the highest classication accuracy. The best dimorphic variables among the 26
parameters studied in preserved skulls were biauricular breadth
and weight.
Since the sex determination using morphometric analysis depends on numerical values, in case of any unidentied skull the rst
and foremost step will be to identify the source population of the
skull and discriminant functions specic for that population group
only should be used in gender determination.
The data of the present study will be of use in forensic science
and anthropology and for further studies on population data and
serve as a future framework for sex determination using craniometric parameters in South Indian population.
Conict of interest
None of the authors have any conict of interest associated with
this study.
Funding
None.
Ethical approval
Institutional ethics committee clearance obtained.
Contribution of authors
All Authors have contributed in the conception, acquisition,
analysis and interpretation of the study. Authors have revised the
article critically for intellectual content. They also have given their
nal approval for the version to be published. Authors agreed to be
accountable for integrity of the article to be published.
Acknowledgments
The authors would like to thank Manipal University for access to
the skull collections in their anatomy museum. They are also
thankful to Dr. Kavya and Dr. Reshma Muhamed for their assistance
during data collection. The authors appreciate Dr. Priya Cinna.T.Durai for critically editing the manuscript. Authors would like to
acknowledge Mr. Gopal M Gokhale, Associate Professor, Department of Statistics, Govinda Dasa College, Surathkal, for his guidance
related to statistics.
References
1. Naikmasur VG, Shrivastava R, Mutalik S. Determination of sex in South Indians
and immigrant Tibetans from cephalometric analysis and discriminant functions. Forensic Sci Int 2010;197:122. e1-6.
2. Gonzalez PN, Bernal V, Perez SI. Analysis of sexual dimorphism of craniofacial
traits using geometric morphometric techniques. Int J Osteoarchaeol 2009;21:
82e91.

3. Sangvichien S, Boonkaew K, Chuncharunee A, Komoltri C, Piyawinitwong S,

Wongsawut A, et al. Sex determination in Thai skulls by using Craniometry:
multiple logistic regression analysis. Siriraj Med J 2007;59:216e21.
4. Suazo GIC, Zavando MDA, Smith RL. Performance evaluation as a diagnostic test
for traditional methods for forensic identication of sex. Int J Morphol 2009;27:
381e6.
5. Nagaoka T, Shizushima A, Sawada J, Tomo S, Hoshino K, Sato H, et al. Sex
determination using mastoid process measurements: standards for Japanese
human skeletons of the medieval and early modern periods. Anthropol Sci
2008;116:105e13.
6. Suazo GIC, Zavando MDA, Smith RL. Evaluating accuracy and precision in
morphologic traits for sexual dimorphism in malnutrition human skull: a
comparative study. Int J Morphol 2008;26:877e81.
7. Suazo GIC, Zavando MDA. Age effect in the morphological traits performance
for sex determination in human skulls and mandibles. Int J Morphol 2012;30:
296e301.
8. Deshmukh AG, Devershi DB. Comparison of cranial sex determination by univariate and multivariate analysis. J Anat Soc India 2006;55:48e51.
9. Rooppakhun S, Piyasin S, Vatanapatimakul N, Kaewprom Y, Sitthiseripratip K.
Craniometric study of Thai skull based on three-dimensional computed tomography (CT) data. J Med Assoc Thai 2010;93:90e8.
10. Sumati, Patnaik VVG, Phatak A. Determination of sex from mastoid process by
discriminant function analysis. J Anat Soc India 2010;59:222e8.
11. Steyn M, Iscan MY. Sexual dimorphism in the crania and mandibles of South
African whites. Forensic Sci Int 1998;98:9e16.
12. Reddy KSN. Identication. In: The essentials of forensic medicine and toxicology.
26thedn. Kolkata: Medical Book Company; 2007. p. 51e88.
13. Measuring adult human remains. Available from: www.redwoods.edu/
instruct/agarwin/anth_6_measurements.htm. [accessed 25.7.14].
14. Scheuer L. Application of osteology to forensic medicine. Clin Anat 2002;15:
297e312.
15. Kranioti EF, Iscan MY, Michalodimitrakis M. Craniometric analysis of the
modern cretan population. Forensic Sci Int 2008;180:110. e1-5.
16. Hsiao TH, Tsai SM, Chou ST, Pan JU, Tseng YC, Chang HP, et al. Sex determination using discriminant function analysis in children and adolescents: a
lateral cephalometric study. Int J Leg Med 2010;124:155e60.
17. Zavando MDA, Suazo GIC, Smith RL. Sexual dimorphism determination from
the lineal dimensions of skulls. Int J Morphol 2009;27:133e7.
18. Gapert R, Black S, Last J. Sex determination from foramen magnum: discriminant function analysis in an eighteen and nineteenth century British sample.
Int J Leg Med 2009;123:25e33.
19. Ahmed AA, Mohammed HA, Hassan MA. Sex determination from cranial
measurements in recent northern Sudanese. Khartoum Med J 2011;04:539e47.
20. Oliveira OF, Tinoco R, Daruge E, Terada ASSD, Silva RHA, Paranhos LR. Sexual
dimorphism in Brazilian human skulls: discriminant function analysis.
J Forensic Odontostomatol 2012;30:26e33.
21. Franklin D, Cardini A, Flavel A, Kuliukas A. Estimation of sex from cranial
measurements in a Western Australian population. Forensic Sci Int 2013;229:
158. e1-8.
22. Uthman AT, Al Rawi NH, Al Timimi JF. Evaluation of foramen magnum in
gender determination using helical CT scanning. Dentomaxillofac Radiol
2012;41:197e202.
23. Gupta D, Banerjee A, Kumar A, Rao SR, Jose J. Discriminant function analysis of
mastoid measurements in sex determination. J Life Sci 2012;4:1e5.
24. Sumati, Patnaik VVG, Phatak A. Determination of sex from hard palate by
discriminant function analysis. Int J Basic Appl Med Sci 2012;2:243e51.
25. Ogawa Y, Imaizumi K, Miyasaka S, Yoshino M. Discriminant functions for sex
estimation of modern Japanese skulls. J Forensic Leg Med 2013;20:234e8.
26. Saini V, Srivastava R, Rai RK, Shamal SN, Singh TB, Tripathi SK. An osteometric
study of northern Indian populations for sexual dimorphism in craniofacial
region. J Forensic Sci 2011;56:700e5.
27. Marinescu M, Panaitescu V, Rosu M, Maru N, Punga A. Sexual dimorphism of
crania in a Romanian population: discriminant function analysis approach for
sex estimation. Rom J Leg Med 2014;22:21e6.
28. Kumar M, Lone MM, Patnaik VVG. Determination of sex by discriminant
function analysis: a cephalometric study. Int J Pure App Biosci 2013;1:18e21.