Content uploaded by Rodolfo F H Melani
Author content
All content in this area was uploaded by Rodolfo F H Melani on Sep 23, 2017
Content may be subject to copyright.
TECHNICAL NOTE
ANTHROPOLOGY
Lara Maria Herrera,
1,2
M.Sc.; Ra
!
ıssa Ananda Paim Strapasson,
1
M.Sc.; Alice Aquino Zanin,
1
D.D.S.;
Jorge Vicente Lopes da Silva,
3
Ph.D.; and Rodolfo Francisco Haltenhoff Melani,
1
Ph.D.
Comparison Among Manual Facial
Approximations Conducted by Two
Methodological Approaches of Face Prediction
ABSTRACT: This study verified the difference between two methods of forensic facial approximation (FFA) regarding recognition and
resemblance rates. Three-dimensional models of skulls were obtained from computerized tomography (CT) scans of two subjects (targets). Two
manual FFAs were performed for each target, by applying two different guidelines for the facial structures (what we called “American method”
(AM) and “Combined method”(CM)). Unfamiliar assessors evaluated the sculptures by recognition and resemblance tests. The AM was that
which allowed more correct responses of recognition and higher resemblance’s scores for the male target (p<0.001). Regarding guidelines for
modeling characteristics of the face, the ones that are practical and easily performed for sculptures, such as the length of the anterior nasal spine
multiplied by 3 for nose prediction, may offer better results in terms of resemblance.
KEYWORDS: forensic science, forensic anthropology, forensic dentistry, skull, face, forensic facial reconstruction, forensic facial approxi-
mation, sculpture
Forensic facial approximation (FFA) is a technique that uses
the analysis of the skull and specific guidelines to estimate the
most likely face of an unknown individual.
The literature often distinguishes the technique according to
the space (two-dimensional (2D) or three-dimensional (3D)), the
tool (pencils, modeling clay, or computer), and the method
adopted. The latter refers to the three different schools of FFA:
American, Russian (or anatomical method), and British (also
called combined, mixed, or Manchester’s method).
The American method, created by Krogman, relies on facial
soft tissue thickness (FSTT) tables. Once the skull is positioned
in the Frankfurt horizontal plane (FHP), markers with FSTT
average values are glued onto the skull at specific landmarks.
Therefore, they can act as guides for the deposition of the mod-
eling material (1,2).
The Russian method, which was developed by Mikhail
Gerasimov, consists of modeling the muscles of the face and the
neck one by one, considering signs or traces on the skull.
Besides this, Gerasimov believed that facial components such as
eyes, nose, mouth, and ears could be predicted from specific
areas of the skull (1,2).
In the U.K., Richard Neave created the combined method,
which is a fusion of the American and the Russian methods. In
this method, Neave used the knowledge of facial muscle anat-
omy and the FSTT mean values (2–4).
Some authors dispute the conception of these schools. There
are records showing that Gerasimov documented the Russian
technique with drawings without soft tissue depth indicators (2).
On the other hand, researchers (5–8) point out that the Russian
method is not solely based on the anatomical knowledge of the
face. They say Gerasimov obtained FSTT data from five land-
marks and used them in some of his sculptures. These data were
not references at isolated sites but “lines of the profile,”built
with strips of clay onto the skull. The interpretation of these
authors is that the Russian method would follow the same prin-
ciple as the combined method. Ullrich and Stephan (6) also
report that Gerasimov considered the determination of the areas
of insertion of the muscles to be questionable and that it was
unusual to represent the muscles in his sculptures, except for the
temporal and the masseter (6).
Stephan (5,8) also questions the American method by claim-
ing that it would have emerged in Germany and that it would
follow the principle of the combined method as Krogman has
stated anatomical knowledge and notions of facial proportions to
be necessary to achieve a good result in the approximation.
Despite these controversies, studies usually designate one of
the three methods to conduct FFAs (see references [9–17]).
Guidelines for predicting eyes, nose, mouth, and ears—struc-
tures that carry important information for the process of recogni-
tion—complement the methods. A variety of guidelines are
1
Department of Community Dentistry, School of Dentistry, University of
S~
ao Paulo, Avenida Professor Lineu Prestes, 2227, S~
ao Paulo, SP 05508-
000, Brazil.
2
Department of Community Dentistry, School of Dentistry, Araraquara,
S~
ao Paulo State University (UNESP), Rua Humait!
a, 1680, Araraquara, SP
14801-903, Brazil.
3
Three-dimensional Technologies Division, Renato Archer Information
Technology Center, Rodovia Dom Pedro I, km 143.6, Jardim Santa M^
onica,
Campinas, SP 13069-901, Brazil.
Received 8 July 2016; and in revised form 23 Nov. 2016; accepted 29
Nov. 2016.
1©2017 American Academy of Forensic Sciences
J Forensic Sci, 2017
doi: 10.1111/1556-4029.13435
Available online at: onlinelibrary.wiley.com
available, developed from the observation of the relationship
between soft and hard tissues in different populations. Some of
these subjective guidelines were assessed in photographs and
CTs, demonstrating uncertainty regarding morphology and
dimensions (4,18–20).
The comparison among individuals and their FAs has been
widely used to evaluate the accuracy of the technique. However,
studies evaluating different methodologies applied on the skull,
especially regarding the three aforementioned schools and differ-
ent guidelines of facial features, received little attention (21).
Due to the lack of studies comparing methods for face predic-
tion through manual FA assessments and the lack of studies with
Brazilian subjects, this study aimed to verify whether there is
difference between two methods of FA regarding recognition
and resemblance rates and which one best helps in the process
of recognition.
Materials and Methods
Two Brazilian individuals—one female and one male, aged
45–60 and 18–30, respectively—with normal nutritional status,
living in the city of S~
ao Paulo, donated copies of their CT scans.
Photographs of their faces were obtained in frontal and profile
views as suggested by authors (22–24). These two individuals
were the targets of the study.
At the Renato Archer Information Technology Center (CTI),
CT scans were used to print full-size prototypes of the target’s
skulls. A recent study by the authors has described the acquisi-
tion and the accuracy of those replicas (25).
At the Laboratory of Forensic Anthropology and Forensic
Dentistry (School of Dentistry, University of S~
ao Paulo), each
replica was fixed at a base with a vertical stem and positioned in
the FHP. The first author, who performed the FAs, was unaware
of the identities of the targets, but had access to information
related to their biological profile.
For each replica, two manual FAs were conducted: one
with the “American”method (AM) and another with the
“combined”method (CM). FSTT data obtained from living
individuals by magnetic resonance imaging (MRI) (26) were
used according to authors (17) considering sex and normal
nutritional status. The FSTT for “trichion”landmark was not
used, as it is an anatomical landmark in soft tissue with no
corresponding site on the underlying bone. Markers were
obtained with vinyl cylinders, cut to different thicknesses with
the aid of a digital caliper, and glued to their respective land-
marks on the replicas.
Both FAs (AM and CM, respectively) were initially conducted
for the female target (FA1 and FA2) and then for the male target
(FA3 and FA4). We decided to switch methods so that the
researcher’s practice would not influence them. The first author
has been previously trained for this research by attending a
forty-hour workshop in the field.
Table 1 shows the parameters we used to compose each
method.
TABLE 1–– Guidelines used to make FAs.
Structure American Method (AM) Combined Method (CM)
Filling Clay strips were used for the filling of the face respecting the
thicknesses given by each marker. Filling began across the front
part, followed by the masseter muscle region and chin area (27)
The main muscles of the face were sculpted, based on the origins and
insertions on bone tissue: temporalis, orbicularis oculi, orbicularis
oris, masseter, buccinator, frontalis, corrugator, procerus, compressor
naris, levator labii alaeque nasi, levator labii, levator anguli oris,
zygomaticus major, zygomaticus minor, risorius, mentalis, depressor
labii, and depressor anguli oris
Eyes Artificial eyes (eyeball of 25 mm diameter and 12 mm iris, according
to the findings of Woff and Tian et al. (cited in [28]) positioned
according to Krogman and Gatliff: In frontal view, the top of the
cornea was placed on the junction of two lines—one designed from
the medial margin to the lateral margin of the orbit and the other
forming a bisector, leaving the top edge toward the bottom; in
lateral view, the tip of the cornea was at the center of a line drawn
between the top and lower margins of the orbit, so that to form a
tangent projected 2 mm (27)
Artificial eyes (eyeball of 25 mm diameter and 12 mm iris, according
to the findings of Woff and Tian et al. (cited in [28]) positioned
according to Whitnall (29): closer to the lateral and to the top
margins of the orbit in a front view. The eyeball protrusion relative
to the orbit was 4 mm (28)
Mouth The depth of lips was determined by the “infradentale”tissue depth
marker; height was obtained by measuring the distance between the
cementum–enamel junction lines of upper and lower central incisors;
the mouth width was determined by measuring the distance between
the distal surfaces of the canine teeth (measured in arc). Finally, the
parting line was performed with a marking at the midpoint of the
vertical thickness of lips (27)
The intercanine distance is three-fourths of the total mouth width
(30). The parting line of the lips was located in a middle line
on the crowns of upper teeth according to Angel (cited in
[14])
Nose In frontal view, the nasal aperture was measured at its widest point
and this value was added of 10 mm for Caucasoids (5 mm for each
side) and 16 mm for Negroids (8 mm for each side). In lateral view,
for obtaining the nasal projection, the length of the anterior nasal
spine was measured and multiplied by 3. This value was added to
the depth of the tissue depth marker “P5—MidPhiltrum.”The
direction and morphology of the anterior nasal spine were also used
to sculpture the shape of the nose (27)
The nose was sculpted according to formulas of Rynn et al. (31)—
according to them, the method restricts the subjectivity of the
method of two tangents of Gerasimov and takes into account the
anatomical variations of the nasal region. The linear distances X, Y,
and Z were obtained, and regression formulas used were as follows:
(i) 0.83 Y !3.5 (mm); (ii) 0.9 X !2 (mm); (iii) 0.93 Y !6 (mm)
Ears Ears were constructed according to the study of Krogman and Iscan
(cited in [20,27]): The length is approximately 50 mm, positioned
behind the mandibular branch, with a slope of 15 degrees
Ears were sculpted and positioned according to Gerasimov (cited in
[20]), who proposes the height corresponding to the distance
between glabella and subnasale points added of 2 mm. The author
also proposes the width as approximately half the length, and also
says that the ear axis is parallel to the jaw branch axis. The format
and positioning were also based on information found in the mastoid
process, which indicated the degree of protrusion of the ear and
whether the lobe was united or free
2JOURNAL OF FORENSIC SCIENCES
For both methods, the neck was modeled with soft clay sup-
ported on the support rod. A thin layer of clay was applied on
the face to cover the markers without interfering with the face
shape or the tissues depth. A sponge was used to create the
pores of the skin. Eyebrows were also sculpted (32), and a stan-
dard hairstyle for females and another for males were adopted
regardless of methods. When each sculpture was finished, pho-
tographs were taken in front and profile standards.
Sixteen other volunteers (eight females and eight males) acted
as “false targets”by providing photographs of their faces for
comparison with FAs. These volunteers presented similar biolog-
ical profiles to the targets (e.g., sex, age, and nutritional status).
Forty adult volunteers were invited to assess FAs. No assessor
was familiar with the targets and the false targets, as described
by Stephan e Cicolini (33).
The assessments were conducted by recognition and resem-
blance tests, as performed by the authors in a recent study (25):
for the recognition test, each FA (FA1, FA2, FA3, and FA4)
was presented to examiners along with a pool of pictures (eight
false targets and the target individual) and assessors performed
the recognition by pointing out the one they believed to belong
to the FA (they also could choose none of the individuals); for
the resemblance test, the pictures of the FAs and their respective
targets were presented together, side by side and examiners had
to point out the level of resemblance among faces (1 =no
resemblance, 2 =slight resemblance, 3 =approximate resem-
blance, 4 =close resemblance, or 5 =great resemblance). Given
the lowest attributed similarity (score 2), the examiner indicated
the most similar characteristics among those listed: face shape,
eyes, nose, mouth, and/or ears. In addition, FAs were presented
with pictures of two false targets to compose a fake test. Evalua-
tions were conducted on-site (printed survey at the Laboratory of
Anthropology and Forensic Dentistry (OFLab), School of Den-
tistry, University of S~
ao Paulo) and using an online survey, pre-
pared with the exact same characteristics as the on-site
assessment.
For each FA (FA1, FA2, FA3, and FA4), the absolute (n) and
relative frequencies (%) of faces identified by the assessors as a
target (none or from 1 to 9) were obtained, as well as the fre-
quencies of correct recognition (success). A binomial model for
generalized estimating equations (GEE) (34) with a logistic func-
tion (generalized linear logistic model) was employed to com-
pare the proportions of successes between methods.
For resemblance assessments, frequencies of responses in each
category and measures of central tendency and position were
obtained. For the cases evaluated with any degree of resem-
blance (score >1), the frequencies of the regions of the face of
greatest resemblance were also obtained.
For each FA, three resemblances were evaluated by the same
assessors, one for the target and two others for the false targets.
Thus, the scores among these three evaluations were compared
by the nonparametric Friedman test (35). In addition, to consider
the dependence among observations of the same assessor and
check whether there is effect of method on resemblance propor-
tions, a nonparametric analysis of ordinal data for repeated mea-
sures (36) was made. The level of significance was 5%; that is,
significant differences were considered when the test descriptive
level (p-value) was less than 0.05. Cases in which p-values are
greater than 0.05 and smaller than 0.10 should also be consid-
ered, suggesting a difference with weaker evidence.
SPSS software (Statistical Package for Social Sciences) ver-
sion 19, R version 3.1.1, Macro Excel on: http://www.ime.usp.b
r/~jmsinger/, Microsoft Excel and Word were used.
All participants in this study provided informed and written
consent, and the research was approved by the Research Ethics
Committee (CEP) of the Faculty of Dentistry, University of S~
ao
Paulo (protocol number 37709314.0.0000.0075).
Results
Figure 1 shows FAs of the two targets with the AM (FA1 and
FA3) and CM (FA2 and FA4). Approximately fourteen hours
was needed to make each sculpture.
Descriptive statistics of recognition frequencies are shown in
Table 2. When evaluating the behavior of the differences
between proportions of success by both methods, considering
the dependence of observations, the GEE model showed effect
of method in the proportions of success only for male target, the
AM being that which allowed more correct responses in this
case (female target: Wald statistic =1.05, degrees of free-
dom =1, p-value =0.305; male target: Wald statistic =78.78,
degrees of freedom =1, p-value =<0.001).
More than a half of assessors considered FAs to have some
resemblance (at least slight resemblance) with the targets as seen
in Table 3. When FAs were compared with false targets, more
than a half of assessors also pointed at least a slight resemblance
in most cases.
Difference between proportions of at least some resemblance
(score 2) for the AM and CM (Fig. 2) was not significant for
the female target (p=0.47), while for the male target, the differ-
ence between proportions was significant (p<0.001). A more
detailed analysis of differences among the similarity scores for
the targets in relation to both methods considering the depen-
dence of observations showed effect of method for the male tar-
get, in which the AM allowed higher resemblance’s scores
(female target: Wald statistic =0.06, degrees of freedom =1, p-
value =0.807; male target: Wald statistic =33.64, degrees of
freedom =1, p-value =<0.001).
The frequencies of the facial regions considered by assessors
to be of great resemblance were obtained only for cases where
there was at least a slight resemblance (Table 4).
Discussion
3D manual FA has been the most popular approximation
method used in Forensics (15). High rates of success in recogni-
tion, although controversial, have been achieved by this method
(3,9,27). Authors (21,37) have stated that 3D manual methods
are superior to others because they provide a more detailed and
closer appearance, even requiring longer working time. They
have also argued that computerized methods produce facial
images that are “unacceptable for identification”due to their
inanimate appearance, “lifeless and limited details.”On the other
hand, other researchers (38–40) have criticized the manual
method, claiming that it is highly subjective, requiring artistic
ability of the practitioner.
Performed either manually or by computer, advantages and
disadvantages are inherent in both procedures. The metric com-
parison between manual and computerized FA with the face of
the target individual conducted by Decker et al. (41) enabled
viewing areas of greater and lesser accuracy. A comparative
method as such used by the authors does not evaluate the capac-
ity of a face to be recognized or its similarity to the target and
only quantifies morphological details. The authors found that
both techniques (manual and computerized) had areas of inaccu-
racy. The present study used 3D manual methods to predict
HERRERA ET AL. .APPROACHES OF FACE PREDICTION 3
faces, as computerized methods represent mimicry of the tradi-
tional ones but in another work environment and with other
resources.
Overall, the results suggested that the AM allows greater pro-
portion of correct recognition and higher rates of some resem-
blance compared to the CM. But, at this point, we have no
explanation for the significant results found only for the male
target from the GEE model and from the nonparametric analysis
of ordinal data for repeated measures. Stephan and Hennenberg
(21) verified higher frequencies of recognition by their AM
compared to the their CM, also significant for just one target.
The authors stated that a method can be better than another
depending on the characteristics of the skull. However, they do
not expose which characteristics they refer to and how the char-
acteristics could help choose one method of FA. We think it is
important to mention that recognition is a subjective process and
assessors may respond differently to the same survey of recogni-
tion at different times.
Regarding scores, medians were smaller than 3, indicating that
at least half of the evaluations did not exceed the approximate
TABLE 2–– Frequency of responses using recognition test to assess FAs.
Individual
Female Male
FA1 (AM) FA2 (CM) FA3 (AM) FA4 (CM)
n%n%n%n%
0 (none) 2 5.0 8 20.0 4 10.0 5 12.5
1 6 15.0 4 10.0 0 0.0 1 2.5
2 5 12.5 14 35.0 4 10.0 1 2.5
3 9 22.5 0 0.0 1 2.5 2 5.0
4 1 2.5 0 0.0 0 0.0 0 0.0
5 1 2.5 2 5.0 2 5.0 1 2.5
6 3 7.5 3 7.5 14 35.0 717.5
7820.0 410.0 7 17.5 9 22.5
8 0 0.0 0 0.0 8 20.0 13 32.5
9 5 12.5 5 12.5 0 0.0 1 2.5
Total 40 100.0 40 100.0 40 100.0 40 100.0
Highlighted values represent correct recognitions (targets).
TABLE 3–– Descriptive statistics of scores of the rating scale for resem-
blance assessment and p-value from comparison tests among suspects.
1st. Quartile Median 3rd. Quartile p-Value*
FA1A 1.00 2.00 3.00 0.760
FA1B 1.00 2.00 3.00
FA1C 2.00 2.00 3.00
FA2A
†
1.00 1.00 2.00 <0.001
FA2B
‡
1.00 2.00 2.00
FA2C
‡
2.00 2.00 3.00
FA3A
‡
2.00 3.00 4.00 <0.001
FA3B
†
1.00 2.00 2.75
FA3C
†
2.00 2.00 3.00
FA4A 1.00 2.00 2.00 0.334
FA4B 1.00 1.00 2.00
FA4C 1.00 2.00 2.75
*Friedman test.
‡
Shows higher significant values than
†
.
Bold values represent the targets.
FIG. 1–– FAs of the targets. FA1: American method (female); FA2: combined method (female); FA3: American method (male); FA4: combined method
(male); T1: female target; T2: male target. Source: adapted from Herrera et al. (25).
4JOURNAL OF FORENSIC SCIENCES
resemblance, showing low resemblance, even in relation to the
targets, for both methods. Assessments have already indicated
that targets do not obtain higher scores compared to false targets
and that approximations made morphologically differently
obtained similar scores by the resemblance tests (10). Maybe this
has happened because FA is a technique that estimates faces
including characteristics that cannot even be predicted solely
from the skull, and assessors may expect portraits. Moreover, in
FFA, professionals deal with a lot of techniques and general
rules developed for specific populations to predict facial appear-
ance (nose, mouth, and so on), which does not mean they will
fit to any skull.
So, we have attributed different techniques for prediction to
each method (AM and CM) for the structures of the face.
Despite the low resemblances pointed to by assessors for com-
pleted faces, it was possible to assess each characteristic alone.
We observed that the overall shape of the face of the FAs was
considered more similar to the face of the targets regardless of
the methods (Table 4). The shape of the face is a very important
parameter for facial assessments, mainly considered by unfamil-
iar assessors (22). The fact that CM has worked better for the
male target than for the female one may be explained by the
FIG. 2–– Frequencies of at least some resemblance (score 2). Letter B (female) and letter A (male) represent the targets.
TABLE 4–– Frequencies of the regions of the face of greatest resemblance.
Methods Regions of the Face
Female
Target Male Target
n%n%
American Face shape 18 72.00 25 67.57
Nose 15 60.00 29 78.38
Eyes 12 48.00 15 40.54
Mouth 6 24.00 10 27.03
Ears 1 4.00 2 5.41
Combined Face shape 8 27.59 15 68.18
Nose 17 58.62 5 22.73
Eyes 16 55.17 8 36.36
Mouth 9 31.03 4 18.18
Ears 1 3.45 0 0.00
HERRERA ET AL. .APPROACHES OF FACE PREDICTION 5
presence of stronger muscles in men, which may have facilitated
configuration of the male face as anatomic principles predomi-
nate in such a method.
Among all the other structures of the face, the nose showed the
highest frequencies of at least some resemblance with the nose of
the targets (Table 4), especially those constructed with the method
of the length of the anterior nasal spine multiplied by 3 (27). This
technique was considered for underestimating the nasal projection
for men and women in a study using radiographs (19), and the
authors stated that it could perform better if applied directly to
skulls. In this paper, the method may have overestimated the nasal
projection, especially for male FA (Fig. 1), but this finding has not
been quantified. Furthermore, it was easily employed, which may
have facilitated the modeling process. The method of linear regres-
sion equations of Rynn et al. (31) is quite objective, coming from
quantitative assessments. However, with regard to practicality, it
has not run easily in 3D manual FAs, presenting great potential for
2D techniques. Recently, a method to predict the nose projection
has been proposed to facilitate sculpting methods (42). It would be
interesting to test other techniques by assessing them by recogni-
tion and resemblance tests.
The ears and mouth were the structures less pointed with some
resemblance. Usually, ears contribute less to identification, except
for cases in which the person has very atypical ears (20). In a
study conducted by Decker et al. (41), a quantitative comparative
analysis (metric comparison) by overlapping manual and computer-
ized FAs to the real face (CT) of a male individual showed that
the areas of lower accuracy in both manual and computerized FAs
were as follows: chin, ears, and nose, especially shape and orienta-
tion of the ears and chin shape. Wilkinson (14) states that the
mouth is the most difficult region to determine the morphology
and heavily relies on the artistic interpretation. Lee et al. (16), after
quantitative analyses of FA accuracy, also reported that the mouth
morphology is quite uncertain to estimate. This factor may have
influenced, even minimally, the final appearance of lips.
To the best of our knowledge, this is the first report to assess
different techniques of FA by means of recognition and resem-
blance tests. However, this study presents some limitations. The
first is the small sample of targets and assessors, which may not
represent real findings. Another limitation refers to the choice of
the different techniques to predict facial components. For this
study, we decided to choose just one set of comparisons (one
that we called AM and another entirely different—the CM) and
this choice was arbitrary. A future approach could involve the
comparison of a set of methods formed by different combina-
tions of techniques. So, by analyzing all the possible combina-
tions—for this study, for example, we would have 32 FAs—it
would be possible to point the best combination for FA. As
there are many other techniques to predict facial structures avail-
able in the literature, more combinations could be addressed,
although a bit impracticable.
The AM and the CM required approximately the same time to
be conducted, but the first author experienced less difficulty in
using the AM. As mentioned above in this section, methods
noted as more objectives are easier to be applied to 2D FAs.
Maybe facial structures (such as nose, mouth, and ears) could be
generated by those techniques and 3D printed for attachment to
manual reconstructions.
Acknowledgments
The authors thank the participation of all research volunteers.
They also thank the entire staff of the “Renato Archer”
Information Technology Center (CTI), Campinas-S~
ao Paulo,
Brazil, for the production of 3D prototypes used in this work.
References
1. Starbuck JM, Ward RE. The affect of tissue depth variation on craniofa-
cial reconstructions. Forensic Sci Int 2007;172(2–3):130–6.
2. Verze L. History of facial reconstruction. Acta Biomed 2009;80(1):5–12.
3. Prag J, Neave R. Making faces using forensic and archeological evi-
dence. London, U.K.: British Museum Press, 1997.
4. Rynn C, Wilkinson CM. Appraisal of traditional and recently proposed
relationships between the hard and soft dimensions of the nose in profile.
Am J Phys Anthropol 2006;130(3):364–73.
5. Stephan CN. Beyond the sphere of the English facial approximation liter-
ature: ramifications of German papers on western method concepts. J
Forensic Sci 2006;51(4):736–9.
6. Ullrich H, Stephan CN. On Gerasimov’s plastic facial reconstruction
technique: new insights to facilitate repeatability. J Forensic Sci 2011;56
(2):470–4.
7. Ullrich H, Stephan CN. Mikhail Mikhaylovich Gerasimov’s authentic
approach to plastic facial reconstruction. Anthropologie 2016;LIV(2):97–
107.
8. Stephan CN. Facial approximation—from facial reconstruction synonym
to face prediction paradigm. J Forensic Sci 2015;60(3):566–71.
9. Wilkinson C, Neave R. Skull re-assembly and the implications for foren-
sic facial reconstruction. Sci Justice 2001;41(3):233–4.
10. Stephan CN, Arthur RS. Assessing facial approximation accuracy: how
do resemblance ratings of disparate faces compare to recognition tests?
Forensic Sci Int 2006;159(1):159–63.
11. Stephan CN, Henneberg M. Recognition by forensic facial approxima-
tion: case specific examples and empirical tests. Forensic Sci Int
2006;156(2–3):182–91.
12. Wilkinson C, Rynn C, Peters H, Taister M, Kau CH, Richmond S. A
blind accuracy assessment of computer-modeled forensic facial recon-
struction using computed tomography data from live subjects. Forensic
Sci Med Pathol 2006;2(3):179–87.
13. Stephan CN, Cicolini J. Measuring the accuracy of facial approxima-
tions: a comparative study of resemblance rating and face array methods.
J Forensic Sci 2008;53(1):58–64.
14. Wilkinson C. Facial reconstruction –anatomical art or artistic anatomy?
J Anat 2010;216(2):235–50.
15. Lee WJ, Wilkinson CM, Hwang HS. An accuracy assessment of forensic
computerized facial reconstruction employing cone-beam computed
tomography from live subjects. J Forensic Sci 2012;57(2):318–27.
16. Lee WJ, Wilkinson CM, Hwang HS, Lee SM. Correlation between aver-
age tissue depth data and quantitative accuracy of forensic craniofacial
reconstructions measured by geometric surface comparison method. J
Forensic Sci 2015;60(3):572–80.
17. Fernandes CM, Serra Mda C, da Silva JV, Noritomi PY, Pereira FD,
Melani RF. Tests of one Brazilian facial reconstruction method using
three soft tissue depth sets and familiar assessors. Forensic Sci Int
2012;214(1–3):211.e1–7.
18. Stephan CN. Anthropological facial ‘reconstruction’–recognizing the
fallacies, ‘unembracing’the errors, and realizing method limits. Sci Jus-
tice 2003;43(4):193–200.
19. Stephan CN, Henneberg M, Sampson W. Predicting nose projection and
pronasale position in facial approximation: a test of published methods
and proposal of new guidelines. Am J Phys Anthropol 2003;122(3):240–
50.
20. Guyomarc’h P, Stephan CN. The validity of ear prediction guidelines
used in facial approximation. J Forensic Sci 2012;57(6):1427–41.
21. Stephan CN, Henneberg M. Building faces from dry skulls: are they rec-
ognized above chance rates? J Forensic Sci 2001;46(3):432–40.
22. Hayes S, Taylor R, Paterson A. Forensic facial approximation: an over-
view of current methods used at the Victorian Institute of Forensic Medi-
cine/Victoria Police Criminal Identification Squad. J Forensic
Odontostomatol 2005;23(2):45–50.
23. Quatrehomme G, Balaguer T, Staccini P, Alunni-Perret V. Assessment
of the accuracy of three-dimensional manual craniofacial reconstruction:
a series of 25 controlled cases. Int J Legal Med 2007;121(6):469–75.
24. Richard AH, Parks CL, Monson KL. Assessment of presentation meth-
ods for ReFace computerized facial approximations. Forensic Sci Int
2014;242:283–92.
25. Herrera LM, Strapasson RA, da Silva JV, Melani RF. Forensic facial
approximation assessment: can application of different average facial
6JOURNAL OF FORENSIC SCIENCES
tissue depth data facilitate recognition and establish acceptable level of
resemblance? Forensic Sci Int 2016;266:311–9.
26. Santos WDF. Mensurac!
~
ao de tecidos moles da face de brasileiros vivos
em imagens multiplicares de Resson^
ancia Magn!
etica Nuclear (RMN)
para fins m!
edico-legais [thesis]. Ribeir~
ao Preto, Brazil: Universidade de
S~
ao Paulo, 2008.
27. Taylor KT. Forensic art and illustration. New York, NY: CRC Press,
2001.
28. Wilkinson CM, Mautner SA. Measurement of eyeball protrusion and its
application in facial reconstruction. J Forensic Sci 2003;48(1):12–6.
29. Whitnall SE. The anatomy of the human orbit and accessory organs of
vision. London, U.K.: Oxford Medical Publications, 1921.
30. Stephan CN, Henneberg M. Predicting mouth width from inter-canine
width –a 75% rule. J Forensic Sci 2003;48(4):725–7.
31. Rynn C, Wilkinson CM, Peters HL. Prediction of nasal morphology from
the skull. Forensic Sci Med Pathol 2010;6(1):20–34.
32. Stephan CN. Position of superciliare in relation to the lateral iris: testing
a suggested facial approximation guideline. Forensic Sci Int 2002;130
(1):29–33.
33. Stephan CN, Cicolini J. The reproducibility of facial approximation accu-
racy results generated from photo-spread tests. Forensic Sci Int 2010;201
(1–3):133–7.
34. Liang KY, Zeger SL. Longitudinal analysis using generalized linear
models. Biometrika 1986;73:13–22.
35. Siegel S, Castellan NJ Jr. Nonparametric statistics for the behavioral
sciences. New York, NY: McGraw-Hill, 1988.
36. Brunner E, Langer F. Nonparametric analysis of ordered categorical data
in designs with longitudinal observations and small sample sizes. Bio-
metr J 2000;42:663–75.
37. Vanezis P, Blowers RW, Linney AD, Tan AC, Richards R, Neave R.
Application of 3-D computer graphics for facial reconstruction and com-
parison with sculpting techniques. Forensic Sci Int 1989;42(1–2):69–84.
38. Miyasaka S, Yoshino M, Imaizumi K, Seta S. The computer-aided facial
reconstruction system. Forensic Sci Int 1995;74(1–2):155–65.
39. Vandermeulen D, Claes P, Loeckx D, De Greef S, Willems G, Suetens
P. Computerized craniofacial reconstruction using CT-derived implicit
surface representations. Forensic Sci Int 2006;159(1):S164–74.
40. Claes P, Vandermeulen D, De Greef S, Willems G, Clement JG, Suetens
P. Computerized craniofacial reconstruction: conceptual framework and
review. Forensic Sci Int 2010;201(1–3):138–45.
41. Decker S, Ford J, Davy-Jow S, Faraut P, Neville W, Hilbelink D. Who
is this person? A comparison study of current three-dimensional facial
approximation methods. Forensic Sci Int 2013;229(1–3):161.e1–8.
42. Tedeschi-Oliveira SV, Beaini TL, Melani RF. Forensic facial reconstruc-
tion: nasal projection in Brazilian adults. Forensic Sci Int 2016;266:123–
9.
Additional information and reprint requests:
Lara Maria Herrera, M.Sc.
Department of Community Dentistry
Araraquara School of Dentistry
S~
ao Paulo State University
Rua Humait!
a, 1680
Araraquara
SP 14801-903
Brazil
E-mail: laraherrera0@gmail.com
HERRERA ET AL. .APPROACHES OF FACE PREDICTION 7
