Accepted Manuscript

Faces of Homo floresiensis (LB1)

Susan Hayes, Thomas Sutikna, Mike Morwood

PII: S0305-4403(13)00234-3
DOI: 10.1016/j.jas.2013.06.028
Reference: YJASC 3743

To appear in: Journal of Archaeological Science

Received Date: 1 February 2013

Revised Date: 30 May 2013
Accepted Date: 30 June 2013

Please cite this article as: Hayes, S., Sutikna, T., Morwood, M., Faces of Homo floresiensis (LB1),
Journal of Archaeological Science (2013), doi: 10.1016/j.jas.2013.06.028.

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to
our customers we are providing this early version of the manuscript. The manuscript will undergo
copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please
note that during the production process errors may be discovered which could affect the content, and all
legal disclaimers that apply to the journal pertain.
 ACCEPTED MANUSCRIPT

Highlights:

• The evidence on which our approximation of LB1’s facial appearance is based

• How our facial approximation compares to other LB1 faces
• Some previous faces bestowed upon LB1 do not appear to accord with the skeletal
evidence

PT
RI
U SC
AN
M
D
TE
C EP
AC
 ACCEPTED MANUSCRIPT
1

Title: Faces of Homo floresiensis (LB1)

Authors: Susan Hayes1*, Thomas Sutikna1,2, Mike Morwood1

Author Affiliations:
1. Centre for Archaeological Science, Faculty of Science, University of Wollongong

PT
2. Indonesian National Centre for Archaeology

*Corresponding author: Susan Hayes

RI
Phone: +61 2 4298 1556
Fax: +61 2 4221 4650
Email: [email protected]

U SC
AN
M
D
TE
C EP
AC
 ACCEPTED MANUSCRIPT
2

1. Introduction
Since 2004 the partial skeleton of the holotype of Homo floresiensis, Liang Bua 1 (LB1)
(Brown, et al., 2004, Morwood, et al., 2005), has been given a number of different faces
using a method popularly referred to as ‘facial reconstruction’. We, along with others (e.g.

PT
George, 1987, Reichs and Craig, 1998, Stephan, 2005, Taylor and Angel, 1998), prefer the
term ‘facial approximation’, both because it is a more accurate description of the results, and

RI
to distinguish the process from skull reconstruction following taphonomic and/or excavation
damage, as is the case here. Regardless of nomenclature, however, this method of depicting

SC
facial appearance is (or at least can be) derived from known relations between the
anatomically modern skull and its soft tissues. Not all of these relations have been verified
scientifically, and those that are verified tend to be statistical averages, so the results can

U
never be a definitive depiction of a unique face. Therefore, as the results are always
approximate, it is only to be expected that facial appearances given to LB1 will differ – and
they do.
AN
M
Here we approximate the facial appearance of LB1 using 2D computer-graphic facial
approximation techniques which have been previously applied to the archaeological remains
D

of anatomically modern humans (Hayes, 2011, Hayes, et al., 2012), and referring to the
relevant findings following nearly a decade of research into this small bodied hominin. We
TE

then apply the geometric morphometric (shape) analysis tool, morphologika2 (v2.5)
(O'Higgins and Jones, 2006) to compare our results with pre-existing faces of LB1 from
EP

Europe (France, Holland, Spain), North America, Australia and Japan (with the images
sourced from Anton, 2012, Balter, 2009, Carr, 2012, Davis and Deak, 2010, Daynès, 2008,
Hall, 2010, Kemp, 2004, Roberts, 2011, Sawyer and Deak, 2007). The results of these
C

analyses indicate that our facial approximation depicts a wider and shorter face, and a
AC

comparatively more modern human nasal morphology. This may be due to our use of facial
approximation methods derived from anatomically modern humans, and/or that many of the
faces given to LB1 include aspects of what appears to be a Pan-like facial morphology,
particularly in regards to upper lip height, even though this is not strongly supported by the
skeletal evidence.

2. Methods and Results

 ACCEPTED MANUSCRIPT
3

Our facial approximation is primarily based on verified, peer reviewed research regarding the
relationship between the skull and its soft tissues. These relationships are derived from
studies of living and deceased modern humans, and are clearly limited in their application to
archaic hominins. However, in the absence of any verified knowledge regarding the likely
facial appearance of archaic hominins, our method does have a distinct advantage in being
evidence-based, and therefore able to be subject to ongoing methodological development,

PT
informed critique and refinement.

RI
The results of a facial approximation are a series of illustrations (see Figures 1 and 2).
Therefore, we present the methods and results as a synthesis, and discuss where these

SC
methods are particular to the evidence provided by the LB1 mandible and cranium, and,
where relevant, why we have chosen one particular method to approximate an aspect of the
face over other published recommendations.

U
2.1 Skull Reconstruction
AN
The LB1 mandible and cranium display a combination of antemortem asymmetry (Baab and
M
McNulty, 2009, Kaifu, et al., 2009, Kaifu, et al., 2010, McNulty and Baab, 2010) and
taphonomic distortions, with further inadvertent damage and distortion occurring both during
D

and after excavation (Brown, et al., 2004, Brown and Maeda, 2009, Brown, 2012, Falk, et al.,
2010, Morwood and van Oosterzee, 2007). A virtual reconstruction was undertaken to best
TE

approximate the original form of both the mandible and aspects of the cranium, but limited to
only those areas of bone that could be justified through information present in preserved areas
EP

of the skull. Our reconstruction of the cranium, therefore, is predominantly the zygomatic and
orbital bones, and does not include, for example, the fronto-glabella region, the nasal bones
or the sub-nasal region of the maxilla (see Table 1 and Sections 2.2, 2.4.2 and 2.4.3 and
C

regarding our approximation of the craniofacial landmarks and the soft tissues relating to
AC

these areas).The reconstructed areas of bone are shown shaded in the frontal view displayed
in Figure 1b, and described in 2.1.1 and 2.1.2 below.

2.1.1 Reconstruction of the mandible

Taphonomic processes contributed to the mandible being broken at the junction of the right
second and third molars, right premolar, and left incisor/canine junction, and the left condyle
is incomplete (Brown and Maeda, 2009). In addition there is lateral twisting to the right
 ACCEPTED MANUSCRIPT
4

mandibular corpus and ramus, which could be due to antemortem positional plagiocephaly
(Kaifu, et al., 2009) rather than taphonomy. Either way, further damage was inflicted post-
excavation, causing an increase in bigonial breadth and an increase in the lateral twisting of
the right corpus and ramus, both of which result in the mandible being unable to be
articulated with either the mandibular fossae or the patterns of dental occlusion (Brown and
Maeda, 2009, Brown, 2012).

PT
In order to locate each condyle within its corresponding fossa, CT scan data (1mm slices) of

RI
the LB1 cranium with the mandible articulated to best fit with the right condyle and right
dental occlusion patterns, were uploaded into Amira 5.2.0, and the skull orientated to the

SC
right Frankfurt Horizontal Plane (FHP) using the right external auditory meatus and inferior
orbital rim. Orthogonal images of the frontal and lateral views were captured, with distortion
to the mandible resulting in the left condyle sitting approximately 10mm below the left

U
mandibular fossa. A high resolution stereolithography cast of the LB1 cranium and mandible

AN
was articulated to best left premolar and molar occlusion when the left condylar process is
orientated within its mandibular fossa, which is located more anteriorally than the right fossa
by approximately 6mm. This articulation results in the left mandibular body sitting at an
M
elevation of 11° to the right FHP. Using Amira 5.2.0, the skull was re-orientated to this left
mandibular elevation, and orthogonal images of the left lateral and frontal view of the
D

mandible were captured. All images were uploaded into Adobe Photoshop CS4 (64bit), and
scaled to life size. The frontal view of the elevated left mandible was extracted from the
TE

midline and relocated within the frontal view of the cranium orientated to the right FHP.
Similarly, the elevated left mandible was extracted from the left lateral view and orientated
EP

within the left mandibular fossa, again with the cranium orientated to the right FHP (see
Figure 1c). This results in a frontal view of the skull with both condylar processes of the
mandible located within their mandibular fossa, and in reasonable accordance with both left
C

and right dental occlusion (see Figure 1b).

AC

Our reconstruction of the mandible is approximate. Although rearticulating the mandible by

separating it at the midline allows for reasonable occlusion and condylar articulation within
very differently orientated mandibular fossae, other reconstructions are possible. For
example, it is not evident to what extent this articulation accords with the antemortem
disharmonic occlusal relationship identified by Kaifu et al. (2009), and as this reconstruction
results in the right and left gonion being symmetrically located despite the rami displaying
 ACCEPTED MANUSCRIPT
5

clear morphological differences (refer Table 1 and Figure 1b). Therefore, it is likely that
aspects of antemortem, as well as postmortem, mandibular asymmetry have been obscured.

2.1.2 Reconstruction of the zygomatic and orbital bones

Overall cranial distortion is more clearly visible when viewed in the transverse plane (i.e.
above or below), forming what Kaifu et al. (2009) describe as horizontal parallelogram

PT
skewing, and which they consider to be symptomatic of antemortem posterior deformational
plagiocephaly (PDP). This cranial asymmetry is still, however, within the normal range for

RI
modern humans, extant African apes, and aspects of asymmetrical skewing is a common
taphonomic process borne by all hominin fossils (Baab and McNulty, 2009). Excavation

SC
damage to the cranium includes inadvertent removal of part of the bregma and left frontal
bone, all of the left supraorbital rim, the glabella region, nasal bones and subnasal part of the
maxilla, and all of the upper incisors. Taphonomic distortion to the right zygomatic arch

U
includes being broken anteriorally and depressed medially, and while the left zygomatic is

AN
also broken, it has been reattached (Brown, et al., 2004, Falk, et al., 2010, Kaifu, et al., 2011).

Comparison of the superior view of the direction and shape of the undamaged section of the
M
right zygomatic root with the left zygomatic root and arch suggests the most lateral point on
the right zygomatic arch is on a par with the maximum protrusion of the inflated right
D

supramastoid region. This zygomatic landmark (refer Table 1) and the morphology of the
undamaged borders of the left zygomatic bone were used to orientate a reflected section of
TE

the left zygomatic, taken from the commencement of the angle of the left temporal margin.
Best fit at the angle of the right temporal margin of the right zygomatic bone involved a 6°
EP

medial rotation. The left temporal margin of the left zygomatic bone was approximated using
a reflected corresponding section of the right zygomatic bone, taken from 4mm superior to
the commencement of the angle of the temporal margin of the right zygomatic bone. This
C

segment needed to be rotated medially by 1.2° to fit with the undulation of the left marginal
AC

tubercle. Again, as with our reconstruction of the mandible, it is likely that this method
reduces LB1’s antemortem as well as postmortem bizygomatic facial asymmetry. However,
as mentioned, LB1’s overall cranial asymmetry is more clearly visible in transverse planes,
and is far less apparent when the skull is viewed from the front (though see Section 2.6).

The left superior orbital rim is largely absent, although there is a small section of preserved
bone evident on a 3D print produced from the original set of LB1 CT scans, which captured a
 ACCEPTED MANUSCRIPT
6

somewhat less damaged skull (refer Morwood and van Oosterzee, 2007), and this appears to
be a fragment of the left supraorbital notch. With the virtual skull orientated in the frontal
view, a section of the right superior rim (from right supraorbital notch to the frontozygomatic
suture) was extracted, reflected, and warped to best fit with the arc indicated by the remains
of the superior orbital shape and the preserved bone fragment. This results in a reconstruction
of the left orbit where the left superior orbital rim is 3.8mm lower than the right (see Figure

PT
1b), which may be in keeping with the overall pattern of asymmetry displayed by the LB1
remains, and is clearly discernible when the skull is viewed from the front.

RI
SC
2.2 Soft Tissue Depths
The soft tissue depths (STDs) applied in this facial approximation are from Stephan and
Simpson (2008) (refer Table 1and Figures1a-c). Over the past 130 years many STD studies

U
have been undertaken and published using a variety of methods (e.g. cadaveric needle-

AN
punctures, radiographs, ultrasound, CT scans), typically with a focus on a particular
population (‘race’) and sub-divided by sex, age, body mass and/or tissue depth range. Two of
the more cited studies include Rhine and colleagues’ African American (N=69) and European
M
(N=63) cadaveric needle-puncture “tissue thicknesses” (reproduced in Taylor, 2001), and
Helmer’s adult, white European (N=124) ultrasound “tissue measurements” taken from living
D

subjects (reproduced in Wilkinson, 2004). As Stephan and Simpson (2008) note, there are a
few problems with STD data collection, including ad hoc assignations of population affinity,
TE

age and body mass; variation in landmarks, their definitions, actual identification and
measurement; and, typographic errors in data set reproductions. Referring to previous studies,
EP

Stephan and Simpson report a typical STD measurement error rate of 10%, which, when
combined with other influences (e.g. methods used, physical identification of landmarks),
produces what they conservatively estimate to be a typical error rate in excess of 2mm.
C

Stephan and Simpson analysed 25 commonly measured landmarks from 55 of the more
AC

reliable studies of adult STDs and found that differences arising from publication year,
method of measurement, population affinity and sex had little practical value as these
differences were consistently less than the conservative STD measurement error value.
Therefore, Stephan and Simpson created a pooled STD data collection, comprising total
weighted means for each landmark, and a subset of weighted means from the studies
reporting standard deviations. The main advantages of the Stephan and Simpson data set is
that it is more robust than individual studies, being derived from very large sample sizes
 ACCEPTED MANUSCRIPT
7

(average 3250 individuals, range 545-6786), and is broadly applicable across both sex and
population affinity.

With the exception of the right zygomatic (refer Section 2.1.2) and the subnasale (refer Table
1), each STD applied to LB1 arises from preserved bone. As mentioned in Section 2.1 we
have not reconstructed the fronto-glabella region, the nasal bones or the sub-nasal region of

PT
the maxilla, all of which were inadvertently removed during excavation. Our choices were,
therefore, to (i) guess the amount of bone removed, (ii) predict projection on the basis of a

RI
possible evolutionary affinity with a different hominin (e.g. Kaifu, et al., 2011), or, (iii) as we
have done, default to the remains of preserved bone, fully aware that this will underestimate,

SC
to an unknown degree, the depth of the flesh covering the glabella, nasion, subnasale, mid-
philtrum and prosthion. However, as all of the underestimated STDs are located on the
cranial midline, the actual depths at these underestimated landmarks are not apparent in a

U
frontal 2D facial approximation (refer Figures 2b and 2c).

AN
As can be seen, each STD displays two depths (Figures 1a-1c). The greater STDs refer to
Stephan and Simpson’s (2008) total weighted means, while the lesser depths are their subset
M
of weighted means, minus one standard deviation (s). We chose STD(-1s) to approximate
LB1’s facial appearance because she is a small-bodied hominin, estimated to be only 106cm
D

in height (Brown, et al., 2004). This does, however, assume an unverified positive correlation
between adult stature and STD1. Both depths are illustrated, and the difference in facial
TE

outline is shown in the right lateral view (Figure 1a). As can be seen, both of the right lateral
facial outlines appear to suggest LB1 has an underbite. As this is very likely an artefact of our
EP

decision to default to STD underestimation, this characteristic is also not emphasised in the
frontal facial appearances (Figures 2b and 2c).
C
AC

2.3 Underlying Anatomy

The lateral and frontal build up of the virtual facial muscles and parotid glands (see Figures
2a and 2b) are based on those initially developed and described for a computer graphic facial
approximation of early Maori remains (Hayes, et al., 2012), which was achieved in
collaboration with two experienced anatomists, and subsequently applied to other
anatomically modern humans. Every application of this suite of virtual anatomy involves
warping each muscle and gland to best fit with the particular morphology of a particular
 ACCEPTED MANUSCRIPT
8

skull, and (where present) the indications of muscle attachments. For this approximation of
LB1 the depressor labii inferioris has been further modified to form a better agreement with
the descriptions provided by Standring (2008). However, and as noted in Standring, the
presence, size and shape of many of these anatomical features vary considerably between
individuals, and our observation is that both anatomical descriptions and anatomical
depictions can also differ between publications, with the latter depending on the degree of

PT
stylisation, the dissector and/or the individuals dissected (e.g. Grant, 1943, Paff, 1973,
Romanes, 1986).

RI
2.4 Facial Features

SC
Approximation of the location, size and shape of the facial features (eyes, nose, mouth, ears)
and overall head shape are illustrated in Figures 1a and 1b, and again broadly follow the
computer graphic methods previously applied to anatomically modern humans (Hayes, 2011,

U
Hayes, et al., 2012). There are a number of published recommendations for approximating
facial features from the skull. Many of these recommendations, however, are one or all of the
AN
following: predominantly experience-based and/or published for a popular audience (e.g.
Gerasimov, 1971, Prag and Neave, 1997); lack the evidence from which they were derived
M
(e.g. Fedosyutkin and Nainys, 1993, Gerasimov, 1955); refer to artistic canons and heuristics
that anthropometric studies (e.g. Farkas and Munro, 1987, Farkas, 1994) largely discount;
D

and/or have since proven to be somewhat less than accurate (e.g. Krogman and Iscan, 1986).
Therefore we provide some explanation, where relevant, of our methodological choices for
TE

approximating the facial appearance LB1, acknowledging that many of these verified skull-
soft tissue relationships are derived from predominantly ‘white’ or ‘European’ population
EP

affinities, and that application of these relationships is still subject to some degree of
interpretation – both of the skull and of the methods themselves.
C

2.4.1 Eyes
AC

According to Wolff (1948) the average human eyeball is 23.5mm horizontally and 23mm in
height, whereas a recent CT scan study involving 375 individuals (204M and 171F, mean age
52.2 years) shows these average distances to be 24.3mm and 24.6mm respectively
(Guyomarc’h, et al., 2012). This approximation uses an average diameter of 24mm, which is
greater than Wolff’s recommendation, yet less than the unverified recommendations of 26mm
(Prag and Neave, 1997) and 25mm (Taylor, 2001, Wilkinson, 2004). A long-standing
forensic recommendation is to locate the eyeball centrally within the orbit (Gatliff and Snow,
 ACCEPTED MANUSCRIPT
9

1979, Krogman and Iscan, 1986). However, two cadaveric studies (Stephan and Davidson,
2008, Stephan, et al., 2009) support earlier anatomical claims that the human eyeball is
displaced from the orbital centre (e.g. Wolff, 1948). The combined results of these cadaveric
studies are that, on average, the eyeball is displaced 1.4 mm superiorly and 2.3 mm laterally
(Stephan, et al., 2009), with these results being further verified by the CT scan study referred
to above (Guyomarc’h, et al., 2012).

PT
Kaifu et al. (2011) comment that LB1 has a “unique, tall orbital shape”, and while are able to

RI
locate LB1’s eyeballs 1.4mm superior to their orbital centres, we found they could only be
reasonably displaced 1mm laterally. It is possible that LB1 had a smaller than average eyeball

SC
diameter to better fit with the orbital widths, however there is an experience-based
recommendation that orbital dimensions bear no relationship to eyeball size (Gerasimov,
1955), at least for anatomically modern humans.

U
AN
With regards to eyeball projection, this has been traditionally located projecting to the
maximum extent of the superior and inferior orbital rims (Krogman and Iscan, 1986). Two
studies, however, show eyeball projection exceeds this point. One study (Stephan, 2002)
M
reviewed published exophthalmometric measures, which included MRI measures (n=79), and
reports that the modern human eyeball, on average, protrudes 3.7mm beyond the tangent
D

formed by connecting the superior and inferior orbital rims. A related study (Wilkinson, et
al., 2003a), examined 39 MRI scans of mature European adults (11M, 28F, 60-90yrs), and
TE

found average eyeball protrusion to be very similar (3.8mm). For this approximation right
lateral eyeball projection is approximately 3.75mm.
EP

Our estimation of the location of LB1’s palpebral fissures is based on a line connecting the
endocanthion and exocanthion. The malar tubercle is often cited as corresponding to the
C

location of the lateral canthal tendon (e.g. Fedosyutkin and Nainys, 1993, Gerasimov, 1955),
AC

which Stephan and Davidson (2008) verify as a consistent finding. For LB1the exocanthion
(lateral canthus) is located at the marginal tubercle, which appears present for both orbits,
while the endocanthion (medial canthus) is inserted into the remains of each lacrimal fossa
1mm below the exocanthion. Stephan and Davidson (2008) offer other orbital landmarks to
identify the average orientation of the endocanthion, however, as it has been observed, many
of LB1’s cranial sutures are difficult to identify even from CT scans (Brown, et al., 2004,
Kaifu, et al., 2011), and this includes the sutures required to identify the dacryon.
 ACCEPTED MANUSCRIPT
10

Other recommendations regarding eye morphology that have been applied in this
approximation include: the average iris diameter is 10-12mm (Larrabee and Makielski,
1993), with 10mm being selected for the surface result (see Figure 2c); the location of the
superior eyelid fold may correspond to the same height of the soft tissue nasion (George,
2007), the shape of the superior eyelid fold may be related to the shape of the superior orbital

PT
rim (Fedosyutkin and Nainys, 1993); and, dominant brow-ridges have been observed to
display thicker soft tissues and a soft transition over the ridges (Gerasimov, 1955).

RI
2.4.2 Nose

SC
Overall, our approximation of LB1’s nasal morphology is highly approximate. Although the
skeletal remains still include the general lower shape and maximum width of the aperture,
and the preserved remains of the ethmoid plate indicate approximate aperture height,

U
excavation damage has removed all accurate locations for the nasion, rhinion and subnasale,

AN
and the anterior edge of the right lateral wall of the aperture. Therefore, other than nasal
width and the general shape of the left nasal wing and nasal tip, different interpretations and
evidence to approximate these missing apsects is both possible and justifiable.
M

Gerasimov’s observation (1955, 1971) is the most well-known and applied skeletal method
D

for estimating the location of the nasal tip. That is, the nasal tip will be found where a line
continuing the general direction indicated by the nasal bones intersects with a line continuing
TE

the general direction indicated by the anterior nasal spine (e.g. Prag and Neave, 1997, Taylor,
2001, Wilkinson, 2004). Unfortunately, in addition to being experience, rather than evidence,
EP

based, this description of the method combines a misleading overgeneralisation with an

unfortunate mistranslation. According to one of Gerasimov’s students, ‘nasal bones’ more
specifically refers to the most anterior undulation of the nasal bones at the rhinion, and nasal
C

spine should instead refer to the right and left base of the nasal aperture, lateral to the nasal
AC

spine, and not the anterior nasal spine itself (Ullrich and Stephan, 2011).

Our approximation uses a subset of the regression equations developed by Rynn et al. (2010)
and derived from 79 North American CT scans and 60 European cephalograms (N=139) of
the adult (< 50 years) head and face and, as the authors note, ‘white’ ancestry dominates the
data (n=110). The results of this study also include some verification of Gerasimov’s other
 ACCEPTED MANUSCRIPT
11

observations regarding the relationship between aspects of the nasal aperture and the soft
tissues of the nose.

To estimate profile nasal dimensions, Rynn et al. (2010) use up to three cranial measures
involving the nasion, acanthion, rhinion and subspinale landmarks (see Table 1, and note that
subspinale is referred to here as subnasale). The regression equations used for our

PT
approximation of LB1’s nasal morphology are those that do not include the acanthion (as the
anterior nasal spine is missing), and are summarised as follows:

RI
• [(rhinion-subnasale 26.8mm)*0.83)] – 3.5 = 18.7mm anterior projection from the
nasion-prosthion plane (valid for all populations studied)

SC
• [(nasion-subnasale 43.4mm)*0.74)] + 3.5 = 34.7mm nasal length from the soft tissue
nasion to nasal tip (valid for ‘white’ populations only)
• [(nasion-subnasale 43.4mm)*0.63)] + 17 = 44.3mm nasal height from the soft tissue

U
nasion to subnasal (valid for female ‘white’ populations only)
• AN
[(rhinion-subnasale 26.8mm)*0.5)] + 1.5 = 15.7mm nasal depth from the nasal tip to
soft tissue subnasal (valid for female ‘white’ populations only).
M
As detailed in Table 1 we approximate the rhinion (end of nasals) in the frontal view through
reference to the preserved remains of the ethmoid plate and a pooled average nasal aperture
D

shape (McDowell, et al., 2012), which is illustrated in Figure 1b. The rhinion, however, is
TE

only used for the Rynn et al. (2010) equations referring to anterior nasal projection and nasal
depth, and not to give shape to the upper nose in profile. As can be seen (Figure 1a), instead
of following the undulation of the nasal bones (which were accidentally removed on
EP

excavation), our depiction of the nasal profile runs straight from nasion to nasal tip. Also
detailed in Table 1, we identify the nasion in reference to Kaifu et al. (2011), and the
subnasale is located 1.5mm superior to the preserved section of the right central incisor fossa.
C
AC

In addition to deriving the regression equations listed above, Rynn et al. (2010) also add
verification to Gerasimov’s (1955, 1971) observations that maximum nasal aperture width is
very close to 2/3 the maximum width of the nasal wings, which for LB1 results in a
maximum nasal width of 35mm. Other observations of Gerasimov that Rynn et al. add
verification to include: that in profile the maximum curve of the nasal wing is approximately
6mm anterior to the most posterior point on the lateral wall of the aperture; maximum nasal
wing height closely corresponds to the christa conchalis on the lateral wall of nasal aperture;
 ACCEPTED MANUSCRIPT
12

maximum nasal wing depth is 4mm lower than the lowest point on the curve of the lateral
aperture border; that the shape of the alar groove generally correlates to the lateral curve of
the aperture, and, overall, that individuals displaying ‘angular’ nasal apertures tend to have a
pointed or angled nasal tip, and ‘rounded’ apertures, rounded tips.

For our approximation of LB1 the shape of the nasal wings and nasal tip are taken to have a

PT
close, rather than more general, correspondence with the shapes of the lateral curves of the
nasal aperture. Although such a literal interpretation of Rynn et al. (2010) is not the published

RI
recommendation and unlikely to be accurate, it does remove the need for a more subjective
interpretation which would be required to produce a somehow more analogous shape

SC
relationship. In addition, and because of excavation damage to the nasal aperture, our capture
of the shape of the right lateral curve of the nasal aperture is more informed by the general
arc of what remains of the lateral wall. Excavation damage also makes it impossible for us to

U
apply a different, verified hard/soft tissue relationship involving the curvature of the superior

AN
nasal bones and the curvature of the nasal tip (Davy-Jow, et al., 2012). A further point is that
in our approximation of LB1’s nasal tip, rather than Rynn et al.’s (2010) reading of
Gerasimov as referring to the anterior nasal spine, we estimate nasal tip direction following
M
the translation supplied by Ullrich and Stephan (2011), and identify nasal tip direction from
two sagittal CT slices taken lateral to the nasal spine (see Figure 1c). As can be seen, this
D

tangent passes through the centre of the nasal tip close to the approximated maximum nasal
length (34.7mm).
TE

2.4.3 Mouth
EP

For many years mouth width was taken to correspond to interpupillary width (Krogman and
Iscan, 1986) and/or at some point radiating out from the canine/premolar junction (Krogman
and Iscan, 1986, Wilkinson, et al., 2003b), while Gerasimov’s observation is that mouth
C

width corresponds to the inter-premolar width of the maxilla (1955). Subsequent research
AC

(Stephan, 2003) has shown that in living subjects the interpupillary rule overestimates
European mouth width (n=61), though is a reasonable approximation for a smaller group of
South East Asian subjects (n=27), while the canine/premolar junction rule consistently
underestimates mouth width for both groups. Stephan’s findings are that the medial border of
the iris is a closer approximation of mouth corner location, but a more consistent guideline
for all of the populations he studied is the following: (intercanine width) +
0.57*[(interpupillary width) – (intercanine width)]. More recently, Stephan and Murphy
 ACCEPTED MANUSCRIPT
13

(2008) further explored the results reported by Song et al. (2007), who’s cadaveric study
included an anthropometric analysis of the relationship between soft tissue landmarks and
skeletal foramina. Song et al. examined 50 embalmed cadavers (32-101 years) of Korean
population affinity, and found: the distance between the infraorbital foramina is close to
identical to mouth width; for approximately half the individuals examined the infraorbital
foramen and cheilion are located on the same saggital plane; and, that the distance between

PT
the cheilion and the mental foramen is, on average, approximately 20mm (mean 20.9 mm, s
3.8mm). Stephan and Murphy (2008) dissected nine cadavers of European population affinity

RI
(62-74 years), and their results add general support to that of Song et al. (2007).

SC
For this facial approximation LB1’s mouth corners are located on the same sagittal plane as
the infraorbital foramina, and the resulting visible mouth width (43.7mm) is very close to the
approximation arising from Stephan’s algorithm reported above (i.e. 39mm +

U
[0.57*(48.5mm-39mm)] = 44.14mm). Interestingly, however, our mouth width does not

AN
agree with a simplified algorithm developed by Stephan and Henneberg (2003), who show
intercanine width to be approximately 75% of mouth width – which in this instance results in
a mouth that is 8mm wider, and were it implemented would result in a mouth that is even
M
wider than the interpupillary rule.
D

Transverse location of the oral fissure typically follows the recommendation of being within
the lower third of the central incisor (Taylor, 2001), which is similar to the anatomical
TE

location of being slightly superior to the incisal edges of the anterior maxillary teeth
(Standring, 2008). Location of the oral fissure for LB1 was taken to be slightly superior to the
EP

inferior edge of what remains of the dentine of the upper right canine, which is in accordance
with the more general anatomical location, and is also approximately 20mm from each of the
mental foramen (i.e. the average distance reported by Song et al. (2007)). Vermilion heights
C

have also been examined in relation the height of the dentine of the central incisors. Digital
AC

calliper measures were taken of living subjects (N=95) (Wilkinson, et al., 2003b), and the
following formulae derived for ‘white’ Europeans:
• upper vermilion height = (0.6*upper central incisor height) + 0.4
• lower vermilion height = (0.4*lower central incisor height) + 5.5
Again due to the absence of the superior central incisors, upper vermilion height was
calculated using the height of the dentine of the upper right canine (7.7mm) and lower
vermilion height was calculated from the lower lateral right incisor (7.6mm) as it appears to
 ACCEPTED MANUSCRIPT
14

display comparatively less tooth wear. This results in our approximation displaying an upper
vermilion height of 5mm and a lower vermilion height of 8.5mm, which is close to the
average lower vermilion height reported in this study (8.7mm), but nearly 1mm less than the
average upper vermilion height (5.8mm). This could be because, in addition to our use of
upper canine dentine, and not upper central incisor, Wilkinson et al.’s (2003b) formulae are
derived from extant Europeans, who presumably did not display extensive tooth-wear –

PT
which is a marked feature of LB1 (Brown and Maeda, 2009).

RI
Other recommendations for the mouth and used in our facial approximation of LB1 are that
prominent canines may suggest lateral vermilion fullness (Wilkinson, 2004), and that philtral

SC
width may be related to the width of the central incisors (Taylor, 2001), which in this instance
was approximated as possibly corresponding with the centre of the upper central incisor roots
rather than the teeth themselves.

U
2.4.4 Ears AN
Recommendations to predict the shape, dimensions and angle of orientation of the ear from
the skull have been evaluated by comparing each with both a CT scan analysis and a raft of
M
published anthropometric data (Guyomarc’h and Stephan, 2012). The results of this study
discredit all recommendations, and in some detail. Guyomarc’h and Stephan conclude that all
D

non-Asian individuals should be approximated with free earlobes (as opposed to attached);
that reliable anthropometric mean ear angles and mean ear width, such as those listed in
TE

Farkas (1994) are preferable, and that ear height is best approximated using their regression
equation, derived from measurements taken from European CT scans (N=78, 43M, 35F; 18-
EP

84 years):
ear length = (4.85*sex) + (0.10*age) + 54.95, where male = 1, female = 0.
What this indicates is that males tend to have an ear that is 4.85mm longer, and that this
C

regression usefully incorporates adult age-related changes to ear length. However, the need to
AC

enter a numerical age into the equation renders application impractical when age is only
generally understood. LB1, for example, is estimated to be adult (Brown and Maeda, 2009),
which is a description spanning decades. Therefore we take ear height to be equivalent to the
soft tissue distance subnasal-menton (50mm), an anthropometric relationship described in
Farkas (1994), and orient the tragus in its relation to the external acoustic meatus (Romanes,
1986). All other aspects of ear morphology depicted in our facial approximation are wholly
subjective (refer Figure 2c).
 ACCEPTED MANUSCRIPT
15

2.4.5 Head shape

Upper head shape (Figure 2) is largely determined by the arc of the cranium, and the
foreshortened forehead height is due to LB1 having a low cranial vault (Brown, et al., 2004).
In the mid-face the maximum width of the cheeks corresponds to the underlying bony
bizygomatic width plus STD(-1s), and upper cheek protrusion and shape follows the

PT
protrusion and shape of the zygomaxillary bones. Bigonial width and overall jaw shape
follows that indicated by the approximately reconstructed mandible plus STD(-1s), and as

RI
Homo floresiensis is without a bony chin (Brown, et al., 2004) we assume this also holds true
for the fleshed face.

SC
2.5 Surface Appearance

U
The general consensus is that human skin is darker at the equator than at greater latitudes

AN
(Barsh, 2003), and Liang Bua (Flores, Indonesia) is located at 8.6°S. Research concerning the
malanocortin 1 receptor (MC1R) locus indicates Homo has been hairless for at least 1.2
million years (Rogers, et al., 2004), which has congruence with the ectoparasite theory, where
M
a lack of thermal fur has an evolutionary advantage in reducing parasites (e.g. lice and fleas)
(Rantala, 2007). We therefore approximate LB1 with fairly dark skin and lacking facial fur.
D

The cranium does not provide indicators regarding head hair colour or shape, and in all of the
TE

previous published research on which our method of approximation is based it has been
argued that baldness is preferable in an evidence-based archaeological facial approximation
EP

(Hayes 2011; Hayes, et al. 2012; Hayes, et al. 2009). Hairless results, however, are typically
interpreted as male (SH, personal observation), and LB1 is most likely female (Brown, et al.,
2004, Brown and Maeda, 2009, Brown, 2012). Therefore the likelihood that LB1 bore head
C

hair is a more reasonable assumption than that she was bald2. As can be seen (Figure 2c), we
AC

approximate an appearance of wet hair to allow for a more indefinite shape, length and
texture, and to avoid an unjustifiable and yet enduring tendency to depict all early hominins,
including anatomically modern humans, as ungroomed and unkempt (Berman, 1999, Kemp,
2004).

2.6 LB1 facial variance

 ACCEPTED MANUSCRIPT
16

A 2D geometric morphometric analysis of an international selection of pre-existing faces of

LB1 (n=9, with 2D images sourced from Anton, 2012, Balter, 2009, Carr, 2012, Davis and
Deak, 2010, Daynès, 2008, Hall, 2010, Kemp, 2004, Roberts, 2011, Sawyer and Deak, 2007)
together with our facial approximation (N=10) was undertaken using the shape analysis tool,
morphologika2 (v2.5) (O'Higgins and Jones, 2006). The 46 homologous soft tissue facial
landmarks applied to these images are based on prior research exploring measurement of

PT
facial morphology in 2D images (Hayes, 2009), which has been further refined to explore
patterns of distortion arising from mimetic portraiture (Hayes and Milne, 2011) and actual

RI
police witness descriptions of suspects (Hayes and Tullberg, 2012). In all of these previous
analyses, the first Principal Component (PC), which captures the greatest variance between

SC
the individuals comprising the data set, concerned variations in head pose. Because we
include Procrustes registration in our analysis of these LB1 faces (i.e. the landmarks are
scaled, and rotated, for comparable fit), head canting (leaning the head towards one shoulder)

U
does not impact on facial shapes, but head turning alters the horizontal dimensions of the face

AN
(and in particular cheek widths), and head pitching results in both vertical and horizontal
changes (i.e. an upwards pose will contract the upper head, widen the jaw, give the
appearance of an upturned nose and downturned mouth, and increase chin height).
M

As can be seen from the Procrustes registered wireframes of each of the LB1 images we
D

analyse (Figure 3), there is some variation in head pose. Six LB1 faces are frontally
orientated (Shapes 1, 2, 5, 7, 8, 10) with one of these displaying an upwards pitch (Shape 1),
TE

three display a head turn to the right shoulder (Shapes 3, 4, 9), and one has a slight head turn
to the left (Shape 6). Furthermore, six of the images are photographs of 3D sculptures taken
EP

from unknown, and likely varying distances (Shapes 2, 3, 4, 6, 7, 8), and therefore, in
addition to the small sample size, our analyses should only be read as indicative of depicted
LB1 face shape variance.
C
AC

When all of the LB1 faces (N=10) are analysed, the thin plate spline (TPS) deformation grids
show, as with previous studies, that the greatest variance is due to head pose (Figure 4: PC1
54.6%). The second greatest variance (PC2 16.6%) separates out the face shapes by relative
eye roundedness, nose width, upper lip height (subnasal to superior vermilion border), and
mouth width. Overall, PC2 indicates some LB1 depictions tend towards displaying LB1 with
rounded eyes, short, narrow noses, narrow vermilion zones of the lips within wide mouths,
and relatively narrow and long faces, with facial length being strongly influenced by upper
 ACCEPTED MANUSCRIPT
17

lip height. The extreme of this group is the very first face given to LB1 (Shape 9 reproduced
in Kemp, 2004), but it is possible that the variance displayed in PC2 is potentially
compromised by including aspects of head pitch (Shape 1).

A second geometric morphometric analysis involving only the frontally orientated LB1 faces
(n=5: Shapes 2, 5, 7, 8 and 10) does not capture variance related to head turning and pitch,

PT
and even without the prototype LB1 face (Shape 9), the results still show a tendency for LB1
faces to be narrow and display a marked upper lip height (Figure 5: PC1 38.1% variance).

RI
The next largest variance (PC2: 30.3%) concerns relative fullness of the vermilion zone of
the lips, eye size, and that Shape 7 (reproduced in Roberts, 2011) depicts a face where

SC
excavation damage to the LB1 right orbital bones may possibly have been interpreted by the
creators as an antemortem injury/pathology affecting both the size and orientation of the right
palpebral fissure.

U
AN
The morphology of the mean LB1 face in this analysis (inset lower left Figure 5) and the
pattern of distribution across PC1 and PC2 (capturing 68.4% of the variance) suggests our
facial approximation of LB1 (Shape 5) is quite different from all of the other LB1 facial
M
shapes. When the Procrustes registered wireframes are extracted from morphologika2 prior to
running the PC analysis (inset lower right Figure 5), the wireframe of our facial
D

approximation displays the widest face, the greatest hemiface asymmetry (the right mid-face
and jaw are distinctly wider than the left), and the shortest upper lip distance.
TE
EP

3. Discussion
The first stage in our facial approximation of LB1 was to reconstruct the mandible and
cranium so as to best approximate its form prior to taphonomic, excavation and post-
C

excavation damage. The actual extent of taphonomic and excavation damage is unknown, and
AC

therefore the remains are open to different reconstructions. For our facial approximation of
LB1 we have applied predominantly robust relationships between the skull and its soft
tissues, and all of the relationships that form our methodology are derived from studies of
anatomically modern humans (Sections 2.2, 2.3 and 2.4.1-2.4.4). The applicability of these
methods to the remains of archaic hominins is limited. LB1 is the holotype specimen of
Homo floresiensis, and while sharing ancestry with Homo sapiens, LB1 is not an
anatomically modern human. To test our results we conducted a geometric morphometric
 ACCEPTED MANUSCRIPT
18

analysis including nine other pre-existing LB1 faces. Because of the influence of head pose
and perspective on facial shapes, and the small number of faces analysed, these analyses are
generally indicative rather than conclusive, and suggest our approximation of LB1’s facial
appearance displays a greater hemiface asymmetry, a proportionally wider face, and a shorter
upper lip (subnasale-vermillion border), than most of the faces given to this archaic hominin.

PT
Our reconstruction of the mandible involved splitting the body at the midline and re-
articulating the left ramus and mandibular body so that both condyles were located within

RI
their mandibular fossae and in reasonable accordance with dental occlusion patterns (Section
2.1.1). Reconstruction of the zygomatic bones and right orbital bones was achieved through

SC
reflection, but with the reflected fragments manipulated so as to agree with information
retained within relevant areas of preserved bone, such as the right zygomatic root, left
marginal tubercle and left supraorbital notch (Section 2.1.2). Other skull reconstructions are

U
possible, and justifiable. For example, it is likely that we have reduced the overall patterns of

AN
asymmetry which may have been an antemortem feature of this individual (see Kaifu, et al.,
2009). However, even though LB1’s asymmetry is a parallelogram skewing that is more
apparent in the transverse plane, the Procrustes registered wireframe of our facial
M
approximation shows a clear hemiface asymmetry that is in agreement with LB1’s pattern of
cranial skewing, and further, that the right and left gonial angles differ (Figure 3, Shape 5) –
D

at least, in so far as the soft tissue landmarks applied capture this aspect of facial morphology.
TE

We apply, where possible, robust relationships to approximate the face of LB1. However,
there are a number of ‘forensic’ recommendations that are still widely circulated and applied,
EP

even though many have proven to be less than reliable. These include STD data sets (Section
2.2) as well as recommendations for estimating the shapes, size and orientation of the eyes,
nose, mouth and ears (Section 2.4). It is likely that the variance displayed between the LB1
C

faces analysed is due, in part, to the application of different skull/soft tissue relationships.
AC

Variance may also be due to some of the faces containing an admixture of extant non-human
primate, as well as anatomically modern Homo, facial characteristics.

A number of the creators of LB1 faces describe their incorporation of “forensic” methods,
such as applying STDs (e.g. as described in Balter, 2009, Daynès, 2008, Koepfer, 2003,
Roberts, 2011, Sawyer and Deak, 2007), and therefore these would be depths derived from
anatomically modern humans. It is likely that the STD data set, landmarks and angles of
 ACCEPTED MANUSCRIPT
19

insertion selected for these LB1 faces will differ from each other, and from ours (i.e. Stephan
and Simpson, 2008), and therefore slight differences in facial morphology is to be expected.
However, because we have applied STD(-1s) (refer Section 2.2), our expectation was that,
overall, our facial approximation would display a proportionally narrower face. However, the
geometric morphometric analysis shows that our facial approximation is very likely the
proportionally widest LB1 face (Section 2.6, Figures 3 and 5).

PT
There are three possible reasons why our LB1 face appears wider. Firstly, our results are

RI
orthogonal (built from CT scans), whereas the other 2D images of LB1 faces probably
include aspects of photographic and/or artistic perspective. Secondly, face shape, including

SC
facial width and facial asymmetry, will be influenced by how the mandible has been
articulated, and the extent to which taphonomic and post-excavation damage has been
addressed and/or incorporated (refer Section 2.1.1). Thirdly, some of the creators of the faces

U
analysed here (Shapes 4, 7, 8) also describe including non-human primate anatomy in their

AN
work (detailed in Koepfer, 2003, Roberts, 2011, Sawyer and Deak, 2007), though these
creators do not specify which aspects, or which primate.
M
Research involving dissection of Pan troglodytes shows that while human and chimpanzee
facial musculature is close to identical, chimpanzee faces have very little adipose tissue
D

(Burrows, et al., 2006). Comparing the average STDs from a preliminary study using
ultrasound measures taken from living, adult Pan troglodytes (n=16-19, Hanebrink, 2006)
TE

with Stephan and Simpson’s (2008) weighted means, chimpanzee STDs are approximately
half that of modern humans in the area of the cheeks (at the landmarks lateral to the superior
EP

and inferior second molars), which is where most of the fat in the human face is located
(Raskin and LaTrenta, 2007). Use of Pan troglodytes STDs would, therefore, produce a
thinner mid-face than a depiction based on anatomically modern human STDs, and perhaps
C

explain why many LB1 faces appear to be proportionately narrower than our facial
AC

approximation. However, the Hanebrink’s (2006) STD data also shows that chimpanzee
STDs are nearly twice as thick as modern humans at the angle of the jaw (gonion), and this
aspect of Pan anatomy does not seem to be a feature that has been incorporated into any of
the LB1 faces analysed here.

Perspective, mandibular articulation and selective incorporation of mid-face Pan STDs may
explain why geometric morphometric analyses suggest a tendency towards depicting LB1
 ACCEPTED MANUSCRIPT
20

with a narrower mid-face, but not the trend towards depicting LB1 with a long upper lip. As
discussed (Sections 2.1 and 2.2), excavation of the LB1 cranium included inadvertent
removal of the surface of the subnasal region of the maxilla, and this damage confounds all of
the upper and mid-face anterior dimensions, such as STDs along the midline when the
skull/face is in profile. This excavation damage does not, however, confound bone orientation
on the saggital plane. LB1’s nasal aperture base, maxilla, alveolar ridges and remaining

PT
maxillary dentition retain their vertical orientations, even though the surfaces have been
inadvertently sheared away. Therefore, the tendency to depict LB1 with a long upper lip is

RI
not strongly justified by the skeletal evidence.

SC
Our approximation of LB1’s nasal dimensions sits within the range of noses depicted for
LB1. Some are narrower, and some are broader, though broader noses could be an artefact of
artistic/photographic perspective. A narrower appearing nose is not related to perspective,

U
and is a feature of three LB1 faces (Shapes 2, 6 and 9). This could be due to an interpretation

AN
of LB1’s nasal aperture as more similar to non-human primates. Research regarding Pan
troglodyte nasal apertures finds they are less elongated (Schmittbuhl, et al., 1998) and
narrower (Neaux, et al., 2013) than modern humans, and in Pongo, Gorilla and Pan, the
M
nasospinale and premaxilla have very different orientations (e.g. McCollum and Ward, 1997).
However, the relationship between the hard/soft nasal tissues of non-human primates does
D

not appear to be as well studied, or verified, as modern Homo. A further point is that of the
creators who include aspects of non-human primates in their work (as described in Koepfer,
TE

2003, Roberts, 2011, Sawyer and Deak, 2007), all depict a broad, and not narrow, LB1 nasal
morphology (Shapes 4, 7, 8). Quite what the justification could be for depicting LB1 with a
EP

narrower nose is therefore unknown at this time.

Other aspects that could influence the facial morphology of LB1 include the quality of the
C

casts on which the faces were based, and the knowledge underpinning interpretation of the
AC

remains. It is likely that an LB1 face would be severely compromised if the skeletal cast was
estimated from photographs and/or the creators did not perceive many of the distortions to the
mandible and cranium as having occurred postmortem (as appears to be the case with Shape
7 – see Section 2.6). It is also likely that some LB1 faces would be discordant with the
remains if they were created without due reference to the existing body of research, and/or
before there was widespread consensus (e.g. Aiello, 2010, Argue, et al., 2009, Kaifu, et al.,
2011) concerning LB1’s status as an archaic hominin (with this being most likely with the
 ACCEPTED MANUSCRIPT
21

first LB1 face, Shape 9). A further possibility is that some of the LB1 faces could also have
been (mis)guided by a common, and well-documented, assumption within archaeology that
archaic hominins fall into the socio-historical categories encompassing “wild men”, “missing
links” and “ape-men” (Berman, 1999, Montagu, 1947, Moser, 1992). This last point could
also be related to why there appears to be a marked tendency to generally, or selectively,
reference or allude to non-human primate morphologies in LB1’s facial features, even though

PT
LB1 is a member of the Homo genus, and much more is known, and verified, regarding the
relationship between the skull and soft tissues of modern Homo sapiens.

RI
SC
3. Conclusion
Our facial approximation of the Homo floresiensis holotype is the result of our virtual
reconstruction of the LB1 mandible and cranium and subsequent application of known and

U
verified data, algorithms and guidelines derived from the faces of anatomically modern

AN
humans. However, although we approximate the face and features of LB1 using statistical
averages, and therefore unavoidably average her facial appearance, these results still
incorporate differences due to the unique bony morphology of the skeletal remains. In the
M
case of LB1 these include a low forehead, prominent brow ridges, distinct cheek bones, facial
asymmetry, a comparatively short upper lip, and a jaw lacking a chin.
D
TE
C EP
AC
 ACCEPTED MANUSCRIPT
22

Footnotes:

1. We are grateful to Mark Collard for alerting us to this

PT
2. Thank you to Debbie Argue for providing this reasoning

Acknowledgements

RI
This research was originally presented at Science & Archaeology: The Australian
Archaeological Association Conference, 9-12 December 2012. We are grateful for the level

SC
of interest and enthusiasm shown, and the generosity and insightfulness of the feedback we
received, though all errors are our own. We are particularly thankful to Debbie Argue for
recommending revisions to an earlier draft of this paper, and to our Reviewers whose careful

U
reading and detailed recommendations were both insightful and encouraging.
AN
M
D
TE
C EP
AC
 ACCEPTED MANUSCRIPT
23

Figure 1: Facial reconstruction and approximation of LB1 (a) right lateral STDs, profile
outlines, feature location/morphology; (b) frontal STDs, feature location/morphology,
reconstructed bone (shaded); (c) left lateral reconstruction (mandible only), STDs and CT
slices for nasal projection.

PT
Figure 2: Approximation of underlying anatomy and surface appearance (a) right lateral

RI
underlying anatomical features; (b) frontal underlying anatomy overlaid with surface
features; (c) final result.

SC
Figure 3: Wireframes of the LB1 faces analysed. Our facial approximation is Shape 5 (top

U
row, far right). The remaining images were sourced as follows: Top row, from left: Shape 1

AN
(Carr, 2012); Shape 2 (Daynès, 2008); Shape 3 (Davis and Deak, 2010); Shape 4 (Balter,
2009). Bottom row, from left: Shape 6 (Hall, 2010); Shape 7 (Roberts, 2011); Shape 8
(Sawyer and Deak, 2007); Shape 9 (Kemp, 2004); Shape 10 (Anton, 2012).
M
D

Figure 4: Geometric morphometric analysis of all LB1 faces (N=10) and Thin Plate Spline
deformation grids showing overall variance for PC1 (x axis) and PC2 (y axis). The Thin Plate
TE

Splines were extracted from the following coordinates: PC1 -0.013, PC 1 +0.014, PC2 -0.09,
PC2 +0.09. The mean face for the group is shown bottom left, and includes some aspects of
EP

head turn facial morphology.

C

Figure 5: Geometric morphometric analysis of only the frontally posed LB1 faces (n=5) and
AC

Thin Plate Spline deformation grids showing overall variance for PC1 (x axis) and PC2 (y
axis). The Thin Plate Splines were extracted from the following coordinates: PC1 -0.07, PC 1
+0.05, PC2 -0.05, PC2 +0.05. The mean face for the group is shown bottom left, the overlay
of Procrustes registered wireframes is shown bottom right.
 ACCEPTED MANUSCRIPT
24

References
Aiello, L.C., 2010. Five years of Homo floresiensis, Am. J. Phys. Anthropol. 142, 167-179.
Anton, M., 2012. Homo floresiensis,
http://www.sciencephoto.com/image/171053/530wm/E4380156-Homo_floresiensis-SPL.jpg.
Argue, D., Morwood, M.J., Sutikna, T., Jatmiko, Saptomo, E.W., 2009. Homo floresiensis: a
cladistic analysis, J. Hum. Evol. 57, 623-639.

PT
Baab, K.L., McNulty, K.P., 2009. Size, shape, and asymmetry in fossil hominins: the status
of the LB1 cranium based on 3D morphometric analyses, J. Hum. Evol. 57, 608-622.

RI
Balter, M., 2009. Bringing hominins back to life, Science 325, 136-139.
Barsh, G.S., 2003. What controls variation in human skin color?, PLoS Biol. 1, e27.

SC
doi:10.1371/journal.pbio.0000027.
Berman, J.C., 1999. Bad hair days in the Paleolithic: modern (re)constructions of the cave
man, American Anthropologist 101, 288-304.

U
Brown, P., Sutikna, T., Morwood, M.J., Soejono, R.P., Jatmiko, Wayhu Saptomo, E., Awe

AN
Due, R., 2004. A new small-bodied hominin from the Late Pleistocene of Flores, Indonesia,
Nature 431, 1055-1061.
Brown, P., Maeda, T., 2009. Liang Bua Homo floresiensis mandibles and mandibular teeth: a
M
contribution to the comparative morphology of a new hominin species, J. Hum. Evol. 57,
571-596.
D

Brown, P., 2012. LB1 and LB6 Homo floresiensis are not modern human (Homo sapiens)
cretins, J. Hum. Evol. 62, 201-224.
TE

Burrows, A.M., Waller, B.M., Parr, L.A., Bonar, C.J., 2006. Muscles of facial expression in
the chimpanzee (Pan troglodytes): descriptive, comparative and phylogenetic contexts, J.
EP

Anat. 208, 153-167.

Carr, K., 2012. Homo floresiensis, http://www.karencarr.com/tmpl1.php?CID=475.
Davis, C., Deak, V., 2010. Welcome to the family, Discovery Channel Magazine 2, 34-41.
C

Davy-Jow, S.L., Decker, S.J., Ford, J.M., 2012. A simple method of nose tip shape validation
AC

for facial approximation, Forensic Sci. Int. 214, 208.e201-208.e203.

Daynès, É., 2008. Homo floresiensis, http://www.daynes.com/fr/reconstructions/flores.php.
Falk, D., Hildebolt, C., Smith, K., Brown, P., Jungers, W., Larson, S., Sutikna, T., Prior, F.,
2010. Nonpathological Asymmetry in LB1 (Homo floresiensis): A Reply to Eckhardt and
Henneberg, Am. J. Phys. Anthropol. 143, 340-342.
Farkas, L.G., Munro, I.R., 1987. Anthropometric Facial Proportions in Medicine, Charles C
Thomas, Springfield.
 ACCEPTED MANUSCRIPT
25

Farkas, L.G., 1994. Anthropometry of the Head and Face, Raven, New York, p. 405.
Fedosyutkin, B., Nainys, J., 1993. The relationship of skull morphology to facial features, in:
Iscan, M.Y., Helmer, R.P. (Eds.), Forensic Analysis of the Skull: Craniofacial Analysis,
Reconstruction, and Identification, Wiley-Liss, New York, pp. 119-213.
Gatliff, B.P., Snow, C.C., 1979. From skull to visage, J. Biocommun. 6, 27-30.
George, R.M., 1987. The lateral craniographic method of facial reconstruction, J. Forensic

PT
Sci. 32, 1305-1330.
George, R.M., 2007. Facial Geometry: Graphic Facial Analysis for Forensic Artists, Charles

RI
C Thomas, Springfield Illinois.
Gerasimov, M.M., 1955. The Reconstruction of the Face from the Basic Structure of the

SC
Skull, Nauka, Moscow.
Gerasimov, M.M., 1971. The Face Finder, J.B. Lippincott, Philadelphia.
Grant, J.C.B., 1943. An Atlas of Anatomy, Williams and Wilkins, Baltimore.

U
Guyomarc’h, P., Dutailly, B., Couture, C., Coqueugniot, H., 2012. Anatomical placement of

AN
the human eyeball in the orbit - validation using CT Scans of living adults and prediction for
facial approximation, J. Forensic Sci., 1271-1275.
Guyomarc’h, P., Stephan, C.N., 2012. The validity of ear prediction guidelines used in facial
M
approximation, J. Forensic Sci., 1427-1441.
Hall, S., 2010. UOW hobbit makes it big in Japan, Illawarra Mercury, Wollongong.
D

Hanebrink, J.R., 2006. Datum is only skin deep: in vivo measurements of facial tissue
thickness in chimpanzees, The Department of Geography and Anthropology, Louisiana State
TE

University and Agricultural and Mechanical College, p. 47.

Hayes, S., 2009. Seeing and Measuring the 2D Face, Life and Physical Sciences, Anatomy
EP

and Human Biology, PhD Thesis. University of Western Australia.

Hayes, S., 2011. A man from San Juan: facial approximation within anthropology (Un
hombre de San Juan: aproximación facial dentro de la antropología), Revista Argentina de
C

Antropologia Biologica 13, 71-81.

AC

Hayes, S., Milne, N., 2011. What's wrong with this picture? An experiment in quantifying
accuracy in 2D portrait drawing, Visual Communication 10, 149-174.
Hayes, S., Buckley, H., Bradley, R., Milne, N., Dennison, J., 2012. Approximating the face of
‘Aunty’: a question of likeness, Journal of Archaeological Method and Theory 19, 306-321.
Hayes, S., Tullberg, C., 2012. Police witness identification images: a geometric
morphometric analysis, J. Forensic Sci. 57, 1487-1494.
 ACCEPTED MANUSCRIPT
26

Kaifu, Y., Baba, H., Kurniawan, I., Sutikna, T., Saptomo, E.W., Jatmiko, Awe, R.D.,
Kaneko, T., Aziz, F., Djubiantono, T., 2009. Brief Communication: "Pathological"
Deformation in the Skull of LB1, the Type Specimen of Homo floresiensis, Am. J. Phys.
Anthropol. 140, 177-185.
Kaifu, Y., Kaneko, T., Kurniawan, I., Sutikna, T., Saptomo, E.W., Jatmiko, Awe, R.D., Aziz,
F., Baba, H., Djubiantono, T., 2010. Posterior deformational plagiocephaly properly explains

PT
the cranial asymmetries in LB1: A reply to Eckhardt and Henneberg, Am. J. Phys. Anthropol.
143, 335-336.

RI
Kaifu, Y., Baba, H., Sutikna, T., Morwood, M.J., Kubo, D., Wayhu Saptomo, E., Jatmiko,
Due Awe, R., Djubiantono, T., 2011. Craniofacial morphology of Homo floresiensis:

SC
description, taxonomic affinities, and evolutionary implication, J. Hum. Evol. 61, 644-682.
Kemp, M., 2004. A hobbit-forming image, Nature 432, 555.
Koepfer, D.L., 2003. Representation and aesthetics in paleo-art: an interview with John

U
Gurche, American Anthropologist 105, 146-148.

AN
Krogman, W.M., Iscan, M.Y., 1986. The Human Skeleton in Forensic Medicine, 2nd edition
ed., Charles C Thomas, Springfield.
Larrabee, W., Makielski, K., 1993. Surgical Anatomy of the Face, Raven, New York.
M
McCollum, M.A., Ward, S.C., 1997. Subnasoalveolar anatomy and hominoid phylogeny:
Evidence from comparative ontogeny, Am. J. Phys. Anthropol. 102, 377-405.
D

McDowell, J.L., L’Abbé, E.N., Kenyhercz, M.W., 2012. Nasal aperture shape evaluation
between black and white South Africans, Forensic Sci. Int. 222, 397.e391-397.e396.
TE

McNulty, K.P., Baab, K.L., 2010. Keeping asymmetry in perspective: A reply to Eckhardt
and Henneberg, Am. J. Phys. Anthropol. 143, 337-339.
EP

Montagu, M.F.A., 1947. A study of man embracing error, Tehnology Review 49, 345-347.
Morwood, M.J., Brown, P., Jatmiko, Sutikna, T., Wahyu Saptomo, E., Westaway, K.E., Awe
Due, R., Roberts, R.G., Maeda, T., Wasisto, S., Djubiantono, T., 2005. Further evidence for
C

small-bodied hominins from the Late Pleistocene of Flores, Indonesia, Nature 437, 1012-
AC

1017.
Morwood, M.J., van Oosterzee, P., 2007. The Discovery of the Hobbit: The scientific
breakthrough that changed the face of human history, Random House, Sydney.
Moser, S., 1992. Visions of the Australian Pleistocene: prehistoric life at Lake Mungo and
Kutikina, Australian Archaeology, 1-10.
 ACCEPTED MANUSCRIPT
27

Neaux, D., Guy, F., Gilissen, E., Coudyzer, W., Ducrocq, S., 2013. Covariation between
midline cranial base, lateral basicranium, and face in modern humans and chimpanzees: a 3D
geometric morphometric analysis, The Anatomical Record 296, 568-579.
O'Higgins, P., Jones, N., 2006. Tools for statistical shape analysis, Hull York Medical
School.
Paff, G.H., 1973. Anatomy of the Head and Neck, W.B. Saunders, Philadelphia.

PT
Prag, J., Neave, R., 1997. Making Faces: Using forensic and archaeological evidence, British
Museum Press, London.

RI
Rantala, M.J., 2007. Evolution of nakedness in Homo sapiens, J. Zool. 273, 1-7.
Raskin, E., LaTrenta, G.S., 2007. Why do we age in our cheeks?, Aesthet. Surg. J. 27, 19-28.

SC
Reichs, K.J., Craig, E., 1998. Facial approximation: procedures and pitfalls, in: Reichs, K.J.
(Ed.), Forensic Osteology: Advances in the identification of human remains, 2nd ed., Charles
C Thomas, Springfield, pp. 491-513.

U
Roberts, A., 2011. Evolution: The human story, Dorling Kindersley: London.

AN
Rogers, A.R., David Iltis, Stephen Wooding, 2004. Genetic variation at the MC1R locus and
the time since loss of human body hair, Current Anthropology 45, 105-108.
Romanes, G.J., 1986. Cunningham's Manual of Practical Anatomy. Volume 3: head and neck
M
and brain, 15th ed., Oxford University Press, Oxford.
Rynn, C., Wilkinson, M., Peters, H., 2010. Prediction of nasal morphology from the skull,
D

Forensic Science, Medicine and Pathology 6, 20.

Sawyer, G.J., Deak, V., 2007. The Last Human: A guide to twenty-two species of extinct
TE

humans, Yale University Press.

Schmittbuhl, M., Le Minor, J.-M., Allenbach, B., Schaaf, A., 1998. Shape of the piriform
EP

aperture in Gorilla gorilla, Pan troglodytes, and modern Homo sapiens: characterization and
polymorphism analysis, Am. J. Phys. Anthropol. 106, 297-310.
Song, W.-C., Kim, S.-H., Paik, D.-J., Han, S.-H., 2007. Location of the infraorbital and
C

mental foramen with reference to the soft-tissue landmarks, Plastic and reconstructive surgery
AC

(1963) 120, 1343-1347.

Standring, S., 2008. Gray's Anatomy: The anatomical basis of clinical practice, Elsevier.
Stephan, C., Henneberg, M., 2003. Predicting mouth width from inter-canine width-a 75%
rule, J. Forensic Sci. 48, 725-727.
Stephan, C., 2005. Facial approximation: a review of the current state of play for
archaeologists, International Journal of Osteoarchaeology 15, 298-302.
 ACCEPTED MANUSCRIPT
28

Stephan, C., Davidson, P.L., 2008. The placement of the human eyeball and canthi in
craniofacial identification, J. Forensic Sci. 53, 612-619.
Stephan, C., Murphy, S., 2008. Mouth width prediction in craniofacial identification: cadaver
tests of four recent methods, including two techniques for edentulous skulls, J. Forensic
Odontostomatol. 27, 2-7.
Stephan, C.N., 2002. Facial approximation: globe projection guideline falsified by

PT
exophthalmometry literature, J. Forensic Sci. 47, 730-735.
Stephan, C.N., 2003. Facial approximation: an evaluation of mouth-width determination, Am.

RI
J. Phys. Anthropol. 121, 48-57.
Stephan, C.N., Simpson, E.K., 2008. Facial soft tissue depths in craniofacial identification

SC
(part 1): an analytical review of the published adult data, J. Forensic Sci. 53, 1257-1272.
Stephan, C.N., Huang, A.J.R., Davidson, P.L., 2009. Further evidence on the anatomical
placement of the human eyeball for facial approximation and craniofacial superimposition, J.

U
Forensic Sci. 54, 267-269.

AN
Taylor, K.T., 2001. Forensic Art and Illustration, CRC Press, Boca Raton.
Taylor, R.G., Angel, C., 1998. Facial reconstruction and approximation, in: Clement, J.G.,
Ranson, D.L. (Eds.), Craniofacial Identification in Forensic Medicine, Arnold, London, pp.
M
177-186.
Ullrich, H., Stephan, C., 2011. On Gerasimov's plastic facial reconstruction technique: new
D

insights to facilitate repeatability, J. Forensic Sci. 56, 470-474.

Wilkinson, C., Caroline, M.W., Sophie, A.M., 2003a. Measurement of eyeball protrusion and
TE

its application in facial reconstruction, J. Forensic Sci. 48, 12.

Wilkinson, C., Motwani, M., Chaing, E., 2003b. The relationship between the soft tissues and
EP

the skeletal detail of the mouth, J. Forensic Sci. 48, 728-732.

Wilkinson, C., 2004. Forensic Facial Reconstruction, Cambridge University Press,
Cambridge.
C

Wolff, E., 1948. The Anatomy of the Eye and Orbit., 3rd ed., H.K. Lewis, London.
AC
 ACCEPTED MANUSCRIPT

Table 1: Craniometric points, definitions and soft tissue depths (STD), from Stephan and
Simpson 2008. Note that the STD less one standard deviation (-1s) are calculated from a subset
supplied by the authors. Refer Figure 1 for location of points and angles of STD.

STD (mm) STD (mm)

Definition Weighted -1s from s Additional notes regarding identification for LB1
Mean Studies
Median Median points were taken primarily from the right

PT
Points lateral, unless otherwise indicated
Highest midline Highest point on the undamaged cranium
Vertex 5 4
ectocranial point corresponding with the midline
According to Kaifu et al. (2011) this projects beyond

RI
Most anterior the preserved mid-supraorbital margin by 0-2mm,
Glabella midline point on the 5.5 4.5 following other adult Homo individuals from the
frontal bone Pleistocene. We locate this at the maximum lateral
projection of the preserved bone (i.e. 0mm)

SC
Using a midsaggital CT scan Kaifu et al. (2011)
identify this as approximately the same height as the
Midline point on the
Nasion 6.5 4.5 right anterior frontomalare. In the lateral view we
naso-frontal suture
locate this at the most anterior section of the
preserved medial border of the left orbit

U
We take this to be correspondent with the most
anterior fragment of the superior nasal aperture where

Midline point at the

3 (not
AN
2 (not
the perpendicular plate of the ethmoid bone
articulates with the nasal bones in the frontal view,
which is lower than that suggested by Kaifu et al.
Rhinion inferior free end of (2011). This location also agrees with the average
applied) applied)
the internasal suture nasal aperture shape for black and white South
M
Africans (McDowell, et al. 2012). Only the frontal
point is approximated, and this is only used for
partial calculation of nasal projection (rhinion-
subnasale 26.8mm)
D

Slightly superior (1.5mm) to the preserved section of

the right central incisor fossa. As this point cannot be
Midline point just
located on preserved bone it is based on an observed
TE

Subnasale below the anterior 12.5 9.5

relationship between the incisor fossa and the inferior
nasal spine
border of the anterior nasal spine in anatomically
modern humans, and which has not been verified.
Midline point In lateral view, midway point on the most anterior
EP

midway between the section of preserved maxillary bone

base of the nasal
Midphiltrum 11 8.5
spine and prosthion
on the anterior edge
of the maxillae
C

Midline point at the Following Kaifu et al. (2011) we locate this as 1mm
most anterior edge superior to the preserved central incisor alveolar
AC

Prosthion of the superior 11.5 8.5 ridge, which also corresponds to their nasion-
alveolar ridge of prosthion distance of 54mm
maxillae
Midline point at the In lateral view, at most anterior point of preserved
most anterior edge alveolar ridge
Lower lip of on the inferior 13 10.5
alveolar ridge of the
mandible
Deepest midline In lateral view, deepest midline point beneath the
point in the groove alveolar ridge of the mandible
Labiomental 11 9
superior to the
mental eminence
 ACCEPTED MANUSCRIPT

STD (mm) STD (mm)

Definition Weighted -1s from s Additional notes regarding identification for LB1
Mean Studies
Most anterior In lateral view, most anterior projection of the
midline point on the mandible beneath the alveolar ridge of the mandible
Pogonion 11 8.5
mental eminence of
the mandible
Midline point
halfway between the

PT
Gnathion 8.5 5.5
pogonion and
menton
Most inferior
midline point at the
Menton 7 4.5

RI
mental symphysis of
the mandible
Bilateral
Points

SC
Point on the lateral Our rearticulation of the left mandible to fit with the
aspect of the border left mandibular fossa and molar occlusion results in
of mandibular angle the right and left gonion being symmetrical in
where a tangent location although the rami differ morphologically
bisects the angle

U
Gonion 10 4
formed by the
posterior ramus
border and the
inferior corpus
border
AN
Most lateral extent See our explanation of our reconstruction of right
of the lateral surface zygomatic (Section 2.1.2)
M
Zygomatic 6 5
of the zygomatic
arch
Point at the centre Morphological difference between the rami results in
Midramus of the mandibular 17.5 13.5 the left midramus point being slightly lower (0.6mm)
D

ramus than the right

TE
C EP
AC
 ACCEPTED MANUSCRIPT

PT
RI
U SC
AN
M
D
TE
EP
C
AC
 ACCEPTED MANUSCRIPT

PT
RI
U SC
AN
M
D
TE
EP
C
AC
 ACCEPTED MANUSCRIPT

PT
RI
U SC
AN
M
D
TE
EP
C
AC
 ACCEPTED MANUSCRIPT

PT
RI
U SC
AN
M
D
TE
EP
C
AC
 ACCEPTED MANUSCRIPT

PT
RI
U SC
AN
M
D
TE
EP
C
AC

View publication stats