Vision Research 49 (2009) 2870–2880

Contents lists available at ScienceDirect

Vision Research
journal homepage: www.elsevier.com/locate/visres

Anchoring gaze when categorizing faces’ sex: Evidence from eye-tracking data
Line Sther a,*, Werner Van Belle b, Bruno Laeng c, Tim Brennen c, Morten Øvervoll a
a
University of Tromsø, Norway
b
Swiss Federal Institute of Technology Zürich, Switzerland
c
University of Oslo, Norway

a r t i c l e i n f o a b s t r a c t

Article history: Previous research has shown that during recognition of frontal views of faces, the preferred landing posi-
Received 11 December 2008 tions of eye fixations are either on the nose or the eye region. Can these findings generalize to other facial
Received in revised form 13 August 2009 views and a simpler perceptual task? An eye-tracking experiment investigated categorization of the sex
of faces seen in four views. The results revealed a strategy, preferred in all views, which consisted of
focusing gaze within an ‘infraorbital region’ of the face. This region was fixated more in the first than
Keywords: in subsequent fixations. Males anchored gaze lower and more centrally than females.
Face perception
Ó 2009 Elsevier Ltd. All rights reserved.
Sex categorization
Eye fixation
Optimal viewing position

1. Introduction
Faces constitute a set of 3D objects which seem to be mentally
represented in a holistical way (e.g., Biederman & Kalocsai, 1997;
Dailey & Cottrell, 1999; Laeng & Caviness, 2001; O’Toole, Millward,
& Anderson, 1988) that is, in a less part-based manner than most
objects. Presumably, holistic facial representations develop along-
side the human expertise in face recognition (Gauthier, Tarr,
Anderson, Skudlarski, & Gore, 1999; Sther & Laeng, 2008).
Although visual expertise can result in expert gaze behavior where
‘diagnostic’ features are prioritized (e.g., Rehder & Hoffman, 2005;
Underwood, Chapman, Brocklehurst, Underwood, & Crundall,
2003; cf. Biederman & Shiffrar, 1987) there is reason to believe that
both the extreme familiarity with the object class, faces, and its
dependence on a template like representation could result in a
gaze strategy where the goal is to get as much information as pos-
sible with just one fixation (Hsiao & Cottrell, 2008). That is, from
the first glance, gaze might be anchored onto a position which pro-
vides a perceptual span that either covers the whole stimulus or
that maximizes the area of the object (especially for large displays
or very near objects) which is included within the region of high
resolution acuity (cf. Hsiao & Cottrell, 2008; Tyler & Chen, 2006).
Interestingly, studies on visual inspection of other shapes than
faces have shown that when participants look at an object or a
group of objects, the preferred landing position of their fixations
occurs near the ‘‘center-of-gravity” (COG) of the display (Coren &
* Corresponding author. Address: University Library, Department of Psychology
and Law, University of Tromsø, N-9073 Tromsø, Norway. Fax: +47 77 64 67 89.
E-mail address: line.sether@uit.no (L. Sther).
Hoenig, 1972; Findlay, 1982; He & Kowler, 1991; Kowler & Blaser,
1995; McGowan, Kowler, Sharma & Chubb,1998; Melcher &
Kowler, 1999; Vishwanath & Kowler, 2004; Vishwanath, Kowler,
& Feldman, 2000). The COG of a shape can also be based on its
3D structure (by spatial weighting according to implied depth;
Vishwanath & Kowler, 2004). Nevertheless, most of this work has
studied saccades to targets with sudden onset. The COG preference
is less obligatory with more voluntary saccadic movements (Find-
lay & Blythe, 2009; He & Kowler, 1991) as would be the case in face
processing.
Studies of fixations on faces have focussed less on the possibil-
ity that a similar COG-anchoring mechanism might be operating.
The available evidence suggests that overt attention appears to
be drawn towards the internal region of the face, particularly to-
wards the eyes (Althoff & Cohen, 1999; Fisher & Cox, 1975;
Henderson, Williams, & Falk, 2005; Langdell, 1978; Minut,
Mahadevan, Henderson, & Dyer, 2000; Schwarzer, Huber, &
Dümmler, 2005; Walker-Smith, Gale, & Findlay, 1977; Yarbus,
1967). Thus, one could speculate that the eyes might be attended
to independently of the task as a side effect of an initial centering
of the gaze (Deaner & Platt, 2003; Grosbras, Laird, & Paus, 2005;
Kingstone, Tipper, Ristic, & Ngan, 2004). In other words, fixations
on the eye region in a particular task may incorrectly suggest a
diagnostic role of this facial region (e.g., Schyns, Bonnar, &
Gosselin, 2002) and might just as well reflect ‘automatic’ anchoring
of the gaze (cf. Vinette, Gosselin, & Schyns, 2004). In fact, Tyler and
Chen (2006) using a forced choice face detection paradigm, and
aperture windows of varying extent, identified a region centered
on the nose bridge in the full face as a zone of specialized percep-
tual processing. Although the other internal features of the face

0042-6989/$ - see front matter Ó 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.visres.2009.09.001
 L. Sther et al. / Vision Research 49 (2009) 2870–2880
provided some improvement in sensitivity, there was no further
advantage for the information around the edge of the face (i.e. hair,
ears, and jaw).
Most interestingly, recent work by Hsiao and Cottrell (2008)
indicates that during recognition of frontal faces, the preferred
landing positions for the first two eye fixations is around the center
of the nose, slightly biased to the left, and that fixations on the eyes
do occur at a later stage (typically after the second fixation). Impor-
tantly, Hsiao and Cottrell showed that after these two fixations,
performance does not improve, that is, no further information ap-
pears to be necessary to perform the identification task. Thus, in
previous studies that used a central starting point for eye fixations,
it would be unnecessary to fixate centrally during recordings, since
recordings would start after the initial informative fixation was al-
ready accomplished and the subsequent gaze behavior (towards
the eyes) would simply be redundant and would not add
information.
One purpose of the present study was to investigate to what ex-
tent the findings in Hsiao and Cottrell (2008) can be generalized.
On the one hand, central fixations on the nose in frontal views
may occur because the nose happens to be in a central position
in the facial image. Alternatively, one could argue that the nose
(and the face region immediately around it) might be a diagnostic
part for face recognition; in this view, anchoring gaze on the nose
would constitute an ‘optimal viewing position’ for several specific
face tasks (cf. O’Regan, Lèvy-Schoen, Pynte, & Brugaillère, 1984).
Such a confound between fixations on the nose as merely an an-
chor point versus as an informationally-rich nexus can be resolved
by exploring fixations with the use of other Angles of View than the
full face view. Therefore, the face stimuli used in the present
experiment were shown in four static views (from full-front to
profile).
One additional point is that the findings of Hsiao and Cottrell
(2008) might be restricted to the identification task. Face identifi-
cation is a complex, visual categorization task involving only famil-
iar faces, whereas other face processing tasks can be performed on
unfamiliar faces as well. Schyns et al. (2002), using the ‘‘Bubbles
technique”, found differing patterns of attended information be-
tween three perceptual tasks (identity, expression and sex catego-
rization). In addition, Malcolm, Lanyon, Fugard, and Barton (2008)
monitored eye fixations while participants made judgments about
the identity or the expression of faces, and found that fixations cor-
related with regional variations of diagnostic information in the
different processing tasks. Finally, we reason that the visual infor-
mation which is most informative for a complex task might not be
the same as the information used in a simpler and more basic task.
To investigate this question we used a very simple face processing
task, i.e. ‘sex categorization’, which can be performed also for unfa-
miliar faces, and which has been little explored with the eye mon-
itoring technique.
Sex categorization can be more efficiently performed than other
processing tasks (i.e., in about 613 ms; as opposed to 897 ms for
familiarity decisions; Bruce, Ellis, Gibling, & Young, 1987). ERP
studies have shown that the brain potential’s latency related to
sex categorizations of faces is remarkably fast (i.e., 150 ms; Schen-
dan, Ganis, & Kutas, 1998). Even with very fast and peripherally lo-
cated exposures (26–75 ms) accuracy of sex decisions is
surprisingly good (O’Toole, Peterson, & Deffenbacher, 1996; Reddy,
Wilken, & Koch, 2004; Sergent & Hellige, 1986). In addition, neuro-
imaging studies reveal that areas of the brain activated by sex cat-
egorizations of faces are more posterior than those activated by
identifying the same faces (Sergent, Ohta, & MacDonald, 1992)
which in turn suggests that such information may be processed
early within the visual pathways.
It might be that even in this simple sex categorization task, the
most dimorphic anatomical difference would correspond to the
2871
part of the face attended to first. Current knowledge of sex differ-
ences in face morphology strongly suggests that the most sexually-
dimorphic facial trait is indeed the nose (Bruce & Young, 1998;
Burton, Bruce, & Dench, 1993; Enlow, 1982, 1990; cf. O’oole, Vetter,
Troje, & Bülthoff, 1997). If so, we can draw some straightforward
predictions about the locus of eye fixations and their priority in a
sex categorization task: (1) if gaze anchors itself onto the most
diagnostic part of the face, then the nose should be attended to
in the first fixation regardless of facial pose (indeed, in the 3=4 and
profile poses, one could expect more fixations on the nose, since
visibility of its size and shape would seem optimal in these per-
spectives; Chronicle et al., 1995). This could be interpreted as a
viewing position that optimally reveals information diagnostic to
the task. (2) If gaze anchors itself onto a central position of the im-
age, the nose should be fixated preferentially in frontal views and
decreasingly so for views increasingly distant from frontal close-
ups. This would indicate the existence of a preferred landing posi-
tion that optimizes the visual inspection (foveally and parafoveal-
ly) of the face as a whole.
However, it is likely that the perceptual process may not be
that straightforward. That is, several dimorphic parts (e.g., the
zygomatic protrusion; Ikeda, Nakamura, & Itoh, 1999) or their
segments might contribute in parallel, and perhaps equally, to
the decision (cf. Bruce et al., 1993; Burton et al., 1993). Also, the
diagnostic value of the parts may not depend on their dimorphic
value alone, but may be influenced by perceptual variables (e.g.,
small parts, or ambiguous parts, like a feminine-looking male
nose, might need foveal vision, and relative size might need to
be computed in a context, by a scan path). Some of these second-
ary diagnostic cues might influence eye fixations subsequent to
the first fixation. One might think that an ‘intelligent’ strategy
could be to anchor gaze onto an intermediate position between
two or more highly diagnostic parts (especially if the viewing con-
ditions allow them to be included within a visual angle of maxi-
mal visual acuity). Moreover, it may be sufficient to view ‘large’
diagnostic parts outside of the foveal- and the parafoveal area,
where low frequencies might provide enough dimorphic informa-
tion for sex categorizations (Schyns et al., 2002; Valentin, Abdi,
Edelman, & O’Toole, 1997). In fact, if few or none of the dimorphic
regions are in need of foveal or central vision, it would be possible
and economical to gather information from more than one part in
one glance.
Current theories of gaze control and oculomotor strategies have
abandoned the idea that the choice of gaze location occurs at ran-
dom (e.g., Kundel, Nodine, Thickman, & Toto, 1987) and recognize
the influence of both stimulus factors (Itti & Koch, 2000) and
expectations (Land & McLeod, 2000; Malcolm et al., 2008; Torralba,
Oliva, Castlhano, & Henderson, 2006; Turano, Geruschat, & Baker,
2003; Wolfe, Cave, & Franzel, 1989). Thus, the present study pro-
vides an opportunity to establish whether scanning of a face during
a sex categorization task follows a specific and replicable pattern,
and also whether any of the sexually-dimorphic Facial Parts are
prioritized in the oculomotor behavior of the viewer. In particular,
by presenting faces in different views we have an opportunity to
assess whether the gaze targets dimorphic parts or perhaps seg-
ments that may not correspond to ‘parts’ that have verbal labels.
Alternatively, gaze may always target the image’s COG so as to fa-
vor the holistic processing of the face. Thus, eye movements of a
group of participants, performing a sex categorization task, on
stimuli shown in four Angles of View, were monitored by use of
an infrared eye-tracker.
Finally, fixational analysis methods (on the basis of a priori and
a posteriori defined parts) were compared in the present study so
as to guide future eye-tracking investigations of face stimuli. Spe-
cifically, a common problem has been that some a priori knowledge
about what portion of the facial surface counts as a part is assumed
2872 L. Sther et al. / Vision Research 49 (2009) 2870–2880
before analyses are completed (Burton et al., 1993; Valentin et al.,
1997). Several computational studies have however derived macro
parts a posteriori from faces learned by classification networks (e.g.,
Abdi, Valentin, Edelman, & O’Toole, 1995; Cottrell & Flemming,
1990; Golomb, Lawrence, & Sejnowski, 1991; Gray, Lawrence, Go-
lomb & Sejnowski,1995; O’Toole et al., 1997; O’Toole et al., 1998).
Thus, in the present experiment these two methods of parsing will
be compared.
2. Methods
2.1. Participants
The participants were 49 naive students (25 females). All partic-
ipants (mean age = 26.7, SD = 7.0) had normal, or corrected to nor-
mal vision by the use of contact lenses.
2.2. Stimulus and apparatus
The stimuli used in this experiment consisted of 96 color photos
of 24 unfamiliar faces (12 female, 12 male) of young Caucasian stu-
dents, randomly presented in four different viewing angles (fron-
tal = 00°, intermediate = 22.5°, three-quarter = 45° and profile =
90° angles as seen from the viewer). The 22.5° angle view was
included as this has been shown to be more diagnostic than the
45° angle view (Laeng & Rouw, 2001; cf. Blanz, Tarr, & Bühltoff,
1999).
The models were all asked to assume a neutral expression when
photographed. The images, including hair, were unaltered so as to
keep the visual stimuli as natural as possible. Half of the images
shown in angles other than the frontal were oriented towards
the left, and the other half towards the right side of the screen.
Each model’s head (including hair) subtended an average visual an-
gle of 14° (vertical dimension; min 13°–max 15°). However, the
face alone (i.e., the internal ‘mask’ from eyebrows to mouth) sub-
tended an average visual angle of 6°.
The experimental editor was SuperLabÒ Pro 1.04, and eye move-
ments were registered by the monocular ‘‘Remote Eye-Tracking
Device” or RED, built by SensoMotoric InstrumentsÒ (SMI, Teltow,
Germany). The initial analysis of the recordings was computed by
the software iViewÓ version 3.0 from SMI. The RED employs the
contrast technique, which determines the center between two
coordinates, by tracking the position of the pupil and the reflection
of the cornea. The tracking device can operate at a distance of 60–
80 cm and the recording eye-tracking sample rate is 50 Hz, with a
tracking resolution of 0.1° (as specified by the manufacturer;
http://www.smivision.com/en/gaze-eye-tracking-systems/prod-
ucts/iview-x-red-red250.html).
2.3. Procedure
Participants were seated in front of the monitor, with their
heads stabilized by a headstand to maintain a constant distance
of 73 cm. A calibration procedure on a 3  3 regularly spaced ma-
trix was performed for each participant. Participants were asked to
discriminate between female and male faces by pressing one of
two keys (labelled $ or #) as quickly as possible. Before starting
the experiment a practice set of six pictures was presented. Before
each stimulus, a fixation point was presented for 1500 ms. Each
fixation cross was positioned 7.85° away from the stimulus’ center
to force eye movements to start exploring the face from a position
external to it. The cross had to be fixated before each recording
started. Each stimulus remained on the monitor until the partici-
pant made a response.
2.4. Analysis procedure
Fixations outside the face were not included in the ‘center-of-
gravity’ or the ‘a priori’ analysis. A fixation was defined as focus
on an area of 60 screen points (the whole screen being
800  600 points) for more than 150 ms. According to Reichle, Ray-
ner, and Pollatsek (2003), duration is related to the difficulty in-
volved in the processing task and can vary between 100–400 ms.
The constraint was chosen on the basis of sex categorization being
an efficient task, and this definition resulted in only 1.97% of trials
without fixations on the face.
2.4.1. Computation of the ‘center-of-gravity’
The 2D center-of-gravity (COG) was computed through a MAT-
LABÒ program that computed the centroid of each stimulus face ta-
ken as a plane region of uniform density. The centroid was
computed through the MATLABÒ function regionprops.
2.4.2. Natural parts ‘a priori’ parsing
For the a priori analyses of fixations, we counted each separate
fixation on the major parts of the face as identified by colloquial
speech (i.e.: hair, forehead, brows, eyes, nose, cheeks, mouth,
chin/jaw and ears). Note that, instead of choosing a few specific
areas of interest, we parsed the whole facial image into its major
anatomical parts. This parsing was based on the major bone and
muscle structures in the face (Putz & Pabst, 2001). The parts’
boundaries were then adjusted to each face and to all four views
of the faces.
2.4.3. Coordinate grid ‘a priori’ parsing
The coordinate grid parsing consisted of 80 parts, based on a grid
that was finer in the internal, than the external, region of the face.
The frontal internal region of the face was thereby divided into 68
approximately equally sized parts. In other views than the frontal,
fewer than 80 segments were visible. The number of segments did
change slightly from face to face, as some segments, in some faces,
would be occluded by other segments or parts. However, the num-
bering system still remained the same. A MATLABÒ program recog-
nized the parts in the grid in the analysis.
2.4.4. A posteriori analysis
For the a posteriori analysis of fixations we developed a Linux
based, Interactive Image Spreadsheet program written in C++
(Stroustrup, 2000) and Qt (http://www.trolltech.com/products/qt/
index.html). As Facial Parts differ in size, shape and position be-
tween faces, the program normalized each face by morphing it
with all other faces (in the same angle) in order to create a proto-
type face, as well as a prototype female and a prototype male face.
Because all of the face stimuli, except the full face, were either ori-
ented towards the left or the right side of the screen, the faces were
first turned towards the left side. All faces were then morphed to
the prototypes, and these were then added, to make sure that no
face was more heavily weighted than others. Sixty-six morphing
points were used, and most points were positioned within the face
itself. The program’s next step was to morph the eye movement
data using the same technique and the same morphing points in
order to create a heat map/spotlight image of the most frequent
fixated facial areas. These images were created by representing
each point in the eye-tracking path with a patch of the same size
as the fovea (23 pixels in diameter) and plot the resulting map onto
the morphed facial images. The morphed prototypes were then
normalized such that the maximum signal was 1 and the lowest
signal was 0. A cut-off value of 5% was used to remove areas that
were of no significance.
 L. Sther et al. / Vision Research 49 (2009) 2870–2880
3. Results
3.1. Accuracy
Participants’ sex categorizations were on average 98.75% cor-
rect. There were no significant differences between face views.
Male and female participants did equally well in this task, but both
sexes made more mistakes when categorizing male than female
models (F(1, 47) = 4.91, p = 0.03). This difference was caused by
one male model (judged as ‘low in masculinity’ by 10 independent
viewers) that triggered 0.31 of the 1.25% mistakes. The remaining
0.93% errors indicated no pattern, and can most likely be consid-
ered erroneous key presses. Despite the near-to-ceiling perfor-
mance, the few erroneous trials were removed from the data.
3.2. RTs
The mean response time (RT) was 715 ms, SD = 155 (frontal view:
726 ms, SD = 162; intermediate view: 700 ms, SD = 136; three-quar-
ter view: 713 ms, SD = 152; profile view: 722 ms, SD = 153). A re-
peated-measures ANOVA with four Angles of View (00°, 22.5°, 45°,
90°)  two Sex of Model (female, male) as the within-participant
factors, and two sex of participant (female, male) as the between-
participants factor, measured participants response time. Results
showed a main effect of View, F(3, 141) = 5.9, p = 0.0008. Post-hoc
tests revealed that, when judging the Sex of faces, participants
were significantly faster with the intermediate view (22.5°) than
when seeing frontal and profile views. There were no other signifi-
cant effects, and there was no relationship between response time
and the eye movement patterns.
3.3. Percent dwell time in the ‘center-of-gravity’
An analysis of the time participants focused on the center-of-
gravity (within 2° of visual angle, corresponding to the size of the fo-
vea) was computed in order to check: (1) whether eye fixations sim-
ply have a tendency to select a central point of gaze, and (2) whether
such centered focusing was as preferred when a dimorphic part was
positioned in the center as when no such part were present in the
center-of-gravity (e.g. ear in profile view). The computation was
done on the head, excluding the neck, as the head may be considered
a natural part or independent object with respect to the rest of the
body. The head is separated from the neck by geometrical properties
based on concavities/convexities, as suggested by several models of
object processing (e.g., Hoffman & Richards, 1985).
A repeated-measures ANOVA with four Angle of View (00°,
22.5°, 45°, 90°)  two Sex of Model (female, male) as the within-
participant factors and two sex of participant (female, male) as
the between-participants factor, measured the percentage of time
that participants’ gaze was within 2° of the COG during each trial.
The results revealed that the gaze was on or near the COG 22.33%
or 159.65 ms (SD = 27.02) of the time. A significant main effect of
Angle of View (F(3, 138) = 102.17, p < 0.0001) demonstrated that
the gaze dwelled longer near the COG when the face was presented
in frontal view (41.12% or 298.53 ms, SD = 30.60) than in all other
views (intermediate view: 28.20%/197.40 ms, SD = 28.27; three-
quarter view: 13.72%/97.82 ms, SD = 19.01; profile view: 6.27%/
45.26 ms, SD = 10.29). The difference between the views was con-
firmed by post-hoc tests (p < 0.0001).
3.4. Analysis based on the natural parts’ a priori parsing
3.4.1. Distribution of fixations
First, the mean number of fixations during a trial (1.87;
SD = 0.39) was calculated. In order to compare the face views
2873
directly, the two sides of the frontal face were collapsed, and the
percentage of fixations within each view and for each of the a priori
defined parts was calculated. Fixation frequency was then sub-
jected to a repeated-measures ANOVA with nine Facial Parts (hair,
forehead, brows, eyes, nose, cheeks, mouth, chin/jaw and
ears)  four view (00°, 22.5°, 45°, 90°)  two Sex of Model
(female, male) as the within-participant factors, and two sex of
participant (female, male) as the between-subject factor. The anal-
ysis revealed a highly significant main effect of Facial Parts
(F(8, 376) = 33.99, p < 0.0001), as eyes (mean = 27.4%, SD = 18.7)
and nose (mean = 23.8%, SD = 12.5), followed by cheeks
(mean = 19.5%, SD = 13.6) elicited significantly more (70.7%) fixa-
tions than other parts. Post-hoc comparisons (Fisher’s PLSD;
p = 0.11) showed no significant differences between eyes, nose
and cheeks. In other words, observers looked repeatedly on the
central area of the face (eyes + nose + cheek).
However, an interaction with View (F(24, 1128) = 26.83,
p = 0.0001; Fig. 1) revealed that the nose was most attended to in
frontal views (nose to eye: p = 0.15, eye to cheek: p < 0.0001,
t(48) = 4.3) and gradually less attended to as the head turned.
The opposite pattern was found for attention towards the cheeks,
with more attention in profile (cheeks vs. eyes; p = 0.0009,
t(48) = 3.54). In other words, the central area of the face that was
specific to each view was the preferentially-attended target.
Additionally, the eye region was highly attended, regardless of
view.
The effect of Parts was also influenced by sex of participant,
F(8, 376) = 2.9, p = 0.003. Specifically, while both sexes attended
more towards the eyes, women looked significantly more on the
eyes than other parts, and more on the brows than men did,
whereas men attended more towards the nose and cheeks than
women. There was also a three ways interaction between Facial
Parts, Participants’ Sex and Angle of View F(24, 1128) = 2.7,
p < 0.0001. This revealed that women looked overall more at eyes,
whereas men looked at noses in frontal view and cheeks in profile
(male participants: frontal view: nose to eye: p = 0.018; nose to
cheek: p < 0.0001; profile: cheek to eye: p < 0.0001; cheek to nose:
p < 0.0001). In other words, males attended more towards the
view-specific center-of-gravity, whereas females attended more
towards the eyes, regardless of the view.
3.4.2. Order of Fixations
In order to assess whether the initial fixations were positioned
more centrally than the subsequent fixations, an analysis of Order
of Fixations was performed. As above, a repeated-measures ANOVA
Fig. 1. The mean percentage of fixations of the three most attended Facial Parts
(eyes, nose and cheeks) for each of the four face views (00°, 22.5°, 45° and 90° view).
The bars indicate 95% confidence intervals.
2874 L. Sther et al. / Vision Research 49 (2009) 2870–2880

was performed, with the additional within-participant factor of Or-
der of Fixations (first, second and third fixation). The results
showed a trivial main effect of Order of Fixations (F(2, 94) =
877.1, p < 0.0001) with more first fixations and less third
fixations.
There was a significant interaction between Order of Fixations
and Facial Parts (F(16, 752) = 16.3, p < 0.0001) that revealed a pref-
erential looking towards eyes, nose and cheek in the first and sec-
ond fixation (see Fig. 2). The pattern of the first and second fixation
was almost identical except for gaze on cheeks (t = 9.2, p < 0.0001)
and hair (t = 5.7, p < 0.0001) where the distance between first and
second fixation was larger than for the other parts. By the third fix-
ation, preferential fixating towards specific parts was not present
to any notable extent.
Furthermore, there was a significant three way interaction be-
tween Angle, Facial Parts and Order of Fixation (F(48, 2256) = 9.1,
p < 0.0001) which showed that the central parts of the face in each
view (e.g. nose in front, eyes/cheek in intermediate view, eyes/
cheek in three-quarter view, and cheek/hair in profile) were fixated
more in the first fixation than in the subsequent fixations (see
Fig. 2).
3.5. Analysis based on coordinate grid a priori parsing
3.5.1. Percentage of dwell time
The natural, but coarse part-based analysis could not provide
information according to whether there were specific locations
within each part that actually captured the participants’ attention.
Also, the time span of the fixations and the size of the areas were
not accounted for. We therefore measured the percentage of time
the gaze dwelled within each small segment of a finer-grid, in each
of the two extreme views (front and profile) and corrected for size
of segment. A ‘heat diagram’ of this analysis is illustrated in Fig. 3.
Dark blue represents 0% dwell time and dark red approximately 5%
dwell time in the frontal view, and approximately 9.5% dwell time
in profile view.
The analysis in terms of dwell time distributions principally
resembled those of the fixation distribution analysis. Differences
could be accounted for by the use of a normalized measure, as well
as the inclusion of fixation pauses in the dwell time distribution,
but not in the distribution-of-fixation analysis. Specifically, the
main effect of Facial Parts was confirmed (F(8, 368) = 340.58,
p < 0.0001) but with more attention towards the nose (29.65%,

Fig. 2. The percentage of first, second and third fixations on the listed Facial Parts within frontal, intermediate view (22.5°), three-quarter view (45°) and profile views. The
bars indicate 95% confidence intervals. The results indicate that the central parts of the face in each view (e.g. nose in front, eyes/cheek in intermediate view, eyes/cheek in
Three-quarter view, and cheek/hair in profile) were fixated more in the first fixation than in the subsequent fixations.
 L. Sther et al. / Vision Research 49 (2009) 2870–2880 2875

Fig. 3. An illustration of percentage of time (corrected for segment size) spent in each pre-selected facial segment within front and profile view, where dark red represents the
highest amount of dwell time. The most attended part in frontal view was the right lower nasal eye (4.99% dwell time), and in profile: the lower central eye (9.52% dwell
time). The green circle illustrates 2° of visual angle, corresponding to the fovea, and the bottom left image shows the numbers given to the central segments.

SD = 13.5) followed by the eyes (24.32%, SD = 6.86) and the cheek Table 1
(24.32%, SD = 13.04). The interactive effect between Facial Parts The means and standard deviations of the most attended segments within the most
attended Facial Parts.
and View (F(8, 368) = 69.94, p < 0.0001) was also confirmed.
As the dwell time distribution depended upon the exact choice of Part Segment Overall Frontal Profile
segments included within each part, an analysis over all central seg- Mean SD Mean SD Mean SD
ments was also run. A repeated-measures ANOVA with 25 segments
Eyes No. 6 2.85 1.28 5.71 2.56 0.00 0.00
(five nose segments, Nos. 1–5 in Fig. 3; nine eye segments, Nos. 6–14; No. 8 2.34 1.33 0.50 0.41 4.19 2.26
and 11 cheek segments, Nos. 15–25)  two Views (frontal and pro- No. 9 7.50 2.06 8.13* 2.13 0.00 0.00
file) showed a main effect of segments (F(24, 1104) = 33.50, No. 10 3.75 1.06 6.88 2.00
No. 11 4.32 4.28 1.10 1.03 7.55 7.53
p < 0.0001) with significantly more dwell time overall on the seg-
No. 12 9.16 2.99 8.80* 2.53 0.00 0.00
ment between the eye and the nose (infraorbital margin, segment No. 13 2.86 1.07 9.52* 3.14
No. 15) than any other segment (mean = 7.14, SD = 3.84). This was No. 14 4.22 1.86 0.79 0.68 7.66 3.05
mainly caused by the fact that this segment was one of the few highly Nose No. 1 4.13 1.62 6.01 1.40 2.25 1.85
attended segments that was visible in both views. No. 2 5.43 2.12 7.15 1.68 3.70 2.56
An interaction between Segments and View (F(24, 1104) = No. 3 4.62 1.90 6.58 2.03 2.67 1.78
54.28, p < 0.0001) was more informative according to specific loca- No. 4 3.76 1.88 5.23 2.06 2.29 1.70

tions within each attended facial part. Table 1 presents the means Infra-orbital No. 15 7.14 3.84 7.30 3.05 6.97 4.55
of the most attended segments in both presented Angles of View. No. 16 3.57 1.78 3.19 1.31 3.96 2.26
No. 17 3.82 1.50 2.28 1.03 5.37 1.98
In frontal view the overall most attended segments was the two
No. 20 2.00 1.18 0.54 0.53 3.47 1.83
lower nasal eye segments (No. 12 and No. 9). In profile presenta-
*
tions the eye dwelled longer on the lower central eye segment Attendance was significantly higher than attendance to any other segment in the
same Angle of View (p < 0.0005).
(No. 13). These two most attended eye segments (No. 12 in the full
face, and No. 13 in profile) did not differ significantly, and they
actually constituted the same segment when the face was turned,
tween the eye, the cheek and the nose. This would seem to
as the lower central eye concealed the lower nasal eye in profile.
constitute a preferred infraorbital region of the face, but biased to-
Therefore a collapsed segment, representing the lower nasal eye
wards the eye.
in both views, was compared to the infraorbital margin (No. 15).
A main effect (F(1, 47) = 9.67, p = 0.003) showed that the collapsed
lower nasal eye segment was significantly more attended to than 3.6. A posteriori analysis of eye movements
the infraorbital margin. To conclude, in frontal views participants
paid most attention towards the lower nasal eye and an area be- The value of the previous analysis might be dependent upon the
tween the eyes and the nose. In profile, attention was directed be- level of resolution, shape and number of segments used in the ‘a
2876 L. Sther et al. / Vision Research 49 (2009) 2870–2880

priori’ analysis design. As such, it may also, to some extent, be
based on arbitrary decisions. In this section we present a novel
way of analysing eye movement data which we label a posteriori
eye movement analysis. A computer program, described in the
method section (Analysis Procedure) created a prototypical stimu-
lus face (morph) which was overlaid by a ‘heat map’. In other
words, the eye itself was used, metaphorically, as a ‘paintbrush’
(see Fig. 4) or alternatively as an ‘incremental spotlight’ (see
Fig. 5), both illustrating the most frequent viewed locations of
the direction of gaze over the whole facial area.
As can be appreciated by inspecting these Figures, these dis-
plays are very consistent with the findings based on the fine-
grained a priori part analysis, previously described. However, the
present a posteriori analysis gives a more veridical (point by point)
picture of the participants’ direction of gaze. An even finer-grained
analysis grid than the one we used, might yet display the same re-
sults as an a posteriori analysis would. Fig. 4 presents the probabil-
ity distribution function that we obtained by plotting the eye-track
positions on a prototypical face. The scale ranges from the area
most likely to be attended (red) to the area least likely to be at-
tended (blue), and the maximum and minimum probability densi-
ties for each analysis representation are included in the heat scales
of the figure. It can be determined that red is 20 times more prob-
able than blue, two times more probable than green and 1.3 times
the probability of yellow. Likewise, yellow is 15.2 times more prob-
able than blue; green is 10.4 times more probable than blue and
cyan is 5.6 times more probable than blue.
As Fig. 4 illustrates, only the central (colored) area of the face
actually corresponded to an attended area. In addition the figure
shows that the facial area that received most fixations (red) was lo-
cated below the pupil, and in all cases towards the central merid-
ian, specifically the infraorbital area of the face. Notably, the most
attended area appears to be more directed towards the center of
the face in the present analysis than in the fine-grained a priori
analysis. The os nasale (the nasal bone) and lateral cartilage (nasal
bridge) were captured by these high levels of attention in all views,
whereas the zygomatic area was more attended to in profile view.
The alar cartilage (lower part of the nose) and the upper regio oralis
(beneath the nose) was attended to in frontal view. The surround-
ing central area was much less attended (blue). In the intermediate

Fig. 4. Heat diagrams of the most attended areas (a posteriori analysed) over a morphed face in four different views (00°, 22.5°, 45° and 90°). The left column shows where all
participants placed their overt attention. The attention patterns of female participants (middle column) and male participants (right column) were significantly different from
each other. The scale ranges from the area most likely to be attended (red) to the area least likely to be attended (blue). These maps can be viewed as probability density
functions, and the maximum and minimum probability for each analysis is included in the heat scales. The green circle illustrates 2° of visual angle, corresponding to the
fovea. The left-turned morphs serve only as a consistent illustration, as in the experiment the stimuli faced the left equally often as the right.
 L. Sther et al. / Vision Research 49 (2009) 2870–2880 2877

Fig. 5. Facial areas that participants overtly attended when looking at female (left) or male (right) faces. These images can be described as ‘‘spotlight” versions of the
probability density functions in Fig. 4. Here, black areas indicate regions of the image where gaze was directed with less than 0.05 probability. The brightest areas represent
the regions that were most probable to be attended (analogous to dark red in the heat maps) and the darkest of the illuminated areas represent the minimum probability for
each analysis (corresponding to dark blue in the heat maps). The left face is a morph of 12 females, and the right face is a morph of 12 males. The morphs only serve the
present illustration and were never shown in the experiment.

(22.5°) view, where participants were faster in responding, atten-
tion appeared to be mostly directed or biased towards one eye
(the closest of the eyes) followed by the nasal bridge and zygo-
matic area.
The overall strategies of male and female participants differed
slightly as males focused their visual attention a bit lower, and
more towards the center than females did (p < 0.004). Specifically,
males offered more attention towards the nasal area in all angles
except in profile. Conversely, males attended more towards the
cheek area (regio infaorbitalis and buccalis) than females in all an-
gles except in frontal presentations. In other words, males attended
more towards the view-specific center-of-gravity, whereas females
attended more towards the eyes, regardless of the view.
Fig. 5 illustrates the same findings, but with undisguised facial
features. In this illustration only the illuminated area of the face
corresponds markedly to the direction of the gaze. The illustrations
can be described as ‘spotlight’ versions of the probability density
functions of Fig. 4. The maximum and minimum probability for
each analysis, included in the heat scales of Fig. 4, corresponds to
the brightest and darkest degree of illumination in Fig. 5. Here,
the brightest areas represent the regions that were mostly at-
tended (corresponding to dark red in Fig. 4) and the dark gray areas
represent the least attended areas (corresponding to dark blue in
Fig. 4). Black, on the other hand, represents areas with less than
0.05 likelihood of being attended, corresponding to non-colored
areas in the heat maps. The results indicate that it could be suffi-
2878 L. Sther et al. / Vision Research 49 (2009) 2870–2880
cient to focus within these illuminated face areas to solve the sex
categorization task efficiently, although the remaining face areas
might also contribute peripherally.
4. Discussion
The internal parts of the face (eyes, nose and cheeks) were
overtly attended to more than the rest of the face, and this oc-
curred in all views. Both the distribution and the duration of the
fixations showed that the relative importance of these parts de-
pended upon Angle of View, indicating that visual attention pre-
dominantly targeted a preferred infraorbital region of the face in
addition to the eyes. A finer-grid parsing, and an a posteriori anal-
ysis, revealed that the most attended segments within the central
areas and the eye area were adjacent to each other. Thus, the cen-
ter of the most attended area for frontal presentations was located
between the nose and the eye (infraorbital margin), and in profile
presentations between the eye and the zygomatic bone (infraor-
bital foramen). The three-quarter views (22.5° and 45°) confirmed
the pattern that the most attended area in all views was infraor-
bital. This was particularly true for the first fixation, as central
areas were progressively less attended to resulting in less preferen-
tial fixating towards specific parts in the second and third fixations.
Males used the ‘central’ strategy to a higher degree and with less
weight on the eyes, than females.
The results of Hsiao and Cottrell (2008) indicated that people
fixate the center of the nose in frontal views when recognizing a
face. Correspondingly, Tyler and Chen (2006) identified a region
centered at the nose bridge in the full face as a zone of specialized
perceptual processing in a forced choice detection paradigm. One
could expect the same to be true for sex categorizations of frontal
faces based on the consensual notion that faces are perceived
holistically or at least with less parsing than for other objects. Fol-
lowing this train of thought, fixations would be expected to land
centrally in other perspectives as well instead of landing consis-
tently in a position close the nose. The present results partially
support these expectations, since a ‘central’ position was attended
more than the nose in three-quarter views and even in profile
views.
Previous research has emphasized the diagnostic value of the
eye region (Althoff & Cohen, 1999; Henderson et al., 2005; Minut
et al., 2000). However, attention towards the eyes may be inter-
preted as a strategy of bringing a small, diagnostic central region
of the face into focus, while larger parts can be viewed parafoveal-
ly. The present study also found eye fixations in frontally viewed
faces, but these were shown to be on the nasal eye segment, in
agreement with Schyns et al. (2002) using the Bubbles technique.
This finding may support an interpretation which states that eye
attention not only indicates the eyes’ diagnostic value, but also
functions as a central landmark from which the larger dimorphic
nasal area can be viewed parafoveally.
Hsiao and Cottrell (2008) observed that eyes were only fixated
in the third fixation in full faces and that performance did not im-
prove after the second fixation. In the present study eyes were at-
tended as much as noses in the second, but not in the first, fixation
in full faces. This could have been caused by the change of task, as
sex categorization decisions are simpler and more efficiently per-
formed than identity decisions (Bruce et al., 1987; Schendan
et al., 1998). In addition, for intermediate and 3=4 Angles of View,
the most attended parts in the first fixation were eyes and cheeks,
whereas cheeks were more attended to in the first fixation on pro-
file faces. This finding again indicates the use of a central fixation
strategy.
The main finding in the present study was the increased atten-
tion towards the infraorbital areas in all views. This can be inter-
preted as a gaze strategy that selects a central ‘anchor point’
which is biased towards an intermediate position between the
eye and the nose in front, and gradually more between the eye,
nose and cheek as the head turns. An important remark to make
is that the observed center of overt attention (infraorbital areas)
in the present study was not a sexually dimorphic part in itself.
In fact, it has been proposed that the largest anatomical sexual
dimorphism in the human face is represented by the protuberance
of the nose (Burton et al., 1993; Enlow, 1982; Enlow, 1990; O’Toole
et al., 1997). Crucially, we assumed that sexually dimorphic infor-
mation would be more visible in other views than the frontal,
which however resulted in lower levels of fixation (e.g. the nasal
bridge is more visible in profile, but was more attended to in fron-
tal views; and the zygomatic protrusion is better visible in three-
quarter views, but was more attended to in profile). The position-
ing of gaze on the infraorbital areas might play a role for the spe-
cific sex categorization task so that the fixations might be
considered a ‘perspective anchoring’ which directs the more
dimorphic internal region of the face into parafoveal vision. How-
ever, the present results do not reject the possibility of holistic pro-
cessing that is unspecific to the perceptual task.
Interestingly, the COG was intensively attended in frontal view
but gradually less attended as the head turned, and the least at-
tended in profile view, where the COG did not correspond to the
center of the face or to any dimorphic part. A possible explanation
for this finding is that the head volume or outline is not used by the
visual system to compute its COG, but rather another conceptual-
ization of COG may be employed. Previous eye monitoring re-
search, in object recognition tasks, has shown that the observers’
gaze often lands near a COG of the luminance distribution in the
display, instead of individual elements (e.g., Findlay, 1982; Kowler
& Blaser, 1995; McGowan, Kowler, Sharma, & Chubb 1998). Specif-
ically, when the stimuli are 3-D objects casting shadows (compara-
ble to our photographic face stimuli) the COG of gaze has been
shown to be biased towards the informative parts (Vishwanath &
Kowler, 2004). It has also been shown that the tendency to fixate
on the COG is much less marked on tasks allowing for more volun-
tary saccadic movements (Findlay & Blythe, 2009; He & Kowler,
1991) like the task in the present experiment. Considering that ear-
lier studies have shown that low spatial frequency information is
sufficient for the sex categorization task (Abdi et al., 1995; Schyns
et al., 2002; Valentin et al., 1997) it seems consistent to conclude
that in the present study, fixations were directed to a gravity point
that provided the best sampling of the internal regions of the face
within foveal/parafoveal vision.
The weighting of a central point of gravity towards a position
between the nose and the eye might also be understood as an opti-
mal viewing strategy that brings the smaller parts (e.g. the eye)
into the highest focus; while at the same time larger informative
parts (e.g. the nose) can be viewed parafoveally. This is in line with
the findings of Schyns et al. (2002) that the nose was more diag-
nostic in coarser frequency scales, and the eye in finer scales. Since
the mean fixation position within the eye area in all views was
underneath the eye itself and fixation on the brows, previously
shown to be more diagnostic than eyes (Brown & Perrett, 1993;
Campbell, Benson, Wallace, Doesbergh, & Coleman, 1999; Camp-
bell, Wallace, & Benson, 1996; Sadr, Jarudi, & Sinha, 2003) were ab-
sent, it is unlikely that the mean fixation position in the present
study corresponds to the maximal informational nexus or diagnos-
tic trait of faces. In fact, what is emerging from the present research
is that the locus of eye fixations might generalize over tasks (cf.
Carmel & Bentin, 2002; Deaner & Platt, 2003; Grosbras et al.,
2005; Hsiao & Cottrell, 2008; Kingstone et al., 2004; Schyns, Jen-
tzsch, Johnson, Schweinberger, & Gosselin, 2003; Smith, Gosselin,
& Schyns, 2004) and that task-dependent effects will be revealed
as subtle displacements from a central, general, anchoring point.
 L. Sther et al. / Vision Research 49 (2009) 2870–2880
Intuitively, the eyes, in particular the contrast between the
white sclera and the colored iris, provide a very salient low-level
visual cue that could be used to pilot the initial ballistic eye move-
ment towards the face (cf. Brunelli & Poggio, 1993; Langton, Watt,
& Bruce, 2000). In other words, the eyes may form a horizontal
meridian that facilitates alignment of the facial image to internal
templates, thus constituting a highly informative point for the rec-
ognition of the face and, perhaps, other perceptual judgments to
faces. Indeed, we observed a rather uniform distribution of atten-
tion towards the eye region regardless of the view, even though
it would seem intuitive that the eyes would be less informative
in a profile view where most of its shape is occluded and only
one eye is visible. Additionally, the social significance of gaze
should also be considered a possible explanation for the eye bias.
Previous research has shown that direct gaze leads to longer RTs
in a sex categorization task, especially for faces of opposite sex
from the observer (Vuilleumier, George, Lister, Armony, & Driver,
2005). This could indicate that a social aspect is interfering with
the task, especially in frontally viewed faces where the gaze is
more direct. Eye-tracking studies of monkeys have shown that
gaze dwells on pictures of other monkeys’ faces longer than on
other objects or scenes, and that the eye region receives most of
their fixations (e.g. Guo, Mahmoodi, Robertson, & Young, 2006;
Guo, Robertson, Mahmoodi, Tadmor, & Young, 2003).
One should also add that categorization performance was high-
er for the intermediate view than the other views. The superiority
of this particular view seems well accounted for by the fact that
both the shape and size of the nose are optimally visible in this
view (Chronicle et al., 1995) and that the size and gradient of cur-
vature of the eyes and cheekbones are also easily appreciated. Thus
a strategy of anchoring fixations infraorbitally, between the nose,
the eye and the cheek, might be rather optimal for sex categoriza-
tion and perhaps for other face perception tasks as well (e.g. iden-
tification; cf. Laeng & Caviness, 2001; Laeng & Rouw, 2001).
The present findings also indicated that female and male partic-
ipants used slightly different strategies, because females preferably
attended towards the eyes, whereas males displaced their atten-
tion more towards a lower central location of the face. This may
show that both strategies are useful, since males and females
solved the task with the same accuracy and speed. We surmise that
males use a more global gaze strategy than females. This might be
understood in terms of gender specific experiences. However, fe-
males have been found to have greater interest in social aspects
than males (Kaplan, 1978) and this could be linked to feminine
gender schemata (Bem, 1981) which often involve interest in the
‘‘people dimension” (Lippa, 1998). The models’ gaze could be inter-
preted by the female observer as a social aspect (Vuilleumier et al.,
2005). The present finding could therefore be understood as a so-
cial mechanism external to the specific task demands. Future stud-
ies may clarify whether the two sexes do use gaze in strategically
different ways.
Compared to some other eye-tracking studies on face perception,
our findings did not reveal any marked preference for either the left
or right side of the face (Butler et al., 2005; Leonards & Scott-Samuel,
2005) or for the left eye (Vinette et al., 2004). One possibility is that
this null finding indicates that for a sex categorization task the right
side of the face is as equally important as the left side of the face.
Alternatively, since eye-tracking studies with face stimuli (including
the present one) have used rather small sets of faces, random varia-
tions between faces might have caused spurious asymmetries. How-
ever, Butler et al. (2005) used a sex categorization task and found a
left-sided bias despite the fact that their stimuli consisted of chime-
ric faces where the left and right side of each face were identical. An-
other possibility is that individuals largely differ in their perceptual
biases in facial inspection and, consequently, that a specific direc-
tional bias might not be a particularly robust effect.
2879
Finally, the present experiment suggested that an a priori pars-
ing of the face might give a rather incomplete understanding of the
eye’s scan paths during a perceptual task, since parsing strongly
depends upon an arbitrarily chosen number, size and shape of
the segments, though if the parsing is sufficiently fine-grained,
such a method might still be functional.
Acknowledgments
We would like to thank Øyvind Sther for recruiting and test-
ing some of the participants, and Tore Morten Andreassen for his
assistance with the initial data analysis.
References
Abdi, H., Valentin, D., Edelman, B., & O’Toole, J. A. (1995). More about the difference
between men and women: Evidence from linear, neural networks and the
principal component approach. Perception, 24, 539–562.
Althoff, R. R., & Cohen, N. J. (1999). Eye-movement-based memory effect: A
reprocessing effect in face perception. Journal of Experimental Psychology:
Learning Memory and Cognition, 25(4), 997–1010.
Bem, S. L. (1981). Gender schemata theory: A cognitive account for sex typing.
Psychological Review, 88(4), 354–364.
Biederman, I., & Kalocsai, P. (1997). Neurocomputational basis of object and face
recognition. Philosophical Transactions of the Royal Society of London: B, 352,
1203–1219.
Biederman, I., & Shiffrar, M. (1987). Sexing day-old chicks: A case study and expert
system analysis of a difficult perceptual-learning task. Journal of Experimental
Psychology: Learning Memory and Cognition, 13, 640–645.
Blanz, V., Tarr, M. J., & Bühltoff, H. H. (1999). What object attributes determine
canonical views? Perception, 28(5), 575–599.
Brown, E., & Perrett, D. I. (1993). What gives a face its gender? Perception, 22,
829–840.
Bruce, V., Burton, A. M., Hanna, E., Healey, P., Mason, O., Coombes, A., et al. (1993).
Sex discrimination: How do we tell the difference between male and female
faces? Perception, 22, 131–152.
Bruce, V., Ellis, H., Gibling, F., & Young, A. W. (1987). Parallel processing of the sex
and familiarity of faces. Canadian Journal of Psychology, 41, 510–520.
Bruce, V., & Young, A. (1998). In the eye of the beholder. Oxford, UK: Oxford
University Press.
Brunelli, R., & Poggio, T. (1993). Face recognition: Features versus templates. IEEE
Transactions on Pattern Analysis and Machine Intelligence, 15, 1042–1052.
Burton, A. M., Bruce, V., & Dench, N. (1993). What’s the difference between men and
women? Perception, 22, 153–176.
Butler, S., Gilchrist, I. D., Burt, D. M., Perrett, D. I., Jones, E., & Harvey, M. (2005). Are
the perceptual biases found in chimeric face processing reflected in eye-
movement patterns? Neuropsychologia, 43, 52–59.
Campbell, R., Benson, P. J., Wallace, S. B., Doesbergh, S., & Coleman, M. (1999). More
about brows: How poses that change brow position affect perceptions of
gender. Perception, 28, 489–504.
Campbell, R., Wallace, S., & Benson, P. J. (1996). Real men don’t look down: Direction
of Gaze affects sex decisions on faces. Visual Cognition, 3(4), 393–412.
Carmel, D., & Bentin, S. (2002). Domain specificity versus expertise: Factors
influencing distinct processing of faces. Cognition, 83, 1–29.
Chronicle, E. P., Chan, M.-Y., Hawkings, C., Mason, K., Smethurst, K., Stallybrass, K.,
et al. (1995). You can tell by the nose – Judging sex from an isolated facial
feature. Perception, 24, 969–973.
Coren, S., & Hoenig, P. (1972). The effect of non-target stimuli on length of voluntary
saccades. Perceptual and Motor Skills, 34, 499–508.
Cotrell, G., & Flemming, M. K. (1990). Face recognition using unsupervised feature
extraction. In Proceedings of the international neural network conference, Paris
(pp. 322–325). Dordrecht: Kluwer.
Dailey, M. N., & Cottrell, G. W. (1999). Organization of face and object recognition in
modular neural network models. Neural Networks, 12(7–8), 1053–1173.
Deaner, R. O., & Platt, M. L. (2003). Reflexive social attention in monkeys and
humans. Current Biology, 13, 1609–1613.
Enlow, D. H. (1982). Handbook of facial growth. Philadelphia, PA: W.B. Saunders.
Enlow, D. H. (1990). Facial growth (3rd ed.). Philadelphia, PA: W.B. Saunders.
Findlay, J. M. (1982). Global visual processing for saccadic eye movements. Vision
Research, 22, 1033–1045.
Findlay, J. M., & Blythe, H. I. (2009). Saccade target selection: Do distractors affect
saccade accuracy? Vision Research, 49(10), 1267–1274.
Fisher, G. H., & Cox, R. L. (1975). Recognizing human faces. Applied Ergonomics, 6(2),
104–109.
Gauthier, I., Tarr, M. J., Anderson, A. W., Skudlarski, P., & Gore, J. C. (1999). Activation
of the middle fusiform ‘face area’ increase with expertise in recognizing novel
objects. Nature Neuroscience, 2, 568–573.
Golomb, B. A., Lawrence, D. T., & Sejnowski, T. J. (1991). Sexnet: A neural network
identifies sex from human face. In R. P. Lippman, J. Moody, & D. S. Touretzky
(Eds.). Advances in neural information processing systems (Vol. 3, pp. 572–577).
San Mateo, CA: Morgan Kaufmann.
2880 L. Sther et al. / Vision Research 49 (2009) 2870–2880
Gray, M., Lawrence, D. T., Golomb, B. A., & Sejnowski, T. J. (1995). A perceptron
reveals the face of sex. Neural Computation, 7, 1160–1164.
Grosbras, M.-H., Laird, A. R., & Paus, T. (2005). Cortical regions involved in eye
movements, shifts of attention, and gaze perception. Human Brain Mapping, 25,
140–154.
Guo, K., Mahmoodi, S., Robertson, R. G., & Young, M. P. (2006). Longer fixation
duration while viewing face images. Experimental Brain Research, 171(1),
91–98.
Guo, K., Robertson, R. G., Mahmoodi, S., Tadmor, Y., & Young, M. P. (2003). How do
monkeys view faces? – A study of eye movements. Experimental Brain Research,
150(3), 363–374.
He, P., & Kowler, E. (1991). Saccadic localization of eccentric forms. Journal of the
Optical Society of America A – Optics Image Science and Vision, 8(2), 440–449.
Henderson, J. M., Williams, C. C., & Falk, R. J. (2005). Eye movements are functional
during face learning. Memory and Cognition, 33(1), 98–106.
Hoffman, J., & Richards, W. A. (1985). Parts of recognition. Cognition, 18, 65–96.
Hsiao, J. H., & Cottrell, G. (2008). Two fixations suffice in face recognition.
Psychological Science, 19(10), 998–1006.
Ikeda, T., Nakamura, M., & Itoh, M. (1999). Sex differences in the zygomatic angle in
Japanese patients analyzed by MRI with reference to Moire fringe patterns.
Aesthetic Plastic Surgery, 23, 349–353.
Itti, L., & Koch, C. (2000). A saliency-based search mechanism for overt and covert
shifts of visual attention. Vision Research, 40, 1489–1506.
Kaplan, H. B. (1978). Sex differences in social interest. Journal of Individual
Psychology, 34(2), 206–209.
Kingstone, A., Tipper, C., Ristic, J., & Ngan, E. (2004). The eyes have it!: An fMRI
investigation. Brain and Cognition, 55, 269–271.
Kowler, E., & Blaser, E. (1995). The accuracy and precision of saccades to small and
large targets. Vision Research, 35, 1741–1754.
Kundel, H. L., Nodine, C. F., Thickman, D., & Toto, L. (1987). Searching for lung
nodules. A comparison of human performance with random and systematic
scanning models. Investigative Radiology, 22, 417–422.
Laeng, B., & Caviness, V. (2001). Prosopagnosia as a deficit in encoding curved
surface. Journal of Cognitive Neuroscience, 13(5), 556–557.
Laeng, B., & Rouw, R. (2001). Canonical views of faces and the cerebral hemispheres.
Laterality, 6(3), 193–224.
Land, M., & McLeod, P. (2000). From eye movements to actions: How batsmen hit
the ball. Nature Neuroscience, 3, 1340–1345.
Langdell, T. (1978). Recognition of faces – Approach to study of autism. Journal of
Child Psychology and Psychiatry, 19(3), 255–268.
Langton, S. R. H., Watt, R. J., & Bruce, V. (2000). Do the eyes have it? Cues to the
direction of social attention. Trends in Cognitive Sciences, 4, 50–59.
Leonards, U., & Scott-Samuel, N. E. (2005). Idiosyncratic initiation of saccadic face
exploration in humans. Vision Research, 45(20), 2677–2684.
Lippa, R. (1998). Gender related individual differences and the structure of
vocational interest: The importance of the people-things dimension. Journal of
Personality and Social Psychology, 74(4), 996–1009.
Malcolm, G. L., Lanyon, L. J., Fugard, A. J. B., & Barton, J. J. S. (2008). Scan patterns
during the processing of facial expression versus identity: An exploration of
task-driven and stimulus-driven effects. Journal of Vision, 8(8), 1–9. 2.
McGowan, J. W., Kowler, E., Sharma, A., & Chubb, C. (1998). Saccadic localization of
random dot targets. Vision Research, 38, 895–909.
Melcher, D., & Kowler, E. (1999). Shapes, surfaces and saccades. Vision Research,
39(17), 2929–2946.
Minut, S., Mahadevan, S., Henderson, J. M., & Dyer, F. C. (2000). Face recognition
using foveal vision. Lecture Notes in Computer Science, 1811, 424–433.
O’oole, A. J., Vetter, T., Troje, N. F., & Bülthoff, H. H. (1997). Sex classification is better
with three-dimensional head structure than with image intensity information.
Perception, 26, 75–84.
O’Regan, J. K., Lèvy-Schoen, A., Pynte, J., & Brugaillère, B. (1984). Convenient
fixation location within isolated words of different length and structure.
Journal of Experimental Psychology: Human Perception and Performance, 10(2),
250–257.
O’Toole, A. J., Deffenbacher, K. A., Valentin, D., McKee, K., Huff, D., & Abdi, H. (1998).
The perception of face gender: The role of stimulus structure in recognition and
classification. Memory and Cognition, 26, 146–160.
O’Toole, A. J., Millward, R. B., & Anderson, J. A. (1988). A physical system approach to
recognition memory for spatially transformed faces. Neural Networks, 1, 179–199.
O’Toole, A. J., Peterson, J., & Deffenbacher, K. A. (1996). An ‘‘other race effect” for
classifying faces by sex. Perception, 25, 669–676.
Putz, R., & Pabst, R. (2001). Head, neck, upper limb. Sobotta, atlas of human anatomy
(Vol. 1). Germany: Appl, Wemding, Urban & Fischer.
Reddy, L., Wilken, P., & Koch, C. (2004). Face-gender discrimination is possible in the
near-absence of attention. Journal of Vision, 4, 106–117.
Rehder, B., & Hoffman, A. B. (2005). Eyetracking and selective attention in category
learning. Cognitive Psychology, 51, 1–41.
Reichle, E. D., Rayner, K., & Pollatsek, A. (2003). The E-Z Reader model of eye-
movement control in reading: Comparisons to other models. Behavioral and
Brain Sciences, 26, 445–526.
Sadr, J., Jarudi, I., & Sinha, P. (2003). The role of eyebrows in face recognition.
Perception, 32(3), 285–293.
Sther, L., & Laeng, B. (2008). On facial expertise: Processing strategies of twins’
parents. Perception, 37, 1227–1240.
Schendan, H. E., Ganis, G., & Kutas, M. (1998). Neurophysiological evidence for
visual perceptual categorization of words and faces within 150 ms.
Psychophysiology, 35, 240–251.
Schwarzer, G., Huber, S., & Dümmler, T. (2005). Gaze behavior in analytical and
holistic face processing. Memory and Cognition, 33, 344–354.
Schyns, P. G., Bonnar, L., & Gosselin, F. (2002). Show me the features! Understanding
recognition from the use of visual information. Psychological Science, 13(5),
402–409.
Schyns, P. G., Jentzsch, I., Johnson, M., Schweinberger, S. R., & Gosselin, F. (2003). A
principled method for determining the functionality of ERP components.
NeuroReport, 14, 1665–1669.
Sergent, J., & Hellige, J. B. (1986). Role of input factors in visual-field asymmetries.
Brain and Cognition, 5, 174–199.
Sergent, J., Ohta, S., & MacDonald, B. (1992). Functional neuroanatomy of face and
object processing. Brain, 115, 15–36.
Smith, M. L., Gosselin, F., & Schyns, G. (2004). Receptive fields for flexible face
categorizations. Psychological Science, 15(11), 753–761.
Stroustrup, B. (2000). The C++ programming language (3rd ed.). Addison-Wesley.
The iView XTM RED system from SMI. <http://www.smivision.com/en/gaze-eye-
tracking-systems/products/iview-x-red-red250.html>.
Torralba, A., Oliva, A., Castlhano, M. S., & Henderson, J. M. (2006). Contextual
guidance of eye movements and attention in real-world scenes: The role of
global features in object search. Psychological Review, 113, 766–786.
The QT product by trolltech. <http://www.trolltech.com/products/qt/index.html>.
Turano, K. A., Geruschat, D. R., & Baker, F. H. (2003). Oculomotor strategies for the
direction of gaze tested with real-world activity. Vision Research, 43, 333–346.
Tyler, C. W., & Chen, C. C. (2006). Spatial summation of face information. Journal of
Vision, 6(10), 1117–1125.
Underwood, G., Chapman, P., Brocklehurst, N., Underwood, J., & Crundall, D. (2003).
Visual attention while driving: Sequences of eye fixations made by experienced
and novice drivers. Ergonomics, 46, 629–646.
Valentin, D., Abdi, H., Edelman, B., & O’Toole, A. J. (1997). Principal component and
neural network analyses of face images: What can be generalized in gender
classification? Journal of Mathematical Psychology, 41, 398–413.
Vinette, C., Gosselin, F., & Schyns, P. G. (2004). Spatio-temporal dynamics of face
recognition in a flash: It’s in the eyes. Cognitive Science, 28, 289–301.
Vishwanath, D., & Kowler, E. (2004). Saccadic localization in the presence of cues to
three-dimensional shape. Journal of Vision, 4, 445–458.
Vishwanath, D., Kowler, E., & Feldman, J. (2000). Saccadic localization of occluded
targets. Vision Research, 40(20), 2797–2811.
Vuilleumier, P., George, N., Lister, V., Armony, J., & Driver, J. (2005). Effects of
perceived mutual gaze and gender on face processing and recognition memory.
Visual Cognition, 12(1), 85–101.
Walker-Smith, G. J., Gale, A. G., & Findlay, J. M. (1977). Eye-movement strategies
involved in face perception. Perception, 6(3), 313–326.
Wolfe, J. M., Cave, K. R., & Franzel, S. L. (1989). Guided search: An alternative to the
feature integration model for visual search. Journal of Experimental Psychology:
Human Perception and Performance, 15, 419–433.
Yarbus, A. L. (1967). Eye movements and vision. New York: Plenum Press.