Brain Research 1644 (2016) 88–97
Contents lists available at ScienceDirect
Brain Research
journal homepage: www.elsevier.com/locate/brainres
Research report
The relationship between visual word and face processing
lateralization in the fusiform gyri: A cross-sectional study
Jodie Davies-Thompson a,b, Samantha Johnston a, Yashar Tashakkor a, Raika Pancaroglu a,
Jason J.S. Barton a,n
a
Human Vision and Eye Movement Laboratory, Departments of Medicine (Neurology), Ophthalmology and Visual Sciences, University of British Columbia,
Vancouver, Canada
b
Crossmodal Perception and Plasticity Laboratory, Center of Mind/Brain Sciences, University of Trento, Italy
art ic l e i nf o
a b s t r a c t
Article history:
Received 14 October 2015
Received in revised form
26 April 2016
Accepted 5 May 2016
Available online 10 May 2016
Visual words and faces activate similar networks but with complementary hemispheric asymmetries,
faces being lateralized to the right and words to the left. A recent theory proposes that this reflects
developmental competition between visual word and face processing. We investigated whether this
results in an inverse correlation between the degree of lateralization of visual word and face activation in
the fusiform gyri. 26 literate right-handed healthy adults underwent functional MRI with face and word
localizers. We derived lateralization indices for cluster size and peak responses for word and face activity
in left and right fusiform gyri, and correlated these across subjects. A secondary analysis examined all
face- and word-selective voxels in the inferior occipitotemporal cortex. No negative correlations were
found. There were positive correlations for the peak MR response between word and face activity within
the left hemisphere, and between word activity in the left visual word form area and face activity in the
right fusiform face area. The face lateralization index was positively rather than negatively correlated
with the word index. In summary, we do not find a complementary relationship between visual word
and face lateralization across subjects. The significance of the positive correlations is unclear: some may
reflect the influences of general factors such as attention, but others may point to other factors that
influence lateralization of function.
& 2016 Elsevier B.V. All rights reserved.
Keywords:
fMRI
Hemispheric specialization
Fusiform face area
Visual word form area
1. Introduction
Face and visual word recognition are both examples of expert
visual processing, requiring fine discriminations between highly
similar stimuli. Neuroimaging studies show that the processing of
visual words and faces involves networks that include regions in
inferior temporal cortex that respond more to faces or visual
words than to any other categories of objects. The ‘fusiform face
area’ (FFA) in the mid portion of the fusiform gyrus (Kanwisher
et al., 1997a, 1997b), is thought to be involved in the processing of
unique facial identity (Haxby et al., 2000a, 2000b), while the ‘visual word form area’ (VWFA) is another mid-fusiform region that
shows a selectivity for visually presented words (Cohen et al.,
2000).
A consistent observation about these networks and regions is that
they show a lateralized asymmetry. The FFA is more often identified
n
Correspondence to: Neuro-ophthalmology Section K, VGH Eye Care Centre,
2550 Willow Street, Vancouver, BC Canada V5Z 3N9.
E-mail address: [email protected] (J.J.S. Barton).
http://dx.doi.org/10.1016/j.brainres.2016.05.009
0006-8993/& 2016 Elsevier B.V. All rights reserved.
and larger in both size and magnitude of response in the right than in
the left hemisphere (Davies-Thompson and Andrews, 2012; Rossion
et al., 2012), while the VWFA (Cohen et al., 2000) is more often
identified in the left than the right hemisphere (Cohen et al., 2000,
2002a, 2002b; Dehaene and Cohen, 2011; Szwed et al., 2011). These
anatomic asymmetries have functional parallels. Tachistoscopic studies show a right visual field bias for words and a left field bias for
faces (Leehey and Cahn, 1979; Levine and Banich, 1982; Levine and
Koch-Weser, 1982), and the latter has been correlated with the degree of face activation in the right hemisphere on functional MRI
(Yovel et al., 2008). Studies using evoked-potentials consistently
show greater N170 potentials for words in the left occipital cortex
(Maurer et al., 2008; Mercure et al., 2011) and for faces in the right
occipital cortex (Scott, 2006). Neuropsychologically, the impaired face
recognition of prosopagnosia typically follows damage to bilateral or
right occipito-temporal cortex (de Renzi, 1986; Haxby et al., 2000a,
2000b; Sergent and Villemure, 1989), while the impaired reading of
alexia is associated with left occipito-temporal damage (Kawahata
et al., 1988; Sakurai et al., 1994).
Besides their complementary lateralization, there are additional relevant observations about the neural bases of face and
J. Davies-Thompson et al. / Brain Research 1644 (2016) 88–97
word processing. For one, both faces and words activate similar
bilateral networks, involving fusiform, lateral temporal and inferior frontal regions, among others (Barton et al., 2010a, 2010b;
Haxby et al., 2000a, 2000b). For another, given that the lateralization of face and word regions is only partial, there remains
significant overlap in each hemisphere between the regions activated by faces and those activated by words (Nestor et al., 2013).
These points have led to recent speculations that the lateralized
hemispheric specialization for words and faces evolves through
competition between face and word processing for the neural
resources in these object-recognition networks (Behrmann and
Plaut, 2013; Dundas et al., 2013). Thus, while the recycling hypothesis initially proposed that the acquisition of literacy is accompanied by co-opting of object-recognition resources by word
processing (Dehaene et al., 2010), recent accounts suggest that this
specifically targets neural substrates that might otherwise be devoted to face processing (Plaut and Behrmann, 2011). Efficiency
constraints to maximize local over long-range connections may
exert pressure to lateralize visual word processing to the left
hemisphere, where non-visual language processing is situated. In
this view, the right lateralization of face processing may follow as a
consequence of the left lateralization of word processing.
Other work has disputed the competition hypothesis. First,
there are observations that a right hemisphere preference to faces
is already present in newborns (de Heering and Rossion, 2015) and
in monkeys (Zangenehpour and Chaudhuri, 2005), and that visual
input to the right hemisphere during infancy is necessary for face
expertise to develop (Le Grand et al., 2003). These suggest that a
right hemisphere bias for faces may be innate.
Nevertheless, the presence of an innate right hemisphere bias
and the competition hypothesis are not necessarily mutually
exclusive phenomena. A competition for resources between
words and faces would predict that the left dominance of word
processing and the right dominance of face processing have a
non-arbitrary relationship in human subjects. If the neural resources of the left and right hemispheres for which faces and
words compete are finite, then one possible outcome would be
that the degree of lateralization of one would be inversely correlated with that of the other across a sample of the population.
That is, given that the degree of lateralization for words and faces
varies across subjects, one might expect that a competition that
resulted in more left and fewer right hemispheric resources being
devoted to visual words would also create a strongly lateralized
face processing system, with more right and less left hemispheric
activation by faces. On the other hand, a subject with more balanced visual word activation across left and right fusiform regions would also be expected to have a less asymmetric face
processing system.
Some suggestive evidence for this predicted relationship between word and face activation has been produced. A study reported that the magnitude of the N170 potential for faces in the
right hemisphere were positively correlated with those for words
in the left hemisphere (Dundas et al., 2014). However, another
study using functional MRI did not find a correlation between the
lateralization of the FFA and VWFA (Pinel et al., 2015). Hence the
issue is not yet settled. To test this prediction further, we examined
a cohort of right-handed literate subjects with functional neuroimaging. We used standard localizers of face and word activation to
determine if there was an inverse correlation across the cohort
between the left/right balance of visual word activity and that for
face activity.
89
2. Results
2.1. Same-ROI analysis
There was a positive correlation between the MR responses of
the peak voxels activated for faces and those for words in both the
left FFA (r¼ 0.68, p o0.001) and left VWFA (r ¼0.55, p o0.005).
Thus subjects with a greater response to faces had a greater response to words in these regions (Fig. 1A). Face and word MR responses were not correlated in the right FFA (r ¼0.17, p ¼0.42) but
there was a trend to a positive correlation in the right VWFA
(r ¼0.36, p ¼0.09).
When we examined the lateralization index (Fig. 1B), there was
a positive correlation between the degree of lateralization of face
activity and that for words within FFA regions (r ¼0.49, p o0.05),
but not within VWFA regions (r¼ 0.02, p ¼0.93). Including subjects’ handedness scores, cortical thickness, and cortical volume as
additional regressors did not significantly change the model
(P's 40.24) and showed again a positive correlation for the lateralization indices within the FFA (r¼ 0.45, p o0.05) but not in the
VWFA (r ¼ 0.02, p ¼0.93).
2.2. Different-ROIs analysis
This analysis examined the relationship between regions in
their response to their preferred stimulus (i.e. the response to faces in FFA regions versus the response to words in VWFA regions).
For the numbers of voxels activated by the localizers, there was no
correlation between the activation by words in VWFA regions and
the activation by faces in the FFA regions, in either the left
(r ¼ 0.14, p ¼ 0.49) or right hemisphere (r ¼0.09, p ¼0.65). For the
peak MR response, there was a positive correlation in the left
hemisphere between the scores for faces in FFA regions and those
for words in VWFA regions (r ¼0.55, p o0.005), but not in the
right hemisphere (r ¼0.32, p ¼ 0.11) (Fig. 2A).
The lateralization indices for the numbers of voxels did not
show any correlation between face activation in FFA regions and
word activation in VWFA regions. Including handedness and cortical volume thickness did not change the model (P's 40.52), but
including cortical thickness did, though the correlation was still
not significant (r ¼0.20, p ¼0.37). For the peak MR responses,
however, there was a positive correlation between the lateralization index for word activation in the VWFA regions and face activation in the FFA regions (r¼ 0.41, p o0.05) (Fig. 2B). Including
subjects’ handedness scores, cortical thickness, and cortical volume as additional regressors did not significantly change the
model (P's 40.27) and a similar positive correlation was still found
(r ¼0.47, p o0.05). However, this was driven by a single data-point
(in the lower-left corner of Fig. 2B); repeating the correlation
without this data-point resulted in a non-significant correlation
(r ¼0.27, p ¼ 0.21).
Finally, we performed an additional correlational analysis, to
investigate parallels with a prior report that the magnitude of the
N170 potential for faces in the right hemisphere were positively
correlated with those for words in the left hemisphere (Dundas
et al., 2014). Thus we studied the relationship between the response to words in the left VWFA and the responses to faces in the
right FFA. This showed no correlation for the number of voxels
(r¼ 0.01, p ¼0.93), but a positive correlation for the peak MR response (r¼ 0.44, p o0.022). The positive correlation was even
stronger when fusiform cortical thickness (r¼ 0.57, p o0.003) and
cortical volume (r ¼0.54, po 0.006) were taken into account.
At first glance, these last results might be taken as supporting
the hypothesis that stronger lateralization for words is associated
with stronger lateralization for faces. However, greater activation
for faces in one hemisphere cannot be taken as indicating a more
90
J. Davies-Thompson et al. / Brain Research 1644 (2016) 88–97
Fig. 1. Correlations in “same regions-of-interest” analysis. (A) The peak responses to words and faces within each region of interest. Graphs of the left are from the FFA
regions, and graphs on the right from the VWFA regions. Top graphs are from the left hemisphere and middle graphs are from the right hemisphere. Lines indicate linear
regression, with correlation coefficients (r) and significant shown, with bold lines and numbers indicating significant correlations. Units on the y-axis are z-scores (for
faces 4scrambled faces or words 4scrambled words). (B) The lateralization indexes for these peak responses. Positive values indicate greater activation in the left hemisphere. Left graph is for activity identified within the boundaries of the FFA regions, and right graph is for activity within the VWFA regions. Full correlations are shown in
black ink; numbers in grey indicate correlations after controlling for handedness, cortical thickness, and cortical volume. Bold lines and numbers indicate significant
correlations.
J. Davies-Thompson et al. / Brain Research 1644 (2016) 88–97
91
Fig. 2. Correlations in “different regions-of-interest” analysis, comparing the response to words in VWFA regions and faces in FFA regions. (A) The peak responses to words
and faces. Lines indicate linear regression, with correlation coefficients (r) and significant shown, with bold lines and numbers indicating significant correlations. Units are
z-scores (faces 4scrambled faces or words 4scrambled words). (B) The lateralization indices for these responses, in which positive values indicate greater activation in the
left hemisphere. Full correlations are shown in black ink; numbers in grey indicate correlations after controlling for handedness, cortical thickness, and cortical volume. Bold
lines and numbers indicate significant correlations.
asymmetric lateralization of faces. To support that inference one
also needs to know how activation for faces in one hemisphere is
related to activation for faces in the other, and likewise for words.
The inference would be supported if there were either a negative
or no correlation. However, for faces we find instead a positive
correlation between the peak MR responses in the right FFA and
those in the left FFA (r¼ 0.65, po 0.0003) and a trend for words
between the right and left VWFA (r ¼ 0.37, p ¼0.06). Again, these
positive correlations are stronger when we take into account
cortical thickness (faces, right versus left FFA: r ¼0.59, p o0.002;
words, right versus left VWFA: r ¼0.49, p o0.013) or cortical volume (faces, right versus left FFA: r ¼0.54, p o0.002; words, right
versus left VWFA: r¼ 0.42, p o0.036). Thus, the positive correlations we find between words and faces both within and between
the hemispheres in our ‘between ROI’ analysis reflect the fact that
subjects with greater activation for faces also had greater activation for words, but do not point to an inverse relationship between
the lateralization of faces and words.
2.3. Inferior temporal analysis
For the total number of voxels activated in inferotemporal
cortex, there was no correlation between face and word activation
in either the left (r ¼ 0.27, p ¼0.90) or right hemisphere (r ¼0.08,
p ¼0.73) (Fig. 3A). For the average of the peak MR responses from
all activated clusters, there was a trend to a positive correlation
between face and word activity in the left (r ¼ 0.34, p ¼0.08) but
not the right hemisphere (r ¼ 0.00, p ¼0.99).
The lateralization indices (Fig. 3B) also showed no correlation between words and faces for either the total number of
voxels responding (r ¼ 0.13, p ¼0.59) or the averaged peak
responses (r ¼ 0.08, p ¼0.73). Including subject's handedness
scores, cortical volume, and cortical thickness as additional
variables did not improve the model (P's 4 0.38), and led to
similar insignificant correlations for the number of voxel analysis (r ¼ 0.10, p ¼ 0.66) and the peak MR response analysis
(r ¼ 0.03, p ¼ 0.91).
92
J. Davies-Thompson et al. / Brain Research 1644 (2016) 88–97
Fig. 3. Correlations in inferior temporal analysis (all voxels). (A) number of voxels activated (left graphs) and average of the peak responses of all clusters (right graphs). Top
graphs are for the left and middle graphs for the right hemisphere. Lines indicate linear regression, with correlation coefficients (r) and significance shown, with bold lines
and numbers indicating significant correlations. Units are z-scores (faces 4scrambled faces or words 4scrambled words). (B) The lateralization indexes for number of voxels
(left graph) and peak responses (right graph), in which positive values indicate greater activation in the left hemisphere. Full correlations are shown in black ink; numbers in
grey indicate correlations after controlling for handedness, cortical thickness, and cortical volume. Bold lines and numbers indicate significant correlations (none were
found).
J. Davies-Thompson et al. / Brain Research 1644 (2016) 88–97
3. Discussion
We hypothesised that, if there is competition between face and
word processing for neural resources in inferior temporal cortex
(Dundas et al., 2013; Plaut and Behrmann, 2011), we would find
that in any given region of a hemisphere, greater activation for one
of these two stimuli would be associated with less activation for
the other, resulting in a negative correlation across subjects. Instead, we found positive correlations for peak MR responses
within left-sided regions of interest, between the left FFA and left
VWFA, and between the maximally active voxels in left and right
inferotemporal cortex, but no other correlations in the right
hemisphere or for the number of voxels activated. Second, we also
hypothesised that the magnitude of hemispheric asymmetry as
represented in a lateralization index would show a complementary relationship between visual word and face activation,
and hence a negative correlation between face and word indices.
Instead, we again found a positive correlation for the lateralization
indices of peak MR responses both within the FFA regions and
between the FFA and VWFA regions. Otherwise, there was little
evidence in other dependent variables or analyses for any correlation, and this did not change whether or not outliers were included or whether the analysis was adjusted for handedness,
cortical thickness and volume.
The positive correlations found within a region or hemisphere,
particularly on the left side, may suggest that the ability or need to
recruit resources to process words is correlated in some fashion
with the ability or need to do the same for face processing.
However, this finding needs to be interpreted with caution. Nonspecific factors such as attention can modulate the degree of activation in visual processing areas (Bressler et al., 2013; Schwartz
et al., 2005) and a positive correlation of word and face activation
across subjects may simply be due to variations in attention across
subjects. The positive correlation found for the lateralization indices may also be a more specific effect, since it is not clear that a
general modulation of signals by attention or other factors would
alter the degree of hemispheric asymmetry. Nevertheless, one
cannot definitively exclude a priori the possibility that general
factors could enhance the activation in a dominant hemisphere
differently than that in the non-dominant one.
We analyzed both the peak MR responses and the number of
voxels activated in a region. At this point it is not clear which is a
better index of the amount of neural resources recruited by a task.
We did not find any correlations when using the cluster size (i.e.
number of voxels) as the response measure. Unlike the case of
peak MR responses, which have the advantage of being unaffected
by the statistical threshold used to define activation, cluster sizes
are heavily dependent upon the choice of threshold, and one can
question whether our use of a variable threshold to set the side of
least activation at 10 voxels may have introduced distortions into
the lateralization index we calculated. However, this method was
necessary to identify a minimum degree of activation in the minor
hemispheres (left for faces and right for words) to enter into the
lateralization index. Failure to do so would have resulted in a lateralization index of 1.00 for a number of subjects, a ceiling effect
that would not accurately capture the magnitude of asymmetry of
their activation. In any case, no significant correlations were observed in the analysis of number of voxels activated in the inferior
temporal cortex, which used a fixed threshold.
The absence of a negative correlation in our study does not
mean that competition between resources for face and word
processing in the occipito-temporal cortex does not exist. It is
important to consider a number of technical points. For one, all our
subjects were right-handed. Previous studies show that handedness can influence the degree of lateralization of word (Cai et al.,
2007) and face processing (Bukowski et al., 2013; Willems et al.,
93
2010). For example, a left-sided lateralization of face areas was
found in 17% of right-handed but 28% of left-handed subjects
(Bukowski et al., 2013; Rossion et al., 2012). It may be that
broadening the handedness spectrum of subjects would generate a
wider range of lateralization indices, and different correlations
may then emerge.
Second, our study is a cross-sectional survey of literate adults.
This is a similar strategy to that used in one prior study that, in
distinction to our results, found a negative correlation between the
response to visual words and that to faces in the left fusiform
gyrus for literate but not illiterate adults (Dehaene et al., 2010).
However, this latter study did not examine the effects in the right
hemisphere or a lateralization index, so one cannot make definitive inferences from this result regarding the emergence of
hemispheric asymmetries in word and face processing. A longitudinal study of the lateralization of word and face activity in
children or adults during the acquisition of literacy would be the
best way of demonstrating the effects of competition during development. Short of that, recent cross-sectional studies of different
age cohorts have found that children possess only a right hemifield
superiority for words but not the left hemifield superiority for
faces that is seen in adults (Dundas et al., 2013). Similarly an
event-related potential study also reported that children showed a
larger N170 potential for words on the left hemisphere but no
hemispheric asymmetry in the N170 potential for faces, with the
latter being found only in adults (Dundas et al., 2014).
Finally, a review of face imaging studies (Dien, 2009) noted that
the use of different subtractive baseline conditions in localizer
protocols may produce different patterns of face lateralization. The
use of non-face visual objects resulted in greater right-lateralization (Gauthier et al., 1999; Haxby et al., 1999; Rhodes et al., 2004),
while less lateralization was observed in studies that used just
fixation or scrambled faces (Gilaie-Dotan and Malach, 2007; Ishai
et al., 2002; Puce et al., 1995). A wide range of subtractive conditions have also been used for visual word studies – for review, see
(Barton et al., 2010a, 2010b): how these affect the apparent lateralization of word processing has not yet been evaluated in a
similar manner. However, we found that the results were very
similar for the response in the FFA to words and faces when using
a conjunction contrast (i.e. (faces 4objects)∩(faces4 scrambled
faces)) as when comparing the response to scrambled stimuli only.
Regardless, the report of (Dien, 2009) indicates that the simple
technical issue of the baseline subtractive condition may have
consequences for the search for a complementary relationship
between the lateralization of word and face processing using fMRI.
Replication of our study with different baseline conditions would
be an important next step.
One previous study (Dundas et al., 2014) reported a correlation
between the magnitude of the N170 for words in the left hemisphere and that for faces in the right hemisphere. In an analogous
fashion, we too found a correlation between the peak MR responses for words in the left VWFA and that for faces in the right
FFA. However, as we discussed in the results above, such results do
not prove the existence of an inverse relationship in the lateralization of activity for words and faces. Rather, we find positive
correlations 1) between peak MR responses for words and faces
within each hemisphere, 2) between faces in the left and faces in
the right hemisphere, and 3) between words in the left and words
in the right hemisphere. All this simply means that a) for a given
stimulus, greater activation in one hemisphere is associated with
greater activation in the other, and b) that greater activation for
faces is associated with greater activation for words. This underlines the importance of a lateralization index to capture the degree
of hemispheric asymmetry of activation.
As discussed above, there may be non-specific reasons why
activation for words and faces may show a positive correlation
94
J. Davies-Thompson et al. / Brain Research 1644 (2016) 88–97
across subjects. However, it is more difficult to attribute a positive
correlation between the lateralization indices for words and faces
to general factors such as attention. One possible reason for a
positive correlation is that there are external factors that influence
the lateralization of both words and faces. Indeed, any inverse
relationship between word and face lateralization may be obscured if the external influences are the dominant factor driving
lateralization. We considered the possibility of handedness (Bukowski et al., 2013; Cai et al., 2007; Willems et al., 2010) and
hemispheric asymmetries in structural neural resources as indexed by the cortical thickness and volume of the fusiform gyri;
however, accounting for these still did not yield an inverse relationship between word and face lateralization. Nevertheless,
other factors remain to be explored. Language dominance may
have an important influence on the lateralization of other processes, as well as competition for visual processing resources by
other objects besides faces and words, particularly those for which
the subject has acquired expertise (Behrmann et al., 2005).
Along these lines are the findings of another study of the lateralization of the response to words and faces in the VWFA and
FFA (Pinel et al., 2014). Like us they also failed to find an inverse
correlation between words and faces; however, they did not observe positive correlations either. (One possible reason is the difference between their study and ours in localising regions of interest: while we identified regions on an individual subject level,
which can take into account individual anatomic variations, they
localized regions from activity observed at the group level.) Perhaps more interestingly, they observed an inverse correlation between the response to faces in the FFA and the response to spoken
language in the STS. This is consistent with our inference from the
positive correlations between visual word and face activity that
other factors such as language may play a role in the hemispheric
asymmetries that develop for both words and faces.
In summary, we did not find an inverse relationship between
the lateralization of word and face activation using functional
neuroimaging in this cross-sectional study of healthy subjects.
Rather, there were positive correlations between lateralization
indices for VWFA and FFA regions: the more right-hemisphere
biased a subject's face activation was, the more right-hemisphere
biased their word activation was also. While these results do not
support a competition between face and visual word processing as
a contributing factor to the complementary lateralization of these
processes, they do not disprove it – more dominant influences that
also drive hemispheric asymmetry may mask an inverse relationship between words and faces. Further work to identify
these factors and to study the evolution of word and face lateralization during development would be important next steps.
4. Methods
4.1. Subjects
Twenty-six healthy participants with no history of neurological
dysfunction, vascular disease or cognitive complaints took part in
the study (16 females, mean age¼ 26, range 21–36). All participants were right-handed with corrected visual acuity of 20/20. A
handedness score was obtained for each subject using the Edinburgh handedness questionnaire (Oldfield, 1971). The protocol was
approved by the institutional review boards of the University of
British Columbia and Vancouver General Hospital, and written
informed consent was obtained for all subjects in accordance with
The Code of Ethics of the World Medical Association, Declaration
of Helsinki.
4.2. Imaging parameters
Subjects were scanned in a Philips 3.0 T scanner at the UBC MRI
Research Centre. T2*-weighted scans using echo planar imaging
were used to collect data from 36 interleaved axial slices (TR
2000 ms, TE 30 ms, FOV ¼240 216 mm, 3 mm thickness with
1 mm gap, voxel size 3 3 mm, 128 mm reconstruction matrix,
reconstructed voxel size 1.88 1.6 mm). These were co-registered
onto a T1-weighted anatomical image (EPI) sequence, 170 axial
slices, FOV ¼ 256 200 mm, voxel size ¼1 1 mm, slice thickness
1 mm, from each participant.
4.3. Localizer scans
Localizer scans were used to identify regions of visual cortex
responding to English words and faces (Fig. 4). The HVEM dynamic
face multiple localizer scan (Fox et al., 2009) was run twice on
each subject to identify face-selective regions of the visual cortex.
The localizer consisted of grayscale video clips of faces, bodies,
objects, Fourier-phase scrambled faces, and Fourier-phase scrambled objects. Each stimulus block included 6 video clips lasting
Fig. 4. Stimuli and protocol used in the face and word localisers.
J. Davies-Thompson et al. / Brain Research 1644 (2016) 88–97
95
Fig. 5. The location of the FFA and VWFA in an individual subject.
the complementary contrast for words ((English words 4 Korean
words)∩(English words 4 scrambled words)) allowed us to identify the left VWFA in 19 subjects but the right VWFA in only
6 subjects. (This is in line with a previous study in which we found
that the VWFA responded equally to several forms of visual symbols, including English words and Korean words (Muayqil et al.,
2015)). As lateralization indices require subjects to have bilateral
responses, using this conjunction contrast only allowed us to examine the response from 6 subjects, and we were therefore unable
to conduct analysis for either 1) the response to faces and words
within the VWFA, nor 2) the more crucial comparison between the
FFA and VWFA. We were however, able to examine the response to
faces and words within the FFA, where we found no significant
positive or negative correlations (Supplementary Fig. 1). In order
to circumvent this issue, word-selective regions were determined
by the contrast English words4 scrambled English words
(Po 0.001, uncorrected), which enabled us to identify the right
and left VWFA in the majority of subjects. For symmetry, we
therefore used the corresponding contrast (faces 4scrambled faces) to identify the face-selective regions.
1.5 s separated by a 500 ms blank screen, resulting in 12 s stimulus
blocks. Stimulus blocks were separated by a 12 s fixation cross.
Each condition was repeated 5 times per run.
A scan to localise word-selective regions of the visual cortex
was also run twice on each subject (Barton et al., 2010a, 2010b).
The localizer consisted of grayscale images of 3- or 4-letter English
words, Korean words, and Fourier-phase scrambled English words.
Each stimulus block included 15 images lasting 500 ms separated
by a 300 ms blank screen, resulting in 12 s stimulus blocks. Stimulus blocks were separated by a 12 s fixation cross. Each condition was repeated 5 times per run.
In all functional scans, attention was monitored by asking
participants to press a button on an MRI-compatible button-box
when the same video clip or image was presented twice in a row.
4.4. fMRI analysis
Statistical analysis of the fMRI data was carried out using FEAT
(http://www.fmrib.ox.ac.uk/fsl; (Smith et al., 2004)). The initial 8 s
of data from each scan were removed to minimize the effects of
magnetic saturation. Motion correction was followed by spatial
smoothing (Gaussian, FWHM 6 mm) and temporal high-pass filtering (cut-off of 0.01 Hz).
Ideally, a conjunction contrast (i.e. (faces 4objects)∩(faces 4
scrambled faces)) would be used to isolate face- and word- selective regions of interest. However, while a conjunction contrast
for faces enabled us to identify the bilateral FFA in all 26 subjects,
4.4.1. Functionally defined region-of-interest (ROI) analysis
This focused on identifying the FFA and VWFA as defined previously, which are regions in the mid-fusiform region of the right
and left hemispheres (Fig. 5). Contiguous voxels activated by the
word contrast above within the Tailarach coordinates of (x ¼ 50
to 30, y ¼ 80 to 30, z o0) were considered the left VWFA
Table 1
Peak MNI coordinates of the region of interest analysis and most significant voxels in the inferior temporal cortex. Also shown are the average peak MR response for all
clusters, along with the average total area showing significant activity (P o 0.001), in the inferior temporal cortex analysis. For comparison across studies, cluster size (as
determined on a 2 2 2 mm standard template) is shown instead of number of voxels. Numbers in brackets show standard error of the mean.
Region
n
Peak coordinates
X
Y
Z
Peak
Cluster size
All peaks
All voxels size
(z-score)
(cm³)
(z-score)
(cm³)
–
–
–
–
–
–
–
–
Region of interest analysis
FFA
L
26
R
26
VWFA
L
25
R
25
41
42
43
44
(0.9)
(0.7)
(1.2)
(1.5)
51
52
50
49
(1.5)
(1.4)
(1.8)
(2.0)
23
22
20
21
(1.0)
(1.0)
(1.2)
(1.2)
8.0
8.5
5.8
4.0
(0.4)
(0.4)
(0.4)
(0.2)
0.8
1.4
6.0
0.3
(0.2)
(0.3)
(1.8)
(0.0)
Inferior temporal cortex
Faces
L
26
R
26
Words
L
26
R
26
41
41
42
38
(0.7)
(0.9)
(1.0)
(2.1)
51
52
51
46
(2.1)
(2.6)
(3.0)
(3.2)
21
22
21
22
(1.2)
(1.3)
(1.4)
(1.8)
8.2
8.7
6.0
4.5
(0.5)
(0.4)
(0.3)
(0.2)
10.8
16.6
7.3
1.2
(1.7)
(2.5)
(1.8)
(0.3)
5.0
5.0
4.3
3.6
(0.2)
(0.1)
(0.1)
(0.1)
14.2
21.9
9.8
2.0
(1.9)
(2.7)
(1.9)
(0.5)
96
J. Davies-Thompson et al. / Brain Research 1644 (2016) 88–97
(Cohen et al., 2002a, 2002b), while a mirror-symmetric location in
the other hemisphere was its right counterpart. Likewise, the right
FFA has been reported to have mean coordinates of about (x ¼40,
y¼ 55, z¼ 10) (Kanwisher et al., 1997a, 1997b). When anterior
and posterior regions were found, the posterior region was assigned as the FFA.
In each subject, the size of the smaller FFA was fixed at 10
voxels (see (Bukowski et al., 2013; Fox et al., 2009; Rossion et al.,
2012), for a similar approach) by varying the statistical threshold
until 10 voxels were significant on the side with less activation.
The same approach was used to identify right and left activation
by words in the VWFA. As no data from any given subject can be
entered into the analysis unless all 4 regions are identified, this
approach allowed us to obtain data from the majority of subjects
while allowing an unbiased comparison between the two different
localisers (Table 1). We extracted the following information from
each of the four identified regions of interest (right and left FFA,
right and left VWFA): 1) the peak MR response to each regions
preferred stimulus (faces in the FFA, words in VWFA); 2) the peak
MR response to each regions non-preferred stimulus (faces in the
peak voxel of the VWFA, words in the peak voxel of the FFA); and
3) the number of voxels above the contrast threshold within each
region.
An alternative to comparing single peak voxels, is to compare
the average z-scores across all voxels within each region-of-interest. This yielded similar results to the peak-voxel analysis, and
are therefore not reported here for reasons of brevity.
4.4.2. Structurally defined inferior temporal cortex analysis
To examine the relationship between word and face activation
in a broader anatomic context, we assessed these in the inferior
temporal cortex, using a combined mask of the temporooccipital
and posterior divisions of the inferior temporal gyrus (HarvardOxford Cortical Structural Atlas; http://www.fmrib.ox.ac.uk/fsl)
that was defined at the group level, and transformed back into the
individual subject's native space. In contrast to the strategy used in
the functionally defined ROI analysis, all voxels above a set statistical threshold (P o0.001, uncorrected) were entered into the
analysis.
4.5. Statistical analysis
We first examined whether word and face activity was correlated within each hemisphere. Next we calculated a lateralization
index, which was the weighted difference (left-right)/(left þright)
of hemispheric activation, first for words and then for faces. As
such indices can occasionally result in large spurious numbers,
values greater than 3 standard deviations from the mean were
defined as outliers and removed before statistical analysis. All our
analyses were re-done in a secondary analysis that included outliers. These did not change the pattern or significance of the results reported. We chose this lateralization index because it expresses the degree of lateralization essentially as a percentage of
the sum amount of activation in both hemispheres; repeating the
analyses with a simple difference (left-right) or a ratio (left/right)
as alternate lateralization indices did not alter the findings. Our
weighted-difference lateralization index results in a positive index
corresponding to greater activity on the left and a negative one to
greater activity on the right hemisphere. A correlation analysis
(pearson's r, two-tailed) was performed for the lateralization index
of words versus faces.
We performed the above at three levels of analysis, which
moved from smaller to larger frames of reference. First, competition may be evident within a functionally defined region. For example, one can ask if, in the right and left FFA, the asymmetry of
face activity is inversely correlated with the asymmetry of word
activity within these face-responsive regions, which were defined
solely by their activity during a face (and not a word) localizer
contrast. Similarly, one can then examine if in the right and left
VWFA, the asymmetry of word activation is inversely correlated
with the asymmetry of face activity within these word-responsive
regions. These we called ‘same-ROI’ analyses. These focused solely
on MR responses of the peak voxels, as the number of voxels for
the non-defining stimulus would be artificially constrained by the
use of the other stimulus type to define the region of interest.
Second, one can ask if the lateralization of face activity within
functionally defined FFA regions is inversely correlated with the
lateralization of word activity within functionally defined VWFA
regions. These we called ‘different-ROI’ analyses. The dependent
variables for these analyses were both peak MR responses and
number of activated voxels.
Third, we examined the lateralization of face and word activity
within the broader context of the inferior temporal cortex, whose
borders we defined structurally rather than by functional localizers. The dependent variables here were the number of all
voxels above the statistical threshold, and the average peak MR
response from all clusters of activation.
Other factors may influence the lateralization of both words
and faces and mask a complementary relationship between the
two. First, to control for a possible influence of handedness, we
repeated the analysis but included subjects' scores on the Edinburgh handedness questionnaire (Oldfield, 1971) as an additional
regressor. Second, we reasoned that asymmetry in the available
neural resources may also influence the appearance of lateralization of both words and faces. Thus we extracted the grey matter
volume and the average cortical thickness of the left and right
fusiform gyri. Cortical reconstruction and volumetric segmentation was performed with the Freesurfer image analysis suite,
(http://surfer.nmr.mgh.harvard.edu/). Lateralization indices were
calculated for the grey matter volume and the average cortical
thickness in the fusiform gyri and entered as additional regressors
in the above analyses. All statistical analysis was carried out using
R (http://www.R-project.org).
Acknowledgements
This work was supported by CIHR grant MOP-106511. JB was
supported by the Marianne Koerner Chair in Brain Diseases and a
Canada Research Chair.
Appendix A. Supplementary material
Supplementary data associated with this article can be found in
the online version at http://dx.doi.org/10.1016/j.brainres.2016.05.
009.
References
Barton, J.J., Sekunova, A., Sheldon, C., Johnston, S., Iaria, G., Scheel, M., 2010a.
Reading words, seeing style: the neuropsychology of word, font and handwriting perception. Neuropsychologia 48 (13), 3868–3877. http://dx.doi.org/
10.1016/j.neuropsychologia.2010.09.012.
Barton, J.J.S., Fox, C.J., Sekunova, A., Iaria, G., 2010b. Encoding in the visual word
form area: an fMRI adaptation study of words versus handwriting. J. Cognit.
Neurosci. 22 (8), 1649–1661. http://dx.doi.org/10.1162/jocn.2009.21286.
Behrmann, M., Marotta, J., Gauthier, I., Tarr, M., McKeeff, T., 2005. Behavioral change
and its neural correlates in visual agnosia after expertise training. J. Cogn.
Neurosci. 17, 554–568.
Behrmann, M., Plaut, D.C., 2013. Distributed circuits, not circumscribed centers,
mediate visual recognition. Trends Cogn. Sci. 17 (5), 210–219. http://dx.doi.org/
10.1016/j.tics.2013.03.007, S1364-6613(13)00072-7 [pii].
J. Davies-Thompson et al. / Brain Research 1644 (2016) 88–97
Bressler, D.W., Fortenbaugh, F.C., Robertson, L.C., Silver, M.A., 2013. Visual spatial
attention enhances the amplitude of positive and negative fMRI responses to
visual stimulation in an eccentricity-dependent manner. Vision. Res. 85,
104–112. http://dx.doi.org/10.1016/j.visres.2013.03.009.
Bukowski, H., Dricot, L., Hanseeuw, B., Rossion, B., 2013. Cerebral lateralization of
the face-selective areas in left-handers: only the ‘FFA’ does not get it right.
Cortex.
Cai, Q., Lavidor, M., Brysbaert, M., Paulignan, Y., Nazir, T.A., 2007. Cerebral lateralization of frontal lobe language processes and lateralization of the posterior
visual word processing system. J. Cognit. Neurosci. 20 (4), 672–681. http://dx.
doi.org/10.1162/jocn.2008.20043.
Cohen, L., Dehaene, S., Naccache, L., Lehéricy, S., Dehaene-Lambertz, G., Hénaff, M.A., Michel, F., 2000. The visual word form area. Spatial and temporal characterization of an initial stage of reading in normal subjects and posterior splitbrain patients. Brain 123 (2), 291–307. http://dx.doi.org/10.1093/brain/
123.2.291.
Cohen, L., Lehericy, S., Chochon, F., Lemer, C., Rivaud, S., Dehaene, S., 2002a. Language-specific tuning of visual cortex? Functional properties of the Visual Word
Form Area. Brain:J. Neurol. 125 (Pt 5), 1054–1069.
Cohen, L., Lehéricy, S., Chochon, F., Lemer, C., Rivaud, S., Dehaene, S., 2002b. Language-specific tuning of visual cortex? Functional properties of the Visual Word
Form Area. Brain 125 (5), 1054–1069. http://dx.doi.org/10.1093/brain/awf094.
Davies-Thompson, J., Andrews, T.J., 2012. Intra- and inter-hemispheric connectivity
between face-selective regions in the human brain. J. Neurophysiol. . http://dx.
doi.org/10.1152/jn.01171.2011
de Heering, A., Rossion, B., 2015. Rapid categorization of natural face images in the
infant right hemisphere. Elife 4, e06564. http://dx.doi.org/10.7554/eLife.06564.
de Renzi, E., 1986. Prosopagnosia in two patients with CT scan evidence of damage
confined to the right hemisphere. Neuropsychologia 24 (3), 385–389. http://dx.
doi.org/10.1016/0028-3932(86)90023-0.
Dehaene, S., Cohen, L., 2011. The unique role of the visual word form area in
reading. Trends Cognit. Sci. 15 (6), 254–262. http://dx.doi.org/10.1016/j.
tics.2011.04.003.
Dehaene, S., Pegado, F., Braga, L.W., Ventura, P., Filho, G.N., Jobert, A., Cohen, L.,
2010. How learning to read changes the cortical networks for vision and language. Science 330 (6009), 1359–1364. http://dx.doi.org/10.1126/
science.1194140.
Dien, J., 2009. A tale of two recognition systems: Implications of the fusiform face
area and the visual word form area for lateralized object recognition models.
Neuropsychologia 47 (1), 1–16. http://dx.doi.org/10.1016/j.
neuropsychologia.2008.08.024.
Dundas, E.M., Plaut, D.C., Behrmann, M., 2013. The joint development of hemispheric lateralization for words and faces. J. Exp. Psychol. Gen. 142 (2),
348–358. http://dx.doi.org/10.1037/a0029503 2012-21157-001 [pii].
Dundas, E.M., Plaut, D.C., Behrmann, M., 2014. An ERP investigation of the co-development of hemispheric lateralization of face and word recognition. Neuropsychologia 61, 315–323. http://dx.doi.org/10.1016/j.neuropsychologia.2014.05.006, S0028-3932(14)00154-7 [pii].
Fox, C.J., Iaria, G., Barton, J.J.S., 2009. Defining the face processing network: optimization of the functional localizer in fMRI. Human. Brain Mapp. 30 (5),
1637–1651.
Gauthier, I., Tarr, M.J., Anderson, A.W., Skudlarski, P., Gore, J.C., 1999. Activation of
the middle fusiform ‘face area’ increases with expertise in recognizing novel
objects. Nat. Neurosci. 2 (6), 568–573. http://dx.doi.org/10.1038/9224.
Gilaie-Dotan, S., Malach, R., 2007. Sub-exemplar shape tuning in human face-related areas. Cereb. Cortex 17 (2), 325–338, bhj150 [pii] 10.1093/cercor/bhj150.
Haxby, Hoffman, Gobbini, 2000a. The distributed human neural system for face
perception. Trends Cognit. Sci. 4 (6), 223–233.
Haxby, J.V., Hoffman, E.A., Gobbini, M.I., 2000b. The distributed human neural
system for face perception. Trends Cogn. Sci. 4 (6), 223–233.
Haxby, J.V., Ungerleider, L.G., Clark, V.P., Schouten, J.L., Hoffman, E.A., Martin, A.,
1999. The effect of face inversion on activity in human neural systems for face
and object perception. Neuron 22 (1), 189–199. http://dx.doi.org/10.1016/
S0896-6273(00)80690-X.
Ishai, A., Haxby, J.V., Ungerleider, L.G., 2002. Visual imagery of famous faces: effects
of memory and attention revealed by fMRI. Neuroimage 17 (4), 1729–1741. http:
//dx.doi.org/10.1006/nimg.2002.1330.
Kanwisher, N., McDermott, J., Chun, M.M., 1997a. The fusiform face area: a module
in human extrastriate cortex specialized for face perception. J. Neurosci. 17 (11),
4302–4311.
Kanwisher, N., Woods, R.P., Iacoboni, M., Mazziotta, J.C., 1997b. A locus in human
extrastriate cortex for visual shape analysis. J. Cognit. Neurosci. 9 (1), 133–142.
http://dx.doi.org/10.1162/jocn.1997.9.1.133.
Kawahata, N., Nagata, K., Shishido, F., 1988. Alexia with agraphia due to the left
posterior inferior temporal lobe lesion – neuropsychological analysis and its
pathogenetic mechanisms. Brain Lang. 33 (2), 296–310.
97
Le Grand, R., Mondloch, C.J., Maurer, D., Brent, H.P., 2003. Expert face processing
requires visual input to the right hemisphere during infancy. Nat. Neurosci. 6
(10), 1108–1112 10.1038/nn1121 nn1121 [pii].
Leehey, S.C., Cahn, A., 1979. Lateral asymmetries in the recognition of words, familiar faces and unfamiliar faces. Neuropsychologia 17 (6), 619–628. http://dx.
doi.org/10.1016/0028-3932(79)90036-8.
Levine, S.C., Banich, M.T., 1982. Lateral asymmetries in the naming of words and
corresponding line drawings. Brain Lang. 17 (1), 34–45. http://dx.doi.org/
10.1016/0093-934x(82)90003-7.
Levine, S.C., Koch-Weser, M.P., 1982. Right hemisphere superiority in the recognition of famous faces. Brain Cogn. 1 (1), 10–22. http://dx.doi.org/10.1016/
0278-2626(82)90003-3.
Maurer, U., Rossion, B., McCandliss, B.D., 2008. Category specificity in early perception: face and word n170 responses differ in both lateralization and habituation properties. Front. Hum. Neurosci. 2, 18. http://dx.doi.org/10.3389/
neuro.09.018.2008.
Mercure, E., Cohen Kadosh, K., Johnson, M.H., 2011. The n170 shows differential
repetition effects for faces, objects, and orthographic stimuli. Front. Hum.
Neurosci. 5, 6. http://dx.doi.org/10.3389/fnhum.2011.00006.
Muayqil, T., Davies-Thompson, J., Barton, J.J.S., 2015. Representation of visual
symbols in the visual word processing network. Neuropsychologia 69, 232–241.
Nestor, A., Behrmann, M., Plaut, D.C., 2013. The neural basis of visual word form
processing: a multivariate investigation. Cereb. Cortex 23 (7), 1673–1684
10.1093/cercor/bhs158 bhs158 [pii].
Oldfield, R.C., 1971. The assessment and analysis of handedness: the Edinburgh
inventory. Neuropsychologia 9 (1), 97–113.
Pinel, P., Lalanne, C., Bourgeron, T., Fauchereau, F., Poupon, C., Artiges, E., Dehaene,
S., 2014. Genetic and environmental influences on the visual word form and
fusiform face areas. Cereb. Cortex, bhu048.
Pinel, P., Lalanne, C., Bourgeron, T., Fauchereau, F., Poupon, C., Artiges, E., Dehaene,
S., 2015. Genetic and Environmental Influences on the Visual Word Form and
Fusiform Face Areas. Cereb. Cortex 25 (9), 2478–2493 10.1093/cercor/bhu048
cercor/bhu048 [pii].
Plaut, D.C., Behrmann, M., 2011. Complementary neural representations for faces
and words: a computational exploration. Cogn. Neuropsychol. 28 (3–4),
251–275. http://dx.doi.org/10.1080/02643294.2011.609812.
Puce, A., Allison, T., Gore, J.C., McCarthy, G., 1995. Face-sensitive regions in human
extrastriate cortex studied by functional MRI. J. Neurophysiol. 74 (3),
1192–1199.
Rhodes, G., Jeffery, L., Watson, T.L., Jaquet, E., Winkler, C., Clifford, C.W., 2004. Orientation-contingent face aftereffects and implications for face-coding mechanisms. Curr. Biol. 14 (23), 2119–2123.
Rossion, B., Hanseeuw, B., Dricot, L., 2012. Defining face perception areas in the
human brain: a large-scale factorial fMRI face localizer analysis. Brain Cogn. 79
(2), 138–157. http://dx.doi.org/10.1016/j.bandc.2012.01.001.
Sakurai, Y., Sakai, K., Sakuta, M., Iwata, M., 1994. Naming difficulties in alexia with
agraphia for kanji after a left posterior inferior temporal lesion. J. Neurol.
Neurosurg. Psychiatry 57 (5), 609–613.
Schwartz, S., Vuilleumier, P., Hutton, C., Maravita, A., Dolan, R.J., Driver, J., 2005.
Attentional load and sensory competition in human vision: modulation of fMRI
responses by load at fixation during task-irrelevant stimulation in the peripheral visual field. Cereb. Cortex 15 (6), 770–786. http://dx.doi.org/10.1093/cercor/
bhh178.
Scott, L.S., 2006. Featural and configural face processing in adults and infants: a
behavioral and electrophysiological investigation. Perception 35 (8), 1107–1128.
Sergent, J., Villemure, J.-G., 1989. Prosopagnosia in a right hemispherectomized
patient. Brain 112 (4), 975–995. http://dx.doi.org/10.1093/brain/112.4.975.
Smith, S.M., Jenkinson, M., Woolrich, M.W., Beckmann, C.F., Behrens, T.E.J., Johansen-Berg, H., Matthews, P.M., 2004. Advances in functional and structural
MR image analysis and implementation as FSL. Neuroimage 23 (Suppl.1),
S208–S219.
Szwed, M., Dehaene, S., Kleinschmidt, A., Eger, E., Valabrègue, R., Amadon, A., Cohen, L., 2011. Specialization for written words over objects in the visual cortex.
Neuroimage 56 (1), 330–344. http://dx.doi.org/10.1016/j.
neuroimage.2011.01.073.
Willems, R.M., Peelen, M.V., Hagoort, P., 2010. Cerebral lateralization of face-selective and body-selective visual areas depends on handedness. Cereb. Cortex
20 (7), 1719–1725. http://dx.doi.org/10.1093/cercor/bhp234.
Yovel, G., Tambini, A., Brandman, T., 2008. The asymmetry of the fusiform face area
is a stable individual characteristic that underlies the left-visual-field superiority for faces. Neuropsychologia 46 (13), 3061–3068. http://dx.doi.org/
10.1016/j.neuropsychologia.2008.06.017.
Zangenehpour, S., Chaudhuri, A., 2005. Patchy organization and asymmetric distribution of the neural correlates of face processing in monkey inferotemporal
cortex. Curr. Biol. 15 (11), 993–1005. http://dx.doi.org/10.1016/j.
cub.2005.04.031.