Foveal Word Reading Requires
Interhemispheric Communication
Zoe¨ R. Hunter
1
, Marc Brysbaert
1
, and Stefan Knecht
2
Abstract
& The left cerebral hemisphere is dominant for language pro-
cessing in most individuals. It has been suggested that this
asymmetric language representation can influence behavioral
performance in foveal word-naming tasks. We carried out two
experiments in which we obtained laterality indices by means
of functional imaging during a mental word-generation task,
using functional transcranial Doppler sonography and func-
tional magnetic resonance imaging, respectively. Subsequently,
we administered a behavioral word-naming task, where par-
ticipants had to name foveally presented words of different
lengths shown in different fixation locations shifted horizon-
tally across the screen. The optimal viewing position for left
language dominant individuals is located between the begin-
ning and the center of a word. It is shifted toward the end of a
word for right language dominant individuals and, to a lesser
extent, for individuals with bilateral language representation.
These results demonstrate that interhemispheric communi-
cation is required for foveal word recognition. Consequently,
asymmetric representations of language and processes of inter-
hemispheric transfer should be taken into account in theoreti-
cal models of visual word recognition to ensure neurological
plausibility. &
INTRODUCTION
One of the striking features of the visual system is the
crossing of the nasal optic fibers in the optic chiasm.
Because of this crossing, stimuli presented in the left
visual field (LVF) are initially projected to the right half
of the brain, and stimuli presented in the right visual
field (RVF) are projected to the left hemisphere. This
characteristic of the visual field has been used in thou-
sands of experiments to investigate brain asymmetry
on the basis of the visual half field (VHF) technique.
The split of the visual field in two halves is also the rea-
son why memories of faces are predominantly based on
information from the LVF (Brady, Campbell, & Flaherty,
2004).
Surprisingly, limited attention has been paid to the
question of what happens at the border where LVF and
RVF meet. For a long time the general assumption was
that the hemifields overlap in the center of the visual
field, so that foveal vision is projected bilaterally and
stimuli have to be presented in parafoveal vision to
ensure unilateral projection (Bradshaw & Nettleton,
1983). This assumption was also shared by psycholin-
guists whose models of visual word recognition did not
include any reference to brain asymmetry or the need of
interhemispheric communication.
Several reviews of the literature have shown, however,
that the assumption of a bilaterally represented fovea
is wrong (e.g., Lavidor & Walsh, 2004; Brysbaert, 1994,
2004). For instance, Corballis and Trudel (1993) exam-
ined whether split-brain patients were able to recog-
nize centrally presented four-letter words that could
not be guessed on the basis of the first or the last two
letters. Two patients were examined (L.B. and D.K.).
They were both unable to recognize foveally presented
words, although their performance was good when the
stimuli were presented in the LVF or RVF. Similar find-
ings were reported by Fendrich and Gazzaniga (1989)
and Fendrich, Wessinger, and Gazzaniga (1996) for the
patients V.P. and J.W.
A second argument that has been made for the con-
jecture that cerebral asymmetry and interhemispheric
transfer do not constrain visual word recognition in
foveal vision is that in healthy participants, interhemi-
spheric communication is so fast and abundant that it
does not limit word processing to a greater extent than
the equivalent intrahemispheric connections. This view
has been phrased most explicitly by Dehaene, Cohen,
Sigman, and Vinckier (2005), who wrote: ‘‘It has been
proposed that ‘foveal splitting,’ whereby the left and
right halves of a centrally fixated word are initially sent to
distinct hemispheres, has important functional conse-
quences for reading. However, beyond V1, callosal pro-
jections have the precise structure required to guarantee
the continuity of receptive fields across the midline and
allow convergence to common visual representations.
1
Royal Holloway University of London, UK,
2
University of Mu¨nster,
Germany
D 2007 Massachusetts Institute of Technology Journal of Cognitive Neuroscience 19:8, pp. 1373–1387
We believe that these connections minimize the func-
tional impact of the initial foveal split’’ (p. 338).
Brysbaert (1994) argued that the discussion about
whether interhemispheric transfer has functional conse-
quences for foveal word recognition can be settled quite
easily on the basis of empirical data. All that is needed is
to compare a group of participants with right-hemisphere
language dominance to a group of participants with left-
hemisphere language dominance. Although language is
lateralized to the left in most individuals (Szaflarski et al.,
2002; Knecht et al., 2000; Pujol, Deus, Losilla, & Capdevila,
1999), there is a small percentage of people with right-
hemisphere language dominance. Comparing the per-
formance of left and right language dominant individuals
in a foveal word-recognition task would reveal to what
extent higher cognitive processes such as reading rely on
interhemispheric transfer and information integration. If
foveal vision is bilateral or if interhemispheric connec-
tions minimize the functional impact of the initial foveal
split, then there should be no difference in the perform-
ance of both groups, at least not for short words that
subtend a visual angle of less than 28 (under most read-
ing conditions there are three to four letters per degree
of visual angle).
Brysbaert (1994) made use of the optimal viewing
position (OVP) effect (O’Regan & Jacobs, 1992) to in-
vestigate the issue. The OVP effect is obtained by ask-
ing participants to read words after initial fixation on
the first, the second, ..., and the last letter. The usual
finding is that participants are fastest in recognizing a
word when they are allowed to fixate a letter within the
first one third of the word (called the OVP) and that
there is a considerable time cost for fixations toward
the end of the word, in particular for long words (see
Figures 3 and 6). Brysbaert recruited a group of nine
participants with atypical brain laterality (i.e., with signs
of right-hemisphere language dominance or bilateral lan-
guage representation) and observed that the OVP was
shifted more toward the end of the words for these par-
ticipants compared to a control group of participants
with left-hemisphere language dominance.
Unfortunately, when Brysbaert (1994) ran his experi-
ments, there were no other noninvasive means of as-
sessing cerebral dominance than VHF tasks. Hence,
participants were classified as left or right dominant on
the basis of their LVF–RVF asymmetry in a VHF task with
parafoveal word presentation. A major weakness of this
approach was that variables other than cerebral domi-
nance could account for the correlation between VHF
asymmetries and the preferred landing position in the
OVP task as observed by Brysbaert. These include, for ex-
ample, an individual bias in attention allocation across the
VHFs (Kim & Levine, 1991), established reading habits,
and asymmetries in the information distribution within
words (Efron, 1990).
In the years since the early 1990s, major break-
throughs have been realized to assess cerebral domi-
nance in a noninvasive way. Two techniques stand out.
The first involves functional transcranial Doppler sono-
graphy (fTCD), through which the differences in blood
flow velocity toward the left and right cerebral hemi-
spheres can be measured while participants are perform-
ing a language-related task, usually word generation.
Individuals with left-hemisphere language dominance
are expected to require a higher blood flow to their
left hemisphere than to their right hemisphere while
doing the task and this asymmetry can be picked up
with fTCD. In a series of studies, Knecht and colleagues
showed that the technique makes it possible to reliably
assess cerebral dominance in a test session of less than
30 min. Knecht et al. (2001), for instance, applied this
technique to a group of 326 healthy participants and
obtained evidence for left-hemisphere dominance in 264
participants (80%), bilateral representation in 31 partic-
ipants (10%), and right-hemisphere dominance in 31
participants (10%). The technique was further validated
by comparing its laterality index (LI) to the LI based on
the well-documented invasive Wada test. Fifteen patients
with epilepsy underwent both tests as part of a presur-
gery evaluation. The LIs of fTCD and WADA agreed in all
patients (11 left dominant and 4 right dominant; Knecht
et al., 1998).
The second technique that has been used to assess
brain dominance in a noninvasive way is functional mag-
netic resonance imaging (fMRI). Knecht et al. (2003)
showed that participants with left and right language
dominance (as assessed by fTCD) showed much higher
activation levels in the expected hemisphere in the areas
related to speech production (Broca’s area and the sur-
rounding regions, including Brodmann’s area (BA) 44 and
BA 45). Similar findings were reported by Pujol et al.
(1999) and Szaflarski et al. (2002).
In the experiments below we will repeat Brysbaert’s
(1994) OVP study with groups of participants whose
brain asymmetry has been assessed with either f TCD or
fMRI. If brain laterality has no functional consequences
for foveal word recognition, we expect to find similar
OVP curves for left-dominant and right-dominant partic-
ipants, at least for short words that subtend a visual
angle of less than 28 of visual angle. In contrast, if
interhemispheric communication constrains foveal word
recognition, we expect to find that participants who are
left dominant for language will perform better than
right-dominant participants after fixating the first letters
of the words, whereas they will perform worse after
fixating the last letters of the words. This is because
fixation of the first letters of a word make the word fall
mainly in the RVF, whereas if the last letters are fixated
the word falls predominantly in the LVF.
EXPERIMENT 1
In experiment 1 we tested the OVP effect for German
words of three, five, and seven letters in participants
1374 Journal of Cognitive Neuroscience Volume 19, Number 8
whose brain laterality had previously been assessed as
typical or atypical by means of f TCD.
Methods
Procedure
Participants were chosen from a cohort of people that
had previously been assessed for language dominance
by fTCD at the Universita¨tsklinikum Mu¨nster (Germany),
such that this information was available for preselection
purpose. All participants gave informed consent and had
to complete a questionnaire on handedness based on
the Edinburgh Handedness Inventory prior to partici-
pation. The experimental paradigm employed in the
Doppler sonography setting is well documented and
has successfully been used in a range of language lat-
eralization studies so far (Knecht et al., 2000, 2001,
2003). All participants were native German speakers.
Because previous laterality assessments had taken
place more than a year ago, hemispheric language dom-
inance was reassessed with fTCD during performance
of a verbal fluency task. Subsequently, participants per-
formed an OVP task in which they had to name three-,
five-, and seven-letter words presented briefly be-
tween two vertically aligned lines at different fixation
locations.
Functional Transcranial Doppler Sonography
Twenty participants were selected from the available co-
hort of people (13 men, 7 women; mean age 28.1 years;
12 left-handed, 8 right-handed). A 2-MHz transcranial
Doppler sonography device (Multidop T; DWL, Sipplingen,
Germany) was used to measure increases in cerebral
blood flow velocity (CBFV) within the left and right
middle cerebral arteries (MCAs) during performance of
a verbal fluency task. Participants were seated in front
of a monitor while a head device, supporting the 2-MHz
ultrasound probes, was fitted and the MCAs were lo-
cated (Ringelstein, Otis, Niggemeyer, & Kahlscheuer,
1990). The verbal fluency experiment started out with
a 15-sec rest period followed by an auditory signal that
indicated the cue phase, during which a random letter
of the alphabet was displayed on screen for 5 sec. A
second auditory signal marked the beginning of the
wordgen phase, lasting 6 sec, which required the par-
ticipant to silently generate as many words as possible
starting with the displayed letter. A third auditory cue
signaled the onset of the speak phase, during which the
words that had been found had to be repeated out loud
(12.5 sec). The end of the first cycle was indicated by
a fourth auditory cue (Figure 1). Twenty of these ex-
perimental cycles were recorded, lasting an entirety of
approximately 20 min. The Doppler signal resulted in
spectral envelope curves that were stored for off-line
analysis.
Data Analysis and LI Calculations
fTCD data were analyzed with the software package
AVERAGE (Deppe, Knecht, Henningsen, & Ringelstein,
1997). After preprocessing and automatic artifact rejec-
tion, the data were integrated over the corresponding
cardiac cycles, segmented into epochs that related to the
different experimental phases (rest period, cue, wordgen,
speak), and averaged. Mean CBFV values from the 15-sec
rest period were taken as baseline value. The relative
CBFV (rCBFV) changes in relation to the baseline value
were calculated and compared for each experimental
phase with the formula:
rCBFV ¼
V
ðtÞ
V
ðrest meanÞ
V
ðrest meanÞ
ð1Þ
where V
(t)
istheCBFVovertimeandV
(rest_mean)
refers to
the mean velocity in the rest period.
The Wilcoxon test was employed to statistically ana-
lyze the differences in blood flow velocity between the
left and right MCAs at each sample point, resulting in
an LI for each participant for the experimentally crucial
wordgen phase. We found eight participants to be right
dominant for language with fTCD_LI values ranging
from 1.17 to 4.93, and 12 participants showed typical
language dominance with values ranging from 1.39 to
7.79 (Table 1). For the current data set and the previ-
ously recorded f TCD LIs a test–retest correlation was
calculated for purpose of comparison (r = .78, p < .01,
t(18) = 5.286), revealing a strong consistency across
time for this measurement technique.
Behavioral OVP
The behavioral word-naming task was performed by the
same group of people who were assessed with fTCD.
Stimuli. Seventy each of three-, five-, and seven-letter
words served as stimuli. The stimulus sample contain-
ed German nouns only, which were selected through
WinWordGen (downloadable online users.ugent.be/
~wduyck/wwgdown.htm) and were controlled for fre-
quency and neighborhood size. Words were displayed in
their common format, with the initial letter capitalized.
Design. Each word, independent of its length, could be
seen at seven possible fixation locations shifted hori-
zontally across the screen. We chose this design (seven
fixation locations even for the shorter words) to be able
to present the same number of three- and five-letter
words as seven-letter words, in equal fixation locations.
This design also allowed us to examine whether there
was a continuity between foveal and parafoveal word rec-
ognition (Brysbaert, Vitu, & Schroyens, 1996). A three-
letter word was presented such that participants were
Hunter, Brysbaert, and Knecht 1375
fixating the blank space two letter positions before
the word (2; i.e., the complete word was in RVF), the
blank space before the word (1), the first letter of the
word (L1), the second letter of the word (L2), the third
letter of the word (L3), the blank space after the word
(1), or two letter positions after the word (2; see Fig-
ure 2). A five-letter word could be fixated on each letter
of the word (L1, L2, L3, L4, L5) or on the space before
Figure 1. Doppler curves for two representative participants. Average curves for CBFV changes in the left and right MCAs throughout the
different experimental phases. Green = right MCA; Red = left MCA. (A) A clear increase in CBFV in the left MCA during the wordgen phase
indicates typical left-hemisphere language dominance for this individual. fTCD_LI +5.84. (B) Atypical language dominance is illustrated
through an increase in CBFV in the right MCA. fTCD_LI 4.25.
1376 Journal of Cognitive Neuroscience Volume 19, Number 8
(1) or after the word (1). A seven-letter word could be
fixated on each possible letter position (L1, L2, L3, L4,
L5, L6, L7). Because it would have taken too many rep-
etitions of the stimuli, each participant did not see each
individual word at every possible fixation location, but at
three different positions only (i.e., the set of 210 stimuli
were repeated in three lists). Therefore, each partici-
pant was eventually exposed to 630 trials. The fixation
Table 1. Sex, Handedness, fTCD Laterality Indices, and OVP Slopes for Three-, Five-, and Seven-Letter Words for Each Participant
(Experiment 1)
Slope
Sex Handedness fTCD_LI Dominance 3-Letter Words 5-Letter Words 7-Letter Words
Par_17 M left 7.79 1 4 5.8 14.3
Par_02 M left 6.8 1 0.1 2.6 9.2
Par_22 M right 5.84 1 1.9 6.7 11
Par_12 F right 3.86 1 0.2 8.2 13.4
Par_04 M left 3.63 1 0.1 5.2 6.9
Par_20 M right 2.52 1 6.2 11.2 13.2
Par_10 M left 2.34 1 2.6 1.2 3.4
Par_11 M right 2.31 1 2.7 3.1 7.5
Par_08 F left 2.27 1 2.3 3.4 8.8
Par_16 M left 2.04 1 0.04 4.6 7.3
Par_19 M left 1.95 1 4.1 4.5 15.8
Par_07 F left 1.39 1 7.1 18.1 21.1
Par_03 M right 1.17 1 3.2 3.2 1.2
Par_21 F left 1.46 1 5.4 3.3 9.1
Par_18 F right 1.91 1 3.9 0.6 3.6
Par_14 M left 2.07 1 4.6 5.2 1
Par_05 M right 2.74 1 4.3 5.5 9.5
Par_15 M right 4.25 1 12.2 7.7 0.1
Par_13 F left 4.89 1 0.6 42.4
Par_01 F left 4.93 1 1.6 4.7 7.8
Figure 2. OVP design. Words
of all lengths were presented
at seven possible fixation
locations shifted horizontally
across the screen. Participants
had to name the word as fast
as possible.
Hunter, Brysbaert, and Knecht 1377
