A Magnetic Stimulation Examination of Orthographic

Neighborhood Effects in Visual Word Recognition

Michal Lavidor1 and Vincent Walsh2

Abstract
& The split-fovea theory proposes that visual word recog- representation of words is split between the cerebral hemi-
nition is mediated by the splitting of the foveal image, with spheres rather than bilateral. Behaviorally, we showed that
letters to the left of fixation projected to the right hemisphere words that have many orthographic neighbors sharing the
(RH) and letters to the right of fixation projected to the left same initial letters (‘‘lead neighbors’’) facilitated lexical
hemisphere (LH). We applied repetitive transcranial magnetic decision more than words with few lead neighbors. This
stimulation (rTMS) over the left and right occipital cortex effect did not apply to end neighbors (orthographic
during a lexical decision task to investigate the extent to neighbors sharing the same final letters). Crucially, rTMS
which word recognition processes could be accounted for over the RH impaired lead-, but not end-neighborhood
according to the split-fovea theory. Unilateral rTMS signifi- facilitation. The results support the split-fovea theory, where
cantly impaired lexical decision latencies to centrally pre- the RH has primacy in representing lead neighbors of a
sented words, supporting the suggestion that foveal written word. &

INTRODUCTION
the bilateral representation of the fovea. Assuming
When a person is fixating centrally, information pre- that the fovea is bilaterally represented in the cerebral
sented to the right of fixation (in the right visual field, cortex, foveal sparing may be due to the foveal repre-
or RVF) is projected to the visual cortex of the left sentation in the undamaged hemisphere, but estimates
cerebral hemisphere (LH) and information presented of the extent of the overlap vary: Bunt and Minckler
to the left of fixation (in the left visual field, or LVF) is (1977), for example, concluded that the bilateral repre-
projected to the visual cortex of the right cerebral sentation covers the whole fovea—a visual area of 38.
hemisphere (RH). However, it is not entirely clear how Others have estimated the overlap to be smaller: Wyatt
the visual half fields flank one another along the vertical (1978), for example, suggested that the overlap covers
meridian (Brysbaert, 1994). The current study aimed to only between 0.58 and 1.58 of the central representation.
look at this question by combining classical methods of Some recent studies have also favored the bilateral,
behavioral studies with the application of repetitive rather than split representation of the vertical midline
transcranial magnetic stimulation (rTMS) over the occip- with normal observers (Brandt, Stephan, Bense, Yousry,
ital cortex, in order to test the impact of a left versus & Dieterich, 2000) and also with hemianopia patients
right lesion on the processing of foveal stimuli. (Trauzettel-Klosinski & Reinhard, 1998).
However, the majority of studies with human subjects
have tended to support a split, rather than bilateral,
The Cortical Representation of the Fovea
cortical representation of the vertical midline. Harvey
Bilateral representation of the fovea has been demon- (1978), for example, presented visual targets to the left
strated in cats (Stone, 1966) and monkeys (Stone, Lei- and right of fixation at various retinal loci and found a
cester, & Sherman, 1973), and the overlap appears separation between crossed and uncrossed responding
primarily to be due to the presence of ipsilaterally pro- at all retinal loci. Targets located 0.258 and 0.58 from
jecting cells in the nasal hemiretina rather than to con- fixation, which should have fallen on the region of
tralaterally projecting cells in the temporal hemiretina nasotemporal overlap, produced a reaction time (RT)
(Leventhal, Ault, & Vitek, 1988). In addition to the difference between crossed and uncrossed responding
anatomical evidence, the foveal sparing phenomenon just as large as at the other loci (e.g., 28 and 38) outside
(e.g., Zihl, 1989) is considered as evidence supporting the overlap. If an area of functionally important overlap
projecting to both cerebral hemispheres exists, then the
need for interhemispheric crossing in word representa-
1
University of Hull, 2University College London tion would be eliminated. Thus, the conclusion from

D 2003 Massachusetts Institute of Technology Journal of Cognitive Neuroscience 15:3, pp. 354 – 363
Harvey’s (1978) study, as well from similar studies (Lines In light of the evidence favoring the split fovea, we
& Milner, 1983; Haun, 1978), was that the human fovea argue that aspects of letter and word recognition that
is not bilaterally represented or at least that the ana- are more characteristic of the right cerebral hemi-
tomical overlap does not appear to have functional sphere will also be reflected in the processing of the
consequences. In a recent study, magnetic resonance initial letters of words, while aspects of letter and
imaging (MRI) and visual field defects in patients with word recognition that are more characteristic of the
occipital lobe injury were correlated, leading Gray, left hemisphere (LH) will apply to the ends of words.
Galetta, Siegal, and Schatz (1997) to conclude that the In this article, the investigated variable was ortho-
foveal region is unilaterally represented in the human graphic neighborhood (N), recently reported as having
primary visual cortex. Fendrich and Gazzaniga (1989; see asymmetric effects in the two cerebral hemispheres
also Fendrich, Wessinger, & Gazzaniga, 1996) asked a (Lavidor & Ellis, 2002a, 2002b; Rubinstein, Henik, &
commissurotomy patient to compare target figures pre- Dronkers, 2001).
sented 18 or less from the retinal vertical midline with
reference figures presented distal to the midline in the
Orthographic Neighborhood
same or opposite visual field. The patient was required
to judge whether the two stimuli were the same or N is defined as the number of same-length words differ-
different. It was found that targets in the same visual ing from a target word by one letter (i.e., neighbors)
field as the reference were readily matched, but accuracy (Coltheart, Davelaar, Jonasson, & Besner, 1977). By this
was only at chance levels for targets in the opposite field. definition, the word marsh, for example, has two neigh-
These results were interpreted as indicating that the bors, harsh and march (Forster & Shen, 1996). In
patient’s fovea was effectively split with respect to the contrast, the word cover has 13 neighbors (including
left and right cerebral hemispheres (see also Sugishita, coven, covet, cower, hover, lover, mover, and rover).
Hamilton, Sakuma, & Hemmi, 1994). In a recent study Previous studies have shown, in general, facilitation
with healthy observers, visual targets presented to the effects for lexical decision for words with larger Ns
LVF and RVF evoked responses in the contralateral (so-called friendly words) (Andrews, 1997; Laxon, Colt-
occipital (and medial parietal) areas, even when the heart, & Keating, 1988; Laxon, Masterson, & Moran,
horizontal eccentricity was 08 (Portin & Hari, 1999). 1994). N effects are robust when frequency, regularity,
Behaviorally, it has been established that, at least for and age of acquisition are controlled (Morrison & Ellis,
brief presentation of foveal stimuli (up to 200 msec), 2000), although frequency of the neighboring words
there is no functional overlap along the vertical meridian may interact with N (Perea & Pollatsek, 1998; Segui &
(Gazzaniga, 2000). Grainger, 1990). Theoretical explanations of N effects
Recently, several behavioral studies with healthy par- generally invoke the notion that a written word can
ticipants have tested directly the predictions of the split- activate not only its own lexical entry but also the entries
fovea theory and concluded that there are no indications for other words of similar appearance. In certain con-
that the center of the visual field is bilaterally repre- ditions at least, activating those other lexical entries can
sented (Brysbaert, 1994; Brysbaert, Vitu, & Schroyens, facilitate the processing of the target word. Andrews
1996). The idea in these experiments was to look at (1997) proposed an explanation for neighborhood size
robust, previously reported visual field differences but effects in terms of McClelland and Rumelhart’s (1981)
with a different definition of visual fields, where RVF interactive activation model of visual word recognition,
starts from 08 to the right of fixation and LVF from suggesting that the processing units responsible for
midline to the left. Employing this definition of visual identifying letters in words receive more top-down sup-
fields, Lavidor, Ellis, Shillcock, and Bland (2001) have port from lexical units in the case of words with many
reported word length effects in the left side of centrally neighbors than in the case of words with few neighbors.
presented words (the initial letters) but not the right
side (the final letters), replicating the well-established
Orthographic Neighborhood Effects in a
finding of word length and hemifield interaction (Ellis,
Split-Fovea Model
Young, & Anderson, 1988).
The split fovea has been successfully implemented in In the split-fovea model, the RH is assumed to reflect
a connectionist model (Shillcock, Ellison, & Monaghan, N effects invoked by the initial letters (represented in
2000). Shillcock et al. (2000) investigated the optimal the LVF), and the LH would, symmetrically, reflect N of
processing of visually presented words in a connec- the end letters, represented in the RVF.
tionist model employing a split functional architecture. However, words are not simple symmetrical stimuli,
The results of their algorithm demonstrated reading and initial letters of words may be more informative
behavior that is observed in psycholinguistic experi- than their endings. For example, informative letter
ments including the optimal viewing position (OVP; sequences at the beginning of a word lead to a devia-
O’Regan, 1981) and the failure to fixate shorter words tion of the landing positions of the saccadic eye move-
(Rayner, 1998). ments toward the word beginning (Doré & Beauvillain,

Lavidor and Walsh 355

1997; Beauvillain, Doré, & Baudouin, 1996). Bryden, Transcranial Magnetic Stimulation and Visual
Mondor, Loken, Ingleton, and Bergstrom (1990) found Word Recognition
that subjects identified significantly more words with
Several types of visual recognition tasks have been
terminal deletions (missing letter or bigram at word’s
disrupted by TMS applied over the occipital pole at
ending, e.g., marc_ ) than words with an initial deletion
various delays from the letter or the word presentation
(missing letters at the word’s beginning, e.g., _arch). It
onset (e.g., Cowey & Walsh, 2000; Corthout, Uttl, Walsh,
has been proposed that the structure of the mental
Hallett, & Cowey, 1999; Corthout, Uttl, Ziemann, Cowey,
lexicon may not easily allow access via the end of the
& Hallett, 1999; Kammer, 1999; Kastner, Demmer, &
word, even if it is highly informative (O’Regan, Levy-
Ziemann, 1998; Amassian et al., 1989, 1998; Amassian,
Schoen, Pynte, & Brugaillere, 1984).
Cracco, et al., 1993, Amassian, Maccabee, et al., 1993).
The different contribution of initial letters to word
For example, single-letter identification was impaired
identification in comparison to final letters implies that
when TMS was applied over the striate visual cortex
‘‘orthographic neighbors are not all equal’’ (Perea,
(Amassian et al., 1989). Recently, we showed that rTMS
1998). Orthographic neighbors that share the initial
over the right occipital cortex has a significant inhibitory
letters with a target (we will term them lead neigh-
effect on lexical decision performance to LVF targets but
bors [LN], e.g., anger – angel ) may facilitate target
no significant effect on RVF targets. The complementary
recognition more than orthographic neighbors that
pattern of rTMS effects was obtained with LH stimula-
share the final letters with a target (end neighbors1
tion, which significantly impaired lexical decision to RVF
[EN], e.g., aloud – cloud ). At the level of processing in
but not LVF targets (Lavidor, Ellison, & Walsh, in press).
the occipital cortex, it is more likely that the RH
activates lead neighbors but the LH does not neces-
sarily activate end neighbors. This hypothesis received RESULTS
some indirect support recently by Lavidor and Ellis
(2002a, 2002b) and Rubinstein et al. (2001), who RTs and the percentage of incorrect responses for words
found robust N effects in the LVF, but not in the are summarized in Table 1.
RVF. The aim of the current study was therefore to
look into the question of orthographic neighborhood
Target Lexicality
effects within the context of the split-fovea theory by
temporarily disrupting processing in the RH and LH Target lexicality had a significant effect on RTs, F(1,7) =
with repetitive TMS. 5.22, p < .05, and accuracy, F(1,7) = 8.3, p < .01, with

Table 1. Mean RTs and Standard Deviations for Word Targets (in msec) and Percentage of Error as a Function of Target
Orthographic Neighborhood Size and Site of Magnetic Stimulation

Many Lead-N, Many Lead-N, Few Lead-N, Few Lead-N,

Many End-N Few End-N Many End-N Few End-N
No-TMS
Mean RT 425 435 462 477
SD 46 57 72 60
% error 0 0 2 2

RH TMS
Mean RT 505 501 485 490
SD 87 55 77 73
% error 5 8 6 0

LH TMS
Mean RT 476 470 508 513
SD 92 55 73 66
% error 7 4 6 10

Lead-N = orthographic lead-neighbors; End-N = orthographic end-neighbors.

356 Journal of Cognitive Neuroscience Volume 15, Number 3

faster and more accurate responses to words (478 msec,
96% correct) than to nonwords (513 msec, 89% correct).
These results are in accordance with previous behavioral
literature (Chiarello, 1988; Ellis et al., 1988). Because the
neighborhood effects for words were the focus of the
current study, the rest of the analyses was performed
only for word targets.

Reaction Times
The application of rTMS had a significant effect,
F(2,14) = 10.5, p < .01, with RTs being slowed when
TMS was applied to the LH (mean = 492 msec) or to the
RH site (mean = 495 msec) compared with the no-TMS
mean of 449 msec (all post hoc differences were Bon-
ferroni corrected to p < .05). Orthographic neighbor-
hood group had a significant effect on response latency,
Figure 1. rTMS effect size for RTs for words as a function of word
F(3,21) = 5.93, p < .01, with faster responses to words orthographic neighborhood size and magnetic stimulation site. RH
with many LN (many LN + many EN = 468 msec, many rTMS significantly impairs responses to stimuli with many lead
LN + few EN = 469 msec) compared with responses to neighbors (LN) but had little effect when words had few LN. LH rTMS
words with few LN (few LN + many EN = 485 msec, few produced an equal (small, yet significant) inhibition for all word types.
LN + few EN = 493 msec).
The interaction between rTMS site (RH, LH) and
orthographic neighborhood was also significant, model. In view of the evidence supporting the split-
F(6,42) = 3.42, p < .05. Separate analyses testing the fovea theory (Gazzaniga, 2000; Shillcock et al., 2000;
neighborhood effect under the three rTMS conditions Portin & Hari, 1999; Gray et al., 1997; Brysbaert, 1994),
revealed significant N effect when rTMS was not applied, the starting point of this investigation was that when a
F(3,21) = 5.1, p < .05, and when the application site was person is fixating centrally, information presented to
in the LH, F(3,21) = 4.8, p < .05. However, the ortho- the right of fixation (in the RVF) is projected to the
graphic neighborhood effect was abolished by RH rTMS, visual cortex of the left cerebral hemisphere (LH) due
F(3,21) = 1.3, ns, that is, the facilitation gained for words to the anatomical arrangement of the human visual
with many LN was impaired only when rTMS was applied system. Conversely, information presented to the left
to the RH. This can be clearly seen in Figure 1, which of fixation (in the LVF) is projected to the visual cortex
presents rTMS effect sizes (the difference in RTs from of the right cerebral hemisphere. Assuming a split-
the no-TMS condition). In the RH site, the rTMS effect foveal representation, we specifically investigated the
(i.e., the inhibition caused by rTMS) is significantly larger importance of the RH in processing the initial letters of
for words with many LN when compared with words a written word and of the LH in representing the final
with few LN. However, the rTMS effect when stimulating letters of centrally presented targets. We presented
the LH site caused equal (small, yet significant) inhi- words that had many or few lead neighbors (same-
bition for all word types. Having few or many EN did not length words that shared the initial letters) and many or
affect performance. few end neighbors (same-length words that shared the
final letters with a target word). Behaviorally, lead-
neighborhood density facilitated performance rather
Error Rates
than end-neighborhood density. However, this facilita-
The application of rTMS had an overall significant effect tion was significantly impaired when rTMS was applied
on error rates, F(2,14) = 6.3, p < .05, with significantly to the RH, but not to the LH. Thus, the ability to
more errors when TMS was applied to the LH (mean = respond to the initial letters is dependent upon RH
6.7%) or to the RH site (mean = 5%) compared with processes. Orthographic neighborhood density was
the no-TMS mean of 1%. Orthographic neighborhood selected as a variable of interest that may characterize
neither interacted with rTMS site nor directly affected the RH processes in word recognition under a split-
error rates to word stimuli. fovea constraint due to the greater orthographic neigh-
borhood effects in the RH but not in the LH (Lavidor &
Ellis, 2002a, 2002b; Rubinstein et al., 2001).
DISCUSSION
One concern that may account for the results is a
In a lexical decision experiment with rTMS applied to possible systematic bias towards a certain visual field.
the left and right visual cortex, we investigated the However, the stimuli were all centrally presented, and
function of the RH in terms of a split-word recognition usually subjects are successful in maintaining fixation to

Lavidor and Walsh 357

foveal stimuli, and only a small percentage of trials are Theoret, & Pascual-Leone, 2001; Fierro et al., 2000;
not properly fixated (less then 1%, see, for instance, Walsh, Ellison, Ashbridge, & Cowey, 1999; Seyal, Ro,
Rayner, 1998). Systematic shifts of eye position to RVF & Rafal, 1995). Thus, the results we obtained do not rule
were reported by one research group ( Jordan, Patching, out the possibility of some functional significance
& Milner, 2000; Patching & Jordan, 1998); however, most behind the anatomical evidence of bilateral representa-
eye-movement researchers would report, if at all, an LVF tion of the fovea. However, interhemispheric interac-
bias due to the OVP (the tendency to fixate slightly to tions cannot explain the significant interaction we
the left of fixation, see O’Regan, 1981). More specifically, obtained between orthographic neighborhood of the
in a companion lexical decision and TMS study (Lavidor initial letters and rTMS site. One other possibility is a
et al., in press), we have monitored fixation during the different model that is a weaker version of the split
TMS sessions and failures to fixate centrally presented model, that is, where both hemispheres maintain a
words occurred in 0.09% of all trials. Therefore, one can strong representation of the contralateral stimuli and a
preclude the possibility that the current data reflect weak representation of the ipsilateral stimuli of centrally
systematic RVF shifts. presented targets. Even if this is the case, our results
Before carrying out a TMS procedure, it was necessary show clearly that the supposedly ‘‘weaker’’ letters that
to obtain reliable baseline effects with the behavioral are ipsilateral to the unstimulated hemisphere did not
task. Word and nonword targets were briefly presented affect performance. They may be there; however, neigh-
in the center of the monitor while right-handed partic- borhood effects of the ‘‘weaker’’ letters were not found,
ipants made lexical decision judgments. The results were so functionally, the split is complete. We have shown
compatible with those of numerous previous lateraliza- how rTMS over the RH can interfere selectively with the
tion experiments (Iacoboni & Zaidel, 1996; Chiarello, processing of the initial letters of centrally presented
1988; Ellis et al., 1988); words were responded to faster words (but not of final letters). This is a strong test of
and more accurately than nonwords, and we found the the split representation of the fovea theory.
predicted orthographic lead-neighborhood effect. In summary, this investigation of the split-foveal rep-
The rTMS effects we found for briefly presented resentation contributes to the emerging literature that is
targets are consistent with other rTMS findings with providing a starting point for the reconceptualization of
occipital stimulation in which TMS is only able to disrupt visual word recognition research in terms of the splitting
perceptual judgments if the relative duration of presen- of the fovea. The results reported here provide support
tation is short (e.g., Amassian et al., 1989), stimuli are for the theory that during fixation, letters to the left of
close to luminance-detection thresholds (Miller, Fen- fixation are projected to and processed in the right
drich, Eliassen, Demirel, & Gazzaniga, 1996) or both cerebral hemisphere and letters to the right of fixation
(see Kammer & Nusseck, 1998; see also Walsh & Cowey, are projected to and processed in the left cerebral
2000; Walsh & Pascual-Leone, 2003, for details of the hemisphere. The challenge for this new area of visual
relationship between stimuli and the temporal duration word recognition research is to integrate this theory
of TMS effects). In the present study, error rates were with the more complex tasks involved in the reading of
affected by rTMS application; however, they were not connected text.
sensitive to the manipulation of orthographic neighbor-
hood. Because performance was nearly errorless (partic- METHODS
ularly for words), the lack of N-related effects for the
accuracy measure may be due to ceiling level perfor- Transcranial Magnetic Stimulation Equipment
mance. Although errors may be a useful measure of The stimulator used was a MagStim (Super Rapid model)
performance close to threshold, it has been difficult to with four external boosters (maximum output 2 T).
obtain errors in cognitive tasks with TMS (see Walsh & Magnetic stimulation was applied at 65% of the maxi-
Cowey, 1998; Walsh & Pascual-Leone, 2003, for meth- mum using a figure-of-eight 70-mm coil (1.3 T). The
odological details). double coil windings in the figure-of-eight coil carry two
The significant behavioral disruption induced by TMS currents in opposite directions and at the midpoint of
during the presentation of targets at the center of the coil where the two loops meet, there is a localized
fixation at first sight suggests it is unlikely that two summation of current. A focal electric current is induced
complete copies of the centrally presented word were in the cortex by the magnetic pulse that undergoes
processed in each hemisphere. However, independent minimal attenuation by the intervening soft tissue and
representations do not preclude interactions between bone. Previous studies have demonstrated that magnetic
the two hemispheres, and when the word is presented stimulation using this type of coil can produce function-
in the region of greatest representational overlap (i.e., at ally dissociable effects when moving the coil by 0.5 – 1 cm
fixation in the midline), it may be the case that any on the scalp (Brasil-Neto, McShane, Fuhr, Hallett, &
disruption to word processing may interfere with either Cohen, 1992). The center of the coil was positioned
competition or cooperation between the hemispheres over the site to be stimulated such that the windings
that is a feature of normal processing (e.g., Hilgetag, were to the left and to the right of it and the handle of

358 Journal of Cognitive Neuroscience Volume 15, Number 3

the coil pointed vertically. The level of stimulation (65%) neighbors (Quinlan, 1993, mean = 1.6 neighbors), so
selected in this article matches the conditions used in that the stimulus groups did not differ on any of these
a previous study of split-foveal processing with TMS categories. The nonwords were generated from another
(Lavidor et al., in press). word pool of six-letter words, with similar frequency
range, by changing one or two letters, such that the
nonwords were pronounceable (and had many and few
Participants
LN and EN as described).
Eight native English-speaking participants, four women Each of the four groups was further divided to three
and four men, took part in the experiment. All of the subgroups (20 stimuli in each subgroup), matched in
participants had normal or corrected-to-normal vision frequency, imageability, and N. The subgroups were
and were aged between 19 and 30 (mean 23.5). All the randomly assigned to one of the three TMS conditions
participants were right-handed and scored at least 85 (no-TMS, TMS to the LH, or TMS to the RH). During the
according to the Edinburgh Handedness Inventory experiment, each stimulus was repeated twice in two
(Oldfield, 1971), with mean score of 96. Participants different TMS conditions.
received £10 for their participation. The stimuli were presented in Times Roman font,
Before taking part in the experiment, participants were size 16, such that target width was 1.68 when viewed
given an information leaflet that explained the procedure from 70 cm. The targets were black on a light gray back-
to be used and were given at least 24 hr to decide ground and were presented in the center of the screen
whether they wished to participate. All participants for 150 msec.
signed a consent form and reported absence of epilepsy
or any other neurological conditions in themselves and
Experiment
their family. The Oxford Research Ethical Committee
granted ethical committee approval for all procedures. We applied rTMS over the left and right occipital
cortex during a lexical decision task to investigate the
extent to which word recognition processes could be
Stimuli
accounted for according to the split-fovea theory. The
The experimental stimuli consisted of 120 words and targets we presented differ in orthographic neighbor-
120 nonwords all containing six letters. The stimuli were hood density. The experiment tested the predictions
selected such that they would match one of four stim- that (1) stimulating the right or left sites would impair
ulus groups created from the four possible combinations lexical decision to centrally presented words, (2) words
of many and few lead and end neighbors according to with many lead neighbors would be processed faster
the following criteria: than words with few lead neighbors, but end neigh-
bors would have a smaller, if any, effect, and, crucially,
1. Lead neighbors (LN): The number of six-letter
(3) stimulating the RH site would decrease the lead-
words that could be made beginning with the same first
neighborhood effect, but LH TMS would not inhibit
three letters of each stimulus word or nonword.2 The
the end-neighborhood effect. This last prediction is
range of this variable was 15 –40 neighbors for the many
based on the split-fovea theory and the asymmetric
LN group and 0 –7 neighbors for the few LN group.
orthographic neighborhood effect.
2. End neighbors (EN): The number of six-letter
A 4
3
2 factorial design (Orthographic Neighbor-
words that could be made ending with the same final
hood: many LN + many EN, many LN + few EN, few
three letters of the stimuli. The range of this variable was
LN + many EN, and few LN + few EN)
(TMS: no
15 –70 neighbors for the many EN group and 0 –10
stimulation, RH stimulation, or LH stimulation)
(tar-
neighbors for the few EN group.
get lexicality: word, nonword) was used in a within-
Group 1 stimuli (30 words and 30 nonwords, e.g., subjects design. Dependent variables were RT and
castle) had many lead neighbors (mean number of percentage of incorrect responses. The application of
LN = 28.9). This group of words also had many end repetitive TMS was presented in alternating blocks of
neighbors (mean number of EN = 39.4). Group 2 single-hemisphere stimulation. The other variables were
stimuli (again 30 words and 30 nonwords, e.g., cheese) randomly applied.
had many LN (mean = 30) but few EN (mean = 9).
Group 3 stimuli (e.g., cement) had few LN (mean = 3) Procedure
and many EN (mean = 41). Group 4 (e.g., coffin) had
few LN (mean = 4) and few EN (mean = 7). Pre-experiment: Induction of Stationary Phosphenes
The word stimuli in each of the four groups were Before attempting to disrupt word processing, the sites
matched according to written frequency (frequency to be stimulated were defined by eliciting phosphenes
norms by Kucera & Francis, 1967, mean frequency = from the occipital cortex to ensure that hemispheric
10.00 per million), imageability (Coltheart, 1981, mean = stimulation selectively affected contralateral visual field
363, from a 100– 700 scale), and number of orthographic processing. Participants wore a latex swimming cap and

Lavidor and Walsh 359

Figure 2. Experimental time
line of the task used in this
study. Stimuli were presented
for 150 msec, and rTMS was
applied for 500 msec at target
onset.

sat with their head supported by a chin rest and head distance to the right of the central point was 1.55 cm. All
strap in order to secure head position and stabilize but one participant reported left to central phosphenes
fixation. Stimulation sites were the right and left corti- when the RH was stimulated and right to central phos-
ces. The upper edge of the inion was marked on the cap, phenes when the LH was stimulated. For the main
and another point (the reference point) was marked experiment, we used the effective phosphene sites for
2 cm above it. The occipital hemispheric sites that were each participant with a fixed 65% of the stimulator
marked on the cap were 1.5 cm to the left of the center output, a level selected based on previous experiments
point (LH site) and 1.5 cm to the right of the central and found to be sufficient to disrupt perception without
point (RH site). The coordinates were selected initially masking stimuli with overt phosphenes. For the partic-
based on previously successful studies with TMS, which ipant who failed to see phosphenes, we used the mean
reported stationary phosphenes and suppression of values of the group as stimulation sites. This subject was
visual perception tasks with TMS at similar sites (Lavidor included because he had taken part in several previous
et al., in press; Pascual-Leone & Walsh, 2001; Cowey & experiments in which visual suppression was induced
Walsh, 2000; Kammer, 1999; Stewart, Battelli, Walsh, & and because visual effects have been coregistered with
Cowey, 1999; Stewart, Ellison, Walsh, & Cowey, 2001; his anatomical MRI scan.
Kastner et al., 1998; Amassian et al., 1989).
At the phosphene stage, we located the stimulation
Main Experiment
sites with TMS delivered at a frequency of 2 Hz and
between 60% and 90% of stimulator output. In a dark- At the experimental stage, we applied rTMS with 65%
ened room, participants closed their eyes while TMS was of the stimulator output at 8-Hz frequency for
delivered to the RH and LH points. Participants were 500 msec from the onset of the target presentation
asked to indicate whenever they saw a phosphene and while participants were performing the lexical decision
to describe its position in space. TMS was applied at task. Stimulus presentation was controlled by an IBM
increasing intensities until participants reported seeing Pentium II computer on a 1700 SVGA display with
phosphenes regularly and reliably. For some partici- 100-Hz refresh rate.
pants, the magnetic stimulation sites were changed in The experimental trials were employed in 8 blocks, 60
a ‘‘win-stay/lose-shift’’ paradigm (see Ashbridge, Walsh, trials in each block, with a total of 480 trials per
& Cowey, 1997) to locate regions on the scalp that participant. While orthographic neighborhood and tar-
resulted in phosphene perception. The mean distance get lexicality (word or nonword) were randomly pre-
to the left of the central point was 1.56 cm and the mean sented, magnetic stimulation sites were fixed per block.

360 Journal of Cognitive Neuroscience Volume 15, Number 3

Thus, we had four blocks with rTMS applied to the right enables the bilateral speeded coding of word length when at
visual cortex and four blocks with rTMS applied to the the initial stages the word is split between the two hemispheres
requires further study. One possibility is the magnocellular
left visual cortex, in alternating order. In each block, pathway, which encodes low spatial frequency information and
rTMS was given in alternating trials to allow 4-sec delay in which neuronal latencies are short in early visual processing
between successive rTMS trials. Thus, in each block, (see Stein & Walsh, 1997; Stuart & Lovegrove, 1992).
every rTMS trial was followed by a non-TMS trial and so
forth. The rationale for this design was to maintain
stimulus presentation at a fast rate typical of lateraliza- REFERENCES
tion studies. Each of the 240 stimuli was presented
Amassian, V. E., Cracco, R. Q., Maccabee, P. J., Cracco, J. B.,
twice (in separate blocks), once in the no-TMS condi- Rudell, A. P., & Eberle, L. (1989). Suppression of visual
tion and once with RH or LH stimulation (order of perception by magnetic coil stimulation of human
stimulation was counterbalanced across participants). occipital cortex. Electroencephalography and Clinical
Each trial began with a fixation cross appearing in Neurophysiology, 74, 458 – 462.
the center of the screen for 75 msec. For the first trial, Amassian, V. E., Cracco, R. Q., Maccabee, P. J., Cracco, J. B.,
Rudell, A. P., & Eberle, L. (1993). Unmasking human
the cross remained visible for 2000 msec until the visual-perception with the magnetic coil and its
onset of the target. The cross would again reappear relationship to hemispheric asymmetry. Brain Research,
for 1800 msec, waiting for the participant’s response. 605, 312 – 316.
Whatever the response, after 1800 msec, a new trial Amassian, V. E., Cracco, R. Q., Maccabee, P. J., Cracco, J. B.,
sequence would begin. Targets were briefly presented Rudell, A. P., & Eberle, L. (1998). Transcranial magnetic
stimulation in the study of the visual pathway. Journal of
for 150 msec at the center of the screen (either a word Clinical Neurophysiology, 15, 288 – 304.
or a nonword) in a random order (see the time line of Amassian, V. E., Maccabee, P. J., Cracco, R. Q., Cracco, J. B.,
the experiment in Figure 2). The participant’s task was Rudell, A. P., & Eberle, L. (1993). Measurement of
to decide, as quickly and as accurately as possible, information-processing delays in human visual cortex with
whether the stimulus was a legal English word or a repetitive magnetic coil stimulation. Brain Research, 605,
317 – 321.
nonword. Participants responded by pressing one of Andrews, S. (1997). The effect of orthographic similarity
two available response buttons with two fingers of the on lexical retrieval: Resolving neighborhood conflicts.
right hand. For half of the participants, the response Psychonomic Bulletin and Review, 4, 439 – 461.
‘‘word’’ was made by pressing the right button, and Ashbridge, E., Walsh, V., & Cowey, A. (1997). Temporal aspects
‘‘nonword’’ by pressing the left button. For half of the of visual search studied by transcranial magnetic stimulation.
Neuropsychologia, 35, 1121 – 1131.
participants, the response keys were reversed. Each Beauvillain, C., Doré, K., & Baudouin, V. (1996). The ‘‘center of
session began with 12 practice trials of centrally pre- gravity’’ of words: Evidence for an effect of the word – initial
sented letter strings, where the task was to perform a letters. Vision Research, 36, 589 – 603.
lexical decision. Participants were informed that central Brandt, T., Stephan, T., Bense, S., Yousry, T. A., & Dieterich, M.
fixation was important. (2000). Hemifield visual motion stimulation: An example of
interhemispheric crosstalk. NeuroReport, 11, 2803 – 2809.
Brasil-Neto, J. P., McShane, L. M., Fuhr, P., Hallett, M., &
Cohen, L. G. (1992). Topographic mapping of the human
Acknowledgments motor cortex with magnetic stimulation: Factors affecting
This study was supported by the European Commission, by a accuracy and reproducibility. Electroencephalography and
Marie Curie fellowship grant (Contract No. HPMF-CT-2000- Clinical Neurophysiology, 85, 9 – 16.
00787), by a Wellcome grant, and by a visit grant from Brain Bryden, M. P., Mondor, T. A., Loken, M., Ingleton, M. A., &
awarded to M. Lavidor. V. Walsh is supported by the Royal Bergstrom, K. (1990). Locus of information in words and the
Society. Related support was granted by the Medical Research right visual field effect. Brain and Cognition, 14, 44 – 58.
Council and the Dr. Hadwen Research Trust. Brysbaert, M. (1994). Interhemispheric transfer and the
processing of foveally presented stimuli. Behavioral Brain
Reprint requests should be sent to Michal Lavidor, Department Research, 64, 151 – 161.
of Psychology, University of Hull, Hull HU6 7RX, England, UK, Brysbaert, M., Vitu, F., & Schroyens, W. (1996). The right visual
or via e-mail: [email protected]. field advantage and the optimal viewing position effect:
On the relation between foveal and parafoveal word
recognition. Neuropsychology, 10, 385 – 395.
Notes Bunt, A. H., & Minckler, D. S. (1977). Foveal sparing:
New anatomical evidence for bilateral representation of
1. End neighbors are also termed body neighbors by Ziegler the central retina. Archives of Ophthalmology, 95,
and Perry (1998). 1445 – 1447.
2. This definition of LN and EN as same-length words that Chiarello, C. (1988). Lateralization of lexical processes in the
share initial or final letters with a target comprises an implicit normal brain: A review of visual half-field research. In H. A.
assumption that word length is encoded initially by both Whitaker (Ed.), Contemporary reviews in neuropsychology
hemispheres, despite the split process. Elsewhere (Lavidor, (pp. 36 – 76). New York: Springer-Verlag.
Ellis, Shillcock, & Hayes, in preparation), we have shown the Coltheart, M. (1981). The MRC psycholinguistic database.
preservation of word length coding when four-letter lead Quarterly Journal of Experimental Psychology, 33A,
neighbors primes were more effective than seven-letter primes 497 – 505.
to four-letter target words. The underlying mechanism that Coltheart, M., Davelaar, E., Jonasson, J. T., & Besner, D.

Lavidor and Walsh 361

 (1977). Access to the internal lexicon. In S. Dornic (Ed.), Kucera, H., & Francis, W. N. (1967). Computational analysis of
Attention and performance VI: The psychology of reading present day American English. Providence, RI: Brown
(vol. 6, pp. 535 – 555). London: Academic Press. University Press.
Corthout, E., Uttl, B., Walsh, V., Hallett, M., & Cowey, A. Lavidor, M., & Ellis, A. W. (2002a). Word length and orthographic
(1999). Timing of activity in early visual cortex as revealed neighborhood size effects in the left and right cerebral
by transcranial magnetic stimulation. NeuroReport, 10, 1 – 4. hemispheres. Brain and Language, 80, 45 – 62.
Corthout, E., Uttl, B., Ziemann, U., Cowey, A., & Hallet, M. Lavidor, M., & Ellis, A. W. (2002b). Orthographic neighborhood
(1999). Two periods of processing in the (circum) effects in the right but not in the left cerebral hemisphere.
striate visual cortex as revealed by transcranial magnetic Brain and Language, 80, 63 – 76.
stimulation. Neuropsychologia, 37, 137 – 145. Lavidor, M., Ellis, A. W., Shillcock, R., & Bland, T. (2001).
Cowey, A., & Walsh, V. (2000). Magnetically induced Evaluating a split processing model of visual word
phosphenes in sighted, blind and blindsighted observers. recognition: Effects of word length. Cognitive Brain
NeuroReport, 11, 3269 – 3273. Research, 12, 265 – 272.
Doré, K., & Beauvillain, P. (1997). Latency dependence of Lavidor, M., Ellis, A. W., Shillcock, R., & Hayes, A.
word – initial letter integration by the saccadic system. (in press). Evaluating a split processing model
Perception and Psychophysics, 59, 523 – 533. of visual word recognition: Effects of orthographic
Ellis, A. W., Young, A. W., & Anderson, C. (1988). Modes of neighborhood size. Brain and Language.
word recognition in the left and right cerebral hemispheres. Lavidor, M., Ellison, A., & Walsh, V. (in press). A virtual lesion
Brain and Language, 35, 254 – 273. examination of a split processing model of visual word
Fendrich, R., & Gazzaniga, M. S. (1989). Evidence of foveal recognition. Visual Cognition.
splitting in a commissurotomy patient. Neuropsychologia, Laxon, V., Masterson, J., & Moran, R. (1994). Are children’s
27, 273 – 281. representations of words distributed? Effects of orthographic
Fendrich, R., Wessinger, C. M., & Gazzaniga, M. S. (1996). neighborhood size, consistency and regularity of naming.
Nasotemporal overlap at the retinal vertical meridian— Language and Cognitive Processes, 9, 1 – 27.
investigations with a callosotomy patient. Neuropsychologia, Laxon, V. J., Coltheart, V., & Keating, C. (1988). Children find
34, 637 – 646. friendly words friendly too: Words with many orthographic
Fierro, B., Brighina, F., Oliveri, M., Piazza, A., La Bua, V., neighbors are easier to read and spell. British Journal of
Buffa, D., & Bisiach, E. (2000). Contralateral neglect induced Educational Psychology, 58, 103 – 119.
by right posterior parietal rTMS in healthy subjects. Leventhal, A. G., Ault, S. J., & Vitek, D. J. (1988). The
NeuroReport, 11, 1519 – 1521. nasotemporal division in primate retina: The neural bases of
Forster, K. I., & Shen, D. (1996). No enemies in the macular sparing and splitting, Science, 240, 66 – 67.
neighborhood: Absence of inhibitory neighborhood effects Lines, C. R., & Milner, A. D. (1983). Nasotemporal overlap in
in lexical decision and semantic categorization. Journal of the human retina investigated by means of simple reaction
Experimental Psychology: Learning, Memory, and time to lateralized light flash. Experimental Brain Research,
Cognition, 22, 696 – 713. 50, 166 – 172.
Gazzaniga, M. S. (2000). Cerebral specialization and inter- McClelland, J. L., & Rumelhart, D. E. (1981). An interactive
hemispheric communication. Brain, 123, 1293 – 1326. activation model of context effects in letter perception:
Gray, L. G., Galetta, S. L., Siegal, T., & Schatz, N. J. (1997). The I. An account of basic findings. Psychological Review, 88,
central visual field in homonymous hemianopia—evidence 375 – 407.
for unilateral foveal representation. Archives of Neurology, Miller, M. B., Fendrich, R., Eliassen, J. C., Demirel, S., &
54, 312 – 317. Gazzaniga, M. S. (1996). Transcranial magnetic stimulation:
Harvey, L. O., Jr. (1978). Single representation of the visual Delays in visual suppression due to luminance changes.
midline in humans. Neuropsychologia, 16, 601 – 610. NeuroReport, 7, 1740 – 1744.
Haun, F. (1978). Functional dissociation of the hemispheres Morrison, C. M., & Ellis, A. W. (2000). Real age of acquisition
using foveal visual input. Neuropsychologia, 16, 725 – 733. effects in word naming. British Journal of Psychology, 91,
Hilgetag, C. C., Theoret, H., & Pascual-Leone, A. (2001). 167 – 180.
Enhanced visual spatial attention ipsilateral to rTMS-induced Oldfield, D. P. (1971). The assessment and analyses of
‘‘virtual lesions’’ of human parietal cortex. Nature handedness. The Edinburgh Inventory. Neuropsychologia,
Neuroscience, 4, 953 – 957. 9, 97 – 113.
Iacoboni, M., & Zaidel, E. (1996). Hemispheric independence O’Regan, J. K. (1981). The convenient viewing hypothesis,
in word recognition: Evidence from unilateral and bilateral In D. F. Fisher, R. A. Monty, & J. W. Senders (Eds.),
presentations. Brain and Language, 53, 121 – 140. Eye movements: Cognition and visual perception
Jordan, R. T., Patching, G. R., & Milner, A. D. (2000). Lateralized (pp. 289 – 298). Hillsdale, NJ: Erlbaum.
word recognition: Assessing the role of hemispheric O’Regan, J. K., Levy-Schoen, A., Pynte, J., & Brugaillere, B.
specialization, modes of lexical access, and perceptual (1984). Convenient fixation location within isolated words of
asymmetry. Journal of Experimental Psychology: Human different length and structure. Journal of Experimental
Perception and Performance, 26, 1192 – 1208. Psychology: Human Perception and Performance, 10,
Kammer, T. (1999). Phosphenes and transient scotomas 250 – 257.
induced by magnetic stimulation of the occipital lobe: Pascual-Leone, A., & Walsh, V. (2001). Fast back projections
Their topographic relationship. Neuropsychologia, 36, from the motion to the primary visual area necessary for
191 – 198. visual awareness. Science, 292, 510 – 512.
Kammer, T., & Nusseck, H. G. (1998). Are recognition deficits Patching, G. R., & Jordan, T. R. (1998). Increasing the benefits
following occipital lobe TMS explained by raised detection of eye-tracking devices in divided visual field studies of
thresholds? Neuropsychologia, 36, 1161 – 1166. cerebral asymmetry. Behavior Research Methods,
Kastner, S., Demmer, I., & Ziemann, U. (1998). Transient visual Instruments and Computers, 30, 643 – 650.
field defects induced by transcranial magnetic stimulation Perea, M. (1998). Orthographic neighbors are not all equal:
over human occipital pole. Experimental Brain Research, Evidence using an identification technique. Language and
118, 19 – 26. Cognitive Processes, 13, 77 – 90.

362 Journal of Cognitive Neuroscience Volume 15, Number 3

Perea, M., & Pollatsek, A. (1998). The effects of neighborhood Stone, J. (1966). The naso-temporal division of the cat’s retina.
frequency in reading and lexical decision. Journal of Journal of Comparative Neurology, 136, 585 – 600.
Experimental Psychology: Human Perception and Stone, J., Leicester, J., & Sherman, S. M. (1973). The
Performance, 24, 767 – 779. naso-temporal division of the monkey’s retina. Journal of
Portin, K., & Hari, R. (1999). Human parieto-occipital visual Comparative Neurology, 150, 333 – 348.
cortex: Lack of retinotopy and foveal magnification. Stuart, G. W., & Lovegrove, W. J. (1992). Visual processing
Proceedings of the Royal Society B, 266, 981 – 985. deficits in dyslexia—receptors or neural mechanisms.
Quinlan, P. (1993). The Oxford psycholinguistic database. Perceptual and Motor Skills, 74, 187 – 192.
Oxford: Oxford University Press. Sugishita, M., Hamilton, C. R., Sakuma, I., & Hemmi, I. (1994).
Rayner, K. (1998). Eye movements in reading and information Hemispheric representation of the central retina of
processing: 20 years of research. Psychological Bulletin, 124, commissurotomized participants. Neuropsychologia, 32,
373 – 422. 399 – 415.
Rubinstein, O., Henik, A., & Dronkers, N. F. (2001). Is the right Trauzettel-Klosinski, S., & Reinhard, J. (1998). The vertical field
hemisphere capable of reading and understanding a word? border in hemianopia and its significance for fixation and
Presented in ESCOP, Edinburgh, 4 – 7 September. reading. Investigative Ophthalmology and Visual Science,
Segui, J., & Grainger, J. (1990). Priming word recognition 39, 2177 – 2186.
with orthographic neighbors: Effects of relative prime target Walsh, V., & Cowey, A. (1998). Magnetic stimulation studies of
frequency. Journal of Experimental Psychology: Human visual cognition. Trends in Cognitive Sciences, 2, 103 – 110.
Perception and Performance, 16, 65 – 76. Walsh, V., & Cowey, A. (2000). Transcranial magnetic
Seyal, M., Ro, T., & Rafal, R. (1995). Increased sensitivity to stimulation and cognitive neuroscience. Nature Reviews
ipsilateral cutaneous stimuli following transcranial magnetic Neuroscience, 1, 73 – 79.
stimulation of the parietal lobe. Annals of Neurology, 38, Walsh, V., Ellison, A., Ashbridge, E., & Cowey, A. (1999). The role
264 – 267. of the parietal cortex in visual attention—hemispheric
Shillcock, R., Ellison, T. M., & Monaghan, P. (2000). asymmetries and the effect of learning: A magnetic
Eye-fixation behavior, lexical storage and visual word stimulation study. Neuropsychologia, 37, 245 – 251.
recognition in a split processing model. Psychological Walsh, V., & Pascual-Leone, A. (2003). Transcranial
Review, 107, 824 – 851. magnetic stimulation: A neurochronometrics of mind.
Stein, J., & Walsh, V. (1997). To see but not to read; The Cambridge: MIT press.
magnocellular theory of dyslexia. Trends in Neuroscience, Wyatt, H. J. (1978). Nasotemporal overlap and visual field
20, 147 – 152. sparing. Investigative Ophthalmology and Visual Science,
Stewart, L., Battelli, L., Walsh, V., & Cowey, A. (1999). 17, 1128 – 1130.
Motion perception and perceptual learning: A magnetic Ziegler, J. C., & Perry, C. (1998). No more problems in
stimulation study. Electroencephalography and Clinical Coltheart’s neighborhood: Resolving neighborhood conflicts
Neurophysiology, 51, 334 – 350. in the lexical decision task. Cognition, 68, B53 – B62.
Stewart, L., Ellison, A., Walsh, V., & Cowey A. (2001). Zihl, J. (1989). Cerebral disturbances of elementary visual
Transcranial magnetic stimulation in studies of vision, functions. In J. W. Brown (Ed.), Neuropsychology of the
attention and language. Acta Psychologica, 107, 275 – 291. visual perception (pp. 35 – 58). Hillsdale, NJ: Erlbaum.

Lavidor and Walsh 363