Visual word recognition: the first half second
Kristen Pammer,aPeter C. Hansen,bMorten L. Kringelbach,bIan Holliday,cGareth Barnes,c
Arjan Hillebrand,cKrish D. Singh,cand Piers L. Cornelissena,*
aDivision of Psychology, School of Biology, University of Newcastle, UK
bUniversity Laboratory of Physiology, University of Oxford, UK
cThe Wellcome Trust Laboratory for MEG Studies, Neurosciences Research Institute, Aston University, UK
Received 22 January 2004; revised 3 April 2004; accepted 3 May 2004
We used magnetoencephalography (MEG) to map the spatiotemporal
evolution of cortical activity for visual word recognition. We show that
for five-letter words, activity in the left hemisphere (LH) fusiform gyrus
expands systematically in both the posterior–anterior and medial–
lateral directions over the course of the first 500 ms after stimulus
presentation. Contrary to what would be expected from cognitive
models and hemodynamic studies, the component of this activity that
spatially coincides with the visual word form area (VWFA) is not active
until around 200 ms post-stimulus, and critically, this activity is
preceded by and co-active with activity in parts of the inferior frontal
gyrus (IFG, BA44/6). The spread of activity in the VWFA for words
does not appear in isolation but is co-active in parallel with spread of
activity in anterior middle temporal gyrus (aMTG, BA 21 and 38),
posterior middle temporal gyrus (pMTG, BA37/39), and IFG.
D 2004 Elsevier Inc. All rights reserved.
Keywords: Reading; Visual word form; Visual analyzer; Inferior frontal
In many cognitive models of reading, the first stage of printed
word processing involves the operation of a ‘‘visual analysis
system’’ (Coltheart, 1981; Ellis, 2004). This converts the symbols
on a page to abstract letter representations that are invariant for
font-type, font-size, and retinal position. In addition, the visual
analysis system extracts information about where letters are posi-
tioned with respect to each other in the string. The task of
identifying letter strings as familiar words is the responsibility of
the ‘‘visual word form’’ processor. This is said to be a mental
word-store that contains representations of the written forms of all
familiar words, and is a ‘‘... stage in the reading process prior to
phonological or semantic analysis’’ (Warrington and Shallice,
1980). Recently, Cohen et al. (2002) have suggested that visual
word form representations ‘‘... are subtended by a restricted patch
of left-hemispheric fusiform cortex [average Talairach coordinates:
x = ?43, y = ?54, z = 12], which is reproducibly activated during
reading’’ (p. 1054). Accordingly, Cohen et al. showed that the
visual word form area (VWFA) responds more strongly to alpha-
betic letter strings than checkerboard stimuli, more strongly to
words than consonant strings and demonstrates invariance with
respect to retinal position. In addition, VWFA shows font-type
invariance (Dehaene et al., 2002).
Others contest the claim that the VWFA is uniquely involved in
representing visual word forms. In a recent critical review, Price
and Devlin (2003) point out that the same area is engaged: when
subjects make manual ‘‘twist’’ or ‘‘pour’’ actions in response to
pictures of familiar objects relative to perceptual judgments on the
same stimuli; when they hear, repeat or think about the meaning of
auditory words; and when congenitally blind subjects read tactile
words with abstract meanings in Braille. None of these acts, so
runs the counter claim, requires access to a visual word form.
We shed light on this controversy by applying synthetic
aperture magnetometry (SAM) to magnetoencephalography
(MEG) data to map the spatiotemporal evolution of cortical activity
during performance in a visual lexical decision task.
Subjects and tasks
Ten adult right-handed skilled readers (six males, four females;
mean age: 34 years, 4 months [range, 28–48 years] with no
recorded history of dyslexia) were required to indicate whether a
presented letter-string was a recognizable word, or an anagram of
one of the words from the test battery. Anagrams were produced by
switching the internal letter position of five-letter words in a
counter-balanced fashion: 1/3 of the anagrams contained second
and third letter position swaps (e.g., HOUSE to HUOSE), 1/3
contained third and fourth letter position swaps (e.g., HOUSE to
HOSUE), and 1/3 contained second and fourth letter position
swaps (e.g., HOUSE to HSUOE) (Cornelissen et al., 1998). The
mean Kucera–Francis frequency of the words was 168.2 (SD =
240.8, range = 42–1815). Systematic bigram frequency differences
1053-8119/$ - see front matter D 2004 Elsevier Inc. All rights reserved.
* Corresponding author. School of Biology, Henry Wellcome Building
for Neuroecology, Framlington Place, Newcastle-upon-Tyne NE2 4HH,
UK. Fax: +44-191-222-5622.
E-mail address: email@example.com (P.L. Cornelissen).
Available online on ScienceDirect (www.sciencedirect.com.)
NeuroImage 22 (2004) 1819–1825
between the three classes of anagram were sought by extracting all
the position-dependent token frequencies of bigrams from the
CELEX psycholinguistic database (Centre for Lexical Information,
Nijmegen, the Netherlands). We calculated a position-sensitive
bigram frequency score for each anagram class and then compared
scores across the three classes of anagram. A one-factor ANOVA
of token frequency determined no statistically significant diffe-
rence between the three bigram groups F(2,213) = 0.65, P > 0.05.
This suggests that there was little information, other than the
location of the position swap, which distinguished the three types
During the task, a fixation cross was presented for 500 ms. This
was replaced by the stimulus letter-string for 100 ms, which in turn
was masked for 100 ms. Subjects responded by a button press
whether they saw a word or not. Their responses were delayed by
1400 ms and were prompted by a briefly flashed spot on the
screen. Behaviorally, mean percentage correct responses for words
and anagrams were 97% (SD, 2.7%) and 83% (SD, 9.4%),
MEG data were collected using a 151-channel CTF Omega
system (CTF Systems Inc., Port Coquitlam, Canada) at Aston
University. Data were sampled at 625 Hz with an antialiasing cut-
off filter of 200 Hz. Subjects viewed the stimuli on a computer
monitor directly, such that word stimuli subtended a visual angle
of approximately 4 ? 1j. All subjects were also scanned with
MRI to get a high resolution T1 volume with typically at least 1 ?
1 ? 1 mm voxel dimensions. Immediately after finishing data
acquisition, a 3-D digitizer (Polhemus Isotrak) was used to
digitize the shape of the subject’s head in the MEG laboratory
and the relative position of the headcoils for the nasion, left and
right ear on the headset, which is then matched to the subject’s
The MEG data were analyzed using synthetic aperture magne-
tometry (SAM), which is an adaptive beam-forming technique for
the analysis of EEG and MEG data (Robinson and Vrba, 1999; Van
Veen et al., 1997; Vrba and Robinson, 2001). SAM has been
previously used in a variety of studies on the functions of the motor
cortex (Taniguchi et al., 2000), the human somatosensory cortex
(Hirata et al., 2002), swallowing (Dziewas et al., 2003), Stroop task
(Ukai et al., 2002), and midline theta rhythms (Ishii et al., 1999). In
addition, SAM has been shown to be able to unveil changes in
cortical synchronization that are spatially coincident with the
hemodynamic response found with functional magnetic resonance
imaging (Singh et al., 2002). This has also been shown to hold true
for combining SAM statistics across individuals (Singh et al.,
2003). Related techniques, such as distributed imaging of coherent
sources (Gross et al., 2001) and source localization using minimum
current estimates (Jensen and Vanni, 2002), have also been used to
study inter-regional coherences within specific frequency bands.
The statistical difference maps that are generated for the whole
brain for an individual are based on the covariance of the data
gathered from this individual, and can thus image changes in
spectral power such as event-related synchronization (ERS) and
event-related desynchronization (ERD) that are not necessarily
phase-locked to a stimulus. There is some debate regarding the
functional meaning of ERS versus ERD, but it has been demon-
strated that ERD is a correlate of increased neural activation
(Pfurtscheller and Lopes da Silva, 1999). Until proven otherwise
from, for example, simultaneous recordings of MEG and local field
potentials in experimental animals, and, in the light of existing
data, in this paper, we assume that synchronization and desynch-
ronization in SAM are equally meaningful correlates of neural
Furthermore, using the appropriate anatomical information
from an individual enables the statistical maps to be transformed
to a standard MNI space and used to make group statistical
inferences. The main limitation of adaptive beam-former techni-
ques is dealing with sources that are perfectly temporally correlat-
ed. Perfect synchrony between two sources in the brain over the
entire course of the experiment is very unlikely, and it has been
shown that the two sources can be resolved even at relatively large
temporal correlation levels (Sekihara et al., 2002; Van Veen et al.,
The SAM analysis links each voxel in the brain to the detection
array using an optimal spatial filter for that particular voxel
(Robinson and Vrba, 1999). The data from the MEG sensors is
then projected through this spatial filter to give a weighted
measure of current density, as a function of time, in the target
voxel, which means that the time series for each voxel has the
same millisecond time resolution as the original MEG signals.
Fourier analysis was used to calculate the total amount of power in
each frequency band within each of the active and passive time
epochs of the time series. The jack-knife statistical method is used
to calculate the difference between the spectral power estimates for
the active and passive states over all epochs to produce a true
t statistic. A three-dimensional image of differential cortical
activity is produced by repeating this procedure for each voxel
in the whole brain.
In this experiment, the SAM analysis created a volume for
covering the whole brain in each individual with a voxel size of 5 ?
5 ? 5 mm. The passive state was defined at the time period
between ?700 and ?500 ms before stimulus onset, and the active
state was defined as a moving 200 ms window starting at ?150 ms
before stimulus onset to 300 ms after. Power changes between the
active and passive states were calculated in the frequency band of
10–20 Hz, which has previously been shown to produce changes
in cortical synchronization that are spatially coincident with the
hemodynamic response found with functional magnetic resonance
imaging (Singh et al., 2002). Furthermore, in the data analysis, we
took care to eliminate eyeblink artefacts.
Group statistical maps were generated by first normalizing the
SAM functional volumes to standard MNI space (Collins et al.,
1994) and then combining these volumes across subjects for each
time window and frequency band. The normalization parameters
were obtained using FMRIB’s Linear Image Registration Tool
(FLIRT; Jenkinson and Smith, 2001) to reslice each individual’s
anatomical MRI to the same orientation and position as the SAM
functional volume and finding the transformation matrix from this
functional space into the standard MNI space. This transformation
matrix was then applied to each of the functional SAM volumes, in
each time window and frequency band, and for each subject. A
simplified mixed-effects model was used to generate group statis-
tical maps by combining volumes across individuals for each
contrast by calculating the sum of individual statistical values
divided by the square root of the number of subjects over each
voxel in the standard brain. These group statistical maps were then
K. Pammer et al. / NeuroImage 22 (2004) 1819–1825
thresholded at t > 2.3, and superimposed on the MNI template
brain with the cerebellum removed.
The analysis revealed that the most salient activity in our
dataset was to be found in the 10–20 Hz frequency band.
Accordingly, Fig. 1 shows a montage of the significant cortical
activation in this frequency band for word and anagram presenta-
tions in four time windows. Event-related desynchronization
(ERD) is represented in blue and event-related synchronization
(ERS) in red. Note that because subjects’ button responses were
delayed by approximately 1.5 s, we did not find any differences in
average reaction times to words versus anagrams.
The early cortical responses to words can be divided into at
least two phases. In the time window between 0 and 200 ms, ERS
activity is present in the lingual gyrus, cuneus, and also predom-
inantly left hemisphere (LH) posterior fusiform gyrus (BA18/19)
[X, Y, Z: ?14, ?88, ?6, and X, Y, Z: 30, ?94, ?6]. At 100–300
ms, an ERD appears in a more anterior part of fusiform gyrus [X, Y,
Z: ?32, ?64, ?6] close to the VWFA as defined by Cohen et al.
(2002). This activity is temporally coincident with activity in the
posterior superior IFG (BA44/6) [X, Y, Z: ?60, 8, 22], which then
spreads inferiorly. In the same time window from 100 to 300 ms,
the presentation of anagrams elicited activation in the IFG but not
in the VWFA. Activity related to anagrams did appear in the
VWFA region, but this appeared later in the 150–350 ms window
(see Fig. 2) and was significantly time delayed with respect to
responses to words. Moreover, the activity in the IFG appears
earlier for anagrams, in the 75–275 ms window, compared to
words. In both cases, the activity in the fusiform gyrus expands
systematically in both the posterior–anterior and medial–lateral
directions over the course of the first 500 ms after stimulus
The third row of Fig. 2 is a direct comparison between word
and anagram responses for each of the 13 time points. Critically,
this analysis shows that there is no difference between the
activation for words and anagrams in the posterior fusiform (BA
18/19) during the 0–200 ms time window. However, this analysis
also reveals a greater degree of synchrony for words compared to
anagrams in more anterior fusiform territory en route to the VWFA
for the same time window.
Later stages of word processing in the 200–400 and 300–
500 ms time windows included co-activation in the anterior middle
temporal gyrus (aMTG, BA 21 and 38). Furthermore, from around
200 ms post-stimulus, we also see activity predominantly in left
posterior middle temporal gyrus (pMTG, BA37/39) that peaks
around 300–500 ms. In its early stages, the pMTG activity is
accompanied by co-activation in the angular and supramarginal
gyri (BA 39/40), and subsequently in the superior temporal
We have presented novel MEG data mapping the spatio-
temporal evolution of cortical activity for visual word recognition
in the first half second. Ideally, data like these should help us to
untangle the sequence of events in the reading network, and
Fig. 1. The brain activity elicited by presenting words and anagrams measured by MEG. At the top of the figure is the SAM group analysis of brain activity in
the 10–20 Hz band to words in four time windows (0–200, 100–300, 200–400, 300–500 ms) superimposed on a canonical brain with the cerebellum
removed. At the bottom of the figure is the brain activity elicited by anagrams in the same time windows.
K. Pammer et al. / NeuroImage 22 (2004) 1819–1825
thereby set limits on the likely functional roles for each of its
components. However, to achieve this, it is fundamentally impor-
tant to know whether the activity we see is related to feedforward
or feedback effects, or some combination of the two.
Feedback versus feedforward
Human electrophysiological studies assume that ERP compo-
nents from 50 to 150 ms post-stimulus represent measures of
signaling through a hierarchical visual organization (Sereno and
Rayner, 2003). That anatomical hierarchies exist in the visual
system is well established (VanEssen et al., 1990). Confirmation
that a sequential progression can be also defined functionally in
humans has come from combined ERP and fMRI studies (Martinex
et al., 1999). In addition, there is a developing argument that the
speed of processing and information flow through the visual
system is more rapid than has traditionally been assumed (Thorpe
et al., 1996). For example, there is evidence that the first afferent
volley reaches frontal cortex 80 ms post-stimulus and continues
through the top-down feedback loops that modulate further pro-
cessing in sensory areas (Foxe and Simpson, 2002). Findings like
these have led to models (e.g., Lamme, 2003) in which stimulus
activation of the visual system produces a rapid fast-forward sweep
followed by a slower set of recurrent interactions operating both
within an activated area and backwards to lower levels of the
system. This may mean that the later an ERP/MEG component is
in time, the more likely it is to be indicative of recurrent feedback-
driven processes rather than the first information sweep through the
system (Buchner et al., 1997).
Two phases of occipitotemporal activation
On the basis of our findings, we suggest that the early stages of
visual word processing can be decomposed into two distinct
phases. The first phase of activity, between 0 and 200 ms, has
the form of an ERS localized to lingual gyrus, cuneus, and also
predominantly left hemisphere (LH) posterior fusiform gyrus
(BA18/19). The posterior fusiform component of this response is
spatially and temporally coincident with the so-called LH Type II
response in fusiform gyrus revealed in previous MEG studies of
reading (Cornelissen et al., 2003; Salmelin et al., 1996; Tarkiainen
et al., 1999, 2002). Critically, this Type II response is in fact
posterior to the VWFA location of Cohen et al. (2002) and it has a
latency of approximately 150 ms (when modeled with equivalent
current dipoles), thereby falling into the 0–200 ms window of the
current SAM analysis. Functionally, Type II responses to written
words in the fusiform gyrus are clearly distinguished from those to
geometrical forms, faces, and other objects. But they cannot be
distinguished from either nonwords (e.g., POLMEX) or random
consonant strings; hence, they are letter sensitive but nevertheless
prelexical. This pattern of responses is consistent with one of two
possibilities. The first is the operation of the ‘‘visual analysis
system’’ in which letter-forms are explicitly encoded in posterior
LH fusiform gyrus, as suggested by many cognitive models. The
alternative is that the Type II response in posterior fusiform gyrus
reflects the activity of a general system for correlation-based
learning whose spatial organization in the cortex of a skilled reader
reflects the temporal and spatial clustering of letters with letters in
the environment (cf. Polk and Farah, 1998), that is, part of a general
Fig. 2. Temporal evolution of left hemisphere and ventral brain activity elicited by visual word and anagram presentation. The figure shows the SAM group
analysis of brain activity measured every 25 ms with MEG (in the 10–20 Hz band) and superimposed on a canonical brain with the cerebellum removed. Rows
1 and 2 show the activity for words; rows 3 and 4 for anagrams; rows 5 and 6 a direct comparison between words and anagrams.
K. Pammer et al. / NeuroImage 22 (2004) 1819–1825
system for extracting the features required for object recognition,
including those for letter identification. Either way, both accounts
lead us to predict that activity in the posterior fusiform (BA 18/19)
should be the same for words and anagrams, because both stimulus
sets contain exactly the same population of letters. The absence of a
significant difference in synchrony between words and anagrams in
BA 18/19 (shown in the third row of Fig. 2) supports this. But the
same comparison also reveals greater synchrony for words relative
to anagrams more anteriorly in fusiform gyrus, en route to the
VWFA. We speculate that this might reflect a faster initial forward
sweep through occipitotemporal cortex for words than anagrams
before the initiation of recurrent feedback activity.
The second phase of occipitotemporal activity we found is in
the region of the VWFA and has the form of an ERD. It does not
appear until the time window 100–300 ms after word presentation.
This fits the temporal profile seen in field potential recordings in
humans (Nobre et al., 1994), as well as the ERP recordings of
Cohen et al. (2000). Both show that nonwords can be distinguished
from real words at around 250 ms post stimulus, but this is almost
100 ms later in time, and is at a site anatomically more anterior
than the prelexical Type II response. However, these findings are
not consistent with other reports in the literature about when lexical
access for visually presented words first occurs. For example, the
word frequency effect represents the difference in responses to
high-frequency (HF) words that are most commonly used and low-
frequency (LF) words that occur much less often. Word frequency
effects are thought to indicate that lexical access has occurred and
have been reported for the N100 response to both reading and
lexical decision tasks (Sereno et al., 1998). Pulvermuller et al.
(2001) also reported that the magnitude of the parieto-occipital
N100 was significantly correlated with a measure of semantic
association (i.e., a score related to the differences between function
words, visual nouns, action verbs, and multimodal nouns). One
way to resolve these apparent discrepancies in the timing of lexical
access is to suggest that the early estimates (i.e., those based on the
N100) might reflect the first sweep of activity through the system.
The later estimates, including the current results, might be based on
signals reflecting recurrent feedback activity. Clearly, further
research is required to disambiguate these possibilities.
Finally, one striking result is the delay we found in VWFA
activation for anagrams relative to words. Recently, Dehaene et al.
(2003) used an unconscious masking paradigm to show that masked
words activated left extrastriate, fusiform, and precentral areas.
Furthermore, masked words reduced the amount of activation
evoked by a subsequent conscious presentation of the same word
in the territory of the VWFA. This repetition suppression effect was
independent of whether the prime and target shared the same case,
indicating that neurons in this region may be tuned for case-
that letter-string tuning in VWFA territory may reflect the frequen-
cies with which particular letter combinations are encountered inthe
real world, with familiar groupings eliciting a faster response.
Accordingly, the fact the average token bigram frequency counts
for the words in our stimulus set (mean = 18,670, SE = 50,134) was
significantly higher (t = 7.99, P < 0.0001) than that for anagrams
(mean = 394,313, SE = 17,503) might explain this delay.
Early IFG activity
The present results show that the response in the VWFA region
is temporally preceded by activity in the posterior superior IFG
(BA44/6), and that this occurs earlier for anagram stimuli (75–
275 ms window) than it does for words (100–300 ms window).
Since subjects had to respond by button press, the early IFG
activation could, in principle, be related to motor preparation or
the readiness potential (Kornhuber and Deecke, 1965). We think
this is unlikely for two reasons. First, there was an approximately
1.5-s delay before subjects were prompted to press a button.
Secondly, there is no reason to assume that motor preparation
alone should be different for button presses in response to words
versus anagrams, and yet we do see differences in latency.
Another possibility is that the IFG activity might relate to the
dynamic control of task switching (Monsell, 2003). When an
anagram is presented, the initial sweep of activity through the
system should fail to elicit a lexical response. As a result, feedback
processes might then initiate a switch from an automated mode of
letter-string processing (i.e., the usual situation for skilled readers
viewing familiar words), to a slower, more analytic processing
mode appropriate to unpacking anagrams. Perianez et al. (2004)
used MEG to map the spatiotemporal sequence of events during
task-switching in an analogue of the Wisconsin card-sorting test
(WCST). They found that IFG was active in the time period 100–
300 ms after a shift cue. However, not only was this activity
bilateral, unlike the left lateralized responses in our data, but also it
localized to a more anterior, inferior region of IFG (BA45/47).
Our preferred interpretation for the left-hemisphere IFG re-
sponse is also the most challenging. The cortical territory in and
around Broca’s Area in the inferior frontal gyrus (IFG) appears to
be associated with fine-grained, speech-gestural, phonological
recoding. This system has been found to function in silent reading
and naming (see Fiez and Petersen, 1998 for review; Pugh et al.,
1996, 1997) and is thought to be more strongly engaged by low-
frequency words and pseudowords than by high-frequency words
(Fiez and Petersen, 1998; Fiebach et al., 2002). Moreover, the
particular region we see in the current data, that is, posterior
superior IFG (BA44/6), fits remarkably well with the territory
associated with phonological processing as revealed in a recent
meta-analysis by Bookheimer (2002). We suggest that the combi-
nation of brief presentation times in combination with backward
masking may well tax the reading network. Consequently, our
lexical decision task may enhance the requirement for early
phonological processing, perhaps to facilitate grapheme–phoneme
mapping. Clearly, we need further SAM data to test whether
contextual reading evokes similar, early left-hemisphere activity
in IFG, and to exclude alternative hypotheses-like task switching.
But if this finding is repeatable, and our interpretation correct, it
would pose a strong challenge to the proposed role of the VWFA—
which is supposed to be a ‘‘stage ... prior to phonological and
In the current data, later stages of word processing in the 200–
500 ms windows included co-activation of cortical areas that have
been associated in previous hemodynamic studies of semantic
processing in the aMTG (BA 21 and 38) (Rossell et al., 2003).
In addition, from around 200 ms post-stimulus, we also see activity
predominantly in left pMTG (BA37/39) which peaks around 300–
500 ms but which in its early stages is accompanied by co-
activation in the angular and supramarginal gyri (BA 39/40) and
subsequently in the superior temporal operculum (Mummery et al.,
K. Pammer et al. / NeuroImage 22 (2004) 1819–1825
Traditionally, nonwords or pseudo-words have been used for
comparison with words. We chose to use anagrams in the lexical
decision task, as we wanted a minimalist intervention which
allowed us to disturb the structure of a letter-string so as to break
any automatic contact between graphemic, phonological, and
semantic representations as early as possible in the chain of
events that underlie visual word recognition. We know from
behavioral evidence that abstract letter identity, independent of
font type and case, represents the basic perceptual unit of visual
word recognition (Besner and McCann, 1987; Grainger and
Jacobs, 1996; Pelli et al., 2003). Thereafter, in order that letter
identities can be mapped onto whole-word representations in
memory, evidence suggests that we also compute letter position
(Humphreys et al., 1990; Mason, 1981, 1982; Peressotti and
Grainger, 1999). Therefore, we argue that anagrams of words,
in which internal letter positions are swapped, provide us with
such a minimalist tool. Only changes in letter position, rather than
letter identity, determine whether or not the subject immediately
perceives a word.
Overall, the pattern of responses to briefly presented, masked
words suggests that current models of visual word recognition may
need constraining. First, our data show activity in parts of IFG
temporally preceding or at the same time as activity in the VWFA.
While we cannot draw firm conclusions from the current data, we
speculate that this may be due to phonological processing rather
than task-switching, for example. If so, this would question the
idea of a functional role of VWFA in word processing solely in the
visual domain, as originally proposed by Warrington and Shallice
(1980). Secondly, the spatiotemporal pattern of widespread activa-
tion in the fusiform gyrus over the course of 500 ms after stimulus
presentation suggests that there could be multiple foci for cortical
integration of, for example, either multimodal information and/or
the influence of top-down processing, but this awaits further
investigation. Thirdly, the widely distributed pattern of responses
between 200 and 500 ms fits better with parallel distributed models
of reading (Plaut et al., 1996) than it does with the (implicitly)
hierarchical structures described by many cognitive models. The
key idea here is that words need not be explicitly represented in a
discrete system of localized units, and access to them does not
depend on an orderly sequence of transformations of the visual
input into the spoken output. Instead, words can be represented in a
parallel distributed way, determined by the weightings of
the connections between, for example, visual, phonological, and
semantic representations. If so, the neurons in the VWFA may be
more tuned to the fast temporal processing of words and not to
anagrams, which provide a fast route for reading. This encoding
appears, however, to rely on co-activations with other brain regions
such as the IFG and thus the implication of the results presented in
this paper is that current models of visual word recognition may
need revision concerning the functional role of the VWFA.
This research was funded by the Wellcome Trust.
Besner, D., McCann, R., 1987. Word frequency and pattern distortion in
visual word identification and production: an examination of
four classes of models. In: Coltheart, M. (Ed.), Attention and Per-
formance XII: The Psychology of Reading. Erlbaum, Hillsdale, NJ,
Bookheimer, S., 2002. Functional MRI of language: new approaches to
understanding the cortical organization of semantic processing.
Annu. Rev. Neurosci. 25, 151–188.
Buchner, H., et al., 1997. Fast visual evoked potential input into human
area V5. NeuroReport 8, 2419–2422.
Cohen, L., Dehaene, S., Naccache, L., Lehericy, S., Dehaene-Lambertz,
G., Henaff, M.A., Micel, F., 2000. The visual word form area:
spatial and temporal characterization of an initial stage of reading
in normal subjects and posterior split-brain patients. Brain 123,
Cohen, L., Lehericy, S., Chochon, F., Lemer, C., Rivaud, S., Dehaene, S.,
2002. Language-specific tuning of visual cortex? Functional properties
of the visual word form area. Brain 125, 1054–1069.
Collins, D., Neelin, P., Peters, T., Evans, A.C., 1994. Automatic 3D inter-
subject registration of MR volumetric data in standardized Talairach
space. J. Comput. Assist. Tomogr. 18, 192–205.
Coltheart, M., 1981. Disorders of reading and their implications for models
of normal reading. Visible Lang. 15, 245–286.
Cornelissen, P.L., Hansen, P.C., Gilchrist, I., Cormack, F., Essex, J., Fran-
kish, C., 1998. Coherent motion detection and letter position encoding.
Vision Res. 38, 2181–2191.
Cornelissen, P., Tarkiainen, A., Helenius, P., Salmelin, R., 2003. Cortical
effects of shifting letter position in letter strings of varying length.
J. Cogn. Neurosci. 15, 731–746.
Dehaene, S., Le Clec, H.G., Poline, J.B., Le Bihan, D., Cohen, L., 2002.
The visual word form area: a prelexical representation of visual words
in the fusiform gyrus. NeuroReport 13, 321–325.
Dehaene, S., Naccache, L., Cohen, L., Le Bihan, D., Mangin, J.-F., Poline,
J.-F., Rivie `re, D., 2003. Cerebral mechanisms of word masking and
unconscious repetition priming. Nat. Neurosci. 4, 752–758.
Dziewas, R., Soros, P., Ishii, R., Chau, W., Henningsen, H., Ringelstein,
E.B., Knecht, S., Pantev, C., 2003. Neuroimaging evidence for cortical
involvement in the preparation and in the act of swallowing. Neuro-
Image 20, 135–144.
Ellis, A.W., 2004. Length, formats, neighbours, hemispheres, and the pro-
cessing of words presented laterally or at fixation. Brain Lang. 88,
Fiebach, C.J., Friederici, A.D., Mueller, K., von Cramon, D.Y., 2002. fMRI
evidence for dual routes to the mental lexicon in visual word recogni-
tion. J. Cogn. Neurosci. 14, 11–23.
Fiez, J.A., Petersen, S.E., 1998. Neuroimaging studies of word reading.
Proc. Natl. Acad. Sci. U. S. A. 95, 914–921.
Foxe, J.J., Simpson, G.V., 2002. Flow of activation from V1 to frontal
cortex in humans: a framework for defining ’early’ visual processing.
Exp. Brain Res. 142, 139–150.
Grainger, J., Jacobs, A.M., 1996. Orthographic processing in visual word
recognition: a multiple read-out model. Psychol. Rev. 103, 518–565.
Gross, J., Kujala, J., Hamalainen, M., Timmermann, L., Schnitzler, A.,
Salmelin, R., 2001. Dynamic imaging of coherent sources: studying
neural interactions in the human brain. Proc. Natl. Acad. Sci. U. S. A.
Hirata, M., Kato, A., Taniguchi, M., Ninomiya, H., Cheyne, D., Robinson,
S.E., Maruno, M., Kumura, E., Ishii, R., Hirabuki, N., et al., 2002.
Frequency-dependent spatial distribution of human somatosensory
evoked neuromagnetic fields. Neurosci. Lett. 318, 73–76.
Humphreys, G.W., Evett, L.J., Quinlan, P.T., 1990. Orthographic process-
ing in visual word identification. Cogn. Psychol. 22, 517–560.
Ishii, R., Shinosaki, K., Ukai, S., Inouye, T., Ishihara, T., Yoshimine, T.,
Hirabuki, N., Asada, H., Kihara, T., Robinson, S.E., Takeda, M., 1999.
Medial prefrontal cortex generates frontal midline theta rhythm. Neuro-
report 10, 675–679.
Jenkinson, M., Smith, S., 2001. A global optimisation method for robust
affine registration of brain images. Med. Image Anal. 5, 143–156.
Jensen, O., Vanni, S., 2002. A new method to identify multiple sources of
K. Pammer et al. / NeuroImage 22 (2004) 1819–1825
Kornhuber, H.H., Deecke, L., 1965. Hirnpotentiala ¨nderungen bei Willku ¨r-
bewegungen und passiven Bewegungen des Menschen: Bereitschaft-
spotential und reafferente Potentiale. Pflu ˆgers Arch. 284, 1–17.
Lamme, V.A.F., 2003. Why visual attention and awareness are different.
Trends Cogn. Sci. 7, 12–18.
Martinex, A., et al., 1999. Involvement of striate and extra-striate visual
areas in spatial attention. Nat. Neurosci. 2, 364–369.
Mason, M., 1981. Recognition time for letters and non-letters as a function
of retinal location and array length. Bull. Psychon. Soc. 18, 75.
Mason, M., 1982. Recognition time for letters and nonletters: effects of
serial position, array size, and processing order. J. Exp. Psychol. Hum.
Percept. Perform 8, 724–738.
Monsell, S., 2003. Task switching. Trends Cogn. Sci. 7, 134–140.
Mummery, C.J., Shallice, T., Price, C.J., 1999. Dual-process model in
semantic priming: a functional imaging perspective. NeuroImage 9,
Nobre, A.C., Allison, T., McCarthy, G., 1994. Word recognition in the
human inferior temporal lobe. Nature 372, 260–263.
Pelli, D.G., Farell, B., Moore, D.C., 2003. The remarkable inefficiency of
word recognition. Nature 423, 752–756.
Peressotti, F., Grainger, J., 1999. The role of letter identity and letter
position in orthographic priming. Percept. Psychophys. 61, 691–706.
Perianez, J.A., Maestu, F., Barcelo, F., Fernandez, A., Amo, C., Alonso,
T.O., 2004. Spatiotemporal brain dynamics during preparatory set
shifting: MEG evidence. NeuroImage 21, 687–695.
Pfurtscheller, G., Lopes da Silva, F.H., 1999. Event-related EEG/MEG
synchronization and desynchronization: basic principles. Clin. Neuro-
physiol. 110, 1842–1857.
Plaut, D.C., McClelland, J.L., Seidenberg, M.S., Patterson, K., 1996.
Understanding normal and impaired word reading: computational prin-
ciples in quasi-regular domains. Psychol. Rev. 103, 56–115.
Polk, T.A., Farah, M.J., 1998. The neural development and organization of
letter recognition: evidence from functional neuroimaging, computation-
al modeling, and behavioral studies. Proc. Natl. Acad. Sci. U. S. A. 95,
Price, C.J., Devlin, J.T., 2003. The myth of the visual word form area.
NeuroImage 19, 473–481.
Pugh, K.R., Shaywitz, B.A., Shaywitz, S.E., Constable, R.T., Skudlarski,
P., Fulbright, R.K., et al., 1996. Cerebral organization of component
processes in reading. Brain 119, 1221–1238.
Pugh, K.R., Shaywitz, B.A., Shaywitz, S.E., Shankweiler, D.P., Katz, L.,
Fletcher, J.M., Skudlarski, P., Fulbright, R.K., Constable, R.T., Bronen,
R.A., Lacadie, C., Gore, J.C., et al., 1997. Predicting reading perfor-
mance from neuroimaging profiles: the cerebral basis of phonological
effects in printed word identification. J. Exp. Psychol. Hum. Percept.
Perform. 2, 1–20.
Pulvermuller, F., Assadollahi, R., Elbert, T., 2001. Neuromagnetic evi-
dence for early semantic access in word recognition. Eur. J. Neurosci.
Robinson, S.E., Vrba, J., 1999. Functional neuroimaging by synthetic aper-
ture magnetometry (SAM). In: Yoshimoto, T., Kotani, M., Kuriki, S.,
Karibe, H., Nakasato, N. (Eds.), Recent Advances in Biomagnetism.
Tohoku Univ. Press, Sendai, pp. 302–305.
Rossell, S.L., Price, C.J., Nobre, A.C., 2003. The anatomy and time course
of semantic priming investigated by fMRI and ERPs. Neuropsychologia
Salmelin, R., Service, E., Kiesila, P., Uutela, K., Salonen, O., 1996.
Impaired visual word processing in dyslexia revealed with magneto-
encephalography. Ann. Neurol. 40, 157–162.
Sekihara, K., Nagarajan, S.S., Poeppel, D., Marantz, A., 2002. Performance
of an MEG adaptive-beamformer technique in the presence of correlat-
ed neural activities: effects on signal intensity and time-course esti-
mates. IEEE Trans. Biomed. Eng. 49, 1534–1546.
Sereno, S.C., Rayner, K., 2003. Measuring word recognition in reading: eye
movements and event-related potentials. Trends Cogn. Sci. 7, 489–493.
Sereno, S.C., Rayner, K., Posner, M.I., 1998. Establishing a time-line of
word recognition: evidence from eye movements and event-related
potentials. Cogn. NeuroReport 9, 2195–2200.
Singh, K.D., Barnes, G.R., Hillebrand, A., Forde, E.M., Williams, A.L.,
2002. Task-related changes in cortical synchronization are spatially
coincident with the hemodynamic response. NeuroImage 16, 103–114.
Singh, K.D., Barnes, G.R., Hillebrand, A., 2003. Group imaging of task-
related changes in cortical synchronisation using nonparametric permu-
tation testing. NeuroImage 19, 1589–1601.
Taniguchi, M., Kato, A., Fujita, N., Hirata, M., Tanaka, H., Kihara, T.,
Ninomiya, H., Hirabuki, N., Nakamura, H., Robinson, S.E., et al.,
2000. Movement-related desynchronization of the cerebral cortex
studied with spatially filtered magnetoencephalography. NeuroImage
Tarkiainen, A., Helenius, P., Hansen, P.C., Cornelissen, P.L., Salmelin, R.,
1999. Dynamics of letter string perception in the human occipito-
temporal cortex. Brain 122 (Pt. 11), 2119–2132.
Tarkiainen, A., Cornelissen, P.L., Salmelin, R., 2002. Dynamics of visual
feature analysis and object-level processing in face versus letter-string
perception. Brain 125, 1125–1136.
Thorpe, S., et al., 1996. Speed of processing in the human visual system.
Nature 381, 520–522.
Ukai, S., Shinosaki, K., Ishii, R., Ogawa, A., Mizuno-Matsumoto, Y.,
Inouye, T., Hirabuki, N., Yoshimine, T., Robinson, S.E., Takeda, M.,
2002. Parallel distributed processing neuroimaging in the Stroop task
using spatially filtered magnetoencephalography analysis. Neurosci.
Lett. 334, 9–12.
VanEssen, D.C., Felleman, D.J., DeYoe, E.A., Olavarria, J., Knierim, J.,
1990. Modular and hierarchical organization of extrastriate visual cortex
in the macaque monkey. Cold Spring Harbor Symp. Quant. Biol. 55,
Van Veen, B.D., van Drongelen, W., Yuchtman, M., Suzuki, A., 1997.
Localization of brain electrical activity via linearly constrained
minimum variance spatial filtering. IEEE Trans. Biomed. Eng. 44,
Vrba, J., Robinson, S.E., 2001. Signal processing in magnetoencephalo-
graphy. Methods 25, 249–271.
Warrington, E.K., Shallice, T., 1980. Word-form dyslexia. Brain 103,
K. Pammer et al. / NeuroImage 22 (2004) 1819–1825