Visual Word Recognition in the Left and
Right Hemispheres: Anatomical and
Functional Correlates of Peripheral Alexias

Introduction
The acquisition of literacy over years of education rests on the
development of elaborate links between vision and language.
On the one hand, the visual system learns how to identify rapidly
letters and letter clusters across a wide range of shapes and
viewing conditions. On the other hand, the verbal system learns
how to segment explicitly the speech stream into discrete units
such as syllables and phonemes (Morais and Kolinsky, 1994).
Associations are established between letter strings and sounds,
and between letter strings and lexical entries in memory,
allowing the sounding out of newly encountered strings, and
access to the stored knowledge attached to familiar words.
The efficiency of the adult reading system rests on its ability to
identify rapidly and in parallel arrays of several letters. Indeed,
word reading latencies are fairly constant irrespective of word
length, at least within a range of ∼3–6 letters (Weekes, 1997;
Lavidor and Ellis, 2002). This perceptual ability takes ∼5 years of
instruction to develop, and an effect of word length persists at
least to the age of 10 years (Aghababian and Nazir, 2000).
Conversely, the breakdown of the expert visual processing of

1Institut de Neurologie, Hôpital de la Salpêtrière, AP-HP, Paris,

France, 2INSERM U562, Service Hospitalier Frédéric Joliot,
CEA/DSV, Orsay, France, 3IFR 49 (Institut d’Imagerie
Neurofonctionnelle), France, 4Service de Neuroradiologie
Fischgold, Hôpital de la Salpêtrière, AP-HP, Paris, France,
5Service des Urgences Cérébro-Vasculaires, Hôpital de la
Salpêtrière, AP-HP, Paris, France and 6Centre d’Anatomie
Cognitive, Hôpital de la Salpêtrière, AP-HP, Paris, France

alphabetic stimuli following brain lesions in literate adults
results in impairments diversely known as pure alexia, alexiawithout-agraphia, word blindness or agnosic alexia. Such
patients typically have entirely preserved production and
comprehension of oral language, and they can write normally
either spontaneously or to dictation. However, they show
various degrees of impairment of word reading. In the most
severe cases, known as global alexia, they cannot identify single
letters, let alone whole words (Dejerine, 1892). Other patients
show preserved letter identification abilities and develop letterby-letter reading strategies, as if only the most finely tuned
mechanisms of word perception were affected, those allowing
for rapid and parallel identification of letter strings.
Beyond the distinction between global and letter-by-letter
alexia, cases of pure alexia may vary along several dimensions. Is
the visual field entirely affected or is alexia restricted to one
hemifield (Dejerine, 1892; Cohen and Dehaene, 2000)? Are
there residual reading abilities in purely implicit reading tasks
such as lexical decision, beyond the apparent inability to identify words (Coslett and Saffran, 1989; Coslett et al., 1993)? Do
patients have access to abstract letter identities, allowing them
to decide that ‘a’ and ‘A’ are instances of the same grapheme
(Miozzo and Caramazza, 1998; Mycroft et al., 2002), etc.? Importantly, those diverse behavioural patterns all result from lesions
confined to the territory of the left posterior cerebral artery. The
crucial brain structures include the ventral occipito-temporal
cortex, the underlying white matter and particularly the posterior
interhemispheric callosal connections (Binder and Mohr, 1992).
Accounting for the diversity of behavioural patterns and for
their anatomical and functional underpinnings is a benchmark
for any model of word reading. The aim of this study is to
confront a simple anatomical and functional model of reading
with multimodal data from patients with various deficits
affecting the visual stage of word processing (i.e. peripheral
dyslexias). It should be clear that our model is relevant only to
the visual stages of reading. It basically stops where so-called
central processes start. Central processes comprise graphemeto-phoneme mapping, lexical and semantic access, and their
impairments correspond to central dyslexias such as surface and
phonological dyslexias. In agreement with the scope of the
model, the present study is restricted to varieties of peripheral
dyslexias, and mainly to varieties of pure alexia.
We will first outline this model of visual word perception, and
consider how it may account for a variety of reading impairments (Fig. 1).
Early Retinotopic Processing
The first stage of visual word processing is the analysis of letters
in retinotopic areas, each half of foveal words being processed
Cerebral Cortex December 2003;13:1313–1333; DOI: 10.1093/cercor/bhg079

Downloaded from https://academic.oup.com/cercor/article/13/12/1313/384468 by guest on 11 April 2024

According to a simple anatomical and functional model of word
reading, letters displayed in one hemifield are first analysed through
a cascade of contralateral retinotopic areas, which compute increasingly abstract representations. Eventually, an invariant representation of letter identities is created in the visual word form area
(VWFA), reproducibly located within the left occipito-temporal
sulcus. The VWFA then projects to structures involved in phonological or lexico-semantic processing. This model yields detailed
predictions on the reading impairments that may follow left occipitotemporal lesions. Those predictions were confronted to behavioural,
anatomical and functional MRI data gathered in normals and in
patients suffering from left posterior cerebral artery infarcts. In
normal subjects, alphabetic stimuli activated both the VWFA and the
right-hemispheric symmetrical region (R-VWFA) relative to fixation,
but only the VWFA showed a preference for alphabetic strings over
simple chequerboards. The comparison of normalized brain lesions
with reading-induced activations showed that the critical lesion site
for the classical syndrome of pure alexia can be tightly localized to
the VWFA. Reading impairments resulting from deafferentation of an
intact VWFA from right- or left-hemispheric input were dissected
using the same methods, shedding light on the connectivity of the
VWFA. Finally, the putative role of right-hemispheric processing in
the letter-by-letter reading strategy was clarified. In a letter-by-letter
reader, the R-VWFA assumed some of the functional properties
normally specific to the VWFA. These data corroborate our initial
model of normal word perception and underline that an alternative
right-hemispheric pathway can underlie functional recovery from
alexia.

L. Cohen1,2,3, O. Martinaud1,2, C. Lemer1,2, S. Lehéricy3,4,
Y. Samson5, M. Obadia5, A. Slachevsky6 and S. Dehaene2,3

in the contralateral hemisphere. Visual input is represented in
increasingly abstract and invariant format through areas V1–V4.
The activation of such areas, located between about Talairach
coordinates (TC) y = –90 and y = –70, is modulated by physical
parameters such as word length, visual contrast (Mechelli et al.,
2000), stimulus degradation (Jernigan et al., 1998; Helenius et
al., 1999), stimulus rate and duration (Price et al., 1996a; Price
and Friston, 1997). Moreover, there are indications, mostly
behavioural, that perceptual processing differs between the
right (RH) and the left hemisphere (LH). Thus the absence of a
word length effect on reading latencies is restricted to optimal
viewing conditions, such as free foveal reading of horizontally
displayed stimuli (Lavidor et al., 2001; Lavidor and Ellis, 2002).
In such conditions, the optimal gaze position for reading is left
of word centre (Nazir, 2000), suggesting that parallel processing
is mostly a property of the right visual field (RVF). Indeed, splitfield studies show that while there is no length effect for words
displayed in the RVF, such an effect emerges whenever words
are displayed in the left visual field (LVF) (Lavidor and Ellis,
2002). Accordingly, when words extend across central fixation,
only their left part induces a length effect (Lavidor et al., 2001).
The left and right visual systems also differ in their tendency to
represent alphabetic characters in a format invariant for specific
case and font. A number of studies, using mostly priming tasks
with split-field stimuli, have shown that alphabetic strings are
encoded in a format less dependent on physical shape and case
when they are viewed in the RVF than in the LVF (Marsolek et
al., 1992; Marsolek et al., 1996; Burgund and Marsolek, 1997),

1314 Pure Alexia and the Visual Word Form Area • Cohen et al.

probably reflecting non-specific processing asymmetries in the
visual system (Marsolek, 1995; Burgund and Marsolek, 2000).
Accordingly, using a masked priming paradigm, Dehaene et al.
(2001) have evidenced case-specific physical repetition priming
in the right more than in the left extrastriate cortex. Finally,
Cohen et al. (2002) identified left occipital activations (TC –24
–78 –12) distinguishing alphabetic strings from chequerboards,
but only for RVF stimulations, a further correlate of the hemispheric asymmetry in word perception. In summary, there are
indications that the left posterior visual system is crucial for
important features of expert reading, namely the fast and
parallel computation of abstract letter identities.
Left Visual Word Form Area
This process culminates in the computation of the visual word
form, a representation of abstract letter identities invariant for
parameters such as spatial position, size, font or case. On the
basis of brain imaging evidence, we have proposed that this
representation is subtended by a definite cortical region in Brodmann’s area 37, the so-called visual word form area (VWFA).
Note that we refer to this region as the VWFA for the sake of
simplicity, although it should be clear that this denomination
rests on functional hypotheses that are still open to debate
(McCandliss et al., 2003). This region, which activates ∼200 ms
after stimulation (Tarkiainen et al., 1999), is reproducibly
located within the occipito-temporal sulcus bordering the left
fusiform gyrus laterally (for a review of activations, see Cohen
and Dehaene, 2000). Following the scheme proposed by Malach

Downloaded from https://academic.oup.com/cercor/article/13/12/1313/384468 by guest on 11 April 2024

Figure 1. Model of the visual word processing system. Letters are first analysed through a cascade of contralateral areas ranging from V1 to V4, which compute increasingly
abstract representations. Eventually, an invariant representation of letter identities is created in the VWFA, located within the left occipito-temporal sulcus. The VWFA then projects
to structures involved in phonological or lexico-semantic processing. Spatial attention (here represented by dark and light gray gradients) modulates the operation of all levels of
word processing up to the VWFA, with a right-hemispheric predominance. The right-hemispheric region symmetrical to the VWFA (R-VWFA) is thought to subtend residual letter
identification in patients with letter-by-letter reading. Letter identities are then transferred to language areas across the corpus callosum.

Afferents and Efferents of the V WFA
The transfer of visual information from lower-order retinotopic
cortices to the VWFA takes place within the LH for stimuli
displayed in the RVF. For LVF stimuli, information is conveyed
from the right to the left ventral occipito-temporal cortex
through interhemispheric fibre tracts that course in the
splenium of the corpus callosum and over the posterior horns of
the lateral ventricles (Binder and Mohr, 1992; Molko et al.,
2002). The output projections of the VWFA to systems involved
in lexical, semantic, motor or phonological processes are less
clearly defined. Following the observation of cases of alexiawith-agraphia, Dejerine (1892) suggested that the next step
following visual word processing would be the angular gyrus,
which he postulated to be the ‘visual center of letters’ (Fig. 2).
Indeed, the angular gyrus is among the regions that may be
activated during reading tasks (Price, 1997), and there are indications of a functional connectivity between the angular gyrus
and the left fusiform gyrus at coordinates matching the VWFA
(Horwitz et al., 1998). There is also a correlated activity in the
region of the VWFA and in left inferior frontal areas (Bokde et
al., 2001). A further potential output pathway is to temporal
regions anterior to the VWFA. These regions, which are difficult
to image with functional magnetic resonance imaging (fMRI)
due to magnetic susceptibility artifacts, are probably involved in
supramodal semantic processing (for a review, see Giraud and
Price, 2001; see also Kreiman et al., 2000; Lambon Ralph et al.,
2001).
Top-down Inf luences
One should remember that the operation of ventral visual
regions depends not only on the features of external stimuli, but

Figure 2. Confrontation of Dejerine’s (1892) original lesion drawings (left column) with fMRI activations in control subjects from the present study (right column). Top row: ventral
occipito-temporal lesion responsible for pure alexia, and activations by alphabetic stimuli versus chequerboards. Bottom row: left angular lesion responsible for alexia-with-agraphia
(dotted), and activations by real words versus consonant strings.

Cerebral Cortex December 2003, V 13 N 12 1315

Downloaded from https://academic.oup.com/cercor/article/13/12/1313/384468 by guest on 11 April 2024

et al. (2002), the VWFA is a subdivision of the ventral occipitotemporal cortex devoted to high-order object perception, other
subdivisions showing preferences for categories such as faces or
buildings (Kanwisher et al., 1997; Ishai et al., 1999). Thus, the
VWFA is activated by alphabetic strings relative to fixation but
also relative to complex non-alphabetic stimuli such as false
fonts, faces or geometrical patterns (e.g. Price et al., 1996b;
Puce et al., 1996; Cohen et al., 2002). The VWFA is activated by
visual but not auditory words (Dehaene et al., 2002), irrespective of their position in the visual field (Cohen and Dehaene,
2000). It represents alphabetic stimuli in an abstract graphemic
format. Using the masked priming paradigm mentioned before,
Dehaene et al. (2001) have shown a reduced activation of the
VWFA whenever a target word is primed by the subliminal presentation of the same word. Importantly, this effect prevails
whether the prime and the target are printed in the same or in
different cases.
Those functional properties of the VWFA may result from
more general features of inferotemporal (IT) neurons, already
present in monkeys, such as preference for semi-complex
shapes analogous to letters (Tanaka, 1996), invariance for
viewing conditions (Booth and Rolls, 1998; Grill-Spector et al.,
1999), or preference for foveal stimuli (Levy et al., 2001). One
may even speculate that the equivalence between upper- and
lower-case letters is afforded by the property of IT neurons to
establish links between stimuli with arbitrary but cognitively
meaningful connections (Sakai and Miyashita, 1991). In the same
vein, the binding of letters into words might be related to the
representation of complex objects in IT cortex through the
coactivation of neurons tuned to their elementary parts
(Tsunoda et al., 2001).

Predictions
This framework generates explicit behavioural, anatomical and
functional predictions on the varieties of reading impairment
that may follow left posterior cerebral artery (PCA) infarcts. In
line with Warrington and Shallice’s hypotheses (Warrington and
Shallice, 1980), the clearest prediction is that the destruction of
the VWFA should result in pure alexia. Indeed, despite
Dejerine’s interpretation of pure alexia as a disconnection
syndrome, cases exist of pure alexia following cortical lesions
(e.g. Beversdorf et al., 1997). We describe a similar case (patient
F).
A second prediction is that a complete loss of input to or
output from the VWFA should also yield pure alexia. The
anatomical case reported by Greenblatt (1973) may be an
instance of deafferentation of an intact fusiform cortex.
Conversely, the so-called subangular alexia may correspond to
the output variant of pure alexia, due to disrupted projections to
lateral language areas during their course in the parietal white
matter (Greenblatt, 1976; Pirozzolo et al., 1981; Iragui and
Kritchevsky, 1991). Another prediction is that partial lesions to
the input pathways to the VWFA should lead to alexia restricted
to part of the visual field. Indeed, alexia restricted to the LVF
results from interhemispheric lesions disconnecting the VWFA
from RH visual regions (Cohen and Dehaene, 2000). Similarly,
the few reported cases of alexia restricted to the RVF may be
due to left intrahemispheric lesions deafferenting the VWFA but
sparing low-level visual regions (Castro-Caldas and Salgado,
1984). Here, we describe two new cases, one with left alexia
(patient D), the other with right hemianopic alexia (patient M).
In cases where the VWFA is spared, right hemianopia by itself
should not yield pure alexia. Nevertheless, lesions affecting left
occipital regions may disrupt the fast and parallel analysis of
letter strings specific to the right hemifield, and induce some
degree of length effect in word reading (Leff et al., 2001),
analogous to the moderate length effect observed in normal
subjects reading LVF words (Lavidor and Ellis, 2002). We test
this prediction by contrasting the reading speed of our patients
D and M.
Among the still unspecified features of the model, one should
mention the role of the right-hemispheric region symmetrical to
the VWFA, which we will call the R-VWFA for the sake of

1316 Pure Alexia and the Visual Word Form Area • Cohen et al.

brevity. This region is activated by visual words relative to fixation, although less strongly and less reproducibly across subjects
than the VWFA proper (Puce et al., 1996; Cohen et al., 2002). It
does not show the same preference for alphabetic stimuli over
geometric shapes as the VWFA (Cohen et al., 2002), and may
not represent letters in a case invariant format (Dehaene et al.,
2001). The fact that right occipito-temporal lesions do not
induce alexia also indicates that the R-VWFA does not play a
necessary role in the reading network. However, it is plausible
that the R-VWFA contributes to residual reading abilities in pure
alexic patients. Bartolomeo et al. (1998) studied a patient with
letter-by-letter reading due to a left PCA infarct. Following a
symmetrical right occipito-temporal infarct, she subsequently
lost most of her residual reading abilities. This observation
agrees with a classical account of letter-by-letter reading
according to which letters would be identified by the intact
right hemisphere and their identity sequentially transferred to
left-hemispheric language areas, in a kind of internal spelling
process (Binder and Mohr, 1992). There are also indications that
residual implicit reading abilities, apparent, for instance, in
lexical or semantic decision tasks, may also rest on righthemispheric word identification (Coslett and Monsul, 1994;
Coslett and Saffran, 1998). These issues are still open, but a
tentative prediction may be that, in case of lesion of the VWFA,
its right-hemispheric homologue assumes at least part of its functional role. We tested this hypothesis in patient F.
In summary, we studied the reading abilities of three patients
suffering from left PCA infarcts, and showing different patterns
of reading impairment, presumably resulting from deafferentation or disruption of the visual word form: left alexia following
interhemispheric disconnection (patient D); alexia resulting
from right hemianopia (patient M); and pure alexia with letterby-letter reading (patient F). To clarify the underlying mechanisms we gathered anatomical and functional imaging data, in
addition to behavioural observations. Complementary anatomical data from three further patients with callosal alexia, letterby-letter reading and global alexia will be included. These multimodal data were confronted to the anatomical and functional
model outlined above. The analysis of activation data concentrated on the VWFA, which is at the core of the model, and on
its symmetrical region in the RH, whose role in residual reading
we attempted to clarify. We also report a similar fMRI study of
reading in a group of control subjects.

General Methods
Subjects
Three right-handed patients suffering from an infarct in the territory of their left PCA were studied. Their medical history will be
described in detail below. Nine healthy subjects (seven females),
aged between 42 and 68 years (mean 60 years), participated in
the study as controls. They were all right-handed according to
the Edinburgh Inventory, except for one ambidextrous man
(Oldfield, 1971). Their vision was corrected to normal. All were
drug free, had no neurological or psychiatric history, and had
normal anatomical MRIs. All controls and patients gave their
written informed consent and the project was approved by the
Ethical Committee of the Hôpital de Bicêtre.
Behavioural Assessment
Acquired dyslexias resulting from a disruption of visual word
processing are characterized by a contrast between impaired

Downloaded from https://academic.oup.com/cercor/article/13/12/1313/384468 by guest on 11 April 2024

also on top-down attentional influences that may impinge on all
processing levels, from striate cortex (Chawla et al., 1999;
Somers et al., 1999) to the ventral occipito-temporal cortex.
Indeed, an increased activation of the VWFA during the perception of written words, probably triggered by frontal control,
predicts a better storage of those words in memory (Wagner et
al., 1998). Finally, as word representations are spatially organized in the visual system, special reference should be made to
spatial attention (Corbetta and Shulman, 1998; Gitelman et al.,
1999). Thus, Hillis and Caramazza (1995) suggested that the varieties of neglect dyslexia may be attributed to spatial attentional
biases acting on one or more of the successive word representations ranging from peripheral viewer-centred representations to the abstract visual word form (Fig. 1) (for a review of
supportive data see Haywood and Coltheart, 2000). There is a
right parietal predominance in the control of spatial attention:
The RH regulates the distribution of attention across the entire
extrapersonal space, while the LH’s influence is predominantly
affected to the contralateral right hemispace (Gitelman et al.,
1999; Vallar, 2001).

processing of printed stimuli and normal performance in nonvisual language tasks. Therefore patients were assessed using (i)
reading tasks, which were expected to be impaired (timed word
and letter reading, story reading); (ii) non-visual tasks involving
alphabetic stimuli, which were expected to be generally spared
(word dictation, identification of orally spelled-out words); (iii)
tests of oral language production and comprehension, which
were also expected to be spared; and (iv) two additional tasks
evaluating the access to abstract letter identities from upper- and
lower-case letters, which may be impaired in some pure alexic
patients.

Other Language Tasks
Subjects were asked to read aloud a short story of ∼15 lines, to
write to dictation a subset of 36 items from the word reading list,
and to identify another subset of 36 words orally spelled-out by
the examiner. Subjects were also asked to name 52 letters (all
upper-case and lower-case letters in random order), using the
same procedure as for word reading.
The access to abstract graphemes was assessed using two
further tasks. First, subjects were presented with 10 upper-case
and 10 lower-case letters and were asked to write down each
letter in the opposite case. Second, they were presented with 80
pairs of letters horizontally displayed on a computer screen. The
two letters in a pair were printed in different cases, and subjects
had to decide whether the two letters corresponded to the same
grapheme (e.g. a A) or not (e.g. a B), and to respond using one
of two keys. In those two tests, only letters with widely different
upper- and lower-case shapes were used.
A short evaluation of oral langage was also performed.
Subjects were asked to describe the Cookie Theft picture from
the Boston Diagnostic Aphasia Examination, to perform the
sentence comprehension subtest no. 6 from the Token Test, and
to name a set of 80 line drawings.
Brain Imaging
Procedure
Subjects were presented with four types of stimuli: a fixation
point, real words, consonant strings, and chequerboards
covering approximately the same field as letter strings. The 128
words were a subset of the list used for the behavioural assessment. Consonant strings were matched one-to-one to real words
in number of letters. The frequency distribution of consonants
was the same in consonant strings as in words. Stimuli were
presented centrally for 1700 ms, followed by a 800 ms fixation
point. Subjects were instructed to pay attention equally to all
types of stimuli and to read real words covertly.
Subjects received four fMRI sequences, each comprising 16
experimental blocks, i.e. four blocks of each condition. Each
block comprised eight trials. Blocks were presented in pseudorandom order within sequences, so as to maximize the variety of

Imaging Parameters
Each sequence consisted of 10 s of initial fixation, followed by
16 stimulation blocks of 20 s each, as described before. In each
sequence, 132 functional volumes sensitive to blood oxygen
level dependent (BOLD) contrast were acquired with a T2weighted gradient echo, echo planar imaging sequence on a 1.5
T Signa Imager [TR (relaxation time) = 2500 ms, α (flip angle) =
90°, TE (echo time) = 60 ms, field of view = 240 × 240 mm, in
plane resolution = 3.75 × 3.75 mm2]. Each volume comprised 17
axial slices of 5 mm thickness covering most of the brain. The
first four volumes were discarded to reach signal equilibrium.
High-resolution T1-weighted images [3-D fast gradient-echo
inversion recovery sequence, TI (inversion time) = 400 ms, TR =
11 ms, TE = 2 ms, α = 20°, field of view = 240 × 240 mm, slice
thickness = 1.5 mm, in-plane resolution = 0.94 × 0.94 mm2] were
also acquired for anatomical localization, as well as T2-weighted
fast spin echo and FLAIR axial slices.
Analysis of MRI data
Functional images were analysed with the Statistical Parametric
Mapping software (SPM99). To correct for motion, functional
scans were realigned using the first image as a reference. The
anatomical image was linearly transformed to TC (Talairach and
Tournoux, 1988) using the standard template of the Montreal
Neurological Institute. In patients, we checked that masking the
lesion and using non-linear normalization parameters did not
improve the eventual match with the template (Brett et al.,
2001). The functional scans were then normalized using the
same transformation. Functional images were smoothed with a
Gaussian spatial filter (5 mm FWHM). The resulting images had
cubic voxels of 3 × 3 × 3 mm3. For single-subject analyses, activation on each of the four types of trials was modelled by a
combination of the standard SPM haemodynamic function and
its temporal derivative. Only the former function was used for
statistical contrasts. Four additional variables of non-interest
modelled constant differences across the four sequences. Longterm signal variations were eliminated with a high-pass filter set
at 240 s. Low-pass filtering was achieved by convolution with
the haemodynamic response function. Data from the group of
normal subjects were submitted to a random-effect group
analysis with subjects as random variable (voxelwise threshold
P < 0.005; cluster-level P < 0.05 corrected). In analyses of individual patients we used a voxelwise threshold of P < 0.001 and
cluster-level P < 0.05 corrected, unless stated otherwise.
The patients’ lesions were drawn manually on the basis of the
anatomical MRIs using the MRIcro software (http://
www.psychology.nottingham.ac.uk/staff/cr1/mricro.html), then
smoothed (4 mm FWHM filter), thresholded (>0.5), and normalized with the same matrix as anatomical images.

Results in Normal Subjects
Behavioural Assessement
As expected, performance was essentially at ceiling on all
subtests (Table 1). In particular, subjects made no errors in the
story, word and single letter reading tasks. The mean reading

Cerebral Cortex December 2003, V 13 N 12 1317

Downloaded from https://academic.oup.com/cercor/article/13/12/1313/384468 by guest on 11 April 2024

Word Reading
Stimuli consisted of 165 lower-case familiar nouns (mean log10
frequency per million = 2.4), 3–9 letters and 1–4 syllables in
length. They were presented centrally on a computer screen,
subtending a maximum angle of 4° on each side of fixation.
Subjects were asked to read each word aloud rapidly, while
minimizing errors. Stimuli remained visible until a response was
produced. Latencies were measured using a voice key, and
sessions were recorded for subsequent scoring of errors.

transitions between conditions while avoiding any repetition of
the same condition in successive blocks. Trials were presented
in random order within blocks. The order of blocks within
sequences and the order of trials within blocks differed across
the four sequences, but were identical for all subjects.

latency for words was 534 ms, and increased by an average 6 ms
per letter (Fig. 3). Actually, latencies were exactly constant for
words of 3–6 letters, and then increased by 12 ms per additional
letter (P < 0.05).

Table 1
Performance of patients and control subjects in language tasks (% correct responses)
Patient D

Patient M

Patient F

Normal subjects (mean)

Cookie theft description

normal

normal

normal

normal

Picture naming

79/80 (99%)

79/80 (100%)

78/80 (97%)

99%

Token test (VI)

12/13 (92%)

12/13 (92%)

10/13 (77%)

87%

Timed story reading (duration)

40 s

52 s

4 min 09 s

35 s

Writing words to dictation

36/36 (100%)

36/36 (100%)

34/36 (94%)

98%

Identification of spelled out words

33/36 (92%)

33/36 (92%)

34/36 (94%)

95%

Cross-case letter matching

77/80 (96%)

79/80 (99%)

76/80 (95%)

96%

Cross-case letter conversion

20/20 (100%)

20/20 (100%)

20/20 (100%)

97%

No. correct (%)

165/165 (100%)

165/165 (100%)

158/165 (96%)

100%

Mean latency (ms)

607

849

2547

539

Free-field word reading

Free-field letter reading
No. correct (%)

52/52 (100%)

52/52 (100%)

52/52 (100%)

100%

Mean latency (ms)

663

640

817

558

Figure 3. Behavioural and fMRI correlates of normal reading. Top: normal controls (black) showed no influence of word length on reading latencies. Patient D, who had an intact
VWFA and no hemianopia showed a normal pattern (blue). Patient M, who had an intact VWFA and right hemianopia, showed a moderate length effect (green). Patients F (magenta)
and A (yellow), whose VWFA was lesioned, read words letter-by-letter, with a slope of over 300 ms per letter. Middle: cortical network activated by alphabetic stimuli relative to
fixation in normal subjects, including bilateral ventral temporal areas (voxelwise P < 0.005; corrected P < 0.05 for cluster extent). Bottom left: bilateral ventral temporal areas
activated by alphabetic stimuli relative to fixation. When contrasting alphabetic stimuli versus chequerboards, activations were restricted to the left-hemispheric VWFA. Bottom
right: left angular and posterior cingulate activations induced by real words relative to consonant strings. The opposite contrast revealed bilateral posterior parieto-occipital
activations.

1318 Pure Alexia and the Visual Word Form Area • Cohen et al.

Downloaded from https://academic.oup.com/cercor/article/13/12/1313/384468 by guest on 11 April 2024

fMRI Study
In order to identify the brain areas involved in alphabetic
processing, we first contrasted alphabetic stimuli (i.e. words and
consonant strings) relative to fixation, revealing a large bilateral
network (Fig. 3 and Table 2). In both hemispheres, the highest
peaks of the parieto-occipito-temporal cluster were ventral
occipital [left: TC –21 –90 –3; t(8) = 9.4; right: TC 48 –75 –12;
t(8) = 12.3] and fusiform [left: TC –42 –63 –15; t(8) = 8.7; right:
TC 36 –63 –18; t(8) = 13.1]. Note that the left fusiform cluster
peaking at TC –42 –63 –15, and extending anteriorily to about
y = –42, matched very closely the coordinates of the VWFA as
identified in a number of previous studies. Note also that the
activation of the symmetrical right-hemispheric region (RVWFA) was comparable to the VWFA itself, with t-values of 13.1
versus 8.7.
In order to identify activations more specific to reading, we
contrasted alphabetic stimuli versus chequerboards. This
comparison revealed a strongly left-lateralized frontal, parietal,
occipital and temporal network (Fig. 3 and Table 2). Contrary to
what prevailed when alphabetic stimuli were contrasted with
fixation, occipital and temporal activations were now restricted
to the left hemisphere. The left occipito-temporal cluster peaked
at TC –36 –75 –12 [t(8) = 10.1], extended anteriorily to y = –33,

and included the region of the VWFA (P < 0.01 at the peak voxel
TC –42 –63 –15). In the right hemisphere, there was no significant occipito-temporal activation. Note however a small right
occipital cluster that did not reach the threshold for cluster
extent [TC 45 –78 –12; t(8) = 8.4; 55 voxels]. In particular, the
R-VWFA, despite its very strong activation by alphabetic stimuli
versus fixation, did not differ between alphabetic stimuli and
chequerboards (P > 0.05 at the peak voxel TC 36 –63 –18).
We then examined the influence of lexical status on brain activations (Table 2 and Fig. 3). The contrast of words versus consonant strings activated a left angular region [TC –45 –63 27; t(8) =
5.7], and a midline cluster in the bilateral posterior cingulate gyri
[TC –3 –33 33; t(8) = 8.40]. Note that for both these regions, this
difference reflected a deactivation by consonant strings relative
to fixation, with no significant change in activation for real
words. For instance at the peak of the angular cluster, t(8) < 1 for
words versus fixation, and t(8) = 4.96 (P = 0.001) for fixation
versus consonants.
Conversely, bilateral posterior occipito-parietal regions were
activated by consonant strings relative to real words (left: TC
–33 –87 12 and TC –21 –51 63; right 27 –72 12 and TC 21 –60
63; Table 2). There was also right postcentral and right lateral
frontal clusters. Note that at a low statistical threshold, the
region of the VWFA and its right-hemispheric counterpart
appeared to be activated by consonant strings more than by real
words (voxelwise P < 0.05; P < 0.05 for cluster extent corrected
within a 12 mm sphere centred on the peak voxels as identified
with the contrast of alphabetic stimuli versus fixation).

Downloaded from https://academic.oup.com/cercor/article/13/12/1313/384468 by guest on 11 April 2024

Cerebral Cortex December 2003, V 13 N 12 1319

Table 2
Group analysis of the fMRI word reading experiment in nine normal subjects
Side

Area

Cluster

Peak voxel

P corrected

No. of voxels

Z score

TC

<0.001

1058

4.87

36 –63 –18

Alphabetic stimuli versus fixation
Right

Fusiform gyrus

Right

Ventral occipital

Left

Ventral occipital

Left

Fusiform gyrus

Left

Thalamus

4.78

48 –75 –12
–21 –90 –3

<0.001

1062

4.35
4.23

–42 –63 –15

0.007

97

4.14

–24 –18 3

Lenticular nucleus

<0.001

364

3.92

–24 0 –3

Left

Inf. Frontal sulcus

<0.001

254

3.78

–45 9 27

Right

Inf. Frontal sulcus

0.005

102

3.65

45 15 18

Left

Dorsolateral prefrontal

0.010

90

3.49

–48 39 15

Right

Lenticular nucleus

0.019

80

3.41

24 18 –6

Left

Supramarginal gyrus

<0.001

143

3.34

–51 –33 39

<0.001

340

Alphabetic stimuli versus chequerboards
Left

Ventral occipital

Left

Fusiform gyrus

4.48

–36 –75 –12

3.44

–36 –45 –27

Left

Inf. Frontal sulcus

<0.001

708

4.33

–36 12 18

Left

Cingulate sulcus

Right

Inf. Frontal sulcus

<0.001

160

3.70

–12 27 33

<0.001

213

3.45

Left

45 36 21

Intraparietal sulcus

<0.001

190

3.41

–18 –60 30

Right

Lenticular nucleus

0.042

76

3.20

24 6 –3

<0.001

761

4.17

–3 –33 33

0.043

62

3.50

–45 –63 27

<0.001

602

4.25

–33 –87 12

3.75

–21 –51 63

3.87

27 –72 12

3.52

21 –60 63

Words versus consonant strings
Bilateral

Posterior cingulate

Left

Angular gyrus

Consonant strings versus words
Left

Right

Occipito-parietal

Occipito-parietal

<0.001

747

Right

Post-central

0.050

60

3.66

48 –21 54

Right

Inferior frontal gyrus

0.006

89

3.54

51 9 12

Discussion
Relative to fixation, the processing of alphabetic stimuli activated a bilateral network, with a left-sided predominance mostly
visible in the frontal lobes. This network included activations in
the left ventral occipito-temporal cortex precisely corresponding to the coordinates of the VWFA. There was a rightsided symmetrical activation, but the functional properties of
those symmetrical regions appeared quite different: the VWFA
was activated by alphabetic stimuli relative to chequerboards,
while the R-VWFA reacted identically to both, confirming that
the VWFA has some degree of specialization for alphabetic
processing. We also observed a more posterior left occipital activation for alphabetic stimuli versus chequerboards. In our
previous study (Cohen et al., 2002) this region was associated
exclusively with the processing of RVF stimuli. In the present
study however, its properties appeared indistinguishable from

1320 Pure Alexia and the Visual Word Form Area • Cohen et al.

those of the VWFA in the present study, as stimuli were spread
over both hemifields.
The VWFA was slightly more activated by consonant strings
than by real words. This could appear as a surprize, as previous
studies have shown the opposite pattern in both blocked-design
and event-related experiments (Price et al., 1996b; Büchel et al.,
1998; Xu et al., 2001; Cohen et al., 2002). However, this
discrepancy likely results from systematic differences in the
experimental paradigms. In our previous study (Cohen et al.,
2002), stimuli were displayed very briefly (200 ms) in a lateral
position (beyond 2° of eccentricity). In such conditions, the fast
identification of familiar words was possible, but meaningless
consonant strings were impossible to scan and their components letters were difficult to identify. In contrast, the present
task setting emphasized the processing of consonant strings, by
providing a long (1700 ms) and central display. As instructions

Downloaded from https://academic.oup.com/cercor/article/13/12/1313/384468 by guest on 11 April 2024

Left

strings would subjects inhibit this basic tonic language process.
Binder et al. (1999) have observed results closely similar to ours
in the auditory modality. Relative to a resting state, the left
angular gyrus was deactivated during an auditory perceptual
task or a phonological task with non-words, while a semantic
retrieval task maintained the activation at the same level as
during rest.
To conclude with a historical point, one may note that the
present activation data converge nicely with Dejerine’s anatomical observations, in pointing to the left fusiform and angular gyri
as key structures for the visual and linguistic processing of
written words, respectively (Fig. 2).
We now turn to the study of patients suffering from an infarct
of the left PCA. Depending on their precise topography, such
lesions should affect to various degrees the left ventral occipitotemporal regions whose activations were evidenced in normal
subjects. Within this area, the VWFA is thought to play a key role
in reading, and it disruption is expected to yield pure alexia.

Left Hemialexia: Patient D
Medical History and Lesion Description
Patient D was a 63-year-old retired male computer engineer. One
year before the present study, he suffered from an infarct in the
territory of the left PCA, revealed by mild right-sided hemiparesis, a right homonymous visual scotoma and transient wordfinding difficulties.
The anatomical MRI showed three small lesions in the territory of the left PCA (Fig. 4), affecting the anterior mesial
temporal lobe, the thalamus and the white matter of the forceps
major, just contiguous to the splenium of the corpus callosum
(approximately centred on TC –18 –33 –17; TC –12 –25 0; and
TC –8 –59 16, respectively). The peak of the VWFA in normal
subjects fell in a region of intact cerebral tissue, posterior and
lateral to the temporal lesion (Fig. 4).
Behaviour
Goldmann perimetry showed a right superior homonymous
peripheral scotoma completely sparing central vision. Oral
language production and comprehension, picture naming, story
reading, writing to dictation, identification of spelled-out words,
and graphemic processing were normal (Table 1). The patient
named flawlessly and rapidly centrally presented words and
single letters. On average, word reading latencies increased by a
marginal 7 ms per additional letter [r(159) = 0.15; P = 0.052],
thus showing a normal pattern.
As his visual field was largely intact, the patient and 15 control
subjects were submitted to a split-field word reading task.
Stimuli consisted of 40 lower-case familiar words, 3–6 letters
and 1–2 syllables in length. They were presented once in the
LVF and once in the RVF, extending from 2° to a maximum of 6°
away from fixation. Subjects were asked to fixate a permanent
central fixation point, while stimuli were flashed during 200 ms.
The patient named flawlessly and rapidly RVF words (0/40
errors; mean latency = 623 ms). With LVF words, he made significantly more errors [16/40 errors; Yates corrected χ2(1) = 17.6; P
< 0.001] and responded more slowly [mean correct latency =
723 ms; t(57) = 2.8; P = 0.007]. The patient’s asymmetry in error
rates far exceeded the average 6.3% RVF reading advantage
observed in controls [Fig. 5; t(14) = 4.16; P < 0.001].
In sum, the patient considered his reading skills as unimpaired
and the experimental assessment of central reading perform-

Cerebral Cortex December 2003, V 13 N 12 1321

Downloaded from https://academic.oup.com/cercor/article/13/12/1313/384468 by guest on 11 April 2024

encouraged subjects to pay as much attention to consonant
strings as to real words, it is likely that subjects attentively
scanned the stimuli, enhancing the activation of the ventral
regions involved in visual processing. This hypothesis is
supported by the major activations elicited by consonant strings
versus words in the bilateral superior parietal lobules. Those
activations fell at coordinates virtually identical to those
observed in tasks of spatial attentional control (Gitelman et al.,
1999; Simon et al., 2002). Tagamets et al. (2000) observed a
pattern very similar to the present one: Activation of the VWFA
did not differ between consonant strings and words, while
consonant strings induced an increased parietal activation relative to words, again at coordinates close to ours. Thus, it appears
that the VWFA is activated by essentially any orthographic
string, with significant modulations that depend both on the
experimental task, and on the orthographic legality of the string.
Despite the strong attentional demands entailed by the
processing of consonant strings, we expected that some
language-related areas should still be activated more strongly by
real words. Indeed, the left angular gyrus showed such a lexicality effect. The angular cortex has long been recognized as a
core supramodal language area. Its central role in reading and
writing has been advocated by Dejerine in his seminal studies
(Dejerine, 1891, 1892). Observing that angular lesions yielded
alexia-with-agraphia, he surmised that the angular gyrus could
be the repository of ‘visual images’ of words, involved in both
reading and writing. Beyond this precise hypothesis, data from
patients and from brain imaging in normals are consistent with
the more general idea that the angular gyrus is a crucial transmodal interface between auditory or visual word forms and
stored lexico-semantic knowledge (Mesulam, 1998). Naturally,
such processes would not be relevant to the perception of
consonant strings, which have no lexical content and do not
even bear a resemblance to real words, explaining the difference
in activation levels.
We noted that the left angular gyrus was activated at similar
levels during word reading and during simple fixation, while its
activation was significantly lower during inspection of consonant strings. This observation does not undermine the reality of
the lexicality effect, but may give further clues on the functional
properties of the angular gyrus. Recent imaging studies and
meta-analyses have shown that a number of cortical areas often
have a higher level of activation during the so-called ‘rest conditions’ than during attention-demanding cognitive tasks (Binder
et al., 1999; Gusnard and Raichle, 2001; Mazoyer et al., 2001;
Raichle et al., 2001). It has been suggested that this may reflect
tonic cognitive processes occupying the ‘rest’ periods; such
processes would be interrupted or inhibited by more attentiondemanding tasks. Although the exact cognitive content of this
‘default brain state’ is still largely unknown, the set of concerned
areas seems relatively well defined. Foremost among them
appear the two areas that were activated by words versus consonant strings, i.e. the angular gyrus and the posterior cingulate
cortex. A favourite hypothesis concerning the posterior cingulate emphasizes its role in the continuous monitoring of the
peripersonal world, an activity which may be interrupted whenever focused attention is required (Vogt et al., 1992; Gusnard
and Raichle, 2001), which is the case during the processing of
consonant string. Regarding the left angular gyrus, one may
speculate that it is involved in some inner speech or conceptual
activity, ongoing both during fixation and during the passive
recognition of familiar words. Only when scanning consonant

Downloaded from https://academic.oup.com/cercor/article/13/12/1313/384468 by guest on 11 April 2024

1322 Pure Alexia and the Visual Word Form Area • Cohen et al.

ance was normal. Nevertheless, split-field reading revealed a
reduced performance with words displayed in the LVF.

fMRI Study of Lateralized Word Reading
Patient D was submitted to an event-related lateralized reading
protocol fully described in Cohen et al. (2002). In summary, he
was presented with randomly mixed words, consonant strings
and chequerboards displayed in the LVF or in the RVF. In order
to assess the comparability with the central reading experiment,
we first run global analyses ignoring the hemifield parameter.
We then study separately the activations induced by RVF and
LVF stimuli. Finally, we assess the significance of the observed
asymmetries.

Summary and Discussion
The case of patient D lends itself to a straightforward account
within the framework that was proposed before. First, reading
performance in ecological conditions was normal, i.e. rapid and
with no significant effect of word length, attesting to the integrity of the core reading pathway, leading from left V1 to the

Figure 4. Reconstruction of the lesions of six patients in Talairach space, compared with the average normal location of the VWFA (white cross). Left column: the lesions of patients
without pure alexia did not affect the critical VWFA region. Right column: the lesions of patients who showed letter-by-letter reading (patients A and F) or global alexia (patient VOL),
encroached on the VWFA, in the lateral occipito-temporal sulcus. The additional callosal lesion may be responsible for the left hemialexia in patient D, and for the lack of letter-byletter reading abilities in patient VOL. Slices are at TC z = –14, z = 17, and y = –60 (for patient RAV z = –6, z = 17, y = –66).

Cerebral Cortex December 2003, V 13 N 12 1323

Downloaded from https://academic.oup.com/cercor/article/13/12/1313/384468 by guest on 11 April 2024

fMRI Study of Central Word Reading
The patient participated in the fMRI reading protocol described
in the General Methods section. We first identified a bilateral
fronto-parieto-temporo-occipital network activated by alphabetic stimuli relative to fixation. It included large bilateral occipitotemporal activations, with bilateral main peaks in the ventral
occipital lobes (left: TC –18 –93 –12; right: TC 21 –84 –12),
extending anteriorily to about y = –40, and encompassing the
normal location of the VWFA and of the R-VWFA.
In order to isolate the VWFA, we contrasted alphabetic stimuli
versus chequerboards (voxelwise P < 10–4; cluster-level P < 0.05
corrected). This revealed left parietal, temporal and rolandic
activations, including a left fusiform peak [TC –39 –51 –18;
t(197) = 5.39; P < 10–6] at coordinates matching the normal position of the VWFA (Fig. 6). At the peak voxel, the activation was
stronger for consonant strings than for real words [t(197) = 2.25;
P < 0.05]. Note that like in normal subjects, the contrast of
words versus consonant strings yielded strong activations in the
left angular gyrus [TC –39 –69 33; t(197) = 6.65] and in the left
posterior cingulate [TC –6 –54 36; t(197) = 5.19].
The contrast of alphabetic stimuli versus chequerboards
revealed no significant activation in the right temporal region.
However, as mentioned before, the comparison of alphabetic
stimuli versus fixation showed large bilateral ventral occipitotemporal activations. In the right hemisphere these activations
comprised three peaks roughly symmetrical to the VWFA (Fig.
6) [TC 42 –54 –12; t(197) = 5.37; TC 33 –66 –18; t(197) = 8.08;
48 –45 –15; t(197) = 5.20; all P < 10–6]. In contrast with the
VWFA, the three peaks were activated by chequerboards to
similar levels as by alphabetic stimuli, relative to fixation [t(197)
= 4.95; 7.19; and 4.34 respectively; all P < 10–4], with no significant difference between alphabetic stimuli and chequerboards
(all P > 0.05).
In summary, as expected on the basis of lesion topography,
central reading elicited a normal pattern of activation. The
VWFA was preserved and responsive to alphabetic stimuli more
than to chequerboards. Conversely, the symmetrical R-VWFA
was equally activated by both types of stimuli. However, this
experiment did not allow us to study the differences between
activations induced by LVF and RVF stimuli, or to study the
mechanism of the patient’s left hemialexia.

The global contrast of alphabetic stimuli versus chequerboards showed the VWFA peaking at coordinates virtually identical to the central reading experiment [TC –39 –48 –18; t(290)
= 3.81; voxelwise P < 0.001; cluster-level P < 0.05 corrected
within a 12 mm sphere centred on the VWFA as identified in the
central reading experiment]. The activation did not differ significantly between real words and consonant strings [t(290) =
0.29]. In agreement with the central reading experiment, this
contrast showed no right temporal activations. These results
were thus closely similar to those of the central reading experiment. We then turned to separate analyses of the activations
induced by RVF and LVF stimuli, expecting to observe an
abnormal pattern in the latter case, underlying the behavioural
asymmetry.
For RVF stimuli, the VWFA was activated by alphabetic stimuli
versus chequerboards, and separately for words versus chequerboards (voxelwise P < 0.001; cluster-level P < 0.05 corrected
within a 12 mm sphere centred on the VWFA as identified in the
central reading experiment). Activation by consonant strings
was not significant (P < 0.05 at the peak voxel). The tendency
for a stronger activation by words than by consonants did not
reach significance [t(290) = 1.47 at the peak voxel]. This pattern
could be considered as normal, with a clear advantage of alphabetic stimuli over chequerboards, and a weaker advantage of
real words over consonants. Note that in normal subjects, the
lexicality effect emerged in group analyses, and was generally
not significant in individual subjects (Cohen et al., 2002).
LVF alphabetic stimuli, LVF words and LVF consonants did not
significantly activate the VWFA relative to chequerboards at the
usual threshold. At a lower voxelwise threshold, the peak voxel
was activated by consonant strings [t(290) = 2.35], but not by
words [t(290) = 1.30]. This pattern was clearly abnormal, as it
showed no activation of the VWFA by words relative to chequerboards.
Before concluding that the difference in the activations
induced in the VWFA by RVF versus LVF words was indeed
abnormal, the patient was compared with a group of nine
normal subjects, with subjects considered as a random factor
(Cohen et al., 2002). At its peak (TC –39 –48 –18), patient D’s
VWFA was more strongly activated by RVF words than by LVF
words [t(290) = 2.67; P < 0.01]. The random effect analysis
showed that this difference was significantly larger in the patient
than in normals [t(8) = 2.89; P = 0.02] (Fig. 5). In this analysis,
the VWFA appeared as a cluster peaking at TC –45 –51 –15 [t(8)
= 7.22; voxelwise P < 0.01; cluster-level P = 0.002 corrected
within a 12 mm sphere centred on the patient’s VWFA]. Thus, it
may be safely concluded that the weak activation of the VWFA
induced by LVF words as compared with RVF words was significantly abnormal in the patient.

Downloaded from https://academic.oup.com/cercor/article/13/12/1313/384468 by guest on 11 April 2024

Figure 5. Lateralized word reading in patient D. Top row: the VWFA was activated more strongly by RVF words than by LVF words (P < 0.005). Middle left: per cent correct
responses in the lateralized reading task, showing flawless performance for RVF words (green), and a high error rate for LVF words (red). Middle right: the RVF advantage in
lateralized reading was larger for the patient (red square) than in 15 controls (blue diamonds). Bottom left: plot of the percent change in BOLD signal for RVF words (green curve)
and LVF words (red curve) (TC –39 –48 –18). Bottom right: the difference between the percent signal change observed to words presented in the RVF and in the LVF was larger in
patient D (red square) than in nine controls (blue diamonds) (TC –45 –51 –15). The patient thus showed an abnormal asymmetry in both his behavioural and his activation pattern.

1324 Pure Alexia and the Visual Word Form Area • Cohen et al.

VWFA and thence to subsequent langage areas. Accordingly, the
temporal lesion was anterior and mesial to the VWFA, which
was activated during reading. Secondly, patient D’s partial
impairment in reading LVF words can be attributed to a partial
interhemispheric disconnection due to a left-sided lesion of the
splenium of the corpus callosum. We have shown previously
using diffusion tensor imaging, albeit in a patient with much
larger lesions, that the fibre tract critically affected by such
lesions connects right and left ventral occipitotemporal regions,
including probably connections from the right V4 area to the
VWFA (Molko et al., 2002). Accordingly, patient D showed an
abnormally weak activation of the VWFA by LVF versus RVF
words. According to our hypotheses, we may assume that the
activations induced by both alphabetic stimuli and chequerboards in the R-VWFA did not play a crucial role in the patient’s
reading abilities.

Right Hemianopic Alexia: Patient M
Medical History and Lesion Description
Patient M was a 54-year-old right-handed male building engineer.
Eight months before the present study, he suffered from an
infarct in the territory of the left PCA, revealed by sudden right
homonymous hemianopia.
The anatomical MRI showed a left ventral and mesial occipitotemporal lesion, destroying both banks of the calcarine sulcus
and the lingual gyrus, and extending to the mesial part of the
fusiform gyrus. The peak of the normal VWFA fell in a region of
intact cerebral tissue, lateral to the ischaemic lesion and to the
enlarged lateral ventricle (Fig. 4).
Behaviour
Goldmann perimetry showed a complete right homonymous
hemianopia with no sparing of the macula. Oral language

production and comprehension, picture naming, writing to
dictation, identification of spelled-out words and graphemic
processing were normal (Table 1). Story reading was flawless
but slower than in normals [52 s in the patient versus a mean of
35 s in normals; t(8) = 3.5; P < 0.01]. The patient named flawlessly centrally presented words and single letters. However,
word reading latencies increased by 64 ms per letter [r(162) =
0.57; P < 0.001], a slope significantly steeper than in normal
subjects [t(8) = 12; P < 0.001].
As the patient had a complete right hemianopia, we compared
this pattern to the performance of the 15 control subjects
reading words displayed in one hemifield, as reported in the
study of patient D. We computed the slope of the patient’s word
length effect restricted to words of 3–6 letters, in order to insure
comparability with the controls. The patient’s slope (53 ms per
letter) did not differ from the slope observed in controls for LVF
words [mean = 22 ms per letter; t(14) = 1.6; P > 0.1], while it
was significantly steeper than the normal slope for RVF words
[mean = 9 ms per letter; t(14) = 3.4; P = 0.005].
In sum, the word length effect exhibited by patient M when
reading stimuli in free-field could be considered as a consequence of his hemianopia, with an otherwise essentially intact
reading system.
fMRI Study of Word Reading
The patient participated in the fMRI reading protocol described
in the General Methods section. We first identified a bilateral
fronto-parieto-temporo-occipital network activated by alphabetic stimuli relative to fixation. It included bilateral occipitotemporal activations, with bilateral main peaks in the ventral
occipital lobes (left: TC –24 –81 0; right: TC 33 –75 –6), and at
the normal coordinates of the VWFA (TC –39 –57 –9). There
was a strong and precisely symmetrical right-sided activation
corresponding to the R-VWFA [TC 39 –57 –15; t(197) = 6.24].

Cerebral Cortex December 2003, V 13 N 12 1325

Downloaded from https://academic.oup.com/cercor/article/13/12/1313/384468 by guest on 11 April 2024

Figure 6. Ventral occipito-temporal activations in patients. The left column shows the contrast of alphabetic stimuli versus fixation, and the right column shows the contrast of
alphabetic stimuli versus chequerboards (P < 0.001). In the two patients without pure alexia (D and M) the VWFA was activated by letter strings relative to chequerboards while
both types of stimuli yielded an equivalent activation of the symmetrical R-VWFA. In the letter-by-letter reader (patient F) the VWFA was partially disrupted, while the R-VWFA
discriminated alphabetic stimuli versus chequerboards, thus assuming functional properties generally specific to the VWFA.

Summary and Discussion
Despite a large cortical lesion and the resulting hemianopia,
patient M had no subjective reading difficulties. Nevertheless,
the study of his reading latencies suggests that those were abnormally dependent on word length, and hence that the mechanisms that normally compute the identities of letter arrays in
parallel were disrupted by the cortical lesion. As suggested in
the introduction, those mechanisms may be specific to the lefthemispheric visual system, explaining why in normals only RVF
words are read with no appreciable effect of word length, while
there is such an effect with LVF words (Lavidor and Ellis, 2002).
Leaving aside this subtle deviation from optimal performance, it
is tempting to relate the sparing of patient M’s reading abilities
to the spared activation of the VWFA by words entering through
the intact LVF. Like in normals and in patient D, we assume that
the activations induced by both alphabetic stimuli and chequerboards in the R-VWFA did not play a crucial role in patient M’s
reading abilities. In short, we suggest that this patient read
roughly like normal subjects read words displayed in their LVF.

Letter-by-letter Reading via the Right Hemisphere: Patient F
Medical History and Lesion Description
Patient F was a 68-year-old right-handed retired male manager.
He suffered from an infarct in the territory of the left PCA,
revealed by sudden right homonymous hemianopia and reading
difficulties. Eight months after the infarct, word reading was still
slow and effortful. When reading a list of 133 words 3–9 letters
long, the patient made 22 errors (16%). Reading latencies on
correct trials, as measured with a manual chronometer, had a
mean of 9200 ms, and showed a major effect of word length,
with a slope of 1140 ms per letter. The patient received rehabilitation by a speech therapist and reading partially improved over
the following months. The present study was carried out 20
months after the stroke.
The anatomical MRI showed a left ventral occipito-temporal
lesion. It affected the inferior occipital cortex and the lingual
gyrus, with no extension dorsal to the calcarine sulcus. It
extended anteriorily to the fusiform and parahippocampal gyri.
There was an additional small right inferior occipital lesion,
probably resulting from a previous overlooked CVA. The region
of normal VWFA activation overlapped substantially with the left
fusiform lesion (Fig. 4).

1326 Pure Alexia and the Visual Word Form Area • Cohen et al.

Behavioural Assessment
Goldmann perimetry showed right superior homonymous quadrantanopia whitout foveal sparing, plus an additional right
inferior scotoma in the left eye. Oral language production and
comprehension, picture naming, writing to dictation, identification of spelled-out words and graphemic processing were essentially normal (Table 1).
Story reading, although correct, was still slow (>4 min). The
patient made few errors (7/165 errors), but responded slowly
(mean latency = 2547 ms) when reading single words. Furthermore, there was a major effect of word length on latencies, with
a slope of 311 ms per letter [r(137) = 0.42; P < 0.001]. This slope
was much steeper than in normal subjects, but also than in
patient M. In summary, patient F qualified as a typical pure
alexic with an efficient letter-by-letter reading strategy.
fMRI Study of Word Reading
The patient participated in the fMRI reading protocol described
in the General Methods section. We first identified a bilateral
fronto-parieto-temporo-occipital network activated by alphabetic stimuli relative to fixation. It included strong right ventral
occipitotemporal activations at the coordinates of the R-VWFA
[TC 39 –60 –18; t(197) = 11.7]. In the left hemisphere, there was
a smaller cluster peaking at TC –45 –69 –21 and extending anteriorily to a group of voxels surrounded by the lesion, at about
the normal coordinates of the VWFA [TC –42 –63 –24; t(197) =
5.17].
The contrast of alphabetic stimuli versus chequerboards elicited a similar pattern. In particular a posterior left-hemispheric
cluster [main peak: TC –24 –72 12; t(197) = 12.9] encompassed
the residual VWFA region identified with the previous contrast
[at TC –42 –63 –24, t(197) = 4.65]. At the peak voxel, this region
was not activated by chequerboards relative to fixation [t(197) =
0.49]. Finally, the activation was marginally stronger for real
words than for consonant strings [t(197) = 1.74; P < 0.1]. Thus,
it was possible to pick out a small residual region, contiguous to
the lesion, with coordinates and functional properties analogous
to those of the VWFA.
Using the contrast of alphabetic stimuli versus chequerboards,
the R-VWFA appeared as a high peak [TC 39 –63 –18; t(197) =
8.47] within a posterior cluster [main peak: TC 21 –69 42; t(197)
= 12.5]. It was also activated by chequerboards relative to fixation, although weakly [t(197) = 2.34]. Moreover, the activation
was stronger for real words than for consonants [t(197) = 2.39].
Thus in patient F, the major part of the normal VWFA was
disrupted, while the right-hemispheric symmetrical region
showed the general functional properties usually attached to the
VWFA.
Finally, we checked that this pattern of R-VWFA activation
deviated significantly from the normal pattern. In an analysis
with subjects as random factor, the activation at the peak voxel
(TC 39 –63 –18) by alphabetic strings and by words relative to
chequerboards was stronger in patient F than in the group of
normal controls [t(8) = 2.87, P < 0.05; and t(8) = 4.31, P < 0.005,
respectively]. We also searched for excessive activations in
patient F, which may reflect the effortful word identification
based on serial letter identification. Indeed, a large part of the
language-related network was activated more strongly by words
versus consonant strings in patient F as compared with normal
subjects (voxelwise P < 0.001, cluster-level P < 0.05 corrected):
left anterior insula/Broca’s area (TC –36 15 3) and supramarginal
gyrus (TC –42 –51 33), but also left-predominant prefrontal

Downloaded from https://academic.oup.com/cercor/article/13/12/1313/384468 by guest on 11 April 2024

In order to isolate the VWFA, we contrasted alphabetic stimuli
versus chequerboards. This revealed left parietal, temporal and
rolandic activations, including a VWFA cluster too small to reach
cluster-level significance with a whole-brain correction [TC –39
–54 –9; t(197) = 3.83; 10 voxels] (Fig. 6). Importantly, the VWFA
was not activated by chequerboards relative to fixation [t(197) =
1.59; NS]. The activation did not differ significantly between
consonant strings and real words [t(197) = 0.86; NS].
The contrast of alphabetic stimuli versus chequerboards
revealed no significant cluster in the right temporal region.
However, as mentioned before, the comparison of alphabetic
stimuli versus fixation showed bilateral ventral occipitotemporal
activations. In contrast with the VWFA, the R-VWFA was also
activated by chequerboards relative to fixation [TC 30 –54 –18;
t(197) = 5.46]. Like in the VWFA proper, the activation did not
differ between consonant strings and real words.

cortex, bilateral rolandic cortex, posterior middle temporal
gyrus and occipital cortex (TC –27 –69 3 and TC 27 –72 12).

Additional Cases and Study of Lesion Overlap
Before studying anatomical–behavioural correlations across
patients, we will enrich our database with three additional
cases, including two previously published ones. Those patients’
behaviour and imaging data will be shortly described.
Letter-by-letter Reading: Patient A
Patient A was a 63-year-old right-handed man, who suffered from
a left PCA infarct responsible for right hemianopia, without
macular sparing, and reading difficulties. Oral language production and comprehension, picture naming, writing to dictation,
identification of spelled-out words and graphemic processing
were normal. The patient read centrally displayed words flawlessly but slowly (mean latency = 2463 ms), with a marked
length effect of 397 ms per letter (P < 0.001). He was also accurate but slow when reading single letters (mean latency = 894
ms). In brief, patient A showed pure alexia with effective letterby-letter reading, and was thus closely similar to patient F.
The anatomical MRI showed a left ventral occipito-temporal
lesion. It affected the inferior cortex of the occipital lobe and
extended to most of its mesial aspect, including both banks of
the calcarine sulcus. It extended anteriorily precisely within the
occipito-temporal sulcus to about TC y = –60. There was no
visible callosal lesion. There was also a small left thalamic lesion.
The region of normal VWFA activation overlapped substantially
with the left fusiform lesion (Fig. 4).

Global Alexia: Patient VOL
Patient VOL was a 66-year-old right-handed woman who
suffered from a left PCA infarct responsible for right hemianopia
with macular sparing, severe alexia and some degree of associative agnosia for objects (Cohen and Dehaene, 2000). She made
as much as 31% errors when trying to name isolated upper-case
letters. She often reached the correct response by outlining the
target letter with her finger. She was almost completely unable
to read any word aloud. She tried to resort to an effortful and
distressing letter-by-letter reading strategy, but made many letter
naming errors, and testing had to be discontinued after the first
few items. In contrast, she did not make a single error in naming
the same words when they were spelled out orally to her. She
could write letters, words and sentences normally, either spontaneously or to dictation.
VOL’s residual word processing abilities were assessed using
lexical decision, language decision and semantic classification
tasks. In summary, the results suggested above chance performance in several lexical decision and language decision tasks, but
no evidence for access to word meaning was found (Cohen and
Dehaene, 2000). There was also ample evidence that she could
accurately access the quantitative meaning of Arabic numerals,
even when she was unable to read those numbers aloud.
A T2-weighted anatomical MRI showed an extensive ventral
occipitotemporal lesion, also affecting the mesial occipital lobe
both below and above the calcarine sulcus (Fig. 4). The infarct
encompassed the left part of the splenium of the corpus
callosum. After normalization in Talairach space, the anterior
boundary of the temporal lesion reached about y = –20. In
summary, patient VOL had global alexia; her lesion included the
VWFA and beyond, and affected interhemispheric connections.
Study of Lesion Overlap
The normalized lesions from the six patients were smoothed
(5 mm Gaussian filter) and then added, with a positive weight
for the three patients with pure alexia (F, A and VOL) and an
equal negative weight for the three patients without pure alexia
(D, M and RAV). The resulting image (Fig. 7, top) is proportional to the log likelihood ratio, log[P(pure alexia|lesion)/P(no
pure alexia|lesion)], which expresses for each voxel the conditional link between the existence of a lesion and the occurrence

Cerebral Cortex December 2003, V 13 N 12 1327

Downloaded from https://academic.oup.com/cercor/article/13/12/1313/384468 by guest on 11 April 2024

Summary and Discussion
Contrary to the previous patients, patient F was substantially
impaired in his daily command of reading, in the absence of
other relevant language deficit, including writing. In the
proposed framework, this pattern is suggestive of an impaired
visual word form. As suggested before, letter-by-letter reading
may result from a kind of serial spelling of stimuli by the right
hemisphere to the left-hemispheric language system. Anatomical and functional data are compatible with this hypothesis
(Figs 4 and 6): first, patient F’s lesion overlapped with the
normal region of the VWFA, explaining the reading deficit.
Secondly, callosal connections appeared to be intact, allowing
for interhemispheric transfer of letter identities. Thirdly, the
right-hemispheric R-VWFA activations showed a strong preference for alphabetic stimuli over chequerboards, a pattern
normally specific to the VWFA, in normals as well as in patients
D and M. This suggests that indeed the right-sided visual system
assumed a function in reading normally devoted to the VWFA.
Still, some residual activations were observed around the VWFA,
suggesting that this structure was partially spared. However, it is
unclear what role, if any, the residual activation of the VWFA
played in the patient’s behaviour, considering that they were
closely surrounded by lesioned cerebral tissue, and possibly
deprived of some of their connections. Finally, letter-by-letter
word reading induced a stronger than normal activation of
language related areas, particularly in a network repeatedly associated with verbal working memory, including Broca’s area (BA
44), dorsolateral prefrontal cortex (BA 9), and left supramarginal
gyrus (BA 40) (for a review see Cabeza and Nyberg, 2000).

Complete Left Hemialexia: Patient R AV
Patient RAV was a 30-year-old right-handed woman who suffered
from a left PCA infarct yielding an isolated interhemispheric
disconnection syndrome in the visual, haptic, and auditory
modalities (Cohen and Dehaene, 1996, 1998, 2000). There was
a right superior homonymous scotoma sparing central vision.
Oral language production and comprehension, and writing to
dictation, were entirely normal. The patient did not report
subjective reading difficulties. However she was almost unable
to read a single word displayed tachistoscopically in her LVF,
while she performed normally with RVF words. With fMRI,
patient RAV’s VWFA was normally activated by RVF words relative to fixation, while LVF words induced no significant activation. She thus displayed a left hemialexia qualitatively similar to
patient D, albeit more severe (Cohen and Dehaene, 2000).
Her lesion affected the left half of the posterior half of the
corpus callosum, extending from TC y = 0 anteriorily to the
posterior end of the splenium at TC y = –50. There was a small
additional lesion in the left occipital cortex, ventral to the
anterior part of the calcarine sulcus.

of pure alexia. The critical area follows closely the left occipitotemporal sulcus and overlaps with the VWFA, as identified with
fMRI in normal subjects by contrasting alphabetic stimuli versus
chequerboards (Fig. 7, bottom).

General Discussion

Critical Lesions and the V WFA
Our first aim was to assess the hypothesis that a necessary and
sufficient condition for the apparion of pure alexia is that the

Figure 7. Correlation between the lesion site associated with pure alexia (top) and the left occipito-temporal activation during normal reading (bottom). The normalized lesions from
six patients were combined, to compute for each voxel an index proportional to the log likelihood ratio, expressing the conditional link between the existence of a lesion and the
occurrence of pure alexia. The critical area follows the left occipito-temporal sulcus, and overlaps neatly with the VWFA, as identified with fMRI in normal subjects by contrasting
alphabetic stimuli versus chequerboards.

1328 Pure Alexia and the Visual Word Form Area • Cohen et al.

Downloaded from https://academic.oup.com/cercor/article/13/12/1313/384468 by guest on 11 April 2024

According to a simple anatomical and functional model of word
reading, letters displayed in one hemifield are first analysed
through a cascade of contralateral retinotopic areas, which
compute increasingly abstract representations. Eventually, an
invariant representation of letter identities is created in the
VWFA, reproducibly located within the left occipito-temporal
sulcus. The VWFA then projects to structures involved in phonological or lexico-semantic processing. This model yields detailed
predictions on the reading impairments that may follow left
occipitotemporal lesions. Those predictions were confronted to
behavioural, anatomical and fMRI data gathered from normals
and from patients suffering from left posterior cerebral artery
infarcts.

VWFA should be destroyed or non-functional. This hypothesis is
directly supported by the spatial congruence of the critical
lesion site, as identified by combining the normalized lesions
from six patients with the activation of the VWFA in normal
subjects (Fig. 7). However, the normalized coordinates of brain
lesions may not always be sufficient to predict accurately
whether pure alexia should be expected in a given patient.
There is some variability in the coordinates of the VWFA even in
normal subjects, and the geometric normalization process is not
flawless, particularly when processing lesioned brains (Brett et
al., 2001). In addition to the Talairach coordinates of the lesion,
an injury to the cortex buried in the occipito-temporal sulcus
may be a useful predictor of pure alexia. Indeed, in spite of the
variability in sulcal anatomy, the normal VWFA activation is
reproducibly located in the depths of the sulcus across subjects
(Cohen et al., 2000, 2002). For instance, this criterion can be
used to contrast the cases of patient M, who had hemianopia
without pure alexia, and patients F and A, who had pure alexia
with letter-by-letter reading. As visible on the coronal images in
Figure 4, which cut through the normal position of the VWFA,
the lesions of patients F and A affected the cortex within the
occipito-temporal sulcus, while in patient M it was confined to
the mesial part of the fusiform gyrus.

methodological differences in the reconstruction of lesions may
be sufficient to account for this discrepancy.
Beside the ventral occipito-temporal cortex, the corpus
callosum is the main structure invoked in most accounts of pure
alexia. In normal subjects, we postulate that its only role in
reading is to convey visual information about letters displayed in
the LVF from the right visual cortex to the VWFA. As a consequence, an isolated callosal lesion only induces pure alexia
restricted to the LVF (Cohen and Dehaene, 2000; Molko et al.,
2002). The case of patient D illustrates this pattern of deficit: as
a result of a small lesion of the left forceps major, patient D’s
VWFA was not normally activated by LVF words. Patient D’s
lesion was similar to that reported by Suzuki et al. (1998) in a
patient with left hemialexia but without anomia for LVF
pictures. A review of published cases of callosal alexia suggests
that the ventroposterior part of the splenium is responsible for
the transfer of visual information about alphabetic stimuli
(Suzuki et al., 1998). In order to induce clinically detectable
pure alexia, a callosal lesion must be associated with right hemianopia, corresponding to a type II lesion, following the typology
of Damasio and Damasio (1983). In non-alexic hemianopic
patients, such as patient M, the callosal route is indeed vital to
reading, as it provides the only input to the VWFA and subsequent stages of reading, and an additional callosal lesion would
be sufficient to yield pure alexia.
However, if we consider only patients who do have pure
alexia, the role of the corpus callosum may go beyond this
normal contribution to reading. If we assume that residual
reading abilities, including letter-by-letter reading and implicit
lexical access (Coslett and Saffran, 1998; Friedman and Nitzberg
Lott, 2000), result from right-hemispheric word processing, the
intervention of the corpus callosum is required whenever the
output of those right-hemispheric computations must be made
available to the left-hemispheric verbal system. This issue will be
addressed in the next sections.
In summary, our main finding is that the VWFA, as defined by
its functional properties in normal subjects, is indeed the critical
lesion locus for pure alexia, an observation which agrees with
previous anatomical studies.
Letter-by-letter Reading
Letter-by-letter reading implies that the normal ability to identifie letter strings in a quasi-parallel fashion is lost. This impairment is tantamount to pure alexia, and we proposed that it
results from the destruction of the VWFA, from its deprivation of
input from both hemifields, or possibly from disrupted projections from the VWFA to subsequent language areas. Why, then,
do some pure alexic patients enjoy effective letter-by-letter
reading abilities, while others suffer from global alexia? What are
the brain mechanisms that underlie letter-by-letter reading?
Clarification on these questions should be gained by reducing
the emergence of letter-by-letter reading to two basic conditions, which will be discussed in turn.
A first condition to letter-by-letter reading is the existence of
residual abilities to identify individual letters, even if with some
inaccuracy and slowness. There is converging evidence that
alphabetic identification is within the reach of the right-hemispheric visual system, even if this process is not needed for
normal reading (for a review see Coslett and Saffran, 1998). Such
evidence comes from studies of split brains (Baynes and
Eliassen, 1998), left-hemispherectomized patients (Patterson et
al., 1989), deep dyslexic patients with extensive LH lesions

Cerebral Cortex December 2003, V 13 N 12 1329

Downloaded from https://academic.oup.com/cercor/article/13/12/1313/384468 by guest on 11 April 2024

The pattern of activation is also compatible with our hypothesis: the VWFA was activated in patient M, and partially
preserved in patient F. The comparison of F and A may suggest a
further source of variability between patients. In patient A, the
lesion destroyed entirely the VWFA cortex, while in patient F, a
small region of cortex was apparently spared, and could be activated. However, this patch was bordered on its mesial, anterior
and posterior flanks by lesioned tissue, and probably did not
fulfil the normal functions of the VWFA, despite some degree of
activation.
Our approach is similar to that put forward by Leff et al.
(2001). During PET scanning, they proposed a word reading task
to normal subjects, hemianopic patients with or without
macular sparing, and to a pure alexic patient with letter-by-letter
reading. The study was focused on a ventral occipito-temporal
region, isolated by virtue of its approximately linear increase of
activation with the rate of stimulus presentation. This functional
criterion was thus correlated with simple physical parameters
such as the amount of visual stimulation, but had no specificity
for alphabetic stimuli. Indeed, the peak coordinates were TC
–42 –72 –16, which is somewhat posterior to the VWFA, and
closer to occipital regions sensitive to presentation rate (Price et
al., 1996a; Price and Friston, 1997) and other physical parameters (Mechelli et al., 2000). Actually, the coordinates reported by
Leff et al. (2001) are close to the main peak of the cluster
isolated in the present study by contrasting alphabetic stimuli
versus chequerboards in normals (TC –36 –75 –12). However, at
this position, which is posterior to the VWFA, previous studies
have shown that alphabetic stimuli have not yet reached a level
of representation invariant for position or case (Cohen et al.,
2000; Dehaene et al., 2001). Importantly, the lesion of the pure
alexic patient AR (reported in Leff et al., 2001) was large,
extended more anteriorily than the normal activation focus
picked out by Leff et al. (2001), and apparently encompassed
the VWFA as defined here. Conversely, the mean image
computed on the basis of the anatomical MRIs of hemianopic
patients suggests that lesions in those patients did not extend
anteriorily to TC y = –70, and hence spared the VWFA. In
summary, we suggest that the activation studied by Leff et al.
(2001) differs from the VWFA and may not be critical to the
genesis of pure alexia, but that their anatomical data are compatible with the account proposed here.
Damasio and Damasio (1983) distinguished three lesion
patterns associated with pure alexia. Type I results from extensive PCA infarcts affecting the mesial and ventral occipitotemporal cortex and the underlying white matter, probably
including the VWFA. Our patient VOL seems to be representative of this pattern. Type II results from right hemianopia, due to
a lesion of the calcarine cortex or of the optic radiations, associated with an interhemispheric disconnection. According to our
framework, this would yield a complete deafferentation of the
VWFA. Type III seems to result from ventral occipito-temporal
cortex lesions smaller than in type I, with partial or complete
sparing of the visual field. Although matching Damasio and
Damasio’s templates with our normalized data can only be
approximate, the critical lesion in type III patients overlaps
nicely with the VWFA as defined here. Binder and Mohr (1992)
went a step further in delineating the critical region, by
comparing the lesions of alexic patients with the lesions of
patients with a left PCA infarct but with no reading impairment.
The critical ventral occipito-temporal cortex region appears to
be slightly more anterior and mesial than the VWFA. However,

1330 Pure Alexia and the Visual Word Form Area • Cohen et al.

between the right ventral occipito-temporal cortex and speech
areas. Di Virgilio and Clarke (1997) studied the brain of a patient
with a right ventral occipito-temporal cortex lesion extending
from TC y = 5 to y = –52 along the antero-posterior axis, overlapping with the putative R-VWFA. By staining anterogradely
degenerating axons, they evidenced monosynaptic projections
from the lesioned area to left-hemispheric language areas,
including Broca’s and Wernicke’s areas and the inferior parietal
lobule. Such data help to flesh out the interhemispheric pathways that might be involved in letter-by-letter reading. In this
context, an interhemispheric disconnection should be sufficent
to shift a patient from letter-by-letter reading to global alexia.
Accordingly, the global alexic patient VOL had a callosal lesion,
while interhemispheric connections were apparently intact in
letter-by-letter readers F and A (Fig. 4). Similarly, Binder and
Mohr (1992) concluded that the only area which is lesioned in
global alexics as compared with letter-by-letter readers includes
the left forceps major, and extends into the left hemispheric
white matter above the posterior ventricular horn. As an aside, if
we assume that the interhemispheric pathways involved in
letter-by-letter reading project from the R-VWFA to language
areas, they should differ from those normally involved in reading
of LVF words, which are thought to project from RH visual
cortex to the VWFA (Molko et al., 2002). The callosal lesions
that induce left hemialexia (patient D) and those that prevent
letter-by-letter reading in alexic patients (patient VOL) appear to
be similar. Whether two anatomically distinct transcallosal pathways can actually be identified remains an open question.
Brain Lesions and Brain Imaging: A Revision of the
Model
We may now summarize the two main conclusions of this study.
First, we propose that, apart from cases of pure disconnection,
the critical brain lesion for pure alexia can be tightly localized to
the VWFA, as defined in normal subjects on the basis of its
anatomical location and functional properties (Fig. 7). Secondly,
we show that when the VWFA is destroyed, letter-by-letter
readers show activations in the right-hemispheric counterpart of
the VWFA, consonant with the idea that letters are identified in
this region. This latter point prompts a revision of the model of
reading proposed in the introduction, in order to insert an alternative pathway subtending letter-by-letter reading (Fig. 1). This
pathway includes the R-VWFA and its putative projections to
left-sided language areas.
Even with this improvement, the model fails to explain the
residual implicit reading abilities observed in some pure alexic
patients such as patient VOL. Implicit reading can be elicited in
tasks with no overt verbal component, such as lexical or
semantic decision (Coslett et al., 1993). Its emergence requires
that explicit reading, particularly through letter-by-letter strategies, be absent or prevented, e.g. through the rapid presentation of masked stimuli. Coslett and Saffran have put forward a
number of arguments supporting the hypothesis that implicit
reading would reflect a lexical and semantic access within the
RH, capable of driving non-verbal responses, and impeded by
concurrent attempts of the LH to engage in verbal strategies
(Coslett and Saffran, 1998). In some patients, implicit word identification may eventually be made available to overt verbal
output, yielding deep dyslexic phenomena, such as fast reading
of short highly imageable words (Buxbaum and Coslett, 1996).
Endorsing the right-hemispheric hypothesis, such deep dyslexic
phenomena also require that semantic information be trans-

Downloaded from https://academic.oup.com/cercor/article/13/12/1313/384468 by guest on 11 April 2024

(Coltheart, 1980) and patients with pure alexia. We mentioned
the case of a pure alexic patient who lost her letter-by-letter
reading abilities following a contralateral RH infarct (Bartolomeo
et al., 1998). Similarly, Coslett and Monsul (1994) transiently
disrupted the residual reading abilities of a pure alexic patient by
means of transcranial magnetic stimulation. There is similar
evidence on the identification of Arabic digits by the RH
(Cohen and Dehaene, 1995, 1996; Gazzaniga and Hillyard,
1971; Gazzaniga and Smylie, 1984; Seymour et al., 1994). The
RH thus appears as a plausible source of alphabetic identification in letter-by-letter reading. Our imaging results support this
hypothesis and clarify its anatomical foundations by suggesting
an involvement of the R-VWFA. First, the right ventral occipitotemporal cortex is strongly activated by alphabetic stimuli even
in normal subjects. Although this activation is not stronger
than for chequerboards, it demonstrates that letters are indeed
processed in the RH up to high-order visual cortices devoted to
object recognition (Gauthier, 2000; Ishai et al., 2000; Malach et
al., 2002). This supports the idea that letters can be identified in
those structures, even if not with the same degree of invariance
as in the VWFA. Secondly, the left ventral occipito-temporal activation by alphabetic stimuli was severely reduced in patient F.
This implies that in those patients, letter-by-letter reading was
not achieved by the LH ventral visual system, although it cannot
be excluded that the residual LH activation in patient F had
some contribution in his behaviour. Moreover, in patient F, the
right-hemispheric ventral occipito-temporal cortex showed
functional properties usually characteristic of the VWFA, i.e.
stronger activation by alphabetic stimuli than by chequerboards.
The existence of alphabetic activations in the R-VWFA in
alexic patients might depend on premorbid functional dispositions. There is indeed some variability among normal subjects in
the functional properties of the R-VWFA. For instance, Cohen et
al. (2002)) found in two out of seven subjects an R-VWFA activation for alphabetic stimuli versus chequerboards, invariant for
position in the visual field, although this activation was always
weaker than in the VWFA proper. One may speculate that such
an RH activation pattern could be a predisposition to the emergence of letter-by-letter reading in case of LH lesion. A related
question is whether the improvement of letter-by-letter reading
over time, as occurred in patient F, could result from changes in
the operation of the R-VWFA. Such adaptations could reflect
changes in the intrinsic properties of the R-VWFA following the
stroke. Alternatively, patients who read letter-by-letter may
merely learn to rely more heavily on their R-VWFA during
reading than do normal subjects. These issues could be tackled
through longitudinal studies of alexic patients during the course
of functional recovery (Behrmann et al., 1990; Small et al.,
1998).
A second precondition to letter-by-letter reading is the ability
to combine the serially identified letters, in order to access the
lexicon and other language-dependent processes. By default,
one may assume that the central mechanisms of letter combination are the same as those involved in recognizing orally spelledout words, and that they are based on verbal and working
memory processes. Indeed, patient F showed increased activation during letter-by-letter reading relative to normal controls in
areas subtending language and verbal working memory. This
additional processing may reflect the stepwise recovery of word
identity in letter-by-letter readers. However, for those mechanisms to receive as input the series of letters presumably identified in the right hemisphere, direct or indirect links must exist

Notes
This work was supported by a ‘Cognitique’ grant from the French
Ministry of Research.
Address correspondence to Laurent Cohen, MD, PhD, Service de
Neurologie 1, Hôpital de la Salpêtrière, 47/83 Bd de l’Hôpital, 75651
Paris Cedex 13, France. Email: laurent.cohen@psl.ap-hop-paris.fr.

References
Aghababian V, Nazir TA (2000) Developing normal reading skills: aspects
of the visual processes underlying word recognition. J Exp Child
Psychol 76:123–150.
Bartolomeo P, Bachoud-Levi AC, Degos JD, Boller F (1998) Disruption of
residual reading capacity in a pure alexic patient after a mirror-image
right-hemispheric lesion. Neurology 50:286–288.
Baynes K, Eliassen JC (1998) The visual lexicon: its access and organization in commissurotomy patients. In: Right hemisphere language
comprehension (Beeman M, Chiarello C, eds), pp. 79–104. Mahwah,
NJ: Lawrence Erlbaum Associates.
Behrmann M, Black SE, Bub D (1990) The evolution of pure alexia: a
longitudinal study of recovery. Brain Lang 39:405–427.
Beversdorf DQ, Ratcliffe NR, Rhodes CH, Reeves AG (1997) Pure alexia:
clinical–pathologic evidence for a lateralized visual language association cortex. Clin Neuropathol 16:328–331.
Binder JR, Mohr JP (1992) The topography of callosal reading pathways.
A case–control analysis. Brain 115:1807–1826.
Binder JR, Frost JA, Hammeke TA, Bellgowan PS, Rao SM, Cox RW (1999)
Conceptual processing during the conscious resting state. A functional MRI study. J Cogn Neurosci 11:80–95.
Bokde AL, Tagamets MA, Friedman RB, Horwitz B (2001) Functional
interactions of the inferior frontal cortex during the processing of
words and word-like stimuli. Neuron 30:609–617.
Booth MC, Rolls ET (1998) View-invariant representations of familiar
objects by neurons in the inferior temporal visual cortex. Cereb
Cortex 8:510–523.
Brett M, Leff AP, Rorden C, Ashburner J (2001) Spatial normalization of
brain images with focal lesions using cost function masking.
Neuroimage 14:486–500.
Büchel C, Price C, Friston K (1998) A multimodal language region in the
ventral visual pathway. Nature 394:274–277.
Burgund ED, Marsolek CJ (1997) Letter-case-specific priming in the right
cerebral hemisphere with a form-specific perceptual identification
task. Brain Cogn 35:239–258.
Burgund ED, Marsolek CJ (2000) Viewpoint-invariant and viewpointdependent object recognition in dissociable neural subsystems.
Psychon Bull Rev 7:480–489.
Buxbaum LJ, Coslett HB (1996) Deep dyslexic phenomena in a letter-byletter reader. Brain Lang 54:136–167.

Cabeza R, Nyberg L (2000) Imaging cognition. II., An empirical review of
275 PET and fMRI studies. J Cogn Neurosci 12:1–47.
Castro-Caldas A, Salgado V (1984) Right hemifield alexia without hemianopia. Arch Neurol 41:84–87.
Chawla D, Rees G, Friston KJ (1999) The physiological basis of attentional modulation in extrastriate visual areas. Nat Neurosci
2:671–676.
Cohen L, Dehaene S (1995) Number processing in pure alexia: the
effect of hemispheric asymmetries and task demands. Neurocase
1:121–137.
Cohen L, Dehaene S (1996) Cerebral networks for number processing:
evidence from a case of posterior callosal lesion. Neurocase
2:155–174.
Cohen L, Dehaene S (1998) Competition between past and present:
assessment and interpretation of verbal perseverations. Brain
121:1641–1659.
Cohen L, Dehaene S (2000) Calculating without reading: unsuspected
residual abilities in pure alexia. Cogn Neuropsychol 17:563–583.
Cohen L, Dehaene S, Naccache L, Lehéricy S, Dehaene-Lambertz G,
Hénaff MA, et al. (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 123:291–307.
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 125:1054–69.
Coltheart M (1980) Deep dyslexia: a right-hemisphere hypothesis. In:
Deep dyslexia (Coltheart M, Patterson K, Marshall JC, eds),
pp. 326–380. London: Routledge.
Corbetta M, Shulman GL (1998) Human cortical mechanisms of visual
attention during orienting and search. Phil Trans R Soc Lond B Biol Sci
353:1353–1362.
Coslett HB, Monsul N (1994) Reading with the right hemisphere:
evidence from transcranial magnetic stimulation. Brain Lang
46:198–211.
Coslett HB, Saffran EM (1989) Evidence for preserved reading in ‘pure
alexia’. Brain 112:327–359.
Coslett HB, Saffran EM (1998) Reading and the right hemisphere:
evidence from acquired dyslexia. In: Right hemisphere language
comprehension (Beeman M and Chiarello C, eds), pp. 105–132.
Mahwah, NJ: Lawrence Erlbaum Associates.
Coslett HB, Saffran EM, Greenbaum S, Schwartz H (1993) Reading in pure
alexia: the effect of strategy. Brain 116:21–37.
Damasio AR, Damasio H (1983) The anatomic basis of pure alexia.
Neurology 33:1573–1583.
Dehaene S, Naccache L, Cohen L, Bihan DL, Mangin JF, Poline JB, et al.
(2001) Cerebral mechanisms of word masking and unconscious
repetition priming. Nat Neurosci 4:752–8.
Dehaene S, Le Clec’H G, Poline JB, 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.
Dejerine J (1891) Sur un cas de cécité verbale avec agraphie suivi
d’autopsie. Mém Soc Biol 3:197–201.
Dejerine J (1892) Contribution à l’étude anatomo-pathologique et
clinique des différentes variétés de cécité verbale. Mém Soc Biol
4:61–90.
Di Virgilio G, Clarke S (1997) Direct interhemispheric visual input to
human speech areas. Hum Brain Mapp 5:347–354.
Friedman RB, Nitzberg Lott S (2000) Rapid word identification in pure
alexia is lexical but not semantic. Brain Lang 00:000–000.
Gauthier II (2000) What constrains the organization of the ventral
temporal cortex? Trends Cogn Sci 4:1–2.
Gazzaniga MS, Hillyard SA (1971) Language and speech capacity of the
right hemisphere. Neuropsychologia 9:273–280.
Gazzaniga MS, Smylie CE (1984) Dissociation of language and cognition:
a psychological profile of two disconnected right hemispheres. Brain
107:145–153.
Giraud AL, Price CJ (2001) The constraints functional neuroanatomy
places on classical models of auditory word processing. J Cogn
Neurosci 13:754–765.
Gitelman DR, Nobre AC, Parrish TB, LaBar KS, Kim YH, Meyer JR, et al.
(1999) A large-scale distributed network for covert spatial attention:

Cerebral Cortex December 2003, V 13 N 12 1331

Downloaded from https://academic.oup.com/cercor/article/13/12/1313/384468 by guest on 11 April 2024

ferred to the LH. Sidtis et al. (1981) studied a patient who underwent a two-stage callosotomy. When only the posterior half of
his corpus callosum was severed, the patient could not read
aloud words presented in his LVF. However, he could still
verbally approach their meaning, presumably owing to the
transfer of semantic information through the intact anterior
callosum. This suggests that beside the transhemispheric
pathway subtending letter-by-letter reading, a more anterior
semantic route, perhaps available only in a few subjects with a
strong premorbid contribution of the right-hemisphere to
reading, may help some patients to circumvent their pure alexia.
Indeed, letter-by-letter reading (Behrmann et al., 1990; Greenwald and Gonzalez Rothi, 1998) and making implicit reading
explicit (Friedman and Nitzberg Lott, 2000) are the two main
rehabilitation methods that can be applied in pure alexia. By
delineating such alternative routes (Price and Friston, 2002), our
work may ultimately support rehabilitation strategies tailored to
individual alexic patients.

1332 Pure Alexia and the Visual Word Form Area • Cohen et al.

McCandliss BD, Cohen L, Dehaene S (2003) The visual word form area:
expertise for reading in the fusiform gyrus. Trends Cogn Sci
7:293–299.
Mechelli A, Humphreys GW, Mayall K, Olson A, Price CJ (2000)
Differential effects of word length and visual contrast in the fusiform and lingual gyri during reading. Proc R Soc Lond B Biol Sci
267:1909–1913.
Mesulam MM (1998) From sensation to cognition. Brain 121:1013–1052.
Miozzo M, Caramazza A (1998) Varieties of pure alexia: the case of
failure to access graphemic representations. Cogn Neuropsychol
15:203–238.
Molko N, Cohen L, Mangin JF, Chochon F, Lehéricy S, Le Bihan D, et al.
(2002) Visualizing the neural bases of a disconnection syndrome with
diffusion tensor imaging. J Cogn Neurosci 14:629–636.
Morais J, Kolinsky R (1994) Perception and awareness in phonological
processing: the case of the phoneme. Cognition 50:287–297.
Mycroft R, Hanley JR, Kay J (2002) Preserved access to abstract letter
identities despite abolished letter naming in a case of pure alexia. J
Neuroling 15:99–108.
Nazir TA (2000) Traces of print along the visual pathway. In: Reading as
a perceptual process (Kennedy A, Radach R, Heller D, Pynte J, eds),
pp. 3–22. Amsterdam: Elsevier.
Oldfield RC (1971) The assessment and analysis of handedness: the
Edinburgh inventory. Neuropsychologia 9:97–113.
Patterson K, Vargha-Khadem F, Polkey CE (1989) Reading with one hemisphere. Brain 112:39–63.
Pirozzolo FJ, Kerr KL, Obrzut JE, Morley GK, Haxby JV, Lundgren S
(1981) Neurolinguistic analysis of the language abilities of a patient
with a ‘double disconnection syndrome’: a case of subangular alexia
in the presence of mixed transcortical aphasia. J Neurol Neurosurg
Psychiatry 44:152–155.
Price CJ (1997) Functional anatomy of reading. In: Human brain function
(Frackowiak RSJ, Friston KJ, Frith CD, Dolan RJ, Mazziotta JC, eds),
pp. 301–328. San Diego, CA: Academic Press.
Price CJ, Friston KJ (1997) The temporal dynamics of reading: a PET
study. Proc R Soc Lond B Biol Sci 264:1785–1791.
Price CJ, Friston KJ (2002) Degeneracy and cognitive anatomy. Trends
Cogn Sci 6:416–421.
Price CJ, Moore CJ, Frackowiak RS (1996a) The effect of varying stimulus
rate and duration on brain activity during reading. Neuroimage
3:40–52.
Price CJ, Wise RJS, Frackowiak RSJ (1996b) Demonstrating the implicit
processing of visually presented words and pseudowords. Cereb
Cortex 6:62–70.
Puce A, Allison T, Asgari M, Gore JC, McCarthy G (1996) Differential
sensitivity of human visual cortex to faces, letterstrings, and textures:
a functional magnetic resonance imaging study. J Neurosci
16:5205–5215.
Raichle ME, MacLeod AM, Snyder AZ, Powers WJ, Gusnard DA, Shulman
GL (2001) A default mode of brain function. Proc Natl Acad Sci USA
98:676–682.
Sakai K, Miyashita Y (1991) Neural organization for the long-term
memory of paired associates. Nature 354:152–155.
Seymour SE, Reuter-Lorenz PA, Gazzaniga MS (1994) The disconnection
syndrome: basic findings reaffirmed. Brain 117:105–115.
Sidtis JJ, Volpe BT, Holtzman JD, Wilson DH, Gazzaniga MS (1981)
Cognitive interaction after staged callosal section: evidence for
transfer of semantic activation. Science 212:344–346.
Simon O, Mangin JF, Cohen L, Le Bihan D, Dehaene S (2002) Topographical layout of hand, eye, calculation, and language-related areas in the
human parietal lobe. Neuron 33:475–487.
Small SL, Flores DK, Noll DC (1998) Different neural circuits subserve
reading before and after therapy for acquired dyslexia. Brain Lang
62:298–308.
Somers DC, Dale AM, Seiffert AE, Tootell RBH (1999) Functional MRI
reveals spatially specific attentional modulation in human primary
visual cortex. Proc Natl Acad Sci USA 96:1663–1668.
Suzuki K, Yamadori A, Endo K, Fujii T, Ezura M, Takahashi A (1998)
Dissociation of letter and picture naming resulting from callosal
disconnection. Neurology 51:1390–1394.

Downloaded from https://academic.oup.com/cercor/article/13/12/1313/384468 by guest on 11 April 2024

further anatomical delineation based on stringent behavioural and
cognitive controls. Brain 122:1093–1106.
Greenblatt SH (1973) Alexia without agraphia or hemianopsia. Anatomical analysis of an autopsied case. Brain 96:307–316.
Greenblatt SH (1976) Subangular alexia without agraphia or hemianopsia. Brain Lang 3:229–245.
Greenwald ML, Gonzalez Rothi LJ (1998) Lexical access via letter naming
in a profoundly alexic and anomic patient: a treatment study. J Int
Neuropsychol Soc 4:595–607.
Grill-Spector K, Kushnir T, Edelman S, Avidan G, Itzchak Y, Malach R
(1999) Differential processing of objects under various viewing
conditions in the human lateral occipital complex. Neuron
24:187–203.
Gusnard DA, Raichle ME (2001) Searching for a baseline: functional
imaging and the resting human brain. Nat Rev Neurosci 2:685–694.
Haywood M, Coltheart M (2000) Neglect dyslexia and the early stages of
visual word recognition. Neurocase 6:33–44.
Helenius P, Tarkiainen A, Cornelissen P, Hansen PC, Salmelin R (1999)
Dissociation of normal feature analysis and deficient processing of
letter-strings in dyslexic adults. Cereb Cortex 9:476–483.
Hillis AE, Caramazza A (1995) A framework for interpreting distinct
patterns of hemispatial neglect. Neurocase 1:189–207.
Horwitz B, Rumsey JM, Donohue BC (1998) Functional connectivity of
the angular gyrus in normal reading and dyslexia. Proc Natl Acad Sci
USA 95:8939–8944.
Iragui VJ, Kritchevsky M (1991) Alexia without agraphia or hemianopia
in parietal infarction. J Neurol Neurosurg Psychiatryry 54:841–842.
Ishai A, Ungerleider LG, Martin A, Schouten JL, Haxby JV (1999) Distributed representation of objects in the human ventral visual pathway.
Proc Natl Acad Sci USA 96:9379–9384.
Ishai A, Ungerleider LG, Haxby JV (2000) Distributed neural systems for
the generation of visual images. Neuron 28:979–990.
Jernigan TL, Ostergaard AL, Law I, Svarer C, Gerlach C, Paulson OB
(1998) Brain activation during word identification and word recognition. Neuroimage 8:93–105.
Kanwisher N, McDermott J, Chun MM (1997) The fusiform face area: a
module in human extrastriate cortex specialized for face perception.
J Neurosci 17:4302–4311.
Kreiman G, Koch C, Fried I (2000) Category-specific visual responses of
single neurons in the human medial temporal lobe. Nat Neurosci
3:946–953.
Lambon Ralph MA, McClelland JL, Patterson K, Galton CJ, Hodges JR
(2001) No right to speak? The relationship between object naming
and semantic impairment: neuropsychological evidence and a
computational model. J Cogn Neurosci 13:341–356.
Lavidor M, Ellis AW (2002) Word length and orthographic neighborhood
size effects in the left and right cerebral hemispheres. Brain Lang
80:45–62.
Lavidor M, Ellis AW, Shillcock R, Bland T (2001) Evaluating a split
processing model of visual word recognition: effects of word length.
Brain Res Cogn Brain Res 12:265–272.
Leff AP, Crewes H, Plant GT, Scott SK, Kennard C, Wise RJ (2001) The
functional anatomy of single-word reading in patients with hemianopic and pure alexia. Brain 124:510–521.
Levy I, Hasson U, Avidan G, Hendler T, Malach R (2001)
Center–periphery organization of human object areas. Nat Neurosci
4:533–539.
Malach R, Levy I, Hasson U (2002) The topography of high-order human
object areas. Trends Cogn Sci 6:176–184.
Marsolek CJ (1995) Abstract visual-form representations in the left cerebral hemisphere. J Exp Psychol Hum Percept Perform 21:375–386.
Marsolek CJ, Kosslyn SM, Squire LR (1992) Form-specific visual priming
in the right cerebral hemisphere. J Exp Psychol Learn Mem Cogn
18:492–508.
Marsolek CJ, Schacter DL, Nicholas CD (1996) Form-specific visual
priming for new associations in the right cerebral hemisphere. Mem
Cognit 24:539–556.
Mazoyer B, Zago L, Mellet E, Bricogne S, Etard O, Houde O, et al. (2001)
Cortical networks for working memory and executive functions
sustain the conscious resting state in man. Brain Res Bull 54:287–298.

Tagamets MA, Novick JM, Chalmers ML, Friedman RB (2000) A parametric approach to orthographic processing in the brain: an fMRI
study. J Cogn Neurosci 12:281–297.
Talairach J, Tournoux P (1988) Co-planar stereotaxicatlas of the human
brain. 3-Dimensional proportional system: an approach to cerebral
imaging (Mark Rayport, trans.). New York: Thieme Medical
Publishers/Stuttgart: George Thieme Verlag.
Tanaka K (1996) Inferotemporal cortex and object vision. Annu Rev
Neurosci 19:109–139.
Tarkiainen A, Helenius P, Hansen PC, Cornelissen PL, Salmelin R (1999)
Dynamics of letter string perception in the human occipitotemporal
cortex. Brain 122:2119–2132.
Tsunoda K, Yamane Y, Nishizaki M, Tanifuji M (2001) Complex objects
are represented in macaque inferotemporal cortex by the combination of feature columns. Nat Neurosci 4:832–838.

Vallar G (2001) Extrapersonal visual unilateral spatial neglect and its
neuroanatomy. Neuroimage 14:S52–S58.
Vogt BA, Finch DM, Olson CR (1992) Functional heterogeneity in cingulate cortex: the anterior executive and posterior evaluative regions.
Cereb Cortex 2:435–443.
Wagner AD, Schacter DL, Rotte M, Koutstaal W, Maril A, Dale AM, et al.
(1998) Building memories: remembering and forgetting of verbal
experiences as predicted by brain activity. Science 281:1188–1191.
Warrington EK, Shallice T (1980) Word-form dyslexia. Brain 103:99–112.
Weekes BS (1997) Differential effects of number of letters on word and
non-word naming latency. Q J Exp Psychol 50A:439–456.
Xu B, Grafman J, Gaillard WD, Ishii K, Vega-Bermudez F, Pietrini P, et al.
(2001) Conjoint and extended neural networks for the computation
of speech codes: the neural basis of selective impairment in reading
words and pseudowords. Cereb Cortex 11:267–277.

Downloaded from https://academic.oup.com/cercor/article/13/12/1313/384468 by guest on 11 April 2024
Cerebral Cortex December 2003, V 13 N 12 1333