NeuroImage 61 (2012) 258–274

Contents lists available at SciVerse ScienceDirect

NeuroImage
journal homepage: www.elsevier.com/locate/ynimg

Cortical networks for vision and language in dyslexic and normal children of variable
socio-economic status
Karla Monzalvo a, b, c, Joel Fluss c, d, e, Catherine Billard c, d,
Stanislas Dehaene a, b, c, f, Ghislaine Dehaene-Lambertz a, b, c,⁎
a
INSERM, Cognitive Neuroimaging Unit, Gif sur Yvette, 91191 France
b
CEA, DSV, I2BM, Neurospin center, Gif sur Yvette, 91191 France
c
University Paris 11, Orsay, France
d
Assistance Publique-Hôpitaux de Paris, 94 Le Kremlin-Bicêtre, France
e
Neurologie pédiatrique, Hôpitaux Universitaires Genève, Suisse
f
Collège de France, Paris, France

a r t i c l e i n f o a b s t r a c t

Article history: In dyslexia, anomalous activations have been described in both left temporo-parietal language cortices and in
Accepted 14 February 2012 left ventral visual occipito-temporal cortex. However, the reproducibility, task-dependency, and presence of
Available online 22 February 2012 these brain anomalies in childhood rather than adulthood remain debated. We probed the large-scale orga-
nization of ventral visual and spoken language areas in dyslexic children using minimal target-detection tasks
Keywords:
that were performed equally well by all groups. In 23 normal and 23 dyslexic 10-year-old children from two
Child
MRI
different socio-economic status (SES) backgrounds, we compared fMRI activity to visually presented houses,
Brain faces, and written strings, and to spoken sentences in the native or in a foreign language. Our results confirm
Speech a disorganization of both ventral visual and spoken language areas in dyslexic children. Visually, dyslexic
Reading children showed a normal lateral-to-medial mosaic of preferences, as well as normal responses to houses
Face and checkerboards, but a reduced activation to words in the visual word form area (VWFA) and to faces in
SES the right fusiform face area (FFA). Auditorily, dyslexic children exhibited reduced responses to speech in
posterior temporal cortex, left insula and supplementary motor area, as well as reduced responses to maternal
language in subparts of the planum temporale, left basal language area and VWFA. By correlating these two
findings, we identify spoken-language predictors of VWFA activation to written words, which differ for
dyslexic and normal readers. Similarities in fMRI deficits in both SES groups emphasize the existence of a
core set of brain activation anomalies in dyslexia, regardless of culture, language and SES, without however
resolving whether these anomalies are a cause or a consequence of impaired reading.
© 2012 Elsevier Inc. All rights reserved.

Introduction (VWFA), that strongly responds to orthographic stimuli in the learned

script, as well as to efficient integration in the superior temporal areas
During reading acquisition, the child's brain must learn to identify a between letters and the sounds they represent (Blomert and Froyen,
new category of visual stimuli, written words, and to connect it with the 2010; van Atteveldt et al., 2004).
already developed spoken language network. Acquisition of this new Dyslexic children exhibit functional deficits in both of these brain
capacity deeply affects anatomical and functional brain organization, systems (Maisog et al., 2008; Richlan et al., 2009, 2011). Impaired
as shown by studies comparing illiterate and literate adults (Carreiras phonological representations have been proposed as the main causal
et al., 2009; Castro-Caldas et al., 1998; Dehaene et al., 2010; Li et al., mechanism of developmental reading impairments (Blau et al., 2009,
2006). In particular, it leads to the development of a specific brain site 2010; Paulesu et al., 2001; Ramus, 2004; Shaywitz, 1998; Temple,
in the left visual occipito-temporal cortex, the “visual word form area” 2002; Torgesen et al., 1994; Vellutino et al., 2004) Brain imaging
studies have indeed revealed hypo-activations in adult and children
dyslexics in regions involved in phonetic computations, particularly
the left temporo-parietal region, often accompanied by an increase
Abbreviations: FFA, fusiform face area; VWFA, visual word form area; PPA, parahip- in the inferior frontal regions which is interpreted as a compensatory
pocampal place area. response (Chandrasekaran et al., 2009; Dufor et al., 2007; Hoeft et al.,
⁎ Corresponding author at: Laboratoire de Neuroimagerie Cognitive INSERM U992,
CEA/SAC/DSV/DRM/NeuroSpin, Bat 145, point courrier 156, F-91191 GIF/YVETTE,
2011; Maisog et al., 2008; Paulesu et al., 2001; Richlan et al., 2009;
France. Fax: + 33 1 69 08 79 73. Ruff et al., 2003; Shaywitz et al., 2003). Particular impressive are
E-mail address: [email protected] (G. Dehaene-Lambertz). recent studies revealing basic anomalies in processing individual

1053-8119/$ – see front matter © 2012 Elsevier Inc. All rights reserved.
doi:10.1016/j.neuroimage.2012.02.035
 K. Monzalvo et al. / NeuroImage 61 (2012) 258–274 259

speech sounds and letter–sound correspondences in the left superior the normal and the dyslexic groups, we constituted two subgroups
temporal gyrus (Blau et al., 2009, 2010). of children with low versus high socio-economic status (SES). The
Nevertheless, a recent meta-analysis (Richlan et al., 2011) indicates dominant view of dyslexia suggests that it reflects mainly a neurobio-
that such activation anomalies in dyslexics' left temporo-parietal cortex logical impairment that should reveal itself identically in all affected
are not always reproducible. In particular, while frequently seen in fMRI children, regardless of their cultural or social background (e.g.
studies of adult dyslexics, they are often absent when scanning dyslexic Paulesu et al., 2001). Yet this issue remains largely untested, since
children (but see Blau et al., 2010 and Maurer 2011). Instead, the most the majority of dyslexia studies have involved middle to high-SES chil-
reproducible site of hypo-activation in adult and children dyslexics dren. Reading difficulties are known to be compounded by a lower SES
appears to be the left occipito-temporal cortex, at the site of the (Noble et al., 2006a), but interestingly meta-phonological competence
VWFA (Richlan et al., 2011). Because this region is unresponsive to remains the most important predictor of reading difficulties even in
written words in non-readers (Dehaene et al., 2010), and quickly low SES children, with SES itself predicting only 9% of the variance in
increases its response during reading acquisition (Maurer et al., reading scores (Fluss et al., 2009; Frederickson and Frith, 1998).
2005b, 2010; Schlaggar and McCandliss, 2007), its reduced activation Noble et al. (2006a) therefore proposed a model where phonological
in dyslexics might reflect a consequence rather than a cause of reading deficits interact with SES, a low SES preventing the natural compensa-
impairment. Nevertheless, across distinct studies, anomalies in the tion available to high-SES children through parental input, early edu-
VWFA area can occur without any superior temporal deficit (McCrory cation and print exposure, larger vocabulary and superior executive
et al., 2005) or vice-versa (Blau et al., 2009). functions (Hackman and Farah, 2009). Here, because we used simple
In this context, the main goal of the present study was to probe and nearly passive tasks, differences in executive functions should
the functional organization of the large-scale maps for vision and minimally influence our results. We tested the prediction that a
spoken language in normal and dyslexic children. Within the same shared cerebral substrate for dyslexia should exist in both high- and
children, we mapped the mosaic of ventral regions responsive to low-SES readers, in spite of the low-SES children's lower vocabularies.
distinct categories of visual stimuli (faces, houses, words, checker-
boards) as well as the entire perisylvian areas responsive to spoken Methods
language, allowing us to evaluate the existence of correlated deficits
in these two domains. Furthermore, we aimed to assess whether Subjects
these functional anomalies may still be observed in the absence of
any complex task. Many previous neuroimaging studies of dyslexia 46 dyslexics and normal readers (25 boys and 21 girls, 8 y 11 m to
relied on reading, metaphonological, or orthographic tasks that 10 y 10 m) were recruited in the Paris area from two different socio-
present difficulties for dyslexic children. Only a handful of studies economic backgrounds (Table 1). Children with co-morbid cognitive,
have revealed persistent anomalies in functional activation during neurological or behavioral disorders, hearing deficit, not corrected
tasks requiring only minimal passive exposure to letter strings visual problem and IQ b 80 were excluded from the study. Half of
(Brunswick et al., 1999a), single letters (Blau et al., 2010), or speech the children were coming from a district classified as an educational
sounds corresponding to single letters (Blau et al., 2009, 2010; priority zone because of multiple social and demographic indicators
Froyen et al., 2011). Our design fell in the latter category: by requiring of a disadvantaged family background such as a high rate of unem-
only a minimal detection task, we aimed to probe functional brain ployment, a weak proportion of native French speakers and low or
architecture in the absence of differences potentially induced by very low family income. The other half came from a highly educated
poorer performance or greater effort in the dyslexic group. area, due to the proximity of a university and of numerous public and
In the first half of our experiment, children were presented with private research laboratories. All children, except one, were born in
visual words, faces, buildings and moving checkerboards while their France and all had been following the normal French academic cursus.
task was to detect a rare target star. This design allowed us to probe All the participants, their parents, school directors and teachers were
the overall organization of the lateral-to-mesial mosaic of activations orally informed about the aims of the project in personal or/and collec-
to visual categories (Golarai et al., 2007; Hasson et al., 2003; Haxby et tive meetings at schools, and received a printed information booklet
al., 2001; Ishai et al., 1999), as well the presence of specific anomalies about MRI scanning. All parents and every child gave their written in-
in the response to each category, either relative to rest, to the checker- formed consent for the behavioral tests and fMRI scanning. The study
board baseline, or to each other. Because face responses were recently was approved by the local ethical committee for biomedical research.
found to be modulated by literacy (Dehaene et al., 2010), we were par- In each SES group, half of the children were dyslexics, as previous-
ticularly interested in the possibility that the fusiform face (FFA) might ly diagnosed by a dedicated learning disability center (Dr. Billard,
be anomalous in dyslexia. CHU Bicêtre) based on extensive behavioral testing with nationally
In the second half of our experiment, the children listened to established criteria following INSERM recommendations (2007)
sentences in their native or in a foreign language, while their task (clinical examination, full-scale IQ, standardized tests for working
was to press a button at the end of each sentence. Following the memory, meta-phonology, spelling, rapid automatic naming, words,
design of our earlier study (Dehaene-Lambertz et al., 2006a), each non-words and passage reading, etc.). At the time of the study, we
sentence was repeated twice at a 14 s interval. This simple design checked their current reading level with “L'alouette”, a standardized
proved powerful in adults (Dehaene-Lambertz et al., 2006a) and in- reading test classically used to detect dyslexia in French speaking
fants (Dehaene-Lambertz et al., 2006a) in order to activate the entire children. It consists in reading as fast and accurately as possible a
perisylvian spoken-language network, including but not restricted to meaningless text of 265 words within 3 min (Lefavrais, 1967). The
regions involved in the phonological processing of connected speech, resulting score is commonly translated in a reading age and a delay
and dissect it into three levels, any of which might be deficient in of more than 18 months is an accepted criterion for dyslexia in
dyslexic children: (1) activations to all speech stimuli relative to children between 8 and 12 years. All the dyslexic children included
rest; (2) activation specific to the native language and absent to in this study had a persisting and pronounced reading delay in their
foreign language; (3) areas responsive to the native language and reading skills (− 50 to −22 months corresponding to −2.77 to
whose activation decreases with sentence repetition, thus probing −0.83 SD, mean delay: − 35 months or -1.95 SD) whereas our nor-
short-term linguistic memory. mal readers were between − 13 and +28 months corresponding to
Finally, one last goal of our study was to probe the invariance of +0.16 to 2.33 SD (mean: + 6 months, or 0.98 SD). Thus there was
functional brain anomalies in dyslexic children in the face of large no overlap between the two groups (Table 1). We also checked the
variations in socio-economic background. To this end, within both number of isolated words read in 1 min (LUM), handedness, verbal
260 K. Monzalvo et al. / NeuroImage 61 (2012) 258–274

Table 1
Characteristics of the four groups. Significant p-values are in bold.

Low-SES Dysl vs NR High-SES Dysl vs NR High vs Dysl vs Reading ∗ SES

Low SES NR (pvalue) (pvalue)
Dyslexics Normal t-tests Dyslexics Normal t-tests
(pvalue)
readers (pvalue) readers (pvalue)

Age (months) 118 (± 6) 115 (± 6) 0.19 120 (±6) 116 (± 6) 0.13 0.57 0.046 0.88
Sex (boys) 7/12 8/11 8/11 4/12
SES Score − 4.0 (± 1.5) − 3.6 (± 2.3) 3.3 (± 1.2) 4.1 (± 0.7)
Bilingualism at home (bilinguals) 10/12 7/11 1/11 0/12
Handedness (Edimburgh) 67.3 83.5 0.82 73.2 77.2 0.83 0.99 0.43 0.64
PIQ 87 91 0.55 98 108 0.26 0.014 0.64 0.2
VIQ 87 104 0.032 105 126 0.046 0.002 0.002 0.78
Number of correctly named images (/48) 30 34 0.13 36 40 0.021 0.001 0.006 0.84
Reading lag relative to age (months and − 38.25 −1 b.0001 − 30.8 + 11.8 b.0001 0.001 b.001 0.35
age normed standard score) (− 2.3 SD) (+ 0.6 SD) (− 1.6 SD) (+1.3 SD)
Reading Speed: Number of words 21 74 b.0001 32 86 b.0001 0.003 b.001 0.88
read in 1 min
RAN (time in sec) 22 17 0.059 20 15 0.002 0.23 0.001 0.85
Phonological awareness (% of correct 69.10% 91.40% 0.006 76.70% 90.70% 0.01 0.44 b .001 0.35
responses)
Forward Digit Span (normalized scores 9 12 0.016 10 14 0.005 0.19 b0.001 0.38
relative to age)
Backward Digit Span (normalized scores 9 11 0.067 10 11 0.4 0.33 0.063 0.62
relative to age)
Sentence span (number of words) 13 16 0.039 15 21 0.002 0.003 b.001 0.2
Visual cancelation task (number of 28 28 0.86 28 30 0.14 0.32 0.25 0.42
bells/35 found in 2 min)
Strings comparisons (% of correct responses) 95% 95% 0.9 95% 95% 0.86 0.31 0.8 0.99
Strings comparisons (time in sec to complete 97 87 0.53 99 81 0.09 0.88 0.13 0.64
the task)

abilities, and other abilities that have been proven to be predictive of was present. More specifically, besides reading, phonological aware-
reading difficulties (phonological awareness, verbal short-term mem- ness (EVALEC), verbal short-term memory (forward digit span and
ory, visual attention, rapid automated naming). Phonological aware- sentence span) and lexical access (RAN) were significantly worse in
ness was tested with a French standardized test (EVALEC) consisting dyslexics than in normal readers, and this was significant within
of the deletion of the first syllable in 10 trisyllabic words, then of the each SES group. In the low SES group, reading lag was significantly
first phoneme in 12 CVC words, finally in 12 CCV words (Sprenger- correlated with phonological awareness (p = 0.002), forward digit
Charolles et al., 2005). Verbal short-term memory was assessed with span (p = 0.02) and RAN (p = 0.04). The same correlations were
forward and backward digit span and a sentence span (correct repeti- observed in the high SES group (phonological awareness, p = 0.006;
tion of sentences of increasing length), visual attention with a cancel- forward digit span, p = 0.001 and RAN p = 0.005).
ation task in which the number of 35 bells found on a page in 2 min As previously noted in numerous studies (Hackman and Farah,
was measured, and with a task in which children detected the differ- 2009; Noble et al., 2007), low SES children showed a lower verbal
ences in 20 pairs of 3 to 5 letters strings. Vocabulary level was deter- performance than their higher SES peers, as reflected here by a
mined with DEN48, consisting in the number (on 48) of pictures weaker verbal IQ, smaller vocabulary and smaller sentence span.
correctly named (Jambaqué and Dellatolas, 2000). Rapid automatic However, none of these scores made this main effect of SES interact
naming (RAN) for pictures was measured. Finally, IQ was estimated with the reading group factor (ps > 0.2). Post-hoc comparisons re-
using the verbal and perceptual subtests of Wechsler's WISC III or IV. stricted to normal readers showed that vocabulary and sentences
Parents were asked to complete a questionnaire about their edu- span were significantly lower in low relative to high SES children
cation and employment level as well as the presence of a second (ps b .01) and that their reading abilities, although in the normal
language spoken at home and reading habits (number of books at range were less developed than their counterparts with higher
home and number of books owned by the child). Children completed SES (− 1 vs + 11.8 months; F(1, 21 ) = 5.88 ,p = 0.024). Note that
another questionnaire concerning their own reading habits, family there was no difference in phonological awareness (F(1,21) b1). In
life and other social and cultural activities. For each child, a composite dyslexics, reading scores were significantly lower in low compared
SES score was determined. This score took into account the school to high SES children (−38.25 vs −30.8 months; F(1, 21 ) = 10.211,
location as a surrogate for neighborhood SES, and the parent's level p = 0.004) but no other verbal difference related to SES was significant
of education because a previous study undertaken in the city of except for vocabulary (p = 0.047): low SES dyslexics had a smaller
Paris demonstrates that these two factors were the most predictive lexical repertoire than their high SES peers.
environmental variables of the reading score (Fluss et al., 2009). Many of the low SES children, although born in France and provided
We added also bilingualism at home, which was present in almost with their entire schooling in France (at least 4 years in our groups),
all families coming from the educational priority zone (but was were also exposed to another language at home. Our groups were too
unconfounded with dyslexia). small to disentangle SES from bilingualism. However, Noble et al.
The children were sorted into four groups according to their SES (2007) studying a group of 168 first-graders reported that SES rather
background and reading scores, as shown in Table 1, and each score than exposure to a second language, was the pertinent variable to
was submitted to a 2 × 2 ANOVA with factors of SES and reading explain reading competencies. Three distinct studies, by Frederickson
group. Dyslexic children were slightly older than normal readers. As and Frith (1998), Fluss et al. (2009), and Lesaux and Siegel (2003), all
can be seen in Table 1, the profile of impairment of the dyslexic chil- note that bilingualism does not have any negative consequences
dren in the low SES group was very similar to that of the high SES on reading per se: even within migrant bilingual populations, what
dyslexia group as proven by the absence of any significant reading matters to reading difficulties is phonological competence in the target
by SES interaction whenever a main effect of the reading group factor language. In fact, Frederickson and Frith (1998) note that, beyond two
 K. Monzalvo et al. / NeuroImage 61 (2012) 258–274 261

years of exposure to English, mono and bilingual populations exhibited each child was trained in a mock MRI scanner: the real experiment
similar phonological competence. Here too, phonological awareness did was simulated using recorded MRI noises and shortened versions of
not differ between the low-SES, primarily bilingual children, versus the the two different types of functional runs, and the child was trained
high-SES, primarily monolingual ones (p= 0.44). to remain absolutely still.

Stimuli and task MR imaging parameters

Magnetic resonance was performed on a Siemens Tim Trio 3.0
Two functional experiments, one assessing the organisation of the Tesla scanner. Stimulus presentation and behavioral responses collec-
visual ventral pathway, the second assessing the spoken language tion were done with E-prime1. Children were protected with noise-
network, were performed. protection ear-phones and a mirror system above the child's head
allowed them to see the visual stimuli presented on a screen at the
1) Visual experiment
end of the tunnel. T1 images were acquired for anatomical reference.
Four categories of visual stimuli (houses, faces, words and a For functional imaging, 84 EPI volumes (TR = 2.4 s, TE = 30 ms matrix
revolving checkerboard) were presented in a block paradigm. 30 64 × 64 × 40, voxel size = 3 × 3 × 3mm) were acquired for each visual
different black and white pictures of unknown people and places run and each 80 volumes for each auditory run. A diffusion tensor
and 30 four-letter regular French words were used. The words were magnetic resonance imaging was finally acquired but these data are
frequent regular words encountered by young readers, as specified not presented here. During the structural acquisition, children were
in Manulex, a lexical database compiling the frequency of occurrence looking at cartoons. To reduce head motion, the quality of the MRI
of words in 54 French scholar reading books (Lété et al., 2004). images was checked after each acquisition and feed-back was given
10 pairs of different images of the same category (200 ms presen- to the child.
tation for the first picture/word, 200 ms inter-stimulus, 500 ms pre-
sentation for the second picture/word) were successively presented Statistical analyses
(inter-pairs interval: 600 ms) followed by a cross fixation image Preprocessing and analyses of the data were done using SPM5.
during 10.5 s (total bloc duration 28.5 s). The stimuli were projected Images were first realigned to the first functional image. During the
onto a translucid screen, 640 pixel wide (VGA mode), subtending functional visual runs, the children moved in an average of 0.9 mm
42.5 cm width and viewed through a mirror from a distance of in translation and 1° in rotation (1.2 mm and 1.8° for the auditory
125 cm, for an overall angular size of 19.3°. For an entire block, runs). There was no group difference in the auditory experiment. In
images in a pair were either repeated or different, with a size-factor the visual experiment, the low-SES kids moved significantly more
difference for pictures, and an upper-lower case transformation for than the high-SES group (mean maximal translation amplitude in
words (the goal was to study repetition effects, but this manipulation the x direction: 0.5 vs 0.6 mm, F(1,42) = 17, p b .001; z direction: 1.4
gave no significant results and therefore is not reported further). For vs 3.2 mm, F(1,42) = 12, p = .001; mean maximal rotation angle: 1.2
the checkerboard category, a round black and white checkerboard vs 1.5 d°, F(1,42) = 13, p b .001). The most restless group was the
picture was smoothly rotated during 18 s followed by a cross fixation normal-readers low-SES group. When the child's movement during
image during 10.5 s. In each block, two stars were randomly inserted a run was superior to 2 mm in translation and/or 1° in rotation, all
during the visual stimuli presentation, either on the right or the left functional images were visually screened to detect volumes with
side of the screen. The child was instructed to press the left button movement present during the volume acquisition detected as hyper-
as soon as detected. This incidental task was planned to keep the or hypo-intense slices in the image. Artrepair5, an SPM5 toolbox
child's attention toward the visual stimuli. In a run, two blocks of (Mazaika et al., 2007) was used to further detect outlier volumes in
each category were randomly presented except the checkerboard which the global intensity of the volume was greater than 2.5% of
block that was presented only once. The total run duration was 3′ the global mean during the time-series. Outlier volumes were
21″ and the children saw 4 runs. replaced by interpolation between the preceding and following cor-
rect images. This procedure was applied to 7 children for the visual
2) Language experiment
runs (in the low-SES groups: 3 normal readers and 2 dyslexics, in
40 short sentences in French (native language) and Japanese the high-SES groups: one child in each reading group), and to 13 chil-
(a foreign language that none of the children understood) were ran- dren for the auditory runs (in the low-SES groups: 4 normal readers
domly presented every 12 s in a slow-event design. Each sentence and 5 dyslexics, in the high-SES groups: two in each reading group).
was repeated one time in a row to study repetition effect (Dehaene- Images were then coregistered with the individual anatomical
Lambertz et al., 2006a). Children were asked to press a right bottom image, normalized toward the adult MNI brain space, and spatially
at the end of the sentence. This incidental task was planned to keep smoothed using a 5 mm Gaussian kernel. The data were modeled,
the child's attention toward the auditory stimuli. For comparison within each fMRI run and for each experiment, using the canonical
purposes, we presented the same set of French sentences previously SPM hemodynamic response function and its time derivative con-
used in infants (Dehaene-Lambertz et al., 2006b), and thus produced volved with the experimental conditions (5 in the visual experiment:
by a native female speaker with a highly intonated voice. The Japanese Words, Faces, Houses, Checkerboard and Targets; 4 in the auditory
sentences were produced with the same characteristics. The mean experiment: French first and second presentation of the sentences,
sentence duration was similar in both languages (2707 vs 2724 ms). Japanese first and second presentation of the sentences). The 6
Each functional run comprised 4 different sentences in each language, movement parameters were entered as regressors of non-interest.
repeated one time for a total duration of 3′12″ (16 sentences). Children For testing left-right differences, we computed the spatial transfor-
listened to 4 runs. mation appropriate in each child to align his/her flipped anatomical
image to the MNI template. We then applied this transformation to
Procedure each original contrast image, and computed a left-right difference
image by subtracting the flipped from the contrast image. These
Before MRI acquisition, the children's reading performance for images were submitted to the same ANOVA model as with the origi-
all the words presented during the fMRI experiment was assessed. nal fMRI images, in order to estimate the hemispheric differences for
Children were instructed to read as fast and accurately as possible each effect of interest.
the words presented on a computer screen either in upper or in A second-level group ANOVA was performed with a between-
lower case. Words appeared 500 ms after the vocal response. Then factor of group (4 levels, defined by the combinations of impaired
262 K. Monzalvo et al. / NeuroImage 61 (2012) 258–274

versus normal readers, and normal versus low SES), and within- Results
subject factors of experimental conditions, using the individual
contrast images smoothed with a 8 mm Gaussian kernel. To better un- 1) Visual experiment
derstand the effect of reading on brain activations, we also computed
analyses with reading performance entered as a regressor. All reading Behavioral results
performances being correlated, we choose the advance or delay of the
reading age relative to the chronological age on the “Alouette” test Before the MRI experiment, we tested reading performance for the
(hereafter called “reading lag regressor”) as representative of the level words used in the paradigm. The mean reading time was significantly
of reading proficiency attained. In independent fMRI analyses, we also faster (894 vs 2506 ms: F(1,42) = 22.4, p b .001) and the percentage
performed regressions with the other behavioral performance mea- of errors lower (1 vs 7% F(1,42) = 32, p b .001) in normal-readers
sures that predict reading proficiency such as RAN, metaphonological compared to dyslexic children confirming the reading impairment.
performance, and memory spans. Reading time was also significantly correlated with the children's
Unless otherwise specified, significant results are reported when reading level as measured with standardized tests (e.g. with the
voxels were significant at p b 0.001 and formed a contiguous cluster number of words read in one minute: R² = 0.47, p b .001). Reading
whose extent was significant at p b 0.05, corrected for multiple com- time was not affected by SES (Fb 1) but the low SES children tended
parisons across the entire brain volume (the cluster-level corrected to make more reading errors (12 vs 7% F(1,42) = 3.2, p = 0.07). The
p value is denoted as pc_cor). Occasionally, we report focussed ana- SES× reading interaction was not significant, neither for reading time
lyses where the search for significant voxels was restricted to a pre- nor for errors.
specified region of interested, e.g. the ventral occipito-temporal Within the scanner, no difference between groups nor interaction
cortex. was observed in the RT to the target star (normal readers: 713 ms;
For the visual experiment, we first examined the ventral maps of dyslexics: 716 ms) confirming that all subjects were similarly engaged
preferential activations to one category relative to all others. While in the task.
these contrasts are useful in order to produce figures of categorical
preference in ventral visual cortex, for group comparison purposes, FMRI results
they do not properly test each category independently of the others.
For instance, in the faces-others contrast, a superior response in nor- Neither the main effect of SES nor the interactions of SES with
mal readers relative to dyslexics might indicate a greater response to reading were significant in any analyses of the visual runs. We thus
faces, or a lower response to one or more of the categories mixed present the results with both SES groups merged.
within the ‘other’ group, e.g. words. To perform truly independent
tests, we thus first compared normal readers and dyslexics on the Group analysis
contrasts of each visual category relative to the blank screen. Second, We first examined the mosaic of ventral visual activations to each
since we were specifically interested in testing, independently, category relative to all others. As shown in Fig. 1, the different catego-
whether dyslexia affects fMRI response to words and to faces, we con- ries elicited activations localized successively along a medial–lateral
trasted each of these categories to the two remaining ones (houses axis from checkerboard in the primary visual and surrounding areas,
and checkerboards). lined externally with houses then faces. In the left hemisphere, only in
For the auditory experiment, we compared the two groups on normal readers, a response to words bordered externally the response
their responses to all stimuli versus silence, native versus foreign to faces (Table 2).
language, and first versus second sentence presentation. We again We compared the two groups on the responses to each visual
complemented the categorical distinction between poor and good category relative to the blank screen (see Methods). No difference be-
readers by a continuous regressor (reading lag). We characterized tween groups was observed for checkerboards or for houses minus
the role of the significant clusters by intersecting these analyses blank. The words minus blank and the faces minus blank comparisons
first with a mask of the native language network (Native–Foreign identified respectively a left and a right cluster more activated in con-
language contrast computed across all children and thresholded trols than in dyslexics in the fusiform regions. These clusters were not
at voxel p b .05) and second, with a mask of the repetition suppres- significant when corrected for repeated measures across the whole
sion effect in the native language (First–Second presentation of a brain, but became significant when the analyses were restricted to a
native sentence computed across all children and thresholded at mask comprising the fusiform and inferior temporal gyri based on the
p b .05). AAL atlas (Tzourio-Mazoyer et al., 2002): We observed a difference
Finally we explored the correlations between the activation to normal readers > dyslexics both for words minus rest ([−45 -45 -15],
written words in the visual experiment and the speech responses z = 4.3, 25 vox, pc_cor = .048) and for faces minus rest ([39 -42 -15],
in the auditory experiment. We first probed the relations be- z = 3.9, 57 vox, pc_cor = .007).
tween written and spoken language networks in young readers Whole brain asymmetry analyses revealed only one significant
by intersecting the contrast Words versus others categories dur- difference between the two groups in the lateralization of the re-
ing the visual runs, and the contrast native versus foreign lan- sponses to faces (Table 3 and Fig. 2). The response to faces minus
guage during the spoken runs. Second, we examined whether a blank was significantly more right-lateralized in normal than in im-
cortical marker of inter-individual variability in reading achieve- paired readers ([42 -39 -18], z = 4.22, 49 vox, pc_cor = .012). No differ-
ment, the level of activation of the VWFA to written words, was ence of lateralization between groups was observed for the other
correlated with the fMRI activation evoked by spoken language— conditions (houses, checkerboards, and even words minus blank).
and, crucially, whether this correlation with VWFA activation To examine whether normal readers and dyslexics specifically dif-
arose from the same brain regions in controls and in dyslexics. To fered within ventral visual regions specialized for words and for faces,
this end, we extracted, for each subject, the activation of the five we also examined the responses to words and faces relative to the
best voxels responding more to words than to houses and checker- mean of houses and checkerboard stimuli, with small volume correc-
boards within the left fusiform region, and entered this activation as tion over the bilateral fusiform and inferior temporal gyri. For the
a regressor of interest in the speech experiment. To check for differ- Word minus [House,Checkerboard] contrast, a significant difference
ences between groups, we entered this variable separately for each favoring normal readers over dyslexics was observed around the peak
reading group, and used SPM to test whether the slope of the regression of the VWFA ([−42 −45 −15], z = 4.6, 46 voxels, small-volume
differed in the two groups. pc_cor = .038). For the Face minus [House,Checkerboard] contrast, a
 K. Monzalvo et al. / NeuroImage 61 (2012) 258–274 263

z=-12 Normal Readers

color
x=-48 y=-60
legend:
Words NR > Dysl
L R
L R
Dyslexics

L R

L R
L R

Normal Readers
fMRI signal change
Dyslexics
(arbitrary units)
16

Left 8
PPA
VWFA
hemisphere

16 FFA

PPA
Right
8
hemisphere

0
-80 -37
antero-posterior
location (mm)

Fig. 1. Mosaic of preference for different visual categories in ventral visual cortex: the three orientation slices on top show the activation differences between a given category and
all the others in normal readers (NR) and dyslexics (Dysl) (voxel p b 0.001, cluster p b 0.05 corrected for multiple comparison at the whole-brain level). Graphs below shows the
fMRI signal change relative to rest in both groups in successive spheres of 5 mm regularly spaced along an anterior–posterior (x axis, − 80 to − 37 mm) and a medial–lateral
axis (color code on the coronal slice: ± 27 to ± 50 mm). Note the differences between normal (plain lines) and impaired readers (dotted lines) in their response to words
(red lines) on the left side, but also to faces (blue lines) on the right side, whereas the responses to houses (green lines) and checkerboard (cyan lines) are similar in the two groups.

significantly greater activation in normal readers relative to dyslexics normal) (see Fig. 3). In the analyses of words minus blank, reading
was observed in a right cluster around the peak of the FFA ([39–51 scores were well predicted by fMRI activation only in the left VWFA
−15], z = 4.1, 15 voxels, small-volume pc_cor = .026). No suprathres- ([− 42 −45 −18], z = 5.34, 85 vox, and a similar correlation was
hold cluster was detected in the converse direction (dyslexics> normal found when analyzing the Word minus [House,Checkerboard] con-
readers). No significant lateralization differences were observed with trast ([−42 − 48 −15], z = 5.64, 85 vox, pc_cor = .022). For faces
these contrasts. minus blank, the regression with reading scores identified a right clus-
While the above analyses were based on a categorical distinction ter within the FFA ([45–57 −18], z = 4.12, 88 vox, pc_cor = .020), and a
between normal readers and dyslexics, they were also replicated, second more dorsal and posterior cluster putatively corresponding to
with slightly increased significance, when replacing the binary the occipital face area (OFA; [21–81 15], z = 3.98, 74 vox,
group distinction by a continuous variable of reading lag for each pc_cor = .039). When using the Face minus [House,Checkerboard] con-
child (where a negative value indicates retardation relative to trast, only the anterior FFA cluster remained ([39–51 −12], z = 4.34,
264 K. Monzalvo et al. / NeuroImage 61 (2012) 258–274

Table 2
Regions of significant activations for each visual category vs all others in normal readers and dyslexics groups.

Area MNI coordinates SPM t-tests

x y z Cluster-level No. of voxels Z value at local

P value (corrected) in cluster maximum

Normal readers Checkerboard > others Right calcarine sulcus 9 − 75 −3 b.001 1807 Inf
Left calcarine sulcus −9 − 75 0 Inf
Left cuneus −3 − 84 39 5.27
Right cuneus 6 − 84 39 4.23
Houses > others Right fusiform gyrus 30 − 45 −9 b.001 883 Inf
30 − 63 −6 5.95
Right middle occipital gyrus 33 − 81 9 6.25
Left fusiform gyrus − 27 − 48 −9 b.001 693 7.75
Left middle occipital gyrus − 27 − 90 9 5.95
Left inferior occipital gyrus − 24 − 84 −9 5.77
Faces > others Right fusiform gyrus 39 − 51 − 15 b.001 722 Inf
Right Inferior occipital gyrus 42 − 81 −9 6.5
Left fusiform gyrus − 39 − 60 − 15 b.001 457 6.42
Left inferior occipital gyrus − 39 − 84 −3 6.31
Words > others Left Inferior temporal gyrus − 51 − 57 − 12 b.001 384 5.71
Left fusiform gyrus − 45 − 48 − 15 5.49
Left superior temporal sulcus − 54 − 33 3 5.59
Left Precentral − 45 9 30 0.006 115 4.31
Impaired readers Checkerboard > others Right calcarine sulcus 9 − 75 0 b.001 1856 Inf
Left calcarine sulcus −6 − 81 0 Inf
Right cuneus 12 − 87 15 7.39
Left cuneus − 12 − 90 15 6.89
Houses > others Right fusiform gyrus 30 − 48 −9 b.001 1288 Inf
27 − 81 −9 6.76
Right middle occipital gyrus 36 84 12 6.6
Left fusiform gyrus − 27 − 48 −9 b.001 1103 7.32
Left middle occipital gyrus − 30 − 90 12 6.19
Left inferior occipital gyrus − 24 − 84 −9 6.09
Faces > others Right Inferior occipital gyrus 42 − 78 −9 b.001 363 7.04
Right fusiform gyrus 39 − 48 − 21 6.12
Left fusiform gyrus − 39 − 48 − 18 b.001 336 6.02
Left inferior occipital gyrus − 39 − 84 − 12 5.72
Words > others No suprathreshold clusters

42 vox, pc_uncor = .024, small-volume pc_cor = .045). No correlation metaphonological awareness, rapid automatic naming, digit and sen-
with reading score was observed in the fMRI activation to houses or tence spans).
checkerboards.
Similar results were observed when the regressor was the number Individual subject analyses
of words read in one minute. No significant area was identified by SPM group analyses leave open the possibility of a selective but
regression with the other behavioral measures (vocabulary size, spatially more variable response to words and faces in dyslexics

Table 3
Regions of significant asymmetries for each visual category vs blank in normal readers and dyslexics groups.

Area MNI coordinates SPM t-tests

x y z Cluster-level No. of voxels Z value at local

p value (corrected) in cluster maximum

Normal readers Houses vs blank Right middle occipital gyrus 33 − 78 9 b.001 608 Inf
Right parahyppocampal gyrus 33 − 33 − 15 5.28
Right fusiform gyrus 36 − 45 − 18 5.14
Faces vs blank Right fusiform gyrus 48 − 57 − 15 b.001 453 7.35
Right fusiform gyrus 42 − 39 − 21 5.66
Right middle occipital gyrus 33 − 81 9 b.001 189 5.66
Right lingual gyrus 24 − 87 −9 4.65
Words vs blank Right lingual gyrus 24 − 87 −9 0.005 58 4.57
Left precentral − 39 0 33 b.001 113 4.17
Impaired readers Checkerboard vs blank Right lingual gyrus 9 − 72 0 b.001 118 4.90
Right middle occipital gyrus 36 − 81 3 4.81
Houses vs blank Right middle occipital gyrus 39 − 78 9 b.001 336 7.31
Right fusiform gyrus 33 − 48 −9 0.001 73 5.73
Faces vs blank Right inferior occipital gyrus 33 − 90 9 b.001 203 6.90
Right lingual gyrus 21 − 87 −3 6.22
Words vs blank Right middle occipital gyrus 36 − 87 9 b.001 86 4.89
Normal > Impaired readers Checkerboard vs blank No suprathreshold clusters
Houses vs blank No suprathreshold clusters
Faces vs blank Right fusiform Gyrus 42 − 39 − 18 0.012 49 4.22
Right inferior temporal gyrus 48 − 54 − 15 4.19
Words vs blank No suprathreshold clusters
 K. Monzalvo et al. / NeuroImage 61 (2012) 258–274 265

Fig. 2. Hemispheric asymmetry in functional activation pattterns. Left, analyses of each category relative to rest (voxel p b 0.001, cluster p b 0.05 corrected for multiple comparison at
the whole-brain level). Right, analysis of words and faces relative to the mean of houses and Checkerboards (Ck). In each case, images show regions where activation was larger in
the left hemisphere than in the right (left-hemisphere voxels) or vice-versa (right-hemisphere voxels), superimposed over axial and sagittal slices of an individual child. The
threshold was lowered to p b .01 for images in the bottom right row, as no difference between normal and impaired readers was present at p b .001.

than in normal readers, which would explain the weaker activation position (4 levels) and condition (3 levels corresponding to House-
found in dyslexics at the group level. We thus searched, within each Checkerboard, Face-Checkerboard and Words-Checkerboard contrast)
child, for the location of the most responsive voxel to Words versus as within-subject factors.
[Houses,Checkerboards] and, separately, to Faces versus [Houses, If categorical specificity varies along the medial-to-lateral axis,
Checkerboards], within 10 mm of the left VWFA and right FFA peaks then a significant interaction of Categories × Lateral position should
obtained across all children (similar results were obtained with be found. Such was the case at each of the eight anterior–posterior
20 mm). For words, the Euclidean distance between each individual y coordinates (−80 to − 37 mm) and within each hemisphere (all
peak and the group mean did not differ significantly between the ps b .001). To determine if this gradient of categorical preferences
two groups (7.98 vs 7.96 mm; F(1,44) b 1). Thus, the spatial disper- differed between normal readers and dyslexics, we then probed the
sion of the responses was not greater in dyslexics. Furthermore, triple interaction of Group × Categories × Lateral Position. This triple
even after searching for the best voxel in the vicinity of the VWFA, interaction reached significance only in the left hemisphere at four
the amplitude of the peak activation to words remained systematical- y coordinates extending from −65.5 to − 44.5 mm (respectively
ly correlated with reading proficiency (p b .0001). Similarly, in the p = .02, 001, b.001, .021). This effect was primarily due to a reduced
vicinity of the right FFA, although the Euclidean distance between response to words in dyslexic children around the x coordinate of
each individual peak and the mean peak tended to be more important the VWFA (see Fig. 1). Nevertheless, even after removing the word
in the dyslexic group (6.4 vs 7.8 mm), this effect did not quite reach condition, a triple interaction was still observed at y = − 37 mm in
significance (F(1,44) = 3.54; p = .07), and the amplitude of the peak the left hemisphere (F(6,132) = 3.69 p = 0.014). Post-hoc analyses
activation to faces remained positively correlated with reading profi- revealed that this effect was due to stronger activation to faces in
ciency (p= .046). These analyses suggest that our results could not be the innermost sphere at x = −27 mm, (F(1,44) = 5.22,p = 0.027)
imputed to a group difference in inter-individual variability: reading in normal readers relative to dyslexics. Thus, besides their reduced
proficiency still affected the amplitude of the response to words in the activation to words, dyslexics had an essentially similar medial–
left fusiform, as well as the amplitude of the face response in the right lateral gradient than normal readers, but the face response extended
fusiform, even when the best voxel was selected in each child. slightly more medially in normal than in impaired readers in the left
hemisphere.

Regions of interest 2) Spoken language experiment

Finally, we probed whether the antero-posterior and medial–
lateral organization of the extrastriate maps was similar in the two Behavioral results
groups. To this aim, we averaged activations for the three contrasts
House, Face and Words minus Checkerboard within 5 mm spheres As expected, the detection of the end of the sentence was faster for
regularly spaced along the anterior–posterior axis of the ventral re- the second presentation of the same sentence (573 vs 739 ms:
gions (eight y coordinates, ranging from − 80 to −37 mm) at each F(1,45) = 139, p b .001). This facilitation with repetition tended to be
of four medial–lateral x locations (±27 to ±50 mm) on each side. stronger for the foreign than for the native language (− 190 vs
These regions corresponded to the pattern of activations for houses, −143 ms: F(1,45) = 4, p = .051). There was no difference between
faces and words when all children were considered. We submitted normal-readers and dyslexics (F(1,42) b 1), but the low SES children
them to separate ANOVAs at each of the 8 y positions, with Group tended to be slower (main effect of SES: F(1,42) = 3.98, p = 053).
as a between-subjects factor and Hemisphere (left, right), lateral The latter effect was due to a few very slow subjects. Four children
266 K. Monzalvo et al. / NeuroImage 61 (2012) 258–274

Activation to FMRI results

Faces Words
Both the dyslexic and the normal-readers groups presented a
Right Face peak bilateral response to speech stimuli, with significantly stronger
responses to the native than to the foreign language, as well as to the
[42 -54 -18] first rather than the second presentation of the sentence (Table 4). In
16
Faces both groups, the network of native-language regions identified by
the native minus foreign language subtraction was significantly asym-
metric along the STS and inferior frontal areas.
When the responses to all stimuli vs silence were considered,
8 normal readers showed significant stronger activations than impaired
readers in both SMAs ([0 6 66], z = 6.07, 263 vox, pc_cor = .003) and in
the right temporal region ([48 0 51], z = 5.84, 1025 vox, pc_cor b .001).
Two marginally significant clusters were also observed on the left
0 side, in the planum temporale ([−39 − 39 18], z = 5.33, 118 vox,
F H W Ck Target
pc_cor = .059) and in the insula ([− 48 6–6], z = 5.23, 123 vox,
pc_cor = .052). The converse comparison (dyslexic > normal readers)
isolated two occipital clusters, due to a lesser deactivation in dyslexic
Left Face peak relative to normal readers ([− 18 −81 30] z = 4.58, 157 vox, pc_cor =
10 [-39 -51 -18] .025 and [− 45 −72 0] z = 4.24, 130 vox, pc_cor = .045).
The two groups also differed in their activation asymmetries dif-
Words ferences: normal readers were significantly more right lateralized
than dyslexics in Heschl's gyrus ([54–12 12], z = 7.57, 230 vox,
5 pc_cor b .001) and in a posterior region going from the superior tem-
poral gyrus till the supramarginal gyrus ([66–36 12], z = 5.48, 64
vox, pc_cor = .037). These two clusters surrounded the left planum
0 temporale cluster observed in the previous analysis, which tended
F H W Ck Target to be more activated on the left than the right in normal readers
r2=0.34,
than in dyslexic ([− 42 −39 18], z = 4.75, 33 vox, pc_uncor = .027).
p=0.0002
Last but not least, controls showed a greater left-hemispheric lateral-
ization than dyslexics in a left ventral temporal cluster coinciding
Left Word peak with the basal language area and extending posteriorly towards
10 [-42 -48 -15] the VWFA (controls > dyslexics: [− 42 −30 −21], z = 5.25, 286 vox,
pc_cor b .001). Indeed, this cluster of abnormal lateralization to speech
Words in dyslexia overlapped significantly with the cluster of reduced activa-
tion to written words in dyslexics reported above (intersection = 21
5 voxels, peak coordinates at [−39 −51 −12] and [− 42 − 42 − 15]).
Analyses where the binary distinction between normal-readers
and dyslexics was replaced by a continuous reading-lag regressor
confirmed a relation between reading lag and activations to speech
0 in the same clusters: SMAs, right precentral, right posterior temporal
F H W Ck Target
region and three clusters in the left hemisphere: planum temporale
Word Reading lag
([− 42 −39 18], z = 4.96 and [−45 −51 12], z = 3.88, 126 vox,
(months)
pc_cor = .048), insula ([− 48 6–6], z = 5.59, 239 vox, pc_cor = .005)
and ventral inferior temporal cortex, corresponding to the basal tem-
Reading: Normal SES Low SES
poral language area and again extending posteriorily to a point just
Normal R anterior to the VWFA ([−45 −27 −21], z = 4.60 to [−36 −45 −27],
z = 4.23, 137 vox, pc_cor = .037). To clarify the role of these regions in
Dyslexics
speech processing, we first intersected this image with a mask of the
regions significantly more active for the native than for the foreign
Fig. 3. Correlation between reading performance and responses to words and faces.
language (pb .05) across all children. Within this mask, clusters signifi-
Left: Mean percentage of signal change in the four groups at the peak of the face and
the word responses minus [Houses-Checkerboard]. Note that the coordinates were cantly more activated in controls were now limited to the inferior tem-
determined by a contrast orthogonal to reading score (peak computed across all poral cluster just anterior to the VWFA (same coordinates as above, 73
subjects). Right: Scatter plots illustrating the correlation between reading performance vox, pc_cor = .026), the left planum temporale ([−42 −42 18], z = 4.59,
and the fMRI response to faces and words at these peaks. In the left hemisphere, the 49 vox, pc_cor = .053) and the right posterior STG ([60–33 9], z = 5.77,
response to words develops laterally to faces and increases with reading performance
(bottom rows). In the right hemisphere, the response to faces also increases with
48 vox, pc_cor = .054). Second, when we intersected the initial contrast
reading performance (top row). Arrows indicate the condition which is significantly image with a mask of the regions significantly more active for the
correlated with reading performance at each location. first than for the second presentation of the native sentences (pb .05)
across all children, the only remaining significant region was the cluster
located in the inferior temporal cortex (same coordinates as above, 103
vox, pc_cor = .016). These effects were nevertheless weak, as no signifi-
(three low-SES normals and one high-SES dyslexic) had a mean reac- cant difference between groups was observed, in either direction, on
tion time of 1340 ms whereas the mean reaction of the remaining the native versus foreign contrast, and on the first versus second pre-
children was 590 ms (±174 ms). When these four children were sentation contrast computed across the whole brain (Tables 4 and 5).
removed, neither main effect of reading, SES, nor any interaction Contrary to the visual experiment, several areas also showed sig-
between reading, SES, repetition and language were present. nificant correlations with other behavioral variables for the speech
 K. Monzalvo et al. / NeuroImage 61 (2012) 258–274 267

Table 4
Regions of significant activations in the linguistic network in normal and impaired readers.

Area SPM t-tests

x y z Clusterlevel p value No. of voxels in Z value at local

(corrected) cluster maximum

Normal readers Native vs foreign language Left anterior STS − 57 −6 − 12 b.001 1477 Inf
Left posterior STS − 48 − 39 3 Inf
Left temporal pole − 48 15 − 21 7.02
Left IFG − 57 30 9 5.84
Right anterior STS 57 −6 − 12 b.001 530 6.28
Right temporal pole 51 18 − 21 5.82
Right posterior STS 45 − 33 0 5.50
Foreign vs native language Right inferior parietal lobe 54 − 57 36 0.054 125 4.35
First > repeated Left anterior STS − 57 −6 −9 b.001 604 6.63
Left posterior STS − 51 − 42 3 4.31
Right anterior STS 57 −3 − 12 b.001 513 6.56
Right posterior STS 48 − 36 3 4.14
Repeated > first No suprathreshold clusters
Impaired readers Native vs foreign language Left temporal pole − 54 6 − 18 b.001 1125 7.60
Left posterior STS − 54 − 42 6 7.51
Right temporal pole 48 15 − 24 b.001 532 6.51
Foreign vs native No suprathreshold clusters
First > repeated Left anterior superior temporal sulcus − 51 −9 − 12 0.012 198 5.25
Right anterior middle temporal gyrus 57 9 − 15 0.015 188 3.34
Repeated > first no suprathreshold clusters
Normal vs impaired readers and impaired vs normal readers/all contrasts: no suprathreshold clusters

versus silence contrast. Regression with vocabulary level isolated observed for sentence span. Thus the planum was the main structure
significant clusters in the left planum temporale ([− 39 −36 12], whose activation correlated with several behavioral performance
z = 5.66, 143 vox, pc_cor = .032) and in the left ventral region adjacent measures, with a partial segregation into subregions differently sensi-
to the VWFA ([− 33 − 42 − 45], z = 5.09, 118 vox, pc_cor = .057). Both tive to the different competences targeted by these measures.
planum temporale (Right: [57–12 9], z = 6.67, 1475 vox, pc_cor b .001; Finally, we probed the effect of SES on spoken language proces-
Left: [−54 −24 9], z = 5.78, 343 vox, pc_cor = .001) and bilateral sing. On the speech minus silence contrast, high-SES children showed
SMA ([3 6 63], z = 5.08, 158 vox, pc_cor = .024) were correlated with stronger activation than low-SES children in the right superior
metaphonological performance. The left planum region related to temporal gyrus encompassing planum temporale and Heschl's gyrus,
metaphonological performance was anterior and lateral to the cluster bilateral putamen extending on the right side into the right insula,
correlated with vocabulary level and reading lag. Finally, forward and and right middle frontal gyrus (Table 6). None of these clusters were
backward digit span performance was again correlated with planum part of the native-language network, as they disappeared when the
clusters (Right: [48–18 6], z = 5.18, 342 vox, pc_cor = .001; Left: analysis was restricted to regions showing more activation to native
[−45 −33 12], z = 5.71, 192 vox, pc_cor = .012) and a right precentral than to foreign sentences. Furthermore, no effect of SES was observed
cluster ( [30 0 51], z = 4.57, 137 vox, pc_cor = 0.039). These clusters for the native language versus silence contrast. By contrast, the same
were embedded in the clusters correlated with metaphonological clusters plus the left SMA were observed for the foreign language
performance. Similar but smaller clusters than for digit span were versus silence contrast (Table 6). In all these analyses, the SES effect

Table 5
Regions of significant asymmetries in the linguistic network.

Area MNI coordinates SPM t-tests

x y z Clusterlevel p value No. of voxels Z value at local

(corrected) in cluster maximum

Normal readers Native > foreign language Left posterior middle temporal gyrus − 48 − 45 3 b.001 157 5.26
Left superior frontal gyrus −9 57 27 0.046 59 4.99
Left inferior frontal gyrus − 54 21 18 4.21
First > repeated No suprathreshold clusters
Impaired readers Native > foreign language Left inferior frontal gyrus − 51 27 3 b.001 307 5.31
Left putamen − 18 9 9 4.70
Left posterior middle temporal gyrus − 66 − 30 6 3.23
First > repeated No suprathreshold clusters
Normal vs impaired readers Native > foreign language No suprathreshold clusters
First > repeated No suprathreshold clusters
Low SES Native > foreign Left middle temporal gyrus − 57 − 45 6 b.001 219 5.87
Left inferior frontal gyrus − 48 12 9 b.001 255 4.65
First > repeated No suprathreshold clusters
Normal SES Native > foreign language Left posterior middle temporal gyrus − 51 − 42 3 0.026 69 4.71
Right middle frontal gyrus 30 45 24 0.029 67 4.50
Right superior frontal sulcus 21 36 30 3.61
Left inferior frontal gyrus − 48 24 3 b.001 162 4.48
Left anterior superior temporal sulcus 54 −3 − 12 0.050 57 4.07
First > repeated Right planum temporale 45 − 33 12 0.013 81 5.11
Normal vs low SES Native > foreign No suprathreshold clusters
First > repeated Right planum temporale 42 − 30 9 0.039 62 4.31
268 K. Monzalvo et al. / NeuroImage 61 (2012) 258–274

Table 6
Regions of significant differences between low and high SES groups.

Area SPM t-tests

x y z Clusterlevel No. of voxels Z value at local

p value (corrected) in cluster maximum

High > low SES Speech vs silence Right planum temporale 54 − 27 9 0.003 276 6.81
Right putamen 24 15 3 b.001 496 5.12
Right insula 33 30 3 4.51
Right middle frontal gyrus 48 45 6 3.66
Left putamen − 21 9 6 0.010 200 4.04
Native language vs silence No suprathreshold clusters
Foreign language vs silence Right planum temporale 54 − 27 9 0.041 134 5.63
Right putamen 24 15 3 b.001 440 5.21
Right insula 30 27 6 4.29
Left putamen − 24 9 6 0.002 301 4.63
Left SMA −6 − 18 60 0.041 134 4.37
Low > high SES No suprathreshold clusters for any of the above contrasts

remained similar when analyses were restricted to the dyslexic spoken and written language at this age. To this aim, we intersected
children, but disappeared when restricted to the normal readers the contrast “words versus others categories” during the visual runs,
group. It was not related to the children's lexical level (which was one and the contrast “native versus foreign language” during the spoken
of the main behavioral differences between the two SES groups), as runs (Fig. 5). This analysis was performed only in normal readers, be-
the same clusters remained significant when lexical level was entered cause there were no significant voxels for the visual comparison in
as a regressor of non-interest in the analyses. As shown in Fig. 4, at dyslexics. It isolated three clusters in the left hemisphere (Fig. 5):
the relevant cortical sites, the high-SES dyslexic group showed similar one in the posterior STS/MTG region ([-48 −42 3]), in the inferior
or even stronger activation than the normal readers group whereas temporal region corresponding to the basal temporal language area
the low-SES dyslexic group showed weaker activation for all speech ([− 42 − 33 −21]) and in the left precentral region ([− 42 15 24]).
conditions. These results are consistent with a reduced activation in Thus, these regions emerge as essential points of connection between
dyslexia that was partially compensated in high-SES children. written and spoken language in young readers, and which fail to ap-
pear in dyslexics.
3) Relations between spoken and written language networks
To clarify the relations between the anomalies observed in visual and
Because the same children participated in the visual and auditory spoken language processing, we also examined whether a cortical mark-
experiments, we could examine which regions were common to er of inter-individual variability in reading achievement, the level of

14 1: [63, -36, 9] 6 2: [51, -3, 48] 3: [4, 6, 66] 11 4: [-48, 6, -6]

8

0
2 -1 0 2
Nat1 Nat2 For1 For2 Nat1 Nat2 For1 For2 Nat1 Nat2 For1 For2 Nat1 Nat2 For1 For2

5: [-42, -39, 18] 6: [-45, -51, 12] 2 7: [-45, -27, -21]

10 6 2.5 8: [-42, -51, -21]
1
0 0

-1
2 0 -2
Nat1 Nat2 For1 For2 Nat1 Nat2 For1 For2 Nat1 Nat2 For1 For2 Nat1 Nat2 For1 For2

Reading: Normal SES Low SES

Normal R
Dyslexics

Fig. 4. Brain regions where activations to speech, relative to silence, are correlated with reading lag (voxel p b 0.001, cluster p b 0.05 corrected for multiple comparison at the whole-
brain level). The SPM map is projected on a sagittal slice of an individual child. Plots present the fMRI response in each group for the four conditions at the peak of these clusters
(Nat1 Nat2 = first and second presentation of native sentences, For1 and For2 = first and second presentation of foreign sentences). At peaks 1–5 activation is similar in all condi-
tions, and the responses are globally increased in normal readers relative to impaired readers. By contrast, at peaks 6–8 (posterior STS, basal language area and VWFA), the response
is significantly increased for the first presentation of sentences in the native language, but only in normal readers. Note also at peaks 1 and 3, the effect of SES in the dyslexic groups.
The low-SES dyslexic children have weaker activations than the other groups, in particular the high-SES dyslexics.
 K. Monzalvo et al. / NeuroImage 61 (2012) 258–274 269

A B
-39 30 -40

-42 6 -50

-48 -15 -60

L R
written words Normal readers > Dyslexics
spoken sentences Dyslexics > Normal readers
intersection Common to all children

Fig. 5. Relations between written and spoken language networks. A, written and spoken language in normal readers. The figure shows, in red, the contrast “written words > other
visual categories” and in yellow, the contrast “first spoken sentence in native language > all other spoken sentences” (voxel p b 0.001, cluster p b 0.05 corrected for multiple com-
parison at the whole-brain level) in normal readers. The regions common to both contrasts appear in orange. The map is projected on sagittal (x = − 39/−42/−48) and axial slices
(z = 30/6/−15 MNI space) of an individual child. B, spoken-language predictors of VWFA activation to written words. The figure shows areas where the speech vs silence contrast
was significantly correlated with the VWFA activation to written words (5 best VWFA voxels, individually extracted; voxel p b 0.001, cluster p b 0.05 corrected for multiple compar-
ison at the whole-brain level). The correlation was performed separately for normal readers and dyslexics, and compared across groups. Results are projected on sagittal slices of an
individual child (x = − 40/−50/−60 MNI space).

activation of the VWFA to written words, was correlated with the fMRI Discussion
activation evoked by spoken language—and, crucially, whether this cor-
relation with VWFA activation arose from the same brain regions in nor- In this experiment, we studied two age-homogeneous groups of
mals and in dyslexics. To this end, we extracted, for each subject, the dyslexic and normal readers, between 9 and 11 years, during their
activation of the five best voxels responsive to words within the left fusi- 4th year of reading training. Both groups were further separated
form region. We first entered this activation as a regressor of interest in by their low versus high socio-economic status (SES). Our main
the speech experiment across all children to find a common effect be- goal was to examine whether the ventral visual and superior tempo-
tween reading groups. This identified a single cluster in the left precen- ral anomalies in fMRI activation, which have been frequently but
tral cortex ([−39 −6 36], z =5.50, 165 vox, pc_cor =.021). This cluster not systematically observed in dyslexia, could be replicated with
remained when the analysis was restricted to the normal readers minimally demanding tasks that did not directly tax reading or meta-
([−39 −6 36], z =5.70, 141 vox, pc_cor =.007), but an additional cluster phonological skills. We also questioned to what extent a low-SES
emerged in the left anterior STS ([−63 −15 −3], z= 6.01, 220 vox, modulates the reading deficit.
pc_cor =.006). The same analysis in dyslexics isolated again a left precen- The results revealed impairments in both visual and spoken-
tral cluster, but at a more dorsal location ([−42 −6 60], z= 5.27, 117 language systems, largely independently of SES. In the visual sys-
vox, pc_cor =.046). To test whether these differences were significant, tem, beyond the well-known reduced activation to words, a re-
we entered the individual VWFA regressor separately for each group duced activation to faces was observed in the right fusiform
(Fig. 5). Stronger correlations between written word and spoken lan- response in dyslexics relative to normal readers, resulting in a sig-
guage responses in normal readers than in dyslexics were observed in nificant increase in the asymmetry of face responses toward the
a wide set of regions, predominantly in classical left-hemispheric lan- right hemisphere in normal readers relative to dyslexics. In the
guage areas of the left superior temporal gyrus, extending laterally and spoken language domain, dyslexics showed reduced activation,
towards the ventral temporal cortex ([−54 0–9], z =5.91; [−45 −36 first in speech regions responding similarly to both languages (bi-
12]; z =4.62; [−51 -72 −9], z= 4.02, 529 vox, pc_cor b .001), but also lateral posterior temporal cortex, left insula and supplementary
the right superior temporal region ([51–15 3], z = 6.54; [54 12–24], motor area), second in regions more involved in native language
z =3.90; 339 vox, pc_cor=.001), both vermi ([−3 −54 −42], processing (right STG and left planum temporale, left basal lan-
z =4.78; 300 vox, pc_cor =.002) and finally a cluster in the left insula guage area extending to the VWFA), and third in regions more
that fell short of significance ([−30 39 9], z= 4.59; 117 vox, recruited for the first presentation of a native sentence (left basal
pc_cor =.060). The converse comparison isolated clusters in the left pre- language area extending to the VWFA). A lower SES amplified the
central region ([−42 −9 57], z = 5.05; 124 vox, pc_cor = .051), and impairment seen in dyslexic children, notably in the right hemi-
the left supra-marginal gyrus ([−57 −36 30], z = 4.38; pFDR_cor = sphere (right superior temporal region, insula, and middle frontal
.017, 63 vox, pc_cor n.s.). gyrus).
270 K. Monzalvo et al. / NeuroImage 61 (2012) 258–274

Crucially, these visual and auditory activations were correlated, tasks, MEG signals did not differ between groups before 250 ms. In par-
differently for dyslexic and normal readers. In both groups, the activa- ticular the specific response to faces at 150 ms was unaffected, suggest-
tion to written words in the VWFA was well predicted by the activa- ing a normal occipito-temporal encoding of faces.
tion to speech in the precentral gyrus, but two other sets of regions The most plausible hypothesis therefore is that both script- and
emerged as unique to each group. In normal readers, VWFA activation face-related activation anomalies are an indirect consequence of the
was predicted by the activation of the left superior and lateral tempo- functional reorganization of the ventral visual system required by
ral regions, the left insula, and both vermis, whereas in dyslexics reading acquisition, and which unfolds abnormally in dyslexics.
the predictive regions comprised the left supramarginal gyrus and Development fMRI studies have observed a slow maturation of ven-
upper precentral region (at a site compatible with Exner's area). tral face responses: the peak location of the FFA is reproducible across
This suggests that reading acquisition in dyslexics might rely on a age, culture, SES, language, expertise and training level, but its size in-
different network than in normal children. We now discuss each of creases with age (Aylward et al., 2005; Golarai et al., 2007; Joseph et
these results in turn. al., 2011; Passarotti et al., 2003; Peelen et al., 2009; Scherf et al., 2007,
2011). Similarly, the left fusiform response to written words emerges
Dyslexia and the development of the ventral visual system rapidly around the age of six, the typical age for reading acquisition
(Maurer et al., 2005a, 2010; McCandliss et al., 2003). We propose
Our first goal was to probe the organization of ventral visual cor- that this emerging VWFA competes with the representation of faces
tex in normal and dyslexic children, beyond the categories of strings in the left fusiform, thus partially displacing face responses toward
and letter-like stimuli that were used in most prior studies. In adults, the right fusiform gyrus as well as more mesially in the left fusiform
a mosaic of areas preferentially responds to different categories of gyrus. Indeed, in adults with variable levels of literacy, the right-
visual stimuli, with a medial–lateral gradient (Golarai et al., 2007; hemispheric fusiform response to faces increases with reading scores
Hasson et al., 2003; Haxby et al., 2001; Ishai et al., 2000). Medially, (Dehaene et al., 2010), at a location very similar to the present study.
a region in the parahippocampal gyrus named the parahippocampal This observation suggests that changes in face responses can be a con-
place area (PPA) responds more to buildings and landscapes than sequence of literacy (and therefore of dyslexia). Developmental stud-
to other stimuli. This area is externally bordered by a region in the ies also suggest a process of competition for cortical territory between
fusiform gyrus responding more to faces (fusiform face area, FFA) faces and other visual categories (Cantlon et al., 2011; Joseph et al.,
then by an area in the lateral occipital cortex (LOC) responding 2011).
more to objects than to scramble images. Dehaene et al. (2010) observed that the increase in VWFA re-
Our results indicate that this basic gradient is present in 9-year- sponse to written words with literacy was associated with a slightly
olds and is unaffected in dyslexia. As typical in adult studies, checker- reduced response to faces at the same location, strongly indicative
boards elicited more activation around the calcarine scissure, buildings of a cortical competition process. Here, no such reduction was
in the parahippocampal gyri, and faces more laterally in the fusiform found, as the amplitude of the left-hemispheric faces responses was
gyrus. Furthermore, the dyslexics' responses to checkerboards and to similar in both groups although the topography was slightly medially
houses were strictly normal. For these categories, in both groups, a shifted in normals. This null result might however reflect that
small but significant right lateralization was observed (Figs. 1 and 2), dyslexics, contrary to illiterates, have been trying to learn to read
similar to that seen in adults using the same non-verbal visual stimuli for several years. Although not fully successful, such learning may
(Dehaene et al., 2010). These findings suggest that ventral visual cortex have partially changed the left fusiform region.
is not disorganized as a whole in dyslexia. Note that a displacement of the face response because of a compe-
With written words, we replicated the strong reduction in fMRI tition with words in the left hemisphere is only one of the possible
activation in dyslexics within the left ventral occipito-temporal cor- interpretations for the observed increase in right fusiform face re-
tex, at the site of the VWFA, as now observed in numerous studies sponses in dyslexia. Another factor may be the enhanced efficiency
(Brunswick et al., 1999b; Maisog et al., 2008; McCandliss and Noble, at processing foveal stimuli brought about by reading acquisition.
2003; Richlan et al., 2009; van der Mark et al., 2009 ). Importantly, Visual response enhancements due to literacy have been observed
our findings confirm that this region differs between controls and in occipital and even primary visual cortex (Dehaene et al., 2010;
dyslexics even when reading is implicit and not required by the task Szwed et al., 2011), and such changes may putatively generalize to
(see also Brunswick et al., 1999a, 1999b). other non-reading stimuli that rely heavily on high-resolution foveal
More surprisingly perhaps, we discovered that the response to faces processing, the most prominent of which is the category of faces
is weaker in the right hemisphere in dyslexics relative to controls either (Hasson et al., 2002). A third possibility worth mentioning is that
when compared to rest or when compared to houses and checker- alphabetic reading requires the discovery of phonetic information
boards. Indeed, the rightward asymmetry in response to faces, present embedded in speech, and might therefore enhance attention to
in normal readers, is not observed in dyslexics (Fig. 2). Finally, the re- facial speech movements. Strikingly, there is a substantial increase
sponse to faces in the left hemisphere extends more medially in normal of audio–visual integration, measured through the McGurk effect, be-
readers than in dyslexics (Fig. 1). It seems unlikely that these results re- tween 6 and 8 years of age in English children but not in Japanese chil-
flect a bilateral disorganization of the fusiform gyrus in dyslexia. Al- dren (Sekiyama and Burnham, 2008). The authors did not consider
though neuronal migration anomalies were occasionally reported in reading acquisition as a potential cause of this change, but the greater
the ventral visual cortex of dyslexics (Galaburda et al., 1985, 2006), an- phonetic demands of alphabetic reading might have enhanced atten-
atomical MR analyses have been inconsistent, with Silani et al. (2005) tion to articulation during reading acquisition in English compared
observing increased gray matter in the left inferior temporal region, to Japanese children. Although audiovisual integration has been main-
while Kronbichler et al. (2008) reported a bilateral fusiform decrease ly related to the superior temporal sulcus (Beauchamp et al., 2004),
in gray matter and others in different temporal or occipital regions while FFA is thought to be more involved in speaker identification
(e.g. Eckert et al., 2005, in the lingual gyrus; Brambati et al., 2004, in (von Kriegstein et al., 2005), several experiments using audio–visual
the anterior fusiform gyri). Also note that in the present results, the speech reported greater activations for speaking relative to still
left-hemispheric activation to faces was similar in normal readers and faces in both left and right fusiform areas (Campbell et al., 2001;
dyslexics at the VWFA site (Fig. 3), confirming that this region is able Dick et al., 2010; Kawase et al., 2005; Stevenson et al., 2010; Nath et
to sustain a normal level of activation. Similarly, using MEG in dyslexic al., 2011). By causing an enhanced attention to articulation, literacy
and control adults, Tarkiainen et al. (2003) reported that, although the might thus benefit to both word and face responses. Future research
dyslexics were slower and more prone to error in facial recognition should try to disentangle these three non-exclusive possibilities.
 K. Monzalvo et al. / NeuroImage 61 (2012) 258–274 271

Dyslexia and the spoken language system The SMA, right posterior temporal region and left insula also
showed a decreased response to all speech stimuli in dyslexics.
The second half of our fMRI study was dedicated to spoken Although we cannot completely exclude that the normal readers
language processing. Phonological competence is a key predictor were more attentive, this interpretation seems unlikely given that
of reading acquisition (Lyytinen et al., 2004; Maurer et al., 2009) and both groups showed the same performance in detecting the end of
relates primarily to the left temporo-parietal region (Dehaene- the sentences, and in particular the same acceleration of response
Lambertz et al., 2005; Jacquemot et al., 2003), a region frequently times for the second presentation of the sentences. Dyslexics and con-
hypoactivated in dyslexics (Blau et al., 2009, 2010; Paulesu et al., trols also did not differ in their amount of motion during fMRI (the
2001; Richlan et al., 2009; Rumsey et al., 1992; Temple, 2002). Howev- more restless participants, and thus possibly the less attentive ones,
er, in their meta-analysis of fMRI dyslexia studies, Richlan et al. (2011) were the low-SES normal-readers). Thus, rather than inattention
note that this hypo-activation is typical of adult dyslexics, but is less in dyslexics, these results are compatible with a broad auditory
frequent in children (Blau et al., 2010 and more recently Maurer et enhancement in normal readers, leading to an expansion of speech
al., 2011). We therefore re-examined this issue by probing whether responses to bilateral temporal cortices and to regions involved in
deficits would be seen even in a minimal sentence-listening task. articulatory coding (SMA, left anterior insula) (Dronkers, 1996).
The results indeed revealed anomalies in activation to spoken lan- These effects may tentatively relate to the enhanced phonological
guage at several sites predominantly located in the left temporal lobe awareness in literates (Morais et al., 1979; Perre et al., 2009) and,
(see Fig. 4). In particular, a salient hypoactivation correlated with conversely, to the documented difficulties of dyslexic children in
reading performance was seen in the left planum temporale. At this perceiving speech in noise (Ziegler et al., 2009). It appears that,
site, normal participants but not dyslexics showed a greater activa- during speech listening in the aversive conditions caused by magnet
tion to native than to foreign language (see Fig. 4, region #6). This noise, normal readers are able to recruit additional brain regions
hypo-activation is likely to be a consequence of impaired reading involved in articulation.
acquisition because activation to speech, at this site, increases with
literacy in adults (Dehaene et al., 2010). Indeed, this region, together Predictors of VWFA development
with the nearby left superior temporal and supramarginal cortices,
hosts native-language phonetic representations (Chang et al., 2010; By studying both visual and spoken language responses in chil-
Jacquemot et al., 2003) which are known to be refined by reading ac- dren, our study offered a unique opportunity to evaluate the relations
quisition (Morais et al., 1979; Perre et al., 2009). As letter–sound cor- between these two systems. In particular, we evaluated whether the
respondences are learned, this region becomes sensitive to letter- inter-individual variability in VWFA activation to written words
sound congruity (Brem et al., 2010; Perre et al., 2009; van Atteveldt could be predicted by the activations to spoken language. Crucially,
et al., 2004; Yoncheva et al., 2010a), an effect which is reduced or ab- our results suggest that the predictive regions differ for dyslexics
sent in dyslexic subjects (Blau et al., 2009, 2010; Perre et al., 2009). compared to normal readers. In normal readers, the spoken-language
The present research shows that this anomaly is detectable in a predictors of VWFA activation during reading were found primarily in
mere speech-listening task. Furthermore, activation in the planum the left superior temporal gyrus and sulcus (Fig. 5). This correlation is
temporale was correlated with several behavioral scores beyond read- hardly surprising since these sites activate amodally immediately
ing performance: verbal memory, phoneme deletion and vocabulary. after the VWFA during word reading (Marinkovic et al., 2003) and
Regression with these different scores isolated slightly different sub- contribute decisively to letter–sound integration (Blomert and
regions of the planum as well as distinct distant cortical areas (verbal Froyen, 2010; van Atteveldt et al., 2004) as well as lexical and senten-
memory with the precentral region, possibly due to articulatory tial analysis (Jobard et al., 2003, 2007). It seems logical that children
planning; metaphonology with SMA, possibly related to articulatory with the most developed spoken-language networks also show the
motor selection and execution; and vocabulary with VWFA, possibly most responsive VWFA during reading. This finding also confirms
related to lexical learning). This observation points to the possibility previous research in adults which indicates that the lateralization of
of a fragmentation of different subparts of the planum into specialized the VWFA is highly correlated with that of temporal lobe spoken-
subterritories, as previously pointed out (Zheng, 2009). language areas, particularly in the STS (Cai et al., 2008, 2010; Pinel
We also observed a reduced activation of the left ventral temporal and Dehaene, 2009).
cortex in dyslexics during spoken language processing, particularly to The left precentral cortex, another essential area of convergence be-
the first presentation of a sentence in the native language. The poste- tween the oral and written language networks, also showed an activa-
rior part of this cluster overlaps with the anterior part of the VWFA, tion to speech that was correlated with the response to written words
and thus may relate to the top-down recruitment of an orthographic in the VWFA, for both normals and dyslexics (at z = +36 mm). This
code only in good readers. Such a top-down activation has been precentral activation extends over regions controlling mouth and
observed by fMRI during selective attention to auditory rhymes larynx movements (Price, 2010), and might thus connect visual words
in adults (Yoncheva et al., 2010b) and is present in literates but with an articulatory representation. In adults, this motor representation
not illiterate adults during a difficult spoken lexical decision task is accessed automatically, even for subliminal words (Dehaene et al.,
(Dehaene et al., 2010). Desroches et al. (2010) observed it in 2001; Nakamura et al., 2007), emphasising the fast connection between
9–15 years-old normal readers but not dyslexics during an auditory the VWFA and this region. The observed correlation suggests that, in
rhyming task. Here, we further observed that this area of top-down ac- complement to the cross-modal integration of letters and sounds
tivation extends anteriorily towards the basal temporal language area, which occurs in the posterior superior temporal cortex (Blomert and
a multi-modal region involved in spoken and written language pro- Froyen, 2010; van Atteveldt et al., 2004), the articulatory component
cessing (Büchel et al., 1998; Burnstine et al., 1990 ; Luders et al., of speech may play an important role in reading acquisition.
1991; Mani et al., 2008 ; Papathanassiou et al., 2000). The role of this Beyond this common site in precentral cortex, dyslexics differed
region remains uncertain, but may concern amodal lexico-semantic from normal readers by exhibiting a reduced correlation of the
processing of single words, as part of the lexical route for reading VWFA activation with lateral and superior temporal regions, and an
(Jobard et al., 2003). Since we found its activation during language increased correlation with a dorsal supra-sylvian network comprising
listening to correlate with reading scores, but also with vocabulary the left supramarginal gyrus and a cluster in the upper precentral cor-
size, we suggest that the vocabulary increase induced by literacy tex (z = +60 mm). The latter site overlaps with a region involved in
might account for the increased activation to spoken language in this hand movement planning (e.g. finger tapping, Meister et al., 2009)
region in normals compared to dyslexics. and especially writing (Exner's area) (Longcamp et al., 2003; Roux
272 K. Monzalvo et al. / NeuroImage 61 (2012) 258–274

et al., 2009). A tentative interpretation is that the most successful We are grateful to Dr. Florence Delteil-Pinton for her help in contacting
dyslexics, who eventually develop a VWFA, rely in part on manual the dyslexic children, the directors of the schools, the children and
gestures to compensate their phonological deficit. Indeed, practice their parents for their interest and participation in this research,
in handwriting has been shown to enhance letter and written word INSERM, CNRS, Collège de France, University Paris 11, the city of Paris,
recognition, both in normal adults (Longcamp et al., 2008), in brain- CONACYT, NERF and the Bettencourt-Schueller Foundation for their
lesioned adults with pure alexia (Seki et al., 1995), and in normal chil- financial support.
dren (Bara et al., 2004; Longcamp et al., 2005). It would be interesting
to examine whether a similar improvement can be seen in dyslexic
children. References

Aylward, E.H., Park, J.E., Field, K.M., Parsons, A.C., Richards, T.L., Cramer, S.C., Meltzoff,
Socio-economic status, dyslexia, and brain activation A.N., 2005. Brain activation during face perception: evidence of a developmental
change. J. Cogn. Neurosci. 17, 308–319.
Bara, F., Gentaz, E., Colé, P., Sprenger-Charolles, L., 2004. The visuo-haptic and haptic
Finally, our results revealed a strong reproducibility of the neural
exploration of letters increases the kindergarten-children's understanding of the
correlates of impaired reading in the two SES groups (Figs. 3 and 4). alphabetic principle. Cogn. Dev. 19, 433–449.
As expected from the literature (Hackman and Farah, 2009), low- Beauchamp, M.S., Argall, B.D., Bodurka, J., Duyn, J.H., Martin, A., 2004. Unraveling mul-
SES children showed lower verbal performances and reading level, tisensory integration: patchy organization within human STS multisensory cortex.
Nat. Neurosci. 7, 1190–1192.
even in the normal-readers group. Nevertheless, we observed almost Blau, V., van Atteveldt, N., Ekkebus, M., Goebel, R., Blomert, L., 2009. Reduced neural in-
no difference related to SES. In particular, we did not replicate Noble tegration of letters and speech sounds links phonological and reading deficits in
et al.'s (2006b) observation of a multiplicative effect of SES and pho- adult dyslexia. Curr. Biol. 19, 503–508.
Blau, V., Reithler, J., van Atteveldt, N., Seitz, J., Gerretsen, P., Goebel, R., Blomert, L., 2010.
nological awareness on VWFA activation. When phonological abilities Deviant processing of letters and speech sounds as proximate cause of reading
were similar, as was the case here in each reading group, there was no failure: a functional magnetic resonance imaging study of dyslexic children.
supplementary impairment related to a lower SES in the ventral areas Brain 133, 868–879.
Blomert, L., Froyen, D., 2010. Multi-sensory learning and learning to read. Int. J. Psycho-
(see Fig. 3). The only observed difference related to SES was observed physiol. 77, 195–204.
during the auditory runs and, as discussed above, might relate to a Brambati, S.M., Termine, C., Ruffino, M., Stella, G., Fazio, F., Cappa, S.F., Perani, D., 2004.
weaker possibility of compensation in the low-SES dyslexics. Relative Regional reductions of gray matter volume in familial dyslexia. Neurology 63,
742–745.
to their high-SES peers, they were less able to recruit right-
Brem, S., Bach, S., Kucian, K., Guttorm, T.K., Martin, E., Lyytinen, H., Brandeis, D.,
hemispheric regions (superior temporal region, insula, and middle Richardson, U., 2010. Brain sensitivity to print emerges when children learn letter–
frontal gyrus). In particular, the reduced activation in the right frontal speech sound correspondences. Proc. Natl. Acad. Sci. U. S. A. 107, 7939–7944.
Brunswick, N., McCrory, E., Price, C.J., Frith, C.D., Frith, U., 1999a. Explicit and implicit
region might be a negative predictive factor for reading abilities in
processing of words and pseudowords by adult developmental dyslexics: a search
these children, as suggested by Hoeft et al. (2011). Further studies for Wernicke's Wortschatz? Brain 122, 1901–1917.
in these populations are needed to understand why they are not Brunswick, N., McCrory, E., Price, C.J., Frith, C.D., Frith, U., 1999b. Explicit and implicit
able to recruit these regions and how adequate rehabilitation can be processing of words and pseudowords by adult developmental dyslexics: a search
for Wernicke's Wortschatz? Brain 122 (Pt. 10), 1901–1917.
proposed. Büchel, C., Price, C.J., Friston, K., 1998. A multimodal language region in the ventral vi-
sual pathway. Nature 394, 274–277.
Conclusion Burnstine, T.H., Lesser, R.P., Hart Jr., J., Uematsu, S., Zinreich, S.J., Krauss, G.L., Fisher, R.S.,
Vining, E.P., Gordon, B., 1990. Characterization of the basal temporal language area
in patients with left temporal lobe epilepsy. Neurology 40, 966–970.
Our results demonstrate that near-passive fMRI, using a minimal Cai, Q., Lavidor, M., Brysbaert, M., Paulignan, Y., Nazir, T.A., 2008. Cerebral lateralization
detection task, suffices to detect localized activation differences be- of frontal lobe language processes and lateralization of the posterior visual word
processing system. J. Cogn. Neurosci. 20, 672–681.
tween normal readers and dyslexics. In the visual system, we ob- Cai, Q., Paulignan, Y., Brysbaert, M., Ibarrola, D., Nazir, T.A., 2010. The left ventral
served a hypoactivation in the VWFA to written words but also to occipito-temporal response to words depends on language lateralization but not
speech listening. In the auditory system, the planum temporale on visual familiarity. Cereb. Cortex 20, 1153–1163.
Campbell, R., MacSweeney, M., Surguladze, S., Calvert, G., McGuire, P., Suckling, J.,
showed a reduced response to spoken sentences in the native lan-
Brammer, M.J., David, A.S., 2001. Cortical substrates for the perception of face ac-
guage. These differences between normal and impaired readers tions: an fMRI study of the specificity of activation for seen speech and for mean-
were common to both SES backgrounds, and are coherent with the ingless lower-face acts (gurning). Brain Res. Cogn. Brain Res. 12, 233–243.
Cantlon, J.F., Pinel, P., Dehaene, S., Pelphrey, K.A., 2011. Cortical representations of sym-
numerous studies testing dyslexic adults and children. At the theoret-
bols, objects, and faces are pruned back during early childhood. Cereb. Cortex 21,
ical level, they remain consistent with two alternative interpretations. 191–199.
One possible scenario, in agreement with the phonological hypothesis Carreiras, M., Seghier, M.L., Baquero, S., Estevez, A., Lozano, A., Devlin, J.T., Price, C.J.,
for the origin of dyslexia, is that before reading, children at risk of 2009. An anatomical signature for literacy. Nature 461, 983–986.
Castro-Caldas, A., Petersson, K.M., Reis, A., Stone-Elander, S., Ingvar, M., 1998. The illit-
dyslexia already present a reduced activation in planum temporale, erate brain. Learning to read and write during childhood influences the functional
which causes difficulties in reading acquisition and therefore impedes organization of the adult brain. Brain 121, 1053–1063.
the normal development of an orthographic representation in the Chandrasekaran, B., Hornickel, J., Skoe, E., Nicol, T., Kraus, N., 2009. Context-dependent
encoding in the human auditory brainstem relates to hearing speech in noise: im-
VWFA. Another scenario, however, acknowledges that the observed plications for developmental dyslexia. Neuron 64, 311–319.
areas of hypo-activation in dyslexic children coincide with the sites Chang, E.F., Rieger, J.W., Johnson, K., Berger, M.S., Barbaro, N.M., Knight, R.T., 2010. Cat-
of reduced activation in illiterate relative to literate adults (Dehaene egorical speech representation in human superior temporal gyrus. Nat. Neurosci.
13, 1428–1432.
et al., 2010). As such, they may solely be a consequence of the abnor- Dehaene, S., Naccache, L., Cohen, L., Bihan, D.L., Mangin, J.F., Poline, J.B., Riviere, D.,
mal acquisition of literacy, rather than a cause of dyslexia. Further ex- 2001. Cerebral mechanisms of word masking and unconscious repetition priming.
periments, possibly using a longitudinal design with the same fMRI Nat. Neurosci. 4, 752–758.
Dehaene, S., Pegado, F., Braga, L.W., Ventura, P., Filho, G.N., Jobert, A., Dehaene-
methods, will be needed to understand the causal relation between Lambertz, G., Kolinsky, R., Morais, J., Cohen, L., 2010. How learning to read changes
reading proficiency and brain organization. the cortical networks for vision and language. Science 330, 1359–1364.
Dehaene-Lambertz, G., Pallier, C., Serniclaes, W., Sprenger-Charolles, L., Jobert, A.,
Dehaene, S., 2005. Neural correlates of switching from auditory to speech percep-
Acknowledgments
tion. NeuroImage 24, 21–33.
Dehaene-Lambertz, G., Dehaene, S., Anton, J.L., Campagne, A., Ciuciu, P., Dehaene, G.P.,
We thank the NeuroSpin infrastructure groups and particularly the Denghien, I., Jobert, A., Lebihan, D., Sigman, M., Pallier, C., Poline, J.B., 2006a. Func-
nurses (Véronique Joly-Testault, Laurence Laurier, Gaelle Mediouni) tional segregation of cortical language areas by sentence repetition. Hum. Brain
Mapp. 27, 360–371.
and radio technicians (Katel Ripert, Severine Roger, Severine Desmidt, Dehaene-Lambertz, G., Hertz-Pannier, L., Dubois, J., Meriaux, S., Roche, A., Sigman, M.,
Chantal Ginisty) for their help in welcoming and testing the children. Dehaene, S., 2006b. Functional organization of perisylvian activation during
 K. Monzalvo et al. / NeuroImage 61 (2012) 258–274 273

presentation of sentences in preverbal infants. Proc. Natl. Acad. Sci. U. S. A. 103, Longcamp, M., Boucard, C., Gilhodes, J.C., Anton, J.L., Roth, M., Nazarian, B., Velay, J.L.,
14240–14245. 2008. Learning through hand- or typewriting influences visual recognition of
Desroches, A.S., Cone, N.E., Bolger, D.J., Bitan, T., Burman, D.D., Booth, J.R., 2010. new graphic shapes: behavioral and functional imaging evidence. J. Cogn. Neurosci.
Children with reading difficulties show differences in brain regions associated 20, 802–815.
with orthographic processing during spoken language processing. Brain Res. Luders, H., Lesser, R.P., Hahn, J., Dinner, D.S., Morris, H.H., Wyllie, E., Godoy, J., 1991.
1356, 73–84. Basal temporal language area. Brain 114 (Pt. 2), 743–754.
Dick, A.S., Solodkin, A., Small, S.L., 2010. Neural development of networks for audiovi- Lyytinen, H., Ahonen, T., Eklund, K., Guttorm, T., Kulju, P., Laakso, M.L., Leiwo, M.,
sual speech comprehension. Brain Lang. 114, 101–114. Leppanen, P., Lyytinen, P., Poikkeus, A.M., Richardson, U., Torppa, M., Viholainen,
Dronkers, N.F., 1996. A new brain region for coordinating speech articulation. Nature H., 2004. Early development of children at familial risk for dyslexia—follow-up
384, 159–161. from birth to school age. Dyslexia 10, 146–178.
Dufor, O., Serniclaes, W., Sprenger-Charolles, L., Demonet, J.F., 2007. Top-down processes Maisog, J.M., Einbinder, E.R., Flowers, D.L., Turkeltaub, P.E., Eden, G.F., 2008. A meta-
during auditory phoneme categorization in dyslexia: a PET study. NeuroImage 34, analysis of functional neuroimaging studies of dyslexia. Ann. N. Y. Acad. Sci.
1692–1707. 1145, 237–259.
Eckert, M.A., Leonard, C.M., Wilke, M., Eckert, M., Richards, T., Richards, A., Berninger, Mani, J., Diehl, B., Piao, Z., Schuele, S.S., Lapresto, E., Liu, P., Nair, D.R., Dinner, D.S.,
V., 2005. Anatomical signatures of dyslexia in children: unique information from Luders, H.O., 2008. Evidence for a basal temporal visual language center: cortical
manual and voxel based morphometry brain measures. Cortex 41, 304–315. stimulation producing pure alexia. Neurology 71, 1621–1627.
Fluss, J., Ziegler, J.C., Warszawski, J., Ducot, B., Richard, G., Billard, C., 2009. Poor reading Marinkovic, K., Dhond, R.P., Dale, A.M., Glessner, M., Carr, V., Halgren, E., 2003. Spatio-
in French elementary school: the interplay of cognitive, behavioral, and socioeco- temporal dynamics of modality-specific and supramodal word processing. Neuron
nomic factors. J. Dev. Behav. Pediatr. 30, 206–216. 38, 487–497.
Frederickson, N., Frith, U., 1998. Identifying dyslexia in bilingual children: a phonolog- Maurer, U., Brandeis, D., McCandliss, B.D., 2005a. Fast, visual specialization for reading
ical approach with inner London Sylheti speakers. Dyslexia 4, 119–131. in English revealed by the topography of the N170 ERP response. Behav. Brain
Froyen, D., Willems, G., Blomert, L., 2011. Evidence for a specific cross-modal associa- Funct. 1, 13.
tion deficit in dyslexia: an electrophysiological study of letter–speech sound pro- Maurer, U., Brem, S., Bucher, K., Brandeis, D., 2005b. Emerging neurophysiological spe-
cessing. Dev. Sci. 14, 635–648. cialization for letter strings. J. Cogn. Neurosci. 17, 1532.
Galaburda, A.M., Sherman, G.F., Rosen, G.D., Aboitiz, F., Geschwind, N., 1985. Develop- Maurer, U., Bucher, K., Brem, S., Benz, R., Kranz, F., Schulz, E., van der Mark, S.,
mental dyslexia: four consecutive patients with cortical anomalies. Ann. Neurol. Steinhausen, H.C., Brandeis, D., 2009. Neurophysiology in preschool improves be-
18, 222–233. havioral prediction of reading ability throughout primary school. Biol. Psychiatry
Galaburda, A.M., LoTurco, J., Ramus, F., Fitch, R.H., Rosen, G.D., 2006. From genes to be- 66, 341–348.
havior in developmental dyslexia. Nat. Neurosci. 9, 1213–1217. Maurer, U., Blau, V.C., Yoncheva, Y.N., McCandliss, B.D., 2010. Development of visual ex-
Golarai, G., Ghahremani, D.G., Whitfield-Gabrieli, S., Reiss, A., Eberhardt, J.L., Gabrieli, J.D., pertise for reading: rapid emergence of visual familiarity for an artificial script.
Grill-Spector, K., 2007. Differential development of high-level visual cortex correlates Dev. Neuropsychol. 35, 404–422.
with category-specific recognition memory. Nat. Neurosci. 10, 512–522. Maurer, U., Schulz, E., Brem, S., der Mark, S., Bucher, K., Martin, E., Brandeis, D., 2011.
Hackman, D.A., Farah, M.J., 2009. Socioeconomic status and the developing brain. The development of print tuning in children with dyslexia: evidence from longitu-
Trends Cogn. Sci. 13, 65–73. dinal ERP data supported by fMRI. Neuroimage 57, 714–722.
Hasson, U., Levy, I., Behrmann, M., Hendler, T., Malach, R., 2002. Eccentricity bias as an Mazaika, P., Whitfield-Gabrieli, S., Reiss, A., 2007. Artifact repair for fMRI data from
organizing principle for human high-order object areas. Neuron 34, 479–490. high motion clinical subjects. 13th Annual Meeting of the Organization for
Hasson, U., Harel, M., Levy, I., Malach, R., 2003. Large-scale mirror-symmetry organiza- Human Brain Mapping, Chicago,Il.
tion of human occipito-temporal object areas. Neuron 37, 1027–1041. McCandliss, B.D., Noble, K.G., 2003. The development of reading impairment: a cogni-
Haxby, J.V., Gobbini, M.I., Furey, M.L., Ishai, A., Schouten, J.L., Pietrini, P., 2001. Distrib- tive neuroscience model. Ment. Retard. Dev. Disabil. Res. Rev. 9, 196–204.
uted and overlapping representations of faces and objects in ventral temporal cor- McCandliss, B.D., Sandak, R., Beck, I., Perfetti, C.A., 2003. Focusing attention on decoding
tex. Science 293, 2425–2430. for children with poor reading skills: design and preliminary tests of the Word
Hoeft, F., McCandliss, B.D., Black, J.M., Gantman, A., Zakerani, N., Hulme, C., Lyytinen, H., Building intervention. Sci. Stud. Read. 7, 75–105.
Whitfield-Gabrieli, S., Glover, G.H., Reiss, A.L., Gabrieli, J.D., 2011. Neural systems pre- McCrory, E.J., Mechelli, A., Frith, U., Price, C.J., 2005. More than words: a common neural
dicting long-term outcome in dyslexia. Proc. Natl. Acad. Sci. U. S. A. 108, 361–366. basis for reading and naming deficits in developmental dyslexia? Brain 128,
INSERM, Collective expertise, 2007. Dyslexie, dysorthographie, dyscalculie : bilan des 261–267.
données scientifiques. Synthèse et Recommendation. Institut National de la Sante Meister, I.G., Buelte, D., Staedtgen, M., Boroojerdi, B., Sparing, R., 2009. The dorsal pre-
Et de la Recherche Medicale, Paris. 860 pp. motor cortex orchestrates concurrent speech and fingertapping movements. Eur. J.
Ishai, A., Ungerleider, L.G., Martin, A., Schouten, J.L., Haxby, J.V., 1999. Distributed represen- Neurosci. 29, 2074–2082.
tation of objects in the human ventral visual pathway. Proc. Natl. Acad. Sci. U. S. A. 96, Morais, J., Cary, L., Alegria, J., Bertelson, P., 1979. Does awareness of speech as a se-
9379–9384. quence of phones arise spontaneously? Cognition 7, 323–331.
Ishai, A., Ungerleider, L.G., Martin, A., Haxby, J.V., 2000. The representation of objects in Nath, A.R., Fava, E.E., Beauchamp, M.S., 2011. Neural correlates of interindividual differ-
the human occipital and temporal cortex. J. Cogn. Neurosci. 12 (Suppl. 2), 35–51. ences in children’s audiovisual speech perception. J. Neurosc. 31, 13963–13971.
Jacquemot, C., Pallier, C., LeBihan, D., Dehaene, S., Dupoux, E., 2003. Phonological gram- Nakamura, K., Dehaene, S., Jobert, A., Le Bihan, D., Kouider, S., 2007. Task-specific
mar shapes the auditory cortex: a functional magnetic resonance imaging study. change of unconscious neural priming in the cerebral language network. Proc.
J. Neurosci. 23, 9541–9546. Natl. Acad. Sci. U. S. A. 104, 19643–19648.
Jambaqué, I., Dellatolas, G., 2000. Epreuve de fluence verbale et de dénomination chez Noble, K.G., Farah, M.J., McCandliss, B.D., 2006a. Socioeconomic background modulates
l'enfant d'âge scolaire (6 ans 6 à 14 ans 6). A.N.A.E., pp. 13–16. cognition–achievement relationships in reading. Cogn. Dev. 21, 349–368.
Jobard, G., Crivello, F., Tzourio-Mazoyer, N., 2003. Evaluation of the dual route theory of Noble, K.G., Wolmetz, M.E., Ochs, L.G., Farah, M.J., McCandliss, B.D., 2006b. Brain–behavior
reading: a metanalysis of 35 neuroimaging studies. NeuroImage 20, 693–712. relationships in reading acquisition are modulated by socioeconomic factors. Dev. Sci.
Jobard, G., Vigneau, M., Mazoyer, B., Tzourio-Mazoyer, N., 2007. Impact of modality and 9, 642–654.
linguistic complexity during reading and listening tasks. NeuroImage 34, 784–800. Noble, K.G., McCandliss, B.D., Farah, M.J., 2007. Socioeconomic gradients predict indi-
Joseph, J.E., Gathers, A.D., Bhatt, R.S., 2011. Progressive and regressive developmental vidual differences in neurocognitive abilities. Dev. Sci. 10, 464–480.
changes in neural substrates for face processing: testing specific predictions of Papathanassiou, D., Etard, O., Mellet, E., Zago, L., Mazoyer, B., Tzourio-Mazoyer, N.,
the Interactive Specialization account. Dev. Sci. 14, 227–241. 2000. A common language network for comprehension and production: a contri-
Kawase, T., Yamaguchi, K., Ogawa, T., Suzuki, K., Suzuki, M., Itoh, M., Kobayashi, T., Fujii, bution to the definition of language epicenters with PET. NeuroImage 11,
T., 2005. Recruitment of fusiform face area associated with listening to degraded 347–357.
speech sounds in auditory-visual speech perception: a PET study. Neurosci. Lett. Passarotti, A.M., Paul, B.M., Bussiere, J.R., Buxton, R.B., Wong, E.C., Stiles, J., 2003. The
382, 254–258. development of face and location processing: an fMRI study. Dev. Sci. 100–117.
Kronbichler, M., Wimmer, H., Staffen, W., Hutzler, F., Mair, A., Ladurner, G., 2008. De- Paulesu, E., Demonet, J.F., Fazio, F., McCrory, E., Chanoine, V., Brunswick, N., Cappa, S.F.,
velopmental dyslexia: gray matter abnormalities in the occipitotemporal cortex. Cossu, G., Habib, M., Frith, C.D., Frith, U., 2001. Dyslexia: cultural diversity and bio-
Hum. Brain Mapp. 29, 613–625. logical unity. Science 291, 2165–2167.
Lefavrais, P., 1967. Test de l'Alouette: Manuel. Les Éditions du Centre de Psychologie Peelen, M.V., Glaser, B., Vuilleumier, P., Eliez, S., 2009. Differential development of se-
Appliquée, Paris. lectivity for faces and bodies in the fusiform gyrus. Dev. Sci. 12, F16–F25.
Lesaux, N.K., Siegel, L.S., 2003. The development of reading in children who speak Perre, L., Pattamadilok, C., Montant, M., Ziegler, J.C., 2009. Orthographic effects in spo-
English as a second language. Dev. Psychol. 39, 1005–1019. ken language: on-line activation or phonological restructuring? Brain Res. 1275,
Lété, B., Sprenger-Charolles, L., Colé, P., 2004. MANULEX: a grade-level lexical database 73–80.
from French elementary-school readers. Behav. Res. Methods Instrum. Comput. 36, Pinel, P., Dehaene, S., 2009. Beyond hemispheric dominance: brain regions underlying
156–166. the joint lateralization of language and arithmetic to the left hemisphere. J. Cogn.
Li, G., Cheung, R.T., Gao, J.H., Lee, T.M., Tan, L.H., Fox, P.T., Jack Jr., C.R., Yang, E.S., 2006. Neurosci. 22, 48–66.
Cognitive processing in Chinese literate and illiterate subjects: an fMRI study. Hum. Price, C.J., 2010. The anatomy of language: a review of 100 fMRI studies published in
Brain Mapp. 27, 144–152. 2009. Ann. N. Y. Acad. Sci. 1191, 62–88.
Longcamp, M., Anton, J.L., Roth, M., Velay, J.L., 2003. Visual presentation of single letters Ramus, F., 2004. Neurobiology of dyslexia: a reinterpretation of the data. Trends Neurosci.
activates a premotor area involved in writing. NeuroImage 19, 1492–1500. 27, 720–726.
Longcamp, M., Zerbato-Poudou, M.T., Velay, J.L., 2005. The influence of writing practice Richlan, F., Kronbichler, M., Wimmer, H., 2009. Functional abnormalities in the dyslexic
on letter recognition in preschool children: a comparison between handwriting brain: a quantitative meta-analysis of neuroimaging studies. Hum. Brain Mapp. 30,
and typing. Acta Psychol. (Amst) 119, 67–79. 3299–3308.
274 K. Monzalvo et al. / NeuroImage 61 (2012) 258–274

Richlan, F., Kronbichler, M., Wimmer, H., 2011. Meta-analyzing brain dysfunctions in Szwed, M., Dehaene, S., Kleinschmidt, A., Eger, E., Valabregue, R., Amadon, A., Cohen, L.,
dyslexic children and adults. NeuroImage 56, 1735–1742. 2011. Specialization for written words over objects in the visual cortex. NeuroImage
Roux, F.E., Dufor, O., Giussani, C., Wamain, Y., Draper, L., Longcamp, M., Demonet, J.F., 56, 330–344.
2009. The graphemic/motor frontal area Exner's area revisited. Ann. Neurol. 66, Tarkiainen, A., Helenius, P., Salmelin, R., 2003. Category-specific occipitotemporal acti-
537–545. vation during face perception in dyslexic individuals: an MEG study. NeuroImage
Ruff, S., Marie, N., Celsis, P., Cardebat, D., Demonet, J.F., 2003. Neural substrates of im- 19, 1194–1204.
paired categorical perception of phonemes in adult dyslexics: an fMRI study. Temple, E., 2002. Brain mechanisms in normal and dyslexic readers. Curr. Opin. Neurobiol.
Brain Cogn. 53, 331–334. 12, 178–183.
Rumsey, J.M., Andreason, P., Zametkin, A.J., Aquino, T., King, C., Hamburger, S.D., Pikus, Torgesen, J.K., Wagner, R.K., Rashotte, C.A., 1994. Longitudinal studies of phonological
A., Rapoport, J.L., Cohen, R.M., 1992. Failure to activate the left temporoparietal cor- processing and reading. J. Learn. Disabil. 27, 276–286 (discussion 287–291).
tex in dyslexia. Arch. Neurol. 49, 527–534. Tzourio-Mazoyer, N., Landeau, B., Papathanassiou, D., Crivello, F., Etard, O., Delcroix, N.,
Scherf, K.S., Behrmann, M., Humphreys, K., Luna, B., 2007. Visual category-selectivity Mazoyer, B., Joliot, M., 2002. Automated anatomical labeling of activations in SPM
for faces, places and objects emerges along different developmental trajectories. using a macroscopic anatomical parcellation of the MNI MRI single-subject brain.
Dev. Sci. 10, F15–F30. NeuroImage 15, 273–289.
Scherf, K.S., Luna, B., Avidan, G., Behrmann, M., 2011. “What” precedes “which”: devel- van Atteveldt, N., Formisano, E., Goebel, R., Blomert, L., 2004. Integration of letters and
opmental neural tuning in face- and place-related cortex. Cereb. Cortex. 21, speech sounds in the human brain. Neuron 43, 271–282.
1963–1980. van der Mark, S., Bucher, K., Maurer, U., Schulz, E., Brem, S., Buckelmuller, J.,
Schlaggar, B.L., McCandliss, B.D., 2007. Development of neural systems for reading. Kronbichler, M., Loenneker, T., Klaver, P., Martin, E., Brandeis, D., 2009. Children
Annu. Rev. Neurosci. 30, 475–503. with dyslexia lack multiple specializations along the visual word-form (VWF) sys-
Seki, K., Yajima, M., Sugishita, M., 1995. The efficacy of kinesthetic reading treatment tem. NeuroImage 47, 1940–1949.
for pure alexia. Neuropsychologia 33, 595–609. Vellutino, F.R., Fletcher, J.M., Snowling, M.J., Scanlon, D.M., 2004. Specific reading dis-
Sekiyama, K., Burnham, D., 2008. Impact of language on development of auditory- ability (dyslexia): what have we learned in the past four decades? J. Child Psychol.
visual speech perception. Dev. Sci. 11, 306–320. Psychiatry 45, 2–40.
Shaywitz, S.E., 1998. Dyslexia. N Engl J. Med. 338, 307–312. von Kriegstein, K., Kleinschmidt, A., Sterzer, P., Giraud, A.L., 2005. Interaction
Shaywitz, S.E., Shaywitz, B.A., Fulbright, R.K., Skudlarski, P., Mencl, W.E., Constable, R.T., of face and voice areas during speaker recognition. J. Cogn. Neurosci. 17,
Pugh, K.R., Holahan, J.M., Marchione, K.E., Fletcher, J.M., Lyon, G.R., Gore, J.C., 2003. 367–376.
Neural systems for compensation and persistence: young adult outcome of child- Yoncheva, Y.N., Blau, V.C., Maurer, U., McCandliss, B.D., 2010a. Attentional focus during
hood reading disability. Biol. Psychiatry 54, 25–33. learning impacts N170 ERP responses to an artificial script. Dev. Neuropsychol. 35,
Silani, G., Frith, U., Demonet, J.F., Fazio, F., Perani, D., Price, C., Frith, C.D., Paulesu, E., 423–445.
2005. Brain abnormalities underlying altered activation in dyslexia: a voxel Yoncheva, Y.N., Zevin, J.D., Maurer, U., McCandliss, B.D., 2010b. Auditory selective at-
based morphometry study. Brain 128, 2453–2461. tention to speech modulates activity in the visual word form area. Cereb. Cortex
Sprenger-Charolles, L., Colé, P., Béchennec, D., Kipffer-Piquard, A., 2005. French normative 20, 622–632.
data on reading and related skills from EVALEC, a new computerized battery of tests Zheng, Z.Z., 2009. The functional specialization of the planum temporale. J. Neurophy-
(end Grade 1, Grade 2, Grade 3, and Grade 4). Eur. Rev. Appl. Psychol. 55, 157–186. siol. 102, 3079–3081.
Stevenson, R.A., Altieri, N.A., Kim, S., Pisoni, D.B., James, T.W., 2010. Neural processing Ziegler, J.C., Pech-Georgel, C., George, F., Lorenzi, C., 2009. Speech-perception-in-noise
of asynchronous audiovisual speech perception. NeuroImage 49, 3308–3318. deficits in dyslexia. Dev. Sci. 12, 732–745.