Dehaene dmcppmll 1997_anatomical_variability_first_&_second_languages.neuroreport

Introduction
Do similar cerebral networks support first and second
language acquisition? When a polyglot suffers a brain
lesion, aphasia is occasionally observed in only one
of the languages originally mastered.1,2
This dissoci-
ation, together with evidence from electrical cortical
stimulation,3
suggests that different brain areas are
recruited for learning and processing the first
language (L1) and the second language (L2). Brain
imaging studies in groups of bilingual subjects have
also revealed differences between L1 and L2 repre-
sentation.4–7
Neuropsychological and imaging studies
have failed to pinpoint a consistent neuronal substrate
for second language acquisition, however, perhaps
because they were obtained in various languages,
using variable tasks, and with subjects of varying
levels of proficiency. Inter-subject variability may
have prevented the emergence of consistent results,
particularly in PET studies based on group averaging.
Functional magnetic resonance imaging (fMRI) is
a method ideally suited for single-subject analyses,
because it permits the assessment of significantly
activated areas in individual subjects without requir-
ing group averaging. The present study aimed at
assessing inter-subject variability in the cortical repre-
sentation of language comprehension processes in L1
and L2. We used whole-brain 3-Tesla echo-planar
fMRI to explore the cerebral networks underlying
comprehension of stories in L1 and L2, in line
with our previous previous work on story listening
using PET.5,8
Eight subjects were imaged while they
listened to different stories in their native language
(L1 = French) or in a second language that they had
learned at school after the age of seven, and for which
they had a moderate proficiency level (L2 = English).
In both languages, short stories were recorded by a
native speaker, digitally edited and cut into three
blocks of 36 seconds. These blocks were then
presented in alternation with a control condition
Brain Imaging
111
0111
0111
0111
0111
0111
111p
© Rapid Science Publishers Vol 8 No 17 1 December 1997 3809
FUNCTIONAL magnetic resonance imaging was used to
assess inter-subject variability in the cortical represen-
tation of language comprehension processes. Moderately
fluent French–English bilinguals were scanned while
they listened to stories in their first language
(L1 = French) or in a second language (L2 = English)
acquired at school after the age of seven. In all subjects,
listening to L1 always activated a similar set of areas in
the left temporal lobe, clustered along the left superior
temporal sulcus. Listening to L2, however, activated a
highly variable network of left and right temporal and
frontal areas, sometimes restricted only to right-hemi-
spheric regions. These results support the hypothesis
that first language acquisition relies on a dedicated left-
hemispheric cerebral network, while late second
language acquisition is not necessarily associated with a
reproducible biological substrate. The postulated contri-
bution of the right hemisphere to L2 comprehension1
is
found to hold only on average, individual subjects
varying from complete right lateralization to standard
left lateralization for L2.
Key words: Bilingualism; Language; Lateralization; Speech;
Temporal lobe
Anatomical variability
in the cortical
representation of first
and second language
Stanislas Dehaene,1
Emmanuel Dupoux,1
Jacques Mehler,1,CA
Laurent Cohen,2
Eraldo Paulesu,3
Daniela Perani,3
Pierre-Francois van de Moortele,4
Stéphane Lehéricy4,5
and
Denis Le Bihan4
1
Laboratoire de Sciences Cognitives et
Psycholinguistique, EHESS/CNRS URA 1198,
54 Boulevard Raspail, 75270 Paris cedex 06;
2
Service de neurologie 1, Hôpital de la
Salpêtrière, 47 Boulevard de l’Hôpital, 75651
Paris cedex 13, France; 3
Institute of
Neuroscience and Bioimaging, C.N.R., HSR
Milan, Italy; 4
Service Hospitalier Frédéric
Joliot, Commissariat à l’Energie atomique,
4 Place du général Leclerc, 91401 Orsay cedex;
5
Service de Neuroradiologie, Hôpital de la
Salpêtrière, France
CA
Corresponding Author
Website publication 13 November 1997 NeuroReport 8, 3809–3815 (1997)

consisting of three blocks of 36 seconds of backward
speech (the tape of a story in Japanese, played back-
wards). This design subtracted out brain activity
related to early auditory processing and isolated the
regions specifically involved in the processing of real
speech, either in L1 or L2.
Method
Subjects. Eight right-handed male French students,
native French speakers, aged between 21 and 25, gave
informed consent to participate in this study. All
were born from French parents and had not been
exposed to English before the age of seven. They had
learned English as a second language at school and
none had lived in an English-speaking country for
more than one year. Good understanding of spoken
English was verified, prior to the experiment, using
word translation and sentence comprehension tests,
and during the experiment, by asking difficult factual
questions about the stories immediately after each
scan (one-tailed paired t-test for a greater number of
errors in English than in French: t(7) = 1.00, p = 0.18,
not significant).
Stimuli: Stories were recorded on digital audiotape
by two native French (F) or English (E) speakers and
were digitally cut at sentence boundaries to obtain
fragments of 36 seconds each. Three such fragments
were alternated with three fragments of a control
condition consisting of Japanese speech played back-
wards (B). Two experimental conditions were run in
counterbalanced order: F-B-F-B-F-B and E-B-E-B-
E-B. English and French versions of the same stories
were counterbalanced across subjects such that if one
subject heard story A in French and story B in
English, another subject heard story A in English and
story B in French. Five subjects also listened to a
replication of these conditions using three indepen-
dent short stories instead of a single story cut into
three fragments. Stimuli were presented over stan-
dard headphones customized for fMRI experiments
and inserted in a noise-protecting helmet that
provided isolation from scanner noise.
Image acquisition: Experiments were performed on
a 3T whole-body system (Bruker, Germany). The
study was approved by a National Ethics Committee
for Biomedical Research. The subject’s head was
fixed by foam cushions and bands to limit motion
artifacts. Shimming on the selected slices was carried
out before each acquisition. Eighteen axial contiguous
slices of 5 mm thickness and 22 cm field of view
were scanned every 6 seconds for 216 s (36 volumes)
using a gradient-echo echo-planar imaging (EPI)
sequence (repetition time/echo time/flip angle =
6000 ms/40 ms/90 °, 64 ´ 64 pixel matrix) sensitive to
BOLD contrast (local increases in blood flow and
oxygenation). To improve the anatomical identifica-
tion of the activated regions, high-resolution 2-D
T1-weighted slices matching the EPI slices were also
acquired using an inversion-recovery sequence
(inversion time/repetition time/echo time/flip angle
= 800 ms/3000 ms/8 ms/90 °, 256 ´ 256 pixel matrix).
Data analysis: Images were initially processed using
custom software written under IDL (Interactive
Data Language, Research System Inc., Boulder,
CO). Images were first checked for absence of head
motion, and evidence of motion exceeding one pixel
implied no further analysis. Activation maps were
calculated on a pixel-by-pixel basis based on the
correlation coefficient9
between the MR signal time
course for each pixel and a waveform derived from
the processing paradigm, taking into account the
hemodynamic nature of the response. The first three
images of each series were discarded from analysis,
because the magnetization was not steady at the
beginning of the experiment. A cluster of pixels was
considered active when it consisted of at least 3 con-
tiguous pixels, each with a correlation coefficient
above 0.45. Assuming independent observations, the
threshold of 0.45 corresponds to a Student’s t of 3.14
and to a pixel-based one-tailed probability of 0.00185
(with 31 degrees of freedom). We estimate that
requiring at least 3 contiguous such pixels brings the
probability of finding a cluster at any given location
down to p < 3.6 10–5
(uncorrected for multiple
comparisons across the brain volume). Correction for
temporal correlation between successive images10
was
not done given the long repetition time between
images.
Because the correlation procedure allows only
for within-condition comparisons (i.e. F vs B or
E vs B), but not for between-conditions comparisons
(e.g. F vs E), the data were also reanalyzed using
SPM96. Images were corrected for subject motion,
normalized to Talairach coordinates using a linear
transform calculated on the TI images, and smoothed
(FWHM = 5 mm). The intensity level of each pixel
was then modeled using a linear regression with
8 variables, namely two temporal activation functions
for each of the four block types: French story (F),
backward speech control for the French condition
(BF), English story (E) and backward speech control
for the English condition (BE). Active areas in L1
and L2 were determined using the main effect terms
F-BE and E-BE, using a voxelwise significance level
of 0.001 corrected for multiple comparison across the
brain volume to p < 0.05. Differences between L1 and
L2 were tested using the interaction term (F-BF)–
S. Dehaene et al.
1111
2
3
4
5
6
7
8
9
10111
1
2
3
4
5
6
7
8
9
20111
1
2
3
4
5
6
7
8
9
30111
1
2
3
4
5
6
7
8
9
40111
1
2
3
4
5
6
7
8
9
50111
1
2
3
4
5
6111p
3810 Vol 8 No 17 1 December 1997

(E-BE), with an uncorrected significance level of
0.001. Correlation- and SPM-based analyses gave
highly consistent results, thus testifying to the relia-
bility of the statistical measures used and indicating
that the intersubject variability observed for L2 was
not due only to the intrinsically noisy nature of fMRI
data.
Results
When listening to L1, there was a remarkable consis-
tency in the observed activated areas in the left hemi-
sphere (Fig. 1). All subjects showed activity in the
left temporal lobe all along the superior temporal
sulcus (STS) as well as in neighboring portions of the
superior and middle temporal gyri (STG, MTG),
often extending forward into the temporal pole (TP;
4 subjects) and backward into the left angular gyrus
(AG; 4 subjects). Although similar activity was occa-
sionally found in the right temporal lobe, including
the right STS (6 subjects) and TP (2 subjects), it was
always weaker, highly variable from subject to
subject, and never extended backward into the right
AG. In six out of eight subjects, a significantly larger
number of active pixels were found in the left than
in the right temporal lobe (␹2 tests, all ps < 0.005).
Outside the temporal lobe, the only consistent focus
of activation was found near the intersection of the
inferior frontal sulcus (IFS) and the precentral sulcus
(PrS), bordering Brodman’s areas 8, 9, 44 and 6
(superior to Broca’s area proper; see Fig. 1). This
region was active in the left hemisphere of 6 subjects,
and in the right hemisphere of 3 subjects.
Listening to L2 uncovered much greater inter-
subject variability. No single anatomical area was
found active in more than six subjects (Fig. 1). Six
subjects showed activation foci in the left temporal
lobe (STS, STG, MTG), but the active pixels showed
considerable dispersion contrasting with their tight
localisation to the banks of the STS when listening
to L1 (Fig. 1). Furthermore, no activity was found
Anatomical variability in bilinguals
111
0111
0111
0111
0111
0111
111p
Vol 8 No 17 1 December 1997 3811
FIG. 1. Intersubject variability in the cortical representation of language is greater in L2 than in L1. (a) Activity observed in single subjects
when listening to L1 or L2 plotted on a standardized atlas of cerebral neuroanatomy27
for purposes of inter-subject comparison. Each subject
was assigned a distinct color. Each disk indexes one active cluster of voxels, with size proportional to the volume of activation. Note the
tight overlap of activation from all 8 subjects in the left temporal lobe (including the temporal pole and posterior language area) when
listening to L1, which is not observed in the right hemisphere in L1 or in either hemisphere in L2. While listening to L2, no activation was
observed in the temporal poles or the angular gyrus, but some subjects showed additional activation in the left inferior frontal gyrus, and
in the anterior cingulate (not shown). In both L1 and L2, left and right temporal activation spares the region surrounding the primary audi-
tory area, which was presumably active in both the stories and the backward speech control. (b) Average active volume (mm3
) in anatom-
ical regions of interest individually defined on anatomical MR images. Only anatomical regions that were found active in at least three
subjects are reported. Abbreviations: STG: superior temporal gyrus; STS: superior temporal sulcus; MTG: middle temporal gyrus; TP: temporal
pole; AG: angular gyrus; SFG: superior frontal gyrus; MFG: middle frontal gyrus; PrG: precentral gyrus; Cin: anterior cingulate gyrus.

S. Dehaene et al.
1111
2
3
4
5
6
7
8
9
10111
1
2
3
4
5
6
7
8
9
20111
1
2
3
4
5
6
7
8
9
30111
1
2
3
4
5
6
7
8
9
40111
1
2
3
4
5
6
7
8
9
50111
1
2
3
4
5
6111p

in the left TP and AG. The remaining two subjects
showed a striking absence of activations in the left
temporal region while listening to L2. Only their
right temporal lobe was active. Hence those subjects
showed left-hemispheric dominance for comprehen-
sion in L1, but right-hemispheric dominance for
comprehension in L2. One such subject is depicted
in Fig. 2 (subject B). When listening to L1, this
subject showed intense activity in the left STS and
the left anterior temporal region, together with some
activation in the right anterior and middle temporal
lobe. When listening to L2, there was a complete
disappearance of activity in the left temporal lobe at
the chosen level of significance; the only significantly
active areas were found in the right STG, MTG and
TP, in a region anterior to, though partially over-
lapping with, that found when this subject listened
to L1.
Even when subjects showed left temporal activity
in L2, its volume was often smaller than in L1, and
listening to L2 activated additional small subregions
in the right temporal lobe (mostly the right STG
and STS; see e.g. Fig. 2, subject A, slice 6). To assess
whether this trend towards a reduced left-lateraliza-
tion for L2 than for L1 was significant despite the
considerable inter-subject variability, an analysis of
variance was performed on the volume of active tissue
in the left and right STS. A significant interaction
between hemisphere and language was found
(F(1,6) = 15.6, p = 0.0075). While listening to L1, the
average active volumes were 1378 mm3
in the left STS
and 456 mm3
in the right STS, a highly significant
asymmetry (F(1,6) = 20.6, p = 0.004). While listening
to L2, the asymmetry decreased to a much smaller,
though still significant value (left STS: 666 mm3
; right
STS: 327 mm3
; F(1,6) = 7.18, p = 0.037).
Variable activation while listening to L2 was also
observed in cerebral regions outside the temporal
lobe. Three subjects showed a highly specific activa-
tion of the left inferior frontal gyrus (Broca’s area)
and of the inferior precental sulcus in L2 which was
not found in L1 (see Fig. 1; Fig. 2, subject F). Four
subjects also showed activity in the left and right
anterior cingulate when listening to L2, but not when
listening to L1. Finally, six subjects showed a frontal
activation at the intersection of the IFS and the PrG
in L2, in a region very similar to that found in L1.
The qualitative differences between L1 and L2 seen
with correlation analysis were confirmed statistically
using an individual analysis with SPM. For instance,
the greater activation of the left STS during L1 than
during L2 in subject B (see Fig. 2) was confirmed
(two foci with Z = 4.50 and Z = 4.27), as was the L2-
specific activation of the left IFG in subject F (see
Fig. 2; Z = 4.66). Seven subjects showed discrete foci
of activation in the left or right temporal lobe for
which listening to L1 yielded significantly greater
activation (p < 0.001) than listening to L2, relative to
backward speech. Four subjects showed foci with the
converse difference (L2 > L1, p < 0.01). In many
cases, these language-specific foci were very near one
another. For instance, subject D showed one left
posterior STS focus where L1 > L2 (Talairach coor-
dinates –54, –57, 15; Z = 5.16), and another focus,
only 15 mm more anterior, where L2 > L1 (Talairach
coordinates –63, –42, 12; Z = 4.02). The exact anatom-
ical location of these foci, however, varied from
subject to subject.
Discussion
The pattern of activation in L1 replicated and
extended the results of previous PET studies of brain
activity while listening to single words11–15
and to
continuous speech in the first language.5,8,16
In the
latter studies, activity averaged across subjects was
reported in the left STG and MTG, the left posterior
temporal lobe, and the bilateral TP. The increased
anatomical accuracy afforded by the present method
shows that temporal lobe activity while processing
continuous speech in L1 is mostly concentrated
in the banks and depth of the STS, once activity in
primary and secondary auditory areas related to
acoustic processing is subtracted. The bilateral
temporal poles, which showed intense activity in
PET, were only modestly activated here probably
because of a loss of signal due to magnetic field in-
homogeneities in the inferior anterior temporal
region. Previous PET studies of listening to contin-
111
0111
0111
0111
0111
0111
111p
FIG. 2. Transaxial slices through the temporal and inferior frontal cortices of three subjects listening to French or English sentences alter-
nating with backward speech. The right side of images corresponds to the subject’s left. While listening to their native language, French,
subjects A (top panel) and B (middle panel) showed highly similar intense left-lateralized activations along the left superior temporal sulcus
(STS) extending anteriorily into the left temporal pole (TP) and posteriorily into the left angular gyrus (AG) (slices not shown). The right
STS and TP were also active. In subject F (bottom panel), activity was more strictly left-lateralized in the vincinity of the STS, with little or
no posterior and anterior extensions. When subjects listened to their second language, English, highly variable activation patterns were
found. In subject A, activation decreased considerably relative to the French condition in the left STS and failed to reach significance in the
left AG and left and right TP. An additional focus of activation was however found in the right STS (slice 6). In subject B, active pixels were
only found in the right temporal lobe, in sectors close to but not necessarily identical with those previously found active in French; no signif-
icant pixels were found in the left hemisphere. In subject F, activity was found in the left STS (in sectors partially non-overlapping with
those found in French), but also in the right STS and in the pars triangularis sector of the inferior frontal gyrus (IFG). In this subject, two
replications of each condition, each time with different stories in French and in English, demonstrated the reliability of the IFG activation
despite some variability in the number and exact localization of active pixels.

uous speech in L1 also reported weak activation of
the left inferior frontal gyrus (Brodman’s area 45)5,8,16
and of the left Brodman area 88
while listening to L1.
While these exact regions were not observed in the
present study with higher spatial resolution, perhaps
due to differences in subjects, languages and/or task
demands, a frontal activation was found at the inter-
section of IFS and PrS, bordering the regions activ-
ated in previous studies.
The highly reproducible left temporal activations
when listening to L1 confirm that a dedicated net-
work of left hemispheric cerebral areas, mostly
localized to the left STS, underlies native speech
comprehension.17
Our data indicate that, in late and
moderately proficient learners of L2, this network
fails to be consistently recruited for second-language
comprehension. Indeed, some subjects showed only
right-hemispheric activations in L2 in temporal areas
homologous to those observed in their left hemi-
sphere while listening to L1. Other subjects showed
activity in the left superior temporal sulcus in L2,
but with greater dispersion than in L1 and with occa-
sional differences in exact topography. A similar
distinction between subregions for L1 and L2 within
the same general anatomical areas has been observed
in a recent fMRI study of language production.18
Our
work shows that a similar dissociation is found
between speech comprehension in L1 and L2 in left
and right temporal areas.
In our work, several subjects showed significant
activations of the left inferior frontal gyrus and of
the anterior cingulate only when listening to L2, but
not when listening to L1. The anterior cingulate
region has been implicated in attentive, controlled or
‘central executive’ processing tasks19,20
suggesting that
some subjects had to engage greater attentional
resources for processing L2 than for the more autom-
atized processing of their maternal language. The
inferior frontal activation may have reflected a
strategy of internally rehearsing the English words
using an articulatory loop21
to maintain L2 sentences
in working memory while processing them.
Our results may contribute to clarify the complex
issue of aphasia in bilinguals. Based on a review of
neuropsychological cases, Albert and Obler1
origi-
nally suggested a decreased degree of leftward later-
alization for the second language. This conclusion
remains controversial, however, because counter-
examples are frequent.22,23
Our data provide some
grounds for reconciliation by showing both how, on
average, lateralization of activity to the left temporal
lobe is significantly reduced in L2 compared to L1,
and how individual subjects show considerable vari-
ation in their lateralization pattern for L2, anywhere
from complete right lateralization to standard left
lateralization.
In a previous PET study of a group of bilingual
subjects comparable to the present population, we
had failed to observe activation while listening to L2
above and beyond that observed for an unknown
language.5
The present results suggest that brain
activity related to lexical, syntactic and semantic
processing in L2 failed to emerge in that group study
because it was inconsistently localized from subject
to subject. Thus our observations emphasize an im-
portant methodological difficulty for brain imaging
studies based on group averaging. Such studies may
fail to identify active areas that are relevant to cogni-
tive function, but whose localization varies consid-
erably across subjects.
What factors might be responsible for inter-subject
variability in the anatomical representation of L2? All
our subjects were right-handed male bilinguals who
had acquired L2 after the age of seven and who had
achieved only a moderate level of proficiency in L2.
The dissimilarity between languages, which may be
an important factor in bilingualism,2
was also kept
constant (French vs English). Nevertheless, the
residual variability among subjects might be due to
the exact conditions under which they acquired L2.
Different methods of teaching L2 might favor
different strategies for language processing, and hence
distinct cerebral circuits. Acquisition of L1, on the
other hand, proceeds under very similar conditions
for all subjects.
Another potential factor of variability is a puta-
tive intrinsic difference in brain organization.
Interestingly, even in L1, the extent of right hemi-
spheric activation varied considerably across subjects,
and a significant correlation was found between the
volumes of right temporal activation in L1 and in L2
(r2 = 47.3%, p < .02). Hence, different subjects vary
in their use of the right hemisphere for language, be
it for L1 or L2. This intrinsic variability in language
lateralization, whose origins remain to be discovered,
appears to contribute to variability in L2 representa-
tion.
A third important factor, finally, may be the exact
age at which L2 was acquired. Indeed, previous
behavioral24
and brain-imaging6,18
evidence suggests
that maturational changes affect the ability to acquire
a second language. Further work is clearly needed to
clarify the respective influences of initial biological
architecture, timing of exposure to various languages,
acquisition method, and eventual proficiency in each
language, on the cortical organization of bilinguals.
In particular, studies of individuals who achieve a
high level of proficiency in a late-acquired second
language should help test whether there is a perma-
nent loss of plasticity in left temporal language areas
with age and maturation, or whether these areas can
be eventually recruited in fluent individuals following
S. Dehaene et al.
1111
2
3
4
5
6
7
8
9
10111
1
2
3
4
5
6
7
8
9
20111
1
2
3
4
5
6
7
8
9
30111
1
2
3
4
5
6
7
8
9
40111
1
2
3
4
5
6
7
8
9
50111
1
2
3
4
5
6111p

intense learning. Brain-imaging studies of motor
learning25,26
have revealed increased distributed
activity in prefrontal and parietal cortices in the initial
stages of acquisition of a motor skill, followed by a
concentration of activation to sensori-motor and
supplementary motor areas once skillful performance
is achieved. If second-language acquisition follows a
similar time course, one might expect a progressive
concentration of activations while listening to L2 to
the classical left perisylvian language network as
subjects move from a moderate to a high level of
proficiency in L2.
Conclusion
This study illustrates the feasibility of using fMRI to
study the cerebral networks involved in first and
second language processing in the human brain.
Reliable left temporal activations were found while
listening to L1, reproducing previous PET results.5,8,16
While listening to L2, activations were strikingly
more variable, with decreased left-lateralization or
even complete right lateralization. The present study
was not designed to determine whether the observed
differences between L1 and L2 were imputable to
phonological, prosodic, syntactic or semantic levels
of speech processing. Nevertheless, the results
support the hypothesis that first language acquisition
relies on a dedicated left-hemispheric cerebral net-
work, while late second language acquisition is not
necessarily associated with a reproducible biological
substrate.
References
1. Albert M and Obler L. The Bilingual Brain: Neuropsychological and
Neurolinguistic Aspects of Bilingualism. New York: Academic Press, 1978.
2. Paradis M. Aspects of Bilingual Aphasia. Oxford: Pergamon Press, 1995.
3. Ojemann GA. The Behavioral and Brain Sciences 2, 189–203 (1983).
4. Klein D, Zatorre RJ, Milner B et al. In: Paradis M. ed. Aspects of Bilingual
Aphasia. Oxford: Pergamon Press, 1995, 23–36.
5. Perani D et al. NeuroReport 7, 2439–2444 (1996).
6. Weber-Fox CM and Neville HJ. Journal of Cognitive Neuroscience 8,
231–256 (1996).
7. Yetkin O, Yetkin Z, Haughton VM and Cox RW. American Journal of
Neuroradiology 17, 473–477 (1996).
8. Mazoyer BM et al. Journal of Cognitive Neuroscience 5, 467–479 (1993).
9. Bandettini PA, Jesmanowicz A, Wong EC and Hyde JS. Magnetic Resonance
in Medicine 30, 161–173 (1993).
10. Friston KJ, Jezzard P and Turner R. Human Brain Mapping 1, 153–171
(1994).
11. Binder JR et al. Journal of Neuroscience 17, 353–362 (1997).
12. Démonet J-F et al. Brain 115, 1753–1768 (1992).
13. Petersen SE, Fox PT, Posner MI et al. Journal of Cognitive Neuroscience
1, 153–170 (1989).
14. Price C et al. Neuroscience Letters 146, 179–182 (1992).
15. Wise RJ et al. Brain 114, 1803–1817 (1991).
16. Bottini G et al. Brain 117, 1241–1253 (1994).
17. Chomsky N. Aspects of the Theory of Syntax. Cambridge: MIT Press, 1965.
18. Kim K, Hirsch J, Relkin N and Lee KM. Nature 388, 171–174 (1997).
19. Pardo JV, Pardo PJ, Janer KW and Raichle ME. Proc Natl Acad Sci USA
87, 256–259 (1990).
20. Posner MI and Dehaene S. Trends Neurosci 17, 75–79 (1994).
21. Paulesu E, Frith CD and Frackowiak RSJ. Nature 362, 342–345 (1993).
22. Berthier ML, Starkstein SE, Lylyk P and Leiguarda R. Brain and Language
38, 449–453 (1990).
23. Rapport RL, Tan CT and Whitaker HA. Brain and Language 18, 342–366
(1983).
24. Johnson JS and Newport EL. Cognitive Psychology 21, 60–99 (1989).
25. Seitz RJ and Roland PE. European Journal of Neuroscience 4, 154–165
(1992).
26. Jenkins IH, Brooks DJ, Nixon PD et al. Journal of Neuroscience 14,
3775–3790 (1994).
27. Duvernoy HM. Le Cerveau Humain. Surface, Coupes Sériées Tridimension-
nelles et IRM. Paris: Springer–Verlag, 1992.
ACKNOWLEDGEMENTS: We thank Olivier Crouzet, Elie Lobel and Eric Giacomini
for their help in data collection and analysis.
Received 5 September 1997;
accepted 11 September 1997
111
0111
0111
0111
0111
0111
111p

Dehaene dmcppmll 1997_anatomical_variability_first_&_second_languages.neuroreport

More Related Content

Similar to Dehaene dmcppmll 1997_anatomical_variability_first_&_second_languages.neuroreport (20)

Recently uploaded (20)

Dehaene dmcppmll 1997_anatomical_variability_first_&_second_languages.neuroreport