Content uploaded by Monica Baciu
Author content
All content in this area was uploaded by Monica Baciu on Feb 20, 2014
Content may be subject to copyright.
Chapter
FMRI Evidence for ACV98
Connectionist Model for Reading of
Mono and Multi-Syllabic Words and
Pseudo-Words
Author: Monica Baciu
1
Co-authors: Bernard Ans, Serge Carbonnel, Sylviane Valdois, Alexandra Juphard
Author and co-authors affiliation:
Laboratoire de Psychologie et Neurocognition, UMR CNRS 5105
Université Pierre Mendès-France,
BP 47, 38040 Grenoble Cedex 09
France
Abstract
According to predictions made by ACV98 connectionist model of reading, behavioral
experiments have shown that syllabic length affects naming latencies for pseudo-
words but not for words. ACV98 postulates the existence of two successive reading
procedures. According to it, any orthographical object (word or pseudo-word) is first
submitted to a global processing which aims to find a “lexical familiarity” computed
from previously experienced words. If it is found, the corresponding phonological
form of the stimulus is activated and reading is performed. If it is not found, a
subsequent analytical processing is performed within the input stimulus using the
same route, in order to extract familiar orthographic components (typically, syllables)
as well as their corresponding phonological forms. The syllabic phonological forms
are successively and temporarily maintained in a phonological buffer, before being
assembled into a whole phonological form. Overall, this model predicts that words
are read by using the global procedure, while the pseudo-words are read by using the
analytical procedure. The present event-related fMRI study aimed to assess the effect
of syllabic length on cerebral activity during reading, in order to obtain additional
1
Corresponding author :Monica Baciu, MD, PhD,Laboratoire de Psychologie et Neurocognition, UMR CNRS
5105,Université Pierre Mendès-France,BP 47, 38040 Grenoble Cedex 09, France, Phone : 00 33 4 76 82
78 07,Fax : 00 33 4 76 82 78 34,Email : mbaciu@upmf-grenoble.fr
Monica Baciu 2
anatomo-functional information as support to this model as well as to behavioral
results. Based on ACV98 predictions and in terms of cerebral network, we
hypothesized (1) a lexicality effect: there should be no differences between words and
pseudo-words, they are processed within a common network of cerebral regions, and
(2) a syllable length effect: the length influences cerebral activity only during pseudo-
words reading because only long pseudo-words are processed following an analytical
procedure involving supplementary visual analysis, visuo-spatial attention and
working memory processes. Eight right-handed volunteers performed a silent reading
task on French printed words and pseudo-words. A pseudo-randomized event-related
paradigm with 6 types (words=W and pseudo-words=PW composed of 1, 2 and 3
syllables) of stimuli was used. Data processing was performed using SPM’99. Our
results have shown that (1) Words and pseudo-words involved common mechanisms
such as visuo-orthographic, phonological, attentional and motor processes. No
process was significantly more involved for one or the other of two types of stimuli
(word or pseudo-word). This result is a priori more in agreement with “single way
models” and particularly with ACV’98 than with dual-route models. Nonetheless the
dual-route model cannot be excluded; (2) A length effect on cerebral activity was
obtained only during pseudo-word reading suggesting an analytical procedure
involvement during reading of long pseudo-words, as ACV’98 predicts.
Introduction
Understanding cerebral mechanisms of reading means understanding how the
orthographic form of visual linguistic stimuli is translated into their phonological form.
Several theoretical models were developed and tried to answer this question. The classical
cognitive “dual-route” model (Coltheart et al., 2001) for reading postulates the existence
of two completely isolated routes that work in parallel and rely on radically different
computational principles: one “lexical-semantic” used for the processing of words and
another “phonological” used mainly for the sequential processing of pseudo-words
(grapheme-to-phoneme rules). According to dual-route model, two different networks of
cerebral regions will be expected to be activated, one for words and another one for
pseudo-words, independently on the item length. These differences are explained in terms
of sub-lexical phonological processing required by pseudo-word reading with respect to
word (even for monosyllabic stimuli). Although the psychological models of reading do
not directly predict that length has an effect on cerebral activation, we can nonetheless
formulate hypotheses from effects expected at the behavioural level and from cerebral
imaging data on reading. Therefore, within the framework of dual-route models
(Coltheart et al., 2001), a length effect could be predicted during pseudo-word reading
only. Indeed, the phonological form of pseudo-words can only be obtained from the
sequential processing using the non lexical route (grapheme-to-phoneme translation).
Furthermore, reading words would mainly be performed by using the lexical route. As
this route allows to access directly the global phonological form of the word, no length
effect is predicted for words. Based on this functional architecture of the dual-route
model, we can assume that the reading of words and pseudo-words would involve
different cerebral regions underlying each of the two routes. Cerebral regions allowing
accessing orthographical and phonological, as well as semantic representations, such as
occipito-temporal (Dehaene et al., 2002; Fiebach et al., 2002; Jobard et al., 2003; Price et
al., 1994), temporo-parietal and inferior frontal cortex (Fiez and Petersen, 1998, St
George et al., 1999; Xu et al., 2002) would mainly be related to the lexical route. As for
FMRI Evidence for ACV8 Connectionist Model for Reading … 3
the non lexical route, it may implicate regions related to grapheme-phoneme
transformation, such as the left superior temporal region (Price et al., 1994; Jobard et al.,
2003). The non-lexical route would additionally require the working memory processes in
order to store, maintain and manipulate extracted phonological segments. The
left supramarginal gyrus and the dorsal portion of the inferior frontal cortex were shown
as being related to these processes (Paulesu et al, 1993; Gold and Buckner, 2002; Jobard
et al., 2003; Thierry et al., 1999). Although the dual-route model did not make
predictions on syllabic length, one could suppose that increasing the pseudo-words length
should more intensively activate regions inherent to the non lexical route, whereas
increasing the words length would have no influence on the degree of activation of
regions underlying the lexical route.
Several connectionist models of reading were subsequently developed and they all
proposed a unitary theory for reading. These models postulate that all orthographic items
are read in the same way, i.e. the pronunciation is always computed from knowledge
about previously experienced exemplars (familiarity). For instance, within PDP (Parallel
Distributed Processing) models (Plaut, et al., 1996; Seidenberg and McClelland, 1989),
the reading of words and of pseudo-words is performed via a unique global processing
way, without involving any serial process In consequence, a common network of cerebral
regions involved in orthographic (occipito-temporal cortex), phonological (posterior
superior temporal and supramarginal cortex, as well as the dorsal part of the left inferior
prefrontal gyrus) and semantic (middle and inferior temporal cortex and angular gyrus, as
well as the anterior inferior part of the inferior prefrontal gyrus) processes, should be
involved, whatever the lexical nature of the items or their length is. Depending on
orthographic and phonological characteristics of the items, a length effect could emerge.
It would then consist in a more significant activation of this cerebral network for the
longest items in comparison with the shortest items. However, the increased activation of
this cerebral network, due to the increase in length, should not differ between words and
pseudo-words, the latter being processed by a common mechanism.
Among the unitary conception of reading, the model developed by Ans et al. (Ans, et
al., 1998) (thereafter ACV98), is the only one that can simulate reading of multi-syllabic
items, the others being restricted to a monosyllabic processing. It postulates the existence
of two sequential reading procedures. According to it, any orthographical object (word or
pseudo-word) is first submitted to a global processing which aims to find a “lexical
familiarity” computed from previously experienced words. If it is found, the
corresponding phonological form of the stimulus is activated and reading is performed. If
it is not found, a subsequent analytical processing is performed within the input stimulus
using the same route, in order to extract familiar orthographic components (typically,
syllables) as well as their corresponding phonological forms. These syllabic phonological
forms are successively and temporarily maintained in a phonological buffer, before being
assembled into a whole phonological form. This model predicts that words are read by
using the global procedure, while pseudo-words are read by using the analytical
procedure. The simulations done with the ACV98 model have shown that monosyllabic
pseudo-words also required a short analytical processing but the differences between
monosyllabic words and monosyllabic pseudo-words have not been detected in
behavioural experiments, in terms of time responses. This can be explained by the weak
analytical procedure required by monosyllabic pseudo-words (only one syllable to be re-
analyzed after the prior unsuccessfully global analysis) with respect to multi-syllabic
pseudo-words. In terms of cerebral neuroanatomy of language and based on ACV98
Monica Baciu 4
predictions, the reading of the two types of items should generate a common network
activation of cerebral regions related to the global procedure, mainly related to
orthographic (occipito-temporal cortex) and phonological (posterior superior temporal
and supramarginal cortex, as well as the dorsal part of the left inferior prefrontal gyrus)
processes. Although not predicted by the model, semantic processing (middle and inferior
temporal cortex and angular gyrus, as well as the anterior inferior part of the inferior
prefrontal gyrus) could be also involved. Furthermore, as the model predicts, the
analytical procedure is activated when long pseudo-words are presented. It implies
additional processes such as a supplementary visuo-orthographic analysis (striate and
extrastriate occipital areas) and visuo-spatial attention (superior parietal lobule). Those
two types of processes would be indicative respectively of the reduction of the size of the
visuo-attentional window and its shift to the different segments of the input item
(typically syllables). This analytical procedure would also involve the verbal working
memory (left inferior frontal cortex, Broca's area, BA 44 and temporo-parietal cortex,
Wernicke's area, BA 22 ; supramarginal gyrus, BA 40), more particularly related to the
subvocal articulation, phonological process and storage.
In this chapter, we present an fMRI study which aims to assess the cerebral regions
that underlie silent reading of words and of pseudo-words of different syllabic lengths.
According to our theoretical frame of reference, the ACV98 multi-trace model, the
syllabic length effect should impact on cerebral activation only during pseudo-word
reading.
More specifically, according to the ACV98 model and in terms of cerebral activation,
we formulated the following hypotheses:
Firstly, independently of the syllabic length, as words and pseudo-words are
processed within a unique system, we expect to observe the activation of a common
network of cerebral regions for both types of items. More precisely, this network should
include the occipito-temporal cortex (related to orthographic processing) as well as the
temporo-parietal cortex and the anterior part of the inferior frontal cortex, Broca's area
(related to phonological processing). The semantic processing is not predicted by
ACV’98 model as being essential for reading, but regions classically related to this
process, as well as to other auxiliary processes (such as motor processes related to the
preparation and the control of the motor act of speech, even if speech was inner in this
study), could be observed and included within the common network. We will refer at this
first assumption, as the lexicality effect.
Secondly, we expect to observe a length effect on the cerebral activation for pseudo-
words only. Increasing syllabic length of pseudo-words, implies the analytical procedure,
meaning, as described previously, the reduction of size and the shift of visuo-attentional
window, sequentially from one segment (typically the syllable) of the visual input, to the
next one, till the whole visual item is processed. Verbal working memory will be involved
and responsible for generating phonological form of each syllable and maintaining them
within the phonological buffer in order to constitute the whole phonological form of the
visual input. Consequently, the length effect during pseudo-word reading, should be
illustrated by stronger involvement of visual associative (~ BA 18,19), of superior parietal
(~ BA 7), supramarginal (~ BA 40) and Broca's (~ BA 44, 45 left) regions for long
pseudo-words when compared to short pseudo-words and long words. We will refer at
this second assumption as the length effect.
FMRI Evidence for ACV8 Connectionist Model for Reading … 5
Material and Methods
Subjects
Eight healthy volunteers (5 women, average age 23.6 years) gave their informed
consent to participate at this experiment that was approved by the local ethics committee.
All the subjects of this experience were right-handed as assessed by the Edinburgh
Handedness inventory (Oldfield, 1971) and they were all French native speakers and had
no visual or reading problems.
Stimuli and Procedure
A pseudo-randomized ER-fMRI paradigm with six types of stimuli (words composed
of 1, 2 and 3 syllables and pseudo-words composed of 1, 2 and 3 syllables), was used.
Overall, 144 words and pseudo-words composed of one (3-5 letters), two (6-8 letters) and
three (9-11 letters) syllables were presented during a single fMRI session. The words
were mid-frequency regular words while the pseudo-words have been obtained by
manipulating words. In addition to words and pseudo-words, 34 null-events (ten of them
at the end of the session) composed of a blank screen and a fixation cross on the center of
the screen were also included. A fixation cross was presented between stimuli on the
center of the screen. The presentation time of each stimulus was 200 ms and the average
interval between the stimuli of the same type was 14 seconds. The stimuli were written in
black lower-case letters on a white screen (« Courier new 45 points » bold). The stimuli
were generated by means of Psyscope V.1.1 (Carnegie Mellon Department of
Psychology) on a Macintosh computer (Power Macintosh 9600). They were transmitted
into the magnet by means of a video projector (Eiki LC 6000), a projection screen
situated behind the magnet and a mirror centered above the patient’s eyes.
The subjects were instructed to read each item silently, without articulating and
vocalizing.
MR acquisition
Functional MR imaging was performed on a 1.5 Tesla MR imager (Philips NT) with
echo-planar (EPI) acquisition. Twenty-three adjacent, axial slices (thickness 5 mm each)
were imaged. The imaging volume was oriented parallel to the bicommissural plane.
Positioning of the image planes was performed on scout images acquired in the sagittal
plane. An EPI MR pulse sequence was used. The major MR acquisition parameters of the
EPI sequence were: TR = 2000 ms, TE = 45 ms, flip angle = 90°, field-of-view =
256x256 mm2, imaging matrix = 64x64, reconstruction matrix = 128x128. Subsequent to
the functional scan, a high resolution 3D anatomical MR scan was obtained from the
volume previously examined.
Data Processing
Monica Baciu 6
Data analysis was performed based on the general linear model (Friston, et al., 1995)
for event-related designs, implemented in SPM’99 software (Wellcome Department of
Imaging Neuroscience, London, UK) running on a Unix workstation under the MATLAB
environment (Mathworks, Sherbon, USA).
A. Spatial Pre-Processing
MR images were processed using the following steps. First, during the slice timing
step, the functional volumes were corrected for sampling bias effects caused by the
different time acquisition of each slice composing the functional volume, relative to the
haemodynamic response. In the second step, the realignment, motion correction was
applied by using rotations and translations in order to realign each functional volume to
the first acquired one. In a third step, the anatomical volume was spatially normalized
into the Talairach and Tournoux reference space (Talairach and Tournoux, 1988) using as
template a representative brain from the Montreal Neurological Institute. The anatomical
normalization parameters were subsequently applied to functional volumes. Finally, in
conformity with the assumption that data are normally distributed, the functional images
were spatially smoothed by using a Gaussian filter (8 mm width).
B. Statistics
For each type of individual events, regressors of interest were created by convolving
a delta function at each event onset with a canonical haemodynamic response function. A
fixed effect group analysis was performed. Clusters of activated voxels were then
identified, based on the intensity of the individual responses (p<0.05 corrected and
p<0.001 uncorrected for multiple comparisons) and the spatial extent (clusters composed
of at least 5 voxels).
B.1. Evaluation of cerebral substrates for word and pseudo-word
reading (lexicality effect)
In order to assess the neural substrates involved during word and pseudo-word
reading, we calculated the following contrasts: W vs. fixation (regions related to word
reading, PW vs. fixation (regions related to pseudo-word reading), W vs. PW (regions
significantly more activated during word than during pseudo-word reading) and PW vs. W
(regions significantly more activated during pseudo-word than during word reading).
B.2. Evaluation of Cerebral Substrates for Mono and Three-Syllabic
(Length Effect) Word and Pseudo-Word Reading
We focused only on the extreme stimuli, meaning words and pseudo-words
composed of one and three syllables. In order to assess the neural substrate presenting a
length effect, we calculated the following contrasts: W1 vs. W3 (regions significantly
more activated for short word reading with respect to long words), W3 vs. W1 (regions
significantly more activated for long word reading with respect to short word), PW1 vs.
PW3 (regions significantly more activated for short pseudo-word reading with respect to
long pseudo-word reading) and PW3 vs. PW1 (regions significantly more activated for
long pseudo-word reading with respect to short pseudo-word reading). By taking into
FMRI Evidence for ACV8 Connectionist Model for Reading … 7
account the lexical type (word or pseudo-word), the contrast W3 vs. PW3 was performed
in order to assess regions significantly more activated for long word than long pseudo-
word reading, while the contrast PW3 vs. W3 has highlighted regions significantly more
activated for long pseudo-word than long word reading. We also calculated the following
contrasts (interactions): (PW3 vs. PW1) vs. (W3 vs. W1) in order to assess if a possible
length effect for pseudo-words is significantly greater than a possible length effect for
words, and (W3 vs. W1) vs. (PW3 vs. PW1) in order to assess if a possible length effect
for words is significantly greater than a possible length effect for pseudo-words.
Results
All activations reported in this sub-section were obtained based on a fixed effect
group analysis.
A. Lexicality Effect
As shown in Figure 1, independently of the item length, the same cerebral regions
were activated during word (W vs. fixation) and pseudo-word (PW vs. fixation) reading.
Overall, these regions mentioned in detail in Table 1, were activated bilaterally: the
infero-medial occipito-temporal gyrus, the superior parietal lobule, the left inferior
parietal lobule (supramarginal and angular gyrus), the motor and premotor cortices, as
well as the right cerebellum. The direct contrast between words and pseudo-words W vs.
PW and PW vs. W did not show any cerebral region activated significantly more by one
type of item with respect to the other one.
Figure 1 Lexicality effect: The images (lateral views of the left hemisphere on the right and of the right
hemisphere on the left) represent in yellow the commonly activated cerebral regions during word and
pseudo-word reading, independently of their length. More precisely, the contrast W vs. Fixation should
induce activation represented in « green », while the contrast PW vs. Fixation should induce activation
represented in « red ». All activations represented on these images are « yellow » suggesting that all
regions were commonly activated by words and pseudo-words. The activated regions are shown in Table
1. LH = left hemisphere, RH = right hemisphere.
Monica Baciu 8
B. Syllabic Length Effect
During reading, three-syllabic words (W3) did not induce significantly more
activation when contrasted to mono-syllabic words (W1). Three-syllabic pseudo-words
(PW3) induced significantly greater cerebral activity when contrasted to monosyllabic
pseudo-words (PW1) bilaterally within middle and inferior occipital gyrus, inferior and
superior parietal lobule and cerebellum. The premotor cortex was activated to the left
whereas the middle and inferior frontal gyrus were activated to the right (Figure 2, Figure
3 and Table 2).
Table 1. Significantly ctivated regions obtained during word and pseudo-word reading (W vs. fixation
and PW vs. fixation). H corresponds to the hemisphere with L for activation lateralized within the left
and R for activation lateralized within the right hemisphere; BA represents the approximate Brodmann
Area corresponding to the activated region; p=significance value; x, y and z indicate for each activated
region, the Talairach coordinates.
Table 1
FMRI Evidence for ACV8 Connectionist Model for Reading … 9
Contrast Location
Talairach coordinates
Region H
~ BA
p x y z
[W vs. Fixation]
Cerebellum R
_
5.04
4 -71 -17
Middle occipital gyrus R
19
7.20
44 -74 -5
L
19
9.61
-40 -66 -5
Fusiform gyrus L
37
7.9 -44 -55 -14
Inferior occipital gyrus L
18
7.43
-32 -82 -0
Cuneus L
17
7.34
-24 -85 4
Superior parietal lobule L
7
6.35
-24 -59 58
Inferior parietal lobule L
40
6.13
-48 -40 53
Precuneus R
7
5.24
28 -56 40
Somesthesic primary cortex
R
1, 2, 3
4.81
40 -25 43
Primary motor cortex L
4
5.82
-44 -1 51
Premotor cortex L
6
5.07
-4 7 60
[PW vs. Fixation]
Middle occipital gyrus L
19, 37
9.98
-40 -67 -9
Superior parietal lobule L
7
8.96
-24 -59 58
Inferior occipital gyrus L
18
8.27
-32 -82 -0
Inferior parietal lobule L
40
7.20
-48 -40 53
Cuneus L
17
6.65
-16 -85 4
Cerebellum R
_
7.50
32 -59 -18
Middle occipital gyrus R
19
7.05
40 -74 -5
Lingual gyrus R
18
5.58
24 -78 -0
Superior parietal lobule R
7
6.37
28 -55 58
Premotor cortex L
6
7.56
-51 -2 37
Primary motor cortex L
4
6.16
-44 -1 51
Figure 2 Syllabic length effect during pseudo-word reading: The images (lateral views of the left
hemisphere, LH on the right and of the right hemisphere, RH on the left) show regions (in red)
significantly more activated during reading long pseudo-word reading with respect to short ones (PW3
vs PW1).
Monica Baciu 10
Figure 3 Syllabic length effect during pseudo-word reading: Functional maps showing some
activated regions (in red) provided by the contrast PW3 vs. PW1 during reading. The activation was
projected onto 2D axial anatomical slices, at different levels (Z) with respect to the bicommissural plane.
Functional maps are shown in neurological convention (left hemisphere, LH on the left side of the
image). The activated regions are mentioned in Table 2 and they are indicated on the images by using
numbers: 1= visual primary and associative cortex, 2 = Broca’s area (left inferior frontal gyrus), 3 = left
supramarginal gyrus, 4 = superior parietal lobule, 5 = right dorsolateral prefrontal cortex.
The length effect obtained during pseudo-word reading was reinforced by results
provided by the interaction between lexical nature and syllabic length ([PW3 vs. PW1] vs.
[W3 vs. W1]). This contrast has highlighted significantly activated regions bilaterally
within the middle and inferior occipital gyri, the superior parietal lobule, as well as the
right middle frontal, the left inferior frontal (Brodmann's area, BA 44) and left
supramarginal (BA 40) gyri (Figure 4 and Table 2). These activations suggest that the
length effect for pseudo-word reading is significantly higher than a possible, non-
detectable length effect for words.
Figure 4 Interaction between lexical nature and syllabic length during reading:
The
images (lateral views of the left hemisphere, LH on the right and of the right hemisphere, RH on the left)
show activated regions for the contrast ([PW3 vs.PW1] vs. [W3 vs. W1]) during reading. The activated
regions are mentioned in Table 2.
Table 2. Significantly activated regions obtained by contrasting PW3 vs. PW1 and ([PW3 vs. PW1] vs.
[W3 vs. W1]). H corresponds to the hemisphere with L for activation lateralized within the left and R for
activation lateralized within the right hemisphere; BA represents the approximate Brodmann Area
corresponding to the activated region; p=significance value; x, y and z indicate for each activated region,
the Talairach coordinates.
Table 2
FMRI Evidence for ACV8 Connectionist Model for Reading … 11
Contrast Location
Significance
Talairach coordinates
Region H
~ BA
p
x y z
[PW3 vs. PW1]
Lingual gyrus L
18
6.68
-12 -81 4
L
19
5.51
-20 -70 -5
Cerebellum L
_
6.00
-32 -67 -13
Cuneus L
18
4.86
-8 -69 17
R
18
6.17
16 -73 17
Lingual gyrus R
18
5.57
12 -74 -1
Cerebellum R
_
5.27
36 -67 -18
Superior parietal lobule
R
7
6.85
36 -48 53
L
7
5.31
-24 -56 53
Inferior parietal lobule L
40
5.32
-48 -40 48
R
40
4.87
51 -33 43
Middle frontal gyrus R
46
5.43
44 47 16
Premotor cortex L
6
5.42
-48 2 41
Inferior frontal gyrus R
47
5.00
48 46 -7
[PW3 vs. PW1] vs. [W3 vs. W1]
Cerebellum R
18, 19
4.56
4 -62 -1
Middle occipital gyrus R
19
4.36
36 -78 -9
L
19, 37
3.87
-40 -63 -9
R
39, 19
3.84
44 -70 8
Lingual gyrus L
18
3.87
-28 -70 -1
Cerebellum L
_
3.74
-12 -70 -9
Cuneus L
17
3.83
-12 -81 9
R
18
3.74
16 -73 17
Precuneus R
7
3.95
8 -44 57
R
31
3.45
4 -72 27
Paracentral lobule R
5
3.78
16 -21 47
Middle frontal gyrus R
9
3.97
32 41 35
R
8
3.25
36 33 40
R
10, 46
3.38
32 55 6
Discussion
The aim of the present event-related fRMI was to evaluate the pattern of neural
activation induced by reading mono and multi-syllabic words and pseudo-words, in order
to assess the effect of the syllabic length on the cerebral activation, based on ACV’98
Monica Baciu 12
predictions. According to ACV98 model predictions, we evaluated the lexicality effect
(we expected to obtain the activation of a common network of cerebral regions during
both word and pseudo-word reading), as well as the syllabic length effect (we expected
that the syllabic length influences only pseudo-word reading, related to the supplementary
analytical procedure required for reading long pseudo-words).
Lexicality Effect
Overall, words and pseudo-words activated a common network of cerebral regions.
This network included the left primary visual area (BA 17), the inferior and middle
temporo-occipital gyrus (BA 18, 19, 37) as well as the inferior (SMG, BA 40) and
superior (precuneus, BA 7) parietal lobule, bilaterally. As stimuli were visually presented,
the primary visual area, responsible for visual perceptive processing, was activated
(Cornelissen et al., 2003; Petersen et al., 1988, 1990, Tarkainien et al., 1999; Wise et al.,
1991). The activation of the middle and inferior temporo-occipital gyrus could be related
to orthographic processing (Fiez and Petersen, 1998; Petersen et al., 1988, 1990; Rumsey
et al, 1997), whereas the inferior parietal lobule (supramarginal gyrus) would reflect the
phonological processing (Devlin et al., 2003; Mc Dermott et al., 2003). The activation of
the superior parietal lobule could be related to visuo-spatial attention processes (Chen et
al., 2002; Corbetta et al., 1993; ; Corbetta et al., 1998; Simon et al., 2002) while the right
cerebellum and the left premotor cortex (BA 6) activation could be related to motor
aspects, such as the coordination and preparation of articulation in relation to the inner
speech (Fiez, 1997; Fiez and Petersen, 1998; Haggort et al., 1999; Wise et al., 1991).
Moreover, it was hypothesized that the cerebellum could be responsible for the control of
attention (Ravizza and Ivry, 2001). Overall, the activation of the cerebellum obtained here
could be attributed quite as much to the motor control of linguistic production (notably its
medial part) as to the cognitive processes (notably its lateral part), such as attention and
phonology (Ravizza and Ivry, 2001; Fiez, et al., 1996).
The activation of a common network of cerebral regions during the reading of words
and of pseudo-words is in line with previous studies showing that similar regions are
activated during the processing of those two types of stimuli (Chen, et al., 2002; Rumsey
et al., 1997; Tagamets, et al., 2000). In return, no cortical region was significantly more
implicated in the processing of one of the two types of item. This result is in disagreement
with the idea suggesting that words and pseudo-words activate different regions (Chen et
al., 2002; Fiez et al., 1997; Fulbright et al., 1999; Hagoort et al., 1999; Herbster et al.,
1997; Rumsey et al., 1997; Xu et al., 2001). However, we can not completely exclude
other reasons which could, even partially, have explained the absence of differential
activation between words and pseudo-words. For instance, the very conservative
statistical threshold of significance (p corrected < .05) used for this comparison,
combined with the poor sampling of subjects having participated to this experiment,
could explain this result. Moreover, even if the event-related paradigm, used in the
present study, offers the advantage of allowing a randomized presentation of different
types of stimuli, it is less sensitive in detecting activation than the « block » paradigm.
Also, the limited number of items of each type could have also played a role in the
absence of activation between words and pseudo-words.
The common network of regions activated for words and pseudo-words suggest
common mechanisms to process the two types of stimuli. With respect to various models
FMRI Evidence for ACV8 Connectionist Model for Reading … 13
of reading, the intervention of common processes for word and pseudo-word reading, is
rather in agreement with the “one way models” for reading, such as PMSP96 (Plaut, et
al., 1996) and ACV98 (Ans, et al., 1998), that postulates the existence of a single system
to process both, words and pseudo-words. However, it does not invalidate the dualistic
approach for reading, underlying dual-route models, according to which words and
pseudo-words would be processed within distinct systems. In fact, the fMRI spatial
resolution, although relatively satisfying by comparing with other techniques such as the
MEG or the ERP, techniques is not sufficient to guarantee the absence of intervention of
different neuronal populations within this common cerebral network, specific to each
types of item.
Syllabic Length Effect
Our main result is that the comparison of the longest items (3 syllables) with the
shortest ones (1 syllable) showed an influence of the increase of the syllabic length on the
cerebral activity only during the reading of pseudo-words. This result is in agreement with
the ACV98 reading model. It is also in agreement with the dual-route model (Coltheart et
al., 2001), insofar as the number of letters increased with the number of syllables. On the
other hand, within the framework of the PMSP96 model (Plaut et al., 1996), as words and
pseudo-words are processed by a common procedure within a unique way, if a length
effect appears, it should affect words as much as pseudo-words.
Compared to short pseudo-words, long pseudo-words induced an increased cerebral
activity in a certain number of regions that are classically attributed to phonological,
visuo-orthographic and visuo-spatial attention processing. More precisely, the bilateral
inferior parietal lobule (supramarginal gyrus, BA 40) was more intensively activated
during the reading of long than of short pseudo-words. This region, especially to the left,
is typically associated with phonological processing, more particularly involved in
phonological storage (Démonet et al., 1992; Fiez and Petersen, 1998; Fiez et al., 1999;
McDermott et al., 2003; Gold and Buckner, 2002; Paulesu et al., 1993; Xu et al., 2002).
The increased activation of the bilateral inferior parietal lobule suggests that phonological
analysis was more intensively involved while processing long pseudo-words than short
pseudo-words. Such an increase in activation, predominantly within the left hemisphere,
could reflect a stronger involvement of phonological storage component of the verbal
working memory during long pseudo-words reading (Gold and Buckner, 2002; Paulesu et
al., 1993; Thierry et al., 1999). This activation is therefore in line with the ACV98 model
(Ans et al., 1998) that postulates that the different syllables successively generated during
the analytical procedure of long pseudo-words are maintained within the working
memory in order to be assembled into a whole phonological form. According to this
model, the working memory is not much involved during reading of short pseudo-words
and of short and long words. The activation of the supramarginal gyrus is also in
agreement with the dual-route model. In fact, this model postulates the short term
maintenance of different phonemes extracted following the grapheme-to-phoneme rules,
during the non lexical processing required to read pseudo-words.
In addition, long pseudo-words induced significantly greater activation, than short
pseudo-words, within extrastriate visual cortex (BA 18, 19). The bilateral activation of
the extrastriate visual cortex is typically associated with visual processing of orthographic
sequences (Fiez and Petersen, 1998). Indefrey et al. (1997) have explicitly studied the
Monica Baciu 14
length effect of pseudo-words and of false characters of the same length, on the activation
pattern of this region. Activation of the medial extrastriate cortex was found for each type
of stimuli with respect to the baseline. On the other hand, those activations disappeared
when pseudo-words were compared to sequences of false characters of same length. The
authors concluded that the length difference in itself was sufficient to account for the
extrastriate activation, suggesting that this region was mainly sensitive to physical
characteristics of visual input. However, in the present study, the increased activation of
extrastriate regions was obtained only by contrasting long pseudo-words with short
pseudo-words. Nevertheless, it was not observed when we compared the long words with
short words. This result therefore contradicts the hypothesis according to which the
medial extrastriate activations are generated by the differences in physical length between
items. In fact, in agreement with this hypothesis, increased activations of those regions
should have been obtained for all the comparisons between the longest and the shortest
items. According to other studies, the increased activation of extrastriate region could
rather be related to greater visual complexity of the input stimuli (Hagoort et al., 1999;
Tagamets et al., 2000; Tarkainien et al., 2002), rather than merely their physical
characteristics. Such an interpretation has proven more consistent with the ACV98 model
insofar as it postulates a visual processing, the degree of complexity of which increases
with the increase in the pseudo-word length. Actually, according to this model, long
pseudo-words would be processed syllable by syllable following the analytical procedure.
The processing of each new syllable would require a new visuo-attentional capture. On
the other hand, a strong implication of extrastriate regions during reading of long pseudo-
words is not directly predicted by the DVC2001 model. This latter hypothesizes that
visual analysis component (i.e., the layers of visual features and of letters) is common to
lexical and non-lexical reading procedures. In consequence, a similar visual processing is
supposed to apply to both words and pseudo-words. However, the non lexical route
involves a segmentation mechanism that fractionates the internal representation of the
orthographic input sequence into its different letters before the application of the
grapheme-to-phoneme rules. If we hypothesize that those areas are activated during the
manipulation of internal orthographic representations, such an orthographic segmentation
could probably then result in an increased activation of the visual extrastriate areas. The
increased activation of the extrastriate cortex could also result from the increase in visual
attention (Clark et al., 1997; Corbetta et al., 1993). In fact, the increase in concomitant
activation of the bilateral superior parietal lobule (BA 7) for long pseudo-words compared
with short pseudo-words suggests a more extensive involvement of visuo-spatial attention
processes during the analytic procedure using for long pseudo-words. The superior
parietal area would be involved in the visuo-spatial analysis and attention (Behrmann, et
al., 2004; Corbetta et al., 1998; Corbetta, et al., 1993; Posner and Petersen, 1990) and
would be sensitive to the attention level allocated to a particular task (Fiez et al., 1995).
Tagamets et al. (2000) have obtained parietal activations for sequences of false characters
and of illegal letters but not for words and for pseudo-words composed of four letters.
They argued that the activation of the superior parietal cortex (BA 7) could reflect the
spatial processing necessary for organizing symbols in the unfamiliar sequences and/or an
additional cognitive cost to process the new stimuli. The increased activation of this
region, during three-syllabic pseudo-word reading, could therefore be related to the task
properties that require a spatial processing allowing serial shifts of attention from one
syllable to the other, within a sequence of letters. The activation of the superior parietal
lobule could also be explained by the relative increase in eye movements in order to
FMRI Evidence for ACV8 Connectionist Model for Reading … 15
process long and unfamiliar sequences (Chen et al., 2002). However, since eye
movements have not been measured, it is impossible for us to know if the activation of
parietal regions reflects the direction of attention or the eye movements. In both cases, the
activation of the superior parietal lobule for long pseudo-words reading in comparison
with short pseudo-words is in line with the ACV98 model. The latter assumed that the
analytical reading of polysyllabic pseudo-words implies a serial shift of the visuo-
attentional window from the first to the last syllable of the input. In consequence, the
analytical procedure would involve a redirection of both the attention and the eye
movements.
Activations of the superior parietal lobule (BA 7) have also been reported in studies
related to working memory, in particular when the task requires visual processing
(Cabeza and Nyberg, 2000). The activation of the superior parietal lobule and of the
extrastriate regions could then reflect an increase in visual attention within the context of
short term memory processes. This latter alternative is in line with the additional
activations of the inferior parietal lobule (BA 40) typically involved in phonological
storage (Gold and Buckner, 2002; Paulesu et al., 1993 ; Thierry et al., 1999).
The left premotor cortex (BA 6), probably associated with the programming of
articulation was activated. Its activation could also be interpreted as related to the
articulatory loop (together with Broca’s area) within the framework of verbal working
memory. The activation of the right middle (BA 46) and inferior frontal (BA 47) gyri is
more difficult to interpret. This region could be related to “central administrator” within
the framework of working memory but this region was generally activated to the left
during verbal working memory. Furthermore, the ACV98 model predicts a “comparison”
between external visual input and an internal visual input generated after recognizing the
visual item within the lexicon. This comparison has a major role in triggering the
analytical procedure: when external and internal visual input is the same, the analytical
procedure is not activated, when they are different, this procedure is activated. The
“comparison” process should depend on one or several cerebral regions. It is well known
that the frontal/prefrontal cortex is very much involved in control and strategy processes,
that is why this region could be involved, based on ACV’98 predictions, as “comparator”
for triggering the analytical procedure. The increase in item’s length should solicit more
intensively the “comparator”, especially when long items are pseudo-words.
Conclusion
The main results of the present fMRI study were:
a) Word and pseudo-word reading involves common processes
such as visuo-orthographic, phonological, attentional and motor
processes. No process was significantly more involved for one
or the other of two types of stimuli (word or pseudo-word). This
result is a priori more in agreement with “single way models”
than with “two distinct processing ways models”. Nonetheless
the dual-route model cannot be excluded, insofar as we can
consider that lexical and non lexical procedures involve
different neural populations within the commonly activated
regions obtained in this study.
b) We have obtained a syllabic length effect on cerebral activity
only during pseudo-word reading and it represents the main
Monica Baciu 16
result of the present study. Within the framework of the ACV98
model, the activation of such regions can easily be attributed to
analytical procedure involvement, which is predominantly
activated while reading pseudo-word.
We will evaluate, in a subsequent study, a larger number of subjects in order to carry
out randomized effect analysis. In addition, the amplitude, the form and the time of the
hemodynamic response will be evaluated within the different activated cerebral regions,
so as to detect more subtle differences, with respect to both, the lexical nature and the
syllabic length.
References
Ans, B., Carbonnel, S. and Valdois, S. (1998). A connectionist multiple-trace memory
model for polysyllabic word reading. Psychological Review 105(4), 678-723.
Behrmann, M., Geng, J. J. and Shomstein, S. (2004). Parietal cortex and attention.
Current Opinion in Neurobiology, 14(2), 212-217.
Cabeza, R. and Nyberg, L. (2000). Imaging cognition II: An empirical review of 275 PET
and fMRI studies. Journal of Cognitive Neuroscience, 12, 1-47.
Chen, Y., Fu, S., Iversen, S. D. et al. (2002). Testing for dual brain processing routes in
reading: A direct contrast of Chinese character and pinyin reading using fMRI. Journal of
Cognitive Neuroscience, 14(7), 1088-1098.
Clark, V.P., Parasuranam, R., Keil, K. et al. (1997). Selective attention to face identity
and color studied with fMRI. Human Brain Mapping, 5, 293-297.
Coltheart, M., Rastle, K., Perry, C. et al. (2001). DRC: A Dual Route Cascaded model of
visual recognition and reading aloud. Psychological Review, 108(1), 204-256
Corbetta, M., Akbudak, E., Conturo, T. et al. (1998). A common network of functional
areas for attention and eye movements. Neuron, 21,761-773.
Corbetta, M., Miezin, F.M., Shulman, G.L. and Petersen S.E. (1993). A PET study of
visuospatial attention. Journal of Neuroscience, 13(3), 1202-1226.
Cornelissen, P., Tarkiainen, A., Helenius, P. and Salmelin, R. (2003). Cortical effects of
shifting letter position in letter strings of varying length. Journal of Cognitive Neuroscience,
15(5), 731-746.
Dehaene, S., Le Clec'H, G., Poline, J.-B. et al.(2002). The visual word form area: A
prelexical representation of visual words in the fusiform gyrus. NeuroReport, 13(3), 321-325.
Démonet, J.-F., Chollet, F., Ramsay, S. et al. (1992).The anatomy of phonological and
semantic processing in normal subjects. Brain, 115(6), 1753 1768.
Devlin, J. T., Matthews, P. M. and Rushworth, M. F. S. (2003). Semantic processing in
the left inferior prefrontal cortex: A combined functional magnetic resonance imaging and
transcranial magnetic stimulation study. Journal of Cognitive Neuroscience, 15(1), 71-84.
FMRI Evidence for ACV8 Connectionist Model for Reading … 17
Fiebach, C. J., Friederici, A. D., Müller, K. and von Cramon, D. Y. (2002). fMRI
evidence for dual routes to the mental lexicon in visual word recognition. Journal of Cognitive
Neuroscience, 14(1), 11-23.
Fiez, J. A. (1996). Cerebellar contributions to cognition. Neuron, 16, 13-15.
Fiez, J. A. (1997). Phonology, semantics, and the role of the left inferior prefrontal cortex.
Human Brain Mapping, 5, 79-83.
Fiez, J. A., Balota, D. A., Raichle, M. E. and Petersen, S. E. (1999). Effects of lexicality,
frequency, and spelling-to-sound consistency on the functional anatomy of reading.
Neuron,24, 205-218.
Fiez, J. A. and Petersen, S. E. (1998). Neuroimaging studies of word reading. Proceeding
National Academic Science, 95, 914-921.
Fiez, J. A., Raichle, M. E., Balota, D. A. et al. (1996). PET activation of posterior
temporal regions during auditory word presentation and verb generation. Cerebral Cortex, 6,
1-10.
Friston, K. J., Holmes, A. P., Worsley, K. J. et al. (1995). Statistical parametric maps in
functional imaging: A general linear approach. Human Brain Mapping, 3, 189-210.
Fulbright, R. K., Jenner, A. R., Mencl, W. E. et al. (1999). The cerebellum's role in
reading: A functional MR imaging study. American Journal of Neuroradiology, 20, 1925-
1930.
Gold, B. T. and Buckner, R. L. (2002). Common prefrontal regions coactivate with
dissociable posterior regions during controlled semantic and phonological tasks. Neuron, 35,
803-812.
Hagoort, P., Indefrey, P., Brown, C. et al. (1999). The neural circuitry involved in the
reading of German words and pseudowords: A PET study. Journal of Cognitive Neuroscience,
11(4), 383-398.
Herbster, A. N., Mintun, M. A., Nebes, R. D. and Becker, J. T. (1997). Regional cerebral
blood flow during word and non-word reading. Human Brain Mapping, 5, 84-92.
Indefrey, P., Kleinschmidt, A., Merboldt, K.-D. et al. (1997). Equivalent responses to
lexical and nonlexical visual stimuli in occipital cortex: A functional magnetic resonance
imaging study. NeuroImage, 5(1), 78-81.
Jobard, G., Crivello, F., and Tzourio-Mazoyer, N. (2003). Evaluation of the dual route
theory of reading: A metanalysis of 35 neuroimaging studies. NeuroImage, 20(2), 693-712.
McDermott, K. B., Petersen, S. E., Watson, J. M. and Ojemann, J. G. (2003). A procedure
for identifying regions preferentially activated by attention to semantic and phonological
relations using functional magnetic resonance imaging. Neuropsychologia, 41, 293-303.
Oldfield, R. C. (1971). The assessment and analysis of handness: The Edinburgh
inventory. Neuropsychologia, 9, 97-113.
Paulesu, E., Frith, C. D. and Frackowiak, R. S. J. (1993). The neural correlates of the
verbal component of working memory. Nature, 362, 342-345.
Monica Baciu 18
Petersen, S. E., Fox, P. T., Posner, M. I. et al. (1988). Positron emission tomograpihc
studies of the cortical anatomy of single-word processing. Nature, 331, 585-589.
Petersen, S. E., Fox, P. T., Snyder, A. Z. and Raichle, M. E. (1990). Activation of
extrastriate and frontal cortical areas by visual words and word-like stimuli. Science, 24, 1041-
1043.
Plaut, D. C. (1997). Structure and function in the lexical system: Insights from distributed
models of word reading and lexical decision. Language and Cognitive Processes, 12(5/6),
765-805.
Plaut, D. C., McClelland, J. L., Seidenberg, M. S. and Patterson, K. (1996).
Understanding normal and impaired word reading: Computational principles in quasi-regular
domains. Psychological Review, 103, 56-115.
Posner, M. I. and Petersen, S. E. (1990). The attention system of the human brain. Annual
Reviews of Neuroscience, 13, 25-42.
Price, C. J., Wise, R., Watson, J. et al. (1994). Brain activity during reading: The effect of
task and exposure duration. Brain, 117, 1255-1269.
Ravizza, S.M. and R.B. Ivry. (2001). Comparison of the Basal Ganglia and Cerebellum in
Shifting Attention. Journal of Cognitive Neuroscience, 13(3), 285-297.
Rumsey, J. M., Horwitz, B., Donohue, B. C. et al. (1997). Phonological and orthographic
components of word recognition. A PET-rCBF study. Brain, 120, 739-759.
Seidenberg, M. S. and McClelland, J. L. (1989). A distributed, developmental model of
word recognition and naming. Psychological Review, 96, 523-568.
Simon, O., Mangin, J. F., Cohen, L., Le Bihan, D. and Dehaene, S. (2002). Topographical
layout of hand, eye, calculation, and language-related areas in the human parietal lobe.
Neuron, 33(3), 475-487.
St George, M., Kutas, M., Martinez, A. and Sereno, M. I. (1999). Semantic integration in
reading: Engagement of the right hemisphere during discourse processing. Brain, 122(7),
1317-1325.
Tagamets, M.-A., Novick, J. M., Chalmers, M. L. and Friedman, R. B. (2000). A
parametric approach to orthographic processing in the brain: An fMRI study. Journal of
Cognitive Neuroscience, 12, 281-297.
Talairach, J. and Tournoux, P. (1988). Co-planar stereotaxic atlas of the human brain. 3-
Dimensional proportional system: An approach to cerebral imaging. New York: Thieme
Medical.
Tarkiainen, A., Cornelissen, P. L. and Salmelin, R. (2002). Dynamics of visual feature
analysis and object-level processing in face versus letter-string perception. Brain, 125, 1125-
1136.
Tarkiainen, A., Helenius, P., Hansen, P. C. et al. (1999). Dynamics of letter string
perception in the human occipitotemporal cortex. Brain, 122, 2119-2131.
FMRI Evidence for ACV8 Connectionist Model for Reading … 19
Thierry, G., Boulanouar, K., Kherif, F. et al. (1999). Temporal sorting of neural
components underlying phonological processing. NeuroReport, 10, 2599-2603.
Wise, R., Hadar, U., Howard, D. and Patterson, K. (1991). Language activation studies
with positron emission tomography. Ciba Foundation Symposia, 163:218-28; discussion 228-
34. Review.
Xu, B., Grafman, J., Gaillard, W. D. et al. (2001). Conjoint and extended neural networks
for the computation of speech codes: The neural basis of selective impairment in reading
words and pseudowords. Cerebral Cortex, 11(3), 267-277.
Xu, B., Grafman, J., Gaillard, W. D. et al. (2002). Neuroimaging reveals automatic speech
coding during perception of written word meaning.
NeuroImage, 17, 859-870.