Neuroscience Letters 290 (2000) 6165

www.elsevier.com/locate/neulet

Brain activation proles in dyslexic children during non-word

reading: a magnetic source imaging study
Panagiotis G. Simos a,*, Joshua I. Breier a, Jack M. Fletcher b, Barbara R. Foorman b,
Eldo Bergman c, Kristen Fishbeck a, Andrew C. Papanicolaou a
a

Department of Neurosurgery, University of Texas Health Science Center, Medical School,

6431 Fannin Street, Suite 7.149, Houston, TX 77030, USA
b
Department of Pediatrics, University of Texas Health Science Center, Medical School,
6431 Fannin Street, Suite 7.149, Houston, TX 77030, USA
c
Texas Reading Institute, Houston, Texas, USA

Received 26 May 1999; received in revised form 19 October 1999; accepted 21 October 1999

Abstract
The purpose of the study was to identify spatiotemporal brain activation proles associated with phonological decoding in dyslexic children using magnetic source imaging. For this purpose maps of regional cerebral activation were
obtained from eleven children diagnosed with dyslexia and ten children without reading problems during engagement
in a pseudoword rhyme-matching task. All dyslexic children showed aberrant activation maps consisting of reduced
activity in temporoparietal areas in the left hemisphere (including the posterior part of the superior temporal, angular
and supramarginal gyri) and increased activity in the right homotopic region. In contrast, the two groups of children did
not differ in the degree of activity in basal temporal areas that typically precedes temporoparietal activation. This is the
rst study to demonstrate the existence of distinct activation proles associated with phonological decoding in individual dyslexic children. q 2000 Published by Elsevier Science Ireland Ltd.
Keywords: Functional imaging; Magnetic source imaging; Magnetoencephalography; Dyslexia; Reading; Phonological decoding;
Temporal lobe

There are many proposals regarding the nature of decits

underlying developmental reading disorder (dyslexia)
[4,5,8,2325]. Regardless of the nature of this decit, it
remains an indisputable fact that phonological decoding
problems are perhaps the most ubiquitous signs of dyslexia
[6,23,25]. The preferred approach for investigating directly
phonological decoding skills is to examine reading performance of pronounceable non-words (pseudowords).
Further, investigating brain activation proles associated
with reading and phonological decoding, in particular, is
expected to help clarify the neurophysiological substrate
of dyslexia. Such proles can be obtained indirectly in the
form of regionally elevated levels of blood ow or metabolism using positron emission tomography or functional
magnetic resonance imaging [4,1214,18], or more directly
in the form of regionally elevated levels of intracellular
* Corresponding author. Tel.: 11-713-500-6128; fax: 11-713500-7787.
E-mail address: asimos@heart.med.uth.tme.edu
(P.G. Simos).

electrical currents in large neuronal aggregates using

magnetic source imaging (MSI, also known as magnetoencephalography [12]). Brain areas that have been shown to
display reduced engagement in a variety of reading tasks in
dyslexics are those located in the vicinity of the temporoparietal junction and typically include the posterior part of
the superior and middle temporal gyri, the angular, and
supramarginal gyri predominantly in the left hemisphere
[14,15,18]. Unfortunately, functional brain imaging methods that rely on measures of blood ow/metabolism are
largely capable of providing proles of relative activation
that are meaningful primarily on a group basis. On the other
hand, MSI can provide detailed information regarding the
temporal course as well as the spatial extent of task-related
regional activation. Moreover, the sensitivity of this method
for identifying areas involved in language-specic
processes has been established in a series of studies
[11,19,22]. Unlike other functional imaging techniques,
the validity of MSI-derived maps has been corroborated
by the results of standard invasive mapping procedures

0304-3940/00/$ - see front matter q 2000 Published by Elsevier Science Ireland Ltd.
PII: S03 04 - 394 0( 0 0) 01 32 2- 7

62

P.G. Simos et al. / Neuroscience Letters 290 (2000) 6165

such as the Wada procedure [2,11]. In addition, temporal

lobe areas shown by MSI to be active during word recognition tasks were found to be associated with transient receptive language decits when stimulated electrically in
patients undergoing awake craniotomies for resection of
tumors or epileptogenic tissue [11,20,22].
It should also be noted that most existing functional
imaging studies on dyslexia have examined adults with
developmental reading disability. However, there is
evidence that dyslexics may develop alternative reading
strategies in order to compensate for their poor phonological
decoding skill [10]. Little is known at present of how this
compensation may affect brain activation proles of adult
dyslexics, creating the need for studies of younger individuals with reading problems.
Recently, MSI was used to obtain direct measures of
regional cerebral engagement during reading from children
diagnosed with dyslexia [21]. In that study ten dyslexic and
eight non-dyslexic children were tested on a visual and an
auditory version of a word recognition task. The most notable nding was that, in every case, temporoparietal areas of
the left hemisphere (including the posterior third of the
superior and middle temporal gyri, the angular and supramarginal gyri) failed to become engaged during reading in
children with dyslexia, whereas homologous areas in the
right hemisphere showed strong activation. Non-impaired
readers showed the opposite pattern of strong left hemisphere activation coupled with signicantly weaker right
hemisphere activation. Although a distinct activation prole
characterized all the dyslexic children in that study, it was
unclear whether the prole was specic to phonological
decoding or reected a variety of other linguistic operations,
such as orthographic and semantic analysis of printed words
and memory that are also involved in word recognition. The
present study was designed to examine spatiotemporal activation proles in individual dyslexic children and agematched non-impaired readers associated specically with
phonological decoding in the context of a pseudoword
rhyme-matching task [3].
Eleven children (seven males, age range 1017 years)
who presented with mild to severe reading difculties and
ten non-impaired readers (seven males, age range 816
years) were examined. All dyslexic children showed severe
impairment in phonological decoding skills (scoring under
the 20th percentile on the Word Attack subtest of the WJ-R
Battery [26] as compared to scores .80th percentile on this
test for each of the non-dyslexic children. All subjects had
IQ scores (WISC III-R composite scores) of 85 or above,
were right-handed with English as their primary language,
and had no history of hearing decit, neurological injury or
disease, or visual impairment.
Each child was tested on a pseudoword rhyme-matching
tasks during which they were presented with two pseudowords (sequentially) and asked to decide whether they
rhymed or not. Stimuli were 45 letters long and were
always orthographically dissimilar (e.g., kume- noom) to

discourage performing comparisons on the basis of orthographic information. Event-related magnetic elds (ERFs)
were recorded to the rst stimulus of each pair, to ensure
that the brain activity recorded corresponded to phonological decoding operations and did not reect the additional
cognitive operations that matching of the stimuli entails.
The responding hand was counterbalanced across children.
MSI data collection and analysis methods developed and
used in our laboratory have been described in detail elsewhere [11]. Briey, recordings were made in a magnetically
shielded room with a whole-head neuromagnetometer
(Biomagnetic Technologies). The intracranial generators
of the observed ERFs were modeled as single equivalent
current dipoles (ECDs) that were tted at successive 4 ms
intervals using the non-linear LevenbergMarquardt algorithm [16]. Although a variety of source modeling
approaches have been proposed by different investigators
[9,17] we decided to use a single-ECD source model that
is part of the BTi software. Alternative algorithms hold
many promises as tools for magnetic source localization.
However, they have not yet been validated against invasive
localization procedures. In contrast, there is currently a
wealth of data testifying to the validity of the single-ECD
model for reliably localizing and lateralizing neurophysiological activity associated with language functions
[2,11,20,22]. On the basis of this evidence, the singe-ECD
source model is part of the standard analysis protocol in
essentially all clinical applications of magnetoencephalography. For a given point in time, the ECD dipole tting
algorithm was applied to the magnetic ux measurements
obtained from a group of 3438 magnetometers, always
including both magnetic ux extrema. ECD computation
was restricted to those latency periods during which a single
pair of magnetic ux extrema dominated the left and/or the
right half of the head surface. ECD solutions were considered as satisfactory upon meeting two criteria: (1) a correlation coefcient of at least 0.90 between the observed and the
`best' predicted magnetic eld distribution, and (2) a 95%
condence volume of 3 cm 3 or smaller. Only ECDs
computed during the `late' portion of ERFs to the printed
pseudowords were considered in the analyses. Previous
studies have shown that these activity sources originate in
association cortices specialized for the analysis of linguistic
stimuli [1,11,22]. Early ERF components (i.e. those
recorded up to approximately 150 ms after stimulus onset)
originate in modality-specic (in this case visual) cortex and
probably reect early sensory analysis of the stimuli. The
procedure for precise coregistration of ECD coordinates on
structural, T1-weighted, MRI scans (TR 13.6 ms,
TE 4.8 ms, recording matrix 256 256 pixels, 1 excitation, 240 mm eld of view and 1.4 mm slice thickness) has
been described in detail elsewhere [19,27].
As expected, performance on the task was signicantly
higher for the non-dyslexic group (79.2 vs. 67.4% correct,
P , 0:05). The number of reliably localized activity
sources in temporoparietal (TMP) and basal temporal

P.G. Simos et al. / Neuroscience Letters 290 (2000) 6165

cortices (BTC) in each hemisphere served as the dependent measure in the statistical analyses. These regions
were selected because they consistently displayed activity
sources in every child, in at least one hemisphere. TMP
areas included the posterior third of the superior (STGp)
and middle temporal gyrus (MTGp), the angular (ANG)
and supramarginal gyri (SMG). Selection of areas was
based on the results of our previous study [27], which
did not reveal any interactions involving these four
temporal and inferior parietal areas. Basal temporal
cortices included the fusiform and lingual gyri. As
previously mentioned, sources were computed during
the later portion of the evoked magnetic response thereby
excluding those which typically originate within modality-specic cortices. This measure has been found to be
the most reliable and valid index of the degree of regional, language-specic cerebral activation in several
studies involving neurologically intact volunteers and
patients [2,11,17,22,27].
Visual inspection of individual brain activation proles
indicated that all dyslexic children showed strong activation
in TMP regions in the right hemisphere with visibly weaker
activity in homotopic areas in the left hemisphere. The
opposite pattern was apparent for all non-dyslexic children
(with the exception of S20 who displayed slightly stronger
right than left temporoparietal activation). This highly

Fig. 1. The distribution of individual hemispheric asymmetry

indices ([Left 2 Right]/[Left 1 Right]) for the number of activity
sources located in temporoparietal (TMP) areas for dyslexic
(lled diamonds) and non-dyslexic children (open circles). All
dyslexic children had negative scores indicating greater right
than left TMP activation, while the opposite was true for 9/10
non-dyslexic children. Asymmetry indices are plotted against
behavioral performance (percent correct detection of rhyming
pseudoword pairs) during the MSI session to demonstrate a
modest linear association between the two variables.

63

Fig. 2. 3-D renderings of MRI scans from a representative

dyslexic (lower set of images) and a non-dyslexic child (top set
of images). Clusters of activity sources computed at 4 ms intervals after the presentation of the pseudoword stimuli were
projected on the brain surface for easier visualization. Sources
occurring between 100 and 300 ms after stimulus onset (shown
in yellow) were typically localized in basal temporal cortices.
Temporoparietal (TMP) sources (shown in orange) usually
became active later between 300 and 1200 ms after stimulus
onset. Note the clear preponderance of activity sources in left
TMP cortices in the non-dyslexic subject and in homotopic right
hemisphere areas in the case of the dyslexic child.

consistent pattern across children can be seen in Fig. 1,

which displays the distribution of individual hemispheric
asymmetry indices ([Left 2 Right]/[Left 1 Right]) for
activity sources located in TMP regions. The precise anatomical distribution of activity sources associated with the
analysis of the pseudowords in two representative cases
(one dyslexic and one non-dyslexic) can be seen in Fig. 2.
The data were further analyzed using a multivariate
approach to ANOVA with two within- (Area: TMP, BTC,
and Hemisphere: Left, Right), and one between-subjects
variable (Group: Dyslexic, Non-dyslexic). All subsequent
pairwise comparisons were evaluated using the Bonferroni
method. This analysis revealed a signicant Group by Task
by Hemisphere interaction, F1; 19 41:04, P , 0:0001
(Fig. 3). Further tests (independent sample t-tests for
samples of unequal size) conrmed the initial observation
that dyslexic children showed signicantly less activation in
the left TMP region compared to non-dyslexic children,
t15:15 4:18, P 0:001, while the opposite was found
for homotopic regions in the right hemisphere,
t18:47 3:58, P 0:002. In addition, hemispheric asymmetries in the number of activity sources located in TMP
regions were of opposite direction in the two groups of
children. Non-dyslexic children showed signicantly
greater activation in the left versus the right TMP regions,

64

P.G. Simos et al. / Neuroscience Letters 290 (2000) 6165

Fig. 3. Mean number of activity sources in temporoparietal

(TMP) and basal temporal cortices in each hemisphere in the
two groups of children (Standard error values in bars).

t9 4:36, P 0:002, while dyslexic children displayed

greater activation in right compared to left TMP regions,
t10 6:52, P , 0:00001. Further, the degree of activity in
TMP regions was a good predictor of behavioral performance on the rhyme-matching task (see Fig. 1). Signicant
positive correlations were found between the percent correct
detection of rhyming pseudowords and both left TMP activation (r 0:75, P , 0:0001) and hemispheric asymmetry
indices for TMP activity sources (r 0:74, P , 0:0001). In
addition, a signicant negative correlation was found
between accuracy scores and right TMP activation
(r 0:49, P , 0:025). In contrast negligible association
was found between performance and degree of activity in
BTC (r , 0:08).
In addition to information regarding the degree of regional activation, MSI is uniquely suited to provide data
regarding the temporal course of activation associated
with the presentation of an external event in the context of
an experimental task. This information may then be used to
indicate potential regional cortico-cortical interactions that
take place during engagement in the task in question (in this
case pseudoword reading). Thus, as in our previous study
using real words as stimuli, there was, in both groups, a
Table 1
Mean onset latency of activation in temporoparietal and basal
temporal areas (in ms after stimulus onset) in the two groups of
children a
TMP

Dyslexics
Non-Dyslexics
a

BTC

Left

Right

Left

Right

474 (51)
297 (39)

351 (75)
415 (66)

268 (25)
260 (11)

436 (34)
248 (44)

Standard error values in parentheses. Abbreviations: TMP,

temporoparietal regions; BTC, basal temporal cortices.

regular progression of activation from occipital, to basal

temporal regions (posterior part of the fusiform and lingual
gyri), and nally to temporoparietal areas (Table 1). The
only signicant group differences with respect to the mean
onset latencies were for clusters of activity sources in the
left TMP, t2:75 6:52, P 0:013, and the right BTC
regions, t17:88 4:45, P , 0:0001, with dyslexics
displaying signicantly longer onset latencies in both
regions (474 and 436 ms, respectively) than non-dyslexics
(300 and 248 ms, respectively). In addition, a signicant
correlation was apparent between the onset of left TMP
activity and performance on the rhyme matching task
(r 0:52, P , 0:023), indicating that earlier onset of activity in this area was associated with better performance.
The results reported here corroborate earlier ndings
from our laboratory [21] indicating that the pronounced
differences between dyslexic and non-dyslexic children in
the degree of activation in temporoparietal regions are
indeed attributable to phonological decoding demands
imposed by reading both words and pronounceable pseudowords. In addition to reduced activation in left TMP areas,
dyslexic children showed a marked increase in activation in
homotopic areas of the right hemisphere. Although both
ndings have been suggested by previous research using
other imaging techniques [13,14,18], the consistency with
which this phenomenon was obtained in every dyslexic individual in both studies (in a total of 21 consecutive cases) is
striking. Such results were not, however, surprising given
that compelling evidence to the validity of individual MSIderived maps of regional activation has been obtained by
comparison with standard invasive techniques in large
series of consecutive neurosurgical patients [2,11,20,22].
The nding that the amount of activity in the left temporoparietal regions predicts performance on the experimental
task from which the activation proles were obtained further
supports the notion that activity in these regions detected by
MSI reects the engagement of neurophysiological
processes that are critically involved in reading.
It is interesting that the degree of activation in basal
temporal regions did not differ systematically between the
two groups of children. In contrast, it was activation (both
in terms of degree and timing) in areas known to be
involved in word recognition (such as the angular gyrus
[7,18]) and in phonological analysis of both auditory and
printed language stimuli [3,18] that clearly differed across
groups. Moreover, analyses on the timing of activity
argue against the hypothesis that dyslexic's activation
proles reect delayed, yet otherwise normal, decoding
of letter stings. Specically, the right TMP regions in
dyslexics rst became active at about the same time (on
average) as left TMP areas in the non-impaired readers.
Similarly, there were no signicant differences in the
onset of activity in left basal temporal regions across
groups.
These results suggest the existence of a functional decit
involving neurophysiological operations normally supported

P.G. Simos et al. / Neuroscience Letters 290 (2000) 6165

by left TMP cortices in children with dyslexia. These operations may either be directly involved in the conversion of
print to sound (phonological decoding), specialized for
phonological analysis of speech sounds, or be common to
both processes [3,6,8,2325]. We are planning studies to
examine the developmental course that leads to the establishment of the activation prole seen in non-impaired readers.
These studies will address directly the proposition that the
distinct activation prole seen in older children with dyslexia
reects a functional decit involving left temporoparietal
areas or, alternatively, represents a more immature brain
mechanism supporting reading (we thank one of the
reviewers of an earlier version of this paper for suggesting
this possibility).
Regardless of the precise nature of this decit, the interindividual consistency of this phenomenon as revealed by
MSI underlines a potential use of this technique as a tool for
evaluating the effectiveness of educational intervention strategies that specically target the `phonological core' of
developmental reading disability.
[1] Breier, J.I., Simos, P.G., Zouridakis, G. and Papanicolaou,
A.C., Relative timing of cortical activation during a word
recognition task, J. Clin. Exper. Neuropsychol., 20 (1998)
782790.
[2] Breier, J.I., Simos, P.G., Papanicolaou, A.C., Zouridakis, G.,
Wilmore, L.J., Wheless, J.W., Constantinou, J.C., Maggio,
W.W., Cerebral laterality assessment using magnetoencephalography: a comparison with the Intracarotid Amobarbital Procedure, Neurology, 22 (1999) 938945.
[3] Breier, J.I., Simos, P.G., Zouridakis, G. and Papanicolaou,
A.C., Temporal course of regional activation associated
with phonological decoding: a MEG study, J. Clin. Exp.
Neuropsychol., 21 (1999) 465476.
[4] Eden, G.F. and Zero, T.A., Neural systems affected in
developmental dyslexia revealed by functional neuroimaging, Neuron, 21 (1997) 279282.
[5] Fletcher, J.M., Shaywitz, S.E., Shankweiler, D.P., Katz, L.,
Liberman, I.Y., Stuebing, C.K., Francis, D.J., Fowler, A.,
Shaywitz, B.A., Cognitive proles of reading disability:
comparisons of discrepancy and low achievement denitions, J. Edu. Psychol., 86 (1994) 623.
[6] Foorman, B.R., Francis, D.J., Fletcher, J.M. and Lynn, A.,
Relation of phonological and orthographic processing to
early reading: comparing two approaches to regressionbased, reading-level matched designs, J. Edu. Psychol., 4
(1996) 639652.
[7] Greenblatt, S.H., Subangular alexia without agraphia or
hemianopsia, Brain Lang., 3 (1976) 229245.
[8] Lyon, G.R., Toward a denition of dyslexia, Ann. Dysl., 45
(1995) 327.
[9] Mosher, J.C., Lewis, P.S. and Leahy, R., Multiple dipole
modeling and localization from spatio-temporal MEG
data, IEEE Trans. Biomed. Eng., 39 (1992) 541557.
[10] Olson, R.K., Kliegl, R., Davidson, B.J. and Foltz, G., Individual and developmental differences in reading disability, In
G.E. MacKinnon and T.G. Waller (Eds.), Reading Research:
Advances in Theory and Practice, Vol. 4, Academic Press,
Orlando, 1985, pp. 164.
[11] Papanicolaou, A.C., Simos, P.G., Breier, J.I., Zouridakis, G.,

[12]
[13]

[14]

[15]
[16]
[17]
[18]

[19]

[20]

[21]

[22]

[23]

[24]
[25]
[26]
[27]

65

Wilmore, L.J., Wheless, J.W., Constantinou, J.C., Gormley,

W., Maggio, W.W., Magnetoencephalographic mapping of
the language-specic cortex, J. Neurosurg., 90 (1998) 8593.
Papanicolaou, A.C., Fundamentals of Functional Brain
Imaging, Swets & Zeitlinger, Lisse, 1998.
Richards, T.L., Dager, S.R., Corina, D., Serani, S., Heide,
A.C., Steury, K., Strauss, W., Hayes, C.E., Abbott, R.D., Craft,
S, Shaw, D., Posse, S., Berninger, V.W., Dyslexic children
have abnormal brain lactate on a reading-related language
task, Am. J. Neuroradiol., 20 (1999) 13931398.
Rumsey, J.M., Andreason, P., Zametkin, A.J., Aquino, T.,
King, A.C., Hamburger, S.D., Pikus, A., Rapoport, J.L.,
Cohen, R.M., Failure to activate the left temporoparietal
cortex in dyslexia: an oxygen 15 positron emission tomographic study, Arch. Neurol., 49 (1992) 527534.
Salmelin, R., Service, E., Kiesila, P. and Salonen, O., Impaired
visual word processing in dyslexia revealed with magnetoencephalography, Ann. Neurol., 40 (1996) 157162.
Sarvas, J., Basic mathematical and electromagnetic
concepts of the biomagnetic problem, Phys. Med. Biol.,
32 (1987) 1122.
Scherg, M. and Buchner, H., Somatosensory evoked potentials and magnetic elds: separation of multiple source
activities, Physiol. Meas., 14 (1993) A35A39.
Shaywitz, S.E., Shaywitz, B.A., Pugh, K.R., Fulbright, R.K.,
Constable, R.T., Menci, W.E., Shankweiler, D.P., Liberman,
A.M., Skudlarski, P., Fletcher, J.M., Katz, L., Marchione,
K.E., Lacadie, C., Gatenby, C., Gore, J.C., Functional disruption in the organization of the brain for reading in dyslexia,
Proc. Natl. Acad. Sci. USA, 95 (1998) 26362641.
Simos, P.G., Breier, J.I., Zouridakis, G. and Papanicolaou,
A.C., Identication of language-related brain activity using
magnetoencephalography, J. Clin. Exp. Neuropsychol., 20
(1998) 705722.
Simos, P.G., Breier, J.I., Maggio, W.W., Gormley, W., Zouridakis, G., Wilmore, L.J., Wheless, J.W., Papanicolaou, A.C.,
Atypical temporal lobe language representation revealed
by MEG and intraoperative stimulation mapping, NeuroReport, 10 (1999) 139142.
Simos, P.G., Breier, J.I., Fletcher, J.M., Bergman, E., Papanicolaou, A.C., Cerebral mechanisms involved in word
reading in dyslexic children: A Magnetic source Imaging
approach, Cereb Cortex, in press.
Simos, P.G., Papanicolaou, A.C., Breier, J.I., Wilmore, L.J.,
Wheless, J.W., Constantinou, J.C., Gormley, W., Maggio,
W.W., Localization of language-specic cortex using MEG
and intraoperative stimulation mapping, J. Neurosurg., 91
(1999) 787796.
Stanovich, K.E. and Siegel, L.S., Phenotypic performance
prole of children with reading disabilities: a regressionbased test of the phonological-core variable-difference
model, J. Edu. Psychol., 86 (1994) 2453.
Tallal, P., Miller, S. and Fitch, R.H., Neurobiological basis of
speech: a case for the preeminence of temporal processing,
Ann. N.Y. Acad. Sci., 682 (1993) 2747.
Wagner, R.K. and Torgesen, J.K., The nature of phonological processing and the acquisition of reading skills,
Psychol. Bull., 101 (1987) 192212.
Woodcock, R.W. and Johnson, M.B.,Woodcock-Johnson
Psycho-Educational Battery - Revised, DLM Teaching
Resources, Texas, 1989-1990
Zouridakis, G., Simos, P.G., Breier, J.I. and Papanicolaou,
A.C., Cerebral dominance assessment using magnetoencephalography, Brain Topogr., 11 (1998) 5765.