A

aiofsaB7oti©

K

1

sfdpssiFc

0d

Neuropsychologia 46 (2008) 640–648

Neural correlates of working memory performance inadolescents and young adults with dyslexia

Nenad Vasic a,∗,1, Christina Lohr b,1, Claudia Steinbrink b,Claudia Martin c, Robert Christian Wolf a

a Department of Psychiatry, University of Ulm, Leimgrubenweg 12-14, 89075 Ulm, Germanyb Transfer Center for Neuroscience and Learning, University of Ulm, Germany

c Department of Psychology, University of Wuerzburg, Germany

Received 28 February 2007; received in revised form 4 September 2007; accepted 6 September 2007Available online 12 September 2007

bstract

Behavioral studies indicate deficits in phonological working memory (WM) and executive functioning in dyslexics. However, little is knownbout the underlying functional neuroanatomy. In the present study, neural correlates of WM in adolescents and young adults with dyslexia werenvestigated using event-related functional magnetic resonance imaging (fMRI) and a parametric verbal WM task which required the manipulationf verbal material. Dyslexics were not significantly slower than controls; however, they were less accurate with the highest WM demand. Theunctional analysis excluded incorrectly performed and omitted trials, thus controlling for potential activation confounds. Compared with controlubjects, both increased and decreased activation of the prefrontal cortex were found in the dyslexic group. Dyslexics showed significantly morectivation than controls with increasing WM demand in the left superior frontal gyrus (BA 8), as well as in the inferior frontal gyrus includingroca’s area (BA 44) and its right homologue. Less activation was found in the middle frontal gyrus (BA 6) and in the superior parietal cortex (BA

). A positive correlation between activation of prefrontal regions and verbal WM performance (as measured by digit span backwards) was foundnly in the dyslexic group. Accuracy deficits at the highest cognitive demand during the verbal WM task and the digit span backwards suggesthat manipulation rather than maintenance is selectively impaired in dyslexics. The fMRI data provide further evidence for functional differencesn cortical regions associated with language processing and executive function in subjects with dyslexia.

2007 Elsevier Ltd. All rights reserved.

x; Br

aJ2Laet

d

eywords: Reading impairment; Executive dysfunction; fMRI; Prefrontal corte

. Introduction

Working memory (WM) refers to the ability to tran-iently store and manipulate information held ‘on-line’ forurther behavioral guidance (Baddeley, 1996, 2003). In his firsteveloped model of WM, Baddeley (1996) separated three com-onents of WM: a control system of limited attentional capacityubserving cognitive control (central executive) and two sub-idiary storage systems known as the phonological loop, which

s based on sound and language, and the visuospatial sketchpad.unctional neuroimaging studies have shown that the prefrontalortex is critical for several component processes of WM, such

∗ Corresponding author. Tel.: +49 731 500 61568; fax: +49 731 500 61412.E-mail address: nenad.vasic@uni-ulm.de (N. Vasic).

1 These authors contributed equally to this study.

2lieadit

028-3932/$ – see front matter © 2007 Elsevier Ltd. All rights reserved.oi:10.1016/j.neuropsychologia.2007.09.002

oca’s area

s executive control and active maintenance (Bunge, Klingberg,acobsen, & Gabrieli, 2000; D’Esposito, Postle, & Rypma,000; Kane & Engle, 2002; Smith & Jonides, 1999; Zhang,eung, & Johnson, 2003). Other brain regions, including lateralnd medial premotor areas (BA 6 und 8) and the posterior pari-tal cortex (BA 40 and 7), are also consistently activated duringasks that require WM (see Wager & Smith, 2003 for a review).

WM deficits have been associated with language-relatedevelopmental disorders like dyslexia (e.g. Brambati et al.,006). Developmental dyslexia is a specific impairment inearning to read which does not result from general cognitivempairment, sensory deficits or inadequate schooling (Shaywitzt al., 1998). The disorder is characterized by difficulties with

ccurate and fluent word recognition and by poor spelling andecoding abilities. Longitudinal studies indicate that dyslexias a persistent and chronic condition and does not represent aransient “developmental lag” (Shaywitz & Shaywitz, 2005).

cholo

AfAtiiiwaP2la2cii

riiladidwta

Pima

afwVusccv2eol

2

2

1

TS

A

S

R

P

A

N

N. Vasic et al. / Neuropsy

number of behavioral studies report deficits in executiveunctioning in dyslexia (e.g. Brosnan et al., 2002; Helland &sbjornsen, 2000; Reiter, Tucha, & Lange, 2005), showing

hat developmental dyslexia is associated with deficits regard-ng executive control of WM. Deficits in executive functioningn individuals with dyslexia reported by a number of behav-oral studies suggest that developmental dyslexia is associatedith deficits in executive control of WM (Swanson, 1993). Onlyfew studies report deficits in visual WM in dyslexics (e.g.

oblano, Valadez-Tepec, de Lourdes Arias, & Garcia-Pedroza,000; Reiter et al., 2005), whereas many show deficits in phono-ogical WM functions (Baddeley & Wilson, 1993; Brosnan etl., 2002; Chiappe, Hasher, & Siegel, 2000; Jeffries & Everatt,004; Nelson & Warrington, 1980; Poblano et al., 2000). Thus,urrently available behavioral evidence suggests a selectivempairment of storage and manipulation of verbal informationn dyslexia.

However, little is known about the neural correlates of WM-elated processes in dyslexia. At present, only one study hasnvestigated verbal WM functions in dyslexics using functionalmaging techniques. Paulesu et al. (1996) have used a phono-ogical WM task (recognition of consonants from item-lists)nd positron emission tomography (PET) to study adults withyslexia and control subjects. Dyslexics showed less activationn several cortical regions that were activated by normal controls

espite similar cognitive performance. The largest differencesere found bihemispherically in Broca’s area (BA 44) and in

he insula. The authors interpreted their results as evidence fordisconnection between brain areas in dyslexics. Furthermore,

wtTif

able 1ubject characteristics

Dyslexics (n = 12)

Mean S.D.

ge (years) 18.3 1.7Laterality scorea 90.4 7.0Intelligence quotientb 117.4 13.8

pelling testc

Percentage rank 6.7 5.5Errors (max. 60) 33.3 8.4

eal-word readingd

Time (s) 58.8 14.2Errors (max. 48) 5.1 3.1

seudo-word readingd

Time (s) 118.0 36.7Errors (max. 48) 15.2 8.2

lphabet retrievalTime (s) 4.4 1.1Errors 0.4 0.7Digit span backwarde—score (max. 12) 5.5 1.5Digit span forwarde—score (max. 12) 8.0 2.4

ote: ns = not significant.a Oldfield (1971).b Weiß (1997).c Kersting and Althoff (2004).d Schulte-Korne (2001).e Tewes (1994).

gia 46 (2008) 640–648 641

aulesu et al.’s results suggest that different brain activationsn dyslexics might not solely reflect limitations in WM perfor-

ance, since they did not find any significant differences in taskccuracy in the dyslexic group.

To further clarify WM function in adolescents and youngdults with dyslexia, we used the advantages of event-relatedMRI to control for the confound of impaired task accuracyithin a previously validated parametric verbal WM task (Wolf,asic, & Walter, 2006; Wolf & Walter, 2005). We were partic-larly interested in the circuitry of the lateral prefrontal cortex,ince prefrontal circuitry has been recognized to play a cru-ial role in WM, especially during processes requiring executiveontrol. Given the non-linear relationship between cerebral acti-ation and increasing WM load (Callicott et al., 2003; Manoach,003), we hypothesized that behavioral and activation differ-nces in prefrontal regions during the manipulation/maintenancef memoranda would preferentially occur at higher WM loadevels.

. Materials and methods

.1. Subjects

We studied young monolingual German adolescents and young adults aged6–21 years (for details on subject characteristics, see Table 1). All subjects

ere strongly right-handed, as assessed by the Edinburgh Handedness Inven-

ory (Oldfield, 1971) and had no history of neurological or psychiatric disorders.he IQ of all subjects was average or above average according to a nonverbal

ntelligence test (Weiß, 1997). The dyslexic group consisted of 12 subjects (3emales), all of which received a diagnosis of developmental dyslexia in child-

Controls (n = 13) Group differences

Mean S.D.

18.3 1.7 ns91.1 6.9 ns

122.7 15.5 ns

50.6 19.3 p < 0.0110.6 5.1 p < 0.01

33.7 7.7 p < 0.010.8 1.0 p < 0.01

61.2 13.8 p < 0.014.3 2.5 p < 0.01

5.5 2.3 ns0.1 0.3 ns7.6 2.6 p < 0.058.9 1.5 ns

6 ycholo

ha

TcmDaal(aurlw

owSb

co

2

(gOpdasaLtswlafpt

Fpmtf

ap

2

Gp6srp1wqdt&tie

2

2

rtvaT

2

ifv3ww

42 N. Vasic et al. / Neurops

ood and reported continuing problems with reading and spelling in adolescencend adulthood. The control group was matched for age, gender and IQ.

All subjects completed a reading test for adults (Schulte-Korne, 2001).he dyslexic group read real words and pseudo-words significantly slower andommitted more reading errors than the control group. Spelling abilities wereeasured with a standardized spelling test for adults (Kersting & Althoff, 2004).yslexics achieved a significantly higher error score than controls. In both tests,

ll subjects in the dyslexic group met criteria for substandard reading and spellingbilities. Since the activation paradigm used in this study tested verbal WM usingetters, an alphabet retrieval task and the digit span forward and backward testTewes, 1994) were additionally administered. Both groups showed equal speednd error scores in the alphabet retrieval task, indicating comparable prereq-isites for accomplishing the fMRI task. Controls achieved significantly betteresults in the digit span backward task. In the dyslexic group, a significant corre-ation between digit span backwards performance and reading or spelling scoresas not found (p < 0.05).

After the scanning procedure all subjects completed a brief questionnairen the individual encoding and maintenance strategies used during fMRI, asell as on difficulties in attention and concentration while performing the task.emi-qualitative analyses of this questionnaire did not reveal any differencesetween dyslexics and controls subjects.

The study was approved by the local Institutional Review Board. After aomplete description of the study to the subjects, written informed consent wasbtained from all participants.

.2. Cognitive activation paradigm

The cognitive activation task has been described elsewhere in full detailWolf et al., 2006; Wolf & Walter, 2005; see Fig. 1). In brief, three capitalrey letters appeared on a black screen during the stimulus period (1500 ms).ne, two or three of these letters were highlighted at the end of the stimuluseriod for 500 ms. Subjects were instructed that during the subsequent 6000 mselay period they were to focus only on those letters which were highlightednd to memorize the letters which followed them in the alphabet (manipulatedet). By emphasizing the shifting of memoranda towards other letters of thelphabet, we introduced a manipulation component during the delay period.ow manipulation demand required that only one letter had to be identified as

he one which followed next in the alphabet and had to be maintained for ahort time period (load level 1). Intermediate and high manipulation demandsere characterized by manipulation of two and three letters, respectively (load

evels 2 and 3). Stimulus and target presentation were held almost identicalcross all conditions, thus minimizing potential confounding effects originatingrom these periods. In the probe period of 2000 ms, a lower case letter wasresented, and subjects had to indicate whether this letter was or was not part ofhe manipulated set. The control condition displayed three grey X’s and required

ig. 1. Activation paradigm, shown for a trial of load level 2 (L2). In this exam-le, the letters S and G were highlighted, and subjects had to subsequentlyemorize the letters T and H (manipulated set). The probe-letter t is a part of

he previously manipulated set, i.e. a positive probe (see also Section 2.2 forurther details). Red line: period-of-interest for the fMRI analysis.

&as(daawlmoat1m

dtM(gf(taco

gia 46 (2008) 640–648

stereotype button press in response to the presentation of a small x during therobe period, thus forming a motor task without mnemonic requirements.

.3. Data acquisition

Data were acquired using a 3.0 T Magnetom ALLEGRA (Siemens, Erlangen,ermany) head MRI system. T2*-weighted images were obtained using echo-lanar imaging in an axial orientation (TR = 2400 ms, TE = 40 ms, FoV 192 mm,4 × 64 matrix, 28 slices, slice thickness 3 mm, gap 1 mm). Stimuli were pre-ented via LCD video goggles. Both reaction times and accuracy indices wereecorded. Head movement was minimized using padded ear phones. The fMRI-rotocol was a rapid event-related design with a pseudorandomized time-jitter of.5 ± 0.5 TR inter-trial-interval (mean trial duration of 10 s + 2.4–4.8 s). Stimuliere pseudorandomized and counterbalanced for the relative appearance fre-uency of each letter per load, highlighted position or target. The experimentalesign avoided the appearance of recent negative trials in order to avoid proac-ive interference effects during retrieval (Jonides, Smith, Marshuetz, Koeppe,

Reuter-Lorenz, 1998). There were three sessions in total, each including 28rials (7 trials per condition), comprising 164 volumes pro session (492 volumesn total). The first eight volumes of each session were discarded to allow forquilibration effects.

.4. Data analysis

.4.1. Behavioral data analysisTask performance was recorded as percentage of correct responses (accu-

acy) as well as the reaction times (RT) of correctly performed trials. Changes inask accuracy and RT with increasing load were assessed separately using uni-ariate analyses of variance (ANOVA) with group as the between-subjects factornd repeated measures for memory load (control condition and load levels 1–3).o avoid �-error accumulation, all post hoc tests were Bonferroni-corrected.

.4.2. fMRI data analysisAnalyses were performed with SPM2 (http://www.fil.ion.ucl.ac.uk/spm/),

mplemented and executed in MATLAB 6.1 (MathWorks, Natick, MA). Theunctional images were slice-timed, motion-corrected by realignment to the firstolume of each session, spatially normalized to the EPI standard template ofmm × 3 mm × 3 mm voxels and then spatially smoothed with a 9-mm fullidth at half maximum (FWHM) isotropic Gaussian kernel. Data were analyzedithin the framework of the General Linear Model (Friston, Frith, Frackowiak,Turner, 1995) using the canonical-hrf-function as a predictor. For correctly

nswered trials, we modeled the stimulus and probe periods of all loads as aingle regressor. In contrast, the delay phase was modeled for each conditioncontrol, loads 1–3) separately, as an event of six seconds spanning the wholeelay period. Thus, there were four delay regressors, and only one stimulusnd one probe regressor. With this model we minimized the degree of event-utocorrelation between the three trial phases. For incorrectly answered trials,e defined one regressor of no interest in which all incorrect trials of all load

evels were collapsed, excluding error trials from activation analyses. First orderotion parameters were individually modeled as regressors of no interest. In

rder to further exclude motion-related bias, we calculated an ANOVA includingll realignment parameters for each group and session. These analyses revealedhat the average range of motion in single subjects was between 10−4 mm and0−2 mm (x, y, z axis translation). Significant between-group differences in headotion were not found (all p > 0.20).

Analyses were performed at two levels. At the first level, subject specificelay effects of conditions were compared using linear contrasts, resulting in a-statistic for each voxel (highpass filter with a cutoff of 137 s, lowpass filter 4 s).

ain effects of load were calculated for the delay period of all three load levelsL1, L2, and L3) and the control condition (cc). We performed mixed effectsroup analysis at the second level, entering the calculated delay specific contrastsor the four conditions of each subject into a multi-group analysis of variance

ANOVA). For within-group comparisons, we contrasted each load with the con-rol condition, i.e. L1 > cc, L2 > cc and L3 > cc. Main effects of load within-groupre reported at an a priori chosen threshold of p < 0.05 family-wise error (FWE)orrected. For the between-group analysis, we calculated pair-wise comparisonsf loadn+1 > control condition. Between-group comparisons are reported at an

N. Vasic et al. / Neuropsychologia 46 (2008) 640–648 643

Table 2fMRI task performance of subjects with dyslexia and control subjects

Condition Dyslexics (n = 12) Controls (n = 13)

Reaction time Accuracy Reaction time Accuracy

Mean S.D. Mean S.D. Mean S.D. Mean S.D.

Control 592 141 97.4 3.2 562 176 99.6 1.3Load level 1 782 122 90.5 10.4 735 190 97.1 3.1LL

R d leve

ae

pteptdllTs

3

3

pbHsmp

6a(iie

3

adspa

3

tsia

vierlIlaedcTww

3

ipfaese(gbiftlec

3

dmrrF

oad level 2 936 129 90.0oad level 3 1173 143 72.3

eaction times are given in ms, accuracy in percent of correct answers. Red: loa

priori chosen threshold of p < 0.001 at the voxel level (uncorrected) with anxtent threshold correction of p < 0.05 at the cluster level.

We further conducted a linear regression analysis including the behavioralerformance indices obtained during digit span performance (backward) andhe subject specific images obtained on 1st level using the contrast for the mainffect of WM load (L3 > cc). We used digit span performance as a regressionarameter, since WM impairment of dyslexic subjects was most evident whenhe manipulation of verbal material was required. Due to the reduction of theegrees of freedom, we chose an a priori threshold of p < 0.005 at the voxelevel (uncorrected) with an extent threshold correction of p < 0.05 at the clusterevel. All anatomical regions are reported according to the atlas of Talairach andournoux (1988). Coordinates are maxima in a given cluster according to thetandard MNI-Template.

. Results

.1. Performance during the fMRI activation task

We found a main effect of factor load (F(3, 69) = 133.40,< 0.001) on reaction time, with increasing reaction times inoth groups (Table 2). There was no main effect of group.owever, there was a trend in the dyslexic group toward being

lower than the control group with increasing load, revealing theost significant difference at load level 3: (F(23) = 3.75, t = 1.94,= 0.065). No group by load interaction was found.

We found a significant main effect of both load (F(3,9) = 37.71, p < 0.001) and group (F(1, 23) = 6.41, p < 0.05) onccuracy. The group by load interaction was also significantF(3, 69) = 6.22, p < 0.001). Post hoc t-tests showed that thisnteraction was due to performance differences between groupsn L3: with highest memory load, controls committed fewerrrors than dyslexics (F(23), t = −3.08, p < 0.01).

.1.1. Functional imaging resultsBoth groups showed a main effect of WM load on BOLD

ctivation in a widely distributed network including the bilateralorso- and ventrolateral prefrontal cortex, premotor cortex, theupplementary motor area, bilateral striatum, cerebellum andarietal cortex (Fig. 4, tables with detailed activation coordinatesvailable on request).

.1.2. Between-group comparisonsThe lowest load level (load level 1, L1) did not yield activa-

ion differences in any direction. At load level 2 (L2), dyslexicshowed relatively more activation in the occipital cortex andn the parietal cortex (for detailed information on coordinates,natomical localization, Z-value and spatial extent of group acti-

arpu

9.3 864 184 91.2 5.45.4 1020 236 87.2 9.0

l with worse performance in dyslexics compared to healthy controls (p < 0.01).

ation differences, see Table 3). In contrast, controls showed anncreased activation of the left precentral gyrus. At the high-st load level 3 (L3), increased activation in dyslexic subjectselative to controls was found in the left prefrontal cortex,eft Broca’s area and its homologue in the right hemisphere.ncreased activation of the parietal cortex (left inferior parietalobule, BA 40; x = −39, y = −54, z = 39; Z = 3.70; number ofctivated voxels = 37) was found when the threshold was low-red to p < 0.001, uncorrected for spatial extent. Compared toyslexics controls showed more activation in the left precentralortex and in the superior parietal cortex (Fig. 2, upper panel;able 3). Furthermore, all functional between-group analysesere recalculated using reading scores as covariates of interestithout altering the main findings.

.1.3. Activation changes associated with WM loadFor further characterization of the relative fMRI-signal

ncreases with increasing WM load, we extracted mean effectarameters (corresponding to the percent signal change dif-erences) per subject and WM load at the most significantlyctivated voxels in the parietal and prefrontal cortical regionsmerging from the between group comparisons (Fig. 2). Ashown by the mean activation size, both groups exhibited a lin-ar relationship in the activation of the left inferior frontal gyrusBA 44) with increasing WM load. In the right inferior frontalyrus (BA 44), the dyslexic group showed a linear relationshipetween activation and WM load, whereas the activation patternn the control group proved to be non-linear and independentrom WM load (see Fig. 2, lower panel; a similar activation pat-ern was found in the left superior frontal gyrus, BA 8). In theeft inferior parietal cortex, the dyslexic group exhibited a lin-ar relationship between activation and WM load, whereas theontrol group showed deactivation with increasing WM demand.

.1.4. Simple regression analysesIn the dyslexic group, we found a positive correlation between

igit span backwards performance and activation in bilateraliddle frontal gyri (left BA 9; x = −45, y = 33, z = 27; Z = 3.73;

ight BA 46; x = 39, y = 21, z = 24; Z = 3.28) and the right supe-ior frontal gyrus (BA 8; x = 9, y = 30, z = 54; Z = 3.44; see alsoig. 3). In order to further investigate the impact of reading

bility on prefrontal and parietal activation, we calculated cor-elation analyses between the mean BOLD activation size inrefrontal and parietal regions of interest and the reading scoressed in this study. However, a significant correlation between

644 N. Vasic et al. / Neuropsycholo

Tabl

e3

Coo

rdin

ates

and

anat

omic

allo

caliz

atio

nof

regi

ons

inw

hich

cont

rols

ubje

cts

show

eddi

ffer

ence

sin

cere

bral

activ

atio

nco

mpa

red

with

dysl

exic

s

Con

trol

s>dy

slex

ics

Dys

lexi

cs>

cont

rols

Ana

tom

ical

regi

onx

yz

ZN

o.of

activ

ated

voxe

lsA

nato

mic

alre

gion

xy

zZ

No.

ofac

tivat

edvo

xels

Loa

dle

vel2

Lef

tpre

cent

ralg

yrus

−48

339

5.40

59L

eftl

ingu

algy

rus

−6−9

30

3.84

74L

efti

nfer

ior

pari

etal

gyru

s−4

2−6

948

3.71

77R

ight

lingu

algy

rus

21−7

8−3

3.33

77

Loa

dle

vel3

Lef

tmid

dle

fron

talg

yrus

−45

339

6.14

101

Lef

tsup

erio

rfr

onta

lgyr

us−2

727

514.

4048

Lef

tsup

erio

rpa

riet

algy

rus

−33

−60

605.

3811

9L

efti

nfer

ior

fron

talg

yrus

−42

1215

4.12

46R

ight

infe

rior

fron

talg

yrus

5718

124.

1166

Res

ults

ofth

ese

cond

leve

lbet

wee

n-gr

oup

AN

OV

A;p

<0.

001

unco

rrec

ted

atth

evo

xell

evel

,p<

0.05

corr

ecte

dfo

rsp

atia

lext

ent.

as

4

acslcrogtdraiiddp

dalrWtdlndejfinwro

ftDdsi(rffhivfio

gia 46 (2008) 640–648

ctivation level and reading scores was not found at the chosenignificance threshold (p < 0.05).

. Discussion

In this study, monolingual German adolescents and youngdults with developmental dyslexia and age and gender-matchedontrols performed an fMRI task that required manipulation andubsequent storage of verbal material in WM. Three differentevels of task difficulty (WM load levels) and a non-mnemonicontrol condition were included. Behaviorally, dyslexics showededuced task-accuracy as well as a tendency for being slowernly during the most demanding WM load level. Functionally,roup activation differences were found only for the two condi-ions with higher memory load. At the level of highest cognitiveemand (L3), dyslexics showed increased activation in left andight ventrolateral prefrontal regions (including Broca’s area),s well as in the left superior frontal gyrus. Controls showedncreased activation during L3 in the left precentral cortex andn the left superior parietal cortex compared with subjects withyslexia. Furthermore, a positive correlation was found betweenigit span backwards performance and activation in dorsolateralrefrontal areas in the dyslexic group only.

The behavioral performance during the fMRI activation taskemonstrated that dyslexic subjects exhibit worse accuracy andtendency for slower RT at the most demanding WM load

evel only. Since our fMRI paradigm requires a very distinctiveequirement of manipulation capacity (increasing with higher

M load), these results suggest that dyslexics might be par-icularly impaired in phonological WM tasks when cognitiveemand is high and phonological information has to be manipu-ated, rather than when the short-term storage is needed. Thisotion is further supported by significant differences duringigit span backward performance, where subjects with dyslexiaxhibited worse performance than controls. Accordingly, sub-ects did not perform worse on the digit span forward task. Bothndings suggest that impairments in the manipulation compo-ent make an important contribution to the reduced accuracyith high cognitive demand and are in accordance with the

esults of Reiter et al. (2005), who showed an equivalent patternf results in 11-year-old dyslexic children.

The positive correlation between digit span backward per-ormance and cerebral activation in prefrontal regions sensitiveo WM load (Altamura et al., 2007; Rypma, Prabhakaran,esmond, Glover, & Gabrieli, 1999; Wolf & Walter, 2005)espite reduced task accuracy suggests that dyslexics activateimilar prefrontal regions as control subjects during tasks involv-ng the manipulation of verbal material and executive functionsee also Fig. 4). Since alphabet retrieval performance, self-eported cognitive strategy and concentration difficulties duringMRI were comparable between both groups, we infer that theMRI task was appropriate for the dyslexic subjects and mayave not been biased by any task-specific conditions. Predom-

nant executive deficits, such as rehearsal and manipulation oferbal material held in WM, are in accordance with previousndings in dyslexic children and adults of impaired inhibitionf distracting stimuli and sequencing of events (Brosnan et al.,

N. Vasic et al. / Neuropsychologia 46 (2008) 640–648 645

Fig. 2. Upper panel: Regions in which control subjects showed differences in cerebral activation compared with dyslexics. Results of the second level between-groupANOVA at the load levels 2 and 3, p < 0.001 uncorrected at the voxel level, p < 0.05 corrected for spatial extent. Right: Mean activation effects (estimated betaparameters and standard error) in the inferior parietal lobule (left BA 40: x = −42, y = −69, z = 48; Z = 3.70) in the dyslexic and the control group. Lower panel:M feriory rametd evel, p

2RtpipTdi

ev

maflcawb

Fa

ean activation effects (estimated beta parameters and standard error) in the in= 18, z = 12; Z = 4.11) in the dyslexic and the control group. Averaged beta paifferences between dyslexics and controls, p < 0.001 uncorrected at the voxel l

002) as well as problem-solving (Jeffries & Everatt, 2004;eiter et al., 2005). However, the behavioral data derived from

he fMRI task performance cannot be attributed to manipulationrocesses alone, since the manipulation component in our tasknvolves several other cognitive functions including attentionalrocesses, prepotent probe selection and response preparation.herefore, a behavioral impairment in subjects with dyslexiauring this fMRI task cannot be precisely reduced to a single

mpaired executive domain such as manipulation of memoranda.

We found that increasing WM load was associated with lin-arly increasing cerebral activation in the bilateral dorso- andentrolateral prefrontal cortex, the premotor cortex, the supple-

aNat

ig. 3. Brain regions showing a significant positive correlation with digit span (backnalyses, p < 0.005 uncorrected at the voxel level, p < 0.05 corrected for spatial exten

frontal gyri (left BA 44: x = −42, y = 12, z = 15; Z = 4.12; right BA 44: x = 57,ers were extracted from the between-group comparisons (significant activation

< 0.05 corrected for spatial extent).

entary motor area, the parietal cortex, the bilateral striatumnd cerebellum in both groups (Fig. 4). Significant group dif-erences were present during intermediate and high WM loadevels only: controls showed increased activation of the left pre-entral cortex (BA 6) at L2 as well as at L3, and increasedctivation of the superior parietal cortex (BA 7) at L3 comparedith subjects with dyslexia. These WM related brain areas haveeen previously associated with verbal WM storage functions

nd subvocal rehearsal mechanisms (Baddeley, 2003; Cabeza &yberg, 2000; Paulesu, Frith, & Frackowiak, 1993). The mean

ctivation size revealed that both groups exhibited a linear rela-ionship between increasing WM load and cortical activation.

ward) performance in dyslexics. Results of the second level simple regressiont.

646 N. Vasic et al. / Neuropsychologia 46 (2008) 640–648

F with dl sis of

Bad

wshiasgetiTp

rtrmaaoBtwefii

lfrfniis2

fgroFaiwijratipom

ig. 4. Main activation effects of increasing WM manipulation load in subjectsevel within-group ANOVA (p < 0.05, FWE corrected), including only the analy

oth controls and dyslexics may therefore use similar brain areasssociated with verbal WM function, although to a differentegree and extent in activation.

In the dyslexic group, we identified three prefrontal areashich showed an increased activation compared to control

ubjects. These activation differences were present during theighest cognitive demand only. An increased activation of thenferior frontal cortex has been described for tasks using letterss stimuli (Cohen et al., 1994; de Zubicaray et al., 1998). Othertudies have emphasized the relevance of the left inferior frontalyrus in verbal executive functioning (Bunge et al., 2000; Osakat al., 2004; Zhang et al., 2003). As shown by the mean activa-ion size, both groups exhibited a linear relationship betweenncreasing WM load and cortical activation in Broca’s region.he dyslexic group additionally showed an increased activationattern in the right BA 44.

Previous findings have associated BA 44 with subvocalehearsal during WM (Zhang et al., 2003) and have speculatedhat this mechanism might be a way of enhancing or biasingelevant information (Smith et al., 2001). In our study, thisechanism is likely to occur when verbal information tends to

ccumulate, i.e. during higher WM load levels, and may serve asn additional speech-based neural input during the manipulationf memoranda. The left superior frontal gyrus (BA 8) as well asroca’s area and its corresponding right homologue are known

o be crucial for verbal controlling and monitoring processes, asell as for rehearsal and maintenance of material during verbal

xecutive functioning (Wolf et al., 2006; Zhang et al., 2003). Ourunctional results suggest that the relatively increased activationn Broca’s area in dyslexics during verbal executive function-ng is foremost associated with task accuracy at higher WM

tsda

yslexia and healthy controls. The activation effects are derived from the secondcorrect trials.

oad levels. The linear relationship of activation and WM loadurther indicates that dyslexics additionally recruit the right infe-ior frontal gyrus at high WM load levels, possibly compensatingor worse task accuracy by recruiting additional right-lateralizedeural resources. This cortical region seems to be of particularnterest in dyslexia, since there is increasing evidence for thenvolvement of right-lateralized prefrontal regions in subvocalpeech processes and reading ability (Larsen, Baynes, & Swick,004), which are overtly impaired in subjects with dyslexia.

The finding of increased bilateral activation in the inferiorrontal gyri is also in accordance with findings of other researchroups suggesting compensatory over-activation of the left andight inferior frontal gyrus in dyslexics during the performancef phonological tasks (Brunswick, McCrory, Price, Frith, &rith, 1999; Pugh et al., 2000; Shaywitz et al., 1998). Shaywitz etl. (2002) found increased activation of the left and right BA 44n adolescents with developmental dyslexia performing pseudo-ord and real-word reading tasks. Since the activation in the

nferior frontal areas correlated positive with the age of the sub-ects, the authors hypothesized that dyslexics might increasinglyecruit inferior frontal areas over time in order to compensate fordysfunction of speech-related left parietotemporal and occipi-

otemporal regions. Although the functional activation task usedn Shaywitz et al.’s study required phonological and semanticrocessing rather than manipulation of memoranda during WM,ur functional results again indicate that a compensatory recruit-ent of inferior frontal areas in dyslexics may occur in order

o optimize task performance. Temple et al. (2003) also demon-trated increased activity in dyslexics in the inferior frontal gyrusuring a phonological processing task. However, this increasedctivation was observed after remediation, indicating that

cholo

i(

ibtpmddiddcdeatspasab

avWiSsvri

aadidrtfBaphradiwpMcr(

ntwtdraoicnd

lhdpdaaa

A

prHwPo

R

A

B

B

B

B

B

B

B

N. Vasic et al. / Neuropsy

nferior frontal activation in dyslexics may change over timeTemple et al., 2003).

At first sight, our findings of increased prefrontal activationn the dyslexic group seem to contradict the results of the studyy Paulesu et al. (1996). In this study, dyslexics exhibited aask related hypoactivation of Broca’s area and the temporo-arietal cortex during tasks involving rhyming and a short-termemory. Since an activation of the left insula was absent in the

yslexic group, the authors hypothesized that a phonologicaleficit in dyslexics may result from an impaired connectiv-ty between anterior and posterior language areas. Despite theecreased prefrontal activation found in Paulesu et al.’s study,yslexics’ performance did not differ from that of controls, indi-ating that a different prefrontal activation pattern in dyslexicsoes not necessarily reflect limitations in task accuracy. How-ver, this study investigated a relatively small sample size (n = 5)nd did not overtly require manipulation processes. In contrast,he activation task employed in the present study included atrong alphabetical manipulation component with a sufficientarametric variation. A strong emphasis on executive function,s provided by our activation paradigm, is more likely to revealignificant differences in task accuracy at higher WM load levelsnd neural compensation mechanisms associated with markedehavioral accuracy decreases.

The absence of a significant correlation between cerebralctivation level and reading scores further suggests that the acti-ation differences in our study might primarily reflect impairedM function in dyslexics rather than the status of reading abil-

ty. As it has been hypothesized previously (Swanson, 1993;wanson & Sachse-Lee, 2001), a domain-unspecific executiveystem may partially contribute to poor WM function in indi-iduals with reading disability, possibly independent of overteading deficits. However, this hypothesis was not directly testedn our study and warrants further investigation.

One objection to our functional findings is that the increasedctivation found in the bilateral inferior frontal gyri might ben artifact of reduced task accuracy, which was present in theyslexic group at L3, i.e. simply reflecting task difficulty. Fornstance, Hoeft et al. compared brain activations of dyslexic chil-ren aged 9–14 years with two control groups (age-matched andeading-matched subjects) using a visual word rhyme judgementask in order to distinguish activations related to dyslexia per serom those simply reflecting task difficulty (Hoeft et al., 2007).ased on the differential between-group activation patterns, theuthors concluded that hyperactivation in dyslexia might reflectrocesses related to the current level of reading ability, whereasypoactivation might be associated with functional alterationselated to dyslexia itself. However, this interpretation of hypo-nd hyperactivation does not fully apply to our findings, sinceifferent cognitive processes and different age groups were stud-ed: Hoeft et al. measured phonological awareness (via a visualord rhyme judgement task), whereas the task employed in theresent study assessed increasing WM manipulation demand.

oreover, the participants in the study of Hoeft et al. were

hildren (aged 9–14 years), who were still in the process ofeading development, while ours were adolescents and adultsmean age 18 years). Still, task difficulty is an important determi-

C

gia 46 (2008) 640–648 647

ant for the findings of the present study. Activation differenceshat emerged between groups during higher cognitive demandsere not present at the less challenging task conditions. The pat-

ern of activation differences also changed with increasing taskemand. However, our event-related analysis excluded incor-ectly performed and omitted trials, thus controlling for potentialctivation confounds arising from these trials. The phenomenonf increased prefrontal activation and impaired task accuracyn subjects with dyslexia might therefore be related to ineffi-ient cortical processing, as it has been previously proposed foreuropsychiatric conditions with both subtle and manifest WMeficits (Callicott et al., 2003).

In conclusion, our data point to an executive deficit in ado-escents and young adults with dyslexia when verbal materialas to be manipulated in WM. The functional data suggest thatyslexics might recruit additional predominantly speech-relatedrefrontal regions with increasing verbal WM processing. Thisifferential activation can be interpreted both as a signature ofneural inefficiency within a WM related prefrontal network

s well as a compensation mechanism involving an increasedctivation of parietal and inferior prefrontal regions.

cknowledgements

The authors are grateful to all subjects for their kind partici-ation in our study, to Margret Linner for her commitment to theecruitment of the subjects and to Manfred Spitzer and Katrinille for supporting the realization of this project. The authorsould like to thank Timothy Laumann, Genes, Cognition andsychosis Program, NIMH, Bethesda, for insightful commentsn a previous version of this manuscript.

eferences

ltamura, M., Elvevag, B., Blasi, G., Bertolino, A., Callicott, J. H., Weinberger,D. R., et al. (2007). Dissociating the effects of Sternberg working memorydemands in prefrontal cortex. Psychiatry Research, 154(2), 103–114.

addeley, A. (1996). The fractionation of working memory. Proceedings ofthe National Academy of Sciences of the United States of America, 93(24),13468–13472.

addeley, A. (2003). Working memory: Looking back and looking forward.Nature Reviews Neuroscience, 4(10), 829–839.

addeley, A., & Wilson, B. A. (1993). A developmental deficit in short-termphonological memory: Implications for language and reading. Memory, 1(1),65–78.

rambati, S. M., Termine, C., Ruffino, M., Danna, M., Lanzi, G., Stella, G., etal. (2006). Neuropsychological deficits and neural dysfunction in familialdyslexia. Brain Research, 1113(1), 174–185.

rosnan, M., Demetre, J., Hamill, S., Robson, K., Shepherd, H., & Cody, G.(2002). Executive functioning in adults and children with developmentaldyslexia. Neuropsychologia, 40(12), 2144–2155.

runswick, N., McCrory, E., Price, C. J., Frith, C. D., & Frith, U. (1999).Explicit and implicit processing of words and pseudowords by adult devel-opmental dyslexics: A search for Wernicke’s Wortschatz? Brain, 122(Pt 10),1901–1917.

unge, S. A., Klingberg, T., Jacobsen, R. B., & Gabrieli, J. D. (2000). A resource

model of the neural basis of executive working memory. Proceedings ofthe National Academy of Sciences of the United States of America, 97(7),3573–3578.

abeza, R., & Nyberg, L. (2000). Imaging cognition II: An empirical review of275 PET and fMRI studies. Journal of Cognitive Neuroscience, 12, 1–47.

6 ycholo

C

C

C

D

d

F

H

H

J

J

K

K

L

M

N

O

O

P

P

P

P

R

R

S

S

S

S

S

S

S

S

T

T

T

W

WW

W

48 N. Vasic et al. / Neurops

allicott, J. H., Mattay, V. S., Verchinski, B. A., Marenco, S., Egan, M. F.,& Weinberger, D. R. (2003). Complexity of prefrontal cortical dysfunctionin schizophrenia: More than up or down. American Journal of Psychiatry,160(12), 2209–2215.

hiappe, P., Hasher, L., & Siegel, L. S. (2000). Working memory, inhibitorycontrol, and reading disability. Memory & Cognition, 28(1), 8–17.

ohen, J. D., Forman, S. D., Braver, T. S., Casey, B. J., Servan-Schreiber, D., &Noll, D. C. (1994). Activation of prefrontal cortex in a non-spatial workingmemory task with fMRI. Human Brain Mapping, 1, 293–304.

’Esposito, M., Postle, B. R., & Rypma, B. (2000). Prefrontal cortical con-tributions to working memory: Evidence from event-related fMRI studies.Experimental Brain Research, 133(1), 3–11.

e Zubicaray, G. I., Williams, S. C., Wilson, S. J., Rose, S. E., Brammer, M.J., Bullmore, E. T., et al. (1998). Prefrontal cortex involvement in selectiveletter generation: A functional magnetic resonance imaging study. Cortex,34(3), 389–401.

riston, K. J., Frith, C. D., Frackowiak, R. S., & Turner, R. (1995). Characterizingdynamic brain responses with fMRI: A multivariate approach. Neuroimage,2(2), 166–172.

elland, T., & Asbjornsen, A. (2000). Executive functions in dyslexia. Neuropsy-chology, Development, and Cognition. Section C Child Neuropsychology,6(1), 37–48.

oeft, F., Meyler, A., Hernandez, A., Juel, C., Taylor-Hill, H., Martindale, J. L., etal. (2007). Functional and morphometric brain dissociation between dyslexiaand reading ability. Proceedings of the National Academy of Sciences of theUnited States of America, 104(10), 4234–4239.

effries, S., & Everatt, J. (2004). Working memory: Its role in dyslexia and otherspecific learning difficulties. Dyslexia, 10(3), 196–214.

onides, J., Smith, E. E., Marshuetz, C., Koeppe, R. A., & Reuter-Lorenz, P. A.(1998). Inhibition in verbal working memory revealed by brain activation.Proceedings of the National Academy of Sciences of the United States ofAmerica, 95(14), 8410–8413.

ane, M. J., & Engle, R. W. (2002). The role of prefrontal cortex in working-memory capacity, executive attention, and general fluid intelligence: Anindividual-differences perspective. Psychonomic Bulletin & Review, 9(4),637–671.

ersting, M., & Althoff, K. (2004). Orthography and spelling test (RT).Gottingen: Hogrefe.

arsen, J., Baynes, K., & Swick, D. (2004). Right hemisphere reading mecha-nisms in a global alexic patient. Neuropsychologia, 42(11), 1459–1476.

anoach, D. S. (2003). Prefrontal cortex dysfunction during working memoryperformance in schizophrenia: Reconciling discrepant findings. Schizophre-nia Research, 60(2/3), 285–298.

elson, H., & Warrington, E. K. (1980). An investigation of memory functionsin dyslexic children. British Journal of Psychology, 71, 487–503.

ldfield, R. C. (1971). The assessment and analysis of handedness: The Edin-burgh inventory. Neuropsychologia, 9(1), 97–113.

saka, N., Osaka, M., Kondo, H., Morishita, M., Fukuyama, H., & Shibasaki,H. (2004). The neural basis of executive function in working mem-ory: An fMRI study based on individual differences. Neuroimage, 21(2),623–631.

aulesu, E., Frith, C. D., & Frackowiak, R. S. (1993). The neural corre-

lates of the verbal component of working memory. Nature, 362(6418),342–345.

aulesu, E., Frith, U., Snowling, M., Gallagher, A., Morton, J., Frackowiak,R. S., et al. (1996). Is developmental dyslexia a disconnection syndrome?Evidence from PET scanning. Brain, 119(Pt 1), 143–157.

Z

gia 46 (2008) 640–648

oblano, A., Valadez-Tepec, T., de Lourdes Arias, M., & Garcia-Pedroza,F. (2000). Phonological and visuospatial working memory alterations indyslexic children. Archives of Medical Research, 31(5), 493–496.

ugh, K. R., Mencl, W. E., Jenner, A. R., Katz, L., Frost, S. J., Lee, J. R.,et al. (2000). Functional neuroimaging studies of reading and reading dis-ability (developmental dyslexia). Mental Retardation and DevelopmentalDisabilities Research Reviews, 6(3), 207–213.

eiter, A., Tucha, O., & Lange, K. W. (2005). Executive functions in childrenwith dyslexia. Dyslexia, 11(2), 116–131.

ypma, B., Prabhakaran, V., Desmond, J. E., Glover, G. H., & Gabrieli, J. D.(1999). Load-dependent roles of frontal brain regions in the maintenance ofworking memory. Neuroimage, 9(2), 216–226.

chulte-Korne, G. (2001). Dyslexia and speech perception. Munster: Waxmann.,pp. 138–139, 273

haywitz, B. A., Shaywitz, S. E., Pugh, K. R., Mencl, W. E., Fulbright, R. K.,Skudlarski, P., et al. (2002). Disruption of posterior brain systems for read-ing in children with developmental dyslexia. Biological Psychiatry, 52(2),101–110.

haywitz, S. E., & Shaywitz, B. A. (2005). Dyslexia (specific reading disability).Biological Psychiatry, 57(11), 1301–1309.

haywitz, S. E., Shaywitz, B. A., Pugh, K. R., Fulbright, R. K., Constable, R.T., Mencl, W. E., et al. (1998). Functional disruption in the organization ofthe brain for reading in dyslexia. Proceedings of the National Academy ofSciences of the United States of America, 95(5), 2636–2641.

mith, E. E., Geva, A., Jonides, J., Miller, A., Reuter-Lorenz, P., & Koeppe, R.A. (2001). The neural basis of task-switching in working memory: Effectsof performance and aging. Proceedings of the National Academy of Sciencesof the United States of America, 98(4), 2095–2100.

mith, E. E., & Jonides, J. (1999). Storage and executive processes in the frontallobes. Science, 283(5408), 1657–1661.

wanson, H. L. (1993). Working memory in learning disability subgroups.Journal of Experimental Child Psychology, 56(1), 87–114.

wanson, H. L., & Sachse-Lee, C. (2001). A subgroup analysis of workingmemory in children with reading disabilities: Domain-general or domain-specific deficiency? Journal of Learning Disabilities, 34(3), 249–263.

alairach, J., & Tournoux, P. (1988). Co-planar stereotaxic atlas of the humanbrain. New York: Thieme.

emple, E., Deutsch, G. K., Poldrack, R. A., Miller, S. L., Tallal, P., Merzenich,M. M., et al. (2003). Neural deficits in children with dyslexia amelioratedby behavioral remediation: Evidence from functional MRI. Proceedings ofthe National Academy of Sciences of the United States of America, 100(5),2860–2865.

ewes, U. (1994). Hamburg-Wechsler Intelligenztest fur Erwachsene—Revision.Bern: Hans Huber.

ager, T. D., & Smith, E. E. (2003). Neuroimaging studies of working mem-ory: A meta-analysis. Cognitive, Affective & Behavioral Neuroscience, 3(4),255–274.

eiß, R. H. (1997). Basic intelligence test scale 2 (CFT 20). Gottingen: Hogrefe.olf, R. C., Vasic, N., & Walter, H. (2006). Differential activation of ventrolat-

eral prefrontal cortex during working memory retrieval. Neuropsychologia,44(12), 2558–2563.

olf, R. C., & Walter, H. (2005). Evaluation of a novel event-related paramet-

ric fMRI paradigm investigating prefrontal function. Psychiatry ResearchNeuroimaging, 140(1), 73–83.

hang, J. X., Leung, H. C., & Johnson, M. K. (2003). Frontal activations asso-ciated with accessing and evaluating information in working memory: AnfMRI study. Neuroimage, 20(3), 1531–1539.