bilateral parieto-frontal network for verbal working memory: an interference approach using...

SHORT COMMUNICATIONBilateral parieto-frontal network for verbal workingmemory: an interference approach using repetitivetranscranial magnetic stimulation (rTMS)

Felix M. Mottaghy,1,2 Tobias DoÈring,3 Hans-Wilhelm MuÈ ller-GaÈrtner,1,2 Rudolf ToÈpper3 and Bernd J. Krause1,2

1Department of Nuclear Medicine (KME), Research Center JuÈ lich, 52426 JuÈ lich, Germany2Department of Nuclear Medicine, Heinrich-Heine University DuÈsseldorf, Moorenstr. 5, 40225 DuÈsseldorf, Germany3Department of Neurology, Technical University Aachen, Pauwelsstr. 30, 52057 Aachen, Germany

Keywords: cognition, human, mapping

Abstract

Verbal working memory has been attributed to a left-dominant neuronal network, including parietal, temporal and prefrontal

cortical areas. The current study was designed to evaluate the contribution of these brain regions to verbal working memory

processes and to assess possible hemispheric asymmetry. The effect of repetitive transcranial stimulation (rTMS) onperformance in a verbal working memory task both during, and after an rTMS train (110% of individual motor threshold, 4 Hz)

over nine different scalp locations was studied [bilateral middle frontal gyrus (MFG), bilateral supramarginal gyrus (SMG), bilateral

inferior parietal cortex (IP) and three different midline control sites]. Signi®cant performance deterioration was observed duringrTMS over the left and right MFG and left and right IP. There was no consistent interference effect across subjects over the left

or right SMG and the three different midline control sites. The interference effect with the given stimulation parameters did not

last beyond the rTMS train itself. The data provide evidence for a symmetrical, bilateral parieto-frontal verbal working memorynetwork. The data are discussed with respect to the competing ideas of a parieto-frontal central executive network vs. a network

that processes the inherent semantic and object features of the visually presented verbal stimuli in parallel.

Introduction

Neuroimaging studies on verbal working memory have revealed

bilateral parietal and prefrontal activations predominantly in the left

hemisphere (for review see Smith & Jonides, 1999). The ability to

perform working memory tasks is strongly correlated with the ability

to control attention and sustain its focus on a particular active mental

representation in the presence of distracting in¯uences (Miyake &

Shah, 1999). Several models on working memory have been proposed

(for review see Miyake & Shah, 1999). At present the model proposed

by Baddeley & Hitch (1974) is the most extensively investigated

theoretical construct of working memory. It is based on behavioural

experiments in normal and brain-damaged patients as well as on

functional neuroimaging studies. Baddeley & Hitch (1974) argue for a

model with a modality-independent central executive with ancillary

storage and buffer systems for visuospatial or verbal stimuli, the

visuospatial sketchpad and the phonological loop, respectively. The

latter was further subdivided into the phonological buffer and

the subvocal rehearsal system. The central executive is assumed to

be an attentional control system being responsible for strategy

selection, as well as for control and coordination of various processes

involved in short-term storage and general processing tasks. In most

previous functional imaging studies, central executive processes were

shown to be related to prefrontal activations (for review see Smith &

Jonides, 1999). More recently Collette & Van der Linden (2002)

reviewed imaging studies of numerous different cognitive tasks

involving executive processes. They proposed that the theoretically

de®ned central executive might rather be the result of an interaction of

prefrontal and parietal areas, in the sense of an interlinked network

rather than a function of circumscribed prefrontal regions (Collette &

Van der Linden, 2002). The buffer systems have been localized in

more posterior cortical regions (for review see Smith & Jonides, 1999)

while the subvocal rehearsal system has been assigned to inferior

prefrontal cortical regions (Paulesu et al., 1993). More speci®cally

areas within the left supramarginal gyrus (SMG) and the left inferior

parietal cortex (IP) were found to be active in previous imaging studies

on verbal working memory (for review see Smith & Jonides, 1999).

Activations in the left SMG (Paulesu et al., 1993) were attributed to the

phonological buffer (Baddeley & Hitch, 1974) but this area has been

discussed in the context of motor attention, the control of limb

movements and number processing (Deiber et al., 1996; GoÈbel et al.,

2001; Rushworth et al., 2001). Activations in IP were attributed to

control and storage of task-related information and again a left

hemispheric dominance for verbal stimuli was observed (for review

see Smith & Jonides, 1999).

Beyond the neuroimaging techniques that allow one to correlate an

activation in a given area with behaviour in a given task, transcranial

magnetic stimulation (TMS) as an interference method can disen-

tangle whether a cortical region is epiphenomenally or crucially

Correspondence: Dr Felix M. Mottaghy, Department of Nuclear Medicine(KME), as above. E-mail: [email protected]

Received 12 July 2002, revised 29 July 2002, accepted 5 August 2002

doi:10.1046/j.1460-9568.2002.02209.x

European Journal of Neuroscience, Vol. 16, pp. 1627±1632, 2002 ã Federation of European Neuroscience Societies

involved in the cognitive task under study (for review see Pascual-

Leone et al., 2000).

The aim of the present study was to deliver rTMS to the most

consistently involved regions in verbal working memory in order to

test their contribution to the studied task (namely left and right middle

frontal gyrus (MFG), left and right IP, and left and right SMG).

Materials and methods

Ten right-handed healthy male volunteers (mean age 24.8 6 2.3

years) participated in the study, which was approved by the ethical

committee of the University of Aachen and which was performed in

accordance with the Declaration of Helsinki. Subjects were recruited

from the local student population and gave written informed consent.

They were naive to the intention of the study. All had a normal

neurological exam and no contraindications for rTMS according to

current guidelines (Wassermann, 1998). We employed a 2-back

verbal working memory task using the ®rst four letters of the alphabet

as stimuli (Smith & Jonides, 1999). Each block consisted of 20

stimuli with an interstimulus interval of 1.5 s. Subjects had to

respond to every single letter by using a response box with two keys.

One key was assigned to the matching response (i.e. the letter shown

two before in the sequence was the same as the actual letter) and the

other key to the negative response (i.e. the letter shown two before in

the sequence was not the same). Repetitive TMS was applied using a

Magstim Rapid (The Magstim Company Ltd, Whitland, UK)

stimulator equipped with a commercially available 9 cm ®gure-of-

eight coil centred over F3, F4, Fz, P3, P4 and Pz of the international

10±20 EEG system (Jasper, 1957) and midway between Pz and Cz

(CPz), midway between P3 and C3 (CP3) and midway between P4

and C4 (CP4) for stimulation of the left or right MFG (F3, F4),

midline prefrontal cortex (Fz), midline parietal cortex (CPz, Pz), left

or right SMG (CP3, CP4) and left and right IP (P3, P4), respectively.

The rTMS stimulation intensity was set at 110% of the individual

active motor threshold (MT). The active motor threshold was

FIG. 1. (A) The experimental design was established to determine the performance of the subjects before, during and after a 4 Hz 30 s rTMS train at 110%motor threshold. The ®rst four letters of the alphabet served as stimuli and were randomly presented every 1.5 s for 30 s (i.e. 20 stimuli). Subjects had toindicate with two ®ngers of the right hand whether the letter they saw two before the presented stimulus was the same or not. After a baseline task, subjectsperformed the task during the rTMS train, than at 15 and 60 s after the end of the train. (B) Positioning of the coil over the left or right middle frontal gyrus(MFG; F3 and F4), midline prefrontal cortex (Fz), midline parietal cortex (CPz and Pz), the left or right supramarginal gyrus (CP3 and Cp4) and left or rightinferior parietal cortex (P3 and P4) was con®rmed in 3 of 10 subjects using a 3D MRI sequence with vitamin E capsules in place.

1628 F. M. Mottaghy et al.

ã 2002 Federation of European Neuroscience Societies, European Journal of Neuroscience, 16, 1627±1632

determined as the minimum percentage of the magnetic stimulator

output that can evoke a visually detectable twitch in a tense muscle of

the subject's hand, contralateral to the stimulated motor cortex

(Rothwell et al., 1999). Initially subjects underwent several training

sessions until their performance was in the range between 85 and

95% correct answers. Then two baseline 30-s long blocks of data

were acquired, before rTMS was applied at a frequency of 4 Hz

during one block. The next block was started 15 s after completion of

the rTMS train, which was followed by another block at 60 s after the

rTMS train (Fig. 1A). We intended to stimulate during the whole

time of task performance (30 s) to ensure we achieved on-line

interference at any time during task performance. It was decided

to use 4 Hz to remain within currently proposed safety criteria

(Wassermann, 1998). A behavioural effect would have been less

likely with lower frequencies and beyond current safety standards

with higher frequencies (Wassermann, 1998). It has been demon-

strated in other studies that this is a reasonable approach to use, as

every area has its own task-dependant time-point for a single pulse

TMS effect (Mottaghy et al., 2002), therefore it would have been very

likely that these experiments would have missed this `vulnerable'

point when using rTMS at a ®xed time in the delay phase of the

working memory task.

The order of the nine different locations was randomized across

subjects. In three of the 10 subjects precise anatomical information

about the brain area stimulated was determined by a 3-dimensional

reconstructed brain MRI, in which vitamin E capsules were taped

onto the scalp locations (F3, F4, Fz, CPz, CP3, CP4, P3 and P4) to be

stimulated (Fig. 1B). In all cases the capsules were well above the

middle third of the MFG (F3 and F4), midline prefrontal cortex or the

parietal cortex (Fz, CPz and Pz), above the SMG (CP3 and CP4) and

the inferior parietal cortex (IP) close to the intraparietal sulcus (P3

and P4). The inter-individual variability of the anatomical targets in

the subset of subjects for whom we had an MRI scan was well within

the presumed spatial resolution of the coil used (1±1.5 cm). Therefore

in the following results the anatomical termini for the stimulated

areas are used. Other groups have also demonstrated a good overlap

of the EEG electrode placement and the intraparietal sulcus (Oliveri

et al., 1999), the supramarginal gyrus or the middle frontal gyrus

(Homan et al., 1987).

The accuracy and the mean reaction time (RT) were calculated for

each block. The two baseline blocks before rTMS were merged to

form an overall baseline. The trials both during and after the rTMS

train were regarded as different time-points given the likely transient

effects of rTMS. First an overall mean RT for each subject was

established. Individual outliers were de®ned when RT was beyond

two standard deviations and were eliminated (5% of the responses).

To determine if there was a RT error-rate trade-off, responses were

correlated, and it was found that there were no signi®cant correlations

across any of the conditions used (all P > 0.05). RT and errors were

then analysed separately employing a 9 3 4 repeated measures

ANOVA (9 sites ± F3, F4, Fz, CP3, CP4, CPz, P3, P4 and Pz; and 4

times ± baseline, and during rTMS, post-rTMS 1 and post-rTMS 2).

FIG. 2. The mean reaction time across the 10 subjects is shown for the 6 different time-points of the experiment. There was no consistent effect acrosssubjects at any of the 9 different stimulation points (left MFG, left middle frontal gyrus; right MFG, right middle frontal gyrus; MF, midline prefrontalcortex; MP1, midline parietal cortex anterior; left SMG, left supramarginal gyrus; right SMG, right supramarginal gyrus, left IP, left inferior parietal cortex;right IP, right inferior parietal cortex; MP2, midline parietal cortex posterior).

rTMS interference effects on working memory 1629


Furthermore post-hoc planned comparisons (LSD test) were per-

formed comparing the during rTMS block with the baseline and post-

rTMS blocks for each stimulation site separately.

Results

The mean RT across all sites was 492 6 35 ms (mean 6 SEM). The

analysis of RT revealed no signi®cant effects in the two-way ANOVA

(F24,216 = 1.07; P > 0.38). Also one±way interactions were not

signi®cant (Fig. 2). The repeated measures two-way ANOVA for the

error-rates however, was signi®cant (F24,216 = 2.43; P < 0.0004;

Fig. 3). There was no signi®cant difference in the all-over accuracy

rate across the nine different stimulation sites (F8,72 = 0.75;

P > 0.42). The mean accuracy, irrespective of the stimulation site,

was 90.8% 6 3.0%. The one-way ANOVA of the accuracy for the

different time-points, irrespective of the stimulation site, was

signi®cant (F3,27 = 8.55; P < 0.0004). In the post-hoc planned

comparisons rTMS over the left or right MFG revealed a signi®cant

interference effect on accuracy (left, 79.9% 6 4.3%; right,

76.7% 6 5.1%; Fig. 3) during rTMS with respect to the baseline

and the two following task blocks (left, P < 0.003; right, P < 0.003).

The ®rst block after the rTMS train over the left and right middle

frontal gyrus showed a slight tendency to a deteriorated accuracy

(left, 88.9% 6 2.5%; right, 89.9% 6 2.7%) but this was no longer

signi®cant with regard to the baseline (P > 0.1). Also over the

inferior parietal cortex rTMS during the task resulted in a signi®cant

interference with accuracy, bilaterally (left, 84.4% 6 5.4%,

P < 0.03; right, 84.9% 6 5.9%, P < 0.02) that was no longer

detectable 15 s after the end of the rTMS-train. rTMS over the left

or right SMG as well as over the three different midline stimulation

sites had no effect on task performance.

Discussion

Using rTMS it was possible to interfere with the ability to perform a

verbal working memory task by stimulating either the MFG or the IP

of either hemisphere. This effect did not last beyond the rTMS train.

Stimulation over the left or right SMG did not result in interference

with task performance. rTMS at the reference regions over the

midline parietal cortex or the midline prefrontal cortex also did not

FIG. 3. The mean accuracy (in percentage correct) across the 10 subjects is shown for the 6 different time-points with regard to the 9 different corticalstimulation sites. The stimulation sites are superimposed on a rendered surface MRI of one of the subjects. **rTMS to bilateral MFG as well as bilateral IPled to a signi®cant (P < 0.05) deterioration of the accuracy, that was no longer detectable already 15 s after the end of the rTMS train; (1±6 time of taskperformance with respect to the rTMS train).



result in a signi®cant deterioration of performance, excluding the

possibility of unspeci®c rTMS effects due to the discharge click of

the machine or the tactile sensation. A consistent effect on RT across

subjects was not found at any stimulated brain site.

In the present study it was dif®cult to state which speci®c function

of working memory showed an interference induced by rTMS at the

prefrontal or the parietal cortex. Both areas have been discussed in

the context of verbal working memory, however, a left hemispheric

dominant activation especially in the parietal cortex was observed in

most studies (for review see Smith & Jonides, 1999 or Collette & Van

der Linden, 2002). It is possible that the bilateral involvement of

prefrontal and parietal regions observed in this study was correlated

to the cognitive load and the dif®culty of the task. It cannot be

excluded that with a 1-back condition, a more pronounced left

hemisphere lateralization for verbal stimuli could have been demon-

strated.

The n-back paradigm requires storage processes and the necessity

to continuously modify the content of working memory. These

processes are accomplished by executive processes, in line with

Collette & Van der Linden's (2002) proposal of a central executive

anterior-posterior network, it could be argued that with interference

of components of the central executive network it is possible to

interfere with task performance. According to the results discussed

here, components of this network would be bilateral to MFG as well

as bilateral to IP. Lesion studies that showed behavioural impairments

in executive functions in patients with parietal lesions (Andres & Van

der Linden, 2000) comparable to behavioural features after prefrontal

lesions (Shallice, 1982) support the idea of this fronto-parietal

network. Functional cortico-cortical circuits that could support the

idea of a bilateral fronto-parietal network of brain regions contrib-

uting to central executive functions in humans have also been

demonstrated in nonhuman primates (e.g. Chafee & Goldman-Rakic,

2000).

A previous study on verbal working memory used the combination

of rTMS and positron emission tomography and revealed comparable

behavioural results with left or right MFG stimulation (Mottaghy

et al., 2000). The regional cerebral blood ¯ow changes induced by

rTMS were most pronounced with right MFG rTMS. An explanation

for the effects on working memory accuracy with right MFG rTMS

could be that there is an interference with attentional resources that

have been shown to be lateralized to the right hemisphere (Coull et al.,

1996). The same explanation could be applied for the interference

with accuracy by bilateral IP stimulation found in the current

experiment. A pure effect on attentional resources seems unlikely, as

usually when attention or alertness is in¯uenced by an external

stimulus one would expect to also see a speed effect (e.g. Pascual-

Leone et al., 1994; Ashbridge et al., 1997; Rushworth et al., 2001;)

which was not observed in the present study.

Beside the earlier proposed distributed fronto-parietal central

executive network (Collette & Van der Linden, 2002) it is also

possible that the two MFG together form the central executive for

working memory (Mottaghy et al., 2000). If central executive

functions are attributed to bilateral MFG activity, the left IP could be

the neuronal substrate for the phonological buffer (semantic process-

ing), while the right IP could be involved in the parallel storage/

processing of the visually presented letter as an object, i.e. right IP

could be the neuronal equivalent of the visuospatial sketchpad. This

stimulus-speci®c lateralization of involved brain regions has been

discussed in some previous papers (for review see Smith & Jonides,

1999).

The lack of interference with SMG rTMS might be explained by

the capacity of the brain to switch buffer functions into the so called

subvocal rehearsal system that has been attributed to the inferior

prefrontal cortex (Paulesu et al., 1993). It could also be postulated

that the SMG is not crucially involved in the studied 2-back verbal

working memory task. Based on the current study we cannot exclude

SMG involvement in verbal working memory studies using letter

strings or words. Several previous studies on motor attention using

rTMS showed an interference with task performance stimulating the

SMG (Rushworth et al., 2001). The fact that no interference was

observed with SMG stimulation but with inferior parietal cortex

stimulation excludes the possibility of a pure disruption of the motor

response required in this task (active ®nger movement to press the

key) as this area was seen to be activated in relation with control of

limb movements (Deiber et al., 1996).

Conclusion

One may propose two possible explanations for the observed bilateral

involvement of MFG and IP in verbal working memory. First it is

possible that left and right IP contribute to central executive functions

that are conjointly executed by the MFG's, which would be in favour

of a distributed central executive neuronal network. Second the

results could be explained by an interference with central executive

functions in the MFG's whereas the left and right IP effects could be

due to interferences in posterior buffer systems. In this scenario the

inherent semantic and object features of the visually presented verbal

stimuli would be processed in parallel.

Acknowledgements

The study was supported by the START-program of the Medical Faculty of theUniversity of Aachen and the `InterdisziplinaÈres Zentrum PraÈvention undKompensation von StoÈrungen des ZNS'.

Abbreviations

IFG, inferior frontal gyrus; IP, inferior parietal cortex; MFG, middle frontalgyrus; rTMS, repetitive transcranial magnetic stimulation; RT, reaction time;SMG, supramarginal gyrus.

References

Andres, P. & Van der Linden, M. (2000) Age-related differences insupervisory attentional system functions. J. Gerontol. B. Psychol. Sci.Soc. Sci., 55, P373±P380.

Ashbridge, E., Walsh, V. & Cowey, A. (1997) Temporal aspects of visualsearch studied by transcranial magnetic stimulation. Neuropsychologia., 35,1121±1131.

Baddeley, A.D. & Hitch, G. (1974) Working memory. In Bower, G.A., (ed).The Psychology of Learning and Motivation. Academic Press, New York,pp. 8±79.

Chafee, M.V. & Goldman-Rakic, P.S. (2000) Inactivation of parietal andprefrontal cortex reveals interdependence of neural activity during memory-guided saccades. J. Neurophysiol., 83, 1550±1566.

Collette, F. & Van der Linden, M. (2002) Brain imaging of the centralexecutive component of working memory. Neurosci. Biobehav. Rev., 26,105±125.

Coull, J.T., Frith, C.D., Frackowiak, R.S. & Grasby, P.M. (1996) A fronto-parietal network for rapid visual information processing: a PET study ofsustained attention and working memory. Neuropsychologia, 34, 1085±1095.

Deiber, M.P., Ibanez, V., Sadato, N. & Hallett, M. (1996) Cerebral structuresparticipating in motor preparation in humans: a positron emissiontomography study. J. Neurophysiol., 75, 233±247.

GoÈbel, S., Walsh, V. & Rushworth, M.F. (2001) The mental number line andthe human angular gyrus. Neuroimage, 14, 1278±1289.

Homan, R.W., Herman, J. & Purdy, P. (1987) Cerebral location of

rTMS interference effects on working memory 1631


international 10±20 system electrode placement. Electroencephalogr. Clin.Neurophysiol., 66, 376±382.

Jasper, H.H. (1957) Report of the committee on methods of clinicalexamination in electroencephalography. Electroencephalogr. Clin.Neurophysiol., 10, 370±375.

Miyake, A. & Shah, P., (eds) (1999) Models of Working Memory: Mechanismsof Active Maintenance and Executive Control. Cambridge University Press,Cambridge, MA

Mottaghy, F.M., Gangitano, M., Krause, B.J. & Pascual-Leone, A. (2002)Chronometry of TMS interference on verbal working memory. Neuroimage,16 (Suppl. 1), A10364. [Abstract].

Mottaghy, F.M., Krause, B.J., Kemna, L.J., ToÈpper, R., Tellmann, L., Beu, M.,Pascual-Leone, A. & MuÈller-GaÈrtner, H.W. (2000) Modulation of theneuronal circuitry subserving working memory in healthy human subjectsby repetitive transcranial magnetic stimulation. Neurosci. Lett., 280, 167±170.

Oliveri, M., Rossini, P.M., Pasqualetti, P., Traversa, R., Cicinelli, P., Palmieri,M.G., Tomaiuolo, F. & Caltagirone, C. (1999) Interhemisphericasymmetries in the perception of unimanual and bimanual cutaneousstimuli. A study using transcranial magnetic stimulation. Brain, 122, 1721±1729.

Pascual-Leone, A., Gomez-Tortosa, E., Grafman, J., Always, D., Nichellim, P.

& Hallett, M. (1994) Induction of visual extinction by rapid-ratetranscranial magnetic stimulation of parietal lobe. Neurology, 44,494±498.

Pascual-Leone, A., Walsh, V. & Rothwell, J. (2000) Transcranial magneticstimulation in cognitive neuroscience ± virtual lesion, chronometry, andfunctional connectivity. Curr. Opin. Neurobiol., 10, 232±237.

Paulesu, E., Frith, C.D. & Frackowiak, R.S. (1993) The neural correlates of theverbal component of working memory. Nature, 362, 342±345.

Rothwell, J.C., Hallett, M., Berardelli, A., Eisen, A., Rossini, P. & Paulus, W.(1999) Magnetic Stimulation: motor evoked potentials. Electroencephalogr.Clin. Neurophysiol.Supplement, 52, 97±103.

Rushworth, M.F., Ellison, A. & Walsh, V. (2001) Complementary localizationand lateralization of orienting and motor attention. Nature Neurosci., 4,656±661.

Shallice, T. (1982) Speci®c impairments of planning. Philos. Trans. R. Soc.Lond. B. Biol. Sci., 298, 199±209.

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

Wassermann, E.M. (1998) Risk and safety of repetitive transcranial magneticstimulation: report and suggested guidelines from the InternationalWorkshop on the Safety of Repetitive Transcranial Magnetic Stimulation,June 5±7, 1996. Electroencephalogr. Clin. Neurophysiol., 108, 1±16.



bilateral parieto-frontal network for verbal working memory: an interference approach using...

Documents