New insights into the Stroop effect

COGNITIVE NEUROSCIENCE NEUROREPORT

0959-4965 & Lippincott Williams & Wilkins Vol 11 No 9 26 June 2000 1849

New insights into the Stroop effect: a spatio-temporal analysis of electric brain activity

Asaid Khateb,1,2,CA Christoph M. Michel,2,3 Alan J. Pegna,1,2 Theodor Landis4 and Jean-Marie Annoni1

1Neuropsychology Unit, 2Functional Brain Mapping Laboratory, and 4Department of Neurology, University Hospital, 24 rueMicheli-du-Crest, CH-1211 Geneva 14; 3Plurifaculty Program of Cognitive Neuroscience, University of Geneva, Geneva,

Switzerland

CACorresponding Author

Received 7 March 2000; accepted 31 March 2000

Recent clinical and imaging studies suggest the involvement ofanterior brain regions in the Stroop effect without providingconsensus on the hemisphere being involved. Here, weinvestigated the dynamics of brain activation during a modi®edStroop task using behavioural, event-related potential mapseries, and source localization analysis. Behavioural analysisshowed an increased RT in the interference (IC) as comparedto the neutral (NC) and congruence conditions (CC). Map

series analysis in these conditions displayed a similar sequenceof 10 stable segments. From these, only segment S6, occurringat �300 ms and displaying a dominant right anterior activation,was of increased duration in IC. Furthermore, in IC only, RTwas shown to correlate with S6 duration. These results arediscussed in terms of increased duration of an attentionalprocess needed to solve the con¯ict. NeuroReport 11:1849±1855 & 2000 Lippincott Williams & Wilkins.

Key words: Brain mapping; Event-related potentials; Selective attention; Source localization; Stroop task; Temporal segmentation

INTRODUCTIONThe Stroop colour-naming task is a classic paradigm [1]used in psychology to test concepts such as interferenceand automaticity. In this task, subjects who are asked toreport the colour in which a word is displayed, arein¯uenced by word meaning even though it is irrelevant tothe task. Reaction times and error rate increase in theincongruent condition or the so-called interference condi-tion as, for example, when the word `red' is displayed inthe colour blue as opposed to other congruent (the word`blue' displayed in blue) or neutral stimuli (series of Xs orinfrequent words such as `helot' displayed in blue, see [2±5]). Numerous variations of the Stroop task were used andcon®rmed the robustness of the interference effect [4,6±8].

Both clinical and functional imaging studies stronglysuggest the involvement of frontal brain regions, in parti-cular the anterior cingulate cortex in the Stroop task,supporting the importance of this area in cognitive inter-ference, attention, and response selection [8]. Lesions of leftfrontal areas have been showed to produce large inter-ference effect in the Stroop task [9]. However, recent stud-ies report that patients with right hemisphere (RH) lesionsare more sensitive to the interference condition than pa-tients with left hemisphere (LH) lesions [10,11]. Further-more, while some functional imaging studies report thespeci®c activation of the left inferior frontal gyrus [12] inthe interference condition, other studies demonstrate theinvolvement of mainly the right anterior cingulate cortex[3,13]. Since the robust effect of interference is an increasein reaction time, methods that allow to correlate the dura-

tion of activity in certain brain regions with this increasedtime to solve the task might give more direct answers onthe speci®city of the involved cortical areas.

Event-related brain potential (ERP) recordings combinedwith recent methods for source localization in the brainoffer this possibility by directly assessing brain activityrelated to the mental chronometry. In previous ERP studiesusing the Stroop tasks, it has been shown that the inter-ference condition modulated brain activity mainly in thetime window between �300 and 600 ms post-stimulus[4,5,7,14]. However, in all of these studies, using either fewchannels [4,7,14] or multi-channels recordings [5], noattempt was undertaken to determine the neural basis ofthe Stroop effect. Our aim in this study was to investigatethe dynamics of brain activation during a modi®ed (rela-tively simple) Stroop task using ERP map series and sourcelocalisation analysis. Based on previous studies using thisanalysis method [15±18], we expected to ®nd time periodsof brain electric ®elds that are speci®c for the interferencecondition and that source localisation procedures wouldindicate the brain regions responsible for this activity.

MATERIALS AND METHODSSubjects: Ten right-handed (mean � s.d. laterality quoti-ent of �80� 27) native French speaking subjects (®vefemales and ®ve males, mean age 22.25 years) participatedin this experiment. All had normal or corrected-to-normalvision. No history of neurological diseases was noted inany individual and all were medication free. Before theexperiment, all subjects gave their written consent.

Stimuli and procedure: The set of stimulus pairs wascomprised of three conditions that required a `yes' re-sponse and one condition requiring a `no' response. In eachtrial, a reference stimulus (coloured rectangle) and acoloured test stimulus were sequentially presented on acomputer screen each for 150 ms (ISI� 300 ms). The coloursused were red, green, blue, yellow and black. Test stimuliwere either strings of ®ve Xs or the French names of thecolours (respectively: rouge, vert, bleu, jaune and noir)written in lower-case letters. In the neutral condition (NC),coloured rectangles were followed by Xs presented in thesame colour as the reference rectangle (e.g. blue rectanglewith blue `XXXXX', 40 trials). In the interference condition(IC), coloured rectangles were followed by the names ofother colours but written in the same colour as therectangles (e.g. blue rectangle followed by the word `red'written in blue, 40 trials). In the congruence condition(CC), coloured rectangles were followed by the name ofthe colour used for the rectangle and written in the samecolour (e.g. blue rectangle with the word `blue' written inblue, 40 trials). Finally, in the `no' condition, colouredrectangles were followed by the name of the displayedcolour but written in a different colour (e.g. blue rectanglewith the word `blue' written in red, 80 trials). This lastcondition was introduced to keep subjects' attention butwas not analysed further. Altogether, these conditionsprovided a set of 200 randomized stimulus pairs of which120/200 required `yes' responses.

Experiments were carried out in an isolated, electricallyshielded room. The stimuli were presented by a PowerMacintosh computer (17" screen, refresh rate 67 Hz) usingcommercial software [15]. From a central ®xation crosslocated vertically at 1.58 and horizontally at 08, the stimuliwere presented centrally with their borders extending upto 1.58 laterally. The sequence (�3 s) of events was asfollows: ®xation cross for 600 ms, reference stimulus for150 ms, inter-stimulus interval for 300 ms, test stimulus for150 ms, and 1500 ms for response. Subjects (seated 120 cmfrom the screen) had to decide as quickly and accurately aspossible using one of two buttons (right hand middle®nger for `yes', the index for `no') whether the test stimuluswas displayed in the same colour as the reference stimulus.Before the experimental block, subjects underwent a train-ing session consisting of 14 randomized trials in order toensure a perfect comprehension of the task demands.

EEG recordings and analysis: The EEG was recorded(using a 64-channel system [15]) from 47 equidistantelectrodes placed manually over the scalp according to theextended 10/10 system (see electrode positions in Fig. 1).The signals recorded against the Cz site at 500Hz were re-computed off-line against the average reference [19]. Filtersettings were between 0.15 and 250 Hz and the impedanceswere kept below 5 kÙ. A bipolar EOG monitored eye-movement artifacts. The exact 3-dimensional positions ofelectrodes were recorded for each subject using a 3-Ddigitizer (ISOTRACK, Polhemus Inc., Clochester, VT). ERPswere averaged separately for the three `yes' conditions(NC, IC and CC) from the onset of the test stimulus to600 ms post-stimulus. After eliminating sweeps exceeding�100 ìV in any of the channels and all eye-movementartifacts by visual inspection, �33/40 artifact-free trials

were obtained in each condition (F(2,18)� 1.07; p , 0.4;34� 5 in the NC; 32� 5 in the IC and 34� 4 in the CC).Before computing grand-mean ERPs, the individual datawere standardized by interpolating the subjects' electrodepositions to a general 47-channel electrode array using thenearest neighbour method.

Grand-mean ERP map series were then temporallysegmented into periods of quasi-stable electric ®eld con®g-urations, each hypothesized to represent a certain func-tional microstate of the brain [15±17,19]. A spatialclustering procedure [20] was used to determine theoptimal number of template maps that represent thedominant ®eld con®gurations in the ERPs and their time ofoccurrence [16,17]. Once the template maps were de®ned,their speci®city for a given condition was tested. For that, aspatial correlation coef®cient [15] was calculated betweenthe template maps and each subject's ERP map series. Eachtime point in the subject's map series was then labelledwith the template map with which it is most highlycorrelated [15,17]. This procedure results in the followingtwo parameters for each individual ERP map series: (1) theduration for which each template map appeared in theindividual data and (2) how well each template mapexplained the data (referred to as global percentage ofexplained variance) [15]. Since the different template mapsare not randomly distributed in the ERP but cluster eachwithin certain time period, this time period and its dura-tion can be determined and used for statistical testing. Wehave shown that systematic variations of these parameterswith experimental conditions and behavioural responsescan be found, allowing to determine statistically those mapcon®gurations that are relevant for a certain task and theirtime of appearance [15±18].

Source localisation: To identify brain regions activatedduring the time periods of interest, we used LORETA [21],a source localization procedure that estimates the smooth-est current density (CD) distribution in the whole brainwithout pre-determining the number, orientation or loca-tion of the sources [15±17]. The CD distribution wasestimated for the ERP of each individual subject andcondition at the time point that was best explained by thegrand-mean template map within a given time segment.For statistical analysis, mean CD in four regions of interest(ROIs, anterior, anterior-temporal, posterior-temporal andposterior) in the two hemispheres (LH vs RH) served aswithin-subject factors.

Behavioural analysis: Percentage of errors, mean reactiontime (RT) and RT standard deviation for each subject werecalculated in each condition. These parameters were statis-tically compared using 1 3 3 analysis of variance.

RESULTSBehavioural data: Error analysis showed a trend towardsa higher rate in IC (3.75%) than in NC (1.75%) and CC (2%)but the ANOVA did not show signi®cant difference be-tween conditions (F(2,18)� 1.24; p , 0.3124). The globalANOVA performed on RT showed a nearly signi®canteffect (F(2,18)� 3.34; p , 0.058, see Fig. 2a). Post-hoc LSDtests showed signi®cantly longer RT for IC (482.3 ms) thanNC (463.9 ms, ps, 0.028) and CC (466.4 ms, p , 0.054).

NEUROREPORT A. KHATEB ET AL.

1850 Vol 11 No 9 26 June 2000

(a) AF7 AFz AF8

F5 F3 F1 Fz F2 F4 F6

FT7 FC5 FC3 FC1 Fz FC2 FC4 FC6 FT8

T3 C5 C3 C1 Cz C2 C4 C6 T4

TP7 CP5 CP3 CP1 CPz CP2 CP4 CP6 TP8

P5 P3 Pz P4 P6

PO7 PO3 POz PO4 PO8

010

020

030

040

050

060

0

Time (ms)

Neutral

Interference

Congruence

840

2428

Voltage (µV)

(b)

Segment maps found in the grand mean ERPs

Segments

Potential

maps

S1 S2 S3 S4 S5 S6 S7 S8 S9 S10

Fig. 1. (a) Grand-mean ERP waveforms for the three experimental conditions from the onset of test stimuli to 600 ms. Vertical scale is equal to� 8 mV. Black thick lines are ERPs for neutral, thin lines for interference and dashed lines for congruence conditions. (b) Scalp equipotential templatemaps (labelled as S1 to S10) for the 10 sequential segments of quasi-stable map con®gurations revealed by a cluster analysis of the ERP map series of thethree conditions. The maps are shown from above with left ear left. Black values are negative, white positive potentials. All maps have been recalculatedagainst the average reference.

DYNAMICS OF BRAIN ACTIVATION IN A STROOP TASK NEUROREPORT

Vol 11 No 9 26 June 2000 1851

Considering that the interference effect may be morere¯ected in response variations than in mean RT, wecompared the individual standard deviations of reactiontimes in each condition. This showed a highly signi®canteffect (F(2,18)� 6.35; p , 0.008) due to the fact that thevariance in subjects' responses was higher in IC (meanSD� 126 ms) than in NC (93.7 ms, p , 0.002) and in CC(106.2 ms, p , 0.044).

Grand-mean ERP and map series temporal segmenta-tion: The grand-mean average-reference ERP waveshapesfrom 47 channels, separately averaged for NC, IC and CCare superimposed in Fig. 1a. Although the visual inspec-tion of these waveshapes may indicate the presence ofsome differences in amplitude and latency of certain com-ponents (see frontal electrodes), it does not allow toconclude about differences in map con®guration which ismandatory if clues about changes of active brain areas areof interest. Applying the temporal segmentation procedureto the grand-mean ERP map series to determine periods ofmap changes revealed a similar sequence of 10 segments inall conditions. Figure 1b illustrates the average-referencetemplate maps representing the dominant ®eld con®gura-tion for each of the 10 time segments (labelled S1±S10)found in the three conditions.

Segment analysis in individual subjects ERPs: The nextstep of analysis sought to characterize those segmentswhich may speci®cally be related to the prolongation ofreaction time in the IC. Only segments S2±S10, startingwith the P100 visual component (S2, with the potential

distribution showing a posterior positivity and anteriornegativity) were considered for further analysis. The tem-plate maps were compared with the individual ERP mapseries [15,17] in the time window 80±600 ms post-stimulus.The mean duration and global percentage of explainedvariance for each segment per condition are reported inTable 1 and Table 2. The 1 3 3 ANOVAs computed on eachsegments' duration revealed a signi®cant effect for segmentS6 (F(2,18)� 4.24; p , 0.0310). This effect was due to aprolongation of the duration of S6 in IC as compared toNC and CC (see Table 1 and Fig. 2b). This result wasfurther supported by the ANOVAs performed on globalpercentage of explained variance which again showed onlyone signi®cant effect that concerned S6 (F(2,18)� 3.93;p , 0.0383): S6 explained better the data in IC than in NC

Fig. 2. (a) Mean reaction time (� s.d. and s.e., in ms) in the threeexperimental conditions. (b) Mean duration (� s.d. and s.e., in ms) of thesegment S6 in the same conditions. Note that both mean and variance ofRT and S6 duration increased in IC compared with NC and CC. (c)Correlation between S6 duration and individual RT in IC (r and p valuesare given on top of this ®gure together with RT equation ®t).

580

540

500

460

420

380

(a)RT

(ms)

NC IC CC

6s.d.6s.e.Mean

110

90

70

50

30

10

(b)

S6 D

urat

ion

(ms)

NC IC CC

6s.d.6s.e.Mean

700

640

580

520

460

400

340

(c)

RT in

IC (m

s)

RT 5 377.4 1 1.6 * S6Correlation: r 5 0.69, p , 0.026

0 20 40 60 80 100 120 140

S6 Duration in IC (ms)

Regression95% confid.

Table 1. Mean (� s.d.) duration in ms for segments S2±S10 in theneutral (NC), interference (IC) and congruence (CC) conditions.

NC IC CC

S2 56� 31 65� 28 61� 22S3 108� 46 96� 40 108� 51S4 33� 27 33� 23 33� 25S5 65� 41 69� 32 67� 42S6� 45� 20 66� 38 38� 22S7 62� 33 46� 33 43� 23S8 65� 36 66� 27 84� 33S9 33� 27 19� 13 30� 18S10 53� 45 60� 51 56� 49

� p , 0.05.


1852 Vol 11 No 9 26 June 2000

and CC (see Table 2). These ®ndings strongly suggest thatthe map con®guration of S6 re¯ects the brain processinvolved in the Stroop effect, a hypothesis supported byadditional analysis: Given the fact that the duration of S6,the only segment differentiating conditions, showed apattern mimicking RT in the different conditions, wesought to determine whether a correlation exists betweenS6 duration in individual ERPs and subjects' RTs. Asillustrated in Fig. 2c, this analysis revealed indeed that,only in IC, S6 duration correlated positively with RT, thuscon®rming the relation between this segment and theinterference process.

Source localisation analysis: This analysis sought toestimate the brain regions that were activated during the

segment S6. For that, S6 best-®t maps (the individual mapsbeing the most highly correlated with segments' maps ofthe grand-mean) from each subject's ERP were determinedin the time window 280±380 ms, as proposed by the grand-mean ERP map series (mean time of occurrence359� 15 ms in NC, 336� 24 ms in IC and 344� 18 ms inCC). These maps were then subjected to LORETA [21] toestimate brain regions activated during this crucial timeperiod. For the purpose of comparison, the same analysiswas performed on S2 which corresponds to the P100 visualcomponent. Mean LORETA solutions for these two seg-ments (illustrated in Fig. 3a) show that in S2 posteriorregions were activated bilaterally while in S6 the activationwas dominant in right anterior regions. To verify thestatistical signi®cance of these localisations, CD was aver-aged in four ROIs (anterior, temporal-anterior, temporal-posterior and posterior, see schema in Fig. 3c2) in eachhemisphere over the lowest four slices where the majoractivity was observed (see schema for slice levels inFig. 3c1). An ANOVA was performed on mean CD usingconditions as a grouping factor and ROIs (4 from anteriorto posterior, A-P) and hemispheres (L vs R hemisphere) aswithin-subject factors. No main effect was observed forconditions and no interaction between this and the otherfactors was observed. Mean CD and the ANOVA resultsfor the two other factors are illustrated in Fig. 3b. For S2,the results only show a highly signi®cant main effect forROIs, con®rming the dominant activation of posteriorregions. In S6, in addition to a highly signi®cant maineffect for ROIs, a L-R hemisphere effect is observedtogether with a marginally signi®cant interaction. This wasdue to the fact the L-R hemisphere differences were ex-

Table 2. Mean (� s.d.) global percentage of explained variance forsegments S2±S10 in the neutral (NC), interference (IC) and congruence(CC) conditions.

NC IC CC

S2 7.0� 4.1 7.7� 3.4 7.6� 2.7S3 14.4� 6.2 12.6� 5.1 14.6� 6.9S4 3.7� 3.2 3.6� 2.5 3.9� 3.0S5 8.6� 5.8 8.7� 4.2 8.7� 5.9S6� 5.7� 2.7 8.3� 5.1 4.7� 3.2S7 8.5� 4.6 6.2� 4.8 5.8� 3.3S8 8.8� 5.1 9.1� 4.6 11.3� 4.6S9 3.7� 3.4 2.4� 1.7 3.6� 2.2S10 6.1� 5.2 6.9� 6.1 6.8� 6.2

� p , 0.05.

Fig. 3. (a) Linear inverse solutions estimated for the segments S2 and S6. The current density distribution (10ÿ3 ìA/mm2) was estimated usingLORETA, a general linear inverse solution. The inverse solution depicts the activity in eight slices with the deepest slice corresponding to the T3-T4-FPzplane in the left and the upper slice in the right (see thick lines in C1 for slice levels, maps are seen from top with left ear left). Black values indicate highdensity. The solutions are individually scaled to their maximum. Note the dominant activity in the lowest four slices in the posterior regions during S2and the dominant activity in right anterior regions during S6. (b) Graphics and the 4 3 2 ANOVA results on mean CD distribution estimated from best-®t maps over subjects as a function of four ROIs (A, TA, TP, P) and hemispheres (LH-RH). For the statistics, A-P indicates anterior to posterior factor;L-R indicates left to right factor and INT indicates the interaction between the two factors. In the graphics, solid lines for RH and dashed lines for LH.Notice that the major activity was found in S2 in the posterior regions and in S6 in the right anterior regions (see text). (c) The scheme in (c1)illustrates the height of each brain slice (in the Z axis, the lowest slice� 0, the highest� 1, see horizontal lines) where CD distribution was estimated.(c2) Represents one of the four slices with ROIs where A for anterior; TA for temporal anterior; TP for temporal posterior; and P for posterior.

(a)

S2

S6

Max.

Min.

54321

Mea

n C

D

(b)

A TA TP PROI

LHRH

S2 54321

Mea

n C

D

A TA TP PROIs

LHRH

S6

A-P: F(3,87) 5 36.2; p , 0.0000L-R: nsInt.: ns

A-P: F(3,87) 5 12.9; p , 0.0000L-R: F(1,29) 5 13.2; p , 0.001Int.: F(3,87) 5 2.5; p , 0.07

(c)1 2

Z 5 1.00Vertex

L RZ 5 0.00

A

TA

TP

P


Vol 11 No 9 26 June 2000 1853

pressed more at the anterior than at the posterior levelcon®rming the fact that the major activity during S6 waslocalized in right anterior regions.

DISCUSSIONThe aim of this study was to characterize brain dynamicsin the Stroop task using the combination of behaviouraland electrophysiological measures. Like other studies[5,7,8], we used a modi®ed Stroop task in which subjectsjudged stimuli as being displayed in the same colour ornot, instead of naming colours [3,4]. Behavioural measuresand ERPs were only compared between conditions (NC, ICand CC) requiring comparable output (`yes' responses).Our results showed an interference effect in that RT in-creased in IC compared with NC and CC. This ®ndingcon®rms that this modi®ed Stroop task allows measure-ment of the interference effect [1]. In some previous stud-ies, a congruence effect has been described [2,4]. In ourtask, a congruence effect was not expected due to the factthat neutral stimuli were simply coloured Xs and theresults indeed con®rm that NC and CC produced compar-able RT and errors.

At the electrophysiological level, the temporal segmenta-tion of ERP map series, during the period between percep-tion and response (0±600 ms), revealed 10 segments ofquasi-stable map con®gurations, similar in their potentialcon®gurations and temporal succession between the threeconditions. This similarity strongly suggests that the neuro-cognitive networks used to solve the task demands did notdiffer between conditions. To determine the electrophysio-logical correlates of the Stroop effect, segments wereanalysed in each subject's ERP. This did not show acondition effect for segments S2±S5 (up to �300 ms)neither in terms of duration nor in terms of explainedvariance. The fact that this period has been involved instimulus encoding (particularly S2 and S3 correspondingrespectively to the P100 and N150 visual components [22±24]) and stimulus evaluation including lexical analysis[25,26] suggest that the Stroop effect occurs at latter stagesof information processing.

In fact, we observed a condition effect for the segmentS6, occurring between �300 and 400 ms. S6 duration in-creased and correlated positively with RT only in IC. Thefact that this segment was only modulated signi®cantly inIC suggests a close relationship between this time periodand the interference process underlying the Stroop effect.Similar to our study, previous studies using visual selectiveattention tasks report that ERP traces generally show afrontal negativity (�300 ms) [27], similar to that character-izing the S6 map. More speci®cally, ERP studies usingStroop tasks and manual responses [4,14] report a P300component of increasing amplitude from the frontal toparietal sites as illustrated here by the S6 map con®gura-tion and shown on the ERP traces (Fig. 1). Although, inthese studies, P300 latency was not modulated by inter-ference, the authors suggested that the RT differencesoriginated more from response selection [4], probably inthe motor system [14], rather than from stimulus evalua-tion. In another study where subjects had to mentally namethe colour of stimuli, an N400 component was observed inthe interference condition [7]. The authors concluded thatthe automatic reading of the name of a colour would in

this case correspond to priming, interfering with access tothe target word. The fact that S6 duration was increasedonly in IC without differentiating NC and CC arguesagainst an interference at the semantic level [2]. In fact, wereported in a previous language study a speci®c semanticsegment in the period between 300 and 400 ms [15] whoseduration correlated with RT and showed a potentialdistribution characterized by an anterior positivity and aposterior negativity. The map con®guration was thusdrastically different from the S6 map found here. As aresult, the semantic map found in the previous study wasexplained by activity in LH anterior regions and bilateralposterior areas while the present segment S6, modulatedby IC, was shown to be mainly explained by activity in RHanterior regions.

In contrast to previous studies suggesting the implica-tion of LH anterior regions in the Stroop effect [9], our®nding support recent clinical [10,11] and imaging studieswhich report the implication of the right prefrontal, parti-cularly right anterior cingulate cortex [3,8,13] in the resolu-tion of the interference con¯ict. The spatial and temporalcharacteristics of S6 strongly suggest that the Stroop effectoriginates from increased demands on RH attentionalresources [3] necessary for response selection rather thanfrom interference on the perceptual and semantic level.

CONCLUSIONMulti-channel event-related potential recordings combinedwith spatio-temporal analysis allow to unravel the tempor-al dynamics of large-scale neural networks involved in thedifferent steps of information processing. This study givenew insights into the Stroop effect by providing additionalelements to the question when and where the interferenceoccurs during the relatively long process of con¯ict resolu-tion. The ®ndings reported here suggest that increasedreaction time in the interference condition results morefrom the prolonged activation of a particular module with-in the complex network rather than from the activation ofnew condition-speci®c modules. In the Stroop task, thisinvolved the prolonged activation of right anterior regionsclassically proposed to play a major role in attentionprocessing.

REFERENCES1. Stroop JR. J Exp Psychol 18, 643±661 (1935).

2. MacLeod CM. Psychol Bull 109, 163±203 (1991).

3. Bench CJ, Frith CD, Grasby PM et al. Neuropsychologia 31, 907±922

(1993).

4. Ilan AB and Polich J. Clin Neurophysiol 110, 367±373 (1999).

5. West R and Alain C. Cogn Brain Res 8, 157±164 (1999).

6. Sugg MJ and McDonald JE. J Exp Psychol Hum Percept Perform 20, 647±

675 (1994).

7. Rebai M, Bernard C and Lannou J. Int J Neurosci 91, 85±94 (1997).

8. Bush G, Whalen PJ, Rosen BR et al. Hum Brain Mapp 6, 270±282 (1998).

9. Perret E. Neuropsychologia 12, 323±330 (1974).

10. Kingma A, La Heij W, Fasotti L et al. Neuropsychologia 34, 273±281

(1996).

11. Vendrell P, Junque C, Pujol J et al. Neuropsychologia 33, 341±352

(1995).

12. Taylor SF, Kornblum S, Lauber EJ et al. Neuroimage 6, 81±92 (1997).

13. Pardo JV, Pardo PJ, Janer KW et al. Proc Natl Acad Sci USA 87, 256±259

(1990).

14. Grapperon J, Vidal F and Leni P. Neurophysiol Clin 28, 207±220 (1998).

15. Khateb A, Annoni JM, Landis T et al. Int J Psychopysiol 32, 215±231

(1999).


1854 Vol 11 No 9 26 June 2000

16. Michel CM, Seeck M and Landis T. News Physiol Sci 14, 206±214 (1999).

17. Pegna AJ, Khateb A, Spinelli L et al. Hum Brain Map 5, 410±421 (1997).

18. Thut G, Hauert CA, Morand S et al. Exp Brain Res 128, 256±261 (1999).

19. Lehmann D. Principles of spatial analysis. In: Gevins AS and Remond

A, eds. Handbook of Electroencephalography and Clinical Neurophysiology.

Vol 1: Methods of Analysis of Brain Electrical and Magnetic Signals.

Amsterdam: Elsevier; 1987, pp. 309±354.

20. Pascual-Marqui RD, Michel CM and Lehmann D. IEEE Trans Biomed

Eng 7, 658±665 (1995).

21. Pascual-Marqui RD, Michel CM and Lehmann D. Int J Psychopysiol 18,

49±65 (1994).

22. Onofrj M, Bazzano S, Malatesta G et al. EEG Clin Neurophysiol 80, 167±

180 (1991).

23. Skrandies W. Cogn Brain Res 6, 173±183 (1998).

24. Wijers AA, Mulder G, Okita T et al. Psychophysiology 26, 529±547 (1989).

25. Rugg MD. Brain Lang 19, 142±152 (1983).

26. Koenig T and Lehmann D. Brain Lang 53, 169±182 (1996).

27. Karayanidis F and Michie PT. EEG Clin Neurophysiol 99, 38±56 (1996).

Acknowledgements: This research was supported by the Swiss National Science Foundation grant No. 31-52923-97 to J.M.A.We thank Denis Brunet for providing the computer programs used in this study.


Vol 11 No 9 26 June 2000 1855

Documents

New insights into the Stroop effect