451BASILE Biol Res 40, 2007, 451-470Biol Res 40: 451-470, 2007 BRComplex, multifocal, individual-specificattention-related cortical functional circuits

LUIS F. H. BASILEa, b

a Laboratory of Psychophysiology, Faculdade de Psicologia e Fonoaudiologia, UMESP Division of FunctionalNeurosurgery, Institute of Psychiatryb University of São Paulo Medical School, Brazil

ABSTRACT

Recent studies focusing on the analysis of individual patterns of non-sensory-motor CNS activity maysignificantly alter our view of CNS functional mapping. We have recently provided evidence for highlyvariable attention-related Slow Potential (SP) generating cortical areas across individuals (Basile et al., 2003,2006). In this work, we present new evidence, searching for other physiological indexes of attention by a newuse of a well established method, for individual-specific sets of cortical areas active during expectingattention.We applied latency corrected peak averaging to oscillatory bursts, from 124-channel EEG recordings, andmodeled their generators by current density reconstruction. We first computed event-related total power, andaveraging was based on individual patterns of narrow task-induced band-power. This method is sensitive toactivity out of synchrony with stimuli, and may detect task-related changes missed by regular Event-RelatedPotential (ERP) averaging. We additionally analyzed overall inter-electrode phase-coherence.The main results were (1) the detection of two bands of attention-induced beta range oscillations (around 25and 21 Hz), whose scalp topography and current density cortical distribution were complex multi-focal, andhighly variable across subjects, including prefrontal and posterior cortical areas. Most important, however,was the observation that (2) the generators of task-induced oscillations are largely the same individual-specific sets of cortical areas active during the resting, baseline state. We concluded that attention-relatedelectrical cortical activity is highly individual-specific (significantly different from sensory-related visualevoked potentials or delta and theta induced band-power), and to a great extent already established duringmere wakefulness.We discuss the critical implications of those results, in combination with other studies presenting individualdata, to functional mapping: the need to abandon group averaging of task-related cortical activity and torevise studies on group averaged data, since the assumption of universal function to each cortical area appearsdeeply challenged. Clinical implications regard the interpretation of focal lesion consequences, functionalreorganization, and neurosurgical planning.

Key terms: Attention, cortical electrical activity, high-resolution electroencephalography, slow potentials,source localization, functional mapping.

Address Correspondence to: L.F.H. Basile, M.D., PhD., Division of Functional Neurosurgery, Institute of Psychiatry,University of São Paulo Medical School, Av. Dr. Ovidio Pires de Campos 785, P.O.Box 3671, São Paulo, SP, 05403-010,BRAZIL, Phones: 55-11-30697284, 55-11-32846821 FAX: 55-11-2894815, e-mail: lbasile@usp.br

Received: October 24, 2007. Accepted: March 3, 2008

INTRODUCTION

Studies that aim at localizing thephysiological changes that follow passivesensory stimulation or simple movementexecution in cortical areas of individualhuman subjects, generally lead touncontroversial, expected results. Thus, forinstance, Electroencephalography andMagnetoencephalography (EEG and MEG)

in such cases of highly focal sensoryevoked potentials and fields can distinguisheven the sources corresponding tostimulation of separate sections of thevisual field or to tactile stimulation ofdifferent fingers. This precision may forinstance be used for protection of sensory-motor areas during surgical interventions,when edema prevents visualization of thecentral sulcus (Wheless et al., 2004).

BASILE Biol Res 40, 2007, 451-470452

Metabolic tracing studies, particularlyfMRI, also allow the analysis of singlesubject sensory-motor data, and suchparadigms are commonly used for testingand validation of the programs of analysisthemselves. As opposed to this situation, avery different panorama arises from studiesthat directly interest Psychophysiology orPsychiatry, that is, those using paradigmsincluding experimental tasks that involveany psychological variable beyondsensation and simple movement, fromsimple behavioral decisions, comparison ofstimuli, expecting attention, memory oremotion. In spite of the great efforts tofunctionally map non-primary cortical areas(whose activity supposedly shouldcorrespond to various ‘cognitive functions’)and the resulting vast literature of the lastdecades, one can notice a clear lack ofconsensus across authors or laboratories.When one starts from the functional end,any unrestricted literature search regardingtask-related cortical activity to any givenpsychological variable (e.g., memoryevocation) will include changes, acrossstudies, in most if not all cortical non-primary areas. The same is true if oneperforms the literature search restricted to agiven architectonic area, which will resultin studies claiming a number of functionsimpossible to reconcile with each other.

Almost all studies on task-related corticalactivity present data collapsed acrosssubjects (group or grand averages), projectedover an abstract, ‘average brain’. Theirobvious limitations include the anatomicalindividual variability in the structure of and,more critically, in the distribution ofcytoarchitectonic areas onto gyri and sulci(Uylings et al., 2005). Recently, however,metabolic tracing studies emphasizing andpresenting individual physiological andanatomical data are emerging, all of whichdemonstrate a large degree of interindividualvariability in the encephalic distribution oftask-related activity (e.g., Cohen et al., 1996;Herholz et al., 1996; Fink et al., 1997; Daviset al., 1998; Brannen et al., 2001; Tzourio-Mazoyer et al., 2002), and in some cases,even to passive stimulation (Davis et al.,1998; Hudson, 2000). We do not refer hereto any form of interindividual variability

(such as exact coordinates of a center, orextension of physiological changes in onearea; e.g. Vandenbroucke et al., 2004), buton the actual set of cortical areas activeduring a given task. Our own recent studies,following the conventional attempt touncover universal areas across subjects,active during expecting attention to specificperceptual domains, have encountered thismajor shortcoming: First, by modeling thegenerators of a Slow Potential (SP; a class ofevent-related potentials, ERPs, whichinclude the contingent negative variation,CNV) occurring in anticipation to task-performance feedback stimuli, and expectingto find some pattern of prefrontal activity,we found a multifocal distribution ofgenerators, both prefrontal and posteriorcortical, although including two commonprefrontal areas. But the actual set of areaswas highly variable across subjects (Basileet al., 2002). Second, by comparing the SPgenerators corresponding to verbal, pictorialand spatial visual selective attention, such acomplex and individually variable set ofactive areas was also found, and precludedany conclusion based on visual inspection:we developed an activity scoring method toallow statistical analysis but simultaneouslypreserving individual anatomical andphysiological information (Basile et al.,2003). Since the task then used wascomplex, and included memorization andcomparison of stimuli, the commonlyclaimed variability in task executionstrategies could not be ruled out as a sourceof interindividual variability. We thusdeveloped a task to allow the recording ofSPs during (temporal) attention to simplevisual stimuli, by simplifying the paradigmof spatial attention cueing of Posner (1980;Posner et al., 1980). However, the same highdegree of interindividual variability incortical activity distribution was once moreobserved (Basile et al., 2006).

Taken together, those studies on task-related cortical activity that preserveindividual data, indicate that interindividualvariability appears to be an inherent aspectof the normal functioning of the centralnervous system. An interesting theoreticalconcept of biological “degeneracy”(Edelman and Gally, 2001), when applied

453BASILE Biol Res 40, 2007, 451-470

to task performance and recovery offunction (Noppeney et al., 2004), explicitlyaccounts for the inexistence of a simple,one-to-one isomorphic mapping betweenfunction and structure. We believe that theinterindividual variability in sets of corticalareas engaged in any given task can accountfor most of the lack of consensus in corticalpsychophysiology, since it is mainly basedon group data averaging, which mayemphasize areas that are strongly active infew subjects, and deemphasize areas whichcould be common to many subjects, butweakly active.

In this work, we present results from thesearch of some new electrophysiologicalcorrelate of expecting attention. Weinvestigated more thoroughly the samesimple task and subjects for whom SPswere analyzed (Basile et al., 2006). Mostresearch on the electrophysiology of humanattention has centered on the endogenouspotentials of the P300 class, and on themodulation of sensory evoked potentials.However, such event-related potentials(ERPs) and their modulations may best bedescribed as correlates of stimulusdetection, and are only enhanced byexpecting attention as an antecedentcondition. So far, SPs are rigorously thedirect correlates of attention. ERPs sufferfrom the limitation of time-locking tostimuli, and if one allows, in principle, forthe occurrence of oscillatory activity burstswith no fixed relation to task events, theywould thus be canceled out in regularERPs. It is exactly this limitation of ERPsthat has brought an increasing interest instudying electrical power measures that arenot synchronized to stimuli, variouslynamed as induced band-power (IBP), orEvent-Related De- or Synchronization(ERD, ERS) for decreases and increases,respectively (Klimesh et al. , 2000;Pfurtscheller, 2001). We here used asystematic IBP analysis to guide theaveraging of oscillatory activity in narrowfrequency bands, by a new application of awell known extension of stimulus-lockedaveraging: peak or corrected latencyaveraging. We first planned to computetask-induced EEG power (which lacksprecise time and phase-locking to stimuli),

and if some oscillatory activity wereobserved to increase during the pre-S2 timewindow, where SPs are maximal inamplitude, we would compute correctedlatency averages (centered on peakamplitude of bursts instead of stimulusonset). Then, we intended to focus on thetopographic and cortical generator analysisby current density reconstructionconstrained by individual MRI data, such aswe established for SP analysis, but stillsearching for a correlate of attention whichwould originate from a simpler, universalset of cortical areas across subjects. Inparticular among all frequency bands, thereis a return of interest in theta (3-7 Hz)oscillations as a putative correlate ofexpecting attention. Theta rhythm may beobserved in the ongoing EEG of waking,healthy individuals during tasks such asovert calculation with pen and paper(Mizuki et al., 1980), but in a smallproportion of subjects (Takahashi et al.,1997). Although it has also been reported incue-target (S1-S2) paradigm similar to theones used for recording SPs (Nakashimaand Sato, 1993), they could be alternativelyaccounted for by sustained mental effort,the uncontrolled stimulus presentation orovert movement.

METHODS

Subjects

Twelve healthy individuals with normalvision and hearing, 9 male and 3 female,participated in the study. They ranged inage between 20 and 45 years, with nohistory of drug or alcohol abuse, and nocurrent drug treatment. Eleven subjectswere current or former medical students.All subjects signed consent forms approvedby the Ethics Committee of the Universityof São Paulo Hospital.

Stimuli and Task

A commercial computer program (Stim,Neurosoft Inc.) controlled all aspects of thetask. Visual stimuli composing the cue-

BASILE Biol Res 40, 2007, 451-470454

target pairs (S1-S2) consisted in smallrectangles (eccentricity ±0.8o, S1: 100 msduration, S2: 17 ms). In half of the trials,the S2 rectangle contained a grey circle –the task target - with ±0.3o of eccentricity.S1 was followed by S2, with onsetsseparated in time by 1.6 seconds. Weinstructed the subjects that a rectanglewould be presented to indicate that 1.6seconds later it would flash again butquickly, containing or not the target circle.The subject decided whether there was atarget inside the S2 rectangle, and indicatedpresence of target by pressing the rightbutton with the right thumb or absence oftarget by pressing the left button with theleft thumb. We explicitly deemphasizedreaction time in the instructions andmeasured performance exclusively by thepercent correct trials, from the total of 96trials comprising the experiment. An eyefixation dot was continually present on thecenter of the screen, as well as a stimulus-masking background, to prevent after-images. To confirm the expecting attentionindex role of some possibly found EEGrhythm, we also used a passive stimulationcontrol condition (S2s never containedtargets), during which subjects were onlyrequired to fixate and relax.

EEG Recording and acquisition of MRIs

We used a fast Ag/AgCl electrodepositioning system consisting of an extended10-20 system, in a 124-channel montage(Quik-Cap, Neuromedical Supplies‚), and animpedance-reducing gel which eliminatedthe need for skin abrasion (Quick-Gel,Neuromedical Supplies‚). Impedancesusually remained below 3 kOhms, andchannels that did not reach those levels wereeliminated from the analysis. To know theactual scalp sampling or distribution ofelectrodes in each individual with respect tothe nervous system, we used a digitizer(Polhemus‚) to record actual electrodepositions with respect to each subject’sfiduciary points: nasion and preauricularpoints. After co-registration with individualMRIs, the recorded coordinates were usedfor realistic 3D mapping onto MRIsegmented skin models, and later used to set

up the source reconstruction equations(distances between each electrode and andeach dipole supporting point). Two bipolarchannels, out of the 124-channels in themontage were used for recording bothhorizontal (HEOG) and verticalelectrooculograms (VEOG). Left mastoidserved as reference only for data collection(common average reference was used forsource modeling) and Afz was the ground.We used four 32-channel DC amplifiers(Synamps, Neuroscan Inc.) for datacollection and the Scan 4.3 software package(Neurosoft Inc.) for initial data processing(until computation of averages). The filtersettings for acquisition were from DC to 30Hz, and the digitization rate was 250 Hz.Thus, the gamma band was not collected,due to concerns with noise from thelaboratory environment and absence ofFaraday shielding. The EEG was collectedcontinuously, and epochs for averagingspanned the interval from 700 ms before S1to 400 ms after S2 presentation. Baselinewas defined as the 400 ms preceding S1.Artifact elimination was automatic: epochscontaining signals in either HEOG or VEOGchannels above +50 or below –50 mV wereeliminated. In our montage, the VEOGdetected blinks as deflections above 130 mVin the positive direction.

MRIs were obtained by a 1.5 Tesla GEmachine, model Horizon LX. Image setsconsisted in 124 T1-weighed sagittal imagesof 256 by 256 pixels, spaced by 1.5 mm.Acquisition parameters were: standard echosequence, 3D, fast SPGE, two excitations,RT=6.6 ms, ET=1.6 ms, flip angle of 15degrees, F.O.V = 26 x 26 cm. Totalacquisition time was around 8 minutes.

Frequency-Time analysis (Task-inducedBandPower) and Power Scalp Topography

After artifact rejection, the signal from eachchannel was spectrally analyzed by means ofa Short Time Fourier Transform (STFT), toobtain frequency-time charts of the induced(stimulus related, but not stimulus-locked)and evoked (stimulus-locked) spectrum ofthe interval from 700 ms previous to S1, to400 ms after S2. To obtain the inducedpower spectrum (Tallon-Braudry; 1996), the

455BASILE Biol Res 40, 2007, 451-470

time-frequency decomposition was made foreach electrode and each trial, from DC to 30Hz, and the resulting charts were thenaveraged. The evoked power spectrum wasobtained applying the spectraldecomposition to the averaged signal.Recently, it has been demonstrated that thismethod is mathematically equivalent toothers like the Hilbert transform, or waveletdecomposition, and that each of them yieldsequivalent results in practical applications toneuronal signals (Bruns and Eckhorn, 2004).The decomposition was computed on theEEG tapered by a sliding Hamming window,256 points in size for frequencies over 5 Hz,and 512 points between 2 and 5 Hz, with atemporal resolution of N/10 (N being thenumber of temporal points of the rawsignal), and a frequency resolution of 4 binsper Hertz. Then, we normalized the averagepower for each electrode to obtain Z-scoresof increments or decrements in eachfrequency bin with respect to the power inthe same frequency during the baseline (<Pj>= (Pj - μj)/ σj; given Pj = spectral power ateach time point in electrode j), μj and σj arethe mean and standard deviation,respectively, of the average power during thebaseline at this electrode).

We computed realistic three-dimensionaltopographic maps of the scalp distribution ofnormalized power, at each frequency bandthat demonstrated task-induced changes, foreach subject, over the reconstructed scalpanatomy. To this purpose, we used acommercial sotfware (Curry V 4.6,Neurosoft Inc.), which co-registeredindividual MRI sets (skin model, see below),with the actual position of each electrodewith respect to common landmarks, andlinearly interpolated the instantaneous valuesof power to obtain continuous maps.

Computation of corrected latency burstaverages

According to the observed inducedfrequency bands and time windows wherethey occurred for each individual, theoriginal artifact-free EEG epochs (rangingfrom 700 ms before S1 through 500 ms afterS2) from each subject were filtered aroundthe bands of interest (Butterworth, 96dB

rolloff, typically 1-3 Hz for delta, 3-7 fortheta, 7-9 for alpha1, 9-12 for alpha2, around25 Hz for beta2, a wider range for beta1, andin some subjects for obtaining an additionallow beta band, between 13 and 15 Hz). Theresulting filtered epochs were then subject toan algorithm developed by us to search forthe peaks of bursts within the task-timewindows of interest. Filtered epochs werethus cut again starting from positive voltagepeaks, resulting in new epochs, ranging from400 ms before to 400 ms after the peaks. Aminimum of 60 epochs was averaged foreach individual and frequency band, usingeach channel in the search for peaks. Eachchannel at a time thus served for peakdetection, to create a multi-channel average,in which all remaining channels simplyfollowed the latency correction of theleading channel. In this way, all systematicphase relations were preserved. Then, agrand average was computed using theaverages obtained by guidance from eachchannel. Since this method would inprinciple suffer from the limitation ofconfounding any possible systematic time(direction) relations between active areas,for instance if groups of areas were active insequence in a given frequency band, we alsocomputed partial averages using groups ofguiding channels ranked for latency ofoccurrence of peaks. That is, using only thefirst one fourth of channels (those withoverall shorter peak latencies), and second,third and last fourth of channels. We alsoperformed independent analyses to studypossible time relations between groups ofelectrodes (next section). Finally, in all caseswe also computed pre-S1 burst averages(representing the baseline topography foreach frequency band), where the programsearched peaks from -400 to 0 ms before S1,for comparison with the task-induced bursts.

Inter-electrode phase-synchrony analysis

Since a systematic and complete phaseanalysis of the present data would consist ina separate and major work, we decided topresent only a first approach: we computedonly the overall pattern of phase relations, inthe form of averages across all pairs ofchannels. This suffices to answer the

BASILE Biol Res 40, 2007, 451-470456

question about whether there are task-relatedphase changes in correspondence to inducedpower. The practice of separately computingphase is becoming common in event-relatedpower studies (Fell et al., 2004; Hanslmayret al., 2006). Due to volume-conductioneffects, we selected a group of 25 regularlyinterspaced electrodes from the originalarray to compute this index. A similarprocedure used to obtain the power spectrumof the signal was used to compute the phase-locking value between electrodes (Lachauxet al., 1999). That is, a STFT of the signalfrom each electrode and trial was computedto obtain the instantaneous angular phase foreach frequency, during a time windowcentered at time t. Then, after subtracting theconstant angular phase, a complex vector ofunitary value was constructed for eachchannel, trial, frequency and time. With thisvalue, a matrix of differences of phasevalues between electrodes, in each trial, wascomputed for each frequency-time, and thenaveraged over all the trials. Using themodulus of this complex value, we obtainedfor each pair of electrodes, in each frequencyand time point, a phase-difference valuebetween 0 (random phase relation) and 1(constant phase relation). That is, Φi (f, t, k)being the phase value of electrode i, atfrequency f, time t, and trial k, and Φj (f, t,k) the phase value of electrode j, in the samefrequency, time and trial, the phase-lockingvalue was computed as

Φij (f, t) = 1/N | Σk=1N Φi - Φj |.

The phase-locking values obtained forthe time interval posterior to stimuluspresentation was then Z-normalized by thevalues obtained during the baseline intervalin the same way as time-frequency spectralmatrices. Finally, we verified whether therewere statistically significant correlationsbetween task-related power and phase, infrequency bands where we could visualizeany such systematic relations.

Intracranial source reconstruction

The computed averaged bursts, MRI sets andelectrode position digitization files were the

raw data for all further analysis (Curry V4.6, Neurosoft Inc.). A detailed descriptionof the reconstruction procedure, and adiscussion on the criteria for method choiceand shortcomings, as well as on criticalsteps, may be found in the methods sectionof our previous publications (Basile et al.,2002; 2006). Noise in the data was definedas the variance of the 20% lowest amplitudepoints in each average. For the inclusion of a‘noise component’ into the source model, thephysical unit-free or ‘standardized’ data(with retained polarity) were decomposed byIndependent Component Analysis (ICA),which searches for the highest possiblestatistical independence or redundancyreduction between components (in this case,space-time averaged data patterns), a robustmethod of blind signal decomposition/deconvolution (for a review see, e.g.Hyvarinen and Oja, 2000). ICA was appliedto each individual’s whole space-time dataset, i.e., to the m x n data matrix (m usedchannels times 201 time samplescorresponding to the 800 ms composing theaveraged bursts). Finally, we fed thereconstruction algorithm with the main ICAcomponent(s) as data to be fitted. Thus, the‘noise component’ of the model was definedas the sum of remaining components (withloadings below SNR=1), all of which addedtogether lead invariably to negligible scalppotentials when compared to the maincomponents. In practically all cases, onlytwo space-time ICA components were thenmodeled. MRI sets were linearly interpolatedto create 3-dimensional images, and semi-automatic algorithms based on pixelintensity bands served to reconstruct thevarious tissues of interest. A BoundaryElement Model (BEM) of the headcompartments was implemented, bytriangulation of collections of pointssupported by the skin, skull andcerebrospinal fluid (internal skull) surfaces.Mean triangle edge lengths for the BEMsurfaces were, respectively, 10, 9 and 7 mm.Fixed conductivities were attributed to theregions enclosed by those surfaces,respectively, 0.33, 0.0042 and 0.33 S/m.Finally, a reconstructed brain surface, withmean triangle side of 3 mm, served as themodel for dipole positions, corresponding to

457BASILE Biol Res 40, 2007, 451-470

a minimum of 20 thousand points. Theelectrode positions were projected onto theskin’s surface following the normal lines tothe skin. The detailed description of theassumptions and methods used by the “Curry4.6” software for MRI processing and sourcereconstruction may be found elsewhere(Curry 4.0 User Guide, 1999; or e.g.,Buchner et al., 1997; Fuchs et al., 1998;Fuchs et al., 1999). The analysis programthen calculated the lead field matrix thatrepresents the coefficients of the set ofequations which translate the data space(SNR values in the set of channels per timepoint) into the model space (above 20thousand dipole supporting points). Thesource reconstruction method itself was Lpnorm minimization, with p=1.2 both for dataand model terms. The regularization factor,or λ values to be used, which typicallyconverged after repeating the fitting processthree to four times (λ gives the balancebetween goodness of fit and model size).Resulting foci of current density wereinspected with respect to the individualanatomy directly, a straightforwardprocedure allowed by our software, in termsof which estimated cytoarchitectonic areasthey covered (and scored for relativeintensity; Basile et al., 2003). The Brodmannareas containing current foci were tabulatedafter verification by comparison withclassical illustrations and the conventionalTalairach and Tournoux atlases (1993,1997).

RESULTS

Task Performance

All subjects reported that performance wasrelatively easy, given that effort of attentionto the S1-S2 pairs was made. The overallaverage performance was 88.5 % correctresponses (standard deviation 8.3 %).

Temporal and topographic characteristicsof task-Induced power and Inter-electrodephase analysis

We first describe the overall findings,common to all subjects, regarding the

behavior of each induced power band withrespect to task events, and their scalptopography. Task-induced theta power wasnot present in the ISI interval, showing aclear post-stimulus increase pattern in allsubjects, practically returning to baselinelevel during the ISI. The two peakscorresponded to around 180 ms post S1 orS2, thus coinciding with the latency of theN200 evoked potential component. Thesame purely stimulus dependent behaviorwas observed for induced delta power, butin this case the peaks occurred later, around350 ms, and in almost all subjects withmuch higher amplitude after S2 (suggestingat least partial relation with the P300component, due to the latency and moretask-relevant S2). Figure 1(1) shows theoverall task-time behavior of the inducedpower, with data collapsed acrosselectrodes and subjects. Induced thetapresented an overall increase in peakamplitude with respect to baseline of 13.8%(±11.5%), but ranging from no change intwo subjects (increases around 6% in foursubjects), to 34%. Delta presented anoverall increase of 39.2% (±42.3%), butranging from reduction in two subjects (to62 and 91% from baseline), virtually nochange in one subject, to 89%. Regardingthe scalp topography of delta and theta,both bands had a clearly posteriordistribution of task-induced power maxima,typically with a double occipital peak, in allcases either or both power peak distributionsbeing coincident with the N200 peak voltagedistribution. Figure 2 shows the inducedpower and corresponding topography of allbands (at peaks of task-induced powerincreases), grand average across subjects,collapsed across the montage and projectedon median head size individual. The timepattern of induced power was virtuallyidentical to the evoked pattern for theta anddelta bands. A more critical finding, whichwe will resume in the discussion inconnection with other general results (validfor all frequency bands), was the identicaltopography of evoked or induced power.Alpha desynchronization is not clearly seenpossibly due to the collapsing acrosselectrodes for time-frequency plotcomputation.

BASILE Biol Res 40, 2007, 451-470458

Figure 1: Schematic representation of the main steps of the method: (1) Computation of task-induced band-power. In this figure, real data is presented, collapsed across channels and subjects(numbers indicate z-score – power relative to baseline). One may appreciate the overall time courseof task-induced power changes, which were fairly common across subjects (see text for fewfrequency bands where exceptions occur, low beta and alpha-1). (2) EEG Epochs were filtered inindividual-specific narrow frequency bands. (3) Based on each channel, positive voltage peaks wereautomatically detected in time windows of power increases, and (4) multi-channel latency correctedaverages computed. (5) A final multi-channel average was computed, between multi-channelaverages guided by each channel. (6) Independent Component Analysis (ICA) separated the (7)main and second space-time components to feed the (8) Current Density Reconstruction (CDR)algorithm.

459BASILE Biol Res 40, 2007, 451-470

The induced alpha range, on its turn,presented a complex multiphasic patternduring the task time, with an inflectionshowing a relative decrease in poweraround 200 ms after the stimuli (alphadesynchronizat ion) , another opposi teinflection reflecting a small increasearound 400 ms after stimuli (which is thelatency of the peak of evoked alpha anddel ta power) , and a major increasepeaking roughly around 700 ms. Inducedalpha oscillations were increased in allsubjects, ranging from 2 to 57% frombaseline (overall 30,4% ±16,3%). Thescalp topography of induced alpha poweralso showed the expected poster iordistribution of highest power. However,the actual isocontour map shapes weredif ferent f rom the s imple pat ternresembling the N200 evoked potentialcomponent that we observed for theta anddelta ranges (and also with no trivialrelation to the P300 topography, whichappear to combine mainly delta, theta and

Figure 2: Task-induced band-power of one example individual, collapsed across all channels, andcorresponding topographic distribution of the main points of change, that were common to allsubjects. Color scale: extreme of power changes (yellow and magenta) correspond to z-score equalto 9.8 standard deviations, ‘hot’ colors indicate increase and ‘cold’ decrease relative to baseline.Below, time course of stimuli and mean event-related potential global field power (bar 5 μV).

alpha frequencies; see e.g., Cacace andMcFar land, 2003) . Each individualpresented a fairly complex and specificmap shape. Since for alpha, evokedactivity occurred exactly where the leastof total induced power was observed(‘post-stimulus’ time window), it served asthe best example for comparison betweentopographies of the two methods ofcomputat ion: in this case also, thetopographic rendering of the data resulted,in all subjects, in identical evoked andinduced maps. Another interesting finding,which was also previously observed in ourlaboratory, was the clear presence ofevoked alpha throughout the ISI (peakingin the ISI with an overall 71 % of themaximum post-stimulus, evoked alpha).Given the long ISI (1.6 seconds) withrespect to alpha wavelength, it is curiousthat so many alpha cycles could besynchronized with the task events. That is,in no other frequency band do we see suchlasting synchronization.

BASILE Biol Res 40, 2007, 451-470460

We may here summarize the results ofphase analysis, given that the proposedcorrelative analysis was applied only to thebeta range, discussed below. In allfrequency bands but beta, the task-relatedphase coherence changes were complex andvariable across subjects, to various degrees,depending on the sub-band inconsideration. In all such variable cases,except for delta, a computation of phasecoherence collapsed across subjects lead tovirtually flat waveforms. For delta,however, at least one common aspect wasretained in the group average: a peak ofinter-electrode coherence increase at 450ms (of around z-score=0.6), and an equallylow amplitude decrease during the ISI,peaking at 1200 ms. To our surprise, giventhe highly regular induced-power pattern intime across subjects, theta inter-electrodecoherence was the most variable. Subjectseven presented opposite results during thepeaks of post-stimulus power, with onlythree subjects presenting parallel increasesin power and phase coherence, and threeother presenting increases only during theISI. Alpha 1 was the most variable in timepatterns of coherence changes. For alpha 2,five subjects presented overall coherencereduction during the period of increasedpower, in the center of the ISI, where threesubjects presented increases. The betarange, however, presented a relativelysimpler, more common aspect acrosssubjects, in the form of changes incoherence roughly parallel with power. Wethus proceeded to analyze intra-individualcorrelations between the two variables, andthe results supported the visual impression:beta 1 had significant non-parametriccorrelations in 11 subjects, and beta 2,highly significant correlations in 10subjects. The implications of those findingswill also be considered in connection withthe other results to allow for a morepanoramic view of oscillatory activity, inthe discussion.

There were two new and main findingsin this work. The first regarded the betarange. In all subjects, a narrow induced betaband around 25 Hz was observed,throughout the ISI, peaking during the pre-S2 time range, following a time pattern that

we originally expected to fit a putativeattention-related induced theta band. Inaddition, all subjects presented a broaderbeta band roughly around 21 Hz, and somesubjects another narrow band close to 15Hz, but more variable in frequency and timepattern. Given the task-time distribution ofbeta, mainly the 25 Hz band, and the vaguebut old association between beta activityand behavioral arousal (that impelled thewidespread but still controversial beta-enhancement by biofeedback; Ramirez etal., 2001), we immediately thought betacould fulfill our expectation of a new indexof expecting attention-induced activity.When computing the scalp distribution ofbeta power, this suspicion increased, due tothe qualitative similarity with our findingsregarding SPs: The topography of inducedbeta power changes was complex,multifocal, including frontal and temporal,as well as more posterior peaks, and highlyvariable across subjects. Finally, there wasa statistically significant increase in betapower during the ISI, when the task wascompared to the passive stimulation controlcondition: beta mean global field powerfrom 500 through 1600 ms, differedsignificantly between conditions both inparametric paired samples t-test (for beta1,p=0,004; beta 2, p=0,012), as in non-parametric tests (Wilcoxon test, p=0,005for beta1 and p=0,012 for beta2; Sign test,p=0,006 for beta1 and p=0,039 for beta2).Figure 3 shows beta mean global fieldpower collapsed across individuals, and z-score of power scatter-plot in bothconditions and bands. This alsocorroborated our idea of the role ofinduced-beta as an index of expectingattention. We thus focused on the sourcereconstruction of the task-induced betabursts, presented in the next section.

Source reconstruction

The second new and important finding ofthis work was evident after computation ofour corrected latency burst averages forsource reconstruction: Since we computedburst averages for the pre-S1 baseline aswell as for the task periods proper, inalmost all cases (subjects and frequency

461BASILE Biol Res 40, 2007, 451-470

Figure 3: (left) Overall pattern of task-induced power increases in the beta range, collapsed acrosselectrodes and individuals (upper curves), compared to passive stimulation control condition(bottom curves). (right) Individual induced beta power Z-score distribution in both conditions.

bands), the topography of baseline activitywas identical to the main component oftask-induced oscillations. That is, it wasclear that the same sources already activebefore stimuli composing the trials,presumably task-independent, were themain sources active during task execution.Moreover, very similar topography ofbaseline oscillations across frequency bandswas observed within subjects. The onlyexception was the alpha band in half of thesubjects, which had a peculiar topography,

different and prevailing over the patternsimilar across all remaining frequencybands, but that also remained the mainalpha component during the task timewindow as well. It must be emphasized herethat the topography observations wereindependent from amplitude of oscillations,whose task changes were variable acrossbands. Since those results were absolutelyunexpected, we performed a comparison,using four subjects, between the averagescomputed for the pre-S1 baseline and

BASILE Biol Res 40, 2007, 451-470462

similar averages computed for a restingcondition of a few minutes which precededthe start of the experiment (fil teredcontinuous EEG was marked in localvoltage peaks using a refractory period of800 ms, to avoid overlap between epochs,since no task events were present). Thesame topography of baseline activity wasthus observed, indicating that the pre-S1baseline indeed reflects task-unrelatedactivity. Otherwise, it would be conceivablethat baseline activity could still reflect taskengagement, due to the cyclical nature ofthe task, i.e., if trial expectation werephysiologically identical to relevant S2expectation. In one subject, we alsoreplicated the experiment after one month,and the same topography of baselineactivity was observed. It is critical forfuture studies, however, that in threesubjects who participated in previousexperiments, four and six years before, weobserved different topographies of restingcondition oscillations, thus suggesting aspontaneous ‘migration’ of the mainlyactive areas in a given subject in this timerange.

Corresponding to the complex scalpdistribution of beta induced power, thesource reconstruction results indicated thesame complexity: Multifocal corticalcurrent distribution, was highly variableacross subjects, including frontal andposterior cortical sources in all subjects. Assimple means of quantifying highervariability of beta topography, obvious tovisual inspection (figure 4), when comparedto evoked potentials or stimulus-relatedinduced power (theta was chosen forcomparison), we computed a deviationindex. This type of index has provedsuccessful in distinguishing the morecomplex and variable topography of SPs ofa group of schizophrenia patients fromhealthy subjects (Basile et al., 2004). It is aquadratic norm or Euclidian distancebetween each individual’s data set and thegroup averages collapsed on the montage(square root of electrode-by-electrodesquared difference in voltage, divided bythe number of channels): i.e., a scalarmeasure indicating individual distance fromthe norm, proportional to electrical power.

To emphasize topography and not to allowfor absolute power to influence themeasure, we first normalized beta and thetapower across individuals into a commonvalue. Results show clearly the largerdispersion of individuals from the (leastrepresentative) beta average, as comparedto the dispersion from the group thetaaverage (figure 4, where deviation (power)values are z-score normalized). Thedifference between beta and theta deviationindexes is statistically highly significant(Wilcoxon: p=0.002; Sign test: p<0.001).The partial averages computed from subsetsof electrodes as explained in methods wereidentical to each other, and indicate absenceof systematic sequential activation betweenbeta generating areas. Added to the inter-electrode phase analysis results, theysuggest a tight phase synchrony between allsuch areas. Figure 5(a) shows the currentdistribution in each subject, accounting forthe main ICA space-time data component,that is, the baseline activity component,which was enhanced (by 33.5% in averagepeak amplitude; std=17.1%; range=6 to67%) and synchronized across channelsduring the ISI, pre-S2 period. Figure 5(b)shows the second component (exclusivelyor task-induced ‘proper’, i.e., not presentduring the pre-task period), of an overallrelative intensity of 10.7% of the maincomponent in electrical power (std= 12.2%,ranging from less than 1% in one subject to36% in two subjects, but within 3 to 13% inthe remaining subjects), but in all cases ofsufficient SNR for source reconstruction(average SNR=1.8; std=0.7; range=1.05 to3.3). We may notice the individual-specificpattern of relative current distribution,especially conspicuous for the second,exclusively task-related component. Onlyparietal area 7 shows some level of activityin all subjects corresponding to the main,baseline component (although highlyvariable in intensity relative to maximumcurrent). The reconstruction results for thebeta-1 (around 21 Hz band) resulted incurrent distributions of main componentsidentical to beta-2 in all subjects. The scalptopography of the second component ofbeta-1 burst averages, on its turn, in most ofthe subjects demonstrated a partial overlap

463BASILE Biol Res 40, 2007, 451-470

Figure 4: On top, examples of individuals presenting similar topography of task-induced thetaactivity, for whom beta distribution seen from the same angle is clearly more variable acrosssubjects. Below, topographic deviation of each individual from normalized mean, between beta 2and theta bands. Deviation was defined as the quadratic norm of the electrode-by-electrodedifference between individual and group averaged data (across the montage, see text for details), assimple means to quantify the higher beta variability depicted by visual inspection.

with the second component of the 25 Hzband. Accordingly, reconstruction resultsshowed very similar patterns in most cases,typically with the beta-1 set of current foci

representing part of the set seen for beta-2.In some subjects, however, few additional(i.e., complementary to beta-2) weaksources were also observed.

BASILE Biol Res 40, 2007, 451-470464

DISCUSSION

The first conclusion from this work regardsthe lack of theta power enhancement duringthe pre-S2 time window, where we wouldexpect a direct correlate of expectingattention to have its maximum amplitude.Indeed, both theta and delta induced band-power behaved clearly as post-stimulusphenomena, with peak amplitudes at

Figure 5: Current density reconstruction results for all subjects. Current density indicated by smallred arrows, (arrow size proportional to local current density). (a) Main component, identical withmain component obtained for baseline activity. (b, overleaf) Second, task-exclusive component.Numbers indicate relative power, in percentage, with respect to the main component from eachindividual, presented in the same order as in (a).

posterior scalp regions, very similar intopography to the N200 component of thevisual evoked potential. In all subjects,theta induced power peaked in coincidencewith the peaks of the N200, and delta in thetime window occupied by the P300component. In ten subjects, the topographyof theta power was practically coincidentwith the one from the N200 (typicallydouble peaks at the occipital region); in the

(5a)

465BASILE Biol Res 40, 2007, 451-470

subjects and showed very small increases infour subjects. This type of finding iscommonly reported in the literature, anddiscussed below in conjunction with resultsvalid for all bands.

With respect to the alpha range inducedband(s), a thorough consideration wouldrequire an independent and major work dueto complexity, the same being the case for acomplete phase analysis. We thus discusshere only a few points which support someviews under current discussion in theliterature, especially our contribution thatregards its topography. Half of the subjectspresented a distinct alpha-1 band - around 8

two remaining subjects, delta topographywas closer to N200. Thus, both powerbands appear simply to contribute to thecomposition of the evoked potential, andthe method here used should not add muchinformation with respect to theconventional ERP: induced and evokedpower are coincident in this case, and themost important fact that seems to occur inthese bands is a synchronization betweentheir generators with respect to the stimuli,or phase resetting: In two subjects, deltaburst averages were actually decreased inamplitude and unchanged in one, whereastheta maintained its amplitude in two

(5b)

BASILE Biol Res 40, 2007, 451-470466

Hz - in addition to the alpha-2 bandcentered around 11 Hz, and the remaininghalf, exclusively a broad or narrow alpha-2band. Induced alpha behavior in relation totask time was multiphasic, with a relative(seen as an inflection) reduction/desynchronization around 200 ms after thestimuli, a relative peak (almostunnoticeable when compared to themaximum induced alpha) overlapping withthe delta peak and the P300 evokedpotential, and maximum power at the centerof the ISI, overall around 700 ms. Theobservation of desynchronization versussynchronization of alpha, evensimultaneously during the immediate post-stimulus time, depends on the method ofpower analysis (Klimesh et al., 2000): ourmethod emphasizes the induced part, whichincludes the evoked alpha (contributing tothe evoked potential, see Cacace andMcFarland, 2003) and overrides the peri-and post-stimulus desynchronization. Moreinteresting, however, was to notice that asmall proportion of (evoked) alpha powerwas phase-locked to stimuli throughout theISI, a phenomenon that we had observedpreviously in various different experiments,in spite of the short alpha wavelength ascompared to the ISI. One may speculatethat this part of the alpha generators couldserve as a kind of task-time marker orestimator, and believe that competinghypotheses on alpha functional role can bereconciled: alpha synchronization mayindex cortical idling (Pfurtscheller, 2001), aconcept that accounts for most of alphabehavior including the Berger effect and indrowsiness, but when occurring in the‘reference interval’ in cyclical tasks such asthe present, will also mean preparation fordetection of forthcoming stimuli (Knyazevet al., 2006), around and immediately afterwhich both (de-) and synchronization maybe observed (Klimesh et al., 2000). Alphascalp distribution included the expectedposterior, occipital-parietal power maximain all subjects, but with fine details in scalpdistribution peculiar to each individual.However complex the individual mapshapes, based on our own unpublishedobservations (for instance, one of thesubjects presented an identical complex

alpha-2 map during a visual verbal taskperformed 6 years before), we believe thatalpha topography is fairly stable in the longterm and possibly task-independent. This isin agreement with a few studies on stabilityof qEEG, although using fewer electrodes,that depict alpha as the most stable bandwithin individuals (see Neuper et al., 2005).Our contribution, however, is theconclusion based on alpha analysis, butverified in all frequency bands, that thecomputation of corrected latency averagesresulted in undistinguishable topographyfor the evoked or induced power bands.One very recent study using the standard10-20 montage clearly corroborates thisfinding, for the alpha band (Hanslmayr etal., 2006).

The central finding of this work was thepresence of beta band power increasesthroughout the ISI, peaking close to the S2stimulus, in all subjects. Since the betarange is traditionally associated withwakefulness and behavioral arousal, wewere interested in verifying the beta bandgenerators, whether they would have asimple and common distribution acrosssubjects. However, once we computed thetopographic maps, we immediately noticedthe qualitative similarity between inducedbeta and the SPs: common task-timebehavior, multifocal, complex topography,highly variable across subjects, andsignificantly enhanced during the task ascompared to the passive stimulation controlcondition. Thus, proceeding to computebeta burst averages centered in peaksoccurring within the 700 to 1600 ms tasktime window, and model their generators byCDR, we obtained analogous results: Betagenerators showed an equally complexpattern, comprising prefrontal and posteriorcortical areas (as do SPs), highly variableacross subjects, with only parietal area 7demonstrating some, but variable, degree ofrelative current density in all subjects. Area7 was also the only common SP generatorregion across subjects (Basile et al., 2006),a fact that may be attributed to the merewakeful state, or that in other primates hasbeen attributed to interested attention to theenvironment (Lynch et al., 1977; Yin andMountcastle, 1978). We thus considered to

467BASILE Biol Res 40, 2007, 451-470

have obtained a new index of attention,which could not have been observed byregular ERP averaging. The comparisonbetween beta and SP generators by visualinspection (the reader may refer to the SPgenerator figure in Basile et al., 2006)revealed a largely complementary set ofactive areas between the two indexes:current foci were typically different fromeach other, forming mostly adjacent setswithin subjects, and whenever there was anoverlap of SP and beta generating areas,they did not correspond to the maingenerators in each case; that is, strong withweak in almost all cases. Moreover, SPgenerators were overall more spread overthe cortical surface, and given the manycases of adjacent generator positions, it wasunavoidable for us to speculate that SPscould represent a fringe effect stemmingfrom the beta generating areas. SPs are fora long time known to be microscopicallygenerated by a major contribution from thepotassium buffering function of glia(Skinner and Molnar, 1983; Roitback,1993; Mitzdorf, 1993), in situations ofincreased overall neural firing, as seems tooccur areas active in the beta range. Finally,our phase analysis results present weakevidence that cortical areas active in thebeta range also oscillate in phase synchronywith each other, at least roughlyaccompanying the power changes.However, the stability of partial averages(with respect to groups of electrodes rankedby peak latency), as well as independentstudies using single-cell, extra-cellularrecordings and model simulations offerstronger support to this view (Bibbig et al.,2002). This suggests that such areasbecome co-recruited, either reciprocally, orby some common subcortical projection(s).

A surprising and, if replicated, mostimportant finding, regarded all frequencybands: their baseline or pre-S1 topographicdistribution. In all cases and subjects,almost all of the task-induced powerdistribution was already manifested duringbaseline, and verified even during rest inthree subjects. That is, individual-specificgenerators of each band (very similar,however, between delta, theta and betaranges) are already active during resting

wakefulness. This finding, also veryrecently observed in the alpha range byHanslmayr and colleagues (2006), whencombined with the indistinguishabletopography between induced and evokedactivity, and with the phase analysis results,may be interpreted into a general panoramaregarding oscillatory activity: thegenerators of each sub-band are largelyfixed in space and continuously active; forall bands, almost all of the task-inducedpower is accounted for by the samegenerators active during the baselineperiod. And during the task, combinationsof three additional phenomena may occur,depending on the sub-band of oscillations:(1) increases and decreases in amplitude,that in some cases appear to beaccompanied by roughly parallel (2)changes in coherence between varyingproportions of the cell populationscomposing the generators, and (3) phase-resetting with respect to task-events. In thepresent study, for instance, it appears thatdelta and theta amplitude changes are leastimportant, given that half of the subjectsshowed very low or no change in theta, oreven decreases (2 subjects) in delta burstaverage amplitude. Alpha generatorssuffered resetting partly (that may beexplained by being restricted to aproportion of cells within the samemacroscopic areas), but most of their powerincrease was out of synchrony with taskevents. In the case of beta, the task-inducedpower increases was appear to be generatedby synchronous areas, but out of phase withtask events. The relative contribution byeach of such phenomena thus depends onfrequency band, and task, as concluded inother studies, some of which explicitlyrelate them to the composition of ERPssimultaneously obtained (Gruber et al.,2005; Hanslmayr et al., 2006; Valencia etal., 2006; and the interesting invasive studyby Fell et al., 2004).

Taken together, our main findings leadus to the following implication topsychophysiology: as opposed to corticalactivity linked to sensory stimulation(evoked potentials, delta, theta and partlyalpha rhythms), which is simpler indistribution and more preserved across

BASILE Biol Res 40, 2007, 451-470468

subjects, electrical activity directly relatedto expecting attention, namely the SlowPotential component of the ERP(corresponding to the DC-1 Hz inducedbandpower) and the induced beta rhythm, ismultifocal, complex in distribution, andhighly variable across subjects. Moreover,it appears that when one engages in thetask, it is largely the same individual-specific set of cortical areas, continuouslyactive during simple resting wakefulnessbut without phase synchrony, that enter inphase and increase in power, and mayrecruit a few other, equally individual-specific areas. Our data, together with theresults from the not many fMRI and PETstudies that present individual data onevent-related metabolic changes (Cohen etal., 1996; Herholz et al., 1996; Fink et al.,1997; Davis et al., 1998; Hudson, 2000;Brannen et al., 2001; Tzourio-Mazoyer etal., 2002), are compatible with the view ofa “degenerate” mapping between presumedfunction and function implementing corticalareas (Noppeney et al., 2004). The mainpiece of knowledge that has guided ourconventional hopes for a universalfunctional mapping are the fairly specificpatterns of cortico-cortical connections inmammals (Pandya et al., 1988), especiallyso between visual cortices (Macko andMishkin, 1985) but known to applythroughout the neocortex, includingprefrontal areas (Pandya and Yeterian,1990; Barbas, 1992). However, we believethat the mere complexity and number ofpossible cortico-cortical functionalpathways are sufficient to allow theformation of variable sets of interconnectedcortical areas across individuals, before andduring execution of any given task.Therefore, we forecast the completeabandonment of the search forpredetermined and unique functions to beattributed to given non-sensory-motor areas.

The implications of our results andconclusions to our own line of investigationare a reversal of focus: to aim exactly atunderstanding task-related and task-specificcortical activity, we will need to bettercomprehend the individual-specific corticaltopography of the resting, baselinecondition. Among critical aspects, we need

to understand how the prevalently activesets of areas are formed during ontogenyand particularly during learning. We haveonly one subject as evidence for thestability in the order of one month, butthree indicating a clear change in two orfour years. It is fundamental to know howspontaneous is the formation of anindividual resting pattern, as opposed tobeing subject to experimental or naturaltraining influences. The possibility ofdirectly influencing and expanding theresting pattern to apparently inactive areaswould substantiate restorative neurologyand neuropsychological procedures for acontrolled reorganization of corticalcircuits. On the other hand, if some corticalareas in one individual prove to be indeed‘silent’, supposing that an exhaustivebattery of experimental tasks can ever beimplemented, such battery would become aprocedure to complement the Wada andrelated tests to aid in neurosurgicalplanning. Finally, these overall conclusionsare naturally compatible with the fact ofhigh variability in clinical symptomatologyafter focal lesions: similar lesions incortical ‘association’ areas lead to criticalimpairments in some individuals and mayonly be noticed in others by incidentallaboratory examinations. Although we haverecently used a method of scoring corticalactivity by estimated corticalcytoarchitectonic area in each individual,previously to group or condition statisticalcomparisons (Basile et al., 2003), stillsearching for commonalities acrosssubjects, equivalent attempts may be usefulin the reconsideration of individual datafrom studies that have only presented groupaverages. We propose, as some otherauthors also explicitly do, that functionalclaims regarding cortical areas never bemade based on group averaged data(Steinmetz and Seitz, 1991; Davis et al.,1998; Noppeney et al. , 2004). Thechallenge of dealing with individual results,from the large number of studies alreadypublished, may at least help us to decidewhether individual patterns are indeedarbitrary or follow rules, in which case newpsychophysiological theories shall bedeveloped.

469BASILE Biol Res 40, 2007, 451-470

ACKNOWLEDGMENTS

This research was supported by the grants03/02297-9 and 02/13633-7 from FAPESP,São Paulo, Brazil. We wish to thank DrCláudia Leite, Dr Edson Amaro Jr. and thestaff from the Department of Radiology ofthe University of São Paulo MedicalSchool, for kindly acquiring and preparingthe MRI sets, and Márcio Costa for hisvaluable technical support.

REFERENCES

BARBAS H (1992) Architecture and cortical connectionsof the prefrontal cortex in the rhesus monkey.Advances in Neurology 57: 91-115

BASILE LFH, BALLESTER G, CASTRO CC and GATTAZWF (2002) Multifocal slow potential generators revealedby high-resolution EEG and current densityreconstruction. Int J Psychophysiol 45: 227-240

BASILE LFH, BALDO MV, CASTRO CC and GATTAZWF (2003) The generators of slow potentials obtainedduring verbal, pictorial and spatial tasks Int JPsychophysiol 48: 55-65

BASILE LFH, YACUBIAN J, FERREIRA BLC, VALIMAC, and GATTAZ WF. Topographic abnormality ofslow cortical potentials in schizophrenia. BrazilianJournal of Medical and Biological Research, 2004, 37:97-109

BASILE LFH, BRUNETTI EP, PEREIRA JR JF,BALLESTER G, AMARO JR E, ANGHINAH R,RIBEIRO P, PIEDADE R, GATTAZ W F. Complexslow potential generators in a simplified attentionparadigm. Int J Psychophysiology (in Press)

BIBBIG A, TRAUB RD, WHITTINGTON MA (2002)Long-range synchronization of gamma and betaoscillations and the plasticity of excitatory andinhibitory synapses: a network model. J Neurophysiol88: 1634-54

BRANNEN JH, BADIE B, MORITZ CH, QUIGLEY M,MEYERAND ME, HAUGHTON VM. (2001)Reliability of functional MR imaging with word-generation tasks for mapping Broca’s area. AJNR Am JNeuroradiol 22: 1711-8

BRUNS A, ECKHORN R (2004) Task-related couplingfrom high- to low-frequency signals among visualcortical areas in human subdural recordings. Int JPsychophysiol 51: 97-116

BUCHNER H, KNOLL G, FUCHS M, RIENACKER A,BECKMAN R, WAGNER M, SILNY J, PESCH J(1997) Inverse localization of electric dipole currentsources in finite element models of the human head.Electroenceph Clin Neurophysiol, 102: 267-278

COHEN MS, KOSSLYN SM, BREITER HC, DIGIROLAMO GJ, THOMPSON WL, ANDERSON AK,BROOKHEIMER SY, ROSEN BR, BELLIVEAU JW(1996) Changes in cortical activity during mentalrotation. A mapping study using functional MRI. Brain,119 (Pt 1): 89-100

DAVIS KD, KWAN CL, CRAWLEY AP, MIKULIS DJ(1998) Functional MRI study of thalamic and corticalactivations evoked by cutaneous heat, cold, and tactilestimuli. J Neurophysiol 80: 1533-46

EDELMAN GM, GALLY JA (2001) Degeneracy andcomplexity in biological systems. Proc Natl Acad SciUSA 98: 13763-8

FELL J, DIETL T, GRUNWALD T, KURTHEN M,KLAVER P, TRAUTNER P, SCHALLER C, ELGERCE, FERNANDEZ G (2004) Neural bases of cognitiveERPs: more than phase reset. J Cogn Neurosci 16:1595-604

FINK GR, FRACKOWIAK RS, PIETRZYK U,PASSINGHAM RE (1997) Multiple nonprimary motorareas in the human cortex. J Neurophysiol 77: 2164-74

FUCHS M, WAGNER M, WISCHMANN HA, KOHLERT, THEISSEN A, DRENCKHAHN R, BUCHNER H(1998) Improving source reconstructions by combiningbioelectric and biomagnetic data. ElectroencephalogrClin Neurophysiol 107: 93-111

FUCHS M, WAGNER M, KOHLER T, WISCHMANN HA(1999) Linear and nonlinear current densityreconstructions. J Clin Neurophysiol 16: 267-95

GRUBER WR, KLIMESCH W, SAUSENG P,DOPPELMAYR M (2005) Alpha phasesynchronization predicts P1 and N1 latency andamplitude size. Cereb Cortex 15: 371-7

HANSLMAYR S, KLIMESCH W, SAUSENG P, GRUBERW, DOPPELMAYR M, FREUNBERGER R,PECHERSTORFER T, BIRBAUMER N (2006) AlphaPhase Reset Contributes to the Generation of ERPs.Cereb Cortex. (in Press)

HERHOLZ K, THIEL A, WIENHARD K, PIETRZYK U,VON STOCKHAUSEN HM, KARBE H, KESSLER J,BRUCKBAUER T, HALBER M, HEISS WD (1996)Individual functional anatomy of verb generation.Neuroimage 3: 185-94

HUDSON AJ. Pain perception and response: centralnervous system mechanisms (2000) Can J Neurol Sci27: 2-16

HYVARINEN A, OJA E (2000) Independent componentanalysis: algorithms and applications. Neural Netw 13:411-30

LACHAUX JP, RODRIGUEZ E, MARTINERIE J,VARELA FJ (1999) Measuring phase synchrony inbrain signals. Hum Brain Mapp 8: 194-208

LYNCH JC, MOUNTCASTLE VB, TALBOT WH, YINTC (1977) Parietal lobe mechanisms for directed visualattention J Neurophysiol 40: 362-89

MACKO KA, MISHKIN M (1985) Metabolic mapping ofhigher-order visual areas in the monkey. Res PublAssoc. Res Nerv Ment Dis 63: 73-86

MCCALLUM WC (1988) Potentials related to expectancy,preparation and motor activity. In: Handbook ofElectroencephalography and Clinical Neurophysiology.Human Event-Related Potentials (revised series vol. 3)TW Picton (Ed.). Elsevier Science Publishers, 427-534

MITZDORF U (1993). Contributions of extracellularpotassium increases to transient field potentials (reviewof data). In: Slow Potential Changes in the Brain. Chap16. Haschke W, Roitbak AI, Speckmann, EJ (Eds.).Birkhäuser. Boston

NOPPENEY U, FRISTON KJ and CATHY J. (2004) PriceDegenerate neuronal systems sustaining cognitivefunctions Journal of Anatomy 205: 433-442

PANDYA DN, YETERIAN EH (1990) Prefrontal cortex inrelation to other cortical areas in rhesus monkey:Arquitecture and connections. In: Progress in BrainResearch, vol.85. Uylings HBM, Van Eden CG, DeBruin JPC, Corner MA, Feenstra, MGP (Eds.). ElsevierScience Publishers BV. 63-94

PANDYA DN, SELTZER, B. AND BARBAS, H (1988)Input-output organization of the primate cerebralcortex. Comparative primate biology, 4: 39-80

BASILE Biol Res 40, 2007, 451-470470

POSNER MI (1980) Orienting of attention. Q J ExpPsychol 32: 3-25

POSNER MI, SNYDER CR, DAVIDSON BJ (1980)Attention and the detection of signals. J Exp Psychol109: 160-74

RAMIREZ PM, DESANTIS D, OPLER LA (2001) EEGbiofeedback treatment of ADD. A viable alternative totraditional medical intervention? Ann N Y Acad Sci931: 342-58

ROITBAK AI (1993). Cortical slow potentials ,depolarization of glial cells , and extracellularpotassium concentration. In: Slow Potential Changes inthe Brain. Chap 14. Haschke W, Roitbak AI,Speckmann, E J. (Eds.). Birkhäuser. Boston

SKINNER JE, MOLNAR M (1983). Event-relatedextracellular potassium ion activity changes in frontalcortex of the conscious cat. J Neurophysiol 49: 204-215

Steinmetz H, Seitz RJ (1991) Functional anatomy oflanguage processing: neuroimaging and the problem ofindividual variability Neuropsychologia 29: 1149-61

TALAIRACH J, TOURNOUX P (1993). Referentiallyoriented cerebral MRI anatomy. Atlas of stereotaxicanatomical correlations for gray and white matter.Thieme Medical Publishers, Inc. New York

TALAIRACH J, TOURNOUX P (1997). Co-planarstereotaxic atlas of the human brain. Thieme MedicalPublishers, Inc. New York

TALLON-BAUDRY C, BERTRAND O, DELPUECH C,PERNIER J (1996) Stimulus specificity of phase-

locked and non-phase-locked 40 Hz visual responses inhuman. J Neurosci 16: 4240-9

TZOURIO-MAZOYER N, JOSSE G, CRIVELLO F,MAZOYER B (2002) Interindividual variability in thehemispheric organization for speech. Neuroimage 16:765-80

UYLINGS HB, RAJKOWSKA G, SANZ-ARIGITA E,AMUNTS K, ZILLES K (2005) Consequences of largeinterindividual variability for human brain atlases:converging macroscopical imaging and microscopicalneuroanatomy. Anat Embryol (Berl) 210: 423-31

VALENCIA M, ALEGRE M, IRIARTE J, ARTIEDA J(2006) High frequency oscil lat ions in thesomatosensory evoked potentials (SSEP’s) are mainlydue to phase-resetting phenomena. J Neurosci Methods154: 142-148

VANDENBROUCKE MW, GOEKOOP R, DUSCHEK EJ,NETELENBOS JC, KUIJER JP, BARKHOF F,SCHELTENS P, ROMBOUTS SA (2004)Interindividual differences of medial temporal lobeactivation during encoding in an elderly populationstudied by fMRI. Neuroimage 21: 173-80

WHELESS JW, CASTILLO E, MAGGIO V, KIM HL,BREIER JI, SIMOS PG, PAPANICOLAOU AC (2004)Magnetoencephalography (MEG) and magnetic sourceimaging (MSI). Neurologist 10: 138-53

YIN TC, MOUNTCASTLE VB (1978) Mechanisms ofneural integration in the parietal lobe for visualattention. Fed Proc 37: 2251-7