The generators of slow potentials obtained during verbal, pictorial and spatial tasks

International Journal of Psychophysiology 48 (2003) 55–65

The generators of slow potentials obtained during verbal, pictorial and spatial tasks ´ Luis F.H. Basile*, Marcus V.C. Baldo, Claudio C. de Castro, Wagner F. Gattaz ˜ Paulo, Laboratory of Neurosciences (LIM-27), Department of Psychiatry, Faculty of Medicine, University of Sao Av. Dr. Ovidio Pires de Campos syn, P.O. Box 3671, Sao Paulo, SP 05403-010, Brazil Received 22 August 2002; received in revised form 10 December 2002; accepted 19 December 2002

Abstract The purpose of this study was to test whether slow cortical electrical activity is specific to performance on verbal, pictorial and spatial tasks. Twenty-nine healthy subjects were required to compare pairs of visual stimuli separated by a delay of 2.5 s in a S1–S2 contingent negative variation-type paradigm. Slow potentials (SPs) were recorded by high-resolution EEG (123 channels) and their generators modeled by current density reconstruction using individual MRIs as source space models. Activity in each architectonic area of Brodmann was scored with respect to individual maximum current by a percentile method. Results showed a multifocal pattern of current density foci comprising the SP generators, including frontal and posterior cortices in all subjects, with the most active areas being common to the three tasks. In spite of the intersubject variability in the sets of active areas for each given task, a few cortical areas were observed to discriminate between tasks in a statistically significant way: the verbal task corresponded to stronger electrical activity in right area 45 than the other tasks; the spatial to weaker activity in right area 38 and left area 5 than the other tasks; the pictorial, compared to the spatial task, to stronger activity in left area 39; the verbal, compared to the spatial task, to stronger activity in left area 10, and compared to the pictorial, to weaker activity in right area 20. The present method of SP analysis may aid in the functional mapping of human association cortices in individual cases. We discuss our results emphasizing intersubject variability in cortical activity patterns and the possibility of finding more universal patterns. 䊚 2003 Elsevier Science B.V. All rights reserved. Keywords: Prefrontal cortex; High-resolution electroencephalography; Slow potentials; Contingent negative variations; Source localization; Functional mapping

1. Introduction In a recent study, we obtained evidence for generation of slow potentials (SPs) in multiple cortical association areas (Basile et al., 2002). The general rationale of our present line of investiga*Corresponding author. Tel.: q55-11-30697284y32846821; fax: q55-11-2894815. E-mail address: [email protected] (L.F. Basile).

tion is to use such analysis of SPs obtained during various conditions, by emphasizing on their spatial distribution, as an additional, noninvasive tool for functionally mapping the human association cortex. The particular aim of this study was to test for task-specific generators of SPs. SPs are a class of potentials that includes the well-studied contingent negative variations (CNVs, for a review,

0167-8760/03/$ - see front matter 䊚 2003 Elsevier Science B.V. All rights reserved. doi:10.1016/S0167-8760(03)00004-7

56

L.F.H. Basile et al. / International Journal of Psychophysiology 48 (2003) 55–65

McCallum, 1988). CNVs were originally provoked in situations, where a given stimulus forewarned the experimental subject about upcoming stimuli requiring motor responses (Walter et al., 1964; Walter and Crow, 1964). However, it is now well established that motor responses are not necessary to provoke SPs, which also occur during mere stimulus anticipation (Ruchkin et al., 1986; Brunia, 1988). It is here relevant to mention that a particular type of task, using paired-associate stimuli separated in time, are also accompanied by SPs (e.g. Rugg, 1984), as would be expected from the point of view of EEG-SP literature. That is, in spite the fact that most such studies focus on evoked potentials and do not use DC recordings, given that the first stimulus of a pair serves as a warning to the next, SPs are still observed, usually reported as by-products of the studies. Regarding SP generators, the frontal cortex has been suspected as their main or exclusive source since the CNV discovery (Walter et al., 1964; Walter and Crow, 1964; Fuster, 1989). By modeling the generators of non-motor SPs and their slow magnetic counterparts (EEG and magnetoencephalography (MEG)), using one or few equivalent current dipoles (ECDs), we and others have found various fields in the prefrontal cortices as the centers of such activity (Bocker et al., 1994; Basile et al., 1994, 1996, 1997). Although ECDs are by definition unrealistic point-like models of activity that is sometimes known to be extended in space, and must be interpreted as unevenly weighed centers of current density, we obtained some evidence for task-specific generators, by comparing their positions in different tasks. More recently, with the availability of extended source modeling (current density reconstruction (CDR)), we obtained evidence for non-exclusive origins of SPs in prefrontal cortices (Basile et al., 2002). Those results agree with evidence from invasive intracranial studies (Ikeda et al., 1996; Hamano et al., 1997), which by themselves are inconclusive due to the use of monopolar electrodes. In this study, by using pairs of stimuli in three different visual association tasks (verbal, pictorial and spatial), a first stimulus (S1) from each category would serve as a task-specific warning. That is, they prepare for selectively attending to

stimuli belonging to either domain. However, to insure that S2 stimuli, presented 2.5 s afterwards, would be attended to, relevant to the task, we also instructed subjects to categorically compare the stimuli forming the pairs. We also instructed the subjects that their learning of relations between stimuli would be tested after each block. Thus, S1 stimuli, as in most real life situations in which meaningful sequences are attended to such as in reading, have also to remain in short term memory to be associated with S2. Thus, SPs in such situations are probably akin to sustained single cell activity such as observed in delay matching-tosample tasks, as it is long suspected to be a main microscopic substrate of CNVs in general (Fuster, 1989). The choice of the three classes of stimuli within the visual domain was based on traditional or well founded anatomical distinctions: given that, we used 29 right-handed subjects, one could expect that engagement in a verbal task, as opposed to two types of nonverbal ones would correspond to activity in verbal specific areas, in particular hemispheric asymmetries; the use of pictorial vs. spatial material among the nonverbal domain was based on the anatomical distinction between the dorsal and ventral visual systems, known to extend itself into the prefrontal cortices, with predominant connectivity between particular prefrontal areas and either visual stream (Pandya and Yeterian, 1990). Most studies searching for task-specific brain activity, now comprising a vast literature, use group averaged data, projected onto a model, ‘average brain’ for their conclusions, such as the work using PET and most fMRI studies. On the other hand, there are now reports on individual case analysis using fMRI and EEG or MEG (e.g. Cohen et al., 1996; Fink et al., 1997; Brannen et al., 2001; Basile et al., 1994, 1996, 2002), indicating intersubject variability in the ensembles of cortical areas active during any given task. The present work is an attempt to preserve intersubject variability but to simultaneously describe cortical activity beyond a simple tabulation of active cortical areas (presence vs. absence of current density foci).

L.F.H. Basile et al. / International Journal of Psychophysiology 48 (2003) 55–65

57

2. Methods 2.1. Subjects Twenty-nine healthy individuals with normal vision and hearing, 15 male and 14 female, participated in the study. They ranged in age between 20 and 55 years and were all right-handed, with no history of drug or alcohol abuse, and no current drug treatment. All subjects signed consent forms approved by the Ethics Committee of the Univer˜ Paulo Hospital. sity of Sao 2.2. Stimuli and task A commercial computer program (Stim, Neurosoft Inc.) controlled all aspects of the task. Visual stimuli composing the paired-associates were taken from pools of 40 items belonging to each of three modalities, verbal, pictorial and spatial. They were presented on a computer screen, and subtended less than 38 in the visual field. Experiments alternated 18 memorization with corresponding test blocks. Stimuli belonging to a given pair (S1 followed by S2) had onsets separated in time by 2.5 s (interstimulus interval) (bottom of Fig. 1); each stimulus lasted for 0.5 s and pairs were separated by 6 s (intertrial interval). Approximately half of the pairs consisted in a categorical match and half in a mismatch during a given block (food or nonfood for the verbal, abstract pictures or traces for the pictorial and same direction or not for spatial pairs). Eight different pairs were presented twice (reversed order on the second time) during a block, which thus consisted in 16 trials. Performance was measured after each block, by the presentation of the same sequence, but in this case only S1 was presented, remaining on the screen until the subject decided whether its recently learned pair (S2 presented on the immediately preceding sequence) belonged or not to the same category as S1. They had to indicate a match by pressing a button with the right index finger or a mismatch by using the middle finger. We measured performance by the percent correct trials, from the total of 288 trials comprising the experiment.

Fig. 1. Example from one individual, of average SPs obtained for each task. Waveforms from all channels are superimposed, with common baselines. Task events are indicated below (S1– S2 interval). Vertical bars indicate 5 mV.

2.3. Recordings, computation of average potentials and acquisition of MRIs We used a fast AgyAgCl electrode positioning system consisting of an extended 10–20 system, in a 123-channel montage (Quick-Cap, Neuromedical Supplies䉸), and an impedance-reducing gel which eliminated the need for skin abrasion (Quick-Gel, Neuromedical Supplies䉸). Impedances usually remained below 3 kV, and channels that did not reach those levels were eliminated from the analysis. We used a digitizer (Polhemus䉸) to record actual electrode positions with respect to each subject’s fiduciary points: nasion and preauricular points. Two bipolar channels, out of the 123-channels in the montage were used for recording both horizontal electrooculograms (HEOG) and vertical electrooculograms

58

L.F.H. Basile et al. / International Journal of Psychophysiology 48 (2003) 55–65

(VEOG). Linked mastoids served only as reference for data collection (common average reference was used for source modeling) and Afz was the ground. We used four 32-channel DC amplifiers (Synamps, Neuroscan Inc.) for data collection and the Scan 4.0 software package (Neurosoft Inc.) for initial data processing (until computation of averages). The filter settings for acquisition were from DC to 50 Hz, and the digitization rate was 250 Hz. The EEG was collected continuously, and epochs for averaging spanned the interval from 300 ms before S1 to 400 ms after S2 presentation. Baseline was defined as the 300 ms preceding S1. Artifact elimination was automatic: epochs containing signals in either HEOG or VEOG channels above q50 mV or below y50 mV were eliminated. In our montage, the VEOG detected blinks as deflections above 130 mV in the positive direction. On a pilot study, we verified the magnitudes of small saccades to various directions, directed to stimuli of approximately 38 of eccentricity, to be associated to deflections of above 70 mV in one or both EOG channels. Therefore, given our 50 Hz lowpass filter and epoch size of 5 s, the 50 mV criterion in the EOG channels proved empirically sufficient for the total elimination of epochs containing blinks and small eye movements. We used a minimum of 40 epochs, from the total 96 from each task, to compute the average SPs. MRIs were obtained by a 1.5 T GE machine, model Horizon LX. Image sets consisted in 124 T1-weighed saggital images of 256 by 256 pixels, spaced by 1.5 mm. Acquisition parameters were: standard echo sequence, 3D, fast SPGE, two excitations, RTs6.6 ms, ETs1.6 ms, flip angle of 158, F.O.V.s26=26 cm2. Total acquisition time was approximately 8 min. 2.4. Intracranial source reconstruction The computed average SPs, MRI sets and electrode position digitization files were the raw input into the software package that performed all further analysis (Curry V 4.5, Neurosoft Inc.). The first critical step in the source analysis was the estimation of noise in the data, whose criterion was the S.D. in the amplitudes of points comprising

the 300 ms baseline. In all cases the noise thus computed was below 1 mV (typically ;0.6 mV). For all subsequent analysis, the electrical data were converted to absolute numbers, i.e. signal to noise ratios, based on individual channel noise statistics. For the inclusion of a ‘noise component’ into the source model itself, the physical unit-free or ‘standardized’ data were decomposed by independent component analysis (ICA). Before deciding to use ICA, we have also applied singular value decomposition (SVD) to a subset of the data (ns 10), in a pilot test. In all those cases the main components were very similar but the resulting SVD- or ICA-filtered maps (for components above SNRs1) were virtually identical, visually indistinguishable. Thus, we chose ICA, which searches for the highest possible statistical independence or redundancy reduction between components, for its increasing acceptance as a robust method of blind signal decompositionydeconvolution (for a review see, e.g. Hyvarinen and Oja, 2000). ICA was applied to the remaining cases, to each individual’s whole space-time data set, i.e. to the m=n data matrix (m used channels times 625 time samples corresponding to the 2.5 s interval from the onset of S1 to that of S2). Finally, we fed the reconstruction algorithm with the main ICA component(s) as data to be fitted, thus defining the ‘noise component’ of the model as the sum of remaining components (with loadings below SNRs1), which at the time point of interest lead invariably to negligible scalp potentials. Such noise component is required by any CDR method for setting up the inconsistent system of equations that relate sources with measured data (see below). Otherwise, a system identical to zero (vector) would result in solutions that rely excessively on the actual measurements, being subject to important distortions by outlying values and any form of experimental error. MRI sets were linearly interpolated to create three-dimensional images, and semi-automatic algorithms based on pixel intensity bands served to reconstruct the various tissues of interest. A boundary element model (BEM) of the head compartments was implemented, by triangulation of collections of points supported by the skin, skull

L.F.H. Basile et al. / International Journal of Psychophysiology 48 (2003) 55–65

and cerebrospinal fluid (internal skull) surfaces. Mean triangle edge lengths for the BEM surfaces were 10, 9 and 7 mm, respectively. Fixed conductivities were attributed to the regions enclosed by those surfaces 0.33, 0.0042 and 0.33 Sym, respectively. Finally, a reconstructed brain surface, with mean triangle side of 3 mm, served as the model for dipole positions, corresponding to approximately 20 000 points. The electrode positions were projected onto the skin’s surface following the normal lines to the skin. The detailed description of the assumptions and methods used by the ‘Curry 4.5’ software for MRI processing and source reconstruction may be found elsewhere (Curry 4.0 User Guide, 1999; or e.g. Buchner et al., 1997; Fuchs et al., 1998, 1999). ‘Curry’ then calculated the lead field matrix, that represents the coefficients of the set of equations which translate the data space (SNR values in the set of channels per time point) into the model space (the approximately 20 000 dipole supporting points). The source reconstruction method itself was Lp norm minimization, with ps 1.2 both for data and model terms. The regularization factor, or l values to be used, typically converged after repeating the fitting process 3–4 times. Lambda is the numeric factor that weighs the model term, one of the two terms of the reconstruction problem: the second or data term is the set of equations whose solution approaches the model (intracranial current densities) to the measured data. Since both terms are minimized simultaneously, l gives the balance between goodness of fit and model size: D2sZD( j)ZpqlZM( j)Zp, where D2 is the extended variance (variable to be minimized), D( j) or data term equals Ljym (L is the lead field matrix, j is the current vector per time point and m is the measured data vector), and M( j) is the model term, which in this case is a measure of the ‘size’, or scalar sum of all current strengths. A short discussion regarding our choice of reconstruction method is presented in the methods section of a closely related work (Basile et al., 2002).

59

Resulting models were inspected with respect to the individual anatomy directly, a straightforward procedure allowed by our software. Foci of current density were analyzed in terms of which cytoarchitectonic areas they covered, after being separately defined or isolated by the application of various percentile cutoff values (see below). The Brodmann areas containing current foci were tabulated after verification by comparison with classical illustrations and the conventional Talairach and Tournoux atlases (1988, 1993). 2.5. Scoring of relative current density and statistical analysis We implemented a semi-quantitative method to compare electrical activity in each cortical area and task across subjects, without collapsing the original data into an ‘universal’, model brain, thus preserving individual anatomical and functional (electrical) information. We scored relative current densities by cortical cytoarchitectonic area into four levels (negligible, low, moderate and high), using percentile bands relative to maximum current, defined by fixed cutoff levels. We inspected current density results in each cortical area after applying a set of percent cutoff values from each data set’s maximum current value. Cortical areas containing current density foci of intensity above 70% of the maximum were given a relative electrical activity score 3 (high). Activity between 40 and 70% was considered of moderate intensity and given a score 2, between 10 and 40% low and given a score 1, and below 10% of the maximum was considered negligible and given a score 0. We are aware that our transformation of actual current distributions into relative activity scores greatly reduces the originally obtained ‘experimental resolution’ of the electrical activity distributions for each subject and condition. However, we must insist that this method allowed the preservation of individual results, by keeping observed relative activity tied to subject-specific anatomy. This is an alternative to keeping the richer, ‘continuous’ distribution of current strengths from the original data, but loosing individual anatomical meaning, if individual or group averaged data were projected onto an abstract ‘anatomical’ space. Then, scores

60

L.F.H. Basile et al. / International Journal of Psychophysiology 48 (2003) 55–65

in each cytoarchitectonic area were tabulated and became the dependent variables of the study. As a first, exploratory approach to the global behaviour of our data, we performed a multivariate analysis of variance (MANOVA), with tasks considered as repeated measures. We performed planned comparisons, searching for significant differences among levels (3=2 MANOVA, three tasks and two hemispheres). We thus tested for the task effects by comparing them by pairs, searching for activity in cortical areas that could be specifically related to given tasks. We also collapsed activity from same cortical areas across both hemispheres as a second test for hemisphere independent task-

Fig. 3. Example of CDR results in one subject. Upper row, verbal task, middle row, pictorial task, and bottom, spatial task. Arrow size proportional to local current density.

specific activity; and by collapsing the tasks and comparing hemispheres, we searched for taskunspecific lateralization effects. Then, as a more rigorous approach and test for consistency of results, in which the electrical activity scores were treated as ordinal variables, we performed nonparametric analysis for comparisons between tasks and nonparametric (Spearman) correlations between electrical activity and task performance. 3. Results

Fig. 2. Left: Isopotential contour maps corresponding to the SP distribution from the individual from Fig. 1, at S2 onset. Right: Isopotential maps corresponding to group averaged data (collapsed across the montage), projected over the scalp of an individual of median head size. Potentials calculated with respect to ‘common average reference electrode’. Notice maximum voltages of group averaged data reduced when compared to individual data, which is due to intersubject variability in the complex voltage distributions.

All subjects reported that effort for concentration during learning the pairs was necessary, but performance was variable across subjects, in agreement with reports regarding actual execution. The overall average performance was 70.2% correct responses (S.D. 11.8%). Average performances and S.D. were also computed separately for each task, and were 71.8% (14.3%) in the verbal, 71.5% (11.4%) in the pictorial and 68.9% (15.5%) in the spatial task. Moreover, there was no significant difference in performance between tasks. Three subjects performed at the level of chance (-56%), and were later excluded from the analysis for a

L.F.H. Basile et al. / International Journal of Psychophysiology 48 (2003) 55–65

confirmation of results with the remaining 26 subjects. Performance remained not significantly different across tasks for this subgroup of subjects. All subjects presented SPs to all tasks, of typical maximum amplitudes of approximately 5 mV at the onset of S2. Fig. 1 shows an example of individual electrical data set. Fig. 2 shows topographic isopotential contour maps corresponding both to the same individual presented in Fig. 1 and to group average data, separated by task. Notice the complexity of voltage distribution on the scalp, as compared with more familiar sensory evoked potentials, for instance. In the case of group averaged data, potentials were computed by collapsing the data across subjects in the common montage, and are presented only for illustration purposes. Notice the reduction of amplitudes with respect to the typical, individual data, which is expected given the intersubject variability in the already complex topography of the SPs presently described. It is interesting to mention here that, although ICA was used only as a necessary step for defining ‘noise components’ for CDR, in all cases the first extracted component had the time course and scalp distribution virtually identical to the SPs. Typically, the second component corresponded to P300-like deflections and had negligible effect over the total potential distribution at the time point used for CDR (S2 onset). CDR results also had a complex form. In all subjects and conditions, we observed a complex, multifocal pattern of current density foci. Fig. 3 presents an example, from one individual, of the CDR results for the three tasks. Given such complexity and intersubject variability, as previously observed in our lab (Basile et al., 2002), visual inspection alone does not allow for easy conclusions regarding a common sets of active cortical areas across subjects, for each task. In any case, a few areas were clearly observed to most frequently contain non-negligible ()10% of maximum density) current density foci, when considering all tasks collapsed. For instance, extrastriate areas 18 and 19 and parietal area 7 bilaterally and left frontal area 6 were active in over 80% of the subjects (right area 18 in over 90%). Prefrontal areas 8, 9 and 10 bilaterally and right 6 were observed to contain current density foci in between

61

60 and 80% of subjects. This is itself by an indication of intersubject variability, since we are not even considering each task separately. It is interesting, however, that the most commonly active areas or areas with highest current densities were not necessarily the ones which differentiate between tasks, as will be seen below. The intersubject variability in active areas during task performance is illustrated in Fig. 4, where a randomly chosen subset of individual results are schematically shown. The planned comparisons testing for the effect of the three tasks, part of our MANOVA, resulted in the distinction of a few cortical areas as specifically related to each task. Thus, the verbal task corresponded to relatively increased electrical

Fig. 4. Schematic representation of CDR results in eight randomly chosen subjects (left hemisphere, verbal task), to illustrate intersubject variability, observed in all tasks. Black circle size is proportional to local current density (to activity scores of 1, 2 and 3). Circles at approximate centers of current foci in each cytoarchitectonic area.

62

L.F.H. Basile et al. / International Journal of Psychophysiology 48 (2003) 55–65

the overall small average scores in cases such as right areas 20 and 38, when compared with right extrastriate area 18 (not shown), which was active in 91.1% of subjects (all tasks collapsed), and had average scores 1.9 for verbal and pictorial tasks and 1.8 for the spatial task. The overall average scores, represented in shades of grey in each cytoarchitectonic area, are shown in Fig. 6, where the areas that distinguished the tasks are depicted by red circles. When more rigorous assumptions regarding our activity scores were used and a nonparametric analysis of variance (Friedman ANOVA and Kendall coefficient of concordance) was performed, similar results were obtained. However, by a change in statistical power, only right areas 45 and 38 were overall distinct in activity across the three tasks. Moreover, we performed a Mann–Whitney test on the pairs of task to verify the specificity of effects with respect to tasks. Whereas the right prefrontal area 45 was confirmed as specifically more active during the verbal task than the other

Fig. 5. Average current density scores (y-axis), computed across subjects, from the cytoarchitectonic areas, where scores significantly differed across tasks. Tasks indicated in x-axis: 1, verbal; 2, pictorial; 3, spatial.

activity in right area 45 as compared to the other tasks (F1,29s10.2, Ps0.003—pictorial; F1,29s 5.29, Ps0.028—spatial); the spatial task to decreased activity in right area 38 (F1,29s4.46, Ps0.043—pictorial; F1,29s8.43, Ps0.007—verbal) and left area 5 as compared to the other tasks (F1,29s4.7, Ps0.038—pictorial; F1,29s4.04, Ps 0.05—verbal); the pictorial task had stronger activity in left areas 9 and 39 than the spatial task (F1,29s3.77, Ps0.06; F1,29s4.37, Ps0.045); the verbal task had stronger activity in left 10 than the spatial (F1,29s5.2, Ps0.03) and weaker in right 20 than the pictorial task (F1,29s3.95, Ps0.05). Fig. 5 depicts the cortical areas that significantly differed across tasks, and shows their overall activity scores, averaged across subjects. Notice

Fig. 6. Average current density scores (on arbitrary grey-scale), computed across subjects, from all cytoarchitectonic areas; red circles indicate the areas presented in Fig. 4.

L.F.H. Basile et al. / International Journal of Psychophysiology 48 (2003) 55–65

two, right temporal pole area 38, significantly reduced in activity during the spatial task, distinguished only the spatial from the verbal task. Finally, left area 39, while not detaching itself across the three tasks, retained its quality of being significantly more active during the pictorial than the spatial task. Regarding the comparisons between activity in Brodmann areas across hemispheres, the verbal task corresponded to higher activity scores in the right hemisphere in frontal area 45 and extrastriate area 18; the pictorial task corresponded to higher activity on the left hemisphere, in area 39. Finally, when tasks were collapsed, i.e. the search for hemispheric asymmetries in activity common to the visual tasks resulted only in two areas: relatively higher activity in right area 18 and in left area 37. We also performed a nonparametric analysis of correlations (Spearman) between task performance and electrical activity scores in cytoarchitectonic areas, after eliminating the 3 subjects who performed at chance levels. Activity in right extrastriate area 19 and parietal area 7 had low positive correlations with performance on the verbal task (rs0.43; Ps0.029 and rs0.39; Ps0.049, respectively), activity in left frontal area 45 and right primary area 17 had low negative correlations with performance on the pictorial task (rsy0.41, Ps0.038 and rsy0.43, Ps0.027, respectively), whereas activity in left temporal area 21 correlated positively with performance on the pictorial task (rs0.52; Ps0.007). However, the significance of such correlations should be taken with caution, given the large number of comparisons (areas) performed. 4. Discussion The SPs obtained in all subjects and tasks were complex in distribution over the scalp, even more than what we recently observed for a task performance feedback stimulus anticipation SP (Basile et al., 2002). Both in that study and in the present case, the generators of SPs have a multifocal distribution comprising mainly association cortices from prefrontal and posterior areas. Qualitatively, this type of result agrees with the typical results

63

from metabolic tracing studies, when individual cases are analyzed, such as fMRI and occasionally PET (e.g. Cohen et al., 1996; Fink et al., 1997; Brannen et al., 2001). To test for a direct relation between metabolic changes such as in regional blood flow or in oxygenation levels, SPs would have to be simultaneously recorded on the same subjects performing a common task. Part of the complexity of our reconstruction results is due to intersubject variability in the sets of active cortical areas during each task. This type of variability is a common observation in studies using most types of tasks, excepting only the cases of simple voluntary sensorymotor activity (Fink et al., 1997). An unavoidable variability in task performance strategy across subjects is a commonly claimed explanation for variable cortical ensembles during any given task. If we conceive of nonsensorymotor visual tasks where a simple categorical comparison or domain-specific association has to be performed, it is hard to think of simpler cases than the present ones. Moreover, the choice of verbal vs. nonverbal domains, or pictorial vs. spatial domains among nonverbal material, is based on classical clinical observations and well established anatomical studies, instead of subtle psychological hypotheses. Since even here the intersubject variability in results is outstanding, either we are unable to devise association tasks that universally and specifically activate given association areas, which would mean that a profound revolution on task design is required, or ‘strategy’ variability is indeed an unavoidable human attribute. If the latter case were true, it would be important to observe some degree of variability among nonhuman primates as well, to parallel the continuity in transitions among species common to any physiological subsystem, including cortico-cortical interactions. Otherwise, the possibility will remain also that the available methods of imaging brain activity and metabolic changes during nonsensorymotor tasks suffer from limitations beyond our understanding. One other extreme possibility in the present type of study, however, is that the physiological correlates of each type of task do not really correspond to brain mechanisms specifically involved in task performance at all. In the present case, although SPs are known to reflect some form of ‘expecta-

64

L.F.H. Basile et al. / International Journal of Psychophysiology 48 (2003) 55–65

tion’ or preparation of the central nervous system, they could in principle have no relation to constructs such as association between stimuli, task– goal setting, memorization or anticipation of stimuli specific to a perceptual domain, for instance. Except for some recent studies such as the ones cited above, the last decades have produced a vast literature on task-related brain changes using results averaged across subjects. This literature review is beyond the scope of the present work, but it is important to mention here that even in those cases where interindividual variability is eliminated, there is usually no agreement on definitive functions of particular cortical association areas. As opposed to literature research defined by some hypothetical function, any unrestrained literature survey that centers on any given cortical association area will result in functional hypotheses difficult to reconcile with each other, and especially so with a general biological theoretical framework. Thus, even though we were able to observe two areas which consistently differed in activity, distinguishing themselves across the three presently used tasks after an analysis that respects interindividual variability, we believe it is best for the moment not to claim particular functions to those areas. Higher activity in right frontal area 45 distinguished the verbal, reduced activity in right temporal area 38 distinguished the spatial task, and higher activity in left area 39 the pictorial (from the spatial). For the sake of exemplifying the difficulties in reconciling functional hypotheses, we may take right area 45: in occasional individual cases, activity in this area was observed during a word-generation task (Brannen et al., 2001), whereas other authors claim that this area is related to episodic memory retrieval exactly when deemphasis of verbal strategies is adopted (McIntosh, 1999; Lepage et al., 2000); whereas it has also been observed in the Wisconsin card sorting test (Mentzel et al., 1998), it has been attributed to ‘discrepancies between signals from sensory systems in Luria’s bimanual coordination task’ (Fink et al., 1999). On the other hand, a minimum can be said regarding some of the most prevalently active areas across subjects, common to the three visual tasks:

the strong, bilateral activity in parietal area 7 and extrastriate areas 18 and 19 can be more easily reconciled with the literature, possibly representing the selective (spatial) attention to unspecific visual stimuli, which is a main correlate of the present SPs. Although visual attention appears to correspond to enhanced activity in various modality specific and unspecific areas, the extrastriate areas are the most consistently observed foci of such changes (Duncan et al., 1997). Studies on the role of parietal area 7 in selective attention (to interesting objects) have also been frequently replicated, since the first single cell electrophysiological studies on monkeys (Lynch et al., 1977; Yin and Mountcastle, 1978). The remaining areas with strong activity, prevalent across subjects and common to the tasks, in frontal areas 6, 8, 9 and 10, are examples, where uncontroversial statements on function cannot be offered. There is a possibility that a more universal pattern of electrical activity across individuals may be observed, in (some set of) association areas, if simpler tasks with respect to neural representation of stimuli are used. That is, it is possible that written words or the abstract figures here used are sufficiently variable in perceptual representation (posterior cortices) as is known to occur for language functions in brains of bilinguals, which would lead to more variability when corresponding prefrontal activity is considered, given the complexity of cortico-cortical connections. We are currently implementing a series of experiments, where we systematically control for this aspect, starting with detection of light dots or pure tones and even the mere temporal order of such type of stimuli, warned by equally simple cues. In case some universal activity pattern is found, we intend to verify at which step of task complexity (until the present type of matching-to-sample or pairedassociates task, where memory encoding, categorical anticipations and comparisons are performed), the intersubject variability is manifested. Finally, it is conceivable that in the near future, if large numbers of individuals are used for performance of given paradigms, the control for simple subject variables such as gender, age and education may also prove to reduce some of the variability as presently observed.

L.F.H. Basile et al. / International Journal of Psychophysiology 48 (2003) 55–65

Acknowledgments This research was supported by fellowship 98y 07640-3 from FAPESP, and grant 97y11083-0 ˜ Paulo, Brazil. from FAPESP, Sao References Basile, L.F.H., Rogers, R.L., Bourbon, W.T., Papanicolaou, A.C., 1994. Slow magnetic fields from human frontal cortex. Electroencephalogr. Clin. Neurophysiol. 90, 157–165. Basile, L.F.H., Simos, P.G., Brunder, D.G., Tarkka, I.M., Papanicolaou, A.C., 1996. Task-specific magnetic fields from the left human frontal cortex. Brain Topogr. 9(1), 31– 37. Basile, L.F.H., Brunder, D.G., Tarkka, I.M., Papanicolaou, A.C., 1997. Magnetic fields from human prefrontal cortex differ during two recognition tasks. Int. J. Psychophysiol. 27, 29–41. Basile, L.F.H., Ballester, G., Castro, C.C., Gattaz, W.F., 2002. Multifocal slow potential generators revealed by high-resolution EEG and current density reconstruction. Int. J. Psychophysiol. 45(3), 227–240. Bocker, K.B., Brunia, C.H., van den Berg-Lenssen, M.M., 1994. A spatiotemporal dipole model of the stimulus preceding negativity (SPN) prior to feedback stimuli. Brain Topogr. 7(1), 71–88. Brannen, J.H., Badie, B., Moritz, C.H., Quigley, M., Meyerand, M.E., Haughton, V.M., 2001. Reliability of functional MR imaging with word-generation tasks for mapping Broca’s area. Am. J. Neuroradiol. 22(9), 1711–1718. Brunia, C.H., 1988. Movement and stimulus preceding negativity. Biol. Psychol. 26(1–3), 165–178. Buchner, H., Knoll, G., Fuchs, M., et al., 1997. Inverse localization of electric dipole current sources in finite element models of the human head. Electroencephalogr. Clin. Neurophysiol. 102(4), 267–278. Cohen, M.S., Kosslyn, S.M., Breiter, H.C., et al., 1996. Changes in cortical activity during mental rotation. A mapping study using functional MRI. Brain 119(Pt. 1), 89– 100. Duncan, J., Humphreys, G., Ward, R., 1997. Competitive brain activity in visual attention. Curr. Opin. Neurobiol. 7(2), 255–261. Fink, G.R., Frackowiak, R.S., Pietrzyk, U., Passingham, R.E., 1997. Multiple nonprimary motor areas in the human cortex. J. Neurophysiol. 77(4), 2164–2174. Fink, G.R., Marshall, J.C., Halligan, P.W., et al., 1999. The neural consequences of conflict between intention and the senses. Brain 122(Pt. 3), 497–512. Fuchs, M., Wagner, M., Wischmann, H.A., et al., 1998. Improving source reconstructions by combining bioelectric and biomagnetic data. Electroencephalogr. Clin. Neurophysiol. 107(2), 93–111.

65

Fuchs, M., Wagner, M., Kohler, T., Wischmann, H.A., 1999. Linear and nonlinear current density reconstructions. J. Clin. Neurophysiol. 16(3), 267–295. Fuster, J.M., 1989. The Prefrontal Cortex. second ed.. Raven Press, New York. Hamano, T., Luders, H.O., Ikeda, A., Collura, T.F., Comair, Y.G., Shibasaki, H., 1997. The cortical generators of the contingent negative variation in humans: a study with subdural electrodes. Electroencephalogr. Clin. Neurophysiol. 104, 257–268. Hyvarinen, A., Oja, E., 2000. Independent component analysis: algorithms and applications. Neural Networks 13(4–5), 411–430. Ikeda, A., Luders, H.O., Collura, T.F., et al., 1996. Subdural potentials at orbitofrontal and mesial prefrontal areas accompanying anticipation and decision making in humans: a comparison with Bereischaftspotential. Electroencephalogr. Clin. Neurophysiol. 98, 206–212. Lepage, M., Ghaffar, O., Nyberg, L., Tulving, E., 2000. Prefrontal cortex and episodic memory retrieval mode. Proc. Natl. Acad. Sci. USA 97(1), 506–511. Lynch, J.C., Mountcastle, V.B., Talbot, W.H., Yin, T.C., 1977. Parietal lobe mechanisms for directed visual attention. J. Neurophysiol. 40(2), 362–389. McCallum, W.C., 1988. Potentials related to expectancy, preparation and motor activity. In: Picton, T.W. (Ed.), Handbook of Electroencephalography and Clinical Neurophysiology. Human Event-Related Potentials, vol. 3. Elsevier Science Publishers, pp. 427–534 revised series. McIntosh, A.R., 1999. Mapping cognition to the brain through neural interactions. Memory 7(5–6), 523–548. Mentzel, H.J., Gaser, C., Volz, H.P., et al., 1998. Cognitive stimulation with the Wisconsin Card Sorting Test: functional MR imaging at 1.5 T. Radiology 207(2), 399–404. Pandya, D.N., Yeterian, E.H., 1990. Architecture and connections of cerebral cortex: implications for brain evolution and function. In: Scheibel, A.B., Wechsler, A.F. (Eds.), Neurobiology of Higher Cognitive Function. The Guilford Press, pp. 53–84. Ruchkin, D.S., Sutton, S., Mahaffey, D., Glaser, J., 1986. Terminal CNV in the absence of motor response. Electroencephalogr Clin Neurophysiol 63(5), 445–463. Rugg, M.D., 1984. Event-related potentials in phonological matching tasks. Brain Lang. 23(2), 225–240. Walter, W.G., Crow, H.J., 1964. Depth recording from the human brain. Electroenceph. Clin. Neurophysiol. 16, 68–72. Walter, W.G., Cooper, R., Aldridge, V.J., McCallum, W.C., Winter, A.L., 1964. Contingent negative variation: an electric sign of sensorimotor association and expectancy in the human brain. Nature 203, 380–384. Yin, T.C., Mountcastle, V.B., 1978. Mechanisms of neural integration in the parietal lobe for visual attention. Fed Proc. 37(9), 2251–2257.