Knowing Left from Right: Characterizing Right Hemisphere Dominance for Spatial Attention via Combined EEG/fMRI

NeuroImage 54 (2011) 3010–3020

Contents lists available at ScienceDirect

NeuroImage

j ourna l homepage: www.e lsev ie r.com/ locate /yn img

Spatio-temporal indications of sub-cortical involvement in leftward bias ofspatial attention

Hadas Okon-Singer a,1, Ilana Podlipsky a, Tali Siman-Tov b, Eti Ben-Simon a,c, Andrey Zhdanov a,2,Miri Y. Neufeld c,d, Talma Hendler a,c,⁎a Functional Brain Center, Wohl Institute for Advanced Imaging, Tel Aviv Sourasky Medical Center, Israelb Cognitive Neurology Unit, Rambam Health Care Campus, Haifa, Israelc Sackler Faculty of Medicine, Tel Aviv University, Israeld EEG and Epilepsy Unit, Department of Neurology, Tel Aviv Sourasky Medical Center, Israel

⁎ Corresponding author. Functional Brain Center,Imaging, Tel Aviv Sourasky Medical Center, 6Weizmann+972 3 6973080.

E-mail address: [email protected] (T. Hendl1 Hadas Okon-Singer is now at the Department of Neu

Human Cognition and Brain Sciences, Leipzig, Germany2 Andrey Zhdanov is now at the Biomag Laboratory

Central Hospital, Helsinki, Finland.

1053-8119/$ – see front matter © 2010 Elsevier Inc. Aldoi:10.1016/j.neuroimage.2010.10.078

a b s t r a c t

Article history:Received 21 June 2010Revised 25 October 2010Accepted 27 October 2010Available online 5 November 2010

Keywords:Cross-correlationDynamic causal modeling (DCM)Inferior parietal sulcusInter-hemispheric transfer timePulvinarSimultaneous ERP–fMRI

A leftward bias is well known in humans and animals, and commonly related to the right hemispheredominance for spatial attention. Our previous fMRI study suggested that this bias is mediated by fasterconduction from the right to left parietal cortices, than the reverse (Siman-Tov et al., 2007). However, thelimited temporal resolution of fMRI and evidence on the critical involvement of sub-cortical regions inorienting of spatial attention suggested further investigation of the leftward bias using multi-scalemeasurement. In this simultaneous EEG–fMRI study, healthy participants were presented with face picturesin either the right or left visual fields while performing a central fixation task. Temporo-occipital event relatedpotentials, time-locked to the stimulus onset, showed an association between faster conduction from the rightto the left hemisphere and higher fMRI activation in the left pulvinar nucleus following left visual fieldstimulation. This combined-modal finding provides original evidence of the involvement of sub-corticalcentral attention-related regions in the leftward bias. This assertion was further strengthened by a DCManalysis designated at cortical (i.e., inferior parietal sulcus; IPS) and sub-cortical (pulvinar nucleus) attention-related nodes that revealed: 1. Stronger inter-hemispheric connections from the right to left than vice versa,already at the pulvinar level. 2. Stronger connections within the right than the left hemisphere, from thepulvinar to the IPS. This multi-level neural superiority can guide future efforts in alleviating attention deficitsby focusing on improving network connectivity.

Wohl Institute for AdvancedSt., Tel Aviv, 64239, Israel. Fax:

er).rology, Max Planck Institute for., HUSLAB, Helsinki University

l rights reserved.

© 2010 Elsevier Inc. All rights reserved.

Introduction

Healthy individuals show a reliable bias to the left visual field (LVF)in the perception of length, size, brightness, and numerosity known as‘pseudo-neglect’ (Bowers and Heilman, 1980). People with attentiondeficits show an opposite bias to the right (Hari et al., 2001), suggestinga role for the leftward bias in effective attention allocation. The leftwardbias was long related to the well-documented right hemisphere (RH)superiority for spatial attention (McCourt and Jewell, 1999), knownfrom lesion studies in humans showing higher prevalence and severityof spatial neglect following RH – compared to left hemisphere (LH) –

lesions (Heilman et al., 1987; Kim et al., 1999) and from functional

magnetic resonance imaging (fMRI; Corbetta et al., 1998) andtranscranial magnetic stimulation (Fuggetta et al., 2006) studies. Arecent fMRI study (Siman-Tov et al., 2007) combined with dynamiccausal modeling (DCM) analysis showed enhanced bilateral activationfor LVF stimulation that is best explained by stronger connectivity fromright to left parietal cortices than vice versa.

Classical models of spatial attention bias have underscored the roleof cortical rather than sub-cortical pathways (Corbetta and Shulman,2002). However, the fact that the leftward bias was discovered even inbirds (Diekamp et al., 2005) suggests an early evolutionary role, andthus possible involvement of sub-cortical regions in its operation. Twomain sub-cortical structures are likely to take part in such amechanism: (i) the mid-brain superior colliculus (SC), a criticalregion for orienting of attention and gaze (Boehnke and Munoz,2008). (ii) The thalamic pulvinar nucleus, involved in variousattention-related operations such as search, selection and engage-ment, and in the occurrence of spatial neglect in patients (Fairhallet al., 2009). Clearly, for effective attention operation, one expects aconcerted activation between high (i.e., cortical) and low (i.e., sub-cortical) levels of processing, especially for implicit processing that

3011H. Okon-Singer et al. / NeuroImage 54 (2011) 3010–3020

might be critical for survival. Animal studies (e.g., Burton and Jones,1975; Petersen et al., 1987) and recent diffusion tensor imaging (DTI)studies in humans point to substantial anatomical connectionsbetween these sub-cortical and attention-related cortical regionssuch as the parietal cortex (Leh et al., 2008), and lesion studies pointto their critical role in spatial attention operations (Karnath et al.,2002).

In order to further explore the role of sub-cortical regions in theformation of the leftward bias, we focused on two questions related toright hemisphere superiority: (i) Is the superiority confined to theattention-related cortical network or also evident at its sub-corticalnodes such as the SC and pulvinar nuclei? (ii) Is the superiority drivenonly by fast inter-hemispheric conduction from right to left, or also bya fast conduction within the right hemisphere between sub-cortical tocortical attention-related regions?

To answer these questions, we applied a combined electroenceph-alography (EEG)–fMRI approach. Simultaneous acquisition of EEG andfMRI was essential in the current study for two reasons. One istheoretical; it assured that the quality of stimulus processing (i.e.,amount of attention allocation) and subjective experiencewas equal forthe multi-scale brain measurements. Since our task involved implicitprocessing that might have changed with repeated stimulation, suchsimultaneous acquisition ensured the functional validity of our data.The other reason is methodological; sub-cortical regions cannot betraced by scalp EEG recording, and neural conduction time cannot bemeasured precisely with fMRI (see Debener et al., 2007 for details).Specifically, we used the superior temporal resolution of the EEG tounveil the exact temporal characteristic of inter-hemispheric transfertime based on a cross-correlation procedure while using the superiorspatial resolution of the fMRI to designate activation in sub-corticalregions that is best explained by the EEG-based timing of connectivity.

In order to examine the effective connectivity in the aforemen-tioned attentional network, a model-driven analysis was pursuedwith a DCM procedure (Friston et al., 2003). DCM is based on Bayesianrules to model task-independent intrinsic connectivity betweenregions, task-dependent modulations of these regions, and directinputs to the system. Hence, DCM assessesmutual influences of neuralregions, and how these influences are affected by the experimentalconditions (Friston et al., 2003; Penny et al., 2004). As an extension ofour previous work, the model tested inter-hemispheric conductionbetween the right and left pulvinar nuclei in addition to the right andleft intraparietal sulcus (IPS) (Siman-Tov et al., 2007). Furthermore,the model tested the intra-hemispheric conduction between thepulvinar nucleus and the IPS of each hemisphere separately.

Materials and methods

Subjects

Fifteen volunteers participated in this simultaneous EEG–fMRIstudy. Three participants were excluded from the final analysis afterreporting neurological disorders such as attention deficit hyperactivedisorder (ADHD) or due to left-handedness, and hence 12 right-handed healthy subjects were included in the analysis [3 males, meanage: 24.5 y (range: 21–30), mean education: 13.7 y (range: 12–16)].All had normal or corrected-to-normal vision. None had anyneurological or psychiatric symptoms, or any structural brainabnormality. Due to technical problems in EEG data acquisition, twosubjects were removed from the EEG study prior to data analysis. TheEthical Committee of the Tel Aviv Sourasky Medical Center approvedthe study and all subjects signed an informed consent form.

Visual stimuli and experimental design

Stimuli were black and white pictures of neutral faces andgeometric patterns of 3.7° (width)×4.7° (height). The face stimuli

were derived from those used by Siman-Tov et al. (2007), andincluded faces with a neutral expression taken from two databases:The Averaged Karolinska Directed Emotional Faces (KDEF) database(Lundqvist et al., 1998) and the Pictures of Facial Affect (Ekman andFriesen, 1976). The pattern stimuli were simple black and whitepatterns taken from a pool of pictures used in our laboratory forclinical purposes and the Internet. There were 40 pictures of faces and39 pictures of patterns, presented randomly with Psychtoolboximplemented in MATLAB 7.0.4 software.

The experiment included 4 consecutive sessions, each consistingof 181 repetitions (6.6 min), composed of blocks of left or right visualfields (LVF/RVF)×stimuli type (face/pattern) presented in a pseu-dorandom order (Fig. 1A). Each session started with 6 blank trials, afirst short block of house pictures that was discarded afterwards, and12 stimuli blocks of 10 events, separated by rest blocks of 3–5 blanktrials.

Fig. 1B illustrates the order of an experimental trial. The eventduration was 2.2 s, starting with a central gray fixation dot, presentedfor 500 ms. Following this, a picture of a face or a pattern waspresented parafoveally (5° angle from the center of the picture) to theright or left of the fixation, for 200 ms. Then, the color of the centralfixation dot was changed to either blue or red for 500 ms, andreturned to gray for 1400 ms for the remaining time of the event. Thepicture onset was shifted by a random jitter of 0–200 ms after the500 ms fixation; however, it always disappeared before the colorchange of the central fixation dot. In addition, the color change of thecentral dot was shifted by a random jitter of 0–300 ms after thepicture disappeared, to prevent time locking of the evoked neuralresponses to the gradient switching. The inter-trial interval wasadjusted according to the jitters of the picture onset and color changeonset, in order to achieve an event duration of 2.2 s.

Participants were asked to report the color change of the centralfixation dot via a response box using their right thumb for a red dotand left thumb for a blue dot. To achieve visual field segregation,participants were explicitly instructed to ignore the pictures and tomaintain fixation throughout the experiment.

Data acquisition

fMRI acquisitionMRI scans were conducted using a 3-Tesla GE scanner (Signa excite,

Milwaukee,WI, USA). All imageswere acquired using a GE four channelhead coil. The scanning session included conventional anatomical MRimages (T1-WI, T2-WI, T2-FLAIR), and three-dimensional spoiledgradient (3D-SPGR) echo sequence [field of view (FOV) — 250 mm;matrix size — 256×256; voxel size — 0.98×0.98×1]. Functionalimages included T2*-weighted images at the same locations as theanatomical images [FOV — 200 mm, matrix size— 64×64, voxel size—3.1×3.1×3.5, TR/TE/FA=2200/35/90, 33–34 axial slices (depending oninitial alignment) of 3.5 mmwithout gap].

EEG acquisitionContinuous EEG data was recorded simultaneously with fMRI

acquisition throughout the experimental sessions. EEG was collect-ed using an MR-compatible system including a 32-channel BrainCapelectrode cap with sintered Ag/AgCl ring electrodes (30 EEGchannels, 1 ECG channel, and 1 EOG channel under the left eye;Falk Minow Services, Herrsching- Breitbrunn, Germany; Fig. 2A) andBrainAmp-MR EEG amplifier (Brain Products, Munich, Germany).The reference electrode was placed between Fz and Cz. Raw EEGwas sampled at 5 kHz and recorded using Brain Vision Recordersoftware (Brain Products). Previous studies at our laboratoryshowed good signal-to-noise ratio of the EEG data in the combinedapproach (Ben-Simon et al., 2008; Sadeh et al., 2008).

Fig. 1. General description of the experiment. (A) Experimental design. Each experiment included 4 sessions, each consisting of 181 repetitions (6.6 min). Each session included 12stimuli blocks of 10 events, separated by rest blocks of blank trials. (B) Example of an experimental event depicting a picture of a face at the left visual field (2.2 s). Each trial startedwith a central gray fixation dot, presented for 500 ms; followed by a picture of a face or a pattern presented parafoveally to the right or left of the fixation for 200 ms; then the color ofthe central fixation dot was changed to either blue or red for 500 ms, and returned to gray for 1400 ms for the remaining time of the event.

3012 H. Okon-Singer et al. / NeuroImage 54 (2011) 3010–3020

Data analysis

fMRI analysisFunctional data were processed and analyzed using Statistical

Parametric Mapping software package (SPM5;Wellcome Departmentof Imaging Neuroscience, London, UK) with MATLAB 7.6.0. Preproces-sing included the following: removal of the first 22 s (first 6repetitions and the first short block) in order for the magnet toreach steady state, as well as to allow the subject to get used to thetask. Co-registered functional and anatomical images were normal-ized into standard Montreal Neurological Institute (MNI) space. Inaddition, we applied motion correction (realignment to the firstvolume), slice time correction (to the middle slice), temporalsmoothing (high-pass filtered at 1/128 s) and spatial smoothingwith a Gaussian kernel (FWHM=6 mm).

A basic statistical analysis was performed in order to locateRegions of Interest (ROIs) for the following DCM analysis and ERP–fMRI interactions. For this purpose, a low statistical threshold wasdetermined at a significance threshold of pb0.05 uncorrected and aminimum cluster size of 10 voxels. Statistical analysis was based onindividual maps of activation obtained from a general linear model,followed by a group analysis computed with random effect. Epochswere time-locked to onset of stimulus presentation. Regressorsmodeling stimulus events were convolved with a canonical hemody-namic response function (HRF). T-statistical maps were obtained bycontrasting hemodynamic responses during epochs of LVF and RVFstimuli presentation. As the pattern stimuli resulted in significantlylower activations compared to the face stimuli, we focused on theanalysis of the face stimuli. Note that subjects were asked to ignorethe pictures and respond according to a central color change. Faces areknown to receive prioritized processing (Kanwisher et al., 1996), and

hence it is reasonable that it was more difficult to ignore the facepictures than the pattern pictures, resulting in differences inactivation.

The statistical maps of face-LVF vs. face-RVF contrast of eachsubject were used to define our pre-determined ROIs in the rightpulvinar nucleus and the right IPS, while the statistical maps of theface-RVF vs. face-LVF contrast were used to define the equivalent ROIsin the LH. The selection of the ROI was based on the center ofactivation in these contrasts, which was closer to the coordinatesgiven by Siman-Tov et al. (2007). In cases where no such activationwas found, we defined an anatomical ROI. Themagnitude of activationwithin each ROI was based on the voxel of maximal activation, whichserved as the center of a 6 mm spherical volume. Although weconsidered the superior colliculus as part of the visual attention sub-cortical pathway, it was not defined as an ROI as no reliable activationwas found in this region across all subjects, probably due to the smallsize of this region. Percent signal change values of each ROI wereextracted for each subject. These percent signal changes in each ROIserved for the following analyses: (i) test for LVF superiority bycomparison of percent signal change values in the RH and the LH ineach ROI (i.e., the pulvinar nucleus and the IPS), (ii) calculatingcorrelations between the fMRI and the event related potentials(ERPs), and (iii) examining an effective connectivity model with DCM.

ERP analysisERP analysis was done offline with EEGLAB 6.01 software package

(Schwartz Center for Computational Neuroscience, University ofCalifornia, San Diego), MATLAB software and FMRIB plug-in forEEGLAB. We preprocessed the data in the following stages: (i) MRgradient artifacts were removed using a FASTR algorithm (Universityof Oxford Centre for Functional MRI of the Brain; (Christov, 2004; Kim

Fig. 2. ERP analysis. (A) Schematic drawing of the 30-electrode cap used in the study. The three electrodes overlying the posterior temporo-parietal scalp in the RH and the LH arecircled in solid and dashed lines, respectively. (B) and (C) Example of the cross-correlation calculation between ERP in the RH and the LH from an individual subject. Maximumcorrelation was extracted from a graph of ERP correlation as a function of the lag between the ERPs. The maximum correlation is marked with an arrow. Information regarding LVFstimuli is transferred from the RH to the LH (A) while information regarding RVF stimuli is transferred from the LH to the RH (B). (LH— left hemisphere, LVF— left visual field, RH—

right hemisphere, RVF — right visual field).

3013H. Okon-Singer et al. / NeuroImage 54 (2011) 3010–3020

et al., 2004). (ii) Cardio-ballistic artifacts were removed. (iii) Datawere filtered using a 30 Hz low-pass filter. Following the preproces-sing, the data was manually inspected for the presence of eyemovements and blinks at electrodes F7, F8, FP1 and FP2. Eyemovements and blinks were detected in less than 2% of the trials.The number of trials containing eye movements and blinks did notdiffer between LVF and RVF stimulation.We also verified that the ERPsin each electrode showed a reliable and expected pattern of P1, N1and P300 peaks (see example of grand averages in Fig. 1, Supple-mental material).

ERP signals were segmented to an interval of 0–400 ms followingstimulus onset, focusing on the face stimuli. Baseline correction wasperformed by subtracting the average of the signal in a time windowof 250 ms before stimulus onset from each epoch. For each subject,electrode and experimental condition, segmented ERPs were aver-aged across all trials. The preprocessed waveforms were averagedover three electrodes overlying the posterior temporo-parietal scalpin each hemisphere; P4, P8, CP6 for RH, and P3, P7, CP5 for LH (seeFig. 2A). These electrodes were chosen since they showed a distinctand reliable ERP waveform that has previously been shown to berelated to perception of visual stimuli, including N1 and P1components (Hopf and Mangun, 2000; Martínez et al., 2006; Nataleet al., 2006).

In order to estimate the propagation of neuronal processing,lagged cross-correlations were computed across electrodes. Fouraveraged ERP waveforms were prepared per subject, representing thecombination of two possible visual fields (RVF/LVF) and two hemi-spheres (LH/RH) (e.g., LVF/LH, LVF/RH, RVF/RH, RVF/LH; see Figs. 2Aand B). Thus, we prepared 40 single waveforms (10 subjects⁎4waveforms for each subject), and each waveform consisted of anaverage of 120 trials (4 sessions⁎30 trials each). The generalwaveform was similar in these four conditions; hence, it was possibleto use a cross-correlation for each two ERPs to reveal the delaybetween the responses. Cross-correlation estimates the time delaybetween twowaves, based on the entiremorphology of the wave. Thisapproach does not rely on a specific peak latency or amplitude, andthus was ideal for the current study in which we did not have aspecific hypothesis regarding a particular ERP, such as the P1 or N1.Importantly, results based on a specific peak may be less reliable thana combination of several peaks, especially in a simultaneous ERP–fMRIdesign in which the signal-to-noise ratio is lower compared to astand-alone ERP design (Fries et al., 2007; König et al., 1995;Womelsdorf et al., 2007).

Cross-correlation was applied between RH and LH following eitherLVF stimulation or RVF stimulation. The lag which had the maximumcross-correlation between the ERPs was taken to be the transfer time

3014 H. Okon-Singer et al. / NeuroImage 54 (2011) 3010–3020

between hemispheres (i.e., inter-hemispheric transfer time, IHTT).IHTT for RH to LH was measured following LVF stimulation and for LHto the RH following RVF stimulation. In order to overcome individualdifferences in conduction time, we further calculated an index forconduction advantage based on: (lagged cross-correlation from RH toLH following LVF stimulation)−(lagged cross-correlation from LH toRH following RVF stimulation). This index represents the individualadvantage in transfer time from the RH to the LH, as opposed to thetransfer time from LH to RH. The greater this index is, the faster thetransfer time from the RH to the LH following LVF stimulationcompared to the opposite transfer from LH to RH following RVFstimulation. This index was used for the correlations between thefMRI and the ERP findings. Our analysis aimed to detect temporalcharacteristics based on phase differences between the experimentalconditions. One should note that spatial characteristics may also differbetween experimental conditions, due to differences in the sourcesthat contribute to the ERP signal. These spatial changes are not known,and may also result in phase differences (Srinivasan et al., 2006).

ERP–fMRI correlations

We correlated the ERP and fMRI measures of each subject in orderto examine the hypothesis that accelerated conduction from the RH tothe LH is related to enhanced activation in attention-related regions inthe LH. For that, we calculated the correlation between the ERP-basedconduction index and the fMRI-based percent signal change in pre-defined ROIs in the LH that included the pulvinar nucleus and the IPS.We further computed the correlation between averaged ERP laggedcross-correlation delays following RVF stimulation and fMRI percentsignal changes in the right hemisphere. In order to overcome theproblem of multiple comparisons, we divided the p value by 4, sincewe compared two hemispheres×two regions.

DCM analysis

DCM (Friston et al., 2003) was used to estimate the strength ofconnections between neural regions involved in the proposed modelfor spatial attention bias. Functional time-series of the BOLD signalextracted from the right and the left IPS and from the right and leftpulvinar nuclei of each session and each subject were used in the DCManalysis. Two alternative models were compared: (i) contralateraland ipsilateral input to the right pulvinar, and only contralateral inputto the left pulvinar (i.e., the ‘classical’ model), (ii) selectivecontralateral input, i.e., RVF input to left pulvinar, and LVF input toright pulvinar (as was suggested by Siman-Tov et al., 2007; Fig. 3A).The latter model is based on a strong a priori assumption that theinput to each pulvinar nucleus is contralateral (please see discussionfor an alternative model based on similar input to each node). Bothmodels assumed bidirectional connections between the right and theleft pulvinar, as well as between the right and the left IPS andconnections from the pulvinar to the IPS within each hemisphere(Fig. 3B). Thus, themodels differed only in their hypothesized input tothe RH: contralateral and ipsilateral input assumed in the classicmodel, compared to merely contralateral input in the revised model.Modulatory effects were not hypothesized.

Model comparisonwas performed for each session separately witha Bayesian model selection procedure (Penny et al., 2004). Modelpreference was computed based on Bayesian and Akaike's informa-tion criteria, using the ratio between probabilities of the measureddata, given eachmodel. Bayes factor (BF)was defined as theminimumof these two criteria. When BFs were N1, the data favored model 1(selective contralateral input) over model 2 (bi-hemifield represen-tation in the right IPS); when BFs were b1 then the data favoredmodel 2. A BF of at least e (2.7183) was regarded as consistentevidence in favor of model 1 (Penny et al., 2004).

In addition, for the model that was preferred by the DCM analysis,we examined the difference in the strength of connections betweenhemispheres. This was done in the two following steps; first, weexamined that each connection exists significantly at a group level.For that, the probabilistic values of each modeled connection wereentered into a T-test aimed at verifying that the probability of theconnection across subjects and sessions is higher than a chance levelof 0.5. Connections that did not reach significance at a group levelwere not included in the following analysis. Subsequently, for thesignificant connections, values of intrinsic connectivity wereextracted for each session from the subject-specific DCM, and wereentered into two-way repeated-measures ANOVA (factors: directionof connection, session number). For the purpose of the ANOVAanalysis, specific connections that received a probability value thatwas lower or equal to 0.5 in the subject's specific DCM were assumedto represent a connection that is not reliable in a session, and hencewere not included in the following ANOVA. In these cases, bothconnections in the specific comparison received a value of zero inorder to avoid creating a matrix with empty cells.

Results

fMRI

As expected from our previous study (Siman-Tov et al., 2007),whole brain analysis of the face-LVF vs. face-RVF contrast revealedcontralateral activation in low visual areas, including Brodmann Areas(BAs) 17 and 18 (Fig. 4A), and higher bilateral activation in anattention-related network for LVF than for RVF. Specifically, LVFstimulation activated bilateral cortical and sub-cortical regions, mostsignificantly in the following regions: superior parietal lobule, IPS,pre-supplementary and supplementary motor area, dorsolateralprefrontal cortex, supramarginal gyrus, insula, pulvinar nucleus,basal ganglia, and brainstem (see Fig. 4B and Table 1).

ROI analysis was performed in order to estimate the magnitude ofthe observed LVF superiority in cortical and sub-cortical ROIs (withthe selected regions being those responding to contralateral visualfield stimulation in each hemisphere). As expected from our previousstudy (Siman-Tov et al., 2007), LVF advantage was found bilaterally inthe IPS, but here we additionally show that it already exists at thepulvinar nuclei level. Please note that the ROIs were localizedindependently of the main effect of hemisphere. The contrasts thatwere chosen to pick ROIs were different for left and right hemi-spheres. For the right hemisphere, the contrast was LVFNRVF, whilefor the left hemisphere, the reverse contrast of RVFNLVF wasperformed. Thus, the main effect of visual field did not result fromthe contrast used. We focused on the pulvinar nuclei and not the SCsince we did not find activation in the SC at the group level. Two-wayrepeated-measures ANOVA (factors: hemisphere, visual field) doneseparately for each ROI revealed a visual field main effect {LVFdominance: [F(1,11)=7.54, pb0.01] and [F(1,11)=6.13, pb0.03], forthe IPS and the pulvinar, respectively}. No significant interaction wasfound between hemisphere and visual field for all these regions (allFsb1). See example for the LVF advantage in the pulvinar in Fig. 4C.

ERP

Using cross-correlation with temporo-parietal ERPs to faces, weexamined the transfer time from the RH to LH following stimulation inthe LVF (average: 19.6 ms, SE=6.8), and from the LH to RH followingRVF stimulation (average: 25.7 ms, SE=4.2). Although the conduc-tion times seem numerically faster from right to left than vice versa,the comparison between these measures did not reach statisticalsignificance (Figs. 2A and B Supplemental material).

Fig. 3. Schematic models for spatial attention: (A) model for attention asymmetry based on current findings: both hemispheres receive visual input from the contralateral hemifield.(B) For both models tested in the DCM, we hypothesized bidirectional connections between the right and the left IPS, and between the right and the left pulvinar. In addition, bothmodels assume connections between the pulvinar and the IPS in each hemisphere. These connections were common for both models, which differed only in the modeled input. Forvisualization purposes, we drew these connections in a different sub-figure. (C) DCM analysis on the fMRI data showed privileged inter-hemispheric conduction from RH to LH(0.04 Hz compared to 0.004 Hz), already existing in the sub-cortical pulvinar nuclei, and stronger transfer within the RH than the LH (0.04 Hz compared to 0.0005 Hz). The numbersrepresent the strength of the intrinsic connections (in Hz) based on the DCM analysis, and are calculated as an average of the strength of all the subjects in all the sessions fordemonstration purposes. (DCM — dynamic causal modeling, IPS — intra parietal sulcus, LH — left hemisphere, RH — right hemisphere).

3015H. Okon-Singer et al. / NeuroImage 54 (2011) 3010–3020

fMRI and ERP

In order to further explore the relation between the timing ofinter-hemispheric connectivity and sub-cortical activation, we corre-lated the ERPs latencies and fMRI activation in the pulvinar and IPSfollowing LVF or RVF stimulation. A significant correlation wasrevealed only between the BOLD activation in the left pulvinarfollowing LVF stimulation, and the ERP-based IHTT for conductionfrom RH to LH (r pearson=0.82, pb0.004; see Fig. 4D). Ashypothesized, the opposite correlation between activation in theright pulvinar and the IHTT of LH to RH was not significant (rpearson=−0.2, pN0.5). Thus, activation located by fMRI in the leftpulvinar nucleus was uniquely related to faster conduction from RH toLH as measured by the scalp ERP.

Model testing

Following our previous results with DCM based on fMRI activation(Siman-Tov et al., 2007), we applied a similar analysis, but with a newset of ROIs that include cortical and sub-cortical regions: the IPS andthe pulvinar nuclei. The comparison between the classical model andthe revised model for each subject confirmed our hypothesis in thisstudy: each hemisphere receives input only from the contralateralvisual field, in contrast to the classical model that is based on bilateral

input to the RH. The DCM analysis favored the revisedmodel in 35 outof 48 sessions (Bayes factor range: 5.83–13.67; mean BF=7.93), 10sessions showed no consistent evidence for either model, and theremaining 3 sessions favored the classical model. The overall BF acrosssessions and subjects significantly supported the first model (one-tailed sign test, pb0.001).

T-tests examining probability of connection being higher than 0.5(chance level for a random event with two equally probableoutcomes) showed significant (pb0.05) connections from the rightpulvinar to the left pulvinar, from the left pulvinar to the rightpulvinar, from the right pulvinar to the right IPS, and from the leftpulvinar to the left IPS. The connections from the right IPS to the leftIPS were not significant (all psN0.1) and thus were not included infurther analyses. Specific connections that received a probability valuethat was lower or equal to 0.5 in the subject's specific DCM (19 out of192 connections) were assumed to represent a connection that is notreliable in a session, and hence were not included in the followingANOVA. Values of intrinsic connections between the right and the leftpulvinar nuclei were entered to 2-way repeated-measures ANOVA(factors: direction of connection, session number). The resultsrevealed stronger connection from the right to the left pulvinar thanvice versa [F(1,11)=13.75, pb0.003]. Finally, the DCM was used toestimate the strength of connections from the pulvinar to the IPS ineach hemisphere. T-tests examining probability of connection that is

Fig. 4. fMRI analysis showing leftward bias in cortical and sub-cortical regions. (A) and (B) fMRI statistical maps of the LVF faces stimulation (shown in red) vs. RVF faces stimulation(shown in blue) contrast (N=12, random effects analysis, pb0.05). The maps show contralateral activation in low-level visual areas (A), and striking advantage of LVF stimulation inbilateral attention-related regions (B), replicating Siman-Tov et al.'s (2007) findings. (C) ROI analysis of bilateral activation in the pulvinar nucleus revealed significant LVF advantagein both hemispheres. The contrasts used to localize the ROIs were independent from the main effect of visual field. For the right pulvinar, we used the contrast LVFNRVF, and for theleft pulvinar, we used the reverse contrast of RVFNLVF. (D) Faster conduction from the RH to the LH was correlated with higher activation in the left pulvinar nucleus. Note that theconduction time is based on ERP data, while the activation is based on fMRI data. Each data point represents a single subject. (BOLD — blood oxygenation level dependent, EEG —

electroencephalography, L — left, LVF — left visual field, OTC — occipito-temporal cortex, ROI — region of interest, R — right, RVF — right visual field, V1 — primary visual area).

3016 H. Okon-Singer et al. / NeuroImage 54 (2011) 3010–3020

higher than chance level (0.5) for each session showed significantconnections from the right pulvinar to the right IPS (all psb0.02) andfrom the left pulvinar to the left IPS (all psb0.03). A two-wayrepeated-measures ANOVA on the values of intrinsic connectionsfrom the pulvinar to the IPS within the RH or the LH (factors:hemisphere, session number) revealed stronger connection in the RHcompared with the LH [F(1,11)=4.67, pb0.05] (Fig. 3C).

Discussion

Simultaneous EEG–fMRI showed a positive correlation betweenright to left IHTT measured by scalp ERPs, and neural activation in the

Table 1Brain regions showing significant activation in the LVF vs. RVF contrast (n=12, random effectsMontreal neurological institute, RH — right hemisphere, SMA — supplementary motor area, SM

Location RH

MNI coordinates

x y z p value t va

SMA 6 −15 63 5×10−3 3.06Cingulate 3 9 33 1×10−3 4.11IPS 12 −63 57 1×10−4 5.37SMG 60 −21 24 1×10−3 4.10STC 63 −27 6 1×10−3 3.89Insula 39 0 6 1×10−3 4.25Hippocampus 24 −21 −15 1×10−4 5.72Pulvinar 18 −30 −3 1×10−4 4.97Pons 9 −39 −36 1×10−3 4.40Cerebellum 6 −63 −33 2×10−3 3.54

left pulvinar estimated by fMRI. Furthermore, analysis based on DCMof fMRI activation in cortical and sub-cortical nodes of the attentionnetwork demonstrated that a conduction advantage already occurredat the left pulvinar; a major sub-cortical node of visual attention inhumans. In addition, intra-hemispheric connections from the pulvinarto the IPS were stronger in the right than in the left hemisphere. Thesefindings replicate and further expand our prior fMRI-based DCMresults, which were based only on the cortical node (i.e., IPS) of thisattention network in the two hemispheres (Siman-Tov et al., 2007).Therefore, we suggest that the known leftward bias is mediated by: 1)faster information transfer between the right and the left hemispherethan vice versa, already at a sub-cortical level, resulting in effective

, pb0.05, cluster sizeN10 voxels). IPS— intraparietal sulcus, LH— left hemisphere, MNI—G — supramarginal gyrus, STC — superior temporal cortex.

LH

MNI coordinates

lue x y z p value t value

−3 9 45 1×10−4 6.30−3 −15 36 1×10−3 3.87

−39 −42 45 2×10−3 3.57−60 −33 24 1×10−3 4.41−63 −9 6 1×10−4 6.67−39 3 9 1×10−3 4.14−18 −18 −24 2×10−3 3.75−9 −36 21 6×10−3 3.05−3 −33 −30 1×10−4 4.50

0 −63 −39 5×10−3 3.06

3017H. Okon-Singer et al. / NeuroImage 54 (2011) 3010–3020

recruitment of left attention-related regions even for stimulipresented in the LVF; and 2) faster information transfer within theright hemisphere than the left hemisphere, resulting in effectivedistribution of information from sub-cortical to cortical nodesfollowing LVF stimulation.

The trend for the biased right to left IHTT, while not significant,was still similar inmagnitude towhat has usually been shown in otherstudies (e.g., Barnett and Corballis, 2005) using different types ofstimuli and modality, including checkerboards (Barnett and Corballis,2005), 3-letter words (Nowicka and Fersten, 2001) and single letters(Moes et al., 2007). Further, various stimuli were presented in visualor auditory modalities (Elias et al., 2000) and involved several tasks,including letter matching or reaction to the location of the target.Thus, the IHTT effect can neither be regarded as specific to one stimulior another, nor to a task. Taken together, we believe that the weakIHTT bias might be due to noise introduced to the ERPs by thechallenging conditions of our study (i.e., a simultaneous EEG–fMRIapproach and a task that is not directed to focus on the stimuli inquestion). This challenging approach was, however, necessary forexamining the possibility that a relatively low-level connectivity biasunderlies the IHTT bias effect.

The role of sub-cortical structures in intact spatial attention

The evolutionary role of low-level neural processing in attentionbias is implied by the finding of a leftward bias in birds (Diekamp etal., 2005), which lack a corpus callosum. In humans, recent DTI studiesshowed evidence of anatomical connections between the SC, thepulvinar nucleus and cortical visual and attention-related regions,including the IPS, the posterior parietal cortex, the frontal eye fieldsand the prefrontal cortex (Leh et al., 2008; Rushworth et al., 2006).These connections further support our claim regarding the centralinvolvement of sub-cortical regions such as the pulvinar nuclei inspatial attention bias. In line with our findings, neuro-imaging studiessuggest that the pulvinar nucleus is related to serial search (Fairhallet al., 2009), covert shifts of attention (Nobre et al., 1997), andattentional selection (Snow et al., 2009). Moreover, lesion studies inhumans showed the critical role of the pulvinar in engagement ofattention (Rafal and Posner, 1987), spatial and temporal componentsof feature binding (Arend et al., 2008), and visual search (Ward andArend, 2007). The critical role of the pulvinar in these different aspectsof spatial attention is in line with its suggested role in the leftwardattention bias. These findings suggest that the pulvinar may be criticalfor the initiation of efficient orienting of attention in space.Interestingly, precise topographic encoding was recently demonstrat-ed in the right pulvinar (Fischer and Whitney, 2009), confirming thatthis nucleus may be involved in basic spatial coding crucial to spatialorienting. Importantly, in line with the neurological model of theleftward spatial attention bias, lesions in the left pulvinar resulted incontralateral spatial neglect (Arend et al., 2008; Karnath et al., 2002).Clearly, the pulvinar may also receive top-down cortical modulation,in addition to being an important node in the attentional pathway tothe cortex in a bottom-up fashion. Importantly, our DCM analysispointed to stronger connections from the pulvinar to the IPS, and notvice versa. Thus, our findings point to the causal role of the pulvinar inmediating the leftward bias in spatial attention.

It should be mentioned that our DCM analysis did not findsignificant functional connections between right and left IPS areas, butonly between the right and left pulvinar nuclei. However, a DCManalysis of our results while modeling only the two IPS replicated thefindings of Siman-Tov et al. (2007) showing stronger connectionsfrom the right to the left IPS. Thus, our new DCM results based on anextended network reveal the importance of sub-cortical rather thancortical transfer in the leftward spatial attention bias.

In the current study, we focused on the pulvinar nucleus, sincethere are proven direct anatomical connections between the pulvinar

and the IPS in animals (Matsuzaki et al., 2004; Schmahmann andPandya, 1990). Thus, it was reasonable to model these two attention-related nodes in the DCMmodel and further analyses. In addition, thepulvinar showed significant activation at a group level in our study.Although not included in the current model, the SC may play a crucialrole in the leftward bias. The SC is the first node in visual perception,and is known to have an essential role in the allocation of spatialattention (Boehnke and Munoz, 2008; Leh et al., 2006; Müller et al.,2005; Sapir et al., 1999; Schneider and Kastner, 2009). Thus, it isreasonable to assume that the leftward bias originates at thisbrainstem node. This issue may be examined in fMRI studies withan optimized design for activating the SC, for example usingemotional stimuli. Indeed, in a previous fMRI study, we presentedfaces with a fearful or neutral expression in the left or right fields. Wefound a LVF advantage for fearful faces compared to neutral faces inthe amygdala, the pulvinar and the SC. This differential activation wasnot found following RVF stimulation. In line with the current findings,these results suggest that the leftward bias in the processing of fearalready starts in sub-cortical regions (Siman-Tov et al., 2009).

The role of connectivity biases in spatial attention leftward bias

Our DCM analysis suggested stronger connections from the RH tothe LH, as well as within the RH. These findings emphasize the role ofefficient connectivity within the attentional neural networks for intactspatial orienting. Recent DTI studies point to the importance of whitematter connections in spatial biases, showing correlation betweendisconnection of right fronto-parietal white matter bundles and theseverity of neglect (He et al., 2007). In addition, conduction deficits inventral and dorsal visual pathways were found in spatial neglect(Committeri et al., 2007; Gabrieli and Whitfield-Gabrieli, 2007; He etal., 2007). The importance of intact connectivity to intact attentionaloperation is strengthened by evidence that neglect may result fromdysfunction of distributed spatial attention networks at cortical andsub-cortical levels within the right hemisphere (Heilman et al., 1993;Mesulam, 1981; Posner and Petersen, 1990). Furthermore, structurallesions in a ventral attention network, centered around the temporo-parietal junction and ventral frontal cortex, resulted in spatial neglectand functional imbalance in bilateral dorsal parietal regions that werestructurally intact (Corbetta et al., 2005; He et al., 2007). Finally, thereis evidence that intra-operative direct stimulation of the rightsuperior occipito-frontal fasciculus resulted in deviation in a line-bisection task (deSchotten et al., 2005).

Connectivity within the attentional system has been investigatedfor decades, and asymmetric inter-hemispheric conduction wassuggested long ago (Poffenberger, 1912). In the classic behavioraltask, participants are required to respond with either the left or righthand to visual stimuli appearing in either the LVF or the RVF. It hasbeen shown that the difference between the RVF-right hand and LVF-right handwas less than the difference between the LVF-left hand andRVF-left hand, implying faster transfer from the right to lefthemisphere than vice versa. ERP studies with better temporalresolution replicated these findings (Barnett and Corballis, 2005).Several differences between our task and the Poffenberger task allowfor generalization of the known inter-hemispheric transfer timeadvantage of left targets: (i) we presented face pictures whileprevious studies mainly used checkerboard stimuli, thus our designbest fits a higher-level natural visual processing. (ii) The implicitaspect of our task might be more sensitive to reflect sub-corticalprocessing as indicated indeed by the simultaneous fMRI.

Interestingly, a recent ERP study (Verleger et al., 2008) comparedrapid presentations of targets in the right and left visual fields, andshowed earlier latencies and larger amplitudes of peaks in the RHcompared to the LH. These findings suggest that the RH is faster thanthe LH in singling out a relevant target. In addition, sustainedcontralateral slow shifts were longer at the RH compared to the LH,

3018 H. Okon-Singer et al. / NeuroImage 54 (2011) 3010–3020

suggesting better visual working memory. Differences in speed oftransfer between hemispheres can be related to more neuronsprojecting to the left hemisphere than vice versa (Marzi et al.,1991), to hemispheric specialization (Nowicka et al., 1996), or togreater activity in the RH resulting from a larger number of fast-conducting, myelinated axons in the RH compared to the LH (Barnettand Corballis, 2005). It is yet to be revealed whether these strongerconnections as described in our study by the fMRI–DCM result from alarger number of neurons or from faster conduction in a similarnumber of neurons.

Studies using DCM analysis are often based on similar inputentering several nodes in the network. In contrast, our model is basedon a strong a priori assumption that the input to each hemisphere iscontralateral. A more traditional model would depict similar input(composed of both LVF and RVF input) entering both the left and rightpulvinar nuclei. Intrinsic connections should be modeled between thetwo pulvinar nuclei and between each pulvinar and the ipsilateral IPS,and internal connections of each pulvinar nucleus. In order to allowmodulations depending on the experimental condition, this modelincludes modulatory connections from one visual field to each of thepulvinar nuclei, to the intrinsic connections from the right to the leftpulvinar, and to the intrinsic connections from the left to the rightpulvinar. This model should account for the effect of diminishedconnections from each visual field to the ipsilateral pulvinar byintroducing a modulatory effect of the corresponding visual field. Thismodeling approach is more traditional; however, it would force us tomodel connections from each visual field to the correspondingipsilateral region. Such connections, which do not necessarily existin the data would weaken the model. Indeed, Bayesian comparison ofthis alternative model to the one we used in the data analysis stronglysupported the originalmodel (BF=8E12, in favor of themodel used inthis study).

One should note that the DCM analysis is a hypothesis-drivenmethod that is used to test a specific hypothesis that is derived fromthe paradigm (Friston et al., 2003). While we focused on attentionalmechanisms and relevant theories, it is possible that other theoriesthat were not tested in the current study would also explain theresults. Finally, as pointed out by Siman-Tov et al. (2007), one canargue that the faster conduction found in our study demonstratesprocesses related to disengagement of attention from a centralfixation point, inhibition of reflexive disengagement, preparation ofeye movements, and changes in alertness. However, as theseprocesses overlap anatomically and functionally with mechanismsof covert spatial attention, our suggested model can further accountfor each of these processes. Future studies may disentangle these sub-processes via different tasks.

Possible shortcomings and future directions

One possible limitation of our study is the fact that our analysisfocused on the faces condition. Thus, it can be argued that our resultsreflect face-specific processes. In particular, there is evidence of a leftvisual field advantage for face stimuli (Geffen et al., 1971; Leeheyet al., 1978). However, several arguments support the suggestion thatthe findings are related to attention and not specific to face-processing. Most importantly, the LVF advantage was also found inour data for the pattern's condition (see Methods). Specifically, abilateral LVF advantage was found in the IPS for pictures of patterns(Supplemental material, Fig. 3A). As hypothesized, contralateralactivation was shown in primary visual areas (Supplemental material,Figs. 3B and C). Our analysis was based on the face condition since thepatterns evoked less consistent ERP features (i.e., P1 and N1) relativeto faces (i.e., N170). It should be noted that there were several visual–perceptual differences between the face and pattern stimuli that canaccount for the less robust results, for example, visual diversity (oneface compared to several small figures or lines), variability between

items (faces vs. completely different types of patterns), and the factthat faces are more interesting and relevant than patterns. Furthersupport for the view that LVF is not a face-specific phenomenoncomes from previous work with fMRI that showed a leftward bias inneural activation to house pictures (Siman-Tov et al., 2007).Additionally, the LVF advantage in the current and previous studywas found in several typical attention-related regions, including thebilateral intraparietal sulcus, superior parietal lobule and superiortemporal sulcus, rather than in face-specific processing areas.Importantly, LVF bias was found at these attention-related regionsin a high threshold (minimum p=2×10−3, uncorrected, clustersize=10 voxels, see Table 1). Thus, although the threshold used forthe individual DCM and correlation analyses was not high, LVF biaswas shown at a high threshold in attention-related regions at a grouplevel, replicating previous findings (Siman-Tov et al., 2007, 2009). TheDCM analysis further showed stronger connections to the IPS, which isknown to be involved in orienting of attention. Accordingly, we choseto perform the EEG analysis for electrodes located in parietal sitesrather than temporal electrodes that could have been closer to face-processing regions such as the fusiform face area.

Eye movement could count for some aspect of the findings,however we believe that they were not significant in the case ofour study for several reasons: (i) our design was similar to that ofSiman-Tov et al. (2007), who monitored eye movements online usingan eye-tracking device, and did not find evidence of significantdifference in eye movements between LVF and RVF presentationconditions. (ii) As mentioned above, we manually inspected electro-des F7, F8, FP1 and FP2 for the presence of eye movements and blinks,which were detected in less than 2% of the trials. (iii) Previous studiesshowed that central fixation can be maintained if subjects areinstructed to avoid eye movements to peripheral items, particularlywhen items are irrelevant to the experimental task (Gitelman et al.,1999). (iv) The bilateral advantage for LVF stimulation was not foundin low visual areas that mainly enhanced activation to the contralat-eral field stimulation. These low visual areas should be most sensitiveto eye movement differences.

In summary, our suggested model for spatial attention bias relieson both faster inter-hemispheric bias already at a sub-cortical leveland efficient intra-hemispheric transfer within the right hemisphere.This revised model adds to the model suggested by Siman-Tov andcolleagues, suggesting the involvement of sub-cortical structures inthe leftward bias, and showing intra-hemispheric transfer bias.Although we examined hemispheric bias with regard to spatialattention, our findings may be relevant to other cognitive functionscharacterized by laterality, including language and emotional proces-sing (Davidson et al., 2004; Springer et al., 1999). In addition, thedesign we used is similar to the brief appearance of stimuli inperipheral locations in standard paradigms examining spatial atten-tion (Posner, 1980). However, as we used a block design, our findingscan also be related to a voluntary orientation of attention to a certainenvironmental location. Interestingly, it has long been proposed thatthe RH is superior for global processing of objects while the LH issuperior for local processing (Fink et al., 1997). Based on our results, itis conceivable to speculate that the RH is dominant for rapid, globaland rough detection of stimuli, as well as rapid transfer of theinformation to the LH for more detailed processing. Further studieswill test this assertion by comparing verbal with non-verbal stimuliusing a similar measurement approach.

Importantly, abnormal lateralization has been reported in neuro-psychiatric syndromes, such as ADHD (Ashtari et al., 2005; Reid andNorvilitis, 2000), autism spectrum disorders (Escalante-Mead et al.,2003), eating disorders (Martins et al., 2001; Smeets and Kosslyn,2001) and schizophrenia (Crow, 1997; Klar, 1999; Sommer et al.,2003). Interestingly, these syndromes have been reported to becharacterized by attention deficits, and left-sided inattention wasreported in children with ADHD (Manly et al., 2005). Thus, abnormal

3019H. Okon-Singer et al. / NeuroImage 54 (2011) 3010–3020

brain lateralization and attention deficits may be related to thepathogenesis of neurobehavioral disorders, the nature of which hasbeen remained enigmatic so far.

Supplementarymaterials related to this article can be found onlineat doi:10.1016/j.neuroimage.2010.10.078.

Acknowledgments

This research was supported by the National Institute forPsychobiology in Israel (HOS), Adams Super Center for Brain Studiesat Tel Aviv University (HOS), Israel Science Foundation ConvergingTechnologies Program (TH) and the Smith's Center Grant of theNational Institute for Psychobiology in Israel (TH).

References

Arend, I., Rafal, R., Ward, R., 2008. Spatial and temporal deficits are regionallydissociable in patients with pulvinar lesions. Brain 131, 2140–2152.

Ashtari, M., Kumra, S., Bhaskar, S., Clarke, T., Thaden, E., Cervellione, K., et al., 2005.Attention-deficit/hyperactivity disorder: a preliminary diffusion tensor imagingstudy. Biol. Psychiatry 57 (5), 448–455.

Barnett, K.J., Corballis, M.C., 2005. Speeded right-to-left information transfer: the resultof speeded transmission in right-hemisphere axons? Neurosci. Lett. 380, 88–92.

Ben-Simon, E., Podlipsky, I., Arieli, A., Zhdanov, A., Hendler, T., 2008. Never restingbrain: simultaneous representation of two alpha related processes in humans. PLoSONE 3 (12), e3984.

Boehnke, S.E., Munoz, D.P., 2008. On the importance of the transient visual response inthe superior colliculus. Curr. Opin. Neurobiol. 18 (6), 544–551.

Bowers, D., Heilman, K.M., 1980. Pseudoneglect: effects of hemispace on a tactile linebisection task. Neuropsychologia 18, 491–498.

Burton, H., Jones, E.G., 1975. Cortical relationships of posterior thalamic region in NewWorld and Old World primates. Anat. Rec. 181 (2), 322.

Christov, I.I., 2004. Real time electrocardiogram qrs detection using combined adaptivethreshold. Biomed. Eng. Online 3 (28).

Committeri, G., Pitzalis, S., Galati, G., Patria, F., Pelle, G., Sabatini, U., et al., 2007. Neuralbasis of personal and extrapersonal neglect in humans. Brain 130, 431–441.

Corbetta, M., Shulman, G.L., 2002. Control of goal-directed and stimulus-drivenattention in the brain. Nat. Rev. Neurosci. 3, 201–215.

Corbetta, M., Akbudak, E., Conturo, T.E., Snyder, A.Z., Ollinger, J.M., Drury, H.A., et al.,1998. A common network of functional areas for attention and eye movements.Neuron 21, 761–773.

Corbetta, M., Kincade, M.J., Lewis, C., Snyder, A.Z., Sapir, A., 2005. Neural basis andrecovery of spatial attention deficits in spatial neglect. Nat. Neurosci. 8 (11),1603–1610.

Crow, T.J., 1997. Schizophrenia as failure of hemispheric dominance for language.Trends Neurosci. 20, 339–343.

Davidson, R.J., Shackman, A.J., Maxwell, J.S., 2004. Asymmetries in face and brain relatedto emotion. Trends Cogn. Sci. 8, 389–391.

Debener, S., Strobel, A., Sorger, B., Peters, J., Kranczioch, C., Engel, A.K., et al., 2007.Improved quality of auditory event-related potentials recorded simultaneouslywith 3-t fmri: removal of the ballistocardiogram artefact. Neuroimage 34, 587–597.

deSchotten, M.T., Urbanski, M., Duffau, H., Volle, E., Lévy, R., Dubois, B., et al., 2005.Direct evidence for a parietal–frontal pathway subserving spatial awareness inhumans. Science 309 (5744), 2226–2228.

Diekamp, B., Regolin, L., Güntürkün, O., Vallortigara, G., 2005. A left-sided visuospatialbias in birds. Curr. Biol. 15, 372–373.

Ekman, P., Friesen, W.V., 1976. Pictures of Facial Affect. Consulting Psychologists Press,Palo Alto, CA.

Elias, L.J., Bulman-Fleming, M.B., McManus, I.C., 2000. Linguistic lateralization andasymmetries in interhemispheric transmission time. Brain Cogn. 43, 181–185.

Escalante-Mead, P. R.,Minshew,N. J., Sweeney, J. A., & Volume33,N., October 2003, pp. (5).(2003). Abnormal brain lateralization in high-functioning autism. J. Autism Dev.Disord. 33(5), 539–543.

Fairhall, S.L., Indovina, I., Driver, J., Macaluso, E., 2009. The brain network underlyingserial visual search: Comparing overt and covert spatial orienting, for activationsand for effective connectivity. Cereb. Cortex 19 (12), 1946–1958.

Fink, G.R., Marshall, J.C., Halligan, P.W., Frith, C.D., Frackowiak, R.S., Dolan, R.J., 1997.Hemispheric specialization for global and local processing: the effect of stimuluscategory. Proc. R. Soc. Biol. Sci. 264 (1381), 487–494.

Fischer, J., Whitney, D., 2009. Precise discrimination of object position in the humanpulvinar. Hum. Brain Mapp. 30, 101–111.

Fries, P., Nikolic, D., Singer, W., 2007. The gamma cycle. Trends Neurosci. 30 (7),309–316.

Friston, K.J., Harrison, L., Penny, W., 2003. Dynamic causal modelling. Neuroimage 19,1273–1302.

Fuggetta, G., Pavone, E.F., Walsh, V., Kiss, M., Eimer, M., 2006. Cortico-corticalinteractions in spatial attention: a combined erp/tms study. J. Neurophysiol. 95,3277–3280.

Gabrieli, J.D.E., Whitfield-Gabrieli, S., 2007. Attention to neglect. Neuron 53, 776–777.Geffen, G., Bradshaw, J.L., Wallace, G., 1971. Interhemispheric effects on reaction time to

verbal and nonverbal visual stimuli. J. Exp. Psychol. 87 (3), 415–422.

Gitelman, D.R., Nobre, A.C., Parrish, T.B., LaBar, K.S., Kim, Y.H., Meyer, J.R., et al., 1999. Alarge-scale distributed network for covert spatial attention: further anatomicaldelineation based on stringent behavioral and cognitive controls. Brain 122,1093–1106.

Hari, R., Renvall, H., Tanskanen, T., 2001. Left minineglect in dyslexic adults. Brain 124,1373–1380.

He, B.J., Shulman, G.L., Snyder, A.Z., Corbetta, M., 2007. The role of impaired neuronalcommunication in neurological disorders. Curr. Opin. Neurol. 20, 655–660.

Heilman, K.M., Bowers, D., Valenstein, E., Watson, R.T., 1987. In: Jeannerod, M. (Ed.),Neurophysiological and neuropsychological aspects of spatial neglect. TheNetherlands, North Holland, Amsterdam, pp. 115–150.

Heilman, K.M., Watson, R.T., Valenstein, E., 1993. Neglect and related disorders, In:Heilman, K.M., Valenstein, E. (Eds.), Clinical Neuropsychology, 3rd ed. OxfordUniversity Press, New York (NY), pp. 279–336.

Hopf, J.-M., Mangun, G.R., 2000. Shifting visual attention in space: an electrophysio-logical analysis using high spatial resolution mapping. Clin. Neurophysiol. 111,1241–1257.

Kanwisher, N., Chun, M.M., McDermott, J., Ledden, P.J., 1996. Functional imaging ofhuman visual recognition. Cogn. Brain Res. 5, 55–67.

Karnath, H.-O., Himmelbach, M., Rorden, C., 2002. The subcortical anatomy of humanspatial neglect: putamen, caudate nucleus and pulvinar. Brain 125 (2), 350–360.

Kim, Y.H., Gitelman, D.R., Nobre, A.C., Parrish, T.B., LaBar, K.S., Mesulam, M.M., 1999. Thelarge-scale neural network for spatial attention displays multifunctional overlapbut differential asymmetry. Neuroimage 9, 269–277.

Kim, Y.H., Yoon, H., Park, H., 2004. Improved ballistocardiac artifact removal from theelectroencephalogram recored in fmri. J. Neurosci. Meth. 135, 193–203.

Klar, A., 1999. Genetic models for handedness, brain lateralization, schizophrenia, andmanic-depression. Schizophr. Res. 39, 207–218.

König, P., Engel, A.K., Roelfsema, P.R., Singer, W., 1995. How precise is neuronalsynchronization? Neural Comput. 7, 469–485.

Leehey, S., Carey, S., Diamond, R., Cahn, A., 1978. Upright and inverted faces: the righthemisphere knows the difference. Cortex 14 (3), 411–419.

Leh, S.E., Johansen-Berg, H., Ptito, A., 2006. Unconscious vision: new insights into theneuronal correlate of blindsight using diffusion tractography. Brain 129 (7),1822–1832.

Leh, S.E., Chakravarty, M.M., Ptito, A., 2008. The connectivity of the human pulvinar: adiffusion tensor imaging tractography study. Int. J. Biomed. Imaging Article ID789539, 5 pp.

Lundqvist, D., Flykt, A., Öhman, A., 1998. The karolinska directed emotional faces [CDROM]. Stockholm: Karolinska Institute KDEF.

Manly, T., Cornish, K., Grant, C., Dobler,V.,Hollis, C., 2005. Examining the relationshipbetweenrightward visuo-spatial bias andpoor attentionwithin thenormal child populationusinga brief screening task. J. Child Psychol. Psychiatry Allied Discip. 46 (12), 1337–1344.

Martínez, A., Teder-Sälejärvi,W., Vazquez,M.,Molholm, S., Foxe, J.J., Javitt, D.C., et al., 2006.Objects are highlighted by spatial attention. J. Cogn. Neurosci. 18 (2), 298–310.

Martins, J.M., Trincaa, A., Afonso, A., Carreiras, F., Falco, J., Nunes, J.S., et al., 2001.Psychoneuroendocrine characteristics of common obesity clinical subtypes. Int. J.Obes. 25, 24–32.

Marzi, C.A., Bisiacchi, P., Nicoletti, R., 1991. Is interhemispheric transfer of visuomotorinformation asymmetric? Evidence from a meta-analysis. Neuropsychologia 29,1163–1177.

Matsuzaki, R., Kyuhou, S.I., Matsuura-Nakao, K., Gemba, H., 2004. Thalamo-corticalprojections to the posterior parietal cortex in the monkey. Neurosci. Lett. 355 (1–2), 113–116.

McCourt, M.E., Jewell, G., 1999. Visuospatial attention in line bisection: stimulusmodulation of pseudoneglect. Neuropsychologia 37, 843–855.

Mesulam, M., 1981. A cortical network for directed attention and unilateral neglect.Ann. Neurol. 10, 309–325.

Moes, P.E., Brown, W.S., Minnema, M.T., 2007. Individual differences in interhemi-spheric transfer time (IHTT) as measured by event related potentials. Neuropsy-chologia 45 (11), 2626–2630.

Müller, J.R., Philiastides, M.G., Newsome, W.T., 2005. Microstimulation of the superiorcolliculus focuses attention without moving the eyes. Proc. Natl Acad. Sci. 102 (3),524–529.

Natale, E., Marzi, C.A., Girelli, M., Pavone, E.F., Pollmann, S., 2006. Erp and fmri correlatesof endogenous and exogenous focusing of visual–spatial attention. Eur. J. Neurosci.23 (9), 2511–2521.

Nobre, A.C., Sebestyen, G.N., Gitelman, D.R., Mesulam, M.M., Frackowiak, R.S., Frith, C.D.,1997. Functional localization of the system for visuospatial attention using positronemission tomography. Brain 120 (3), 515–533.

Nowicka, A., Fersten, E., 2001. Sex-related differences in interhemispheric transmissiontime in the human brain. NeuroReport 12 (18), 4171–4175.

Nowicka,A., Grabowska,A., Fersten, E., 1996. Interhemispheric transmissionof informationand functional asymmetry of the human brain. Neuropsychologia 34, 147–151.

Penny, W.D., Stephan, K.W., Mechelli, A., Friston, K.F., 2004. Modelling functionalintegration: a comparison of structural equation and dynamic causal models.Neuroimage 23, 264–274.

Petersen, S.E., Robinson, D.L., Morris, J.D., 1987. Contributions of the pulvinar to visualspatial attention. Neuropsychologia 25 (1A), 97–105.

Poffenberger, A.T., 1912. Reaction time to retinal stimulation with special reference tothe time lost in conduction through nervous centres. Arch. Psychol. 23, 1–73.

Posner, M.I., 1980. Orienting of attention. Q. J. Exp. Psychol. 32, 3–25.Posner, M.I., Petersen, S.E., 1990. The attention system of the human brain. Annu. Rev.

Neurosci. 13, 25–42.Rafal, R.D., Posner, M.I., 1987. Deficits in human visual spatial attention following

thalamic lesions. Proc. Natl Acad. Sci. 84, 7349–7353.

3020 H. Okon-Singer et al. / NeuroImage 54 (2011) 3010–3020

Reid, H.M., Norvilitis, J.M., 2000. Evidence for anomalous lateralization across domain inadhd children as well as adults identified with the wender utah rating scale.J. Psychiatr. Res. 34, 311–316.

Rushworth,M.F.S., Behrens, T.E.J., Johansen-Berg,H., 2006. Connectionpatternsdistinguish3 regions of human parietal cortex. Cereb. Cortex 16 (10), 1418–1430.

Sadeh, B., Zhdanov, A., Podlipsky, I., Hendler, T., Yovel, G., 2008. The validity of the face-selective erp n170 component during simultaneous recording with functional mri.Neuroimage 42, 778–786.

Sapir, A., Soroker, N., Berger, A., Henik, A., 1999. Inhibition of return in spatial attention:direct evidence for collicular generation. Nat. Neurosci. 2, 1053–1054.

Schmahmann, J.D., Pandya, D.N., 1990. Anatomical investigation of projections fromthalamus to posterior parietal cortex in the rhesus-monkey — a Wga-Hrp andfluorescent tracer study. J. Comp. Neurol. 295 (2), 299–326.

Schneider, K.A., Kastner, S., 2009. Effects of sustained spatial attention in the humanlateral geniculate nucleus and superior colliculus. J. Neurosci. 29 (6), 1784–1795.

Siman-Tov, T., Mendelsohn, A., Schonberg, T., Avidan, G., Podlipsky, I., Pessoa, L., et al.,2007. Bihemispheric leftward bias in a visuospatial attention-related network.J. Neurosci. 27, 11271–11278.

Siman-Tov, T., Papo, D., Gadoth, N., Schonberg, T., Mendelsohn, A., Perry, D., Hendler, T.,2009. Mind your left: spatial bias in subcortical fear processing. J. Cogn. Neurosci. 21(9), 1782–1789.

Smeets, M.A.M., Kosslyn, S.M., 2001. Hemispheric differences in body image in anorexianervosa. Int. J. Eat. Disord. 29, 409–416.

Snow, J.C., Allen, H.A., Rafal, R.D., Humphreys, G.W., 2009. Impaired attentionalselection following lesions to human pulvinar: evidence for homology betweenhuman and monkey. Proc. Natl Acad. Sci. 106, 4054–4059.

Sommer, I.E., Ramsey, N.F., Mandl, R.C., Kahn, R.S., 2003. Language lateralization infemale patients with schizophrenia: an fMRI study. Schizophr. Res. 60, 183–190.

Springer, J.A., Binder, J.R., Hammeke, T.A., Swanson, S.J., Frost, J.A., Bellgowan, P.S., et al.,1999. Language dominance in neurologically normal and epilepsy subjects: afunctional MRI study. Brain 122, 2033–2046.

Srinivasan, R., Winter, W.R., Nunez, P.L., 2006. Source analysis of EEG oscillations usinghigh resolution EEG and MEG. In: Klimesch, N. (Ed.), Progress in Brain Research,Vol. 19. Elsevier B.V., pp. 29–42.

Verleger, R., Sprenger, A., Gebauer, S., Fritzmannova,M., Friedrich,M., Kraft, S., et al., 2008. Onwhy left events are the right ones: neural mechanisms underlying the left-hemifieldadvantage in rapid serial visual presentation. J. Cogn. Neurosci. 21 (3), 474–488.

Ward, R., Arend, I., 2007. An object-based frame of reference within the humanpulvinar. Brain 130, 2462–2469.

Womelsdorf, T., Schoffelen, J.-M., Oostenveld, R., Singer, W., Desimone, R., Engel, A.K., etal., 2007. Modulation of neuronal interactions through neuronal synchronization.Science 316, 1608–1612.