Attentional selection of relative SF mediates globalversus local processing: Evidence from EEG

Department of Psychology, University of California atBerkeley, USA, &

Veterans Administration Center, Martinez, CA, USAAnastasia V. Flevaris

Department of Psychology and the Center of NeuralComputation, Hebrew University of Jerusalem,

Jerusalem, IsraelShlomo Bentin

Department of Psychology, University of California atBerkeley, USA, &

Veterans Administration Center, Martinez, CA, USALynn C. Robertson

Previous research on functional hemispheric differences in visual processing has associated global perception with lowspatial frequency (LSF) processing biases of the right hemisphere (RH) and local perception with high spatial frequency(HSF) processing biases of the left hemisphere (LH). The Double Filtering by Frequency (DFF) theory expanded thishypothesis by proposing that visual attention selects and is directed to relatively LSFs by the RH and relatively HSFs by theLH, suggesting a direct causal relationship between SF selection and global versus local perception. We tested this idea inthe current experiment by comparing activity in the EEG recorded at posterior right and posterior left hemisphere sites whileparticipants’ attention was directed to global or local levels of processing after selection of relatively LSFs versus HSFs in aprevious stimulus. Hemispheric asymmetry in the alpha band (8–12 Hz) during preparation for global versus localprocessing was modulated by the selected SF. In contrast, preparatory activity associated with selection of SF was notmodulated by the previously attended level (global/local). These results support the DFF theory that top-down attentionalselection of SF mediates global and local processing.

Keywords: global versus local processing, hemispheric asymmetry, spatial frequency, alpha reduction

Citation: Flevaris, A. V., Bentin, S., & Robertson, L. C. (2011). Attentional selection of relative SF mediates global versuslocal processing: Evidence from EEG. Journal of Vision, 11(7):11, 1–12, http://www.journalofvision.org/content/11/7/11,doi:10.1167/11.7.11.

Introduction

Our environment is hierarchically organized in thatscenes are composed of objects that exist at different levelsof detail. Efficient visual perception requires the ability torepresent such multiple levels so that attention can bedirected to the level of the whole (“global processing”) or tothe level of its components (“local processing”). Thisability relies on neural mechanisms that allow for theconcurrent segregation and integration of object features inorder to represent objects that are embedded within oneanother rather than being located in distinct locations.These mechanisms have been the focus of ample research,but their characterization remains incomplete. For exam-ple, previous neuropsychological, cognitive, and imagingresearch has converged to establish functional hemisphericasymmetries in global versus local perception, with theright hemisphere (RH) biased toward global processing andthe left hemisphere (LH) biased toward local processing(e.g., Delis, Robertson, &Effron, 1986; Fink et al., 1997; Hanet al., 2002; Lamb, Robertson, & Knight, 1989; Martin,

1979, Martinez et al., 1997; Robertson & Delis, 1986;Robertson & Lamb, 1991; Robertson, Lamb, & Knight,1988; Robertson, Lamb, & Zaidel, 1993; Sergent, 1982,1983, Volberg & Hubner, 2004; Weissman & Woldorff,2005). However, global and local are inherently relative(i.e., they only exist in relation to the other level(s)) andthe cognitive and neural processes that yield theseasymmetric hemispheric biases continue to be debated.Several theories of global versus local perception have

proposed selective processing of spatial frequency (SF)ranges to underlie the hemispheric asymmetry. Thesetheories rely on evidence that the visual system filtersincoming information into channels that are tuned fordifferent SF bands (De Valois & De Valois, 1988). Otherstudies have shown that SFs are flexibly used to influenceperformance depending on task demands (e.g., Loftus &Harley, 2004; Morrison & Schyns, 2001; Schyns & Oliva,1999).Broadbent (1977) was among the first to suggest that

the important functional distinction between attending toglobal and local levels of a hierarchical display is in theirSF content; low spatial frequencies (LSFs) attract more

Journal of Vision (2011) 11(7):11, 1–12 http://www.journalofvision.org/content/11/7/11 1

doi: 10 .1167 /11 .7 .11 Received October 2, 2010; published June 13, 2011 ISSN 1534-7362 * ARVO

attention to the global than local levels, and high spatialfrequencies (HSFs) attract more attention to the local thanglobal levels. The importance of SFs in global and localprocessing has more recently been supported by severalbehavioral (e.g., Goffaux, Hault, Michel, Vuong, & Rossion,2005; Goffaux & Rossion, 2006) and ERP studies of faceprocessing (e.g., Boeschoten, Kemnera, Kenemansc, & vanEngeland, 2005; Flevaris, Robertson, & Bentin, 2008).Other studies have found an analogous hemispheric

asymmetry in processing SFs, with the RH biased indiscriminating LSFs and the LH biased in discriminatingHSFs (Christman, Kitterle, & Hellige, 1991; Kitterle,Christman, & Hellige, 1990; Kitterle, Hellige, & Christman,1992), as well as a direct link between LSF versus HSFdiscrimination and attending to global versus local objects,respectively (Flevaris, Bentin, & Robertson, 2011; Shulman,Sullivan, Gish, & Sakoda, 1986; Shulman &Wilson, 1987).Furthermore, the asymmetry occurs whether the stimuli aresmall or large (Robertson & Lamb, 1991), making the roleof SFs more complex than a simple bottom-up analysis. Toaccount for these and other data, Ivry and Robertson (1998;Robertson & Ivry, 2000) developed the Double Filtering byFrequency (DFF) theory of hemispheric specialization forSF processing. According to this theory, there are two stagesof SF filtering, and the two hemispheres differ in how theyprocess SF information after the initial visual stage.Specifically, attention first selects an SF range from theincoming spectrum that is most suited for the current task(e.g., the range of spatial scales that is most likely to containthe target; see Watt, 1988 for a model of how this mightwork). This SF range is then fed forward in both cerebralhemispheres, but it is at this second stage that the hemi-spheres differ, with the LH acting as a relatively high-passfilter, emphasizing information from the higher SFs withinthe initially selected range, and the RH as a relatively low-pass filter, emphasizing the lower SFs within the range.Despite evidence associating SF processing with global

ersus local perception, a number of alternative proposalshave also been made to account for these hemisphericasymmetries. Among those are theories proposing hemi-spheric differences in sensitivity to the saliency of thestimulus, with the right hemisphere biased toward moresalient objects and the left hemisphere biased toward lesssalient objects (Mevorach, Humphreys, & Shalev, 2005,2006), or in shape-level integration, with the right hemi-sphere involved in binding shapes to the global leveland the left hemisphere involved in binding shapes to thelocal level (Hubner & Volberg, 2005). Moreover, there issome disagreement even among theories positing SFprocessing differences. In contrast to the flexible mecha-nism posited by DFF theory, some researchers haveassumed that the use of SF scales reflects a lower levelmechanism by which the two hemispheres receive differ-ent input (or asymmetrically emphasize different input)during lower level visual processing and, thus, differ inabsolute SF biases (e.g., Jacobs & Kosslyn, 1994; Sasakiet al., 2001; Sergent, 1982).

We recently reported evidence in support of a flexiblemechanism, based on relative SF, to underlie hemisphericasymmetry in global versus local perception (Flevaris et al.,2011). In that study, we found that allocating attention toglobal or local levels of hierarchical displays (Navonletters) biased selection of LSFs or HSFs, respectively, insubsequently presented compound gratings. Importantly,this bias was determined by the relationship between thetwo SFs in the compound grating (i.e., their relativefrequency) rather than the absolute SF values. That is,global attention biased selection of a 1.8 cycles/degreegrating when it was paired with a higher SF (5.3 cycles/degree), but local attention biased selection of the same1.8 cycles/degree grating when it was paired with a lowerSF (0.9 cycle/degree) in the compound. Although thesedata are consistent with DFF theory in the sense thatglobal/local selection is based on relative SF, neuro-biological measures of hemispheric asymmetries were notdirectly assessed in that study. In a different studyexamining hemispheric differences in integrating shapeswith hierarchical level, we found evidence that attention toLSFs facilitates binding of shapes to the global level bythe RH and that attention to HSFs facilitates binding ofshapes to the local level by the LH (Flevaris, Bentin, &Robertson, 2010). These data suggest that attentionalselection of relative SF scales could provide the mecha-nism for shape-to-level integration proposed by Hubnerand Volberg (2005), thereby integrating the DFF andshape-to-level integration theories of global versus localprocessing in a unified framework.However, the behavioral priming effects in the previous

studies may have been observed due to a bidirectionaladditive model, in which the right hemisphere is biasedtoward relative LSFs and global objects whereas the lefthemisphere is biased toward relative HSFs and localobjects. This is in contrast to the DFF theory proposal thatselection of SF induces global/local processing. To thisend, in the current study, we adapted the behavioral experi-ment described in Flevaris et al. (2011) to an EEG study inorder to examine the direction of modulation of responsesin the two cerebral hemispheres while participants werepreparing to respond to relative SF and hierarchical level.We investigated preparatory neural activity by examiningoscillatory brain responses in the scalp EEG in alpha band(8–12 Hz).Modulation of the alpha band has been previously

established as a measure of brain activity and used as a directindex of attention (Bastiaansen, Bocker, Brunia, deMunck, &Spekreijse, 2001; Thut, Nietzel, Brandt, & Pascal-Leone,2006; Volberg, Kliegl, Hansimayr, & Greenlee, 2009; Ward,2003; Worden, Foxe, Wang, & Simpson, 2000). Specifi-cally, since the alpha band has been considered as “idlingrhythms” of the brain (Pfurtscheller, Stancak, & Neuper,1996), reduction in alpha amplitude (Event-RelatedDesynchronizationVERD) is assumed to reflect elevatedbrain activation, whereas augmentation of alpha amplitude(Event-Related SynchronizationVERS) is assumed to reflect

Journal of Vision (2011) 11(7):11, 1–12 Flevaris, Bentin, & Robertson 2

inhibitory processes (Klimesch, Doppelmayr, & Hanslmayr,2007). In the case of directing attention to global versuslocal objects, Volberg et al. (2009) showed that, at least fortrials in which behavioral performance was optimal (i.e.,trials in which response times were fastest), alpha ampli-tudes at centro-parietal sites were greater over the righthemisphere (i.e., higher left than right hemisphere activa-tion) before participants reported local targets and greaterover the left hemisphere (i.e., higher right than left hemi-sphere activation) before they reported global targets. Thesedata confirmed that the hemispheric asymmetry in globalversus local processing occurs not only during perceptualprocessing but also during attentional orienting in preparingfor one level or the other and that EEG oscillations in thealpha band can index relative attentional orienting across thetwo cerebral hemispheres. In the current study, we examinedthe modulation of this orienting response after selection ofan SF in a complex grating and before the global or the localstimulus actually appeared. In this way, we could examinehow HSF versus LSF selection influenced the preparatoryresponse to the global or local level. We also looked at alphamodulation at the LH and RH sites prior to the processing ofgratings following a global or local response and prior to

global or local processing following the selection of eitherLSF or HSF in the gratings. This allowed us to determinethe directionality or symmetry of the SF level of processingrelationship.Participants performed two tasks in alternating sequence:

an SF grating orientation task, in which they wereinstructed to report the orientation of either the relativelyLSF or HSF stripes in a compound SF grating, and a hierar-chical level same/different task, in which they reportedwhether the global or local letters in two hierarchicalNavon displays were the same or different (see Figure 1).There were approximately 2 s in between each stimulus(grating/Navon). The long inter-stimulus interval allowedenough time to sufficiently measure amplitudes in thealpha band. In our previous behavioral study (Flevariset al., 2011), the SF grating appeared immediately afterresponse to the Navon display, and as mentioned above,we found priming effects such that attention to the globallevel facilitated responses to relatively LSFs and atten-tion to the local level facilitated responses to relativelyHSFs. Since, in the present study, we extended theinterval between the Navon and grating stimuli, we didnot necessarily expect a priming effect on reaction times

Figure 1. Sample trial. Participants performed two tasks in alternating sequence: a Navon task and a grating task. For the Navon task, twoNavon letters appeared on the screen side by side and participants indicated if they were the same or different at the global or local level(in separate blocks of trials). For the grating task, a single compound grating appeared at fixation and participants reported the orientationof the lower spatial frequency (LSF) or higher spatial frequency (HSF) grating (in separate blocks of trials). A 1000-ms fixation crosspreceded each stimulus. Additionally, the fixation cross appeared on the screen after each stimulus for a variable interval between 1024and 1224 ms (resulting in 2024–2224 ms between each stimulus).

Journal of Vision (2011) 11(7):11, 1–12 Flevaris, Bentin, & Robertson 3

(RTs); instead, we focused on preparatory activity asreflected in modulations of alpha amplitude during theinter-task intervals relative to a baseline period. Specifi-cally, we compared the alpha ERD over the left and righthemispheres directly prior to the onset of the Navonletters as a function of the relative SF that was previouslyattended, as well as the alpha ERD directly prior to thecompound SF grating as a function of the level that waspreviously attended in the Navon task. If the behavioralpriming effects associating relative LSF processing withglobal perception and relative HSF processing with localperception were due to the bidirectional additive modeldescribed previously, then we would expect a similarpattern of alpha ERD in preparation for the response to thegratings as in preparation for response to the Navon.Conversely, if the DFF hypothesis is correct, and SFselection is the medium by which hierarchical perceptionoccurs, then we would expect attended SF to modulatepreparatory alpha activity unidirectionally. Moreover, inline with our hypothesis that LSF selection mediatesglobal perception by the right hemisphere and HSFselection mediates local perception by the left hemisphere,we predicted a second-order interaction between task(global/local), SF, and hemisphere (which was observed).

Methods

Participants

Sixteen undergraduates (ten women) from the Universityof California, Berkeley participated in the experiment formonetary compensation. All were right-handed and hadnormal or corrected-to-normal vision. All gave informedconsent as approved by the Committee for the Protection ofHuman Subjects at the University of California, Berkeley.

Stimuli

The Navon patterns were created using Adobe Photoshopby arranging a series of (local) letters in black Helveticabold font to form a larger (global) letter. The letters usedwere A, C, D, E, F, and H, in all their global and localcombinations, with the exception of congruent combina-tions (e.g., a global A made up of local A’s). This resultedin 30 distinct Navon displays. Seen from a distance of 1 m,each local letter subtended 0.7- of visual angle and eachglobal letter was 4.0- wide by 5.7- high.The compound gratings were generated in Matlab

(Mathworks, Natick, MA) with a sinusoid function foreach SF. Each compound grating subtended 5.4- of visual

angle overall and was composed of a 0.7 cycle/degree(4 cycles/image) grating (the relatively LSF component)and a 2.4 cycle/degree (13 cycles/image) grating (therelatively HSF component). Both SF gratings were at100% contrast. The relatively LSF component wasoriented either at 45- (tilted to the right) or at 135- (tiltedto the left) and the relatively HSF component was orientedin the opposite direction.

Procedure

The stimuli were shown on a Dell 17-inch color CRTmonitor with a vertical refresh rate of 60 Hz and aresolution of 1024 � 768 pixels. Participants were seated1 m from the screen in a dark room. Trial timing wascontrolled by Presentation (Neurobehavioral Systems,Albany, CA) and is depicted in Figure 1. Each trial beganwith a central fixation cross presented for 1000 ms. TwoNavon patterns were then presented simultaneously for300 ms, one in the left and one in the right visual field, withthe medial edge 1- from fixation. A variable inter-trialinterval (ITI) between 1024 ms and 1224 ms followed theNavon patterns, during which time participants indicated viabutton press if the two patterns were the same or different atthe global level (during half of the blocks) or local level(during the other half of the blocks). A 1000-ms fixationperiod followed the ITI, after which a compound SF gratingwas presented at fixation for 300 ms. A variable ITI between1024 ms and 1224 ms followed the grating, during whichtime participants indicated via button press whether therelatively LSF component was tilted to the left or right(during half of the blocks) or whether the relatively HSFcomponent was tilted to the left or right (during the other halfof the blocks). At the end of this ITI, a new trial began.Hence, the Navon letters were separated from the grating by arandom interval between 2024 and 2224 ms and a similarinterval separated the grating from the Navon letters. Thisensured that the two tasks were perceived as alternatingrather than one preceding the other. The response mappingsfor the four response buttons (i.e., two for the Navon task andtwo for the grating task) were counterbalanced acrossparticipants. Participants performed four blocks of 120 trialseach, in counterbalanced order: attention to the global level inthe Navon patterns/LSFs in the grating, global/HSF, local/LSF, and local/HSF. This block design was used to eliminateconfusion about the dual task as well as keep the experimentas close as possible to our previous behavioral version(Flevaris et al., 2011). A 20-trial practice block precededeach experimental block. Additionally, EEG was recordedwhile participants closed their eyes for 2 min prior to theexperiment and 2 min following the experiment. Alphaactivity during this period was averaged and used as abaseline for calculating the ERD.

Journal of Vision (2011) 11(7):11, 1–12 Flevaris, Bentin, & Robertson 4

EEG recording and analysis

The EEG analog signals were recorded continuously by64 Ag–AgCl pin-type active electrodes mounted on anelastic cap (ECI) according to the extended 10–20 systemand from two additional electrodes placed at the right andleft mastoids, all referenced to a common-mode signal(CMS) electrode between POz and PO3 and were sub-sequently re-referenced digitally to the average reference.Eye movements and blinks were monitored using bipolarhorizontal and vertical EOG derivations via two pairs ofelectrodes: one pair attached to the external canthi and theother to the infraorbital and supraorbital regions of the righteye. Both EEG and EOG were sampled at 256 Hz using aBiosemi Active II digital 24-bit amplification system withan active input range of j262 2V to +262 2V per bitwithout any filter at input. The digitized EEG was savedand processed offline.Data were analyzed using Matlab (Mathworks, Natick,

MA). To remove EOG and EEG artifacts, a change involtage of more than 100 2V during a 100-ms epoch in anyof the channels was considered an artifact and the EEGrecorded in the interval 200 ms before and after the artifactwas removed. Following this artifact removal procedure,more than 90% of the trials were preserved. To analyzeoscillatory cortical activity associated with preparing forthe global/local and SF tasks, the single trial EEG signal oneach channel was convolved with 6-cycle Morlet waveletsover a 2950-ms window beginning 2700 ms prior tostimulus onset. To examine oscillatory cortical activityduring and following each stimulus, the EEG signal oneach channel was also convolved as over a 2950-mswindow beginning 250 ms prior to stimulus onset.Instantaneous power and phase were extracted at each timepoint (at 256-Hz sampling rate) over frequency scales from0.7 to 60 Hz incremented logarithmically (Lakatos et al.,2005). The square roots of the power values (the sum of thesquares of the real and imaginary Morlet components) werethen averaged over single trials to yield the total averagedspectral amplitudes (in 2V) for each condition. The averagedspectral amplitude at each time point and frequency wasbaseline-corrected by dividing the mean spectral amplitudefound during the baseline period (Tallon-Baudry, Bertrand,Peronnet, & Pernier, 1998). The ratio was used to reduce theimpact of the variability in absolute EEG power caused byindividual differences such as scalp thickness and electrodeimpedance (Pineda & Oberman, 2006). Further, since ratiodata are inherently not normally distributed as a result oflower bounding, a log transform (of that ratio) was used foranalysis. Hence, a log ratio of less than zero indicates ERD,a value of zero indicates no change, and values greater thanzero indicate ERS.Analyses of variance (ANOVAs) with repeated measures

were performed on the mean signal change in the EEGalpha band amplitudes (8–12 Hz). Following inspection of

the alpha distribution, we focused on the parieto-occipitalsites (see below), in which the alpha power was greatest(see Figure 2).

Results

Behavioral results

Accuracy was high for both the SF grating and the Navontasks (96% and 92%, respectively). Reaction times (RTs) toaccurate responses that were within 2 standard deviationson the mean were analyzed with separate 2 � 2 ANOVAsfor the Navon task and the SF grating task. The factors werelevel (global versus local) and SF (LSF versus HSF). Asexpected, these analyses did not yield significant effects.

Global/local preparation

To assess the alpha activity associated with preparationfor global versus local target discrimination, the mean

Figure 2. Scalp distribution of power at 10 Hz. Alpha power isgreatest in the occipital sites (PO7/PO8, PO3/PO4, and O1/O2).

Journal of Vision (2011) 11(7):11, 1–12 Flevaris, Bentin, & Robertson 5

alpha signal change (relative to baseline, see above) in the1-s period preceding the Navon display was analyzed witha 2 � 2 � 2 � 3 ANOVA with repeated measures.The factors were to-be-attended level (global versuslocal), previously attended SF (LSF versus HSF),

hemisphere (LH versus RH), and site (PO7/PO8, PO3/PO4, O1/O2). As described above, these sites wereselected because they yielded the greatest power in thealpha band when inspecting the distribution across theentire scalp.

Figure 3. (A) Alpha reduction (relative to baseline) during highlighted interval (shown in gray) for global versus local preparation, following LSFversus HSF selection, shown for occipital electrodes (PO7/8, PO3/4, O1/O2). In preparation for the global task, there is significantly greateralpha reduction in the right hemisphere than in the left hemisphere following LSF selection but no significant hemispheric difference followingHSF selection. There were no significant hemispheric differences in preparation for the local task. (B) Spectral amplitudes in the righthemisphere (electrode PO8)minus amplitudes in the left hemisphere (electrode PO7) shown for each condition for the 1-s period preceding theNavon. The greater alpha reduction in the right hemisphere is absent for global preparation following HSF selection.

Journal of Vision (2011) 11(7):11, 1–12 Flevaris, Bentin, & Robertson 6

The overall ANOVA revealed a significant level � SF �hemisphere second-order interaction [F(1,15) = 6.29, p G0.05, MSE = 0.002, partial )2 = 0.30], and no other effectswere statistically significant. To examine this interaction, weconducted SF (LSF versus HSF) by Hemisphere (LH versusRH) ANOVAs for the global and local preparation con-ditions separately, and these data are depicted in Figure 3.When participants prepared for the global task, there was asignificant SF � hemisphere interaction [F(1,15) = 11.84,p G 0.01, MSE = 0.001, partial )2 = 0.28]. Follow-up t-testsrevealed significantly greater alpha reduction in the righthemisphere (j0.897) than in the left hemisphere (j0.827)when attention was previously allocated to LSFs [t(15) =2.14, p G 0.05] and a smaller, non-significant hemisphericdifference (j0.875 in the right hemisphere and j0.840 inthe left hemisphere)when attention was previously allo-cated to HSFs [t(15) = 1.07, p = 0.30]. When participantsprepared for the local task, the ANOVA did not yieldsignificant results.

To provide evidence that the alpha ERD preceding theNavon display was indeed preparatory activity associatedwith global versus local processing, we also analyzed themean alpha signal change during the 1-s period followingthe compound SF grating, prior to the 1-s windowpreceding the Navon display. This analysis did not yieldsignificant results, suggesting that the SF modulation ofalpha activity preceding the Navon stimuli was associatedwith preparing to respond to the global or local levelsrather than residual activity from processing the grating(see Figure 4B).

LSF/HSF preparation

To assess the alpha activity associated with preparationfor LSF versus HSF selection, we performed the same 2 �2 � 2 � 3 repeated measures ANOVA on the mean alphasignal change in the 1-s period preceding the compound

Figure 4. Spectral amplitudes in the right hemisphere (electrode PO8) versus amplitudes in the left hemisphere (electrode PO7) shown foreach condition for (A) the 1-s period preceding the SF grating (shown in gray) and (B) the 1-s period following the grating (shown asdotted). There were no significant hemispheric differences for any condition in either period.

Journal of Vision (2011) 11(7):11, 1–12 Flevaris, Bentin, & Robertson 7

SF grating. This analysis did not yield any significantresults. Importantly, there was no SF� level� hemisphereinteraction in the analysis of the 1-s period preceding thegrating [F(1,15) G 1], consistent with our hypothesis thatSF mediates the hemispheric asymmetry in global versuslocal processing.

Statistical analysis of preparatory timewindow

We also conducted an overall ANOVA comparing bothpreparatory time windows (i.e., pre-Navon and pre-SFgrating). The factors were Analysis window (pre-Navonvs. pre-grating, Stimulus 1 (HSF vs. LSF, or local vs.global), Stimulus 2 (local vs. global, or HSF vs. LSF),Hemisphere, and Site. This analysis revealed a significant4-way interaction between Analysis window, Stimulus 1,Stimulus 2, and Hemisphere [F(1,15) = 5.4, p G 0.05,MSE = 0.004, )2 = 0.27], indicating that the modulationin preparation for global stimuli (that we found in the pre-Navon analysis) is in fact significantly different frommodulation in preparation to LSF stimuli.

Theta band

Although we had no a priori predictions about prepar-atory oscillations in other frequency ranges, we conductedsimilar analyses for the mean signal change in the theta(4–7 Hz) band.1 As we did for the alpha band, we focusedour analyses on electrode sites in which the theta powerwas the greatest (Figure 5). For the theta band, thisincluded the same 3 sites as the alpha band analysis withthe two additional sites P1/P2 and P3/P4.In the theta band, the to-be-attended level (Global versus

Local) � previously attended SF (LSF versus HSF) �hemisphere (left versus right) � site (PO7/PO8, PO3/PO4,O1/O2, P3/P4, and P1/P2) ANOVA for the 1-s periodpreceding the Navon showed a significant main effect ofhemisphere [F(1,15) = 4.9, p G 0.05, MSE = 0.022, partial)2 = 0.25], with greater theta reduction in the righthemisphere (j0.747) than in the left hemisphere (j0.721).There was also a significant level � hemisphere interaction[F(1,15) = 4.9, p G 0.05, MSE = 0.002, partial )2 = 0.25].Follow-up t-tests revealed significantly greater theta reduc-tion in the right hemisphere (j0.748) than in the lefthemisphere (j0.715) preceding global attention [t(15) =2.54, p G 0.05] and a smaller non-significant differencepreceding local attention [t(15) = 1.65, p = 0.12]. Thisanalysis also revealed a main effect of site [F(1,15) = 70.7,MSE = 0.043, p G 0.001, partial )2 = 0.84], demonstratinggreatest theta reduction at the more medial sites P1/P2(j0.849) and P3/P4 (j0.851), followed by more anteriormedial sites PO3/PO4 (j0.730; t(15) = 4.7, p G 0.001),which in turn demonstrated greater alpha reduction than the

more lateral posterior sites PO7/PO8 (j0.613) and O1/O2(j0.627; t(15) = 4.56, p G 0.001).The ANOVA for the 1-s period preceding the SF

grating only revealed a main effect of site [F(1,15) =70.7, p G 0.001], demonstrating a similar pattern as thatfound for the period preceding the Navon. The greatesttheta reduction was seen at the more medial sites P1/P2(j0.910) and P3/P4 (j0.908), followed by more anteriormedial sites PO3/PO4 (j0.804; t(15) = 13.7, p G 0.001),which in turn demonstrated greater alpha reduction thanthe more lateral posterior sites PO7/PO8 (j0.693) andO1/O2 (j0.683; t(15) = 14.9, p G 0.001).

Discussion

This experiment provides direct evidence in support ofDFF theory by demonstrating that attentional selection ofSF modulates preparatory activity in global versus localperception. Consistent with previous findings (Volberget al., 2009), we found greater alpha reduction in the righthemisphere than in the left hemisphere in preparation for

Figure 5. Scalp distribution of power at 6 Hz. Theta power isgreatest in the parieto-occipital sites (PO7/PO8, PO4/PO3, P3/P4,P1/P2, and O1/O2).

Journal of Vision (2011) 11(7):11, 1–12 Flevaris, Bentin, & Robertson 8

global targets. This hemispheric asymmetry was foundfollowing attentional selection of LSF but not followingattentional selection of HSF, supporting the theory thatLSF selection in the right hemisphere mediates globalprocessing. We did not find similar SF modulation duringthe interval preceding the compound SF grating orimmediately following it. If attentional selection of SFinduces a bias for a particular SF range, one might wonderwhy we did not see significant alpha modulation in the 1-speriod following the grating. Interestingly, when compar-ing the alpha reduction in the period immediatelyfollowing the SF grating (Figure 4B) with that in theperiod immediately preceding the Navon (Figure 3A),there is far greater alpha reduction overall in the periodimmediately following the grating. As time passed, thealpha reduction dissipated across both the left and righthemispheres. However, when participants were preparingto attend to the global level, the reduction dissipated in theleft hemisphere to a significantly greater degree than it didin the right hemisphere in the post-LSF condition. In thepost-HSF condition, on the other hand, the reductiondissipated equally across the two hemispheres.The unidirectional pattern of hemispheric modulation by

SF suggests that the alpha modulation was not induced purelyby processing low and high SFs but rather as an interactionbetween the relative frequency range selected in the gratingand the preparation for the global or local task. This wasfurther corroborated in the analysis comparing the preparatoryinterval before the Navon and the interval before the grating,in which we found a significant Analysis window (pre-Navonversus pre-grating) � Stimulus 1 (global versus local/LSFversus HSF) � Stimulus 2 (LSF versus HSF/global versuslocal) � hemisphere interaction. Thus, these data support acausal relationship between attentional selection of SF bythe right hemisphere and the perception of hierarchicalobjects at the global level.We did not find a significant opposite pattern, that is,

alpha suppression was not larger over left hemisphere sitesin preparation for the local level (e.g., Yamaguchi,Yamagata, & Kobayashi, 2000). Rather, although the maineffect of hemisphere did not reach statistical significancein the analyses of preparatory alpha activity, there wasgreater alpha reduction in the right hemisphere than in lefthemisphere in all conditions. Right hemisphere constrainedmodulation of alpha seems to be a robust phenomenon as ithas been recently replicated in new studies in which attendingto global or local levels in Navon stimuli modulated thecongruency effect in the composite face illusion (Gao,Flevaris, Robertson, & Bentin, 2011). As in the present study,regardless of whether the Navon stimuli were processed atthe local or the global level, the alpha reduction whilepreparing for the composite-face-matching task was largerover the right hemisphere than the left hemisphere. Notethat this asymmetrical distribution of neural activity is inline with evidence suggesting right hemisphere dominance invisuospatial attention (e.g., studies of patients with unilaterallesions show more prolonged and severe attentional

deficits following right hemisphere than left hemispheredamage; Becker & Karnath, 2007; Bowen, McKenna, &Tallis, 1999; Heilman, Watson, & Valenstein, 1985;Mesulam, 1999; Ringman et al., 2004). Further corroborat-ing this idea, a number of neuroimaging studies have foundgreater right hemisphere activity associated with shifts ofspatial attention and target detection (e.g., Gitelman et al.,1999; Nobre et al., 1997; Shulman et al., 2010). Finally, weshould note that although not statistically significant in thealpha range, theta reduction was significantly greater overthe right than the left hemisphere sites during thepreparatory period preceding the Navon in the currentexperiment.Although we did not find significantly greater alpha

reduction in the left hemisphere than in the right hemi-sphere while preparing for local targets, the importantpoint is that we observed a significant SF � level �hemisphere interaction. This is actually consistent withfindings across the literature in global versus localprocessing as a whole, in which the simple effects areoften not found to be significant for either the global orlocal task or both (e.g., Robertson et al., 1988; Weissman& Banich, 1999; Weissman & Woldorff, 2005). Instead,task � hemisphere interactions such as the ones foundhere have become the standard measures of hemi-spheric asymmetry (Hellige, 1983; Robertson et al., 1993;Weissman & Banich, 1999). In addition, it is consistentwith proposals that posit a small bias one way or the otherin terms of relative SFs, which appears in turn to operatein terms of a balance of activity between the twohemispheres (see Sergent, 1982).Preparing for SF selection did not modulate theta

activity. Akin to the alpha analyses, the only significanteffects in the theta band occurred for the time intervalpreceding the Navon. We found greater overall thetareduction in the right than the left hemisphere, especiallyin the global preparation condition, as evidenced by thesignificant level � hemisphere interaction. This is con-sistent with findings associating theta oscillations withperceiving a global object in the context of an ambiguousimage and, specifically, with the integration of LSF visualfeatures in the experience of that global percept (Smith,Gosselin, & Schyns, 2006). However, that study did notexamine preparatory EEG activity, and future research isneeded to clarify the role of preparatory theta oscillationsas they relate to global versus local processing.In conclusion, the data from the current experiment are

consistent with previous research showing that the hemi-spheric asymmetry in global versus local processingoccurs not only during “bottom-up” perceptual processingof global versus local objects but also during “top-down”preparation for selecting those global/local objects (e.g.,Volberg et al., 2009; Yamaguchi et al., 2000). The presentdata extend previous work by suggesting that attentionalselection of spatial frequency modulates the relativeneural activity in right and left hemispheres wheninvolved in preparing for global or local processing.

Journal of Vision (2011) 11(7):11, 1–12 Flevaris, Bentin, & Robertson 9

Acknowledgments

This work was supported by an NIMH Grant to Lynn C.Robertson and Shlomo Bentin (Ro1 MH 64458). Lynn C.Robertson is a Senior Research Career Scientist in VAClinical Sciences Research Service, Department of Vet-erans Affairs Medical Center, Martinez, CA. Anastasia V.Flevaris is now at Department of Neurosciences, UCSD.

Commercial relationships: none.Corresponding author: Anastasia V. Flevaris.Email: ani.flevaris@gmail.com.Address: Department of Neurosciences, University ofCalifornia, San Diego, 9500 Gilman Drive #0608, La Jolla,CA 92093, USA.

Footnote

1Given that our primary focus was on preparatory neural

activity, it did not make sense to analyze oscillations in thebeta (14–30 Hz) and gamma (30–60 Hz) bands, as activity inboth of these frequency ranges generally occurs after stimulusonset and has been used as an index of visual perception (inthe case of beta activity; Piantoni, Kline, & Eagleman, 2010;Smith et al., 2006) and is associated with higher levelcognitive processes, including the integration of features intoa coherent percept (in the case of gamma activity; Rodriguezet al., 1999; Tallon-Baudry & Bertrand, 1999; Zion-Golumbic & Bentin, 2007). Indeed, visual inspection ofthe spectral plots in the beta and gamma ranges did not showvery much power in these ranges.

References

Bastiaansen, M. C., Bocker, K. B., Brunia, C. H., deMunck, J. C., & Spekreijse, H. (2001). Event-relateddesynchronization during anticipatory attention for anupcoming stimulus: A comparative EEG/MEG study.Clinical Neurophysiology, 112, 393–403.

Becker, E., & Karnath, H. O. (2007). Incidence of visualextinction after left versus right hemisphere stroke.Stroke, 38, 3172–3174.

Boeschoten, M. A., Kemner, C., Kenemans, J. L., & vanEngeland, H. (2005). The relationship between localand global processing and the processing of high andlow spatial frequencies studied by event-related poten-tials and source modeling. Cognitive Brain Research,24, 228–236.

Bowen, A., McKenna, K., & Tallis, R. C. (1999). Reasons forvariability in the reported rate of occurrence of unilateralspatial neglect after stroke. Stroke, 30, 1196–1202.

Broadbent, D. E. (1977). The hidden preattentive pro-cesses. American Psychologist, 32, 109–118.

Christman, S., Kitterle, F. L., & Hellige, J. (1991). Hemi-spheric asymmetry in the processing of absolute versusrelative spatial frequency. Brain and Cognition, 16,62–73.

Delis, D. C., Robertson, L. C., & Efron, R. (1986).Hemispheric specialization of memory for visualhierarchical stimuli. Neuropsychologia, 24, 205–214.

De Valois, R. L., & De Valois, K. (1988). Spatial vision.New York: Oxford University Press.

Fink, G. R., Halligan, P. W., Marshall, J. C., Frith, C. D.,Frackowiak, R. S., & Dolan, R. J. (1997). Neuralmechanisms involved in the processing of global andlocal aspects of hierarchically organized visualstimuli. Brain, 120, 1779–1791.

Flevaris, A. F., Bentin, S., & Robertson, L. C. (2010).Local or global? Attentional selection of spatialfrequencies binds shapes to hierarchical levels.Psychological Science, 21, 424–431.

Flevaris, A. F., Bentin, S., & Robertson, L. C. (2011).Attention to hierarchical level influences attentionalselection of spatial scale. Journal of ExperimentalPsychology: Human Perception and Performance, 37,12–22.

Flevaris, A. V., Robertson, L. C., & Bentin, S. (2008).Using spatial frequency scales for processing facefeatures and face configuration: An ERP analysis.Brain Research, 1194, 100–109.

Gao, Z., Flevaris, A. V., Robertson, L. C., & Bentin, S.(2011). Revisiting processing-bias effects on thecomposite-face illusion: EEG evidence for predom-inant right-hemisphere involvement. Paper presentedat the Visual Science Society 11th Annual Meeting,Naples, Florida, USA.

Gitelman, D. R., Nobre, A. C., Parrish, T. B., LaBar, K. S.,Kim, Y.-H., Meyer, J. R., et al. (1999). A large-scaledistributed network for covert spatial attention. Brain,122, 1093–1106.

Goffaux, V., Hault, B., Michel, C., Vuong, Q. C., &Rossion, B. (2005). The respective role of low andhigh spatial frequencies in supporting configural andfeatural processing of faces. Perception, 34, 77–86.

Goffaux, V., & Rossion, B. (2006). Faces are “spatial”VHolistic face perception is supported by low spatialfrequencies. Journal of Experimental Psychology:Human Perception and Performance, 32, 1023–1039.

Han, S., Weaver, J. A., Murray, S. O., Kang, X., Yund, E.W.,& Woods, D. L. (2002). Hemispheric asymmetry inglobal/local processing: Effects of stimulus position andspatial frequency. Neuroimage, 17, 1290–1299.

Journal of Vision (2011) 11(7):11, 1–12 Flevaris, Bentin, & Robertson 10

Heilman, K. M., Watson, R. T., & Valenstein, E. (1985).Neglect and related disorders. In K. M. Heilman &E. Valenstein (Eds.), Clinical neuropsychology(pp. 243–293). New York: Oxford.

Hellige, J. B. (1983). Feature similarity and lateralityeffects in visual masking. Neuropsychologia, 21,633–639.

Hubner, R., & Volberg, G. (2005). The integration ofobject levels and their content: A theory of global/local processing and related hemispheric differences.Journal of Experimental Psychology: Human Percep-tion and Performance, 31, 520–541.

Ivry, R. B., & Robertson, L. C. (1998). The two sides ofperception. Cambridge, MA: The MIT Press.

Jacobs, R. A., & Kosslyn, S. M. (1994). Encoding shapeand spatial relations: The role of receptive field size incoordinating complementary representations. CognitiveScience, 18, 361–386.

Kitterle, F. L., Christman, S., & Hellige, J. B. (1990).Hemispheric differences are found in the identi-fication, but not the detection, of low versus highspatial frequencies. Perception & Psychophysics, 48,297–306.

Kitterle, F. L., Hellige, J. B., & Christman, S. (1992).Visual hemispheric asymmetries depend on whichspatial frequencies are task relevant. Brain & Cognition,20, 308–314.

Klimesch, W., Doppelmayr, M., & Hanslmayr, S. (2007).EEG alpha oscillations: The inhibition-timing hypoth-esis. Progress in Brain Research, 159, 151–165

Lakatos, P., Shah, A. S., Knuth, K. H., Ulbert, I., Karmos, G.,& Schroeder, C. E. (2005). An oscillatory hierarchycontrolling neuronal excitability and stimulus pro-cessing in the auditory cortex. Journal of Neuro-physiology, 94, 1904–1911.

Lamb, M. R., Robertson, L. C., & Knight, R. T. (1989).Attention and interference in the processing of global andlocal information: Effects of unilateral temporal–parietaljunction lesions. Neuropsychologia, 27, 471–483.

Loftus, G. R., & Harley, E. M. (2004). How differentspatial-frequency components contribute to visualinformation acquisition. Journal of ExperimentalPsychology: Human Perception and Performance,30, 104–118.

Martin, M. (1979). Hemispheric specialization for localand global processing. Neuropsychologia, 17, 33–40.

Martinez, A., Moses, P., Frank, L., Buxton, R., Wong, E.,& Stiles, J. (1997). Hemispheric asymmetries inglobal and local processing: Evidence from fMRI.Neuroreport, 8, 1685–1689.

Mesulam, M. M. (1999). Spatial attention and neglect:Parietal, frontal and cingulate contributions to the

mental representation and attentional targeting of salientextrapersonal events. Philosophical Transactions of theRoyal Society B: Biological Sciences, 354, 1325–1346.

Mevorach, C., Humphreys, G. W., & Shalev, L. (2005).Effects of saliency, not global dominance, in patientswith left parietal damage. Neuropsychologia, 44,307–319.

Mevorach, C., Humphreys, G. W., & Shalev, L. (2006).Opposite biases in salience-based selection for the leftand right posterior parietal cortex. Nature Neuro-science, 9, 740–742.

Morrison, D. J., & Schyns, P. G. (2001). Usage of spatialscales for the categorization of faces, objects, andscenes. Psychonomic Bulletin & Review, 8, 434–469.

Nobre, A. C., Sebestyen, G. N., Gitelman, D. R.,Mesulam, M. M., Frackowiak, R. S. J., & Frith, C. D.(1997). Functional localization of the system forvisuospatial attention using positron emission tomog-raphy. Brain, 120, 515–553.

Pfurtscheller, G., Stancak, A., & Neuper, C. (1996). Event-related synchronization (ERS) in the alpha bandVAnelectrophysiological correlate of cortical idling: Areview. International Journal of Psychophysiology, 24,39–46.

Piantoni, G., Kline, K. A., & Eagleman, D. M. (2010). Betaoscillations correlate with the probability of perceivingrivalrous visual stimuli. Journal of Vision, 10(13):18,1–11, http://www.journalofvision.org/content/10/13/18,doi:10.1167/10.13.18. [PubMed] [Article]

Pineda, J. O., & Oberman, L. M. (2006). What goadscigarette smokers to smoke? Neural adaptation andthe mirror neuron system. Brain Research, 1121,128–135.

Ringman, J. M., Manly, R., Beschin, N., Daini, R.,Haeske-Dewick, H., Homberg, V., et al. (2004).Frequency, risk factors, anatomy, and course ofunilateral neglect in an acute stroke cohort. Neurology,63, 468–474.

Robertson, L. C., & Delis, D. C. (1986). ‘Part–whole’processing in unilateral brain-damaged patients: Dys-function of hierarchical organization. Neuropsychologia,24, 363–370.

Robertson, L. C., & Ivry, R. (2000). Hemisphericasymmetries: Attention to visual and auditory primi-tives. Current Directions in Psychological Science, 9,59–63.

Robertson, L. C., & Lamb, M. R. (1991). Neuropsycho-logical contributions to theories of part/whole organ-ization. Cognitive Psychology, 23, 299–330.

Robertson, L. C., Lamb, M. R., & Knight, R. T. (1988).Effects of lesions of temporal–parietal junction on

Journal of Vision (2011) 11(7):11, 1–12 Flevaris, Bentin, & Robertson 11

perceptual and attentional processing in humans.Journal of Neuroscience, 8, 3757–3769.

Robertson, L. C., Lamb, M. R., & Zaidel, E. (1993).Interhemispheric relations in processing hierarchicalpatterns: Evidence from normal and commissurotom-ized subjects. Neuropsychology, 7, 325–342.

Rodriguez, E., George, N., Lachaux, J. P., Martinerie, J.,Renault, B., & Varela, F. (1999). Perception’sshadow: Long-distance synchronization of humanbrain activity. Nature, 397, 430–433.

Sasaki, Y., Hadjikhani, N., Fischl, B., Liu, A. K., Marrett, S.,Dale, A. M., et al. (2001). Local and global attentionare mapped retinotopically in human occipital cortex.Proceedings of the National Academy of Sciences of theUnited States of America, 98, 2077–2082.

Schyns, P. G., & Oliva, A. (1999). Dr. Angry andMr. Smile:When categorization flexibly modifies the perceptionof faces in rapid visual presentation. Cognition, 69,243–265.

Sergent, J. (1982). The cerebral balance of power:Confrontation or cooperation? Journal of ExperimentalPsychology: Human Perception & Performance, 8,253–272.

Sergent, J. (1983). Role of input in visual hemisphericasymmetries. Psychological Bulletin, 93, 481–512.

Shulman, G. L., Pope, D. L. W., Astafiev, S. V.,McAvoiy, M. P., Snyder, A. Z., & Corbetta, M.(2010). Right hemisphere dominance during spatialselective attention and target detection occurs outsidethe dorsal frontoparietal network. Journal of Neuro-science, 30, 3640–3651.

Shulman, G. L., Sullivan, M. A., Gish, K., & Sakoda,W. J. (1986). The role of spatial-frequency channelsin the perception of local and global structure.Perception, 15, 259–273.

Shulman, G. L., & Wilson, J. (1987). SF and selectiveattention to local and global information. Perception,16, 89–101.

Smith, M. L., Gosselin, F., & Schyns, P. G. (2006).Perceptual moments of conscious visual experienceinferred from oscillatory brain activity. Proceedings ofthe National Academy of Sciences, 103, 5626–5631.

Tallon-Baudry, C., &Bertrand, O. (1999). Oscillatory gammaactivity in humans and its role in object representation.Trends in Cognitive Sciences, 3, 151–162.

Tallon-Baudry, C., Bertrand, O., Peronnet, F., & Pernier, J.,(1998). Induced gamma-band activity during thedelay of a visual short-term memory task in humans.Journal of Neuroscience, 18, 4244–4254.

Thut, G., Nietzel, A., Brandt, S. A., & Pascal-Leone, A.(2006). Alpha-band electroencephalographic activityover occipital cortex indexes visuospatial attentionbias and predicts visual target detection. Journal ofNeuroscience, 26, 9494–9502.

Volberg, G., & Hubner, R. (2004). On the role of responseconflicts and stimulus position for hemispheric differ-ences in global/local processing: An ERP study.Neurophsychologia, 42, 1805–1813.

Volberg, G., Kliegl, K., Hansimayr, S., & Greenlee, M. W.(2009). EEG alpha oscillations in the preparation forglobal and local processing predicts behavioral per-formance. Human Brain Mapping, 30, 2173–2183.

Ward, L. M. (2003). Synchronous neural oscillations andcognitive processes. Trends in Cognitive Sciences, 7,553–559.

Watt, R. J. (1988). Visual processing: Computational, psycho-physical, and cognitive research. London: Erlbaum.

Weissman, D. H., & Banich, M. T. (1999). Global-localinterference modulated by communication betweenthe hemispheres. Journal of Experimental Psychology:General, 128, 283–308.

Weissman, D. H., & Woldorff, M. G. (2005). Hemisphericasymmetries for different components of global/localattention occur in distinct temporo-parietal loci.Cerebral Cortex, 15, 870–876.

Worden, M. S., Foxe, J. J., Wang, N., & Simpson, G. V.(2000). Anticipatory biasing of visuospatial attentionindexed by retinotopically specific alpha-band elec-troencephalography increases over occipital cortex.Journal of Neuroscience, 20, RC63.

Yamaguchi, S., Yamagata, S., & Kobayashi, S. (2000).Cerebral asymmetry of the “top-down” allocation ofattention to global and local features. Journal ofNeuroscience, 20, RC72.

Zion-Golumbic, E., & Bentin, S. (2007). Dissociated neuralmechanisms for face detection and configural encod-ing: Evidence from N170 and induced gamma-bandoscillation effects. Cerebral Cortex, 17, 1741–1749.

Journal of Vision (2011) 11(7):11, 1–12 Flevaris, Bentin, & Robertson 12