B R A I N R E S E A R C H 1 3 1 9 ( 2 0 1 0 ) 9 2 – 1 0 2

ava i l ab l e a t www.sc i enced i r ec t . com

www.e l sev i e r . com/ loca te /b ra i n res

Research Report

Functional differences of low- and high-frequency oscillatorydynamics during illusory border perception

Conrado A. Bosmana,⁎, Francisco Zamoranob, Francisco Aboitizb,⁎aDonders Institute for Brain, Cognition and Behaviour, Radboud University Nijmegen, Kapittelweg 29, 6525 EN Nijmegen, The NetherlandsbLaboratorio de Neurociencia Cognitiva, Departamento de Psiquiatría, Centro de Investigaciones Médicas,Pontificia Universidad Católica de Chile, Marcoleta 387 2

dopiso. Casilla 114-D, Santiago 1, Chile

A R T I C L E I N F O

⁎ Corresponding authors. F. Aboitiz is to beDepartamento de Psiquiatría, Escuela de MeDonders Institute for Brain, Cognition and Be6525 EN Nijmegen, The Netherlands. Fax: +3

E-mail addresses: c.bosman@donders.ru.nURL: http://www.ru.nl/fcdonders (C.A. BoAbbreviations: ERP, event-related potentia

chart

0006-8993/$ – see front matter © 2010 Elsevidoi:10.1016/j.brainres.2010.01.004

A B S T R A C T

Article history:Accepted 1 January 2010Available online 11 January 2010

Neuronal oscillations are an essential part of several brain processes like perception,memory and attention, but little is known about the simultaneous dynamics of differentfrequency bands in such processes. Here, we studied the low- and high-frequency dynamicsof neuronal oscillations in a task that involves covert searching and visual perception of anillusory contour figure (Kanisza square). Subjects were cued to a particular region of thespace, and then a matrix of illusory contour inductors appeared. Illusory contour couldappear at the cued position, in a different position or not appear at all. We analyze thecondition in which an illusory contour appears and those in which the illusory borders donot appear. During the visual perception of the illusory contour, we found a significantincrease in a narrow and sustained gamma frequency band (55–65 Hz). We also found atransient increase in the theta band (4–10 Hz), 0.3 s after stimuli onset. This low-frequencypower increase correlates strongly with the detection of the illusory contour and it isinversely correlated with the gamma-band response. When illusory contours were absent, asustained low-frequency response in the same frequency range was present untilappearance of the motor response. The intensity of this low-frequency power responsewas also related to the speed of the motor response. We hypothesize that theta and gammainteract to process different aspects of the task. Gamma would be related to perception ofthe stimuli, while theta would be involved in searchmechanisms or the initiation of amotorresponse.

© 2010 Elsevier B.V. All rights reserved.

Keywords:Illusory border perceptionGamma power modulationTheta power modulationFrequency dynamicsFourier analysisVisuomotor interactionCross-frequency analysis

contacted at Laboratorio de Neurociencia Cognitiva, Centro de Investigaciones Médicas,dicina, Pontificia Universidad Católica de Chile, Chile. Fax: +56 2 638 0443. C.A. Bosman,haviour, Centre for Cognitive Neuroimaging, Radboud University Nijmegen, Kapittelweg 29,1 24 36 10989.l (C.A. Bosman), faboitiz@puc.cl (F. Aboitiz).

sman).ls; IC, illusory contours; NS, non-square; S, square; RTs, reaction times; TFC, time-frequency

er B.V. All rights reserved.

93B R A I N R E S E A R C H 1 3 1 9 ( 2 0 1 0 ) 9 2 – 1 0 2

1. Introduction

A prominent characteristic of brain electrophysiologicalrecordings is the presence of ongoing, rhythmic and periodicfluctuations of the electric potential (Berger, 1929; Buzsáki andDraguhn, 2004; Buzsáki, 2006). These neuronal oscillationsreflect intrinsic network properties, which define periodicwindows of neuronal excitability and subsequently, aneffective mechanism of neuronal communication betweenbrain areas that oscillate synchronously (Fries, 2005; Womels-dorf et al., 2007).

The role of neuronal oscillations in different sensory, motorand cognitive processes has been extensively reviewed in theliterature (Buzsáki andDraguhn, 2004; Buzsáki, 2002; Engel et al.,2001; Fries, 2009; Jensenetal., 2007; Singer andGray, 1995;Varelaet al., 2001;Womelsdorf and Fries, 2007). Particularly interestingis the finding that low-frequency neuronal rhythms modulatethe strength and precision of oscillations at higher frequencybands (Bragin et al., 1995; Canolty et al., 2006; de Lange et al.,2008; Händel and Haarmeier, 2008; Jensen and Colgin, 2007;Lakatos et al., 2005, 2008; Osipova et al., 2008). In this sense, ithas been recently described that sensory stimulation mightreset the phase of low-frequency oscillations widespread in theneocortex, rendering higher frequency oscillations particularlysuitable to process new incoming information (Lakatos et al.,2005, 2007, 2008; Rizzuto et al., 2003; Shah et al., 2004). Thus, theinteractionbetween low-andhigh-frequencyoscillationsmightimplement a neuronal mechanism of searching and sensoryselection (Fries, 2009; Schroeder and Lakatos, 2009). However,we do not yet fully understand the relation between differentneuronal oscillations during search and how this might affectsensory processes and motor behavior.

A candidate paradigm to study these interactions is thevisual processing of illusory contours (IC) (Kanizsa, 1976). ICseem to be processed primarily at areas V1 and V2 (Ffytcheand Zeki, 1996; Stanley and Rubin, 2003), but also in extra-striate regions (Halgren et al., 2003; Mendola et al., 1999),suggesting that IC processing is the result of an activeintegration of these areas. Related to this is the early findingof Tallon-Baudry et al. (1996), who showed a power increaseat gamma frequency band at fixed latencies after the IC

Fig. 1 – Task paradigm. Numbers below panels indicate time of pthe screen, and then a square appears at the cued position. Thenon-square (NS) and theywere displayed for 0.4 s. Thewindow reafter stimulus onset.

presentation, a phenomenon that has been replicated in otherstudies (Csibra et al., 2000; Herrmann et al., 1999; Lachaux etal., 2000; Schadow et al., 2009; Spencer et al., 2003). Further-more, IC processing is modulated by localization in the visualfield, attention task difficulty and expectancy among othercognitive processes (Herrmann et al., 1999; Murray et al., 2002;Senkowski and Herrmann, 2002; Senkowski et al., 2005),making this phenomenon a suitable model to evaluatedifferent frequency interactions.

In this study, we focused in the oscillatory dynamics duringan IC perception task using a paradigm that involved searchand perception of an IC square between distracters (ICinducers at different orientations) (Fig. 1). Subjects were cuedto a particular position on amonitor screen and therefore theyhad to report the presence or absence of an IC square amongdistracters. If the IC did not appear at the cued position, theyhad to covertly search for it until responding whether thesquare was present in another position or absent. Weassessed two conditions, namely, “square” (S) condition inwhich subjects found the IC square at a cued position in thecenter of themonitor and “non-square” (NS) condition inwhichno IC was presented. The aim of this design was to put the Sstimuli in a predictable but uncertain context that might elicitthe event-related component (ERP) P300 recognizable as an “endof task signal” (Sutton et al., 1965), and on the other handtriggered an active search during the absence of the square atcued position. This task allowed us to compare the oscillatorydynamicsof ICperceptual processingwith theeliciteddynamicsof object search in which the IC processing is absent.

We observed a rich neuronal oscillatory dynamics betweenlow and high frequencies during both the perception of the Scondition and the searching process at NS condition thatsupport the notion of a relation between active searching andlow-frequency oscillations (Schroeder and Lakatos, 2009).

2. Results

2.1. Behavioral results

We first evaluated the general performance of the subjects inthe task. Percentages of correct responses were high and they

resentation. Task starts with a fixation cross in the center of2 conditions assessed in this experiment are square (S) orsponse starts immediately after stimulus onset and ends 2.2 s

Fig. 2 – (A) Percentages of correct responses (average acrosssubjects±SEM). (B) Reaction times (RTs) in seconds (averageacross subjects±SEM) of correct responses. Asteriskrepresents significant differences (p<0.01 two-tailed t-test).

94 B R A I N R E S E A R C H 1 3 1 9 ( 2 0 1 0 ) 9 2 – 1 0 2

did not statistically differ between square (S) and non-square(NS) conditions (Fig. 2A, S: 96.6±2.0% versus NS: 93.4±1.9%,two-tailed t-test, t=1.57, p=0.144). On the other hand, whenillusory borders were present (S condition), reaction times(RTs) were significantly faster than RTs obtained at theabsence of illusory borders (NS condition) (Fig. 2B, S: 0.64±0.06 s versus NS: 0.80±0.07 s, two-tailed t-test, t=3.75,p<0.01). Thus, subjects correctly recognize the presence orabsence of illusory borders and their responses are fasterwhen they detect the pre-cued illusion than when searchingfor the illusion among the distracters.

2.2. Event-related potentials

As expected, both conditions elicited a series of visuallyevoked event-related potentials (ERPs), widely distributedacross the scalp (Fig. 3A). Fig. 3B shows the topographicaldistribution of the most prominent ERPs elicited by theparadigm. Stimulus onset at both conditions evoked earlyN100/P100 ERP. Maximum amplitudes of these componentsare found at posterior electrodes (Fig. 3B, first row). In addition,the S condition triggered a late P300 response whose

Fig. 3 – (A) Event-related potentials for S (blue) and NS (red) conditShaded regions represent the windows of averaging for the topoand NS conditions. Above: averages over gray region (0.1–0.2 po(0.3–0.4 post-stimulus latencies). (C) Statistical plot of significant d(red color shows channels with p<0.001). The p values were calcExperimental procedures). Electrodes were ordered according to

maximum peak is projected over the central region of thescalp (Fig. 3b, second row). Point-wise statistical comparisonsbetween S and NS conditions show significant differences asearly as 70ms after stimulus presentation, coincident with thepeak of the P1 component (Fig. 3C). This is consistent withprevious reports (Murray et al., 2002) and reveals that illusoryborders are processed at early stages in sensory cortex(Stanley and Rubin, 2003).

2.3. Time-frequency analysis

We then focused at the power amplitudes induced by stimulusonset. Fig. 4 shows the respective time-frequency charts(TFCs) for S and NS conditions, obtained from the electrodesthat show significant differences between conditions (seebelow).

The S condition was characterized by a sustained andnarrow band increase in the gamma frequency range (55–65 Hz) that started around 0.2 s and lasted until 0.8 s afterstimulus on. This increase is located mainly at occipitalelectrodes. At low frequencies (below 10 Hz), there were twocontinuous peaks of power increase. One started immediatelyafter the stimulation and reached a peak at 0.1 s. At 0.2 s afterstimulus onset, a second peak at the same frequency bandwas observed. Immediately before the averagemotor response(around 0.64 s), the power amplitude came back to pre-stimulus levels. A transient decrease in the amplitude powerof frequencies between 10 and 20 Hz starting at 0.4 s afterstimulus onset was also observed. On the other hand, the NScondition elicited a transient increase in gamma-band powerat 0.4 s after stimulus onset, a 10- to 20-Hz power reduction atlatencies similar as those observed in the S condition and,more importantly, a sustained post-stimulus power increasebelow 10Hz that started at∼0.1 s and covered the time intervalbetween the average RT response±1 SD (See SupplementaryFig. 1).

A detailed characterization of these findings in terms offrequency dynamics, significance level (p<0.05 using a point-wise Wilcoxon rank sum test) and topography are provided in

ions (Channel 41 in Fig. 9). Time 0 represents stimulus onset.graphic plots displayed in panel B. (B) Topographic plots of Sst-stimulus latencies). Below: averages over light blue regionifferences between S andNS during the time course of the ERPulated across subjects according a two-tailed t-test (seethe region of interest presented in Fig. 7.

Fig. 4 – Time-frequency chart of power increase relative to pre-stimulus baseline (Z-values). Average across electrodes thatshow significant differences. S (left) and NS (right) conditions. Continuous black line represents stimulus onset. Broken linerepresents RT average response for each condition.

95B R A I N R E S E A R C H 1 3 1 9 ( 2 0 1 0 ) 9 2 – 1 0 2

Fig. 5. Fig. 5A shows the average time course of gamma-bandactivity (from 55 to 65 Hz) of electrodes with significantdifferences between conditions. Consistent with previousreports (Csibra et al., 2000; Herrmann et al., 1999; Lachaux etal., 2000; Spencer et al., 2003; Tallon-Baudry et al., 1996), the Scondition, which contain the cued IC, shows significantdifferences at gamma-band activity that arises ∼0.4 s afterstimulus onset compared with the NS condition. Fig. 5B showsthe topography of the gamma power differences between bothexperimental conditions (S–NS) when they are statisticallysignificant. Significant differences are located at occipitalelectrodes, also consistent with previous reports (Tallon-Baudry et al., 1996). Fig. 5C shows the power dynamics of thelow-frequency oscillations. Consistent with the onset of theP300, the S condition displays a significant power increase at

Fig. 5 – (A) Statistical comparison of averaged gamma-band powelectrodes shown in panel B. Power values are Z-transformed andregion represents significant differences between conditions (p<plot of the subtraction between S and NS average gamma-band duin panel A). Highlighted electrodes represent electrodes with siglow-frequency band power (4–10 Hz). (D). Same as panel B, but fo

0.25 s respecting NS condition, which rapidly drops to baselinelevels at 0.5 s. On the other hand, activity in the NS conditionmaintains higher than baseline levels until the end of theanalysis window. The late NS-S difference reaches statisticalsignificance before and after ∼0.8 s. The topographic map ofthe condition differences (NS-S in this case) shows a patternsimilar to the differences in gamma frequency band (Fig. 5D),significant electrodes concentrating over right temporal-parietal regions.

2.4. Behavioral effects of low-frequency dynamics

The analysis so far demonstrated a different low-frequencydynamics during the perceived IC condition (square) com-paredwith the non-perceived IC condition (non-square). It has

er (55–65 Hz) between conditions. Average across significantbaseline corrected. Blue: S condition. Red: NS condition. Gray0.05 point-wise Wilcoxon rank sum test). (B) Topographicalring significant differences between conditions (gray regions

nificant values. (C) Same as panel A, but for average of ar the low-frequency band.

96 B R A I N R E S E A R C H 1 3 1 9 ( 2 0 1 0 ) 9 2 – 1 0 2

been shown that efficiency on a visuomotor task requiresthe precise engagement of the visual cortex to a particularlynarrow gamma frequency band (Bosman et al., 2009;Womelsdorf et al., 2006). Here, we evaluate whether randompower fluctuations of the low-frequency oscillation couldimpair the performance of the task. To that effect, we usedthe RTs as a measure of a visuomotor transformation andwe compared the low-frequency power associated withfast responses compared with those associated to slowresponses (Fig. 6).

Fig. 6A shows the low-frequency power modulation (4–10 Hz) of all the S trials, pooled across subjects. As expected,both the first and the second peaks of increased power arelocked to the stimulus onset. Trials with faster RTs elicit morepower at a latency of 0.3 s post-stimulus onset. Fig. 6B showsthe significant difference between the fastest (first quartile)and slowest (fourth quartile) trials of the power increaseelicited around 0.3 s post-stimulus onset (two-tailed t-test,t=6.44 p<0.001). On the other hand, the low-frequencyactivation at the NS condition is not stimulus locked since itfollows the RTs (Fig. 6C). Therefore, we compared a 0.1-ssegment window immediately before themotor response. Thecomparison between quartiles also shows a relation betweenRTs and the signal power (Fig. 6D, two-tailed t-test, t=3.38p<0.001). We subsequently test if this relation is also

Fig. 6 – (A) Low-frequency power dynamics (4–10 Hz) across all S cvalues were baseline corrected and Z-transformed. Solid line: stimVertical smoothingwindow: 50 trials. (B) Power (mean±SEM) comin S condition. Analysis based on a 0.1-s window between 0.25 aobtained with p<0.001, two-tailed t-test. (C) Same as panel A, buAnalysis based on a 0.1-s segment window that precedes the RT

manifested in the time domain (Fig. 7). In both conditions,the temporal sequence of the ERPs is preserved independentlyof the velocity of the motor response, but amplitudes at 0.3 slatencies are severely affected by RTs (Figs. 7A and C). Theamplitude of the P300 response, clearly distinguishable in theS condition (Fig. 3), is significantly bigger than those responsestriggered by trials with slow RTs (Fig. 7B, two-tailed t-test,t=17.77 p<0.001). Interestingly, faster trials in the NS condi-tion also elicit a significant major amplitude at the samelatency (Fig. 7D, two-tailed t-test, t=13.42 p<0.001).

Then, we focused at the gamma frequency band elicited bythe S condition (Fig. 8A). Previous studies had been showndifferent functional modulations between low- and high-frequency rhythms (Bragin et al., 1995; Canolty et al., 2006; deLange et al., 2008; Händel and Haarmeier, 2008; Jensen andColgin, 2007; Lakatos et al., 2005, 2008; Osipova et al., 2008).Here, we extend such analyses by performing a power–powercorrelation between low (4–10 Hz)- and high (55–65 Hz)-frequency bands as a function of RTs (Fig. 8b). We found astrong and significant inverse correlation between bothfrequency bands as a function of RTs (R=−0.45, p<0.05permutation test). Thus, in those trials in which responsesare faster, gamma frequency power response elicited by thetask is lower compared by the power response observed attheta band, while the opposite is observed at lower trials.

ondition trials collected over 12 subjects, sorted by RTs. Powerulus onset. Broken line: time latency of behavioral response.

parison between trials (1st and 4th quartile) sorted across RTs.nd 0.35 s post-stimulus onset. Significant differences weret for NS condition. (D) Same as panel B, but for NS condition.values.

Fig. 7 – (A) Event-related potentials response across all S condition trials collected over 12 subjects, sorted by RTs. Single-trialresponses were normalized by their baseline activity. Solid line: stimulus onset. Broken line: time latency of the behavioralresponse. Vertical smoothing window: 50 trials. (B) P300 signal amplitude (mean±SEM) comparison between RTs sorted trials(1st and 4th quartile). Significant differences were obtained with p<0.001, two-tailed t-test. (C) Same as panel A, but for NScondition. (D) Same as panel B, but for NS condition.

97B R A I N R E S E A R C H 1 3 1 9 ( 2 0 1 0 ) 9 2 – 1 0 2

3. Discussion

We found a rich oscillatory neuronal dynamics associatedwith IC processing and visual search mechanisms. These

Fig. 8 – (A) High-frequency power dynamics (55–65 Hz) across allPower values were baseline corrected and Z-transformed. Solid lresponse. Vertical smoothing window: 50 trials. (B) Scatter plot ohigh (55–64)-frequency band power estimates (R=−0.45, p<0.05 pbetween conditions (time latency: 0.35–0.45 s post-stimulus onsRTs.

results appears in the context of a task that requires theoperation of different cognitive processes. First, the conditionin which the IC is presented at the cued location requires acovert orienting of attention towards the cued location and atarget “match” response. By contrast, the condition in which

S condition trials collected over 12 subjects, sorted by RTs.ine: stimulus onset. Broken line: time latency of behavioralf trial-based Pearson correlation between low (4–10 Hz)- andermutation test) across channels with significant differenceset). Each dot represents an average of 50 trials sorted across

98 B R A I N R E S E A R C H 1 3 1 9 ( 2 0 1 0 ) 9 2 – 1 0 2

the IC is not presented requires a covert orienting of attentiontowards the cued location, a memory-match comparison witha mismatch outcome and subsequent visual search mechan-isms. It could be argued that due to these differences in taskrequirements, both conditions are difficult to compare.However, the processing of IC as well as the visual searchmechanisms are well contrasted in S and NS conditions.Therefore, our subsequent analysis was focused on theneuronal responses triggered by these processes.

Previous studies had shown a power increase of gammafrequency band after IC presentation (Csibra et al., 2000;Herrmann et al., 1999; Lachaux et al., 2000; Spencer et al., 2003;Tallon-Baudry et al., 1996). Here, besides the gamma frequen-cy power increment, we also describe a low-frequencyoscillation dynamic. In the context of this paradigm, thepower modulation of this low-frequency oscillation mightengage brain resources into a searching task or might providean efficient visuomotor transformation.

The original description of IC gamma-band processinginvolves two responses (Tallon-Baudry et al., 1996). First, atransient, stimulus-locked and broadband (30–80 Hz) powerincrease at 0.1 s after stimulus onset, and second a broadband,not stimulus-locked power increase at latencies around 0.3 spost-stimulus onset. These two power increments had beencalled “evoked” and “induced” gamma, respectively (Tallon-Baudry and Bertrand, 1999), and it has been claimed that thedynamics of these two components might underlie differentfunctions during cognitive processing (Herrmann et al., 1999,2004; Tallon-Baudry et al., 1998; Tallon-Baudry and Bertrand,1999). However, recent studies suggest that the “induced”gamma-band activitymight be the result of artifacts caused bythe EEG representation of extraocular muscles during micro-saccades, together with the use of a nose reference (Yuval-Greenberg et al., 2008). Despite this, we consider improbablethat in our study the gamma activity elicited by the ICprocessing be attributed to microsaccades for several reasons.We use as a reference the electrode Cz, instead of the nose.Even though the Cz reference may hide some significantdynamics at central and parietal regions, it allows us to showthat significant increases in gamma are more posterior thatprevious studies and are therefore less sensitive to ocularmovement artifacts. More importantly, we found a sustainedband-specific increase of gamma power during IC processing.This sustained increase is similar to what has been describedin human MEG studies (Hoogenboom et al., 2006; Vidal et al.,2006; Wyart and Tallon-Baudry, 2008) and monkeys (Fries etal., 2008; Rols et al., 2001). Finally, post-stimulus latencies inthis study are different to those described previously (Herr-mann et al., 1999; Spencer et al., 2003; Tallon-Baudry et al.,1996). We observed an increase of power that starts at ∼0.15 sand last until 0.8 s.We are tempted to suggest that this gammaincrease is related to motor preparation (see below), althoughthere was no condition without a motor response to test thispossibility. Despite the above controversy, it is important tonote that this study contributes with evidence supporting theoriginal idea that gamma power increase is relevant for ICprocessing (Tallon-Baudry and Bertrand, 1999).

In addition, a rich dynamics emerged from 4 to 10 Hz low-frequency band, a range of frequencies that mostly encom-passes the theta frequency band. Theta oscillations have been

classically observed in the rodent hippocampus duringlocomotion, spatial navigation, location, memory and othervoluntary behaviors (Buzsáki, 2005). Theta rhythm might alsoplay a role in other cognitive tasks such as memory, itemrecognition, retrieve lexical information and task switching(Klimesch et al., 1996; Osipova et al., 2006; Jensen and Tesche,2002; Sauseng et al., 2006; Bastiaansen et al., 2008). Ourfindings also suggest a more general involvement of thetaband in cognitive processing. The observed late increment ofpower in theta and the P300 ERP component have the samepost-stimulus latency (around 0.3 s) during the presentation ofthe IC (S condition). Thus, both components might befunctionally related (Mazaheri and Jensen, 2008; Shah et al.,2004) and can be considered as “end of task signals” thatreflect the match between the expected and the observedfigure (Herrmann et al., 2004). Similar paradigms have shownthe occurrence of a theta burst synchronization in concomi-tance with a P300 potential (Makeig et al., 2004), which mightreflect the context update of memory events, following goal-directed actions. Furthermore, our results might suggest thatpower increase at theta frequency could in turn be associatedwith the preparation of a motor response. In this context, wefound a significant difference on amplitude power betweenthe fastest and slowest trials and an inverse relation betweenthe power of theta and gamma-band as a function of the speedof the motor response. This later finding is particularlyinteresting and might support the idea of an interplaybetween high and low-frequency oscillations that mightcontrol the efficiency of sensory selection and visuomotortransformations (Bosman et al., 2009; Schroeder and Lakatos,2009; Womelsdorf et al., 2006).

On the other hand, theta dynamics in the NS conditiondisplayed a different pattern.We found a sustained incrementof power amplitude at theta following the first burst, whichremained after 1.0 s post-stimulus onset, in a time intervalcovered by the variance of RT across trials. It is noteworthythat some studies had also shown the involvement of thetarhythm in the generation of searching strategies and sensoryselection (Schroeder and Lakatos, 2009; Fries, 2009). Recentresearch suggests the involvement of low-frequency neuronaloscillations during search. In monkeys, Bichot et al. (2005)described the coexistence of theta and gamma frequencybands during a serial search task. More indirectly, a recentlydescribed low-frequency rhythm of saccades and microsac-cades in humans and monkeys might reveal a rhythmicexploration of the visual environment (Bosman et al., 2009;Maldonado et al., 2008; Otero-Millan et al., 2008). Furthermore,this low-frequency rhythm modulates the incoming visualgamma oscillation during stimuli processing and affects theresponse behavior (Bosman et al., 2009).

Taking all this evidence together, we speculate a possiblemechanism that might explain our findings. Gamma powerseems to be relevant in the IC perceptual processing andperhaps in the initiation of a motor response due to the latelatencies involved. On the other hand, theta rhythm plays arelevant role in the two processes: (1) the generation of anactive search processing that engages the visual system in anobject search mode, possibly by adjusting the gain of thesensory system (Lakatos et al., 2005, 2008; Schroeder andLakatos, 2009), and (2) a power resetting of theta band that

99B R A I N R E S E A R C H 1 3 1 9 ( 2 0 1 0 ) 9 2 – 1 0 2

allows the visual system to start a motor response and makethe visual system suitable to process new information orrepresents a signal that initiates a motor neuronal assembly(Rizzuto et al., 2003; Shah et al., 2004). Additional studies arerequired to test this hypothesis.

4. Experimental procedures

4.1. Subjects

Twelve subjects (6 females), with a mean age of 22 years(range, 18–27 years) provided a written informed consent toparticipate in the study. All were right handed, with normal orcorrected-to-normal vision and without history of neurolog-ical or psychiatric disorders.

4.2. Experimental task

Sequence task was implemented in STIM software (Neuros-can, Compumedics) and presented in a computer monitorlocated 45 cm in front of the subjects. All stimuli were blackcolored and presented over a gray Gaussian noise backgroundto decrease contrast and afterimage effects. The stimuli weregenerated in Matlab (Mathworks Inc.) using custom maderoutines. They consist in a matrix of 77 illusory bordersinducers (pac-men), organized in 7 rows and 11 columns, eachone with a diameter of 2°, separated by 3° from the next one.The orientation of the aperture of each inducer was definedrandomly among the matrix. Subjects had to detect thepresence or absence of an illusory border square formed byfour inducers, among the entire matrix of distracters. Thepossible position was previously cued by a black solid squareof 5° diameter (Fig. 1). Subjects had to orient their attention tothe cued position in the space, but keeping the eyes fixed in afixation point (cross in the center of the screen), presentedduring the complete task. After the presentation of the stimuli(0.4 s), subjects had to respond as fast as possible if the illusoryborder square was present or absent pressing one of tworesponse buttons with the right index finger (Fig. 1).

Illusory border squares could appear or not with equalprobability in 3 positions: center (1° of eccentricity taking thecenter of the monitor as a reference), right and left lateralpositions (12° of eccentricity). A block of this task consisted in30 trials of each condition, randomly presented. In a usualsession, subjects performed two blocks of the task. Wesubsequently studied the two conditions presented in thecenter, defined as square (S) or non-square (NS).

4.3. EEG data acquisition

Continuous EEG was acquired at a sampling rate of 1000 Hzthrough a Neuroscan Nuamps amplifier (Neuroscan, Compu-medics) using an 80-electrode cap arranged according to theinternational 1020 system. During recordings, 2 electrodeswere used as references (two mastoids) and 3 additionalelectrodes to monitor eye movements (2 periocular electrodesand 1 in the left malar to monitor horizontal and vertical eyemovements). All electrode impedances were set below 5kΩand referenced to Cz. All subsequent analyses were performed

using Matlab toolbox FieldTrip (http://www.ru.nl/neuroima-ging/fieldtrip/) and custom made routines.

4.4. Data analysis

Epochs of continuous EEG data were obtained for eachcondition and subjects, ranging from 0.5 s before theappearance of the inducer matrix to 1.3 s after stimulusonset. Epochswithmotor artifactswere visually inspected andsubsequently rejected. Eye blinks were identified by athreshold criterion of ±60 μV and their contribution removedfrom each data set using an Independent Component Analysisfilter (Jung et al., 2000).

Relevant segments containing no artifacts for both condi-tions (S and NS) were subsequently used to calculate event-related potentials (ERPs) and time-frequency estimates (TFEs)for each electrode and subject. Statistical analyses (see below)were performed at inter-subject level. Figures were obtainedby averaging ERPs and TFEs from different subjects.

ERPs waveforms were obtained using standard procedures(Picton et al., 2000). Data were digitally filtered with a band-pass filter between 0.5 and 25 Hz to remove 50-Hz line noisecomponents and subsequently averaged separately for eachcondition.

TFEs analyses were obtained using a short-time windowFourier analysis (Makeig, 1993). Relevant epochs were digitallyfiltered between 0.5 and 100 Hz. Line noise component wasremoved using a DFT filter centered at 50 and 100 Hz(Schoffelen et al., 2005). For each trial (and electrode andsubject), data were convolved with a 256-point sliding Han-ning window (with 90% of overlapping through the entire datasegment), padded with zeros until 1024 points. A fast Fouriertransform (FFT) algorithmwas applied to each epoch to obtainfrequency estimation. Frequency analyses were performedbetween 2 and 80 Hz with a frequency bin of 1 Hz. Thenormalization procedure was performed applying a Z-trans-formation of the data, using the mean baseline activity (from0.5 s to stimulus onset) for each data set. Electrodes werecompiled in region of interests (ROIs) for subsequent statisticalanalysis. The resulting Z-values were averaged across trialsfor each electrode (and subject). Thus, according to thisprocedure, we preserve both the oscillatory perturbationsthat are phase locked (i.e. evoked) and non-phase locked (i.e.induced) to the stimulus onset (Tallon-Baudry et al., 1996;Tallon-Baudry and Bertrand, 1999; Vidal et al., 2006).

Power estimation at theta frequency (4 to 10 Hz) was alsoused tomeasure the efficiency of visuomotor transformations.We adjusting a general procedure detailed elsewhere (Lakatoset al., 2008; Makeig et al., 2004; Womelsdorf et al., 2006). Inthose channels that show significant differences at thetafrequency band between conditions, baseline-corrected singletrials for both conditions were compiled across subjects,sorted according to RTs and finally smoothed with a slidingaverage window of ±25 trials. Then, we extracted the 25%fastest/slowest trials and compared the Z-transformed poweramplitude elicited by those trials at the frequency of interest.The same procedure was applied to the gamma frequencyband and the resulting ERPs. Finally, we evaluated thecorrelation between low (4 to 10 Hz)- and high (55 to 65 Hz)-frequency power as a function of RT in the S condition. To that

Fig. 9 – Electrode distribution. Regions of interest used in Fig.3C are defined by the horizontal lines.

100 B R A I N R E S E A R C H 1 3 1 9 ( 2 0 1 0 ) 9 2 – 1 0 2

effect, we calculated the Pearson correlation coefficientbetween the baseline-corrected power estimates at 0.35–0.45 s post-stimulus latencies across averages of 50 trialssorted by RTs.

4.5. Statistical analyses

ERPs significant differences between conditionswere assessedusing the approach of Murray et al. (2002). Briefly, point-wisepaired t-tests between S and NS conditions were calculatedacross 75 electrodes. Differenceswere considered significant ifthe obtained p value was under the α value of 0.01. Electrodeswere grouped in 3 ROIs: posterior, central and frontal.Corresponding electrodes were as follows: posterior, electro-des 4 5 17 18 19 20 21 22 23 25 39 40 41 42 43 44 45 46 47 48 62 6364 65 66 67; central, electrodes 1 2 3 6 7 8 9 13 14 15 16 26 27 28 2935 36 37 38 49 50 51 52 58 59 60 61 68 69 70 71 73 74 75; andfrontal, electrodes 10 11 12 30 31 32 33 34 53 54 55 56 57 72(Fig. 9).

Evaluation of statistical differences was assessed using anonparametric point-wise Wilcoxon rank sum test acrosselectrodes (Lachaux et al., 1999, 2005). To avoid loss ofstatistical power due to multiple comparisons, band-basedfrequencies of interest were pooled before the statisticalcomparison. We set an α value of 0.05 to consider significantdifferences between conditions. Channels that show statisti-cal differences were grouped and average for single-trialanalysis. The statistical significance of the power–powercorrelation was evaluated using a permutation approach(Maris and Oostenveld, 2007). We randomly shuffled thepower estimates of one frequency bin and then we calculatedthe Pearson correlation coefficient. This procedure wasrepeated 10.000 times. Thus, we obtained a distribution ofPearson coefficients under the null hypothesis of no power–power correlation across trials. From this distribution, wedetermined the 2.5% and 97.5% points and we evaluated theobserved correlation coefficient again these thresholds. Thisprocedure corresponds to a two-sided test which a false-positive rate of 5%.

Acknowledgments

The authors want to thank Pablo Gaspar, Eugenio Rodríguez,Vladimir López and Javier López-Calderón for their help andcommentaries of a previous version of this manuscript.Supported by: MIDEPLAN-MECESUP-PUC005 (Conrado Bos-man) and Millenium Center for Integrative Neuroscience(F. Aboitiz).

Appendix A. Supplementary data

Supplementary data associated with this article can be found,in the online version, at doi:10.1016/j.brainres.2010.01.004.

R E F E R E N C E S

Bastiaansen, M.C., Oostenveld, R., Jensen, O., Hagoort, P., 2008. Isee what you mean: theta power increases are involved in theretrieval of lexical semantic information. Brain Lang. 106,15–28.

Berger, H., 1929. Über das Elektroenkephalogramm des Menchen:1. Mitteilung. Arch. Psych. 87, 527–570.

Bichot, N.P., Rossi, A.F., Desimone, R., 2005. Parallel and serialneural mechanisms for visual search in macaque area V4.Science 308, 529–534.

Bosman, C.A., Womelsdorf, T., Desimone, R., Fries, P., 2009. Amicrosaccadic rhythm modulates gamma-bandsynchronization and behavior. J. Neurosci. 29,9471–9480.

Bragin, A., Jando, G., Nadasdy, Z., Hetke, J., Wise, K., Buzsaki, G.,1995. Gamma (40–100 Hz) oscillation in the hippocampus of thebehaving rat. J. Neurosci. 15, 47–60.

Buzsáki, G., Draguhn, A., 2004. Neuronal oscillations in corticalnetworks. Science 304, 1926–1929.

Buzsáki, G., 2002. Theta oscillations in the hippocampus. Neuron33, 325–340.

Buzsáki, G., 2005. Theta rhythm of navigation: link between pathintegration and landmark navigation, episodic and semanticmemory. Hippocampus 15, 827–840.

Buzsáki, G., 2006. Rhythms of the Brain. Oxford University Press,Oxford ; New York.

Canolty, R.T., Edwards, E., Dalal, S.S., Soltani, M., Nagarajan, S.S.,Kirsch, H.E., Berger, M.S., Barbaro, N.M., Knight, R.T., 2006. Highgamma power is phase-locked to theta oscillations in humanneocortex. Science 313, 1626–1628.

Csibra, G., Davis, G., Spratling, M.W., Johnson, M.H., 2000. Gammaoscillations and object processing in the infant brain. Science290, 1582–1585.

de Lange, F.P., Jensen, O., Bauer, M., Toni, I., 2008. Interactionsbetween posterior gamma and frontal alpha/beta oscillationsduring imagined actions. Front. Hum. Neurosci. 2, 7.

Engel, A.K., Fries, P., Singer, W., 2001. Dynamic predictions:oscillations and synchrony in top-down processing. Nat. Rev.Neurosci. 2, 704–716.

Ffytche, D.H., Zeki, S., 1996. Brain activity related to the perceptionof illusory contours. NeuroImage 3, 104–108.

Fries, P., 2005. A mechanism for cognitive dynamics: neuronalcommunication through neuronal coherence. Trends Cogn.Sci. 9, 474–480.

Fries, P., 2009. Neuronal gamma-band synchronization as afundamental process in cortical computation. Annu. Rev.Neurosci. 32, 209–224.

101B R A I N R E S E A R C H 1 3 1 9 ( 2 0 1 0 ) 9 2 – 1 0 2

Fries, P., Womelsdorf, T., Oostenveld, R., Desimone, R., 2008. Theeffects of visual stimulation and selective visual attention onrhythmic neuronal synchronization in macaque area V4. J.Neurosci. 28, 4823–4835.

Halgren, E., Mendola, J., Chong, C.D., Dale, A.M., 2003. Corticalactivation to illusory shapes as measured withmagnetoencephalography. NeuroImage 18, 1001–1009.

Händel, B., Haarmeier, T., 2008. Cross-frequency coupling of brainoscillations indicates the success in visual motiondiscrimination. NeuroImage 45, 1040–1046.

Herrmann, C.S., Mecklinger, A., Pfeifer, E., 1999. Gamma responsesand ERPs in a visual classification task. Clin. Neurophysiol. 110,636–642.

Herrmann, C.S., Munk, M.H., Engel, A.K., 2004. Cognitive functionsof gamma-band activity: memory match and utilization.Trends Cogn. Sci. 8, 347–355.

Hoogenboom, N., Schoffelen, J.M., Oostenveld, R., Parkes, L.M.,Fries, P., 2006. Localizing human visual gamma-band activity infrequency, time and space. NeuroImage 29, 764–773.

Jensen, O., Tesche, C.D., 2002. Frontal theta activity in humansincreases with memory load in a working memory task. Eur. J.Neurosci. 15, 1395–1399.

Jensen, O., Colgin, L.L., 2007. Cross-frequency coupling betweenneuronal oscillations. Trends Cogn. Sci. 11, 267–269.

Jensen, O., Kaiser, J., Lachaux, J.P., 2007. Human gamma-frequencyoscillations associated with attention and memory. TrendsNeurosci. 30, 317–324.

Jung, T.P., Makeig, S., Humphries, C., Lee, T.W., McKeown, M.J.,Iragui, V., Sejnowski, T.J., 2000. Removingelectroencephalographic artifacts by blind source separation.Psychophysiology 37, 163–178.

Kanizsa, G., 1976. Subjective contours. Sci. Am. 234, 48–52.Klimesch, W., Doppelmayr, M., Russegger, H., Pachinger, T., 1996.

Theta band power in the human scalp EEG and the encoding ofnew information. Neuroreport 7, 1235–1240.

Lachaux, J.P., Rodriguez, E., Martinerie, J., Varela, F., 1999.Measuring phase synchrony in brain signals. Hum. BrainMapp.8, 194–208.

Lachaux, J.P., Rodriguez, E., Martinerie, J., Adam, C., Hasboun, D.,Varela, F.J., 2000. A quantitative study of gamma-band activityin human intracranial recordings triggered by visual stimuli.Eur. J. Neurosci. 12, 2608–2622.

Lachaux, J.P., George, N., Tallon-Baudry, C., Martinerie, J.,Hugueville, L., Minotti, L., Kahane, P., Renault, B., 2005. Themany faces of the gamma band response to complex visualstimuli. NeuroImage 25, 491–501.

Lakatos, P., Shah, A.S., Knuth, K.H., Ulbert, I., Karmos, G.,Schroeder, C.E., 2005. An oscillatory hierarchy controllingneuronal excitability and stimulus processing in the auditorycortex. J. Neurophysiol. 94, 1904–1911.

Lakatos, P., Chen, C., Oconnell, M., Mills, A., Schroeder, C., 2007.Neuronal oscillations and multisensory interaction in primaryauditory cortex. Neuron 53, 279–292.

Lakatos, P., Karmos, G., Mehta, A., Ulbert, I., Schroeder, C., 2008.Entrainment of neuronal oscillations as a mechanism ofattentional selection. Science 320, 110–113.

Makeig, S., 1993. Auditory event-related dynamics of the EEGspectrum and effects of exposure to tones. Electroencephalogr.Clin. Neurophysiol. 86, 283–293.

Makeig, S., Delorme, A., Westerfield, M., Jung, T.P., Townsend, J.,Courchesne, E., Sejnowski, T.J., 2004. Electroencephalographicbrain dynamics following manually responded visual targets.PLoS Biol. 2, 747–762.

Maldonado, P.E., Babul, C., Singer, W., Rodriguez, E., Berger, D.,Grun, S., 2008. Synchronization of neuronal responses inprimary visual cortex of monkeys viewing natural images. J.Neurophysiol. 100, 1523–1532.

Maris, E., Oostenveld, R., 2007. Nonparametric statistical testing ofEEG- and MEG-data. J. Neurosci. Methods 164, 177–190.

Mazaheri, A., Jensen, O., 2008. Asymmetric amplitudemodulations of brain oscillations generate slow evokedresponses. J. Neurosci. 28, 7781–7787.

Mendola, J.D., Dale, A.M., Fischl, B., Liu, A.K., Tootell, R.B., 1999.The representation of illusory and real contours in humancortical visual areas revealed by functional magneticresonance imaging. J. Neurosci. 19, 8560–8572.

Murray, M.M.,Wylie, G.R., Higgins, B.A., Javitt, D.C., Schroeder, C.E.,Foxe, J.J., 2002. The spatiotemporal dynamics of illusorycontour processing: combined high-density electricalmapping, source analysis, and functional magnetic resonanceimaging. J. Neurosci. 22, 5055–5073.

Osipova, D., Takashima, A., Oostenveld, R., Fernandez, G., Maris,E., Jensen, O., 2006. Theta and gamma oscillations predictencoding and retrieval of declarative memory. J. Neurosci. 26,7523–7531.

Osipova, D., Hermes, D., Jensen, O., Rustichini, A., 2008. Gammapower is phase-locked to posterior alpha activity. PLoS ONE 3,e3990.

Otero-Millan, J., Troncoso, X.G., Macknik, S.L., Serrano-Pedraza, I.,Martinez-Conde, S., 2008. Saccades and microsaccades duringvisual fixation, exploration, and search: foundations for acommon saccadic generator. J. Vis. 8, 1–18.

Picton, T.W., Bentin, S., Berg, P., Donchin, E., Hillyard, S.A., JohnsonJr., R., Miller, G.A., Ritter, W., Ruchkin, D.S., Rugg, M.D., Taylor,M.J., 2000. Guidelines for using human event-related potentialsto study cognition: recording standards and publicationcriteria. Psychophysiology 37, 127–152.

Rizzuto, D.S., Madsen, J.R., Bromfield, E.B., Schulze-Bonhage, A.,Seelig, D., Aschenbrenner-Scheibe, R., Kahana, M.J., 2003. Resetof human neocortical oscillations during a working memorytask. Proc. Natl. Acad. Sci. U. S. A. 100, 7931–7936.

Rols, G., Tallon-Baudry, C., Girard, P., Bertrand, O., Bullier, J., 2001.Cortical mapping of gamma oscillations in areas V1 and V4 ofthe macaque monkey. Vis. Neurosci. 18, 527–540.

Sauseng, P., Klimesch, W., Freunberger, R., Pecherstorfer, T.,Hanslmayr, S., Doppelmayr, M., 2006. Relevance of EEG alphaand theta oscillations during task switching. Exp. Brain Res.170, 295–301.

Schadow, J., Dettler, N., Paramei, G.V., Lenz, D., Frund, I., Sabel, B.A.,Herrmann, C.S., 2009. Impairments of Gestalt perception in theintact hemifield of hemianopic patients are reflected in gamma-band EEG activity. Neuropsychologia 47, 556–568.

Schoffelen, J.M., Oostenveld, R., Fries, P., 2005. Neuronal coherenceas a mechanism of effective corticospinal interaction. Science308, 111–113.

Schroeder, C.E., Lakatos, P., 2009. Low-frequency neuronaloscillations as instruments of sensory selection. TrendsNeurosci. 32, 9–18.

Senkowski, D., Herrmann, C.S., 2002. Effects of task difficulty onevoked gamma activity and ERPs in a visual discriminationtask. Clin. Neurophysiol. 113, 1742–1753.

Senkowski, D., Rottger, S., Grimm, S., Foxe, J.J., Herrmann, C.S.,2005. Kanizsa subjective figures capture visual spatialattention: evidence from electrophysiological and behavioraldata. Neuropsychologia 43, 872–886.

Shah, A.S., Bressler, S.L., Knuth, K.H., Ding, M., Mehta, A.D., Ulbert,I., Schroeder, C.E., 2004. Neural dynamics and the fundamentalmechanisms of event-related brain potentials. Cereb. Cortex14, 476–483.

Singer, W., Gray, C.M., 1995. Visual feature integration and thetemporal correlation hypothesis. Annu. Rev. Neurosci. 18,555–586.

Spencer, K.M., Nestor, P.G., Niznikiewicz, M.A., Salisbury, D.F.,Shenton, M.E., McCarley, R.W., 2003. Abnormal neuralsynchrony in schizophrenia. J. Neurosci. 23, 7407–7411.

Stanley, D.A., Rubin, N., 2003. fMRI activation in response toillusory contours and salient regions in the human lateraloccipital complex. Neuron 37, 323–331.

102 B R A I N R E S E A R C H 1 3 1 9 ( 2 0 1 0 ) 9 2 – 1 0 2

Sutton, S., Braren, M., Zubin, J., John, E.R., 1965. Evoked-potentialcorrelates of stimulus uncertainty. Science 150, 1187–1188.

Tallon-Baudry, C., Bertrand, O., 1999. Oscillatory gamma activity inhumans and its role in object representation. Trends Cogn. Sci.3, 151–162.

Tallon-Baudry, C., Bertrand, O., Delpuech, C., Pernier, J., 1996.Stimulus specificity of phase-locked and non-phase-locked40 Hz visual responses in human. J. Neurosci. 16, 4240–4249.

Tallon-Baudry, C., Bertrand, O., Peronnet, F., Pernier, J., 1998.Induced gamma-band activity during the delay of a visualshort-term memory task in humans. J. Neurosci. 18,4244–4254.

Varela, F.J., Lachaux, J.P., Rodriguez, E., Martinerie, J., 2001. Thebrainweb: phase synchronization and large-scale integration.Nat. Rev. Neurosci. 2, 229–239.

Vidal, J., Chaumon, M., O'Regan, J.K., Tallon-Baudry, C., 2006.Visual grouping and the focusing of attention inducegamma-band oscillations at different frequencies in human

magnetoencephalogram signals. J. Cogn. Neurosci. 18,1850–1862.

Womelsdorf, T., Fries, P., 2007. The role of neuronalsynchronization in selective attention. Curr. Opin. Neurobiol.17, 154–160.

Womelsdorf, T., Fries, P., Mitra, P.P., Desimone, R., 2006.Gamma-band synchronization in visual cortex predicts speedof change detection. Nature 439, 733–736.

Womelsdorf, T., Schoffelen, J.M., Oostenveld, R., Singer, W.,Desimone, R., Engel, A.K., Fries, P., 2007. Modulation ofneuronal interactions through neuronal synchronization.Science 316, 1609–1612.

Wyart, V., Tallon-Baudry, C., 2008. Neural dissociation betweenvisual awareness and spatial attention. J. Neurosci. 28,2667–2679.

Yuval-Greenberg, S., Tomer, O., Keren, A.S., Nelken, I., Deouell,L.Y., 2008. Transient induced gamma-band response in EEG asa manifestation of miniature saccades. Neuron 58, 429–441.