This article appeared in a journal published by Elsevier. The attached copy is furnished to the author for internal non-commercial research and education use, including for instruction at the authors institution and sharing with colleagues. Other uses, including reproduction and distribution, or selling or licensing copies, or posting to personal, institutional or third party websites are prohibited. In most cases authors are permitted to post their version of the article (e.g. in Word or Tex form) to their personal website or institutional repository. Authors requiring further information regarding Elsevier’s archiving and manuscript policies are encouraged to visit: http://www.elsevier.com/copyright
Transcript
This article appeared in a journal published by Elsevier. The attachedcopy is furnished to the author for internal non-commercial researchand education use, including for instruction at the authors institution
and sharing with colleagues.
Other uses, including reproduction and distribution, or selling orlicensing copies, or posting to personal, institutional or third party
websites are prohibited.
In most cases authors are permitted to post their version of thearticle (e.g. in Word or Tex form) to their personal website orinstitutional repository. Authors requiring further information
regarding Elsevier’s archiving and manuscript policies areencouraged to visit:
Kai Kaspara,⁎, Uwe Hassler b, Ulla Martensb, Nelson Trujillo-Barreto c, Thomas Gruberb
aUniversity of Osnabrück, Institute of Cognitive Science, 49069 Osnabrück, GermanybInstitute of Experimental Psychology I, University of Osnabrück, GermanycCuban Neuroscience Center, Havana, Cuba
A R T I C L E I N F O A B S T R A C T
Article history:Accepted 28 April 2010Available online 5 May 2010
In present high density electroencephalogram (EEG) study, we examined steady-state visualevoked potential (SSVEP) correlates of object recognition. In SSVEP tasks a visual stimulus ispresented repetitively at a specific flickering rate and typically elicits a continuous oscillatorybrain response. This response is characterized by the same fundamental frequency as theinitiating stimulus. The stimulus material consisted of a series of pictures depicting familiarand unfamiliar objectswhich have been successfully applied in previous EEG studies on objectrecognition. In particular, we presented familiar and unfamiliar objects at rates of 7.5, 12 and15 Hz. At all three driving frequencies, we found specific SSVEPs that furthermore showedsignificant amplitude differences between familiar and unfamiliar objects. The familiar/unfamiliar effects were localized to early occipital, lateral occipital and temporal areas bymeans of VARETA (Variable Resolution Electromagnetic Tomography). Interestingly, themorphology of the familiar/unfamiliar effect differed between flicker rates. The 12 and 15 Hzconditions revealed higher SSVEP amplitudes for familiar as opposed to unfamiliar objects,whereas in the 7.5 Hz condition the effect was reversed. We concluded that SSVEPs aresensitive to stimuli's semantic content. Thus, SSVEP paradigms open new venues to studyobject recognition. Nonetheless, selecting appropriate driving frequencies is non-trivial,because flicker rate might have an influence on the observed effects.
Object recognition plays a central rolewithin the scope of visualinformation processing. During the last decade it has becomeone of the central research issues in neuroscience (see e.g.Martin et al., 1996; Riesenhuber and Poggio, 2000; Koutstaalet al., 2001; Simons et al., 2003; Bar et al., 2006; Peissig and Tarr,2007; Martin, 2007).
To examine the corticalmechanisms underlying the ability toclassify incoming sensory information as a meaningful object,
variousneuroscientificmethodshavebeenused. The timecourseof object recognition was examined by means of event-relatedpotentials (ERPs) (e.g. Doniger et al., 2000; JohnsonandOlshausen,2005; Thorpe et al., 1996). Invasive studies in animals as well asimaging studies in humans were used to identify the corticalregions involved in object recognition (e.g. Shmuelof and Zohary,2005; Grill-Spector et al., 2001; Tanaka, 2003). These investigationsrevealed an involvement of numerous spatially separated brainregions whenever a cortical object representation was activated.Using the human electro- and magnetoencephalogram (EEG/MEG), it was hypothesized that high-frequency oscillations above∼30 Hz (gamma band activity) mirror the activation of theseobject representations (Tallon-Baudry and Bertrand, 1999). To
B R A I N R E S E A R C H 1 3 4 3 ( 2 0 1 0 ) 1 1 2 – 1 2 1
⁎ Corresponding author.E-mail address: [email protected] (K. Kaspar).
find further support for this hypothesis, in a series of MEG/EEGstudies (e.g. Gruber et al., 2008a; Supp et al., 2007) the activationpatterns elicited by images of meaningful (i.e. familiar) stimuliwere compared to the activation patterns to meaningless (i.e.unfamiliar) pictures (these stimuli varied in the psychologicaldimension of stimulus familiarity but not in physical parameters;see also Busch et al. (2006) and Fig. 1 for examples). These studiesrevealed an augmentation of gamma band activity for familiar asopposed to unfamiliar stimuli. Based on these results it wasconcluded that familiar/unfamiliar designs are well suited toexamine electrophysiological correlates of object recognition.
Besides the above mentioned neuroscientific techniques,the so-called steady-state visual evoked potential (SSVEP) is afurther fruitful method to investigate cortical informationprocessing. The SSVEP is the electrophysiological oscillatoryresponse of the cortex to a flickering stimulus, which has thesame temporal frequency as the initiating stimulus (Regan,1989). This approach has been used in order to search fororientation and size detectors (Campbell and Maffei, 1970) instudies with primates (Nakayama and Mackeben, 1982) or inhuman EEG studies in the context of visual attention (Mülleret al., 1998; Di Russo and Spinelli, 1999; Müller et al., 2003, 2006;Andersen et al., 2009) andmemory processes (Silberstein et al.,2001).
The SSVEP has certain advantages over conventional EEGstudies: It is rapidly quantifiable in the frequency domain andprovides a continuous measure of cortical activation patternsassociated with the processing of a frequency-tagged stimulus(Müller and Hillyard, 2000). Furthermore, SSVEPs are charac-terized by excellent signal-to-noise ratios, a crucial pre-requisite for reliable source reconstructions.
Thus, the aim of the present study was to apply theadvantages of the steady-state technique to the familiar/unfamiliar design mentioned above. Specifically, we intendedto clarify if SSVEPs allow for a differentiation of corticalactivity evoked by flickering familiar as opposed to unfamiliarobjects. Furthermore, we aimed at elucidating the influence ofdifferent driving frequencies on SSVEP correlates of objectrecognition.
Significant differences between both types of stimuli wereevaluated separately for each flicker frequency by means ofwavelet transforms. Data were analyzed at the scalp level (i.e.in electrode space) as well as in the source space by means ofVARETA (Variable Resolution Electromagnetic Tomography;Bosch-Bayard et al., 2001).
2. Results
2.1. Behavioral data
In order to uphold subject's attention to the presented stimuli,they had to detect occasionally presented dots (targets). Onaverage it took participants 557 ms to detect a target (range402ms to 764 ms, standard deviation 78 ms). A STIMULUS TYPE(familiar versus unfamiliar) by FLICKER FREQUENCY (7.5, 12,15 Hz) repeated measures ANOVA neither revealed a statisti-cally significantmain effect of stimulus type [F(1,19)=.42; P=.52;ηp2=.02] nor amain effect of flicker frequency [F(2,38)=.19; P=.82;ηp2=.01]. Furthermore, the interaction termwasnot significant [F(2,38)=.40; P=.67; ηp2=.02], i.e. target detection performance wasneither influenced by stimulus familiarity nor by drivingfrequency.
2.2. SSVEPs in electrode space
Fig. 2 depicts the mean potentials (averaged across posteriorelectrodes) for all conditions showing clearly visible SSVEPsignatures at the related driving frequencies (7.5, 12, and 15 Hz).
In correspondence to the SSVEPs in the time domain, thespectrally decomposed signal revealed an amplitude increase atthe respective driving frequencies (and their first harmonic)starting at about 500ms after stimulus onset (see time byfrequency (TF) plot in Fig. 3). Additionally, the TF plots indicatethe typical reciprocal relationship between frequency andamplitude with higher driving frequencies resulting in loweramplitudes.
Based on the TF plots, the time window from 500–2500 msafter stimulus onset was chosen for subsequent analyses. Thetopographical amplitude distributions within the selected timewindow (atwavelets designed for 7.5, 12, and15 Hz respectively)were characterized by a maximum over posterior corticalregions (see Fig. 4, left panels: averages across familiar andunfamiliar stimuli are presented).
Difference topographies between familiar and unfamiliarpictures are depicted in the right panels of Fig. 4. Electrodesfulfilling the criterion of an activity difference of two or morestandard deviations above or below the average amplitude of allelectrodes were identified. (On the bottom-left of all differencetopographies we present the predicted topographies based onforward solutions of the related generators. The similarity
Fig. 1 – Excerpt of stimulus sequence. Familiar (meaningful) and unfamiliar (meaningless) colour pictures were presented inrandomized order at rates of 7.5, 12, or 15 Hz, respectively. Note: the rightmost unfamiliar picture was derived from the pictureof the giraffe on the left.
113B R A I N R E S E A R C H 1 3 4 3 ( 2 0 1 0 ) 1 1 2 – 1 2 1
between original and predicted topography speaks for the closecorrespondence between SSVEP source configurations (seebelow) and the original topographies).
In the 15 Hz condition only the electrodes indicating higheramplitudes for familiar as opposed to unfamiliar picturesresulted in significant differences [t(19)=2.27; P=.035; d=.51].The corresponding electrodes are marked by solid disks inFig. 4.
Fig. 5 depicts the time course of the wavelet designed forthe 15 Hz-band at selected electrodes (selection based on thecriterionmentioned above). Higher amplitudes for the familiaras opposed to the unfamiliar condition can be appreciatedthroughout the analyzed time window.
The 12 Hz condition revealed a very similar pattern of results(see Figs. 4 and 5) with a slightly different topographical
distribution of the effect [familiar/unfamiliar effect at selectedelectrodes: t(19)=3.25; P=.004; d=.72].
A different picture emerged for the 7.5 Hz condition:Although the topographical distribution of the effect is similarto the 15 and 12 Hz condition (see Fig. 4), here only theelectrodes indicating higher amplitudes for unfamiliar asopposed to familiar pictures showed a significant difference[t(19)=−2.84; P=.010; d=.64]. In other words, in the 7.5 Hzcondition unfamiliar pictures elicited higher SSVEP ampli-tudes as compared to familiar pictures throughout theanalyzed time window (see also Fig. 5). Please note that theline plots (Fig. 5) are derived from a more focused set ofelectrodes as compared to the TF plots in Fig. 3. Thus, theamplitude values in Fig. 5 are higher than in Fig. 3.
2.3. SSVEPs in source space
Sources of SSVEP effects were determined by means ofVARETA. The 15 Hz-flicker condition showed statisticallysignificant differences in SSVEP generators (familiar versusunfamiliar) in bilateral posterior regions (P<0.01). The centreof gravity of activity differences was located in the rightlingual gyrus (MNI coordinates X: 14, Y: −91, Z: −10). Furtherareas of statistically significant activity differences are listedin Table 1 and are marked in Fig. 6. Please note that VARETAcalculates volumetric solutions. Having a local tomographicalmaximum does not mean having point-like activations.Table 1 describes local and global centres of gravity. However,the activity is not restricted to these specific areas.
Bilateral sources of SSVEP differences between stimulusconditions were also found for both 12 Hz- and 7.5 Hz-flickerfrequencies. The centre of gravity at 12 Hz-flicker frequencywas located in the right lateral occipitotemporal gyrus (MNIcoordinates X: 21, Y: −91, Z: −10) and activity differences weresymmetrically spatially distributed. The centre of gravity at7.5 Hz-flicker frequency was found in the left occipital pole(MNI coordinates X: 76, Y: 70, Z: 188) with a slightly morepronounced distribution of statistically significant SSVEPdifferences in the right hemisphere.
Furthermore, Table 1 describes the generators of theSSVEPs itself (independent of familiar/unfamiliar differences).Notably, the generators of the SSVEP itself and of familiar/unfamiliar effects are very similar.
3. Discussion
In current SSVEP study we presented pictures of familiar andunfamiliar objects at different flicker frequencies to answerthe following two questions: (1) does the SSVEP morphologyallow for a differentiation between stimuli's semantic content(i.e. familiar versus unfamiliar objects) and (2) does the SSVEPmorphology differ between different driving frequencies.
(1) We found clear SSVEPs for those frequencies in whichour stimuli were presented. In the 15 Hz condition the SSVEPreached a maximum at approximately 500 ms after stimulusonset and remained on a similar level during the wholestimulation period. Both, the 7.5 Hz and 12 Hz condition,revealed a similar pattern characterized by slightly increasing
Fig. 2 – Baseline-corrected grand mean potentials (SSVEPs)for all flicker frequencies (15 Hz, 12 Hz, and 7.5 Hz) and bothstimulus types (familiar/unfamiliar) at posterior electrodesites.
114 B R A I N R E S E A R C H 1 3 4 3 ( 2 0 1 0 ) 1 1 2 – 1 2 1
Author's personal copy
amplitudes throughout stimulation. These findings are in linewith previous SSVEP studies (e.g. Müller et al., 1998) and showthat SSVEPs— although stimuli were presented for 3000 ms—are not prone to habituation.
Importantly, we found statistically significant amplitudedifferences between the SSVEP to familiar and unfamiliarobjects at all utilized driving frequencies. All effects werecharacterized by a more than moderate effect size (from .51 to.72; cf. Cohen, 1988). Hence independent of sample size andinter-subject variance, activity differences showed practicalsignificance. Thus, our results clearly show that SSVEPscannot be regarded as a simple natural resonance phenom-enon of the brain, which is not influenced by higher cognitivefunctions. In contrast, SSVEPs are highly sensitive to stimuli'ssemantic content (as shown in present study), but also byattention (as shown e.g. by Müller et al., 2003) and memory (asshown e.g. by Silberstein et al., 2001).
The topographical distribution of the familiar/unfamiliareffect was restricted to posterior electrode sites for all drivingfrequencies. This distribution is similar to SSVEP studiesexaminingmechanisms of attention bymeans of less complexflickering stimuli (e.g. flickering LEDs, Müller et al., 1998; orflickering dot clouds, Andersen et al., 2009). Furthermore, thescalp distribution of present familiar/unfamiliar effects is inline with EEG studies focusing on early electrophysiologicalmarkers of object recognition, namely high-frequency oscilla-tions >30 Hz at around 100 ms after stimulus onset (so-calledearly gamma band oscillations, Hermann et al., 2004a,b).
In close concordancewith the topographical SSVEP distribu-tion, source reconstructions of the observed effects revealedrelatively focal generators in bilateral occipital and temporalareas. This stands in contrast to a previous claim stating thatcortical object representations involve widespread networks inposterior, parietal and frontal areas (e.g. Gruber et al., 2006; Suppet al., 2007). This claim was based on source localizations of so-called late gamma band oscillations starting at around 250 ms
after stimulus onset— another prominent electrophysiologicalmarker of object recognition (cf. Tallon-Baudry and Bertrand,1999).However, present results are incloseagreementwith fMRIfindings by Kourtzi and Kanwisher (2000) and Grill-Spector et al.(2001). They describe an involvement of extra-striatal visualcortex in the representation and perception of objects: thelateral occipital complex (LOC). Comparing our source localiza-tions to e.g. Grill-Spector et al. (2001), it seems reasonable toconclude that our observed activation pattern can be ascribed tothe LOC. Importantly, fMRI findings by Marlach et al. (1995)revealed that LOCmight be “an intermediate link in the chain ofprocessing stages leading to object recognition in human visualcortex”. Based on this notion it can be speculated that SSVEPsmirror a similar “intermediate link”, i.e. a complementaryprocess as ascribed to late representational gamma bandresponses mentioned before.
We want to point out that EEG source analyses are spatiallyimprecise and one should be cautious when interpreting theresults. Nonetheless, it was demonstrated that the localizationerror of cLORETA (a technique highly comparable to VARETA)results in localization errors of about 1–2 cm (Trujillo-Barretoet al., 2004). One might argue that this only holds for localiza-tions based on individual MRI scans and digitized electrodepositions. However, across subject statistics based on individualMRI scans imply computational deformations of all images inorder to make them comparable. By doing so, one introduceserrors that are comparable to theuncertainty of thehereapplied‘average brain’ (particularly in the case of 20 participants).
(2) Most interestingly, the morphology of the familiar/unfamiliar effect differed between the 12 and 15 Hz conditionon the one hand and the 7.5 Hz condition on the other hand. Inthe 7.5 Hz case, unfamiliar objects elicited higher SSVEPamplitudes as opposed to familiar objects, whereas in thetwo other conditions the effect was found to be reversed.
Higher amplitudes for familiar objects elicited by the 12 and15 Hz driving frequencies are not surprising. As mentioned
Fig. 3 –Grandmean TF plots averaged across posterior electrodes for all flicker frequencies (averaged across familiar/unfamiliarstimuli). Note: different scaling for high and low flicker frequencies.
115B R A I N R E S E A R C H 1 3 4 3 ( 2 0 1 0 ) 1 1 2 – 1 2 1
above, LOC activation is usually higher in response to mean-ingful (i.e. familiar) as opposed to meaningless pictures. Since afamiliar stimulus activates more cortical representations invisual memory, one has to expect higher activity within therespective cortical areas. However, a second explanation is notunlikely: The amplitude of the 12 and 15 Hz SSVEP responsemight overlap with internally generated alpha oscillations.Usually, stimulus onset leads to a phenomenon known as
alpha-blocking or -suppression (Başar et al., 1999). The amountof alpha suppression is proportional to the attentional state ofthe participant, with stronger suppression being related to ahigher cortical activation (Pfurtscheller, 1989). Furthermore,previous studies revealed stronger alpha suppression forunfamiliar as compared to familiar material (Gruber et al.,2006). Thus, unfamiliar objects might attract more attentionthan familiar objects (resulting in stronger alpha suppression).
Fig. 4 – Left: Topographical distributions of 15 Hz, 12 Hz and 7.5 Hz SSVEP responses (500–2500 ms) averaged across familiar andunfamiliar pictures. Right: Difference topographies between familiar and unfamiliar pictures. Electrodes used for statisticalanalyses are marked by solid disks (see text for details on the selection procedure). Predicted topographies based on forwardsolutions of the SSVEP generators are presented to the bottom-left of the original difference topographies.
116 B R A I N R E S E A R C H 1 3 4 3 ( 2 0 1 0 ) 1 1 2 – 1 2 1
In turn, the externally driven increase in alpha activity (theSSVEP) might be less pronounced in response to unfamiliarpictures.
Obviously, both explanations are not mutually exclusive,because “attention” and “visual memory” are highly inter-mingled concepts (a less attended stimulus activates lesscortical representations; cf. Desimone, 1996). Interestingly,targets were detected equally well during the presentation offamiliar and unfamiliar pictures. Thus, target detectionperformance was not influenced by the speculative differentattentional demands required by the processing of familiarand unfamiliar pictures, respectively. Future studies, in whichattention is explicitly manipulated, have to clarify this issue.
Surprisingly, the 7.5 Hz condition resulted in higher SSVEPamplitudes for unfamiliar as opposed to familiar objects. Wecan only speculate that this effect is caused by a similar“overlap phenomenon” described before. It was suggested thatinternally generated theta responses (∼7 Hz) reflect access to
declarativememory (e.g. Jensen and Colgin, 2007; Gruber et al.,2008b). In this sense, the 7.5 Hz SSVEP response might overlapwith internally generated theta rhythms elicited by anongoing semantic integration attempt for unfamiliar stimuli.As argued before future studies have to clarify this issue; inthis case by explicitly manipulating mnemonic processes.
It has to bementioned that our approach to select appropriatesensors for statistical analyses in electrode spacemight fall underthe suspicion of “double dipping”. In EEG double dipping isdefined as the restriction of statistical analyses to a subset ofsensors that show expected responses (Kriegeskorte et al., 2009).However, two arguments speak against this confound. First, wehave not restricted the analysis to sensors showing the expectedresponsebutwehavealsoanalyzedsensorsshowingamaximumopposite effect. These analyses revealed no significant results.Secondandmore importantly, our results in electrode spacewereconfirmed by restriction-free analyses in source space. Although,the topographies of familiar/unfamiliar effects were slightlydifferent between the three flicker frequencies (15 Hz: two lateralfoci; 12 and 7.5 Hz: single posterior focus) and the tomographicaldistributions looked very similar for all driving frequencies, thisdoes not pose a contradiction. Even if the sources have the samemagnitude they can point in different directions, and thus,generate different topographies. To make this point clear, we
Fig. 5 – Grand mean response for the 15, 12, and 7.5 Hzwavelets for familiar and unfamiliar pictures averagedacross electrodes indicated in Fig. 4.
Table 1 – Areas of statistically significant SSVEPgenerators (P<.01). Areas tagged with (2) refer todifferences in SSVEP generators between familiar andunfamiliar conditions. Areas tagged with (3) refer togenerators which reveal significant effects of the SSVEPitself (i.e. averages across familiar and unfamiliarconditions versus baseline). Tag (1) indicates the Centresof Gravity of SSVEP generators (a = familiar versusunfamiliar, b = SSVEP versus baseline; in cases of bilateralactivations all centres of gravity were located in the righthemisphere).
SSVEP Anatomical descriptions of significant areas
have calculated the predicted topographical distribution of theeffects based on the related sources (forward solution). Theseanalyses revealed that the observed topographies can be wellexplained by the predicted topographies. In other words, sourceanalyses provided support for the appropriateness of the sensorsselected for statistical analyses in electrode space. Consequently,we are confident that the problem of double dipping does notapply to present analyses in electrode space.
Furthermore, the following point seems noteworthy: Thescalp topographies of the SSVEP itself (Fig. 4, left column) andthe difference topographies (familiar minus unfamiliar, Fig. 4,right column) look slightly different. Whereas the SSVEPtopography itself is characterized by a focal, posterior midlinedistribution, the familiar/unfamiliar effects are reflected atslightly more lateral electrodes. Nonetheless, the respectivesource solutions of the SSVEP itself and the familiar/unfamil-iar effect are very similar (cf. Table 1). This can be taken asevidence that different networks within similar activatedareas support the generation of the “pure” SSVEPs and effectsof object recognition.
To summarize, the present experimental design (presen-tation of familiar and unfamiliar pictures) in combinationwiththe SSVEP approach provides a useful tool to study mechan-isms of object recognition. However, one should be carefulwhen selecting appropriate driving frequencies, because thefrequency, in which a stimulus is presented, might have aninfluence on the observed effects. Before future studies furtherelucidate the plausible influence of internal ongoing oscilla-tions (e.g. theta and alpha) on externally driven SSVEPs, itmight be advisable to select driving frequencies which do notdirectly fall into these frequency bands.
4. Experimental procedures
4.1. Participants
Twenty healthy university students (8 male) participated inthe study. The average age was 25.2 years (±3.50 years). Allvolunteers agreed with the conditions of the experiment, had
Fig. 6 – Statistically significant SSVEP sources (familiar versus unfamiliar; P<0.01) in axial, sagittal and coronal planescontaining the centre of gravity marked by the intersection of vertical and horizontal lines. X-, Y-, and Z-coordinates representthe location of the slices in MNI space. The lightest shades of gray indicate the highest T2 values. L = left; R = right; A = anterior;P = posterior.
118 B R A I N R E S E A R C H 1 3 4 3 ( 2 0 1 0 ) 1 1 2 – 1 2 1
Author's personal copy
normal or corrected-to-normal visual acuity and no registeredneurological or psychiatric disorders. The study conformed tothe Code of Ethics of the American Psychological Associationand the World Medical Association.
4.2. Stimuli and procedure
Hundred and eighty six pictures of real world objects — takenfrom a standard picture library (Hemera Technologies, 1997) —served as familiar (meaningful) stimuli. Unfamiliar (meaning-less) pictures were created by randomly distorting the originalimages by using optical filters und effects until they were notrecognizable as meaningful objects anymore (for examples seeFig. 1). In order to make both familiar and unfamiliar stimulisimilar regarding physical parameters the spatial frequencies ofunfamiliar pictures were matched to their familiar counterpartby combining the amplitude spectrumof a familiar picture withthe phase spectrum of the corresponding unfamiliar picture.This procedure is described in detail in Busch et al. (2006).
Every image was presented only once in randomized orderand covered a visual angle of approximately 7×7°. Imageswere centrally shown on a light-gray background. Pictureonset was synchronized to the vertical retrace of the monitor.Each trial consisted of a randomized 500 to 800 ms baselineperiod during which a fixation cross (0.3×0.3°) was presented,followed by one of the images that was presented for 3000 ms.Stimuli were presented at a rate of 15 Hz, 12 Hz or 7.5 Hz,respectively. Themonitor was running at a frame rate of 60 Hzand the following duty-cycles were used: 15 Hz/1on-3off,12 Hz/1on-4off and 7.5 Hz/1on-7off. The assignment of aspecific picture to a flicker frequency was fully randomizedacross pictures and across participants. After presentation,stimuli were replaced by a signal cross which remained onscreen for another 800 ms and allowed participants to blink ifnecessary. Participants were asked to avoid eye movementsand blinking during the display of the fixation cross and thestimulus. Prior to the main experiment, participants per-formed a practice block of 20 trials. The experimental setupresulted in six experimental conditions, namely familiarity(familiar/unfamiliar) by flicker frequency (7.5, 12 and 15 Hz).Each condition consisted of 62 trials (overall 372 trials).Furthermore, in 20% of the trials a magenta-coloured dotwas superimposed on the pictures. The dot's detection had tobe signalled by pressing a key. These target detection trialswere introduced to uphold subject's attention to the pictures.The target appeared at a random position for 120 ms in a timewindow from 250 ms to 2750 ms after object onset. Targetdetection trials were excluded from subsequent EEG analyses.To allow for breaks, the experiment was subdivided into sixblocks, each consisting of 62 trials. At the end of each blockparticipants were provided with feedback regarding theirtarget detection rate.
4.3. Electrophysiological recordings
The EEGwas recorded continuously from 128 active electrodesusing a BioSemi Active-Two amplifier system (sampling rate512 Hz). Eye movements and blinks were monitored byrecording the horizontal and vertical electrooculogram (EOG).Two additional electrodes (CMS: Common Mode Sense and
DRL: DrivenRight Leg; cf. www.biosemi.com/faq/cms&drl.htm)served as recording reference and ground. For further analysisthe average reference was used. In line with several previousstudies (e.g. Keil et al., 2001; Busch et al., 2006; Gruber et al.,2006) artifact correctionwasperformedbymeansof “statisticalcorrection of artifacts in dense array studies — SCADS”(Junghöfer et al., 2000). In brief, this procedure uses acombination of trial exclusion and channel approximationbased on statistical parameters of the data. If less than 20channels are contaminated by an artifact the information ofeliminated electrodes is replaced with a spherical interpola-tion from the full channel set. With respect to the spatialarrangement of the approximated sensors, it was ensured thatthe rejected sensors were not located within one region of thescalp, since this would have made interpolation for this areainvalid. Single epochs with excessive eye movements andblinks or more than 20 channels containing artifacts werediscarded from further analyses.
Additionally, some rare artifacts that were not detected bySCADS were eliminated after visual inspection. The averagerejection rate of EEG-data after artifact correction wasapproximately 20% of the epochs.
4.4. Data analysis (A): behavioral data
Reaction times to targets (magenta dots) were analyzed toevaluate whether stimulus type or flicker frequency affectedtarget detection performance. To that end, we performed arepeated measures ANOVA with the factors STIMULUS TYPE(familiar versus unfamiliar) and FLICKER FREQUENCY (3)(Greenhouse–Geisser correction was applied). Only responsetimes from 100 to 1200 ms were taken into account. Effectsizes (partial eta squared, ηp2) were calculated.
4.5. Data analysis (B): SSVEPs in electrode space
In a first step, SSVEPs were visualized by averaging all trialsseparately for each condition. In the time domain, SSVEPswere plotted from 200 ms prior to stimulus onset (baseline) to2500 ms after stimulus onset. For these plots an averageacross posterior electrodes was used. The selection of electro-des was based on previous EEG studies on object recognition(e.g. Gruber and Müller, 2005; Gruber et al., 2006) and severalSSVEP findings (Müller et al., 1998, 2006, 2008).
To examine amplitude differences between familiar andunfamiliar stimuli at each driving frequency, in a next step thesignal was spectrally decomposed by means of wavelet trans-forms. Wavelet analysis provides a time-varying magnitude ofthesignal ineach frequencyband (BertrandandPantev, 1994). Inparticular, complex Morlet wavelets g can be generated in thetime domain for different analysis frequencies f0 according to
g t; f0ð Þ = A′e− t 2
2σ 2t e2iπf0t; ð1Þ
with A′ depending on the parameter σf, specifying the width ofthe wavelet in the frequency domain, the analysis frequency f0and the user-selected ratiom:
In order to achieve good time and frequency resolution inthe target frequency range, the used wavelet family is definedby a constant m= f0 /σf=12 with f0 ranging from 1 to 30 Hz (theso-generated wavelets can be regarded as “prototypes” of theto-be-detected oscillations). In a subsequent step, the familyof wavelets g(t,f0) was convoluted with the SSVEP (i.e. theaveraged EEG epochs across all trials). A convolution can bedescribed as the integral, which expresses the amount ofoverlap of g and the SSVEP. The absolute value of thisconvolution results in time×frequency (TF) representationsof the EEG signal. Based on the TF plots we extractedappropriate time×frequency windows for further analysesand plotted spherical spline interpolated topographical dis-tributions of the respective TF windows (topographies weregenerated within the software package EEGLAB; Delorme andMakeig, 2004). TF plots and topographical distributions wereaveraged across familiar and unfamiliar stimuli but separatelyfor each driving frequency. Furthermore, difference topogra-phies (familiar minus unfamiliar) were used to determineappropriate electrodes for further statistical analyses. Inparticular, we used electrodes showing an activity differenceof two or more standard deviations from the averageamplitude of all electrodes. Afterwards mean activity differ-ences between familiar and unfamiliar stimuli were comparedby means of planned t-tests for paired samples. Please notethat wewere reluctant to apply an ANOVA including the factorFLICKER FREQUENY, since SSVEP amplitudes are known toshow a reciprocal relationship between amplitude andfrequency. In addition, effect sizes as an indicator of practicalsignificance of activity differences were calculated accordingto Faul et al. (2007).
4.6. Data analysis (C): SSVEPs in source space
In order to localize the cortical generators of the statisticalsignificant activity differences between familiar and unfamil-iar stimuli, we applied VARETA (Variable Resolution Electro-magnetic Tomography, Bosch-Bayard et al., 2001). Thisprocedure provides the spatial intracranial distribution ofprimary current densities (PCD) in source space, which is bestcompatible with the amplitude distribution in electrode space(cf. Gruber et al., 2006). In particular, the SSVEP was trans-formed into the frequency domain as described above(wavelet analysis) and VARETA was applied to the complexwavelet coefficients. Due to the linear relationship betweenEEG and PCD, the complex source reconstructions can beinterpreted as an estimate of the wavelet coefficients of thePCD (complex inverse solution; see Trujillo-Barreto et al.,2004). As possible sources of the signal 3244 grid points(‘voxels’) of a 3D grid (7 mm grid spacing) were used. This gridand the arrangement of 128 electrodes were placed in
registration with the average probabilistic MRI atlas (‘averagebrain’) produced by the Montreal Neurological Institute (MNI;Evans et al., 1993). Statistical comparisons were carried out bymeans of Hotelling t2-tests against zero in order to localizedifferences in activation between familiar and unfamiliarstimuli (familiar minus unfamiliar separately for each drivingfrequency). Activation threshold corrections, accounting forspatial dependencies between voxels, were calculated bymeans of Random Field Theory (Worsley et al., 1996).Regarding all statistical parametric maps (SPMs), the resultswere thresholded at a significance level of P<0.01. Finally, theoutcomes were depicted as 3D activation images, constructedon the basis of the MNI average brain.
In order to document the close relationship betweenoriginal topographical distributions and the generators of theeffects, we calculated predicted topographical distributionsbased on separate forward solutions for each source config-uration. Forward solutions can be regarded as re-constructedvoltage matrices due to primary current densities at eachvoxel (for a detailed description see Gruber et al., 2006).
Additionally, we have calculated the generators of theSSVEP response itself by means of the same approach asdescribed above. However, in this case the average of bothstimulus conditions was tested against baseline (i.e. zero).This was done separately for each driving frequency.
Acknowledgment
We would like to thank Linda Zurlutter for the help in dataacquisition.
R E F E R E N C E S
Andersen, S.K., Müller, M.M., Hillyard, S.A., 2009. Color-selectiveattention need not bemediated by spatial attention. J. Vis. 9, 1–7.
Bar, M., Kassam, K.S., Ghuman, A.S., Boshyan, J., Schmid, A.M.,Dale, A.M., Hämäläinen, M.S., Marinkovic, K., Schacter, D.L.,Rosen, B.R., Halgren, E., 2006. Top-down faciliation of visualrecognition. PNAS 103, 449–454.
Başar, E., Başar-Eroğlu, C., Karakaş, S., Schürmann, M., 1999. Arecognitive processes manifested in event-related gamma, alpha,thetaanddeltaoscillations in theEEG?Neurosci. Lett. 259, 165–168.
Bertrand, O., Pantev, C., 1994. Stimulus Frequency Dependence ofthe Transient Oscillatory Auditory Evoked Response (40 Hz)Studied by Electric and Magnetic Recordings in Humans.Plenum Press, New York.
Bosch-Bayard, J., Valdés-Sosa, P., Virués-Alba, E., Aubert-Vázquez,E., John, E.R., Harmony, T., Riera-Diaz, J., Trujillo-Barreto, N.,2001. 3D statistical parametric mapping of variable resolutionelectromagnetic tomography (VARETA). Clin. Electroencephal.32, 47–66.
Busch, N.A., Herrmann, C.S., Müller, M.M., Lenz, D., Gruber, T.,2006. A cross-laboratory study of event-related gamma activityin a standard object recognition paradigm. Neuroimage 33,1169–1177.
Campbell, F., Maffei, L., 1970. Electrophysiological evidence for theexistence of orientation and size detectors in the human visualsystem. J. Physiol. 207, 635–652.
Cohen, J., 1988. Statistical Power Analysis for the BehaviouralSciences, Second ed. Erlbaum, Hillsdale.
Delorme, A., Makeig, S., 2004. EEGLAB: an open source toolbox foranalysisof single-trial EEGdynamics. J.Neurosci.Meth. 134, 9–21.
120 B R A I N R E S E A R C H 1 3 4 3 ( 2 0 1 0 ) 1 1 2 – 1 2 1
Faul, F., Erdfelder, E., Lang, A.-G., Buchner, A., 2007. G*Power 3: aflexible statistical power analysis program for the social,behavioral, andbiomedical sciences.Behav.Res.Meth. 39, 175–191.
Grill-Spector, K., Kourtzi, Z.,Kanwisher,N., 2001. The lateral occipitalcomplexand its role inobject recognition.Vis. Res. 41, 1409–1422.
Gruber, T., Müller, M.M., 2005. Oscillatory brain activity dissociatesbetween associative stimulus content in a repetition primingtask in the human EEG. Cereb. Cortex 15, 109–116.
Gruber, T., Trujillo-Barreto, N.J., Giabbiconi, C.-M., Valdés-Sosa, P.A.,Müller, M.M., 2006. Brain electrical tomography (BET) analysis ofinduced gamma band responses during a simple object recog-nition task. Neuroimage 29, 888–900.
Gruber, T., Maess, B., Trujillo-Barreto, N.J., Müller, M.M., 2008a.Sources of synchronized induced Gamma-band responsesduring a simple object recognition task: a replication study inhuman MEG. Brain Res. 1196, 74–84.
Gruber, T., Tsivilis, D., Giabbioconi, C.-M., Müller, M.M., 2008b.Induced electroencephalogram, oscillations during sourcememory: familiarity is reflected in the gamma band, recollectionin the theta band. J. Cogn. Neurosci. 20, 1043–1053.
Johnson, J.S., Olshausen, B.A., 2005. The earliest EEG signatures ofobject recognition in a cued-target task are postsensory. J. Vis.5, 299–312.
Junghöfer, M., Elbert, T., Tucker, D.M., Rockstroh, B., 2000.Statistical control of artifacts in dense array EEG/MEG studies.Psychophysiology 37, 523–532.
Keil, A., Müller, M.M., Gruber, T., Wienbruch, C., Elbert, T., 2001.Human large-scale oscillatory brain activity during an operantshaping procedure. Cogn. Brain Res. 12, 397–407.
Kourtzi, Z., Kanwisher, N., 2000. Cortical regions involved inperceiving object shape. J. Neurosci. 20, 3310–3318.
Koutstaal, W., Wagner, A.D., Rotte, M., Maril, A., Buckner, R.L.,Schacter, D.L., 2001. Perceptual specificity in visual object priming:functional magnetic resonance imaging evidence for a lateralitydifference in fusiform cortex. Neuropsychologia 39, 184–199.
Kriegeskorte, N., Simmons, W.K., Bellgowan, P.S.F., Baker, C.I.,2009. Circular analysis in systems neuroscience: the dangers ofdouble dipping. Nat. Neurosci. 12, 535-540.
Marlach, R., Reppas, J.B., Benson, R.R., Kwong, K.K., Jiang, H., Kennedy,W.A., Ledden, P.J., Brady, T.J., Rosen, B.R., Tootell, R.B.H., 1995.Object-related activity revealed by functional magnetic resonance
imaging in human occipital cortex. Proc. Natl Acad. Sci. U. S. A. 92,8135–8139.
Martin, A., 2007. The representation of object concepts in thebrain. Annu. Rev. Psychol. 58, 25–45.
Martin, A., Wiggs, C.L., Ungerleider, L.G., Haxby, J.V., 1996. Neuralcorrelates of category-specific knowledge. Nature 379,649–652.
Müller, M.M., Hillyard, S., 2000. Concurrent recording ofsteady-state and transient event-related potentials as indicesof visual–spatial selective attention. Clin. Neurophysiol. 111,1544–1552.
Müller, M.M., Teder-Sälejärvi, W., Hillyard, S.A., 1998. The timecourse of cortical facilitation during cued shifts of spatialattention. Nat. Neurosci. 1, 631–634.
Müller, M.M., Malinowski, P., Gruber, T., A., H.S., 2003. Sustaineddivision of the attentional spotlight. Nature 424, 309–312.
Müller, M.M., Andersen, S.K., Trujillo, N.J., Valdés-Sosa, P.,Malinowski, P., Hillyard, S.A., 2006. Feature-selective attentionenhances color signals in early visual areas of the humanbrain. PNAS 103, 14250–14254.
Müller, M.M., Andersen, S.K., Keil, A., 2008. Time course ofcompetition for visual processing resources betweenemotional pictures and foreground task. Cereb. Cortex 18,1892–1899.
Nakayama, K., Mackeben, M., 1982. Steady state visual evokedpotentials in the alert primate. Vis. Res. 22, 1261–1271.
Peissig, J.J., Tarr, M.J., 2007. Visual object recognition: do we knowmore now than we did 20 years ago? Annu. Rev. Psychol. 58,75–96.
Pfurtscheller, G., 1989. Spatiotemporal analysis of alpha frequencycomponents with the ERD technique. Brain Topogr. 2, 3–8.
Regan, D., 1989. Human Brain Electrophysiology: Evoked Potentialsand Evoked Magnetic Fields in Science and Medicine. Elsevier,New York, Amsterdam, London.
Riesenhuber, M., Poggio, T., 2000. Models of object recognition.Nat. Neurosci. 3, 1199–1204.
Shmuelof, L., Zohary, E., 2005. Dissociation between ventral anddorsal fMRI activation during object and action recognition.Neuron 47, 457–470.
Silberstein, R.B., Nunez, P.L., Pipingas, A., Harris, P., Danieli, F.,2001. Steady state visually evoked potential (SSVEP) topographyin a graded working memory task. Int. J. Psychophysiol. 42,219–232.
Simons, J.S., Koutstaal, W., Prince, S., Wagner, A.D., Schacter, D.L.,2003. Neuralmechanisms of visual object priming: evidence forperceptual and semantic distinctions in fusiform cortex.Neuroimage 19, 613–626.
Supp, G.G., Schlögl, A., Trujillo-Barreto, N., Müller, M.M., Gruber, T.,2007. Directed cortical information flow during human objectrecognition: analysing induced EEG gamma-band responses inbrain's source space. PLoS ONE 8, 1–11.
Tallon-Baudry, C., Bertrand, O., 1999. Oscillatory gamma activityin humans and its role in object representation. Trends Cogn.Sci. 3, 151–162.
Tanaka, K., 2003. Colums for complex visual object features in theinferotemporal cortex: clustering of cells with similar butslightly different stimulus selectivities. Cereb. Cortex 13, 90–99.
Thorpe, S., Fize, D., Marlot, C., 1996. Speed of processing in thehuman visual system. Nature 381, 520–522.
Trujillo-Barreto, N.J., Aubert-Vazquez, E., Valdés-Sosa, P.A., 2004.Bayesian model averaging in EEG/MEG imaging. Neuroimage21, 1300–1319.
Worsley, K.J., Marrett, S., Neelin, P., Evans, A.C., 1996. Searching scalespace for activation in PET images. Hum. Brain Mapp. 4, 74–90.
121B R A I N R E S E A R C H 1 3 4 3 ( 2 0 1 0 ) 1 1 2 – 1 2 1
