NeuroImage 60 (2012) 170–178

Contents lists available at SciVerse ScienceDirect

NeuroImage

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

It's only in your head: Expectancy of aversive auditory stimulation modulatesstimulus-induced auditory cortical alpha desynchronization

Thomas Hartmann a,⁎, Winfried Schlee b, Nathan Weisz a,c

a Department of Psychology, University of Konstanz, PO Box D25, 78457 Konstanz, Germanyb Institute for Psychology and Education, University of Ulm, 89069 Ulm, Germanyc Zukunftskolleg, University of Konstanz, PO Box D25, 78457 Konstanz, Germany

⁎ Corresponding author.E-mail address: thomas.hartmann@uni-konstanz.de

1053-8119/$ – see front matter © 2011 Elsevier Inc. Alldoi:10.1016/j.neuroimage.2011.12.034

a b s t r a c t

a r t i c l e i n f o

Article history:Received 6 September 2011Revised 9 December 2011Accepted 13 December 2011Available online 23 December 2011

Keywords:Auditory alphaTau rhythmAuditory cortexExpectancyEEG

Increasing evidence underlines the functional importance of non-phase-locked cortical oscillatory rhythms.Among the different oscillations, alpha (8–12 Hz) has been shown to be modulated by anticipation or atten-tion, suggesting a top-down influence. However, most studies to date have been conducted in the visual mo-dality and the extent to which this notion also applies to the auditory cortex is unclear. It is furthermore oftendifficult to dissociate bottom-up from top-down contributions in cases of different stimuli (e.g., standards vs.deviants) or stimuli that are preceded by different cues. This study addresses these issues by investigatingneuronal responses associated with intrinsically fluctuating perceptions of an invariant sound. Sixteen partic-ipants performed a pseudo-frequency-discrimination task in which a “high-pitch” tone was followed by anaversive noise, while the “low-pitch” tone was followed by silence. The participants had to decide whichtone was presented even though the stimulus was actually kept constant while pseudo-randomized feedbackwas given. EEG data show that auditory cortical alpha power decreased by 20% in “high-pitch” trials relativeto trials in which a “low pitch” was perceived. This study shows that expectancy of aversive feedback mod-ulates perception of sounds and these fluctuating perceptions become manifest in modulations of sound-related alpha desynchronizations. Our findings extend recent evidence in the visual and somatosensory do-main that alpha oscillations represent the excitatory/inhibitory balance of sensory cortical cell assemblies,which can be tuned in a top-down manner.

© 2011 Elsevier Inc. All rights reserved.

Introduction

An increasing amount of empirical evidence underlines the func-tional importance of oscillatory rhythms that are not phase-locked toan external stimulus (Bollimunta et al., 2008; Buzsáki and Draguhn,2004; Min and Herrmann, 2007; Thut et al., 2006). It has been shownthat these induced rhythms have a significant impact on the processingof the stimulus and have the potential to explain – at least in parts – thebehavioral trial-by-trial variability (Linkenkaer-Hansen et al., 2004;Romei et al., 2008b; Sauseng et al., 2009). In this study, we show thatexpectancy shapes perception as well as auditory cortical alpha oscilla-tions (8–12 Hz) during the processing of an invariant sound.

Experiments in cognitive neuroscience usually compare neuronalresponses to well-defined experimental conditions while discardingtrial-to-trial fluctuations as noise. However, an increasing amount ofevidence underlines the functional relevance of fluctuating oscillatorybrain activity, for instance in the motor, somatosensory and visualcortices. This relationship has been researched in the pre-stimulus

(T. Hartmann).

rights reserved.

period (e.g., Linkenkaer-Hansen et al., 2004; Romei et al., 2008a;Sauseng et al., 2009) as well as in the post-stimulus period (e.g.,Gross et al., 2007). In the presented study, we were interested inhow distinct expectations to an identical auditory stimulus influencethe alpha desynchronization pattern in the auditory cortex.1

Among the different kinds of oscillations, those with a frequencybetween 8 and 12 Hz have received an increasing amount of atten-tion. These so-called alpha oscillations were the first oscillations tobe noticed in humans (Berger, 1929) and are ubiquitous in all mainsensory and motor brain regions (for a review see e.g., Hari andSalmelin, 1997). They are very prominent in resting EEG and are re-duced (‘desynchronized’) as soon as the brain engages in the proces-sing of information (e.g., external input) (Mimura et al., 1962). Thishas led to the notion that alpha oscillations represent an idling or in-active state of those brain areas in which they are expressed. More re-cent publications, however, have shown that “idling” is a misleading

1 We are aware that alpha-like oscillations in the auditory cortex are conventionallycoined “tau” (Lehtelä et al., 1997). Since, however, we assume functionally similar pro-cesses associated to these rhythms across modalities (Weisz et al., 2007), we prefer re-ferring to them generally as alpha.

171T. Hartmann et al. / NeuroImage 60 (2012) 170–178

term. On the contrary, a modulation of alpha oscillations seems toplay an essential role in information processing. For example,Tuladhar et al. (2007) showed that alpha activity in occipito-parietalareas increases in a visual working memory task with increasingload. Additionally, alpha oscillations appear to increase in visualareas that are involved in processing distracting information (Thutet al., 2006). Anticipation of a target stimulus leads to a desynchroni-zation of alpha oscillations, which has been demonstrated in the visu-al (Thut et al., 2006) and somatosensory (Babiloni et al., 2004) areasinvolved in processing an upcoming stimulus and concomitantly asynchronization of alpha in sensory regions involved in processing adistractor (Rihs et al., 2007; Worden et al., 2000). These findingsclearly speak against the conception that alpha is merely an idlingrhythm that reflects inactivity. Based on the current knowledge,alpha appears to reflect the excitatory–inhibitory balance within sen-sory and motor brain regions that can be modulated in a top-downmanner (Klimesch et al., 2007; Weisz et al., 2007, 2011).

In contrast to the visual and somatosensory domains, studies ex-ploring the functional relevance of alpha oscillations in the auditorydomain are scarce. This may be related to the fact that auditoryalpha activity appears less obvious from resting recordings than, forinstance, the dominant visual alpha rhythm (Weisz et al., 2011). Nev-ertheless, early reports by Lehtelä et al. (1997) and a recent reviewwork by our group (Weisz et al., 2011) clearly demonstrate sound-induced desynchronizations in the alpha range located in the audito-ry cortex, unequivocally showing that such a rhythm indeed exists inthe auditory modality analogous to other modalities. An open ques-tion is, however, whether auditory alpha oscillations also have a sim-ilar functional role as that reported for the visual or somatosensorysystems. To our knowledge, no study exists to date that scrutinizesthe relationship between attention towards an auditory stimulus orits saliency and the respective modulations of auditory cortical oscil-lations. However, some evidence exists that auditory alpha oscilla-tions are involved in auditory memory processes (Pesonen et al.,2006; Van Dijk et al., 2010), in which enhancements were reportedthat bore striking similarities to findings reported in the visual do-main (Tuladhar et al., 2007). These increases during working memorytasks have been interpreted as a disengagement of putatively conflict-ing regions. So-called top-down effects are normally induced usingdiffering stimuli or by differing the instructions or cues given in anexperiment. In the current EEG study, we investigated whether intrin-sically fluctuating top-down processes themselves accompany modula-tions of auditory cortical alpha oscillations. The physical properties ofthe auditory stimulation were kept constant while the participants'expectancy of whether this tone would be followed by an aversivestimulus fluctuated from trial to trial. Modulation of expectancy waspromoted by the fake context of a frequency-discrimination task. Par-ticipants were told that they would hear one of two possible tones,each followed by an auditory feedback (aversive noise or silence)after the forced choice (high- vs. low-pitch sound). While the soundactually remained unchanged, the feedback was pseudorandomizedaccording to a procedure shown to modulate expectancy (Perruchet,1985).

We compared the level of alpha when participants perceived thesound to carry salient information (expectancy of aversive feedback)to instances when the sound was believed to carry non-salient infor-mation (expectancy of silent feedback). As most experiments study-ing conscious perception, the current experiment relies on thesubjective report of the participant, in this case which sound was per-ceived. We are aware of the ongoing discussion in the consciousnessliterature with regards to the validity of self-reports to reflect veridi-cal contents of conscious perception (Lamme, 2006). Even thoughdebriefings following the experiment are suggestive that participantsindeed perceived different pitches associated with differently salientconsequences and data by others (e.g., Carrasco et al., 2004) showthat expectations can induce modulations of perception equivalent

to tuning basic physical properties of a stimulus, from our presentstudy we cannot make statements about the concrete contents ofconscious perception when the participants were confronted by thesound. An alternative explanation could be that the subjects' reportswere based solely on their expectation of hearing two differentsounds which did not influence the concrete perception of the pitch.In short, qualia and their assessment is a highly controversial issuenot within the scope of the present study, so when we speak of “per-ception” then this means “perception reported by the subject” with-out any strong claims with regards to the accompanying qualia.

Based on findings from the somatosensory modality (Babiloni etal., 2004), we hypothesized that the perception of salient soundswould induce stronger alpha desynchronization in auditory corticalregions. While this is our main hypothesis, we also analyzed ourdata with respect to possible pre-stimulus effects. Such an effectcould not apriori be excluded, as the history of past feedback (i.e.,whether salient or not) could have lead to expectations that wouldhave manifested themselves in EEG activity prior to the actualsound onset. We furthermore investigated the period between theonset of perceptual decision and prior to receiving the feedback inorder to control for late effects. Although the main focus of thisarticle lies particularly upon possible auditory cortical alpha effects,we also analyzed evoked time-domain measures (the ERP and theslow cortical potentials). Our results show that altered expectationsof the same stimulus shape auditory cortical alpha activity duringthe presentation of the sound; this underscores that the variabilityof neuronal responses is not necessarily noise but nonetheless carriesfunctionally important information.

Methods

Participants

The participants were 16 right-handed, healthy volunteers be-tween the ages of 19 and 27 (mean age±SD: 22.3±2.35; fivemen). All participants reported normal hearing and had normal orcorrected-to-normal vision. Two participants were excluded fromfurther analysis because of too many artifacts in the EEG data. Twofurther subjects reported hearing the same tone throughout the ex-periment and thus were excluded from further analysis (see later).This left 12 participants for the analysis (four men; mean age±SD:22.25±2.49). The participants were recruited at the University ofKonstanz, gave written informed consent before the experimentbegan and were compensated 10 € for their participation.

Stimuli and procedure

Participants took part in a pseudo-frequency discrimination taskin which they had to judge whether they perceived a high- or low-pitch tone by pressing a button. In reality, the same tone (1000 Hz;40 Hz amplitude modulation) along with a weak background noise(7000–10,000 Hz; 4 dB below the intensity of the tone) was alwayspresented to the participant. Because responses to auditory steady-state stimuli are reported to be dominant in the right hemisphere(Ross et al., 2005), stimuli were delivered to the left ear only. To cre-ate the impression of a real frequency discrimination task, each tonewas followed by a feedback indicating whether the high- or low-pitch tone was presented. Feedback was pseudorandomized so thatequal feedback was given one to four times in a row. The feedbackstimulus was a 3000–6000 Hz bandpass filtered noise of one-secondduration. The intensity of the feedback noise was individually adjust-ed by the subjects, who were instructed to make it as loud as possiblewithout it being painful. The design is an adaptation of Perruchet(Perruchet, 1985) on expectations in classical conditioning. The ratio-nale is that the participant's expectation that the sound and thus thefeedback are different between runs increases alongside the number

172 T. Hartmann et al. / NeuroImage 60 (2012) 170–178

of repeated presentations of the same feedback. Two of the 16 partic-ipants were aware that there were no differences between the pre-sented sounds and were excluded from further analysis. Followingthe experiment, all other participants reported perceiving differencesin pitch.

Participants were seated in a comfortable chair situated in an elec-trically shielded and sound-attenuated room and were prepared forthe EEG recording. Participants were then instructed that they weretaking part in a psychoacoustical study investigating the influenceof background noise on frequency discrimination abilities. Theywere told that two tones of dissimilar pitch would be presented andthat they had to decide which of them they had heard. As feedback,the higher-pitch sound would then be followed by a loud and un-pleasant high-pitch noise, while the low-pitch sound would be fol-lowed by silence. In fact, the sounds presented in both conditionswere always the same.

To familiarize the participants with the experiment and to maketheir scenarios as realistic as possible, they were first presentedwith two different tones (975 Hz and 1025 Hz; 40 Hz amplitude mod-ulation; no added background-noise) three times each. The investiga-tor ensured that the participants were able to discriminate betweenthese tones in the test trials. Following this, participants underwentten trials of the task with the two different tones without backgroundnoise or corresponding feedback. After the practice trials, subjectswere told that the same tones would be presented in the real taskwith added background noise. This would make the task very diffi-cult, perhaps resulting in difficulty discriminating between the twotones.

Participants then received 140 trials of the pseudo-discriminationtask. The sound was presented for 8 s. Six seconds following soundonset, participants had to decide within 2 s which tone they werehearing by clicking on one of two icons presented on the screen. Im-mediately following the sound presentation, the feedback was givenfor 1 s followed by an inter-trial interval of 10 s (see Fig. 1)

After the experiment, we asked the participants if they had theimpression that they were able to solve the task to confirm theyremained naive about the deceptive setup throughout the experi-ment. Afterwards, the participants were debriefed.

Stimulus presentation, presentation of the instructions and ratingscheme as well as behavioral response acquisition were conductedusing the open source PsyscopeXenvironment (Macwhinney et al.,1997; http://psy.ck.sissa.it/) running on an Apple Macintosh iBook(900-MHz PowerPC G3) using Mac OS X version 10.3.9. The instruc-tions and the rating scheme were presented on a 17-in. CRT-monitor situated approximately 1.5 m from the subject. Auditorystimulation was delivered to the left ear through stereo headphones(Sennheiser HD pro 180).

Fig. 1. One trial of the experiment. An identical AM sound was presented for 8 s in alltrials. Participants had to decide whether they heard a low or high tone. Feedback im-mediately followed the offset of the sound. This was followed by an ITI of 10 s.

Data acquisition

EEG recordings were carried out in a dimly lit, sound-attenuatedroom using a 64-channel EEG system (Neuroscan Synamps). Thesampling rate was 250 Hz. Online filtering with a low-pass of100 Hz and DC filtering was applied. Impedances between the elec-trodes and the skull were kept below 5 kΩ.

Data analysis

Behavioral data were analyzed using R (R Development Core Team,2008), an open-source application for statistical computing. EEG datawere preprocessed and analyzed using fieldtrip (Oostenveld et al.,2011), an open-sourceMatlab Toolbox for analyzing EEG andMEG data.

Behavioral analysisA two-way ANOVA was performed on the data using a linear

mixed-effect model (Pinheiro and Bates, 2002). This approach allowsthe specification of classical fixed effects in addition to introducingrandom effects, taking into account that the data are not taken fromthe whole population but only from a sample. In this case, the fixedeffects were condition (high/low feedback in the previous trial) andstring length (how often the same feedback was given consecutively;one to four times). The random effect was defined as the contributionto the variance by the inter-subject variability.

Spectral analysisAfter re-referencing to average reference and application of a FIR

high-pass filter (cutoff frequency: 0.5 Hz, order: 300), data were cor-rected for eye movements and blinks using the FastICA algorithm(Hyvarinen and Oja, 1997). Epochs of artifact-free data were defined7 s pre- and 9 s post-stimulus onset and grouped according to the re-sponse given by the subject (perception of “high” or “low” pitchsound). Due to the (deliberately) strong imbalance in trial numbersacross the amount of consecutive equal feedback (i.e., same feedbackgiven four times in a row should be perceived as more rare (fourtimes per condition) than a single feedback (18 times per condi-tion).), we refrained from more specifically analyzing the questionof how this may have influenced brain activity. A multi-taper time-frequency transformation using Slepian tapers (Mitra and Pesaran,1999) was applied to each trial of the data at 10 Hz as well as afixed time-window of 500 ms with an overlap of 250 ms and onetaper for each frequency, resulting in a frequency smoothing of~2 Hz. This approach resulted in only one time course, which com-prised the entire frequency band of interest between 8 and 12 Hz.The power values were averaged over all trials of each condition.The same transform was also applied to the average of all trials,which was then subtracted from the mean of the averaged powervalues to remove purely time-locked responses. Since the history ofreceiving a series of contingent feedback could in principle alterpre-stimulus activity, no baseline correction was performed. Thedata were divided into 3 time-bins of interest (see Introduction):stimulus anticipation (500 ms to 100 ms pre-stimulus), stimulus pro-cessing (1000 ms to 5000 ms post-stimulus) and pre-feedback(6000 ms to 7500 ms).

Statistical analysisThe non-baseline corrected spectral data of the conditions were

compared using a cluster-based non-parametric, permutation-basedstatistic (Maris and Oostenveld, 2007) that controls the Type I Errorwith respect to multiple comparisons. The maximum distance be-tween neighboring electrodes for the cluster algorithm was set to45 mm, meaning that each electrode was on average linked to fourneighbors. All electrodes were included in the statistical analysis.

First, ordinary two-sided t-statistics (high vs. low; dependentsamples) were calculated for all sensors and time points. This was

Fig. 2. Behavioral results: response=1 means participants reported the high tone, re-sponse=0 means participants reported the low tone. “Previous Position in String” re-fers to the amount of consecutive equal feedback the participant received beforehand.Participants were more likely to rate a tone as the high one if they had received thelow-tone feedback four times in a row before and vice versa. The linear mixed-effectmodel statistic revealed a strong interaction between condition (high/low) and stringlength (recurrent equal feedback) that reached significance (F=6.74; p=0.011).

173T. Hartmann et al. / NeuroImage 60 (2012) 170–178

done independently for each time bin. The cluster-finding algorithmthen identified clusters of contiguous sensors and time points thathad a pb0.05. The test statistic for the permutation test was thesum of all t-values in a cluster. This statistic was repeated for shuffleddata, in which data were randomly reordered across conditions (nullhypothesis stating that it made no difference whether the participantperceives the high- or low-pitch tone). Upon each permutation, thecluster with the highest sum of t-values was kept. By these means, anull distribution was created from 1000 permutations and the p-values for the empirically derived clusters could be calculated.

Source analysisFor the identification of neuronal sources contributing to the elec-

trode level effects, a beamforming algorithm using an adaptive spatialfilter (Dynamic Imaging of Coherent Sources (DICS)) was applied(Gross et al., 2001). To define the timeframe used for source projec-tion, the t-values of all electrodes not belonging to the significantcluster were first set to zero. For each time-point, the mean t-valueover all electrodes was calculated. The time-point yielding the highestmean t-value in each of the previously defined time-bins with a sig-nificant cluster (stimulus processing and pre-feedback) was thenchosen for further source analysis. The resulting time points were3752 ms for the stimulus processing bin and 6752 for the pre-feedback bin. Each trial was cut into a segment ±250 ms aroundthe respective time point. To keep the analysis comparable to ourelectrode level analysis, the time series were averaged over eachtime bin and response category (high/low) and the evoked responseswere subtracted from each trial. The cross-spectral density matriceswere then separately calculated for the two time bins and the high/low trials using a multitaper FFT approach with four tapers, resultingin a frequency-smoothing of ±2 Hz. The FFT was only calculated forthe frequency of interest (10 Hz). The leadfield matrices were calcu-lated using a standard BEM model (Oostenveld et al., 2001) and cor-rected for the subtracted ICA components using subspaceprojection. The standard brain was divided into a grid with 1-cm res-olution. To achieve numerically more stable results, the cross-spectraldensity matrix was regularized by 10% of the mean of the diagonalvalues. In conjunction with the leadfields, spatial filters were calculat-ed for all grid points. This allowed the spatial distribution of power ineach condition and subject to be estimated. Paired, two-sided t-testswere computed for each grid point between the two conditions ineach time bin and in each subject. These were subsequently interpo-lated onto a standard MRI and then converted into Talairach space forlocalizing significant regions using the Talairach atlas provided by theAFNI software package (Cox, 1996). The t-values were used to maskareas with little and/or unstable differences (p>0.1).

Time-domain analysisAlthough not of primary interest to our research question, we also

investigated possible phase-locked effects in the transient cortical re-sponse: the fast transient ERPs as well as the slow cortical potentials(SCP). For the ERP Analysis, a bandpass filter (2–30 Hz passband, filterorder: 6, twopass) was applied to the re-referenced and artifact-freedata. The data were then averaged between 500 ms pre-stimulusand 1000 ms post-stimulus. The period from 500 ms to 100 ms pre-stimulus was used for baseline correction. The same cluster-basedstatistic was used as for the time-frequency data in order to correctfor the Type I error; a two-sided, dependent sample t-test wasemployed for each permutation. Since the ERPs were not of central in-terest in this study and the statistical analysis did not suggest any sig-nificant effects, there shall be no further descriptions of this in thismanuscript. Interested readers should refer to Fig. 2 in the Supple-mentary materials.

For the SCP analysis, a low-pass filter was applied to the re-referenced and artifact-free data (10 Hz, filter order: 6, two-pass).We averaged the data between 500 ms pre-stimulus and 9000 ms

post stimulus and used the period between 500 ms and 100 ms pre-stimulus for baseline correction. We again used the cluster-based sta-tistic described earlier between stimulus onset and 8000 ms post-stimulus for our statistical analysis. To scrutinize the effects at sourcelevel, we applied an LCMV beamformer (Veen and Drongelen, 1997)to the time-bin of stimulus processing using the same corrected lead-fields as for the frequency-domain source analysis.

Results

Behavioral task

To achieve sufficient variability in the subjects' expectations ofwhich tone they would hear and thus what feedback was to beexpected, we adapted a classical conditioning design by Perruchet(1985) by consecutively presenting the same feedback one to fourtimes. The analysis of the behavioral data of the 14 participants whoreported hearing two different tones showed that participants failedto give a response in 2 – 5 trials (mean: 3 trials), resulting in an averageof 137 trials that could be used for the behavioral analysis. In 48% of thetrials, participants reported having heard the low-pitched sound (stan-dard deviation: 8.7; min: 40%; max: 56%; difference not significant). Tofurther account for biases introduced by individual response prefer-ences of the participants, we also calculated the number of “errors”,i.e. the number of trials in which the participants' responses did notmatch the corresponding feedback. Mismatch between response andfeedback occurred in 48% of the trials (standard deviation: 4%; min:38%; max: 56%; difference not significant) which is in line with the pre-sumption that amatch and amismatch between response and feedbackshould be distributed randomly at least over the group. Box-and-whiskers plots show one outlier for the amount of “errors” (see Supple-mentary Fig. 1). The data showed that the number of consecutivelyequal feedbacks had a strong effect on the expectancy of the partici-pants (see Fig. 2). As expected from Perruchet's (1985) results, the like-lihood of participants to perceive a high tone linearly decreased with

174 T. Hartmann et al. / NeuroImage 60 (2012) 170–178

the amount of consecutive low-tone feedback and vice versa for thehigh-tone feedback. The linear-mixed effect model statistic revealed astrong interaction between condition (high/low) and string length (re-current equal feedback) that reached significance (F=6.74; p=0.011).

Spectral analysis

In addition to our hypothesis of a modulation of alpha powerin contralateral auditory regions during the processing of the stimu-lus, we were also interested in determining whether differences inthe interval before stimulus presentation could be found, since antic-ipation effects (e.g., during visual spatial attention) often manifestthemselves in the pre-stimulus period (Romei et al., 2010). We thusanalyzed the data in three time bins representing the pre-stimulusperiod, the stimulus processing period and the pre-feedback period.

In the period before stimulus presentation, lasting from 500 ms to100 ms pre-stimulus, the algorithm did not find a significant cluster.In order to confirm that the sensor statistic was not insensitive tosmall but stable changes in auditory cortical alpha activity in thepre-stimulus period, we explored differences at source level (−500to 0 ms; data not shown). However, no systematic differences couldbe identified in the auditory cortex, even at liberal masking levels(pb0.1). Based on these results, we conclude that the history of con-tingent feedback does not significantly influence auditory corticalalpha activity at the pre-stimulus stage.

During the time of stimulus processing and pre-feedback, numer-ous electrodes over extended time periods exhibited significant differ-ences between high- and low-pitch trials. To account for Type I

Fig. 3. The top row (a, b) shows the topography for the stimulus processing and pre-feedbfollowed by an aversive feedback. a) In the stimulus-processing time bin, we found a signithe aversive noise over right temporal areas, contralateral to the presented sound betweenyielding the highest sum of t-values in the cluster. Electrodes belonging to the cluster are m(p=0.02) in centro-parietal regions, showing an enhanced synchronization if an aversiveThe plot shows the topography at 7 s, the time point with the highest sum of t-values ishows the time course of the relative difference between aversive noise expected and silenthe plots denotes the time bin of interest. The red part of the line signifies the time in whic

errors due to multiple comparisons, we applied a cluster-basedpermutation-based statistic. In the period of the stimulus processing(1000 ms to 5000 ms post-stimulus), we found a significant negativecluster (p=0.04) at right temporofrontal electrodes (contralateral tothe stimulus), showing on average a relative desynchronization of~20% for trials expected to be followed by an aversive noise between3252 ms and 4000 ms post-stimulus (see Figs. 3a and c).

During the time period of pre-feedback (6000 ms to 7500 mspost-stimulus), the analysis revealed a significant positive cluster(p=0.02) at centro-parietal electrodes, showing a synchronizationof alpha oscillations of ~23% when an aversive noise was expectedto be imminent (see Figs. 3b and d).

Source analysis

For the effect identified during the time of stimulus processing,DICS analysis revealed a comparatively stronger desynchronizationfor sounds expected to be followed by an aversive noise in temporalregions mainly contralateral to stimulus presentation (a smaller ipsi-lateral source can be seen in BA 21). The effect encompasses the rightprimary auditory cortex (BA41) as well as secondary auditory corticalregions (BA 21 and BA 22). (see Fig. 4a). Note that no concomitantdesynchronizations were observed in posterior (visual regions) that“spread out” to auditory regions.

No differences were found in the auditory cortex after decision-making and before the feedback was given. Instead, we found an in-crease in alpha synchronization in occipital regions encompassingsecondary visual areas and the precuneus, lateralized ipsilateral to

ack time bins. Positive values denote higher synchronization for trials expected to beficant cluster (p=0.04) showing a desynchronization in trials rated to be followed by3.25 s and 4 s post-stimulus. The plot shows the topography at 3.75 s, the time pointarked with bold circles. b) In the time of pre-feedback, we found a significant clusterfeedback was expected. The cluster was found between 6.75 s and 7 s post-stimulus.n the cluster. Bold circles mark the electrodes of the cluster. The bottom row (c, d)ce expected trials of the electrodes of the respective clusters. The black bar on top ofh the cluster showed a significant difference between the two conditions.

Fig. 4. Source localization using DICS, projecting the differences between both conditions for the three time bins of interest. Only voxels with a p-valueb0.1 are plotted. Positivevalues denote higher synchronization for trials expected to be followed by an aversive feedback. a) In the time of stimulus processing, the enhanced desynchronization was locatedat right temporal areas including the primary auditory cortex, BA21 and BA 22. A small difference was also found at the side, ipsilateral to stimulus presentation. b) At the time ofdecision-making, we found a higher synchronization at secondary visual areas if the aversive noise was expected. Temporal sources are absent, as expected from the topography.

175T. Hartmann et al. / NeuroImage 60 (2012) 170–178

stimulus presentation (see Fig. 4b). This result is in line with the re-sults at sensor level depicted in Fig. 3b.

Time-domain analysis

In order to gain a more complete impression of our data, we alsoanalyzed whether phase-locked effects in the time domain were pre-sent in the data. As already mentioned in the Methods section, wedid not find any effects in the transient ERPs shortly after stimulus pre-sentation (see Supplementary Fig. 2). There is, however, an effect in theslow cortical potentials (see Fig. 5). In trials in which sounds wereexpected to be followed by an aversive noise, the SCP steadily rises atright fronto-temporal sensors, while the SCP in trials expected to beneutral goes in the other direction. The polarity is reversed in the lefthemisphere thus suggestive of a medial frontal generator being differ-entially activated during the two conditions. The differences on bothsides are identified as significant clusters by the cluster-based statistic(left cluster: p=0.003, right cluster: p=0.001). The time window ofsignificant differences basically extended the entire trial period, how-ever stabilizing after approximately 1.5 s (i.e., approximately 1–2 sprior to the difference in auditory cortical alpha activity) post stimulusonset and increasing in strength during the rest of the trial. Sensor-level analysis was followed up by source analysis (2000–6000 ms),showing a pronounced effect at the anterior cingulum and neighboringregions (see Fig. 5c). The direction of the source level difference sug-gests an enhanced activation of these medial prefrontal structureswhen participants are expecting an aversive feedback to follow thesound presentation.

Discussion

The aim of the current study was to investigate whether auditoryalpha desynchronization elicited by auditory stimulation can be pure-ly modulated by intrinsic top-downmechanisms. The expectation of a(physically not present) high-pitch tone was made salient by associ-ating its occurrence with an aversive consequence (uncomfortablenoise). We contrasted trials in which such an expectation occurredwith trials in which the participants reported the perception of alow-pitch sound, which was associated with a silent feedback. Any ef-fect that we report can only be accounted for by fluctuations in expec-tancy, as all other properties, especially all physical properties of thestimulus, were kept constant.

We first validated our design by analyzing the behavioral resultsand comparing them to the findings of Perruchet (1985). Our results

show a similar trend of changing expectations, as repeated presenta-tion of an identical feedback (e.g., the loud noise) led to increased ex-pectancy of the other feedback. This shows that the history offeedback clearly influences what a person expects to hear and thatthis expectation in fact influences perception. Of the 16 investigatedparticipants, only two realized the deceptive setup of the experiment,while all others reported having actually perceived higher- andlower-pitch sounds.

By contrasting the neuronal responses based on the individuals'overt behavior, our major finding was reduced alpha power in contra-lateral auditory areas while the sound was being processed and per-ceived as the salient (high-pitch) relative to the non-salient (low-pitch) sound. At an electrode level, non-parametric cluster-based per-mutation statistics revealed a significant cluster at right-temporalelectrodes, in which desynchronization was 20% stronger when par-ticipants reported having heard a salient sound. Source localizationssuggest that this difference originated in the primary and secondaryauditory cortex contralateral to the stimulated ear. These structuresmatch the results from the sensor-level analysis. However, we alsofound an enhanced desynchronization for high-pitch trials in ipsilat-eral auditory regions. This overall weaker desynchronization cannotbe seen on the sensor topography perhaps due to an overlap of syn-chronization from neighboring regions that obscured this effect atsensor level. An interesting aspect is the time course of thestimulus-induced effect: the contrast between both “perceptions” be-came significant relatively late (from ~3 s to ~2 s before the responsehad to be given). Fig. 3c illustrates that the differentiation in alphapower between the two conditions started ~1 s following stimulusonset and reached its maximum at ~2.5 s. This extended period of rel-atively increased desynchronization of alpha in the auditory cortexmay be required in order for one to determine which sound was per-ceived in this seemingly challenging task, meaning that the influencesof top-down expectations gradually (rather than quickly) modulateauditory alpha activity.

Contrary to the stimulus-processing period, we did not find anyeffect in auditory areas just before feedback was given. Interestingly,however, we found an increase in alpha power in secondary visualareas, including the precuneus, prior to the anticipated aversive feed-back. We are aware that the validity of this effect is debatable becauseparticipants are presented with the visual prompt to respond at thattime (which however is identical on all trials). So, for instance, thealpha increase could reflect a visual evoked potential, although thisinterpretation would not explain why alpha power would be signifi-cantly higher if an aversive stimulus was anticipated. We are alsoaware that general contaminations from visual and motor processing,

Fig. 5. a) Time course of the slow cortical potential at sensor on the right hemisphere showing the largest difference. Significant sections of the time-course are marked with solidlines, non-significant sections are shown transparent. b) Topography of the difference between high and low rated trials (high–low) between 1200 ms and 7600 ms post-stimulus.c) Source localization using LCMV for the effect found in the sensory domain. As expected from the topography, the difference was located at the anterior cingulum and neighboringregions.

176 T. Hartmann et al. / NeuroImage 60 (2012) 170–178

both surely present at the respective time, could be too high to leavethe data interpretable. There are, however, several findings in the lit-erature that can be related to this result. Although the location of thiseffect does not fit, its similarity to a study by Freunberger et al.(2009), which showed that alpha oscillations exert top-down controlon incoming stimulation, is striking. As in the Freunberger et al. study,in which participants tried to actively inhibit words they were toldnot to remember later, participants in our study might have tried toinhibit the upcoming aversive noise, though in this case we wouldhave expected to find a relative increase in the auditory cortex,which was not the case. Two alternative interpretations of thisexist: Richter et al. (2010) found increased activation of the precu-neus during anticipation of pain-related (thus more salient) com-pared to non pain-related (non-salient) words in an fMRIexperiment. Furthermore, the precuneus plays an important role inthe so-called “default mode network” (Fransson and Marrelec,2008), a prominent model in the area of resting-state fMRI and PET(see, e.g. Raichle and Snyder, 2007). Although evidence bridging thegap between the “default mode network” and electrophysiologicaldata is still sparse, a recent study was able to show that in case ofself-referential processing (Raichle and Snyder, 2007) high correla-tion to EEG alpha oscillations in the precuneus could be found(Knyazev et al., 2011). This might be significant because the pre-feedback period is the time when participants expect the evaluationof the performance. At a more basic level and conforming to the inhi-bition hypothesis described earlier, relative increases in alpha powerare often attributed to the inhibition of task-irrelevant areas in orderto save (the putatively limited) resources for the processing of moreimportant or attended stimuli (e.g., van Dijk et al., 2008; Babiloni etal., 2004). It is therefore possible that the expectation of an upcomingsalient acoustic event (aversive noise) could reduce processing re-sources in non-auditory modalities such as the visual cortex, althoughwe would have expected alpha power increase in primary rather thansecondary visual areas. However, these hypotheses are drawn post-

hoc and should be elucidated in greater detail in future experiments.With overall regards to our research question, our findings imply thatauditory cortical regions are modulated particularly during soundprocessing but neither at the pre-stimulus nor at the pre-feedbackstage. The stronger relative desynchronization when participants per-ceived the more salient sound indicates an increased activation of au-ditory regions, signifying the increased salience of this sound incomparison to the neutral sound.

The present findings are interesting for various reasons. First of all,they contribute to a better understanding of auditory alpha oscilla-tions, for which there remains weak empirical support (Barry et al.,2004; Bastiaansen et al., 2001; Fujioka and Ross, 2008; Pfurtscheller,2003; Schürmann and Basar, 2001; Weisz et al., 2011). Our studynot only confirms that sound-induced desynchronization originatesin auditory cortical areas, but also shows that top-down inducedchanges of the alpha desynchronization are found at low-level corti-cal areas of auditory processing. This supports findings in the visual(van Dijk et al., 2008) and somatosensory (Babiloni et al., 2004) mo-dality, either localizing top-down related changes of alpha-desynchronization at or in close proximity to primary or secondaryareas of the respective modality or interpreting results at sensor-level that originate in these areas. Furthermore, all of these studies at-tribute the functional meaning of alpha modulations to the amplifica-tion of relevant and the inhibition of irrelevant properties of thesensory input. Given that the perceived sound (high or low) inducedthe expectation of a salient or non-salient feedback, our study is inline with results reported in similar research in other domains (e.g.,Summerfield and Egner, 2009; Babiloni et al., 2004). Our studyshows that this principle also applies to the auditory domain, eventhough no direct manipulation concerning the task-relevance of thestimuli was used.

As stated earlier, any of the effects described in the present studycannot be attributed to low-level physical features and must neces-sarily involve top-down processes. However, there are different

177T. Hartmann et al. / NeuroImage 60 (2012) 170–178

alternatives how these may be interpreted. Trivial explanations suchas a response bias can be ruled out since the behavioral data indicatesthat perceiving low- vs. high-pitch sounds (as well as the behavioralcategories of giving a “false” vs. “correct” response) was equiproba-ble. Furthermore, it is worth emphasizing that during the stimulusprocessing period the alpha effects were strictly located to relevantauditory cortical regions and not e.g. motor/premotor regions, speak-ing for a sensory effect. Another possibility would be that the historyof feedback has an influence on the modulation of alpha oscillationsreported in this article. This would be a plausible point since the be-havioral data clearly speak for the history of feedback having some in-fluence and it is likely that this should also manifest itself inneurophysiological processes. At this point however we cannot statethat it is indeed reflected in the level of auditory cortical alpha activitysince the amount of trials decreases strongly with increases of occur-rences of the identical feedback and could not be taken into consider-ation in the present study. It is however worth reemphasizing thatthe experimental manipulation indeed lead to different “qualia” (inthis case perceived pitches) as also confirmed by reports of the partici-pants following the experiment, opening up the possible questionwhether this would be sufficient to generate the observed differences,i.e. a higher pitch sound generally going along with stronger alpha de-creases. Even though this cannot be completely excluded based on thepresent design of the experiment, we think this to be an unlikely expla-nation since the actual pitch differences even in the training was verylow (975 Hz vs. 1025 Hz). On plausibility grounds the observed alphadifferences (>20% at relevant electrodes) appear too great to be causedby such a small pitch difference alone. Based on the present design weassume that themost likely explanation driving the effect is the salience(i.e. aversive vs. non-aversive) consequence associated with the pro-cessed sound, which is likely to be to some extentmodulated by expec-tations based on internal models of the probability which sound shouldoccur. Alpha desynchronization while expecting an aversive (more sa-lient) stimulus has already been previously reported, thus corroborat-ing this interpretation (see, e.g. Babiloni et al., 2004). Even though wewill continue to discuss our effects along this line of reasoning, we ac-knowledge that future studies will need to a) balance the aversive vs.non-aversive feedback between the low- and high-pitch sounds andb) significantly increase the total amount of trials in order to substanti-ate also on a neurophysiological level that the history of the previousfeedback influenced the auditory cortical alpha effect.

We also did not find any significant differences in the event-related potentials between stimulus onset and 500 ms afterwards. Iftop-down modulation would have caused an instant effect, it shouldhave influenced the early time-locked components. We, however, ob-served a change in the slow cortical potentials that built up until thetime of feedback reception. The source at the anterior cingulate cortexas well as the time course of the SCP is different from our results fromthe alpha-band. While we saw a steadily increasing difference from anon-auditory source for the SCP, the modulation of the alpha bandcame into play only at distinct periods and from different corticalareas during stimulus processing and pre-feedback. At this point, itis important to state that the choice of the baseline—needed for con-trolling slow (non-brain-related) drifts of the EEG—could critically in-fluence the time course of the SCP; in principle, it is conceivable thatsuch slow changes take place prior to the onset of the stimulus. Eventhough pre-stimulus effects (here −100 to 0 ms before stimulusonset) were not seen in the present study, it could be stated thatthe choice of the baseline (here −500 to −100 ms before stimulusonset) was not ideal for revealing such an effect. In an additionalanalysis (not shown here; see Supplementary Fig. 3), we shifted thebaseline to −1500 to −1100 and were able to show that theprinciple effect remained unchanged and, importantly, that again nopre-stimulus effect emerged. We are therefore confident in statingthat the current SCP effect requires the processing of a stimulus andreflects a slowly augmenting process in the anterior cingulate.

One possible explanation for our findings is the activation of a “con-text-dependent” network including the anterior cingulate cortex thatinduces a top-down effect on auditory areas that depend on the sa-lience of the upcoming stimulus. In fMRI studies, the anterior cingulatehas been shown to belong to a network that includes the temporopar-ietal junction, the precuneus, the insula and the thalamus and whichtends to respond to salient auditory and visual stimuli with higher ac-tivation (Downar et al., 2001, 2002). Interpreting reduced alpha activ-ity as reflecting an excitatory state, our results also match with astudy by Jaramillo and Zador (2010) that showed an increase infiring-rate and local field potentials for expected items in the auditorycortices of rats. This interpretation is also in line with results showingvisual context-dependent top-down feedback in the visual domain(Bar et al., 2006). An alternative explanation might be that thehigher-order processes associated with anticipation—that is, beforethe actual decision of high or low-pitch sound has been formed—maybe more related to the generator underlying the SCP effect. Medial pre-frontal regions may gradually accumulate sensory evidence over timefor the pre-existing but not necessarily conscious expectation of a sa-lient or unsalient sound (Wyart and Tallon-Baudry, 2009). A gradualincrease in evidence accumulation may then be reflected in the gradualrise of the SCP. This interpretation fits well with the observation thatthe onset of the SCP seems to be triggered by the actual sound onset,regardless of which baseline period was chosen. Within such a frame-work, once a certain threshold is exceeded the anterior cingulate maythen drive auditory cortical excitability in a top-down manner in thecase of a perceived high-pitch (and thus salient) sound. This interpre-tation is in line with recent theories concerning the top-down impactof expectancy on visual processes. For instance, Summerfield andEgner highlight that expectancy leads to top-down-driven modulationof visual perception (Summerfield and Egner, 2009). Since this studywas not designed to actually investigate the role of the anterior cingu-late, both explanations are necessarily in the realm of speculation andwould need to be followed up by further studies.

One possible shortcoming of the reported study is the fact that wedid not counterbalance the relationship between the type of perceivedsound (high- vs. low-pitch) and the type of feedback (aversive vs. non-aversive). Namely, the “difference” between the high and low tone inthe practice runs were only separated by 5% (975 Hz vs. 1025 Hz)and all participants reported that the task was extremely difficult,thus implying a considerable amount of uncertainty. Based on this ar-gument and the fact that the two tones in the real runs were physicallyequal, we consider the feedback's contribution to the reported effectsto be stronger than that of the sound-type. Another shortcoming ofthe study is that interpretations such as the enhanced excitation of au-ditory cortex are indirect and based on plausibility assumptions. With-in this experiment, there are no behavioral data to support that, forexample, an enhanced desynchronization in the auditory cortexwould go along with a superior processing of sounds. Extensions ofthis basic paradigm including, for instance, an attentional task andalso distractors (to provoke synchronizations in task-irrelevant re-gions) will be useful in advancing our understanding of the functionalrelevance of auditory cortical alpha activity.

Final conclusion

Disregarding specific underlying physiological mechanisms, ourwork adds to the increasing evidence that at least certain alpha oscil-lations reflect the excitatory–inhibitory balance in cortical cell assem-blies. We extend this notion to the auditory domain, for which onlylittle (mostly clinical) evidence exists. The most important contribu-tion of this work is in showing that levels of desynchronization duringa single presentation of the same sound are directly related to the ex-pectation of which sound (salient/nonsalient) might follow.

Supplementary materials related to this article can be found on-line at doi:10.1016/j.neuroimage.2011.12.034.

178 T. Hartmann et al. / NeuroImage 60 (2012) 170–178

Acknowledgments

The authors wish to thank Christiane Wolf for her help in acquir-ing the data. This work was supported by the DFG (grant numberWE 4156/2-1).

References

Babiloni, C., Brancucci, A., Arendt-Nielsen, L., Babiloni, F., Capotosto, P., Carducci, F.,Cincotti, F., et al., 2004. Attentional processes and cognitive performance duringexpectancy of painful galvanic stimulations: a high-resolution EEG study. Behav.Brain Res. 152 (1), 137–147 (Elsevier. doi:15135977).

Bar, M., Kassam, K.S., Ghuman, a S., Boshyan, J., Schmid, a M., Schmidt, a M., Dale, a M.,et al., 2006. Top-down facilitation of visual recognition. Proc. Natl. Acad. Sci. U. S. A.103 (2), 449–454. doi:10.1073/pnas.0507062103.

Barry, R.J., Rushby, J.A., Johnstone, S.J., Clarke, A.R., Croft, R.J., Lawrence, C.A., 2004.Event-related potentials in the auditory oddball as a function of EEG alphaphase at stimulus onset. Clin. Neurophysiol. 115 (11), 2593–2601. doi:10.1016/j.clinph.2004.06.004.

Bastiaansen, M.C., Böcker, K., Brunia, C., 2001. Event-related desynchronization duringanticipatory attention for an upcoming …. Clinical … Retrieved from http://linkinghub.elsevier.com/retrieve/pii/S138824570000537X.

Berger, H., 1929. Uber das Elektrenkephalogramm des Menschen. Eur. Arch. PsychiatryClin. Neurosci. 87 (1), 527–570. doi:10.1007/BF01797193.

Bollimunta, A., Chen, Y., Schroeder, C.E., Ding, M., 2008. Neuronal mechanisms of corti-cal alpha oscillations in awake-behaving macaques. J. Neurosci. 28 (40),9976–9988 doi:28/40/9976.

Buzsáki, G., Draguhn, A., 2004. Neuronal oscillations in cortical networks. Science (NewYork, N.Y.) 304 (5679), 1926–1929 doi:15218136.

Carrasco, M., Ling, S., Read, S., 2004. Attention alters appearance. Nat. Neurosci. 7 (3),308–313. doi:10.1038/nn1194.

Cox, R.W., 1996. AFNI: software for analysis and visualization of functional magneticresonance neuroimages. Comput. Biomed. Res. 29 (3), 162–173 doi:8812068.

Downar, J., Crawley, A.P., Mikulis, D.J., Davis, K.D., 2002. A cortical networksensitive to stimulus salience in a neutral behavioral context across multiplesensory modalities. J. Neurophysiol. 87 (1), 615–620. Retrieved from http://scholar.google.com/scholar?hl=en&btnG=Search&q=intitle:A+cortical+network+sensitive+to+timulus+salience+in+a+neutral+behavioral+context+across+multiple+sensory+modalities#0.

Downar, J., Crawley, A., Mikulis, D., 2001. The effect of task relevance on the cortical re-sponse to changes in visual and auditory stimuli: an event-related fMRI study.Neuroimage 14 (6), 1256–1267. doi:10.1006/nimg.2001.0946.

Fransson, P.,Marrelec, G., 2008. Theprecuneus/posterior cingulate cortex plays a pivotal rolein the default mode network: evidence from a partial correlation network analysis.Neuroimage 42 (3), 1178–1184. doi:10.1016/j.neuroimage.2008.05.059 (Elsevier).

Freunberger, R., Fellinger, R., Sauseng, P., Gruber, W., Klimesch, W., 2009. Dissociationbetween phase-locked and nonphase-locked alpha oscillations in a working mem-ory task. Hum. Brain Mapp. 30 (10), 3417–3425. doi:10.1002/hbm.20766.

Fujioka, T., Ross, B., 2008. Auditory processing indexed by stimulus-inducedalpha desynchronization in children. Int. J. Psychophysiol. 68 (2), 130–140.doi:10.1016/j.ijpsycho.2007.12.004.

Gross, J., Salmelin, R., Timmermann, L., Schnitzler, A., Kujala, J., Hämäläinen, M., 2001.Dynamic imaging of coherent sources: studying neural interactions in the humanbrain. Proc. Natl. Acad. Sci. 98 (2), 694 National Acad Sciences. Retrieved fromhttp://www.pnas.org/cgi/content/abstract/98/2/694.

Gross, J., Schnitzler, A., Timmermann, L., Ploner, M., 2007. Gamma oscillations inhuman primary somatosensory cortex reflect pain perception. PLoS Biol. 5 (5),e133. doi:10.1371/journal.pbio.0050133.

Hari, Riitta, Salmelin, R., 1997. Human cortical oscillations: a neuromagnetic view throughthe skull. Trends Neurosci. 20 (1), 44–49. doi:10.1016/S0166-2236(96)10065-5.

Hyvarinen, A., Oja, E., 1997. A fastfixed-point algorithm for independent component analysis.Neural Comput. 9 (7), 1483–1492. doi:10.1162/neco.1997.9.7.1483 (MIT Press).

Jaramillo, S., Zador, A.M., 2010. The auditory cortex mediates the perceptual effects ofacoustic temporal expectation. Nat. Neurosci. 14 (2), 246–251. doi:10.1038/nn.2688 (Nature Publishing Group).

Klimesch,W., Sauseng, P., Hanslmayr, S., 2007. EEG alpha oscillations: the inhibition-timinghypothesis. Brain Res. Rev. 53 (1), 63–88. doi:10.1016/j.brainresrev.2006.06.003.

Knyazev, G., Slobodskoj-Plusnin, J., Bocharov, A., Pylkova, L., 2011. The default modenetwork and EEG alpha oscillations: an independent component analysis. BrainRes. 1402, 67–79. doi:10.1016/j.brainres.2011.05.052 (Elsevier B.V.).

Lamme, V., 2006. Towards a true neural stance on consciousness. Trends Cogn. Sci. 10(11), 494–501. doi:10.1016/j.tics.2006.09.001.

Lehtelä, L., Salmelin, R., Hari, R., 1997. Evidence for reactive magnetic 10-Hz rhythm inthe human auditory cortex. Neurosci. Lett. 222 (2), 111–114. doi:10.1016/S0304-3940(97)13361-4 (Elsevier).

Linkenkaer-Hansen, K., Nikulin, V.V., Palva, S., Ilmoniemi, R.J., Palva, J.M., 2004. Presti-mulus oscillations enhance psychophysical performance in humans. J. Neurosci.24 (45), 10186. doi:10.1523/JNEUROSCI.2584-04.2004 (Soc Neuroscience).

Macwhinney, B., Provost, J., Cohen, J., 1997. The PsyScope experiment-building system.Spatial Vision 11 (1), 99–101. doi:9304758. http://www.ncbi.nlm.nih.gov/pubmed/9304758.

Maris, E., Oostenveld, R., 2007. Nonparametric statistical testing of EEG-and MEG-data. J.Neurosci. Methods 164 (1), 177–190. doi:10.1016/j.jneumeth.2007.03.024 (Elsevier).

Mimura, K., Sato, K., Ozaki, T., Honda, N., Masuya, S., 1962. On the physiological signif-icance of the EEG changes caused by sonic stimulation. Electroencephalogr. Clin.Neurophysiol. 14, 683.

Min, B.-K., Herrmann, C.S., 2007. Prestimulus EEG alpha activity reflects prestimulus top-down processing. Neurosci. Lett. 422 (2), 131–135 doi:S0304-3940(07)00701-X.

Mitra, P., Pesaran, B., 1999. Analysis of dynamic brain imaging data. Biophys. J. 76 (2),691–708 Elsevier. Retrieved from http://linkinghub.elsevier.com/retrieve/pii/S000634959977236X.

Oostenveld, R., Fries, P., Maris, E., Schoffelen, J.-M., 2011. FieldTrip: open source soft-ware for advanced analysis of MEG, EEG, and invasive electrophysiological data.Comput. Intell. Neurosci. 2011, 1–9. doi:10.1155/2011/156869.

Oostenveld, R., Praamstra, P., Stegeman, D., 2001. Overlap of attention and movement-related activity in lateralized event-related …. Clinical … Retrieved from http://linkinghub.elsevier.com/retrieve/pii/S13882457010046062001.

Perruchet, P., 1985. A pitfall for the expectancy theory of human eyelid conditioning.Pavlov. J. Biol. Sci. 20 (4), 163–170. Retrieved from http://www.ncbi.nlm.nih.gov/pubmed/4069791.

Pesonen, M., Björnberg, C.H., Hämäläinen, H., Krause, C.M., 2006. Brain oscillatory1–30 Hz EEG ERD/ERS responses during the different stages of an auditory memorysearch task. Neurosci. Lett. 399 (1–2), 45–50. doi:10.1016/j.neulet.2006.01.053.

Pfurtscheller, G., 2003. Induced oscillations in the alpha band: functional meaning. Epi-lepsia 44 (Suppl 1), 2–8 doi:14641556.

Pinheiro, J.C., Bates, D.M., 2002. Mixed Effects Models in S and S-Plus. 528 pp. Springer.Retrieved from http://www.amazon.com/dp/0387989579.

R Development Core Team, 2008. R: a language and environment for statistical com-puting. Retrieved from http://www.r-project.org.

Raichle, M.E., Snyder, A.Z., 2007. A default mode of brain function: a brief history ofan evolving idea. Neuroimage 37 (4), 1083–1090. doi:10.1016/j.neuroimage.2007.02.041 (Elsevier).

Richter, M., Eck, J., Straube, T., Miltner, W.H.R., Weiss, T., 2010. Do words hurt? Brainactivation during the processing of pain-related words. Pain 148 (2), 198–205.doi:10.1016/j.pain.2009.08.009.

Rihs, T., Michel, C.M., Thut, G., 2007. Mechanisms of selective inhibition in visual spatialattention are indexed by alpha-band EEG synchronization. Eur. J. Neurosci. 25 (2),603–610. doi:10.1111/j.1460-9568.2007.05278.x.

Romei, V., Brodbeck, V., Michel, C., Amedi, A., Pascual-Leone, A., Thut, G., 2008a. Spon-taneous fluctuations in posterior alpha-band EEG activity reflect variability in ex-citability of human visual areas. Cereb. Cortex (New York, N.Y.: 1991) 18 (9),2010–2018. doi:10.1093/cercor/bhm229.

Romei, V., Gross, J., Thut, G., 2010. On the role of prestimulus alpha rhythms overoccipito-parietal areas in visual input regulation: correlation or causation? J. Neu-rosci. 30 (25), 8692–8697. doi:10.1523/JNEUROSCI.0160-10.2010.

Romei, V., Rihs, T., Brodbeck, V., Thut, G., 2008b. Resting electroencephalogram alpha-power over posterior sites indexes baseline visual cortex excitability. Neuroreport19 (2), 203–208. doi:10.1097/WNR.0b013e3282f454c4.

Ross, B., Herdman, A., Pantev, C., 2005. Right hemispheric laterality of human 40 Hz audi-tory steady-state responses. Cereb. Cortex 15 (12), 2029–2039. doi:10.1093/cercor/bhi078.

Sauseng, P., Klimesch, W., Gerloff, C., Hummel, F.C., 2009. Spontaneous locally restrict-ed EEG alpha activity determines cortical excitability in the motor cortex. Neurop-sychologia 47 (1), 284–288. doi:10.1016/j.neuropsychologia.2008.07.021.

Schürmann, M., Basar, E., 2001. Functional aspects of alpha oscillations in the EEG. Int. J.Psychophysiol. 39 (2–3), 151–158. doi:10.1016/S0167-8760(00)00138-0.

Summerfield, C., Egner, T., 2009. Expectation (and attention) in visual cognition.Trends Cogn. Sci. 13 (9), 403–409. doi:10.1016/j.tics.2009.06.003.

Thut, G., Nietzel, A., Brandt, S.A., Pascual-Leone, A., 2006. Alpha-band electroencepha-lographic activity over occipital cortex indexes visuospatial attention bias and pre-dicts visual target detection. J. Neurosci. 26 (37), 9494–9502 doi:26/37/9494.

Tuladhar, A.M., Huurne, N., Schoffelen, J.M., Maris, E., Oostenveld, R., Jensen, O., 2007.Parieto-occipital sources account for the increase in alpha activity with workingmemory load. Hum. Brain Mapp. 28 (8), 785–792. doi:10.1002/hbm.20306(New York: Wiley-Liss, c1993-).

Van Dijk, H., Nieuwenhuis, I.L.C., Jensen, O., 2010. Left temporal alpha band activity in-creases during working memory retention of pitches. Eur. J. Neurosci. 31 (9),1701–1707. doi:10.1111/j.1460-9568.2010.07227.x (Wiley Online Library).

Veen, B.V., Drongelen, W.V., 1997. Localization of brain electrical activity via linearlyconstrained minimum …. IEEE transactions on …Retrieved from http://scholar.google.de/scholar?hl=de&lr=&client=firefox-a&hs=LMH&rls=org.mozilla:de:official&channel=s&q=author:“Van+Veen”+intitle:“Localization+of+brain+electrical+activity+via+linearly+…”+&um=1&ie=UTF-8&oi=scholarr#0.

Weisz, N., Hartmann, T., Müller, N., Obleser, J., 2011. Alpha rhythms in audition:cognitive and clinical perspectives. Front. Psychol. 2. doi:10.3389/fpsyg.2011.00073.

Weisz, N., Muller, S., Schlee, W., Dohrmann, K., Hartmann, T., Elbert, T., 2007. The neu-ral code of auditory phantom perception. J. Neurosci. 27 (6), 1479. doi:10.1523/JNEUROSCI.3711-06.2007 (Soc Neuroscience).

Worden, M.S., Foxe, J.J., Wang, N., Simpson, G.V., 2000. Anticipatory biasing of visuo-spatial attention indexed by retinotopically specific alpha-band electroencepha-lography increases over occipital cortex. J. Neurosci. 20 (6), RC63 Retrieved fromhttp://www.ncbi.nlm.nih.gov/pubmed/10704517.

Wyart, V., Tallon-Baudry, C., 2009. How ongoing fluctuations in human visual cortexpredict perceptual awareness: baseline shift versus decision bias. J. Neurosci. 29(27), 8715–8725. doi:10.1523/JNEUROSCI.0962-09.2009.

van Dijk, H., Schoffelen, J.M., Oostenveld, R., Jensen, O., 2008. Prestimulus oscillatoryactivity in the alpha band predicts visual discrimination ability. J. Neurosci. 28(8), 1816. doi:10.1523/JNEUROSCI.1853-07.2008 (Soc Neuroscience).