G

B

M

SZa

b

c

a

ARAA

KFNI

1

co2ftcanlFbwHlo

of

0h

ARTICLE IN PRESS Model

IOPSY-6710; No. of Pages 11

Biological Psychology xxx (2013) xxx– xxx

Contents lists available at SciVerse ScienceDirect

Biological Psychology

journa l h o me page: www.elsev ier .com/ locate /b iopsycho

odulation of frontal-midline theta by neurofeedback

tefanie Enriquez-Gepperta,b,1, René J. Hustera,c,1, Robert Scharfenorta,acharais N. Mokoma, Jörg Zimmermannb, Christoph S. Herrmanna,c,∗

Department of Experimental Psychology, Carl von Ossietzky University, Oldenburg, GermanyKarl-Jaspers Clinic, Oldenburg, GermanyResearch Center Neurosensory Science, Carl von Ossietzky University, Oldenburg, Germany

r t i c l e i n f o

rticle history:eceived 1 September 2012ccepted 20 February 2013vailable online xxx

eywords:m-thetaeurofeedback

ndividualized training

a b s t r a c t

Cortical oscillations demonstrate a relationship with cognition. Moreover, they also exhibit associationswith task performance and psychiatric mental disorders. This being the case, the modification of oscilla-tions has become one of the key interests of neuroscientific approaches for cognitive enhancement. Forsuch kind of alterations, neurofeedback (NF) of brain activity constitutes a promising tool. Concerningspecific higher cognitive functions, frontal-midline theta (fm-theta) has been suggested as an importantindicator of relevant brain processes. This paper presents a novel approach for an individualized, eight-session NF training to enhance fm-theta. An individual’s dominant fm-theta frequency was determinedbased on experiments tapping executive functions. Effects of the actual NF training were compared to a

pseudo-NF training. Participants of the pseudo-NF training experienced a comparable degree of motiva-tion and commitment as the subjects of the actual NF training, but found the “training” slightly easier. Incomparison to the pseudo-NF training, proper NF training significantly enhanced fm-theta amplitude inthe actual training sessions, as well as during the whole course of training. However, unspecific changesin the alpha and beta frequency ranges found with both the actual NF and the pseudo-NF training groups

of ac

emphasize the relevance

. Introduction

Cognitive functions are related to synchronous neuronal pro-esses as reflected in electroencephalographic (EEG) oscillationsf specific frequencies (Basar and Güntekin, 2008; Engel et al.,001; Herrmann and Knight, 2001; Herrmann et al., 2004). In therequency domain, these oscillations appear as “peaks” in spec-ral analyses (e.g., Klimesch, 1999). Synchronization of oscillationsan occur locally between neurons within the same area, butlso between neural populations of different areas within a wideretwork (e.g., Ward, 2003). Thus, synchronous oscillations out-

ine a possible mechanism for communication within the brain.or instance, Varela et al. (2001) postulate that communicationetween widespread brain areas relies on lower frequency bandshereas higher frequency bands implement local communication.

Please cite this article in press as: Enriquez-Geppert, S., et al., Modulatiohttp://dx.doi.org/10.1016/j.biopsycho.2013.02.019

ence, the size of a functional brain network determines its oscil-atory frequency: the more distributed, the slower the underlyingscillation (Von Stein and Sarnthein, 2000). Especially for complex

∗ Corresponding author at: Department of Experimental Psychology, Universityf Oldenburg, 26111 Oldenburg, Germany. Tel.: +49 441 798 4936;ax: +49 441 798 3865.

E-mail address: christoph.herrmann@uni-oldenburg.de (C.S. Herrmann).1 These authors contributed equally to this work.

301-0511/$ – see front matter © 2013 Elsevier B.V. All rights reserved.ttp://dx.doi.org/10.1016/j.biopsycho.2013.02.019

tive control groups for neurofeedback studies.© 2013 Elsevier B.V. All rights reserved.

cognitive tasks more widespread brain regions are supposed toact as networks. Therefore, it is suggested that cognitive eventsrely heavily on such long-range low frequency mechanisms (e.g.,Canolty and Knight, 2010).

The specific relationship between oscillations and cognition isa highly debated topic (Gohse and Maunsell, 1999; Shadlen andMovshon, 1999; Treisman, 1999); however, causality has alreadybeen demonstrated in an animal study with rats (McNaughtonet al., 2006). McNaughton et al. (2006) demonstrated that pre-cise, restored hippocampal rhythmicity, introduced by electricalstimulation, is crucial for appropriate hippocampal functioningconcerning spatial learning and memory. In 2006, slow poten-tial oscillations were reported to have a causal role on memoryconsolidation during sleep (Marshall et al., 2006). Marshall et al.(2006) instructed their subjects to perform a declarative anda procedural learning task right before sleep. After waking up,only those subjects who received transcranial electrical stimu-lation at 0.75 Hz during REM sleep showed enhanced memoryretrieval. Causality was furthermore demonstrated for the alphafrequency and perception (Neuling et al., 2012). Neuling et al.(2012) applied oscillating transcranial direct current stimulation

n of frontal-midline theta by neurofeedback. Biol. Psychol. (2013),

(tDCS) at 10 Hz and revealed that detection thresholds in anauditory detection task depended on the phase of the entrainedoscillation. These examples suggest that oscillations in general areof functional relevance. Therefore, there is a high potential for the

ING Model

B

2 gical

io

tmr2Wemec((tv

mhAMtbtttatreetCMf

weeIrtacMb1v

wfGi2tnBttFrLtpTs

ARTICLEIOPSY-6710; No. of Pages 11

S. Enriquez-Geppert et al. / Biolo

mprovement of cognitive functions by modulating specific neuralscillations.

Currently, two main approaches are used to modulate the ampli-ude of EEG oscillations: neurostimulation (including transcranial

agnetic stimulation (TMS), tDCS, and transcranial alternating cur-ent stimulation (tACS)) and neurofeedback (NF; e.g., Egner et al.,002; Demos, 2004; Hanslmayr et al., 2005; Zoefel et al., 2010).hile neurostimulation directly stimulates the brain by means of

lectrodes mounted on the scalp, NF requires the active engage-ent of the participant and relies on operant conditioning. Active

ngagement might facilitate long-term retention as suggested byonstructivist learning theories applied in school environmentse.g., Narli, 2011). Moreover, self-efficacy might also be enhancede.g., Carlson-Catalano and Ferreira, 2001; Linden et al., 2012) dueo the experience of success on regulating one’s own brain activityia continuous feedback.

Within the last years, NF has been successfully applied as treat-ent for patients with intractable epilepsy or attention deficit

yperactivity disorder (ADHD; Birbaumer et al., 2009). RegardingDHD, NF has been proven to be highly efficient (Arns et al., 2009;onastra et al., 2003) with induced effects lasting for more than

wo years (Gani et al., 2008). The protocols used for ADHD are oftenased on the modulation of the alpha-to-theta ratio. It is importanto note that effects can result from either a change in the alpha, inhe theta or from a change in both frequency bands. With respecto alpha band NF, Hanslmayr et al. (2005) have demonstrated anssociation between NF-induced amplitude increases and cogni-ive performance enhancements in a mental rotation task. Theseesults received further support by a study conducted by Zoefelt al. (2010). However, as ADHD is a disorder showing disruptedxecutive functions (e.g., Barkley, 1997), and frontal-medial (fm)heta is suggested as a correlate of executive functioning (e.g.,avanagh et al., 2011; Nigbur et al., 2011; Trujillo and Allen, 2007;oor, 2005), one might assume that at least part of the effects found

or NF in ADHD are due to the influence of theta.The identification of a particular oscillation reliably associated

ith cognition is of crucial importance for the development offfective neurocognitive training. For a putative modulation ofxecutive functions, fm-theta might serve as an ideal parameter.n accordance with the nomenclature of Klimesch (1999), fm-thetaepresents a phasic oscillation in terms of a task-related modula-ion of the EEG, in contrast to tonic theta, that is not task-related andssociated with a rather diffuse topography. Most often, enhancedognitive processing is associated with an increase of fm-theta (e.g.,itchell et al., 2008). Furthermore, high fm-theta amplitude has

een linked to improved task performance (e.g., Klimesch et al.,996). In addition, fm-theta shows a high degree of inter-individualariability (e.g., Mitchell et al., 2008).

More specifically, fm-theta oscillations have been associatedith specific event-related brain potentials (ERPs), the so called

m-negativities in cognitive control (e.g., Cavanagh et al., 2011;ruendler et al., 2011). Regarding response inhibition tasks, for

nstance, increased fm-theta amplitude can be observed at around00–600 ms post stimulus presentation, thereby falling well intohe time range of the pronounced N200/P300-complex as seen ino-go and stop as trials (for a review refer to Huster et al., 2012).esides response inhibition, Nigbur et al. (2011) probed variousasks involving interference and showed that fm-theta relates tohe N200 and the response-locked error-related negativity (ERN;alkenstein et al., 1991). The ERN was further demonstrated to cor-espond to fm-theta generated in the midcingulate cortex (MCC;uu and Tucker, 2001). More specifically, recent results support

Please cite this article in press as: Enriquez-Geppert, S., et al., Modulatiohttp://dx.doi.org/10.1016/j.biopsycho.2013.02.019

he view that the ERN and the N200 are generated by partialhase resetting and amplitude enhancement of theta activity (e.g.,rujillo and Allen, 2007; Cohen et al., 2008). Cavanagh et al. (2011)uggested fm-theta to be the universal source of fm-negativities,

PRESSPsychology xxx (2013) xxx– xxx

indicating a general biophysical processing mechanism for thecoordination of performance-relevant information associated withMCC functioning.

A (further) prerequisite for the modulation of oscillations inorder to affect cognition is to demonstrate its trainability whichrefers to testing if the spectral effects within the trained frequencyband are modulated by NF or neurostimulation. In general, onlyoscillations at biologically relevant frequencies can be modified(Hutcheon and Yarom, 2000). Beyond that, studies have demon-strated that individual peaks within specific frequency bands varyacross subjects as a result of age, neurological diseases, task perfor-mance or brain volume (e.g., Klimesch, 1999; Moretti et al., 2004).Hence, it has been suggested to estimate individually determinedfrequency bands. Based on such an individualized procedure, train-ability should be even more enhanced (e.g., Hanslmayr et al., 2005;Zoefel et al., 2010).

Here, we report a novel approach of an individualized, eight-session, gap-spaced NF training to enhance an individual’s fm-thetaamplitude. NF incorporates several advantages in contrast to neu-rostimulation. An active engagement of the participants, as well asassociated long term effects (e.g., Gani et al., 2008; Monastra et al.,2003), are among the potential advantages. Here, we calculatedthe individual fm-theta peak frequency from four executive func-tions (task-switching, memory updating, response inhibition, andconflict monitoring), known to be the important and independentrepresentatives of executive functions (Miyake et al., 2000; Fisk andSharp, 2004). Each 30-min training session was further subdividedinto six training blocks with brief gaps (similarly as with Van Boxtelet al., 2012; Zoefel et al., 2010). To control for repetition-related, aswell as non-specific effects, we included an active control group.This so-called pseudo-NF group received a pseudo-feedback notrelated to the actual EEG activity, which was matched in its basiccharacteristics to those of the actual training group. Self reportsof all subjects were used to assess the comparability of both sub-ject groups with respect to motivation, commitment and perceivedtraining difficulty. The aim of the study is to vigorously assess thetrainability of fm-theta, focusing on its enhancement. We expectedthat fm-theta amplitude should be enhanced via the actual NF train-ing as compared to the pseudo-NF training.

2. Materials and methods

2.1. Participants

Thirty-one healthy participants (15 men, mean age: 25 years; standard devia-tion: 3 years) took part in the NF experiment. All were right-handed, as indicated bythe Edinburgh Handedness Inventory (Oldfield, 1971) and had normal or corrected-to-normal vision. Prior to the measurements, all participants were informed aboutthe schedule and goals of the study and gave written informed consent to the pro-tocol approved by the ethic committee of the University of Oldenburg. The studywas conducted in accordance with the Declaration of Helsinki. Sixteen participantswere randomly assigned to the experimental NF group (7 men), and the other 15 tothe pseudo-NF group (8 men). Subjects received a monetary reward (8D per hour)for their participation.

2.2. Calculation of the individualized fm-theta

For the individualized fm-theta NF training, the participants’ task-induced EEGwas measured during processing of a cognitive test battery the day before the actualtraining started. The dominant fm-theta frequency of each individual was estimatedfor four executive functions and respective tasks (task-switching, memory updating,response inhibition; Miyake et al., 2000; Fisk and Sharp, 2004). (1) For task-switchinga visual number-letter task was used. Dependent on the background color, partic-ipants were instructed either to classify the number (even vs. odd) or the letter(vowels vs. consonants) via button presses. That is, dependent on the switch cueparticipants had to change between number and letter task processing (switch con-dition) or to continue with the same task set (no switch condition). (2) For memory

n of frontal-midline theta by neurofeedback. Biol. Psychol. (2013),

updating, a visual three-back task was performed. Subjects were instructed to indi-cate per button press whenever a letter was presented three trials before the currentone. Whenever this was the case, the trials were assigned to the updating condition,otherwise to the no-updating condition. (3) For response inhibition, a visual stop-signal task was utilized. On the majority of trials, participants had to react as fast

ARTICLE IN PRESSG Model

BIOPSY-6710; No. of Pages 11

S. Enriquez-Geppert et al. / Biological Psychology xxx (2013) xxx– xxx 3

F t conda

aaccnoHa

eR(atooRcoc

2f

wtfiktr

miat

haSe

ig. 1. Electrophysiological responses of a single subject as seen in the incongruenddition, the scalp topography of fm-theta is exhibited.

nd accurate as possible with a right or left button press to indicate the presentedrrow direction (go condition). However, whenever a stop-signal (a change of arrowolor) followed a go-signal, subjects had to abort the already initiated response (stopondition). (4) For conflict monitoring, subjects performed a Stroop task. Here, colorames were either printed in the named color (congruent condition) or in anotherne (incongruent condition). In the latter case, name and ink color did not match.owever, subjects had to respond to every trial by indicating the print color as fastnd accurate as possible.

Mean RTs and mean accuracy scores were calculated for correct trials of allxperimental conditions and for each group. (1) In the task switching experimentTs and accuracy scores were calculated for the switch and no-switch conditions.2) For the memory updating task accuracy scores are computed for the three-backnd the zero-back task conditions. (3) In the stop-signal task RTs are given for the gorials, and stop signal reaction times (SSRT) for the stop trial. The SSRT is an estimatef the time needed to abort an already initiated response; its calculation is basedn the independent race model by Logan and Cowan (1984). (4) For the Stroop task,Ts as well as accuracy scores are calculated for the incongruent and the congruentonditions. To exclude that differences between groups in NF-success are basedn differences in the performance of the EF experiments, two-sample t-tests werealculated for all conditions.

.3. EEG recordings and preprocessing for the calculation of the individualizedm-theta

All EEG recordings (for the cognitive test battery and the NF training)ere performed in an electrically shielded and sound attenuated room using

he Brain Vision Recorder software in combination with BrainAmp EEG ampli-ers (Brain Products GmbH, Gilching, Germany). Electrode impedances wereept below 5 k�. Data was recorded continuously, sampled at 500 Hz and fil-ered online with a low-pass filter of 250 Hz. The nose was used as an onlineeference.

For the calculation of the individualized fm-theta frequency, EEG activity waseasured at 32 electrodes placed in accordance with the extended version of the

nternational 10–20 system (easycap, Falk Minow Services, Munich, Germany). Inddition, the electrooculogram (EOG) was recorded from one electrode placed belowhe right eye to aid in quantification of ocular artifacts.

Please cite this article in press as: Enriquez-Geppert, S., et al., Modulatiohttp://dx.doi.org/10.1016/j.biopsycho.2013.02.019

Recordings were further processed offline using EEGLAB (freely available fromttp://www.sccn.ucsd.edu/eeglab/). Data was filtered with an 80 Hz low-pass and

0.5 high-pass filter, and down-sampled to 250 Hz. An infomax ICA (Bell andejnowski, 1995; Makeig et al., 2004, 1996) was applied to detect and to correct forye artifacts. Epochs were defined from 1250 pre to 1250 ms post-stimulus onset

ition in the Stroop task. Shown is the event-related spectral perturbation at Fz. In

and baseline-corrected. Trials with incorrect behavioral responses were discardedfrom the analyses. Then, the event-related spectral perturbation (ERSP) was calcu-lated for a given condition. Peaks in the time-frequency matrix of electrode Fz wereidentified in the theta range (4–8 Hz). Instead ERSP calculations, Fourier transfercan be performed as well. However, if a Fourier transform is computed for 1 s ofdata that contains only 200 ms of theta activity, theta will be dampened by a fac-tor of five, making it harder to detect in a 1/f spectrum. EPSP analysis was carriedout for the switch condition in task switching, the updating condition in the three-back task, the stop condition in the stop-signal task and the incongruent conditionin the Stroop task (see Fig. 1). The individual fm theta frequency was then calcu-lated as the mean of the frequency peaks as extracted from these four conditions±1 Hz.

2.4. EEG recordings during neurofeedback

During NF, EEG signals were obtained from five electrodes at locations Fz, FC1,FCz, FC2 and Cz. Recordings were referenced against the nose. In addition, Fp1 andFp2 were used for the detection of ocular artifacts. For NF, EEG data was read out inreal-time and processed by an in-house software programmed in Matlab 7.14 (TheMathWorks, Natrick, USA). During training, fast Fourier-transforms (FFT; using ahamming window) were computed every 200 ms based on 2 s data windows; hence,analysis windows showed an overlap of 1800 ms. This setup was chosen to providethe participants with a rather smooth appearance of the visual feedback while avoid-ing large “jumps” in feedback colors. Data windows that showed contamination ofocular artifacts were discarded from feedback according to the procedure specifiedbelow.

During training blocks the FFT-derived amplitude values were used for the indi-vidualized theta-NF (more details below).

2.5. Neurofeedback training procedure

Participants received either a neurofeedback or pseudo-NF training over thecourse of eight training sessions within two consecutive weeks. Training sessionswere conducted from Tuesday to Friday in the first, and from Monday to Thurs-day in the second week (see Fig. 2). Each training session consisted of six 5-min

n of frontal-midline theta by neurofeedback. Biol. Psychol. (2013),

training blocks, separated by short breaks. During these breaks, participants wereinstructed to write down the strategies they applied within the last NF block. Theusage of 5-min training blocks and the report on strategies were implementedto encourage continuous application of strategies and to prevent concentrationdeclines.

ARTICLE IN PRESSG Model

BIOPSY-6710; No. of Pages 11

4 S. Enriquez-Geppert et al. / Biological Psychology xxx (2013) xxx– xxx

Fig. 2. Overview of the neurofeedback schedule. (A) Schematic of the eight-session neurofeedback training starting on Mondays with task-related EEG and the cognitive testb essionw Schemb ed.

2

rwmttntpwtwpbwpUa

lrsascaThaabrsswppnrbcNp

td

attery for the calculation of the individual fm-theta. On Tuesday, the first training seek. In the second week, sessions 4–8 (Monday to Thursday) are completed. (B)

lock are the resting EEG measurements, in between six training blocks are perform

.6. Neurofeedback training and its implementation

First, an EOG calibration measurement was performed to aid the detection andejection of eye blinks. From the continuous EEG, time frames with a length of 2 sere extracted every 200 ms (overlap: 1800 ms) in real-time and a zero-mean nor-alization was applied to these data windows. Whenever a threshold of 75 �V in

he time-domain was exceeded at fronto-polar channels during the EOG calibration,he relevant time frame was marked as containing an eye artifact. The appropriate-ess of this amplitude threshold was verified by the experimenter who could adapthe threshold to a participant’s EEG whenever needed. Time points corresponding toeak amplitudes were then identified within these marked time frames and 2 s timeindows centered around these peaks were extracted. These 2 s time frames were

hen subjected to an FFT and the frequency exhibiting the highest peak amplitudeas detected in the frequency spectrum. An interval (±1 Hz) was created around thiseak and the mean amplitude for this individualized eye artifact-related frequencyand was calculated. During the following EEG measurements a trial was rejectedhenever the amplitude in the artifact-related frequency band was larger than thereviously computed mean amplitude of the artifact minus 1.5 standard deviations.pon detection of an ocular artifact, feedback of the actual EEG was suppressed and

gray square was presented instead (see further specifications below).After EOG calibration resting EEG was measured for 5 min (start baseline), fol-

owed by the six training blocks (block 1–6). At the end of a training session a secondesting EEG measurement was conducted (end baseline). For the training sessionsubjects were instructed to apply strategies to increase their individual fm-thetamplitude during training sessions relative to the amplitude during resting EEG. Totart, subjects were given a list of strategies which earlier studies reported as suc-essful in this context (e.g., mental operations, emotions, imagination, memories,nd thoughts of movements). Feedback was given by means of a colored square.he color ranged from a highly saturated red (with steps of 40 colors) over gray to aighly saturated blue (with 40 color steps as well). Depending on the actual fm-thetamplitude, the color was changed to red whenever the amplitude was enhancednd to blue when it was attenuated relative to the baseline measurement. Red andlue values corresponded to amplitudes above and below the actual start baseline,espectively. 95% of the amplitude range was covered by the feedback saturationcale. Values above 97.5% or below 2.5% were indicated by maximal red or blueaturation, respectively. If no difference from baseline was present, or eye blinksere detected, the square was gray. Feedback was updated every 250 ms. Partici-ants were informed to use those strategies that would favor a highly saturated androlonged red-coloration of the square. To control for repetition-related as well ason-specific effects, a pseudo-NF group was included. Whilst the NF group receivedeal-time feedback of their own brain activity, the pseudo-NF group received a play-ack of the feedback of a matched participant of the NF group recorded during theorresponding training session and block. Nevertheless, the feedback of the pseudo-

Please cite this article in press as: Enriquez-Geppert, S., et al., Modulatiohttp://dx.doi.org/10.1016/j.biopsycho.2013.02.019

F group was halted by the presentation of a gray square, when eye blinks duringseudo-NF were detected, increasing the credibility of the manipulation.

During the baseline measurements the square also changed its color from grayo red and blue. Participants were instructed to count the red gradients to avoidrowsiness and to ensure comparability with the actual NF regarding the degree of

(Session 1) is performed, followed by sessions 2–4 on consecutive days in the firstatic of a single training session consisting of eight 5-min blocks. The first and last

visual stimulation. During the start baseline of every training session, the amplitudeof the individual fm-theta was calculated as the mean over all artifact-free FFT-windows as reference for the feedback during the training blocks (see Fig. 3). Themean over FFT-windows of the end baseline was used for statistical comparisons.During the six training blocks, feedback was given based on the fm-theta measuredat five previously specified electrodes relative to the fm-theta amplitude of the startbaseline of the corresponding training session.

2.7. Assessment of motivation, commitment and task difficulty

To compare the NFT and the pseudo NFT group concerning the plausibility ofthe intervention, a subject self-report was utilized. Subjects reported on motivationto participate in the study, commitment to the study (before each session), anddifficulty of the session (right after each session) using a seven-point Likert-scale(1 = not at all to 7 = very strong).

2.8. Responders vs. non-responders

As previous studies showed that a subset of subjects does not respond to NFtraining (e.g., Fuchs et al., 2003; Lubar et al., 1995; Hanslmayr et al., 2005) we addi-tionally report descriptive statistics separately for responders and non-responders.Training results were inspected visually by two independent raters concerning theincrease of fm-theta amplitude during training sessions and baseline measures ona five-point Likert-scale (1 = no increase to 5 = strong increase). A participant wasclassified as responder for values of three to five, and as non-responders for valuesof one to two by each rater. In 93.3% of the classified participants, there was a perfectrater agreement. For the descriptive analysis in the manuscript, a participant wasfinally grouped as non-responders when there was perfect agreement, otherwise asresponder.

2.9. Statistical analyses: training effects on frequency amplitudes

To test if there were any differences at individual fm-theta peak frequencies, anindependent samples t-tests were calculated. To assess variations between groupsalready at the first baseline measurement, independent samples t-test was calcu-lated for the fm-theta, beta, or alpha frequency, respectively.

For the analyses of NF success, the relative change in fm-theta amplitude acrossall six NF blocks for each session (1–8) was quantified as change in �V and percentrelative to the corresponding values of the first training session. To investigate thespecificity of training success this calculation was not only performed for fm-theta,but also for alpha and beta activity. Furthermore, resting EEG was calculated asthe mean of the start and end baseline measurements per session (1–8) relative totheta amplitude observed during the baseline measurements of the first session.

n of frontal-midline theta by neurofeedback. Biol. Psychol. (2013),

This calculation was performed for the alpha and beta resting amplitude as well.Training effects were analyzed by a repeated-measures ANOVA with the factorssession (1–8) and group (NF vs. pseudo NF) for training amplitude. To investigate thecourse of fm-theta amplitude increase during training, a regression line was fittedfor each subject. To test if gradients were different between groups (NF vs. pseudo

ARTICLE IN PRESSG Model

BIOPSY-6710; No. of Pages 11

S. Enriquez-Geppert et al. / Biological Psychology xxx (2013) xxx– xxx 5

Fig. 3. Implementation of the neurofeedback protocol. (A) A continuous time-frequency analysis of the EEG is computed for five electrode positions. This moving FFT windowhas a width of 1 s and is updated every 200 ms. Results of this analysis are visually represented as a colored square for the subjects as feedback. The feedback is used by thesubjects to find and use strategies to color the square as red as often as possible, thereby influencing their own brain activity. (B) Rough overview of three possible categoriesof the feedback. During feedback, the fm-theta amplitude is calculated in relation to the amplitude of the start baseline measurement. Whenever there is no difference fromt squarl e squa on th

Nia(c

2

sVabtElv

2

sc

3

3t

pin

he start baseline amplitude or whenever eye blinks are detected, the color of theower than during start baseline the square’s color will be blue (left). On the right, thmplitude in relation to the start baseline. Here, the exact color saturation depends

F) a one-tailed independent-samples t-test was calculated for the slope and thentercept (III). As last step, training effects on resting EEG were analyzed as well,gain utilizing a repeated-measures ANOVA with factors session (1–8) and groupNF vs. pseudo NF) for (IV). In cases of sphericity violations, Greenhouse–Geisserorrections were performed; corrected p-values as well as ε-values are reported.

.10. Statistical analyses: dynamical changes within sessions

A further method to identify changes due to NF is the analysis of changes withinessions compared to the baseline measurements (see for example Dempster andernon, 2009). Thus, training amplitude for each experimental block was extractednd averaged across all sessions (start baseline, block 1, block 2, block 3, block 4,lock 4, block 6, end baseline) for fm-theta, alpha, and beta frequencies relative tohe amplitude observed during the first start baseline as change in �V and percent.ffects were analyzed by a repeated-measures ANOVA with factors block (start base-ine, block 1, block 2, block 3, block 4, block 4, block 6, end baseline) and group (NFs. pseudo NF).

.11. Statistical analyses: motivation, commitment and task difficulty

Training effects were analyzed by a repeated-measures ANOVA with factorsession (1–8) and group (NF vs. pseudo-NF) for the dependent variables motivation,ommitment and task difficulty as reported by the participants.

. Results

.1. Assessment of behavioral performance measures prior toraining

Please cite this article in press as: Enriquez-Geppert, S., et al., Modulatiohttp://dx.doi.org/10.1016/j.biopsycho.2013.02.019

RTs (mean, ±1 standard deviation), and accuracy scores inercent (mean, ±1 standard deviation) are given for all exper-

mental conditions in Table 1. Two-sample t-tests resulted inon-significant tests for each measurement. Thus, there were no

e is gray as shown in the middle. However, when the current fm-theta activity isare’s color is red, which represents the intended goal, the enhancement of fm-thetae magnitude of the difference from the start baseline.

performance differences between the NF and the pseudo-NF groupin any task.

3.2. Training effects on amplitude (NF vs. pseudo)

I. The mean (and standard deviation) of the individual fm-thetapeaks calculated before training were 5.67 Hz (0.51 Hz) for theNF group and 5.133 Hz (0.83 Hz) for the pseudo NF group. Sta-tistically, they did not differ. At the beginning of the trainingin the first start baseline measurement, the mean and standarddeviation of the absolute theta amplitude was 2.08 �V (0.45 �V)for the NF group and 2.63 �V (0.59 �V) for the pseudo NFgroup. Values for the alpha amplitude were 1.78 �V (0.55 �V)for the NF group and 1.93 �V (0.65 �V) for the pseudo NF group;whereas the corresponding beta values were 1.00 �V (0.23 �V)for the NF group and 0.99 �V (0.22 �V) for the pseudo NF group.Statistical assessment suggests that the NF and the pseudogroup did not differ concerning amplitude in any frequencybefore NF training.

II. The increase of training-associated fm-theta amplitude duringthe course of training is shown in Fig. 4, that for alpha and betain Fig. 5. The frequency spectra are depicted in Fig. 6 for the first(S1) and the last training session (S8). The proper NF training,compared to the pseudo-NF training, led to a stronger gain onlywith fm-theta as indicated by a significant main effect of group

n of frontal-midline theta by neurofeedback. Biol. Psychol. (2013),

(changes in �V: F(1, 29) = 6.396, p < 0.001; changes in percent:F(1, 29) = 6.671, p < 0.05). This effect was completely absent forbeta, and it was found for alpha only when testing relativedifferences in percent (main effect group: F(1, 29) = 7.272,

ARTICLE IN PRESSG Model

BIOPSY-6710; No. of Pages 11

6 S. Enriquez-Geppert et al. / Biological Psychology xxx (2013) xxx– xxx

Table 1Performance in tasks on executive functioning.

Task Condition Group RT mean (SEM) Accuracy in % (SEM)

(1) Task switching Switch NF 1183 (54) 96.9 (1)Pseudo-NF 1171 (54) 91.3 (7)

No switch NF 1421 (34) 91.2 (3)Pseudo-NF 1321 (67) 84.5 (4)

(2) Three-back task 3-back NF n.a. 81.3 (2)Pseudo-NF n.a. 77.6 (3)

0-back NF n.a. 99.7 (0)Pseudo-NF n.a. 99.7 (0)

(3) Stop-signal task Stop NF 269 (8) n.a.Pseudo-NF 286 (7) n.a.

Go NFPseudo-NF

(4) Stroop task Incongruent NF 608 (29) 82.1 (8)PseNFPse

Ffiaitbsg

Congruent

p < 0.05). This group difference is depicted in Fig. 4 for the thetarange.

In addition, an increase of amplitude in all frequency bandscan be observed as well in both type of measurement (�Vand percent) which is confirmed by a significant main effectof session. Both the actual and the pseudo-NF group showan enhancement of amplitude in all three frequencies (fm-theta changes in �V: F(7, 203) = 6.224, p < 0.001 and changes

Please cite this article in press as: Enriquez-Geppert, S., et al., Modulatiohttp://dx.doi.org/10.1016/j.biopsycho.2013.02.019

in percent F(4.227, 122.575) = 6.144, ε = 0.604, p < 0.001; alphachanges in �V: (F(4.669, 135.41) = 9.018, ε = 0.667, p < 0.001 andchanges in percent F(3.222, 93.425) = 7.561, ε = 0.460, p < 0.001;beta changes in �V: F(2.331, 67.598) = 3.247, ε = 0.333, p < 0.01

ig. 4. Neurofeedback training gain: fm-theta enhancement during training. Thegure depicts fm-theta amplitude changes for each training session (S1–S8) as aver-ged over all corresponding blocks (block 1–6) relative to the first training sessionn �V. The training course of fm-theta amplitude is shown for the NF (black) andhe pseudo NF training group (gray). Additionally, fm-theta amplitude changes inaseline measurements are shown for the course of the training relative to the firsttart baseline measurement for the NF (black dashed line) as well as the pseudo NFroup (gray dashed line).

udo-NF 591 (28) 82.7 (8) 542 (24) 96 (1)udo-NF 523 (18) 96.5 (1)

and changes in percent: F(4.488, 130.151) = 2.652, ε = 0.244,p < 0.05).

III. Over the training sessions, fm-theta and alpha amplitudes alsochanged with resting EEG measurements as indicated by a maineffect of session (fm-theta changes in �V: F(7, 203) = 2.881,p < 0.01; changes in percent: F(7, 203) = 2.788, p < 0.01; alphachanges in �V: F(4.958, 143,789) = 4.285, p < 0.001; changes inpercent F(7, 203) = 4.474, p < 0.001), but not beta. No group dif-ferences were found.

IV. When testing for differences with respect to the strengths oflinear increases, regression slope coefficients show that theactual NF participants exhibit a stronger enhancement of fm-theta amplitude than the pseudo NF group (changes in �V:t(29) = 2.07; p < 0.05; changes in percent: t(29) = 1.74; p < 0.05)

3.3. Dynamical changes within sessions

Fig. 7 depicts the dynamical changes within sessions across alltraining days for fm-theta, alpha and beta. Fm-theta is stronglyenhanced during training blocks with actual NF participants ascompared to the pseudo NF group. This is confirmed by significantmain effects of group (changes in �V: F(1, 29) = 25.821, p < 0.001;changes in percent: F(1, 29) = 28.492, p < 0.001) and block (changesin �V: F(3.867, 112,133) = 10.208, p < 0.001; changes in percent:F(3.54, 102.648) = 6.357, p < 0.01), as well as the interaction of groupand block (changes in �V: F(3.867, 112.133) = 10.503, p < 0.001;changes in percent F(3.54, 02.648) = 10.233, p < 0.01).

Post hoc tests demonstrated that fm-theta was stronglyenhanced in all training blocks (block 1 to block 6) compared tothe start baseline measurement in the NF group (changes in �V:start baseline vs. block 1: t(15) = −4.787; p < 0.001; start baselinevs. block 2: t(15) = −4.837; p < 0.001; start baseline vs. block 3:t(15) = −5.912; p < 0.001; start baseline vs. block 4: t(15) = −5.283;p < 0.001; start baseline vs. block 5: t(15) = −5.677; p < 0.001; startbaseline vs. block 6: t(15) = −5.183; p < 0.001). No differencesbetween training blocks and the start baseline measurement wereapparent in the pseudo NF group except for the last block and theend baseline measurement (start baseline vs. block 6: t(14) = 2.701;p < 0.05; start baseline vs. end baseline t(14) = −3.664; p < 0.05).

Alpha amplitude also changed within training sessions, exhibit-ing a linear increase across blocks (main effect of block, changesin �V: F(2.541, 29) = 7.331, p < 0.001; changes in percent: F(1.815,

n of frontal-midline theta by neurofeedback. Biol. Psychol. (2013),

52.645) = 5.364, p < 0.001; interaction of group × block, changes in�V: F(2.541, 73.683) = 3.354, p < 0.01; changes in percent F(1.815,52.645) = 3.73, p < 0.01). Post hoc tests revealed the onset of thisincrease with the NF group to be at block 4 (start baseline vs. block

ARTICLE IN PRESSG Model

BIOPSY-6710; No. of Pages 11

S. Enriquez-Geppert et al. / Biological Psychology xxx (2013) xxx– xxx 7

F depics e firsn

4pbo

3

jiseNasoTdt

Fgf

ig. 5. Neurofeedback training effects on alpha and beta frequency bands. The figureession (S1–S8) as the mean over all corresponding blocks (block 1–6) relative to theurofeedback (black) and the pseudo neurofeedback training group (gray).

: t(15) = −2.156; p < 0.05; start baseline vs. block 5: t(15) = −2.36; < 0.05; start baseline vs. block 6: t(15) = −2.762; p < 0.05; startaseline vs. end baseline: t(15) = −4.736; p < 0.001). No effects werebserved for the beta frequency.

.4. Assessment of motivation, commitment and task difficulty

Fig. 8 provides the mean and standard deviations of the sub-ects’ self-report for both groups over all training sessions. Visualnspection reveals similarly high motivation to participate in thetudy, similarly high commitment to the study, but slight differ-nces regarding the experienced training difficulty between theF and the pseudo-NF training group. Concerning the statisticalssessment of motivation, a significant main effect was found foression (F(7, 27) = 3.353, p < 0.01). In other words, during the course

Please cite this article in press as: Enriquez-Geppert, S., et al., Modulatiohttp://dx.doi.org/10.1016/j.biopsycho.2013.02.019

f training, motivation did change similarly in both groups (seeable 2). No further effects were found. Commitment was stableuring training session and did not differ between the NF andhe pseudo-NF group (no significant effects were found). However,

ig. 6. Frequency spectra. This figure shows frequency spectra during the first training sroup (left) and the pseudo NF training group (right). The alpha amplitude was enhancem-theta amplitude, however, was enhanced only in the NF training group.

ts changes in the alpha amplitude (left) and beta amplitude (right) for each trainingt training session in �V. The time course of oscillatory amplitudes is shown for the

with regard to difficulty the NF group rated the training as moredifficult than the pseudo-NF group (F(1, 26) = 10.245, p < 0.05).

3.5. Descriptive analysis of training gain in responders andnon-responders

Four of the 16 participants belonging to the NF traininggroup were identified as non-responders that were not ableto modulate their fm-theta amplitude during the whole train-ing time course (see Fig. 9). This seems not to result from alack of motivation or commitment as responders and non-responders show comparable results with respect to the subjectself-reports (mean motivation responders = 5.59, std = 0.95;

n of frontal-midline theta by neurofeedback. Biol. Psychol. (2013),

non-responders = 4.94, std = 1.1; mean commitment respon-ders = 5.06, std = 1.64; non-responders = 5.06, std = 0.89; meandifficulty (responders: mean = 4.48; std = 0.85; non-responders:mean = 4.37; std = 0.8)).

ession (S1, gray) and during the last training session (S8, black) for the NF trainingd after training as compared to before training unspecifically in both groups. The

ARTICLE IN PRESSG Model

BIOPSY-6710; No. of Pages 11

8 S. Enriquez-Geppert et al. / Biological Psychology xxx (2013) xxx– xxx

Fig. 7. Dynamical changes within sessions. The figure depicts changes in fm-theta (black l(start baseline, block 1, block 2, block 3, block 4, block 5, block 6, end baseline) as the methe NF group (left) and the pseudo NF group (right). For fm-theta a clear enhancement ca

Fig. 8. Subjects’ self-reports. This figure compares results on subjects’ self-reports inthe NFT and pseudo NFT group. Depicted are the means and corresponding standarderrors based on a seven-point scale (1 = not at all, to 7 = very strong) concerning moti-vt

4

a

functions are known to be involved in several aspects of everyday

TS

ation to participate at the study, the commitment to the study and the difficulty ofhe training session.

. Discussion

Please cite this article in press as: Enriquez-Geppert, S., et al., Modulatiohttp://dx.doi.org/10.1016/j.biopsycho.2013.02.019

Fm-theta amplitude was shown to be enhanced by means ofn individualized, eight-session NF training. In general, there is a

able 2ubjects’ self-reports on motivation, commitment and task difficulty.

Session 1 Session 2 Session 3 Se

Motivation (NFT) 5.67 (1.05) 5.5 (1.16) 5.21 (1.12) 5.2Motivation (pseudo) 6 (0.76) 5.8 (0.86) 5.4 (1.06) 5.2Commitment (NFT) 4.93 (1.49) 4.86 (1.46) 4.93 (1.54) 5.0Commitment (pseudo) 5.27 (1.59) 5.07 (1.53) 5.13 (1.51) 5.2Difficulty (NFT) 4.82 (1.35) 4.71 (dr 1.64) 4.64 (1.5) 4.4Difficulty (pseudo) 3.67 (1.5) 3.73 (1.58) 3.67 (1.4) 3.5

ine), alpha (gray line), and beta (dotted gray line) amplitude for each training blockan over sessions (S1–S8) relative to the first start baseline measurement in �V forn be observed in all six training blocks.

considerable linear increase of fm-theta amplitude with stableeffects in-between sessions. Correspondingly, dynamical changeswithin sessions were characterized by a specific increase of fm-theta in the NF group during training blocks not exhibited by thepseudo-NF group. In line with earlier studies, a subsample of theactual neurofeedback group was not able to deliberately enhancetheir fm-theta amplitude. In addition, our data also reveal substan-tial non-specific effects as seen with a pseudo-NF group matchedfor time and effort invested. In the following, we will discuss aspectsconcerning the trainability of fm-theta, the inclusion of an activecontrol group, possible reasons for NF non-responsiveness andaspects concerning the optimization of NF protocols.

4.1. Trainability of fm-theta

The current study shows that the main prerequisite for enhanc-ing cognition by fm-theta NF holds: the oscillations of interestcan indeed be enhanced. Additional changes in the alpha and toa lower degree also for the beta frequency were further observedin both the NF and pseudo-NF feedback groups. These might con-stitute unspecific alpha amplitude effects already reported earlier(Zaehle et al., 2010). We found that baseline theta activity wasalso modulated by training. This finding is in line with previousreports of increasing alpha baseline activity from day to day (e.g.,Cho et al., 2007; Zoefel et al., 2010). The next crucial step is toinvestigate whether fm-theta NF trainings affect cognition, as forinstance executive functions. Specifically these higher cognitive

n of frontal-midline theta by neurofeedback. Biol. Psychol. (2013),

life and are associated with psychiatric disorders (e.g., Goldberg andSeidmann, 1991). In a recent article by Johnson (2012) good execu-tive skills were proposed to mediate compensatory mechanisms in

ssion 4 Session 5 Session 6 Session 7 Session 8

1 (0.98) 5.29 (1.27) 5.33 (1.05) 5.2 (1.37) 5.47 (1.13) (1.27) 5.33 (1.11) 5.4 (1.06) 5.47 (1.06) 5.6 (1.06)7 (1.44) 5 (1.52) 5.2 (1.37) 5.33 (1.5) 5.07 (1.49)

(1.37) 5.27 (1.39) 5.07 (1.39) 5.27 (1.28) 5.33 (1.4)3 (1.5) 4.43 (1.45) 4.43 (1.52) 4.53 (1.55) 3.73 (1.62)7 (1.34) 4.27 (1.79) 3.07 (1.33) 3.3 (1.22) 3 (1.6)

ARTICLE IN PRESSG Model

BIOPSY-6710; No. of Pages 11

S. Enriquez-Geppert et al. / Biological Psychology xxx (2013) xxx– xxx 9

Fig. 9. Responders vs. non-responders. This figure depicts the training gain for responders vs. non-responders of the neurofeedback training group. On the left, fm-thetaa ng bloa the mp

i(b

4

io2tgeElecn

iamitpvates

4

r2nmr

mplitude is shown for each training session (S1–S8) as mean over all correspondind non-responders of the neurofeedback training group in percent. On the right,

lotted concerning motivation, commitment and training difficulty.

ndividuals at genetic risk for psychiatric disorders. Thus, Johnson2012) suggested using executive function training to enhance therain’s system-level resilience to mal-development.

.2. Inclusion of an active control group

NF studies have been criticized for not allowing an unambiguousnterpretation or generalization of the results, particularly becausef the omission of appropriate control groups (Gevensleben et al.,009; Gruzelier and Egner, 2005). Although some studies con-rol for repetition effects by means of including a passive controlroup, these miss to control for unspecific effects. These unspecificffects can be substantial as already was shown for expectancy.xpectancy is known to be a major cause of placebo effects whichead to improvement in clinical drug study outcomes (e.g., Pricet al., 2008). The magnitude of non-specific effects observed in theurrent study underlines the importance for adequately controllingot only for repetition-related but also for such non-specific effects.

A further general point of criticism concerns all types of train-ng or enhancement procedures, including NF, behavioral trainings,nd neurostimulation procedures. Too often motivation and com-itment of the control group compared to the experimental group

s not assessed (Morrison and Chein, 2011). Our data demonstratehat the current pseudo-NF training, with its features of usinglaybacks of matched subjects of the same training level, but pro-iding real-time feedback based on eyeblinks, indeed serves as andequate control procedure. Furthermore, differences in fm-thetaraining success cannot simply be deduced from pre-training differ-nces in executive functioning between groups, as cognitive testinguggests the absence of such differences in our subject samples.

.3. Non-responsiveness to neurofeedback

EEG-neurofeedback studies repeatedly reported non-esponders to NF (e.g., Fuchs et al., 2003; Hanslmayr et al.,

Please cite this article in press as: Enriquez-Geppert, S., et al., Modulatiohttp://dx.doi.org/10.1016/j.biopsycho.2013.02.019

005; Lubar et al., 1995; Zoefel et al., 2010). The proportion ofon-responders identified in the current study (25%) fits prettyuch to observations made earlier studies (e.g., Zoefel et al. (2010)

eport about 27%; Lubar et al. (1995) about one-third). Based

cks (block 1–6) relative to the first training session (S1) separately for respondersean and standard error of the mean over all participants and training sessions are

on the results of the subjects’ self-reports, the non-responders’absence of training success was not caused by a lack of motivationor commitment. Two possible reasons will be further discussedand need to be targeted in future studies.

The first concerns the usage of ineffective strategies or theirinappropriate utilization by non-responders. The effectiveness ofmental strategies has rarely been investigated. As a result, par-ticipants typically receive a collection of possible strategies withexamples and the instruction to adapt and test them before train-ing. An exception represents the study of Nan et al. (in press). Theauthors let their subjects make notes of their mental strategies andscored them according to their valance. As a result, positive think-ing was suggested as a possibly successful strategy for individualalpha NF training.

A second aspect might be of specific importance to fm-thetaNF-responsiveness. In the domain of behavioral trainings, for exam-ple, regional differences in brain structure have been investigatedconcerning the predictability of training effectiveness in complexcognitive tasks and language learning (e.g., Basak et al., 2011;Erickson et al., 2010; Flöel et al., 2009; Loui et al., 2011; Wonget al., 2011). With respect to NF and associated generators of EEG-oscillations, such associations have not yet been investigated. Tobegin, several studies come to the conclusion that fm-theta origi-nates from midcingulate cortex (MCC). For instance, Wang et al.(2005) reported intracranial recordings in different MCC layersduring task performance of epileptic patients. Task-related thetawas found in the superficial cingulate layers that might possi-bly communicate with medial, lateral frontal and temporal brainareas. Similarly, several studies estimate the dominant source offm-theta in healthy subjects to be in the MCC (e.g., Gevins et al.,1997; Ishii et al., 1999; Sauseng et al., 2007). Exactly this struc-ture is known to show a high degree of structural variability andis known to have a specific structure function relation (e.g., Yücelet al., 2001; Huster et al., 2007). The degree of midcingulate fis-surization is known to be associated with differences in behavior,

n of frontal-midline theta by neurofeedback. Biol. Psychol. (2013),

as well with neuropsychological functioning in executive tasks(e.g., Huster et al., 2009, 2011, 2012). Thus, differences of midcin-gulate fissurization might also account for known interindividualdifferences in fm-theta, and might also facilitate the possibility to

ING Model

B

1 gical

mmlimiidmgtioo

4

btnoiNsCtthttttoaHba2eawr

bcc

R

A

BB

B

B

B

C

C

ARTICLEIOPSY-6710; No. of Pages 11

0 S. Enriquez-Geppert et al. / Biolo

odulate this oscillation. However, in an unfavorable case thisight imply deteriorated chances for NF training success, at

east with short interventions. Nevertheless, another possible NFntervention might be to train participants to directly control

idcingulate activity by means of functional magnetic resonancemaging (e.g., Johnston et al., 2011). Ultimately, the ongoing debaten the frame of personalized medicine can be transferred to theomain of cognitive trainings as well. Effort should be put toake predictions possible if an individual would belong to the

roup of responders or non-responders to assign them to a properraining type. With respect to non-responsiveness to NF, thisncludes testing whether these individuals would benefit fromther approaches, such as neurostimulation, behavioral trainings,r their combination due to possible cumulative effects.

.4. Optimizing neurofeedback protocols

It has been shown that additional training gains can be inducedy increased temporal lags within training as well as betweenraining sessions. This is a well established and long-known phe-omenon in experimental psychology (e.g., Ebbinghaus, 1964). Tour knowledge, investigations focusing on the exploration of train-ng gaps of different lengths have not been conducted so far forF. Generally, training gaps can be determined based on two time

cales: gaps of several days and gaps within a training session.oncerning the induced neuronal changes it is well possible thathe nature of neural reorganization differs between training gapshat last only minutes within a single training session on the oneand, and prolonged training breaks in the order of hours or days onhe other hand. For instance, with training, two types of consolida-ion can be distinguished: synaptic vs. system consolidation. Afterhe first hours of training, synaptic plasticity takes place includinghe formation of new connections and the restructuring of existingnes (e.g., Dudai, 2004). Also, sleep contributes to consolidations during sleep a so called “replay” of memory takes place (e.g.,uber et al., 2004). In contrast, system consolidation is known toe much slower including the reorganization of brain regions, prob-bly reflecting memory stabilization (see Frankland and Bontempi,005, for an overview). The exploration of training effects of differ-nt long-lasting training lags (from minutes up to several days) on

systematic level is important concerning at least to two aspects:ith respect to the investigation of neuronal correlates and in

egard to find the optimal repetition interval for NF.In conclusion, we clearly showed that a modulation of fm-theta

y means of an individual, eight-session NF training, adequatelyontrolling for unspecific effects with the aid of a pseudo-NF pro-edure, is very well feasible.

eferences

rns, M., de Ridder, S., Strehl, U., Breteler, M., Coenen, A., 2009. Efficacy of neu-rofeedback treatment in ADHD: the effects on inattention, impulsivity andhyperactivity: a meta-analysis. Clinical EEG and Neuroscience 40 (3), 180–189.

arkley, R.A., 1997. ADHD and the Nature of Self-Control. Guilford Press, New York.asar, E., Güntekin, B., 2008. A review of brain oscillations in cognitive disorders and

the role of neurotransmitters. Brain Research 1235, 172–193.asak, C., Voss, M.W., Erickson, K.I., Boot, W.R., Kramer, A.F., 2011. Regional differ-

ences in brain volume predict the acquisition of skill in a complex real-timestrategy videogame. Brain and Cognition 76 (3), 407–414.

ell, A.J., Sejnowski, T.J., 1995. An information-maximization approach to blind sep-aration and blind deconvolution. Neural Computation 7 (6), 1129–1159.

irbaumer, N., Ramos Murguialday, A., Weber, C., Montoya, P., 2009. Neurofeed-back and brain–computer interface clinical applications. International Reviewof Neurobiology 86, 107–117.

Please cite this article in press as: Enriquez-Geppert, S., et al., Modulatiohttp://dx.doi.org/10.1016/j.biopsycho.2013.02.019

anolty, R.T., Knight, R.T., 2010. The functional role of cross-frequency coupling.Trends in Cognitive Sciences 14 (11), 506–513.

arlson-Catalano, J., Ferreira, C., 2001. Linking self-efficacy theory to neurofeedback:a conceptual framework for practice and research. Applied Psychophysiologyand Biofeedback 26 (3), 242.

PRESSPsychology xxx (2013) xxx– xxx

Cavanagh, J.F., Zambrano-Vasquez, L., Allen, J.J.B., 2011. Theta lingua franca: a com-mon mid-frontal substrate for action monitoring processes. Psychophysiology49 (2), 220–238.

Cho, M.K., Jang, H.S., Jeong, S.H., Jang, I.S:, Choi, B.J., Lee, M.G., 2007. Alpha neuro-feedback improves the maintaining ability of alpha activity. Neuroreport 19 (3),315–317.

Cohen, M.X., Ridderinkhof, K.R., Haupt, S., Elger, C.E., Fell, F., 2008. Medial frontalcortex and response conflict: evidence from human intracranial EEG and medialfrontal cortex lesion. Brain Research 1238, 127–142.

Demos, J.N., 2004. Getting Started with Neurofeedback. Norton, New York.Dempster, T., Vernon, D., 2009. Identifying indices of learning for alpha neurofeed-

back training. Applied Psychophysiology and Biofeedback 43 (4), 309–328.Dudai, Y., 2004. The neurobiology of consolidations, or, how stable is the engram?

Annual Reviews of Psychology 55, 51–86.Ebbinghaus, H., 1964. Memory: a Contribution to Experimental Psychology (H.A.

Ruger, C.E. Bussenius, E.R. Hilgard, Trans.). Cover Publications, Inc., New York(Original work published in 1885).

Egner, T., Strawson, E., Gruzelier, J.H., 2002. EEG signature and phenomenology ofalpha/theta neurofeedback training versus mock feedback. Applied Psychophys-iology and Biofeedback 27 (4), 261–270.

Engel, A.K., Fries, P., Singer, W., 2001. Dynamic predictions: oscillations and syn-chrony in top-down processing. Nature Reviews Neuroscience 2 (10), 704–716.

Erickson, K.I., Boot, W.R., Basak, C., Neider, M.B., Prakash, R.S., et al., 2010. Striatalvolume predicts level of video game skill acquisition. Cerebral Cortex 20 (11),2522–2530.

Falkenstein, M., Hohnsbein, J., Hoormann, J., Blanke, 1991. Effects of crossmodaldivided attention on late ERP components. II. Error processing in choice reactiontasks. Electroencephalography and Clinical Neurophysiology 78 (6), 447–455.

Fisk, J.E., Sharp, C.A., 2004. Age-related impairment in executive functioning:updating, inhibition, shifting, and access. Journal of Clinical and ExperimentalNeuropsychology 26 (7), 874–890.

Flöel, A., de Vries, M.H., Scholz, J., Breitenstein, C., Johansen-Berg, H., 2009. Whitematter integrity in the vicinity of Broca’s area predicts grammar learning success.Neuroimage 47 (4), 1974–1981.

Frankland, P.W., Bontempi, B., 2005. The organization of recent and remote memo-ries. Nature Reviews 6 (2), 119–129.

Fuchs, F., Birbaumer, N., Lutzenberger, W., Gruzelier, J.H., Kaiser, J., 2003. Neuro-feedback treatment for attention-deficit/hyperactivity disorder in children: acomparison with methylphenidate. Applied Psychophysiology and Biofeedback28 (1), 1–12.

Gani, C., Birbaumer, N., Strehl, U., 2008. Long term effects after feedbackof slow cortical potentials and of theta-beta-amplitudes in children withattention-deficit/hyperactivity disorder, ADHD. International Journal of Bioelec-tromagnetism 10 (4), 209–232.

Gevins, A., Smith, M.E., McEvoy, L., Yu, D., 1997. High-resolution EEG mapping ofcortical activation related to working memory: effects of task difficulty, type ofprocessing, and practice. Cerebral Cortex 7 (4), 372–385.

Gevensleben, H., Holl, B., Albrecht, B., Vogel, C., Schlamp, D., Kratz, O., et al.,2009. Is neurofeedback an efficacious treatment for ADHD? A randomized con-trolled clinical trial. The Journal of Child Psychology and Psychiatry 50 (7),780–789.

Gohse, G., Maunsell, J., 1999. Specialized representations in visual cortex: a role forbinding? Neuron 24 (1), 79–85.

Goldberg, E., Seidmann, L.K., 1991. Higher cortical functions in normals and inschizophrenia: a selective review. In: Steinhauser, S.R., Gruzelier, Z.J. (Eds.),Handbook of Schizophrenia. Elsevier, Amsterdam, pp. 397–433.

Gruendler, T.O., Ullsperger, M., Huster, R.J., 2011. Event-related potential correlatesof performance monitoring in a lateralized time-estimation task. PLoS One 6(10), e25591.

Gruzelier, J., Egner, T., 2005. Critical validation studies of neurofeedback. Child andAdolescent Psychiatric clinics of North America 14 (1), 83–104.

Hanslmayr, S., Sauseng, P., Doppelmayr, M., Schabus, M., Klimesch, W., 2005.Increasing individual upper alpha power by neurofeedback improves cognitiveperformance in human subjects. Applied Psychophysiology and Biofeedback 30(1), 1–10.

Herrmann, C.S., Knight, R.T., 2001. Mechanism of human attention: event-relatedpotentials and oscillations. Neuroscience and Biobehavioral Reviews 25 (6),465–476.

Herrmann, C.S., Munk, M.H., Engel, A.K., 2004. Cognitive functions of gamma-bandactivity: memory match and utilization. Trends in Cognitive Science 8 (8),347–355.

Huber, R., Ghilardi, F., Massimini, M., Tononi, G., 2004. Local sleep and learning.Nature 430, 78–91.

Hutcheon, B., Yarom, Y., 2000. Resonance, oscillation and the intrinsic frequencypreferences of neurons. Trends in Neuroscience 23 (5), 216–222.

Huster, R.J., Westerhausen, R., Kreuder, F., Schweiger, E., Wittling, W., 2007. Mor-phologic asymmetry of the human anterior cingulate cortex. NeuroImage 34(3), 888–895.

Huster, R.J., Wolters, C., Wollbrink, A., Schweiger, E., Wittling, W., Pantev, C., et al.,2009. Effects of anterior cingulate fissurization on cognitive control duringstroop interference. Human Brain Mapping 30 (4), 1279–1289.

n of frontal-midline theta by neurofeedback. Biol. Psychol. (2013),

Huster, R.J., Westerhausen, R., Herrmann, C.S., 2011. Sex differences in cognitive con-trol are associated with midcingulate and callosal morphology. Brain Structureand Function 215 (3–4), 225–235.

Huster, R.J., Enriquez-Geppert, S., Lavallee, C.F., Falkenstein, M., Herrmann, C.S., 2012.Electroencephalography of response inhibition tasks: functional networks and

ING Model

B

gical

I

J

J

K

K

L

L

L

L

L

M

M

M

M

M

M

M

M

M

ARTICLEIOPSY-6710; No. of Pages 11

S. Enriquez-Geppert et al. / Biolo

cognitive contributions. International Journal of Psychophysiology (Epub aheadof print).

shii, R., Shinosaki, K., Ukai, S., Inouye, T., Ishihara, T., Yoshimine, T., et al., 1999.Medial prefrontal cortex generates frontal midline theta rhythm. Neuroreport10 (4), 675–679.

ohnson, M.H., 2012. Executive function and developmental disorders: the flip sideof the coin. Trends in Cognitive Science 16 (9), 454–457.

ohnston, S., Linden, D.E.J., Healy, D., Goebel, R., Habes, I., Boehm, S.E., 2011. Upregu-lation of emotion areas through neurofeedback with a focus on positive mood.Cognitive, Affective and Behavioral Neuroscience 11 (1), 44–51.

limesch, W., Doppelmayr, M., Russegger, H., Pachinger, T., 1996. Theta band powerin the human scalp EEG and the encoding of new information. Neuroreport 7(7), 1235–1240.

limesch, W., 1999. EEG alpha and theta oscillations reflect cognitive and memoryperformance: a review and analysis. Brain Research Reviews 29 (2–3), 169–195.

inden, D.E.J., Habe, I., Johnston, S.J., Linden, S., Tatineni, R., Subramanian, L., Sorger,B., Dealy, D., Goebel, R., 2012. Real-time self-regulation of emotion networks inpatients with depression. PLoS One 7 (6), e38115.

ogan, G.D., Cowan, W.B., 1984. On the ability to inhibit thought and action: a theoryof an act of control. Psychological Reviews 91 (3), 295–327.

oui, P.s., Li, H.C., Schlaug, G., 2011. White matter integrity in right hemispherepredicts pitch-related grammar learning. Neuroimage 55 (2), 500–507.

ubar, J.F., Swartwood, M.O., Swartwood, J.N., O’Donnel, P.H., 1995. Evaluation ofthe effectiveness of EEG neurofeedback training for ADHD in a clinical set-ting as measured by changes in T.O.B.A. scores, behavioral rating, and WISC-Rperformance. Biofeedback and Self-Regulation 20 (1), 83–89.

uu, P., Tucker, D.M., 2001. Regulating action: alternating activation of humanprefrontal and motor cortical networks. Clinical Neurophysiology 112 (7),1295–1306.

arshall, L., Helgadóttir, H., Mölle, M., Born, J., 2006. Boosting slow oscillationsduring sleep potentiates memory. Nature 44 (7119), 610–613.

akeig, S., Delorme, A., Westerfield, M., Jung, T.P., Townsend, J., Courchesne, E., 2004.Electroencephalographic brain dynamics following manually responded visualtargets. Plos Biology 2 (6), 747–762.

akeig, S., Bell, A.J., Jung, T.P., Sejnowski, T.J., 1996. Independent component analysisof electroencephalographic data. In: Touretzkey, D., Mozer, M., Hasselmo, M.(Eds.), Advances in Neural Information Processing Systems, vol. 8, pp. 145–151.

cNaughton, N., Ruan, M., Woodnorth, M.-A., 2006. Restoring theta-like rhythmicityin rats restores initial learning in the Morris water maze. Hippocampus 16 (12),1102–1110.

itchell, D.J., McNaughton, N., Flanagan, D., Kirk, I.J., 2008. Frontal-midline thetafrom the perspective of hippocampal theta. Progress in Neurobiology 86 (3),165–185.

iyake, A., Friedman, N.P., Emerson, M.J., Witzki, A.H., Howerter, A.M., Wagner, T.,2000. The unity and diversity of executive functions and their contributions tocomplex frontal lobe tasks: a latent variable analysis. Cognitive Psychology 41(1), 49–100.

onastra, V.J., Monastra, D.M., George, S., 2003. The effects of stimulant therapy,EEG biofeedback, and parenting style on the primary symptoms of attention-deficit/hyperactivity disorder. Applied Psychophysiology and Biofeedback 27(4), 231–249.

Please cite this article in press as: Enriquez-Geppert, S., et al., Modulatiohttp://dx.doi.org/10.1016/j.biopsycho.2013.02.019

oor, N.C., 2005. The neurotherapy of anxiety disorders. Journal of Adult Develop-ment 12 (2–3), 147–154.

oretti, D.V., Babiloni, C., Binetti, G., Cassetta, E., Dal Forno, G., Ferreric, F., et al.,2004. Individual analysis of EEG frequency and band power in mild Alzheimer’sdisease. Clinical Neurophysiology 115 (2), 299–308.

PRESSPsychology xxx (2013) xxx– xxx 11

Morrison, A.B., Chein, J.M., 2011. Does working memory training work? The promiseand challenges of enhancing cognition by training working memory. Psycho-nomic Bulletin and Review 18 (1), 46–60.

W. Nan, J.P. Rodrigues, J. Ma, X. Qu, F. Wan, P.-I. Mak, et al. Individual alphaneurofeedback training effect on short term memory. International Journal ofPsychophysiology (article ahead of print), in press

Narli, S., 2011. Is constructivist learning environment really effective on learningand long-term knowledge retention in mathematics? Example of the infinityconcept. Educational Research and Reviews 6 (1), 36–49.

Neuling, T., Rach, S., Wagner, S., Wolters, C.H., Herrmann, C.S., 2012. Good vibrations:oscillatory phase shapes perception. Neuroimage 63 (2), 771–778.

Nigbur, R., Ivanova, G., Stürmer, B., 2011. Theta power as a marker for cognitiveinterference. Clinical Neuropyhsiology 49 (2), 220–238.

Oldfield, R.C., 1971. The assessment and analysis of handedness: the Edinburghinventory. Neuropsychologia 9 (1), 97–113.

Price, D.D., Finniss, D.G., Benedetti, F., 2008. A comprehensive review of the placeboeffect: recent advances and current thought. Annual Review of Psychology 58,565–590.

Sauseng, P., Hoppe, J., Klimesch, W., Gerloff, C., Hummel, F.C., 2007. Dissociation ofsustained attention from central executive functions: local activity and interre-gional connectivity in the theta range. European Journal of Neuroscience 25 (2),587–593.

Shadlen, M.N., Movshon, J.A., 1999. Synchrony unbound: a critical evaluation of thetemporal binding hypothesis. Neuron 24 (1), 67–77.

Treisman, A., 1999. Solutions to the binding problem: progress through controversyand convergence. Neuron 24 (1), 105–110.

Trujillo, L.T., Allen, J.J.B., 2007. Theta EEG dynamics of the error-related negativity.Clinical Neurophysiology 118 (9), 645–668.

Van Boxtel, G.J.M., Denissen Ad, J.M., Jäger, M., Vernon, D., Dekker, M.K.J., Mihajlovic,V., Sitskoorn, M.M., 2012. A novel self-guided approach to alpha activity training.International journal of Psychophysiology 83 (3), 282–294.

Varela, F., Lachaux, J.-P., Rodriguez, E., Martinerie, J., 2001. The brainweb: phasesynchronization and large-scale integration. Nature Reviews Neuroscience 2 (4),229–239.

Von Stein, A., Sarnthein, J., 2000. Different frequencies for different scales of corti-cal integration: from local gamma to long-range alpha/theta synchronization.International Journal of Psychophysiology 38 (3), 301–313.

Wang, C., Ulber, U., L’Schomer, D.L., Marinkovic, K., Halgren, E., 2005. Responsesof human anterior cingulate cortex microdomains to error detection, conflictmonitoring, stimulus-response mapping, familiarity, and orienting. The Journalof Neuroscience 25 (3), 604–613.

Ward, L.W., 2003. Synchronous neural oscillations and cognitive processes. Trendsin Cognitive Science 7 (12), 553–559.

Wong, F.C.K., Chandrasekaran, B., Garibaldi, K., Wong, P.C.M., 2011. White matteranisotropy in the ventral language pathway predicts sound-to-word learningsuccess. The Journal of Neuroscience 31 (24), 878–8785.

Yücel, N., Stuart, G.W., Maruff, P., Velakoulis, D., Crowe, S.F., Savage, G., et al., 2001.Hemispheric and gender-related differences in the gross morphology of theanterior cingulated/paracingulate cortex in normal volunteers: an MRI morpho-metric study. Cerebral Cortex 11, 17–25.

Zaehle, T., Rach, S., Herrmann, C.S., 2010. Transcranial alternating current stim-

n of frontal-midline theta by neurofeedback. Biol. Psychol. (2013),

ulation enhances individual alpha activity in human EEG. PLoS One 5 (11),e13766.

Zoefel, B., Huster, R.J., Herrmann, C.S., 2010. Neurofeedback training of the upperalpha frequency band in EEG improves cognitive performance. Neuroimage 54(2), 1427–1431.