International Journal of Psychophysiology 72 (2009) 212–216

Contents lists available at ScienceDirect

International Journal of Psychophysiology

j ourna l homepage: www.e lsev ie r.com/ locate / i jpsycho

A spectralanalytic approach to emotional responses evoked throughpicture presentation

Rene J. Huster b,c,⁎, Stephan Stevens a, Alexander L. Gerlach a, Fred Rist a

a Department of Clinical Psychology, University of Münster, Germanyb Institute for Biomagnetism and Biosignalanalysis, University of Münster, Germanyc Department of Psychiatry and Psychotherapy and Interdisciplinary Center for Clinical Research (IZKF), University of Münster, Germany

⁎ Corresponding author. Institute for Biomagnetismdyweg 15, 48149 Münster, Germany. Tel.: +49 251 83 5

E-mail address: rhuster@uni-muenster.de (R.J. Huste

0167-8760/$ – see front matter © 2008 Elsevier B.V. Adoi:10.1016/j.ijpsycho.2008.12.009

a b s t r a c t

a r t i c l e i n f o

Article history:

Frontal EEG asymmetry ha Received 4 July 2008Received in revised form 11 November 2008Accepted 5 December 2008Available online 24 December 2008

Keywords:Alpha asymmetryIAPSReliabilityAffectPictureEEG

s been linked to emotional and motivational reactivity. A frequently appliedmethod to provoke specific asymmetry profiles is the presentation of affective film clips. Although thesefilms might elicit strong emotional reactions, the exact time course and peak of an affective response remainsunclear. In an alternative attempt, stimuli from the International Affective Picture System (IAPS), known toreliably alter emotional states, are utilized. These stimuli are less likely to cause excessive variations inaffective responding. However, relevant studies have most often been unable to find the predicted effects.One reason for such failures might be the inadequate knowledge about the minimum number of stimulineeded for psychometrically stable results. In the present study, an adequate split-half reliability for theexperimental procedure was assured and substantial effects of affective picture category were found. Thispattern of results was robust for both Cz and linked ears as reference. Thus, presenting pictures with anadequate recording length might be a reliable alternative for inducing affective reactions in alpha asymmetryresearch.

© 2008 Elsevier B.V. All rights reserved.

1. Introduction

In recent years much research has been devoted to the under-standing of the biological systems mediating the motivationalcomponents of emotional responding. In this context, two coredimensions of behavioral relevance, approach and avoidance, haveextensively been studied using emotional film clips to alter theparticipants' affective states and motivated behavior. By means ofelectroencephalographic investigations functional asymmetries havebeen observed indicating a greater relative left frontal activityassociated with approach motivation and greater relative right frontalactivity associated with withdrawal motivation (e.g. Davidson, 2004;Coan and Allen, 2004). Most often these studies infer the degree ofcortical activation from the power in the alpha band (8–12 Hz) atseveral electrodes, with lower values in the alpha range indicating ahigher degree of activity. Not surprisingly, such findings of functionalasymmetries have increased the interest in the lateralization ofmotivational states and traits. Studies with both healthy subjects (e.g.Jackson et al., 2003) and patients with emotional disorders (Allenet al., in press) support the notion that left and right frontal regions areinvolved in approach- and withdrawal-related affect, respectively.

and Biosignalanalysis, Malme-6884; fax: +49 251 83 56874.r).

ll rights reserved.

However, past research did not always produce the predicted results(for reviews, see Murphy et al., 2003; Pizzagalli et al., 2003).

One factor that might have added to such inconsistencies is the factthat the distinct characteristics of film clips with the duration of about60 s or more are hardly accessible and will not stay constant duringpresentation. Although these clips might elicit relatively strongemotional reactions, it is unclear when exactly an affective respondingoccurs or whether the characteristics of the stimulus are stable acrosstime. Collecting data throughout the entire film period mighttherefore not necessarily be appropriate to find associations betweenan experimental condition and concurrent cortical asymmetriesreflecting emotional responses.

An alternative approach is the use of pictures with known stimuluscharacteristics. Although pictures do not elicit as strong emotionalresponses as film clips, it is less likely that excessive variations ofaffective responding during their presentation will occur. Further-more, stimuli of the International Affective Picture System (IAPS; Langet al., 2001) are established and reliable elicitors of emotionalexperiences thoroughly tested in numerous studies using startlemethodology (e.g. Schupp et al., 2004). In a recent survey, Gable andHarmon-Jones (2008) found an association of individual differencesin the asymmetry of frontal activations to neutral and dessert pictureswith the desire for a dessert on the one and the time since the lastdessert on the other hand. Despite this successful approach of late,several prior studies, which were presented at international

213R.J. Huster et al. / International Journal of Psychophysiology 72 (2009) 212–216

conferences (e.g. Elgavish et al., 2003), failed to reliably associatefrontal asymmetries with emotional states induced by affectivepictures.

Such divergent findings are likely linked to psychometricalcharacteristics of a given procedure, most importantly the reliabilityof EEG measures which usually increases as a function of the totalrecording length. Several studies showed at least good internalconsistency (Cronbach's alpha of about 0.80 and above) with restingEEG recorded for 4 min or more (Reid et al., 1998; Hagemann et al.,1998). However, although a variety of studies assessed corticalasymmetries during experimental manipulations via film clips toalter emotional or motivational responding, to our knowledge thereliability of these EEG recordings has not yet been evaluated. Priorstudies (e.g. Zinser et al., 1999; Davidson et al., 1990) required as fewas 10 to 15 s of artefact free data as a minimum criterion. Given theabove mentioned findings with respect to the internal consistency ofbaseline EEG, longer intervals seem desirable.

Another relevant factor in the interpretation of EEG data concernsthe choice of an appropriate reference electrode, which holdsespecially true for all kinds of topographical analyses. The effects ofdifferent reference schemes on measurements of frontal traitasymmetry have been in the focus of evaluation before (Hagemannet al., 1998, 2001; Reid et al., 1998). Critically, measurements againstthe most often used mathematically linked ears/mastoids andcommon vertex references did not show strong associations. Thislack of convergent validity was not attributable to poor reliability.Most likely, such difficulties have added to the inconsistenciesregarding the empirical patterns of relationships between corticalasymmetries and motivational and emotional variables. Therefore,results comparing multiple reference montages should be reported(Allen et al., 2004).

Consequently, this study was meant to test the suitability of thepicture presentation paradigm for research on alpha asymmetriesmeasured via EEG. We consider this paradigm to be an alternative tothe presentation of film clips loaded with the above mentionedmethodological obstacles. By comparing positive and negative stimulitaken from the IAPS and, for now, omitting the otherwise advanta-geous neutral category, we hoped to maximize differences in corticalresponding to the affective pictures. It was expected that theasymmetry metric would indicate a relative right hemisphericactivation for negative pictures and vice versa. Models assuming ahemispheric specialization of frontal systems in the mediation ofmotivational and/or emotional tendencies received empirical supportover the last two decades (see special issue of Biological Psychology 1,2004). Therefore, we expected an effect of the picture condition to bemost pronounced at frontal electrode sites. Furthermore, we wereinterested in the differential effects of reference montages on theevoked cortical asymmetries. We compared the most widely usedmathematically linked ears (A1A2) and vertex (Cz) references byincluding them as repeated measures factor in the statistical analysis.Additionally, reliability measures of the asymmetry metric andactivations at single electrode sites were analyzed.

2. Methods

2.1. Design and procedure

Data were collected from 28 right-handed female and malestudents. The protocol was approved by the Ethical Committee ofthe University of Münster. Participants gave their informed consent.Participants were screened for symptoms of neurological or psychia-tric disorders that could have interfered with the purpose of thisstudy. Handedness was validated using the Edingburgh HandednessInventory (Oldfield, 1971).

After parameterization of the physiological data including artifactdetection and exclusion, the final sample consisted of 16 participants

(13 female, 3 male) for statistical analysis. The high exclusion rate ofsubjects resulted from our rather conservative criterion regarding theminimum of required epochs for averaging (compare sectionparametrization).

During the experimental session participants were shown a total of36 photographs from the IAPS (Lang et al., 2001) with positive andnegative emotional and motivational content. We selected pictureswith high values of arousal, half of them rated as highly positive, theother half as highly negative in valence. The two categories havecarefully been matched regarding the arousal ratings and themagnitude of the valence ratings. The final compilation consisted ofthese pictures: 1300, 3000, 3060, 3102, 3150, 3170, 3180, 3530, 6212,6230, 6313, 6560, 6570, 7380, 9140, 9570, 9910, 9921,1710, 4599, 4641,4660, 5480, 5621, 5700, 5190, 7270, 8030, 8080, 8190, 8200, 8210,8370, 8380, 8420 and 8470.

The pictures were presented in a restricted randomized order. Tomaximize the participants' emotional reactivity, three pictures fromthe same affective category were always presented in succession.Apart from this restriction, the assignment of stimuli of one affectivecategory to these short blocks (three pictures each) and the sequenceof such blocks (each single block containing positive or negativepictures only) were random. An interstimulus interval of 2500 msseparated pictures within a block of 3 stimuli having the same valence.This interval consisted of a fixation cross, shown for 1000 ms,preceded and followed by blank screens lasting for 1000 ms and500 ms, respectively. Transitions from one affective block to the nextconsisted of a similar sequence of events, but with an extendedduration of the first blank screen summing up to an averageinterstimulus interval of 6000 ms.

Having viewed the stimulus material while EEG activity wasrecorded concurrently, participants were exposed to the picturesagain and had to rate their emotional responses to the stimuli. Ratingswere obtained with the Self Assessment Manikin (SAM, Bradley andLang, 1994) to measure self-report pleasure and arousal. The SAMdepicts each dimension with graphic characters along a nine pointscale. For pleasure, the endpoints indicate positive and negative affect(low and high scores respectively). Strong arousal is indicated byhigher values on the scale. We deliberately refrained from letting thepictures being rated immediately after their presentation during therecording of EEG activity. On the one hand, higher associationsbetween physiological measures and picture ratings might beobtained with shorter intervals between these tasks. On the otherhand, intermingling ratings and presentations during the recording ofphysiological activation would have interfered with our primaryaffective task.

2.2. Parametrization and statistical analysis

Brain activity was recorded from 28 positions according to the 10-10 System using sintered Ag/AgCl electrodes mounted on a flexiblelycra-electrocap (SynAmps amplifier, Compumedics Neuroscan, USA;easycap, Falk Minow Services, Germany). Cz was used as onlinereference and a ground electrode was placed on the forehead. Twoelectrode-clips at the earlobes served for later rereferencing of thedata to mathematically linked ears. Additional electrodes recorded thebipolar horizontal and vertical electrooculogram (EOG). Impedanceswere below 5 kΩ and matched for homologous sites with a maximumdeviation of 500 Ω. Data were collected at a sampling rate of 500 Hzwith filters set to 0.05 and 70 Hz. Epochs with a length of 1024 datapoints and an overlap of 50% were extracted resulting in a total of 90segments per category. All segments with artifacts (including thosefrom ocular and muscular sources) were rejected. If less than 50epochs remained in any of the categories the participant was excludedfrom further analysis. Epochs were then extracted through a Hanning-window (10%). A Fast-Fourier-Transformation (FFT) was performed toderive estimates of spectral power in the alpha-band (8–13 Hz)

Table 1Alpha power density at single electrode sites.

F3 F4 P3 P4 T7 T8

Positive M 2.931 2.931 3.231 3.220 3.173 3.097SE 0.208 0.197 0.240 0.225 0.222 0.237

Negative M 2.833 2.803 3.042 2.982 3.127 3.002SE 0.202 0.191 0.223 0.211 0.230 0.212

214 R.J. Huster et al. / International Journal of Psychophysiology 72 (2009) 212–216

computed as power density (µV2/Hz). These values were averagedacross the segments of each stimulus category and logarithmized (ln)to approximate a normal data distribution. The EEG was processedusing the software package BrainVision Analyzer (Brain ProductsGmbH, Germany).

In addition to measures of single site activity, indices of corticalasymmetry for homologous pairs of electrodes were computed forreliability analyses by subtracting left-sided transformed powerdensity from right-sided transformed power density. Assuming aninverse relationship between alpha power density and corticalactivation, higher values of this asymmetry metric correspond togreater relative left hemispheric activity. Asymmetry values werecomputed for frontal (F3/4), parietal (P3/4) and temporal (T7/8)electrode positions.

Satisfactory estimates of cortical asymmetry have to be based on asufficient amount of data. Therefore, we computed and compared thesplit-half reliabilities with all available segments (minimum criterionof 50 epochs) and with the medium 20 segments of each subjectssession. The latter number of segments approximates the amount ofdata used in prior studies to evoke motivational-affective states.Reliabilities were computed for all electrode positions, references,hemispheres and emotional conditions separately. Correlationsbetween the test-halves were corrected according to the Spearman–Brown formula and converted via Fischer's z-transformation forfurther analysis. Reliability analyses were conducted for both singleelectrode sites and asymmetry metrics, as the latter are regularly usedin studies on frontal alpha asymmetries.

EEG data and the picture ratings were subjected to repeatedmeasure ANOVAs (GLM) with Greenhouse–Geisser Epsilon correc-tions applied when appropriate. Electrophysiological activity at singleelectrode positions were used for the statistical comparisons ratherthan asymmetry metrics, as such an analysis scheme may help todisentangle a hemispheres' relative contribution to cortical asymme-tries. Tukey HSD tests were used for post-hoc tests.

3. Results

3.1. EEG data analysis

Regarding emotion, amain effect (F(1,15)=12.72, pb0.01) emergedas well as two-way interactions with hemisphere (F(1, 15)=19.06,p≤0.001) and position (F(1.38, 20.74)=13.96, pb0.001). Compared topositive stimuli, the exposure to negative affective stimuli resulted inlower alpha power density. This effect was more pronounced for theright hemisphere than for the left, leading to significantly different

Fig. 1. Means and standard errors (SE) of evoked alpha asymmetries. Depicted aremeans and standard errors of the asymmetries in alpha power density as measured atfrontal (F3/4), parietal (P3/4) and temporal (T7/8) electrodes. The alpha powerdensity of the left is subtracted from the alpha power density of the corresponding rightelectrode. Thus, more positive values reflect a greater relative left hemispheric activity.

levels of activation between the hemispheres for negative emotionalstimuli (Tukey HSD, pb .01). Also, the difference between stimulusconditions was larger at parietal positions than at frontal and centralelectrode sites and reached post hoc significance at these electrode sites(Tukey HSD, pb .001). A depiction of these effects in terms ofasymmetries is given in Fig. 1; means and standard errors for singleelectrodes can be found in Table 1.

Furthermore, an interaction of reference and position wasobserved (F(2, 30)=79.35, pb0.001). The lowest power density ofall positions was measured at temporal electrodes against linked ears.It differed significantly from power at parietal electrodes (Tukey HSD,pb .01). In contrast, when referencing against Cz, alpha power washighest at temporal positions resulting in a significant differencebetween frontal and temporal electrode positions (Tukey HSD,pb .001). See Table 2 for means and standard errors.

3.2. Reliability

Computed parameters (means) of reliability are based on calcula-tions of Fischer's z-transformed values. Statistical comparisons fordependent samples were made following the procedure proposed byOlkin (1967). Excellent reliabilities were obtained for single electrodepositions, both with 20 and with 50 segments. The lowest estimate was0.94. Because no significant differences between factor categories werefound, reliabilities were aggregated. Mean reliability based on 20segments was only marginally lower than when based on 50 or moresegments (mean reliabilities: 0.97 vs. 0.98) and did not differstatistically (z=0.047; p≤0.05). Because no differences between factorcategories were found with correlations of asymmetry scores either,again mean reliabilities were calculated. Split-half reliability for 20segments was significantly lower than for 50 segments (0.73 vs. 0.828;z=−2.53, pb0.01).

3.3. Picture ratings

This analysis was conducted on the final sample of participants forwhom a sufficient number of segments with recorded electrophysio-logical activity was available. First, the middle value of the scale (five)was subtracted from individual rating scores of the pleasure dimensionand the absolute values were computed. Thus, a direct comparison ofratings on the pleasure dimension between picture conditions ispossible. A repeated measure MANOVA (GLM) on the ratings ofpleasure (transformed) and arousal indicated a significant overalleffect of the picture condition (positive/negative; F(2, 10)=36.99;pb .001) with negative stimuli inducing higher arousal (M=6.9,SD=1.3 vs. M=4.5, SD=1.8) as well as more pronounced ratingsregarding pleasure (M=3.05, SD=0.65 vs. M=2.4, SD=0.82).

Table 2The position×reference interaction.

Frontal Parietal Temporal

A1A2 M 3.177 3.411 2.858SE 0.216 0.242 0.283

Cz M 2.573 2.827 3.341SE 0.195 0.215 0.184

215R.J. Huster et al. / International Journal of Psychophysiology 72 (2009) 212–216

4. Discussion

The main purpose of this study was to demonstrate the applic-ability of stimuli from the IAPS in studies of emotion and motivationutilizing quantitative analysis of spectral EEG parameters. The analysisof activations at single electrode sites revealed differences betweenthe emotional conditions. The exposure to negative stimuli resulted ina global reduction of alpha power density at all regions. This effect wasmore pronounced in the right hemisphere and at the parietal region.Therefore, we found the expected modulation of hemisphericactivation, but it was not restricted to frontal sites as we had expected.Although these asymmetries are most often found at medial andlateral frontal regions, such patterns of activation at temporalelectrode positions are also reported. For example, Coan, Allen, andHarmon-Jones (2001) found less left-sided activation at lateral–frontal, midfrontal and fronto-temporal regions with facial poses ofwithdrawal emotions compared to control and approach states.Interestingly, their manipulation did not result in significant changesof relative hemispheric activation in the parietal region.

In general, asymmetric activations at electrodes P3 and P4 arenormally not found in the literature on frontal alpha asymmetries.What might be an explanation for our finding? According to Helleret al. (1997) the right parietal lobe takes part in the mediation ofemotional arousal. Given the differences between the affectivepicture categories in ratings of arousal in our study, the observedparietal asymmetry may reflect the greater arousal associated withnegative stimuli. However, the increase in cortical activity waspronounced at the parietal area in both hemispheres. We thereforeconclude that the higher cortical activation found at this region likelyis caused by a greater allocation of attentional resources to the morearousing negative stimuli. Bilateral activation during functionalimaging has frequently been found in the superior parietal regionin attention demanding tasks (e.g. Fan et al., 2005). Because priorstudies as well as our own often did not include a neutral categorywhile matching arousal for the emotional evocative conditions, theymight have been insensitive to arousal related effects at this specificregion.

Additionally, a differential effect of the emotion provocationprocedure on a specific dependent variable cannot be ruled out. In aninteresting approach, Simons et al. (1999) investigated the effect ofmotion on the emotion-response profile. The authors compared filmclips and pictures (derived via the motion capture technique fromthe former) that were matched for other physical properties. Theyfound higher ratings of valence and arousal in case of moving stimuli.More interestingly, differential effects occurred in physiologicalmeasures of emotional reactivity. Skin conductance and activity ofthe corrugator muscle indicated an interaction of motion andarousal, with more pronounced responses to film clips only in caseof high arousing stimulus material. This was not true for changes ofthe heart rate, where any interaction between stimulus type, valenceand arousal reached statistical significance. Unfortunately, Simons etal. did not include an electroencephalographic measure. Hence, theeffect at parietal electrodes in our study may be due to themethodology utilized to provoke an affective response. A directcomparison of the most often used materials to induce affectivestates, the IAPS and the standardized film clips from the Davidsonlaboratory, has not been made. Such an investigation could furtherclarify the impact of the stimulation procedure and specific stimuluscharacteristics on self-report and physiological indicators of affectiveexperiences.

Furthermore, a puzzling finding was that the participants ratedthe negative pictures as more affective than the positive pictures,despite the fact that we hadmatched the two categories based on theIAPS ratings. One reason may be cultural differences in theevaluation of these pictures, depicting a possible limitation of thepresent study.

One quite classical topic in the analysis of EEG data is the choiceof an appropriate reference electrode. As each EEG channelnecessarily reflects the difference in electrical activity of twosites, a change of the reference electrode might result in quitedistinct data. For example, the not uncommon reference Cz has theserious disadvantage of providing a high degree of activity byitself, thereby potentially distorting effects of interest. The moresimilar the effects are at a target electrode and the chosenreference, the smaller will the outcome at the rereferenced targetprobably be. Rather, relevant effects will be mirrored at electrodesshowing less correlation with the reference site, because theinformation is inherent in the reference itself. Not surprisingly, ithas been shown before, that the convergent validity of differentreference montages for the measurement of alpha asymmetries israther low (Hagemann et al., 2001). With the present results it wasfound that differing reference montages did not dramatically alterthe effects of interest. While power in the alpha band at frontaland parietal electrodes was higher referenced against mathema-tically linked ears compared to vertex, the opposite patternoccurred at the temporal region. The observed effect can beattributed to different interelectrode distances between the targetelectrodes and the respective references. Higher similarity inactivity at target and an nearby reference electrode leads to loweroverall power in the alpha band compared to reference electrodesat higher distances. Importantly, with the present results thereference scheme did not interact with any of the emotion relatedstatistical effects. In contrast to the comparison of Hagemann et al.(2001), we did not include the so called average reference in ourdesign as relevant assumptions were not met. The averagereference reflects the mean of the activity at the measuredelectrodes, which results in an approximation of an inactivereference because the average surface potential of a sphere shouldbe zero. However, this holds true only if a sufficiently high numberof electrodes is used. Although there seems to be no clearconsensus concerning the exact number of electrodes, a minimumof 60 for sure is desirable. The recording sites additionally shouldbe equally spaced and cover the whole head. Otherwise biases mayoccur as the average reference does not adequately approach aninactive reference (see for example Junghöfer et al., 1999). Inaddition, the average reference is not regularly used in the contextof frontal asymmetry research as most often rather few recordingelectrodes are utilized. Still, it seems to be recommendable to moreregularly compare reference schemes.

In summary, we were able to induce distinct emotional statesthrough the well established picture presentation paradigm usingstimuli from the IAPS. The results are in agreement with contem-porary theories of emotion and motivation. We compared the mostcommonly used reference schemes and found emotional alterna-tions to be independent from reference effects. Given a sufficientrecording length (60 s or more of artifact free data) reliableasymmetry indices can be derived. Therefore, this procedure is anacceptable alternative to the use of film clips for research on alphaasymmetries. To guarantee an appropriate reliability for indices ofcortical asymmetry a minimum of 50 segments were required forsignal averaging. Future studies should regularly assess the relia-bility of their physiological dependent variables until the lowerlimits of necessary recording lengths for relevant paradigms areknown. Also, different reference schemes should more often beincluded as a factor in an analysis to test for the independence ofrelevant experimental manipulations.

Acknowledgements

This research was partially funded by the Association for theAdvancement of Clinical Psychology and of Training in Psychotherapy,Münster.

216 R.J. Huster et al. / International Journal of Psychophysiology 72 (2009) 212–216

References

Allen, J.J., Coan, J.A., Nazarian, M., 2004. Issues and assumptions on the road from rawsignals to metrics of frontal EEG asymmetry in emotion. Biological Psychology 67,183–218.

Allen, J.J., McKnight, K.M., Moreno, F.A., Demaree, H.A., Delgado, P.L. (in press).Alteration of frontal EEG asymmetry during tryptophan depletion predicts futuredepression. Journal of Affective Disorders.

Bradley, M.M., Lang, P.J., 1994. Measuring emotion: the self-assessment manikin and thesemantic differential. Journal of Behavior Therapy and Experimental Psychiatry 25,49–59.

Coan, J.A., Allen, J.J., 2004. Frontal EEG asymmetry as a moderator and mediator ofemotion. Biological Psychology 67, 7–49.

Coan, J.A., Allen, J.J., Harmon-Jones, E., 2001. Voluntary facial expression andhemispheric asymmetry over the frontal cortex. Psychophysiology 38, 912–925.

Davidson, R.J., 2004. What does the prefrontal cortex “do” in affect: perspectives onfrontal EEG asymmetry research. Biological Psychology 67, 219–233.

Davidson, R.J., Saron, C.D., Senulis, J.A., Ekman, P., Friesen, W.V., 1990. Approachwithdrawal and cerebral asymmetry— emotional expression and brain physiology.Journal of Personality and Social Psychology 58, 330–341.

Elgavish, E., Halpern, D., Dikman, Z., Allen, J.J.B., 2003. Does frontal EEG asymmetrymoderate or mediate responses to the international affective picture system(IAPS)? Psychphysiology 40 (Suppl. 1), 38.

Fan, J., McCandliss, B.D., Fossella, J., Flombaum, J.I., Posner, M.I., 2005. The activation ofattentional networks. Neuroimage 26, 471–479.

Gable, P., Harmon-Jones, E., 2008. Relative left frontal activation to appetitive stimuli:considering the role of individual differences. Psychophysiology 45, 275–278.

Hagemann, D., Naumann, E., Becker, G., Maier, S., Bartussek, D., 1998. Frontal brainasymmetry and affective style: a conceptual replication. Psychophysiology 35,372–388.

Hagemann, D., Naumann, E., Thayer, J.F., 2001. The quest for the EEG reference revisited:a glance from brain asymmetry research. Psychophysiology 38, 847–857.

Heller, W., Nitschke, J.B., Lindsay, D.L., 1997. Neuropsychological correlates of arousal inself-reported emotion. Cognition and Emotion 11, 383–402.

Jackson, D., Mueller, C., Dolski, I., Dalton, K., Nitschke, J.B., Urry, H., Rosenkranz, M., Ryff,C., Singer, B., Davidson, R.J., 2003. Now you feel it, now you don't: frontal brainelectrical asymmetry and individual differences in emotion regulation. Psycholo-gical Science 14, 612–617.

Junghöfer, M., Elbert, T., Tucker, D.M., Braun, C., 1999. The polar average reference effect: abias in estimating thehead surface integral in EEG recording. Clinical Neurophysiology110, 1149–1155.

Lang, P., Bradley, M.M., Cuthbert, B.N., 2001. International Affective Picture System (IAPS):Instruction Manual and Affective Ratings. The Center for Research in Psychophysiol-ogy, University of Florida. Technical Report A-5.

Murphy, F.C., Nimmo-Smith, I., Lawrence, A.D., 2003. Functional neuroanatomy ofemotions: ameta-analysis. Cognitive, Affective,&BehavioralNeuroscience3,207–233.

Oldfield, R.C., 1971. The assessment and analysis of handedness: the Edinburghinventory 853. Neuropsychologia 9, 97–113.

Olkin, J., 1967. Correlations revisited. In: Stanley, J.C. (Ed.), Improving ExperimentalDesign and Statistical Analysis. Rand McNalley, Chicago.

Pizzagalli, D., Shackman, A.J., Davidson, R.J., 2003. The functional neuroimaging ofhuman emotion: asymmetric contributions of cortical and subcortical circuitry. In:Hugdahl, K., Davidson, R.J. (Eds.), The Asymmetrical Brain. MIT Press, Cambridge,MA, pp. 511–532.

Reid, S.A., Duke, L.M., Allen, J.J., 1998. Resting frontal electroencephalographicasymmetry in depression: inconsistencies suggest the need to identify mediatingfactors. Psychophysiology 35, 389–404.

Schupp, H.T., Cuthbert, B.N., Bradley, M.M., Hillman, C.H., Hamm, A.O., Lang, P.J., 2004.Brain processes in emotional perception: motivated attention. Cognition andEmotion 18, 593–611.

Simons, R.F., Detenber, B.H., Roedema, T.M., Reiss, J.E., 1999. Emotion processing in threesystems: the medium and the message. Psychophysiology 36 (5), 619–627.

Zinser, M.C., Fiore, M.C., Davidson, R.J., Baker, T.B., 1999. Manipulating smokingmotivation: impact on an electrophysiological index of approach motivation.Journal of Abnormal Psychology 108, 240–254.