One of the most intriguing questions in neuroersecvieonlvces around how and where emotions are psreodciens the brain. More than 100 years ago, studies in penatits with unilateral right-hemispheric lesions demnostrated that emotional processing seems to be lateraliznedthi e brain as these patients were impaired in threaibility to express their emotion1.sVese results were complemented by a large numbeor f behavioral results both from healthy and patient cohorts supporting the nohtiaot nemtotions are asymmetrically processed in threaibn235. With the emergence of brain recording and neur oinimgatgechniques such as the EEG or fM6R38I, more specifc statements about where afective states are processdewithin cortical and subcortical regions coulmdbadee. For cortical emotional processing, especially aflraolpnhta power has been strongly associated withanchges in afective state and emotional regulati9o,10n.For example, Hannesdóttir et a1l1. investigated relative leg frontal asymmetry (rLFA) and found that reduced rLFA in lcdhrien was predictive of impaired emotional reguolanti and stronger physiological responses to emottiiomnulail. sFurthermore, rLFA serves as a suitable predictor for individual diferences in emotional expression allsaws eregulatio n12,13.

Regarding the nature of emotional lateralizatwioont,hteories have been dominant in research on aseytrmiems of emotion processing. Vese theories are knowhenriagshtt hemisphere hypothesis (RHH) and the valenec model (VM) of emotional processing. Ve RHH poesstthualattall emotions regardless of valence are persosecd in the right hemisphere14. In contrast, the VM claims that positive emotiaorendsominantly processed in the leg hemisphere whereas negative emotions are prosceedsin the right hemisphere15. Both theories have received much support from behavioral, electrophysiologasicwalell as neuroimaging studi1e6.sVe nature of emotional lateralization in the brain therefore remains rratihneconclusive to this day. Despite the richnesstohefneuroscientifc literature on emotions, the vast majooriftypublished papers in this feld share a commonsueis: It is largely unclear to what extent the used paradigmlsiceit neuronal processes that actually resemble thperocesses Scienti昀؀c Reports | (2021) 11:1142 during real-life emotional encounters. A poterenatsioaln for this heterogeneity in the literaturehmtbige the lack of ecological validity. Recent systematic reewviarticles have pointed out the necessity for eocgoilcally valid research in neuroscien1c7e,18. Vis holds especially true for emotion researchthaes most prevalent method of positive or negative emotional induction is viiaesm, poicvtures or music19322. Vus, emotions are largely only perceived during experimental paradigms. However,emrely perceiving emotions might not be suocienat for valid measurement of the underlying neural subsetsr.aRteal-life emotions comprise both the feelisnelgf iats well as a preparation for and performance of an adaecqtuioaten associated with the felt emot2io3. Vnis for example involves a behavioral expression such as avoidabnecheavior if someone experiences fear, or approeahcahvibor if someone experiences happin24e.sVse lack of a behavioral component in laboraetottirnygss challenges the external validity of experimental results and makitesdiocult to transfer fndings to, for examploe, dmdoisorders for which the associated behavior is of paramomunpotirtance.

In conventional settings, this behavioral comp oisnunenftortunately diocult to realize either duetthoe experimental design or the environment in whticahkeist place, for example in an fMRI scanner orndgurEi EG recordings. Ve past decade however brought fortvhenltoechniques such as mobile EEG2s5,26, mobile fNIR2S7,28 and less constraining ME2G933s1. Vese technological advancements have been develdoapnd optimized to allow for the investigation of neural correlates in settings of high ecological validity as the parnticmipoa nvets ca freely and be tested outside the lab. While the nubmer of publications using these techniques arel stpialrse, a growing body of research is being generated coingvenrgeuroscientifc research with real-life actievsitsiuch as cycling32,33, walking34, navigating over obstac3l5e,sreal world drivin3g6, viewing real-life faces in natural settin37 gs or skateboardin38g. Recently, Packheiser and colleagu3e9sinvestigated the neural basis of hand and foowthuisle the participants were wearing a mobile EEG. Vey fnodu that both alpha and beta frequency asymmetrieesrwe predictive of the participants9 handedness andefdonoests and that the neural signals could distinhgubiestween limb preferences of individuals. Importantly, tfhoeyund that the neural signals were unafected by memoevnt parameters during activities such as jumping or tohrwing and kicking balls. Vus, mobile EEGs provivdaeluaable tool to investigate the neural basis of human eomnostaind their associated behaviors in more nastuetrtailngs.

A very prominent human behavior that is usually ceuxeted in emotional settings is social touch. Toveycoonur social intentions or emotional states to othenrhs,uwmeastrongly rely on the use of a variety otifletiancteractions, especially in very intimate social relatsiohinps40,41. Touch is the earliest sensory modality to fulelyvedlop during the lifespan42[and is experienced from birth onwards by beingdlcerdain the mother9s arm43s. For that reason, social touch has been strongly associaittehdhwuman development, shaping attachment, emotilona regulation and cognitive maturat4i4o.Anfective social touch has been demonstrated tohbigehly benefcial for the well-being and physical as well as mental heahltin humans as it reduces stress, blood pressure dancan even protect from viral infections and allergic reseps4o53n48.sStudies on the neural basis of social touch hinavdeicated that somatosensory corti4c9ebsut also limbic or orbitofrontal brain regioancstiavraeted when an experimenter or the romantic partner applies non-sexual pleastanctitle stimulation50352. Vus, there seems to be a strong overlap in cerebral processing of social touche manodtions indicating that social touch carriesoa nstgr afective component. However, as for studies investigatinogtioemns, the experimental designs investigatingntehueral basis of social touch lack ecological validityhaesatpplication of gentle touch to for exampleegtshwehlile lying perfectly still rarely occurs in real-life settin. gs

Ve aim of the present study was to investigate e mioont al lateralization in a setting with high ecocal olgi validity. To this end, we tested romantic partneinrs their home during both embracing and kissinglewt hei participants were recorded using a mobile EEG symste.We also investigated the neural correlates ofoetmional speech as this type of social interaction is funednatmal to maintain a healthy and long-lastingiorenlasthi5p3. We focused on diferences in asymmetrical processinintghe alpha frequency band due to the pronounrocleed of frontal alpha asymmetries in emotional progc.eSssinince beta power asymmetries have been demonstetrda to be highly comparable in function to alpha paoswymermetries in studies investigating motor prenfecrees39 and resting state oscillatio5n4, swe also included beta power asymmetries as a depnedent variable in our study. For each behavioral task (embracing, kissing andeescph), we employed a positive and a neutral conodinti. We hypothesize that emotional lateralization difertswbeeen the emotional and neutral condition. Iifghtht-e r hemisphere hypothesis holds true, we expect stronrgreight-hemispheric activity in the emotional comrepdato the neutral condition. If the valence hypotheslidsshtorue, we expect stronger leg-hemispheric activtyi in the emotional compared to the neutral condition.

Participants. A total of 32 individuals (16 females) took part in this study. Ve sample size was determined based on prior studies investigating frontal alpha asymmetries using within-subject dersoigdnuscipng re-li able and large efects (Cohen9s d = 1.505;4· p2 = 0.3656). Vere was no restriction regarding the sexuality of the participants, but all couples were heterosexual in the present study. Age ofptahreticipants ranged between 19 and 63  years (mean age = 29  years, SD = 14  years). Participants with neurological or psycihciadtirsorders were excluded from the study. Ve study was conducted in accordance with the declaration of Helsdinki an was approved by a local ethics committee of the psychological faculty at Ruhr UniversithyuBmo.Acll participants gave written informed consent. Five participants were excluded from the PANAS analyesiesmoofttiohnal induction and one participant was excluded from the EEG analysis of thvieobreahlatasks due to technical issues. In total, this leg 27 data sets for the PANAS and 31 data sets eligible for EEG data analysis in the fnal analysis sample. Vere was no overlap between excluded participants from the PANAS and the EEG data. Experimental task. Ve experimental paradigm consisted of conducting three behavioral tasks, i.e. embracing, kissing and listening to speech in both an emotional and neutral condition. Ve neuntral condi served as a control condition and was conducted using identical movements to control for mohteor efects i experiment. Testing took place in the participants9 homes to provide as much nseattuinragl and ecological validity as possible. One participant was set up with the mobile EEG system while the other participant flled out demographic questionnaires, the relationship assessment scale (RAS) and pedreapnar emotional text about fond memories and experiences with their partner. Importantly, the partner, thsantowat recorded, was also equipped with an electrode cap to reduce the awkwardness of only one partner wearinpg. tNhexct,athe participants were instructed about the behavioral tasks and how to perform them apptreolypbriyathe experimenter and an illustrating photograph. Ager instructing the participants, the experimenter leg the roooallmo wt for privacy for the entire experimental procedure. Ve behavioral tasks looked as follows: 1. During the embracing condition, the participawnertes either asked to embrace each other from threofnt in the emotional condition (1FAig),. or embrace a body pillow in the neutral conodnitiwith the partner being absent (Fig.1  B). Ve participants were further instructed to avdotiouching the electrode cap and move as little as possible during the embrace. 2. During the kissing condition, the participanertes ewither asked to kiss each other on the lipstihne emotional condition (Fi1gC. ), or kiss their own hand during the neutral c otionndi by forming a lip-like structure with the thumb and index fnger with the partner bnegi absent (Fig1.D ). Importantly, the participants were instructed not to use their tongue during thei nkisesither condition to avoid strong motor a. rtifacts 3. During the speech condition, the participant rwi neag the mobile EEG system was either listening tthoe partner reading the previously prepared emotionexatl tin the emotional condition 1(FEi)g,.o r was listening to a weather report that was recorded priorthtoeexperimental session in the neutral conditi oithnthwe partner being absent (Fig1F. ).

Ve experimental design was counterbalanced so thaetexperiment could start with the neutral or emiontal condition and with the male or female partner breeicnogrded initially. Furthermore, during the emontiaoland neutral condition, the order of behavioral tasskfsuwllay randomized. Each behavioral task was perfmoerd for 1 min in total and the tasks were separated by antiertrial interval (ITI) of 30 s in accordance wthitehprocedures used in Packheiser et al3.9. During the ITI, the participant(s) received an additional pre-recorded auditory and visual instruction about the upcoming behaviosrka.l Vta is was necessary as it was unknown to the picairptant(s) if the procedure started with embracing, kissing soperech due to the randomization procedure. Ve riuncs-t tions were presented using the Presentation sog(wNaeruerobehavioral Systems Inc., CA, USA). Vere waas 5 min break between the emotional and neutral ctoionndito fll out the Positive and Negative AfecteSdcuhle (PANAS), a questionnaire evaluating their positaivned negative afective states by indicating the cuenrrt emotional state on 10 positive and 10 negative itenmsa oscale from 1 (Very slightly or not at all)(Etoxt5remely)57. Ve break was also used to allow for the partnerletaove or to join (depending whether the recorderdtnpear was in the emotional or neutral condition, revseplye)c.tWie deliberately chose to let the partner levae the room in the neutral conditions to account for poctoenntfiaol und variables such as social support and nrobnal-ve communication between the partners.

Ager one participant had completed the experimenptarlocedure (three emotional and three neutral ta),sks the recorded partner again flled out the PANAS shoattthe afective state was measured ager both the emotional and neutral condition. Agerward, the roles ofatrhtnepers switched, and the non-recorded partner wthenrotugh the identical experimental protocol.

EEG recording, preprocessing and analysis. EEG signals were obtained with a mobile EEG recording system (LiveAmp 32, Brain Products GmbH, Gilching, Germany). Ve LiveAmp 32 comprises 32 Ag3AgCL electrodes arranged in the international 10320 system (C3/C4, FP1/FP2, Fz, F3/F4, F7/F8, FCz, FC1/FC2, FC5/ FC6, FT9/FT10, T7/T8, CP1/CP2, CP5/CP6, TP9/TP10, Pz, P3/P4, P7/P8, Oz and O1/O2). Ve FCz electrode served as reference signal during data recording. All signals were amplifed using a wireless amplifer (analogto-digital conversion: 24-bit) and recorded using the Brain Vision Analyzer sogware at a sampling rate of 1 kH Impedances were lowered to under 10 kHz prior to the recording session to ensure good signal quality. Ve EEG system furthermore comprised three acceleration sensors in the X (mediolateral axis), Y (anteroposterior axis) and Z (dorsoventral axis) direction located at the rear of the skull that recorded movementsaortfictiphaenpts9 head.

Following data acquisition, the EEG signals wereeprporcessed onine in Brain Vision Analyzer (Broaidn- Pr ucts GmbH, Gilching, Germany). Ve raw data fles weerband-pass fltered from 0.1 Hz (high pass) to 30z  H(low pass) at 24 dB (octave). All signals were manualilynspected for technical artifacts and channelsooofrprecording quality. Systematic artifacts, i.e. horizontal oerrtvical eye movements as well as pulse-associatedigsnals, were removed via the application of an infomax independent component analysis (ICA). Ve reference channel (F and channels of insuocient signal quality werelcruelcaated via topographic interpolation.

Ager preprocessing, the individual tasks were frstegmented across the entire trial duration (60 sn) da then baseline corrected. Ve 500 ms prior to tasksoetnwere used as baseline signal. Ve large trial sgement was then further divided into 58 non-overlapping setgs moefn1024 ms duration each. Individual segmentsrewe excluded via an automatic artifact rejection if aonfythe following criteria were met: ( 1 ) voltaegpes sot f 50 µV / ms, ( 2 ) amplitude diferences of more than 200 µV within a 200 ms interval and ( 3 ) signal strength below 0.5 µV within a 100 ms interval. In a next step, we appdliea current source density (CSD58)transformation to remove the reference potential from the fltered and segtmede ndata. Finally, we used a Fast-Fourier transfoartmion to decompose the oscillatory data into its difereneqtuferncy bands (Hammond window of 10%). Alpha aotisocinll s were defned in the 8313 Hz range. Beta frequencwieesre defned in the 13330 Hz range. We then calcueladtthe average power density (power per unit bandwidth)r peleectrode with a bilateral arrangement (C3/C4, F1P/FP2, F3/F4, F7/F8, FC1/FC2, FC5/FC6, FT9/FT10, T7/T8, CP1/CP2, CP5/CP6, TP9/TP10, P3/P4, P7/P8 and O1/O2) and extracted it for the three tasks across bothncdoitions individually. In a fnal step, asymmetrdyiicnes (AIs) were computed between the electrode pairs using tfhoellowing formula in accordance with Ocklenb aul.r5g4: et

AI = ln power right − ln power left Statistical analysis. Statistical analyses were conducted using SPSS (version 21, Chicago, Ilinois, USA). Ve PANAS scores were evaluated using a two-factorial repeated measures ANOVwitAh the factor valence (two levels: average score for all positive and all negative items) and the factor condition (two foacttiorns:ael mand neutral). Post hoc testing was performed using a Bonferroni correction. Neural data was analyezleydfsoeprarat the three behavioral tasks. We investigated diferences in AIs in the alpha and beta frequency band onc-all ele trode-pairs for which they could be computed, i.e. for all non-central electrodes. We computed a otwrioal-fact repeated measures ANOVA with each individual electrode pair as the frst factor (14 levels: C3/C4, FP1/FP2, F3/ F4, F7/F8, FC1/FC2, FC5/FC6, FT9/FT10, T7/T8, CP1/CP2, CP5/CP6, TP9/TP10, P3/P4, P7/P8 and O1/O2) and the experimental condition as second factor (two levels: emotional and neutral). Again, postrih-oc compa sons were conducted using a Bonferroni correction. For all analyses, we also used sex as a bejetwcteveanri--sub able to identify sex-related interactions. Since partners were tested consecutively rather thaanralilnel,pwe also used the sequence of testing as a between-subject variable to exclude any efects of testing order. If a nsi gtnifca diference between the emotional and neutral condition could be detected on a specifc electrode pa-ir, we fur thermore correlated this diference with the afectivity score from the RAS question.nTaoiriedentify movementrelated diferences between the conditions, we extracted the acceleration sensor signals on the X-,Y- and Z-axis during the emotional and neutral condition. A two-factorial repeated measures ANOVA waesdpewrifthorm the factor orientation (three levels, X,Y and Z) and condition (two levels: emotional and neutral).

Emotional induction. First, we investigated whether our emotional condition elicited more positive afective states compared to the neutral conditions using the PANAS scores. Ve refsutlhtseoPANAS questionnaire were evaluated by comparing the average value of all positive and all negative items between tohetieomnal and neutral condition in a 2 × 2 ANOVA. We found signifcant main efects of v(1a,2l6e)n=c3e29(.F72, p < 0.001, · p2 = 0.93) and condition(1(,2F6) = 72.42, p < 0.001, · p2 = 0.74) with the positive items being rated higher (mean score = 2.75) than the negative items (mean score = 1.24) and the emotional condition receiving higsher ratin (mean score = 2.35) compared to the neutral condition (mean score = 1.64). We found a signifcant interactio between item valence and the experimental conditio n(1,s26()F= 54.07, p < 0.001, · p2 = 0.68, see Fig. 2). Bonf-er roni-corrected post-hoc tests revealed signifcantly higher positive afect in the emotional (mree=an3.4s6co, SEM = 0.14) compared to the neutral condition (mean score = 2.05, SEM =p0<.103.,001). Negative afect did not difer between the conditiopn>s0(.250) and was basically absent in both the emotional (mean score = 1.23, SEM = 0.05) and the neutral condition (mean score = 1.25, SEM = 0.07).

Alpha power asymmetries. To investigate diferences in neural processing between the emotional and neutral condition, we investigated changes in AIs between the conditions for all three behskasvionrdailvtidaually in a 2 (factor condition) × 14 (factor electrode pair) ANOVA. For the embracing condition, we found neit a signifcant main efect of conditio(1n,30)(=F2.39, p = 0.133, · p2 = 0.07), nor a signifcant interaction between condition and electrode pair(s13(,3F90) = 0.61, p > 0.250, · p2 = 0.02, Fig. 3A). Ve interaction between condition, electrode pairs and sex did not reach signifcanc(e13(,3F77) = 0.73, p > 0.250, · p2 = 0.03). For the kissing condition, we found no signifcant main efect of cond(i1t,3io0) =n0.(3F8, p > 0.250, · p2 = 0.01), but a signifcant inter-ac tion between condition and electrode pa(1i3r,s39(0F)= 1.76, p = 0.048, · p2 = 0.06). Bonferroni corrected post hoc testing revealed a signifcantly higher asymmetry index on the FP1/FP2 electrode pair in the emotional (mean µV2/Hz = 0.24, SEM = 0.10) compared to the neutral condition (mea n2/HµzV= 2 0.05, SEM = 0.10, p = 0.043, Fig. 3B). Ve interaction between condition, electrode pairs and sex did not reach signif(c13a,3n77c) e= (1F.55, p = 0.096, · p2 = 0.05). For the speech condition, we found no signifcant main efect of co(1n,30d)i=ti0o.9n4,(F p > 0.250, · p2 = 0.03), but a signifcant interaction between condition and electrode(13p,3a9i0r) =(F2.28, p = 0.007, · p2 = 0.07). Bonferroni corrected post hoc testing revealed a signifcantly lower asymmetry index on the P7/P8 electrode pair in the emotional (mean2/µHVz = 2 0.06, SEM = 0.08) compared to the neutral condition (mean µV2/Hz = 0.18, SEM = 0.06, p = 0.022, Fig. 3C). Ve interaction between condition, electrode pairs and sex did not reach signifcance ((F13,377) = 1.57, p = 0.091, · p2 = 0.05). Vere were no signifcant results for sequence. Co-rrela tions with RAS scores did not reach signifcance for any behavioral task.

Beta power asymmetries. We repeated the analysis conducted for the alpha frequency band in the beta frequency band. For the embracing condition, we found neither a signifcant main efec-t of con tion ((F1,30) = 3.58, p = 0.068, · p2 = 0.11), nor a signifcant interaction between condition and electrode pairs (F( 13,390 ) = 1.49, p = 0.118, · p2 = 0.05, Fig. 4A). Ve interaction between condition, electrode pairs and sex reached signifcance ((F13,377) = 1.98, p = 0.021, · p2 = 0.06). Here, males showed a higher asymmetry index on the F7/F8 and the FT9/FT10 electrode pairs in the emotional (mean2/HµVz= 0.21, SEM = 0.18) compared to the neutral condition (mean 2µ/VHz = 2 0.40, SEM = 0.17, p = 0.005). Similarly, males showed a higher asymmetry index on the FT9/FT10 electrode pair on the emotional (mean2/HµVz= 0.17, SEM = 0.13) compared to the neutral c o-n dition (mean µ2V/Hz = 2 0.07, SEM = 0.09, p = 0.017). No diference could be detected for female participants. For the kissing condition, we found no signifcant main efect of c o(1,n30)d=it1io.8n9, p(F= 0.189, · p2 = 0.06), but a signifcant interaction between condition and electrode(p13a,3i9r0s) =(F2.57, p = 0.002, · p2 = 0.08). Bonf-er roni corrected post hoc testing revealed a signifcantly higher asymmetry indehxe oFPn1/tFP2 electrode pair in the emotional (mean 2µ/VHz = 0.25, SEM = 0.09) compared to the neutral condition (mea n2/HµzV= 2 0.10, SEM = 0.09, p = 0.007). Furthermore, there was a signifcantly higher asymmetry index on the F7/F8 electrode pair in the emotional (mean 2µ/VHz = 2 0.04, SEM = 0.10) compared to the neutral condition (mean2/ µV Hz = 2 0.49, SEM = 0.14, p = 0.011, Fig. 4B). Finally, we found a signifcantly higher asymmetry index on the O1/ O2 electrode pair in the emotional (mean2/µHVz = 0.11, SEM = 0.12) compared to the neutral condition (mean µV2/Hz = 2 0.25, SEM = 0.15, p = 0.024). Ve interaction between condition, electrode pairs and sex did not reach signifcance ((F13,377) = 1.57, p = 0.092, · p2 = 0.05).For the speech condition, we found no signifcant main efect of condition((1,F30) = 0.92, p > 0.250, · p2 = 0.03), but a signifcant interaction between condition and electrode pair (F( 13,390 ) = 2.41, p = 0.004, · p2 = 0.07). Bonferroni corrected post hoc testing revealed a signifcantly high-er asym metry index on the FP1/FP2 electrode pair in the emotional (mea2n/Hµz V=0.22, SEM = 0.10) compared to the neutral condition (mean2/µHVz = 2 0.08, SEM = 0.11, p = 0.016). Furthermore, there was a signifcantly higher asymmetry index on the F7/F8 electrode pair in the emotional (mea2n/Hµz V=2 0.06, SEM = 0.13) compared to the neutral condition (mean2/µHVz = 2 0.34, SEM = 0.12, p = 0.047, Fig. 4C). Ve interaction between condition, electrode pairs and sex did not reach signifcanc(e13(,3F77) = 1.09, p > 0.250, · p2 = 0.04). Vere were no signifcant results for sequence. Correlations with RAS scores did not reach signifcance for any behsakv.ioral ta Acceleration sensors. To identify whether the emotional condition was associated with strong-er move ment, we compared the acceleration sensor signals between the emotional and neutral condition for ea behavioral task. To this end, we computed 2 × 3 ANOVA with the factors condition (two levteilos:ne maloand neutral) and orientation (three levels: X, Y, and Z-axis). For embracing, we found no main efect of conditi (F( 1,30 ) = 4.15, p = 0.051, · p2 = 0.12), nor an interaction with the movement orientat(i2o,60n)= (1F.93, p = 0.154, · p2 = 0.06, see Fig. 5A). For kissing, the results were comparable as we also did not detect a main efect of-condi tion ((F1,30) = 1.01, p > 0.250, · p2 = 0.03), nor an interaction with the movement orientat(2i,o60)n= (0F.08, p > 0.250, · p2 = 0.003, see Fig. 5B). Finally, the speech condition also did not yield any signifcant main efect of conditio (F( 1,30 ) = 0.57, p > 0.250, · p2 = 0.02), nor an interaction with the movement orientat(i2o,60n)= (1F.20, p > 0.250, · p2 = 0.04, see Fig. 5C). Grand averages of the movement signals split by frequency band (alpha, beta, gamma, delta) for the three behavioral tasks are depicted in SI Fig. 1.

In the present study, we used a mobile EEG to recdobrrain activity of romantic partners during afvecstoicial touch and emotional speech in their everyday ennvimroent to provide high ecological validity. We sfpceaclily focused on asymmetries in our analysis due tortohenpounced lateralization of emotional procetshsienbgrain. We found that the participants were in a more piovesimtood ager they executed the behavioral tasks wthi their respective partner. On the neural level, we founhdigaher alpha AIs on the FP1/FP2 electrode pair itnhe emotional compared to the neutral condition during kissing. Forwsepefeochu,nd a lower alpha AI in the emotional compared to the neutral condition on the P7/Pt8roeldece pair. In the beta frequency band, we foungdhehri AIs in the emotional compared to the neutral c oionndiotn the F7/F8 and FT9/FT10 electrodes only faloesr m in the embracing condition. Across both sexeso,wuenfd higher Ais in the emotional compared to tuhteranl e condition on the FP1/FP2 and F7/F8 electrode puariirndg both kissing and emotional speech. Furtheerm,or there was an increased AI on the O1/O2 electrodeirpdauring the emotional compared to the neutradlictoionn during emotional kisses. Movement signals did nifoertbdetween the emotional and neutral condition.

Increases in oscillatory alpha power have been nstrgoly associated with functional inhibition, fmoprl exa during visuospatial attenti o59,n60, face recogniti o61nand working memory tas62k.sAlpha oscillations are hypo-th esized to be generated by rhythmic burst of locnalhiibitory GABAergic interneuro62n.Isncreases in AIs are therefore indicative of stronger right-hemisphi nerhicibition or increased leg frontal activity w hsedreecareases in AIs refect stronger leg hemispheric inhibitiornrioght-hemispheric activation. Vus, the frontaclrienase in alpha power asymmetries during emotional comparoednteutral kisses indicates that frontal regio nthseolefg hemisphere were more strongly activated in the perensce of strong positive afect. Vese results arlei nine with the VM of emotional lateralization which postulatepsothsaitive emotions are processed in the leg hemhiesrpe and oppose predictions made by the RHH claiming that all emotions are processed in the right hemisphepreci-rres tive of valence. Interestingly, previous behaviorersaelarch on the efects of emotional context olantertahleization of social behavior has indicated that the RHH prdoevsithe overall best prediction to explain chanigneslaterality in emotional compared to neutral situat6i3o365n.Psrete and colleague6s6 similarly found that behavioral and neural fndings regarding hemispheric asymmetries were inncgoruent and do not necessarily correspond. Idtsbheoul noted however that the VM and RHH are not necelyssmaruitually exclusive67. Killgore and Yurgelun-To6d8d have proposed an integrative model postulatingththeaVtM provides accurate predictions for anteorirofrrontal asymmetries whereas posterior or parietal asymmetersi are more in line with predictions of the RHHn.cSei we could fnd lower asymmetry scores during emotiloconmapared to neutral speech on the P7/P8 electrode pair indicating stronger right-hemispheric activoanti in the emotional condition, our results c oatrerothbaotr emotional lateralization does not seem to be umnaicfroorss cortical brain regions but is rather reng-iospecifc.

For asymmetries in the beta frequency band, we fodmunostly comparable and sometimes even larger efsect compared to the alpha band. Opposed to oscillataolprhya power, the functional role of the beta fnrecqyubeand has been rather inconclusive. While some studiesvheasuggested that beta activity is indicative of gconitive activation69,70, beta power has also been suggested to be assoecdiawtith the function of inhibitory interneuron networks indicating that alpha and beta activityarsehsimilar characteristics71. In a previous mobile EEG study investigating alpha and beta asymmetries during mooretxecution, we also found that alpha and betma maseytries were functionally similar and associated with ibnithioin39. Ocklenburg et al5.4 even found that alpha, beta, delta and theta oscillations were all signifcantly coatrerdelindicating that there might be some commonctfiuonn underlying rhythmic brain activity in general. Ionnctrast to alpha power however, beta asymmetriesddni ot reveal decreased AIs on occipital electrodes duri nthge emotional kiss, but rather increased AIs comrapbale to frontal electrodes. Vus, while there are evidleynstimilarities between alpha and beta power, theayre not simply functionally identical. Vese fndings sugtgheasttbeta power asymmetries should be investigatemd ore thoroughly in the future to clearly identify thnecftiuonal role of beta asymmetries in the brain.

Opposed to the kissing and speech condition, wledcnoout fnd any overall efects in the alpha or beta frequency band for the embracing condition. Oennetipaoletxplanation for this lack of a fndings rteolathee fact that in contrast to long lasting kisses otrioemnoal speech, embraces take place frequently bet wene platonic friends and even unfamiliar individ4u3.alVsus, the emotional condition might have lacksetdroang afective component as embraces are not a partner-specifecraincttion eliciting strong emotional responshese.rFmuortre, embraces are usually shorter than the 1-min interlveamployed in our experimen72,twhich was necessary for reliable data acquisition. Ve unusually long duoranti could have negatively infuenced our resulthsiisnextperiment. Interestingly, we however found a sex-speecfifect in the beta frequency band with only maliesspdlaying higher asymmetry indices over frontal electrodees.siIt could be speculated that males experiencedhte embrace more emotionally because they engage in embracsess olegen in everyday life as compared to femalepse,ceisally in close male-male relationshi7p3.s

An important issue that needs to be addressed aroetpential efects of eye movements that might havecatfed our present results since eye movements were noptucraed by the acceleration sensors. While ICA algitohrms can reduce eye movement artefacts, results over F/PF1P2 electrodes still raise suspicions. We howevfermly believe that eye movements did not infuence thespernet fndings for two major reasons: frst, thelactaiolcun of asymmetry indices likely cancels out all ocualratrefacts due to eye movements and occurring ftohrebyeos simultaneously during blinks. Since signals froe mle gthhemisphere are subtracted from right hemisphree signals, all residual artefacts should have been elimniated by this computation. Second, our individeugamlsents were not time-locked to a trial start since we instivgeated 1 min of oscillatory data for each behravlitoask in each condition. We then segmented this data andraagveed it. Vis procedure likely averages out any reamining ocular signals as they occur randomly across t hiael tlrength.

Given the relative novelty of the approach andetxhpeerimental paradigm, there are several limitatiosnassociated with the present study. First, the sample size of the study was rather low even for a withinbj-escut design.

Vus, it might have prevented the detection of smeralelfects due to insuocient power. Second, the pernest paradigm employed no negative emotional conduitei otonbdoth practical reasons (dioculty to artiflcyiainlduce negative emotions during embraces and kisses) and ethical reasons (possible tension arising innthsheirpe)l.atio Unfortunately, the lack of a negative emotiondaitlicoonndoes not allow to conclusively embed ouultrs rinesto theories of emotional lateralization as the valheynpcoethesis distinguishes between positive and ntievgearather than positive and neutral afective states. Finalwlye, could not investigate brain-to-brain synchbroetwneyen both partners as there was only one mobile EEGesmysatvailable.

In conclusion, we found diferences in alpha p osywmemraetries during emotional compared to neutral conditions in highly ecological situations thcaot nargeruent with models of emotional lateralizinatitoegnrating frontal valence and posterior right-hemisphericepsrsoing. To provide conclusive evidence in thiasrrde,gfuture studies should however conceive a similar experimteannd include a negative emotional conditio nlis(tein.gg. annoying habits of the partner) that was not ptriensetnhis study. Additionally, future studies shoduinlvestigate beta frequencies in more detail as the functioonlael droes not seem to be identical compared to aflrpehqauencies. Furthermore, the ecological validity of the pressetnudty is rather limited to couples from Westernciseoties. For kissing, there are notable diferences in both ferneqcuy and lateralization based on the co n743te76x,tespecially for cultural context7s7. Similar results have been found for the frequyeonbtlserved leg cradling bias of childr7e8,n79. Vus, it would be interesting to see if cultural edriefnces in social behavior are refected in alternedeurophys-i ological processing of these types of interacFtionnalsly., there was little variance in relationshsiaptisfaction in our sample. Future studies could replicate simairl experimental designs, but specifcally invite ptaircipants from both happy and unhappy relationships to ifdyenpoti ssible diferences due to the diference in atfiveicty.

Scienti昀؀c Reports | (2021) 11:1142 | https://doi.org/10.1038/s41598-020-80590-w 11

Vis work was supported by the Deutsche Forschungsmgeinschag Grant Number OC127/9-1 and the Research Training Group <Situated Cognition= (GRK 2185/e1a)c. kWnowledge support by the Open Access Publicanti os funds of the Ruhr-Universität Bochum.

J.P. and G.B. conceived the study, analyzed the daaatnd wrote the manuscript. N.R. consulted on asnesaalynd reviewed the manuscript. L.S. and C.B. collected dta and reviewed the manuscript. O.G. provided mateiarls and reviewed the manuscript. S.O. conceived the studyc,onsulted on data analyses and reviewed the maniupstc.r

Open Access funding enabled and organized by PtrDoEjeAkL.

Ve authors declare no competing interests.

Supplementary Information Ve online version contains supplementary material availlabhlettpast://doi. org/10.1038/s41598-020-80590- w.

Correspondence and requests for materials should be addressed to J.P.

Reprints and permissions information is available at www.nature.com/reprin.ts Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional aoliations.

Open Access Vis article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the souvirdce,aplrionk to the Creative Commons licence, and indicate if changes were made. Ve images or other third party material in this article are included in the article9s Creative Commons licence, unless indicahteedrwoiste in a credit line to the material. If material is not included in the article9s Creative Commloicnens ce and your intended use is not permitted by statutory regulation or exceeds the permitted use, you wielel dnto obtain permission directly from the copyright holder. To view a copy of this licence,hvtitspit://creativecommons.org/licenses/by/4..0/ Scienti昀؀c Reports |