Contents lists available at ScienceDirect

Biological Psychology

journal homepage: www.elsevier.com/locate/biopsycho

Affect dynamics of facial EMG during continuous emotional experiences

Yulia Gollanda,⁎, Adam Hakimb, Tali Alonia, Stacey Schaeferc, Nava Levit-Binnuna

a Sagol Center for Brain and Mind, Baruch Ivcher School of Psychology, Interdisciplinary Center Herzliya, Herzliya, Israelb Sagol School of Neuroscience, Tel Aviv University, Israelc Center for Healthy Minds, University of Wisconsin Madison, Israel

A R T I C L E I N F O

Keywords:Affective dynamicsFacial EMGContinuous measuresInter-subject variability

A B S T R A C T

Emotional experiences are complex, multi-component processes that unfold over time. Accordingly, it is crucialto understand the temporal dynamics of these constituent components. Here we studied the dynamics of one ofthe core emotional systems, i.e. facial muscle activity, during continuous emotional experiences, elicited bymovies. We found that transient zygomatic fluctuations were narrowly tuned to a positive emotional experience.During a positive but not a negative movie, zygomatic response patterns were consistent across participants,tracked with subjective ratings and co-varied with emotional dynamics. Corrugator activity evidenced a broaderaffective profile and larger individual variability. It was coordinated with tonic changes in emotional negativityand inversely coupled with transient changes in positive affect. Taken together, our results confirmed previousfindings on the affective profiles of facial responses and extended them to temporal dynamics. They furtheruncovered important differences in temporal response characteristics of zygomatic and corrugator measures.

1. Introduction

Contemporary approaches highlight the dynamic nature of emo-tions, suggesting that emotional experiences emerge as a result of in-teractions between multiple components and continuously change overtime (Barrett, Mesquita, Ochsner, & Gross, 2007; Boiger & Mesquita,2012; Frijda, 1988; Scherer, 2009). A growing research effort is in-vested in understanding the temporal dynamics of emotional processesand its constituent components (Hollenstein & Lanteigne, 2014;Kuppens, 2015; Sander, Grandjean, & Scherer, 2005; Schuyler et al.,2014; Waugh, Shing, & Avery, 2015). We add to this effort by in-vestigating the dynamics of facial activity, a prominent emotional re-sponse system, during continuous emotional experiences. While abun-dant previous research has demonstrated that facial activity measuredvia electromyography (EMG) serves as a reliable and valence specificmeasure of affect (Cacioppo, Petty, Losch, & Kim, 1986; Lang,Greenwald, Bradley, & Hamm, 1993; Larsen, Norris, & Cacioppo, 2003;Tassinary & Cacioppo, 1992), this research has been mainly confined toshort-term emotional stimuli and measures aggregated across time.Consequently, systematic investigation of EMG dynamics and theirtemporal and affective characteristics is lacking.

1.1. Affect-specificity of facial responses

Facial expressions have long been considered to serve an essentialemotional function (Darwin, Ekman, & Prodger, 1998; Ekman &Rosenberg, 1997; Fridlund & Cacioppo, 1986), accompanying emo-tional experiences and conveying them to others (Ekman & Scherer,1984; Ekman, 1992). Assessment of electrical activity from facialmuscles (EMG) provides a reliable and valence specific measure ofemotional expressions (Dimberg, 1990; Fridlund & Cacioppo, 1986),capable of revealing activity too subtle to detect visually (Tassinary &Cacioppo, 1992). Specifically, activation of the zygomaticus majormuscle, which pulls the corners of the mouth into a smile, is con-sistently associated with positive emotions, while activation of thecorrugator supercilii muscle, which draws the brows into a frown, ispotentiated by unpleasant and attenuated by pleasant emotional states(Cacioppo et al., 1986; Dimberg, 1990; Lang et al., 1993; Larsen et al.,2003; Sato, Fujimura, Kochiyama, & Suzuki, 2013; Tan et al., 2012).

The vast majority of previous EMG research has been based onquantifying average EMG responses elicited by short emotional inputs,such as pictures or sounds. While recent studies have begun to unravelthe temporal characteristics of EMG responses to short emotional events(Chołoniewski et al., 2016; Heerey & Crossley, 2013; Korb, Grandjean,& Scherer, 2010; Lapate et al., 2014; Van Reekum et al., 2011), thedynamics of EMG activity during complex emotional experiences

https://doi.org/10.1016/j.biopsycho.2018.10.003Received 17 January 2018; Received in revised form 31 July 2018; Accepted 3 October 2018

⁎ Corresponding author at: Baruch Ivcher School of Psychology, Interdisciplinary Center (IDC), Herzliya, Kanfei Nesharim St., P.O.Box 167, 46150, Israel.E-mail address: ygolland@idc.ac.il (Y. Golland).

Biological Psychology 139 (2018) 47–58

Available online 06 October 20180301-0511/ © 2018 Elsevier B.V. All rights reserved.

T

unfolding over time have not been systematically studied.

1.2. Reliability of facial dynamics

To elicit naturalistic emotional experiences in the lab, researchfrequently employs emotional films (Rottenberg, Ray, & Gross, 2007).Excerpts of cinematic movies have been shown to evoke powerfulemotional states (Rottenberg et al., 2007) as well as consistent phy-siological changes (Kreibig, Wilhelm, Roth, & Gross, 2007; Kreibig,2010), which are shared across observers. For instance, studies haveshown that positively valenced movies evoke enhanced zygomatic ac-tivation across participants, while negative movies lead to increasedcorrugator responses (Codispoti, Surcinelli, & Baldaro, 2008; Hess,Banse, & Kappas, 1995; Wilhelm et al., 2017). This research has beenconducted from a state-oriented perspective on emotions, using mea-sures aggregated across time. Yet, movies involve a series of succeedingemotional events at varying temporal scales, providing a continuousstream of affective information with unique temporal characteristics.Currently, little is known about whether and how these emotional dy-namics are reflected in continuous, moment-to-moment changes ofEMG response patterns.

Reactivity measures, commonly used to study emotions, lack thetemporal information necessary to investigate dynamics. As an alter-native, previous research on complex physiological dynamics has oftenemployed measures of response similarity across individuals exposed toidentical stimuli. Such temporal alignment of physiological fluctuationsacross individuals has been used as an index of stimulus-response re-liability (i.e. reproducibility), when assessing activity over time (Hasson,Malach, & Heeger, 2010). For example, imaging studies have demon-strated that naturalistic emotional stimuli, such as movies or musicalpieces, elicit common neural dynamics in participants’ emotional braincircuitry (Jääskeläinen et al., 2016; Nummenmaa et al., 2012, 2014;Singer et al., 2016; Trost, Frühholz, Cochrane, Cojan, & Vuilleumier,2015). Similar results were obtained for the autonomic nervous systemmeasures (ANS), demonstrating significant temporal alignment of ANSresponses across observers while watching emotional movies (Golland,Keissar, & Levit-Binnun, 2014; Golland, Arzouan, & Levit-Binnun,2015). EMG activity reflects the end result of the nervous system ac-tivity, being shaped by multiple emotional and cognitive factors as wellas by social display rules (Hess et al., 1995; Scherer & Ellgring, 2007).In addition, spontaneous facial responses seem to exhibit significantinter-individual variability (Duran, Reisenzein, & Fernandez-Dols,2017; Fernandez-Dols, Sanchez, Carrera, & Ruiz-Belda, 1997). Ac-cordingly, whether a continuous stream of complex affective processingelicits consistent fluctuations in EMG responses and whether thesefluctuations are similar across participants remains an open question.

1.3. Temporal alignment with affect dynamics

Although there is a common consensus that facial expressions re-liably represent subjective emotional experiences (1993, Ekman, 1992),literature on this matter is inconclusive. While some studies reportstrong correlations between emotional reports and facial EMG activa-tions (Bradley & Lang, 2000; Lang et al., 1993; Larsen et al., 2003; Satoet al., 2013; Tan et al., 2012), recent meta-analyses fail to find con-sistent links (Kappas, 2003; Reisenzein, Studtmann, & Horstmann,2013). Given the dynamic nature of emotional experiences and facialbehavior (Cunningham, Dunfield, & Stillman, 2013; Krumhuber,Kappas, & Manstead, 2013; Kuppens, 2015), investigation of the tem-poral correspondence between facial behavior and fluctuations of affectmay contribute significantly to the aforementioned issue (Hollenstein &Lanteigne, 2014; Krumhuber & Scherer, 2011). This approach has beenpioneered by the early works in facial emotional responses (Cacioppo,Martzke, Petty, & Tassinary, 1988; Rosenberg & Ekman, 1994) whichprovided initial evidence for the temporal coherence between the ex-pressive and the experiential emotional systems. A recent study has

directly assessed temporal correspondence between facial expressionsand subjective experience during a mixed emotion movie, containingdistinctive sadness and amusement episodes (Mauss, Levenson,McCarter, Wilhelm, & Gross, 2005). Using a clever approach in whichmoment-by-moment changes in the intensity of amusement and sadnessfacial expressions were coded by external raters, this study has de-monstrated that facial behaviors show within-person correlations withthe momentary changes in corresponding subjective feelings. This studyhas provided evidence that EMG activity is linearly coupled with fluc-tuations of affect within participants. However, the robustness of thisresult is unclear as it could reflect dichotomic switches between highlydistinctive emotional states (that is sadness and amusement), ratherthan a continuous unfolding of emotional dynamics. In addition, thisstudy was based on a holistic encoding of sadness and amusement ex-pressions, potentially involving multiple facial and body features. Ac-cordingly, whether zygomatic and corrugator response patterns reliablyrepresent continuous affective dynamics is unclear. Finally, althoughthe most consistent evidence for the EMG-affect correspondence hasarrived from the dimensional studies of emotion (Bradley & Lang, 2000;Lang et al., 1993; Larsen et al., 2003; Sato et al., 2013; Tan et al., 2012),the correspondence of continuous facial activity with changes on thevalence dimensions (as opposed to distinct emotional categories, suchas sadness) has not been tested.

1.4. Methodological considerations in the temporal domain

Emotional dynamics can occur at multiple temporal scales, spanningfrom seconds to minutes and even hours (Levenson, 2014; Rottenberget al., 2007). When measured over time, short-term distinct emotionalevents are expected to elicit transient fluctuations in affect dynamicswhile more gradual emotional changes, unfolding over minutes, mayinduce tonic components in the designated measures (Hamaker,Ceulemans, Grasman, & Tuerlinckx, 2015; Waugh et al., 2015). Ac-cordingly, prolonged physiological dynamics elicited in affective con-text may reflect phasic fluctuations, tonic changes or a combination ofboth (Cacioppo et al., 1988; Wang, Liu, Yianni, Aziz, & Stein, 2004).While these time-related considerations are acknowledged in emotionresearch (e.g. Butler, 2011), they are rarely assessed systematically inphysiological studies. Centrally, the temporal sensitivity of facial EMGdynamics to phasic and tonic affective fluctuations is unknown.

1.5. The present study

In the present study we aimed to conduct a systematic investigationof the facial EMG dynamics, elicited by continuous emotional experi-ences. To study affect dynamics, consistent fluctuations in emotionalresponses should be reliably elicited to begin with. Previous researchhas demonstrated the efficiency of commercial movies in producingpotent emotional responses, which are similar across observers(Golland et al., 2014; Nummenmaa et al., 2012; Rottenberg et al.,2007). Here we employed pre-tested excerpts from emotional movies,which targeted a positive and a negative emotional experiences. Thepositive movie, depicting a performance of a child, elicited warmfeelings and joy, strongly associated with zygomatic activations. Thenegative movie, taken from a horror film, induced fear and distress. Aseries of previous studies reported consistent corrugator activations inresponse to fear stimuli (Codispoti et al., 2008; Dimberg, 1986; Magnée,Stekelenburg, Kemner, & de Gelder, 2007). Fear was chosen over sad-ness, since it is more prone to evoke phasic emotional events and wasshown to elicit higher corrugator activation as compared to sadness in aprevious study employing movies (Kreibig et al., 2007). Finally, wevalidated that both emotional movies elicited transient changes as wellas tonic modulations of emotional experience (Fig. 1, SupplementaryFig. 1).

As a first step, we assessed the temporal response characteristics ofcontinuous EMG time-courses, while differentiating between its phasic

Y. Golland et al. Biological Psychology 139 (2018) 47–58

48

and tonic components. The dynamics of EMG responses may manifestslow modulations, i.e. general tonic increases, or phasic changes, i.e.transient fluctuations on a time scale of seconds. Since previous re-search on the temporal properties of facial EMG is lacking, we em-ployed a data-driven approach, which defies the need for a priori hy-potheses. To assess the susceptibility of zygomatic and corrugatorsignals to slow, tonic changes we computed their autocorrelationfunctions, i.e. correlation of a signal with itself as a function of timeshift. Higher presence of slow components induces higher dependencyof consecutive observations, i.e. slower declines of autocorrelationfunction in time. Autocorrelation width was computed for both emo-tional movies and for non-emotional physiological baseline. To examinewhether zygomatic and corrugator signals differed in their tonic pre-dominance we compared the width of their autocorrelation functionswithin each experimental condition. To examine whether tonic changeswere linked with distinctive emotional states we compared the auto-correlation width across experimental conditions. In addition, we in-vestigated whether phasic EMG fluctuations reliably reflect emotionaldynamics. For that aim we performed the below described analysesboth on the original EMG signals and on the de-trended signals, lackingthe slow, tonic components.

Central to the present study, we investigated whether EMG fluc-tuations are reliably driven by the emotional movies. Past research hasdemonstrated that emotional movies evoke consistent neural and au-tonomic temporal responses patterns which are time-locked acrossparticipants (Golland et al., 2014; Golland, Levit-Binnun, Hendler, &Lerner, 2017; Hasson et al., 2010; Nummenmaa et al., 2012). As evi-denced by a previous study (Golland et al., 2014) and by a preliminarytest (Supplementary Fig. 1), the emotional movies, employed in thecurrent research, elicited robust emotional dynamics, which were si-milar across participants. Here we examined whether these movies in-duced reliable facial response patterns over the zygomatic and corru-gator sites. For that aim, we conducted an inter-subject correlation

(interSC) analysis, which quantifies response similarity across partici-pants viewing the same movie (Golland et al., 2014; Hasson, Nir, Levy,Fuhrmann, & Malach, 2004; Nummenmaa et al., 2012). Given theconsistent links of zygomatic responses with positive affect, we ex-pected to find enhanced reliability, i.e. interSC, of the zygomatic re-sponse fluctuations during the positive movie as compared to the ne-gative movie or non-emotional baseline. The corrugator responses tendto exhibit a linear correlation with valence, being activated by thenegative affect and inhibited by the positive affect (Lang et al., 1993;Larsen et al., 2003). However, whether such response patterns arepreserved during continuous emotional experiences was not studiedbefore. We expected both the negative and the positive emotional dy-namics to induce reliable corrugator fluctuations, leading to higherinterSC during the emotional movies as compared to non-emotionalbaseline.

Finally, we examined whether continuous changes in zygomatic andcorrugator EMG measures were temporally coordinated with the movie-driven fluctuations of emotional experience on a valence dimension.Specifically, we asked whether EMG fluctuations were linearly alignedwith the movie-driven emotional fluctuations. The emotional timelinesof the movies were assessed in an independent sample of participants(Fig. 1, Supplementary Fig. 1). We expected to find enhanced zygomaticalignment with fluctuations of positive affect and enhanced corrugatoralignment with fluctuations of negative affect. Furthermore, we ex-pected the corrugator dynamics to be negatively correlated with posi-tive valence fluctuations.

2. Methods

2.1. Participants

Forty four female students (age: M=23, SD=2.94) participated inthe study for course credits (the current experiment was part of a larger

Fig. 1. A. Dynamic timelines of EMGand valence during emotional movies.Upper panel shows facial EMG measure(zygomatic: EMGZYG in orange, corru-gator: EMGCORR in blue), corrected byimmediately preceding baseline (a tenseconds countdown screen). Lowerpanel shows dynamic ratings of valence(black), elicited by emotional movies.B. Autocorrelation of EMG signals.Average autocorrelation as a functionof time plots of the zygomatic (orange)and corrugator (blue) measures duringemotional movies and physiologicalbaseline. As can be seen, the corrugatorautocorrelation functions declinedmore slowly, in particular during thenegative movie. C. Autocorrelationwidth, computed for each experimentalcondition. Autocorrelation width wasassessed by computing the temporal lagat which autocorrelation dropped to0.5. The autocorrelation width was in-creased for the corrugator (blue) ascompared to zygomatic (orange) signalsacross conditions (p < 0.001). In ad-dition, it was highest during the nega-tive movie, with significant differencesacross all experimental conditions.Error bars signify 2*SE. *p < 0.05;**p < 0.001. (For interpretation of thereferences to colour in this figure le-gend, the reader is referred to the webversion of this article).

Y. Golland et al. Biological Psychology 139 (2018) 47–58

49

research, which prescribed a reliance on female participants).Experimental procedures were approved by the institutional ethicscommittee. Written informed consent was obtained after the procedureshad been fully explained.

The current research was part of a larger project (details are spe-cified in a Procedure subsection), which determined the collectedsample size, to ensure 80% power to detect effect size, f, of 0.30 at p <.05 (Faul, Erdfelder, Buchner, & Lang, 2009), while taking into account∼ fifteen percent dropout, due to corrupted physiological signals.

2.2. Emotional movies

Two movie excerpts, taken from the horror film “ParanormalActivity” (394 s), used in our previous study (Golland et al., 2014) andthe popular entertainment TV show "Britain's Got Talent" (364 s) wereused in this study to elicit negative and positive emotions, respectively.As was validated in a pre-test, the emotional timelines of these movieswere comparable in their temporal characteristics, containing bothgradual increases of affect and distinctive phasic emotional events(Fig. 1).

2.3. Dynamics ratings of valence in pre-test participants

The emotional timelines of the chosen positive and negative movieswere assessed using continuous ratings of valence, obtained in an in-dependent sample of participants. Fifteen female students arrived to thelab to participate in a behavioral rating test. They were explained thatthe goal of the experiment is to assess momentary changes in theiremotional experience while viewing movie excerpts. Following a shortpractice trial, the positive and the negative movies were presented in arandom order. Participants continuously rated how positive or negativethey have felt on a horizontal sliding valence scale, ranging from verynegative to very positive. The scale was positioned at the bottom of thescreen and allowed participants to update their affect ratings at anypoint by moving a visible marker with a mouse. A Matlab-based in-house software was used to present the movies and collect ratings (2 HZsampling rate) (for a similar method see Raz et al., 2012; Zaki, Bolger, &Ochsner, 2009). Individual ratings are presented in SupplementaryFig. 1. As can be seen from the figure, raters showed substantialagreement among them (positive: r= 0.3 ± 0.29; negative:0.68 ± 0.12), providing further support for the existence of a commonemotional timeline in the presented movies.

2.4. Procedure

Participants were introduced to the lab, connected to physiologicalsensors and sat quietly with their eyes open for five minutes while thephysiological baseline measures were collected. After that, they parti-cipated in an implicit emotional experiment (similar to Van Reekumet al., 2011), lasting for ∼15min, during which they viewed emotionalIAPS pictures (Lang, Bradley, & Cuthbert, 1999). This experiment wasnot part of the current study and is not reported here. Following a shortrest, the participants viewed a positive and a negative movie. Eachemotional movie was preceded by a ten seconds countdown screen andthe order of the movies was counterbalanced across participants. Mo-vies were presented using the Matlab’s Psychophysics Toolbox exten-sions, integrated with the physiological recording system. Followingeach movie, participants rated their emotional experience using self-report questionnaires. This procedure yielded physiological and facialmeasures from three experimental conditions: non-emotional physio-logical baseline (baseline), positive emotional movie (positive) andnegative emotional movie (negative).

2.5. Measures

2.5.1. Behavioral measures2.5.1.1. Self-reports of emotional experience. Participants rated theintensity to which they experienced distinct emotions (joy, warmth,fear, and distress), using a 9-point Likert scale. They also reported theirlevel of arousal, positive valence and negative valence, using unipolarscales. The reports on the positive and negative items after viewing thepositive and the negative movies, respectively, exhibited high levels ofreliability (Positive movie: positive valence, joy, warmth – Cronbach’sα=0.88; Negative movie: negative valence, fear, distress - Cronbach'sα=0.89). We therefore combined them into two composite scores ofpositive and negative emotions.

2.5.1.2. Movies’ emotional time-lines. The movies’ emotional timelineswere assessed using continuous ratings of valence in pre-testparticipants (N=15, all females), as described above. This approachcapitalizes on the fact that emotional movies elicit distinctive emotionaltimelines, which are commonly shared across participants, as shownboth in previous research (Golland et al., 2014; Nummenmaa et al.,2012, 2014, 2017) and for the current raters. It thus allows to assessstimulus-response regularities, beyond individual differences inemotional responses. Since the majority of ratings during the positivemovie were limited to the positive portion of the scale and those duringthe negative movie to the negative portion of the scale (SupplementaryFig. 1), individual ratings time-series were converted to unipolarpositive and negative scores, respectively. Ratings obtained duringthe negative movie were further multiplied by −1 to easeinterpretation and comparisons. For further analysis, positive andnegative emotional time-lines were computed by averaging individualratings across participants for the positive and the negative movies(Fig. 1). Notably, the averaging procedure allowed for highlightingmoments of common agreement among raters and diminishing theimpact of individual differences in ratings.

2.5.2. Physiological measures2.5.2.1. Collection. Continuous physiological measures were recordedwith a Bionomadix (BIOPAC Systems Inc., Santa Barbara, CA) MP150data acquisition system and Biopac® AcqKnowledge 4.4 software at1000 Hz sampling rate.

Facial electromyography (EMG) data from the zygomaticus majorand corrugator supercilii sites were collected from the left facial sites inaccordance with published guidelines (Fridlund & Cacioppo, 1986). Thesensor regions were cleaned using 70% isopropyl alcohol, then slightlyabraded using ELPREP skin preparation gel prior to sensor placement toreduce skin impedance to an acceptable level (below 20 kΩ). Raw EMGsignals were filtered with a 50 Hz notch filter. Manual inspection forartifacts was made using custom software (Beniczky et al., 2013). Allartifact-free data was subjected to a Fast Fourier Transform analysis toderive estimates of spectral power density (μV 2/Hz) in the 45–200-Hzfrequency band in 1 s windows (Heller, Lapate, Mayer, & Davidson,2014; Lapate et al., 2014; Van Reekum et al., 2011). Finally, the re-sulting values were log-transformed to normalize the data. This analysisyielded continuous 1 Hz EMGZYG and EMGCORR time-series. The first20 s and the last 5 s were cropped from the resulting time series to re-move non-specific accommodation effects and preprocessing artifacts.

For the mean level analysis, we computed reactivity scores bysubtracting the mean activity of the immediately preceding countdownscreen from the mean activity during each movie.

2.5.2.2. Slow trend removal. To examine the temporal resolution offacial EMG dynamics, we performed the analysis routines describedbelow both for the original time-courses and after removal of slowcomponents. A slow trend was assessed by fitting a polynomal of 2nddegree to the original signal, using a ‘least-squared’ method via polyfitMatlab function (Barbosa, Barbosa-Filho, de Sa, Barbosa, & Nadal,

Y. Golland et al. Biological Psychology 139 (2018) 47–58

50

2003; Friman, Borga, Lundberg, & Knutsson, 2004). De-trending wasdone by subtracting the slow trend from the original data. De-trendedEMG and emotional time-courses are shown in Supplementary Fig. 2.

2.5.2.3. Missing data. In the preprocessing stage, data was monitoredfor 1) low signal quality across conditions 2) excessive artifacts, such asmultiple non-specific spikes or motion artifacts in specific conditions.The overall percent of missing data points for each channel was 18% forEMGZYG and 5% for EMGCORR. While analysis was always conducted onall available data points, the number of participants who had allmeasures in all experimental conditions was thirty-five.

2.6. Data analysis

2.6.1. Assessment of tonic componentWe assessed tonic predominance in EMG signals by 1) computing

autocorrelations for individual zygomatic and corrugator measureswithin each experimental condition and 2) quantifying the temporal lagat which signals’ autocorrelation drops to half (i.e. r= 0.5). Larger lagssignify slower declines of autocorrelation, i.e. higher level of toniccomponent. The lag statistic was compared across EMG sites and acrossexperimental conditions. Physiological baseline condition was used toassess tonic predominance during a non-emotional time period.

2.6.2. Sliding windows cross-correlation approachAn assessment of temporal similarity between response time-courses

is central to the current work. To assess temporal correspondence weapplied a cross-correlation analysis, which provides a measure of si-milarity between two series as a function of temporal displacement ofone relative to the other. This approach has allowed us to examine thecross-correlation functions for a typical form of interdependency,peaking around zero and subsiding as the time shift between signalsincreases. In addition, we validated that no consistent temporal shiftsexisted between the designated time-series, allowing to extract zero lagcorrelations as a reliable measure of temporal correspondence. Sincethe degree of temporal correspondence during fluctuating emotionalstate might vary in time, we applied a sliding windows approach,computing cross-correlations in short (30 s) partially overlapping (15 s)temporal windows. Notably, sliding windows approach does not as-sume stationarity of correlations in time (Boker, Rotondo, Xu, & King,2002) and focuses on local temporal dynamics. As such it serves as aconservative estimate of correspondence between signals. The slidingwindows cross-correlation approach was applied in reliability analysisand in correspondence with emotional timeline analysis, as specifiedbelow.

2.6.3. Assessment of EMG reliability through inter-subject correlation(interSC)

To assess the reliability (i.e. reproducibility) of EMG dynamicsacross participants, we employed inter-subject correlation measures(interSC), by quantifying similarity between individual responses andthe average response pattern. The interSC approach was successfullyused by us (Golland et al., 2007,2014, 2017) and others (Hasson et al.,2010; Jääskeläinen et al., 2008; Nummenmaa et al., 2012, 2014; Trostet al., 2015) in previous studies investigating stimulus-driven temporaldynamics of physiological signals. Importantly, this approach does notassume a canonical response form, and thus accounts for the fact thatresponse patterns can be specific to a given physiological channel andto a particular movie timeline.

All response time-courses were z-normalized and then divided intoshort (t= 30 s), partially overlapping (Δt= 15 s) time windows. Foreach participant j, within each temporal window t, we computed cross-correlation between xjt , the response time-course of participant j, and

= ∑− ≠

x xt N i j it1

1 , the response time-course averaged across all partici-pants except j. Individual cross-correlation functions were then

averaged across temporal windows. Performing the cross-correlationanalysis on a variety of window sizes (ranging from 120 s to 6 s) al-lowed us to validate that the main results of the cross-correlationanalysis are not distinctively dependent upon window size parameters(Supplementary Fig. 3).

For further statistical analyses, we extracted individual zero lagcorrelation values (hereafter, interSC) from these averaged cross-cor-relation functions. Individual interSC scores were used in analyses ofvariance, after applying r-to-z Fisher transformations. The above pro-cedure has yielded interSC indexes for the zygomatic and corrugatormeasures (interSCZYG, interSCCORR), during three experimental condi-tions (baseline, positive, negative).

2.6.4. Temporal correspondence of EMG signals with the emotional timelineof the movies

To assess temporal correspondence of EMG dynamics with themovie-driven emotional dynamics, we computed similarity of in-dividual EMG responses with dynamic ratings of valence. EMG responsetime-courses as well as valence time-courses were z-normalized anddivided into short (t= 30 s), partially overlapping (Δt = 15 s) timewindows. For each participant j, within each temporal window t, wecomputed cross-correlation between xjt , the EMG response time-courseof participant j, and vt the average valence time-course. The resultingcross-correlation functions were averaged across windows. Individualzero-lag correlations (EMG-valence correlations) were then extractedand r-to-z transformed. Using repeated measures ANOVAs, we com-pared the temporal correspondence of corrugator and zygomatic re-sponses with the positive and the negative emotional timelines.

In addition to the group analysis, described above, we examined thestatistical likelihood of individual EMG-valence correlations. To controlfor various factors (i.e. autocorrelations) which can significantly inflatecorrelation values, we conducted a non-parametric bootstrapping pro-cedure. Specifically, we quantified control chance correlations for eachindividual subject by computing the above described EMG-valencecorrelations using emotional time-lines, in which temporal windowswere shuffled in time. Experimental correlations signify correspondencewith the presented emotional timeline. In contrast, control chancecorrelations signify spurious correlations with random emotional time-lines with similar temporal characteristics. This procedure was per-formed for 1000 iterations, allowing to assess the statistical likelihoodof obtaining experimental EMG-valence fluctuations by chance.

3. Results

3.1. Manipulation check: positive and negative movies elicited distinctiveemotional and facial profiles

To validate that the presented emotional movies elicited distinctiveemotional profiles, we conducted a repeated measures ANOVA on thereported scores of positive and negative valence. As expected, we foundsignificant valence (positive, negative) by condition (positive movie,negative movie) interaction (F(1,47)= 575.8, p < 0.001). High levelsof positive feelings (M=6.7 ± 1.55) and low levels of negative feel-ings (M=0.5 ± 1.1) was reported for the positive movie. High levelsof negative feelings (M=6.64 ± 1.59) and low levels of positivefeelings (M=0.75 ± 1.2) was reported for the negative movie.Positive movie elicited joy (M=6.42 ± 1.43) and warmth(M=6.13 ± 1.92), while negative movie elicited fear(M=6.33 ± 1.93) and stress (M=6.21 ± 2.03).

To validate that the presented emotional movies differentially af-fected the zygomatic and the corrugator activations, we conducted arepeated measures ANOVA on EMG reactivity scores and found an ex-pected EMG site (zygomatic, corrugator) by condition (positive, nega-tive) interaction (F(1,35)= 30.49, p < 0.001). EMGZYG was po-tentiated by the positive (M=0.32 ± 0.73) movie and attenuated bythe negative movie (M=−0.31 ± 0.52, p < 0.001), while EMGCORR

Y. Golland et al. Biological Psychology 139 (2018) 47–58

51

was higher during the negative (M=0.33 ± 0.47) as compared to thepositive movie (M=0.02 ± 0.55, p < 0.001).

3.2. Temporal characteristics of the zygomatic and corrugator dynamics

Central to the present study, we assessed whether emotional movieselicited consistent EMG response patterns. Fig. 1A presents averagezygomatic and corrugator response time-courses (upper panel) anddynamic ratings of valence (lower panel) for the emotional moviescondition. As evident in the figure, zygomatic and corrugator dynamicsshowed profoundly different temporal response characteristics. Thezygomatic activity evoked by the positive movie exhibited clear phasicmodulations along the timeline of the movie. In contrast, the corrugatoractivity during the negative movie was predominantly tonic, showingslow increases as the movie progressed. To further explore the apparentdiscrepancy in the temporal characteristics of zygomatic and corrugatoractivity we computed autocorrelations for these signals (Fig. 1B), andassessed the autocorrelation width by calculating the temporal lag atwhich signals’ autocorrelation dropped to half (Fig. 1C). A larger widthsignifies slower declines of autocorrelation as a function of lag, i.e.higher level of slow tonic changes. Repeated measures ANOVA on thelag statistic with site (corrugator, zygomatic) and condition (baseline,positive, negative) factors, showed significant main effects of site (F(1,66)= 27.3, p < 0.001), experimental condition ((F(2,66)= 18.66,p < 0.001) as well as site by condition interaction (F(2,66)= 11.3,p < 0.001). As can be seen in Fig. 1C, corrugator as compared to zy-gomatic responses contained more slow components across all condi-tions (all ps< 0.02). In addition, autocorrelation width of corrugatorsignals was significantly larger during the negative compared to thepositive movie (t(1,41)= 4.83, p < 0.001), rising from 10.5 ± 9.4 sto 23.7 ± 20.8 s, suggesting that tonic increases in corrugator activityare linked with increases in negative affect. No such pattern was evi-dent for the zygomatic autocorrelation width which wasn’t differentbetween the positive (6.5 ± 2.7) and the negative (5.7 ± 4.6) movies(p > 0.5).

3.3. Across-subjects reliability of EMG dynamics

To assess whether emotional movies elicited reliable EMG dy-namics, which were time-locked across participants, we employed theinter-subject correlation (interSC) analysis. The interSC approach as-sesses the extent to which the same response pattern is reliably ob-served across all participants, by quantifying similarity between in-dividual responses and the average response pattern (Methods). Fig. 2Apresents average cross-correlation functions, as well as zero lag corre-lations (interSCs) for the zygomatic (left panel) and the corrugator(right panel) responses. As can be seen in the figure, the zygomaticcross-correlation during the positive, but not during the negative movieor baseline, exhibited the typical interdependency form, peaking

around lag zero and subsiding as the temporal lags increased. As for thecorrugator, cross-correlation was highest during the negative movieand lowest during baseline. Descriptive statistics for the interSC scoresare presented in Table1.

Repeated measures ANOVAs on the interSC scores, revealed a sig-nificant effect of condition both for the zygomatic (F(2,66)= 84.8,p < 0.001) and for the corrugator (F(2,66)= 18.8, p < 0.001) in-dexes. Post-hoc analyses showed clear affective patterns in the zygo-matic and corrugator dynamics. The reliability of zygomatic responses(interSCZYG) was significantly larger during the positive movie com-pared to both negative movie and baseline (all ps< 0.001), which werenot significantly different. The reliability of corrugator responses(interSCCORR) was highest during the negative movie and lowest duringbaseline with significant differences among all three conditions (allps< 0.001).

We next asked whether the above affective profiles persist for phasicEMG dynamics, after a removal of slow components from EMG signals.We repeated the interSC analysis for the de-trended EMG signals(Supplementary Fig. 2). InterSCs for the de-trended signals are pre-sented in Fig. 2A (dashed bars). As can be seen in the figure, removal ofslow components did not affect neither the degree (all ps> 0.25) northe affect-specificity of zygomatic dynamics, which still yielded higherinterSC during the positive, as compared to the negative movie (t(35)= 10, p < 0.001). In contrast, it significantly reduced the interSCsof the corrugator signals across conditions (F(1, 38)= 29.3,p < 0.001). Both positive and negative movies elicited significantlyhigher corrugator interSCs as compared to non-emotional baseline(ps< 0.001). However, de-trending disrupted the corrugator’s specifi-city for negative affect, as interSCCORR during the negative movie(r= 0.22 ± 0.16) was now lower than the interSCCORR during thepositive movie, at marginally significant level (t(41)= 1.9, p=0.061).

Taken together, the above analyses demonstrate that transientfluctuations in zygomatic EMG were reliably controlled by the positivemovie. Zygomatic reliability (i.e. interSC) was specific to the positivecondition and was driven by phasic response modulations. The relia-bility of corrugator EMG fluctuations was higher during the negativemovie but was significant during the positive movie as well. An en-hancement of corrugator reliability by negative emotions was ex-clusively driven by tonic modulations during the negative movie. Aremoval of slow changes from the corrugator responses resulted insignificant drop of interSCs in general and lower interSC scores duringthe negative as compared to the positive condition.

3.4. Controlling for facial mimicry

The emotional movies used here involved human protagonists,which are known to elicit strong emotional states in observers(Codispoti et al., 2008; Hess et al., 1995; Kreibig et al., 2007; Wilhelmet al., 2017). It could be suggested that protagonists’ facial expressions

Fig. 2. InterSC analysis for the zygo-matic (left panel) and corrugator (rightpanel) measures. A. Average cross-cor-relation functions for each experi-mental condition (baseline (grey), po-sitive (green), negative (red)) B. Zerolag interSCs, for the original (full bars)and de-trended (dashed bars) signals.Error bars signify 2*SE (For interpreta-tion of the references to colour in thisfigure legend, the reader is referred tothe web version of this article).

Y. Golland et al. Biological Psychology 139 (2018) 47–58

52

elicited facial mimicry (Dimberg, Thunberg, & Elmehed, 2000; Hess &Bourgeois, 2010; Sato et al., 2013), hence biasing the results of EMGinterSC analysis. To control for the effects of facial mimicry we codedthe movie times in which positive and negative facial expressions wereclearly evident in the positive and the negative movies, respectively.We then conducted the interSC analysis in narrow temporal windows(window size= 10 s, overlap=5 s) and computed average interSCscores, limiting the computation to the temporal windows which didn’tinclude vivid emotional expressions (Supplementary Fig. 4 presents thefull details of facial exclusion analysis). The analysis was done for thede-trended EMG signals, focusing on the reliability of phasic responses.The results demonstrated that the affective profiles of EMG dynamicswere not exclusively driven by facial mimicry, since a removal of facialexpressions didn’t modify the interSC patterns observed for the fullmovies’ timelines. Specifically, the zygomatic interSC was significantlyhigher during the positive (M=0.47 ± 0.28) as compared to the ne-gative movie (M=0.12 ± 0.1, t(35)= 6.8, p < 0.001). The corru-gator interSC was still lower during the negative movie(M=0.17 ± 0.13) as compared to the positive movie(M=0.36 ± 0.31, t(41)= 3.33, p < 0.002).

3.5. Association with subjective ratings

We assessed whether individual differences in average interSCscores were associated with subjective emotional ratings, reported fol-lowing each movie (Table 1, lower part). We found that individualzygomatic interSC scores significantly tracked subjective intensity ofpositive (but not negative) emotions. Correlations with subjective rat-ings were evident both for the original (p < 0.008) and for the de-trended (p < 0.003) zygomatic measures. In other words, participantswho reported higher levels of positive feelings following the positivemovie, also evidenced highly reliable zygomatic dynamics. Interest-ingly, higher positive feelings during the positive movie were correlatedwith lower corrugator interSC scores. This association was found for theoriginal (p < 0.02) but not for the de-trended (p=0.36) corrugatormeasures. No consistent associations were found between the corru-gator reliability indexes and negative feelings.

3.6. Temporal correspondence between EMG dynamics and emotionaltimelines

We next examined whether EMG dynamics were temporally co-ordinated with the movie-driven emotional fluctuations. For that aimwe quantified correlations between individual EMG time-courses andthe within-movie fluctuations of valence (Methods). Fig. 3A presentstemporal response patterns of z-normalized EMG measures and con-tinuous ratings of valence during the positive and the negative movies.Fig. 3B presents average cross-correlation functions and zero lag cor-relations for the zygomatic-valence correspondence (left panel) and forthe corrugator-valence correspondence (right panel). Descriptive sta-tistics of zero lag EMG-valence correlations are shown in Table 2.

Zygomatic correspondence with valence was different for the posi-tive and the negative movies (t(35)= 4.2, p < 0.001), showing sig-nificant positive correlations with positive emotional fluctuations (t(35)= 5.8, p < 0.001) but not with the negative ones (t(35)=−0.7,p > 0.25). Corrugator correspondence showed a complementary pat-tern (t(41)=−3.47, p < 0.001), demonstrating high positive correla-tion with fluctuations of negative (t(41)= 7.02, p < 0.001) but notpositive affect (t(41)= 1.1, p > 0.25).

We next examined whether a removal of tonic components fromEMG dynamics modified its alignment with emotional fluctuations(Supplementary Fig. 2 presents the de-trended signals). De-trending hadprofoundly different impact on the zygomatic and corrugator corre-spondence (Fig. 3B, dashed bars, Table 2). The zygomatic correspon-dence with positive emotional fluctuations wasn’t modified by de-trending (t(35)=−1.62, p=0.113), and was still higher than its cor-respondence with negative affect (t(35)= 5.9, p < 0.001). In contrast,de-trending significantly modified corrugator's correspondence withboth emotional timelines. First, it significantly reduced corrugator'scorrespondence with negative emotional fluctuations (t(41)= 5.1,p < 0.001). Second, it revealed a negative correspondence with posi-tive emotional fluctuations, which was significantly different from zero(t(41)=−4.53, p < 0.001).

Taken together, these results suggest that continuous changes inemotional experience elicited corresponding fluctuations in EMG dy-namics. Zygomatic activity was time-locked to the positive emotionaltimeline and this correspondence was driven by phasic fluctuations. Incontrast, corrugator’s coupling with negative emotional timeline un-folded at a slower temporal scale, such that gradual changes in nega-tivity elicited tonic changes in corrugator activity. Furthermore, tran-sient corrugator fluctuations were inversely correlated with changes inpositive affect, during the unfolding of the positive movie.

3.7. Individual differences in EMG-valence correspondence

The above analysis suggested that the coordination of phasic cor-rugator activity with transient changes in negative affect is diminished.However, such group-level lack of correspondence might be driven by asubset of participants, showing negligible or even negative correlationswith fluctuations of negative affect. To examine this possibility, weassessed the distribution of individual EMG-valence correlations, andcomputed their statistical significance, using non-parametric boot-strapping procedure (Methods). Fig. 4A presents the individual EMG-valence correlations, rank ordered within each experimental conditionand marked for statistical significance. As can be seen in the figure, thisanalysis revealed substantial sub-groups in EMG response patterns.Descriptive statistics of EMG-valence subgroups are presented inTable 3.

Individual zygomatic responses evidenced highly consistent tem-poral correspondence with positive affect fluctuations, with 86% ofparticipants exhibiting significant EMG-valence correlations during thepositive movie. Interestingly, a sub-group of participants (36%) also

Table 1Upper part presents descriptive statistics for the interSC scores in each experimental condition.

Original signals De-trended signals

baseline positive negative baseline positive negative

Descriptive statistics for interSC scoresInterSCZYG 0.03±0.13 0.5±0.20 0.12± 0.11 0.02±0.11 0.52±0.20 0.13± 0.10InterSCCORR 0.11±0.2 0.4± 0.38 0.61±0.36 0.01±0.14 0.31±0.28 0.22± 0.16

Correlation with subjective emotional ratingsInterSCZYG 0.42 (0.008) 0.26 (0.11) 0.46 (0.003) 0.06 (0.7)InterSCCORR −0.34 (0.02) −0.13 (0.42) −0.14 (0.36) 0.02 (0.9)

Conditions which elicited the highest interSCs within each EMG channel are highlighted with bold. Lower part presents correlations of interSC scores with subjectiveemotional ratings, reported by participants.

Y. Golland et al. Biological Psychology 139 (2018) 47–58

53

showed positive correspondence with negative valence fluctuationsduring the negative movie. In sum, individual level analysis of zygo-matic measures confirmed the group-level results showing consistentzygomatic correspondence with positive valence for the majority ofparticipants and extended them to show significant zygomatic corre-lations with negative valence in a subset of sample participants.

Individual corrugator correlations exhibited substantial individualvariability, with 40% of participants showing significant positive cor-relations with phasic fluctuations of negative affect and the remaining60% showing either negligible or negative (16%) correlations. Fig. 4Bshows mean corrugator response time-courses computed for these twosubgroups of participants. As for the corrugator’s correspondence withpositive affect, it was more uniform across participants with the largerproportion (60%) showing significant negative correlations with phasic

fluctuations of positive affect and only 9% showing positive corrugator-valence correlations. In sum, individual level analysis of corrugatorcorrespondence with emotional fluctuations confirmed the consistentinhibitory effect of positive affect fluctuations on corrugator responses,which was evident in the larger portion of our sample. It also shedadditional light on the apparent lack of corrugator correspondence withnegative valence, showing that 40% of our sample did show such cor-respondence while the remaining part didn’t.

4. Discussion

The main goal of the present work was to investigate facial EMGdynamics during continuously unfolding emotional experiences. Ourresults indicated that zygomatic and corrugator dynamics exhibitedclearly distinctive affective profiles as well as different temporal char-acteristics.

4.1. Affect dynamics in the zygomatic response system

The dynamics of EMG activity measured over zygomaticus majorfacial area showed robust links with positive affect. First, the positivemovie elicited reliable zygomatic fluctuations, which were time-lockedacross participants. Individual differences in such response reliabilitywere associated with the intensity of reported positive experience.Second, transient zygomatic fluctuations were significantly coupledwith changes in positive affect as the movie unfolded, and such cou-pling was significant in the majority of sample participants. Taken

Fig. 3. A. EMG and valence dynamics. Figure presents z-normalized time-courses of EMG measures and valence ratings during the positive and the negative movies.B. Average cross-correlations between zygomatic EMG (left panel), corrugator EMG (right panel) and valence dynamics for the positive (green) and negative (red)movies (A.) as well as zero lag correlations computed for original (full bars) and de-trended (dashed bars) signals (B.). Error bars signify 2*SE (For interpretation ofthe references to colour in this figure legend, the reader is referred to the web version of this article).

Table 2Temporal correspondence between facial and emotional dynamics.

original signals de-trended signals

positive negative positive negative

EMGZYG-valence

0.22 ± 0.23 −0.03 ± 0.26 0.28 ± 0.14 0.08 ± 0.14

EMGCOR-valence

0.08 ± 0.3 0.3 ± 0.27 −0.13 ± 0.19 0.06 ± 0.18

Descriptive statistics for the correspondence of facial signals (EMGZYG;EMGCORR) with dynamic ratings of valence.

Y. Golland et al. Biological Psychology 139 (2018) 47–58

54

together, these results demonstrate that momentary changes in positiveexperience elicited corresponding phasic fluctuations in zygomatic ac-tivity and these fluctuations were similar across participants. Our re-sults extend the previous findings on affective zygomatic profile (Larsenet al., 2003, Bradley & Lang, 2000; Lang et al., 1993) to dynamicmeasures, showing that transient zygomatic fluctuations systematicallyreflect the unfolding of positive emotions over time. Notably, whilenegative emotional experience didn’t elicit consistent zygomatic re-sponse patterns at the group-level, thirty six percent of participantsshowed significant zygomatic alignment with negative affect fluctua-tions. These findings come in line with previous research showing zy-gomatic activation during extremely negative emotional stimuli (Langet al., 1993; Larsen et al., 2003). They further highlight the importanceof individual differences in facial expressions.

The robust reliability and affect specificity of transient zygomaticfluctuations suggest that this response system can provide a viablephysiological measure of positive affective dynamics during complexemotional experiences. Given the important role played both by

positive affect and by emotional dynamics in psychological well-beingand psychopathology (Fredrickson & Losada, 2005; Houben, Van DenNoortgate, & Kuppens, 2015, 2015; Kuppens, Stouten, & Mesquita,2009; Wichers, Wigman, & Myin-Germeys, 2015), dynamic measures ofzygomatic EMG hold a potential to significantly advance the under-standing of temporal regularities and individual differences in positiveaffect dynamics.

4.2. Affect dynamics in the corrugator response system

The dynamics of EMG activity measured over corrugator superculliishowed a broader affective profile. Corrugator activity exhibitedhighest reliability during the negative movie. Furthermore, it wastemporally aligned with slow changes in emotional negativity, hencedemonstrating the links of corrugator dynamics with negative affect.

In addition, corrugator activity evidenced a substantial reliabilityduring the positive movie as compared to the resting baseline. Previousstudies have suggested that corrugator responses are systematicallyattenuated by positive emotional events (Lang et al., 1993; Lapate et al.,2014; Larsen et al., 2003). In particular, Larsen and colleagues havedemonstrated a linear effect of positive valence on event-related cor-rugator de-activations. Our results fully replicate this previous researchand extend it to temporal emotional processes. Specifically, we foundthat phasic fluctuations of corrugator activity during the positive moviewere both reliable and inversely correlated with the fluctuations ofpositive affect. In other words, the ups and downs of positive affectexerted dynamic cycles of inhibition and recovery upon corrugatorresponses, leading to a significant time-lock across participants. Suchinverse correlation with positive affect was evident in sixty percent ofour sample participants.

Taken together, these results suggest that in contrast to zygomaticactivity, which was narrowly tuned to positive affect dynamics, theactivity of corrugator facial muscles was driven by tonic increases inemotional negativity as well as by inhibitory phasic effects of positive

Fig. 4. A. Individual EMG-valence cor-relations scores, rank ordered withinexperimental conditions (green-positivemovie, red-negative movie), of the zy-gomatic (left panel) and corrugator(right panel) responses. Full circles re-present significant and empty circlesrepresent non-significant individualcorrelations in the predicted direction.B. Corrugator response patterns duringthe negative movie in two subsets ofparticipants. Left panel shows averageresponse pattern in participants whoshowed significant correspondencewith negative valence (40%, N=17).Right panel represents the rest of thesample (60%, N=25). (For inter-pretation of the references to colour inthis figure legend, the reader is referredto the web version of this article).

Table 3Sub-groups showing significant positive and negative correspondence withdynamic ratings of valence.

Positive Movie Negative Movie

positivecorrelation

negativecorrelation

positivecorrelation

negativecorrelation

EMGZYG-valence 31 (86.1%)0.3±0.1

1 (2.8%)–

13 (36.1%)0.23± 0.06

3 (8.3%)–

EMGCOR-valence 4 (9.5%)–

25 (59.5%)−0.25± 0.09

17 (40.5%)0.23± 0.09

7 (16.6%)–

Table presents the number and the proportion of sample participants showingsignificant correlations, as well as means and standard deviations of the sub-groups larger than 20% of the sample.

Y. Golland et al. Biological Psychology 139 (2018) 47–58

55

affect dynamics.

4.3. Temporal characteristics of the corrugator and the zygomatic responses

Emotional processes can unfold over different timescales, includingtransient responses to specific events alongside more gradual dynamicsreflecting tonic affective processes. Systematic investigation of thetemporal response characteristics, undertaken in this work, revealedprofound differences in the temporal resolution of the zygomatic andthe corrugator affective dynamics.

Zygomatic responses elicited by the positive movie exhibited clearphasic modulations which evolved over seconds. These transient mod-ulations were coupled with momentary changes in positive affect eli-cited by the unfolding of the movie. These findings suggest that zygo-maticus major is a phasic response system reflecting transient changes inindividuals’ positive feelings and expressions. The phasic capabilities ofthe zygomatic system resonate with the profound communicativefunction of positive facial expressions, which signal social intents toothers (Fischer & Manstead, 2008; Fridlund, 1991; Frijda & Mesquita,1994; Hess et al., 1995). Since abrupt perceptual changes are betterdetected than gradual ones (David, Laloyaux, Devue, & Cleeremans,2006; Simons & Rensink, 2005), the prominence of phasic responses inzygomatic activity, might be beneficial for a swift detection of positivefacial responses by social environment. Indeed, happy facial expres-sions are faster detected and more accurately recognized in comparisonto other emotions (Ambadar, Schooler, & Cohn, 2005; Calvo &Lundqvist, 2008). Moreover, such superiority is driven by the mouthregion, due to its saliency and diagnostic value (Calvo & Nummenmaa,2008; Calvo, Nummenmaa, & Avero, 2010; Calvo, Fernández-Martín, &Nummenmaa, 2014).

In comparison with the zygomatic responses, corrugator dynamicsdemonstrated clear susceptibility to tonic increases across all experi-mental conditions, including non-emotional baseline. Previous researchhas suggested that beyond its affective properties, corrugator responsesare influenced by such factors as cognitive effort and goal conducive-ness (Aue & Scherer, 2008; Kappas, 2003), which might underlie theoverall propensity of corrugator responses to tonic changes observed inthe current study. Our results also showed tonic effects which werespecific to negative affect. Thus, during the negative movie, corruga-tor’s tonic modulations were particularly pronounced and coordinatedwith the movie-driven changes of negativity. These results highlight theneed to carefully differentiate between affect-driven tonic corrugatorresponses and drifts arising from non-affective sources in future re-search of facial affect dynamics. In addition, further research is neededto understand whether the observed differences in corrugator and zy-gomatic susceptibility to tonic modulations is grounded in the phy-siology of this responses system or is carved by higher-order factors,such as display rules or social functions.

4.4. Phasic corrugator dynamics during negative emotional experience

The analysis of phasic corrugator dynamics evidenced a complexpicture. First, while the reliability of phasic corrugator responses duringthe negative movie was significant, it was somewhat lower as comparedto the positive movie. Second, corrugator responses showed diminishedcorrespondence with transient changes in negative affect. Taken to-gether, these findings could suggest that phasic corrugator responsesexhibited impoverished representation of transient changes in negativeemotional experience. As such, they come at odds with previous re-search showing linear effects of valence on corrugator activation(Larsen et al., 2003). However, assessment of individual differences incorrugator correspondence with valence dynamics shed additional lighton these findings. Specifically, we found that forty percent of partici-pants exhibited significant positive correlations with transient fluctua-tions of negative affect while the remaining of sample didn’t. Thus,rather than showing a generally reduced correspondence with negative

emotional fluctuations, corrugator activity evidenced a substantialinter-subject variability in such correspondence.

Given that the negative movie used in the current study elicitedstrong feelings of fear, several plausible explanations of the observedvariability in corrugator responses can be suggested. First, people mayvary in the degree to which their corrugator muscles are involved in amanifestation of transient fear dynamics. Notably, the frowning ex-pression, subserved by the corrugator activation, is only one of the fa-cial features representing the prototypical configuration of fear (Friesen& Ekman, 1978). Accordingly, different individuals may enact fearfulexpressions via different facial muscles. In support, studies in-vestigating spontaneous facial responses have reported significantvariability among individuals in the facial expressions of negativeemotions (Duran et al., 2017; Fernandez-Dols et al., 1997; Namba,Kabir, Miyatani, & Nakao, 2017; Reisenzein et al., 2013). Moreover, astudy focusing on the dynamics of prototypical facial expressions didn’tfind frowning among facial configurations characterizing the apex offear (Krumhuber & Scherer, 2011). Second, highly negative scenes inthe movie could have elicited other subjective feelings, such as con-tempt, anger or surprise, hence increasing facial variability. Finally,previous research employing the same negative movie (Golland et al.,2014) has shown that its high negative scenes were marked by auto-nomic arousal. While negative arousal could have significantly shapedthe subjective ratings, corrugator responses are weakly associated withthe arousal dimension (Lang et al., 1993; Sato et al., 2013; Tan et al.,2012). Taken together, the above scenarios highlight the complexity offactors potentially affecting spontaneous facial responses during emo-tional experiences. In particular, they suggest that future studies offacial affective dynamics can benefit from combining both dimensional(such as valence) and categorical (such as fear, disgust) measures ofemotions.

4.5. Limitations

The current research taps into a largely unchartered territory of thefacial production dynamics (Scherer, Mortillaro, & Mehu, 2013). Whileit revealed a series of novel results, their generalizability should betaken with caution and awaits future replications, accounting for themethodological limitations of the present study. First, we employedphysiological baseline measures to approximate an emotionally neutralcondition. However, several non-emotional factors could potentiallyimpact the observed differences, including timing of measurement anddemand characteristics. Neutral movies, presented together with emo-tional ones, should be used as a control condition in future studies.Second, due to methodological constraints, the current study was basedon female participants. Since gender differences might play a significantrole in emotional responding (Codispoti et al., 2008; Dimberg &Lundquist, 1990; Hall, Carter, & Horgan, 2000), it is necessary to re-plicate current results in mixed gender samples. Third, an employmentof a broader range of emotional stimuli can significantly improve thegeneralizability of current findings. Here we found that fear inducingfilm elicited substantial variability in corrugator response patterns.Whether such variability exists for other negative emotions, such asanger and sadness, remains to be uncovered. In addition, future re-search should control for the effects of facial mimicry on EMG dy-namics. The emotional states of other people elicit powerful emotionalstates in observers, and are thus frequently employed as emotionalelicitors in facial research, including the current one (Codispoti et al.,2008; Hess et al., 1995; Kreibig et al., 2007; Wilhelm et al., 2017).However, such strategy hinders the interpretation of results, whichreflect both emotional and social processes. While it might be chal-lenging to elicit powerful phasic modulations of affect without em-ploying human protagonists, such an effort should be made, to differ-entiate between these two processes. Finally, in the current research weemployed a between-subjects approach, asking whether facial dynamicsare consistently modulated by the emotional timelines of the movies,

Y. Golland et al. Biological Psychology 139 (2018) 47–58

56

above and beyond individual differences. However, given the varia-bility in emotional responding (Stemmler & Wacker, 2010) as well asindividual differences in corrugator dynamics observed in the currentstudy, within-subject measures of facial and subjective emotionalfluctuations (Mauss et al., 2005) may be essential to fully uncover thedegree to which facial dynamics reflect subjective feelings.

4.6. Conclusions

Using continuous measures of facial EMG we demonstrated thatfluctuations of facial activity, elicited by naturally unfolding emotionalexperiences, exhibit consistent, affect specific temporal response pat-terns. Across analyses, transient zygomatic dynamics were narrowlytuned to positive emotional experiences. A subset of participants alsoshowed consistent zygomatic correlations with fluctuations of negativeaffect. The corrugator responses showed a clear susceptibility to tonicchanges. Overall, corrugator dynamics showed temporal alignmentwith slow changes in emotional negativity and inverse correlation tophasic changes in positive affect. Corrugator’s correspondence withtransient fluctuations of negative affect showed substantial differencesbetween participants. Taken together, the findings and the methodo-logical approach employed in the current research open novel possi-bilities for studying the involvement of facial response system in dy-namic processes of affect generation, expression and communication.

Funding

This work was supported by Israel Science Foundation Grant 698/15 to Yulia Golland.

Acknowledgement

We thank Dana Mevorah and Roni Nadav for a tremendous help inrunning this study

Appendix A. Supplementary data

Supplementary material related to this article can be found, in theonline version, at doi:https://doi.org/10.1016/j.biopsycho.2018.10.003.

References

Ambadar, Z., Schooler, J. W., & Cohn, J. F. (2005). Deciphering the enigmatic face: Theimportance of facial dynamics in interpreting subtle facial expressions. PsychologicalScience, 16(5), 403–410. https://doi.org/10.1111/j.0956-7976.2005.01548.x.

Aue, T., & Scherer, K. R. (2008). Appraisal-driven somatovisceral response patterning:Effects of intrinsic pleasantness and goal conduciveness. Biological Psychology, 79(2),158–164.

Barbosa, P. R. B., Barbosa-Filho, J., de Sa, C. A. M., Barbosa, E. C., & Nadal, J. (2003).Reduction of electromyographic noise in the signal-averaged electrocardiogram byspectral decomposition. IEEE Transactions on Biomedical Engineering, 50(1), 114–117.

Barrett, L. F., Mesquita, B., Ochsner, K. N., & Gross, J. J. (2007). The experience ofemotion. Annual Review of Psychology, 58(1), 373–403. https://doi.org/10.1146/annurev.psych.58.110405.085709.

Beniczky, S., Aurlien, H., Brøgger, J. C., Fuglsang-Frederiksen, A., Martins-da-Silva, A.,Trinka, E., ... Herman, S. T. (2013). Standardized computer-based organized re-porting of EEG: SCORE. Epilepsia, 54(6), 1112–1124.

Boiger, M., & Mesquita, B. (2012). The construction of emotion in interactions, re-lationships, and cultures. Emotion Review, 4(3), 221–229.

Boker, S. M., Rotondo, J. L., Xu, M., & King, K. (2002). Windowed cross-correlation andpeak picking for the analysis of variability in the association between behavioral timeseries. Psychological Methods, 7(3), 338.

Bradley, M. M., & Lang, P. J. (2000). Affective reactions to acoustic stimuli.Psychophysiology, 37(2), 204–215.

Butler (2011). Temporal interpersonal emotion systems: The “TIES” that form relation-ships. Personality and Social Psychology Review, 15(4), 367–393. https://doi.org/10.1177/1088868311411164.

Cacioppo, J. T., Martzke, J. S., Petty, R. E., & Tassinary, L. G. (1988). Specific forms offacial EMG response index emotions during an interview: From Darwin to the con-tinuous flow hypothesis of affect-laden information processing. Journal of Personalityand Social Psychology, 54(4), 592.

Cacioppo, J. T., Petty, R. E., Losch, M. E., & Kim, H. S. (1986). Electromyographic activityover facial muscle regions can differentiate the valence and intensity of affectivereactions. Journal of Personality and Social Psychology, 50(2), 260.

Calvo, M. G., & Lundqvist, D. (2008). Facial expressions of emotion (KDEF): Identificationunder different display-duration conditions. Behavior Research Methods, 40(1),109–115.

Calvo, M. G., & Nummenmaa, L. (2008). Detection of emotional faces: Salient physicalfeatures guide effective visual search. Journal of Experimental Psychology. General,137(3), 471–494. https://doi.org/10.1037/a0012771.

Calvo, M. G., Fernández-Martín, A., & Nummenmaa, L. (2014). Facial expression re-cognition in peripheral versus central vision: Role of the eyes and the mouth.Psychological Research, 78(2), 180–195. https://doi.org/10.1007/s00426-013-0492-x.

Calvo, M. G., Nummenmaa, L., & Avero, P. (2010). Recognition advantage of happy facesin extrafoveal vision: Featural and affective processing. Visual Cognition, 18(9),1274–1297. https://doi.org/10.1080/13506285.2010.481867.

Chołoniewski, J., Chmiel, A., Sienkiewicz, J., Hołyst, J., Küster, D., & Kappas, A. (2016).Temporal Taylor’s scaling of facial electromyography and electrodermal activity inthe course of emotional stimulation. Chaos, Solitons & Fractals, 90, 91–100. https://doi.org/10.1016/j.chaos.2016.04.023.

Codispoti, M., Surcinelli, P., & Baldaro, B. (2008). Watching emotional movies: Affectivereactions and gender differences. International Journal of Psychophysiology, 69(2),90–95. https://doi.org/10.1016/j.ijpsycho.2008.03.004.

Cunningham, W. A., Dunfield, K. A., & Stillman, P. E. (2013). Emotional States fromaffective dynamics. Emotion Review, 5(4), 344–355. https://doi.org/10.1177/1754073913489749.

Darwin, C., & Prodger, P. (1998). The expression of the emotions in man and animals. USA:Oxford University Press.

David, E., Laloyaux, C., Devue, C., & Cleeremans, A. (2006). Change blindness to gradualchanges in facial expressions. Psychologica Belgica, 46(4), 253–268.

Dimberg, U. (1986). Facial reactions to fear-relevant and fear-irrelevant stimuli. BiologicalPsychology, 23(2), 153–161. https://doi.org/10.1016/0301-0511(86)90079-7.

Dimberg, U. (1990). Facial electromyography and emotional reactions. Psychophysiology,27(5), 481–494. https://doi.org/10.1111/j.1469-8986.1990.tb01962.x.

Dimberg, U., & Lundquist, L.-O. (1990). Gender differences in facial reactions to facialexpressions. Biological Psychology, 30(2), 151–159.

Dimberg, U., Thunberg, M., & Elmehed, K. (2000). Unconscious facial reactions toemotional facial expressions. Psychological Science, 11(1), 86–89. https://doi.org/10.1111/1467-9280.00221.

Duran, J., Reisenzein, R., & Fernandez-Dols, J.-M. (2017). Coherence between emotionsand facial expressions. The Science of Facial Expression, 1849.

Ekman, P. (1992). Facial expressions of emotion: New findings, new questions.Psychological Science, 3(1), 34–38.

Ekman, P. (1993). Facial expression and emotion. American Psychologist, 48(4), 384.Ekman, P., & Rosenberg, E. L. (1997). What the face reveals: Basic and applied studies of

spontaneous expression using the facial action coding system (FACS). USA: OxfordUniversity Press.

Ekman, P., & Scherer, K. (1984). Expression and the nature of emotion. Approaches toEmotion, 3, 19–344.

Faul, F., Erdfelder, E., Buchner, A., & Lang, A.-G. (2009). Statistical power analyses usingG* power 3.1: Tests for correlation and regression analyses. Behavior ResearchMethods, 41(4), 1149–1160.

Fernandez-Dols, J.-M., Sanchez, F., Carrera, P., & Ruiz-Belda, M.-A. (1997). Are sponta-neous expressions and emotions linked? An experimental test of coherence. Journal ofNonverbal Behavior; New York, 21(3), 163–177.

Fischer, A. H., & Manstead, A. S. (2008). Social functions of emotion. In M. Lewis, J.Haviland-Jones, & L. F. Barrett (Eds.). Handbook of emotions (pp. 456–468). (4th ed.).New York, NY: Guilford.

Fredrickson, B. L., & Losada, M. F. (2005). Positive affect and the complex dynamics ofhuman flourishing. The American Psychologist, 60(7), 678–686. https://doi.org/10.1037/0003-066X.60.7.678.

Fridlund, A. J. (1991). Sociality of solitary smiling: Potentiation by an implicit audience.Journal of Personality and Social Psychology, 60(2), 229.

Fridlund, A. J., & Cacioppo, J. T. (1986). Guidelines for human electromyographic re-search. Psychophysiology, 23(5), 567–589.

Friesen, E., & Ekman, P. (1978). Facial action coding system: A technique for the measure-ment of facial movement. Palo Alto: Consulting Psychologists Press.

Frijda, N. H. (1988). The laws of emotion. American Psychologist, 43(5), 349.Frijda, N. H., & Mesquita, B. (1994). The social roles and functions of emotions. In K.

Shinobu, & H. R. Markus (Eds.). Emotion and culture: Empirical studies of mutual in-fluence (pp. 51–87). Washington, DC, US: American Psychological Association.

Friman, O., Borga, M., Lundberg, P., & Knutsson, H. (2004). Detection and detrending infMRI data analysis. NeuroImage, 22(2), 645–655.

Golland, Y., Arzouan, Y., & Levit-Binnun, N. (2015). The Mere Co-presence:Synchronization of autonomic signals and emotional responses across Co-presentindividuals not engaged in direct interaction. PLOS ONE, 10(5), e0125804. https://doi.org/10.1371/journal.pone.0125804.

Golland, Y., Bentin, S., Gelbard, H., Benjamini, Y., Heller, R., Nir, Y., & Malach, R. (2007).Extrinsic and intrinsic systems in the posterior cortex of the human brain revealedduring natural sensory stimulation. Cerebral Cortex, 17(4), 766–777.

Golland, Y., Keissar, K., & Levit-Binnun, N. (2014). Studying the dynamics of autonomicactivity during emotional experience. Psychophysiology, 51(11), 1101–1111. https://doi.org/10.1111/psyp.12261.

Golland, Y., Levit-Binnun, N., Hendler, T., & Lerner, Y. (2017). Neural dynamics under-lying emotional transmissions between individuals. Social Cognitive and AffectiveNeuroscience. https://doi.org/10.1093/scan/nsx049.

Y. Golland et al. Biological Psychology 139 (2018) 47–58

57

Hall, J. A., Carter, J. D., & Horgan, T. G. (2000). Gender differences in nonverbal com-munication of emotion. Gender and Emotion: Social Psychological Perspectives, 97–117.

Hamaker, E. L., Ceulemans, E., Grasman, R., & Tuerlinckx, F. (2015). Modeling affectdynamics: State of the art and future challenges. Emotion Review, 7(4), 316–322.

Hasson, U., Malach, R., & Heeger, D. J. (2010). Reliability of cortical activity duringnatural stimulation. Trends in Cognitive Sciences, 14(1), 40–48.

Hasson, U., Nir, Y., Levy, I., Fuhrmann, G., & Malach, R. (2004). Intersubject synchro-nization of cortical activity during natural vision. Science, 303(5664), 1634–1640.https://doi.org/10.1126/science.1089506.

Heerey, E. A., & Crossley, H. M. (2013). Predictive and reactive mechanisms in smilereciprocity. Psychological Science, 24(8), 1446–1455. https://doi.org/10.1177/0956797612472203.

Heller, A. S., Lapate, R. C., Mayer, K. E., & Davidson, R. J. (2014). The face of negativeaffect: trial-by-trial corrugator responses to negative pictures are positively asso-ciated with amygdala and negatively associated with ventromedial prefrontal cortexactivity. Journal of Cognitive Neuroscience, 26(9), 2102–2110. https://doi.org/10.1162/jocn_a_00622.

Hess, U., & Bourgeois, P. (2010). You smile–I smile: Emotion expression in social inter-action. Biological Psychology, 84(3), 514–520. https://doi.org/10.1016/j.biopsycho.2009.11.001.

Hess, U., Banse, R., & Kappas, A. (1995). The intensity of facial expression is determinedby underlying affective state and social situation. Journal of Personality and SocialPsychology, 69(2), 280–288. https://doi.org/10.1037/0022-3514.69.2.280.

Hollenstein, T., & Lanteigne, D. (2014). Models and methods of emotional concordance.Biological Psychology, 98, 1–5. https://doi.org/10.1016/j.biopsycho.2013.12.012.

Houben, M., Van Den Noortgate, W., & Kuppens, P. (2015). The relation between short-term emotion dynamics and psychological well-being: A meta-analysis. PsychologicalBulletin, 141(4), 901–930. https://doi.org/10.1037/a0038822.

Jääskeläinen, I. P., Koskentalo, K., Balk, M. H., Autti, T., Kauramäki, J., Pomren, C., ...Sams, M. (2008). Inter-subject synchronization of prefrontal cortex hemodynamicactivity during natural viewing. Open Neuroimaging Journal, 2(1), 14–19.

Jääskeläinen, I. P., Pajula, J., Tohka, J., Lee, H.-J., Kuo, W.-J., & Lin, F.-H. (2016). Brainhemodynamic activity during viewing and re-viewing of comedy movies explained byexperienced humor. Scientific Reports, 6. Retrieved from https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4914983/.

Kappas, A. (2003). What facial activity can and cannot Tell us about emotions. In M.Katsikitis (Ed.). The human face (pp. 215–234). Boston, MA: Springer US. https://doi.org/10.1007/978-1-4615-1063-5_11.

Korb, S., Grandjean, D., & Scherer, K. R. (2010). Timing and voluntary suppression offacial mimicry to smiling faces in a Go/NoGo task—an EMG study. BiologicalPsychology, 85(2), 347–349.

Kreibig, S. D. (2010). Autonomic nervous system activity in emotion: A review. BiologicalPsychology, 84(3), 394–421. https://doi.org/10.1016/j.biopsycho.2010.03.010.

Kreibig, S. D., Wilhelm, F. H., Roth, W. T., & Gross, J. J. (2007). Cardiovascular, elec-trodermal, and respiratory response patterns to fear- and sadness-inducing films.Psychophysiology, 44(5), 787–806. https://doi.org/10.1111/j.1469-8986.2007.00550.x.

Krumhuber, E. G., & Scherer, K. R. (2011). Affect bursts: Dynamic patterns of facial ex-pression. Emotion, 11(4).

Krumhuber, E. G., Kappas, A., & Manstead, A. S. (2013). Effects of dynamic aspects offacial expressions: A review. Emotion Review, 5(1), 41–46.

Kuppens, P. (2015). It’s about time: A special section on affect dynamics. EmotionReview1754073915590947.

Kuppens, P., Stouten, J., & Mesquita, B. (2009). Individual differences in emotion com-ponents and dynamics: Introduction to the Special issue. Cognition and Emotion,23(7), 1249–1258. https://doi.org/10.1080/02699930902985605.

Lang, P. J., Bradley, M. M., & Cuthbert, B. N. (1999). International affective picture system(IAPS): Instruction manual and affective ratings. The Center for Research inPsychophysiology, University of Florida.

Lang, P. J., Greenwald, M. K., Bradley, M. M., & Hamm, A. O. (1993). Looking at pictures:Affective, facial, visceral, and behavioral reactions. Psychophysiology, 30(3), 261–273.

Lapate, R. C., van Reekum, C. M., Schaefer, S. M., Greischar, L. L., Norris, C. J.,Bachhuber, D. R. W., ... Davidson, R. J. (2014). Prolonged marital stress is associatedwith short-lived responses to positive stimuli: Marital stress positive affect.Psychophysiology, 51(6), 499–509. https://doi.org/10.1111/psyp.12203.

Larsen, J. T., Norris, C. J., & Cacioppo, J. T. (2003). Effects of positive and negative affecton electromyographic activity over zygomaticus major and corrugator supercilii.Psychophysiology, 40(5), 776–785.

Levenson, R. W. (2014). The autonomic nervous system and emotion. Emotion Review,6(2), 100–112. https://doi.org/10.1177/1754073913512003.

Magnée, M. J., Stekelenburg, J. J., Kemner, C., & de Gelder, B. (2007). Similar facialelectromyographic responses to faces, voices, and body expressions. Neuroreport,18(4), 369–372.

Mauss, I. B., Levenson, R. W., McCarter, L., Wilhelm, F. H., & Gross, J. J. (2005). The tiethat binds? Coherence among emotion experience, behavior, and physiology.Emotion, 5(2), 175–190. https://doi.org/10.1037/1528-3542.5.2.175.

Namba, S., Kabir, R. S., Miyatani, M., & Nakao, T. (2017). Spontaneous facial actions map

onto emotional experiences in a Non-social context: Toward a component-basedapproach. Frontiers in Psychology, 8. https://doi.org/10.3389/fpsyg.2017.00633.

Nummenmaa, L., Glerean, E., Viinikainen, M., Jaaskelainen, I. P., Hari, R., & Sams, M.(2012). Emotions promote social interaction by synchronizing brain activity acrossindividuals. Proceedings of the National Academy of Sciences, 109(24), 9599–9604.https://doi.org/10.1073/pnas.1206095109.

Nummenmaa, L., Saarimäki, H., Glerean, E., Gotsopoulos, A., Jääskeläinen, I. P., Hari, R.,... Sams, M. (2014). Emotional speech synchronizes brains across listeners and en-gages large-scale dynamic brain networks. NeuroImage, 102, 498–509. https://doi.org/10.1016/j.neuroimage.2014.07.063.

Raz, G., Winetraub, Y., Jacob, Y., Kinreich, S., Maron-Katz, A., Shaham, G., ... Hendler, T.(2012). Portraying emotions at their unfolding: A multilayered approach for probingdynamics of neural networks. NeuroImage, 60(2), 1448–1461. https://doi.org/10.1016/j.neuroimage.2011.12.084.

Reisenzein, R., Studtmann, M., & Horstmann, G. (2013). Coherence between emotion andfacial expression: Evidence from laboratory experiments. Emotion Review, 5(1),16–23. https://doi.org/10.1177/1754073912457228.

Rosenberg, E. L., & Ekman, P. (1994). Coherence between expressive and experientialsystems in emotion. Cognition & Emotion, 8(3), 201–229.

Rottenberg, J., Ray, R. D., & Gross, J. J. (2007). Emotion elicitation using films. Handbook ofemotion elicitation and assessment. London: Oxford University Press9–28.

Sander, D., Grandjean, D., & Scherer, K. R. (2005). A systems approach to appraisalmechanisms in emotion. Neural Networks, 18(4), 317–352.

Sato, W., Fujimura, T., Kochiyama, T., & Suzuki, N. (2013). Relationships among facialmimicry, emotional experience, and emotion recognition. PLoS ONE, 8(3), 1–8.https://doi.org/10.1371/journal.pone.0057889.

Scherer, K. R. (2009). The dynamic architecture of emotion: Evidence for the componentprocess model. Cognition & Emotion, 23(7), 1307–1351. https://doi.org/10.1080/02699930902928969.

Scherer, K. R., & Ellgring, H. (2007). Are facial expressions of emotion produced by ca-tegorical affect programs or dynamically driven by appraisal? Emotion, 7(1),113–130. https://doi.org/10.1037/1528-3542.7.1.113.

Scherer, K. R., Mortillaro, M., & Mehu, M. (2013). Understanding the mechanisms un-derlying the production of facial expression of emotion: a componential perspective.Emotion Review, 5(1), 47–53.

Schuyler, B. S., Kral, T. R. A., Jacquart, J., Burghy, C. A., Weng, H. Y., Perlman, D. M., ...Davidson, R. J. (2014). Temporal dynamics of emotional responding: Amygdala re-covery predicts emotional traits. Social Cognitive and Affective Neuroscience, 9(2),176–181. https://doi.org/10.1093/scan/nss131.

Simons, D. J., & Rensink, R. A. (2005). Change blindness: Past, present, and future. Trendsin Cognitive Sciences, 9(1), 16–20.

Singer, N., Jacoby, N., Lin, T., Raz, G., Shpigelman, L., Gilam, G., ... Hendler, T. (2016).Common modulation of limbic network activation underlies musical emotions as theyunfold. NeuroImage, 141, 517–529.

Stemmler, G., & Wacker, J. (2010). Personality, emotion, and individual differences inphysiological responses. Biological Psychology, 84(3), 541–551. https://doi.org/10.1016/j.biopsycho.2009.09.012.

Tan, J.-W., Walter, S., Scheck, A., Hrabal, D., Hoffmann, H., Kessler, H., ... Traue, H. C.(2012). Repeatability of facial electromyography (EMG) activity over corrugatorsupercilii and zygomaticus major on differentiating various emotions. Journal ofAmbient Intelligence and Humanized Computing, 3(1), 3–10.

Tassinary, L. G., & Cacioppo, J. T. (1992). Unobservable facial actions and emotion.Psychological Science, 3(1), 28–33 0956-7976.

Trost, W., Frühholz, S., Cochrane, T., Cojan, Y., & Vuilleumier, P. (2015). Temporal dy-namics of musical emotions examined through intersubject synchrony of brain ac-tivity. Social Cognitive and Affective Neuroscience, 10(12), 1705–1721. https://doi.org/10.1093/scan/nsv060.

Van Reekum, C. M., Schaefer, S. M., Lapate, R. C., Norris, C. J., Greischar, L. L., &Davidson, R. J. (2011). Aging is associated with positive responding to neutral in-formation but reduced recovery from negative information. Social Cognitive andAffective Neuroscience, 6(2), 177–185.

Wang, S.-Y., Liu, X., Yianni, J., Aziz, T. Z., & Stein, J. F. (2004). Extracting burst and toniccomponents from surface electromyograms in dystonia using adaptive waveletshrinkage. Journal of Neuroscience Methods, 139(2), 177–184. https://doi.org/10.1016/j.jneumeth.2004.04.024.

Waugh, C. E., Shing, E. Z., & Avery, B. M. (2015). Temporal dynamics of emotionalprocessing in the brain. Emotion Review1754073915590615.

Wichers, M., Wigman, J. T. W., & Myin-Germeys, I. (2015). Micro-level affect dynamics inpsychopathology viewed from complex dynamical system theory. Emotion Review,7(4), 362–367.

Wilhelm, F. H., Rattel, J. A., Wegerer, M., Liedlgruber, M., Schweighofer, S., Kreibig, S.D., ... Blechert, J. (2017). Attend or defend? Sex differences in behavioral, autonomic,and respiratory response patterns to emotion–eliciting films. Biological Psychology,130(Supplement C), 30–40. https://doi.org/10.1016/j.biopsycho.2017.10.006.

Zaki, J., Bolger, N., & Ochsner, K. (2009). Unpacking the informational bases of empathicaccuracy. Emotion, 9(4), 478–487. https://doi.org/10.1037/a0016551.

Y. Golland et al. Biological Psychology 139 (2018) 47–58

58