Cognition and Emotion ISSN: 0269-9931 (Print) 1464-0600 (Online) Journal homepage: http://www.tandfonline.com/loi/pcem20 A sad thumbs up: incongruent gestures and disrupted sensorimotor activity both slow processing of facial expressions Adrienne Wood, Jared D. Martin, Martha W. Alibali & Paula M. Niedenthal To cite this article: Adrienne Wood, Jared D. Martin, Martha W. Alibali & Paula M. Niedenthal (2018): A sad thumbs up: incongruent gestures and disrupted sensorimotor activity both slow processing of facial expressions, Cognition and Emotion, DOI: 10.1080/02699931.2018.1545634 To link to this article: https://doi.org/10.1080/02699931.2018.1545634 View supplementary material Published online: 15 Nov 2018. Submit your article to this journal Article views: 16 View Crossmark data Full Terms & Conditions of access and use can be found at http://www.tandfonline.com/action/journalInformation?journalCode=pcem20 COGNITION AND EMOTION https://doi.org/10.1080/02699931.2018.1545634 A sad thumbs up: incongruent gestures and disrupted sensorimotor activity both slow processing of facial expressions Adrienne Wood a , Jared D. Martinb, Martha W. Alibalib and Paula M. Niedenthalb a Department of Psychological and Brain Sciences, Dartmouth College, Hanover, NH, USA; bDepartment of Psychology, University of Wisconsin-Madison, Madison, WI, USA ABSTRACT ARTICLE HISTORY Recognising a facial expression is more difficult when the expresser’s body conveys incongruent affect. Existing research has documented such interference for universally recognisable bodily expressions. However, it remains unknown whether learned, conventional gestures can interfere with facial expression processing. Study 1 participants (N = 62) viewed videos of people simultaneously producing facial expressions and hand gestures and reported the valence of either the face or hand. Responses were slower and less accurate when the face-hand pairing was incongruent compared to congruent. We hypothesised that hand gestures might exert an even stronger influence on facial expression processing when other routes to understanding the meaning of a facial expression, such as with sensorimotor simulation, are disrupted. Participants in Study 2 (N = 127) completed the same task, but the facial mobility of some participants was restricted, which disrupted face processing in prior work. The hand-face congruency effect from Study 1 was replicated. The facial mobility manipulation affected males only, and it did not moderate the congruency effect. The present work suggests the affective meaning of conventional gestures is processed automatically and can interfere with face perception, but does not suggest that perceivers rely more on gestures when sensorimotor face processing is disrupted. Received 16 May 2018 Revised 28 October 2018 Accepted 5 November 2018 The human body provides multiple channels for communication. In addition to information communicated verbally, faces display expressions and hands produce meaningful gestures. However, much research on the perception of facial expression neglects the communicative nature of the human body beyond the face. This is true in spite of the fact that observers perceive communicative bodies and faces as parts of a single whole (Aviezer, Trope, & Todorov, 2012b): a facial expression may take on new meaning as a function of the information conveyed by the body (Aviezer, Trope, & Todorov, 2012a), or vice versa (Burgoon, Buller, Hale, & de Turck, 1984). Since most faces encountered in everyday life are accompanied by fully or partially visible bodies, faces and bodies must be studied together in order to fully understand how meaning is communicated nonverbally (de Gelder & Hortensius, 2014). Emotion perception; gesture; sensorimotor simulation Research on how perceivers process expressive bodies, separately or in combination with expressive faces, has focused on body postures or movements with evolved affective or functional meaning, such as cowering in fear or jumping for joy (Atkinson, Tunstall, & Dittrich, 2007; de Gelder, de Borst, & Watson, 2014; Tamietto et al., 2009). But much communicative body and hand movement is learned, culturallyspecific, and at least partly symbolic (i.e. not derived from a functional motion). We will refer to these movements as conventional gestures. Some conventional gestures have emotional connotations (e.g. thumbsdown; Redcay & Carlson, 2015), and should in theory facilitate or interfere with processing of an accompanying facial expression, just as emotional body expressions do (Meeren, van Heijnsbergen, & De Gelder, 2005). On the other hand, given that CONTACT Adrienne Wood adrienne.wood@wisc.edu Supplemental data for this article can be accessed at 10.1080/02699931.2018.1545634 © 2018 Informa UK Limited, trading as Taylor & Francis Group KEYWORDS 2 A. WOOD ET AL. conventional gestures are learned and culturallyspecific, they may not be strongly correlated with facial expressions of emotion in people’s social environments. Thus, observers may not process conventional gestures as a part of a cohesive unit along with the face. To date, it is unknown whether conventional hand gestures are processed separately from – or holistically with – facial expressions of emotion. In the current work, we ask whether conventional gestures, like bodily expressions of emotion, automatically influence the processing of facial expressions, and vice versa, whether facial expressions automatically influence the processing of conventional hand gestures. We first review evidence that observers combine the messages conveyed by faces and bodies, processing them as a cohesive whole. We then build the argument that such holistic processing might also include conventional hand gestures, which play an important role in daily communication but have been largely neglected by emotion perception research. We then step back to review evidence for a psychological process thought to support face and body perception generally, known as sensorimotor simulation. Given the evidence that sensorimotor simulation assists in facial expression perception, conventional gestures may have even greater influence on the automatic processing of facial expressions when sensorimotor simulation is compromised. In Study 1 we ask whether people are slower to categorise the meaning of a dynamic facial expression accompanied by an incongruent, versus congruent, hand gesture. Study 2 tests whether information conveyed by an expresser’s hands may exert even stronger influence on the perception of their facial expression when sensorimotor contributions to face perception are disrupted. Combined processing of face and body cues Expressive faces and bodies are attention-grabbing and difficult to ignore, even when they are task-irrelevant (Van den Stock & de Gelder, 2012; 2014). Facial expressions and bodily signals that are universal and related to functional actions – such as the erratic, domineering movements associated with anger or the cowering, submissive movements associated with fear – are known to elicit early preferential neural and physiological responses in observers (De Gelder, 2006; Eimer & Holmes, 2002) whether or not people are consciously aware of them (De Gelder & Hadjikhani, 2006; Tamietto et al., 2009). Since emotional faces and bodies are difficult to ignore, each influences processing of the other when they are combined. When the signals are congruent (e.g. an angry face with an angry body) they are both better-recognized and judged to convey a more intensely emotional message (de Gelder et al., 2014; Martinez, Falvello, Aviezer, & Todorov, 2016). Emotions expressed by a congruent face-body stimulus are also more evocative of neural and physiological responses than emotions expressed by either the face or the body in isolation (Kret, Roelofs, Stekelenburg, & de Gelder, 2013; Poyo Solanas et al., 2018). When the body and face convey distinct or even competing messages, the perceived meaning of each signal changes. A smile, for instance, is interpreted differently when paired with an engaged, forward-leaning body posture compared to a disengaged posture (Burgoon et al., 1984). Indeed, people are worse at recognising facial or bodily emotion expressions when paired with an incongruent body or face, respectively (Aviezer et al., 2012b; Meeren et al., 2005). Conventional gestures can be learned affective signals The work on bodily contributions to facial expression processing reviewed above has focused largely on body gestures that resemble functional behaviours (e.g. cowering in fear), but much nonverbal human communication involves the exchange of symbolic conventional hand gestures. These gestures are learned and vary in meaning from culture to culture: for instance, the forefinger-to-thumb gesture that means “A-OK” and signals approval in the United States might start a fight in Turkey. A metaphoric origin for some conventional gestures can be identified – thumbs-up as good and thumbs-down as bad presumably map onto the metaphor of higher as better – but these gestures are sufficiently removed from functional actions that the relationship between the gesture and its referent is largely arbitrary (Archer, 1997). Despite the differences between bodily expressions grounded in functional behaviours and conventional hand gestures, people learn that certain conventional gestures constitute positive or negative social feedback and should be prioritised similarly to emotional expressions. Symbolic gestures like a thumbs-up or a middle finger capture and sustain visual attention, much like biologicallygrounded emotion expressions (Flaisch, Häcker, COGNITION AND EMOTION Renner, & Schupp, 2011; Flaisch, Schupp, Renner, & Junghöfer, 2009). As with facial expressions of emotion, people process the affective meanings of hand gestures extremely rapidly. Magnetoencephalography recordings of people viewing hand gestures indicate they encode the emotional valence and selfrelevance of the gestures just milliseconds after onset (70 and 100 ms after presentation, respectively; Redcay & Carlson, 2015). Together, such studies suggest conventional gestures are meaning-laden social signals that are processed early and are difficult to ignore. The learned emotional relevance of conventional gestures is underscored by their potential to serve as “unconditioned” positive and negative stimuli. In one study, participants were conditioned to associate pictures of neutral faces with pictures of negative (middle finger), neutral (pointing), or positive (thumbs-up) gestures (Wieser, Flaisch, & Pauli, 2014). Faces associated with the negative gesture were rated as more unpleasant and arousing. Those faces also elicited larger steady-state visually evoked potentials (measured with electroencephalography), which reliably occur when people view aversively conditioned stimuli (Moratti, Keil, & Miller, 2006). This study suggests valenced hand gestures automatically influence how associated faces are encoded. However, the study paired gestures with neutral faces, and these faces were not physically attached to the images of the hands. To our knowledge, no work has examined how valenced hand gestures influence processing of the whole-person signal. Sensorimotor processing supports visual perception of facial expressions We have proposed that conventional hand gestures, like biologically-grounded body expressions of emotion, can modulate processing of facial expressions, and vice-versa. However, it stands to reason that the relative importance of the hand gesture and the facial expression will depend on which signal is clearer in its meaning or more accessible to the observer. In accordance with this logic, bodily expressions override facial expressions when faces are expressing extreme emotions, as people’s faces tend to collapse into ambiguous scream-like expressions during intense emotions (Aviezer et al., 2012a). The relative influence of the body expression on the perception of face-body stimuli can also be decreased by administering oxytocin to participants, 3 causing them to attend more to the face (Perry et al., 2013). It is possible that we could increase the influence of gestures on face perception by making it more difficult for participants to process facial displays. One technique for disrupting facial expression processing is to experimentally compromise the observer’s face-related sensorimotor processes (e.g. Ipser & Cook, 2016; Rychlowska et al., 2014; Wood, Lupyan, Sherrin, & Niedenthal, 2016). Substantial behavioural, clinical, and neural evidence suggests that when people perceive an action, like a facial expression, they engage their own sensorimotor systems to simulate the experience of producing the action (Blakemore & Decety, 2001; Niedenthal, Mermillod, Maringer, & Hess, 2010). Such sensorimotor simulation contributes to their ability to process the meaning and intention underlying the perceived action. When the ability to engage in sensorimotor simulation is disrupted – for instance, by having participants generate incompatible facial movements or distorting the somatosensory feedback from their faces – emotion perception speed and accuracy is reduced (for a recent review, see Wood, Rychlowska, Korb, & Niedenthal, 2016). Disruptions to face-specific sensorimotor simulation processes may then increase observers’ reliance on other sources of emotional information, such as hand gestures, a hypothesis we test here. The present studies We examine how perceivers process dynamic hand gestures paired with dynamic facial expressions that are either congruent or incongruent. Participants in two studies categorised stimuli according to valence (whether they convey negative or positive emotions) and we therefore operationalised “congruent” facehand pairs as positive face-positive hand or negative face-negative hand.1 Studies 1 and 2 both tested the prediction that perceivers are slower and less accurate in categorising the valence of either the hand or face in a hand-face pair with messages that differ in valence (Hypothesis 1). Confirmation of this prediction would suggest that observers automatically process the affective meanings of conventional gestures, interfering with categorisation of the accompanying facial expressions (and vice-versa). Study 2 added a between-subjects manipulation of facial mobility to address two goals. First, we sought to replicate previously-observed effects of facial sensorimotor disruption on speed and accuracy in categorising facial 4 A. WOOD ET AL. expressions. Second, we predicted that participants are slower and less accurate in judging the valence of facial expressions when their facial mobility is restricted (Hypothesis 2). The mobility manipulation also allowed us to test a further prediction, that incongruent hand gestures have a greater influence on judgments of facial expressions when sensorimotor simulation is disrupted (Hypothesis 3). Finally, considering prior evidence for gender differences in emotion perception (Scherer & Scherer, 2011) and sensitivity to disruptions to sensorimotor processes (Niedenthal et al., 2012), we included gender as a moderator in all of our statistical models; however, we did not make directional predictions about the potential moderating influence of gender. Study 1 Method In the following sections, we report how we determined our sample size, all data exclusions, all manipulations, and all measures. The experiment and stimuli files are available online (https://osf.io/9xs48/). Video stimuli Four White actors (two female, two male) were recruited from the University of Wisconsin–Madison undergraduate theatre programme and paid to be filmed. Actors signed appropriate release forms and were coached on how to produce each facial and hand movement. The facial expressions were happiness, positive surprise, fear, sadness, disgust, and anger, and the hand gestures were thumbs-up, AOK, thumbs-down, and a fist raised as if in anger (see Figure 1). The dynamic facial expressions began from neutral and ended at the apex of the expression. The dynamic hand gestures began with the hand offcamera, then the actor raised their hand, emphasised the gesture with 2 superimposed beats (i.e. pulses), and held the position still for the remainder of the video. The facial expressions and hand gestures for each actor were filmed separately on green screen backgrounds and were then overlaid on one another with a black background, creating all possible hand gesture-facial expression combinations for each actor (4 actors * 6 facial expressions * 4 hand gestures = 96 unique videos). “Congruent” face-hand pairs were combinations of positive facial expressions and positive hand gestures or negative facial expressions and negative hand gestures. “Incongruent” facehand pairs were combinations of a negativelyvalenced channel (either face or hand) and a positively-valenced channel (hand or face, respectively). For use in the Baseline Phase, we also created handonly (16 in total) and facial expression-only (24 in total) videos for each actor. The face-only stimuli showed the actors’ heads (with hands not visible), and the hand-only stimuli showed the actors’ hands (with faces not visible). Participants and procedure Figure 1. Still frames from Actor 1’s video stimuli. The dynamic facial expressions began from neutral and ended at the apex of the expression. All 4 actors posed 4 negative facial expressions, sadness (A), disgust (B), fear (C), anger (D); and 2 positive expressions, positive surprise (E), and happiness (F). For the dynamic hand gestures, all 4 actors started with their hand off-camera, then raised it, emphasised the gesture with 2 pulses, and held it still for the remainder of the video. The negative gestures were thumbs-down (G) and fist (H), and the positive gestures were A-OK (I) and thumbs-up (J). Sample congruent (K) and incongruent (L) stimuli for the Face-Hand Phase are also shown. We aimed for 60 participants for the fully withinsubject design. 62 participants were recruited from the University of Wisconsin–Madison undergraduate Introduction to Psychology subject pool to participate in exchange for course credit (nmale = 25, nfemale = 37). The study was conducted with approval from the Institutional Review Board. All participants gave their informed consent. The design included a Baseline Trial phase followed by a Face-Hand Trial phase. On all trials, participants watched a video of a dynamic hand, face, or facehand combination video and were instructed to use a keypress to categorise the target stimulus as positive or negative. Participants pressed the “f” and “j” keys to indicate “positive” or “negative”, and the key assignment was counterbalanced across participants. In the Baseline Trial phase, participants categorised stimuli that contained facial expressions only and COGNITION AND EMOTION stimuli that contained hand gestures only, presented in randomised order in a single block (40 trials in total). After the Baseline Trial phase, participants completed 6 blocks consisting of 32 Face-Hand trials each (192 trials in total).2 Before each block participants were instructed to attend to and categorise only the faces or only the hands in that block, ignoring the other signal. Instructions alternated across blocks. Each Face-Hand video appeared twice during this phase, once per attention instruction condition. Participants were told to “respond as fast as you can while still being accurate.” Results We hypothesised that participants instructed to categorise the valence of either a hand gesture or facial expression would be unable to ignore the other channel, resulting in slower reaction times (RTs) when the valences of the hand and face were incongruent (e.g. a positive gesture with a negative facial expression) compared to congruent. Because we expected participants to have near-ceiling performance on what is a straightforward task–categorizing a clear and recognisable hand or face signal as negative or positive–we did not anticipate effects on accuracy, although we did analyze the accuracy data. An R Markdown document (RStudio and Inc, 2014) containing complete code and output from our data processing and analyses for both studies is available online (https://osf.io/9xs48/). In all our analyses we used linear mixed-effects models with the lme4 package in R (Bates, Mächler, Bolker, & Walker, 2015). Degrees of freedom and p values were estimated using Satterthwaite approximations with the lmerTest package (Kuznetsova, Brockhoff, & Christensen, 2017). We adhered to the “keep it maximal” approach to random effects structures and therefore included by-item (stimulus) and by-subject random intercepts and slopes for all predictors that were repeated within-item or within-subject (Barr, Levy, Scheepers, & Tily, 2013). In the case of convergence failures, we removed the random slopes for main effects, leaving the random slopes for interaction terms, as recommended by Brauer and Curtin (2018). In all models, we allowed participant gender to interact with the experimental variables of interest, since previous work documents gender effects on emotion perception (e.g. Donges, Kersting, & Suslow, 2012; Korb et al., 2015; Montagne, Kessels, Frigerio, de Haan, & Perrett, 2005). Before conducting our analyses, 5 we removed trials on which participants’ RTs were 3 standard deviations above the mean (1.56% of trials) or faster than 150 ms (0.44% of trials). Accuracy Descriptive statistics for participant accuracy in the Baseline Phase suggest that the valence (positive versus negative) of the facial expressions and hand gestures was unambiguous. Participants made very few errors in categorising the valence of the facial expressions (proportion correct M = 0.99, SD = 0.11) and the hand gestures (proportion correct M = 0.94, SD = 0.24). We did not have any predictions regarding the Baseline Phase, which was included to adjust the Face-Hand Phase RTs. To examine whether participants were more accurate in the Face-Hand Phase when the valence of the face and hand matched, we calculated each participants’ proportion correct for congruent and incongruent face-attend and gesture-attend trials (resulting in 4 accuracy scores per participant). We regressed these Accuracy scores on unit-weighted, centred variables for Attention Instructions (attend to the hand gesture = −.5, attend to the facial expression = .5), Face-Hand Congruency (incongruent stimuli = −.5, congruent stimuli = .5), Participant Gender (male = −.5, female = .5), the three-way interaction between the variables, the three two-way interactions, and all main effects. We initially included all relevant random effects: the by-participant random intercept and random slopes within participants for the interaction between Attention Instructions and FaceHand Congruency and the two main effects. By-item random effects were not applicable since accuracy scores are aggregated for each participant in each cell of the experimental design, and not calculated on an item-by-item basis. Statistically controlling for all other variables, accuracy was significantly greater on face-attend trials (proportion correct M = 0.98, SD = 0.15) compared to gesture-attend trials (proportion correct M = 0.93, SD = 0.26), b = 0.051, t(180.00) = 7.093, p < .001. Supporting Hypothesis 1, the main effect of Face-Hand Congruency was also significant, with greater accuracy on congruent (M = 0.96, SD = 0.20) compared to incongruent (M = 0.94, SD = 0.23) trials, b = .015, t(180.000) = 2.139, p = .034.3 Neither Participant Gender nor any of the interaction terms were significant. In post hoc Supplementary Analyses, we checked for emotion and gesture-specific effects and found 6 A. WOOD ET AL. that the Face-Hand Congruency effect on accuracy was stronger for trials involving positive facial expressions and negative gestures, compared to trials involving negative facial expressions and positive gestures. Given the small number of different actors we included in our stimuli, we also checked for moderating effects of Actor and found, among all possible comparisons between actors, that only Actors 2 and 4 differed in the size of their Congruency effect, with a stronger effect of Face-Hand Congruency for Actor 4 (see Supplementary Analyses, https://osf. io/9xs48/). Reaction times In the RT analyses, we included only correct-response trials. Since the stimuli were dynamic, some facial expressions and gestures have slower onset times than others, resulting in stimulus effects on RTs. To remove some of this variance, we calculated each participant’s average RT for each hand or face stimulus during the Baseline Phase. We then adjusted the RTs on the Face-Hand Phase by subtracting the average Baseline Trial RTs for the relevant participant and stimulus. These adjusted RTs now indicate how much faster (negative value) or slower (positive Figure 2. Study 1 model estimates from linear mixed-effects model in which adjusted RTs in the Face-Gesture Phase were regressed on the interaction between Attention Instructions, Face-Gesture Congruency, and Participant Gesture. RTs were adjusted for participants’ Baseline Phase RTs for each stimulus. Points are individual participants’ average adjusted RTs. Participants were faster when categorising the valence of the face compared to the hand, and they were also faster when the valences of the face and hand were congruent compared to incongruent. The significant 3-way interaction indicates that female participants were particularly disrupted by incongruent facial expressions when categorising the valence of a hand gesture. value) a participant was in categorising a given gesture or facial expression when accompanied by a signal from the other channel, compared to when they categorised the gesture or expression in isolation. Note that the by-item random effects account for the variability in RTs across face-hand stimuli (some facehand pairs may tend to be processed more quickly than others), whereas subtracting the Baseline RTs removes some unexplained variance due to the properties of the isolated target face or hand stimulus. These two steps, subtracting Baseline Trial RTs from Face-Hand RTs and including by-item random effects, account for unique sources of otherwise unexplained variance in our analyses. We regressed the adjusted Face-Hand Trial RTs (in seconds) for correct-response trials on unit-weighted, centred variables for Attention Instructions (attend to the hand gesture = −.5, attend to the facial expression = .5), Face-Hand Congruency (incongruent stimuli = −.5, congruent stimuli = .5), Participant Gender (male = −.5, female = .5), the three-way interaction between the variables, the three two-way interactions, and all main effects. We included all relevant random effects: the by-participant random intercept, the by-participant random slopes for the interaction between Attention Instructions and Face-Hand Congruency and the two main effects, the by-item random intercept, and the by-item random slopes for the interaction between Attention Instructions and Participant Gender and the two main effects. The intercept of the model revealed that participants were, on average, 258 ms slower in the FaceHand Phase compared to Baseline Phase, controlling for all predictor variables, b = 0.258, t(73.750) = 10.578, p < .001. The main effect of Attention Instructions was significant, controlling for all other variables, such that participants responded more quickly on face-attend trials (adjusted RT M = 136 ms, SD = 378 ms) compared to gesture-attend trials (adjusted RT M = 383 ms, SD = 323 ms), b = –0.241, t(118.72) = – 9.942, p < .001. The main effect of Face-Hand Congruency was also significant, controlling for all other variables, such that participants responded more quickly when the face and hand valences were congruent (adjusted RT M = 226 ms, SD = 365 ms) compared to incongruent (adjusted RT M = 285 ms, SD = 380 ms), b = –0.059, t(106.74) = –3.354, p = 0.001. Finally, the 3-way interaction between Attention Instructions, Face-Hand Congruency, and Participant Gender was significant, such that incongruent facial expressions slowed the valence judgments of hand COGNITION AND EMOTION gestures more for female participants than for males, b = 0.073, t(89.03) = 2.210, p = 0.030 (see Figure 2). As for accuracy, we conducted several post hoc analyses to check for potential moderating effects of Facial Expression and Gesture categories, as well as actor identity, on RTs (see Supplementary Analyses, https://osf.io/9xs48/). To summarise the additional analyses, the effect of incongruent Face-Hand pairs on RTs was stronger for some negative facial expressions compared to positive facial expressions. Specifically, the Face-Hand Congruency effect was stronger for anger, fearful surprise, and sadness compared to positive surprise, and stronger for anger compared to happiness. The only moderating effect of Gesture that emerged indicated that the Congruency effect was weaker for the A-OK gesture compared to the Thumbs-Down gesture. No moderating effects of actor identity on the Face-Hand Congruency effect emerged. We also examined whether the Congruency effect might be due, least in part, to “carry-over” effects of task demands from one block to the next (e.g. the possibility that incongruent gestures interfered with face categorisation because in the prior block, participants were instructed to attend to the gesture). There was no support for the existence of such carry-over effects. Study 1 summary Despite the near-ceiling accuracy across trials, Study 1 participants were more accurate and faster on the face-attend (compared to gesture-attend) trials. Supporting Hypothesis 1, participants were more accurate and faster on trials on which the face and hand signals were congruent compared to incongruent. Finally, the unexpected significant 3-way interaction in the RT model provides suggestive evidence that female participants are less able to ignore task-irrelevant facial expressions than are male participants. Study 2 Study 1 provided evidence in favour of Hypothesis 1, that individuals are slower and less accurate in categorising the valence of faces and hands accompanied by incongruent compared to congruent hands or faces, respectively. Study 2 sought to replicate this finding and added a between-subjects manipulation of facial muscle mobility to test our other two hypotheses. We predicted that individuals would be slower and less accurate in judging the valence of facial 7 expressions when their facial mobility is restricted (Hypothesis 2), and that incongruent hand gestures would have a greater influence on judgments of facial expressions when sensorimotor simulation is disrupted (Hypothesis 3). Method In the following section, we report how we determined our sample size, all data exclusions, all manipulations, and all measures. The experiment and stimuli files are available online (https://osf.io/9xs48/). Study 2’s procedure was the same as Study 1, with the addition of a between-subjects manipulation of facial mobility, which allowed us to test Hypotheses 2 and 3. Participants and procedure We aimed for 60 participants in each of the two Facial Mobility conditions, or 120 participants in total. We ultimately recruited 128 participants from the University of Wisconsin–Madison undergraduate Introduction to Psychology subject pool to participate in exchange for course credit (nfemale = 88, nmale = 40). One male participant did not complete the study and was excluded from analyses. The study was conducted with approval of the Institutional Review Board. After participants heard instructions and gave their informed consent, they were randomly assigned to one of two Facial Mobility conditions: the face tape condition or the control condition. Participants then completed the 40 trials in the Baseline Trial phase, followed by the 192 trials of the Face-Hand Trial phase. See Study 1 Method for details. Facial mobility manipulation Participants were randomly assigned to a face tape condition or control condition. The present facial mobility manipulation is identical to a procedure developed and employed in previous research (Carpenter & Niedenthal, under review). In the face tape condition, three kinds of stiff and inflexible medical tape were applied across the width of participants’ foreheads in three layers: the first layer was 1/2 inches wide, extending vertically from the bridge of the nose to the hairline; the second layer was 2 inches wide, extending horizontally across the forehead from one temple to the other; the third layer consisted of two 1 inches wide pieces of tape applied on top of the second layer, to maximally 8 A. WOOD ET AL. disrupt upper facial mobility. In order to disrupt mobility of the lower part of their faces, participants in the face tape condition received boil-and-bite mouth guards, a manipulation shown to be effective in previous research (Rychlowska et al., 2014). Participants prepared their own, single-use mouth guards by first submerging their new mouth guard in boiling water for 7 s. Then, participants fitted the now soft mouth guard to their teeth for 10 s in order to conform to the shape of their mouths. Finally, the mouth guard was placed in cold water for 20 s to ensure that it maintained the shape of the participant’s mouth. In the control condition, participants also had tape applied to their faces but in a location and extent so as to minimally impact facial mobility. Specifically, small 1/2 inches wide pieces of tape were placed on the participants’ temples. By also having participants in the control condition receive some form of facial taping, we experimentally controlled for the potential confound of receiving face tape. This allows us to draw stronger inferences regarding the experimental effect of facial mobility restriction. Participants in the control condition also made a boil-and-bite mouth guard but were told to place it on a paper towel beside the experimental computer for later use. In fact, participants in the control condition never wore their mouth guards during the experiment. Results We tested 3 hypotheses in Study 2: (1) participants are slower when the valences of the face and hand signals are incongruent compared to congruent (replicating Study 1), (2) participants whose facial movements are disrupted by the face tape are slower on faceattend trials compared to control condition participants, conceptually replicating past work (Ipser & Cook, 2016; Wood, Lupyan, et al., 2016), and (3) the effect of Face-Hand Congruency depends on facial mobility, such that incongruent hand gestures are even more disruptive for participants whose facial mobility is restricted than for participants whose facial mobility is not restricted. As in Study 1, we did not have specific predictions about effects on categorisation accuracy, since we expected participants to perform near ceiling. Besides the addition of another moderator – Facial Mobility – the models we report for Study 2 are identical to those used in Study 1. Here we also included analyses of the Baseline Phase, since these trials now pertain to Hypothesis 2. The complete code and output for the following analyses can be found online in the same R Markdown file as the Study 1 analyses (https://osf.io/9xs48/). As in Study 1, before conducting our analyses, we removed trials on which participants had RTs that were 3 standard deviations above the mean (2.23% of trials) or faster than 150 ms (.79% of trials). Accuracy in the baseline phase Figure 3. Study 2 Baseline Phase. Model estimates from linear mixedeffects model in which RTs in the Baseline trials were regressed on the interaction between Stimulus Type (facial expression vs. gesture), Participant Gender, and Facial Mobility. Points are individual participants’ average RTs. The 3-way interaction term is significant, suggesting male participants were slower to categorise facial expressions compared to gestures when their facial mobility was disrupted, but this was not the case for female participants. Using a linear mixed-effects model, we first regressed participants’ proportion correct scores for the Baseline Phase on the 3-way interaction between Stimulus Type (hand gesture = −.5, facial expression = .5), Facial Mobility (control condition = −.5, face tape condition = .5), and Participant Gender (male = −.5, female = .5), the three 2-way interactions, and all main effects. We included the by-participant random intercept and the random slope within participants for Stimulus Type. The only variable to significantly predict accuracy in the Baseline Phase was Stimulus Type, such that participants were more accurate on facial expression trials (M = 0.97, SD = 0.11) than on hand gesture trials (M = 0.93, SD = 0.15), b = 0.042, t(368.30) = 3.786, p < 0.001. Reaction times in the baseline phase We next ran an identical model with RT (in seconds) on correct-response trials as the outcome variable. COGNITION AND EMOTION The only difference in the specified model was the inclusion of by-stimulus random effects: the random intercept and random slopes for the 2-way interaction between the Facial Mobility and Participant Gender, and the two main effects. Stimulus Type was again a significant predictor, with faster RTs on facial expression trials (M = 1213 ms, SD = 471 ms) than on gesture trials (M = 1144 ms, SD = 420 ms), b = 0.086, t(46.46) = 2.045, p = 0.047. The 3way interaction between Stimulus Type, Facial Mobility, and Participant Gender was significant, b = – 0.169, t(77.63) = –2.411, p = 0.018 (see Figure 3). The key test of Hypothesis 2, if it were not moderated by Participant Gender, is the Stimulus Type * Facial Mobility interaction term, which was not significant here, b = 0.042, SE = 0.035, t(114.30) = 1.21, p = 0.229. This null effect is unlikely to be due to a lack of power. A power analysis using the simr package for R (Green & MacLeod, 2015) suggested that we had acceptable power (94.80%) to detect a Stimulus Type * Facial Mobility interaction term, averaged, over males and females, of equivalent effect size to the male-specific interaction effect (see below; b = 0.127). Given the significant 3-way interaction, we next consider the Stimulus Type * Facial Mobility interaction for males and females separately. To unpack the 3-way interaction and test the hypothesis that the face tape specifically slowed RTs on facial expression trials (Hypothesis 2), we recentered the Stimulus Type variable (facial expression = 0, gesture = 1) so that we could interpret the effects of Participant Gender and Facial Mobility for facial expression trials specifically. In this recentered model, the interaction between Participant Gender and Facial Mobility was significant for facial expression trials, b = –0.296, t(94.45) = –2.219, p = 0.029. The difference in RTs between the face tape and control conditions for males (M = 1291 ms, SD = 516 ms vs. M = 1184 ms, SD = 436 ms) was greater than the difference between the face tape and control conditions for females (M = 1160 ms, SD = 409 ms vs. M = 1171 ms, SD = 475 ms) in females. Recentering Stimulus Type over gesture (gesture = 0, facial expression = 1) revealed that the Gender * Facial Mobility interaction was not significant for gesture trials, meaning males’ and females’ RTs for gesture stimuli were not differentially affected by the face tape condition, b = –0.127, SE = 0.105, t(93.23) = –1.213, p = .228. Finally, we recentered the Participant Gender variable over males (males = 0, females = 1) and then over 9 females (females = 0, males = 1) to examine the 2way interaction between Facial Mobility * Stimulus Type for each gender. For male participants, the face tape manipulation slowed RTs on facial expression trials significantly more than on gesture trials, b = –0.127, SE = 0.057, t(114.73) = –2.206, p = 0.029. The Facial Mobility * Stimulus Type interaction was not significant for female participants, however, b = 0.043, SE = 0.038, t(113.34) = 1.114, p = .268. To summarise, in the Baseline Phase, the face tape manipulation slowed RTs on facial expression trials specifically for male participants. Accuracy in the face-hand phase Since the Face-Hand Phase included an additional within-subjects variable – Face-Hand Congruency – we ran the same model as for Baseline Trial accuracy scores, but with the addition of Face-Hand Congruency as a moderator. Thus, the highest-order term was a 4-way interaction between Face-Hand Congruency, Attention Instructions, Facial Mobility, and Participant Gender, and there were also by-participant random slopes for the 2-way interaction between Attention Instructions and Face-Hand Congruency and the two main effect terms. Once again, participants were significantly more accurate on face-attend trials (M = 0.96, SD = 0.06) than on gesture-attend trials (M = 0.91, SD = 0.11), b = 0.050, t(124.30) = 6.794, p < .001. In support of Hypothesis 1, participants were significantly more accurate when the valences of the face and hand were congruent (M = 0.95, SD = 0.09) than when they were incongruent (M = 0.92, SD = 0.09), b = 0.028, t(131.24) = 4.721, p < .001. Unexpectedly, the Facial Mobility×Attention Instructions×Participant Gender three-way interaction term was also significant, b = −.059, t(124.30) = –1.986, p = .049. To unpack the three-way interaction, we recentered Participant Gender to look at the two-way interaction between Facial Mobility and Attention Instructions specifically for males, which was just above the threshold for conventional statistical significance, b = .047, t(124.47) = 1.928, p = .056. This interaction can be interpreted as indicating that male participants were more accurate on face-attend than gesture-attend trials, and this difference was even stronger when they were in the face tape condition. Since this two-way interaction was not moderated by Face-Hand Congruency, was unexpected, and was not below the threshold for significance, we do not draw conclusions from it, but simply note it as 10 A. WOOD ET AL. something to follow up on in future work. No other variables significantly predicted accuracy in the FaceHand Phase. Reaction times in the face-hand phase Recall that in Study 1 we adjusted RTs in the FaceHand Phase using participants’ RTs from the Baseline Phase trials. In Study 2, adjusting the Face-Hand Phase RTs using participants’ own Baseline Phase RTs would essentially subtract out any between-participant differences due to the Facial Mobility manipulation. For instance, if a participant were slower due to the face tape, they would be slower on both Baseline and Face-Hand Trials and subtracting the former from the latter would remove the effect of face tape on their Face-Hand Trial RTs. We therefore computed average RTs on correct trials for each of the 96 stimuli using Study 1 participants’ Baseline Phase RTs, and then subtracted these bystimulus RT averages from the Study 2 Face-Hand Phase correct-trial RTs. Although an imperfect solution, this reduced some variance in RTs due to differences in stimulus onset speed. We initially regressed the adjusted RTs on the same 4-way interaction as in the Baseline Phase analysis, with by-participant random intercept and random slopes for the interaction between Attention Instructions and Face-Hand Congruency, and the two main effects. We also included the by-item random intercept and random slopes for the 3-way interaction between Attention Instructions, Participant Gender, and Facial Mobility, and all main effects. As before, participants were significantly faster on face-attend (M = 237 ms, SD = 433 ms) compared to gesture-attend trials (M = 473 ms, SD = 357 ms), b = –0.229, t(122.50) = –19.280, p < .001. Replicating Study 1’s support of Hypothesis 1, they were also significantly faster on trials on which the valences of the face and hand were congruent (M = 325 ms, SD = 406 ms) compared to incongruent, (M = 379 ms, SD = 422 ms), b = –0.051, t(95.90) = –2.309, p = 0.023. The 2-way interaction between Attention Instructions and Face-Hand Congruency was also significant, with the Congruency effect being somewhat weaker on face-attend trials (i.e. incongruent gestures did not slow RTs for facial expressions as much as incongruent facial expressions slowed RTs for gestures), b = 0.028, t(752.90) = 2.903, p = 0.003. No other main effects or interaction terms were significant; thus, we did not find support for the hypothesis that the face tape manipulation would exaggerate the face-hand congruency effect (Hypothesis 3; see Figure 4). Study 2 summary We replicated the finding from Study 1 that participants were less accurate and slower to correctly categorise the valence of a facial expression or hand gesture when it was accompanied by an incongruent signal from the other channel (Hypothesis 1). We found mixed support for the prediction that, relative to controls, participants would be slower and less accurate in categorising facial expressions when their facial movements were disrupted by the face tape (Hypothesis 2). On Baseline Phase trials, on which participants saw facial expressions or hand gestures in isolation, male participants’ performance was slowed significantly more by the face tape manipulation compared to females, whose performance appears to have been unaffected by the face tape. However, this gender-specific effect was not present on Face-Hand Phase trials. Thus, we partially replicated previous work that disrupted facial mobility results in poorer facial expression perception (Wood, Lupyan, et al., 2016). Finally, we did not find evidence supporting the hypothesis that the face-hand congruency effect would be exaggerated when participants’ facial mobility was disrupted (Hypothesis 3). Discussion In two studies, we asked how participants perceive actors producing dynamic facial expressions (sad, angry, disgusted, afraid, happy, positive surprise) along with congruent or incongruent conventional hand gestures (thumbs-up, thumbs-down, A-OK, and a fist). When instructed to categorise the valence of either the face or hand and ignore the other expressive channel, participants’ judgment accuracy and reaction time were nonetheless influenced by the task-irrelevant channel. Such effects had previously been observed in face-body combinations involving functional, as opposed to symbolic and conventional, body postures (e.g. cowering in fear; Aviezer et al., 2012b). The current work suggests that, like body postures, learned symbolic hand gestures interfere with facial expression processing, and vice-versa. The current work also increases the ecological validity of prior work on face-body expression processing by using video stimuli rather than still photos (cf. Martinez et al., 2016). COGNITION AND EMOTION 11 Figure 4. Study 2 Face-Hand Phase. Model estimates from a linear mixed-effects model in which Face-Hand Phase RTs (adjusted using Study 1 Baseline Phase RTs) were regressed on the 4-way interaction between Attention Instructions (face-attend vs. gesture-attend), Face-Hand Congruency (incongruent vs. congruent), Participant Gender, and Face Tape Manipulation (face tape vs. control). Points represent individual participants’ average adjusted RTs, which were negative in the event that a participant’s responses were faster in the Face-Hand Phase than the Baseline Phase (presumably due to practice effects). The observed face-hand congruency effect, which was replicated in the second study, has implications for how observers process a whole expressive person. Since conventional hand gestures are presumably more volitional and controllable than facial, vocal, and bodily expressions of emotion, they may more regularly contradict the other, otherwise cohesive, expressive channels in daily life. If your friend notices you seem sad, you may not be able to suppress the sadness conveyed by your face or voice, so you may give them a thumbs-up to indicate that they do not need to take care of you. The current work suggests your friend will not ignore or prioritise one of the communicative channels but will instead combine the channels to infer your nuanced affective and intentional state. Study 2 was identical to Study 1, with the addition of a facial mobility manipulation: the facial movements of half of the participants were restricted with the application of medical face tape and a mouthguard, which previous work has used successfully to reduce facial movements and produce disruptive somatosensory feedback (Carpenter & Niedenthal, under review; Rychlowska et al., 2014). This manipulation allowed us to replicate previous studies in which this and similar manipulations reduced accuracy and response times in facial expression perception tasks (e.g. Wood, Lupyan, et al., 2016). Here we partially replicated the effect, specifically in male but not female participants, and the effect appeared sensitive to repeated exposure to the stimuli, as it was statistically significant in the Baseline Phase but not the Face-Hand Trials. The exposure effect is not surprising, as the task (judge whether an expressive face is positive or negative) is already easy to begin with, and should only get easier with practice, reducing the need to recruit iterative, cross-modal perceptual processes like sensorimotor simulation. The interaction between the facial mobility manipulation and participant gender, however, was unexpected. A cautious, post hoc interpretation of the current finding is that males, who generally 12 A. WOOD ET AL. perform worse on emotion recognition tasks (Donges et al., 2012; Korb et al., 2015; Montagne et al., 2005), experience cumulatively less emotion-related socialisation and their emotion recognition is therefore more vulnerable to disturbances, such as disruptions to the sensorimotor simulation process. Indeed, previous work has shown that extended pacifier use – which inhibits facial mobility – negatively predicts future emotional intelligence in boys but not girls (Niedenthal et al., 2012). The present male-specific effect should be interpreted with extreme caution and will be examined in future work. Finally, we did not find support for the prediction that incongruent hand gestures interfere with facial expression judgments even more strongly when facial mobility is restricted. While the context of the expresser’s body and the perceiver’s sensorimotor representation contribute separately to facial expression processing, perceivers do not increase reliance on the former when the latter is disrupted. Whether this null result is due to the generally weak effect of the facial mobility manipulation or to the hypothesis being incorrect is unclear. We have suggested previously that sensorimotor simulation will contribute more to emotion perception when facial expressions are subtle or difficult to interpret (Wood, Rychlowska, et al., 2016). Future work should therefore employ subtler facial displays to create a more difficult perceptual task that is more vulnerable to disruption. Future directions The current work involved novel, information-rich videos of actors producing all possible combinations of hand gestures and facial expressions. These stimuli could be used to address many research questions besides the ones we addressed here. For instance, how does ambiguous social feedback (a sad thumbsup) influence behaviour in a decision-making context? We have made the stimuli openly available (https:// osf.io/9xs48/) and we hope that other researchers will use them to address other research questions. In Study 2, we inhibited people’s ability to move their facial muscles, and we examined how this influenced their ability to process facial expressions. A parallel question is whether restricting hand movements would restrict participants’ ability to process gestures. We did not include a manipulation to restrict hand movements, because participants used their hands to respond with keypresses in the current work. A future study could use a vocal response format and could test whether restricting hand movements influences processing of hand gestures that convey valence, as in this study, or gestures that convey semantic information. Male and female participants in Study 1 differed in how strongly incongruent facial expression influenced their processing of gestures. This leads us to ask whether other individual difference factors might also matter. Some evidence suggests that people high in social anxiety fixate more on the hands of expressive bodies than people low on social anxiety, perhaps as a way of avoiding eye contact but nevertheless extracting information from an expressive body (Kret, Stekelenburg, de Gelder, & Roelofs, 2017). For socially anxious people, hand gestures may be particularly important sources of social feedback, an idea that could be explored using the current paradigm. Future work could also study users of signed languages, such as ASL, in which facial expressions can modulate the information conveyed by hands. Whereas participants in this study showed difficulty integrating “incongruent” signals, fluent signers use the face and body as interactive channels of communication (Wilbur, 2000). For instance, an inner eyebrow raise – usually associated with negative emotions like sadness – acts as an adverb that intensifies the meaning of a sign, even if the sign’s meaning itself is not negatively valenced. Future work should explore whether ASL users process the meanings of seemingly contradictory hand and face signals differently than non-ASL users. Future work should also explore whether valence-incongruent face and body signals sometimes combine to produce a new emergent signal in non-sign communication, as they can in ASL – for instance, a sad facial expression with a thumbs-up might convey a nuanced message not communicable with just a signal expressive channel. In conclusion, the present work explores the complementary questions of how perceivers extract meaning when faces and bodies communicate divergent messages, and whether the process of meaning extraction is affected when simulation processes are disrupted. Although research suggests that communicative information from the face and body is processed holistically, only limited research in facial expression perception has considered the effect of bodily signals that are inarguably learned (i.e. conventional gestures). Extending previous findings on combined face-body signals, the present work suggests the affective meaning of COGNITION AND EMOTION conventional gestures is processed holistically along with information from the face. Notes 1. What constitutes a “congruent” face-hand pairing depends on the salient feature dimension; in the current study participants focused on the valence of the facial expressions and hand gestures. The facial expressions and gestures might also be re-categorized according to other dimensions, such as approach-avoidance, resulting in a different set of face-hand stimuli pairs being considered “congruent” vs. “incongruent.” We operationalised congruency by valence, rather than by specific emotions (e.g. angry faces and angry gestures), because it is unclear what conventional gestures, if any, would convey the same specific meanings as discrete facial expressions. 2. A coding error that was not discovered until both studies were completed resulted in all participants repeating a block of gesture-attend trials at the end of the session. Since this occurred at the end of the session, we simply excluded these redundant trials from our analyses. Including them in the analyses did not change any of our conclusions. 3. The means and standard deviations reported are averaged across participants and are therefore not sensitive to within-subject changes in reaction times. This is why the effect can be statistically significant despite the means being so close together. Acknowledgments We thank the individuals who participated in this study, and we thank Magdalena Rychlowska, Nolan Lendved, Leah Schultz, Nathanael Smith, Crystal Hanson, Olivia Zhao, Emma Phillips, Holden Wegner, Alicia Waletzki, Mathias Hibbard, and Jay Graiziger for their help with stimuli creation and validation and data collection. We also thank John Lendved for filming and editing the video stimuli, and Mark Koranda for his useful insights on ASL. Disclosure statement No potential conflict of interest was reported by the authors. Funding A.W. and J.D.M. were supported by NIH Emotion Research Training Grant (T32MH018931-24). Further support for this research was provided by the Office of the Vice Chancellor for Research and Graduate Education at the University of Wisconsin– Madison with funding from the Wisconsin Alumni Research Foundation; National Institute of Mental Health. ORCID Adrienne Wood http://orcid.org/0000-0003-4773-4493 13 References Archer, D. (1997). Unspoken diversity: Cultural differences in gestures. Qualitative Sociology, 20(1), 79–105. doi:10.1023/ A:1024716331692 Atkinson, A. P., Tunstall, M. L., & Dittrich, W. H. (2007). Evidence for distinct contributions of form and motion information to the recognition of emotions from body gestures. Cognition, 104 (1), 59–72. Aviezer, H., Trope, Y., & Todorov, A. (2012a). Body cues, not facial expressions, discriminate between intense positive and negative emotions. Science, 338(6111), 1225–1229. Aviezer, H., Trope, Y., & Todorov, A. (2012b). Holistic person processing: Faces with bodies tell the whole story. Journal of Personality and Social Psychology, 103(1), 20–37. Barr, D. J., Levy, R., Scheepers, C., & Tily, H. J. (2013). Random effects structure for confirmatory hypothesis testing: Keep it maximal. Journal of Memory and Language, 68(3), 255–278. Bates, D., Mächler, M., Bolker, B., & Walker, S. (2015). Fitting linear mixed-effects models using lme4. Journal of Statistical Software, 67(1), 1–48. doi:10.18637/jss.v067.i01 Blakemore, S.-J., & Decety, J. (2001). From the perception of action to the understanding of intention. Nature Reviews Neuroscience, 2(8), 561–567. doi:10.1038/35086023 Brauer, M., & Curtin, J. J. (2018). Linear mixed-effects models and the analysis of nonindependent data: A unified framework to analyze categorical and continuous independent variables that vary within-subjects and/or within-items. Psychological Methods, 23(3), 389–411. doi:10.1037/met0000159 Burgoon, J. K., Buller, D. B., Hale, J. L., & de Turck, M. (1984). Relational messages associated with nonverbal behaviors. Human Communication Research, 10(3), 351–378. doi:10. 1111/j.1468-2958.1984.tb00023.x Carpenter, S. M., & Niedenthal, P. M. (under review). Inhibiting facial action increases risk taking. De Gelder, B. (2006). Towards the neurobiology of emotional body language. Nature Reviews Neuroscience, 7(3), 242–249. de Gelder, B., de Borst, A. W., & Watson, R. (2014). The perception of emotion in body expressions. Wiley Interdisciplinary Reviews: Cognitive Science, 6(2), 149–158. De Gelder, B., & Hadjikhani, N. (2006). Non-conscious recognition of emotional body language. Neuroreport, 17(6), 583–586. de Gelder, B., & Hortensius, R. (2014). The many faces of the emotional body. In J. Decety, & Y. Christen (Eds.), New frontiers in social neuroscience (pp. 153–164). Cham: Springer International Publishingdoi:10.1007/978-3-319-02904-7_9 Donges, U.-S., Kersting, A., & Suslow, T. (2012). Women’s greater ability to perceive happy facial emotion automatically: Gender differences in affective priming. PLoS ONE, 7(7), e41745. Eimer, M., & Holmes, A. (2002). An ERP study on the time course of emotional face processing. NeuroReport, 13(4), 427–431. Flaisch, T., Häcker, F., Renner, B., & Schupp, H. T. (2011). Emotion and the processing of symbolic gestures: An event-related brain potential study. Social Cognitive and Affective Neuroscience, 6(1), 109–118. doi:10.1093/scan/nsq022 Flaisch, T., Schupp, H. T., Renner, B., & Junghöfer, M. (2009). Neural systems of visual attention responding to emotional gestures. NeuroImage, 45(4), 1339–1346. doi:10.1016/j. neuroimage.2008.12.073 Green, P., & MacLeod, C. J. (2015). SIMR: An R package for power analysis of generalized linear mixed models by simulation. 14 A. WOOD ET AL. Methods in Ecology and Evolution, 7(4), 493–498. doi:10.1111/ 2041-210X.12504 Ipser, A., & Cook, R. (2016). Inducing a concurrent motor load reduces categorization precision for facial expressions. Journal of Experimental Psychology: Human Perception and Performance, 42(5), 706–718. Korb, S., Malsert, J., Rochas, V., Rihs, T. A., Rieger, S. W., Schwab, S., … Grandjean, D. (2015). Gender differences in the neural network of facial mimicry of smiles – an rTMS study. Cortex, 70, 101–114. doi:10.1016/j.cortex.2015.06.025 Kret, M. E., Roelofs, K., Stekelenburg, J., & de Gelder, B. (2013). Emotional signals from faces, bodies and scenes influence observers’ face expressions, fixations and pupil-size. Frontiers in Human Neuroscience, 7. doi:10.3389/fnhum.2013.00810 Kret, M. E., Stekelenburg, J. J., de Gelder, B., & Roelofs, K. (2017). From face to hand: Attentional bias towards expressive hands in social anxiety. Biological Psychology, 122, 42–50. doi:10.1016/j.biopsycho.2015.11.016 Kuznetsova, A., Brockhoff, P. B., & Christensen, R. H. B. (2017). lmerTest package: Tests in linear mixed effects models. Journal of Statistical Software, 82(13), 1–26. doi:10.1863 Martinez, L., Falvello, V. B., Aviezer, H., & Todorov, A. (2016). Contributions of facial expressions and body language to the rapid perception of dynamic emotions. Cognition and Emotion, 30(5), 939–952. doi:10.1080/02699931.2015.1035229 Meeren, H. K., van Heijnsbergen, C. C., & De Gelder, B. (2005). Rapid perceptual integration of facial expression and emotional body language. Proceedings of the National Academy of Sciences, 102(45), 16518–16523. Montagne, B., Kessels, R. P. C., Frigerio, E., de Haan, E. H. F., & Perrett, D. I. (2005). Sex differences in the perception of affective facial expressions: Do men really lack emotional sensitivity? Cognitive Processing, 6(2), 136–141. doi:10.1007/ s10339-005-0050-6 Moratti, S., Keil, A., & Miller, G. A. (2006). Fear but not awareness predicts enhanced sensory processing in fear conditioning. Psychophysiology, 43(2), 216–226. doi:10.1111/j.1464-8986. 2006.00386.x Niedenthal, P. M., Augustinova, M., Rychlowska, M., Droit-Volet, S., Zinner, L., Knafo, A., & Brauer, M. (2012). Negative relations between pacifier use and emotional competence. Basic and Applied Social Psychology, 34(5), 387–394. Niedenthal, P. M., Mermillod, M., Maringer, M., & Hess, U. (2010). The simulation of smiles (SIMS) model: Embodied simulation and the meaning of facial expression. Behavioral and Brain Sciences, 33(6), 417–433. Perry, A., Aviezer, H., Goldstein, P., Palgi, S., Klein, E., & ShamayTsoory, S. G. (2013). Face or body? Oxytocin improves perception of emotions from facial expressions in incongruent emotional body context. Psychoneuroendocrinology, 38(11), 2820–2825. doi:10.1016/j.psyneuen.2013.07.001 Poyo Solanas, M., Zhan, M., Vaessen, M., Hortensius, R., Engelen, T., & de Gelder, B. (2018). Looking at the face and seeing the whole body. Neural basis of combined face and body expressions. Social Cognitive and Affective Neuroscience, 13 (1), 135–144. doi:10.1093/scan/nsx130 Redcay, E., & Carlson, T. A. (2015). Rapid neural discrimination of communicative gestures. Social Cognitive and Affective Neuroscience, 10(4), 545–551. doi:10.1093/scan/nsu089 RStudio and Inc. (2014). rmarkdown: R Markdown Document Conversion, R package (Version 0.1.90). Retrieved from github.com/rstudio/rmarkdown. Rychlowska, M., Canadas, E., Wood, A., Krumhuber, E. G., Fischer, A., & Niedenthal, P. M. (2014). Blocking mimicry makes true and false smiles look the same. PLoS ONE, 9(3), e90876. Scherer, K. R., & Scherer, U. (2011). Assessing the ability to recognize facial and vocal expressions of emotion: Construction and validation of the emotion recognition index. Journal of Nonverbal Behavior, 35(4), 305–326. doi:10.1007/s10919-0110115-4 Tamietto, M., Castelli, L., Vighetti, S., Perozzo, P., Geminiani, G., Weiskrantz, L., & De Gelder, B. (2009). Unseen facial and bodily expressions trigger fast emotional reactions. Proceedings of the National Academy of Sciences, 106(42), 17661–17666. Van den Stock, J., & de Gelder, B. (2012). Emotional information in body and background hampers recognition memory for faces. Neurobiology of Learning and Memory, 97(3), 321–325. doi:10.1016/j.nlm.2012.01.007 Van Den Stock, J., & de Gelder, B. (2014). Face identity matching is influenced by emotions conveyed by face and body. Frontiers in Human Neuroscience, 8. doi:10.3389/fnhum.2014.00053 Wieser, M. J., Flaisch, T., & Pauli, P. (2014). Raised middle-finger: Electrocortical correlates of social conditioning with nonverbal affective gestures. PLoS ONE, 9(7), e102937. Wilbur, R. (2000). Phonological and prosodic layering of nonmanuals in American sign language. In K. Emmorey, & H. Lane (Eds.), The signs of language revisited: An anthology to honor Ursula Bellugi and Edward Klima (pp. 215–244). Mahwah, NJ: Lawrence Erlbaum Associates. Wood, A., Lupyan, G., Sherrin, S., & Niedenthal, P. (2016). Altering sensorimotor feedback disrupts visual discrimination of facial expressions. Psychonomic Bulletin & Review, 23(4), 1150–1156. doi:10.3758/s13423-015-0974-5 Wood, A., Rychlowska, M., Korb, S., & Niedenthal, P. (2016). Fashioning the face: Sensorimotor simulation contributes to facial expression recognition. Trends in Cognitive Sciences, 20 (3), 227–240. doi:10.1016/j.tics.2015.12.010