Most people have a keen awareness of how others express their feelings, particularly for negative emotions (Rozin et al.,
2005). When a friend seems sad (Yoo & Noyes,
2016), or when a colleague gets angry (Tiedens,
2001), we are often able to infer how they feel. Yet our current understanding of how positive emotions are expressed is much less developed, because work tends to either focus on one or two positive emotions (e.g., Tracy & Robins,
2007; Vritcka et al.,
2014), or to examine emotions within a particular modality or two (e.g., Dael et al.,
2012a,
2012b; Ekman et al.,
1987; Laukka,
2005). However, theoretical accounts suggest that positive emotions would be expected to be expressed in ways that are congruent with their functions (Kitayama et al.,
2006; Sauter,
2017). The goal of this paper is to provide an empirical test of these theoretically informed predictions.
Two Frameworks of Positive Emotions
The interest in understanding positive emotions is growing with advances in the fields of affective science (Keltner,
2019), and social psychology (Fredrickson,
2001). Previous theorizing and research in which negative and positive emotions were compared has mainly focused on joy (or happiness), which is conceived as a singular emotion argued to be primarily expressed using the face (e.g., Ekman,
1992). However, recent theorizing posits that there are multiple discrete positive emotions (see Shiota et al.,
2017) that are expressed through channels that go beyond the face, including postural cues (Dael et al.,
2012a,
2012b), the use of words (Campos et al.,
2013), displays via touch (Schirmer & Adolphs,
2017), and vocalizations (Sauter & Scott,
2007).
Several accounts have sought to map out how different positive emotions relate to one another and what functions they serve. Whereas none of these accounts yield straightforward predictions of how different positive emotions are expressed, we drew on this work to inform our selection of emotions as well as in the formation of exploratory initial hypotheses. Given the long-running debate in the field with regard to the universality of emotion (e.g., Russell,
1994, and reply by Ekman,
1994), we selected two frameworks that each stem from a different research tradition: the emotion families approach (Sauter,
2017) and the arousal-engagement matrix (Kitayama et al.,
2006; Tsai et al.,
2006). Emotion families are based on evolutionary functions that are shared across human beings (Tooby & Cosmides,
2008), with a focus on specialised functions for different emotions. The arousal-engagement framework emphasises cross-cultural differences (Boiger & Mesquita,
2012), and posits that emotions can be classified according to underlying dimensions. Although these taxonomies are inspired by divergent lines of work, they are not mutually exclusive, because they each focus on different elements and functions. We first outline these frameworks and the predictions we derived from them, and next specify the modality-specific hypotheses for each of the four positive emotions that are the focus of the present paper.
The emotion families approach (Sauter,
2017) postulates that positive emotions can tentatively be classified into one of four clusters, which diverge in terms of evolutionary functions (for similar suggestions, see Shaver et al.,
1987, and App et al.,
2011). This classification scheme is informed by work relating to expressions (e.g., Simon-Thomas et al.,
2009) and appraisals (e.g., Roseman,
1996). The modalities in which different positive emotions are expressed is thought to map onto the purported function of the cluster in which a given emotion is thought to belong.
Epistemological emotions (e.g.,
interest) involve a change or shift in one’s knowledge state. Congruent with the cognitive nature of such emotions, expressing an epistemological emotion would be likely to involve the use of higher order cognitive processes, such as language (LeDoux & Brown,
2017).
Prosocial emotions (e.g.,
gratitude) focus people towards the welfare of others and help foster social relationships. Expressing prosocial emotions should hence involve affiliative gestures that bring people closer together, which can be expressed through physical contact or verbal signals.
Savoring emotions (e.g.,
feeling moved) stem from experiencing enjoyable, calming, or heartwarming stimuli. Given that such emotions are purportedly linked to unconditioned stimuli that make one feel good, expressions would be likely to include smiling. Finally,
agency-approach emotions (e.g.,
triumph) are characterized by approach tendencies towards potential rewards. Given the need to explicitly display movement towards an objective, such emotions may be likely to involve bodily movement.
In the arousal-engagement matrix, we combine the complementary dimensions of physiological arousal (Tsai et al.,
2006) and social engagement (Kitayama, et al.,
2006), such that positive emotions could theoretically fit into one of four potential quadrants. Arousal refers to the extent to which an emotion elicits a heightened physiological response in an individual (Larsen & Diener,
1992), while engagement denotes the degree to which an emotion is socially oriented and thereby brings people closer together (Kitayama et al.,
2000). Based on this framework,
high arousal emotions may involve the recruitment of multiple nonverbal modalities that act as amplifiers to convey the heightened arousal an expresser feels. In contrast,
low arousal emotions might involve less nonverbal signals, and a greater reliance on words to express how one feels instead.
Socially engaging emotions should be likely to involve the use of signals that clearly display affiliative motives—such as smiling, touch, and words—to bring people closer together and to minimize interpersonal misunderstandings. Conversely,
socially disengaging emotions, which increase social distance between oneself and others, should make less use of the abovementioned channels.
Based on these theoretical taxonomies, we selected four positive emotions that varied in terms of emotion family, arousal, and social engagement: feeling moved, gratitude, interest, and triumph. Table
1 lists the four positive emotions we selected in relation to the theoretical frameworks. We also added the modalities through which we expected each emotion to be expressed. These modalities are based on both the characteristics and functions of each of these emotions, according to the theoretical frameworks outlined above, as well as a review of the empirical literature on nonverbal expressions. In the ensuing section, we lay out the rationale for our predictions.
Table 1
List of positive emotions with corresponding categories based on emotion family, physiological arousal, and social engagement, as well as predicted expressive modalities
Feeling moved | Savouring | High | Other focused | Face |
Gratitude | Prosocial | Low | Other focused | Words, voice |
Interest | Epistemological | Low | Malleable | Words, face, voice |
Triumph | Agency-approach | High | Self focused | Body, face, voice |
Nonverbal Expressions of Positive Emotions
A growing body of empirical work has shown that some specific positive emotions have uniquely distinguishable expressive signals (Keltner & Cordaro,
2017), occurring across multiple channels (Cowen et al.,
2019; Elfenbein et al.,
2007). In production studies, researchers typically examine how people express positive emotions using one or two modalities (most commonly the face, and to a lesser extent voice). Expressions can be produced spontaneously in ecologically valid scenarios (Tracy & Matsumoto,
2008), with poses in controlled laboratory settings (Cordaro et al.,
2018), or elicited with stimuli that are meant to induce specific emotions (Levenson et al.,
1991). The expressive cues, such as facial muscle movements or vocalizations, are then empirically mapped using established coding schemes (Dael et al.,
2012a,
2012b; Ekman et al.,
2002) and automatic detection software using algorithms (Krumhuber et al.,
2019; Schuller & Schuller,
2020). A broad range of positive emotions have been found to have specific nonverbal cues (for a review, see Sauter,
2017). For example, amusement is expressed with an open-jaw smile (Ambadar et al.,
2009) and vocalizations with multiple amplitude onsets (Sauter et al.,
2010), while awe is signaled with raised eyebrows, eye-widening, parted lips (Campos et al.,
2013), and visible inhalations (Shiota et al.,
2003).
Yet, little research has systematically compared across modalities (see Kessous et al.,
2010, and Schirmer & Adolphs,
2017). Our present understanding of positive emotional expressions is based on scattered bodies of work, and many modalities have yet to be considered in relation to specific positive emotions. The empirical findings we review, and our hypotheses for the present research, are consequently constrained by the limitations of this literature. With regards to our four selected positive emotions (
feeling moved,
gratitude,
interest,
triumph), some empirical evidence has been put forth that allows predictions to be made about the non-verbal cues used to express each of them. We introduce each of these emotions in turn, and outline the predictions we made for each, along with theoretical and empirical rationales for our hypotheses. In addition, where applicable, we also distinguish between signals and signs (see Sauter & Russell,
2020). Signals are expressive behaviors that are inherently communicative; they purposefully (though not necessarily consciously) convey social information to observers (Fridlund,
1994). In contrast, signs are expressions that are simply “given off” (Goffman,
1959); neither communicative functions nor intentions are necessary components of signs.
Feeling moved is an intense emotion that is triggered when partaking in (or observing) communal sharing relationships, such as an unexpected reunion (Cova & Deonna,
2014). This emotion is thought to be mixed in terms of valence (e.g., bittersweet), in that mostly positive but also negative feelings are elicited when feeling moved (Menninghaus et al.,
2015; Vuoskoski & Eerola,
2017). Possibly as a reflection of this negative valence, some empirical evidence points to increased muscular activity around the eyes (e.g., corrugator) when people feel moved (Wassiliwizky et al.,
2017a,
2017b; but see Zickfeld et al.,
2020). Robust cross-cultural evidence also suggests that feeling moved is displayed and identified via tear droplets and crying (Schubert et al.,
2018; Seibt et al.,
2018; Zickfeld et al.,
2019b). While other modalities have also been suggested to be involved in the expression of feeling moved, such as vocalisations and hand gestures (Fiske et al.,
2019; Zickfeld et al.,
2019a), these modalities have not consistently been found across studies. The most robust empirical evidence has been found for tears.
In terms of the arousal-engagement matrix, feeling moved is a high-arousal high-engagement positive emotion. Clear communicative signals are to be expected for conveying feeling moved in order to bring others closer to the expresser. This is consistent with empirical work indicating that tears signal a desire for help, and trigger approach orientations in others (Gracanin et al.,
2018). In the emotion family framework, being moved is considered a savoring positive emotion. Savoring emotions typically have quite clear displays on the face (Sauter,
2017). In sum, both theoretical and empirical evidence led us to postulate that the face should be the main modality through which feeling moved is expressed.
Gratitude is defined as a feeling of thankfulness and indebtedness for the positive actions and contributions of another person or group (Algoe,
2012). Much of the work on gratitude has focused on verbal rather than nonverbal expressions (Algoe & Haidt,
2009; Algoe et al.,
2013), with research showing that when indebted, people express being thankful with words of appreciation (Williams & Bartlett,
2015). There are also findings pointing to gratitude being expressed via handshakes, meaning that touch could be employed in some contexts (Hertenstein et al.,
2009). A clear facial expression for gratitude has however yet to be surfaced (Campos et al.,
2013), and given the scant attention paid to vocalisations of gratitude (see also Yoshimura & Berzins,
2017), it remains theoretically possible that people may reduce the volume of their voice when they feel grateful.
Gratitude is a low-arousal high-engagement positive emotion. It is thus similar to feeling moved in terms of triggering social orientations. However, gratitude displays would be expected to be less intense in nature, reflecting the comparatively lower state of arousal of the expresser. In term of emotion families, gratitude is a prosocial emotion, which entails communication via touch in order to stimulate closeness and social attachment. This aligns with empirical work as described above (App et al.,
2011). However, expressing gratitude through touch (e.g., clasping someone’s hand) also intrudes on the physical space of the perceiver, and so would only be appropriate to use between people who already have a close relationship with one another (Lee & Guerrero,
2001). Taking theory and the contextual nature of empirical findings into account, we therefore expected only words and the voice to be used for expressing gratitude.
Interest is conceptualised as a feeling that arises when new and relevant stimuli are encountered in the environment (Silvia,
2008). It is a multi-faceted emotion, such that expression modalities may depend on the domain of interest (Silvia,
2005). Past work has demonstrated interest to be expressed using a quickened speech rate, and a widened vocal frequency range (Banse & Scherer,
1996). The use of phrases and questions also signal a keenness to engage (Silvia,
2008), indicating that both the voice and words could be key communicative channels for expressing interest. For the facial expression of interest, both signals (constriction of the eyebrows to communicate a deep concentration on a specific topic: Campos et al.,
2013; smiling: Mortillaro et al.,
2011) and signs (parted lips: Reeve,
1993; widened eyes: Shiota et al.,
2003) are postulated to be to be involved. A case has also been made for a body movement sign associated with interest, namely, leaning forward (Dael et al.,
2012a,
2012b; Dukes et al.,
2017).
Interest is thought to be a low-arousal emotion. This would suggest that interest is
not likely to be expressed using the full range of body movements, which would be more likely to characterise emotions that are more intense. In terms of engagement, interest is thought to be malleable; being interested draws one’s attention inwards towards a specific topic (Sung & Yih,
2016), but at the same time, one could also share their interest with others (Yoon et al.,
2012; Rime,
2009). The epistemological nature of interest points to the potential use of higher order cognitive processes like language, and so words could be an important communicative tool for expressing interest. Given that a direct comparison between modalities has yet to be made for interest, we postulated the use of modalities that both empirical work and theorising concur with. As such, words, the face, and voice, were all expected to be used to express interest.
Triumph is the feeling elicited upon victory (Matsumoto & Hwang,
2012). Empirical work has suggested that in the immediate aftermath of winning, specific signals emerge: people adopt a straightened body posture with the chest protruding, and make guttural sounds signalling their victory (Tracy & Matsumoto,
2008). There is also evidence from two divergent cultural samples examining athletes immediately post victory, which found that winners tended to show open mouthed smiles and upward head tilting (Hwang & Matsumoto,
2014). Even in work suggesting that victory displays are regulated so as to spare perceivers’ feelings, an expansive body posture comes through as a partially suppressed signal related to winning (Van Osch et al.,
2019).
From a theoretical standpoint, triumph is deemed a high-arousal low-engagement emotion. Given high arousal, its expressions should involve the recruitment of multiple expressive channels to demonstrate a larger social presence via expansiveness of the self (App et al.,
2011). The dominance focused nature of triumph signals highlights the socially disengaging nature of this emotion (Kalokerinos et al.,
2014), and suggests that multiple channels should be involved in its expression. Similar expressive patterns are expected based on triumph being an agency-approach emotion. To signal one’s agency and possibly even power, triumph should involve body postures and head movements that seek to make the self look taller and larger (Hwang et al.,
2016). Both theory and empirical work thus point to the hypothesis that triumph should be expressed via body movements, facial expressions, and the voice.
Emotion Expressions Across Cultures
A major caveat to the patterns of expression described above is that the degree to which they generalize to people from most parts of the world is at present unestablished. Much of the empirical research on emotional expressions has thus far been conducted with North American samples (see Keltner & Cordaro,
2017), and even genuine attempts at cross-cultural work are often hampered by the fact that most respondents are university students. Student samples share many common characteristics regardless of which part of the world they come from (Henrich et al.,
2010) and are typically not representative of the populations they are thought to reflect. While some researchers have recruited community samples to draw a contrast between modernized and culturally isolated populations, such lines of work have tended to either focus on one specific positive emotion (e.g., Tracy & Robins,
2008), or a particular modality in isolation (e.g., Ekman et al.,
1987; Sauter et al.,
2010). The need to study community samples across a broad range of cultures, while examining various modalities of expression for different positive emotions, remains a worthwhile endeavour for the advancement of theory in affective science.
For the present research, we adopted an established method from cross-cultural psychology to examine self-reported expressions of positive emotions: the intersubjective approach (Chiu et al.,
2010). When applied to emotion expressions, the intersubjective approach entails asking people how
other members of their culture in general express specific positive emotions (
feeling moved,
gratitude,
interest,
triumph), with commonly used modalities presented as response options (
face,
voice,
body,
touch,
words).
Self-report instruments provide many advantages for cross-cultural work, especially when contrasted with classic methods from production studies. Many of the existing tools used in expression research are not only costly and time intensive, but also involve the practical constraints associated with setting up on-site data collection in divergent cultural contexts. As a means of navigating such difficulties, other researchers too have employed variants of self-report: respondents choose from photographs that depict various stages of an expression (Matsumoto,
1990; Van Osch et al.,
2019), indicate on a mannequin or avatar which parts of the body are most implicated when an emotion is felt (Nummenmaa et al.,
2014; Van Cappellen & Edwards,
2020), or recollect emotional experiences and textually describe all components they associate with each specific emotion, including expressions (Campos et al.,
2013). These self-report formats facilitate cross-cultural research, because surveys can be conducted wholly online, and thus be used to reach community samples with relative ease and at lower cost than traditional lab studies.
The intersubjective approach deals with two common methodological concerns associated with self-reported emotion expressions: the expectation that participants have accurate insight into how they express different emotions, and the assumption that aggregated individual responses can be taken as a valid reflection of a cultural group’s response patterns. By using the intersubjective approach, we measure expressions from the perspective of the perceiver, thereby capturing estimates of expressions from commonly encountered targets in everyday life. Such items mitigate the need for introspective evaluation by shifting the focus away from one’s own behaviors to observed displays in others. Furthermore, by presenting categorical response options with clear and familiar labels, this approach eliminates culturally-dependent response biases that may occur on Likert-type scales (Van Herk et al.,
2004). Because this approach is based on the principle of shared consensus (Wan et al.,
2010), culture-level conclusions are drawn based on agreement between raters, rather than simply assuming a given sample’s demographic features and response patterns to be representative of a larger nation unit.
In sum, existing theorizing and empirical work suggest that positive emotions are nonverbally expressed through several different channels, yet to date studies have tended to examine only singular (or maximally two) modalities at a time. Here, we sought to provide a first step towards systematically mapping out which modalities are associated with particular positive emotions across countries that differ on cultural values (Minkov & Hofstede,
2012; Schwartz et al.,
2001).