Emotion and its Management: The Lens of Language and Social Psychology

Forty Years of Emotion in JLSP

We reviewed all the research articles in JLSP since its first issue. First, we searched for research articles whose titles clearly referenced emotion or affect. Using an automatic text search, we found 38 such articles, out of a total of 245 research articles that contained emotion terms (words like emotion, affect, positive, negative, etc.). Second, we conducted a text-mining analysis of all emotion-related research articles with Leximancer (see Prologue, this issue, for more details about Leximancer), which revealed the main themes and concepts in the articles for each decade of JLSP’s existence.

Some clear patterns emerged from these analyses, which also reflect the changes in the larger field since the early 1980s. Emotion has become more prominent over the years: There was an increasing number of emotion-related papers, even given the larger number of papers that have appeared in JLSP since 2012. Of the 38 clear emotion titles, two of the papers came from the 1980s, both concerning the first main theme, affect and nonverbal communication. Eleven papers appeared in the 1990s. These continued the focus on nonverbal communication, adding three more themes: (2) status and power, (3) ethnicity and bilingual expression of emotion (including the equivalence of emotion words across languages), and (4) emotional communication across gender and in marriage. There were eight articles in the 2000s, continuing the theme of affect and marriage, but adding (5) intergroup political communication, (6) emotional discourse, and (7) negative emotions like remorse. Finally, 17 articles have appeared since 2010. They concern in particular negative emotions, including trauma, war and conflict, depression, grief, embarrassment, and taboo words. There is also a significant amount of research on affect expression (8) across languages, and (9) in new media. As a group, the articles also show an increasing focus on the intergroup communication of emotion over time.

The Leximancer analysis we conducted showed similar patterns. There was a larger number of papers as time went on, with most peak years for emotion articles occurring in this century (especially since 2011). Leximancer produces a concept thesaurus, which indicates the most prominent themes and the concepts associated with them. This analysis indicated that articles in the 1980s in JLSP concerned emotion perceptions, speakers, and attitudes, communicated via facial and vocal expression in interpersonal communication. The 1990s continued this focus on emotional messages in interpersonal contexts, but added positive emotions and positive affect expression. In the 2000s, there was more stress on negative as well as positive affect, along with a stronger emphasis on intergroup communication, particularly gender. Emotion studies have blossomed since 2010, as they have in psychology more generally (as shown in dedicated journals like Emotion, which started in 2001). In JLSP , there is also a stronger emphasis on emotional language and words, as well as the intergroup focus (especially on gender). The increasing emphasis on language and words in the past decade likely reflects the brief of the journal, which has become more strongly focused on language per se. These papers also reflect the advent of automated content analysis programs like Linguistic Inquiry and Word Count (LIWC: Pennebaker et al., 2015; see Boyd & Schwartz, 2021), Leximancer, and other text-mining and content-analysis software.

The long-standing interest in language and gender merits special note. Much of it concerns the gender-linked language effect (GLLE: e.g., Mulac, 1989; Mulac et al., 2000; Mulac & Lundell, 1986), which examines gender differences in language use, especially around language intensity and tentativeness, dynamism and aestheticism, use of particular grammatical structures (like adverbs), and so forth. Clearly, this work implicates emotion, and there have been a number of studies finding specific—and sometimes counter-stereotypical—gender differences in affective language (e.g., Mulac et al., 2013; Palomares, 2008, 2009). This work emphasizes intergroup (usually inter-gender) contexts, and fits well with the increasing prominence in JLSP of intergroup emotional language.

Language in Emotion Theory

Theories of emotion have been around for many years, although they have tended to be taxonomies (e.g., Frijda, 2007; Russell, 1980). These theories have elucidated the structure of emotion words and their cross-cultural universality or difference; this work has had a resurgence in recent years. Aside from these taxonomies and the work on nonverbal communication, the major theories of communication, like social identity/self-categorization theory (SIT: e.g., Turner et al., 1987) and even communication accommodation theory (CAT: e.g., Giles, 2016) have been surprisingly independent of the study of emotion. For example, SIT provides a sophisticated analysis of the dynamics of identity and identification, along with contextual influences on them. The impact of affect, however, has largely been left out of this process, even though the contexts often evoke strong emotion (e.g., intergroup conflict). Likewise, CAT has looked in detail at the motivations, antecedents, and consequences of (non)accommodation, which often implicates strong emotion. CAT has long contained a place for emotion (e.g., Williams et al., 1990), but emotion was not taken up by many studies testing CAT until this century. Since then, CAT has included emotional relations as a strategy (see Giles, 2016, for a review and examples). This has meant significant modification to these theories in recent years. For example, CAT now takes explicit account of emotion-based and relationship-based communication strategies, along with strategies of approximation, interpretability, topic management, and interpersonal control.

In the 1980s and 1990s, researchers began to examine the effects of emotion on cognitive processing. Forgas (1995), for example, proposed his Affect Infusion Model to provide a comprehensive explanation of such effects. Affect infusion was defined as the process “whereby affectively loaded information exerts an influence on and becomes incorporated into the judgmental process, entering into the judge’s deliberations and eventually coloring the judgmental outcome” (p. 39). According to the model, the more complex the situation (e.g., deciding whether or not to take drugs), the more the person will experience affect infusion. Although this, and other models like it, did not initially address the impact of emotion on the processing of language, they did lead to distinctions between positive and negative affect promoting assimilative (internally-focused) and accommodative (externally-focused) thinking strategies, respectively (Forgas, 2013). For example, when people were asked to produce verbal captions for cartoon images, participants in the positive mood condition produced more creative and humorous messages than those in the negative or neutral conditions (Forgas & Matovic, 2020).

More recently, Barrett (2017) proposed a Theory of Constructed Emotion, which argues against the notion that emotions are hard-wired and have any specific patterns of brain or physiological activity. This theory also argues that there are no “universal” emotions across people, nations, or cultures. Instead, the central premise of the theory is that emotions are concepts constructed by the brain and, therefore, highly dependent on language. This is based on embodied accounts of cognition (Barsalou, 2008). According to the theory, perceptions of hedonic valence and arousal give rise to emotions physiologically. The anger construct, for example, originates from bodily sensations that may have accompanied early experiences in which another person thwarted one’s goals. With repeated experiences and labelling by others (e.g., “you look angry”), one comes to develop the concept of anger and applies it in new, similar, situations. The appearance of universality across cultures reflects shared conceptualizations of emotion, because of common experiences as well as physiology.

Thus, one might predict that language groups in close proximity will share more emotion concepts than those of more distant groups. Indeed, in an analysis of 2474 spoken languages, “colexification” of emotion terms—when languages named related concepts with the same word—was predicted by geographic proximity (Jackson et al., 2019). Consistent with Barrett’s theory, these words were also predicted by the valence and physiological arousal associated with the term. One implication of constructivist theory is that if a person does not possess a particular emotion concept, the person will not perceive that emotion. For example, people with alexithymia (dysfunction in the ability to identify and describe emotions) typically lack words to describe their emotional experiences. Participants higher in alexithymia tend to judge an affective stimulus, such as negatively-valenced musical stimuli, as more neutral than those low in alexithymia (Hesse & Gibbons, 2019; Larwood et al., 2020).

In turning to methods, we found a discontinuity between the traditional methods for studying LSP and those for studying emotion. Researchers in LSP have relied on questionnaire and other self-report methods, along with observation of detailed language behavior from a sociolinguistic or social-psychological perspective. Research on emotion, however, has tended to use detailed measurement by outsiders to capture physiological or macro- and micro-behavioral expressions.

Emerging Methods of Studying Emotion

Emotional Language “in the Wild”

Much of the research investigating everyday emotional language in naturalistic settings (“in the wild”) has centered on emotion differentiation, also known as emotional granularity (Kashdan et al., 2015). Emotion differentiation involves putting feelings into words in a granular way by using fine-grained emotion labels (Kashdan et al., 2015; Smidt & Suvak, 2015). People low in differentiation may say they feel bad in general, endorsing many different emotion labels (e.g. feeling fearful, guilty, and sad all at once), which is reflected in a high correlation between the specific emotion labels they use. In contrast, high differentiators are likely to be more selective, endorsing some emotions but not others (e.g., feeling sad, but not guilty or fearful), reflected in a lower correlation between specific emotion labels.

Research on emotion differentiation has demonstrated that endorsing less differentiated words in everyday life is associated with a range of clinical disorders, including depression (e.g. Demiralp et al., 2012) and borderline personality disorder (Suvak et al., 2011). Differentiating between emotions also acts as a protective factor against negative outcomes, including binge drinking (Kashdan et al., 2010), self-injury (Zaki et al., 2013), and aggressive behavior (Pond et al., 2012). Differentiation may also confer interpersonal benefits: People who differentiated more between their negative emotions were more accurate in inferring their romantic partners’ emotions in everyday life (Erbas et al., 2016). Finally, the act of completing an experience sampling protocol may improve the ability to differentiate between emotions. Patients with depression demonstrated increased emotion differentiation after completing an experience sampling protocol (Widdershoven et al., 2019). Taken together, the research suggests that differentiating between emotion labels in everyday life is generally beneficial.

Why might this be so? Research has suggested that the simple act of putting our feelings into words can reduce negative emotional experience (Torre & Lieberman, 2018). Labeling our emotions in a granular way may be helpful because it is useful in regulating our everyday emotional experiences. Congruently, recent research has shown via using experience sampling methods that low differentiation may hinder successful emotion regulation (Kalokerinos et al., 2019).

There is also research examining language and emotions using passive sensing methods during everyday conversations. One study had participants wear an electronically activated recorder to record their conversations while they completed a self-report experience sampling protocol assessing emotional experience (Sun et al., 2020). Language in the passive recordings was coded using LIWC. The researchers found that LIWC emotion scores were not significantly associated with the corresponding self-reported emotional experience assessed using ESM. These findings suggest the need to interpret passive measures of emotional language with caution. These measures are probably not a direct index of how emotions are experienced by the person, but rather reflect more outside social and expressive dimensions of emotion.

Nonverbal Language and Emotion

As noted above, emotion researchers have also been interested in nonverbal language. Our faces and bodies convey a great deal of information about our underlying states to others. Although there may not be hardwired universals, as originally proposed by Ekman and others (for a review, see Nelson & Russell, 2013), there is a set of postures and facial expressions whose meaning will be clear to a particular cultural group. Humans are certainly not clueless when they are asked to infer another’s emotions based on facial and bodily cues. Happy facial expressions are the most recognized (Russell & Barrett, 1999), although they do not always reflect the speaker’s actual feelings of happiness. For example, humans have the ability to distinguish between genuine and posed smiles, perhaps because smiles are critical to interpersonal relationships (Williams et al., 2001). For example, genuine smiles of enjoyment are rated higher for cooperation and trust by perceivers than are posed smiles (Johnston et al., 2010; Vanman et al., 1997). Moreover, perceivers are more likely to attend to smile types, particularly in contexts involving trust (Centorrino et al., 2010; Krumhuber et al., 2007). Thus, to be effective in social situations, people often rely on their ability to distinguish the deceptive behavior (i.e., a posed or fake smile) from the genuine one (e.g., see Levine, 2018, for a recent review of this literature). To correctly distinguish between genuine and posed smiles, paying attention to the eye regions may be essential. Contraction of the orbicularis oculi muscle (which raises the cheek and results in crow’s feet around the eyes) has been proposed to be a marker of the genuine enjoyment smile (sometimes called the Duchenne smile: Ekman, 1989).

This highlights how facial expressions convey much more than a small set of discrete emotions. Some have argued that facial behavior is a better term, because these expressions are not simply readouts of emotion (Fridlund, 1994; Seyfarth & Cheney, 2003). In naturalistic studies, people rarely show the prototypical emotional expressions, but instead express multiple cues that are imprecise and context-dependent (Crivelli et al., 2016; Fernández-Dols & Crivelli, 2013).

Another source of information is the body. In fact, interpretation of facial expressions is influenced by body expression (Kret & de Gelder, 2013; Nelson & Mondloch, 2017). For example, when participants viewed images of emotionally congruent and incongruent face-body pairs, threatening cues were looked at longer than happy cues, regardless of whether they appeared in the face or the body, and whole-body expressions were influenced by the surrounding social scene (Kret et al., 2013). Like facial expressions, there appears not to be a set of communicative universals for the body. Instead, bodily expressions are also imprecise cues that are interpreted with context and ongoing facial expressions. Emotional meaning is thus constructed from multiple cues.

Finally, the voice carries a wealth of information about the speaker’s affect, as well as other aspects of the speaker (Kappas et al., 1991; Pittam, 1993). Vocal features (prosody, tone of voice, and paralanguage) enable listeners to identify the speaker’s gender, age, ethnic group, and status. In addition, listeners are also able to determine the speaker’s emotional state, whether presented in a few sentences or in a single shout or cry. As the reviews above indicate, a large body of research on emotion and the voice extends over many years, much of it describing the soundwaves conveying emotional information. More recently, specific brain regions have been identified as differentially involved when adults listen to vocal expressions of anger or happiness (Johnstone et al., 2006) or when infants are presented emotionally positive, neutral, or negative vocalizations (Blasi et al., 2011). Furthermore, algorithms and machine learning have been used to recognize emotion in voices at call centers and by social robots.

In sum, these multiple nonverbal channels provide multimodal convergence of emotion information that is integrated into a holistic, more reliable percept of the other person’s emotional state (Schirmer & Adolphs, 2017). Words are not required for such inferences, and it is likely these channels of emotional communication predate the evolution of human language, as they appear to be shared across other mammals—a point Darwin stressed in The Expression of the Emotions in Man and Animals (1872). However, nonverbal emotional signals may not convey much more information than the valence and intensity of the sender’s affective state. How the listener categorizes the information is most likely influenced by one’s language (e.g., “he’s jealous!”). This research highlights the need for integrated multi-channel studies of emotional communication, as well as the importance of integrating work on nonverbal behavior with language.

Conclusion: Language and Emotional Communication in Future Research

We have highlighted ways in which the social-psychological study of emotion resonates with the research done in LSP and with the long history of JLSP . We have canvassed themes around affect and emotional language and communication over the years, and looked at some new trends in theory and methodology with potential to bring emotion to the center of LSP. We conclude by making some modest suggestions about how researchers might approach this task.

Emotion is ubiquitous in relationships, and researchers have studied the ways in which emotion is communicated in real-world settings like health, the workplace, the courts, or interactions between police and citizens. These interactions are intergroup, and often involve the expression of negative emotions around conflict, rivalry, and aggression. They also frequently involve great anxiety; the emotional stakes are higher in ongoing relationships.

For example, Hewett et al. (2009) looked at how doctors express conflict across specialties, where groups that are invisible to outsiders can be sources of great tension to insiders. Watson et al. (2016), reviewing communication accommodation in health settings, noted that CAT’s emotional support strategy was mainly developed from communication between patients and health professionals. This context can produce even greater anxiety when the patient or health professional is speaking a second language—no one wants to be ill in a second language (Zhao et al., 2019). It would be hard to find a context where emotion is more important than health, and the potential for further research here is very large.

Likewise, there has been extensive work (see Gnisci et al., 2016) on the impact of (non)accommodative (friendly to hostile) communication between police and citizens, especially in the fraught and dangerous context of driving stops (see Lowrey-Kinberg, in press ). This is probably the most difficult of all workplace settings, as lives are literally at stake. Watson and Soliz (2019) provide a review of accommodation, including the communication of emotion, in institutional settings, noting the similarities between settings, but also highlighting the need to study each context in great depth. Once again, the potential for research here is large, and LSP has barely begun to tap it.

One key direction for future research on language and emotion is to turn more often to ongoing real-world settings. Much research to date has been conducted in the laboratory, which has been essential for revealing the structure of emotional expression and communication. As our section on emotion in the wild shows, however, things are changing, and there is increasing interest in ongoing relationships and applied settings. To do this, it will be important to combine the study of language with nonverbal behavior. For at least 40 years, researchers have called for more attention to the whole package of emotional expression, including the ways in which various channels do—or do not—bring redundancy to a message. Nevertheless, there is only a small minority of studies which do this, and vanishingly few in applied contexts or with ongoing relationships. This is a difficult task, but it is well worth undertaking.

Beyond this, the study of emotion in new media and social media (often around health and welfare) is a burgeoning area. For example, Rains (2015) looked at the perception of support for health issues from the affective words used in specialized blogs and computer-mediated fora. Similarly, Seabrook et al. (2018) examined the use of affective words on Twitter and its association with depression. Recently, Vine et al. (2020) examined emotion vocabulary in written essays (to determine psychometric soundness) and public blogs; they found that more negative emotion vocabulary was associated with several markers of lower psychological adjustment and well-being. These are word-based media (barring emojis), which determines the linguistic focus of this work. Furthermore, Facebook, Twitter, and blog users are not a representative population sample, but they are still worth close study. There is also a major opportunity for researchers to look at more complex intergroup contexts, which is where emotion often turns to open hostility and conflict. We must do this if we are to make significant inroads into solving the wicked problems around conflict and cultural deficiencies, not to mention the fraught interplay of identities in interaction.

A second opportunity is to make more use of existing LSP theory, and to extend it to include explicit propositions about emotional communication. While a few theories (like CAT) have begun to do this, the majority have left emotion as a black box in a social context, with the focus on larger intergroup or interpersonal variables. We hope that in the next decades, all major theories in this area will include emotional communication at their core.

The study of emotion is now well-consolidated in social psychology and communication. The moment has come to include emotion in most research, to take it into the field, and to look at the complexity with which it is played out in all our interactions. Just as emotions can be and often are communicative, that act of communication can be and often is emotional. The study of the one is incomplete without an understanding of the other, and language and social psychology as a field is incomplete without a detailed understanding of language and emotion.