Cross-Cultural Calibration of Words and Emotions: Referential,Constructionist,and Pragmatic Perspectives

Investigating linguistic distinctions

If different “emotion” words sometimes refer to the same or similar emotions but also carry implications about extrinsic nonemotional considerations (such as purity, duration, intensity, cause, context, regulation, evaluation of the emotional person, normative appropriateness of the response, and so on), how might researchers determine which words pick out genuinely distinct emotions in the cleanest possible way? Various methods are available to help with this task, many of which rely on informants’ perceptions of semantic relations. For example, Shaver et al. (1987) collected 213 words collated from various lists of “emotions” and asked 112 US undergraduate psychology students to rate how good an example of the general “emotion” category each of those words was. Based on this procedure, the researchers selected a smaller set of 135 words representing relatively better examples of emotion. Next, another 100 psychology students sorted cards showing these words into piles containing words with similar meanings. The researchers then computed semantic similarity scores based on how many students put each pair of emotions in the same pile and derived a hierarchical classification from these scores. They concluded that the most basic level of the conceptual hierarchy providing the clearest set of distinctions divided the emotion words into six categories. Labels for these categories (determined by identifying the words which participants included in corresponding piles most frequently, and excluded from these piles least frequently) were “affection,” “happiness,” “depression,” “anger,” “fear,” and “amazement.” However, the researchers noted that the sixth category, which contained only three items, was potentially an artifact of relaxing the good-example inclusion criterion at the initial word-screening stage and leaving “surprise” in the list only because prior researchers had considered its referent to be a universal basic emotion for different reasons (e.g., Ekman, 1972, and see below).

Shaver et al.’s (1987) study provides evidence about how US undergraduate students taking psychology courses distinguish the meanings of English-language “emotion” nouns, thereby offering a provisional answer to the question of which of these words pick out distinct concepts. If participants’ similarity judgments were systematically informed by their shared understanding of the specifically emotional meanings conveyed by the presented words (implicitly factoring out any emotionally extrinsic connotations), then the derived clusters may help to clarify which emotions are cleanly distinguished in this part of the Anglophone world.

However, even then the findings do not tell us whether “emotion” words fall into similar categories in other languages and across other cultural groups. Although Shaver et al. (2001) reported similar clusters of Indonesian “emotion” words, more radical differences have been reported in societies that are less similar to North American ones (including those with languages without any word for the general category of “emotion,” e.g., Lutz, 1982). A recent study by Jackson et al. (2019) directly compared a much wider range (2474) of languages using available data about colexification—the use of the same word to denote semantically related concepts (e.g., “upset” to denote both “anger” and “distress”). The investigators found that languages used in geographically closer societies had relatively more similar patterns of colexification of “emotion” terms. However, the only consistent colexificatory tendencies across all sampled languages corresponded to distinctions between positive and negative emotions and between emotions involving high and low levels of arousal (see also Russell, 1983; Russell et al., 1989). In other words, languages do not usually colexify positive and negative emotions (e.g., by using a single word representing both “anger” and “joy”) or high-arousal and low-arousal emotions (e.g., by using a single word representing both “calmness” and “excitement”). However, several languages do colexify pairs of specific emotion categories that English speakers represent as distinct from one another (e.g., the word-form dard in the Sirkhi dialect of Dargwa colexifies “anxiety” and “grief”).

One possible implication is that speakers of different languages experience different sets of distinct emotions. Alternatively, it may be that all humans experience pleasure and arousal but specific emotion categories are culture-specific linguistic artifacts (e.g., Russell, 1991, and see below). Despite growing challenges to their universalist assumptions, however, a number of contemporary psychologists still work from the assumption that some human-wide discrete emotions genuinely exist but get lumped together and formulated in different ways by different languages (e.g., Keltner et al., 2019). A plausible reason for such cross-language differences is that different emotional distinctions are differentially useful and relevant in different cultural and ecological contexts. For example, being able to deploy finer-grained concepts representing subcategories of anger-like emotion might serve important functions in societies where there is greater variety in the actions and goals pursued by different individuals and groups, leading to wider diversity in interpersonal and intergroup conflicts. An emotion (or set of emotions) with greater cultural significance may become hypercognized (Levy, 1973) so that a richer vocabulary develops around it. Correspondingly, emotions with less cultural significance may be hypocognized and consequently end up underrepresented in a society's lexicon.

Investigating nonverbal distinctions

Researchers seeking to develop a neutral scientific language for distinguishing putative universal emotions need to find a way of circumventing local cultural factors that have shaped their linguistic representation. This usually involves taking account of nonlinguistic criteria when formulating distinctions. For example, Ekman et al. (1972) arrived at their initial list of discrete “basic” emotions partly on the basis of evidence about perceived relations between linguistically marked Anglo-American emotion categories and photographs of facial expressions. Words were retained as names for basic emotions if participants viewing the photographs consistently selected the predicted verbal label. This procedure resulted in the inclusion of the questionably emotional word “surprise” (Ortony, 2022, and see above) because a distinctive facial expression was consistently perceived as indicating its presence, but the omission of some other candidate “emotions” initially proposed by the researchers (e.g., “shame” and “contempt”) because the available facial pictures were not allocated to the predicted category reliably enough by participants of both genders (Ellsworth, 2014).

In fact, the use of facial markers to distinguish emotion categories brings similar problems to the use of linguistic data. Like words, facial expressions may only acquire consensual emotional meanings when the emotions that they denote have sufficient cultural significance to merit symbolic representation. When people use these facial expressions to convey emotion-related meanings, other emotionally extrinsic connotations may be communicated too. This means that we can't count on cross-cultural consistency in the emotional meanings of faces either. Indeed, current evidence suggests that perceived associations between preselected photographs of many “emotion” expressions and “basic emotion” labels decrease as researchers sample populations which are increasingly remote from Anglophone societies (e.g., Crivelli et al., 2017; Gendron et al., 2018; Nelson & Russell, 2013).

If facial expressions don't provide markers of putatively universal emotions, where else can researchers look to ground their concepts? Perhaps other measurable emotion-related variables that are less directly susceptible to normative influences or semantic indeterminacy might provide a more solid basis for establishing which words are associated consistently with distinct emotional meanings. However, patterns of neither autonomic (e.g., Siegel et al., 2018) nor central nervous system activity (e.g., Barrett & Wager, 2006; Lindquist et al., 2012) seem to correlate strongly with the induced or observed states associated with English-language labels that supposedly represent basic emotions. Pinning down a consistent referent for these words presents continuing problems.

Reformulating linguistic distinctions

Where does the apparent variability in response profiles leave those researchers who remain committed to the identification of universal and discrete basic emotions? Scarantino (2015, 2018) has argued that a scientifically coherent emotion category need not pick out a consistently distinctive pattern of bodily responses. The integrity of the associated concept instead derives from the single underlying psychological process that is activated whenever the putative emotion occurs. For example, formulating emotions as modes of action readiness (e.g., Frijda, 1986) implies that the particular way that each one prepares the body and mind to address the same kind of functional concern differs depending on the specific object at which the readied action is oriented and the specific context in which that object is confronted. Indeed, many theorists believe that a crucial adaptive benefit of emotional processes is precisely that they are more flexible than reflexes or fixed action patterns, thereby permitting relatively fast but context-sensitive adjustments (e.g., Smith & Ellsworth, 1987). Thus, a consistent and distinctive emotion of “anger” may involve different patterns of neural and bodily activity when the goal blockage that it is designed to push against comes from an antagonist standing directly in front of the angry person rather than a recalcitrant operator at the other end of a phone line. Attributing the range of situationally sensitive response profiles to a common causal process provides an alternative basis for establishing the distinctness of the underlying emotion.

Scarantino (2015) acknowledges that even adopting this criterion may not be enough to rehabilitate the status of English-language words such as “anger,” “fear,” and “disgust” as names for discrete “basic” emotions. A further issue concerns those ethnolinguistic concepts themselves (see also Fiske, 2020; Griffiths, 1997). Everyday language has evolved to serve different purposes to those necessitated by scientific enquiry and has consequently arrived at normatively biased and potentially overinclusive categories. Just as Anglophone laypeople once believed that whales and dolphins belonged in the category of “fish,” the contemporary vernacular concepts denoted by words such as “anger,” “fear,” and “disgust” include some instances that are not actually attributable to the same underlying process as other instances. To make scientific progress, researchers therefore need to refine the available folk concepts so that they only include exemplars that are governed by a common generative mechanism.

For Scarantino (2015), it is a mistake to presuppose that everyday English-language “emotion” words pick out genuine basic emotions (see also Fiske, 2020). To get a fix on scientifically coherent psychological processes, we need to develop more tightly bounded basic emotion concepts that should be labeled by neologisms such as “defensive basic anger” and “unconditioned basic fear” (Scarantino, 2015, p. 363). Arriving at the right concepts requires researchers to identify which instances of the more inclusive folk-language category are generated by a common causal mechanism.

This recommendation raises a number of issues. First, how specific would the refined basic emotion categories need to get before they could provide the requisite level of conceptual coherence? Unless the generative causal mechanisms are themselves specified in sufficiently broad and abstract terms, there is a risk that the resulting range of reformulated basic emotions would approach the unmanageable number of existing folk names for emotions. Second, the reformulated concepts would only retain a loose relation to vernacular “emotion” categories, making it debatable whether we’d really end up with a taxonomy of “emotions” at all. Third, Scarantino's starting point for the re-specification procedure is still the particular folk language of English, with all of its ideological biases and normatively motivated omissions (e.g., Blasi et al., 2022). Choosing any other single language as the initial reference point for identification of basic emotions would clearly bring similar problems (Fiske, 2020). Then again, it is not clear that sampling concepts from a wider range of human languages would yield a consistent set of discrete emotion categories in the first place (Jackson et al., 2019, and see above).

A different set of concerns arises from Scarantino's focus on emotion names as descriptive representations of pre-existing nonlinguistic phenomena. If applying folk concepts to experience also makes a difference to how psychological episodes unfold and contributes to perceptions of their discreteness, then abandoning those folk concepts risks losing purchase on a range of significant “emotional” processes operating both within and across cultures (Parkinson, 2017). The possible language-dependence of emotions is discussed further in the next section.

Psychological construction of emotion perceptions

Russell (2003) draws an analogy between how people perceive emotions and how they identify constellations in the night sky. Lay astronomers are confronted with a scattering of points of light without any obvious configural organization, but our ancestors developed representations of animals and mythical figures that could be superimposed onto this astral array, allowing their descendents to join some of the dots. If we know where the imaginary lines should be drawn, we can see patterns that provide a simplifying structure to our perceptions. However, the resulting constellations are not really out there in the places that we are looking; they are a product of the way in which we mentally represent what we see. Because different representations are available in different societies and at different stages of history, different groups of people end up seeing the same things in different ways.

So too with emotions, according to Russell (2003). What is available for perception is core affect constituted by pleasure and arousal, the unfolding situations in which we experience it, the objects to which it is attached, and our changing thoughts and inclinations about them. The emotion concepts provided in any shared language allow us to package together these diverse phenomena in ways that produce an apparent structure. But that structure is not governed by any integrative set of emotional processes and varies depending on the specific concepts that language makes available. “Emotions” thus reside in the eyes of English-speaking perceivers not anywhere else in their psychological environment.

In Russell's (2003; Fehr & Russell, 1984) view, the concepts encoded by emotion words are not defined by necessary and sufficient criteria that allow sharp lines to be drawn around them. Instead, they are represented by prototypes (e.g., Rosch, 1973) that identify a central exemplar with which putative instances of the category are compared. Emotional prototypes take the form of scripts (e.g., Schank & Abelson, 1977) that set out the usual chronology of the events that they represent. If whatever is happening corresponds more closely to this prototype across its many represented sequential features, classification as a member of the category is more likely. Emotion concepts have fuzzy boundaries because each of the inclusion criteria are probabilistic and not all of them need to be satisfied for an episode to be classified as an instance of the relevant category.

Application of prototypical representations to affective experiences does not directly bring a corresponding emotion into being in Russell's (2003) model. The construction process operates at the level of labeling and perception, producing an “emotional meta-experience” (Russell, 2003, 2015) rather than any differentiated emotional state. However, classifying your experience as “anger” or “fear” etc., may also encourage activate regulatory strategies that shape future behavior (Russell, 2003). It also seems possible that activated emotion concepts may prime related thoughts and attune the cognitive system to aspects of the situation that are more prototypical, thereby shaping the ensuing episode so that its prototypicality increases. Both these processes potentially lead to more differentiated emotional episodes as well as more differentiated emotion perceptions.

Psychological constructionist accounts rarely emphasize the interpersonal elements of representational processes, but these too may contribute to the production of differentiated concept-congruent conduct. For example, someone else's perception that you are “angry” may encourage them to be defensive, retaliatory, and antagonistic in return. To the extent that another person's antagonism features in the “anger” prototype, your detection of it may then lead you to see yourself as “angry.” It may also encourage you to engage in reciprocal antagonism, thereby reinforcing the other person's initial categorization of you as “angry” and making the interpersonal emotion cycle (Hareli & Rafaeli, 2008) a more vicious one. Indeed, emotion categorization may operate as a socially distributed process (e.g., Hutchins, 1995), with meaning consolidating as a function of interactions between more than one person. This kind of co-constructed emotional categorization process is difficult to disentangle from the relational pattern that it both represents and influences as the interpersonal episode develops and unfolds in real time (see also Parkinson, 2020).

Psychological construction of emotional experiences

Unlike Russell (2003), Barrett (2017) contends that emotion categorization produces differentiated emotions and not just changed perceptions of affect (i.e., distinct emotional experiences rather than merely emotional meta-experiences). This deeper form of construction is made possible by her reformulation of emotion representation as a context-dependent embodied simulation of multimodal perceptual information rather than an exclusively mental process that compares semantic features with an abstractly specified emotion prototype. More specifically, Barrett (2017) argues that emotion categories are represented as sets of situation-specific memory traces of the internal and external signals associated with use of the relevant emotion word. Correspondingly, emotions themselves are treated as perceptions of what the body is doing (in particular, its levels of pleasure or arousal) in a particular context, thus similarly combining interoception and exteroception. In Barrett's (2017, p. 30) own words: “An emotion is your brain's creation of what your bodily sensations mean, in relation to what is going on around you in the world.”

These mutually compatible formulations of categorization and emotion help to close the theoretical gap between these two processes making it seem more plausible that categorization constitutes emotion itself. However, they also raise the question of where the input for the categorization process that generates emotion comes from in the first place. Barrett's (2017) view is that many of the interoceptively registered internal changes that contribute to emotional experience arise as a consequence of the brain's predictions about requirements for metabolic energy (body budgeting or allostasis). But if the categorization process is detecting the body's capacity to deal with specific challenges and threats presented by the environment, then categorization starts to sound like a form of appraisal, which is widely believed to cause emotion in models that don't make specifically constructionist assumptions (see also Moors, 2022). Indeed, many appraisal theories (e.g., Lazarus, 1991) have also proposed that emotions are activated by perceptions of challenges and threats and the individual's capacity to cope with them (see Moors, 2022). If the bodily changes that cue the emotion category and hence the emotional experience are in fact produced by a form of appraisal, then categorization itself stops being the primary cause of the emotion. What additional work does categorization then need to do in the process of emotion generation, apart from adding a linguistic label to the action preparation that has already been initiated?

Barrett's (2017) view is that specific emotions have less internal integrity than most appraisal theories imply. Instances are not linked by any common essence or core relational theme (Lazarus, 1991), but rather by the culturally provided concept associated with the emotion word. This concept may lump together a diverse set of context-marked memory traces that lack any distinctive interconnections apart from the consistent use of a word to represent them. In Barrett's (2017, p. 138) terms, “any two emotion concepts, like ‘Surprise’ and ‘Fear,’ need no consistent fingerprints to distinguish them. So we, as a culture, introduce mental similarity using words.” But if there were no consistent perceivable differences between the events picked out by different emotion categories prior to the application of the different emotion words, how did these words get differentially associated with those events in the first place? Why would any society come up with a name for a diverse set of experiences without any overlapping features? In response to this question, Barrett might argue that the word's meaning consolidates around the interrelated goals served by the behavioral tendencies it encodes rather than any consistent patterns of bodily response. In that case, however, the key factor needed to produce the emotion again seems to be activation of those goals rather than detection of one of the various patterns of interoceptive and exteroceptive information associated with the emotion word.

Psychological constructionist theories mainly focus on the effects of emotion categorization and pay less attention to the sociocultural origins of the concepts on which that categorization depends (Lindquist et al., 2022). Although Barrett's and Russell's writings acknowledge the relevance of these prior processes, neither author provides a complete account of how they might operate as part of a more extended emotion construction process. In this regard, it seems clear that understanding the point of using emotion words within any society can help to make sense of the cultural forces that have shaped the differentiation of emotion concepts, emotion perceptions, and potentially emotions themselves. In the next subsection, I discuss Averill's (1980) social constructionist theory of emotion, which more directly addresses this complementary part of the story.

Social construction of enacted emotions

Averill's (1980) theory focuses squarely on the cultural functions served by emotions, formulated as transitory social roles. From his perspective, a person follows a cultural script when getting emotional rather than representing a prior affective experience in culturally scripted terms. And the point of the script is not to impose perceptual order on otherwise loosely arranged components but rather to navigate a path through situations where societal norms clash with one another.

Averill has applied this analysis to a range of different emotional syndromes in different historical and cultural contexts (e.g., Averill, 1985; Averill et al., 1990; Averill & Nunley, 1988), but his most extensive work addresses the Anglo-American phenomenon of “anger” (Averill, 1982). In his view, getting angry is a strategy for resolving the conflict between pacifistic and retributional norms in the United States and many other Western societies. These norms encourage mutually incompatible behaviors when an individual is insulted or threatened. On the one hand, pacifistic norms encourage people to turn the other cheek or not dignify the provocation with any response. On the other hand, retributional norms encourage people to stand up for themselves, retaliate, or at least threaten to do so. So any deliberate adoption of a retaliatory stance is warranted by one set of norms but discouraged by a conflicting set that specifically prescribes an opposite conciliatory stance. To resolve this conflict, the transitory “anger” script allows people to behave in a way that demonstrates their willingness to fight back without that action seeming intentional. Their “anger” is perceived by socialized members of their culture (including themselves) as something that has come over them in the heat of the moment rather than being any premeditated act.

In my view, Averill's argument that emotions are “disclaimed actions” (Griffiths, 1997, p. 140) tends to overstate the level of deliberation behind adoption and enactment of emotional roles under many circumstances. If patterns of angry behavior are orchestrated not only by the actor's internalized script but also by the ongoing responsiveness of the other people around who share similar culturally derived perceptions and expectations (see Parkinson, 2012), then there is less need to see all anger instances as premeditated yet self-deceptively disowned. When an Anglo-American adopts an angry stance, other people make complementary adjustments to their own relational positions that are partly shaped by their own representations of how anger normatively unfolds. The emerging mutual positions of interactants thus provide structure to the angry response in a socially distributed way without the need for any prior strategic intention on the part of the actor (see also Mesquita & Parkinson, in press; Parkinson, 2019).

According to this reformulation of Averill's theory, the words that any society uses to represent emotions capture a set of shared expectations about how people should conduct themselves when facing normative conflicts. Thus, the “anger” script not only specifies individual response components and the prototypical contexts in which they occur, but also prescriptive guidelines about how people should conduct themselves when angry or when dealing with others who are angry.

Interim conclusions

I have argued that constructionist accounts of language-emotion relations provide important insights into the effects of linguistic categorization. Representing your own and other people's experience in emotional terms is not a self-contained perceptual response to what is happening but something that changes the way in which the emotional episode subsequently unfolds. However, a full understanding of this postrepresentational process also requires attention to the normative factors that have helped shape the cultural evolution of the emotion representation in the first place. These normative considerations produce concepts that have prescriptive as well as descriptive content, meaning that they provide guidelines for appropriate emotional conduct as well as an interpretative framework for perceiving emotions. The prescriptions carried by emotion representations are also likely to differ across languages and cultures meaning that linguistic differences may reflect as well as affect differences in how emotional episodes are enacted.

Appraisal functions of emotion statements

Words can perform a variety of social actions in addition to merely describing objects and events. When the representative of the jury says “guilty” at the appropriate stage of a British trial in response to the judge's properly formulated question, the delivery of that word transforms the defendant into a criminal (Austin, 1962). When two people take marriage vows at an official wedding ceremony in front of the necessary witnesses in the United Kingdom and elsewhere, the two people are legally bound together. When I promise to meet you at a certain time and place without sarcasm, I make a commitment to be there, and give you the right to be offended if I fail to show up.

What social actions are performed when people make emotional statements? According to the ordinary-language philosopher, Bedford (1957), “emotion words are part of the vocabulary of appraisal and criticism” (p. 294) rather than “names for feelings” (p. 281). In his view, then, the principal function of statements using emotion words is not to provide information about internal states but rather to present judgments and evaluations of people, objects, and events (a “judicial function,” p. 298). For example, saying that I am angry with you is an act of blame in many Anglo-American societies. It is a declaration that presents your conduct as negligent or otherwise reprehensible. Correspondingly, saying that I am embarrassed serves to criticize my own conduct and conveys my commitment to social expectations that I was temporarily unable to meet.

How do first-person English-language emotion statements perform these appraisal functions? One possible answer is that they conjure up the semantic representation of the presented emotion concept in hearers, who then make inferences about how the speaker is appraising what is happening. For example, you may interpret my verbal expression of anger as indicating that I am experiencing an unpleasant reaction to something you have done that I consider blameworthy. In other words, you might be reverse engineering an other-blame appraisal from my reported emotion based on your knowledge that anger is caused by seeing someone else as to blame (e.g., Hareli & Hess, 2010). You then need to decide how to react to this inferred appraisal of you as blameworthy, perhaps by apologizing to me or professing your innocence.

One issue with this account is that it requires a separate set of principles to explain the hearer's response. Simply surmising that I perceive you as blameworthy in itself does not seem sufficient to explain why you feel compelled to address my blame by acknowledging or denying its appropriateness. We need to introduce notions of relational management or norms about defending reputation to explain your motivation for addressing the implications of the inferred meaning.

From Bedford's (1957) perspective, emotion words don't name objects in the first place but instead are primarily used to evaluate or criticize someone or something. In this case, the hearer simply needs to respond to the speaker's evaluative act with a complementary act of shared evaluation or the presentation of a counter-evaluation. In the terms of Austin's (1962) analysis, saying “I am angry with you” has the illocutionary force of an act of blame and the perlocutionary effect on the hearer is either to accept that blame or resist and deflect it (see also Parkinson, 2021).

Preference structures in emotional conversations

The normative pressure to respond to illocutionary speech acts in particular ways has been investigated by conversation analysts. Sacks (1992) introduced the notion of an adjacency pair as an organizational unit where one speaker's contribution to a two-person conversation is followed by an interlocking and relevant response from the other speaker. In many cases, however, different possible responses to the first speaker's contribution are equally relevant. For example, an invitation may be either accepted or refused by the person who receives it. In cases, such as this, a preference structure (Pomerantz, 1984) governs the second speaker's response. Because acceptance is the preferred response to an invitation, an accepting utterance can be delivered straightforwardly and without hesitation. By contrast, the dispreferred response of refusal requires additional conversational work such as provision of an excuse, initial deflection of topic, hesitation, and so on. The invitee may say something like: “I would love to come, but unfortunately …” or “As it happens, May 5 is exactly the date when …” Thus, the observed characteristics of the second part of an adjacency pair reveal the preference structure that is in operation and provide evidence about the specific nature of the illocutionary force exerted in the first part of the adjacency pair.

If emotional statements are evaluative, critical, or judicial, as implied by Bedford (1957), what responses are they designed to encourage in hearers? According to Levinson (1983), the preferred response to a judgment is an acceptance of that judgment. For example, if I say “that was a great movie,” the preferred response from you would be along the lines “yes, it was” or “it was terrific, wasn't it?” By extension, the preferred response to the appraisal presented in an emotion statement should acknowledge the validity of that appraisal (Parkinson, 2021). If I say that “I blame you,” the preferred response would be for you to agree that you are to blame. However, saying that “I am angry with you” seems to carry additional force than simply saying “I blame you.” It implies that your blameworthy act has consequences for things that matter to me, and that I care about. It carries the threat that I may be willing to retaliate or seek reparations from you. My anger statement thus encourages more than a simple acknowledgement of blame. Its illocutionary force includes a solicitation of your commitment to make amends, to apologize and ideally to express complementary guilt, thereby demonstrating that you care about the same things that I care about.

The illocutionary force of “anger” statements obviously differs depending on the subject and object of the ascribed anger as well as the broader linguistic and relational context (see also below). For example, when first-person anger is directed at a third-person rather than second-person object (i.e., when I say that I am angry with her rather than you), it is often an appeal for the hearer to take the speaker's side against a putatively blameworthy antagonist, and to agree both that they are to blame and that the consequences of their actions are serious enough to need addressing. Elaboration of the many possible things that emotion words can do when used in different ways requires systematic research across a variety of contexts (see also below).

Culture-specific pragmatics

If the pragmatic account of first-person English-language emotion statements presented above is valid, then it becomes important to consider the specific evaluative, critical, and judicial functions served by using particular emotion words such as “anger” when seeking to understand how their meanings have developed in Anglo-American societies. Correspondingly, if we want to compare the meanings of emotion words from different languages then assessing the pragmatic functions of their usage in local conversations takes precedence over trying to distill any semantic meaning or pin down their putative referential function in naming psychological states. Along these lines, Lutz (1988) conducted an anthropological investigation of “emotions” (or their closest local equivalent) on the Micronesian atoll of Ifaluk. One of the words that Ifaluk people use when talking about “their insides” (“niferash”) is “song,” glossed by Lutz as “justifiable anger.” If we want to understand the similarities and differences between “song” and English-language “anger,” then we need to compare the pragmatic functions served by usage of these two words in real-world conversations.

Lutz's (1988) ethnographic observations provide us with an initial basis for this comparison. Her general conclusion was that the people of Ifaluk use first-person ascriptions of song to express moral judgments of other people who have failed to live up to accepted norms about social relations, sharing resources and so on. Thus, according to Lutz (1988, pp. 155–156), “When this term is used, an assessment is being made; the speaker is saying, ‘Something immoral or taboo has happened here, and it ought not to have happened.’” However, to avoid direct confrontation with others in a community where people are often in direct face-to-face contact with each other, those expressing song usually deliver their condemnation to third parties rather than alleged offenders themselves. The function served by song statements taking a third-person object is to solicit similar condemnation of the target's actions by others. According to Lutz (1988, p. 162): “Each time a person declares ‘I am song’ is a gambit or bid in an effort to install a particular interpretation of events as the definition of that situation to be accepted by others. [In many cases] … the opening bid is accepted and the force of public opinion sides with the person who first asserts that the situation is one of rule violation and hence one of justifiable anger.”

In the Ifaluk language, “song” is distinct from other emotion concepts that also seem comparable to English-language “anger.” For example, “lingeringer” is used when people suffer a mounting series of minor frustrating events, and “nguch” is specifically directed at relatives who fail to fulfill their obligations. English also has distinct anger-related words such as “indignation,” “frustration,” and “dismay,” but differs by also including the more general term “anger” which can be used across a wider range of contexts. It also seems more common for Anglo-Americans to say that they are “angry” with the person that they are talking to than for inhabitants of Ifaluk to express “song” directly at its target.

Do these differences between usages of different emotion words in different languages tell us something about differences in the emotional lives of the people who use these words? From a referential perspective, the associated concepts might simply be making more or less fine distinctions between subvarieties of universally experienced emotions depending on their different levels of cultural relevance in each society (e.g., hypercognition and hypocognition, Levy, 1973). Or the words might encode culturally significant contexts in addition to the universal emotions experienced in those contexts. It is true that there are sufficient parallels between “song” and “anger” to make it plausible that their linguistic differences are about granularity and selective emphasis rather than the fundamental nature of their referents.

From a pragmatic angle, there are also commonalities between the functions and effects of using these words. Both “song” and “anger” avowals deliver accusations and allocate blame. However, “song” avowals serve the purpose of upholding a shared moral order to a greater extent than most “anger” avowals, which are more often oriented to individual rights and expectations within interpersonal relationships. Relatedly, first-person “song” statements are relatively more likely to come from the “chiefs” as uncontested moral authorities on Ifaluk. They also carry more weight when delivered by people occupying more exalted positions in the social hierarchy because those people have greater power to impose sanctions on the moral offenders identified by song ascriptions. Although “anger” exerts a greater influence on other people when expressed by more powerful individuals in Anglo-American societies too (e.g., van Kleef et al., 2006), formal pronouncements of “anger” are less commonly used in public condemnations of those whose conduct is presented as offending the broader community. In this sense, the pragmatic meaning of “song” is bound to be different from that of “anger” because its illocutionary force depends on the specific social and institutional arrangements that prevail on Ifaluk.

Pragmatics and emotional reality

How does emotional language relate to emotions themselves according to the pragmatic perspective outlined in this section? If, as Bedford (1957) argued, emotion nouns are not names for psychological objects, what exactly is it that needs to be calibrated against emotion talk to understand the language-reality connection? One useful way of addressing this question is by considering how appraisals feature in emotional language and other kinds of emotional process.

Above, I outlined how a referential approach might deal with the pragmatic functions of emotion talk by drawing on Hareli and Hess's (2010) reverse engineering model. The idea behind this and other inferential accounts (e.g., de Melo et al., 2014) is that real occurrent emotions are psychological states caused by appraisals (i.e., implicit or explicit judgments and evaluations of what is happening to the person, see Lazarus, 1991). The second assumption, familiar from the referential perspective outlined above, is that emotion words (and facial expressions) are names for psychological states that constitute real emotions. Thus, telling someone that you are (or that somebody else is) “angry,” “elated,” or “hopeful” activates the descriptive concept associated with that emotion word so that they know what kind of psychological state you are referring to. They also know that this psychological state is typically caused by a particular kind of appraisal, and this knowledge allows them to work backwards from their emotion attribution to an inference about what that appraisal might be. Thus, saying “I am angry” leads you to conclude that I must also perceive someone else as to blame for a bad thing that has happened to me.

By contrast, Bedford's (1957) pragmatic approach to emotional language implies that using emotion words is already a process of making an appraisal of something. By saying that I am angry, I am not describing a psychological state caused by a separate appraisal of other-blame but performing an act of blame directed at a particular social object. So where does emotion itself come into this process? One way of answering this question is to extend pragmatic assumptions beyond the realm of language use and into the domain of other kinds of emotional enactment and “expression.” For example, some social functional accounts see emotions as relational actions directed at particular objects that serve the purpose of changing or maintaining other people's orientations toward those objects (e.g., Parkinson, 2008; 2019). From this perspective, emotions, like emotion statements, are acts of appraisal and not separate psychological states caused by appraisal. Similarly, so-called facial “expressions” of emotion may be interpreted as interpersonal moves designed to influence other people rather than as signs or symptoms of internal affective states (Crivelli & Fridlund, 2018; Parkinson, 2021). Taken together, these ideas suggest that emotion talk performs similar social functions to other aspects of the emotional process. Like facial activity, and other aspects of a person's emotional orientation to what is happening, it serves to align relations between people and objects in the shared environment.

According to this relation-alignment approach (e.g., Parkinson, 2008; Parkinson, Fischer, & Manstead, 2005), the kinds of emotion that emerge in any society depend on the local social arrangements that position individual and collective actors with respect to each other and to events in their shared environment. Some abstract aspects of these sociomaterial relations are likely to be universal, such as those pertaining to the regulation of social distance, distribution of resources, and management of social and practical threats. However, each of these abstract concerns is instantiated in culture-specific ways that vary depending on local institutions and ecological factors.

Discursive flexibility

Following Bedford (1957), the provisional account of the pragmatic functions of emotion statements sketched out in the previous sections of this article exaggerates the level of coherence that characterizes actual language usage. This simplification was achieved partly by focusing on concocted and ostensibly typical examples of sentences that people might plausibly utter. When we investigate real-world language delivery, the story becomes more complicated. In fact, emotion words are deployed in a variety of ways (Fiske, 2020) and serve a wide range of possible rhetorical functions in different linguistic and extra-linguistic contexts (Edwards, 1999). For example, Edwards' (1999) discursive analysis of talk during relationship-counselling sessions shows how formulations of a person's conduct using words such as “anger” and “jealousy” can either present it as symptomatic of the emotional person's underlying psychological pathology (e.g., their dispositional “jealousy” produces “angry” over-reactions), or as something justified by the extremity of an external provocation (e.g., their “jealousy” and “anger” arise from repeated betrayals). “Anger” avowals made by English speakers are clearly not always simple acts of blame directed at a specific target. They can also serve as claims or confessions about acting irrationally in the heat of the moment. They can function to counter accusations about acting out of calculated malice. And so on.

According to Edwards (1999), the multiplicity of emotional meaning itself serves a function by allowing speakers to adjust their formulations to meet the specific requirements of the current contextualized conversation: “It is because people's emotion displays (thus categorized) can be treated either as involuntary reactions, or as under agentive control or rational accountability, as internal states or public displays, reactions or dispositions, that emotion discourse can perform flexible, accountability-oriented, indexically sensitive, rhetorical work. For doing talk's business, people require conceptual resources that are inconsistent, contradictory, fuzzy and to-be-indexically-specified” (p. 288).

Rather than seeing abstract emotion concepts as underlying the pragmatic effects of emotion talk, Edwards treats pragmatics as primary and the conceptual representations of emotions that are available to speakers as necessary consequences of the functions that emotion statements are designed to perform. In particular, the absence of firm and fixed boundaries around emotion categories (e.g., Fehr & Russell, 1984, and see above) and the polysemy of emotional words (Fiske, 2020) are not prior causes of inconsistency and looseness in emotion talk, but rather linguistic features that develop precisely because pragmatically oriented emotion talk needs to operate in context-sensitive and adjustable ways. In other words, “Categories have to be like that to do talk's business, not simply reflect how the world is, or how minds work” (Edwards, 1991, p. 537).

The functional flexibility of emotion talk further complicates cross-cultural comparisons. It implies that different societies not only have different words that perform different kinds of appraisal functions, but also different ranges of functional usages for each of those words that are sensitive to different kinds of context. If so, it becomes less surprising that different languages do not consistently include single words corresponding to the generic English-language concept of “emotion” (Wierzbicka, 1999, and see above) to cover all of the more specific varieties of situated utterance that are possible. Societies may well generalize linguistic categories on the basis of pragmatic functions that are more significant for them than those of presenting an appraisal of events to others or naming a putative internal state (e.g., see Fenigsen et al., 2020, on cultural differences in use of “I-language”). And given that the resulting higher-level concepts are also designed for context-sensitive deployment in talk, there is no guarantee that the functions they cover remain unitary or consistent under all circumstances either.

The upshot is that some of the culture-specific words whose English translations have broadly “emotional” meanings may not exclusively serve appraisal functions (if they have those functions at all). It also remains possible that not all Anglophone emotion words are equally fuzzy, polysemic or pragmatically oriented. Some of them may even translate more directly than others into other languages, potentially because they refer to universally experienced psychological processes. So far, none of the proposed candidates for English-language basic emotion terms seem to meet the criteria of translatability or universality, but that need not mean that no words in any language serve as specific markers for consistent, human-wide affective experiences. Whether such experiences would qualify specifically as “emotions” either in vernacular English or in scientific discourse is a different question. Unless the pragmatic and referential parameters of the word in question showed some commonality with those of other words that laypeople and scholars want to call “emotions,” it is hard to see why anyone would want to treat it as an important or central exemplar of the category.