Abstract
Despite decades of challenges to the idea that a small number of emotions enjoys the special status of “basic emotions,” the idea continues to have considerable influence in psychology and beyond. However, different theorists have proposed substantially different lists of basic emotions, which suggests that there exists no stable criterion of basicness. To some extent, the basic-emotions enterprise is bedeviled by an overreliance on English affective terms, but there also lurks a more serious problem—the lack of agreement as to what emotions are. To address this problem, three necessary conditions are proposed as a minimal requirement for a mental state to be an emotion. A detailed analysis of surprise, a widely accepted basic emotion, reveals that surprise violates even this minimal test, raising the possibility that it and perhaps other would-be basic emotions might not be emotions at all. An approach that combines ideas such as undifferentiated affect and cognitive appraisal is briefly proposed as a way of theorizing about emotions that is less dependent on the vagaries of language and incoherent notions of basic emotions. Finally, it is suggested that the perennial question of what an emotion is should be given more serious attention.
Many, indeed perhaps most, emotion theorists—mainly psychologists (e.g., Ekman, 1992; Ekman & Friesen, 1971; Frijda, 1986; Izard, 2007, 2011; Johnson-Laird & Oatley, 1992; Lazarus, 1991; McDougall, 1908/1942; Plutchik, 1962; Tomkins, 1980)—subscribe to the view that there is a small set of emotions that have a special status. This view is also sometimes held by neuroscientists (e.g., Damasio, 1994; Panksepp & Watt, 2011), by philosophers (e.g., Scarantino & Griffiths, 2011), and by researchers in affective computing and sentiment analysis (e.g., Perikos & Hatzilygeroudis, 2018), and it is routinely promoted in the popular press (e.g., Beck, 2015; Burton, 2016). The psychological states endowed with this privileged status are usually called “basic” emotions, but they also go by other names such as “fundamental” or “first-order” emotions (e.g., Izard, 1977, 2011), “primary” emotions (e.g., Damasio, 1994; McDougall, 1908; Plutchik, 1962, 1980), and “distinct” emotions (e.g., Tracy, 2014), and more elaborate labels such as “affect programs” (e.g., Tomkins, 1980) and “evolutionarily built-in primary process emotions” (Panksepp & Watt, 2011) sometimes find their way into the literature. Basic emotions are generally taken to be universal, innate, and hardwired.
Despite the intuitive appeal and widespread acceptance of the general idea, the basic-emotions construct has been increasingly contested over the past 30 years, largely because of significant differences of opinion on several issues, most of which can be reduced to two fundamental questions: What are the criteria of basicness, and exactly which emotions are the basic ones? The first main section of this article includes an account of the principal claims associated with three different views of basic emotions, and the second main section raises several problems relating to those claims. Then follows a discussion of how the language of affective terms can warp thinking and research on emotions. This leads to a characterization of some minimal conditions for a mental state to be an emotion, followed by a detailed analysis (with respect to those conditions) of surprise, a mental state widely held to be a basic emotion. Finally, the last section outlines a strategic alternative to the goal of trying to identify a set of basic emotions. A more in-depth coverage of the topic of basic emotions can be found in various places, including in the October 2011 special issue of the journal Emotion Review and in many chapters of The Science of Facial Expression (Fernández-Dols & Russell, 2017).
What Makes an Emotion a Basic Emotion?
Although emotion theorists all have to confront the problem that emotions are hard to study empirically because they are unobservable, subjective, and ill-defined, there are some observable and measurable phenomena—certain prototypical behavioral expressions, especially facial expressions—that are generally taken to be indicative of the occurrence of emotions. Accordingly, it is perhaps not surprising that the modern empirical approach to the study of emotions began with studies that focused on facial expressions. In the late 19th century, fascinated by photographs published by the French neurologist Guillaume-Benjamin-Amand Duchenne, Charles Darwin (1872/1965) conducted what was perhaps the first “experiment” relevant to the idea of basic emotions, an experiment that he wrote about in The Expression of the Emotions in Man and Animals: Dr. Duchenne galvanized . . . certain muscles in the face of an old man . . . and thus produced various expressions which were photographed on a large scale. It fortunately occurred to me to show several of the best plates, without a word of explanation, to above twenty educated persons of various ages and both sexes, asking them, in each case, by what emotion or feeling the old man was supposed to be agitated; and I recorded their answers in the words which they used. Several of the expressions were instantly recognized by almost everyone. . . . On the other hand, the most widely different judgements were pronounced in regard to some of them. (p. 21)
And so were sown the seeds of a tradition of research into basic emotions—a tradition that places a special emphasis on widely recognizable facial expressions and the language of emotions and one that continues to this day.
Three approaches to basic emotions
The first difficulty that arises in discussions of basic emotions is that different theorists have different conceptions of basicness. In this section, three approaches to basic emotions will be discussed, although one only briefly. Because they focus on different aspects of a concept of basicness, rather than providing competing answers to the question of what it is for an emotion to be a basic emotion, these approaches are not mutually exclusive. At the same time, their differing emphases do constrain, in different ways, answers to that question. For example, the most common conception of basic emotions takes as central the idea that a small set of emotions is rooted in biologically basic mechanisms (vs., e.g., socially constructed mechanisms). This approach, which I refer to as the biologically basic view (although in no way meaning to imply that the other views reject a biological basis of emotions), leads to analyses of basic emotions in terms of presumed pancultural regularities, the most prominent candidate being, as mentioned above, spontaneous facial expressions. The second approach, which I call the elemental view, considers an emotion to be basic if it does not have other emotions as constituents. This view is much less concerned with external expressions of emotions and more concerned with their interrelationships. Finally, the third view, which I consider only very briefly, is the conceptual view. It has a quite different take on the topic, treating emotions as basic if they occupy a special position in the semantic organization of affective space. In the next three sections, I lay out the key ideas associated with each of the three positions—positions that were also discussed in detail by Scarantino and Griffiths (2011) under the rubric of biologically basic, psychologically basic, and conceptually basic views, respectively.
The biologically basic view
As already indicated, a central idea of the biologically basic view, at least of the kind typically embraced by psychologists, is that the most diagnostic feature of a basic emotion is the universally recognizable unique facial expression with which it is associated (e.g., Fridlund et al., 1987). This was already evident in the seminal work of Tomkins (1962) and soon thereafter championed by Paul Ekman (e.g., Ekman & Friesen, 1971). Tomkins identified interest, enjoyment, surprise, fear, anger, distress, shame, contempt, and disgust as his nine basic emotions (which he called “affect programs”). These, he maintained, constitute the driving force behind the human motivational system. Perhaps partly as a reaction to both implicit (Fehr & Russell, 1984) and explicit (Ortony & Turner, 1990) challenges to the general idea of basic emotions, as well as to competing views from behavioral ecology (e.g., Fridlund, 1991), attempts were made to refine the construct, especially by Ekman (e.g., Ekman, 1992). Ekman proposed several key features that all (or at least, most?) basic emotions purportedly share, the most important of which, again, was a distinctive, universally expressed and universally recognizable facial expression. Using his criteria, Ekman identified six basic emotions that have become the default set, a set that Keltner et al. (2019) referred to as the “basic six”: anger, fear, enjoyment, sadness, disgust, and surprise. But while proposing these, Ekman also allowed for the possibility that interest, contempt, shame, guilt, embarrassment, and awe might also be basic, and more recently, Ekman and Cordaro (2011) speculated that more than a dozen other, mainly positive, emotions might turn out to satisfy (most or all of) Ekman’s criteria for being basic emotions. Likewise, Keltner and colleagues, working in the context of what they called basic emotion theory (BET), compiled what appears to be an ever-growing list of emotions (24 at the time of this writing) that have distinct multimodal expressions, the universal signaling value of which appears to be their implicit criterion for basicness.
That facial (and other) expressions are widely viewed as central to the notion of basic emotions is not an accident, and it is not an instance of the proverbial “drunkard’s search effect,” in which the drunk looks for a lost object under a streetlight because the light is good there. The idea is a principled one in that it is rooted in the Darwinian idea that the expressions of emotions are vestiges of once “serviceable habits” for dealing with the vicissitudes of daily life. From this perspective, the biological foundation of basic emotions is paramount. Their vestigial facial and bodily movements could be used by conspecifics and other animals as important information. Furthermore, a biologically based, evolutionary approach to basic emotions carries with it other possibilities. For example, each basic emotion might be expected to be prewired and to have its own distinct physiology, and there have been serious attempts to demonstrate that this is indeed the case (e.g., Ekman et al., 1983; Panksepp, 1998; Rainville et al., 2006; for a review on the neurobiological bases of basic emotions, see also Celeghin et al., 2017). The idea that different basic emotions might be associated with differences in autonomic nervous system (ANS) activity makes sense in that such differences might be expected to contribute to the uniqueness of the associated experienced feelings. And one can imagine proponents of this view speculating that just as distinct facial expressions can provide important information about imminent life-relevant events to nearby others (and distinct vocalizations to more distant others), distinct ANS activity might provide bodily feedback—an associated internal signal of such events together with any necessary mobilization for the animal to act.
The elemental view
For proponents of the biologically basic view, the emphasis is on the identification of those emotions that have a unique evolutionary foundation. By contrast, the elemental approach has a more reductionist flavor, analogous to the idea of basicness in the physical sciences. The two views are not necessarily incompatible, but their starting points are quite different; the elemental view starts from the idea that the basic emotions are those that do not have other emotions as constituents. In this view, someone claiming, for example, that jealousy is some sort of combination of, say, anger, resentment, and disliking, would be prohibited from attributing basicness to jealousy.
Because for the elemental view the basic emotions are the ultimate, irreducible building blocks of all other emotions, an important concern is the question of how nonbasic emotions can be built up from the hypothesized basic ones. The most detailed attempt to address this question is Plutchik’s psychoevolutionary theory of emotion (e.g., Plutchik, 1960, 1980, 2001), although the general idea of combining basic emotions to produce complex ones was also proposed by Watson (1930). Plutchik, although certainly inspired by biological, especially evolutionary, considerations, proposed four pairs of (polar-opposite) “primary” emotions, which he cautiously labeled “joy” and “sadness,” “acceptance” and “disgust,” “surprise” and “expectancy,” and “fear” and “anger.” All other emotions, he claimed, are mixed in that “they can be synthesized by various combinations of the primary emotions” (e.g., Plutchik, 1960, p. 160).
Plutchik’s proposal that nonbasic emotions are mixtures of basic ones is not the only way to address the question of how the two might be related. An alternative way is to view all nonbasic emotions as psychologically elaborated forms of one or another of the hypothesized basic emotions. This is the approach taken by Johnson-Laird and Oatley (e.g., Johnson-Laird & Oatley, 1989; Oatley & Johnson-Laird, 1987). Appealing to the distinct facial expression criterion, these authors characterized five of the basic six emotions (happiness, sadness, fear, anger, and disgust) as biologically basic “emotion modes,” which they viewed as arising at important junctures of ongoing plans. Then, the nonbasic emotions, which they referred to as “complex” emotions, are elaborations of one of these modes “by means of the propositional meanings that are ascribed to [them]” (Oatley & Johnson-Laird, 1987, p. 35). In fact, the Johnson-Laird and Oatley model can also be classified as a biologically based view because they explicitly committed to the idea that their basic emotions (or, emotion modes) “occur universally in the human species” and claimed that an important criterion for a basic emotion is that “the facial expression associated with it should be recognized panculturally” (p. 33).
A third proposal for dealing with the relation between hypothesized basic emotions and nonbasic ones is to treat the basic emotions not as individual distinct emotions but as families of emotions and then to assimilate all the nonbasic emotions into one or other of the families (e.g., Ekman, 1992; Ekman & Friesen, 1975). For example, Ekman (1992) claimed that “amusement, contentment, relief, enjoyment from sensory sources, and enjoyment based on accomplishment” all belong to his basic-emotion family of enjoyment (p. 193). However, this way of conceptualizing the relation between basic and nonbasic emotions is not characteristic of the elemental approach because it does not treat basic emotions as the irreducible building blocks of all other emotions. Instead, it treats them as prototypes of categories around which other emotions cluster on the basis of some sort of similarity metric.
The conceptually basic view
The third sense in which “basic” is used in the context of basic emotions is sometimes referred to as the conceptually basic sense (e.g., Scarantino & Griffiths, 2011) because it pertains to the way in which people’s concepts of emotions are organized. A central idea of this view, which derives from early work in cognitive psychology (e.g., Rosch, 1973) and cognitive semantics (e.g., Lakoff, 1987) in the 1970s and 1980s, is that for many categories, members can be organized into something approximating an inheritance hierarchy in which entries lower in the hierarchy inherit all the properties that are associated with those higher in the hierarchy. Crucially, within such hierarchies, there is often a level, called the “basic level,” that is typically the most salient, cognitively accessible level, and this level exhibits some particularly interesting properties. For example, it tends to be the most concrete level that is mono-morphemically lexicalized (chair vs. armchair, dog vs. Scottish Terrier), and words associated with the basic level are generally of higher frequency and learned earlier than those associated with other levels in the hierarchy. In addition, under normal conditions, the spontaneous response upon encountering an exemplar of the category is to label it at the basic level so that if, for example, a dog suddenly runs into the road in front of us as we are driving, the usual spontaneous response would be to say something like “Watch out for the dog!” rather than “Watch out for the animal!” or “Watch out for the Scottish Terrier!” There are several other interesting characteristics that are often true of basic-level categories, all of which were nicely summarized by Fodor (1983, pp. 94–96).
In the context of emotions, a good example of the conceptually basic approach can be found in the work of Shaver at al. (1987), who had participants sort putative English emotion words into groups on the basis of the perceived similarity of the words’ meanings. 1 The results of a hierarchical cluster analysis revealed six groups of emotions at what the authors considered to be the basic level: love, joy, anger, sadness, fear, and, tentatively, surprise. They also noted that their six basic-level categories agreed moderately well with many other lists of basic emotions, and indeed, although love rarely appears in lists of basic emotions, except for disgust, all of the basic six show up as basic-level emotions in the Shaver et al. study. Furthermore, these basic-level emotions map well onto results emanating from other work (e.g., Fehr & Russell, 1984; Shaver et al., 1987), demonstrating that emotion concepts tend to be structured (i.e., cognitively represented) in terms of the proximity of their instances to some sort of prototype as well as being consistent with the idea that emotion categories are graded, meaning that some emotions (e.g., anger) are judged to be better (i.e., more typical) exemplars than others (e.g., spite).
Some Problems for Basic-Emotions Theories
The differences between the three conceptions of basic emotions just summarized might lead one to conclude that emotion theorists have embraced basic emotions as a theoretical construct without any consensus as to what exactly the construct is. For researchers committed to the biologically basic view, basic emotions are evolutionarily derived, innate, and hardwired and have universally recognizable unique facial expressions and possibly distinct physiology. Meanwhile, proponents of the elemental view, although acknowledging their biological bases, focus on identifying a small set of emotions that constitute the irreducible constituents of all other emotions and seek to specify how these basic emotions are related to the other, nonbasic ones. Finally, the conceptually basic view is more about how people’s emotion concepts are organized in semantic memory, as exemplified by the language used to talk about them. This view of basic emotions seems to carry the least theoretical baggage and makes the fewest assumptions about emotions per se. It is primarily descriptive in nature, in contrast to the biological and elemental views, which are presumably intended to have some explanatory power and predictive value. Not surprisingly, each of the three approaches has its own set of criticisms to deal with, some of which are reviewed in the following sections. I first very briefly review the just discussed conceptually basic view and then, because of its importance in addressing the relation between basic and nonbasic emotions, discuss in somewhat more detail some of the difficulties facing the elemental view. Finally, I point to the main problems associated with the biologically basic view—problems that I do not discuss in detail because they have been extensively covered elsewhere.
Challenges to specific views of basic emotions
The conceptually basic view
The main difficulty facing the conceptually basic view is that it is not really about emotions at all. Instead, it is about how people represent some of their knowledge about emotions—their emotion concepts—as evidenced by the way in which they understand the meanings of words. Knowing that “anger” is a basic-level term, however, tells us nothing about what, if anything, it might mean to say that anger (the state) is a basic emotion. Just as the fact that dog occupies a special, basic-level position in an animal–mammal–dog–Terrier–Scottish Terrier hierarchy fails to tell us much about the nature of dogs, so too does the fact that anger occupies a special position in a hierarchy such as psychological state–emotion–anger–irritation–grumpiness fails to tell us much about the nature of anger.
The elemental view
In contrast to the conceptually basic view, the elemental and biologically basic views really do purport to be about emotions rather than about emotion words and corresponding concepts. The elemental view, in particular, confronts an important and legitimate question pertaining to basic emotions that is largely ignored by proponents of the biologically basic view: the question of how nonbasic emotions are related to the basic emotions.
As already indicated, the most detailed attempt to address this question was that of Robert Plutchik. Modern psychological approaches to emotion tend to pay relatively little attention to Plutchik’s model, but there are reasons why it warrants some discussion here. First, it remains the only systematic attempt to spell out the relationship between hypothesized basic emotions and all the other, nonbasic ones. Even a hint of the kind of difficulties associated with this problem might serve to indicate what a challenging, one might say hopelessly challenging, problem it is. But there is also a quite different kind of reason: The model is frequently used in real-world applications of computing technologies such as social robotics (e.g., Qi et al., 2019) and sentiment analysis (Chafale & Pimpalkar, 2014), which often make assumptions about the objectivity of basic emotions. Moreover, it is frequently referenced in the popular press and on the Internet as an established model of basic emotions. With this kind of impact outside of scientific psychology, if the model has shortcomings, it is fitting that they not be ignored.
A general concern with Plutchik’s approach is its lack of clarity as to the way in which his primary emotions were intended to combine. Plutchik frequently used the analogy of chemical compounds even while vacillating between speaking of them as compounds and speaking of them as mixtures. Furthermore, the incoherence of some of his combinations, only a couple of which will be mentioned here, serves only to reinforce any skepticism one might have about the entire agenda. For example, guilt is said to be a (tertiary) mixture of fear and (curiously) joy (Plutchik, 1962, p. 118), and love results from mixing joy and acceptance (Plutchik, 2001, p. 350). Considering just the second of these, the only way to interpret it is to suppose that joy and acceptance play some sort of causal role in producing (the complex emotion of) love. But the implied causal direction makes little sense compared with the reverse causal direction, in which, for example, at least some forms of love cause acceptance rather than being caused by it. And this is all assuming that love is not itself a basic emotion, an assumption that some emotion theorists (e.g., Parrott, 2001; Shaver et al., 1996) would contest. Issues such as these raise real questions about the viability of Plutchik’s proposals for complex emotions, and this is before one asks the key question of whether Plutchik’s primary emotions of acceptance and expectancy are emotions at all, never mind basic or primary ones.
Generally speaking, the more specific a set of proposals is, the easier it is to find weaknesses, and in this regard, it is almost unfair to take Plutchik to task. However, given that he has been the only theorist to attempt a detailed account of how basic emotions are related to nonbasic ones, and especially given the importance (and general neglect) of that question for any view that embraces the notion of basic emotions, my identification of some typical problematic issues seems warranted.
Regardless of its shortcomings, Plutchik’s account of the relation between his basic emotions and nonbasic emotions is detailed and comprehensive. The Johnson-Laird and Oatley theory is less specific. However, it is possible to make some general observations about it. For example, if one accepts their notion of basic-emotion modes as contentless internal signals, then the idea that complex emotions are propositionally (i.e., cognitively) elaborated versions of them seems quite plausible. On the other hand, treating basic emotions as contentless signals comes at a cost. At the most general level, it is inconsistent with the fundamental premise underlying basic-emotion theories, which is that basic emotions are a subset of emotions proper. It is widely recognized that emotions proper are intentional states—they are about something—and it is their content that distinguishes one emotion from another. This means that Johnson-Laird and Oatley contentless basic-emotion modes, because they lack content, cannot be emotions proper. Perhaps this is why Johnson-Laird and Oatley referred to them as basic emotion “modes” rather than as simply, “basic emotions.” But this leads to another problem because as well as making their case for basic-emotion modes as contentless, Johnson-Laird and Oatley identified five of the big six (happiness, sadness, fear, anger, and disgust) as their basic-emotion modes. So we are left with the confusing situation in which basic-emotion modes are both standard basic emotions, which have content, and contentless internal signals, which by definition do not.
A different kind of difficulty for the Johnson-Laird and Oatley view pertains to their principal criterion for a basic emotion: That its associated facial expression be recognizable panculturally. However, even though this criterion usually results in the inclusion of surprise, these authors excluded surprise (and interest) because these states “are not single emotions.” But then, a few sentences later, they allowed that surprise might be a basic emotion after all “but with properties that allow it to combine with other emotions” (Oatley & Johnson-Laird, 1987, p. 33). In general, as I argue below, skepticism about whether surprise is a (basic) emotion is easy to justify, but not, I think for Johnson-Laird and Oatley. For them, as for Mandler (1984), interruptions to ongoing processes are the sine qua non of emotions. Surprising events interrupt ongoing processes, and they have consequences for plans and goals, and therefore, according to the Johnson-Laird and Oatley view, they should generate a signal to the information-processing system to reassess current plans, to reprioritize goals, and perhaps to do something or to stop doing something. Johnson-Laird and Oatley’s ambivalence about whether surprise is a basic emotion, which, importantly, appears to be due in part to their ambivalence about whether it is even an emotion, illustrates exactly the kind of question that most, if not all, basic-emotion advocates have to confront: What exactly are they talking about when they talk about basic emotions or even nonbasic emotions for that matter?
Finally, although Ekman’s view of basic emotions does not focus on the relation between basic emotions and nonbasic ones, he appears to have taken two positions on the issue. The first position relies on his notion of emotion families and treats nonbasic emotions as being variations on a theme exemplified by the “basic” emotion prototype (Ekman, 1992, p. 173). But Ekman (1992) also wrote about emotion blends, albeit usually in the context of two or more emotions occurring at the same time, and he expressed skepticism about the possibility of blends with respect to physiology and expression (p. 195). However, we get no further than this because a few years later, Ekman (1999) wrote, “It should be clear by now that I do not allow for non-basic emotions” (p. 57). So that would seem to be that.
In summary, with respect to the key question relative to elemental theories, one cannot but concur with Reisenzein’s (1995) conclusion that nobody has proposed any convincing empirically verified or verifiable rules that relate nonbasic emotions to proposed basic ones.
The biologically basic view
The most prominent criticisms of the biologically basic view have to do with challenges to the validity of the criteria for basicness and the empirical methods used to test them, and the most persistent of these challenges target the facial expression criterion of basicness. Some authors (e.g., Fridlund, 1994, 2017; Ortony & Turner, 1990; Russell, 1995) have argued that a distinctive facial expression is not a necessary component of any emotion, let alone of a would-be “basic” emotion, whose existence such authors, in any case, question. Nor does the regular co-occurrence of a facial expression with a particular state ensure the presence of an emotion, again, let alone a basic emotion. For example, puzzlement is often accompanied by a distinctive facial expression, yet it is rarely, if ever, proposed as a basic emotion. Much the same can be said of extreme exertion and severe pain. So some widely recognizable facial expressions appear to be associated with states that are not considered to be emotions and certainly not basic emotions. Meanwhile, some evolutionary psychologists suggest that for an emotion to be a basic emotion, it must have a distinct adaptive function relevant to reproductive success, and on this criterion, as pointed out by Buss (2014), jealousy qualifies as a basic emotion even though it has no associated recognizable facial expression. Other work has raised questions about the methodology used in establishing the often cited empirical evidence for basic emotions, again, especially evidence that purports to establish six or seven basic emotions on the basis of the universality of facial expressions (e.g., Jack et al., 2016; Russell, 1994). Because criticisms of the biologically basic view are ubiquitous and readily accessible (see, e.g., several chapters in Barrett & Russell, 2015, and in Fernández-Dols & Russell, 2017), they will not be elaborated further here. However, note that the various critiques have met with robust responses from advocates of the biological view (e.g., Ekman, 2017; Hutto et al., 2018; Keltner et al., 2019).
Challenges to the general idea of basic emotions
The chief conclusion to be drawn from my brief review of the different approaches to basic emotions and some of the difficulties they have to confront is that there is considerable disagreement about which emotions are the basic ones and why. If there were a well-established and scientifically justified criterion for what it is for an emotion to be a basic emotion, disagreements among theorists might be expected to be more apparent than real, perhaps reflecting only the occasional newly discovered basic emotion to be added to an already established list of verified ones. But this is not the case. To see how widely divergent are the claims about which are the basic emotions, it is worth considering both within-theorist and across-theorist comparisons.
Ekman has for decades been the most ardent advocate of basic emotions. Accordingly, his views provide a good opportunity to make a then-and-now, within-theorist comparison. As already discussed, his early work (e.g., Ekman & Friesen, 1971) resulted in the establishment of anger, fear, enjoyment (i.e., happiness?), sadness, disgust, and surprise as the basic six, although already in the early 1990s, Ekman was open to the idea that there might be other basic emotions, such as contempt, shame, guilt, embarrassment, interest, and awe (Ekman, 1992). Now, in his more recent work (Ekman & Cordaro, 2011), the positive states of sensory pleasures, amusement, relief, excitement, wonder, ecstasy, schadenfreude, and rejoicing are listed, along with naches, characterized as “a Yiddish word for the feeling a parent/caregiver or teacher feels when witnessing the achievement of their offspring,” and fiero, “an Italian term for the emotion felt when meeting a difficult challenge” (p. 365). Ekman and Cordaro (2011) also indicated that guilt, shame, embarrassment, envy, and familial compassion satisfy most but not all of Ekman’s (now 13) criteria for basicness, and they raised the possibility that jealousy, love, and hate do, too. However, they noted that there is as yet insufficient evidence with respect to these latter states and admitted that there is even disagreement as to their status as emotions. In this connection, Ekman and Cordaro now explicitly excluded interest, which, correctly in my opinion, they took to be a cognitive state, 2 although it often features in other theorists’ lists of basic emotions (e.g., Izard, 2011; Levenson, 2011; Tomkins, 1980). Ekman and Cordaro concluded that a determination as to whether the states mentioned in their wide assortment of new candidates are indeed basic emotions will have to await additional cross-cultural evidence. The addition of these new candidates brings the number of putative basic emotions from the original half dozen to close to 20.
Furthermore, this is unlikely to be the final count. Apart from the occasional word (e.g., “naches” and “fiero”) drawn quite unsystematically from just two of the world’s seven thousand or so languages, this new list of candidate basic emotions is based only on a consideration of English. It would not be surprising to see other candidates emerging if one were to explore the affective lexica of other languages. On the other hand, if we take seriously Ekman’s assertion that he did not “allow for nonbasic emotions,” there is no need to wait for a final list of basic emotions. All we need do is decide which states are emotions and which are not because the two sets—emotions and basic emotions—are now coextensive. And, of course, if all emotions are basic emotions, we can dispense with the construct of basic emotions altogether. Again, we are confronted with the spectacle, alluded to in the previous section with respect to Johnson-Laird and Oatley’s vacillation with respect to interest and surprise, that many advocates of basic emotions do not have a stable criterion of basicness or even of what it takes for a state to be an emotion.
Meanwhile, in the same issue of the same journal as the Ekman and Cordaro (2011) article (Emotion Review, Vol. 3, No. 4), Izard (2011) listed interest, enjoyment (again, happiness?), sadness, anger, disgust, and fear as basic emotions while having 2 years earlier also explained that “shame, guilt, and contempt . . . and the pattern of emotions in love and attachment may be considered basic in the sense that they are fundamental to human evolution, normative development, human mentality, and effective adaptation” (Izard, 2009, p. 8). It is true that Izard’s set of basic emotions included all but one (surprise) of the basic six, but it also included many others. Furthermore, as indicated in Table 1, across a wider array of emotion theorists, the commonality is much less encouraging.
Sample of Influential Emotion Theorists and Their Proposed Basic Emotions
Note: In some cases (e.g., Ekman, Izard) the basic emotions ascribed to authors should be taken as snapshots in time rather than as stable across time.
Recently, Cowan and Keltner (2021) summarized the results of large-scale computational analyses and identified 18 emotions that can be reliably distinguished through their distinct multimodal signatures (e.g., facial, bodily, vocal expression). Although not explicitly referring to them as basic emotions, Keltner (personal communication, June 29, 2020) indicated that if he were to apply that term, he would do so to their 18 emotions: amusement, anger, anxiety, awe, confusion, contentment, desire, disgust, elation, embarrassment, fear, interest, love, pain, relief, sadness, surprise, and triumph.
A reasonable conclusion from the above observations is that the lack of a robust criterion supported by a reasonable scientific justification for what is to count as a basic emotion has resulted in a scattered and confusing landscape in which the number of proposed basic emotions ranges from two to more than 20, with promises of more to come. This raises the question of what, if anything, is to be gained by postulating and searching for basic emotions. If you ask a physical scientist to list the basic chemical elements, you get the same answer regardless of who you ask and, these days, when you ask. If you ask an emotion theorist to list the basic emotions, you get a different answer depending on who you ask and when you ask. With respect to the validity of the construct, this is not a promising situation.
A more general line of attack on basic emotions gets its impetus from rejecting the assumption that emotions are discrete categories. The idea is that what we think of as discrete emotions are better thought of as vaguely bounded regions in a two-dimensional (or sometimes three-dimensional) space (e.g., Russell, 1980; Russell & Mehrabian, 1977). Advocates of this kind of view tend not to postulate a small number of universal foundational emotions, implicitly accepting the idea that the only thing that is basic about emotions are the dimensions in terms of which they can be characterized—dimensions such as valence and arousal (Russell, 1980) or perhaps three dimensions such as evaluation, potency, and activity (Osgood et al., 1957; see also Schlosberg, 1954). Such dimensions, of course, are not themselves emotions. Building on Russell’s (2003) notion of two basic dimensions comprising “core affect,” there has now emerged an explicit rejection of the whole notion of basic emotions by theorists who subscribe to constructionism, a view that considers emotions to be psychological constructions built out of culturally shaped conceptual and linguistic information (e.g., Barrett, 2006; Barrett & Russell, 2015). The increasing influence of constructionism has thus become a kind of countermovement to basic emotion theory.
The Language Problem
The vagueness of the concept of emotion
Up to this point, my focus has been on problems relating to different theoretical approaches to the idea of basic emotions. However, the study of basic emotions, and indeed of emotions more generally, has to confront a completely different set of problems: those associated with the meaning and reference of emotion words. Of the myriad experienceable emotions, in any one language, only a handful are lexicalized. Furthermore, different languages lexicalize different emotions, and even when two languages lexicalize what appears to be the same emotion, the cross-linguistic mapping is often quite imperfect. A principal reason that questions relating to the language of emotions are relevant to the basic-emotions issue is that most proposed basic emotions, especially the big six, all seem to be conveniently labeled with very salient, high-frequency, English words.
In discussing the conceptually basic view of basic emotions, I mentioned that emotion categories are graded categories. The same is true of the category emotion itself. So, just as a desk will be judged to be a better example of the category furniture than a piano (which some might judge to not be furniture at all), some putative emotions will be judged to be better examples of the emotion category than others, and some might not be judged to be emotions at all. If you ask people how confident they are that anger is an emotion, they will tell you that they are very confident, but they are likely to be much less confident that hunger is an emotion. This lack of confidence is not rooted in ignorance about what it is to be hungry. It is rooted in uncertainty about what it is for a state to be an emotion. Part of the explanation for this is that our experiences of conditions such as anger and hunger are more homogeneous than our experiences of the conditions that we typically think of as emotions. Of course, this is true of category hierarchies in general: Show people a couple of cows selected at random, and they will judge them to be more similar to each other (because they are drawn from a more homogeneous category) than if they are shown a couple of animals picked at random. So, insofar as we have a problem about whether hunger is an emotion, it is mainly because we have a problem with the concept of emotion rather than with the concept of hunger. Given this, it is perhaps not surprising that even after decades of serious research on the topic of emotions, researchers not only fail to agree about which emotions, if any, are basic and why, but also fail to agree about what emotions are. This fact has long been recognized in the field (e.g., Barrett, 1998; Ekman, 1992; Fehr & Russell, 1984; Gendron & Barrett, 2009; Izard, 2010; Mandler, 1975; McDougall, 1908/1942; Oatley & Jenkins, 1996; Russell, 2003; Scherer, 1984), yet it has not deterred advocates of basic emotions.
The problem is that if there is no agreed-on scientific definition of emotion, if we cannot specify what it takes for something to be an emotion, how can we be sure that all proposed basic emotions are themselves actually emotions? If there is a privileged set of “basic” emotions, the least that one would expect is that every member of that set really is an emotion. That this might not always be the case was already hinted at when I mentioned earlier that there might be an issue about the status of Plutchik’s “primary” emotions of acceptance and expectancy, as well as questions about the status of interest and surprise in the Johnson-Laird and Oatley theory, and some expressed uncertainty by Ekman and Cordaro (2011) concerning some putative candidates for basic emotions “about which there is argument as to whether they are emotions” (p. 366). Whether such states are or are not emotions depends on what it is for something to be an emotion, a question that emotion theorists have rarely been willing to tackle head on. It matters not that for the purposes of everyday language, the word “emotion” is vague. But for scientific purposes, it matters a great deal. The referent needs to be fixed if only so that competing theories about emotions really are theories about emotions and can be evaluated as such. For everyday purposes, looking up at the millions of objects in night sky and referring to them all as stars is not a problem. But for scientific purposes, conflating stars, asteroids, and planets is. It seems preferable that as emotion theorists and researchers, we take a stand and try to reach a consensus on how to characterize the object of study rather than plowing ahead as though there were no problem. Accordingly, in the next major section, A Minimalist Account of Emotion, I shall attempt to specify some, even if not all, of what I take to be the necessary conditions for membership in the category emotion. The joint satisfaction of these conditions would then constitute a prerequisite for something to be an emotion, thereby providing a minimal test of emotionhood that could be applied to putative basic emotions.
The language specificity of emotion words
The vagueness of the concept of emotion is only one of the problems pertaining to the relation between emotions and language. Other problems arise relating to the language specificity of putative emotion words, and these problems have particularly interesting implications for the basic-emotions claim. Consider first the obvious fact that no language has or can have a word for every experienceable emotion. Across different languages, one can find a wide variety of specialized, language-specific, affective terms that are untranslatable in the sense that the phenomena they pick out are not lexicalized in other languages (e.g., Watt-Smith, 2015). Two examples (Yiddish “naches” and Italian “fiero”) were mentioned earlier in connection with Ekman and Cordaro’s (2011) expanded list of basic emotion candidates, and it is probably safe to assume that few, if any, other languages have a word for the Inuit iktsuarpok, which refers to a kind of impatient anticipation for a person’s arrival, complete with periodic checking (Lomas, 2016). Furthermore, even when words are roughly translatable across languages, as tends to happen when the languages are from the same family, the match is often less than satisfactory.
To better understand the implications of these facts for basic-emotions theories, imagine an isolated culture in which the language lacks words that translate into, for example, the English word “sad” and its derivatives. Imagine, furthermore, that psychologists in this culture are engaged in trying to identify basic emotions. A natural first step would be for them to consider the most salient emotions in their culture, which would probably correlate quite well with the emotions most frequently talked about. So these psychologists would start by considering common emotion words in their quest to identify basic emotions. This means that a word designating sadness would not be one of them even though, for some of their English-speaking counterparts, sadness is a basic emotion. But if sadness is a basic emotion, would one not expect it to play a significant role in the emotional lives of the people of our imaginary (and every other) culture? And so would one not expect it to be lexicalized in that language—indeed in all languages? As J. L. Austin (1956) put it, “Our common stock of words embodies all the distinctions men have found worth drawing, and the connexions they have found worth marketing, in the lifetimes of many generations” (p. 8). Yet in the culture we have just imagined, sadness appears to be so unimportant that there is not even a word for it.
In fact, our example of a culture whose language fails to reflect the emotion of sadness is actually not as fanciful as it might seem because according to Levy (1973), the Tahitians do not have a word for sadness. Discussing the consequences of this lexical gap, Wierzbicka (1999) wrote, It is ethnocentric to think that if the Tahitians don’t have a word corresponding to the English word sad (Levy, 1973) they must nonetheless have an innate conceptual category of “sadness”; or to assume that in their emotional experience “sadness”—for which they have no name—is nonetheless more salient and more relevant to their “emotional universe” than, for example, the feelings of tiōaha or pe’ape’a, for which they do have a name (although English does not). (p. 26)
The point here is not to deny that Tahitians might experience sadness but, rather, to suggest that for them, sadness is not sufficiently culturally salient to warrant being lexicalized, a fact that is hardly consistent with the presumed fundamental nature, universality, and biological importance of sadness as a basic emotion. Elsewhere, Wierzbicka (1998) discussed other examples of the failure of high-frequency English “basic” emotion words to translate directly into other languages and noted, for example, that there are no Russian words that have exactly the same meaning as the English words “sadness” and “anger.”
I am discussing sadness in the context of emotions in different cultures and languages because for researchers from English-speaking cultures, sadness is a prime candidate for a basic emotion, appearing as it does in virtually every list of basic emotions. The same is true of disgust, another emotion that appears in many lists of basic emotions but that is not lexically coded in, for example, the South American language, Sarawak (Bennett, 1989, as cited in Hupka et al., 1999). It is worth remembering that lists of basic emotions have mostly been proposed by theorists working in the Anglo tradition, and in that tradition, sadness and disgust are salient emotions. Compared with other emotion words, “sad” and “disgust” and their derivatives occur with relatively high frequency and so are likely to come to mind as candidates for basic emotions in a way that ennui, for example, is not. In the Anglo emotion research tradition, putative basic emotions, which conveniently all have English labels, seem to be treated as though their cultural and linguistic salience represented some sort of universal fact about them (Wierzbicka, 1986). It is as though theorists have implicitly assumed that (at least) English lexicalizes all of the basic emotions, and their favorite candidates are the referents of high-frequency English emotion words such as “anger,” “fear,” “enjoyment” (i.e., “happiness”?), “sadness,” “disgust,” and “surprise.” How fortunate that the search for basic emotions was not undertaken by Tahitian or Sarawak psychologists.
In fact, of course, there is no reason to suppose that (only) English, or any other language for that matter, provides special direct lexical access to a set of universal basic emotions. It is true that Western emotion theorists have sometimes ventured beyond the confines of Western cultures in their quest for basic emotions (e.g., Ekman & Friesen, 1971), but they have generally done so only in an attempt to confirm their already established beliefs about the universality of their own candidates for basic emotions. Meanwhile, less comforting cross-cultural and cross-linguistic findings (e.g., Crivelli et al., 2016; Lutz, 1982; Rosaldo, 1980; Wierzbicka, 1999) seem to have had little influence on mainstream advocates of basic emotions. Dislodging entrenched, ethnocentric beliefs is no easy matter (Henrich et al., 2010). One can but hope that researchers seeking to identify basic emotions in any one language/culture recognize that cross-cultural variation and untranslatability are warning signals of the potential for both errors of omission and errors of commission. For adherents to the idea of basic emotions, one interesting candidate for an error of omission might be the state of amusement. Very few lists of basic emotions in English include amusement, and it appears in only one of the 16 entries in Table 1, yet it is difficult to imagine that the emotion that in English we refer to as “amusement” is not experienced in all cultures.
The above discussion suggests that whether an emotion is lexicalized in some particular language is not determinative of that emotion’s possible status as a basic emotion and that the common practice of attempting to identify basic emotions by starting with high-frequency (usually English) emotion words is likely to embody unavoidable implicit biases. If there are universal patterns of emotional organization (and the existence of a small set of “basic” emotions would be one example of such a pattern), then as Wierzbicka (1999) pointed out, they “cannot be captured by means of English folk categories such as anger, sadness, or disgust, but only in terms of universal human concepts” (p. 28), by which she means concepts such as feel, do, think, want, good, and bad (see also Ortony et al., 2005; Wilt & Revelle, 2015) together with happen and body and, albeit implicitly, cause (Goddard & Wierzbicka, 1994; Wierzbicka, 1996, 1999). Starting with emotion words prevalent in some particular language is probably not the best strategy for theory construction in the realm of emotions in general.
A Minimalist Account of Emotion
I now return to the question of what an emotion is. Because the everyday use of the word “emotion” refers to a vague concept whose boundaries are fuzzy, the question has no easy answer, William James (1884) notwithstanding. However, it is possible to start with clear, paradigmatic cases and to specify what features are necessary for such cases (Reisenzein, 2012). Thus, difficult as it might be to specify the necessary and sufficient conditions for a mental state to be an emotion, we might nevertheless be able to agree on some minimal, necessary conditions. I believe that a strong case can be made that for a mental state to be an emotion, it must possess at least the following three features: It must be intentional (i.e., about something), it must be valenced (i.e., positive or negative), and it must be conscious (i.e., experienced).
Emotions are intentional states
The first requirement is that for a mental state to be an emotion, it must be about something. In other words, emotions are intentional states—they have objects. This is not true of all affective states, but it is true of emotions. When one is angry, one is angry about something or with someone; when one is relieved, one is relieved that something did not happen or stopped happening; and when one is proud, one is proud of someone or of having done something. The object of an emotion can be a direct target, as in the fear of a snake, or a propositional object, as in the joy that a friend was awarded a prestigious prize (Scarantino & de Sousa, 2018), but one way or another, emotions have objects.
In fact, the requirement that emotions are about something is a logical consequence of the most prevalent theories of emotion: cognitive appraisal theories (e.g., Arnold, 1960; Lazarus, 1991; Oatley & Johnson-Laird, 1987; Ortony et al., 1988; Roseman, 1996; Scherer, 1984; Smith & Ellsworth, 1985). The essence of appraisal theories is that our emotions are grounded in (cognitive) appraisals of the personal significance of what is going on around us. But appraisals are themselves intentional because they involve assessments of something or that something (is the case). Emotions necessarily have the same object as the appraisals that underlie them.
An important consequence of the requirement that emotions are intentional states is that it enables us to distinguish emotions from affective states more generally, of which emotions are a subclass. For example, the emotion of anxiety arises when a person is worrying about some particular anticipated or imagined troubling event. It is a mental state about something specific (the event). On the other hand, free-floating anxiety—“a diffuse, chronic sense of uneasiness and apprehension not directed toward any specific situation or object [emphasis added]” (VandenBos, 2015)—is a pathology, not an emotion, but it is certainly an affective state, or condition. Likewise for moods. When one is in a bad (irritable) mood, the mood itself is not about anything in particular (Lamb, 1987; Ortony & Clore, 1989; Solomon, 1993, p. 71), even though one might know why one is in that mood. Nor do we think of the mood itself as an emotion, although we can readily acknowledge that it increases the likelihood of mood-congruent emotions.
Finally, it is important to distinguish emotions from core affect. Core affect, conceived of as one’s background affective state, is characterized as a blend of valence and arousal and is, by definition, objectless (Russell, 2003; Russell & Barrett, 1999). It is postulated to be the foundation of emotions and, in fact, of all affective experiences, but it is not itself an emotion because it is not about anything in particular. So emotions have objects and are therefore distinguishable from amorphous affective conditions such as moods, pathologies, and core affect, which do not.
Emotions are valenced
The second characteristic of emotions that I propose is necessary for a state to be an emotion is that the state be valenced (Kringelbach & Berridge, 2017), a requirement that is consistent with the idea that a person cannot have an emotion about something while maintaining an attitude of indifference with respect to that something. Having an emotion about something entails that one cares about it, whereas being indifferent about it entails that one does not care about it (Ortony et al., 1987). This is not to say that one’s indifference might not itself be the object of an emotion. There is nothing to prevent a mental state such as indifference from being the object of an emotion. For example, there is nothing to prevent me from being pleased that I do not care about the outcome of some sports event (I might see my indifference as benefiting me because I will not waste time watching it on television). However, emotion and indifference cannot coexist with respect to the same object, a claim that accords well with everyday experience. In our daily lives, we encounter all manner of things that are of no real significance to us and therefore have no emotional consequences for us. Emotions arise only when something we cognize touches our value systems—our goals, standards, and tastes or our concerns (Frijda, 1986; Ortony et al., 1988). Thus, emotions involve some sort of positive or negative assessment. Of course, some emotions (e.g., nostalgia) can involve both positive and negative valence in the sense that they are experienced as having both good and bad aspects—experienced, as it were, as a kind of affective Necker cube—but a mental state that is experienced as neither good nor bad is not an emotion.
Emotions are conscious
At first glance, one might think it obvious that emotions are mental states of which we are conscious. However, convincing as this claim is to some, it is equally unconvincing to others. One objection is based on everyday experiences of allegedly unacknowledged emotions, as when, for example, a person accused of being jealous vehemently denies it while those close to them find it blindingly obvious. The same can be said of cases of unacknowledged anger. Surely, the argument goes, the person is indeed jealous or angry but, at least for a while, is simply unaware of it. It might be countered that sometimes allegedly jealous or angry people are not in fact jealous or angry at all but are merely behaving in a way that leads observers to erroneously infer that they are. Alternatively, their denial might be self-delusional, as evidenced by the fact that under pressure, they eventually admit to being in the grip of the emotion. However, even if such explanations of alleged emotions without awareness fail, it remains the case that there is no way of determining conclusively that the emotion denier really did “have” the alleged emotion. Furthermore, it remains true that our normal way of thinking and talking about emotions is in terms of them being experienced, and as Clore (1994) pointed out, echoing Freud (1915/1975, for whom the notion of an emotion without awareness was an oxymoron), emotions involve an experience “and one cannot have an experience that is not experienced” (p. 285).
There are, however, serious arguments in support of the idea of emotions without awareness (e.g., Kihlstrom et al., 2000; Winkielman & Berridge, 2004). In an interesting article titled “Unconscious Emotion,” Winkielman and Berridge (2004) argued in support of “the existence of genuinely unconscious emotions” (p. 120). They pointed to studies in which subjects, after being subliminally exposed to valenced stimuli (happy or sad faces), showed biases in judgments and behavior that were congruent with the valence of the stimuli (positive stimuli being associated with greater liking and negative stimuli with less liking of meaningless stimuli or with more or less consumption of a novel beverage). They interpreted such findings as showing that some intervening unconscious emotion caused the observed changes in behavior. However, there are two ways in which Winkielman and Berridge’s findings can be interpreted without appealing to unconscious emotions. First, if an unconscious emotion is responsible for the observed biases, one can ask what particular emotion it was. One then notices that in reviewing the evidence, Winkielman and Berridge never once wrote of an unconscious emotion, choosing instead to always talk about unconscious emotional, or more often, affective reactions. There is no mention of any specific but inaccessible-to-consciousness emotion influencing subjects’ responses. Nor is it necessary to postulate such a thing because the findings they reported can be explained in terms of undifferentiated affect rather than in terms of some (inaccessible-to-consciousness) emotion. So although the studies they described demonstrate that some sort of valenced information that was inaccessible to consciousness played a role, they do not show that that information was a “genuinely unconscious emotion” as opposed to some objectless, amorphous, affective state.
The second way in which Winkielman and Berridge’s (2004) findings can be interpreted without appealing to unconscious emotions is by focusing on the three-stage process model that their explanation appears to assume. The model starts with the processing of a valenced stimulus of which the subject is unaware. This leads to an affective (or emotional) reaction of which the subject is also unaware, which, in turn, leads to a biased behavior or judgment, the subject being unaware of the bias. However, an alternative model would omit the middle step and propose instead that subliminally presented valenced stimuli influence behavior and judgment through the activation of the punishment or the reward system directly biasing information-processing mechanisms and associated behavioral outputs. In other words, one might imagine a direct-influence mechanism à la LeDoux’s two-system model (e.g., LeDoux & Pine, 2016) that does not require an intervening unconscious affective reaction as a kind of biasing trigger. Affectively biased information-processing mechanisms certainly play a role in all sorts of behavior as well as in felt emotions, but they are not themselves emotions. One might think it preferable to distinguish two constructs, such as undifferentiated affect and emotion, than to conflate them, especially in a context in which one might be playing a role and the other not.
The foregoing discussion suggests that even without a consensual definition of emotion, there are good reasons for at least making a start on trying to reach one, and my proposal that emotions are intentional, valenced, conscious mental states might be considered a step in that direction. At the same time, although I consider these characteristics to be necessary for a state to be an emotion, they obviously do not, and are not intended to, constitute a full answer to James’s question “What is an emotion?” Addressing that question comprehensively would require a detailed analysis of the multicomponential nature of emotions as well as more clarity as to what is necessary for a state to be an emotion as opposed to what is merely typical. Thus, my claim that at a minimum, emotions are intentional, valenced, conscious mental states is metatheoretic, not definitional. Even the most hardened skeptic would surely concur that, at the very least, a mental state that does not possess these characteristics is not a very good example of an emotion.
What Kind of Emotion Is Surprise?
Whereas influential emotion theorists have often subscribed to quite different sets of basic emotions, many have embraced surprise as one of their basic emotions (e.g., Ekman, 1992; Frijda, 1986; Izard, 1977; Plutchik, 1962; Tomkins, 1962), and indeed, it is one of the generally accepted basic six emotions. And although on rare occasions someone raises questions about the status of surprise as an emotion (e.g., Oatley & Johnson-Laird, 1987; Reisenzein, 2000), it is fair to say that most emotion theorists consider surprise to be an emotion and that many proponents of basic emotions consider it to be a basic emotion. Yet I argue that for all its elevated status as a basic emotion, surprise fails to satisfy the minimal requirements that I have proposed for something to be an emotion, and if it is not an emotion, it cannot possibly be a basic emotion. I readily acknowledge that some of my arguments are likely to be more convincing than others; however, taken as a group, they all converge on the same conclusion: that surprise is not an emotion, a conclusion that, if accepted, constitutes a serious difficulty for the basic-emotions construct. At the same time, I hope that the ensuing discussion will shed some light on the nuanced nature of the mental state or states that we refer to as “surprise.”
The valence of surprise
Consistent with the proposed minimalist account of emotion—that emotions are intentional, valenced, conscious mental states—it is clear that a person who is surprised must be surprised about or by something. In other words, the intentionality of surprise is not in dispute. Nor does it make sense to deny that surprise is a conscious state 3 —one cannot be surprised about or by something but yet be unaware of it. What is less clear, however, is that surprise is necessarily valenced.
What is the status of surprise with respect to the requirement that emotions are valenced? Is surprise a particular kind of good feeling or a particular kind of bad feeling? Many authors have answered this question by claiming that surprise is intrinsically negative (e.g., Miceli & Castelfranchi, 2015; Noordewier & Breugelmans, 2013; Topolinski & Strack, 2015). One argument in support of this position is that the interruption of ongoing cognitive processes associated with surprise represents a disruption that is fundamentally unpleasant. For example, according to Topolinski and Strack (2015), “surprising stimuli are cognitively disfluent (since they are schema-discrepant by definition) and thereby elicit immediate negative affect” (p. 2), whereas Noordewier and Breugelmans (2013) claimed that surprise is intrinsically negative because the associated unexpectedness “frustrates people’s need for predictability and structure” (p. 1328). Unfortunately, the force of the proposal that the interruption of ongoing processes is disruptive and therefore unpleasant is weakened by hedges that signal limitations on its generality, hedges such as that surprise “tends [emphasis added] to be unpleasant” (Noordewier & Breugelmans, 2013, p. 1237) and that the attendant disruption (due to belief inconsistency) is “in general [emphasis added] unpleasant” (Miceli & Castelfranchi, 2015, p. 50). But these limitations on the generality of the claim open the door to the conclusion that surprise is negative, except when it is not, thus reducing the plausibility of the claim that surprise is intrinsically negative.
In fact, skepticism about the idea that surprise is intrinsically negative (see also Reisenzein et al., 2017) can be justified on other grounds, such as the counterintuitive explanations it requires of the everyday experience of affectively positive surprises. Defenders of the position that surprise is intrinsically negative explain that in cases of positive surprises, the immediate intrinsic negativity of the surprise is overridden by positive valence once the positive significance of the surprising event is understood (e.g., Topolinski & Strack, 2015). However, a simple gedanken experiment demonstrates the implausibility of this kind of explanation. Imagine that you have an excruciating toothache that you are convinced will necessitate some thoroughly unpleasant dental treatment, which you are dreading. Then, quite unexpectedly, your toothache is gone. Suddenly, you are pain free, and you are amazed (i.e., greatly surprised). It stretches credulity to suppose that the surprise resulting from the sudden alleviation of your pain is intrinsically negative but that the negativity is somehow immediately neutralized by the realization that you no longer have to suffer the pain and that you no longer have to fear some dreaded dental procedure. This same example seems to undermine the idea that prediction failure is necessarily a source of negative valence. Surely, it depends on what, if anything, is being predicted. More generally, the suggestion that the surprising unexpected cessation of a noxious stimulus is fundamentally aversive simply defies logic. If that were so, would not, contrary to fact, relief be a negative emotion?
A second example that might make one wary of concluding that pleasant surprises are initially affectively negative is much simpler than the toothache example. Suppose that you are consciously looking forward to some modestly positive event such as a tax refund of, say, $100, but then when the refund arrives, you find that it is much bigger, say, $500. You would presumably be quite surprised and no doubt delighted too. There must surely be a better explanation of your reaction than that your initial negative response was overridden by your delight once you understood the implications of getting $500 compared with $100. Unexpectedly getting something better than an already expected good thing seems to leave no room at all for negative affect. Examples such as these suggest that the idea that surprise is intrinsically negative is neither parsimonious nor consistent with common sense.
The apparent appeal of the idea that surprise is intrinsically negative might have its origin in a too narrow conception of surprise—a conception that equates surprise with expectation failure. Surprise is complicated and cannot be reduced to a unitary notion of expectation failure. There are many different ways in which it can arise (Ortony & Partridge, 1987). One way is indeed through the failure of an active, consciously entertained expectation—an explicit prediction—as when, for example, one anticipates that one’s league-leading favorite sports team will win its upcoming game against an inferior opponent. However, many cases of surprise involve the failure not of explicit predictions but rather of implicit (i.e., not consciously entertained) expectations. These might be better called background assumptions, and they include the general assumption that under normal conditions, well-established regularities will continue to hold. We do not wake up every day actively expecting the sun to rise, yet we would certainly be surprised were we to wake up one day and dawn were never to come. And at a more mundane level, if upon mindlessly turning on a light switch, the light fails to come on, one would again be surprised by the failure of a background assumption, this time the well-established regularity that when you flip a light switch, typically, the light comes on. Arguments in support of the idea that surprise is intrinsically negative often hinge on examples in which explicit predictions are violated—arguments that carry no weight when this condition is not satisfied.
Surprise is always the result of the registration of a discrepancy 4 between what is encountered and some reference point, but that reference point can be not only what was predicted but also what was assumed or presupposed and never consciously entertained. In many cases, faulty assumptions that lead to surprise can easily be explained away—the light bulb must have blown. But in other cases, violated assumptions cannot be so easily explained, as when, for example, you encounter a dog with five legs; or you walk into a room and see a cat floating, unaided, in midair; or out of the blue, a brick comes crashing through your window. The point is that there are different sources of surprise. Some involve the violation of norms (a nun who turns out to be a con artist), others violate what we perhaps naïvely consider to be the laws of probability (spinning a coin and getting 20 heads in succession), and yet others, were they ever to occur, would violate what we consider to be laws of nature (a failure of the sun to rise or a floating cat). Many sources of surprise do not involve expectation violations at all because there was no (prospective) expectation to be violated.
People can assign some likelihood estimate to the things that they predict or assume. Usually, these estimates are quite imprecise, simply falling into broad subjective categories such as extremely likely or unlikely, very likely or unlikely, quite likely or unlikely, and so on. Furthermore, this kind of general sense of the subjective likelihood of something, although prospective in the case of active predictions, is more often than not retrospectively evaluated after the fact by recruiting some norm or standard in terms of which the event is evaluated (Kahneman & Miller, 1986). But clearly, if one is asked why one is surprised by something, the answer is always in terms of the subjective improbability of the event—“I’m surprised I lost my bet. I was sure my team was going to win” or “Nobody could have predicted that a brick would come crashing through my window.”
When unexpected wonderful or awful things happen to people, their surprise is of course accompanied by the positive or negative affect deriving from the wonderful or awful thing itself. This is because whatever it was that transpired affects them. But when they encounter or hear of an unexpected occurrence that does not affect them, the unexpectedness itself might well lead to surprise, but it is a detached surprise. They might find it interesting, but they really do not care one way or another, and indeed, a great many cases of surprise are just like that, valence-neutral, and the fact that it is possible to have valence-neutral surprise is inconsistent with the idea that surprise is intrinsically negatively valenced. All that is needed for valence-neutral surprise is that the surprising event be of no subjective importance to the person evaluating it. After all, unless the longevity of fish somehow matters to you, you do not really care that some fish live longer than humans. It might be hard for you to believe, and it might be very surprising, but it just does not affect you one way or the other, and as argued earlier, emotion and indifference cannot coexist. Surprise, rather than being intrinsically negative, is actually intrinsically unvalenced, and when it appears to be valenced, it is because of the valence resulting from the evaluation of the surprising event. It does not emanate from the surprise itself.
Note that there is one line of argument that might be proffered against this conclusion, a line of argument that depends on studies that show more activity of the corrugator supercilium muscle (which draws the eyebrows together as in a frown) when viewing pictures of unpleasant scenes as opposed to pleasant scenes (Cacioppo et al., 1986). On the basis of such findings and to show that the immediate reaction to surprising stimuli is negative, Topolinski and Strack (2015) monitored the activity of the corrugator supercilium muscle while subjects read trivia statements that varied in how surprising they were. The data showed greater immediate activity when subjects read statements (that had been independently rated) as more surprising than when they read statements rated as less surprising (e.g., the invention of matches is attributed to Johnny Walker vs. women have a higher life expectancy than men). However, such findings do not establish that the furrowing of the brow (the frown) is itself a reflection of the processing of negative affect. It could be a reflection of, for example, puzzlement or even concentration. In addition, and perhaps more telling, there is no reason to believe that subjects in these experiments experienced any negative feelings in response to the stimuli, and if, as argued above, emotions are valenced and if subjects in these experiments are responding to surprising events but not experiencing any negative affect, then the surprise they are experiencing is not a negatively valenced feeling. Furthermore, because it is certainly implausible to suggest that surprise is intrinsically positive, and indeed, to the best of my knowledge, no scholar has ever suggested that it is, the most reasonable conclusion to be drawn is that surprise has no intrinsic valence.
The idea that surprise is intrinsically unvalenced can be reinforced by an argument about the way in which surprise is linguistically coded. At least in English, when we talk about cases of surprise that are not valence-neutral, they are linguistically marked. We routinely speak of pleasant and unpleasant surprises (which already presupposes that some surprises are neither pleasant nor unpleasant, that is, that some surprises are indeed valence-neutral). Notice also that in everyday uses of language, antonymous affective adjectives cannot freely modify genuine emotions words—we do not speak of unpleasant joy or of pleasant disgust with the same ease with which we can speak of pleasant and unpleasant surprise. The most reasonable inference to be drawn from the fact that the word “surprise” frequently appears with valence-indicating modifiers such as “nice,” “pleasant,” “nasty,” “unwelcome,” and so on is that with respect to its intrinsic valence, surprise is neutral and that when surprise is valenced, it is either linguistically marked or pragmatically inferred (e.g., by implicature).
Evidence from other languages provides further support for the idea that surprise is fundamentally affectively neutral. Mandarin, for example, has a completely separate character, 惊 (jing), for neutral (valence-free) surprise. In fact, Chinese speakers tend not to use single characters, preferring instead to make two-character combinations such as 惊讶 (jingya), which nevertheless conveys essentially the same valence-free meaning as 惊 (jing). The valence-neutral meaning of 惊讶 (jingya) contrasts sharply with 惊喜 (jingxi)—roughly positive surprise—and with 惊悚 (jingsong) and 震惊 (zhenjing), both of which are used to refer to negative surprise. Thus, 惊 (jing) denotes valence-neutral surprise, whereas 喜 (xi) marks it as affectively positive and 悚 (song) and 震 (zhen) mark it as affectively negative. If surprise were intrinsically (negatively) valenced, there would be no need for the language to have a means of expressing the valence-free sense.
If surprise is intrinsically unvalenced, and therefore not an emotion, what is it? A reasonable answer is that surprise is a purely cognitive state (along with states such as alert, confident, confused, and interest), that is, a mental state in which cognition dominates in contradistinction to purely affective states (i.e., emotions), which are mental states in which affect dominates (see Note 2 and Ortony et al., 1987). Alternatively, one might think of surprise as what Reisenzein (2000) called a metacognitive feeling, even though later, Reisenzein et al. (2017, p. 20) asserted explicitly that surprise is an emotion. Some have attempted to salvage surprise as an emotion by relegating it (along with other states such as interest and curiosity) to a special class of epistemic emotions (e.g., Morton, 2010; Valdesolo et al., 2017). However, this move uncritically assumes that such states are indeed emotions, and proponents of the view rarely specify their criteria for something to be an emotion. The position appears to be simply that any reaction (feeling, for Morton) that pertains to knowledge is, ipso facto, an emotion. Indeed, Morton (2010, p. 389) actually acknowledged and discussed “dispassionate” cases of such states but nevertheless appears to assume they are emotions simply because we can speak of “feeling” them. But we can feel all kinds of things that are not emotions. One can feel itchy, or nauseous, or cold, or thirsty, and although one might not like being in those states and although they certainly can be the objects of emotions, the feelings themselves cannot reasonably be called emotions. Furthermore, one can feel like going to the movies or not feel enthusiastic about going away for the weekend. If preferences, desires, and bodily states are emotions because they can be felt, we have opened the flood gates with no hope of reprieve. We need some constraints on what it is for a condition to be an emotion, and intentionality, valence, and conscious awareness seem to do a good job of providing them.
All of this suggests that it might be a mistake to think of surprise as an emotion. Its intrinsic affective neutrality is certainly not a characteristic of good examples of emotions; it violates the minimal requirement that emotions are valenced.
Surprise and emotion intensity
There is one other interesting reason for doubting that surprise is an emotion, a reason having to do with the relation between surprise and emotion intensity. Because this relation is more comfortably expressed in terms of the relation between unexpectedness and emotion intensity rather than between surprise and emotion intensity, I, as have others (e.g., Reisenzein et al., 2017; Valdesolo et al., 2017), treat surprise and (the registration of) unexpectedness as identical. However, caution needs to be exercised when equating surprise and unexpectedness because there is a unidirectional causal relation between the two. The feeling of surprise is caused by unexpectedness, not the other way around. Nevertheless, if we make the simplifying assumption that judgments about the unexpectedness of a stimulus and reported surprise are perfectly correlated, then we can, for practical purposes, legitimately treat unexpectedness as a surrogate for surprise.
There is good reason to believe that unexpectedness (surprise) plays an important role in modulating the intensity of emotions (Ortony et al., 1988). This general idea is the foundation of the uncertainty intensification hypothesis (e.g., Bar-Anan et al., 2009) that uncertainty intensifies affective reactions as well as being discussed in the context of music cognition under the rubric of contrastive valence (Huron, 2007). Convincing direct evidence of the influence of unexpectedness on emotion intensity was demonstrated by Mellers et al. (1997) in their work on decision affect theory. In an experiment (Experiment 1) involving gambles, these authors found that their participants reported more positive affect (elation) given an unexpected small win than given a more expected larger win. Furthermore, for losses, participants reported more negative affect (disappointment) on getting an unexpected smaller loss than on getting an expected large loss. In other words, unexpectedness not only influenced the intensity of both positive and negative emotions but also its influence was greater than that of the value of the outcome.
But here comes a possible conundrum for advocates of the view that surprise is an emotion. If, as the evidence suggests, unexpectedness (and therefore surprise) modulates the intensity of emotions, we have a paradox: The intensity of an emotion, surprise, is modulated by itself. And if, indeed, surprise and the subjective (i.e., the awareness of) unexpectedness are the same thing, the paradox cannot be resolved by arguing that it is the unexpectedness that modulates the surprise. One solution might be to define surprise in terms of the discrepancy between the current situation and some prospective or retrospective reference point. One could then argue that when the intensity of an emotion appears to be influenced by surprise, what is actually happening is that the discrepancy is independently influencing the intensity of both the emotion and the surprise (as it does many other things, such as learning and attention), thereby giving the illusion that the surprise is modulating the intensity of the emotion. Of course, this kind of explanation would not establish that surprise is an emotion, it would merely escape one of the curious consequences of supposing that it is, and continuing to maintain that surprise is an emotion still would require one to abandon even the minimal requirement that emotions are valenced.
There is another strange consequence of the idea that surprise is an emotion and especially that it is a negatively valenced emotion. If surprise modulates the intensity of emotions and if surprise is itself an emotion, then in theory, any emotion whose intensity was being modulated by unexpectedness would always have to be accompanied by some level of a second emotion—surprise. This would be particularly bizarre in the case of positive emotions for those who also insist, as we have seen that many do, that surprise is negatively valenced because it would mean that intense positive emotions would always have to have the intrinsically negative emotion of surprise in tow. This would seem to be a less parsimonious account than to acknowledge that emotion intensity, like emotion differentiation, is generally determined by cognitive factors. If we treat surprise as a cognitive state, it falls nicely into this kind of framework, but if we treat it as an emotion, we have to endow it with some new special status—a “super-emotion” (negative to boot) that often occurs to some degree with all other emotions while at the same time having fundamental characteristics that are importantly different from them.
Taken together, the various problems associated with treating surprise as an emotion suggest that it is both more plausible and theoretically more elegant to acknowledge that surprise is a cognitive state rather than insisting that it is an emotion. And of course, if surprise is not an emotion, it cannot possibly be a basic emotion, even though, like hunger, it is without doubt a basic biological response. Surprise is a basic nonemotion!
Finally, suppose my arguments that surprise is not even an emotion, let alone a basic emotion, succeed, how should we respond to the critic who says “OK, so I accept that surprise is not an emotion. That just means that it was a mistake for me to include it in my list of basic emotions. Everything else is fine”? The counter to such a response would be to point out that with respect to the entire community of emotion theorists who are committed, in many instances, deeply committed, to the basic emotion construct, surprise is not an isolated case. There are many examples of proposed basic emotions that appear to violate one or more of the minimal conditions I have proposed for a mental state to be an emotion—that it be intentional, valenced, and present in consciousness. Were they all mistakes? Reviewing Table 1, we find, at least, courage (Arnold, 1960), confusion (Cowan & Keltner, 2021), subjection (McDougall, 1908/1942), interest (e.g., Frijda, 1986; Izard, 1977; Tomkins, 1984), desire (Arnold, 1960; Cowan & Keltner, 2021; Frijda, 1986), seeking, lust, care, play (Panksepp & Watt, 2011), and acceptance and expectancy (Plutchik, 1962). Of course, someone could salvage his or her favorite candidates by simply rejecting my minimalist account of emotion, but then the onus would be on that person to come up with better criteria. Either way, it is difficult to resist the conclusion that as a community, basic-emotions theorists are, at the very least, muddled.
Conclusion
If lists of basic emotions, including the widely accepted basic six, embrace as basic emotions states that are not even emotions, or at least states whose status as emotions can reasonably be questioned, where does that leave the intuition that there must be a small set of privileged emotions—the basic emotions? One answer to this question would be to acknowledge that the intuition is just that, an intuition, and so to take seriously the possibility that the concept of basic emotions is a shaky theoretical construct that lacks a compelling or coherent evidentiary basis. Then, instead of trying to reconcile the ever-changing diverging lists of alleged basic emotions, one could simply acknowledge that emotions result from the interactions of different kinds of fundamental (yes, basic) biological processes and products such as approach and avoidance motivation, cognitions, and core affect (Kringelbach & Berridge, 2017; Russell, 2003). Thus, for example, one could start by granting, as did Barrett and Russell and others (e.g., Barrett & Russell, 2015; Russell & Barrett, 1999), that there are two basic affective states—two amorphous, generic, undifferentiated, states, one of positive and one of negative affect. From there, it is but a short step to view emotions as being just particular ways of feeling good or bad. Of course, this immediately leads to the objection that there are all kinds of ways of feeling good or bad, such as being in pain or being extremely hungry, that we might not want to call emotions. This is indeed true, but unlike pain and hunger, emotions are first and foremost cognitively elaborated ways of feeling good or bad, and emotion types differ from one another in terms of the specific nature of the cognitive elaboration.
Viewing emotions as cognitively elaborated feelings of pleasure or displeasure is not new. As discussed in detail by Reisenzein (2006), it dates back at least to Aristotle’s Rhetoric and in more recent times was prevalent in the late 19th century through the work of the philosopher Alexius Meinong, and the idea that particular emotions differ from each other in terms of their different underlying cognitions is, as discussed earlier, the hallmark of cognitive appraisal theories of emotion. Treating emotions as cognitive elaborations of undifferentiated positive or negative affect allows one to completely bypass the question of basic emotions while providing another way (distinct from the intentionality requirement) of separating genuine emotions from states such as hunger and pain and pleasant bodily feelings such as the feeling one might have during a pleasant massage (even if one’s language has no word for it). This approach might help move us toward an answer to questions such as the one posed in the section, The Vagueness of the Concept of Emotion, as to why one might want to deny that hunger is an emotion or, more generally, how we might have a principled way of deciding to which category some particular feeling state belongs.
In the end, the important questions about emotion are not about whether this or that affective state is a basic emotion. Questions such as “Are there basic emotions?” and “Which are the basic emotions?” are probably as unanswerable as they are unimportant. The important questions have to do with issues such as the role of emotions in judgment, decision-making, and memory; the management, control, and regulation of emotions in self and others; and the causes and consequences of “abnormal” emotional reactions and affective disorders. For people who study anxiety or depression with a view to developing more effective preventions, treatments, and cures, the concept of basic emotions is of little or no practical, or even theoretical, significance. They care not one iota about whether anxiety (or fear) or sadness or shame are basic emotions or about what it could possibly mean to say that they are.
The fact that lists of basic emotions are widely discrepant, often containing items that might not be emotions at all, puts enormous pressure on the basic-emotions construct, but it also puts pressure on the emotions construct itself. The claim that the intrusion of nonemotions into lists of basic emotions merely reflects “a disagreement between competing theories of emotions” (Scarantino & Griffiths, 2011, p. 446) is not at all persuasive. What such intrusions really reflect is a fundamental problem in the field: Not only do different theorists have different conceptions of basic emotions, but also, more seriously, they have different conceptions of what it is for a mental state to be an emotion, which means that competing theories of emotions are not necessarily competing theories of the same thing. I have proposed that at a minimum, mental states have to satisfy three necessary (albeit not jointly sufficient) conditions to qualify as emotions—they have to be intentional, valenced, and conscious. I believe that this constitutes a good starting point for debate—a debate that is much more important than the moribund issue of basic emotions.
Footnotes
Acknowledgements
I am grateful to Gerald Clore, Jelena Radulovic, Rainer Reisenzein, and two anonymous reviewers for helpful comments on various aspects of this article, and I particularly thank Alan Fridlund, the third reviewer, for his extraordinarily detailed and constructive feedback on the submitted manuscript. Thanks also go to Chinese native speakers Jiajia Liu, Tony Tang, and Yinping Yang for their tireless efforts to explain to me some of the nuances of Mandarin.
Transparency
Action Editor: June Gruber
Editor: Laura A. King
