2.1. The Expressivist Fallacy

The Expressivist Fallacy finds it roots in Darwin’s The Expression of Emotions in Man and Animals (Reference Darwin1872), and it has shaped Ekman’s (Reference Ekman, Dalgleish and Power1999a) basic emotion theory. By emotional expressions, Darwin (Reference Darwin1872, 29) meant “changes in any part of the body,” with primary examples involving facial, postural, and vocal bodily changes, to which I collectively refer as nonverbal bodily changes.

According to Darwin, nonverbal bodily changes need to be involuntary in order to express emotions. Darwin’s most central explanatory principle when it came to the origin of emotional expressions was that of serviceable associated habits.Footnote ¹ It stated that “some actions, which were at first performed consciously” and were serviceable in the evolutionary history of the species, “have become through habit and association converted into reflex actions” (41), despite no longer being serviceable for practical purposes.

Consider the baring of the teeth. For Darwin, it used to be part of a voluntary action—biting—that was serviceable in our ancestral past in conflict situations and became an anger expression when it started being reflexively associated with anger, despite biting having exited the aggressive behavioral repertoire of our species.

Ekman followed Darwin in assuming that facial expressions, the main variety of nonverbal bodily changes he studied, must be involuntary. However, Ekman (Reference Ekman, Dalgleish and Power1999a, 47) acknowledged that facial expressions may have evolved because they were serviceable in our ancestral past for communicative purposes: “I believe it was central to the evolution of emotions that they inform conspecifics, without choice or consideration [i.e., involuntarily], about what is occurring.”

And whereas Darwin thought of emotions as feelings, Ekman (Reference Ekman, Dalgleish and Power1999b) identified them with basic affect programs that may or may not involve feelings and are primarily selected to solve recurrent evolutionary problems like escaping dangers (fear), fighting (anger), responding to losses (sadness), dealing with novelty (surprise), and so on. They provide solutions because they cause a cascade of coordinated physiological, expressive, phenomenological, and behavioral responses that allow agents to quickly and efficiently deal with life challenges (Tooby and Cosmides Reference Tooby, Cosmides, Lewis, Haviland-Jones and Barrett2008).

Ekman (Reference Ekman, Segal, Weisfeld and Weisfeld1997) acknowledged that (facial) emotional expressions can carry several types of information. For example, an anger expression carries the information that “someone insulted/offended/provoked her” (antecedents), that “she is planning to attack that person,” that “she is remembering the last time someone insulted her” (person’s thoughts: plans, expectations, memories), that “she is feeling very tense” (internal physical state), that “she is boiling” (a metaphor), that “she is about to hit someone” (what the expresser is likely to do next), that “she wants the person who provoked her to stop what he/she is doing” (what the expresser wants the perceiver to do), and that “she is angry” (an emotion word; 316, 318).

This is a promising account of the domains of information carried by emotional expressions, but it is not reflected in the empirical work done by basic emotion theorists, who have mostly focused on collecting evidence about how facial snapshots (deprived of any context) carry information about “emotion words” across cultures (Ekman Reference Ekman and Rorty1980, Reference Ekman, Dalgleish and Power1999b). This method has been criticized for a variety of methodological and conceptual reasons (Fridlund Reference Fridlund1994; Russell Reference Russell1994), but this is not my concern here.

My concern is that the snapshot method, with its exclusive focus on pairing faces with emotion words while neglecting all the other types of information carried by emotional expressions, has wrongly suggested that the primary communicative point of emotional expressions is to express emotions. This is nothing other than the Expressivist Fallacy, a mistaken assumption about what emotional expressions are for, which behavioral ecologists have influentially criticized, only to replace it with an equally misguided fallacy of their own.Footnote ²

2.2. The Nonexpressivist Fallacy

The Nonexpressivist Fallacy finds its roots in Fridlund’s critique of basic emotion theory in Human Facial Expression (Reference Fridlund1994), and it has shaped the behavioral ecology view of bodily displays ever since. Largely inspired by Krebs and Dawkins’s (Reference Krebs, Dawkins, Krebs and Davies1984) critique of classical ethology, Fridlund argued that no signal could possibly evolve to involuntarily inform recipients about the inner states of signalers. “Automatic readouts or spillovers of [inner states],” Fridlund (Reference Fridlund1994, 109) stated, “would be extinguished early in phylogeny in the service of deception, economy, and privacy.” This is because in competitive contexts involuntarily revealing one’s inner states may put the signaler at a disadvantage (e.g., broadcasting one’s fear during a confrontation).

Fridlund proposed to replace the notion of an involuntary facial expression of emotion with that of a facial display produced in an audience-dependent fashion when signalers expect benefits from it. Audience dependence entails that signalers tailor their context-sensitive facial displays to their audience and do not produce them indiscriminately upon experiencing a given emotion. Displays are “declarations that signify our trajectory in a given social interaction, that is, what we will do in the current situation, or what we would like the other to do” (Fridlund Reference Fridlund1994, 130).

Fridlund (Reference Fridlund1994) provided a list of “declarations” signified by facial displays, contrasting it with the list provided by Ekman.Footnote ³ The Nonexpressivist Fallacy is exemplified by the contrast between the left and the right columns in table 1, which shows that for behavioral ecologists facial displays only declare social motives and do not express emotions. Since there is no consensual scientific definition of emotion, Fridlund thinks that we are better off avoiding any reference to emotional expressions. “What cannot be done,” he concludes, “is to show that emotions have no role in facial displays, because excluding emotion would require a [scientific] definition that allows it. At present, arguing against ‘emotion’ in any form is shadow boxing” (Fridlund Reference Fridlund1994, 186).

Fridlund is here conflating two varieties of lack of consensus. One is lack of scientific consensus on what emotions are—a distinctive feature of contemporary emotion theory (Scarantino Reference Scarantino, Lewis, Haviland-Jones and Barrett2016). The other is lack of scientific consensus on whether something is an example of emotion. There is no lack of consensus of this second sort: alternative definitions of emotions agree on what count as prototypical examples of emotion (Scarantino Reference Scarantino2017b).

This is common in scientific fields before the emergence of consensual definitions. Chemists disagreed for centuries about how to define water but agreed about which items prototypically qualified as water, which explains why they investigated the chemical structure of a shared set of samples despite their theoretical differences.

Similarly, even though affective scientists disagree on what fear or anger are, they largely agree on whether certain displays are manifestations of fear or anger (or neither). Suppose someone meets a bear and responds to the threat with upper eyelids raised, jaw dropped open, and lips stretched horizontally, while manifesting a higher heart rate, tremors, and an avoidance tendency. In such circumstances, most emotion theorists would conclude the subject is afraid. Thus, despite definitional disagreements, we can establish with the force of scientific consensus whether prototypical instances of any given emotion are instantiated.

Fridlund (Reference Fridlund1991, 39) is also convinced that distinguishing between emotional and nonemotional displays is explanatorily irrelevant because social “motives bear no necessary relation to emotion, and indeed, a range of emotions can co-occur with any social motive.” Since only social motives matter predictively, we should focus on them exclusively.

This is another red herring because the absence of a necessary relation is compatible with the presence of a predictively relevant probabilistic relation. The question is not whether perceiving that the signaler is having a certain emotion necessitates a certain social motive but whether it changes the probability of a certain social motive.

Most scientific theories accept that emotions and action tendencies are constitutively or causally related. For example, it is widely acknowledged that anger involves a tendency to attack, fear a tendency to avoid, happiness a tendency to affiliate, and disgust a tendency to expel (Frijda Reference Frijda1986). It follows that if I realize that my opponent’s pursed lips express anger rather than pain, I can predict that an attack is now more likely than it would have been had my opponent been in pain. If so, failure to tell the difference between angry and pained expressions would likely have been extinguished early in phylogeny in favor of the ability to exploit probabilistic relations between inner states and social motives for predictive purposes.

To sum up, neither lack of scientific consensus on emotions nor lack of a necessary relation between emotions and social motives supports removing the notion of emotional expression from affective science. The Nonexpressivist Fallacy should to be rejected: there is nothing wrong with the idea that displays on occasion express emotions in a certain context. At the same time, we should reject the assumption basic emotion theorists have long held that emotional expressions must be involuntary: behavioral ecologists are right that most nonverbal bodily changes are audience-dependent communicative tools at the service of social motives (Scarantino Reference Scarantino2017a).

3.1. How to Do Things with Words

Speech act theory is the brainchild of Austin’s How to Do Things with Words (Reference Austin1962). Austin noted that “it was for too long the assumption of philosophers that the business of a ‘statement’ can only be to describe some state of affairs” (1). To explain what “businesses” sentences may have other than describing, Austin formulated a distinction between three things we “do” when we utter a sentence “X”:

Locutionary acts are acts of speaking or writing, performed in conformity with syntactic and semantic rules about the sense and reference of the words employed. Illocutionary acts are instead acts done in speaking. Here I borrow Searle’s (Reference Searle1979) taxonomy of illocutionary acts, generally considered to be an improvement over Austin’s own. According to Searle, there are five and only five things we can “do” with words: “We tell people how things are (Assertives), we try to get them to do things (Directives), we commit ourselves to doing things (Commissives), we express our feelings and attitudes (Expressives), and we bring about changes in the world through our utterances ([Proclamatives])” (Searle Reference Searle1979, viii).Footnote ⁴

Searle added that speech acts involve the expression of psychological states. For example, uttering “the cat is on the mat” is engaging in an Assertive and expressing a belief, uttering “pass me the salt!” is engaging in a Directive and expressing a desire, uttering “I will do it” is engaging in a Commissive and expressing an intention, uttering “I apologize” is engaging in an Expressive and expressing regret, and uttering “I declare you husband and wife” is engaging in a Proclamative and expressing the desire that two people be husband and wife and the belief that they will be if you say so.

Perlocutionary acts, finally, are acts done by uttering sentences. Uttering “the cat is on the mat” may convince the hearer that the cat is on the mat, uttering “pass me the salt!” may get the hearer to pass you the salt, uttering “I will do it” may get the hearer to expect that you will do it, uttering “I apologize” may get the hearer to forgive you, and “I declare you husband and wife” may legally bind two people in matrimony.

Searle (Reference Searle1969, 42) emphasized that speech acts involve intentions to communicate: “in speaking I attempt to communicate certain things to my hearer by getting him to recognize my intention to communicate just those things.” This is what Grice (Reference Grice1957) called a reflexive intention, namely, an intention on the part of the speaker to produce an effect in an audience at least in part by virtue of such intention being recognized. Reflexive intentions are involved in the production of nonnatural meaning (a.k.a. nonnatural information), which Grice distinguished from natural meaning (a.k.a. natural information). Compare the following two sentences:

In the first case, the notion of meaning is used in its natural sense: bared teeth in the context of a parking lot confrontation naturally mean anger because they are statistically correlated with anger (but bared teeth while lifting a weight at the gym do not naturally mean anger). In the second case, the notion of meaning is used in its nonnatural sense: “I am angry” uttered in the context of a parking lot confrontation nonnaturally means that I am angry because, on Grice’s view, Tom uttered it intending the hearer to come to believe that Tom is angry in part because of recognizing Tom’s intention to make him believe that he is angry (but “I am angry” said into a microphone to test it does not nonnaturally mean that the utterer is angry).

I define a Speech Act Analog (SAA) as any behavior that naturally means roughly what a speech act nonnaturally means. So Tom’s baring of the teeth is a SAA of the Expressive speech act of uttering “I am angry.” The speech act nonnaturally means that the utterer is angry by offering evidence of his intention to communicate just that. But the baring of Tom’s teeth also naturally means that Tom is angry because being angry and baring one’s teeth are statistically correlated. The question is: Which SAAs do emotional expressions make available?

3.2. How to Do Things with Emotional Expressions

I distinguish between three things we “do” when we express emotions:

I understand emotional expressions as nonverbal bodily changes—either voluntary or involuntary—designed to influence recipients through the transfer of natural information about emotions (Scarantino Reference Scarantino and Stegmann2013).Footnote ⁵ TAP’s key proposal is that in expressing emotions we engage in communicative moves analogous to four of the five types of illocutionary acts (EE stands for Emotional Expression):Footnote ⁶

On this view, emotional expressions have several equally important communicative points (contra the Expressivist Fallacy), one of which is to manifest emotions (contra the Nonexpressivist Fallacy). But emotional expressions do not just express emotions: they also synchronously engage in four communicative moves that make explicit the possible social motives associated with nonverbal communication.

Finally, emotional expressions have communicative effects, which are the effects signalers produce by expressing emotions. Van Kleef (Reference Van Kleef2016) has suggested that such communicative effects are brought about by means of affective reactions and inferential processes. For example, by expressing anger I can elicit in you a fear response (affective reaction) and the expectation that I will behave aggressively toward you (inferential process), which may lead you to give up on a contested resource. I will not further discuss communicative effects in this article, focusing my attention on communicative moves.Footnote ⁷

3.3. A Taxonomy of Communicative Moves

TAP holds that, apart from expressing emotions (Expressives_EE), emotional expressions commit the expresser to future actions (Commissives_EE), try to get the recipient to do things (Imperatives_EE), and represent how things are in the world (Declaratives_EE). In other words, by carrying natural information about emotions, emotional expressions also carry natural information about the action tendencies emotions involve, the demands associated with emotions, and the ways emotions represent the world.

The fact that emotional expressions belong to the class of Expressives_EE is accepted by all researchers on nonverbal communication except sympathizers of the behavioral ecology view, and I have examined their skeptical arguments, finding them wanting. The content of Expressives_EE (i.e., what is expressed) is a description of the emotion the expression provides natural information about. Since different expressions manifest different emotions, there is no unique content for emotional expressions under an expressive communicative point. For example, in the appropriate context bared teeth express that the agent is angry, trembling expresses that the agent is afraid, jumping up and down expresses that the agent is happy, and so on.

Commissives_EE have the communicative point of committing the sender to a future course of action by means of natural information transfer. Their content (i.e., what one commits to) is a description of the conditions of satisfaction for the emotional expression under a signaler-based world-to-mind direction of fit, in the sense that the signaler is responsible for changing the world so as to fit the content.

This is what Commissives_EE share with linguistic Commissives: their contents describe how the world is to be changed by the signaler. For example, the utterance of the sentence “I am going to attack you” and the emotional expression of baring one’s teeth (in the right context) mean, nonnaturally in the first case and naturally in the second, that the signaler will attack the opponent unless the matter of contention is resolved.

As reported earlier, Ekman has characterized emotional expressions as carrying “information about the person’s plans” and “information about what the expresser is likely to do next,” two formulations we can treat as roughly equivalent. Fridlund has emphasized how nonverbal displays are declarations of behavioral “intent” or “states of readiness.” We can now combine these two lines of analysis and say that a happy Duchenne smile (in the right context) is an expression of happiness and is informative about or declares “readiness to affiliate,” that the baring of one’s teeth (in the right context) is an expression of anger and is informative about or declares “readiness to attack,” that the raising of the upper eyelids with jaw dropped open (in the right context) is an expression of fear and is informative about or declares “readiness to submit,” and so on.

The reason why emotional expressions inform about action commitments is that emotional expressions naturally mean that the agent is experiencing a certain emotion, and emotions are associated with action tendencies. Importantly, the action tendencies associated with emotions have distinctive characteristics that allow us to appreciate the nature of the commitments being communicated.

Unlike garden-variety action tendencies like ordering eggplants in Asian restaurants, emotional action tendencies have control precedence: they “lie in waiting for signs that they can or may be executed,” they “tend to interrupt other ongoing programs and actions,” and they “tend to preempt the information-processing facilities” (Frijda Reference Frijda1986, 78). For example, the attack tendency associated with anger brings aggressive behavioral options to the top of the goal hierarchy, preempts access to information not related to the ongoing confrontation, and prepares the body for aggressive action.

It is true, as behavioral ecologists have noted, that one can convey readiness to attack without being angry and that being angry does not entail that one will attack. In fact, an expression of anger does not even signal a full-fledged intention to attack, unlike the utterance of the sentence “I am going to attack you.” Nonetheless, an expression of anger signals a commitment to attacking.

As Schelling (Reference Schelling and Nesse2001, 48) has pointed out, “to commit is to relinquish some options, eliminate some choices, surrender some control over one’s future behavior.” What anger expressions do is to show that the agent is in the grip of a motivational mechanism—anger—which leads to relinquishing options for peaceful interaction while activating options for aggressive interaction. One’s declaration of being ready to attack is possible without being angry, but it will be less credible as a commitment than a declaration of being ready to attack backed up by an anger mechanism that manifests itself through an attack tendency with control precedence.

Let us now turn to Imperatives_EE, whose communicative point is to try and get a recipient to do something by means of natural information transfer. The content of an Imperative_EE (i.e., what is demanded) is a description of the conditions of satisfaction for the emotional expression under a recipient-based world-to-mind direction of fit, in the sense that the recipient is responsible for changing the world so as to fit the content. This is what Imperatives_EE share with linguistic Directives: their contents describe how the world must be changed by the recipient. For example, the utterance of “stop it!” and the emotional expression of baring one’s teeth (in the right context) mean, nonnaturally in the first case and naturally in the second, that the opponent should stop what he or she is doing.

Ekman (Reference Ekman, Segal, Weisfeld and Weisfeld1997, 318) has characterized emotional expressions as carrying “information about what the expresser wants the perceiver to do.” Fridlund has spoken of displays as “requests,” but he has limited the requesting function to a handful of cases (e.g., only “sadness” expressions in table 1). On the contrary, all emotional expressions make demands on others. This follows from the insight that there can be no evolutionary premium on expressing internal states unless such expression leads recipients to behave in ways that are beneficial to signalers (Krebs and Dawkins Reference Krebs, Dawkins, Krebs and Davies1984).

On this view, all communicative signals, the genus of which emotional expressions are a species, emerged because of their ability to generate behaviors on the part of recipients that benefited signalers at least on average. What recipients gain from the communicative exchange is information about the signaler’s future behavior or about the environment, which sustains the signaling system by making communication useful to both signalers and recipients.

So what exactly do emotional expressions “demand”? This is one of the least explored topics in the science of emotional expressions, and one ripe for further investigation. Absent a general framework, I offer a tentative list of demands borrowed from Parkinson’s (Reference Parkinson1995, Reference Parkinson2005) work on the “communicative agendas” of emotions. On Parkinson’s (Reference Parkinson1995, Reference Parkinson2005) view, expressing anger is demanding that the recipient takes the signaler more seriously, expressing fear is demanding that the recipient helps and protects the signaler, expressing happiness is demanding that the recipient celebrates a success with the signaler, and so on.

The similarity of content between linguistic Directives and Imperatives_EE should not blind us to their differences. Compare uttering “celebrate with me!” with a happiness expression like producing a Duchenne smile. In both cases, the signaler provides evidence of being in a mental state that would be satisfied if the recipient celebrated with the signaler. Whereas in the case of a linguistic request I offer evidence about my intention to communicate how I desire the hearer to act in the future, to make a “demand” through a happiness expression, I do not have to intend to communicate anything, let alone intend to communicate through the recognition of my intention to do so.

As long as my expression is designed to elicit a celebratory response in my recipient and carries information about my happiness, it communicates my demand for celebrating together. This is the same sense in which a vervet monkey snake alarm call counts as a demand to conspecifics that they avoid the snake, despite the fact that vervet monkeys produce alarm calls independently of the knowledge states of recipients (Maynard Smith and Harper Reference Maynard Smith and Harper2003). Vervet monkeys are not trying to make their communicative intentions manifest to recipients, but they are nevertheless making demands on them.

The final SAA I consider is that of Declaratives_EE, which have the communicative point of representing how things are in the world by means of natural information transfer. The content of Declaratives_EE (i.e., what is represented) is a description of the conditions of satisfaction for the emotional expression under a mind-to-world direction of fit, in the sense that their content aims to fit what the world is like. This is what Declaratives_EE share with linguistic Assertives: their contents try to describe the world as it is. For example, uttering “you have slighted me” and the emotional expression of baring one’s teeth (in the right context) mean, nonnaturally in the first case and naturally in the second, that the recipient has slighted the signaler.

Ekman (Reference Ekman, Segal, Weisfeld and Weisfeld1997, 318) has characterized emotional expressions as carrying “information about antecedents,” whereas Fridlund (Reference Fridlund1994) did not explicitly include descriptive declarations in his analysis. But some of the table 1 examples fit the descriptive bill. For example, what basic emotion theorists would characterize as a “sad face” is for Fridlund a display of “damage” and what they would characterize as a “contempt face” is a display of “superiority.” Neither declaration is a request or a manifestation of a state of readiness but rather a representation of what the world is like. Roughly, a sadness expression represents the signaler as having suffered damage (or a loss), and a contempt expression represents the signaler as being superior to the recipient.

It is generally recognized that emotions have intentionality or the ability to represent and that Lazarus’s (Reference Lazarus1991) list of core relational themes provides a good preliminary account of what specific emotions represent. But if emotional expressions carry natural information about emotions and emotions represent core relational themes, then emotional expressions carry natural information about how the expresser has represented (or appraised or construed or evaluated) what the world is like. On this view, to express sadness is to represent an irrevocable loss, to express anger is to represent a slight against me or mine, to express shame is to represent a failure to live up to an ego ideal, to express fear is to represent danger, and so on.

Although there are similarities of content between linguistic Assertives and Declaratives_EE, there also are differences. Compare uttering “this picture of a snake is dangerous” with a fear expression like raising the upper eyelids with jaw dropped open upon being presented with the picture of a snake. In both cases, the signaler provides evidence of being in a mental state that would be appropriate if the picture of the snake were indeed dangerous. With a linguistic assertion, the signaler is also expressing the belief that the picture of the snake is dangerous. Not so with a fear expression, which may be accompanied by the conflicting belief that the picture of a snake is not dangerous, giving rise to the well-known phenomenon of emotional recalcitrance (D’Arms and Jacobson Reference D’Arms, Jacobson and Hatzimoysis2003). This is to say that the representations associated with emotional expressions differ in kind from the representations associated with linguistic assertions.

3.4. Emotional Expressions Do Not Separate Content and Force

I have argued that four of the five ways of using language singled out by speech act theorists—to express inner states, to commit to doing things, to try to get others to do things, and to represent how things are—are available through the use of nonverbal emotional expressions. This being said, emotional and linguistic communication differ in one crucial respect: in linguistic communication the same illocutionary force can be freely associated with an open range of contents, whereas in emotional communication unaided by language, communicative moves are wedded to restricted contents. For example, I can assert any proposition whatsoever, and I can commit to any future action whatsoever through language. But through an anger expression I can only represent someone’s conduct as a slight and commit to aggressive future action. Similarly, I can demand any behavior on your part and express any inner state through language, but through an anger expression I can only demand that you stop what you are doing and express my anger.

This tells us that in emotional expressions content and force are not independent of one another. Emotional expressions are analogous to holophrases, or one-word phrases lacking compositional structure: demanding-that-someone-stops-what-they-are-doing, committing-to-aggressive-action, representing-something-as-a-danger, expressing-one’s-disgust, and so on.

If emotional communication faces this structural limitation, why do we keep relying on it even when linguistic communication is available? One reason is that emotional communication has a special degree of communicative strength in getting a certain communicative point across. As Searle and Vanderveken (Reference Searle and Vanderveken1985) argued, “different illocutionary acts often achieve the same illocutionary point with different degrees of strength” (15). Requesting has a higher degree of directive strength than suggesting, and swearing has a higher degree of assertive strength than conjecturing.

By the same token, emotional expressions often display more communicative strength than speech acts of analogous content. For instance, expressing anger to Tom has a higher degree of representational strength than merely saying that Tom slighted you, it has a higher degree of directive strength than merely saying that Tom should stop what he is doing, and it has a higher degree of commissive strength than merely saying that you will engage in aggressive behavior toward Tom. This may be due in part to the fact that emotional expressions are accompanied by general bodily preparations for action that provide evidence that the emoter “means” what the emotional expression “shows,” and in part to the fact that emotional expressions are harder to fake (Frank Reference Frank1988).

This higher degree of communicative strength may explain why emotional expressions are often more successful at producing communicative effects than locutionary acts are at producing perlocutionary effects. For instance, expressing my anger is often a better way to make you understand that I consider your conduct offensive, a better way to make you stop what you are doing, and a better way to make you expect that I will behave aggressively than simply telling you each of these things.