1. The Perceptual Theory and the Action Analogy

According to the cognitive theory of emotions, emotions involve states representing certain evaluative properties of their objects. In feeling fear, one represents the object as dangerous. In feeling angry, one represents the object as offensive. In feeling sad, one represents the object as a loss or lack of some kind. On an important variant of this view, the significance of emotions lies in their constituting a distinctive way of representing value, a kind of representation that is not inherited from further mental representations and thus that cannot exist without emotions. This claim is most clearly at play in the so-called perceptual theory of emotion, arguably the most popular version of the cognitive theory today. As the theory goes, just as a visual experience of blue gives you a representation of the color blue that you cannot get from any other mental state (it is an ‘original’ representation of blue), your fear gives you a representation of danger that you could not get from elsewhere (see, e.g., Cowan Reference Cowan2016; Döring Reference Döring2007; Milona Reference Milona2016; Tappolet Reference Tappolet2016; cf. Ballard Reference Ballard2020; Johnston Reference Johnston2001). For this reason, the perceptualist argues that emotions should be seen as a distinctive way to access value and in that respect as having fundamental epistemological significance.

More generally, the perceptualist typically relies on an analogy between emotion and perception (henceforth: perceptual analogy) to achieve the dual aim of providing an account of emotions’ central features and of revealing the way in which emotions are distinctive, ineliminable aspects of our lives (Tappolet Reference Tappolet2016). The perceptual analogy promises to shed light on features of emotions (intentionality, phenomenology, passivity, etc.) that are apparently shared with perception by arguing that they should be elucidated in the same way, in turn elucidating further, more distinctive features of emotions (such as their intuitive link to value) through a perceptual lens. As a result, the perceptual analogy, if taken seriously (Milona Reference Milona2016), provides an interesting and controversial account of our knowledge of value, whereby emotions provide access to value in much the same way sensory perception provides access to sensory features of the environment. The fundamental significance of emotions might thus be claimed to be epistemological: as distinctive perceptual experiences of value, they can play the sort of foundational role commonly attributed to perception. On this sort of story, emotions are significant in that they are our most fundamental way to access value, the same way vision is our most fundamental way to access colors. (Strictly speaking, the perceptual theory is compatible with the denial of this claim; after all, one could deny that visual experience is our most fundamental way to access colors. But this is, I think, a common assumption among the advocates of the perceptual theory.) A creature without emotions would thus be blind to important aspects of the world in a straightforward sense.

The strategy at play in the perceptual analogy is to model core features of emotions on the features of a familiar type of state (perception) and argue that emotions in general play a role akin to that of perception. There are three main ways to resist the resulting picture (see Naar Reference Naar2022a). First, one might resist the attribution of some of the features ascribed to both emotion and perception or deny their centrality. Second, one might argue that emotions have further features that are in tension with the perceptual theory. (A common complaint is that by contrast with perceptual experiences, emotions are supported by reasons and can be responses to reasons [Brady Reference Brady2013; Deonna and Teroni Reference Deonna and Teroni2012; Dietz Reference Dietz2018; Müller Reference Müller2019; Naar Reference Naar2022a].) Third, one might argue against the way the perceptualist elucidates certain features of emotions. One way to do this is by arguing that the central features of emotion are best modelled on another type of response, a type whose distinctiveness might not be epistemological. A recent incarnation of this last strategy is to resist the perceptual theory by providing a different analogy, namely an analogy between emotion and action (Benbaji Reference Benbaji2013; Naar Reference Naar2022a; cf. Ballard Reference Ballard2021)— henceforth to be called the action analogy.

One might be tempted to characterize anyone positing a robust link between emotion and action as pursuing the action analogy. For instance, one could maintain that an advocate of a motivational theory—in claiming that emotions essentially motivate to action—is automatically drawing an analogy (even if implicitly) between emotion and action. This does not follow, however. The action analogy is an analogy between emotion and action, not an analogy between emotion and motivational states such as desire, and thus there is no clear affinity between the motivational theory and the action analogy. That said, everyone—including the perceptualist—is free to appeal to the action analogy for various purposes. The question that I will be concerned with is whether it can lead to an interesting theory of emotion if taken seriously.

As some philosophers have argued, many central features of emotions are features that are apparently shared with certain types of action and can be understood in a similar way (Ballard Reference Ballard2021; Benbaji Reference Benbaji2013; Müller Reference Müller2019; I provide a systematic development of the action analogy in Naar Reference Naar2022a). Indeed, certain platitudes about emotions find clear analogues in action. Here is a sample (see Naar [2022a] for more details): (1) Both emotions and actions can conflict with judgment in a way that counts as a rational conflict and that falls short of contradiction (Benbaji Reference Benbaji2013); (2) both emotions and actions can be intentionally related to the world; (3) and both emotions and actions can be supported by reasons and can be based on reasons (Dietz Reference Dietz2018; Müller Reference Müller2021). Moreover, (4) both emotions and actions can be appropriate or inappropriate, which in turn suggests that appropriateness is not a representational relation but some other relation, arguably one of normative support (Müller Reference Müller2014: 75f.; Reference Müller2017; Naar Reference Naar2021; cf. Ballard Reference Ballard2021; see Naar Reference Naar2021 for a sustained argument for the claim that appropriateness should be seen as a relation of normative support). If it is further assumed that value is what makes emotions appropriate, an assumption shared with the perceptualist, then the relation between emotion and value will not be epistemological but normative—value favors, rather than is accessed by, emotions—which would make emotion disanalogous to perception in some important respect (after all, colors do not favor visual experiences). If appropriateness is in turn to be understood as implying reasons, then values might be reasons for emotions (see Maguire [Reference Maguire2018] for dissent and Faraci [Reference Faraci2020] for a reply; for a sustained argument for the claim that value can both favor and be accessed by emotion, see Mitchell [Reference Mitchell2021]; for dissent, see Müller [Reference Müller2021]).

Further support for these ways of elucidating central features of emotions comes from the fact that to each emotion type a range of action types seems to correspond that can be made appropriate by the same type of value property (Naar Reference Naar2022a). To fear correspond actions such as running away, staying put, and fighting. To admiration correspond actions such as emulating and promoting. To anger correspond actions such as punishing, insulting, and avoiding. Call the action types that can be made appropriate by the same value property as a given emotion type an ‘action counterpart’ of that emotion type. (For the term, see Church [Reference Church, Paperman and Ogien1995], but she means something different by it. Note that because the question of what value property makes an emotion type appropriate is a substantive question, it is an open question what action types are made appropriate by the same value property.) It is important to stress that although they may sometimes constitute expressions of emotions or be performed out of emotions, action counterparts are still responses that can be independently characterized. Even if they are often motivated by emotions (see below), they can be performed without the presence of the relevant emotion, and they can be appropriate or inappropriate in a way that is independent of the appropriateness of the emotion. For instance, while anger might be an appropriate response to a mildly offensive remark, retaliation—an action counterpart of anger—might not be. To be sure, the relevant action types should be individuated in a rather fine-grained way such that it is not the case that the very same action type (e.g., striking) is an action counterpart of two different emotion types (fear and anger). One way to isolate the relevant types is ostensively: the action counterparts of a given emotion type are those that are made appropriate by the same value, and they differ from those actions that are not made appropriate by the same value. Action counterparts will play an important role below.

One might insist that, though distinct from emotions and not necessarily expressions of them, action counterparts should nonetheless be understood in terms of their relation to emotions. This would then mean that it will not be possible, without circularity, to model emotions on their action counterparts. For on this account, emotions would be elucidated by analogy with entities whose nature essentially involves emotions. There are two reasons to reject this skepticism about the appeal to action counterparts. First, as I said before, it does not seem to be the case that every action type that is made appropriate by the same value as the corresponding emotion type (i.e., every action counterpart) bears an essential relation to that emotion. Although punishing bears some affinity to anger (perhaps because anger disposes one to fight, see section 3), it is unlikely that an adequate account of this action type will have to refer to anger. Second, even if this is the case—even if an adequate account of action counterparts will mention their relation to emotions—it might still be held that emotions are analogous to these actions in those respects that do not make reference to emotions. Recall that as I define it, an action counterpart is an act type that is made appropriate by the same value property as some type of emotion. Unless the explanation of why the action is made appropriate by the same value is that it bears a relation to the corresponding emotion (e.g., fighting is appropriate because it expresses anger, which is itself appropriate), which might be doubted, we can still understand some of emotions’ core features by analogy with their action counterparts. Admittedly, more could be said here. Since the primary aim of this paper is to introduce a novel theory of emotions, I hope it is fair to assume the degree of independence I ascribe to action counterparts. The provision of a sustained argument for the claim that emotions and their action counterparts are normatively on a par—or at any rate that emotions do not come first in the order of normative explanation—is work for another time.

Of course, it is an open question whether the action analogy, suitably developed, can capture all the core features of emotions. Elsewhere, I argue that there is no prima facie reason to think that it cannot (Naar Reference Naar2022a). But even without a complete story, we can see the dialectical significance of the analogy. First, if emotions are on the whole better modeled on action than on perception, then a number of claims made on the basis of the perceptual analogy become unmotivated. We are indeed no longer entitled to understand emotions’ various features through the lens of the perceptual theory, and the perceptualists’ interpretation of certain platitudes about emotions is now something they cannot take for granted, given the superiority of this alternative interpretation. (Of course, I have not given reasons to prefer the action analogy over the perceptual analogy. The following discussion is meant to be conditional on the action analogy being taken seriously. For reasons for taking it seriously, see Naar [Reference Naar2022a].) Furthermore, the distinctiveness of emotions may need to be revisited. The availability of the action analogy, at least in its most systematic form,Footnote ¹ should make us doubt that the distinctive role of emotions lies in their being distinctive representations of value with fundamental epistemological significance.Footnote ² Finally, the action analogy constitutes a fertile territory for further research in that it supports a systematic comparative study of emotion and action informed by the important recent advancements in action theory.

Despite its dialectical significance, however, the action analogy might be criticized for not providing a clear account of the emotions. At first sight, the analogy is compatible with the denial of the claim that emotions are actions—after all, analogies are usually drawn between distinct things. Of course, one might insist that the perceptual analogy is officially compatible with the denial of the claim that emotions are perceptions as well. As some perceptualists have admitted, the perceptual analogy might be an imperfect analogy. But as its advocates are keen to claim, the perceptual analogy suggests a substantive picture according to which emotions are fundamentally perception-like (because being fundamentally perception-like does not imply being literally a perceptual experience, the claim here should be acceptable even to those perceptualists who deny that emotions are perceptual experiences). The question, now, is whether the advocates of the action analogy can similarly go beyond the mere elucidation of commonalities and provide a substantive picture of emotions as fundamentally action-like—what I call an agential theory. If they cannot do so, then the action analogy may not be superior to the perceptual analogy after all when it comes to offering a fundamental explanation of emotion. Of course, this is compatible with finding the action analogy dialectically useful in the debate over perceptualism. Given the perceptualists’ fondness of analogies, why not fight their analogy with another analogy (see Naar Reference Naar2022a)? It should also be emphasized that my rejection of various theories of emotion throughout the discussion should be seen as confined to the particular project of formulating a plausible agential theory of the emotions and a new alternative to the perceptual theory (which relies on a different analogy) and not aiming for the larger project of identifying the best theory of emotions. For all I say here, the motivational theory, for example, might be the right theory of emotions, even if—as I later argue—it does not count as a properly agential theory in my sense. Can the idea that emotions are fundamentally action-like be properly defended?

2. Taking the Action Analogy Too Seriously: Emotions as Actions

As I have characterized it, an agential theory of emotions is a theory that claims emotions to be fundamentally action-like, just as the perceptual theory claims them to be fundamentally perception-like. At this stage, this is deliberately vague. The question is how we should understand the action-like character of emotions. Following the perceptualists’ way of proceeding on the basis of their analogy, one might claim that emotions are either full-blown (mental) actions or else actions in a more liberal or minimal sense. Alternatively, one might claim that emotions are constituted by actions.

Notice that this way of pursuing the agential theory has the advantage of being very straightforward. If emotions are so much like actions, this might be because they are actions. And indeed, this sort of view is not without precedent. Some theorists have indeed gone so far as to claim that emotions are, or are constituted by, actions of some sort (Sartre Reference Sartre and Frechtman1948; Solomon Reference Solomon1973; Slaby and Wüschner Reference Slaby, Wüschner, Roeser and Todd2014). One might take so-called enactivism to involve the claim that emotions are actions, but this is controversial. The view might best be seen as a version of the motivational theory that emphasizes the idea that emotions are bodily ways to prepare for action (e.g., Colombetti Reference Colombetti2014; Shargel and Prinz Reference Shargel, Prinz, Naar and Fabrice Teroni2018). That said, the idea that emotions are not representational (e.g., Hutto Reference Hutto2012), at least in the sense of constituting autonomous, original representations, is something that is shared with the view presented here. It should also be noted that the view introduced below is distinct from that involved in the embodied cognition tradition (which might yield a variant of the motivational theory; see Hufendiek Reference Hufendiek2018). Jennifer Church (Reference Church, Paperman and Ogien1995) has proposed a view of emotions as what she calls internalized actions, namely (bodily) actions that have been rendered internal. On this view, fear is an internalized version of fleeing, anger is an internalized version of hitting, and shame is an internalized version of hiding. If this view can be made good, this looks like a version of the view of emotions as actions. Church is quick to point out, however, that emotions are not actions in a robust sense, instead calling them ‘proto-actions’, claiming that emotions and their action counterparts involve similar bodily changes such as tensed muscles and a rise in adrenaline. To the extent that the latter are involved when one is preparing for action, the resulting view might count as a version of the motivational theory. The agential theory introduced below is neutral on the role of the body in emotions.

Even if some theorists have been happy to declare emotions (perhaps mental) actions, and even if the view might sound attractive to some ears, it is indeed difficult to maintain that emotions are actions like any others (including mental actions). Emotions do not seem to have essential features often attributed to action in standard action theory. Emotions do not seem to be ‘intentional under some description’ (Anscombe Reference Anscombe1957); it is certainly unclear that in merely having an emotion, one is intentionally doing something. Emotions do not seem to be the sort of thing that can be caused by the intention to ‘perform’ them. And the idea that emotions are literally things we can perform at will or voluntarily is highly controversial to say the least; coming to be angry with a friend certainly does not look like a product of choice. At any rate, it is common among many emotion theorists, including perceptualists, to claim that emotions are ‘passive’ in a sense that excludes their being voluntarily formed.

That said, nothing I say here is supposed to cast doubt on the sophisticated arguments we might find in the literature for the claim that emotions are actions (i.e., things that can be performed intentionally, voluntarily, etc.). All I am doing here is noting that the claim that emotions are actions is prima facie difficult to maintain if we confine ourselves to relatively standard conceptions of action. If the view that emotions are literally actions (or constituted by actions) is the only possible version of the agential theory and thus the only substantive account that the action analogy can deliver, this could be seen by the opponent as bad news for us.

Neither would it help to claim that emotions are nonintentional actions, subintentional actions, habitual actions, or more generally automatic actions, for these actions (assuming they are not intentional, voluntary, and so on) are still the sort of thing that could be intentional, voluntary, or caused by an intention to perform them. Moreover, we seem to have a sort of control over habitual actions that we do not have over our emotions—even if I am in the habit of brushing my teeth a certain way, I can intervene over its execution (on intervention control, see Pollard Reference Pollard2003).

One might distinguish between having control over the formation of an emotion—which we arguably do not have—and having control over its subsequent aspects, such as its expressions, duration, and so on. If we have the latter control over our emotions, then we could maintain that we can have intervention control over our emotions. Assuming that intervention control is a direct kind of control (we do not control X by doing something else), it is unclear that we have this sort of control over our emotions, at least with respect to some of their aspects. Take duration. It might look like we can make sure we do not have an emotion for too long. But this is not going to be done by simply deciding to stop the emotion, the same way one might decide to stop brushing one's teeth. The way it is done is typically by attending to other things, breathing, and the like—that is, indirectly. In any case, appealing to the possibility of intervention control over our emotions would not help the advocate of the claim that emotions are actions since having the capacity to intervene in a process is not sufficient for that process to originate in an action of yours (e.g., I can intervene while my computer is doing an automatic update). Having intervention control over our emotions is thus compatible with emotions not being actions.

It does not seem plausible either to claim that emotions are actions in some more minimal sense. For this minimal sense would still have to involve the idea of a doing. Emotions, however, are not things we literally do. Perhaps the suggestion that emotions are literally constituted by actions (actions being proper parts of emotions) can be developed (for a view in the vicinity, see Slaby and Wüschner Reference Slaby, Wüschner, Roeser and Todd2014). But it is unlikely that emotions necessarily have actions as proper parts. At any rate, it looks implausible to say that in feeling angry, I am always performing an action.

Nothing here is meant to suggest that it is not possible to develop a theory of emotion that looks, at least at first sight, radically revisionary (revisionary about emotions and/or actions). But it would be bad news for advocates of the action analogy if this were the only sort of agential theory they could go for. For one thing, we would face the challenge of motivating the radical revisions to our pretheoretical conception of emotions. If we could accept the broad similarities between emotions and actions without adopting the claim that emotions are actions, it seems that we should. For another thing, it would be a serious blow to the action analogy if the only alternative to the perceptual theory it could deliver was the highly controversial view that emotions are actions.

Although highly controversial, the view that emotions are actions (or doings or constituted by actions or doings) might implicitly rely on an intuition that might look plausible. I think that there is a plausible view in the vicinity, according to which emotions—though not themselves actions—fall under certain action types. This seemingly paradoxical view will be introduced and motivated in section 4. In the next section, I briefly consider two views that should be contrasted with the agential theory because they do not take the action analogy seriously enough.

3. Two Ways Not to Take the Analogy Seriously Enough

In order to develop a properly agential theory according to which emotions are action-like in some fundamental sense, one challenge—at least in the context of providing a plausible alternative to the perceptual theory—is to find a view that does not lead to radically revisionary claims about emotions. But there is another challenge, namely, that the view should not be a non-agential view in disguise. It should be the case that the view is one on which (1) emotions bear an intimate relation to certain action types, and (2) there should be something about emotions themselves that is action-like.

In light of (1) and (2), we can exclude two possible ways of pursing the action analogy as not taking the analogy seriously enough. First, on a liberal conception of agency, a response counts as ‘active’ just in case it is a response to reasons. Here is Joseph Raz:

For Raz and others (see, e.g., Hieronymi Reference Hieronymi, O'Brien and Soteriou2009; in the philosophy of emotion, see Müller Reference Müller2021), so long as emotions are responsive to reasons, they can be characterized as active. In this respect, emotions are not different from actions. However, they are also not different from belief, intention, and other responses that can be formed for reasons. There is, on this view, nothing distinctively action-like about emotions. Although emotions might helpfully be compared to actions, they can also helpfully be compared to other ‘rational’ responses such as belief (see Müller Reference Müller2017, Reference Müller2019, Reference Müller2021). The resulting view, therefore, is not an agential theory in my sense—it fails to deliver the claim that emotions are fundamentally action-like—even if it commits itself to the claim that emotions are in some sense manifestations of our agency. In fact, even the perceptualists could accept the claim that emotions are in some sense manifestations of our agency insofar as they allow certain background (e.g., attentional, conceptual) capacities of the subject to play a role in the formation of emotions. At least if emotions are to be modeled on high-level perception, there might be a sense to be made of the claim that both emotion and perception involve activity rather than the passive taking in of stimuli. (For reasons for thinking that the perceptualist should not model emotions on high-level perception, see Milona and Naar Reference Milona and Naar2020). If I am right, however, this claim would fall short of the idea that emotions are fundamentally action-like.

On the reasons-responsive view, emotions do not bear any robust relation to action . Now, there is a view that does posit an essential relation between emotion and action (or at least behavior), namely the so-called motivational theory. Emotions, on this view, are motivational states, ‘where a motivational state broadly understood is an internal cause of behaviors aimed at satisfying a goal’ (Scarantino and de Sousa Reference Scarantino and de Sousa2018: sect. 8) On Nico Frijda's classic approach (Reference Frijda1986), emotions are action tendencies or states of ‘action readiness’, that is, states that dispose us to certain kinds of behavior characteristic of the relevant emotion type (for a recent elaboration of Fridja's approach, see Scarantino Reference Scarantino, D'Arms and Jacobson2014). In a similar vein, Julien Deonna and Fabrice Teroni hold that emotions are experiences of our body getting ready for action (see, e.g., Deonna and Teroni Reference Deonna and Teroni2012, Reference Deonna and Teroni2015; for important critical discussion of Deonna and Teroni's theory, see Dokic and Lemaire Reference Dokic and Lemaire2015; Mitchell Reference Mitchell2020; and Müller Reference Müller2017). Anger, for instance, ‘consists in an experience of one's body prepared to retaliate’ (Deonna and Teroni Reference Deonna and Teroni2015:303).

To the extent that the action analogy supports rejecting the perceptual theory, it might be thought to give direct support to the motivational theory. While the perceptual theory placed emotions on the theoretical side of the traditional theoretical/practical distinction, the action analogy seems to suggest that emotions are on the practical side. However, the motivational theory is not open to the advocate of the agential theory. Although there is in the account an essential reference to action, the relation posited between emotion and action does not look like the sort of relation we are looking for when saying that emotions are action-like in themselves. On the motivational theory, where the actions appealed to are external to emotions (as the actions are caused by emotions), nothing like this is at play, and the motivational theorist can easily deny that emotions are action-like in any interesting sense. Adopting the motivational theory would therefore not be taking the action analogy seriously enough either.Footnote ³ That said, that emotions tend to motivate action—in particular their action counterparts—is difficult to deny. What is easier to deny is the claim that emotions necessarily motivate action. Would accepting the claim that emotions necessarily motivate commit us to the motivational theory? No, it would only commit us to the claim that emotions have a certain feature necessarily. The motivational theory goes beyond this claim by saying that emotions are ‘at bottom’, ‘fundamentally’ motivational states. But perhaps emotions are perception-like or action-like states with motivational powers. As we will see, the fact that emotions tend to motivate action can help us formulate a plausible version of the agential theory.

For now, let us formulate the dilemma we are facing: either there is a straightforward but highly controversial way to formulate the agential theory, or the views we end up with are not genuine agential views (i.e., views that hold emotions to be in themselves action-like). It thus looks like the only way to take the action analogy seriously and thus to provide a coherent formulation of it is to claim that emotions are actions (or doings or constituted by actions or doings). In the next section, I argue that the prospects for a coherent and attractive agential theory are not so bleak.

4. Taking the Action Analogy Seriously Enough

To get to a view of emotions as action-like—to get to an agential theory—we need to get clear on what is meant by ‘action-like’. We have seen that not just any connection or similarity to action will be sufficient for claiming that emotions are action-like in an interesting sense. The idea is that emotions are action-like neither in the sense that they fall under a broader kind of response under which action (but also belief, desire, etc.) falls nor in the sense that they have a robust causal link to action.

I propose that an entity is action-like in the relevant sense if it falls under an act type, where an act type is understood as a type of entity whose instances count, at least under a certain description, as actions in many, but not necessarily all, cases. Take the act type ‘kissing’. Kissing can happen whenever one intentionally performs a certain act of kissing. But it can also happen when one is a victim of compulsion or when one is sleepwalking. The concept of a kiss is flexible enough to apply to a variety of cases, not all of which will involve a genuine action. Indeed, calling kissing an act in all these cases seems to constitute a stretch of ordinary language. Even if we might nonetheless be willing to call all these pieces of behavior ‘actions’, there is another kind of case we can appeal to, namely that of entities that function as proxy for certain act types. Consider a tattoo originally acquired in protest to racism. There is a sense in which the tattoo itself communicates one's disapproval, where communication is an act of a certain type. The tattoo, though not an act (in contrast with the original act of acquiring it), falls under a certain act type—be it communicating, protesting, or ‘sending a message’. In a sense, the tattoo is a standing form of communication.

Coming back to emotions, a view of emotions as forms of communication (a view I will shortly reject in favor of a related but different view of emotions as action-like) can in fact be found in the literature on reactive attitudes. A common claim in that literature is that reactive emotions such as resentment, indignation, and gratitude are fundamentally communicative entities in that they call for a response in a target individual (see, e.g., Darwall Reference Darwall2010; McGeer Reference McGeer, Coates and Tognazzini2013; Smith Reference Smith, Coates and Tognazzini2013; Watson Reference Watson, McKenna and Russell2008). On this sort of view, resentment is a bit like an act of demanding or urging a response in someone else. But since emotions are not acts, they cannot be literally communicative acts. Of course, one might take these emotions to be communicative only in the sense of disposing us to communicative acts. It is sometimes held, however, that a privately held emotion itself can be a form of communication (e.g., McGeer Reference McGeer, Coates and Tognazzini2013: 181; Watson Reference Watson, McKenna and Russell2008: 122). One might wonder though how something that is not an action and that is privately held can be fundamentally communicative. Although suggestive, however, the idea that emotions are in themselves communicative (on top of being communicative by virtue of their expression) might be too metaphorical to do serious philosophical work. To remedy the situation, Coleen Macnamara (Reference Macnamara2015) has recently argued that privately held emotions, such as resentment and indignation, are genuine forms of communication because they incorporate a message whose evolutionary function is to elicit a certain response in a recipient— ‘the reactive attitudes have the same function as the directives, i.e., the function of getting a response’ (Reference Macnamara2015: 565, fn. 27). According to Macnamara, this is sufficient for emotions (even privately held ones) to be communicative entities. Though we might disagree with this picture (for a critical discussion of Macnamara's proposal, see Glazer Reference Glazer2014), we can interpret it as an instance of the agential theory to the extent that it claims that emotions are fundamentally forms of communication.

(If some emotions are best understood on the model of directives or complaints, this casts doubt on the view according to which emotions are attitudes of approval or disapproval (Müller Reference Müller2018, Reference Müller2019: ch. 4; Mitchell Reference Mitchell2021). For although a complaint presupposes, indeed expresses, disapproval, it is not identical to it. Compare ‘I disapprove of what you did’ with ‘I urge you to stop doing this’. Whether or not this view of reactive emotions under discussion here is on the right track, it is plausible that the normative significance of emotions and their action counterparts will not be a matter of them constituting mere forms of approval or disapproval. In fact, a thoroughgoing agential theory might need to deny that emotions are strictly speaking attitudes or stances if they are to be seen as the ‘internal counterparts’ of bodily action (Naar Reference Naar2022a). Although clearly connected to attitudes, bodily actions are not themselves attitudes. I leave open the possibility of further pursuing the agential theory in this way.)

We now have a version of the agential theory on the table. There are three natural problems we can raise against this view, however. First, what is really the cash value of claiming that reactive attitudes are communicative entities on top of the communicative acts they dispose us to perform? If emotions’ fundamental significance has to do with communication, why can we not cash that out in terms of their functional role—the acts that constitute their manifestation—rather than in terms of their intrinsic nature? Second, one might worry that the resulting view strips emotions of their significance by grouping them under the same class of responses as the ones to which they dispose us. It is one thing to undergo indignation, quite another to demand change or protest. One might think that an indignant person is not protesting yet; they are only disposed to protest. Third, one might wonder whether this sort of picture can be generalized to all emotions. It is indeed unclear that fear and shame are forms of communication. The view proposed by Macnamara and others would therefore lead to a rather disjointed picture of emotions.

The view that emotions are directives of a certain kind, gestured at by Macnamara, in fact points to an alternative agential theory that does not face these problems. On the alternative view, emotions are not commands directed at their object—indignation is not a command directed at a wrongdoer. Rather, emotions are commands the agent undergoing them imposes on themselves. (The view introduced here contrasts also with Mitchell's view according to which emotions are responses to commands or directives made by their object [Mitchell Reference Mitchell2021]. It is indeed one thing to say that the world demands of me that I feel a certain way, another thing to say that my emotion demands of me to act in certain ways.) There are two aspects to the view that need some unpacking and motivating: emotions are commands, and emotions are self-imposed commands. Let us start with the former. What could emotions command? A natural way to go is to claim that emotions are commands to perform their action counterparts. Because commands are the sort of thing to motivate generally, we thus have an explanation of why emotions tend to motivate their action counterparts. Commands, furthermore, might be made appropriate by the same object as the actions they enjoin. An act of injustice makes appropriate both protest and a command to protest whenever nobody is protesting. As a result, we also have an explanation of the normative link between emotions and their action counterparts that does not classify them as the same kind of thing, namely that emotions are commands to perform their action counterparts explains why both emotions and their action counterparts are made appropriate by the same object.

This view should be reminiscent of the recent influential imperativist theory of pain (see, e.g., Klein Reference Klein2015). According to this theory, pain functions very much like an imperative to protect a part of one's body. Without going into the details of the view, however, it is not clear that pain is the sort of thing we would want to analogize to action in any interesting way. So even if pain is a kind of command (for Klein, in virtue of having imperative content), it is not a command of a sort that could constitute an act of commanding performed by an agent. It is because, as it were (and to use Klein's metaphor), it is the agent's body that is commanding them to act, rather than the agent commanding themselves to act. The sort of command we are looking for to get to a genuinely agential theory is one that bears much more resemblance to acts of commanding performed by agents. This issue might also be raised against Macnamara's view to the extent that it does not claim that emotions are forms of communication carried out by the agent undergoing them. As far as that view goes, emotions might be purely arational states, on a par with pains, with imperative content. I propose that the notion of a command in question is of an entity that is acquired or brought about by an agent for reasons. The link to the level of the agent and rationality is crucial to allow for the possibility of forms of commands that do not count as actions but are still action-like in falling under the act type of commanding. These forms of commands, though not genuine actions, are still the work of rational agents.

Now, for emotions to constitute commands in the relevant sense, they should be, unlike pains, the ‘work of rational agents’. Though the idea that emotions are the sort of things we acquire for reasons might be challenged in various ways, it is a widely held thesis (see Naar [Reference Naar2022b] for a discussion of arguments against normative reasons for emotions and rebuttals). Suppose that there are normative reasons to which we can respond with emotions. If emotions are commands of some kind, then they are commands that can be acquired on the basis of reasons. And given that reasons-responsiveness is an agent-level phenomenon, we can claim that emotions are commands one (nonintentionally, spontaneously) imposes on oneself. This claim meshes well with the common claim that emotions, unlike pains, are formed partly on the basis of our cares, concerns, and other states that reflect the agent's perspective on the world (Roberts Reference Roberts2003). The view sketched here might explain cases where one is alienated from one's emotions, in the following way: the emotions command one to do certain things without being commands of the agent. This view can thus allow for cases of emotions that are, like pain (as the imperativist construes it), commands imposed from the outside, as it were, and so as not formed by the agent for reasons. (For a defense of a pluralist view of emotions as coming in both rational and arational forms, see Naar [Reference Naar2019].) The broadly imperativist view of emotions sketched here is flexible enough to allow only certain emotions to count as action-like in the relevant sense (the other emotions being arational commands on a par with pain as the imperativist thinks of it). I conclude that an imperativist theory of emotions seems to constitute a promising version of the agential theory that no emotion theorist should ignore.