The Non-Arbitrary Link between Feeling and Value: A Psychosemantic Challenge for the Perceptual Theory of Emotion

Abstract

This essay raises a challenge for the perceptual theory of emotion. According to the perceptual theory, emotions are perceptual states that represent values. But if emotions represent values, something should explain why. In virtue of what do emotions represent the values they do? A psychosemantics would answer this, and that’s what the perceptual theorist owes us. To date, however, the only perceptual theorist to attempt a psychosemantics for emotion is Jesse Prinz. And Prinz’s theory, I argue, faces an important difficulty: It makes the pairing of any given emotion with its respective value entirely arbitrary. But that’s a problem. It seems—and this is a major contention of this essay—that an emotion, in virtue of how it feels, bears a natural or non-arbitrary link to the value it represents. And this datum makes it all the more difficult to provide a viable psychosemantics for the evaluative content of emotion.

1. Introduction

Many theorists of emotion would say that typical emotions, when conscious, feel a distinctive way. For example, consider the unpleasant aversive character of disgust. This feeling is clearly distinct, say, from that of happiness or fear.

In addition to claiming this phenomenological distinctness for emotions, many theorists of emotion would also accept a further thesis, namely, that emotions represent values. For instance, disgust towards a pile of sludge represents the sludge as foul.

Granting these two claims for now, a interesting question appears on the horizon. Suppose these two features of an emotion—its characteristic feeling and its representational content—were to come apart. For instance, imagine experiencing an emotion that feels like disgust but which represents, instead of the foul, the beautiful. Something seems deeply incongruous or mismatched about such a pairing. There is something about disgust that naturally links it to the foul. And I want to argue that if we think carefully about this phenomenon—the natural link between the feeling of an emotion and its associated value—we arrive at an interesting new challenge for the perceptual theory of emotion.1

According to the perceptual theory of emotion, emotional experiences are ways of perceiving values. They attribute evaluative properties to their objects, and they do so as perceptual states. For example, in fearing the serpent as it rears, you perceive the impending harm. In feeling awe toward the mountain-scape, you perceive its sublimity. To be clear, it is not that we perceive these values, then feel those emotions as a result. It is rather that the emotions themselves are ways of perceiving those values.

The perceptual theory, now a dominant view in philosophical work on emotion, has a lot going for it. For one, it can explain the fittingness conditions of different emotions by reducing these to accuracy conditions: A fitting emotion is simply one that correctly attributes an evaluative property to its object.2 For another, the perceptual theory explains the sense in which emotional experience can conflict with our value judgments.3 Just as the stick looks bent even when you know it is straight, so the serpent might feel menacing even when you know it is innocuous. Such conflicts between emotion and judgment are made intelligible by appreciating that it is a general feature of perceptual states that they might conflict with judgments in this way.

Besides this, the perceptual theory suggests an attractive epistemology of value. If emotional experience is a mode of evaluative perception, then perhaps values are revealed to us through our emotions. Put another way, perhaps our emotions can justify evaluative beliefs just as sense perception can justify beliefs about the shapes and colors of things around us. So our experience of shame might justify us in believing what we’ve done is monstrous. Or our experience of admiration toward some artwork might justify us in believing the work is beautiful. Such experiences might make certain values perceptually manifest to us. Indeed, many who have embraced the perceptual theory of emotion have been mindful of this epistemological pay-off.4

In spite of its promise, however, the perceptual theory of emotion faces its share of problems. Notably, it seems there can be reasons for emotions, but it is less plausible there can be reasons for perceptual experiences. You do not have a reason to see the rose as red. You just see it.5 You might, in contrast, have reason to admire the rose.6 And this disanalogy between emotion and perception has seemed to many to raise a significant challenge for the perceptual theory, not to mention a check on the epistemic pay-off many had hoped for. Arguably, this challenge, in some form or other, has been the main focus of work on the perceptual theory of emotion over the past decade.7

I believe, however, that in the shadow of this challenge from the reasons-assessability of emotion, other problems facing the perceptual theory have gone overlooked. In this essay, I aim to remedy this. I shall articulate a challenge for the perceptual theory which seems to me at least as grave as any other. I mean the challenge of providing a viable psychosemantics for the evaluative content of emotion. I aim to lay out this challenge, and show why a certain underappreciated datum makes this challenge especially difficult for the perceptual theorist to meet. Described at its most general level, the datum I have in mind—previewed at the outset of this essay—is that the feeling of an emotion bears some kind of natural or non-arbitrary link to the value associated with it. I shall call this the naturalness datum. This essay is about the naturalness datum and the psychosemantic challenge it raises for the perceptual theory of emotion.

In Section 2, I explain why the perceptual theorist owes us a psychosemantics for the evaluative content of emotion. In Section 3, I argue the naturalness datum raises a problem for the psychosemantics we find in the early work of Jesse Prinz [6]. In Section 4, I address the most tempting response for Prinz to adopt, and I offer a thought experiment to rebut this response. In Section 5, I defend my thought experiment from a series of objections. In Section 6, I argue that the naturalness datum raises a challenge not just for Prinz [6] but for the perceptual theory more generally. In Section 7, I explore one potential solution to this challenge, an appeal to iconicity. Finally, in Section 8, I address Prinz’s more recent work in which Prinz himself raises a concern about his earlier work [6] similar to the concern I raise in this paper.8

2. The Need for an Emotional Psychosemantics

If emotional experiences represent values, presumably there is some explanation as to why. In fact, we can distinguish two questions here. First, why do emotional experiences represent values at all? Second, why do particular emotion types (fear, disgust, etc.) represent the values they do, rather than other values? Why, for example, does fear represent the menacing rather than the beautiful? A psychosemantics for emotion seeks an answer to these questions.

In addition to having evaluative content, emotions also have descriptive content. When I am overjoyed at winning the lottery, the descriptive content of my emotion is “I won the lottery.” And there does not seem to be any deep mystery here as to how my emotion comes to have this content. Presumably, my emotion somehow inherits or derives this content from the state that elicits my emotion, say, my belief that I won the lottery.9 For perceptual theorists, however, my joy at winning the lottery also represents this event as good. That’s the evaluative content of my emotion here. And the perceptual theorist does not think this evaluative content gets inherited from some eliciting state. After all, the perceptual theorist typically denies that emotions depend in any general sense on antecedent evaluative contents, such as evaluative beliefs. Rather, the perceptual theorist thinks emotional experience itself supplies the evaluative content. We do not feel because we judge; we judge because we feel. Moreover, perceptual theorists generally claim that the evaluative content of emotional experience is non-conceptual. When my joy represents my situation as good, it isn’t that my emotion is simply deploying a concept I obtained from some other, non-emotional source. Indeed, many perceptual theorists think emotional experience is the source of our evaluative concepts in the first place.10 Accordingly, for perceptual theorists, there’s no shuffling off the question of how it is emotions represent values.

The question remains, then, What explains the evaluative content of emotion? Of course, it could simply be a brute fact that emotions represent values, and that each emotion represents the particular value it does. But this appeal to bruteness seems like a desperate maneuver. It seems more plausible that something should explain why emotions have evaluative content. After all, identifying the basis of mental content remains one of the central projects in the philosophy of mind and cognitive science.11 It is hard to see why this project should not be extended to the evaluative content of emotion.

Once we appreciate the need for an emotional psychosemantics, we can also appreciate the gaping hole in recent work on the perceptual theory. To date, the only perceptual theorist to even attempt a psychosemantics for the evaluative content of emotion is Jesse Prinz.12 This is a problem. If the perceptual theory is going to be ultimately acceptable, we will need to see an explanation as to why emotions possess their respective evaluative contents.13

Since Prinz [6] does offer an emotional psychosemantics, it’s only natural to begin there. Does his psychosemantic theory succeed? In what remains, I wish to argue that Prinz’s psychosemantics faces an important difficulty in light of what I am calling the naturalness datum, which I shall lay out presently. I will not assess here whether the difficulty is fatal. I argue only that it needs to be addressed somehow. So, for all I say, the psychosemantics in Prinz [6] need not be abandoned, only modified. Indeed, in Section 7, I will suggest how that modification might go.

3. The Problem of Naturalness for Prinz’s Psychosemantics

The emotional psychosemantics in Prinz [6] faces a certain difficulty: It cannot explain why emotional experiences are naturally paired with their respective values. In this section, I wish to explain this problem, and show how it affects Prinz’s theory.

3.1. Prinz’s Psychosemantics for Emotion

Following Dretske [63,64], Prinz holds that “a mental representation is a mental state that is reliably caused by something and has been set in place by learning or evolution to detect that thing” [6] (p. 54). So if a visual experience represents red things, it is because that visual experience is reliably caused by red things, and its function is to detect them.

That is the basic shape of the psychosemantics Prinz wants to apply to emotion. Now, there are well-known issues with this Dretske-style psychosemantics.14 But I am prepared to set these aside. I think that even if this Dretske-style approach works for some mental content—indeed, for a great deal of mental content—it is poorly positioned to explain the evaluative content of emotional experience (unless we subject the theory to serious modification, again, as I will suggest in Section 7).

In the quotation above, Prinz speaks of reliable causation. But elsewhere, he speaks of reliable co-occurrence, and I shall use this term for the sake of uniformity, and because it at least respects the view that evaluative properties are causally inefficacious, which we should not assume to be false out of hand.15 We can see, then, that on Prinz’s view, for emotions to represent evaluative properties, it must be the case that emotions (a) reliably co-occur with values, and (b) have the function of detecting them. For instance, perhaps fear represents the menacing because it reliably co-occurs with menacing things, and has the function of detecting such things.

This much is typical information semantics. But the distinctive thing about Prinz’s view is that he claims emotions detect values by monitoring the subject’s bodily states:

Each emotion is both an internal body monitor and a detector of dangers, threats, losses, or other matters of concern. Emotions are gut reactions; they use our bodies to tell us how we are faring in the world. [6] (p. 70)

Thus, emotions detect values by tracking changes in the body. The bodily changes in question, of course, are not simple processes like the mere racing of the heart. Since such simple processes would fail to individuate emotion types, the idea is rather that each emotion has a unique and therefore suitably complex configuration of bodily symptoms. Each emotion has a somatic signature capable of distinguishing it from other emotions.16 It is this somatic signature that co-varies with the relevant value, enabling us to perceive it.

3.2. The Problem: Emotions as Arbitrary Beeps of Value

The thing that should trouble us about Prinz’s view is this. For Prinz, feelings of disgust just happen to co-occur with the foul. There is no deeper link between the feeling of disgust and the foul. There is no sense in which feelings of disgust are especially apt for representing the foul.

To see the problem, consider an analogy Prinz himself provides. He likens the evaluative content of emotion with the beeps emitted by fuzz busters,

which people place in their cars to determine when they are driving in zones monitored by police radars. A beep emitted from a fuzz buster represents the presence of a police radar. But the beep itself is utterly lacking in structure. It cannot be analyzed into meaningful subbeeps. … [T]he beep emitted by a fuzz buster does not describe what it represents. It represents police radars because it is reliably caused by police radars, and it is set up for that purpose. Likewise, emotions can represent [values] without describing them. [6] (p. 65)

As this analogy makes explicit, the link between police radars and the beeping of fuzz busters is totally arbitrary. The beeps could just as well have represented anything else, even the absence of police radar, as there is no deep or inherent linkage between police radar and the sound of beeping. But is that how emotions are related to the values they represent? This seems to me severely mistaken. Emotions are not arbitrary beeps of value that could have just as well represented anything else.

I mean, emotions could have represented other things. In principle, the feeling of disgust could be deployed by a mental system to represent just about anything. So my point is not a modal one. It is rather that the feeling of disgust is naturally well-suited for representing the foul. Thus, when we imagine the feeling of disgust being deployed by a representational system to represent something other than the foul—say, the good or the beautiful or the blue or the rectangular—we can see there would be something deeply and weirdly arbitrary about this. And this suggests that for disgust to represent the foul (as it does for us) is precisely not arbitrary. Somehow, the feeling of disgust and its associated value bear a natural linkage, a kind of deep congruity. And this does not seem to be some special thing about disgust. As Julien Deonna and Fabrice Teroni put it, “we cannot conceive of the connection between, for instance, the phenomenology of fear and danger as arbitrary. Intuitively, no other emotional experience than that of fear is a suitable candidate for presenting the world in terms of a danger” [70] (p. 86).17 Indeed, as Jan Slaby writes, “Prinz illegitimately cuts the connection between how an emotion feels and what it is about” [71] (p. 443).

Let this stand as an initial statement of the naturalness datum and the problem it raises for Prinz’s psychosemantics. In what remains, I wish to show that this problem constitutes a serious challenge.

4. A Tempting Explanation Met with a Thought Experiment

So far, we have seen that something about the feeling of disgust makes it naturally rather than arbitrarily connected with the foul. At this point, we must consider what is initially the most obvious explanation as to why, an appeal to emotional motivation. This appeal seems to offer a tempting strategy for modifying Prinz’s psychosemantics. I shall argue, however, that the appeal to emotional motivation does not succeed.

4.1. The Tempting Explanation: An Appeal to Emotional Motivation18

The most tempting explanation of the naturalness datum begins with this observation, that the feeling of an emotion is typically part of an emotion’s motivational profile, as noted by Frijda [72] and many others. Part of feeling an emotion is feeling motivated to act in certain ways. What ways are those? Ways that are fitting with respect to the value the emotion represents. So disgust represents the sludge as foul, and motivates us to avoid the sludge, and in feeling disgust, we feel ourselves thus motivated. The feeling of disgust, then, is naturally paired with the foul because the feeling of disgust is part of an unfolding behavioral process that is precisely appropriate when faced with steaming sludge and other such foul items.19 In contrast, were the feeling of disgust to represent, say, the beautiful—as suggested in the remarks that began this paper—then such feelings would represent something as beautiful while motivating us to act in ways that are precisely unfitting with respect to the beautiful, say, by avoiding engagement with it. No wonder, then, such a mismatch seems incongruous to us.

Crucially, this proposal could be embraced by Prinz [6] without in any way departing from a Dretske-style psychosemantics. Thus, if this proposal were defensible, the naturalness datum would not be a problem for Prinz [6]. Indeed, Prinz comes close to asserting this proposal himself. He is clear, for example, that emotions are acquired (through learning or evolution) because they confer an advantage by helping us behave adaptively in response to values.20 Now, adaptive behavior is not just the same thing as fitting behavior—we can imagine cases where the these come apart—but still, Prinz [7] is clearly well-positioned to embrace the appeal to emotional motivation I have outlined here.

As tempting as this proposal may seem, I do not think it succeeds. Presently, I shall argue this by offering a thought-experiment in which the feeling of an emotion is severed from its typical motivational impact.

4.2. The Thought Experiment: The Strange Case of the Emotionally Inverted Creature

It seems logically possible for there to exist a mind in which feelings of disgust represent the beautiful instead of the foul. Surely God could make such a creature. In this creature’s mind, let’s say, the rapturous awe we feel towards a vast mountain-scape actually feels like revulsion. It feels like revulsion but nevertheless represents the mountain-scape as beautiful. When this creature experiences this emotional state, for instance, he judges the mountain-scape to be beautiful, and he immediately and unreflectively takes his emotional experience as a kind of evidence for this judgment.

Let us add now that the creature’s emotion—the mental state which feels to him as revulsion feels to us but which represents its objects as beautiful—retains the same motivational profile that rapturous awe has for people like us. So both we and the creature see the mountain-scape, and feel an emotion that captures our attention, and which focuses our minds on the mountains before us, and disposes us to snap photographs and say things like, “How wonderful! Do you see that, Jim?” But, for the creature we’re imagining, all of this feels as absolute revulsion feels to us.

We’re familiar with the possibility of an inverted spectrum for color vision.21 I am now inviting us to imagine an inverted spectrum for the felt character of emotional experience. And if a complete spectrum inversion for emotions does not make sense for some reason—if, for instance, there is no emotional equivalent of the color wheel—we can at least see that certain pairs of emotions exhibit something like polarity, and we can at least imagine these pairs being inverted in their felt characters, so that the unpleasant character of disgust and the immense hammer-like pleasure of awe toward the staggeringly beautiful trade places, holding all motivational features of these emotions fixed, and even holding fixed their respective evaluative contents.

Perhaps such emotional inversions could not occur in the actual world. But such inversions seem possible in the broadly logical sense. And here is the thing to see: There would still be something profoundly unnatural about emotional inversions like these. In particular, there would be something deeply incongruous about the unsettling character of disgust being deployed to represent the sublime or the beautiful. Conversely, there would be something deeply incongruous about the piercing overmatching delight of awe being deployed to represent the foul. But notice—and here is the crucial thing—this cannot be explained by appealing to inappropriate motivation. By stipulation, the inverted creature is motivated to behave appropriately with respect to the emotion object. So his disgust feelings represent the object as beautiful, and he is motivated to act appropriately towards the beautiful, and yet it remains that disgust feelings make for an unnatural pairing with these other elements. Therefore, something aside from inappropriate motivation must explain the unnaturalness present in this case.

Of course, the emotionally inverted creature is far-fetched, but no more so than any number of outlandish thought experiments philosophers have devised over the centuries. And I maintain our emotionally inverted creature shows something important. For normal human beings like us, the feeling of disgust is paired with both the motivation to avoid the disgusting object, and the representation of the disgusting object as foul. And the inverted creature shows us that this pairing is not just accidental. It is not an arbitrary configuration. The feeling of disgust, by its very phenomenal character, belongs with these other features of the emotion. That is the basic insight of the naturalness datum. And the question is why this would be so. And my point is that Prinz’s psychosemantics seems to lack the resources for explaining this. Indeed, Prinz’s view seems precisely to predict that emotional experiences are just arbitrarily hooked up with the values they represent. If Prinz is right, there should be nothing deeply weird about the kinds of emotional inversions I have described. The inverted creature should be no stranger or more surprising than we are. And the most promising means of resisting this objection—the appeal to emotional motivation—does not succeed, since the emotional inversion case before us is precisely designed to rule out such an explanation.

5. Objections to the Thought-Experiment

There are several objections one might raise against the thought experiment I have presented here. In this section, I address what I take to be the most pressing of these.

5.1. Are Feelings Constitutively Linked to Emotional Motivation?

The first objection one might raise against my thought experiment is to claim that feelings cannot come apart from motivation in the ways I have suggested. The idea here would be that the feeling of an emotion is not just typically but constitutively motivational; that the unpleasant character of disgust, for instance, is inherently and inextricably aversive; that to feel disgust just is to feel motivated to avoid the disgusting object. And so the emotionally inverted creature I have described is simply incoherent.

The claim this objection relies on we may call the strong dissociability thesis:

The strong dissociability thesis. The hedonic valence of an emotion cannot possibly be dissociated from its felt motivational impact. The two elements are constitutively linked.22

I agree that if the strong dissociability thesis were correct, this would spell trouble for my thought experiment. But I believe this thesis is not correct. In its place, we should affirm that the feeling of an emotion and its characteristic motivation can come apart. And while I can hardly settle the matter here, I can offer at least a provisional case.

For starters, the strong dissociability thesis suffers from a dearth of support in its favor. Notice, for instance, that the strong dissociability thesis gains little support from the observation that emotional unpleasantness is typically paired with aversive motivation. Yes, it is true that, as Zagzebski writes, “the emotion of compassion characteristically generates a desire to alleviate suffering; the emotion of anger characteristically generates a desire to ‘get even’ or something similar” [25] (p. 116). But it does not follow that “emotions are potentially motivating because of their intrinsic features” (ibid), if by this it is meant that the motivational impact of an emotion is somehow inextricably bound to its hedonic tone. Thus, we should wonder what might justify the strong dissociability thesis. Certain writers seem to take the thesis or something like it for granted, and treat it as a datum that any theory of pain needs to accommodate.23 But such bare assertions will not help us here. Alternatively, we might appeal to evolutionary considerations24: Should we not expect evolution to produce beings whose hedonic states are linked to their motivational tendencies? For instance, creatures who avoid pain? Granting we should expect this, notice that, again, this only requires that unpleasantness and aversiveness are typically paired. This would not require anything as strong as a constitutive link.

Setting aside the question of support, the strong dissociability thesis faces positive difficulties of its own. I will lay out five.

First, it appears the brain circuits involved in an emotion’s motivational impact, and those involved in an emotion’s being consciously experienced, are two independent systems.25 Often enough, emotional processes in the brain are getting us moving long before we are even aware of undergoing any emotion at all. So it would be a mistake to suppose the feeling of an emotion is what causes us to be motivated, let alone that the feeling somehow is the motivation. Evidently, we are already motivated in advance of undergoing such feelings. Now, that leaves open the idea that emotional feelings are somehow essentially of motivations—that they are essentially directed at an intentional object that has something to do with action. But even if emotional feelings happen to be directed at such an object as a matter of fact, the question remains, Why must they be? At the very least, it isn’t obvious how we could plausibly answer that.

Second, as Christine Tappolet has pointed out, we often have emotions about fictions without any accompanying motivation.26 When we pity Anna Karenina, we are not driven to do something for her. To see this, consider that while pitying Anna Karenina is entirely appropriate, being motivated to help her is entirely benighted. Here’s the question for us presently, then: Does our pity for Anna have any hedonic valence? If it is unpleasant, as pity towards real world objects tends to be, then the answer is Yes. But the answer is also Yes if such pity is experienced as pleasant (just as the horror one feels upon watching a scary movie might be experienced as pleasant). What seems implausible is that the pity one feels for Anna has no hedonic valence whatsoever. And yet, evidently, our pity for Anna does remain devoid of felt motivational impact: From our own perspective as readers, we do not experience our pity for Anna as pushing us to intervene on her behalf in some deluded way. So this seems to be a case of felt hedonic valence without felt motivational impact.27 Third, we can think of cases where the strong dissociability thesis seems to make the wrong prediction. Consider two emotions one might have toward the past: regretting some event and remembering it with fondness. Both of these emotions motivate us to ruminate on the past. Indeed, there does not seem to be anything fundamentally different in the behavior they motivate. Their primary behavior is a cognitive one—actively dwelling on some past event. Phenomenologically, the main difference between these two emotions is nothing motivational at all: It lies in their hedonic valence. So if hedonic valence is somehow constitutively linked to motivational impact, how can these two emotions converge in their motivational profile while differing so wildly in their hedonic valence? This should not be the case, if the strong dissociability thesis is correct.

Fourth, we can press the point from the other direction. Imagine recalling some past event that makes us feel, not regretful, but ashamed. Shame and regret toward some past event clearly both have an unpleasant valence. But they also differ dramatically in their motivational impact. Sure, both might motivate us to avoid repeating such mistakes. And yet, cognitively speaking, regret motivates dwelling on the past, while shame motivates avoiding its recollection. Thus, here again, the strong dissociability thesis seems to make the wrong prediction, though this time, we have it the other way around—two emotions converging in hedonic valence while differing in motivational impact. How could this be, if hedonic valence and motivational impact are constitutively linked?

Fifth, if we think about pleasures more generally, and not just emotional pleasures, we can imagine other cases where hedonic valence dissociates from motivation. To borrow a case from Jennifer Corns, think of a chain smoker who feels compelled to light his fiftieth cigarette for the day but who anticipates no pleasure from the act.28 The thought of another cigarette repulses him, let’s say. Nevertheless, he feels an irresistible urge to reach for his pack. If we can coherently imagine this case, as I think we can, then unpleasantness can dissociate from motivation. And if that’s true about other unpleasant experiences, it’s hard to see why it wouldn’t be true of unpleasant emotional experiences as well.29

In sum, the strong dissociability thesis faces serious difficulties. It seems mistaken, therefore, that the feeling of an emotion and its motivation are constitutively linked. And so we do not have any reason here to think our emotionally inverted creature is not a coherent possibility.

5.2. Am I Begging the Question against Reductive Theories of Consciousness?

If you are a reductive functionalist about consciousness30, you might insist that by giving my emotionally inverted creature an emotional state with the motivational and representational profile of disgust, we thereby guarantee that his disgust experience feels to him as ours does to us. Have I begged the question against the reductive functionalist here?

Note that it is not my aim to offer a sort of inverted qualia case to refute reductive functionalism, or any form of reductive physicalism for that matter. I do not take my case to carry such an implication. In fact, though I have claimed the emotionally inverted creature is possible, this need not be so for my point to go through: Even if I am summoning a counterpossible, the case can still help us see that such an emotional inversion would run afoul of the deep congruity between emotional experiences and the values they represent. And that’s all the case was meant to show. It is an aid to judgment, not a counterexample to a reductive theory of consciousness.

Of course, not all counterpossibles offer an aid to judgment. If I cannot even begin to imagine some scenario, or do not even know what I am being asked to imagine—a largest number, a heaviest color, a bachelor with the happiest marriage—then the counterpossible in question has limited utility as an illustration, perhaps none at all. But the emotional inversion case I am asking us to imagine is not like that. Particularly, because the strong dissociability thesis is mistaken, it isn’t just utter nonsense to picture a creature who has disgust feelings that represent something as beautiful while being motivated to act in ways appropriate to the beautiful.

I expect, however, that not all readers will accept what I have just said about the admissibility of counterpossible cases. So I will bracket this point from here on out, and defend the possibility of emotional inversion as I have envisioned it.

What, then, shall I say to the reductive functionalist? For starters, notice that it is not essential to my emotional inversion case that all of the inverted creature’s functional properties are the same as ours while his emotional phenomenology differs. It may still be that his emotional phenomenology is reducible to certain functional properties. I simply have to maintain that these are distinct from the functional properties that constitute his motivational profile. But this would require that affective feeling and emotional motivation can come apart in a substantial way. But I have just defended this in Section 5.1. Thus, the emotionally inverted creature is not just obviously ruled out by reductive functionalism.

But reductive functionalists are not the only ones I risk offending here. We must also consider representationalists, for I am asking us to imagine that my inverted creature’s disgust experience, which feels just like ours, has a different evaluative content (his disgust experience represents beauty rather than foulness). Therefore, that would mean it is at least logically or metaphysically possible for the phenomenal character of an emotion to come apart from its representational content.

However, it’s not clear why we should resist the possibility of content and phenomenology coming apart, even if we accept an ambitious representational theory of consciousness. After all, it’s just obvious that representational content alone is not sufficient for explaining phenomenal character. Otherwise, merely believing that something is foul would feel exactly as disgust feels. For that matter, merely believing the banana is yellow would have the same color phenomenology as seeing a yellow banana. But that’s clearly not the case. And since representationalists are aware of this obvious pitfall, they tend to avoid making content—specified in this crude way—sufficient for phenomenology.31 And that creates room, I think, for the thought that some creature in some other possible world might represent something as beautiful by way of a mental state whose phenomenal character feels as disgust feels to us.

In light of this, the emotionally inverted creature I have described should be an admissible case for functionalists, representationalists, and anyone else wishing to reduce consciousness to the physical. Our creature does not pose a challenge to these theories but rather shows us something about (at least paradigmatic) emotional experiences; namely, that in virtue of their phenomenal character, they bear some deep congruity to the values they represent.

5.3. Does the Naturalness Datum Hold for All Emotions?

The emotional inversion case considered here was put in terms of awe and disgust. But that just shows the naturalness datum holds for some emotions. Does it hold for all? In other words, for all emotions, is the feeling of that emotion naturally paired with its associated value? If not—if the naturalness datum reflects a quirk about a small subset of emotions rather than a feature of emotions generally—then it is unclear just how serious a problem this is for Prinz [6] or anyone else for that matter.

The naturalness datum seems to bear out wherever we can phenomenally invert emotions differing in hedonic valence. For instance, representing a positive value by way of an unpleasant emotion, no matter the specific emotion type, would seem to exhibit the same incongruity we saw in our earlier inversion cases. Just consider an inverted creature who undergoes an experience that feels as terror feels to us but which represents its object as amusing; or who undergoes an experience that feels as sorrow feels to us but which represents its object as a massive stroke of good fortune. This strikes me as no less incongruous or unnatural.

Conversely, we find the same incongruity with pleasant experiences representing negative values—say, a creature who undergoes an experience that feels as relief feels to us but which represents its object as an injustice. Again, there is something deeply incongruous here.

So inverting valence seems to generate the kind of incongruity we’ve been observing, and it apparently does so for any emotion that has a valence to invert. But what about emotions that have no valence, those which are neither pleasant nor unpleasant? Some authors regard certain cases of surprise as being hedonically neutral in this way. So imagine a creature who undergoes an experience that feels to it as neutral surprise feels to us, but which represents its object as an utter horror. This, too, seems incongruous, though less so than if the experience felt pleasant in some way. It seems, therefore, that at least part of the natural link between an emotion and its associated value arises from hedonic valence: Unpleasant emotions are naturally paired with negative values, pleasant emotions with positive values, and neutral emotions (if such there be) with neither positive nor negative values.

But that’s not all there is to it. Even within the class of unpleasant emotions, phenomenal inversion seems to result in a similar incongruity. So let’s consider a case of “endo-hedonic” inversion, as we may call it—inversion within the same hedonic class. Imagine a creature who undergoes an experience that feels as humiliation feels to us but which represents its object, not as being humiliating, but as being dangerous or menacing. Or imagine a creature who undergoes an experience that feels as rage feels to us but which represents its object, not as outrageous, but as foul. When I reflect on these examples, I still have the sense that these pairings are incongruous. Something about the determinate feel of humiliation—not just its generic unpleasantness, but its determinate phenomenal tone—makes humiliation naturally paired with, well, the humiliating; with unflattering exposure to the gaze of others. Likewise, something about the determinate feel of rage naturally pairs it with the outrageous, and makes it an unnatural pairing with other values, even other negative values.

From this brief discussion, we may draw two lessons. First, the naturalness datum arises from highly determinate features of emotional phenomenology, not just hedonic valence. Second, the naturalness datum applies, if not to all emotions, then at least to a great many of them. It is unclear anyway what the counterexample would be. And so the emotionally inverted creature has a perfectly general lesson to teach us, it seems, and it’s a lesson that cannot be easily accommodated by Prinz [6], or by any perceptual theory, as I shall argue presently.

6. The Challenge for the Perceptual Theory

So far, this essay has focused on the naturalness datum and the challenge it poses for the particular psychosemantics in Prinz [6]. In this section, I want to clarify the more general challenge here facing the perceptual theorist.

6.1. Not So Fast!

It may seem to some that the naturalness datum shows that Prinz’s psychosemantics is mistaken. However, I believe putting the challenge in those terms is too fast. Strictly speaking, all we may conclude from our discussion so far is that Prinz’s psychosemantics does not explain the naturalness datum. But that leaves it open to whether something else explains the naturalness datum, something compatible with Prinz’s psychosemantics.

Let’s think of it this way: A psychosemantic theory points to a content determinant , a feature that grounds an emotion’s evaluative content. But why think this content determinant should also explain why an emotion is naturally paired with that evaluative content? Perhaps instead there is some other feature, a naturalness determinant, that explains the naturalness of this pairing. And perhaps this naturalness determinant is distinct from the content determinant but compatible with it. In light of this possibility, it may turn out that Prinz need not revise his psychosemantics at all. He might simply identify some plausible candidate for a naturalness determinant that is not also a content determinant.

Granted, even if some candidate for naturalness determinant could be found, the naturalness datum would still put some pressure on our psychosemantics. Even if it isn’t mandatory, it remains desirable to explain both the naturalness datum and the content of emotion in one fell swoop. After all, both these features of emotion—that emotions have evaluative contents, and that emotions are naturally paired with the evaluative contents they have—do need to be explained. So if we can tell a single story that accounts for both these bits of data, that story would be especially appealing. To that extent, the naturalness datum gives us reason to revise Prinz’s psychosemantics, if doing so allows us to explain the naturalness datum.

If that’s all there were to it, though, then this challenge for Prinz [6] would not be especially grave. It would simply point to an explanatory desideratum, one factor to weigh in the balance as we compare rival theories of emotion. I believe, however, that the challenge before us is a good deal more serious than that.

6.2. The Real Challenge

The real challenge is that there does not appear to be any feature of an emotion that can explain the naturalness datum without also explaining the evaluative content of emotion.

What Prinz would need here is a content-irrelevant feature of an emotion that explains its natural pairing with its represented value. But what feature might that be? Initially, the most promising answer would appeal to an emotion’s motivational impact: An emotion is naturally paired with its represented value because the emotion motivates us to act in appropriate ways with respect to that value. However, we’ve already been down that road: this explanation would require the strong dissociability thesis. It would require that the feeling of an emotion is constitutively linked with its motivation. Otherwise, we can imagine cases of emotional inversion that still seem unnatural, even though by stipulation the inverted creature is motivated to behave in the appropriate ways. As we have seen, though, the strong dissociability thesis seems mistaken.

In light of this, it is not obvious there are any content-irrelevant features of emotion that could explain the natural link between an emotion and its represented value. Now, that should not prevent us from trying to come up with some other proposals—I am all for creative speculation here—but, absent a viable alternative, it is reasonable to expect a psychosemantic theory to double as an account of the naturalness datum. We have good reason to expect there to be a single story explaining why emotions represent the values they do and why their representing those values rather than others is especially natural.

6.3. Stating the Argument

What we have just seen suggests an argument that’s worth stating more rigorously. We may begin by assuming the perceptual theory for conditional inference:

• The perceptual theory is true.

Then let us assert:

2.

If the perceptual theory is true, then the perceptual theory explains the naturalness datum.

This premise is credible because we should expect the correct theory of emotion to explain the important features of emotional experience. I have argued that the natural fit between an emotion’s feeling and its associated value is just such a feature.

From 1 and 2, we may infer:

3.

The perceptual theory explains the naturalness datum.

Then we can say:

4.

If the perceptual theory explains the naturalness datum, then either it does so by appealing to content-relevant features, or it does so by appealing to content-irrelevant features of emotion.

Premise 4 is true because it effectively states a dilemma, one that is clearly exhaustive: any given explanatory feature will either be relevant to explaining content or not. And thus, we may infer:

5.

The perceptual theory explains the naturalness datum either by appealing to content-relevant features, or by appealing to content-irrelevant features of emotion

Now let us add:

6.

The perceptual theory cannot explain the naturalness datum by appealing to content-irrelevant features.

Certainly, this is a substantive premise. But it is one for which I have argued at length in criticizing the strong dissociability thesis and the appeal to motivation it enables (see Section 4.1). And, as admitted, while there may be other content-irrelevant factors to explain the naturalness datum, it is not immediately obvious what they would be. So at this point, premise 6 seems more plausible than not.

From 5 and 6, it follows that:

7.

The perceptual theory explains the naturalness datum by appealing to content-relevant features.

Using 7, we may discharge the assumption at premise 1, inferring:

8.

If the perceptual theory is true, then the perceptual theory explains the naturalness datum by appealing to content-relevant features.

Effectively, premise 8 says that, if the perceptual theory is correct, the psychosemantics the perceptual theorist adopts should double as an explanation of the naturalness datum. Now, this does not yet create a challenge for the perceptual theorist. It does so only when we add:

9.

The perceptual theory cannot explain the naturalness datum by appealing to content-relevant features.

From which it follows that

10.

The perceptual theory is false.

Clearly, premise 9 is contentious. Basically, it claims that no psychosemantics can do the work premise 8 says it has to. Thus, it seems to me that premise 9 is what the perceptual theorist should target. To reject this premise, the perceptual theorist needs a psychosemantics that can plausibly explain the naturalness datum. Now, I have argued that a traditional Dretske-style psychosemantics will not do the job. But we can easily see that certain other well-known psychosemantic theories will fare no better. For example, functional or conceptual role semantics would make emotions no less arbitrary as representations of value.32 In that case, what should the perceptual theorist say here? Is she doomed?

My aim in this essay has not been to answer this question. My aim has rather been to lay out the challenge. However, there is a psychosemantic theory the perceptual might appeal to here. And it is at least worth seeing how this theory might solve the problem before us, though the merest sketch will have to suffice for now.

7. The Promise of Iconicity

I think there is a way to modify Prinz’s psychosemantics that would also give us an explanation of the naturalness datum. To be clear, I am only canvassing a potential hypothesis. Obviously, I cannot defend my favored idea at any length here. But I can at least set it on the table in a flattering light.

What’s needed, I think, is to consider representations that bear a natural or non-arbitrary link to their contents. Examples come readily to mind. Consider pictorial representations like maps and realistic portraits. These artifacts represent, in part, by replicating certain crucial features of their objects. Maps, for instance, use spatial structure to represent that very spatial structure.33 And portraits use arrays of shape and color to represent a similar array of shape and color. This feature of these representations we may call iconicity.34 And, as we’ve seen, onomatopoetic words also exhibit iconicity: Such words sound like what they mean.

Now, the notion of iconicity has been roundly criticized in the past century by figures as illustrious as Goodman and Quine.35 However, in recent decades, iconicity has made something of a comeback in work on depiction as well as in work on mental content.36 And the notion of iconicity can no longer be dismissed out of hand.

This is good news for the perceptual theorist, who might consider appealing to the notion of iconicity in her emotional psychosemantics, thus enabling her to reject premise 9. So the idea here would be that emotional experiences are iconic with respect to the values they represent; that emotions are icons of value. This would mean that an emotional experience represents the value it does in part by replicating certain features of that value in the emotion’s phenomenal character.37 That’s the idea, at any rate. And this is at least the right kind of idea for developing an emotional psychosemantics that explains why the feeling of an emotion and the value it represents are naturally rather than arbitrarily paired. They are naturally paired because the feeling of an emotion itself reflects crucial features of the value it represents.

Let me be the first to admit that this idea faces serious complications. For perceptual theorists wishing to pursue this line, the two most pressing complications are: First, in what ways exactly do emotional experiences replicate features of the values they represent? In brief, what are the resemblances here? Second, how does the mind come to exploit those resemblances in order to generate representations? However attractive this proposal may seem, appealing to iconicity in an emotional psychosemantics will not be finally acceptable until these questions can be answered.

8. Prinz vs. Prinz

In a recent article with Daniel Shargel [56], Prinz raises criticisms against his earlier view [6]. These criticisms foreground a problem close to my worry about naturalness with regard to Prinz’s earlier psychosemantics. But I doubt Shargel and Prinz quite put their finger on it, and it’s worth seeing why. Doing so will bring further clarity to the naturalness datum, and it will help locate Prinz’s more recent work with respect to the argument I have presented here against his earlier work.

Whereas my concern is that the feeling of an emotion should not be made arbitrary with respect to its evaluative content, Shargel and Prinz are concerned that an emotion’s characteristic bodily state should be made arbitrary in this way. These two worries should be kept distinct, because, as we shall see below, the feeling of an emotion is not the same thing as its characteristic bodily state.

So here is how Shargel and Prinz raise their worry:

Informational semantics claims that content is determined by covariation between mental states and the world. But, on that approach, there is nothing about a given mental state that determines what it refers to—mental states are no more than arbitrary symbols…. Therefore it is irrelevant that emotional components are realized by bodily states and bodily perceptions…. One could easily represent the same content without the bodily change, since the theory appeals to their information-processing characteristics and not their bodily character. [56] (p. 113)

One gets the basic sense of the objection here. This passage expresses a concern about emotions being “arbitrary symbols” with no deeper connection to their evaluative contents. At the risk of nitpicking, however, I think one thing about this passage needs straightening out. This concern about arbitrariness should be kept distinct from the observation that we “could easily represent the same content without the bodily change” (ibid). For this is entirely a modal point. As such, it poses no serious objection to Prinz [6]. It is, after all, plainly true that we “could easily represent the same content without the bodily change.” We do so whenever we believe the relevant content.

Shargel and Prinz rely on this modal language again when discussing Griffiths and Scarantino [117], whose theory, they think, renders “the connection between emotion and embodied responses … contingent” [56] (p. 116, my italics). However, whether the connection here is contingent is one thing. Whether it is arbitrary is another. The opposite of the contingent is the necessary. But the necessary is not the opposite of the arbitrary. So we have two distinctions here, and they do not track each other invariably. In particular, the link between a pair of relata can be contingent but still non-arbitrary. For instance, consider onomatopoetic terms. “Quack” refers to the sound ducks make, and it does so both contingently and non-arbitrarily. Likewise, an emotion’s bodily changes might carry evaluative information only contingently; but that does not mean they do so arbitrarily. In fact, it seems like a strange demand that the connection between bodily changes and evaluative information be something other than contingent. Why should we have expected this?

I take it, then, that Prinz and Shargel are not really trying to make a modal point at all, and are better understood as focusing on the arbitrary role bodily changes are made to play in Prinz [6].38 And they think such an arbitrary role is problematic. In Prinz’s earlier psychosemantics for emotion, the status of bodily states qua bodily states is made “semantically irrelevant” [56] (p. 113). This is mistaken, Shargel and Prinz think. The “bodily character” of an emotion’s bodily state should be more directly and crucially involved in the content of the emotion.

However, we should now observe that it is at least unclear why we should accept that particular demand. Shargel and Prinz emphasize that emotions are embodied. Very well. But to my mind, that only suggests that an emotion’s bodily state—and in virtue of its bodily character—should play some kind of important and non-arbitrary role in our theory of emotion. Why must that role pertain to content determination?

Indeed, we can envision a non-arbitrary role for bodily states in emotion which has nothing to do with content. I offer the following suggestion. Let us observe that the bodily state in an emotion prepares us to act in ways that are appropriate in light of the value the emotion represents. This is something Shargel and Prinz are well aware of. They observe, for instance, that in fear, one’s body “prepares for flight from a predator” [56] (p. 128), and that more generally emotions “have long been recognized to have associated actions” [56] (p. 114).

The perceptual theory in Prinz [6] should avail itself of this insight, making use of what I earlier described as an appeal to emotional motivation. To recall, let’s say that in disgust, our emotion represents the sludge as foul. We should also say that one’s bodily state prepares us to rear back or otherwise avoid contact with the sludge. And this is exactly the right response to sludge when it comes for you. One really should rear back. This suggests a solution to Prinz and Shargel’s worry: Emotions represent values by perceiving bodily states which co-vary with those values, and which prepare us to act in ways that are fitting in light of those values obtaining. That sure sounds like a non-arbitrary role for bodily states to me.

Notice that we cannot assail this proposal by crafting a thought experiment like the one I did earlier. In the case of the emotionally inverted creature, we saw an emotion’s phenomenology severed from its motivational impact; and we found that even when motivation was stipulated to be appropriate to the emotion object, a sense of unnaturalness remained. However, at present, aping this move would mean severing an emotion’s motivational impact from its bodily changes. But what exactly would this mean? Perhaps we are to imagine, say, a disgust experience in which we feel a conscious desire to avoid the sludge, but where this felt desire is divorced from bodily changes that prepare us to avoid the sludge, being replaced by bodily changes that prepare us to engage with it. And there would obviously be something incongruous about that. And here, there is no deep question as to why: In this scenario, one’s bodily changes would be fighting against one’s felt desires.

To my mind, this version of the appeal to emotional motivation takes care of Prinz and Shargel’s worry. But I want to make it clear that it does nothing to address the worry about arbitrariness that I have raised. To see this, we have to appreciate that an emotion’s phenomenal character is not just the same thing as its characteristic bodily changes. The one is not reducible to the other.

Granted, when we experience an emotion we typically experience bodily changes. In feeling disgust, one can become aware of one’s stomach churning. But nothing about an emotion’s bodily changes guarantees they will be consciously experienced at all. And even when they are, an emotion’s phenomenal character is not just reducible to the experience of bodily changes. At least two further elements can be seen in the typical case. First, an emotional experience typically has some hedonic valence. It is pleasant or unpleasant to some degree (though surprise raises complications here, as pointed out earlier). Second, in the typical case, an emotional experience also involves some consciously felt motivation to act in some emotion-congruent way. And these two features, hedonic valence and motivational impact, are distinct from one’s experience of bodily changes. Clearly, one can consciously feel one’s racing heart and trembling hands without experiencing this as pleasant or unpleasant, and without experiencing this as pushing one towards fleeing or biting or embracing or whatever else. And even when we imagine a more complex suite of bodily changes, it isn’t obvious why these must be felt as pleasant or unpleasant. And it isn’t obvious why we must experience these as pushing us towards a certain action. After all, when an emotion makes us feel especially motivated, we do not just experience bodily changes: We experience bodily changes as part of a gestalt in which our bodies prepare us to act in some emotion-congruent way. My point is, this gestalt aspect seems to be something over and above the feeling of the bodily changes themselves.

For these reasons, it seems to me emotional phenomenology cannot be wholly reduced to felt bodily changes. And that gives us a clear way of seeing that the natural fit between an emotion’s evaluative content and its bodily changes is not just the same thing as the natural fit between an emotion’s evaluative content and its phenomenal character.

With this in mind, we can see why the solution I’ve suggested to Shargel and Prinz’s worry—a solution that emphasizes the motivational character of the bodily changes in an emotion—does not address the worry I have raised. This solution could address my worry only by appealing to the motivational character of the feelings in an emotion. The feeling of disgust, we’d have to say, motivates us to act in ways that are fitting with respect to the foul, such as avoiding contact. But this is exactly the appeal to emotional motivation we have already considered, and which we rejected for its reliance on the strong dissociability thesis (see Section 4.1). Thus, it turns out there is a solution to the worry Shargel and Prinz raise, whereas the worry I raise remains a problem for Prinz [6].

9. Conclusions

Let us take stock. According to the naturalness datum, the feeling of an emotion bears some natural or non-arbitrary link to the value the emotion represents. Perceptual theorists need to explain the naturalness datum. In addition, perceptual theorists also owe us a psychosemantics, a theory as to why emotions have the evaluative contents they do. And I have argued we should expect a psychosemantics for emotion to explain the naturalness datum as well. That’s because it seems doubtful any content-irrelevant features of emotion can do so. So we should expect content-relevant features to step in, and these will be precisely the features our emotional psychosemantics will invoke.

That’s a problem for Prinz [6]. Prinz’s psychosemantics depends on a Dretske-style theory that makes the feeling of an emotion totally arbitrary with respect to the values that emotion represents. This flies in the face of the naturalness datum. Thus, the emotional psychosemantics from Prinz [6] will either need to be abandoned for some other psychosemantic theory, or else substantially modified. And yet, it is unclear whether any other psychosemantic theory can do any better, with one intriguing exception: the appeal to iconicity. Iconicity is at least the kind of thing that might explain the naturalness datum. But, of course, relying on iconicity in our emotional psychosemantics is itself highly contentious. And the claim that emotions are icons of value would require serious elaboration. In light of this, what we may conclude from this essay is a disjunction: Either emotions are icons of value, or else the perceptual theory cannot explain the naturalness datum and should thus be left behind in search of a theory that can.