Affect: representationalists’ headache

Abstract

Representationalism is the view that the phenomenal character of experiences is identical to their representational content of a certain sort. This view requires a strong transparency condition on phenomenally conscious experiences. We argue that affective qualities such as experienced pleasantness or unpleasantness are counter-examples to the transparency thesis and thus to the sort of representationalism that implies it.

Vision tells us that the strawberry is spotty red, round, and located a short distance from our hands. While even here controversies abound, there is a plausible and well-supported general account of how vision performs this task: our visual system extracts information about the sensible visual features in the environment, and makes this information consciously available to a subject for further processing and action-preparedness (both motor and epistemic). On this model, simplifying quite a bit, perceptual experience represents features of the environment. One of the main projects of recent philosophy of perception and cognitive science has been to fill in the details of this general account. At present there are many options from which to choose, and no clear winner is on the table.

It is difficult enough to work out the details of such an account for even the most basic sensible features, like colors and shapes. And perhaps this explains why so much effort has been devoted to these features in recent years. This focus is fine so far as it goes, but it can have the unintended consequence of obscuring the full richness and variety of perceptual experience. Perception, after all, not only provides information about objects in the environment but also often seems to pass a kind of affective “judgment” on them. This judgment manifests itself in the pleasant and unpleasant character so common to many of our perceptual encounters. There is something especially pleasant and seemingly good about the taste of a juicy, perfectly sweet strawberry on a warm summer day. And there is something especially unpleasant and seemingly bad about encountering an awful smell, like rotting fish, in a confined space.

In both cases, as in the paradigm case of seeing that the strawberry is red and round, there seems to be something—the strawberry—presented to us being a certain way, as pleasant or unpleasant. But whereas it is initially plausible that there is some objective sensible property—call it redness—that our visual system detects and represents, it’s not at all clear that there is an objective analogue in the case of the pleasant and unpleasant. Accounting for these features poses a major challenge to any extant account of perceptual experience. Indeed, we believe, and shall argue in what follows, that these features prove to be fatal to a popular class of views known as strong representationalism. This fatality stems from a crucial ambiguity in the nature of affective experience, one that no version of strong representationalism can adequately explain in its preferred terms. We will lay out the details of this argument in stages, starting with a minimal framework for thinking about perceptual affect, followed by a characterization of our main target.

1 Locus of affect: a framework

It is clear that we find many things pleasant or unpleasant. A walk on the beach on a warm day is quite pleasant. Grading a stack of exams on a warm day, however, is not. These are examples of secondary affect: cases in which some activity or general state of affairs is deemed in positive instances to be enjoyable or fun, and in negative instances to be unenjoyable or a chore. This is a rough characterization, to be sure, but we mention it only to set it aside in this essay. Such cases are not our intended focus. Our interest lies in a more specific occurrence, in cases of primary affect, in which a token sensory experience seems to present us with the pleasant and unpleasant character of things.¹

We start by noting a curious ambiguity found in attributions of primary affect. Let’s call the qualities like the awfulness of smells, the pleasantness of tastes, the painfulness of pains, and so on, affective qualities.² The ambiguity arises when we ask: what do these qualities qualify in the first instance? It seems obvious that, depending on context, we attribute these qualities both to our perceptual experiences and to their objects. Although it seems to apply in all sensory modalities, the ambiguity is especially robust in the case of smells, tastes, and bodily sensations. Indeed, when we talk about how pleasant the taste of a strawberry is, we may be attributing the pleasantness to our subjective experience, or we may be attributing it to whatever objective sensible qualities of the strawberry are responsible for our experience.³ We will assume that both kinds of attributions can be correct. So suppose, in the case where one takes a good bite from the strawberry, the following express two judgments that are true:

• (1) This strawberry is pleasant.

• (2) My (taste) experience of it is pleasant.

(1) Expresses a de re judgment attributing a positive affective quality, pleasantness, to the object of my perceptual experience (or, perhaps more precisely, to the complex of the strawberry’s sensible qualities objectively understood—we will mostly ignore this for convenience). Hence it is what we will call a first-order or perceptual judgment directly (i.e., non-inferentially) prompted by the experience of the strawberry. (2), on the other hand, is an introspective judgment attributing the affective quality directly to the experience of the strawberry (rather than to the strawberry itself).

Can (1) and (2) attribute the same affective quality? Can strawberries and their experiences be pleasant in exactly the same way or sense? The intuitive—and, as far as we can tell, widely shared—answer is no. The qualities attributed in (1) and (2) are distinct properties. Is there some sort of a theoretically interesting connection between the two properties? Here, again, the intuitive answer is yes, there is. Whatever they are, these properties are not just accidentally related. It is plausible to assume that there is some kind of dependency relation between them.⁴ After all, the instantiation of one is somehow implicated in the instantiation of the other. Putting aside the issue of what these properties are and the exact nature of this dependency, for the moment, we can now define two incompatible views about the direction of the dependency:

the experience view (EV):

• Affective qualities primarily/fundamentally qualify perceptual experiences rather than their objects, and

• For any x and affective quality A, if x is A and x is not an (aspect of an) experience, then x is A only derivatively (i.e., x is A only because its experience is A).

the object view (OV):

• Affective qualities primarily/fundamentally qualify the objects of perceptual experiences rather than the experiences themselves, and

• For any x and affective quality A, if x is A and x is an (aspect of an) experience, then x is A only derivatively (i.e., x is A only because its object is A).

Note that, with some fairly natural assumptions, these views entail another dependency ordering among our ways of epistemically accessing affective qualities. EV entails that introspective access to affective qualities is primary/fundamental. On this view, if we ever perceive affective qualities or make (first-order) perceptual judgments about them, it is only because we can introspect those that qualify the experiences. Call this the introspection-first view (IFV). On the other hand, OV entails that the perception of affective qualities is primary/fundamental, and our ability to introspect them (if we ever do) or make introspective judgments about them is derivative—i.e., depends on our ability to perceive them or make perceptual judgments about them. Call this the perception-first view (PFV). We will return to this distinction later.

The picture we have drawn so far is fairly abstract and somewhat formal. It says nothing about what affective qualities are or what the dependency relation among them is. It does not assume any particular view about how introspection or perception works, or what it takes to make perceptual or introspective judgments. It is neutral on the nature of perceptual (and, for that matter, affective) experiences. It does not say whether EV or OV is correct. It takes a cue from the way we attribute affective qualities and sets up a fairly broad framework with very minimal constraints in it. We are inclined to believe that pretty much anyone who is a phenomenal realist about affect can agree with this framework.

However, we will show that this framework has enough meant to cause a serious headache for the defenders of (strong) representationalism about phenomenal character. In particular, the framework presents a fatal dilemma for defenders of strong representationalism. By their own lights, they should prefer to adopt a version of OV. But as we will show in some detail, OV is a thoroughly implausible and deeply problematic account of affective qualities and cannot be adopted by the strong representationalist. So that leaves some version of EV as the only plausible account of affective qualities, but EV is inherently anathema to both the spirit and the letter of strong representationalism. We show that on both standard and extended versions of the view, EV and strong representationalism cannot be plausibly combined. Given the minimal framework, this leaves the strong representationalist with no way to account for affective qualities and, given the ubiquity and centrality of affective qualities in perceptual experience, this poses a serious problem for the view.

Before we spell out our argument in more detail, we need to make clear the kind of representationalism we will target in this paper.

2 Representationalism and the problem of affect

Intentionalism in philosophy of perception is the view that all perceptual experiences are representational. Intentionalist theories vary widely. They can be reductionist or non-reductionist, naturalist or non-naturalist, pure or impure, wide or narrow, Russellian or Fregean, restricted or non-restricted. Some of these classifications cross-classify the theories, and are not completely independent of the others. Furthermore, many of these distinctions are known under different names in the literature, sometimes with (slightly) different extensions. So to forestall potential confusion we start by clearly stating our target: We are interested in the group of intentionalist theories that are reductionist, naturalist, pure, wide, Russellian, and non-restricted, and sometimes known as strong representationalism (hereafter, SR). Defenders include Harman (1990), Dretske (1981, 1995), Tye (1995, 2000), Byrne and Hilbert (1997, 2003), Byrne and Tye (2006).

According to SR, the entire phenomenal character of an experience is one and the same with, or is entirely determined by, the wide representational content of the experience. The semantic externalism assumed by representationalism thus yields phenomenal externalism that says that the phenomenal contents of the experiences of physical duplicates may differ depending on their environments.

Representationalists defend or assume naturalistic accounts intended to explain how the vehicles of experiences realized in the brain acquire their representational content.⁵ Furthermore, they propose or assume naturalistic constraints on the vehicles (or their content) that are intended to explain why only experiential representation has an essential and distinctive phenomenology in the way it does.⁶ The representationalist project is therefore a metaphysical project: the ultimate aim is to understand conscious experiential phenomenology completely in naturalistic and reductionist terms.

Given that conscious phenomenology is essentially introspectable (at least for creatures of a certain sort), representationalism entails the strong transparency (ST) thesis:

(ST) Any quality that we (can) epistemically encounter in the introspection of an experience is a quality only (widely) represented by this experience, thus not a quality of the experience (a fortiori, not an intrinsic quality of the experience).⁷

Because some version of ST is intuitively deemed self-evident on ordinary phenomenological grounds, many representationalists argue that this entailment is an argument in favor of their representationalism.⁸ So, if, in having an experience, there are any qualities we can introspect that are not attributable to the objects of the experience, representationalism is refuted. This is not something we argue for here, but which is assumed on all extant SR accounts. We intend to show that affective qualities are such qualities. We will show that representationalism is false on account of the fact that it cannot explain affective phenomenology in its preferred terms.

To the extent that the affective aspects of experiences are phenomenal, representationalists are committed to giving a representationalist account of them on a par with accounts they give for other sensory-discriminative aspects of experiences. Yet very few representationalists have taken up the challenge and attempted to develop a detailed account. Similar worries arise for other phenomenal experiences such as the so-called intransitive bodily sensations such as pains, itches, tickles, orgasms, etc.,⁹ as well as for general bodily feelings, moods, and aspects of emotions. Many of these pose a challenge precisely because they don’t seem representational at all. The worry about affect is different, since it arises even if we allow that affective qualities are represented (perhaps derivatively) in perceptual experience. So they pose a novel and especially difficult challenge to SR, and in general representationalists have been coy in directly tackling them.

Michael Tye has been a remarkable exception to this trend among representationalists.¹⁰ Since around mid-1990s until very recently, he has written extensively—significantly more than any other representationalists at any rate—on how to extend representationalism to treat these problematic cases. And more recently, in a work co-authored with Brian Cutter (Cutter and Tye 2011), Tye developed and defended a sustained representationalist account exclusively of experiential affect. Although their account is specifically directed towards the negative affective character of pain experiences (the painfulness), they are explicit that the account is meant to generalize to the positive as well as negative affective phenomenology of all experiences. We believe the account C&T develop is paradigmatic for representationalism. Indeed it is almost a textbook example of what a principled (strong) representationalist would or could say about affective phenomenology. Therefore, our criticism, although it takes some aspects of C&T’s account as its stalking horse, is intended to generalize to the core thesis of representationalism itself; that is, roughly to the intentionalist views that are reductionist, naturalist, pure, Russellian, wide, and non-restricted. Such views say that all experiential phenomenal content (non-restricted) reduces without remainder (reductionist and pure) to wide (Russellian) representational content of a certain sort, of the sort that completely admits of a naturalistic psychosemantics (some version of indicator and/or consumer semantics as applied to mental representations that play a certain kind of role in the informational architecture of creatures capable of conscious experiences).

While these various choice points may make it seem like we are picking out a narrow target, this is not so. Most of these various choices mentioned here hold together naturally, offering a robust, unified, and popular way of accounting for the character of perceptual experience. And there are heavy theoretical penalties, especially for the naturalistically minded, for adopting other options here, like Fregean or narrow representationalism, or a nonreductionist or restricted form of representationalism. So while our particular target here may seem overly specific, the worries we raise are quite general and apply to a major form of intentionalism, and can be generalized to other accounts of perceptual experience.¹¹

3 Representationalism about affect

First, a few preliminaries and some terminology. Our perceptual encounters with our environment typically result in the formation of perceptual beliefs. What makes these beliefs perceptual, in the way we want to highlight for the purposes of this paper, are two factors. One is their perceptual content: their content is the kind of content that perceptual experiences can have. The issues surrounding perceptual content are complicated. But in what follows we will focus on an uncontroversial kind of content: sensible qualities understood objectively. Surely, if experiences represent at all, they represent at least the sensible qualities that characterize the perceptual modality through which they are generated. The other factor stems from the fact that one can form perceptual beliefs in all sorts of ways. Here we are interested in the particular way they are formed in direct causal response to perceptual experiences—their informational etiology. Our perceptual experiences typically non-inferentially prompt conceptual responses to the information they contain about their environmental objects and properties. These responses underwrite our ability to discriminate, identify, and recognize features of our environment. We can regard these direct conceptual responses (direct concept applications) as the non-inferential formation of perceptual beliefs—or, to highlight their occurrent character, perceptual judgments. These non-inferential beliefs or judgments are not only causally/informationally prompted by the experiences, but also, under appropriate conditions, epistemically justified by them. When it is important to differentiate these non-inferential tokenings of the perceptual beliefs/judgments with a particular sort of informational etiology, we will mark them as direct perceptual beliefs/judgments (cf. Dretske 1999). We will sometimes abbreviate these as ‘p-beliefs’ (or ‘p-judgments’), and concepts thereby applied, as ‘p-concepts’. It is important to keep in mind that perceptual beliefs are essentially first-order in the sense that they are about the extra-mental objects of experiences that directly prompt them, and not about the experiences themselves (they are not introspective). In other words, the relevant concepts involved in them are directly applied to what the perceptual experiences represent, and not to the experiences or to their phenomenal aspects.¹²

Before presenting the natural representationalist account of affective qualities, it is useful to start with a non-affective case. The contrast will make it easier both to understand representationalism’s main commitments and our criticism of it. Let e stand for a perceptual experience and P stand for whatever phenomenal aspect of e specifically represents a sensible property of an extra-mental object. Let P(e) refer to e’s being (or having) P.¹³ According to SR, when we see an object as red, for instance, our experience is representing the object as red.¹⁴

The “reddish” phenomenology of our experience, P, is identical to (or entirely determined by) the property represented by our experience, in this case, red. Our experience of the object therefore has correctness conditions:

P(e) represents Red(o), and is veridical IFF o is red.

So, when, on the basis of our experience P(e), we form the direct perceptual belief that the object is red, then—bracketing certain skeptical scenarios—the truth value of our belief matches that of the perceptual experience. In general, if a perceptual experience is non-veridical, then so too will be the p-belief based on it, and vice versa. In other words, according to SR, the truth value of a p-belief generally matches that of the perceptual experience upon which it is appropriately based. Call this semantic parallelism¹⁵:

(parallelism) p-belief(o is red) is true IFF P(e) is veridical.

Finally, note that the sensible qualities of perceptual objects, primarily or fundamentally, qualify the objects of experiences—they are objective properties in this sense. If there is any sense in which they ever qualify their experiences, their so doing is only derivative (i.e., somehow depends on their qualifying the objects primarily).¹⁶ This is the analog of OV as applied to sensible qualities. Given SR+ST, we take it that this is all common ground. Now let’s state representationalism about affect.

3.1 Representationalism and the Object View (OV)

Representationalism with its commitment to the ST thesis naturally invites the adoption of the OV, according to which affective qualities such as pleasantness or unpleasantness primarily qualify the objects of perceptual experiences and are represented by these experiences. The experiences get to have affective phenomenology in virtue of representing these affective qualities. Applied to pain, the affective quality, painfulness, primarily qualifies objective bodily conditions such as physical disorders or disturbances of various kinds occurring or about to occur in parts of the sensitive body that are also represented by pain experiences. So a pain experience typically represents both some kind of objective bodily disorder (sensory-discriminative aspect of pains) and its unpleasantness (pain’s affective-motivational aspect). Let’s use ‘d’ to denote the bodily disorder and ‘P’ to denote the sensory-discriminative phenomenology of pain experiences that represents the disorder.¹⁷ Also, let P* be the negative affective phenomenology of a pain experience, e, that represents d as painful.¹⁸

The most fundamental tenet of SR is that P* represents (just as P does) a non-mental objective property of the objects of experiences. This property, according to Tye (1995, 2000, 2005a, b) and Cutter and Tye (2011), is the property of being bad to an individual to a particular degree¹⁹:

The content of a pain experience of an individual A is something like: there is a bodily disturbance of (physiological) type d in location l, and d is bad for A to degree x. (Cutter and Tye 2011, p. 99)

C&T propose that the naturalistic property (H) to be identified with the property of being bad for A to degree x is

H = the property of being apt to harm individual A to degree x.²⁰

Thus, according to C&T, P* represents H (=harmful, for short):

P*(e) represents H(d), and is veridical IFF d is harmful.

C&T comment on the notion of harm involved thus:

We can understand the notion of harm in relation to the notion of a teleological system. Very roughly, something harms a teleological system to the extent that it hinders that system (or one of its subsystems) from performing its function(s). (Cutter and Tye 2011, p. 99–100)

What is important to keep in mind here is that, in both positive and negative affect, H is a non-intentional physical property or a relation that doesn’t involve e or anything mental as a relatum, and is to be identified with the naturalistic property represented by the affective phenomenology of our experiences. So it is an objective property of the objects of experiences, not a property of the experiences.²¹

So far so good. According to OV, the affective quality, being painful, primarily qualifies the objects of experiences. This object is some kind of bodily disturbance, d, in the case of pains. C&T claim that the affective phenomenology of pain, P*, represents d as being apt to harm its possessor to some specific degree (harm or harmful for short). So being painful is identified with being harmful, and is attributed primarily to d. But adopting OV in this way implies the PFV, which in turn implies that one cannot make the introspective judgment that

(i-judg) one’s experience is painful,

unless one is capable of making the perceptual judgment that

(p-judg) d is painful.

So the attribution in p-judg is primary, just as OV says. Hence, given H, parallelism implies that p-judg is false if d is not harmful—in other words, when P*(e) is not veridical.

But now consider the following situation. Let d be the removal of a scar tissue due to a serious burn by a doctor to prevent further pathological complications. So the removal (d) is not harmful. Then, P*(e) is not veridical.²² But then, given parallelism, p-judg, when made by me, is not true. ²³

But clearly, when sincerely made by me using the ordinary concept—what else?—PAINFUL, p-judg is true! We don’t think we need to argue for this point at length, but one way to see this is to see that there is no (independent) evidence that would convince me that when I judge that:

(contradiction) Boy, this is good for me but so painful!

I am making a semantically contradictory judgment (where, ‘this’ refers to the removal, and good = beneficial/not-harmful). If the property attributed in p-judg were identical to the property of being harmful (=H), contradiction’s truth-conditions couldn’t obtain in any possible world—formally spelled out: ‘this is both harmful and not-harmful’). But clearly I would insist on its being true and balk at the suggestion that I am in fact making a contradictory judgment. And this is the ordinary default situation (not just me). Note that these kinds of judgments aren’t rare in the case of first-order affective judgments—there are plenty of them we encounter in everyday life. Take, for instance, just one tiny portion of the plethora of daily examples: parents’ constant pleadings like “eat/do x, it is good for you”, where x is also judged to be unpleasant). If p-judg and its ilk attribute H to the objects of our experiences, then a lot of the situations we are commanding, pleading, praising, recommending, judging, etc., are literally contradictions.

Can representationalists say that the identity between, say, painfulness and harmfulness is a posteriori? To see that this doesn’t help in this context, consider other contradictory/false beliefs arising out of failure to believe an a posteriori identity: when the reference of the terms/predicates in these identities are independently traced or established, we readily withdraw our judgments on ground of realizing that they are false or contradictory. For instance, when Pierre learns that London = Londres, he would no longer assent to “Londres is pretty but London isn’t” (Kripke 1979). Same with contradictory judgments using indexicals, demonstratives, or names. Same with judgments using natural kind terms. But what evidence could convince me (or, any person who is not a dogmatist about SR) that, when I judge contradiction, I am in fact expressing a contradictory belief? The challenge here for the representationalists who might be tempted to claim aposteriority, then, is to come up with independent evidence for the truth of SR+OV about affective qualities. We don’t see how they can do that. But, in fact, they don’t even try, since, as far as we can tell, they already grant that affective qualities are not to be identified with H (see below).

Furthermore, insisting on the a posteriori identity and therefore the actual falsity of p-judg would amount to changing the topic. A representationalist account that would insist on systematically correcting people’s judgments about what/when they find things painful/pleasant would not be an account of experiential affect as this notion was originally targeted for explanation. Recall that the representationalist project is to reduce all experiential phenomenology to representation (of a certain sort). SR agrees that many experiences have an affective phenomenology. The professed project, therefore, is to understand the pleasantness or unpleasantness of experiences in representational terms. The explanandum here is a phenomenon conceived in a certain way. This way gives the experiencers a certain sort of cognitive and epistemic authority regulated by the phenomenology of their affective experiences. Even if we grant that the affective phenomenology (mis)represents H, our conceptual responses to this phenomenology obviously don’t follow its purported representational norms and function in the way SR envisions. An account like SR+OV that implies that they do is empirically false. An account that suggests that they ought to is at best unmotivated, or at worst radically revisionist in a way that misses its original explanatory target. Anyway, we don’t think we need to elaborate this any further: as said, almost all representationalists reject the identity of affective qualities with H.

3.2 Representationalism and the Experience View (EV)

So contrary to natural expectations, OV isn’t the right view for representationalists. And, indeed, Tye (2005a, b) and Cutter and Tye (2011), along with almost every other representationalist, reject OV about affective qualities. Instead they hold what we have called the Experience View (EV), according to which affective qualities primarily/fundamentally qualify perceptual experiences rather than their objects.

One way to motivate EV is to observe that given H, the properties attributed in i-judg and p-judg can’t exactly be one and the same property: experiences aren’t harmful in the same way their objects can be harmful. We have made the same observation before in the context of pleasantness when we were setting up the framework for the locus of affect attribution. Similarly, it is not implausible to say that ‘painful’ in i-judg and p-judg is ambiguous—irrespective of the nature of H. Indeed, following common sense, Tye finds it quite plausible that:

…when we say that something is painful, what we normally mean is that it causes the feeling of pain. Thus, we speak of a cut, a bruise, a sore, an operation, a cough, a lashing as painful, that is, as causing pain. This is why, if a person feels a pain in his left arm, and his doctor informs him that the cause of the pain lies in a disturbance in his heart, we allow that what is really painful here is the disturbance in my heart. If the doctor has made no mistake, that is what is actually hurting the patient. It is evident that [pain experiences] themselves are not painful in this [derivative] sense. Pain does not cause the feeling of pain. (Tye 2005a, p. 114)

So, given the proper disambiguation, in p-judg, d is painful in the sense that d is causing a painful experience, and in i-judg, the sense of ‘painful’ is primary. So contrary to first-impressions, judgments like p-judg that attribute affective qualities to the objects of affective experiences don’t track a completely objective property of these objects such as their being harmful or beneficial. In this regard, the attribution of affective qualities in the act of experiencing them is radically different from the parallel attribution of sensible qualities to the objects of our experiences.

As should be expected, this view comes with an introspection-first view (IFV), which implies that affective qualities that primarily qualify our experiences (=EV) are introspectively knowable directly, i.e., not by an epistemic detour through our capacity to make direct perceptual judgments about the objective qualities of their objects. In other words, when we make introspective judgments like i-judg, our attribution of affective qualities to our experiences does not depend on our capacity to have perceptual judgments about the relevant range of objective properties of their objects.

However, this seems to be a clear violation of the ST thesis (ST): it implies that we can come to know a quality of our experiences without necessarily coming to know a corresponding quality of the object of our experiences represented by the former quality. But if ST is genuinely counter-exampled, SR is refuted. Given the untenability of OV, EV is the only option for representationalists. And in fact, they routinely seem to take that view. But EV violates ST. And if ST falls, SR falls. As far as we can tell, representationalists, including Cutter and Tye, haven’t yet fully appreciated the true impact of adopting EV about affective qualities. It’s quite puzzling why. Here is a guess. Like many others, C&T correctly think that even if my experience misrepresents d, it is still painful. So, on this basis, it would be natural for me to come to believe that

(feels-bad) d feels bad/painful.

This is not, of course, of any immediate help to explain the difficulty with adopting EV since this is not the content of a first-order p-belief anymore. It is a belief about how d is experienced by me. It’s an introspective belief similar to i-judg in the sense that it is essentially a commentary on my experience—albeit as an experience of d. feels-bad is ambiguous between two readings:

(feels-bad _E) I am painfully feeling d

and

(feels-bad _O) I am feeling d as painful.

These readings would roughly correspond to adopting EV and OV about painfulness respectively. When C&T make claims like feels-bad, they may have feels-bad _E in mind but somehow think that it is compatible with SR on account of the availability of the second, feels-bad _O, reading. The trouble, as we have seen, is that neither reading is of any help to explain how they can hold EV and think that ST is preserved—while they seem to reject OV.

Thus neither OV nor EV is an option for representationalists: they can’t answer the locus question about affective qualities. But as far as we can tell, there isn’t any other option for representationalists. Nevertheless, the next section presents a likely response that they might try.²⁴

4 Reformulating representationalism about affective qualities

Here is how it would go: painfulness is indeed a phenomenal feature of pain experiences. This feature is not to be identified with harmfulness. Indeed experiences are not themselves (let’s grant) harmful. It’s what they represent, d, that is harmful, and thus, d is only derivatively painful. So EV about affective qualities is correct. In our evolutionary history, we have, for some reason, picked up a way of thinking/talking about affective experiences so that we attribute the affective qualities primarily to the experiences, rather than to their objects. But one can still argue that the property that we attribute to experiences when we judge them to be, say, painful is their representing something as harmful to the experiencer. So there is the representational vehicle, the pain experience (e); there is also what it represents, bodily disturbance (d) and its being harmful. Then there is the intentional relation between the two: the vehicle e’s representing d as harmful. Because it is an intentional relation, e can represent something as harmful even in the absence of that “something”. So e’s representing d as harmful is a property of e, and it’s properly intentional as required by SR. To explain how e could come to have this property is the goal of various naturalization projects in psychosemantics. But for now, the proposal is that painfulness (or affective qualities in general) is to be identified with the representational property of e: namely, e’s representing something as harmful. So:

(PH) e’s painfulness = e’s representing-something-as-harmful-to-the-experiencer-A-to-degree-x (e’s representing-harm, for short).

This claim should not be confused with:

d’s painfulness = d’s being apt to harm individual A to degree x (d’s being harmful, for short).

SR does not propose to identify affective qualities with objective (at least, non-mental/non-intentional) properties of objects of experiences (i.e., with H). So, following EV, it might be claimed that affective qualities are special in some way. This is why we have separate concepts and words for them that are attributed/applied to the affective experiences themselves, rather than to their objects. But what is being attributed (an intentional property as per (PH)), a representationalist might claim, is still perfectly in line with SR and naturalistically kosher—assuming we have a successful naturalistic psychosemantics.

For comparison, consider the visual experiences that represent red things. There is a particular visual phenomenology to seeing red. Outside philosophy classes, usually we don’t have ordinary concepts/words to refer to this phenomenology. We say things like: the way red things appear or look to us. But, if we wish, we might introduce quasi-technical terms, like ‘red*’, ‘red-prime’, or ‘reddish’ experiences to refer to the particular phenomenal feature “distinctive of experiences as of red things” (Cutter and Tye 2011: 93). By stipulation, then:

(STIPULATION) The property of being reddish = the phenomenal property distinctive of experiences as of red things.

The representationalist proposal then is this:

(RR) e’s reddishness = e’s representing-something-as-red (say, as having a particular surface spectral reflectance profile).

When we have a naturalistic psychosemantics to explain the representation relation in question, we will solve the problem of experiential phenomenology, and hopefully, the problem of phenomenal consciousness. So what is the problem?

One of the immediate problems with this response is that it seems to violate ST. Recall:

(ST) Any quality that we (can) epistemically encounter in the introspection of an experience is a quality only (widely) represented by this experience, thus not a quality of the experience (a fortiori, not an intrinsic quality of the experience).

If we accept the suggestion that the property of representing-harm is a property of the experience e and thus not widely represented by e, then, to preserve ST, we need to conclude that we don’t (can’t) epistemically encounter this property in introspecting e. But this is absurd: under the proposed new formulation of SR, the point is precisely that this is a property of e and we can and do directly become aware of this property. The spirit of the proposal is that this property is still an intentional property, and so, is not an intrinsic property of e—hence the thought is that it’s kosher for representationalists to advert to such properties in introspection. So, at a minimum, we need to reformulate ST following the above reformulation of SR. But it is not at all clear how to do this in a way that will still support SR over weaker forms of intentionalism. Here is one way it may go consistent with the reformulation of SR (based on PH and RR above)—we may call it weak transparency:

(WT) Any quality that we (can) epistemically encounter in the introspection of an experience is an intentional quality of this experience,

where, for instance, reddishness and painfulness (as fixed by PH and RR above) are among the introspectable intentional qualities of an experience. But this is too weak. It is compatible, for instance, with forms of intentionalism that are not (strongly) representationalist: intentionalist theories about phenomenal character that are Fregean or narrow or impure would endorse WT, as would the (non-reductionist) qualia theories that take a phenomenology-first approach to intentionality. This is not accidental. If ST follows from SR, then, of course, any transparency thesis weaker than ST (that may be compatible with EV) will also follow from SR. Thus WT may be true but irrelevant. Remember that we have reformulated SR in the above way because we were trying to understand how adopting the Experience View about affective qualities is compatible with ST. The representationalist suggestion on the table now is that the affective qualities we attribute to experiences are (as per PH) in fact intentional properties. But this suggestion, or WT for that matter, doesn’t help us to solve the original problem, namely, how ST and EV can be compatible—unless it is a tacit admission that they are not.

The main problem with the suggested reformulation of representationalism is that there is a very crucial difference between (PH) and (RR): grasping the property of being reddish in introspection is “transparent” in a way grasping the property of being painful is not. If so, we still don’t have an explanation of how EV and ST can be compatible. Let us explain.

5 Introspective knowledge of affective qualities

To start, note that the identity proposed in (PH) between e’s painfulness and e’s representing-harm is meant to be a theoretical identity and it is by no means obvious; in fact, in some ways, it is quite perplexing. For, to begin with, painfulness doesn’t seem to be an intentional property at all: if (PH) is true, then when we judge our experiences as painful (pleasant, etc.), we are, as it turns out, attributing an intentional property to them that makes them answerable to semantic evaluation. So in the above example, when I judge my pain experience to be extremely painful when, on SR, it misrepresents the removal of scar tissue on my face as harmful, I am in fact attributing an intentional property to my experience that makes it, on this occasion, non-veridical. Despite my (conceptual level) knowledge that the removal is not harmful, and in fact, beneficial, I am not (unless I am an avid representationalist) in the least bit inclined to attribute falsehood to my painful experience; more dramatically put, I am not in the least bit inclined to blame my experience for deceiving me or telling me a lie. In other words, despite my finding my experience very painful, I don’t at all find it as (mis)representing the removal of scar tissue on my face as harmful. We take this observation to be true for most ordinary people who have the relevant concepts. This observation doesn’t immediately refute SR, of course. For such epistemological or psychological incongruities are not uncommon in theoretical identities that may sometimes involve surprisingly incongruent concepts flanking the identity sign. But we think that this “theoretical” identity is special and requires additional independent evidence given the general motivation behind defending SR and SR’s general commitments.

Consider (RR). ‘Reddish’ is a technical term (just as ‘red*’ or ‘red-prime’ are), not ordinarily used to attribute a phenomenal property to one’s visual experiences. Nevertheless, it is, let’s grant, a well-defined term stipulatively introducing a more or less well-behaving property (well, let’s assume it does). What is the status of (RR)? Defenders of SR intend to take (RR) and the like also as an a posteriori theoretical identity. But clearly there is a difference between (RR) and (PH). If (PH) and (RR) are true, we are introspecting the property of representing-harm and the property of representing-red when we introspect our relevant experiences. But, on SR, we need perceptual concepts to make introspective judgments about these experiences.

Given how we have introduced the concept REDDISH, we can come to know that our experiences are reddish only if we have the concept RED that applies to red things that these experiences represent. Given the general motivation behind SR, this is as it should be: our perceptual experiences are strongly transparent in that anything phenomenal we can come to know about them is, somehow, via knowing what they represent. Indeed, according to SR, when I look at a red bulgy tomato in good light, I know that it (phenomenally) looks red to me—indeed, I know what it is like to experience the redness of the tomato—only if I can apply a first-order perceptual concept, RED, to what my experience represents. The p-concepts like RED involved in this sort of introspective knowledge may be special in that their acquisition may require serious epistemic/psychological constraints on how they depend on the experiences from which they are directly and immediately acquired. Whatever the case may be about how these p-concepts are acquired and deployed, according to SR, they apply not to the experiences but to what these experiences represent. This is all common ground.²⁵ Given all this, there appears to be no special puzzle about how I acquire the introspective knowledge that my experience is reddish or that something looks red to me.²⁶

Given (STIPULATION), when I look at the red tomato in good light, it is a priori that

I know that my experience is reddish only if I know that my experience is of the sort distinctive of experiences as of red things.

So there is literally a conceptual connection between the concepts REDDISH and PHENOMENAL-PROPERTY-DISTINCTIVE-OF-EXPERIENCES-AS-OF-RED-THINGS, and, for that matter, REPRESENTS-RED, in that they can’t be possessed without possessing the concept RED. So,

Necessarily, one has the concept REDDISH only if one has the concept RED.

We take it that what we have said so far about the introspective knowledge of our experiences of sensible qualities is all common ground between us and representationalists.

But, on the representationalist proposal we are now considering, the introspection of affective qualities works radically differently. It is not a priori that

I know that my experience is painful only if I know that my experience is one of those distinctive as of harmful things.

If I don’t have the concept HARMFUL or I have additional empirical information of a certain sort, it may not be even true. As should be obvious, there is no conceptual connection between the concepts PAINFUL and HARMFUL, and for that matter, REPRESENTS-HARM. So it is just not true that:

one has the concept PAINFUL only if one has the concept HARM (or, HARMFUL, or REPRESENTS-HARM).

These concepts are distinct and conceptually unconnected. One is a monadic concept and the other is complex. Their acquisition and application conditions are quite different and independent. But to say that they nevertheless attribute one and the same property doesn’t help to remove the mystery of how EV and ST are compatible.

If SR+ST is true, it is not accidental that the introspection of experiences requires the availability of the very same p-concepts that apply to the qualities of the objects of experiences: REDDISH is, after all, a technical term conceptually manufactured out of the concept RED (see above). So it is not accidental that there is an a priori connection between RED and REDDISH (i.e., EXPERIENCE OF RED or REPRESENTING-RED): it is required by SR+ST. On this view, the only sense in which I know that my experience is reddish is the sense in which I know that my experience is an experience as of red, or that it represents red. If I don’t have the p-concept that applies to the object of my perception (not to say anything about concepts such as EXPERIENCE, REPRESENTS, etc.), then I can’t have the concept required to introspectively believe what experience I’m having. In brief, one can’t track an intentional property qua intentional without representing to oneself what the intentional property represents—indeed without one having any clue whether the property in question is even intentional. But, if so, insisting on (PH) would amount to admitting that there are qualities of experiences that we can directly become aware of (i.e., we can “i-conceptually” track) without, strangely enough, becoming aware of (i.e., without “p-conceptually tracking”) the qualities represented by these experiences. So adopting EV and thus the introspection-first view (IFV) for affective qualities threatens the core of representationalism by making ST optional.²⁷

Here is another way to make our point. Recall that when I make the introspective judgment:

(i-judg) my experience is painful,

I am not attributing the property of being harmful to my experience in exercising my concept PAINFUL—as agreed by all parties, painful experiences aren’t harmful in the intended sense. Furthermore, whatever property I am attributing to my experience, that property is not conceived by me in a way that requires me to have a p-concept that applies to what my painful experiences (by being painful) represent. i-judg is not an inferential or abductive judgment: it is introspective and seems to be the immediate result of directly applying my i-concept, PAINFUL, to an instance of an affective quality of my experience. This means that I am introspectively capable of responding to a quality of my experience without having the slightest clue what that quality may represent, or even whether it represents anything.²⁸ It is a mystery on SR how we can directly apply i-concepts like PAINFUL to our experiences without having the slightest idea that this quality may be representational (let alone what it might represent). This is just the denial of ST that representationalists are committed to.

Indeed, it is very natural to use the following form, instead of i-judg, in expressing my introspective judgment:

(i-judg*) My (pain) experience has this quality,

where the intended demonstration is clearly to an instance of painfulness instantiated in my experience. Here the demonstration can’t possibly be to an instance of harmfulness: as said, my experience isn’t harmful in the way its object (d) might have been harmful. I may go ahead and add to i-judg*, “it’s painful” or “it’s unpleasant”—expressing a fact about how I conceive of this quality. Clearly, this seems to be an act of inner demonstration involving at the same time an act of direct recognition and thus categorization, which suggests that I am directly applying a “recognitional” concept, PAINFUL or UNPLEASANT, to my experience. We have already seen that I cannot be applying this concept to d either—remember we are trying to understand how ST can be compatible with EV.

So the suggestion that EV is true of affective qualities but that affective qualities are a posteriori identical to intentional qualities of experiences in the way proposed by (PH) turns out to be not compatible with the ST claim that representationalists are committed to. But if ST is falsified by affective qualities, so is representationalism. Given that the representationalism under review is non-restricted (i.e., is meant to apply to any kind of phenomenal character), SR is just false.

6 Conclusion

We have argued that representationalism about phenomenal character is false on the ground that representationalists can’t give a satisfactory account of affective qualities consistent with the core tenets of their representationalism: neither OV nor EV provides a home for affective qualities.

There are of course more general considerations against SR about affect. The unpleasantness or pleasantness of an experience, on SR, is just the property of this experience’s representing its object as harmful or beneficial. Clearly these properties have different second-order properties. For instance, affective experiences such as pain are intrinsic motivators—they motivate the experiencer solely on the basis of their presence. This is not true at all about the property of an experience’s representing its object as harmful or beneficial. Representationalists may appeal to the functional role of these experiences, but this would be an appeal to the wrong property. Functional role can explain an experience’s being an intrinsic motivator only qua causal role, not qua representor of harm/benefit. Moreover, making the further claim that that functional role also constitutes the experience’s representing harm or benefit doesn’t add anything to the explanation by functional role alone. It’s redundant at best. (Cf. Aydede 2005, 2013)

There is also the fact that affect attaches to experiences in all sensory modalities that give rise to conscious experiences. A representationalist needs to find a common second-order property for all pleasant or unpleasant experiences to represent across all modalities. It is very doubtful whether she has the resources to do that on the basis of what is harmful or beneficial to organisms (or, what has historically harmed or benefited their ancestors).

Furthermore, affective responses vary enormously across as well as within individuals with very strong conative and behavioral effects even for subtle differences in the representational content of experiences. For instance, while I like dark chocolate, I dislike milk chocolate; while I like cauliflower, I hate broccoli. But my brother’s preferences are just the other way around: He likes milk chocolate, but hates dark chocolate, etc. Further, my preferences were just like my brother’s when we were growing up. Tons of variation! Note that if the representationalist sticks to the second-order properties like apt to harm or benefit, misrepresentation will be the rule (not the exception!) in modern history when explaining affective phenomenology—small wonder then why the truth-values of our first-order (derivative) affective judgments don’t follow those of our affective experiences (even supposing that experiences represent harm or benefit). These properties are just too coarse grained to do justice to our intuitions about affective aspects of our experiences.

We conclude that the headache caused by affect is fatal for representationalists.