Armchair Evaluative Knowledge and Sentimental Perceptualism

: We seem to be able to acquire evaluative knowledge by mere reﬂection, or “from the armchair.” But how? This question is especially pressing for proponents of sentimental perceptualism, which is the view that our evaluative knowledge is rooted in affective experiences in much the way that everyday empirical knowledge is rooted in perception. While such empirical knowledge seems partially explained by causal relations between perceptions and properties in the world, in armchair evaluative inquiry, the relevant evaluative properties are typically not even present. The paper shows how sentimental perceptualists can ultimately provide a broadly causal explanation of our reﬂective evaluative knowledge.


Introduction
How do we learn about value? One historically prominent answer is what I call sentimental perceptualism, a view which aims to demystify evaluative knowledge by modeling it on perception while likewise casting its source as close to motivation. To a first approximation, it says that our evaluative knowledge is rooted in desiderative and emotional experiences (or simply, affective experiences), much as our everyday knowledge about the world around us is rooted in perceptual experiences. I use 'evaluative' here in a broad way that is inclusive of, e.g., reasons for action, rightness/wrongness, and goodness/badness. Furthermore, I focus on versions of sentimental perceptualism that are realist about value. Just as perceptual experiences help us to learn about a reality that exists independently of our experiences, affective experiences help us to learn about an evaluative realm that is no mere projection of those experiences. 1 Sentimental perceptualism faces challenges, however, and one of the most significant is what I call the armchair challenge. Concern arises because we appear to be able to acquire evaluative knowledge by mere reflection, or 'from the armchair'. In this way, evaluative knowledge can seem to have more in common with various sorts of a priori knowledge than it does a posteriori empirical knowledge rooted in perceptual experience. The perceptual analogy at the heart of sentimental perceptualism thus appears to break down in a puzzling fashion. This paper explains how sentimental perceptualists can allow for a priori evaluative knowledge, or at least something near enough. 2 I begin with a detailed description of sentimental perceptualism and the armchair challenge. It turns out that the difficulty for sentimental perceptualists is more peculiar than we might have thought. Most value epistemologists who have attempted to explain mere reflective evaluative knowledge assume that such knowledge is always of necessary truths. While this is an attractive starting point given a rationalist value epistemology (e.g., [12]), it is dubious within a sentimentalist framework. For as I explain, there is a strong case to be made that affective evaluative experiences can ground knowledge of contingent truths if they ground any such knowledge. An attractive sentimental perceptualist theory of armchair evaluative knowledge, then, is going to take a shape that is crucially different from familiar rationalist alternatives. 3 In the face of this difficulty, the sentimental perceptualist may be tempted to reject the possibility of acquiring evaluative knowledge by mere reflection, adopting instead a purely a posteriori evaluative epistemology. This paper, however, explores what happens when sentimental perceptualists stand firm, agreeing with the rationalists that we can obtain evaluative knowledge from the armchair.
I argue that a promising approach to the armchair challenge is ultimately by way of an independent, further challenge, the content challenge. This challenge, which I describe in detail below, is to explain how affective experiences ever came to have evaluative content. Mark Schroeder calls this the 'fundamental problem' for views that ascribe evaluative content to desire [14] (p. 127), and Karl Schafer likewise maintains that it may be 'the most serious problem' for such views [15] (p. 268). I do not believe that there are any easy solutions to the content challenge-whether the focus is only on desires or affect more generally-but the challenge is one that sentimental perceptualists must address, for it threatens the entire psychological framework of the theory. Nevertheless, I argue that we should be cautiously optimistic about a certain strategy for answering the content challenge. I then argue that, insofar as the strategy is successful, sentimental perceptualists secure a novel, straightforward solution to the armchair challenge, one that has some distinctive advantages (and disadvantages) compared to rationalist explanations of armchair evaluative knowledge. In this sense, the conclusion of the paper is conditional. My aim is not to put forward decisive solutions to the armchair or content challenges; it is instead to sketch the framework of a more comprehensive sentimental perceptualist theory, one which faces up to some of the view's most important outstanding issues.

Sentimental Perceptualism and the Armchair
Sentimental perceptualism, as the label suggests, is a blend of sentimentalism and perceptualism. Let us begin with the latter. 'Perceptualism' is defined in terms of an epistemic analogy between evaluative experiences (more on these momentarily) and ordinary perceptual experiences. By 'ordinary perceptual experiences', I mean experiences in the traditional five senses (vision, audition, etc.). The basic idea of perceptualism, then, is as follows: just as an ordinary perceptual experience (e.g., of a dog running) provides some basic justification for believing that the perceptual experience's content is true, so too does an evaluative experience (e.g., of an act being wrong) provide some basic justification for believing that the evaluative experience's content is true. Perceptualists assume these experiences, which are a source of basic justification, can also be a source of basic knowledge. This invites the question of what it is for justification and knowledge to be basic. I leave it open how best to answer this question, though, since answers vary and it ultimately is not important for answering the armchair challenge. 4 The simple, non-technical way of thinking about perceptualist value epistemologies is that there are value experiences that are our ultimate source of data about value. This means that perceptualists can allow, say, that evaluative testimony is a source of knowledge, but as perceptualists, they would qualify that testimony is a source of knowledge because evaluative experiences are.
A central issue dividing perceptualists concerns the nature of the relevant evaluative experiences. 5 Some are rationalists, appealing to intellectual experiences of value. On these theories, perceiving value is akin to perceiving the truth of a mathematical proposition (e.g., [12]). 6 Others maintain that we can literally see and hear things as valuable (e.g., [20]). The sort of perceptualism I am interested in, however, appeals to affective experiences of value. I use 'affect' here to denote desiderative and/or emotional experiences, emphasizing their valenced (positive or negative) felt quality that, according to some theorists, represents value. For example, here is Sabine Döring, describing an agent's negative emotional response to a caretaker's harsh punishment of a toddler: "In experiencing indignation at the harsh punishment of the toddler, it seems to you that the punishment is in fact unjust: your occurrent emotional state puts forward your indignation's content as correct. This is in analogy to the content of a sense perception. In perceiving that the cat is on the mat, it seems to you that the cat is actually there." [21] (p. 377) Similar views are defended about desire. Graham Oddie, for instance, argues that in desiring we experience the object of our desire as 'needing to be pursued' [6] (p. 41) and 'the good just is that which needs to be pursued' [6] (p. 41). Thus, desires are experiences of the good. He then goes on to build a sentimental perceptualist epistemology around these desiderative experiences of value. 7 However, like other sentimental perceptualists (and unlike his intellectual perceptualist cousins), he has nothing to say about the armchair challenge, which I now describe.
Almost anyone who has taken an introductory ethics course has thought about whether it would be right to flip a switch to divert a runaway trolley onto a separate track, thereby saving the lives of five strangers but also killing a single stranger who would not otherwise have been killed [23]. This is a paradigmatic instance of armchair evaluative reflection; there is no actual interaction with a value-laden scenario involving a runaway trolley. It is difficult to overstate just how central mere reflective evaluative inquiry appears to be; we engage in it whenever we investigate how the evaluative landscape might be as opposed to how it is. This can happen not only in the philosophy classroom but also when one contemplates, say, whether a fictional hero made the right choice or whether a certain politician would make for a good president. 8 Here, I assume a version of perceptualism that accepts that we can learn evaluative truths through mere reflection by having evaluative experiences-which, according to sentimental perceptualism, are affective in nature-in response to our imaginings.
An easy way to see how the armchair challenge arises is to notice an apparent asymmetry between evaluative and empirical inquiry. 9 Although we can think about empirical questions from the armchair (e.g., what would happen if I pushed my laptop off the table?) and such reflection can help us to gain knowledge, such armchair inquiry is typically understood as epistemically subordinate to non-armchair investigation. In other words, the ultimate arbiter for empirical questions is (literal) perception, what we observe in the world. However, there does not always seem to be the same kind of asymmetry in the evaluative case. The evaluative experiences we have in response to our perceptions and beliefs about the world-our online evaluative experiences-do not seem always to be privileged over the evaluative experiences we have in response to imaginings-our offline evaluative experiences. 10 Consider the case above of a runaway trolley; it does not seem as if actually witnessing the event would improve our epistemic position in such a way that that experience should (all else equal) take priority over our responses to the vignette. To borrow an observation from Sarah McGrath, while we rely on actual experiments to figure out the nature of the empirical world, evaluative inquiry generally only seems to require thought experiments [27]. 11 The armchair challenge for perceptualists (sentimentalist or otherwise), then, is to develop their theory in a way that respects the intuition that online evaluative experiences are not always privileged over offline ones when it comes to grounding evaluative knowledge. Or, put slightly differently, it is to develop a theory that respects the idea that we can sometimes acquire evaluative knowledge simply by reflecting. Yet, some perceptualists may be tempted to reject this challenge. In that case, they will avoid one difficulty at the cost of incurring a new one, namely that of explaining why we cannot, after all, gain evaluative knowledge simply by reflection. 12 This is not a path that I take here. 13 In this paper, I confront the armchair challenge, and I do so from the perspective of sentimental perceptualism rather than, as is more typical, intellectual perceptualism (e.g., [12]) or non-perceptualist versions of rationalism (e.g., [18]).
Crucially, the challenge is about knowledge rather than mere justification. Some perceptualists focus primarily on justification, but it is arguably much easier to make sense of justification from the armchair. Suppose, following Huemer, we accept the principle that if it seems to you that P, then you are defeasibly justified in believing that P [12]. On this kind of picture, there is no deep puzzle about reflective evaluative justification, so long as we have evaluative experiences from the armchair. But forming a belief on the basis of an accurate experience is certainly not enough for knowledge, since the experience may only be accidentally correct. To take a familiar example, a person suffering from a brain injury might hallucinate that there is a red apple and believe justifiably on this basis that there is one. Still, they do not seem to know that there is a red apple, for they are only right by accident [30]. What is required for a perceptual experience to be non-accidentally correct? The most common idea is that perceptual knowledge requires (among other things) a causal connection between the experience and what the experience is about. The difficulty is that appealing to causation may not seem open to (sentimental) perceptualists, for when we merely reflect, we are not in causal contact with any evaluative properties. 14

Deepening the Challenge for Sentimental Perceptualism
One may be tempted to think that there is no special problem here for value epistemology independent of a general puzzle about how synthetic a priori knowledge is possible (if indeed it is). To illustrate, consider that many ethicists assume the evaluative truths we learn from the armchair are necessary, and there are familiar explanations of how we can know necessary truths, evaluative or otherwise, by mere reflection. Borrowing from David Lewis, one might think that any experience of a necessary truth must be non-accidently correct if it is correct at all [31] (p. 113). Alternatively, intellectual perceptualists may be tempted by a view according to which intellectual experiences arising out of an adequate understanding of necessarily true propositions (whether about value or mathematics) are going to non-accidentally track the truth. 15 However, sentimental perceptualists have not embedded their theories in general accounts of synthetic a priori knowledge. I would argue that there are good reasons to keep things this way. Even if affective experiences can sometimes give us knowledge of a necessary truth (see below for how), it does not seem as if they must. In this section, I briefly review an argument I have given elsewhere for why affective experiences are generally suited for giving us knowledge of contingent evaluative truths, if they are suited for giving us any such knowledge at all [34]. I then explain how this conclusion makes the armchair challenge for sentimental perceptualists distinctive and pressing. (Errol Lord raises the armchair challenge in similar terms, and with reference to an earlier version of the present paper. See [35] (pp. 91-93). In noting this, I do not mean to indicate that he would endorse any of my solutions.) To begin, let us first distinguish between necessary and contingent evaluative truths. The former are true in all metaphysically possible worlds. For example, I suspect that the following is necessarily true: torturing people just for fun is wrong. Contingent evaluative truths, by contrast, are not true in all possible worlds. It is only contingently true, say, that Eduardo ought to attend his daughter Emma's soccer match this afternoon. After all, Eduardo's son might have just gotten into a serious accident, in which case he ought to go to the hospital instead, or it may be that Emma does not have a soccer match this afternoon. Now consider a different sort of contingent evaluative truth: if Emma were to have a soccer match, then it would be the case that Eduardo ought to go. This truth is also contingent, for there are worlds in which the antecedent is true but the consequent false (e.g., ones in which Eduardo's son is in a serious accident). We could also formulate a contingent evaluative truth that is general: normally, parents ought to attend their children's soccer matches. This is true, but we can imagine a world in which it is not normally true that parents should do this (e.g., ones in which children are normally horrified to compete in front of their parents).
The basic explanation of why affect seems suited for providing us with knowledge of evaluative contingencies is as follows. The nature of affective dispositions is not such as to generate affective experiences only given inputs that necessitate the presence of the relevant value. 16 When I say 'inputs' here, I am referring to the contents of the mental states to which we affectively respond. Borrowing an example from [34], suppose that Jacqueline becomes angry at Sydney, a famous athlete who she once adored, when he uncaringly stomps on her toe. The anger here is a response to the content (represented in, e.g., belief or perception) that Sydney uncaringly stomps on her toe. The sentimental perceptualist takes the anger to contribute evaluative content. Suppose anger is best understood as a representation of wrongness. In this case, the content of the anger (including the inputs that trigger it) is something like this: 'Sydney wrongs me by uncaringly stomping on my toe'. This is not a necessary truth, taking for granted that it is true. The reason that it is contingent is not merely that Sydney may never have uncaringly stomped on her toe. It is also possible that Sydney is confused from a brain injury, or alternatively, on a drug that creates an overwhelming urge to toe-stomp. 17 Despite these possibilities, however, Jacqueline's disposition to feel anger may be entirely well-functioning. Furthermore, on a sentimental perceptualist theory, we would expect that such well-functioning affective experiences to be the ones apt to ground knowledge, and such experiences often arise on the basis of information that normally, rather than always, occurs alongside the corresponding value.
Turn now to the armchair. Suppose that Jacqueline's anger responds to a hypothetical version of the same case. 18,19 Because the emotion is a response to imaginings, rather than beliefs and/or perceptions, the relevant truth would be better captured in counterfactual terms (e.g., 'if Sydney were to uncaringly stomp on my toe, then he would wrong me'). 20 This truth is contingent, for there are worlds (distant ones, I am assuming) in which the antecedent is true but the consequent is false. The key observations that I want to make are these: (i) Jacqueline's anger, under many natural specifications of the case, accurately represents an evaluative contingency, and (ii) such responses can be the product of wellfunctioning dispositions to anger. This then indicates that in such cases, her anger is geared to provide knowledge of evaluative contingencies; for if anger ever provides any knowledge at all, it is precisely our well-functioning, accurate evaluative representations that we would expect to play the foundational role in generating knowledge within a sentimental perceptualist theory. 21 While I focused on a case involving a particular emotion, similar examples can easily be constructed for other emotions as well as desires.
In what follows, I develop a sentimental perceptualist explanation of armchair evaluative knowledge of contingent evaluative truths. 22 Before doing so, however, I want to head off some potential misunderstandings. First, I am not claiming that any value epistemologist should allow for knowledge of evaluative contingencies from the armchair. My point only concerns sentimentalist views, and is most notably silent about alternatives that ground evaluative knowledge in the intellect. In fact, I believe that the most attractive rationalist (whether intellectual perceptualist or otherwise) explanations of armchair evaluative knowledge will embed their value epistemology in a general account of synthetic a priori knowledge, one which has the result that synthetic a priori knowledge of any sort is of necessary truths. Second, I am not suggesting sentimental perceptualists should deny that there are metaphysically necessary evaluative truths. The claim is simply that they ought to allow that we can gain knowledge of (at least some) evaluative contingencies from the armchair, and anyone who believes that there are any evaluative truths at all should believe that some such truths are contingent.

The Content Challenge
How should sentimental perceptualists answer the armchair challenge? In this section, I propose an answer on their behalf. I do so by turning temporarily to what I call the content challenge. I argue that one of the most promising ways for addressing the content challenge provides sentimental perceptualists all they need to overcome the armchair challenge.
The content challenge is that of explaining how affective experiences ever came to be about value, if value is not simply explained by or reducible to the affective experiences that represent it. Karl Schafer raises this issue for sentimental perceptualist models of desire, in particular. He argues that 'Perhaps the most serious problem for any account that accepts [desires have evaluative content] is explaining how and why desires involve perceptions with normative or evaluative content' [15] (p. 268). Unfortunately, as Schafer points out, proponents of the view have hardly even acknowledged the problem. Mark Schroeder raises the same issue, focusing specifically on Sergio Tenenbaum's brand of sentimental perceptualism [42], labeled in what follows as the scholastic view. Schroeder asks, '[I]f the good is independent of desire, as the scholastic view claims . . . then how does it get to be about it?' [14] (p. 127). He then helpfully expands on his question: "Compare the question of how your greenish perceptual experiences get to be about green, rather than about some other thing, such as orange or square. There is great disagreement in the theory of content determination about just how this happens. But there is widespread agreement that it doesn't happen simply by magic. Your perceptual state has to somehow latch on to green." [14] (pp. [127][128] In what follows, I explain a tempting way for sentimental perceptualists to overcome the content challenge. I want to emphasize that the answer to the content challenge is not meant to be conclusive but rather cautiously optimistic. This cautious optimism thus extends to the proposed solution to the armchair challenge.

How to Answer the Content Challenge
A natural assumption I make is that sentimental perceptualists should stick close to the perceptual analogy, modeling their story about content determination on that of ordinary perception. However, here we face an immediate roadblock: theories of perceptual content determination are manifold. Still, from a bird's-eye perspective, they tend to fall into one of two broad families (see [43,44]). One grouping of views appeals to some sort of naturalistic relation between a given experience and what the experience is hypothesized to be about. This relation is standardly taken to have informational/covariational, causal, and/or teleological dimensions. Following Uriah Kriegel, I call these tracking theories [43] (p. 167) (see also [45] (pp. 33-55 ff.)). Another approach to content appeals to the phenomenal character of the experience in question. For example, a 'square-ish' perceptual experience counts as representing the property of being square by virtue of its phenomenology, or what it is like. Efforts to explain content by appealing to phenomenology are often called phenomenal intentionality theories [45]. 23 In what follows, I center tracking theories of content. My reasons are dialectical. 24 For one, assuming a tracking approach in addressing the content challenge is the most helpful with the armchair challenge. Another key is that, if perceptual phenomenology explains (or is identical to) perceptual content, it is not unreasonable to think that such phenomenology is undergirded by naturalistic relations of the sort identified by tracking theories. Consider an example from Kriegel: even if my perceptual representation of an apple is explained by the phenomenal character of the experience, as a phenomenal intentionalist would say, there is still plausibly a tracking relation between that type of experience and apples. This everyday co-extensiveness between tracking and phenomenal character is why it helps to avert to science fiction-style thought experiments to identify cases in which the two approaches come apart. 25 For example, a phenomenal intentionalist may find it intuitive that an envatted brain can perceptually represent apples even without tracking anything. 26 Despite such possibilities, however, the important point for our purposes is this: even if we altogether reject tracking theories of representation, the presence of relevant naturalistic tracking relations between an experience and some hypothesized content would be circumstantial evidence in favor of the experience having that content. 27 A related dialectical point goes for another prominent approach to perceptual content. According to Tyler Burge, perceptual representation is not reducible. However, Burge does maintain that tracking relations are present where representation is. Consider the following remarks: "Organisms like bacteria, amoebae, paramecia, worms, molluscs, clams are differentially sensitive to various attributes in the physical environment. They discriminate those attributes. Their sensory capacities carry information. They function to respond in certain ways, given this information . . . These sensory capacities are not perceptual." [47] (p. 315) He later continues: "Where there is perception, there is sensory information registration. That is, where there is perception, there is functional, causally based, usually high statistical correlation, between a type of state impacted by surface stimulation (and that encodes surface stimulation), on one hand, and a type of stimulation, on the other." [47] (p. 317) For Burge, what makes perceptual states special is that they have veridicality conditions, while functional, causal, and correlational relations do not. Nevertheless, the emergence of perceptual states with veridicality conditions depends on the presence of these other kinds of relations (e.g., for fixing reference).
In what follows, I assume that perceptual representation involves some combination of informational, causal, and/or teleological dimensions. I suggest that by adopting a simple, straightforward understanding of each of these notions, we are led to a prima facie attractive theory incorporating all three. Furthermore, for reasons just observed, advocates of alternative theories of content can, and I think should, raise their confidence that an experience of a given type has some hypothesized content, if the sort of tracking relationship I describe holds between the experience and that putative content.
As a starting point, many theorists have thought the notion of information important for perceptual content. 28 When a mental state carries information about something, this intuitively helps us to understand, at least to an extent, how the mental state could come to be about that something. For example, that perceptual representations of objects as square carry information about squareness seems explanatory of how such representations came to be about squareness. What exactly is the information relation, though? A traditional way of cashing out what it is for one thing (mental state or otherwise) to carry information about another is in terms of covariation. 29 One thing, A, carries information about another thing, B, just in case A statistically co-varies, even weakly, with B. More concretely, smoke carries information about fire, because where there is smoke, there is (often) fire. Sentimental perceptualists would do well to accept an information requirement for perceptual content. Although not universally accepted, sentimental perceptualists should not want their response to the content challenge to turn on the hope that information is not important. 30 Furthermore, the absence of an information relation would make sentimental perceptualism a non-starter for explaining evaluative knowledge, never mind the content or armchair challenges.
However, it is widely agreed that we need more than a (simple) information relation to explain perceptual content. The reason for this is that perception is about a much narrower range of things than perception carries information about: perceivers do not perceptually represent everything their perceptions covary with. One ingredient that is natural to add, although it is not enough to complete the picture on its own (or so I would argue), is causation. Causal and covariational requirements can be joined together in an attractive way. To see how, consider that covariational relations often have an underlying explanation. In many cases, the underlying explanation references a causal relation between the two things (objects, properties, etc.) that covary. Thus, a popular view says that perceptions stand in causal-tracking relations with what they are about. That causal-tracking relations would be important for perception is very intuitive; it helps us to see how a mental state latches on to external objects and properties.
Unfortunately, causation and covariation do not seem sufficient. I will once again use vision to illustrate. Visual perceptual experiences are about distal (environmental) particulars, properties, and relations; however, such representations are immediately triggered by proximal stimulations, most notably light arrays of varying intensity that are registered on the retina. Visual perceptions of a certain type covary with certain proximal stimulations and are also caused by them. Yet, a perceiver does not represent the proximal stimulation; what gets represented (at least in the case of vision) is almost always something about how the world is external to the perceiver. So, if we want to understand why an observer represents, say, solidity, rather than the light arrays on the retina that cause the representation of solidity, we need more than causation and covariation.
One natural way to fill the gap is to build a teleological component. Why, we might ask, is a perceptual state about solidity rather than light arrays on the retina? It's because the perception, or underlying perceptual system, has the biological function of detecting solidity but does not have the function of detecting light arrays on the retina (or, if we think that perceptions do have the biological function of detecting such proximal stimulations, then, following Karen Neander, we can add a qualification: our perceptual systems have the function of producing such perceptions ('solidity perceptions') in response to certain light arrays because that is a means of producing the perceptions in response to what is solid and not vice versa [49]). The functional relation helps secure the latch between a perceptual state and what it is about. According to the account I work with, a perceptual state's content gets fixed by what it has the biological function of standing in causal-covariational relations to. I suggest, then, that sentimental perceptualists can give a reasonable answer to their content challenge by arguing that (some sufficient range of) affect has the biological function of standing in causal-covariational relations with value. 31 By taking on board a fairly demanding tracking theory-one that incorporates biological function, covariation, and causation-the sentimental perceptualist puts themselves in a good dialectical position for answering the content challenge, assuming they can argue that (some) types of affect stand in such relations with value. First, it may be that some weaker tracking account turns out to be correct. Karen Neander, for instance, thinks that biological function and causation are enough; strictly speaking, we do not need covariation. If she is right, then sentimental perceptualists following my strategy will have defended a bit more than is necessary to answer the content challenge, but will still have answered the challenge. That said, it may be that the most promising tracking account is not entailed by the theory with which I am working. For example, Christopher Peacocke defends what he calls the action-answerability account. This approach says that '[T]he correct attribution of representational content to perceptual states is constitutively answerable to the range of properties of actual and counterfactually possible actions of their subject that those perceptual states are capable of explaining (in combination with other states)' [54] (pp. 477-478). 32 This is a kind of covariation approach, but what I say above does not directly speak to it, since I say nothing about action explanations. That said, Peacocke would agree (or so it seems to me) that if desires or emotions have the biological function of tracking value, then that would be evidence that the value figures into the content of the relevant experience. 33 So, my suggestion for theorists who favor alternative tracking theories not entailed by the one that I work with is the same as my earlier suggestion for those who reject tracking theories altogether: the presence of the sort of tracking I have described between affect and value should still raise our confidence that affect is about value.

Causation and Covariation
Is it plausible to maintain that affect has the biological function of standing in causalcovariational relations with value? Focus first on causation and covariation, setting aside biological functions for the moment. Graham Oddie has shown that there is a plausible case to be made that affective experiences can be caused by values [6] (pp. 181-210), though he does not make this case in the service of overcoming the content challenge. 34 Furthermore, his argument, if successful, secures a covariation relation, too. It will be helpful to describe Oddie's view in some detail, then. In what follows, I focus on desires, but similar arguments could be marshaled for emotions. 35 Oddie argues that commensurateness is a useful guide to causation. 36 Commensurateness is a matter of adequacy and dependency [6] (pp. 191-195). 37 I will start with the former. The notion of adequacy captures the idea that a cause is sufficient in the circumstances for its effect. Suppose a father walks into a clothing store with the intention to buy his son a shirt. He ultimately opts for the magenta-colored shirt. The magenta color of the shirt is, let us suppose, adequate for his deciding to buy it. Being magenta is a determinate of the determinable, being colored; however, the shirt's being colored is not a cause of his buying it because it is not in the relevant sense adequate. This is indicated by the fact that his father would not buy his son a green, yellow, or blue shirt, because he knows his (rather picky) son does not like those colors. So, being colored is not enough, but being magenta is. Turn to dependency. Dependency captures the idea that causes are necessary in the circumstances for their effects. It turns out that the shirt's magenta color does not cause the father to buy it because he knows his son's favorite color is purple and would thus have bought him any purple-colored shirt (build into the case that there were numerous non-magenta purple shirts in the store). Oddie insists that adequacy and dependency together make up the idea that causes are commensurate with their effects, i.e., that they are in the circumstances necessary and sufficient. 38 In this case, it is the purple color of the shirt, rather than, say, the shirt's being colored (not adequate) or magenta (not dependent) that seems commensurate.
Values are often commensurate with desires, or so Oddie argues. To illustrate, consider a case (similar to certain ones Oddie himself discusses) in which a hiker, Felicity, stumbles upon a fellow hiker suffering from what appears to be a severe injury. Felicity immediately forms a desire to help them. Could it be that her desire is caused by a value that it represents? To consider this question, we need to have in mind a working hypothesis regarding what value desires are about. Oddie's own view is that desires are appearances of goodness. However, since I have argued elsewhere that desires are oriented to the desirer's normative reasons, I focus on this nearby alternative [58]. 39 Felicity's desire to help, on this picture, involves an appearance (perhaps inter alia) of the other hiker's severe injury as a reason to help them. The question then, is whether the desire might be caused by this relation of being a reason. It is important to be careful here. The issue is not whether the desire could be caused by the fact that the hiker is severely injured; it is rather about whether it could be caused by this consideration's favoring helping them (see [59] (p. 38 ff.)).
It is quite plausible that Felicity is sensitive to reasons to help in a way that is indicative of causation. To see why, consider the highly specific way in which the hiker is injured: a severe gash caused by falling over a rock-filled embankment. This specific injury is adequate to the desire. That is, being injured in that way is sufficient in the circumstances for Felicity's desire to help. However, had the hiker been injured in some slightly different way, the desire would still have arisen. For example, they might have a severely twisted ankle. We can even imagine more substantial variations. Perhaps they passed out from dehydration or are suffering from a venomous snakebite. It is not even clear that these are injuries, in an ordinary sense of the term. Furthermore, if they were injured in certain ways (e.g., if their arm were in a cast, or they had a mild cut on their arm), she likely would not have desired to help. In general, then, the way in which her desire patterns in nearby possible worlds indicates that it is dependent on a normative favoring relation (i.e., her reason to help). 40 It is important not to be distracted by the fact that the desire could have arisen even if there were not really a reason to help. 41 For instance, if the other hiker were playing a strange prank whereby they smeared themselves in mud and red paint so as to appear in dire straits, then Felicity would have desired to help (assuming that she did not immediately see through the ruse). In this case, the relation of being a reason to help would not have been present to serve as a cause. However, such an observation no more undermines the view that values can be causes of our affective experiences than do similar observations about ordinary perceptual experiences. For example, a perceptual experience as of a rock could also have arisen if, instead of a rock, there were a holographic image of one. Such distant possibilities do not threaten the claim that rocks can be, and indeed often are, part of the cause of perceptual experiences as of rocks. In the case at hand, the relevant counterfactual is this: had there been no reason to help the other hiker, the desire would not have arisen. This depends on what nearby worlds are like. 42

Biological Function
Even if it is reasonable to maintain that values cause and covary with affective experiences, we still need to know whether it is also reasonable to maintain that affect has the biological function of standing in causal-covariational relations with value (or, as I will often say, detecting value). Given that there is no standard picture about the biological function(s) of affect (and which forms of affect even have such a function), there can be no conclusive argument here. Nevertheless, I offer four reasonable ideas that collectively make it plausible that some affect has the function of detecting value.
The first reasonable claim is that a wide range of affective experiences were selected for. Psychologists standardly refer to this class of affect as basic, and sets of purportedly basic affect often include, for example, joy, parental concern, fear, lust, disgust, and anger. 43 Basic types of affect 'have a fixed set of neural and bodily expressed components, and a fixed feeling or motivational component that has been selected for through longstanding interactions with ecologically valid stimuli' [64] (p. 398). 44 Of course, the key question is why they were selected for. Were they selected for (inter alia) because they help us to detect value of some sort?
Some philosophers balk at the idea that affect (or any mental state) could have such a function. A principal worry expressed by Simon Blackburn is that there is an apparently competing hypothesis about the function of affect that is supposedly more scientifically respectable [72] (p. 363). The basic thought is that it is far more plausible that affect was selected for because of how it caused us to behave rather than because it allowed us to detect some evaluative property or relation. However, this brings us to the second reasonable claim, namely, the hypothesis that affect has the function of detecting value is not in competition with, and may be complementary to, affect's function to get us to behave in certain ways (cf. [73] (p. 268)). The easiest way to see this is just to consider an example from ordinary perception. 45 The ability of many animals to detect predators is almost surely an adaption, but this ability would not have conferred any advantage whatsoever were it not intimately linked to action. 46 That is, the ability to perceptually detect predators would not have been selected for unless it tended to trigger fight or flight responses (or at least not unless it were linked up to action in some evolutionary advantageous way). So, there are at least two functions of a predator-response system: (i) to produce a perceptual state, S, in response to inputs that indicate the presence of a predator, and (ii) to motivate fight or flight when S occurs. The sentimental perceptualist insists that matters are much the same with affect and value. We have value-response systems with at least two important kinds of functions: (i) to produce an affective state, A, in response to inputs that indicate the presence of some value, V, and (ii) to motivate action of the relevant sort when A occurs. 47 Would it promote our reproductive success to be able to detect values, though? Some truths, after all, would confer little advantage to be able to track, e.g., truths about lowfrequency electromagnetic wavelengths, to borrow an example from Street [75] (p. 130). By contrast, being able to detect many other kinds of truths, such as ones about the presence of predators, fires, conspecifics, etc., would confer significant evolutionary advantage. The third reasonable claim, then, is that if values enter into causal relations, then in many cases being able to detect them would promote reproductive success. 48 Just consider some schematic examples which illustrate the kinds of things we typically appeal to values in order to explain: (i) the unreasonable verdict led to the riot, (ii) the good deed brought the community together, and (iii) bad parenting is at the root of their depression. Such examples could easily be multiplied and fleshed out (one might also recall the example of Felicity above: the reasons to help the other hiker led her to desire to help them). This point should not come as a surprise in light of Street [75]. After all, she observes that it is suspicious how close of a connection there often is between what we believe to be right or wrong and what promotes our reproductive success. That may be suspicious given a denial of the causal efficacy of value. However, sentimental perceptualists who accept the causal efficacy of values can attempt to turn this point to their advantage, using it as grounds to raise our credence that we have evolved to detect value. 49 Thus far, we have identified three reasonable claims: (i) a wide range of affective experiences were selected for, (ii) the hypothesis that affect has the function of detecting value is not incompatible with, and appears even complementary to, affect's function to get us to behave in certain ways, and (iii) being able to detect values (e.g., normative reasons) would often promote reproductive success. These three claims perhaps explain why many scientists talk (albeit loosely) about affect as having the function of detecting value. 50 These observations may be enough to make it reasonable for sentimental perceptualists to maintain that affect has the function of detecting value, at least as a working hypothesis. 51 Now, before moving on, consider one last idea.
The fourth reasonable claim is that affect has what we might call an 'evaluative phenomenology'. Such phenomenology is evidenced by the way in which it seems, at least to many theorists working on affect, that describing what affective experiences are like requires evaluative language. Consider the following: "In experiencing indignation at the harsh punishment of the toddler, it seems to you that the punishment is in fact unjust: your occurrent emotional state puts forward your indignation's content as correct. This is in analogy to the content of a sense perception. In perceiving that the cat is on the mat, it seems to you that the cat is actually there." [21] (p. 377) "Very roughly, the phenomenology of feeling sad about something, which involves negative affect, experientially represents that that something is of disvalue." [82] (p. 46) "The person in an emotional state, like the person describing what he sees, largely describes the part of the world that, in her emotion, she is attending to. She says, "The dirty bastard took me for all I had" or "This is a wonderful day!" or "He's always gazing at other women with that famished look" . . . Just as the person who is reporting on his visual impressions most often tells you what he sees, the person reporting on his emotions most often tells you how the world appears to him." [7] (pp. [72][73] These phenomenological observations might be taken as a direct response to the content challenge, if phenomenal character were itself explanatory (or identical to) representational content. 52 However, even once we set phenomenal intentionality theories aside, the phenomenology of affect should lead us to raise our credence that some affect has the biological function of causally covarying with value. To see why, consider a sensory experience appropriately described as having a 'square-ish' or 'predator-ish' phenomenology. Such experiences plausibly arise from perceptual systems which have the function of producing them in response to squares or predators, respectively (cf. [47] (pp. 315-317 ff.)). In the case of affective experiences which were selected for, then, we would expect matters to be much the same. To be sure, we might try to avoid this result by denying one of the other claims already proposed. For example, if values cannot be causes, then affective systems could not have evolved to generate affective experiences in causal response to values. However, once we combine the evaluative phenomenology of affect with other key theses, namely that values can be causes, that (some) affective experiences were selected for, that detecting values would be evolutionary advantageous, it becomes at least a reasonable hypothesis that the affective experiences in question have the function of detecting value. 53

Extending the Story to Offline Affect and Answering the Armchair Challenge
The story about content sketched above extends in straightforward fashion to offline affective experiences. Or at least it does, given the following natural assumption: the types of affective experience (desires and emotions) that occur in response to perceptions and beliefs are of the same type as those which occur in response to imaginings. 54 According to this picture, there are not really two affective systems giving rise to two different kinds of affective experience; there are just single sets of dispositions that trigger in response to value-indicating contents represented both online and offline. It is not difficult to see why it would be to our advantage for affect to work in this way, assuming the above story about online affect is correct. It would help us to plan for the future, by signaling what would be valuable. That is, we would be able (at least in many cases) to learn in advance what would be best, rather than having to wait for the future to come to us (cf. [24,25]).
A key idea is that a core set of the dispositions to affectively represent values in the imagination are native rather than learned. A standard way of thinking about innateness, and the one I have in mind, is that innate characteristics are organized in advance of experience. Innate characteristics need not be universal or unmalleable; the idea is that genetic structure supplies, as Gary Marcus puts it, a first draft [86]. 55 We are prewired in certain ways, though it may be that experience edits, or alters, that prewiring. It is often taken for granted that some 'valuations' or 'proto-evaluative judgments' must be innate. Here is how Cosmides, Tooby, and Barrett make the point: here must be an irreducible core set of initial, evolved, architecture-derived content-specific valuation assignment procedures, or the system could not get started. The debate cannot sensibly be over the necessary existence of this core set. The real debate is over how large the core set must be, and what the proper computational description of these valuation procedures and their associated motivational circuitry is." [88] (p. 317) Psychologists as diverse as Laurence Kohlberg and Jonathan Haidt have taken this innate structure to be wholly, or at least in large part, affective. 56 This theory about the origins of (online and offline) affective evaluative content provides the resources for an attractive answer to the armchair challenge. Remember, the difficulty is to explain how there can be non-accidental connections between offline evaluative experiences and corresponding evaluative truths. This explanation should, moreover, respect the intuition that we can in many cases obtain evaluative knowledge by mere reflection. The additional constraint for sentimental perceptualists (as opposed to, say, intellectual perceptualists) is to make sense of how this knowledge can be of evaluative contingencies (see Section 3). Meeting this constraint is going to require information from the world to somehow shape the dispositions giving rise to offline experiences. I have sketched a non-mysterious story about how this shaping works. Affective dispositions have evolved to respond to sets of conditions that tend, in the actual world, to indicate the presence of value. Since we are prewired to respond affectively to certain contents, whether represented online or offline, we need not always rely on online experience to learn evaluative truths. Furthermore, because well-functioning affective experiences do not typically respond to contents (as given in perceptions, beliefs, imaginings, and so on) that necessitate the presence of a corresponding value, the propositions that they help us to know will typically not be necessary.
There is more to be said with respect to the propositions about which offline affect can help us to learn. In my view, sentimental perceptualists should say that the specifics depend on the mental states which trigger the experience. For instance, returning to an example from above (Section 3), if Jacqueline's experience of anger is triggered by an imagining of her favorite athlete, Sydney, uncaringly stomping on her toe, it is natural to suppose that her evaluative experience helps her to learn the following: 'if Sydney were to uncaringly stomp on my toe, then he would wrong me'. By contrast, if Jacqueline experiences anger in response to the general thought of people uncaringly stomping on other people's toes, then perhaps she is positioned to learn a general contingent truth: 'normally, people uncaringly stomping on other people's toes is wrong'. This is contingent since there are (e.g.,) worlds in which stomping on toes is not painful and is instead a form of greeting (cf. [61] (pp. 222-223)). Sometimes these general truths may turn out to be necessary (e.g., an aversive response to the thought of torturing others just for fun). Still, our affective systems are tailored for this world, and stumbling upon direct knowledge of evaluative truths which are necessary is largely going to be a matter of happenstance.
Readers may notice that I have referred to the proposition learned on the basis of offline affect rather than the proposition represented (or, alternatively, the proposition corresponding to what is represented). This is because I intend to leave open whether these are the same. To illustrate, consider whether Jacqueline's anger really represents that 'if Sydney were to uncaringly stomp on my toe, then he would wrong me'. An affirmative answer has a pleasing consequence: the propositions we learn from the armchair match the contents of the relevant affective experiences. However, despite expressing sympathy for this position in other work [62], I am now skeptical. 57 Part of the difficulty arises because, prior to the affective experience, the imagining does not represent anything about what would be. Instead, the initial imagining of Sydney's uncaring stomping presumably represents the non-existence, or perhaps the possibility, of something. 58 So, the sentimental perceptualist following this path needs a story about how the emotion combines with the imagining to transform the content of her experience into one of what would be the case (cf. [62] (pp. 3092-3093)).
An alternative strategy, one which I tentatively favor, has it that well-functioning offline affective experiences typically involve a subtle illusion. On this approach, Jacqueline's offline anger represents that something is wrong, just as online anger does. Nevertheless, from her perspective, this need not lead to any particular confusion. This is because, in normal cases, people can immediately and effortlessly discern whether or not they are imagining [89] (pp. 187-189). So, instead of believing anything is wrong, Jacqueline, and other similarly situated agents, treat their affective experience as signaling what would be valuable. An intriguing feature of this view is that it hews closer to the perceptual analogy, perhaps even allowing a sentimental perceptualist to maintain that emotions (and desires) are literally perceptual. For just as genuine perceptual experiences seem to involve a representation of certain objects and properties as real, so too does affect [84]. 59 There are potential concerns with this alternative strategy, however. One is that it may seem to make offline affect inevitably unfitting if fittingness is a matter of accurate representation. To be sure, many philosophers of emotion do think of fittingness as accuracy. 60 However, it is arguably more natural for sentimental perceptualists to think of affective fittingness in other terms. For at least at first glance, it seems that sensory perceptual experiences can be both fitting and inaccurate if they arise from well-functioning sensory dispositions. For example, it is in some intuitive sense fitting to experience the lines in the Müller-Lyer illusion as being of different lengths, even if one knows that they are not. If this is right, offline affective experiences can likewise be fitting, at least in one sense, despite being inaccurate, insofar as they arise from well-functioning affective dispositions (i.e., ones that are sufficiently sensitive to contents which indicate the presence of corresponding values). 61 Another potential downside of this approach is that it may have the result that evaluative knowledge from the armchair is never strictly a priori. This is because agents would need to have an at least implicit understanding that affective experiences in response to imaginings should be treated as a signal of what would be the case evaluatively, even if this is not exactly what they represent; this would seem to require a prior grasp of the interface between affect, imagination, and reality. I am skeptical that there is any real cost here, though. Even if the sentimental perceptualist's model of armchair evaluative reflection disallows for knowledge that counts as officially a priori, since it requires the agent to have some highly general a posteriori knowledge, it is not clear that this violates any important pretheoretical datum.
Whichever approach we take, the key idea is this: our affective systems are suited for this world. It may be that our imagination leads us to learn a necessary truth, if we imagine content sufficient to necessitate the presence of the relevant value in all possible worlds. However, this would be to 'stumble into' knowledge of a necessary evaluative truth.
A major payoff of the strategy is that it allows sentimental perceptualists to adopt a causal explanation of how we obtain evaluative knowledge from the armchair. Even though there are not causal relations between offline affective experiences and any instantiated values, causal relations between affect and value still play a role in explaining how we evolved to have affective dispositions that can track value offline. The drawback of this strategy, however, is that it depends on empirical assumptions about the nature and origins of affect. All I can do here is motivate some optimism that the assumptions are true. Ultimately, we should keep in mind an obvious point: value epistemology is difficult and rival theories, too, have major costs. For example, rationalist theories of reflective evaluative knowledge (e.g., [12]) typically maintain that we can acquire synthetic a priori knowledge of necessary evaluative truths through reason alone, and the details of such theories are notoriously controversial (e.g., [33]). The advantages and disadvantages of the sentimental perceptualist model proposed here deserve close evaluation alongside more familiar alternatives. 62

Conclusions
Perceptualists aim to demystify evaluative knowledge by analogizing it with perception. Yet, as I pointed out, the analogy seems to come up short when we turn to reflective evaluative knowledge. 63 This challenge is especially acute for sentimental perceptualists, since affective experiences will often ground knowledge of contingent evaluative truths if they ground any such knowledge. How could affect ground knowledge of contingencies from the armchair? I argued for the following solution. Sentimental perceptualists should answer the armchair challenge by way of the content challenge, an important difficulty that they need to address anyway. If they solve the latter in the way that I have proposed, then they also receive a solution to the former. Additionally, as it turned out, there is evidence motivating cautious optimism that the content challenge can be solved. Ultimately, the proposed answer to the armchair challenge is attractive insofar as it preserves a kind of causal story of evaluative knowledge without giving up the pretheoretical idea that we can learn evaluative truths by mere reflection. Acknowledgments: For feedback on this work, I am especially grateful to Stephen Finlay, Alex Gregory, Janet Levin, Mark Schroeder, and Ralph Wedgwood. I also benefited from conversations with Errol Lord, Hichem Naar, and Karl Schafer as well as audiences at the University of Southern California, the University of Pennsylvania, and the Hebrew University of Jerusalem. I would also like to thank the students in my Winter 2023 graduate seminar at Toronto Metropolitan University. Our explorations of recent philosophy of mind were crucial to helping me to revamp and update this paper. I also thank the anonymous reviewers for numerous helpful comments and suggestions.

Conflicts of Interest:
The author declares no conflict of interest.

1
Sentimental perceptualism became especially prominent in 17th-18th century Great Britain, due especially to the influence of the Third Earl of Shaftesbury [1] and Francis Hutcheson [2]. The view was further developed in the 19th and early 20th centuries by Franz Brentano [3] and Alexius Meinong [4]. For contemporary proponents, see, among others, Mark Johnston [5], Graham Oddie [6], Robert Roberts [7], and Christine Tappolet [8]. Notably, some theorists prefer to think of desires and emotions as responses to perceptual-like feelings of value rather than as perceptual-like themselves [9,10]. Theorists drawn to this model should feel free to translate my arguments accordingly.
Perceptualists may allow for many different kinds of value experience, though considerations of simplicity and unity require caution in doing so [17]. 6 Not all rationalists are perceptualists, however. Some (e.g., [18]) argue that there's no need to appeal to intellectual experiences to explain evaluative knowledge. Instead, evaluative knowledge is grounded in non-experiential understanding of the evaluative propositions that we know. For discussion of this non-perceptualist view, and how it compares to the perceptualist alternative, see [13]. It is also worth noting that Audi has more recently assigned a prominent role to evaluative experiences of various sorts [19]. 7 There is a distinction between those who maintain that affective experiences have evaluative content and those who still treat such experiences as evaluative but locate the evaluation outside the content (e.g., [15,22]). Although proponents of the latter view agree that affect involves an evaluation, their picture is not compatible with sentimental perceptualism as defined here. Still, proponents of this cousin of sentimental perceptualism face the armchair challenge, if they want their theory to be provide a basis for evaluative knowledge, and so they may (in principle) find aspects of the solution I offer congenial. 8 Interestingly, the example of the politician indicates that evaluative inquiry is often an impure mix of armchair and non-armchair inquiry. Suppose that I have a strong reason to vote for the politician and that it would be good if the politician were president. Since the reason is instantiated, a recognition of it would not qualify as armchair inquiry. Learning about this reason might help me to determine whether it would be good if the politician were president. Such impure cases may be less challenging for sentimental perceptualists insofar as the agent has some perceptual-like interaction with a part of the relevant evaluative landscape. In any case, the view that I propose below does not exploit this potential advantage and works even for 'pure' cases. 9 My aim here is to identify a familiar assumption about empirical inquiry that helps us to see the force of the armchair challenge; the asymmetry is in no way essential to it. For important complications about drawing too sharp an asymmetry, see (e.g.,) [24,25]. 10 I have in mind imaginings that are analogues of perception as well as of belief; see [26] (p. 734). 11 I do not mean to indicate that either McGrath or myself maintains that online experiences are never privileged over offline ones. For example, Alison Jaggar's discussion of 'outlaw emotions' offers one illustration of why lived experience should sometimes take epistemic priority [28]. I briefly flag how my view might begin to accommodate this in n62, though I leave a detailed exploration of this matter to another occasion. See also [27] for important and helpful discussion. 12 Such a perceptualist can still maintain that we can acquire analytic evaluative knowledge by mere reflection. Although it often goes without saying, perceptualist theories are meant to explain substantive knowledge [29]. 13 I do not take myself to have said anything definitive against an empiricist model according to which our online evaluative experiences are always privileged over our offline experiences. However, in my view, this should be a fallback position in the event that an adequate solution to the armchair challenge cannot be found. 14 It has been put to me in conversation that one might hold that, while evaluative reflection does not put us into causal contact with any instantiated evaluative property, it does put us into contact with, say, goodness, wrongness, etc. as such. For many existing sentimental perceptualist theories (e.g., [6]), such a view would generate a peculiar asymmetry between armchair and non-armchair evaluative inquiry-since no sentimental perceptualists believe that it is wrongness, goodness, etc. as such that cause our online affective responses. Moreover, I know of no sentimental perceptualists who have ever tried to defend such an asymmetrical theory (though admittedly they are largely silent about armchair inquiry in general). The sentimental perceptualist theory that I propose maintains as close of a symmetry between armchair and non-armchair inquiry as I think can be expected for those who take the armchair challenge seriously. 15 For criticism of the Lewisian idea, see [32]. For a detailed critical discussion of the appeal to adequate understanding (as well as additional references), see [33]. 16 See [36,37] on why our affective experiences would have evolved to respond to conditions that indicate, but often do not (metaphysically) necessitate, the presence of a relevant evaluative property. 17 It is important to be clear that my point here is not that it is epistemically possible from Jacqueline's perspective that she is wrong (though that may be true). My point is that the full content of Jacqueline's anger, which includes evaluative content combined with content supplied by other mental states (cf. [22] on the notion of a "cognitive base"), is only contingently true. One reason for this is that there are metaphysically possible worlds in which Sydney uncaringly stomps on her toe but does not wrong her; in those worlds, it will not be true that Sydney wrongs her by uncaringly stomping on her toe. Thanks to an anonymous reviewer for pushing me to clarify this point. 18 I am not talking about predictions of how one would feel if something happened. For a sentimental perceptualist, such predictions are at best a highly indirect route to evaluative knowledge. 19 Philosophers of mind debate about whether offline affective responses to imaginings and fictions are the same mental state types as their online counterparts [38,39]. I believe my arguments can be adapted to relevant plausible ways in which philosophers have argued that they are not, but I will not be able to explore these matters in depth here. 20 For more detail on what sorts of propositions we might come to know through imaginative reflection, see Section 4.4. 21 One could insist that an affective response only grounds evaluative knowledge if it is responsive to conditions that necessitate the presence of the relevant evaluative property. I suspect that such a picture will lead to the result that, on a sentimental perceptualist theory, we almost never have evaluative knowledge; in any case, I think that this approach sets the bar on evaluative knowledge too high (see [35]). However, the explanation of armchair knowledge that I offer below could be paired with the idea that affective experiences only ground knowledge if they are responsive to conditions that guarantee the presence of the target evaluative property. 22 Kripke famously argues that we can have contingent a priori knowledge that the standard meter stick is one meter long [40], but this does not provide a useful model for sentimental perceptualists. For one, the knowledge in Kripke's case is grounded in our understanding of 'meter', and sentimental perceptualists are not trying to explain semantic evaluative knowledge. For a more detailed discussion about why we should be skeptical about building a value epistemology (cf. [41] (pp. 161-162)). 23 I do not mean to suggest that these two approaches are exhaustive. Most notably, I set aside what are sometimes called "short-arm" and "long-arm" functional role theories [45,46]. Versions of the former do not include the external environment as part of the role and generally appeal to inferential roles, specifically. Such views strike me as better suited for explaining the contents of beliefs rather than perceptions. I am moreover persuaded by Mendelovici that these theories struggle more obviously than others in explaining how contents arise in the first place, even if they can explain how we get more complex contents from already existing contents [45] (pp. [72][73][74][75][76]. One might (controversially) think that long-arm functionalism can overcome this difficulty, but the most natural way to do so is by appealing to the kinds of relations involved in tracking theories [45] (pp. [76][77][78][79]. What I say below will thus be relevant to theorists drawn to this approach. 24 My own leanings are increasingly in the direction of phenomenal intentionality as explaining original (underived) intentionality. See especially the considerations outlined in [45] (pp. 79-80, 90-93). 25 Kriegel makes this point with respect to what he calls objective and subjective representation, a conceptual contrast that I set aside for the purposes of this paper [43] (p. 162). As Kriegel notes, tracking theories appear suited for targeting objective representation while phenomenal intentionality theories appear suited for targeting subjective representation. 26 A phenomenal intentionalist may prefer to say that representation is more precisely of 'the manifest kind property of being "apple-y" (or perhaps that of playing the apple role)' [43] (p. 163). 27 Mendelovici [45] argues that certain perceptual experiences (e.g., color experiences) always misrepresent and so fail to track what they are about. While Mendelovici's view merits careful study, I cannot carry out that project here. I assume that our everyday perceptual experiences are not subject to widespread, systematic error. 28 See, for instance [48][49][50]. 29 See Burge [47]. Neander defines information in terms of causation rather than covariation [49]. Information in her sense is part of my account, too. 30 Millikan [51] and Papineau [52] deny that information is necessary. I should note, however, that what they deny, strictly speaking, is that information features in the analysis of perceptual representation. However, that is compatible with information still playing an evidential role. 31 For views roughly along these lines, see [47,50,53]. As noted above, however, Burge appeals to bio-functions not in order to reduce content, as some others do, but merely to explain how perceptual experiences end up with certain content. Sentimental perceptualists can do something similar. 32 On this approach, the explanation for why an animal represents the shape and color of some fruit appeals to the fact that the animal's behavior is counterfactually sensitive to the presence of such fruit; all else equal, it eats all and only fruit of that shape and color. 33 Peacocke asks, 'Is biological function not merely evidentially or epistemically relevant to the determination of content, but constitutively involved as a matter of the very nature of perceptual representation itself?' [54] (p. 477). The assumption-which seems to me correct-is that biological function is at least evidential. 34 A major forerunner to Oddie's argument is Sturgeon [55]. 35 Prinz [56] offers a helpful starting point for similar arguments centering emotions. 36 I do not mean to commit to any analysis of causation. It is sufficient for my purposes that commensurateness functions as a useful test for causal relations. 37 Oddie is adapting some ideas from Yablo [57] on mental causation. Oddie and Yablo actually use 'contingent' rather than 'dependent'. I tweak their language in order to avoid confusion, since the way in which I use 'contingent' elsewhere in this paper is different. 38 Following Oddie, I have used counterfactuals to characterize commensurateness. This is a heuristic. The best way to understand commensurateness (as well as causation) is presumably in non-counterfactual terms [6] (p. 195). 39 The view that desires are oriented to reasons is also defended in [59,60]. 40 A reader might wonder why Oddie's opponent cannot just construct a disjunction of possible specific ways the natural world could have been so that the disjunction includes all and only the specific ways the world could have been sufficient to produce the desire. The claim would then be that the desire is dependent on this disjunctive property. Oddie is aware of this objection, and his reply is that the grotesque disjunction would not really be a property and so cannot be a cause. This is a reasonable response to the objection, I believe. However, note that even if it is not, it does not immediately follow that values (e.g., reasons) cannot be causes if the long disjunction is just another way of picking out the reason. 41 One might also worry that affect cannot be caused by values because it is always a response to another mental state (see [61]). See [62] for a response to this line of argument. 42 One may worry that the relation of being a reason is not really adequate to the desire, since we can imagine versions of the case in which the hiker's situation favors helping but not in a way that leads to the desire. Strictly speaking, we should speak of reasons within a certain range. However, this is not a surprising qualification. An object's squareness can cause a visual representation as of its being square, but of course, strictly speaking, what we normally mean is squareness within a certain range (the square object needs to be within a certain size range, for example). 43 According to Prinz, some emotions which are blends of basic emotions were selected for, too [53] (p. 76). For example, contempt, which he takes to be a blend of anger and disgust, may be such a case. In fact, some evolutionary psychologists maintain that all emotion types are suspeptible to evolutionary explanation [63] (pp. [45][46][47]. 44 This passage is quoted from Jessica Tracy and Daniel Randles' overview of the contributions to a special issue of Emotion Review on basic emotions [64]. As they observe, each contributor accepts this way of thinking about basic affect and also agrees that such affect exists [65][66][67][68]. For some philosophers who accept a similar thesis, see [37,53,69]. By contrast, social constructivists deny that emotions are rooted in such affect programs (e.g., [70]). However, this would not entail that more primitive, pre-emotional forms of affect are not rooted in the sorts of affect programs others see as the biological basis of emotions (cf. [5]). In any case, social constructivist approaches to emotion face what appear to me to be severe difficulties [71]. 45 Artiga [74] makes this point in a reply to Street [75]. The reply is really a very simple one. The core of the response is simply to point out the typical kinds of descriptions which are offered by psychologists, cognitive neuroscientists, and biologists for how perceptual capacities evolve (e.g., Comer and Leung's description of how cockroaches came to sense predators [76]). 46 On the claim that the ability to perceive predators is a common kind of adaption, see the evidence compiled in Burge and Artiga [47,74]. Both Burge and Artiga also emphasize the important connections between perception and action. 47 The appropriate sort of action will vary depending on the nature of V. 48 Street's main target is non-naturalist moral realism, which includes commitments to moral value's irreducibility and causal inefficacy [75] (p. 111). She only briefly considers the extent to which her arguments extend to a view on which value properties are causally efficacious. Thus nothing that I say here responds to the core of Street's widely cited paper. 49 The basic point that value properties and relations seem to be significant from an evolutionary perspective has been observed by a number of philosophers [74,77,78]. 50 See [79,80]. For philosophers who use similar language, see [56,81]. Many psychologists believe that emotions involve appraisals, which are defined as representations of how something or other bears on well-being, paradigmatically the well-being of the appraiser. In the case of basic emotions, the appraisals have the function of responding to some class of considerations that bear on well-being. 51 For more in-depth discussions of how particular emotions may have evolved to detect certain dimensions of value, see Prinz [56]. His proposals about fear, in particular, are especially illustrative and congenial to the approach outlined here [56] (pp. [60][61][62][63][64]. 52 For doubts about these phenomenological claims, at least concerning emotion, see [22,83]. See [84] for more detailed discussion of this dispute about the phenomenal character of affect (and emotion, in particular). 53 Of course, an opponent might predict that the science will not turn out as the sentimental perceptualist hopes. One complication is that showing this to be the case would seem to require making certain normative assumptions about what is or is not valuable (see [85]). 54 This assumption is natural in light of the way in which we ordinarily talk about offline affect. We do not have two different classes of affect terms, for example. However, even if offline analogues of online affect are a subtly different attitude, it does not follow that the present project is undermined. It would depend on the details of why they are treated as different attitudes. See also n19 above. 55 Marcus' understanding of innateness is also adopted by Haidt and Joseph [87], among others. 56 Haidt and Joseph compare their own view to that of Kohlberg (as well as Piaget before him) [87] (p. 374). 57 In earlier work with Hichem Naar, we briefly raise and explore the intersection of (what I refer to here) as the content and armchair challenges [62] (pp. 3090-3094). The overarching aim of that paper was to address a different problem for sentimental perceptualism (roughly, about how affect could offer perceptual-like access to value insofar as affect responds to other mental states). The present paper takes on the armchair and content challenges in focused and systematic fashion; furthermore, while much of what I say is in the spirit of the earlier remarks, some shifts (as indicated above) have emerged in my thinking. 58 For a detailed exploration of how imagination represents its content, see [89] (pp. 184-194). 59 On the idea of perceptual experiences representing their contents as real, see [89]. 60 See Naar [90] for detailed references. Naar ultimately thinks that treating fittingness as accuracy is misguided. I also note that I have myself proposed that sentimental perceptualists should think of fittingness as accuracy, which is why I previously rejected the sort of suggestion favored here about the content of affect in response to imaginings [62] (p. 3091). 61 I defend this view in greater length elsewhere [91]. 62 The proposal here may allow for a principled explanation of why lived experience can sometimes have epistemic priority over detached evaluative reflection (cf. [28,92]). In the former case, there is the possibility of changing affective tendencies through causal interaction with value. For reasons of space, I am unable to further explore this line of thought here. See also n11 above. 63 A reader might think that we can get some basic empirical knowledge from the armchair (though in that case, 'empirical' could be somewhat misleading). I am sympathetic to this possibility, although I do not consider it here (cf. [24,25]).