Experience and Evidence ( forthcoming in Mind ) SUSANNA SCHELLENBERG

away from its details in order to reach a more general rationale for thinking that perceptual content is determined by the perceiver’s environment in the right way. In the last section, I argued that perceptual capacities function to single out particulars. If the fact that such capacities single out particulars in some situations but not in others has any semantic significance, then the content ensuing from employing these capacities will depend at least in part on the environment in which they are employed. After all, relations to particulars are implicated in the very nature of perceptual content, if perceptual content is determined by employing perceptual capacities and such capacities function to single out particulars. In so far as perceptual capacities function to single out particulars, perceptual experience is fundamentally both relational and representational. On the proposed view, employing perceptual capacities yields representational content. More specifically, employing perceptual capacities determines a content type that subjectively indistinguishable experiences have in common. However, the token content of an experience co-varies with the environment of the experiencing subject. Since the perceptual capacities employed are the very same in subjectively indistinguishable experiences, the content type will be the same. Individuating experiences by a content type amounts to individuating experiences with regard to the experiencing subject’s sensory state. While subjectively indistinguishable experiences are individuated by the same content type, their token content differs to the extent that the environment of the experiencing subject differs. How should we understand these token contents? I argued that employing perceptual capacities determines our sensory states and that by means of employing such capacities we (purport to) single out particulars in our environment. This idea is analogous to the Fregean idea that modes of presentation both have a cognitive significance and are a way of referring to particulars. Corresponding to Frege’s use of modes of presentation as accounting for both these roles, there are two standard ways of understanding modes of 28 One might reject this view by arguing that perceptual evidence merely supervenes on content, such that there could be changes in content without changes in evidence. Any such view would have to account for why there is a difference between perceptual states and perceptual evidence and so would be more complicated than the view suggested. So reasons of parsimony will count against such a view. For a defense of the view that epistemic reasons are mental states, see Turri 2009. Susanna Schellenberg Experience and Evidence 23 presentation. If one focuses on the role of modes of presentation as accounting for cognitive significance, then it is natural to think of them as de dicto. A de dicto mode of presentation can be the very same regardless of what (if anything) the experiencing subject is related to. If one focuses on the role of modes of presentation as a way of referring to a particular, then it is natural to think of them as de re. A de re mode of presentation is relational in that what particular (if any) the subject is related to has repercussions for the token content. This way of thinking about the content of experience recognizes that the mental act of representing a particular is not independent of singling out the particular that is the referent of the sense. Building on this idea, we can distinguish between mode of presentation types that are determined by employing capacities (for instance concepts), on the one hand, and mode of presentation tokens that are determined by employing capacities in a particular environment, on the other. More specifically, the token content of a perception of object o1 that instantiates property P1 can be expressed in the following way: (contentp) where MOPr(o1) is an object-related de re mode of presentation of o1 and MOPr(P1) is a propertyrelated de re mode of presentation of an instance of P1. So the content of any two subjectively indistinguishable perceptions e1 and e1* in which we are perceptually related to the same object o1 in the same way will include the token de re mode of presentation MOPr(o1), where MOPr(o1) ensues from employing a perceptual capacity and being perceptually related to o1. A hallucination that is subjectively indistinguishable from e1 is a matter of employing the same perceptual capacity, but the environmental requirements for successfully singling out a particular are not met. So unwittingly no particular is singled out. As a consequence, the capacity employed remains baseless in the sense that there is no external, mind-independent particular to serve as the target of discrimination and selection. It is key that the failure is not on the level of employing the relevant capacity. The failure is on the level of singling out a particular. Since there is no failure on the level of employing the capacity, there is no reason to think that the mental state of hallucination does not have a token content. After all, the very same perceptual capacities are employed that determine the content of a subjectively indistinguishable perception. Employing perceptual capacities gives sufficient structure for the relevant experience to have a token content and moreover accounts for the fact that we purport to single out a particular. However, since we fail to single out a particular, the token content is defective. One way of understanding the idea that the content is defective is to say that it is gappy. The gap marks the failure to single out a particular. So the content of a hallucination in which we purport to single out an object that instantiates a property is: (contenth) Susanna Schellenberg Experience and Evidence 24 where MOPr(__) in the object-place is a gappy, object-related de re mode of presentation and MOPr(__) in the property place is gappy, property-related de re mode of presentation. For a perceptual capacity to be baseless amounts to the ensuing token content being gappy. Analogously, an illusion is a matter of employing the same capacity that in a subjectively indistinguishable perception is used to single out a property-instance, but since the subject fails to be perceptually related to the relevant property-instance, the capacity employed remains baseless. The token content that ensues from employing that perceptual capacity is gappy since the subject fails to single out a property-instance. So the token content of an illusion includes a non-gappy object place and a gappy property place. It is important to note that the content types are not general contents, but rather potentially particularized token contents. To motivate this, consider that if I have a thought as of a white cup, but there is no white cup present, I fail to refer. In such a case, the content of my thought is not singular. After all, I failed to refer. However, it is not a general content either. After all, I purport to refer to a particular object. So the content has the form of a singular content while failing to be a token singular content. In short, content can have the form of a singular content while failing to be a token singular content. This does not imply that the content is general. There are more options than that, namely being a potentially singular content. As in the case of a failed singular thought, the content of hallucination is not a general content. The content is structured by two levels: the content type and the token content. More specifically, a potentially particularized content type ands a defective or gappy token content. One might object that the content of a hallucination and the content of a perception could never be tokens of the same type. After all, the former is gappy and the latter is not. In response, tokens of the same type can differ significantly. For them to be tokens of the same type they need only to exhibit the feature relevant to be classified under that type. There are many ways to type contents. One way is with regard to whether or not they are gappy. On this criterion, gappy contents and non-gappy contents would be tokens of different types. However, another way to type contents is with regard to the perceptual capacities employed. On the suggested capacity view, employing perceptual capacities in a sensory mode yields a content type that subjectively indistinguishable experiences have in common. The token contents of this type can differ significantly: in the case of hallucination, they are gappy; in the case of perception, they are environment-dependent. Despite these differences, they have features in common, namely the 29 It is important to distinguish this view from Burge’s view. Burge has been read as defending a gappy content view. However, as Burge notes of his view ‘I have heard interpretations ... according to which there is a ‘hole’ in the representational aspects of the proposition, where the hole corresponds to the object (which completes the proposition). I regard these interpretations as rather silly’ (1977/2007, p. 75). Burge argues that there are demonstrative elements in the content of experience that are in place regardless of whether they refer to the object of experience. As he puts it ‘I do not think that a physical re in the empirical world ... is itself ‘part of’ the belief. ... In my view, the Intentional side of a belief is its only side. In many cases, in my view, a belief that is in fact de re might not have been successfully referential (could have failed to be de re) and still would have remained the same belief. Moreover, the belief itself can always be individuated, or completely characterized, in terms of the Intentional content’ (1991, p. 209). The way I am using the terms, what Burge refers to as de re would be more aptly labelled de dicto. Susanna Schellenberg Experience and Evidence 25 perceptual capacities employed in the same sensory mode. In virtue of these common features, they are tokens of the same type. Now while in the good case the perceiver is perceptually related to the particulars that it seems to her she is perceiving, in the bad case she is not so related. Nevertheless, the gappy token content is inherently relational in virtue of being determined by employing perceptual capacities that are asymmetrically dependent on their employment in the good case. The perceptual capacities that determine the content of experience are inherently related to external and mind-independent particulars of the type that they function to single out. What follows from this for perceptual evidence? The view of perceptual content that I have presented provides us with two ways of individuating perceptual experiences. Each perceptual experience can be individuated with regard to the content type that is determined by the capacities employed in perceptual experience. Alternatively, each perceptual experience can be individuated with regard to the environment-dependent token content that ensues from employing perceptual capacities in a particular environment. This account of perceptual content not only puts the notions of phenomenal and factive evidence on firmer footing, but also integrates them into a unified view of perceptual evidence. Phenomenal evidence is determined by the content type of a perceptual experience. Factive evidence is determined by the environment-dependent token content of a perceptual experience. So the factive evidential basis changes as the token content changes—even if one cannot tell. In this sense, factive evidence provides the perceiver with evidence that goes beyond mere phenomenal evidence. So the distinction between phenomenal and factive evidence emerges from two levels of perceptual content. In the next section, I will show how this bilateral view of evidence grounds the internal and external dimensions of perceptual evidence in the perceptual capacities that we employ in perceptual experience. I have presented a way of giving support to the idea that perceptual content is determined by the perceiver’s environment in the right way. In order to reach a more general rationale for this idea, I will now abstract away from the details of the proposal. A more general rationale for the idea is motivated by how best to think of the accuracy conditions of experience. The accuracy conditions of perceptual experience specify the way the world would have to be for the content of experience to be accurate. Given this constraint there are several different ways of understanding accuracy conditions. If the content of experience lays down a condition under which it is accurate in a way that is sensitive to which particulars (if any) are perceived, then the way the experiencing subject’s environment is will make a difference to the content of her experience. The motivation for this way of understanding accuracy 30 For an elaboration of the advantages of this understanding of perceptual content over competing disjunctivist views, see my 2010 and 2011b. 31 For a detailed discussion of the nature of gappy token contents, see my 2010. 32 For an alternative context-sensitive view of factive evidence, see Neta 2003. Susanna Schellenberg Experience and Evidence 26 conditions is that the condition that needs to be met for an experience to be accurate is not just that there is an object in the world that instantiates the properties specified by the content. It is necessary to specify which particular object in a subject’s environment is represented to determine whether the subject’s environment is as it is represented to be. If this is right, then for an experience with the content ‘that coffee cup is white’ to be accurate it is not sufficient that ‘that’ refers to some coffee cup instantiating the right properties. It is necessary that ‘that’ refer to the particular object perceived. If the content of experience lays down the conditions under which the experience is accurate and the accuracy of an experience depends on the environment, then the particulars to which the subject is perceptually related will make a constitutive difference to the token content of her experience. Such a relational view of perceptual content can be contrasted with a non-relational view. On a non-relational view, the content connects with the particulars it is about only in virtue of that particular satisfying the condition laid down by the content. The relation between content and particulars is the semantic relation of satisfaction. The condition to be satisfied does not depend on the particular that satisfies it. This condition may amount to a description such that the particulars that the content is about are those that uniquely fit the description. On the proposed relational view, the content is itself dependent on the particulars it is about. So the content does not remain constant whatever the environment of the experiencing subject. If she were in a different environment and so would single out different particulars or none at all, the content of her experience would be different. 3.3 Coda In order to fully support the idea that experience provides us with more evidence in the good than in the bad case, we need to show that we do not have factive evidence in the bad case. It falls out of the argument for why we have factive evidence in the good case that we do not have factive evidence in the bad case. On the suggested view, the token content of an accurate perception is a singular content, but the token content of hallucination is gappy. The gappy token content of hallucination does not provide evidence, since a gappy content cannot be true. After all, it is defective and so either does not have a truth-value or is necessarily false. It is not rational to heed something that by its very nature could not guide one to the truth. Therefore, a gappy content does not provide evidence. A different way of expressing the underlying idea is that systematic linkage to how things are in the good case is what makes a mental state rational to heed. A gappy token content is never systematically linked to external and mind-independent particulars that it is of in the good case, since in the good case, a mental state will never have a gappy token content. So due to its defectiveness, a gappy content cannot yield factive evidence. More generally, we can say that an environment-dependent content is singular in the case of an accurate perception, but fails to be singular in the case of a hallucination. I argued that only if the content 33 A non-relational view need not, however, amount to a descriptive view. The distinction between relational and non-relational views of content overlaps but does not coincide with the distinction between non-descriptive and descriptive views of reference. Susanna Schellenberg Experience and Evidence 27 of our experience is determined by our environment in the right way, could it yield factive evidence. If we hallucinate, the content of our experience is not determined by our environment in the right way and so cannot yield factive evidence. Now, accepting that perceptual experience provides us with factive evidence in the good case requires denying that our evidence is always accessed or even accessible to us. At the very least, this requires denying that we can always recognize what the content of our experience is by introspection alone. In order to see why, consider the following switching case. At time t1, we perceive a white coffee cup (cup1) on a desk. At time t2, the cup (cup1) is, unbeknownst to us, replaced by a qualitatively identical cup (cup2). Since the two cups are qualitatively identical, it is implausible that we have access to whether we are perceptually related to cup1 or cup2 at t2. But if perceptual content is environmentdependent, then the content of our experience will be different before and after cup1 is replaced with cup2. So accepting that perceptual content is environment-dependent requires accepting that we are not always in a position to have access to all aspects of our perceptual content. While it is necessary to deny that we have access to all aspects of our perceptual content if we accept that perceptual states are factive, there are reasons to deny this for non-factive mental states. In order to see why, consider the following case. We perceive three subtly distinct shades of red: red1, red2, and red3. We cannot perceptually tell the difference between red1 and red2. We cannot perceptually tell the difference between red2 and red3. Yet we can perceptually tell the difference between red1 and red3. In order to analyse this case, it is plausible that there is a difference regarding our phenomenal evidence when we perceive red1 and red2, despite the fact that we cannot distinguish between the two shades of red. Similarly, it is plausible that there is a difference regarding our phenomenal evidence when we perceive red2 and red3. An explanation for how we can distinguish between red1 and red3 draws on the premiss that there is a subjectively indiscernible difference between our phenomenal evidence when we perceive red1 and red2, and there is a difference between our phenomenal evidence when we perceive red2 and red3. The case suggests that there are aspects of our phenomenal evidence to which we do not have access and which moreover are not accessible to us. Given that there are reasons to reject the thesis that we have access to all aspects of our perceptual content for non-factive mental states, we need not be troubled that we must reject it, if we accept that experience provides us with factive evidence. 34 For classical discussions of the problem of knowing or grasping the content of one’s mental states, if that content is externally individuated, see Brueckner 1986, McKinsey 1991, Warfield 1998, and Brown 2004. The idea that we are always in a position to access our evidence has been famously criticized by Williamson (2000). 35 Some internalists have understood the accessibility of evidence as an essential part of the very nature of evidence. Indeed, it has been argued that denying the accessibility of evidence or even our accessing our evidence amounts to changing the subject (Cohen 1984, p. 284). It will lead too far astray to address this issue here. 36 For a more general discussion of the limits of introspection and knowledge of one’s mental states, see Pereboom 1994, Goldberg 2000, and Fumerton 2009. Susanna Schellenberg Experience and Evidence 28 4. The Common Rational Source of Phenomenal and Factive Evidence I have argued that in virtue of its sensory character perceptual experience provides us with phenomenal evidence, and that an accurate perception provides us with additional factive evidence. Phenomenal and factive evidence both have their rational source in the perceptual capacities employed in experience. Phenomenal evidence is determined by the content type of an experience that is in turn determined by the perceptual capacities employed. Factive evidence is determined by the token content of an experience that ensues from employing these capacities in a particular environment. In so far as both kinds of evidence have the same rational source, this capacity view of perceptual evidence shows how the internal and external aspects of perceptual evidence are fundamentally connected in the perceptual capacities that we employ in perceptual experience. As I argued in section 2, it is rational to heed the testimony provided by sensory states, which are determined by employing perceptual capacities, since such states are systematically linked to external and mind-independent particulars of the type that they are of in the good case. After all, if a subject’s environment sensorily seems to contain F particulars, then she is in a sensory state that is systematically linked to external and mind-independent F particulars. She is systematically linked to such particulars in the sense that the perceptual capacities employed in the bad case are explanatorily and metaphysically parasitic on their employment in the good case. If she is in a sensory state that is systematically linked to external and mind-independent F particulars, then she is in a sensory state that provides evidence for the presence of F particulars. After all, if a subject is in a sensory state that is determined by employing perceptual capacities that function to single out F particulars, then the subject is in a sensory state that provides evidence for the presence of F particulars. As I argued in section 3, an accurate perception provides us moreover with factive evidence. The analysis of the epistemic role of phenomenal evidence in virtue of a notion of systematic linkage carries over to an analysis of the epistemic role of factive evidence. After all, in the case of a perception, there is an ideal link between our perceptual state and the environment due to our being perceptually related to our environment. Therefore, we have additional factive evidence in virtue of accurately representing our environment. Factive evidence provides additional evidence that is different from phenomenal evidence. It is evidence of a different kind since the systematic linkage to the environment is stronger than the one governing phenomenal evidence. More specifically, it is evidence of a different kind since it is provided by successfully employing perceptual capacities in a particular environment. So factive evidence provides a rationality boost beyond the rationality boost that one already has from phenomenal evidence. Thus it is explained why the perceiver is in a better epistemic position than the hallucinator. Now from the first person perspective one may not be able to tell the difference between a hallucination, in which one has only phenomenal evidence, and a perception in which one has both phenomenal and factive evidence. However, we need not think that what is accessible from the first person perspective dictates what is rational to heed. A sensory state is rational to heed in virtue of being determined by employing Susanna Schellenberg Experience and Evidence 29 perceptual capacities that function to single out the external mind-independent particulars that the state is of in the good case. There is no need to have access to all aspects of that state. It is worth pausing to consider how the content type and the content token are expressed in natural language. The content type and a singular token content can be articulated in the very same way in natural language. They better be. After all, a perception and a hallucination can be subjectively indistinguishable. Consider an experience of a white cup. The content type and the singular token content can both be articulated with ‘that cup is white’. However, the demonstrative ‘that’ will play a different logical role in the two cases. In the content type, the demonstrative will refer to whatever (if anything) happens to be available to be singled out. By contrast in the singular content, the demonstrative will refer to the very thing that the perceiver is perceptually related to. So unbeknownst to the experiencing subject, the two contents will play different roles in inferences and so have different evidential force. This capacity view of perceptual evidence has several attractive features. First, it is an externalist view of evidence that makes room for hallucinations providing us with evidence without retreating to introspective evidence, a general content, or an existentially quantified content. The view is externalist in so far as the content of factive evidence is an environment-dependent token content and in so far as our phenomenal evidence is determined by our sensory states, which in turn are individuated externally. Sensory states are individuated externally since they are determined by employing perceptual capacities that are by their very nature linked to the particulars that they are of in the good case. While the content of factive evidence is an environment-dependent token content, the content of phenomenal evidence is a content type. No doubt we can articulate a general content or an existentially quantified content to express the content of our sensory states. But the fact that we can articulate such content does not imply that the content of phenomenal evidence is such a general content or an existentially quantified content. It is a potentially particularized content type. Second, the capacity view implies that we can have perceptual evidence, only if we are in a sensory state. Arguably, it is a condition of adequacy for a view of the epistemological role of perceptual experience that we have perceptual evidence only if we are in a sensory state. I argued that employing perceptual capacities yields phenomenal evidence and, if the environment plays along, also factive evidence. So we can have phenomenal evidence without having factive evidence. However, since we have factive perceptual evidence in virtue of employing perceptual capacities and since employing such capacities yields a sensory state, we cannot have factive perceptual evidence without being in a sensory state. If this is right, then monotonicity between factive and phenomenal evidence is guaranteed. Third, the capacity view provides for a way of explaining why it is that a perceiver is in a better epistemic position than a hallucinatory. Consider again Percy who perceives a white cup on a desk and 37 Thanks to Adam Pautz for pressing me on this point. 38 For a discussion of the role of sensory awareness in perceptual justification, see Bergman 2006, Glüer 2009, Lyons 2009, Hellie 2011, Silins 2011, Johnston 2011, Smithies 2011, and Siegel and Silins forthcoming. Susanna Schellenberg Experience and Evidence 30 Hallie who suffers a subjectively indistinguishable hallucination. On the suggested capacity view, Hallie has some evidence for her belief that there is a white cup on the desk, but Percy has more evidence for his belief. More generally, we can say that a person who is perceiving is in a better epistemic position than a person who is hallucinating, since the perceiver has more evidence, where that evidence is a distinct kind of evidence, namely factive evidence. So for any subject S1 and any subject S2, if S1 has all the evidence that S2 has plus an additional bit of evidence that is factive and thereby of a distinct kind, then S1 is in a better epistemic position than S2. Now, an alternative way of arguing that Percy has more evidence than Hallie is to show that there are two distinct facts that can figure as the truthmakers of perceptual content: facts about the experience and facts about the environment in which one is experiencing. After all, we have evidence that consists of true propositions when we are hallucinating, namely, introspective evidence of how it seems to us that the environment is. Such an approach restricts the evidence we get when we experience to factive evidence, however the factive evidence includes not just perceptual evidence, but also introspective evidence. So the evidence we have in perceptual experience is either factive with regard to our environment or with regard to our experience. Williamson defends a version of this view. According to Williamson, evidence is a known proposition and knowledge is a mental state. Evidence is the object of the mental state, namely, a proposition or a set of propositions. Since evidence is a known proposition, there is no room on Williamson’s view for evidence provided directly through experience in the bad case. After all, in the bad case there are no true propositions provided directly through experience. On the Williamsonian view, we have only introspective evidence in the bad case, that is, known propositions about how things seem to us. On both the Williamsonian view and the one I have argued for, perceiving Percy has more evidence than hallucinating Hallie. On the Williamsonian view, Percy has factive perceptual evidence and factive introspective evidence, while Hallie has only factive introspective evidence. On the capacity view, Percy has phenomenal and factive perceptual evidence, while Hallie has only phenomenal perceptual evidence. While the Williamsonian view also accounts for the idea that Hallie has some evidence, but not as much as Percy, it is less attractive than the capacity view for three reasons. First, the Williamsonian view requires positing that we do not get evidence directly through our experience when we hallucinate, but only through introspection. However arguably, experience provides us with evidence directly—even when we hallucinate. The notion of phenomenal evidence that I have developed makes room for experience providing us with phenomenal evidence directly even in the bad case without retreating to introspective evidence. Second, introspection is a sophisticated intellectual activity, yet even subjects who do not have sophisticated intellectual abilities can get evidence through hallucination. By relying on subjects attending to how things seem to them the Williamsonian view over-intellectualizes the way we get Susanna Schellenberg Experience and Evidence 31 evidence in the bad case. A distinct and more pressing over-intellectualization worry is that on the Williamsonian view, the evidence we have in the bad case is an appearance proposition. Appearance propositions involve appearance concepts and some sort of self-reference. However, non-rational animals hallucinate and presumably, they gain evidence in virtue of hallucinating even though they are not capable of being in mental states that are constituted by appearance propositions. The capacity view does not face these over-intellectualization problems, since we have phenomenal evidence in the bad case in virtue of being in a sensory state: there is no need to introspect or attend to our experience to have phenomenal evidence. On the view developed, we can have phenomenal evidence, even if we have no ability to refer to ourselves and do not possess appearance concepts. Finally, a view on which we get evidence only through introspection in the bad case, but directly through perceptual experience in the good case, requires positing that the source of our evidence differs at least in part in the good and the bad case. By contrast the capacity view shows that the source of both factive and phenomenal evidence is our perceptual experience. Indeed, the capacity view provides for a unified account of perceptual evidence by revealing the common rational source of the evidence one has in perception and the evidence one has in a subjectively indistinguishable hallucination. So while I am following Williamson in arguing that we have a kind of evidence in the good case that we do not have the bad case, contra Williamson I am not rejecting the phenomenal conception of evidence. Moreover, the notion of evidence in play is not understood as identified with knowledge. Against Williamson, I have argued that we should not and need not retreat to the idea that experience provides us only with introspective evidence in the bad case. Doing so would undermine the epistemic force of experience. The capacity view makes room for an externalist account of the epistemic role of perceptual experience that does not depend on and does not entail reliabilism (Goldman 1979). One might argue that it is in virtue of perceptual capacities being reliable that the sensory states they determine provide us with evidence. On the defended view, sensory states provide us with evidence since sensory states are systematically linked to the particulars that they are of in the good case. I argued that sensory states are systematically linked to what they are of in the good case in the sense that the perceptual capacities employed in the bad case are explanatorily and metaphysically parasitic on their employment in the good case. More specifically, I argued that if a subject’s environment sensorily seems to contain F particulars, then she is in a sensory state that is determined by employing perceptual capacities that function to single out F particulars. If a subject is in a sensory state that is determined by employing perceptual capacities that function to single out F particulars, then she is in a sensory state that provides evidence for the presence of F particulars. So the notion of systematic linkage in play is understood in terms of a metaphysical and explanatory primacy notion rather than a reliabilist notion. As I argued in section 2, while a sensory state provides phenomenal evidence in so far as it is determined by perceptual capacities that are metaphysically and explanatorily dependent on the good case, such capacities may more often Susanna Schellenberg Experience and Evidence 32 than not be used in ways that fail to produce accurate representations of the world. So such capacities may fail to be reliable. But while the suggested capacity view does not depend on and does not entail reliabilism, it is compatible with such a view. The thesis that experience yields factive and phenomenal evidence is compatible not only with reliabilism, it is compatible also with the basic commitments of virtue epistemology. However again one can accept the capacity view without accepting the basic commitments of virtue epistemology. The capacity view for instance neither entails nor depends on the thesis that epistemology is a normative discipline. Neither the capacities nor the metaphysical and explanatory primacy notions in play need be understood in terms of virtues or any other normative notion. Indeed, the capacity view shows how the epistemic force of experience is grounded in metaphysical facts about experience.