2. INTERNALIST MOTIVATIONS

There are many straightforwardly externalist views about perceptual knowledge in the literature. Why does Pritchard hold on to reflective access? One reason has to do with an ordinary or commonsensical way of thinking about rational support. Pritchard seems to understand accessibilism as a pre-theoretical or natural view, at least when it comes to rational support (though he does talk about “epistemic standings” generally in this context – see quote immediately below). He takes it as an important constraint on epistemology to avoid revisionism whenever possible. I will leave to one side the issue of whether Pritchard is correct that accessibilism is a pre-theoretical or commonsense epistemological outlook.Footnote ⁸

Another motivation for Pritchard's commitment to reflective accessibility turns on considerations about epistemic responsibility. He says:

Pritchard seems to accept the idea that an agent counts as in some sense responsible for one's epistemic standing – and thus, presumably, is in some sense creditable for it – only if the facts in virtue of which one has that epistemic standing meet an internalist access condition. As we'll see below, the connection between epistemic responsibility and access is complicated. As such, I will save discussion of this point for when the details become crucial below. In any case, according to Pritchard, it is in virtue of an ability to remain non-revisionary and to respect intuitions about epistemic responsibility, while accounting for truth-conduciveness, that ED is an interesting view about paradigm cases of perceptual knowledge.

3. THE HIGHEST COMMON FACTOR ARGUMENT

Aside from these very broad considerations, Pritchard does not present much in the way of positive support for ED. Instead, he presents the view as a commonsensical or natural idea that is normally thought too good to be true, and explains why a variety of arguments against it are no good. The most important of these for our purposes is the “highest common factor” argument.Footnote ⁹ It goes as follows:

The point of this conclusion is that, even in the good case, the best one has by way of rational support for one's perceptual beliefs are facts about the way the world seems to one. Thus, ED's claim that there is such a thing as rational support in the form of reflectively accessible factive mental states is challenged.

According to Pritchard there is no obvious move from P2 to C1. He maintains that there is a difference in one's rational support in the good and bad cases, even though the cases are “phenomenologically indistinguishable”. In other words, Pritchard holds that merely pointing out there is no phenomenological difference between the good and bad cases does not settle the question of whether the rational support one has is the same in each case. Thus, Pritchard urges, the proponent of the highest common factor argument owes us a justification for the move from P2 to C1.

4. EXPLANATORY CHALLENGE

What can the proponent of the highest common factor argument say about the move from P2 to C1? I think the best way to understand the support for the move from P2 to C1 is in terms of an explanatory challenge. The challenge, put simply, is to point out that resisting the move from P2 to C1 generates an explanatory burden. In particular, it generates the burden of explaining how a person's reflectively accessible rational support could be different in phenomenologically indistinguishable good and bad cases. To see the force of the challenge, we need to clarify and emphasize a couple of things.

First, we need to clarify the notion of “phenomenological indistinguishability”. This is a fairly obscure term, and making it more precise tends to lead to controversial commitments. However, one way of understanding what Pritchard has in mind, which is neutral enough for present purposes, is the following:

Second, we need to emphasize what we've already said about the notion of reflectively accessible rational support. In particular, we said that having reflective access to one's rational support entails that one can know by reflection alone that one has the rational support one has. The idea is that this is not an ordinary empirical kind of access. Having reflective access to one's rational support is being in some sort of special position to know that one has that rational support (again, more on this below).

We can now see the force of the challenge. If there is no property of case X or Y such that one can know by reflection alone that X and Y are different, how can X and Y differ with respect to a property (the kind of rational support one has) to which one has reflective access, i.e. a property that one can know obtains by reflection alone? The idea that there can be phenomenologically indistinguishable cases involving different kinds of rational support stands in direct tension with the idea that one has reflective access to one's rational support. So, Pritchard's claim that there is a lacuna in the move from P2 to C1 lands him with the task of resolving this tension. Meanwhile, those who find the move from P2 to C1 natural will likewise be sceptical that ED has a plausible way of doing so. The point is not that the challenge provides conclusive support for the move from P2 to C1. The point is that there is a natural line of thought behind it, contrary to Pritchard's portrayal of the dialectical situation.Footnote ^10, Footnote ¹¹

Now, one way Pritchard might push back is by showing that the tension is merely apparent. To do so, he might appeal to his distinction between “discriminating evidence” and “favouring evidence”. According to Pritchard, discriminating evidence is a kind of evidence we acquire for a belief that p “just by looking”. It is evidence that supports a belief that p over some known to be incompatible proposition q by way of perceptual distinguishability. My bike is green and yours is blue: when I see a blue bike I've got discriminating evidence for thinking it's my bike, not yours. Meanwhile, favouring evidence is something else. Pritchard is not entirely forthcoming on this, but the most independently plausible example of favouring evidence is the kind of evidence we get for a belief that p through reasoning processes like deduction, induction and abduction (Pritchard Reference Pritchard2010). Consider the famous cleverly disguised mule case: there is no phenomenologically distinguishable difference between perceiving a real zebra and perceiving a cleverly disguised mule (at a certain distance); but arguably one has favouring evidence for the belief that one is now looking at a zebra rather than a cleverly disguised mule – one has evidence, for example, in the form of abductive reasoning from the fact that zoos don't generally go in for that sort of thing. Pritchard might argue, then, that the explanatory challenge only seems pressing if we restrict the notion of rational support to the notion of discriminating evidence. If we fail to acknowledge favouring evidence, we lose sight of a crucial way in which there can be a rational difference between indistinguishable cases.

This is a plausible distinction. Surely there is some kind of difference between the sort of support we get for a belief “just by looking” at things (perception), and the kind we get from deductive, inductive and abductive reasoning. But how, more specifically is this relevant to the explanatory challenge? To put it to work, Pritchard needs to claim specifically that the kind of rational support provided by seeing that p in the good case – i.e. reflectively accessible factive reasons – is a kind of favouring evidence as opposed to discriminating evidence. After all, presumably that is what the explanatory challenge overlooks when it demands an explanation of how phenomenologically indistinguishable cases could involve different kinds of reflectively accessible rational support. But my response here is that, while I think the general distinction is a good one, I think it is ad hoc to claim that the kind of rational support provided by seeing that p in the good case is a kind of favouring evidence as opposed to discriminating evidence. This is because I see no reason to think it belongs in the group of independently plausible types of favouring evidence other than that saying it does helps Pritchard defend ED. Indeed, seeing that p seems much more naturally understood as (potentially) providing a kind of discriminating evidence.

The proponent of the highest common factor argument avoids the explanatory challenge by simply denying that there is a difference between the kind of rational support one has in the good and bad cases. But we could avoid the challenge in other ways. For example, perhaps rational support is not the sort of thing to which one has reflective access. The thought, then, would be that there is nothing mysterious or surprising about phenomenologically indistinguishable cases involving different kinds rational support, if rational support is not the sort of thing to which one has reflective access. But of course, Pritchard's commitment to reflective access seems to render this way of avoiding the explanatory challenge unavailable.

In other words, I think Pritchard faces a dilemma in responding to the explanatory challenge: on one hand, he can concede the highest common factor conception of rational support after all. But since denying the highest common factor conception is the whole point of ED, this move is obviously unavailable. On the other hand, he can deny that rational support is the sort of thing to which one has reflective access. But because of his distinctive commitment to the reflective accessibility of our rational support, it's not clear that this move is available either.

5. WEAK ACCESS TO THE RESCUE?

At this point, a further distinction – one between types of “accessibility” – may be relevant. We can talk about the accessibility of our rational support in at least two importantly different ways. Call them “weak” and “strong” access conditions (Smithies Reference Smithies2013). Weak access says that one is necessarily in a position to know that one has rational support e when they in fact have rational support e.Footnote ¹² Strong access says that one is necessarily in a position to know that one has rational support e when they have it and that one is necessarily in a position to know that one does not have rational support e when they do not have that rational support.Footnote ¹³ Pritchard can maintain that ED is only committed to weak access. In the bad case, one is not in a position to know that they do not have a factive reason to believe that p. It is important to note that the denial of strong access does not imply a commitment to the claim that agents in bad cases have no access to any rational support for their beliefs. Rather, it implies they do not have access to the fact that their rational support is not factive.Footnote ¹⁴

With this distinction on hand, Pritchard might respond to the explanatory challenge in the following way. He can explain the difference of rational support in phenomenologically indistinguishable good and bad cases in terms of the fact that the agent can fail to be in a position to know what kind of rational support they in fact have for a given empirical belief when they are in the bad case. This might be a good way of responding to the explanatory challenge. But it raises an important question about Pritchard's commitment to accounting for internalist intuitions in general. We have seen that he motivates the commitment in part by claiming that it is good to avoid revisionism where possible. But we have also seen that he explains the motivation for internalism in terms of the notion of epistemic responsibility (Pritchard Reference Pritchard2012: 2). The idea in the background seems to be that a belief can only count as a paradigm case of perceptual knowledge if it is one for which the agent can “take responsibility”. Pritchard explicitly claims that it is hard to see how an agent is able to take responsibility for the epistemic standing of their beliefs unless the facts in virtue of which those beliefs enjoy their epistemic standing meet an internalist access condition. Now, since the agent in the bad case is only in a position to know that they have non-factive rational support for their belief, but is not in a position to know that they do not have factive rational support, it seems that they cannot take the same degree of responsibility for their beliefs as the agent in the good case. The implication is that the belief is in some sense, in addition to not counting as perceptual knowledge, epistemically worse off than the belief in the good case. A standard way of putting this is to say that the agent's belief is not as justified as the agent's belief in the good case. It is not clear what the relationship between being able to “take responsibility for one's beliefs” and being epistemically justified is. But for present purposes that does not really matter. An upshot of the opacity of one's rational support in the bad case, according to Pritchard himself, is that one's beliefs in the good and bad cases are not equally justified. To put it one way, Pritchard explicitly denies the so-called “New Evil Genius” thesis (NEG), which says that agents in good and bad cases can be equally justified in believing that p (Pritchard Reference Pritchard2012).

As various authors point out (Smithies Reference Smithies2013; Madison Reference Madison2014; Goldberg Reference Goldberg2016; Kelp and Ghijsen Reference Kelp, Ghijsen, Chienkuo, Slote and Sosa2016), this commitment puts serious pressure on Pritchard's ability to reconcile externalist and internalist intuitions. After all, NEG is a key internalist thesis. Thus, the present response to the explanatory challenge takes Pritchard out of the frying pan and into the fire of some familiar objections.Footnote ¹⁵ Indeed, this is the illuminating and powerful feature of the explanatory challenge. While many have pointed out that Pritchard's commitment to the denial of NEG is problematic for his ability to account for internalist intuitions, the explanatory challenge shows us why Pritchard must take on that commitment. Embracing a problematic-seeming level of opacity in our rational support looks very much like the only way to meet the challenge.

That said, some independently interesting and, in my view, inadequately addressed issues remain. They concern Pritchard's appeal to blamelessness. The remainder of this paper takes up some of these issues and connects them to the recently much-discussed distinction between justifications and excuses in epistemology.

6. JUSTIFICATION, BLAMELESSNESS, AND EXCUSES

Pritchard draws a distinction between epistemically justified and epistemically blameless belief to address worries about rejecting the NEG thesis. The appeal to epistemic blamelessness is important not only for Pritchard, but for externalist epistemology generally. Various authors have challenged this general sort of appeal to epistemic blamelessness (Lackey Reference Lackey2007; Gerken Reference Gerken2011; Gibbons Reference Gibbons2013). Madison (Reference Madison2014) directly argues that Pritchard's appeal to epistemic blamelessness cannot do the work he needs it to do.Footnote ¹⁶

In particular, Madison argues that the notion of epistemic blamelessness is not fine-grained enough to distinguish between importantly different cases. For example, it seems that the BIV and a recent stroke victim can both be blameless for forming whatever beliefs they do under their respective circumstances. Moreover, we can imagine a BIV who behaves in epistemically irresponsible ways, such that the beliefs they form have an importantly different epistemic standing to the beliefs of the epistemically responsible BIV. The notion of blamelessness does not seem capable of capturing all this (Madison Reference Madison2014: 68). Perhaps that is one reason it is tempting to endorse NEG (although, more on this last claim below).

This is a standard worry about appeals to blamelessness in epistemology (below I will refer to it as the Madison-style argument). But it is not clear that it stands up to more detailed scrutiny. For example, this objection assumes that epistemic blamelessness is a homogeneous normative category. Is there any reason to think that epistemic blamelessness is a homogeneous normative category? Some recent work on excuses in epistemology suggests not. For example, Timothy Williamson (Reference Williamson, Dorsch and DutantForthcoming) offers a principled account of various types of blamelessness in terms of excuses that agents can have in virtue of complying with different types of dispositional norms (Williamson Reference Williamson, Dorsch and DutantForthcoming). Williamson claims that “excuses are inexhaustibly various; one should not expect a neat taxonomy” (fn. 6). But he also claims to provide a well-motivated account of how various problem cases fit the conditions he lays out for at least some important kinds of excuses. Moreover, Clayton Littlejohn (Reference Littlejohn, Dorsch and DutantForthcoming) has recently defended a Strawsonian “trichotomous scheme” (Strawson Reference Strawson1962). Arguably, the trichotomous scheme makes precisely the distinctions that the Madison-style argument calls for. Let me explain.

In the trichotomous scheme, agents such as the stroke victim are exempt. Meanwhile, agents such as the epistemically responsible BIV have an excuse. These are two different ways in which an agent can be blameless yet unjustified. The really interesting idea in the Strawsonian distinction between exemptions and excuses is that excuses are normative standings that require agents to be reasons-responsive in such a way that they can be held accountable for what they do in general (Littlejohn Reference Littlejohn, Dorsch and DutantForthcoming: 11). When an agent is excused with respect to the violation of some norm, this is in part because they have the relevant capacities needed for being held accountable – it's just that certain other circumstances mitigate blame (for example, non-culpable ignorance, or lack of control). Meanwhile, this is not the case when we exempt someone. Agents exempt with respect to some norm are not reasons-responsive such that they can be held accountable for what they do in general. To put it metaphorically, they are not in the market for accountability, in a global way.

With this in mind, we can formulate the following argument against proponents of NEG. BIVs have systematically false beliefs. But precisely because BIVs meet conditions normally thought to support NEG – i.e. being reasons-responsive in such a way that they can generally be held accountable for what they do – these agents have an excuse as opposed to an exemption for being systematically mistaken. After all, had they not exhibited this kind of rational excellence, it is hard to see why they would have an excuse. Importantly, the point is defensive: just because the BIV is blameless in a way that differs from, say, a stroke victim, this does not imply that the BIV is justified. In fact the BIV satisfies precisely those conditions that are necessary for having an excuse as opposed to a mere exemption.Footnote ¹⁷

Given these developments, it seems that ED has resources for dealing with the explanatory challenge. Pritchard can endorse the weak access condition and account for internalist intuitions in a more nuanced way. Rather than mere blamelessness, Pritchard can attribute a well-motivated notion of excuse to agents in the bad case. The Madison-style argument would no longer present an obstacle. This might be a step in the right direction, but there is good reason to be sceptical.

7. PROBLEMS

I find nothing objectionable in the trichotomous scheme per se. There are compelling reasons to draw the distinction between exemptions and excuses. But the argument so far does not do enough to settle the question of whether agents in the bad case are merely excused as opposed to justified. In order to successfully address worries about denying the NEG thesis, we need more positive support for the idea that BIVs are merely excused for believing what they do.

For starters, the argument contains a merely defensive claim. It claims that just because the BIV is blameless in a way that differs from, say, a stroke victim, this does not imply that the BIV is justified. This seems true, however it is doubtful that proponents of the NEG thesis would disagree. No one thinks that because the BIV's blamelessness differs from that of a stroke victim, the BIV must be justified (i.e. that NEG is true). This is simply not how epistemologists inclined to endorse NEG argue. Rather, I think many would say that, when presented with the cases, they simply have an evaluative reaction: “justified” is an appropriate term for the positive epistemic standing of BIVs. The Madison-style argument is wheeled in after the fact, as a way of challenging an alternative interpretation of the cases – i.e. as cases of mere blamelessness (Madison Reference Madison2014). This is not to say that none of this is controversial (Turri Reference TurriForthcoming). Indeed, many epistemologists challenge the idea that anyone should be allowed to draw conclusions about justification simply by reflecting on their “evaluative reactions” (I will discuss this shortly). The present point, however, is that it is important not to make NEG look founded on worse grounds than it really is.

The argument above also contains a positive theoretical claim. We have identified a necessary condition on excuses (as opposed to exemptions). Again, excuses require agents to be reasons-responsive in such a way that they can be held accountable for what they do in general. And this, in combination with the defensive claim, was supposed to explain why it is plausible to attribute an excuse as opposed to a justification to the BIV. But notice that we have not identified any sufficient conditions on excuses. And on closer inspection, it looks like providing a sufficient condition on excuses is a relevant desideratum in the present context. After all, many will find it plausible that our necessary condition on being excused is also a necessary condition on being justified. For example, Littlejohn seems to agree: “A good test to determine whether the subject has exercised these capacities well [i.e. has been reasons-responsive in the relevant sort of way] is by asking whether the subject could be the same on the inside as someone who acts rightly or believes with justification” (Littlejohn Reference Littlejohn, Dorsch and DutantForthcoming: 15).Footnote ¹⁸ If justifications and excuses have a shared set of necessary conditions, but we do not know of any sufficient conditions on excuses, it is hard to see how the above argument provides enough of a decision procedure when it comes to assigning either of these two normative standings to cases (especially in combination with the problem just outlined for the defensive claim).

Perhaps this is why some epistemologists make a further claim. Littlejohn, for example, rejects the methodological assumption I attributed to proponents of NEG above. That is, he rejects the assumption that we can appeal to evaluative intuitions about BIV cases to draw the conclusion that agents in the cases have justified beliefs (Littlejohn Reference Littlejohn, Dorsch and DutantForthcoming: 4).Footnote ¹⁹ On Littlejohn's and others’ preferred way of putting things, a belief is justified only if it complies with the central norm of belief (Williamson Reference Williamson2000, Reference Williamson, Dorsch and DutantForthcoming; Gibbons Reference Gibbons2013; Srinivasan Reference Srinivasan2015). Meanwhile, sometimes agents have an excuse for violating the norm of belief. Thus, even if justifications and excuses share a necessary condition, justified belief comes with at least one additional necessary condition. A justified belief must comply with the norm of belief. Setting aside debate about what the norm of belief is for a moment, it is clear that the trichotomous scheme does not support a rejection of NEG alone.

If Pritchard wants to go down the route of Littlejohn and others and claim that justified belief requires compliance with the norm of belief, it seems Pritchard faces another dilemma. On one hand, he can endorse an internalist norm of belief. But in that case it seems unclear why the BIV does not meet the additional criteria for justified belief over excused belief. On the other hand, he can endorse an externalist norm of belief (for example, a norm cashed out in terms of knowledge, or factive reasons). That this latter option is an unpalatable one for ED, however, hardly needs explaining at this point. An externalist norm is traditionally a norm with which agents are not necessarily in a position to know they comply. That is just a standard way of putting the notion of “external”. To be sure, as we have seen, Pritchard ostensibly rejects “classical” demarcations of the epistemically internal and external. However, we have also seen that the extent to which this involves holding on to an accessibility condition on justification is limited by the explanatory challenge. Pritchard has already been forced to give up on an important kind of accessibility. The move to “weak access” led to the discussion of Madison-style arguments in the first place. So it is not clear that he has resources for some kind of reimagining of the notion of an “externalist” norm, either.