2. Epistemological disjunctivism of perceptual knowledge

According to Pritchard (Reference Pritchard2012: 13), epistemological disjunctivism of perceptual knowledge (henceforth, ED-P) is formulated as follows:

Clearly, ED-P is a thesis providing a necessary condition for perceptual knowledge in paradigm cases. Two technical terms in the formulation should be elaborated. First, reflective accessibility is a requirement according to which the subject has to be capable of being aware of the rational support underlying her belief. Suppose that S believes that there is a tree in front of her on the basis of her visual experience. The reason (namely, her visual experience) for the belief that there is a tree in front of her is reflectively accessible to her if S is able to be aware of the reason why she holds her belief. Second, factivity requires that the rational support S possesses for her belief that p entails that p. For instance, seeing that p is a kind of factive rational support for a belief that p, for no one can see that p unless p is true. Furthermore, seeing that p differs from seeing an object, in that the former requires that the content of p has to be grasped by the subject, yet the latter does not.Footnote ²

In light of the explanation above, what ED-P maintains is that, in order for S to obtain perceptual knowledge that p (in paradigm cases), S must believe that p in virtue of having rational support R which entails that p and is reflectively accessible to S (i.e., R satisfies both the factivity requirement and the reflective accessibility requirement).

Pritchard further argues that ED-P is a preferable position if it is theoretically available. Particularly, the account is preferable in three aspects. Firstly, ED-P can satisfy both internalist and externalist intuitions in epistemology (Pritchard Reference Pritchard2012: 3–5). For reflective accessibility is an internalist requirement which constrains grounds for our beliefs to be knowledgeable to us, on the one hand; factivity is an externalist requirement according to which truth conducive belief formation is necessarily connected with knowledge, on the other hand.

Secondly, common practices in justifying perceptual beliefs can also support ED-P (Pritchard Reference Pritchard2012: 17). For one thing, a common practice in justifying one's perceptual belief that p is to appeal to her seeing that p. For another, since ED-P proposes that one cannot see that p unless p is true, and that one can know by reflection alone that she sees that p, it can account for these common practices. Consequently, these common practices give us grounds for endorsing ED-P.

Lastly, according to Pritchard (Reference Pritchard2016: 133–4), ED-P is a satisfactory way to cope with underdetermination-based radical scepticism, which may be formulated as follows (Pritchard Reference Pritchard2012: 133–4):

For Pritchard, each statement in this paradox is either intuitive or follows from intuitive claims. (UP1) is a reasonable claim in that we would not seem to be able to distinguish a radical sceptical hypothesis (say, the BIV hypothesis) from a normal case of perceptual knowledge. So, it seems hopeless in this situation for us to possess any rational support that favours our perceptual beliefs over the BIV hypothesis. Also, (UP3) is an intuitive statement that most people will endorse, especially with regard to perceptual knowledge.

Though, once (UP2), as a bridging claim, is considered, (UP1) and (UP3) are in conflict. (UP2) relies on a principle called the underdetermination principle (Pritchard Reference Pritchard2016: 30):

According to Pritchard (Reference Pritchard2016: 30–1), (UP) is plausible for it is hard to see how one can know that p if, provided that she knows the incompatibility of p and q, she fails to have a rational basis that favours p over q. Thus, given (UP2) realizes (UP), together with the assumption that S knows the incompatibility of E and H, S acquires perceptual knowledge E only if S has a rational basis that favours E over H. Accordingly, each of the statements in the underdetermination-based sceptical paradox regarding perceptual knowledge is plausible, but there is an inconsistency between them, for the denial of (UP3) is entailed by (UP1) and (UP2).

Nevertheless, Pritchard (Reference Pritchard2016: 133–4) argues that an advocate of ED-P can simply reject (UP1), at least as it applies to perceptual beliefs. According to ED-P, in paradigm cases of perceptual knowledge, S's rational support for her belief that p is factive, which is not available to her, say, envatted counterpart, S*. As the rational support that S has is factive, but the one S* has is not, S's rational support is thereby better than S*'s. In this sense, S has rational support that does favour the belief that p over the brain in a vat hypothesis. Given that other theories such as epistemic externalism and attributer contextualism face difficulties dealing with underdetermination-based scepticism about perceptual knowledge, Pritchard concludes that ED-P has an edge over other theories for it offers a satisfactory way to cope with the paradox.Footnote ⁴

Given the theoretical advantages of ED-P discussed above, we should adopt ED-P unless there is any plausible argument or reason for rejecting it. In fact, Pritchard (Reference Pritchard2012) addresses three potential problems for ED-P – the access problem, the distinguishability problem, and the basis problem (see Pritchard Reference Pritchard2012: 19–22, 2016: 127–32). A brief overview of these problems will be given below.

Firstly, Pritchard (Reference Pritchard2012: 46) formulates the access problem as follows:

(AP1) is clearly a consequence of ED-P: Only through having rational support which is reflectively accessible for her belief that p can S acquire perceptual knowledge that p. Moreover, it seems natural to accept (AP2), for there is little room for doubt that S can know solely by reflection that she cannot see that p unless p is true. Yet, (AP1) and (AP2), along with a form of the epistemic closure principle, (APC) seems to follow. In other words, it appears that S can know by reflection alone that p while p is an empirical fact. The result is that, at first glance, ED-P seems to imply the odd consequence that empirical facts can be known by reflection alone.

Secondly, the distinguishability problem arises from the assumption that S is not able to distinguish by reflection alone illusory or hallucinatory states from states of seeing that p. However, ED-P maintains that S can know by reflection alone that S sees that p. For given that S can know by reflection alone that S sees that p entails that S is not in an illusory or hallucinatory state, with a form of the epistemic closure principle in play, S is able to know by reflection alone that S is not in an illusory or hallucinatory state. If so, ED-P seems to imply that S is able to distinguish by reflection alone illusory or hallucinatory states from states of seeing that p, which follows that the assumption above is not true. Due to the counterintuitiveness of this result, ED-P therefore has to give an account for why one is capable of doing this distinguishing work.

Lastly, the basis problem stems from a thesis which some philosophers (such as Dretske Reference Dretske1969: 78–39; Williamson Reference Williamson2000: Ch. 1; Cassam Reference Cassam2007: 26–34) endorse:

The rationale behind (SK) is that seeing that p seems to be a species of knowing that p. If so, S's perceptual knowledge that p, which is grounded on seeing that p, will be self-supporting. As it is more or less accepted that self-supporting does not produce rational support, it suggests that seeing that p fails to be rational support for S's belief that p. This consequence is however in conflict with ED-P.

With these three problems in mind, let us briefly examine Pritchard's rejoinders to them.Footnote ⁵ As Pritchard (Reference Pritchard2012: 50–2, 2016: 130) argues, the access problem is not a genuine problem for ED-P, for S's perceptual knowledge is constituted partly by virtue of having rational support which is not acquired by means of reflection alone. In turn, the conclusion of the access problem, (APC), does not follow from the combination of (AP1) and (AP2).

As to the distinguishability problem, Pritchard (Reference Pritchard2012: 96, 2016: 131–2) concedes that we are unable to discriminate between states of seeing that p and states of illusion and hallucination solely on the basis of perception. Yet he suggests that, in paradigm cases of perceptual knowledge, S can still know that S sees that p rather than merely seems to see that p. From the fact that S knows the difference between the two sorts of states mentioned above it does not follow that S knows it by a discriminative ability.

But how is it possible? Take the zebra case discussed in Dretske (Reference Dretske1970) to illustrate (see also Pritchard Reference Pritchard2012: 69). Suppose that S is at the zoo and looks at a zebra in the clearly marked zebra enclosure. It is natural to say that S knows what she sees is a zebra in the cage, although she is unable to distinguish a zebra from a cleverly disguised mule. For her visual experience is not the sole evidence that she has for her belief, but rather her expertise and background knowledge (such as that it is not too expensive to buy a zebra, that it is too risky for the reputation of the zoo to paint the mule exactly like a zebra) also provide her evidence for the belief. Given all pieces of evidence S has, it is not unreasonable to say that S has better evidence for her belief in a normal scenario than the one possessed in the ‘cleverly disguised mule’ scenario. It is this favouring evidence that leads S to have the knowledge that S sees that there is a zebra in the cage. In the same vein, even if S does not have the ability to discriminate between states of seeing that p and states of illusion and hallucination, she still has some expertise and background knowledge which serve as evidence to favour the scenario she is in over the illusory or hallucination one. If Pritchard is right, the distinguishability problem will not be a genuine problem for ED-P as well, in that it does not maintain that S can distinguish between the two incompatible scenarios and thus no explanation is required.

In order to dissolve the basis problem, Pritchard (Reference Pritchard2012: 26–8, 2016: 127–8) argues against (SK), a core thesis of the problem, by indicating that it is possible for S to see that p without believing that p. To be more specific, S wrongly believes that her faculty is malfunctioning due to the presence of a misleading defeater. In this situation, S still sees that p while not believing that p. For it is generally accepted that knowing that p entails believing that p, and yet there are cases where S sees that p without believing that p, seeing that p thereby fails to entail knowing that p. Accordingly, given that (SK) is false, perceptual knowledge that p is not self-supporting in ED-P's framework, and the basis problem is therefore dissolved.Footnote ⁶

All in all, in this subsection, we have explored ED-P. Specifically, we have discussed three theoretical advantages of ED-P, three problems for ED-P, and Pritchard's rejoinders to those problems. Keep these in mind as we consider the case for the parity thesis between ED-P and epistemological disjunctivism of memory knowledge.

3. Epistemological disjunctivism of memory knowledge

To examine the prospects of epistemological disjunctivism of memory knowledge (henceforth, ED-M), we have to of course discuss memory. In psychology and philosophy, memory has been classified in different ways. Psychologists have tended to use declarative and non-declarative memory as a basic distinction in the taxonomy of memory, whereas some philosophers have employed a tripartite division for their taxonomy of memory. Furthermore, Budson and Price (Reference Budson and Price2005) claim that declarative memory is explicit memory (or memory with conscious awareness), while Bernecker (Reference Bernecker2010) identifies whether a memory can be expressed as a distinguishing feature of declarative or non-declarative memory. It should be noted, however, as Michaelian and Sutton (Reference Michaelian and Sutton2017) remark, that the taxonomy from psychology is now becoming standard in philosophy.

Declarative memory has in turn been divided into semantic and episodic. Semantic memory is memory of general information, such as Paris is the capital of France. Semantic memory is a sort of memory of information abstracted from the context of the generation of that memory. It can include memory of information about oneself, such as one's birthday, as long as such a memory does not involve the recollection of the episode in which one learns one's birthday. Presumably most readers’ memories of their birthdays are like this. Episodic memory is not such abstracted memory but rather memory of information together with memory of the context or with some aspects of the context of the episode that generated the memory. This produces qualia in the rememberer, whereas no such qualia need accompany semantic memory. It is with the former type of memory that one can engage in a sort of mental time travel or reliving an experience.

Instead of employing the above distinction, some philosophers have employed a distinction that divides memory in three parts (Bernecker Reference Bernecker2010). Experiential (or personal) includes memory of experience of the self which has qualia and the phenomena of reliving; propositional (or factual) is memory of facts which needs not be about oneself and of which one needs not have direct experience. As Bernecker (Reference Bernecker2010) points out, both types of memory seek to represent the world. In contrast, practical (or procedural) memory is not like this. Examples of such memory include how to ride a bike, swim, and so on. As the reader has probably anticipated, experiential memory is sometimes associated with episodic memory and semantic memory is sometimes associated with propositional memory. Milburn and Moon (Reference Milburn, Moon, Doyle, Milburn and Pritchard2019: 282) use “experiential memory” but favour using “stored memory” rather than “propositional memory”. Someone remembering their own name is offered as an example of stored memory and in discussing the example they note that such memory knowledge “does not depend on any kind of memory image or experience”. Of course, it is clear that they are using “stored memory” in contrast to “experiential” or “episodic memory”. While there is clearly some untidiness in all this conflicting terminology, making the case for one set of distinctions over another is not our primary concern. Instead, our position is that epistemological disjunctivism of memory applies to memory knowledge where this implies a belief. So, what is key to our position then is not the appropriateness of the psychological or philosophical taxonomy of memory but simply what is true of memory knowledge that is propositional.Footnote ⁷

Each sort of memory (episodic and semantic; experiential and propositional; experiential and stored) is relevant for propositional memory knowledge, whereas this may not be the case for, say, practical memory.Footnote ⁸ In fact, it is memory that yields belief that is solely of interest here. After all, by parity with ED-P, it is memory knowledge that p which is of interest for an epistemological disjunctivist that follows in the steps of ED-P.Footnote ⁹

ED-M is a claim about memory knowledge that S acquires in virtue of being in possession of rational support R (i.e., remembering that p) for her belief that p, which is both factive (R's obtaining entails p) and reflectively accessible to S. As is the case in ED-P, R is factive if R's obtaining entails p, and R is reflectively accessible to S if S is able to be aware of the rational support underlying her belief – she is aware that she remembers that p. Thus, according to ED-M, if S has memory knowledge that Paris is the capital of France, then S knows this in virtue of being in possession of rational support, remembering that Paris is the capital of France, for her belief. S's rational support is factive in that if one remembers that Paris is the capital of France, then it is true that Paris is the capital of France and of course remembering this is something that is reflectively accessible to S. After all, she is able to be aware of the rational support underlying her belief. Now she may not remember when she was told, anything about the context in which she learnt that Paris is the capital of France, and so on. This does not, however, imply that she cannot remember that Paris is the capital of France, nor that she is unable to be aware of her memory that Paris is the capital of France underlying her belief. The above is an example of semantic memory, although there is nothing about the description of the case that means that episodic memory could not equally be a basis for knowledge.

If all this is right, and we will consider some challenges, then ED-M holds for declarative memory. Can we say it holds for memory in general? Here the essential point is about what can count as propositional memory knowledge in the first place, as that is the sort of memory knowledge that is of interest to us. The most important question here is what sort of memory can produce propositional memory knowledge in the first place. On the face of things, practical memory cannot, although it can produce knowledge-how. The obvious candidates for the sorts of memory that can produce propositional memory knowledge are semantic and episodic memory. Nonetheless, our position is simply that a requirement for any paradigm case of propositional memory knowledge is that the relevant agent has knowledge that p in virtue of being in possession of rational support, R, for this belief that p, which is factive, R's obtaining entails p, and reflectively accessible to the agent.

As discussed, memory can produce knowledge that fits within an ED-M framework. Our claim is that any paradigm case of memory knowledge can be handled by the epistemological disjunctivist framework. We claim this on the basis that paradigm memory knowledge cases are such that the subject meets the disjunctivist requirements for memory knowledge. That is, the subject has propositional memory belief that p by virtue of possessing rational support R, remembering that p, which is factive and the subject is able to be aware of R underlying her belief, just as is the case for perceptual knowledge. Assuming that remembering that p is factive, we do not see a challenge to ED-M in cases in which the subject is unable to be aware of her rational support for her belief but the knowledge is memory knowledge. This brings us to the next point.

There are cases that might at first blush seem to be memory knowledge that are in fact not memory knowledge. These are the sorts of cases where although the agent remembers that p, what underlies their belief is not their memory that p but rather an alternative source such as an inference for p, perhaps informed by memory q, or even informed by memory that p but underlain by the inference, rather than by the memory that p. A similar point holds for deduction. The case in which an agent remembers that the ketchup is either in the fridge or is in the living room, and the agent sees that it is not in the fridge and so believes it is in the living room is not a case of memory knowledge. The relevant proposition has not been remembered, but it has been inferred.

There is also, however, a more difficult challenge that can arise from how we think about the link between memory and belief. One might hold that memory that p does not entail a belief that p. Rather the belief only comes about through a process of inference. Of course one can say the same about perceptual belief. One can hold that having the perception that p is quite separate from the belief that p. After all, sometimes one does not believe one's eyes. One may believe, wrongly, that one is hallucinating. If the equivalent for memory is a problem for ED-M, it is a problem for ED-P.Footnote ¹⁰ Recall, our argument is merely that if ED-P is successful then ED-M can be successful too.

4. The case for parity

Let us consider the case for parity of agreement between ED-P and ED-M. ED-M is the following position:

As stated in this formulation, like ED-P, ED-M provides a necessary condition for memory knowledge in paradigm cases. Correspondingly, in the formulation of ED-M, reflective accessibility is a requirement according to which S should be able to be aware of the rational support underlying her memory belief; factivity requires that the rational support S possesses for her belief that p entails that p.

ED-M is attractive in three aspects just as ED-P. First, if ED-M is true, it can satisfy both epistemological internalist and externalist intuitions: after all, reflective accessibility is an internalist requirement and factivity is an externalist requirement.

Second, ED-M makes sense of common practice regarding justifying memory beliefs. Analogous with ED-P, a common practice in justifying one's memory belief that p is to appeal to one's remembering that p. More importantly, these common practices can be illuminated by ED-M. For one thing, remembering that p is factive as well; S cannot remember that p unless p is true.Footnote ¹¹ For another, in paradigm cases of experiential memory knowledge, S can know by reflection alone that S remembers that p. Those common practices hence provide grounds for us to adopt ED-M.

Third, ED-M can also offer a satisfactory way to cope with Pritchard's radical sceptical paradox as it pertains to memory knowledge:

The plausibility of (UM1) stems from the idea that we are unable to subjectively distinguish normal cases of memory knowledge from radical sceptical hypotheses (say, the scenario in which the world is created five minutes ago, and all of one's memories of events prior to this time are illusory).Footnote ¹² Given this, it seems difficult for us to possess any rational support that favours our memory beliefs over the five-minutes-world hypothesis. At the same time, (UM3) seems generally acceptable.

(UM2) is the very same statement as (UP2). If we accept (UP2), (UM2) should be accepted. (UM2) is of course the source of the conflict between (UM1) and (UM3). While each of the statements in the underdetermination-based sceptical paradox regarding memory knowledge seem plausible, there is an inconsistency between them, for the denial of (UM3) is entailed by (UM1) and (UM2).

Just like the strategy ED-P deploys to deal with the underdetermination-based sceptical paradox regarding perceptual knowledge, we can reject (UM1) by accepting ED-M. According to ED-M, in paradigm cases of memory knowledge, S's rational support for her memory belief that p is factive, which is not available to her counterpart situated in the five-minutes world. Granted that the rational support S possesses is factive, it favours the belief that p over the five-minutes-world hypothesis. Therefore, ED-M is preferable for it presents a satisfactory way to cope with the paradox above.

Based on the considerations above, we have three reasons for adopting ED-M if it is available. As we shall now see, however, ED-M faces three familiar worries.

Being similar to ED-P, ED-M encounters the access problem, the distinguishability problem, and the basis problem. As Milburn and Moon (Reference Milburn, Moon, Doyle, Milburn and Pritchard2019) point out, the presence of both the access problem and the distinguishability problem has nothing to do with what kind of knowledge the version of disjunctivism is about. In other words, what gives rise to these two problems is that disjunctivists maintain that S can form beliefs about contingent truths with rational supports which are factive and reflectively accessible to S. If this is the case, ED-M can deal with these two problems by rejoinders similar to ED-P's.Footnote ¹³

Along with the two problems above, ED-M must have a satisfactory response to the basis problem which is similar to ED-P so as to establish the parity thesis. As mentioned previously, by rejecting (SK), ED-P blocks the presence of the basis problem. In the same vein, if ED-M can reject the claim that remembering that p entails that knowing that p (RK), then ED-M can also block the presence of the basis problem.

As we have seen, Pritchard (Reference Pritchard2012: 26–8, 2016: 127–8) argues against (SK) by making the case that S can see that p without believing that p. The example of this possibility was that S wrongly believes that her faculty is malfunctioning due to the presence of a misleading defeater. So while S sees that p, and p is thus true, she rejects what she sees as grounds for p. The very same approach can be taken with memory. S remembers that p but fails to believe p on the basis, say, that she has mistakenly treated the sort of memory she has as being very unreliable. Remembering that p thereby fails to entail knowing that p, given that knowledge implies belief but remembering does not. Since (RK) is false, the basis problem is therefore dissolved just as ED-P's rejecting (SK) dissolves the basis problem.

5. Objections to ED-M

ED-M faces challenges. As previously discussed, Milburn and Moon (Reference Milburn, Moon, Doyle, Milburn and Pritchard2019) distinguish between experiential memory knowledge and store memory knowledge. While they argue the case for an epistemological disjunctivism of experiential memory, they regard epistemological disjunctivism of store memory knowledge (ED-MS) not to be viable. Below is their articulation of an epistemological disjunctivism of store memory:

ED-MS provides a necessary condition for stored memory knowledge in paradigm cases. According to ED-MS, S has stored memory knowledge that p only if S's belief that p is formed through having rational support which is reflectively accessible and factive. Milburn and Moon (Reference Milburn, Moon, Doyle, Milburn and Pritchard2019: 192) argue that an epistemological disjunctivism of memory does not succeed for what they describe as non-deductive inferential stored memory and non-inferential stored memory knowledge. We will deal with the latter charge here — see section 3 for our response to the former.

The authors offer the example of Michael who remembers that snow is white. Actually this example is essentially the same as the example of knowing that Paris is the capital of France discussed previously. If pressed for rational support for the claim that snow is white, Michael is likely to appeal to having seen or remembering that snow is white. But this would not do according to the authors, after all here we are interested in stored memory rather than experiential memory. Stored memory is often unconscious. What if Michael is taking a nap?

Here the authors identify a general problem for access internalists. Presumably the thought though is that experiential memory by definition never falls foul of such a problem given what is required for experiential memory – qualia and experience of reliving. Someone then cannot be experiential remembering when in dreamless sleep, as she cannot have the experiences that are part of experiential memory in such a case. Still, this also produces the result that I only have memory knowledge of the day I graduated, where I was on 11 September 2001, or whatever, when not asleep, and so on. The challenge here then is a general one for the access internalist, rather than a challenge that is specific to ED-MS.

Perhaps the sleeping example is an unnecessary distraction from a more serious challenge to a general ED-M or an ED-MS in particular. Even at the best of times – for example, when one is awake and alert – one cannot access unconscious memory, whereas this is obviously not the case with experiential memory. Does this then show that there is a problem for ED-MS? After all, how can S be said to be in possession of rational support for her belief that p which is both factive and reflectively accessible? It is true that S does not have reflective access to a particular memorial experience that can serve as rational support for her belief. Nonetheless, it is not obvious that this is what is necessary. S after all has reflective access to the fact that he remembers.

Consider again the case of Michael. Michael remembers that snow is white without, let us suppose, remembering a particular occasion of discovery or even a particular experience of seeing or hearing that snow is white. So it is not as if the memory of discovering that snow is white or a particular experience of being told snow is white is reflectively accessible for his belief that snow is white. Michael does, however, have reflective access to the fact that he remembers. That this is devoid of the experiential component that is part of experiential memory does not lessen the rational support that it provides for his belief that snow is white. When challenged as to how he knows snow is white, Michael can say that he remembers that snow is white.

Following the previous discussion, it might be argued that the problem with ED-M is not that it's overly permissive but rather that it's overly restrictive. Key to our answer above was that Michael has reflective access to the fact that he remembers. Let us consider a different case – one in which he does not have access to the fact that he remembers. Suppose upon being asked a question on a quiz show, he finds himself with an answer. His answer is that Elizabeth I died in 1603.Footnote ¹⁴ He is not sure why he thinks this, perhaps he even thinks it is just a guess. In any case, let us stipulate that he does not have reflective access to the fact that he remembers this. Of course, he is right that Elizabeth I died in 1603. Should we say that he knows in this case? If he does know, then why think ED-M is a necessary condition for memory knowledge?

A natural response is to either deny that he knows or say that this is a non-paradigmatic case. While the former is a natural response for epistemologists with internalist sympathies, it is not for those with externalist sympathies. The alternative answer is that it is not a paradigmatic case of memory knowledge and epistemological disjunctivism only aims to handle paradigmatic cases. Indeed, epistemological disjunctivism requires reflective access for paradigm cases. The sort of case in which paradigmatic perceptual knowledge is gained, is the sort of case in which the agent possesses rational factive support from seeing that p (Pritchard Reference Pritchard2012: 31). Pritchard (Reference Pritchard2012: 37, 46) makes clear that for a case to be a paradigm case, such rational support must satisfy the reflective access requirement. In other words, the agent must be capable of being aware of the rational support underlying her belief. By parity with ED-P, ED-M would deny that the quiz case is a paradigmatic case, as, according to the description of the case, the protagonist lacks reflective access.Footnote ¹⁵

The claim that it is not a paradigmatic case neither commits ED-M to say that it is a case of knowledge or that it's not. Once it is not a paradigmatic case, then it falls outside the scope of epistemological disjunctivism.Footnote ¹⁶ Our account of ED-M is saved from the objection, as are the theoretical benefits that come with adopting that account.