Against overconfidence: arguing for the accessibility of memorial justification

Abstract

In this article, I argue that access internalism should replace preservationism, which has been called “a received view” in the epistemology of memory, as the standard position about memorial justification. My strategy for doing so is two-pronged. First, I argue that the considerations which motivate preservationism also support access internalism. Preservationism is mainly motivated by its ability to answer the explanatory challenges posed by the problem of stored belief and the problem of forgotten evidence. However, as I will demonstrate, access internalism also has the resources to provide plausible solutions to those problems. Second, I argue that preservationism faces a couple of problems which access internalism avoids. Doing so, I present a new scenario which, on the one hand, functions as a counterexample to preservationism, and, on the other hand, provides intuitive support for access internalism. Moreover, I also demonstrate how preservationism, in light of recent research in cognitive psychology, is vulnerable to skepticism about memorial justification, whereas access internalism remains unthreatened.

1 Memorial justification

Many, if not most, of our beliefs are justified by memory. Consider your belief (whether occurrent or dispositional) that Mercury is the smallest planet in our solar system, that Fermat’s Last Theorem has been proved to be true, or that Napoleon lost the battle of Waterloo. Such beliefs appear to be justified, but not by your perceptual experience. Rather, they are justified by your memories.¹ So instead of saying that they are perceptually justified, we should say that they are memorially justified, or that we have memorial justification for them.

Memory is a source of valuable epistemic properties, like justification. But what conditions must be satisfied in order for your memory to confer justification upon your doxastic attitudes? In this paper, I am going to develop a novel argument, inspired by Laurence BonJour’s (1980, 1985) clairvoyance scenario, for the claim that there is an internalist accessibility condition on memorial justification. This I will do by presenting a scenario involving an epistemic agent who exhibits a special sort of overconfidence with respect to the memory beliefs he has. Despite the agent’s memory beliefs being reliably produced by a properly functioning cognitive faculty (as the scenario stipulates), they nevertheless appear to be unjustified. And, I argue, the best explanation for why that is so is that there is an internalist accessibility condition on memorial justification which the agent fails to satisfy. Moreover, after having presented the argument, I will consider and respond to the threat which preservationism—i.e., the position that memory preserves the (positive) justificatory status of the beliefs that enter it—poses to my view.

The paper is structured as follows. In Sect. 2, I present BonJour’s clairvoyance scenario and discuss the way in which it provides the resources for formulating an argument for access internalism. In Sect. 3, I present the overconfidence scenario and argue that there is an indistinguishability problem for anyone who denies the veracity of the intuition it elicits. In Sect. 4, I argue, based on the aforementioned intuition, that there is an internalist accessibility condition on memorial justification and spell out some of the details about how that condition should be understood. In Sect. 5, I argue that access internalism should replace preservationism as the standard view in the epistemology of memory. In Sect. 6, I conclude and discuss the relevance of the paper and its argument for the larger literature on the epistemology of memory and the internalism–externalism distinction.

2 BonJour on clairvoyance and justification

According to simple process reliabilism, most famously endorsed by Goldman (1979), a belief is justified just in case it is formed on the basis of a reliable doxastic disposition (and there are no undefeated defeaters). However, the view is vulnerable to a number of counterexamples, most of which either target the claim that reliability is necessary for justification or the claim that reliability is sufficient for justification.² One of the most influential counterexamples to the sufficiency claim is offered by Bonjour, and it goes as follows:

Clairvoyance

Norman, under certain conditions which usually obtain, is a completely reliable clairvoyant with respect to certain kinds of subject matter. He possesses no evidence or reasons of any kind for or against the general possibility of such a cognitive power or for or against the thesis that he possesses it. One day Norman comes to believe that the President is in New York City, though he has no evidence either for or against this belief. In fact the belief is true and results from his clairvoyant power under circumstances in which it is completely reliable. (Bonjour 1985, p. 41, Cf. 1980.)

Now, as simple process reliabilism would have it, Norman’s belief that the President is in New York City is justified because it is produced by a reliable belief-forming process (and there are no undefeated defeaters). However, the reliabilist verdict clearly seems to be wrong. Intuitively, Norman’s belief is no more justified than a random guess or hunch. Hence, provided that the intuition is correct, simple process reliabilism must be false since reliability isn’t sufficient for justification.

But where exactly does simple process reliabilism go wrong? BonJour helps us to a diagnosis of the issue. Moreover, his diagnosis also provides us with resources for formulating an argument for access internalism based on our intuitive judgments about the Clairvoyance scenario. In The Structure of Empirical Knowledge, BonJour has the following to say about the reliabilist response to the Clairvoyance scenario:

One reason why externalism may seem initially plausible is that if the external relation in question genuinely obtains [i.e., the reliable relation between one’s belief and the truth], then Norman will in fact not go wrong in accepting the belief, and it is, in a sense, not an accident that this is so: it would not be an accident from the standpoint of our hypothetical observer who knows all the relevant facts and laws. But how is this supposed to justify Norman’s belief? From his subjective perspective it is an accident that the belief is true. And the suggestion here is that the rationality or justifiability of Norman’s belief should be judged from Norman’s own perspective rather than from one which is unavailable to him. (BonJour 1985, pp. 43–44.)

Similarly, in his co-authored book with Ernest Sosa, Epistemic Justification, he writes:

[Norman is] being epistemically irrational and irresponsible in accepting beliefs whose provenance can only be a total mystery to [him], whose status is as far as [he] can tell no different from that of a stray hunch or arbitrary conviction. Here again, externalism seems to sunder the concept of epistemic justification entirely from the concept of epistemic rationality or responsibility, leaving the former concept with no clear intuitive content. (BonJour and Sosa 2003, p. 32.)

Here, BonJour tells us that simple process reliabilism and similar externalist positions are forced to give the wrong verdict about the clairvoyance scenario because they fail to recognize that a subject cannot have a justified belief B without B being rational or reasonable from the subject’s first person perspective. Even though Norman’s belief about the President’s whereabouts is both true and reliably produced, he doesn’t have any supporting evidence or reason that is accessible from his first person perspective; as far as he is in a position to tell, there is no reason to think that his belief in any way accurately represents where the President happens to be located. Moreover, BonJour’s diagnosis doesn’t only provide a plausible explanation of where the reliabilist goes wrong, but also explains why the access internalist appears to be correct. His original diagnosis relies on the thesis that if you don’t have privileged access to the justificatory status of a belief B, then B cannot be justified. And by contraposition we get the (equivalent) thesis that if the belief B is justified, then you must have privileged access to the justificatory status of B—which simply is what access internalism states. Thus, BonJour’s diagnosis provides the resources for formulating an abductive argument for access internalism. As he sees it, access internalism offers the best explanation for our intuitive judgments about the clairvoyance scenario.³,⁴

BonJour’s diagnosis and argument have met with a lot of criticism from externalists. Recently, I have responded to several of the externalist’s objections.⁵ In this paper, however, I will simply offer a reconstruction of the argument and present my own, structurally analogous, argument about memorial justification.

3 Overconfidence and the Indistinguishability problem

Allow me to follow BonJour by introducing the overconfidence scenario as a counterexample to Goldman’s (1979) simple process reliabilism, before I present the argument for access internalism about memorial justification in the next section. Applied to memorial justification in particular, rather than justification in general, simple process reliabilism tells us that a belief B is memorially justified just in case B is the result of a reliable memory-based doxastic disposition (and there are no undefeated defeaters). Thus, the view is committed to both a necessity claim—according to which having a reliable memory-based doxastic disposition is necessary for memorial justification —and a sufficiency claim—according to which having a reliable (undefeated) memory-based doxastic disposition is sufficient for memorial justification. However, as we will see, the sufficiency claim is particularly vulnerable. Consider the following scenario:

Overconfidence

Robert is, for the most part, a normal epistemic agent. His ability to reason is more or less like that of other people, his perceptual faculties function normally, and he usually acquires knowledge and justified beliefs by relying on his perception and reasoning. However, Robert’s memory functions in somewhat unusual ways. Whereas his memory is exceptionally reliable—his memory-based doxastic dispositions almost always produce true beliefs with the same or relevantly similar content to the beliefs that entered his memory—he never has any reason to think that his memory beliefs indeed are products of his memory, rather than some other process or source.⁶ Moreover, Robert doesn’t have any reason or evidence for or against the reliability of his memory.⁷ Nevertheless, he always trusts himself with respect to whatever his memory “suggests” to be the case.

One day, Robert is asked by his friend which year Goethe died. Although his interest in literature is miniscule and he doesn’t in any way seem⁸ to remember anything about when Goethe (or any other author for that matter) might have died, his memory prompts him to answer that “Goethe died on the 22^nd of March 1832”, and he believes it. Intrigued by Robert’s quick and confident answer, the friend asks him how he can be so sure. Robert thinks for a while, but, since he can’t come up with any reason for thinking that his answer is true, he simply tells his friend that he doesn’t know why he is so confident in his answer, he just is. Puzzled by the response, the friend asks him whether he is trying to pull a trick on him. Robert, taken aback by the accusation, thinks for himself a bit more, before he reassuredly says that he isn’t trying to pull a trick but that he is absolutely confident that his answer is true. “Indeed, too confident”, the friend thinks to himself.

Just like with the clairvoyance scenario, according to simple process reliabilism, Robert’s belief that Goethe died on the 22nd of March 1832 is memorially justified because it is the result of a reliable memory-based doxastic disposition (and there are no undefeated defeaters). However, yet again, the reliabilist verdict seems to be wrong. Intuitively, Robert is, as the friend’s concluding thought suggests, overconfident, and his belief isn’t justified. Indeed, his belief appears to be no more justified than a random guess or hunch. Hence, provided that the intuition is correct, simple process reliabilism about memorial justification must be false since having a reliable (undefeated) memory-based doxastic disposition isn’t sufficient for memorial justification.

In order to support this intuition (i.e., that overconfident subjects, like Robert, don’t have memorially justified beliefs), I’m going to point out a problem which is yet to be discussed in the literature, and which those who deny its veracity invariably must face. I call it the indistinguishability problem, and it is a deductive argument against anyone who denies the veracity of the aforementioned intuition. The argument can be summarized as follows:

• 1. Overconfident subjects are not in a position to know of their memory beliefs that they are products of memory, rather than some non-reliable process or source.

• 2. If a subject isn’t in a position to know of their memory beliefs that they are products of memory, rather than some non-reliable process or source, then the subject should not believe of their memory beliefs that they are products of some reliable process or source.

• 3. If a subject should not believe of their memory beliefs that they are products of some reliable process or source, then their memory beliefs cannot be memorially justified.

• 4. Hence, overconfident subjects cannot have memorially justified memory beliefs.⁹

Since the argument’s conclusion entails that the intuition elicited by the overconfidence scenario is true, those who want to deny the veracity of said intuition must reject at least one of the argument’s premises. Now, let’s examine how one might go about doing so.

Rejecting premise 1 amounts to the claim that overconfident subjects are in a position to know of their memory beliefs that they are products memory, rather than some non-reliable process or source. Before we evaluate this claim, I need to say a little bit more about what I take the notion of being in a position to know to mean. Following Smithies (2012b, p. 268), I will define the somewhat abstruse notion of being in a position to know as meaning that one satisfies all the epistemic conditions of knowledge—like having unGettiered (propositional) justification to believe a true proposition. Knowing a proposition, by contrast, also requires that one satisfies the psychological conditions of knowledge—like believing a true proposition on the basis of one’s unGettiered justification to believe it. Being in a position to know is therefore a necessary, but not sufficient, condition for knowing.¹⁰

Now, with this definition in mind, rejecting premise 1 is to claim that overconfident subjects have unGettiered justification to believe of their memory beliefs that they are products of memory. But this clearly seems false. If we consider Robert, for example, he doesn’t have any reason or evidence which can justify him in believing that his belief about Goethe’s date of death indeed is a memory belief. Moreover, not only doesn’t he have any reason or evidence to justify such a second-order belief, it wouldn’t be reliable in the circumstances under consideration. If, say, Robert actually were to form the (second-order) belief that his (first-order) belief about Goethe is a memory belief, then, given whatever (plausible) rule for belief-formation that he tries to follow, his best attempt to follow that rule wouldn’t stop him from also believing that other (first-order) beliefs of his which actually amount to nothing more than mere guesswork, but without himself being in a position to know so, also are products of his memory—and this of course undermines the idea that Robert’s (second-order) belief that his (first-order) belief about Goethe is a product of memory is reliable.¹¹ So premise 1 therefore appears to be on solid ground; overconfident subjects are not in a position to know of their memory beliefs that they are products or memory, rather than some non-reliable process or source.

Rejecting premise 2 amounts to the claim that subjects who aren’t in a position to know of their memory beliefs that they indeed are products of memory, rather than some non-reliable process or source, are rationally permitted to believe of their memory beliefs that they are products of some reliable process or source. But this naturally raises the question: why are such subjects permitted to believe of their memory beliefs that they are the result of some reliable process or source? Those who want to reject premise 2 thus face a difficult explanatory challenge; namely, they have to explain how subjects who aren’t in position to distinguish their memory beliefs from random guesses or hunches nevertheless somehow still have grounds for believing of their memory beliefs that they aren’t such guesses or hunches, or indeed products of some other unreliable process.

Given that subjects who aren’t in a position to know of their memory beliefs that they are products of memory don’t have any reason or evidence supporting second-order beliefs to the effect that their first-order memory beliefs actually come from the source that they do, I don’t see any reason why we should think that it would be rationally permitted or reasonable for them to hold second-order beliefs to the effect that their first order memory beliefs are reliably produced. However, one might perhaps try to answer the challenge by following Wright (2004) and arguing that there are certain basic beliefs which we are entitled to hold by default,¹² and that second-order beliefs about the reliability of one’s first-order memory beliefs somehow belong to that category. Although this might be a consistent line of response, it would require a lot of work to argue that second-order beliefs about the reliability of one’s first-order memory beliefs somehow belong to the category of beliefs that are entitled by default (assuming that such a category exists). In his work on entitlement, Wright focuses for the most part on presuppositions (2004) and authenticity-conditions (2014), which function more or less like Wittgensteinian hinge propositions that we are entitled to believe without evidence or argument,¹³ rather than such second-order beliefs about the reliability of one’s first-order memory beliefs. Moreover, this line of response is also problematic insofar as it seems to license as warranted or entitled some sort of bootstrapping argumentation, since forming a first-order memory belief (but without being in a position to know its source) is sufficient for generating entitlement to believe in the reliability of that first-order memory belief.¹⁴

That being said, it seems that premise 2 should be accepted and that subjects who aren’t in a position to know of their memory beliefs that they indeed are products of memory shouldn’t believe of their memory beliefs that they are products of some reliable process or source.

Rejecting premise 3 amounts to the claim that the memory beliefs of subjects who shouldn’t believe of their memory beliefs that they are products of some reliable process or source nevertheless can be memorially justified. But this would be an odd claim to argue for. To see why that so, consider which doxastic attitude such a subject should have toward the proposition that their memory beliefs are products of some reliable process or source. Presumably, for any proposition that you have the concepts to understand, you should either believe it, withhold judgment about it, or disbelieve it, and you should never have more than one of these attitudes toward it at the same time.¹⁵ So if the subject shouldn’t believe in the proposition under consideration, then he should withhold judgment about it or disbelieve it (but not both). But then he would end up being in an absurd quasi-akratic position where he withholds/disbelieves of his memory beliefs that they are reliably produced even though they (supposedly) are memorially justified.¹⁶ If, on the one hand, he disbelieves of his memory beliefs that they are reliably produced, then (equivalently) he believes of his memory beliefs that they are unreliably produced. But if he believes of his memory beliefs that they are unreliably produced, then that defeats any memorial justification he might have for his memory beliefs—in which case they cannot be memorially justified. If, on the other hand, he withholds judgment about the proposition that his memory beliefs are reliably produced, then he is committed to being agnostic or neutral with respect to its truth-value. But if he is committed to being neutral with respect to the reliability of his memory beliefs, then it doesn’t seem that they can be memorially justified. After all, if the subject is equally confident that his memory beliefs are reliably/unreliably produced, why should he hold them?

Following Smithies (2012a), in order to further pump the intuition that a subject cannot justifiably withhold/disbelieve of his memory beliefs that they are reliably produced at the same time that they are memorially justified, let me just briefly point out how such a subject would find himself in a position where it would be completely natural for him to assert Moorean conjunctions. Consider the following conjunctions:

• (1) p and I do not know that p.

• (2) p and it is an open question whether I know that p.

Some philosophers think that asserting either (1) or (2) above is absurd.¹⁷ Why? Plausibly because one cannot rationally believe both conjuncts of such Moorean conjunctions at the same time.¹⁸ If one rationally believes the conjunct (i) that p, then one cannot rationally believe the conjunct (ii) that one doesn’t know that p, or the conjunct (iii) that it is an open question whether one knows that p. In other words, according to some philosophers, one cannot rationally hold a certain belief and at the same time fail to take that belief as amounting to knowledge.¹⁹

Similarly, consider the following conjunctions about justification:

• (3) p and I do not have justification to believe that p.

• (4) p and it is an open question whether I have justification to believe that p.

Just like with conjunctions (1) and (2) above, asserting either (3) or (4) is absurd because one cannot rationally believe both of their conjuncts at the same time. If one rationally believes the conjunct (i) that p, then one cannot rationally believe the conjunct (ii) that one doesn’t have justification to believe that p, or the conjunct (iii) that it is an open question whether one has justification to believe that p.

However, despite the absurdity of asserting either (3) or (4), a subject who justifiably withholds/disbelieves of his memory beliefs that they are reliably produced at the same time that they are memorially justified will be in a position where asserting those conjunctions is entirely natural. For example, he might testify to some proposition p that he has a (justified) memory belief in, while simultaneously (justifiably) disbelieving/withholding judgment that his belief that p has been reliably produced and thereby also asserting that he doesn’t, or might not have, justification to believe that p. So having this sort of lower-level belief, while at the same time failing to also have some higher-level belief to the effect that the lower-level belief is reliably produced, simply seems absurd.

One can of course always try to avoid this dilemma by denying that one always has justification for adopting either of the doxastic attitudes toward any proposition that one has the concepts to understand. Although I remain doubtful about that kind of position, even if it turns out to be true, I don’t think that the kind of proposition that we’re currently considering will be one of those that we shouldn’t always hold either of the doxastic attitudes toward. For example, whereas your evidence might not determine that you should hold any particular doxastic attitude toward the proposition that the Pope currently is lying in his bed, it does seem that it always will determine such an attitude when it comes to the reliability of your memory beliefs. If your evidence on balance neither tells in favor of the truth nor the falsity of such a proposition, then you should withhold judgment in it. And if it does tell in favor of the truth/falsity of the proposition, then you should believe/disbelieve it. Evaluating which doxastic attitude your evidence on balance supports when it comes to the proposition that your memory beliefs aren’t unreliably produced may of course be very difficult, but that doesn’t mean that it is impossible or that your evidence on balance doesn’t tell in favor of either attitude.

I therefore think that premise 3 should be accepted and that the memory beliefs of subjects who shouldn’t believe of their memory beliefs that they are products of some reliable process or source cannot be memorially justified. The indistinguishability problem thus presents a deductive argument against anyone who denies the intuition elicited by the overconfidence scenario. And it does so by highlighting the fact that overconfident subjects aren’t in a position to distinguish bona fide memory beliefs from mere guesswork, and that failing to be in such a position is incompatible with having memorially justified beliefs.

In this section, I have presented a scenario that is analogous to those offered by BonJour and others, but which focuses on memorial justification in particular, rather than justification in general. Moreover, in order to move beyond the discussions by earlier internalists, I have presented a novel argument which supports the intuition elicited by the overconfidence scenario, and which thus bolsters the case for access internalism about memorial justification. With that said, I now turn to the issues of why and how memorial justification is internally accessible.

4 The accessibility of memorial justification

Why exactly does Robert’s belief fail to be justified? I think the best explanation for this intuitive datum is that there is an internalist accessibility condition on memorial justification that he fails to satisfy. A belief can be memorially justified only if the subject has privileged access to the justificatory status of that belief. If the subject cannot even in principle tell what a certain belief B rationally has going for it, then B cannot be memorially justified—even if it is produced by the subject’s reliable, properly functioning faculty of memory. Indeed, if the belief B is wholly indistinguishable from, say, mere guesswork, then B cannot be memorially justified.

The access internalist about memorial justification thus thinks that a belief can be memorially justified only if the subject has privileged access to the justificatory status of that belief. But this naturally raises the question: what kind of memorial justification does the internalist condition attach to? It is common to distinguish between two different kinds of justification, which we may call propositional justification and doxastic justification, following Firth (1978). On the one hand, you can have justification to believe a certain proposition, regardless of whether you actually believe it. For example, you can have justification to believe that Napoleon lost the battle of Waterloo, but without having to believe it. This is what we call propositional justification. On the other hand, you can also have justifiably held beliefs. For example, if you believe the proposition above—that Napoleon lost the battle of Waterloo—in a way that is properly based on that which gives you (propositional) justification to believe it, then your belief is justifiably held. This is what we call doxastic justification. So which of these two kinds of memorial justification does the internalist condition attach to?

I claim that we should be access internalists about memorial justification in the propositional sense, but not the doxastic sense. Why? Because doxastic justification, as presented above, simply is propositional justification plus proper basing, which means that both those conditions must be internally accessible if facts about doxastic justification are to be internally accessible. So access internalism about doxastic justification is indeed a stronger position, and, as a result, it is more implausible. To see why that is so, consider how the facts about whether the beliefs you have are properly based on their justificatory basis are psychological (and not epistemic) in nature; that is, they have to do with the psychological processes and mechanisms that have led you to form and maintain your beliefs. But such psychological processes don’t seem to be internally accessible. Not only did many of them occur in the past, thus making our access to them dependent on our fallible memorial capacities, but many of them also occur entirely below the level of conscious experience, thus making them completely inaccessible to our faculties of reflection and introspection.

So, I claim, we should be access internalists about memorial justification in the propositional sense.²⁰ But another question we also need to ask is how facts about memorial justification are internally accessible. Here I want to make two suggestions. One, they are internally accessible insofar as we are in an a priori position to know them; two, we are in an a priori position to know them insofar the memorial justifiers and the justifying relations are similarly accessible. Let me begin by elaborating on the first point.

Drawing on Smithies’ definition from the previous section, we always have unGettiered a priori justification to believe the facts about what we have memorial justification to believe.²¹ Moreover, it should be noted that I’m using a priori in the traditional sense that a condition is a priori just in case it doesn’t depend on any of the sense modalities. This means that what we have a priori justification to believe shouldn’t only be taken to include what we have reflective or reasoning-based justification to believe, but also, for example, what we have introspective justification to believe—even though introspection clearly involves an experiential aspect.²² Thus, according to the position I advocate, anyone who takes advantage of their epistemic position by also satisfying the psychological conditions of knowledge will know in an a priori manner what they have memorial justification to believe.

When it comes to the second point, we are in an a priori position to know the facts about what we have memorial justification to believe by being in an a priori position to know the memorial justifiers we have and how they justify the beliefs they do.²³ And we are in a position to have the latter knowledge (i) by having unGettiered introspective justification to believe what our memorial justifiers or reasons for belief are, and (ii) by having unGettiered reflective justification to believe that those reasons justify the beliefs they do (and to the degree that they do).²⁴,²⁵

So in a case where you have some memorial justification ^MJ for the belief that p (written ^MJ: p), you will also, in virtue of your faculties of introspection and reflection, have:

• 1. A priori justification ^APJ to believe that you have memorial justification ^MJ for believing that p: ^MJ: p → ^APJ: (^MJ: p).

  (where the a priori justification isn’t vulnerable to Gettier cases.)

And by having this a priori justification, it will be sufficient for:

• 2. Being in an a priori position to know ^APK that you have memorial justification ^MJ for believing that p: ^APJ: (^MJ: p) → ^APK: (^MJ: p).

And this gives us the following internalist accessibility condition on memorial justification:

• 3. Access internalism about memorial justification: ^MJ: p → ^APK: (^MJ: p).²⁶

Although I don’t have the space to go into more detail here, this will do as an initial characterization of the internalist accessibility condition that attaches to memorial justification. According to access internalism about memorial justification, if you have memorial justification to believe that p, then you are in an a priori position to know that you have memorial justification to believe that p.²⁷

5 Preservationism dethroned

Access internalism about memorial justification is, however, incompatible with what recently has been called “a received view” about memorial justification; namely, preservationism (Frise 2017, p. 487). This is what the thesis says:

Preservationism: if a belief B is justified at t₁, and B is retained in memory until t₂, then B is (prima facie) memorially justified at t₂.²⁸

So according to preservationism, any memory belief, whether occurrent or dispositional, that was justified at an earlier point in time is presently (prima facie) memorially justified.²⁹ Preservationism is in conflict with access internalism since a subject can satisfy the antecedent of the conditional, and thereby have a memorially justified belief at t₂, but without satisfying the internal accessibility condition argued for in the previous section at t₂. In other words, on the preservationist picture, you can have a memorially justified belief, but without being in a position to know what that belief rationally has going for it.³⁰

Moreover, preservationism is motivated by its ability to solve a couple of notorious problems in the epistemology of memory. The first is the problem of stored belief, which challenges us to explain how the stored beliefs we have that intuitively seem to be justified indeed are justified.³¹ Preservationism solves the problem by claiming that a belief (whether stored or consciously entertained) will retain its justification as long as it is preserved in memory. And since stored beliefs by definition are preserved in memory, they do retain their justification. The second problem is the problem of forgotten evidence, which challenges us to explain how our justified (memory) beliefs whose original supporting evidence we have forgotten indeed are justified.³² You might, for example, have a justified belief that your best friend’s phone number used to be such-and-such, but without remembering what the evidence was that you first formed your belief on the basis of. Preservationism again solves the problem by claiming that a belief will retain its justification (whether forgotten or not) as long as it is preserved in memory.

Given that preservationism thus is motivated by its ability to avoid the problems above, and that it is in conflict with access internalism, does this provide the dialectical upper hand to preservationism? I am going to argue that it doesn’t. My strategy for doing so will be two-pronged: first, I will argue that access internalism about memorial justification has the resources to provide plausible solutions to the problem of stored belief and the problem of forgotten evidence; second, I will argue that preservationism faces a couple of problems which access internalism avoids, and therefore that internalism actually has the dialectical upper hand. Let me begin with the first point.

Access internalism about memorial justification can provide plausible answers to the explanatory challenges which the problem of stored belief and the problem of forgotten evidence pose. Consider first the problem of stored belief, which challenges us to explain how the stored beliefs we have that intuitively seem to be justified indeed are justified. According to access internalism, stored beliefs are typically justified because we are in an a priori position to know that we have certain justifying reasons and that those reasons support them. Conee and Feldman (2001), cf. Feldman (2004) and McCain (2016) have suggested that what justifies such stored beliefs are “stored justifications”, which, just like the beliefs they justify, also are memorial in nature. Moreover, this suggestion can be plausibly expanded along the lines suggested in the previous section by claiming that what justifies stored beliefs usually are other stored beliefs which we have introspective justification to believe that we have and reflective justification to believe support the stored beliefs that they indeed do justify. So, for example, my stored belief that Napoleon lost the battle of Waterloo is plausibly justified by another stored belief to the effect that a trustworthy source—e.g., my middle school teacher—told me so some time ago.

However, an anonymous referee raised the following objection: there will probably be cases where one isn’t able, from one’s subjective point of view, to distinguish genuine memory beliefs from confabulated beliefs, and this raises the worry that one cannot rationally rely on one’s genuine memory beliefs. Consider, for example, the belief that Napoleon lost the battle of Waterloo. Although I might have another stored belief to the effect that a trustworthy source told me so some time ago, it might nevertheless be the case that I am not able to tell whether or not the second belief is a genuine memory belief.³³ In response, I think we can appreciate the seriousness of this objection, while at the same time demonstrating that it does not undermine access internalism about memorial justification. More specifically, I think we can do so by acknowledging that what the objection says is true, but that it fails to contradict access internalism about memorial justification. According to the access internalist position laid out in the previous section, if one has memorial justification to believe p, then one is in an a priori position to know that this is the case. However, this version of access internalism does not claim that one must be able to take advantage of one’s epistemic position and know that one has memorial justification to believe p. Neither does it claim that one must be able to take advantage of one’s propositional justification by forming doxastically justified beliefs on their proper justificatory basis. So what this means is that even though the subjective indistinguishability of genuine and confabulated memory beliefs might prevent one from taking advantage of the memorial justification one has, it is still true that one is in an a priori position to know about one’s memorial justification. To put the point in more general terms, the access internalist position endorsed in this paper only makes a claim about what one is in an epistemic position to do, in the sense that it might be done by an ideally rational agent, while remaining neutral about people’s actual psychological capacity for reasoning and rationality.

Consider next the problem of forgotten evidence, which challenges us to explain how our justified (memory) beliefs whose original supporting evidence we have forgotten indeed are justified. According to access internalism, justified memory beliefs whose original supporting evidence we have forgotten are justified by other internally accessible evidence or reasons that we haven’t forgotten. In the literature on the epistemology of memory, we find two different kinds of explanation that have been suggested along these lines.³⁴ First, Madison (2014) has argued that just like the perceptual seemings we have justify certain corresponding beliefs, we also have memory-seemings which justify certain corresponding beliefs.³⁵ So whenever you have a justified memory belief whose original evidence you have forgotten, you will typically have some memory-seeming (which differs from the belief) supporting it. Second, Schroer (2008), following Audi (1995), has argued that memory beliefs themselves have certain distinctive phenomenal properties (indicative of their source) and that those properties provide defeasible reasons for belief. According to the view, simply having a memory belief that p necessarily involves a certain feel or phenomenal aspect that justifies believing that p (if I understand the view correctly).³⁶

I think the best way for the access internalist to answer the explanatory challenges posed by the problem of stored belief and the problem of forgotten evidence is to combine the idea that memory beliefs can function as “stored justifications” with Madison’s idea that we typically have memory-seemings that confer justification upon certain beliefs. By combining these ideas, the result is a sophisticated internalist view of memorial justification that has the power to deal with the challenges raised above. Our memorially justified beliefs are supported by other such beliefs or by memory-seemings. In the case of justified stored beliefs, they are justified by other stored beliefs, as is illustrated by the example about Napoleon above. And in the case of beliefs whose original supporting evidence we have forgotten, they will either be supported by memory-seemings or other stored beliefs (or both, of course), insofar as they are justified.

Now, it should be mentioned that whether or not we actually have enough stored beliefs and memory-seemings to justify all the beliefs that we intuitively think of as memorially justified is an empirical question. This means that the internalist, insofar as he relies on the view that I here endorse, runs the risk of committing himself to some form of skepticism, if the empirical evidence were to show that we don’t have a sufficient stock of such memory beliefs and memory-seemings. Moreover, I should note that this is a bullet that I am willing to bite; although I strongly suspect, on the basis of my own introspection, that we do have enough memory-seemings and memory beliefs in order to justify all the beliefs that intuitively seem to be memorially justified (cf. footnote 39).

Access internalism about memorial justification is thus just as suited to answer the explanatory challenges posed by the problem of stored belief and the problem of forgotten evidence as preservationism. Next, I want to argue that the philosopher’s scales should be tipped in favor of access internalism, since preservationism is vulnerable to a couple of objections which access internalism avoids.

The first objection is that there are counterexamples to the preservationist analysis of memorial justification. For example, Frise (2017) has recently provided several such examples formulated as dilemmas for the preservationist. In this paper, however, I will focus on the overconfidence scenario from Sect. 3. The scenario presented a subject named Robert who satisfies the antecedent of the preservationist conditional. By relying on his reasoning and perception, Robert acquires justified beliefs which his faculty of memory reliably retains. According to preservationism, Robert’s memory beliefs should therefore be memorially justified when he consciously entertains them at a later time. However, intuitively, they clearly fail to be justified. Since he doesn’t have any reason or evidence supporting them, Robert’s memory beliefs aren’t memorially justified. Indeed, even though his memory belief that Goethe died on the 22nd of March 1832 is true, reliably produced by a properly functioning cognitive faculty, and may have been justified in the past, it doesn’t seem to be (memorially) justified at the present. Moreover, as we have seen, access internalism offers the correct verdict about the scenario and it appears to receive the most support from the evidence which our intuitions provide. According to access internalism, the reason Robert’s memory belief isn’t memorially justified is that he isn’t in a position to know what that belief rationally has going for it. So whereas preservationism gives the wrong verdict about the scenario, access internalism plausibly accounts for our intuitive judgments about it.

The second objection suggests that preservationism, in light of recent work in cognitive psychology, is vulnerable to skepticism about memorial justification. The reason is simply that preservationism assumes that something like the traditional “storehouse view” of memory is true, when in fact it probably isn’t. According to this traditional view, memory is like a storehouse insofar as the items we retrieve (be they beliefs or episodic events³⁷) are the same ones that we deposited at an earlier time. The idea is simply that memory preserves (propositionally encoded) information and our doxastic attitude toward it. However, as Michaelian (2011) in his discussion and interpretation of recent work in cognitive psychology has demonstrated, the standard scientific view is that memory processing is generative or constructive, not preservative. The information that enters memory drastically changes at various stages of its processing; indeed, so much so that the information we retrieve most plausibly should be considered a construction of our memory processing, rather than simply a preserved informational unit—like a bottle of wine retrieved from a wine cellar.³⁸ According to Michaelian (2016, p. 82), “the basic lesson of constructive memory research is that remembering does in fact routinely involve modifications that introduce information not even implicitly contained in the earlier representation.”

So how does the fact that memory processing is constructive rather than preservative threaten preservationism about memorial justification? The problem is as follows. Given that recent work in cognitive psychology indicates that memory is generally constructive, why, on the preservationist view, should our memory beliefs be (memorially) justified? If our (justified) beliefs tend to change content while they’re processed in memory, why should we think that the beliefs we retrieve from memory are justified? After all, the belief we retrieve might very well be different from the one that entered memory in the first place, in which case the preservationist doesn’t provide us with any reason to think that that belief actually is justified. And without any such reason, skepticism follows. Preservationism, at least in its current form, is therefore unable to explain why our memory beliefs whose content differs from those that originally were encoded in our memory are justified.

The access internalist, on the other hand, avoids the problem. Since memory beliefs usually will be produced by a process with some associated phenomenology, or have some phenomenal property of themselves which is indicative of their source, the subject will usually be in a position to know what their memory beliefs rationally have going for them—which plausibly explains why they are (memorially) justified.³⁹ And if, for whatever reason, a subject has memory beliefs but without any of the relevant phenomenology, access internalism again plausibly explains why those beliefs intuitively won’t be memorially justified (assuming that the subject doesn’t have any other internally accessible ground for the beliefs). So access internalism isn’t only just as suited as preservationism when it comes to answering the explanatory challenges posed by the problem of stored belief and the problem of forgotten evidence, it also avoids important problems which preservationism is vulnerable to. For these reasons, I therefore claim that access internalism should replace preservationism as the standard view in the epistemology of memory.

6 Conclusion

Let me conclude by taking stock. In Sect. 2, I presented BonJour’s clairvoyance scenario and his argument for access internalism. In Sect. 3, I presented the overconfidence scenario and argued that anyone who denies that the memory belief of the subject in that scenario fails to be memorially justified is vulnerable to an indistinguishability problem. In Sect. 4, I used the intuition elicited by the overconfidence scenario as the basis for an abductive argument for access internalism about memorial justification. In Sect. 5, I argued that access internalism should replace preservationism as the standard view in the epistemology of memory, since (i) it is just as able to answer the explanatory challenges posed by the problem of stored belief and the problem of forgotten evidence, and (ii) it avoids a couple of problems which preservationism is vulnerable to.

Access internalism is an intuitively compelling thesis about the nature of memorial justification. By reflecting on the justificatory status of overconfident subjects who aren’t (even in principle) in a position to tell what their memory beliefs rationally have going for them, we can infer that there is an internal accessibility condition on memorial justification. However, this kind of view is naturally threatened by preservationism, which has become the dominant position in the epistemology of memory. Preservationists argue that memory preserves the justificatory status of the beliefs that enter it, and that this explains why most of our stored beliefs and the beliefs whose original evidence we have forgotten are justified. The purpose of this paper has been to challenge the orthodoxy by presenting a new scenario which provides the intuitive basis for argumentation which, on the one hand, supports access internalism, and, on the other hand, undermines preservationism. Moreover, the paper has also argued that preservationism, in contrast to access internalism, is vulnerable to skepticism about memorial justification, insofar as it relies on an outdated folk-theoretic conception of memory processing. The upshot of the paper is thus that a reorientation of the epistemology of memory is in order: the discipline should readjust itself by building its theoretical framework in coherence with the scientific work done on memory processing—and, as I have argued, access internalism provides a plausible framework for doing so.