1 About Time (and Place and Memory)
Saint Louis has great barbecue. I believe it. I also know it. But I am no perfect believer. I don’t know just anything I believe. Some things help metamorphose mere believing into knowing. One of those things is truth. Another – and more important for this Element – is epistemic justification. This is a kind of support I have, or a good quality, in my believing that this or that is true. This good quality springs from many wells. This Element is on how memory provides it. I have justification for my belief that Saint Louis has great barbecue. But I’m not eating that food at the moment. I’m not even near Saint Louis. I’m not reading a travel blog or listening to a friend rave about Missouri cuisine. It may be obvious to many that Saint Louis is the place for brisket and ribs. But it’s not obvious in the way that, obviously, barbecue is food. No support for my belief springs from my thinking about the ideas involved. So how could my belief have this justification? Somehow or other, memory is involved. But exactly how? This probes at something surprisingly complex. There are competing answers about many of the details, each with an air of truth.
I will support a package of views that will have most philosophers either raising their eyebrows or shaking their heads. And I claim that we find this support in the epistemology of memory – that is, in the philosophical study of how memory matters for knowledge and justification. I say the epistemology of memory suggests that one’s justification ultimately depends only on what one is mentally accessing at a time. We’ll see, in other words, that the epistemology of memory supports a time-slice variety of internalism about epistemic justification, specifically one that has an access condition – and we’ll see what these other words mean.
This section will take just the time-slice and internalism bits out of their packaging. Section 2 will hold the time-slice bit up to the light and argue for it. To start, let’s lay out some traditional and majority views in epistemology that contrast mine. One is that our justification from memory in the present directly depends, mainly or in part, on the past. Some call this a historical theory of justification (Goldman Reference Goldman and Pappas1979). This view says, for example, that my belief that Saint Louis has great barbecue depends for its justification directly on, say, my hearing in the past from an apparent connoisseur that Saint Louis has great barbecue, my believing that then as a result, and on my keeping this belief in memory.
But in what sense does the justification here directly depend on the past? To see this, it helps to see first how the past might matter only indirectly. Suppose you acquire at an earlier time some evidence that justifies your belief that p then. And your memory causes you to have this evidence, or something near enough to it, now. And this evidence you have now justifies your belief that p now. The past affects your present justification, sure, but merely by shaping your present evidence. Here, what’s directly determining your justification now is just your evidence now. This follows from a synchronic or time-slice view of justification. That view says, more generally, all that directly matters for justification at a time is how things are at that time (for convenience, I’ll often call that time “now” or “the present”).
A historical theory says, no, the past matters for more. It can directly matter. For example, consider a simple case of forgotten evidence: In the past you had evidence for believing that p, and you formed the belief that p on it. Since then, you’ve forgotten that evidence and learned nothing new for or against believing that p. Yet you still believe that p. A historical theory might say the status of your current belief that p is tied to how you originally acquired the belief. Its status now depends on the status it had from that forgotten evidence. So you can have justification for believing that p now even if you have forgotten your original evidence for p and have no support for p now. How your past was can make this so (Goldman Reference Goldman1999).Footnote 1
You might understand time-slice and historical views to take conflicting stances on either of two sorts of justification: Propositional or doxastic. Propositional justification is the justification a subject has for having a doxastic attitude toward (i.e., for believing, disbelieving, or suspending judgment in) a proposition. The subject may or may not have that attitude. The subject may have it, but for reasons unrelated to her justification. Still, having the attitude is justified for her. Doxastic justification is the justification of a doxastic attitude a subject has. If a subject’s belief that p is doxastically justified, she believes that p. Doxastic justification requires propositional justification. So, if a subject’s belief that p is justified, she has justification for believing that p. Her belief is justified because it is based on what provides the propositional justification.
And it’s plausible and generally agreed that this basing involves causation. This helps us see how to best understand what historical and time-slice views ultimately disagree about. Causation is diachronic. It involves temporal succession. If the basing involved in doxastic justification requires causation, and this causation involves temporal succession, then doxastic justification is not always settled by the non-relational facts at a time. The facts of earlier times matter too. Whether a subject’s belief that p is justified now isn’t just a matter of how things are now. It’s also a matter of whether she, say, had reason for believing that p, and whether that reason caused her to believe that p now.
I won’t take a stance here on whether there is a causal element of basing that involves temporal succession. Perhaps there is no such element. The causal connection in doxastic justification could be one of sustaining, a relation, which might be synchronic. But in case there is such an element, we can simply understand the time-slice view to be about propositional justification, and all parties can agree that doxastic justification is not settled by the time-slice.
Crosscutting the historical and time-slice distinction, we have another popular view: That one’s justification directly depends on contingent features of the world external to one’s mind. Differences in the world that make no mental difference can still make a justificatory difference. This view is externalism about epistemic justification. In contrast, internalism about epistemic justification says, at a minimum, that mentally identical subjects will be equally justified in having the same doxastic attitudes toward the same propositions. A subject’s justification is settled by her mental life.
I’ve called both externalism and historical views popular. A common way they join hands is in a process reliabilist account of justification from memory. Process reliabilism says, roughly, that a belief’s justification depends on the reliability of the process that formed or retained it (Goldman Reference Goldman and Pappas1979). Why is my belief that Saint Louis has great barbecue justified now? The answer has a few parts. It’s because I came to believe in the past, from a reliable process (namely, testimony from an apparent relevant expert), that Saint Louis has great barbecue. And so that belief had justification then. And that past justified belief enters a memory process that produces my belief with the same content now. And that memory process tends to produce true outputs, at least when its inputs are true. So, my belief that Saint Louis has great barbecue is justified now by memory.
On process reliabilism, the past and the external world both directly matter for justification in the present. The past matters in part because the past status of the belief that memory retains matters. The external world matters because it helps determine whether the process of retaining a belief through memory is, in fact, reliable.
At the risk of stating the obvious, I note we’re thinking about two variables or dimensions here: time and space. We’re thinking about the time dimension when asking: What directly matters for how I have justification now from memory? Is it just how things are now? Or do other times matter too? We’re thinking about the space dimension when asking: What contingent facts directly matter for how a subject has justification from memory? Is it just facts about the subject’s mental life? Or do facts about the external world matter too?
It’s no coincidence that both externalism and historical views poll well. The views seem to fit together nicely. But I think it’s the view about time that usually drives the view about space, and not the other way around. That is, if both views appeal to you, odds are not that you think the past matters because you’re an externalist. Odds are you’re an externalist because you think the past matters.
Maybe this claim sounds suspicious. Yet the work of the father of process reliabilism himself, Alvin Goldman, helps make it plain. Goldman (Reference Goldman and Dougherty2011) discusses a particular historical view about how memory justifies. He identifies it as something a theory of justification should make sense of; he claims it’s a mark in favor of a theory of justification if it, like reliabilism, explains how the past directly matters for our having justification now. And Goldman (Reference Goldman2009) appeals to that historical view to develop a case for reliabilism. It would appear that the historical view is more evident than process reliabilism itself. Historical considerations have a long history of motivating externalism.
This time and space combination, this pairing of a historical view with externalism, makes sense. Consider how uncomfortable the main alternative is for externalism. The main alternative is that the past does not directly matter for having justification from memory, yet the external world does. That is, a time-slice externalism. Advocating, say, a time-slice process reliabilism is no one’s idea of a good time. It says external conditions matter, in part because they matter for reliability. But the view says the subject’s past, the history of her doxastic attitudes, their past epistemic status, and the past track record of a belief-formation process, do not directly matter for justification. External past conditions don’t directly bear on present justification. They don’t affect reliability. Reliability supervenes on a set of facts just about present conditions.
On a time-slice process reliabilism, a process’s reliability might undergo dramatic changes, even over short intervals. Reliability could change as the set of facts that determine reliability changes from moment to moment. A process’s reliability is at risk of radical instability. Time-slice process reliabilism is as uninviting as it gets.
Yet there’s reason to bring it up. Suppose the time consideration indeed drives the space consideration when externalism and a historical view team up. In that case, making it plausible that the past does not matter will take the wind from externalism’s sail. If historical views are wrong, externalism’s best hope is a time-slice view. But time-slice externalism is understandably unpopular.
In the next section I offer a new argument that bottles up this wind. I deny that the past directly matters for how memory provides justification. That argument concerns the time dimension for theories of justification. If the argument succeeds, it also reduces support for externalism, the popular option on the space dimension.
2 Trying Times for Memory Justification
History is bunk, as far as justification from memory in the present is concerned. But our journey to this conclusion has an unexpected beginning. I have canvased just two options for the time dimension in how memory provides justification now. There are others. We’ll look at one that offers a plot twist: The future. Section 2.1 claims the future does not matter for justification from memory. Section 2.2 explains that there is no arbitrary asymmetry in what matters for justification. And Section 2.3 argues there is no relevant difference between historical and future considerations. Section 2.4 reinforces this argument with a case. From these sections, it follows that the past does not matter for justification from memory.
First, we’ll look more closely at historical views. They have shown up in two forms. One says that our justification from memory in the present depends primarily on how the past was. Motivating this idea is what I call anti-generativism, which places a limit on memory: It can’t introduce justification.Footnote 2 It can at best retain justification from the past. Whatever justification we have from it now has simply been passed down to us.
These ideas used to be popular. But they have been on the retreat in recent decades, particularly in light of a recent, growing consensus about how human memory works: Memory reconstructs and generates new information as a part of its normal operations, and much more often than we had thought (Michaelian Reference Michaelian2016). And if normally functioning memory often generates new information, there’s a case to be made that we often have justification for believing that information. And so it looks like memory’s justificatory role is not primarily one of retention. It looks more and more plausible that memory can and often does generate justification.
A more temperate historical view says that justification from memory in the present depends in part on how the past was. This view remains popular. And its most popular precisification is one of the views that in the literature goes by the name “preservationism”. Preservationism says that if a subject believes that p with justification at one time, and her belief that p stays in memory later, then her belief that p is prima facie justified at the later time, thanks to memory.Footnote 3 (Justification that is prima facie is some support; having reason to doubt the belief or the justification can prevent the belief from being justified overall). This is just a sufficient condition for having justified belief from memory. Preservationism is not claiming that just the past matters. And it is not claiming that certain conditions in the past are necessary for justification from memory in the present.
Preservationism says the past directly makes some difference to the justification we have now. Some beliefs can in a sense inherit justification. Your belief that p might have justification now in part because your belief that p had it in the past. Someone otherwise just like you now may not have justification for her belief that p now, ultimately because she had a different past. Preservationism faces its share of problems (BonJour Reference BonJour, McLaughlin and Kornblith2016; Frise Reference Frise2017; Kelly Reference Kelly, Kornblith and McLaughlin2016). Still, no other historical theory is as appealing and developed.
Maybe that last claim seems too strong. What about process reliabilism? It’s clearly appealing and well-developed. What’s more, more philosophers identify as reliabilists than as preservationists. These observations are correct. However, reliabilism can be understood as a version of preservationism; reliabilism implies preservationism, provided the plausible assumption that memory tends to produce true outputs when its inputs are true.Footnote 4 Similarly, perhaps epistemic conservatism seems more appealing than preservationism. Epistemic conservatism says, roughly, that if S believes that p, then S is prima facie justified in retaining her belief that p (McCain Reference McCain2008, Reference McCain, Carter and Bondy2019; McGrath Reference McGrath2007, Reference McGrath, McLaughlin and Kornblith2016). And epistemic conservatism is well-developed. However, this theory can be understood as yet another version of preservationism and implies preservationism. If retained beliefs of any stripe have justification, then retained beliefs that had justification have justification.
Preservationism is a flagship historical theory. Yet even some of its critics are nonetheless optimistic about historical theories. For example, Thomas Kelly (Reference Kelly, Kornblith and McLaughlin2016) objects to preservationism. But he offers a pair of cases that he (2016: 63) says “suggest that some historical theory or other must be correct.” Reviewing these cases will help set the stage for my argument for the time-slice theory.
The cases are of temporally extended reasoning, involving multiple inferences. As such, they involve memory too (Burge Reference Burge1993). The subject in each case is calculating how much to tip at a restaurant. Both subjects are trying to calculate 20 percent of the bill. So each first calculates what 10 percent of the bill is. Then, each doubles that amount. In each case, the subject starts from a justified belief about the bill. One subject begins from the justified belief that the bill is $15.50, and eventually deduces competently from it that p: 20 percent of the bill is $3.10. The other subject begins from the justified belief that the bill is $13.50 but, due to a nontrivial error, deduces that p as well. At the time of deducing that p, each subject forgets the amount of the original bill.
Kelly calls the case with no error the “Good Case” and the case with the error the “Bad Case”. For reasons that will become apparent, I’ll call them the “Good Past Case” and the “Bad Past Case”. At the end of their reasoning the subjects in these cases share a “time-slice profile” – they are psychologically identical in all relevant ways at that moment. But Kelly thinks only the subject in the Good Past Case has justification for believing that p. The subject who reasoned erroneously does not have justification for believing that p.Footnote 5
Kelly thinks you need to look at the subjects’s histories, not just their time-slices, to see what they are justified in believing at a moment. A time-slice theory would get things wrong. Since the subjects are alike after their reasoning, the time-slice theory says – incorrectly in Kelly’s view – that they are the same with respect to justification. However, the subject in the Good Past Case has something in her past that suffices for now having justification for believing what she’s calculated the tip to be. Kelly (Reference Kelly, Kornblith and McLaughlin2016: 57) says: “That is, when I reason impeccably from a belief that I’m justified in holding by deploying my mathematical competence, this is sufficient for my being justified in holding the resulting belief, at least so long as I do not acquire any additional evidence in the interim.”
The subject in the Bad Past Case lacks that something in her past that suffices for having justification later. Presumably, she lacks anything else in her past that suffices for having justification later. And, presumably, there’s nothing about how the subjects are when sharing a “time-slice profile” – at the time of completing their deductions – that suffices for their having justification. We find that only in the past of one of the subjects. So, strictly speaking, the error in the past in the Bad Past Case is not itself what explains why the subject lacks justification later. Rather, she lacks it because of the absence of anything that suffices for justification.
Crucially, Kelly doesn’t describe either subject’s phenomenology – what the subject’s experience is like – at the time of completing the reasoning. What matters for him is just that the subject in the Good Past Case alone had a certain sequence of thoughts or performed a sequence of mental actions in the right order, with a justified belief as a starting point.
The omission about the subjects’s phenomenology may be surprising. But it, in a way, helps these cases present a challenge for the time-slice view. The subjects have to be identical phenomenologically at the time we’re evaluating, if they’re to share a psychological profile. And their relevant phenomenology must be minimal. Any evidentially relevant memorial or inferential phenomenology in the case would help the time-slice theory. If, say, p seems true to the subject, or p seems to the subject to follow from thinking of hers that she recalls as true, or p feels familiar to her, then the time-slice theory can cite that as what accounts for the justification in the Good Past Case. Then the time-slice theory could identify that same thing as sufficing for having justification in the Bad Past Case, since the subjects are psychologically alike. The time-slice theory is vindicated. It implies that the subjects in the two cases are justificationally alike. And we can point to something, in each case, that justifies. Minimal relevant phenomenology helps prevent this.
In short, Kelly concludes some historical theory or other must be correct because he thinks all time-slice theories fail. They get his cases wrong. Alvin Goldman reasons similarly when trying to support a version of preservationism (which he calls “PM”). He says:
Consider the consequences of rejecting (PM) and its ilk. If (PM) is rejected, there will be an epistemic phenomenon that I shall call continual clearing of the justificational slate (CCJS). There will be no carry over of a belief’s J-status from one moment to the next … If CCJS holds – which is implied by the rejection of (PM) – it will have catastrophic results for the number of beliefs people are justified in holding at any time.
It isn’t quite right that rejecting this or any version of preservationism leads to a time-slice theory. After all, some historical theory other than preservationism could still be true. But perhaps the point is that denying historical theories more generally implies a time-slice view. And (so the reasoning goes) that’s a flawed view.
These criticisms focus on the justification of beliefs at different phases. Goldman looks at beliefs we’re keeping over time. Kelly looks at beliefs we’re forming over time, through extended reasoning. Goldman thinks the past directly matters because memory can preserve justification from the past. Kelly thinks the past directly matters because reasoning and memory can generate justification over time from the past. The criticism in each case is that a time-slice view cannot explain this justification. So reject the time-slice, embrace the historical.
2.1 Looking Ahead
But this reaches for the chips before all hands have been played. Suppose these challenges are genuine, and the time-slice theory has no way past them. For one thing, historical theories may be just as badly off. That is, we’ve seen alleged problems for time-slice views. Still, that’s no positive case for historical views, not until it’s also plausible that they face no comparably bad problem of their own. Are time-slice theories worse? That’s what we need to see. And so far it’s at best unclear.
For another thing, let’s suppose time-slice theories clearly are worse. Still, it doesn’t follow that some historical theory or other must be correct. We’ve been thinking about the time dimension, about what matters for how we have justification now from memory. And there’s an option here that we haven’t mentioned yet: What if this justification depends on how the future is?
I noted that historical views have said the past matters primarily, or matters in part. If we think analogously about options on which the future matters, even the “in part” version immediately makes us wince. Very roughly, the view would be: Some facts about the future partially and directly determine the epistemic justification a subject has (from memory) for believing that p at t. Call this view futurism.
One potential refinement of futurism is a near analog to preservationism: If S will justifiedly believe that p, then S now has prima facie justification for believing that p (thanks to memory). And this is absurd. It’s unreasonable for you to believe now what you have yet to learn and what you may have never yet had a chance to learn. Suppose tomorrow you will learn something new about a friend. Still, you have no justification for believing that now, regardless of your future! You haven’t learned it yet. Maybe there’s a respectable view about how the justification we now have from memory depends in part on the future. If there is not, this justification does not depend primarily on the future either.
Let’s appreciate just how utterly dismissive we are – and maybe rightly – of futurism. Here’s an anecdote. During a talk at a recent major epistemology conference, after I introduced futurism, I asked if anyone present endorsed it. One in forty philosophers raised a hand. The rest looked mildly appalled. Here’s something more telling. When I introduced the time-slice view earlier in this Element, I suggested it is interchangeable with a synchronic view of justification. The standard contrast for synchronic is diachronic, which looks at matters over time. But many epistemologists, when they (e.g., Naylor Reference Naylor2015; Senor Reference Senor, Bernecker and Michaelian2017; Swinburne Reference Swinburne and Dougherty2011) elaborate on this contrast, explain diachronic theories exclusively as theories on which history or the past directly matters for justification in the present. A diachronic view is simply a historical view. That’s it.
It’s better not to conflate diachronic with historical views, however, as there’s more than one way a theory of justification could be diachronic. For example, it could say the future directly matters for justification in the present. Backward in time from the future is still diachronic – it looks at matters over time, through time, not just at one time. Futurism, in other words, is a diachronic view. But I wager that, like many epistemologists, you too are tempted to think of diachronic theories as historical exclusively. That’s how ridiculous futurism is to us. In other words, we accept, and we think reasonably so, what I call No Futurism – the denial of futurism.
So, why bring up futurism at all, when it immediately looks like a dead end? Because, I claim, the implausibility of the future mattering is evidence that the past doesn’t matter either. If you get to reject being diachronic in one way, you’ve got to reject being diachronic in the other way too. The next section develops this line of reasoning. No Futurism is its first premise.
Perhaps this forewarning leaves you wondering if futurism isn’t so bad after all. If futurism’s irrelevance will be weaponized against historical views, maybe we can scrape together some reason to accept futurism, thereby rescuing historical views? Maybe, despite its initial appeal, No Futurism is false.
However, the best cases for futurism are not promising. Here’s one sort of case. Suppose someone asks you a question about something you once learned. You knew the answer before. But when asked now, you don’t think of it, despite your attempts. An epistemologist might wonder with some interest: While in this state, do you know the answer, or have justification for believing it? And it might appear that the best response to this involves the future. If, for example, you bring the answer to mind in the future without having relearned it, then it might seem plausible that you have justification for believing it now. If instead you never bring it to mind again, it might seem plausible that you do not have justification for believing it now. And here we are looking at the future to determine the epistemic goods of the present. We have a case for futurism. (For a lengthier discussion of a similar case, see Section 4.4.1).
This is not a reason to adopt futurism. The future itself is not determining the epistemic goods of the present. Rather, our current understanding of the future provides us with evidence. It’s helping us discern which epistemic goods are present in the present. Seeing what unfolds in the future helps us figure out what the subject is like now – whether and to what extent the information is mentally inaccessible to the subject now. It’s how the subject is now, though, that really matters for her justification now. Has she indeed forgotten the information now? Certain futures shed light on this without themselves directly mattering for justification. There is no case for futurism if this heuristic explanation is viable. And the heuristic explanation is simpler.
Here’s another case for futurism. If process reliabilism is true, justification depends on how well belief-formation process types tend to perform. And one measure of this is the process’s ratio of true-to-false outputs throughout time, including the future. So, facts about the future performance of a process determine its justificatory power now. If a process tends strongly enough to have true outputs over all time, including in the future, it justifies now.
A key step in this case is independently doubtful, however. It’s the claim that a process’s future outputs help settle its justificatory power now. The view has bad consequences for reliabilism (Frise Reference Frise2018b). For example, consider a belief-forming process that has tended to yield true beliefs in observed cases, such as a basic perceptual process. It is common sense that its belief outputs are justified. But if certain unfortunate futures occur, that process may in fact have a poor ratio of true-to-false outputs throughout all time. On reliabilism, the belief outputs of that process are unjustified even in the present. Similarly, if certain fortunate futures occur, a belief-forming process that has tended to yield false beliefs in observed cases will have a good enough ratio of true-to-false outputs throughout all time. On reliabilism, then, a paradigmatically bad belief-formation process can turn out to be a source of justification now, on account of the future.
Also, if it’s reasonable to suppose a process’s future outputs matter for its justificatory power now, strange inferences become reasonable on reliabilism. Take some beliefs we reasonably believe are justified. If the future matters for a process’s reliability, then in order for those beliefs to be justified, no future will occur that will compromise the reliability of the process that formed them. Recognizing this, on reliabilism we can then reasonably infer that certain futures won’t occur. Now, it may be independently plausible that those futures won’t occur. But it’d be bizarre to learn what the future won’t look like, simply by looking at a theory of justification and at which beliefs we reasonably believe are justified!
Reliabilism fares poorly if a process’s future performance matters for its present justificatory power. It has several unpromising implications. Futurism finds no help here.
We’ve seen No Futurism is plausible, and we’ve seen no plausible case for futurism. I’ll discuss another potential case for futurism in Section 2.4.4. For now I turn to supporting the next premise in my argument against historical views.
2.2 Against Asymmetry
The implausibility of the future mattering is evidence that the past doesn’t matter either. To support this, I propose a principle. Roughly, it’s that there is no arbitrary asymmetry in what matters for justification. If there is an asymmetry in what matters, there is an explanation for it. A bit more precisely:
No Arbitrary Asymmetry. If X-facts directly matter for whether a subject has or lacks justification for believing that p at t, and Y-facts do not, then something about X- or Y-facts (at t) accounts for this difference.
It could be that one kind of fact matters for justification, but another does not. When that is the case, something makes sense of this. Otherwise, it is arbitrary, unmotivated, and implausible, that just one kind of fact matters. The above statement of No Arbitrary Asymmetry could be tightened up, but is serviceable.
Internalists and externalists will both find something like this principle appealing. They might even find it implicit in their thinking. Let X-facts be about a subject’s mental life. Let Y-facts be contingent facts about the external world. The internalist applies No Arbitrary Asymmetry, and then goes on to tell us what she thinks the important difference between the mental and external is. Something explains why just the mental facts matter. The externalist applies No Arbitrary Asymmetry in something like a modus tollens. He denies that there is a relevant difference between the mental and the external. (How a process tends to perform, for example, involves facts about the external world, and is no less important than facts about the subject’s mind). So, the externalist denies that only the mental matters.
No Arbitrary Asymmetry is intuitive. And no reason to deny it creeps in our way. Here’s why the principle matters. Let X-facts be historical facts. Let Y-facts be future facts. We agree it’s absurd that the future directly matters for justification from memory now. But I don’t see why the past is relevantly different from the future here. We agree that how things are, going in one temporal direction from now – the future – does not directly matter for justification from memory now. So why would how things are, going the other temporal direction – the past – directly matter? Put another way: If the past matters, why doesn’t the future matter also? If you accept a historical theory like preservationism, on which you in effect inherit justification from the past, why can’t you also borrow it from the future, take it on loan? Why is only one non-present time a creditor?
An example may put a sharper point on these questions. In the Good Past and Bad Past Cases, two time-slice identical subjects are arriving at the same conclusion, from different chains of reasoning. The subjects have different pasts. According to the historical view, the subject with the good past has justification for believing her conclusion because of that good past. The subject with the bad past does not have justification for believing her conclusion, because of what her past lacks.
Let’s rewind the tapes on these cases. And let’s ask a new question. Look at the subjects, not as they’ve reached the end of their reasoning, but as they are just beginning their reasoning. One subject will go on to competently deduce a conclusion from a justified belief. Here we have, not the Good Past Case, but the Good Future Case. (It’s the same case, but with the spotlight on a different moment of the timeline). The other subject will go on to make a nontrivial error when deducing a conclusion from a justified belief. Here we have, not the Bad Past Case, but the Bad Future Case. Is either subject justified in believing the conclusion they will deduce, before they’ve derived it, and because of their futures? I suspect you think: No. I suspect the typical historical theorist thinks: No. Yet on a historical view like Kelly’s, in the Good Past Case, the subject has justification. So why not in the Good Future Case too? Why doesn’t the subject’s future matter, if her past does matter? A good answer is not obvious.
On my premises, we can tidily account for all four cases – the Good Past, Bad Past, Good Future, and Bad Future Cases. And the account fits well with the time-slice theory. There’s no relevant difference between the past and the future. And the future clearly doesn’t matter. So, the subjects in the Good and Bad Future Cases are equally justified in believing the same propositions, prior to their deductions, despite their different futures. Given No Arbitrary Asymmetry, the past doesn’t matter either. So, the subjects in the Good and Bad Past Cases are equally justified in believing the same propositions, after their deductions, despite their different pasts. You’d expect precisely this, on the time-slice theory.
2.3 Back to the Future
The remaining sticking point is whether the past and future indeed differ in no relevant way. Right now you’re likely enumerating differences. Some might seem to explain why only the past directly matters for justification in the present. In the subsections 2.3.1, 2.3.2 and 2.3.3, I’ll address the three potentially relevant differences that are the most natural to inspect. I’ll argue it’s at best unclear that they are really up for the explanatory job. The arguments will be neutral on whether internalism or externalism is true, and will assume nothing controversial about the metaphysics of the future. This supports:
No Relevant Difference. Nothing about historical or future facts relative to t accounts for a difference in which facts directly matter for whether a subject has justification for believing that p at t.
That is, there is no relevant difference between historical and future facts regarding whether a subject has justification for believing that p in the present.
No Futurism, No Arbitrary Asymmetry, and No Relevant Difference together imply that times other than the present don’t directly matter for justification. But the present does matter. So, only the present matters. This is my main argument for the time-slice view.
2.3.1 Relevant Difference: Openness?
No Relevant Difference concerns facts. The “facts” apparatus is clunky. I’ll mostly drop it, putting things more loosely. The first potentially relevant difference is this: Only the future hasn’t happened yet. Or: there’s an openness to the future alone. Some difference like this explains why the past matters for justification and the future does not.
A perhaps flippant reply notes that a difference this really gives no explanation at all. It does nothing more than tell us what the future is. The relevant difference is that just the future hasn’t happened yet? The relevant difference might as well be that only the future is the future!
A perhaps more helpful reply clarifies: Remember, we aren’t simply thinking about the future relative to our time. When imagining the Good Future Case and the Bad Future Case, we can suppose these cases are entirely in our past. We need not deny that the future is open. The important parts of the future, relative to those cases, have happened, relative to us.
Let’s focus on the outset of each subject’s reasoning in the Good Future Case and the Bad Future Case. Does each subject’s future reasoning – which is in our past – matter for the justification they have at the outset of their reasoning? The answer seems clearly: No. But the subject’s future reasoning is by now no more open than anything in her past. Whether the future has occurred yet is more a distraction than an explanation. The alleged difference in which events matter for a subject’s justification at a time comes down merely to their chronology in the subject’s life. We must hear why the events on one side of a moment, but not those on the other side, matter for justification at that moment. Maybe there’s something special about the events on one side. But we haven’t seen what that is. Instead, we’ve simply reiterated how the sides are distinct.
2.3.2 Relevant Difference: Causation?
The next candidate explanation might correct this. It attempts to identify what’s special about the past. The second potentially relevant difference is this: Particular causal relations obtain only between the past and the present, and memory is part of this causal connection. For example, a belief in the past can stand in a particular causal or sustaining relation with a belief in the present, thanks to memory. Past believing, thanks to memory, can causally contribute to present believing. And, the explanation goes, this causal connection is a justificatory corridor. Justification can travel along or be produced by it. So, some past justified believing can help justify some present believing. Past justified believing can help explain why believing in the present is justified (Goldman Reference Goldman2009, Reference Goldman and Dougherty2011).
Nearly all of this could be correct. It just fails to identify a relevant difference between the past and present. It attributes a notable relational feature to the past, but does not establish an asymmetry. The future might have the feature too. And in fact, it does. A belief in the present can stand in the same particular causal or sustaining relation with a belief in the future. Present believing, thanks to memory, can causally contribute to future believing. If this causal connection is a justificatory corridor, one might think, then future justified believing should be able to travel along it too, into the present. Traffic can flow along a corridor in more than one direction.
Take a case in which you believe that p at time t1, and it causally contributes to your believing that p at t3. Your believing that p at t3 is overdetermined, however. You will gain strong evidence for p that will cause you to believe that p at t2. And your believing that p at t2 will cause you to believe that p at t3, independently of your believing it at t1. You will believe that p with justification at t3. And if a causal connection is a justificatory corridor, your justification from t3 should make a justificatory difference at t1. Your believing that p at t1 should count as justified, directly on account of your future justified believing. After all, there’s a causal connection between believing at those times.
We might also question whether a causal connection is a justificatory corridor. But it’s enough to note that the second potentially relevant difference points out no difference at all between the past and future.
2.3.3 Relevant Difference: Causal Direction?
Perhaps that assessment was obtuse. We skirt it with a more careful identification of the potentially relevant difference. It’s not just that the past and present stand in certain causal or sustaining relations. It’s also that they occupy specific places in these relations, and it’s the nature of these relations, of which memory is a part. A past justified belief can cause a present belief, but not vice versa. And justification travels down or generates from that corridor only in the direction of the causation. So only the present belief can end up justified. But when a present belief is causally sufficient for a future belief, it doesn’t matter if the future belief is justified. The present belief can cause the future belief, but not vice versa. And justification can’t travel or generate against the one-way causal direction. This is why the past but not the future matters for justification from memory in the present.
This is the most promising difference we’ve seen. Here is one objection. Section 2.4 covers a second that merits more ink.
The alleged relevant difference between the future and the past is that only the past can make a causal difference to the present. This is supposed to be why the past and not the future directly matters for justification in the present. The first objection is that we have not heard a case yet for accepting that this difference is relevant, and no case is obvious. We still need to hear more about the magic of causation, such that justification travels at all in its direction, and in no other. Why suppose this is so? It’s unclear why the past’s position in the causal relation makes sense of how it can directly matter for justification, in a way that the future can’t. We need a positive case for accepting the difference as relevant. If causal history matters, why not also causal destiny?
Think analogously with value and identity. Maybe a pen can have special value now because of its past – Abraham Lincoln used it, say, to sign the Emancipation Proclamation. If it can have value for that reason, why can’t it also have value before the signing, because of what Lincoln will go on to use it for? A persistence relation connects the pen across the times before, during, and after the signing. But where an object is across the times in that relation does not obviously explain the direction that value can and cannot travel.
2.4 Causation Is Irrelevant
The past bears a particular causal connection to the present. The first worry in Section 2.3.3 was this: This connection doesn’t obviously make sense of how the past can directly matter for justification in a way that the future can’t. A bigger worry, and the focus of this section, is this: There is reason to deny the causal connection matters for justification.
The causal connection seems significant when we consider how a belief comes about and is retained. A belief formed and kept well looks better justified than one formed or kept poorly. And the past matters for whether a belief is caused or causally sustained well. However, this directs our attention to doxastic justification above all. And that masks the mystery.
As noted before, a traditional view is that doxastic justification requires propositional justification and proper basing. And it’s plausible that this basing involves causation – whatever propositionally justifies believing helps cause belief that is doxastically justified. So it’s appropriate to cite causal relations to the past when making sense of one component of doxastic justification: basing.
But what about that other component – propositional justification? How does bearing a particular causal relation to the past matter for a subject’s present propositional justification? To focus on this, let’s eliminate the causal element that goes hand in hand with justified belief. I claim that when we do so, we strip away a veneer that hid causation’s irrelevance to propositional justification.
The irrelevance is straightforward on certain historical theories. Recall epistemic conservatism, which says approximately that if a subject S believes that p, then S is prima facie justified in retaining her belief that p. This being justified in retaining is an “ex ante status,” or a matter of having propositional justification (McGrath Reference McGrath, McLaughlin and Kornblith2016: 77). And what the subject has justification for – that is, retaining – appears to itself have a causal element. But what’s doing the justifying does not. Simply having a doxastic attitude at time x is sufficient for having propositional justification at time y. Satisfying epistemic conservatism’s antecedent does not require anything at x to cause anything at y. It’s not even obvious why x must precede y. On epistemic conservatism, there is no principled asymmetry between the past and future, such that only the past can propositionally justify. Maybe future believing also suffices for retaining belief earlier.
Causation’s irrelevance to propositional justification hides better on other historical theories. Smoking it out will take some work.
Memory helps causally connect times. A complication, however, is that over time we forget. This forgetting appears to matter to some of our justification. A subject who has propositional justification in the past may not have it later, even when she hasn’t gone on to learn anything that diminishes her justification. This may be because of what she has forgotten.
Of course, forgetting does not always wipe away justification. But usually, when epistemologists think a subject’s justification survives her forgetting, they have in mind a specific kind of forgetting case. The case is one where the subject retains a belief she originally formed with justification, on the basis of evidence that she has forgotten (Goldman Reference Goldman1999; Harman Reference Harman1986). Here memory, as Goldman claims above, preserves the belief’s past justified status.
Let’s look at a different kind of forgetting case, one that makes the headlines less often. Here, as before, the subject forgets her original evidence that helped justify believing that p for her. But here, the subject never formed belief that p. Supposing memory preserves a belief’s justified status does not tell us whether memory preserves justification in the absence of belief. And we should not simply assume that a subject’s justification remains only when her doxastic attitude does. That would get the explanatory relation backward, on the traditional view of doxastic and propositional justification. Keeping propositional justification explains why the kept attitude might be doxastically justified. Retaining the attitude should not fully explain the retention of propositional justification. Yet historical views tend to discuss only doxastic justification, or tend to discuss a subject’s propositional justification for believing that p only when her belief that p is doxastically justified. Historical views, then, can have unclear implications about whether and how memory or a causal connection to the past contributes to having propositional justification specifically.
So we’ll focus on propositional justification in a case where the subject lacks a belief that p, while she is justified in believing that p. Rather than preserve belief, memory could help generate it. Recall the Good Past Case. Kelly says that at the time of completing a competent deduction of the tip amount over several steps, the subject has propositional justification for believing that 20 percent of the bill is $3.10. As we’ve seen in this case, the subject generates a belief that is allegedly justified.
Imagine a case otherwise like the Good Past Case, but in which the subject does not form a belief after reasoning about the tip amount. So she has no doxastic justification here, as she does not form the doxastic attitude she’s justified in having. Call this a stalled generation case.
In a stalled generation case, at the end of her reasoning, the subject has propositional justification for believing that 20 percent of the bill is $3.10. How long afterward, or in what circumstances, does she continue to have it? Does having a good past matter indefinitely? Presumably not. But why not?
Remember Kelly (Reference Kelly, Kornblith and McLaughlin2016: 57) says: “… when I reason impeccably from a belief that I’m justified in holding by deploying my mathematical competence, this is sufficient for my being justified in holding the resulting belief … .” But it’s possible to reason impeccably from a justified belief, deploying mathematical competence, without forming belief in what you reason to. So what’s the expiration date on your having justification for believing what you reason to? In a stalled generation case, at what point does the subject cease to have propositional justification, all else being equal?
In the following subsections, I’ll consider and reject a few answers to this question. Without a good answer, we see no relevant difference between historical and future facts; No Relevant Difference is confirmed. Allegedly, the past matters for present justification, and the future does not, and this is due to the past’s unique causal connection to the present. But this causal account leads to problems in stalled generation cases. These cases are crucibles into which we should throw our theories about memory and justification. Perhaps some theories have hung around so long in the literature only because these cases are also scandalously underexplored.
2.4.1 Ex Ante Justification
Or maybe that’s all hokum and hype. I’ve been working with the traditional view. It says propositional justification is more fundamental than doxastic justification. Perhaps that’s the root of any problem here. Perhaps a better view is that doxastic justification is more fundamental. And perhaps, on that better view, historical accounts have sensible things to say about stalled generation cases. Let’s look at some theories that would subvert tradition and see if they help.
Goldman’s (Reference Goldman and Pappas1979) seminal work on process reliabilism and historical views suggests doxastic justification is more fundamental. John Turri’s (Reference Turri2010) theory of propositional justification develops Goldman’s work while implying neither process reliabilism nor a historical view. It’s neutral across a few categories we’ve discussed so far. His (Reference Turri2010: 320) theory is:
(PJ) Necessarily, for all S, p, and t, if p is propositionally justified for S at t, then p is propositionally justified for S at t BECAUSE S currently possesses at least one means of coming to believe p such that, were S to believe p in one of those ways, S’s belief would thereby be doxastically justified.
I suspect “currently” is also picking out t. If it is, the idea is that whenever a subject has propositional justification for believing that p, it is because at that time she has a way of forming or keeping belief that p, and if she believed in that way it would be doxastically justified. On PJ, it looks like memory could help provide propositional justification in a stalled generation case. It could be thanks to memory that the subject has the means to form doxastically justified belief. And having that means helps explain her having propositional justification. In other stalled generation cases, the subject lacks the relevant means, and so she lacks propositional justification.
Still, PJ does not even half-answer our main question here. PJ states just a necessary condition for having propositional justification. It gives no sufficient condition. So PJ does not teach us when a subject has or lacks propositional justification in a stalled generation case. It only informs us of one situation in which the subject lacks it, namely, when she lacks the means to form doxastically justified belief. There could be other situations when she lacks the justification. And PJ says nothing about exactly when she has the justification. It just explains why she has it when she does.
PJ’s flaw is a departure from its source material. The best hope may be a return home to Goldman’s (Reference Goldman and Pappas1979). Propositional and doxastic justification more or less match what he calls “ex ante” and “ex post” justification, respectively. Goldman (Reference Goldman and Pappas1979: 21) says:
[EAJ] Person S is ex ante justified in believing p at t if and only if there is a reliable belief-forming operation available to S which is such that if S applied that operation to his total cognitive state at t, S would believe p at t-plus-delta (for a suitably small delta) and that belief would be ex post justified.
Ex Ante Justification (EAJ) says just when a subject has propositional justification. Roughly, it’s when she can apply a reliable process that would result in doxastically justified belief. EAJ has seen many revisions over the years.Footnote 6 But the changes do not matter for our purposes. Doxastic justification remains fundamental. Whether a subject has propositional justification in a stalled generation case depends on further details. As the case parallels the Good Past Case, it depends on whether a reliable process is available to the subject, which, once applied, leads to a doxastically justified belief that 20 percent of the bill is $3.10. If she has propositional justification, the right sort of process is available to her. Memory likely makes the process available or is itself part of the process.
But according to Goldman, the past matters too. Right after stating EAJ, he (Reference Goldman and Pappas1979: 21) clarifies: “For the analysans of [EAJ] to be satisfied, the total cognitive state at t must have a suitable causal ancestry. Hence [EAJ] is implicitly an Historical Account of ex ante justification.” I suspect the subject’s total cognitive state has the right causal ancestry in a simple stalled generation case, whatever such a state and ancestry amount to. After all, the subject reasoned well, and there was no funny business. The reasoning just hasn’t yielded belief. So EAJ, a historical account, explains the conditions in which the subject in a stalled generation case has propositional justification. And the conditions explicitly and critically involve a causal connection. If erring in reasoning earlier has shaped the subject mentally now in something like a stalled generation case, her bad causal past prevents her from satisfying EAJ now. She would not have propositional justification.
2.4.2 Still No Relevant Difference
EAJ is not the hero historical theories need. Historical theories need to explain stalled generation cases in such a way that the past but not the future directly matters for how memory provides justification. A major issue keeps EAJ from being such an explanation.Footnote 7
What does EAJ look at when it implies a subject has or lacks propositional justification at a time? EAJ looks at how things are at the time. For example, whether the subject has the right sort of belief-forming operation available, and what the subject’s total cognitive state is then. And, EAJ looks at the past. For example, how that state came about, and which cognitive states preceded it. However, EAJ looks at more.
Remember, for an available belief-forming operation to be the right sort, on EAJ, the following is true: If it’s applied to the subject’s total cognitive state at the time, she would have doxastically justified belief. She would have this belief “at t-plus-delta (for a suitably small delta)” – that is, shortly after applying the belief-forming operation. That is, potentially, in the future. What the subject will go on to believe with doxastic justification matters for whether the subject has propositional justification now. On EAJ it could be that a subject has propositional justification in the present because of what and how she believes in the future.
EAJ implies futurism.
Futurism says facts about the future partially, directly determine the justification a subject has (from memory) for believing that p at t. And on EAJ, this is so. Certain futures directly determine the justification the subject has in the present. It can be that the subject has available a belief-forming operation that, if applied to her total cognitive state at the time, would soon after result in doxastically justified belief. And (as I will show in Section 2.4.3), what makes it true that the belief-forming operation would do this can be that it in fact will do this. Moreover, this belief-forming operation can involve, or be available because of, memory. So EAJ does not just imply that what happens in the future makes a difference for propositional justification in the present. It also implies that what memory does in the future can make this difference in the present.
On EAJ, even the subject in the Good Future Case has propositional justification. Recall that this is simply a case in which the subject is about to competently infer from a justified belief to the conclusion that 20 percent of the bill is $3.10. EAJ says that, before completing this inference, the subject has justification for believing this conclusion. This is because she has the right past and has available the right sort of process that, when applied, would yield justified belief in the conclusion.Footnote 8
Because EAJ implies futurism, it cannot help historical theories here. Yes, on EAJ, the causal connection to the past can make a difference to justification in the present. But we are looking for an account that does more. It must also make sense of why only the past matters, why the future does not also matter. This is where EAJ, by implying futurism, fails.
The problem, in other words, is that EAJ implies No Relevant Difference, the very premise EAJ was recruited to overthrow. Remember, No Relevant Difference says: Nothing about historical or future facts relative to t accounts for a difference in which facts directly matter for whether a subject has justification for believing that p at t., EAJ, however, implies that future facts directly matter too. There’s no difference in which facts matter. So there’s no difference between the facts that accounts for their mattering differently.
Another wrinkle to the problem is that EAJ denies that we’ve found the relevant difference between the past and future. We’ve been trying out the idea that the past’s special causal connection to the present explains why it and not the future makes a difference to justification in the present. This idea is mistaken if EAJ is right. There is nothing to explain here, so the past’s special connection does not do this explaining.
EAJ is no menace to No Relevant Difference. At best, EAJ challenges a distinct premise in my main argument for a time-slice view: No Futurism. EAJ doesn’t make the future paramount, but does make it matter. The dose is light, but the drink is still forbidden. The future doesn’t matter. This was a laughably secure datum our inquiry proceeded from.
The next subsections try to exonerate EAJ. The subsections after explore other potential historical responses to stalled generation cases.
2.4.3 Wait, Does EAJ Really Imply Futurism?
Maybe it’s not obvious that EAJ implies futurism. Here’s why EAJ might seem innocent: It’s really just a kind of counterfactual theory, a theory about how things would be if this or that were so. It says a subject has propositional justification just when she has available to her a belief-forming operation that is both reliable and has a certain counterfactual feature. The feature is this: If the operation were applied, it would lead to belief and the belief would be justified. Whether the operation has this feature is determined by what the nearby possible worlds (other potential realities similar to ours) are like. If, in all nearby worlds in which the operation is applied, it leads to justified belief, then the operation in fact has the feature. EAJ ultimately leans on facts about modal space, then, not on facts about the future. It can help historical theories handle stalled generation cases well.
Not so. As stated, EAJ looks at what a subject would believe “at t-plus-delta (for a suitably small delta).” It’s looking at a future time. If that’s not clear enough evidence of futurism, consider this. EAJ looks at whether a subject would believe with justification if she applied an available belief-forming operation. In some cases, in all the nearest worlds in which a subject applies the operation, she applies it at a future time. And in some cases, she also applies that available belief-forming operation in her own world, at a future time. And it’s plausible that the way her world is in the future determines which worlds are near it. Her world’s future, then, helps shape which counterfactuals are true in it in the present. In short, while EAJ does look at modal facts, the future can make a difference to these modal facts. On EAJ, justification in the present still directly depends in part on the future.
The rest of this subsection drives home the point that EAJ implies futurism. If you’re already convinced, then the next subsection may have greater charms.
Consider two subjects, Al and Ike, who are as similar as can be. Mentally, Al and Ike are exactly alike. Their pasts are alike. What they have had justification for believing is alike. Even their worlds are as alike as can be, past and present. However, their futures differ. While the same reliable belief-forming operations are available to Al and Ike, nearly any given operation can have varied performance. It can be imperfect in a way, not yielding all justified outputs a more powerful process could. Or, at any given time, there is a variety of justified outputs it could yield. Now, both Al and Ike apply a reliable belief-forming operation, R, to their total respective cognitive states at t. For Al, R will yield belief that p at t + 1, and that belief is doxastically justified then. For Ike, R will not yield belief that p at t + 1, nor any other time soon. Perhaps, R will instead yield belief that q, and that belief will be justified. Or perhaps R will yield no belief at all for Ike. And these possibilities are not fantastic. R could be a memory process. Sometimes we don’t find what we look for in memory. Other times we do. Or, sometimes a memory process is too indistinguishable from mere imagination. Other times memory announces itself distinctly. So, sometimes, a memory process does not yield belief. Other times, that same process does.
Al satisfies EAJ at t, with respect to p, thanks to R. EAJ says Al is propositionally justified in believing that p. What about Ike? R does not help him satisfy EAJ with respect to p at t. And let’s add that no other suitable belief-forming operation is available to Ike then. So for Ike and p, EAJ is unmet. EAJ says Ike is not propositionally justified in believing that p.
What explains why EAJ allots propositional justification differently to Al and Ike? Their different futures. It can even be what memory will do for each of them that helps directly determine the justification each subject has in the present. EAJ implies futurism.Footnote 9
I’ve mentioned EAJ has been tweaked over the years. Some refinements may treat Al and Ike alike, implying they have the same propositional justification. Those revisions to EAJ still imply futurism, however. For example, Lyons’s (Reference Lyons, Kornblith and McLaughlin2016: 167) revision says: “S is ex ante justified in believing that p at t if and only if, if S were to come to believe p using only processes that took as inputs mental states S is in at t, then S would be (doxastically) justified in believing p.” Call this LEAJ. An important difference between EAJ and LEAJ is the antecedent of the counterfactual each looks at. EAJ looks at an application of a process, which may or may not lead to belief that p. LEAJ looks at a subject who is indeed coming to believe that p. So in cases like Ike’s, LEAJ ignores the subject’s future if he does not form belief that p. For Ike, LEAJ looks at counterfactual situations in which he does form belief that p. And it could be that the belief is doxastically justified in those situations. So, like EAJ, LEAJ says Ike is propositionally justified in believing that p. Al and Ike’s different futures aren’t making a justificatory difference. There’s no scent of futurism here.
But it’s not far off. Take another subject, Mike, who is like Al mentally and whose world is like his, presently and historically. Al and Mike both will momentarily come to believe that p. Each will do so using only processes that take just their current respective mental states as inputs. However, for Al and Mike, there’s a plurality of processes that are up for this job. And they are unequal in their reliability. As it turns out, Al will use process R, which is reliable. Mike will use process U, which is unreliable. So Al satisfies LEAJ’s counterfactual: If he were to come to believe that p using only processes that take as inputs his current mental states, then Al’s belief that p would be doxastically justified. But Mike does not satisfy LEAJ’s counterfactual. If he were to come to believe that p using only the same sorts of processes, it’s not the case that his belief that p would be doxastically justified. So LEAJ says that Al but not Mike is propositionally justified in believing that p now. Their different futures make this difference. What’s more, memory’s role in R or U can help explain their different reliability. So on LEAJ, even what memory will do makes a justificatory difference in the present. Tinkering with EAJ has not avoided futurism.
2.4.4 Wait, Is Futurism So Bad?
Maybe it’s best simply to accept futurism. The preceding has helped us appreciate how views like EAJ – which were perhaps already plausible – have implied futurism all along. And maybe that should warm us up to futurism. If we must choose between No Futurism and futurist views like EAJ, let’s embrace the future.
We might indeed be attending to an underappreciated futurist feature of some favorite theories. But that itself is not evidence for futurism. It might instead condemn the favorites. Futurism looks as bad as ever. Perhaps we should have sniffed out the futurism sooner. But in philosophy there are no statutes of limitations.
What will happen should not bolster or diminish our justification now. A view that respects this has a leg up on a view that does not. Time-slice views respect this and are otherwise viable. They have an advantage.
2.4.5 Wait, Couldn’t It Be Worse?
Maybe we’re not taking views like EAJ seriously enough. Take them more seriously, and we’ll see they could do worse than imply futurism. We need to keep in mind, specifically, that on views like EAJ, doxastic justification is more fundamental than propositional justification. We’ve seen that, on EAJ, the future directly matters for propositional justification now. Still, the future might not directly matter for doxastic justification now. A view like EAJ implies futurism, but only about what, by EAJ’s lights, is a derivative sort of justification. The more fundamental kind of justification remains free of the future.
This might all be right. But fundamentality is irrelevant to the crime. This Element originally focused on propositional justification, which is traditionally thought of as more fundamental. And perhaps it is more fundamental. But our focusing on it was not because of its being (regarded as) more fundamental. It was because the alternative – focusing on doxastic justification – might miss the heart of a conflict. Doxastic justification might involve causation. So it’s not clearly always settled by the time-slice. We’ve been investigating whether there is a notable kind of epistemic justification that is settled by the time-slice. Historical views say No, history matters too. And how many of us had assumed that was the end of the story? If the historical views with the greatest legacies imply futurism – if that is also why they say must No – let it be clear.
Fundamentality does not feature in the charge against views like EAJ. Any futurism is implausible, regardless of whether it’s about a less fundamental kind of justification. And any view implying futurism is incompatible with No Futurism.
2.4.6 Wait, Can’t We Just Drop the Futurism?
Maybe it is no small problem for views like EAJ that they imply futurism. Maybe we should reject them. Do historical theories have other promising means for addressing propositional justification in stalled generation cases (cases in which a subject does not form belief in a conclusion she reasoned to well, from a justified belief)? On EAJ, the past, present, and future all matter for propositional justification in the present. Historical theories could find hope by looking instead at just the past and present. Or, they could look at just the past. Or, they could look at just how severely stalled the generation is. I’ll consider these options in turn.
On the first option, whether a subject in a stalled generation case has propositional justification for believing the conclusion at a time depends on how the subject is at that time, and what took place before. So, her justification may vary at times depending on what she is like at those times. We have to look at how she is five minutes after her reasoning, an hour after, a day after, and so on, to see whether at each of those times she is still justified in believing that 20 percent of the bill is $3.10. And, of course, she must have a suitable past.
This option is bad for historical views. One way to illustrate the badness is to see how the option dooms Kelly’s argument for historical views. Recall that the argument was from the Good and Bad Past Cases. Allegedly, at the end of her reasoning, the subject in the Good Past Case has justification for believing that 20 percent of the bill is $3.10, because of her good past. The subject in the Bad Past Case allegedly does not have this justification, because of her bad past (or, more accurately, because of her lacking a good past). Nothing in the present directly makes the difference in whether each subject has justification now. Nothing in the present is sufficient for their having justification now. This is why (the argument goes) we must look at each subject’s past to see what justification each has now.
So if historical views address stalled generation cases by citing features in the past and present, the argument for historical views fails. It’s no longer the case that the present is too impoverished to account for the Good and Bad Past Cases. It’s no longer the case that we must rummage through the past for these resources.
Suppose in a stalled generation case, for example, the part that matters in the present is whether the subject has a certain disposition. It’s the disposition to remember that she reasoned that 20 percent of the bill is $3.10. If the subject has this disposition, and has a suitable past, then she has justification for believing that 20 percent of the bill is $3.10. If this disposition matters in a stalled generation case, it should also matter in the Good and Bad Past Cases. But if it matters in those cases, something in the present is making a justificatory difference. So why forage in the past for something further? A time-slice theorist can cite that present feature as doing all the required justificatory work. (And the subjects in both the Good and Bad Past Cases have this feature. The time-slice theory predicts these subjects are alike in their justification. So the time-slice theory’s prediction is confirmed.) The argument for the historical theory fails.
If historical theories cite the present and the past in order to account for propositional justification in stalled generation cases, the present features make the past features look extraneous. A time-slice view is simpler and just as explanatory. And there isn’t otherwise reason to prefer the historical view.
This might pull us toward a second option, which handles stalled generation cases by looking at just the past. How the subject reasoned, and why, and in what circumstances, for example, could contain all that determines the subject’s justification, however long after the reasoning we look. Unless the subject gains new relevant evidence, nothing about how she is moments after her reasoning, hours after, days after, and so on, factors into whether she is justified in believing that 20 percent of the bill is $3.10 at those times. An advantage of this option for historical theories is that the past will not become extraneous. It’s all we’re consulting. Nothing in the present will crowd it out, as the present plays no role here.
On this option it’ll be crucial to identify the specific past factors that allegedly matter in stalled generation cases. Otherwise, the theory has no implications about these cases. Whatever those factors are, there’s a problem. The past isn’t changing. So, the past factors that matter in a particular case – how the subject reasoned, and why, and in what circumstances – aren’t changing. So however long after the reasoning we look, the subject’s justification from the past will be the same. Suppose the reasoning played out in such a way that, five years after, the subject does not have justification for believing that 20 percent of the bill is $3.10. Then she mustn’t have had justification for believing that five seconds after the reasoning. That’s almost draconian, still perhaps true but not obvious. However, suppose the reasoning played out in such a way that, five seconds after, the subject does have justification for believing that 20 percent of the bill is $3.10. Then she’ll have it five years out too. The relevant part of the past isn’t changing, so her justification isn’t either. But this is absurd! The justification should expire at some point, regardless of her past. Looking at just the past over-generates justification, resulting in epistemic goods that endure too long. This option gets the cases wrong.
A third, and perhaps more promising option, does not center on the past. It looks at how much time has elapsed after the reasoning – the duration of the belief generation’s stall. If the subject has not learned anything new and relevant since her reasoning, the specific temporal distance from that reasoning decides whether she has justification that long after her reasoning.
This third option has countless varieties, each specifying a different temporal distance as relevant. The varieties face individual problems, however, and a common major problem. One variety says the justification expires immediately in a stalled justification case. Use it or lose it. That is not just extreme, but also unmotivated. A variety at the other end of the spectrum says that the justification never expires. This is also extreme and equally implausible. All other varieties place the amount of time somewhere between these two extremes. Each of these less extreme varieties seems arbitrary.
Moreover, all varieties of this third option (other than “use it or lose it”) must explain why there is justification within a particular distance only forward in time from the stalled generation. Why not suppose justification is also present at all times within that interval prior to the stalled generation? Suppose the subject has propositional justification for believing that 20 percent of the bill is $3.10 for 5 minutes after her reasoning. That the reasoning is within 5 minutes in the past plays no clear special role here. Nothing about the past does. If temporal proximity to the reasoning matters, we need a rationale for supposing it matters in only one direction. Otherwise, it should matter in both directions, and No Relevant Difference stands.
2.5 No Time Like the Present
We have explored the conditions in which a subject has propositional justification after her reasoning well from justified beliefs ends in stalled generation. We have found no good account of these conditions on which the past but not the future makes a difference for justification in the present. Causation matters for doxastic justification. But we’ve yet to see why it should matter for propositional justification. We can grant that doxastic justification is more fundamental. Still, on this assumption, the available accounts of propositional justification either are flawed or still lead to No Relevant Difference. None of the most obvious differences between the past and future explain well why only one of them matters for justification in the present. We have a case for No Relevant Difference. And No Futurism is common ground. No Arbitrary Asymmetry is also plausible. Memory plays a role in our having propositional justification in the present. But it’s only present memory that plays this role.
The main argument of Section 2 has been more than a mere objection to the historical view. The argument has prompted us to look more carefully at cases in which a past makes an alleged difference to present justification. We’ve also looked for what it is about the past that might make this difference. And we’ve seen the importance of this query: If we can’t identify why the past matters, it looks irrelevant to present justification – as irrelevant as the future. The historical view is no better than futurism.
3 Out of Time, Into Space
We’ve settled the time dimension. How memory is in the present and what it does then, and at no other times, directly matters for present justification. Now on to the space dimension. Maybe what things are like outside the mind, contingently, directly matters for justification. Or maybe not – just the mental directly matters. Externalism is the first option. What makes no mental difference can still matter for justification. The other option is internalism. Mentally alike subjects are alike in their justification, regardless of how their external worlds differ. This section sabotages externalism, leaving only internalism intact.
In Section 1, I noted that externalism’s appeal often has to do with a particular view about the time dimension. A traditional, influential argument for the dominant externalist view – process reliabilism – is from historical, memorial considerations. Allegedly, historical factors such as how a belief came about or has been sustained directly matter for the belief’s justificatory status in the present. And process reliabilism respects this. It says the reliability of the type of process that forms or preserves a belief affects that belief’s justification – as does that belief’s past justificatory status, and more. And reliability is settled in part by the external world. Whether a process type is reliable has to do with how it tends to perform. And the process types that matter are instantiated across multiple subjects. A subject might have a belief now because of a certain kind of remembering, but other subjects have their beliefs because of that kind of remembering too. How a process tends to perform has to do with a broad range of cases. Most of these cases are external to any given subject. Also, often, how the external world is determines whether a belief is true. So, how accurately a process tends to perform – whether it tends to produce true rather than false beliefs – is in general external to any given subject. Reliability matters and depends largely on the nonmental.
Now, memory is in many ways responsible for the beliefs we have. It, for example, seems to preserve some beliefs over time. So how does memory matter for justification? Reliabilism’s answer is both historical and externalist. It looks at the history of the belief and the reliability of the process preserving it.
That answer scores reliabilism no points. As we’ve seen, the past does not directly matter for present justification from memory. So historical considerations do not make sense of this present justification. Not only is reliabilism’s answer wrong, but the traditional argument for reliabilism fails. The argument would confirm reliabilism on the assumption that the past matters. But that assumption is mistaken.
What’s more, for some externalist views, including process reliabilism, that erroneousness spells doom. Reliabilism’s answer to how memory matters for justification not only scores no points, but also costs it the game. Process reliabilism implies that the past matters. If the past is insignificant, process reliabilism is false. Let’s bring this implication out into the light of day. Reliabilism consists of two sufficient conditions for justification, and a further claim that these are the only sufficient conditions.Footnote 10 One condition is:
If a subject’s belief that p at t results from a belief-dependent process that is conditionally reliable and all belief inputs to which were justified, then the subject’s belief that p is justified at t.
The idea is this. Consider a process that, when it outputs a belief, uses beliefs as inputs – for instance, the process of inferring from some believed premises to a belief in a conclusion. Suppose that a type of process tends to output true beliefs when the belief inputs are true. It’s truth-preserving. And suppose all the belief inputs to an instance of that process were justified. The above sufficient condition says that the belief output of this token process is justified.
The sufficient condition applies not just to processes that form new beliefs, but also to processes that sustain or preserve beliefs. Those processes might take as an input a subject’s belief that p at one time, and output belief that p at a later time, perhaps at many later times. Preservative processes make sense of the persistence and continuity of a person’s beliefs over time. And preservative processes include, perhaps consist entirely of, memorial processes. Much of what you believe now, you believed before. Memory has kept what you believed.
Let’s put more flesh on these bones. Suppose a subject forms a belief that p with justification. A memory process preserves this belief over time. And that type of memory process tends to yield true rather than false beliefs when it takes in true beliefs. Reliabilism implies that the preserved belief is also justified. An unjustified belief entering this same preservative process would not have met this (nor potentially any other) sufficient condition for justification. So, whether the belief was justified in the past directly matters for whether the preserved belief is justified later. On reliabilism, a historical difference makes a justificatory difference. But historical differences don’t matter. One of reliabilism’s two sufficient conditions for justification is false. So reliabilism is too.
The remaining, best-known externalist views are recognizable only as historical theories. They lend themselves poorly to the time-slice. Consider, for example, proper functionalism. Put simply, it says a justified belief is one that is produced by cognitive faculties that are functioning properly, aimed at the truth, and reliable in the environments for which they were designed (Bergmann Reference Bergmann2006). One thing this last condition says is that cognitive faculties responsible for justified belief must have been designed for certain contexts. Whether this condition is met can depend on the past. Some faculties may have, once upon a time, been designed for functioning well in a particular environment. Other faculties were never designed that way. That is why some but not other faculties can produce justified beliefs in that environment, according to proper functionalism. Predictions such as these characterize proper functionalism. They also make it incompatible with a time-slice view.
Consider also, for example, a less popular cousin of process reliabilism, indicator reliabilism. Simplified, it says justified belief that p is based on a reliable indicator of p (Alston Reference Alston1988). It’s plausible that the past plays some key role here. Differences in the past make differences in what reliably indicates what. For instance, suppose a memory experience of my locking my front door is a reliable indicator that I locked my front door this morning. Presumably, whether I locked my front door this morning can help determine what reliably indicates that I locked it. Suppose my memory experience is caused in a normal way, involving my consciously locking the door. The fact that this experience comes about from my actually locking the door seems to matter for whether the experience is a reliable indicator of I locked my front door this morning. If the memory experience does not trace back to my locking the door and is instead, say, merely drug-induced, then it does not reliably indicate what I did this morning. Indicator reliabilism, like proper functionalism, is a historical theory. That’s a problem.
Process reliabilism may not be down for the count, however. It does have a second sufficient condition for justification, one that does not so obviously imply that the past matters. Perhaps reliabilism has a chance if it severs its two conditions. Take the one we’ve discussed, and consign it to the flames. Retain just the potentially nonhistorical condition. That condition is, roughly:
If a subject’s belief that p at t results from a reliable process, then the subject’s belief that p is justified at t.
This says that a belief is justified if the process responsible for it is of a type that tends to produce true rather than false beliefs. It does not consider the history or past status of the inputs to the responsible process. The condition here does concern resulting, which is a causal relation. But perhaps we can purge history from the condition by bracketing diachronic resulting here and consider just synchronic resulting.
The Frankensteinian view that emerges here has the upshot of counting as a time-slice externalism. But it has downsides. For one, it’s unlikely to account for all the justification from memory that we presumably have. Many defended reliabilist views invoke the history of a belief to make sense of whether it has justification now (Goldman Reference Goldman2009, Reference Goldman and Dougherty2011; Senor Reference Senor2019). A belief with a proper enough past has justification in the present. A belief with a poor past – acquired from bias or poor reasoning, and lacking evidence – lacks this justification. Without reference to history, it’s doubtful that reliabilism can separate the wheat from the chaff.
For another, as noted in Section 1, on a time-slice reliabilism, the past performance of a process doesn’t help determine whether that process is reliable in the present. Facts about the present carry the whole load. That’s still bad. As time passes, reliability could fluctuate, perhaps wildly, as the facts that determine reliability change over time.
To elaborate: On most reliabilist views, a reliable process is one that performs well enough in a special, multi-membered domain of worlds. That is, in these worlds, the process tends to deliver enough true beliefs relative to false beliefs. That’s why its deliverances are justified. Despite the popularity of this type of reliabilism, the literature elaborates little on which worlds populate this special domain (Frise Reference Frise2018b). Still, all answers on offer locate our world in the domain (Alston Reference Alston1995; Goldman Reference Goldman1986, Reference Goldman and Smith2008; Henderson and Horgan Reference Henderson and Horgan2011). That is, how our world is – including how well a belief-formation process performs in our world – helps settle whether a process is reliable. A time-slice version of this reliabilism would look at just the momentary performance of a process in the special domain of worlds. Whether your belief is justified at a time is a matter of whether it results from a process performing well enough then in the special domain. Again, our world is in that domain. The momentary performance of a process in our world varies. It varies in the other worlds in the domain too. It could vary spectacularly. But even small variation can make the justificatory power of a process hostage to the time. In short, on a time-slice version of the most common sort of reliabilism, whether a given process justifies can be in flux, perhaps even constantly.
There is yet another downside to time-slice reliabilism – at least, many externalists would say so. On this reliabilism, justification from memory is wholly generative, not preservative. For many externalists, this is reason for weeping and gnashing of teeth. On such a view:
There will be no carry over of a belief’s [justificatory] status from one moment to the next. Assuming for simplicity discrete moments in time, after each such moment the epistemic slate is wiped clean. A person may mnemonically retain a belief from one moment to the next, but that retention makes no contribution to the justifiedness or unjustifiedness of the succeeding moment’s belief.
Goldman adds that a purely generative view “would open the door to rampant skepticism about our common sense knowledge. This would bestow a huge and unwarranted gift on the skeptic.” So he would agree that a time-slice reliabilism has the downside of poorly accounting for all the justification from memory we seem to have.
Time-slice internalism has no such downsides, however. It is much friendlier to memory’s justificatory role being purely generative. It does not cite reliability in its account, so it needn’t bludgeon that square peg into the round hole of an ahistorical theory. And time-slice internalism has the resources to make sense of our having the justification from memory we seemingly have.
For example, it could be that a subject has justification for believing that p if she seems to recall that p (Pollock Reference Pollock1986). This may be because of the particular phenomenology of that experience (Feldman & Conee Reference Feldman and Conee2001), or because the experience has a kind of doggedness (Brogaard Reference Brogaard, Bernecker and Michaelian2017), or because the experience justifies the subject in believing that she has already justifiedly believed that p (Schroer Reference Schroer2008). Another resource, compatible with the preceding, is that a subject has justification for believing that p if she is disposed to recall that p (Conee & Feldman Reference Conee, Feldman and Dougherty2011; Ginet Reference Ginet1975; McCain Reference McCain2014). The subject might not be recalling that p or even thinking about p at a particular time. But she can still have the disposition to do so then. And that disposition – a mental feature of hers – can make sense of why believing that p is justified for her even when p is not before her mind.
Time-slice internalism is free to draw on any of these resources. Recalling and being disposed to recall are entirely mental features. They concern only the way a subject is, mentally. So, they fit with an internalist view. Subjects who are mentally alike in all ways recall and are disposed to recall the same things. And they are alike in their justification. Also, the resources concern only how things are at the time. Past and future recalling and dispositions don’t factor in. What the subject recalls or is disposed to recall only then is what matters then.
These resources outfit time-slice internalism well. They equip it to make sense of the justification it seems we have from memory. Someone asks you a difficult trivia question. With some effort, an answer comes to mind. It’s not that you are merely entertaining a potential answer then. You seem to recall one as true. And it’s because you seem to recall it that you have justification for believing it then. Additionally, it’s common sense to suppose that, all the while, there’s plenty else you’re justified in believing – such familiar information as where you were born, who the first U.S. President was, what state you are in, and so on. You are, unsurprisingly, disposed to recall all of that information then. And it’s because you have those dispositions that you are justified in believing all of that information, even when none of it is at the fore of your mind. It does not matter if you’ve forgotten how you originally learned the information. What matters is that you’re recalling it, and that you’re disposed to.
And these resources do not render time-slice internalism reckless, casting justification about without an eye to where it lands. That is, time-slice internalism does not end up overattributing justification from memory. It, for example, respects the right kind of forgetting. Forgetting how you first learned some information does not itself carry weight. But forgetting the very information does. It’s plausible that you are not justified in believing information you’ve long or altogether forgotten – information you can’t recall. And notice: Information you can’t recall is not information you seem to recall or are disposed to recall. (Of course, if you go on to relearn the information, you may then become disposed to recall it. But in that case, the information is no longer forgotten.) Time-slice internalism does not imply a subject has justification for believing completely forgotten information, even if recollection and recollective dispositions can matter.
We’re in a position now to see why time-slice internalism makes no mistake if it attributes justification in the Bad Past Case. How could a subject begin from the justified belief that the bill is $13.50, make an error while reasoning, and still end up having justification for incorrectly believing that 20 percent of the bill is $3.10? Part of the answer is this. By stipulation, at the time of deducing her answer, the subject has forgotten that the bill is $13.50. She is not recalling that the bill is $13.50. She is not disposed to recall that either. A forgotten link in a chain of reasoning may no longer limit which conclusions are legitimate. (More generally, what you forget no longer limits what you may legitimately believe.) By forgetting her starting point, the subject has lost reason to doubt her destination. And look at how she is just while at her destination. Perhaps her conclusion seems correct to her. Perhaps it seems to her to follow from reasoning she seems to recall as correct. And her memory gives her no reason to doubt her conclusion. It’s hard to see why any attitude toward it other than believing could be justified for her, if her memory is this way and things seem this way to her. Time-slice internalism’s resources respect the right kind of forgetting. Attributing justification in the Bad Past Case can be correct.
Denying the historical view confirms internalism in the epistemology of memory. It shows internalism is much better off than its externalist archrival, process reliabilism. Externalism’s champion is slain. Externalism’s best hope is a painful divorce from historical views. I will not survey exotic externalisms compatible with a time-slice view. They have not been developed in much detail, particularly in the domain of memory. The next section concentrates on where within time-slice internalism we should ultimately land, in light of memory.
4 Internalism on the Tip of the Tongue
The rest of this Element addresses a quarrel among time-slice internalist views. It explores which mental conditions in the present matter for our having epistemic justification. The epistemology of memory so far has given us reason to accept that only the present and the mental matter for justification. Reflecting further on memory will suggest that only some of the mental matters. Memory, I claim, supports an extreme view: Just a narrow range of the mental matters. A bit more specifically, memory supports strong internalism. Strong internalism says, roughly, that just a subject’s currently accessed mental life matters for her present justification (Bergman Reference Bergmann2006; Goldman Reference Goldman1999; Moon Reference Moon2012). So, yet more specifically, I’ll show that memory cases involving the tip of the tongue phenomenon support strong internalism over other time-slice internalist views. I do not claim this support is decisive here, or even the best available. I will not weigh here the totality of considerations that would ultimately crown one of the internalist contenders. But it’s worth seeing which view the tip of the tongue phenomenon favors. After all, it is familiar and interesting. Yet it is puzzling and nearly unexplored – in epistemology in general, and in the epistemology of memory in particular. If the phenomenon offers a case for a theory, it’s worth hearing.
In Section 4.1, I’ll state the main time-slice theories that are elbowing for a position on the podium. In Section 4.2, I’ll describe the tip of the tongue phenomenon. In Section 4.3, I’ll explain how tip of the tongue cases appear to present a problem for strong internalism. Finally, in Section 4.4, I’ll show that the only sensible epistemic evaluation of these cases is one that supports strong internalism over its main internalist rivals. Strong internalism is by far the least popular of these options. It’d be quite the upset if it comes out on top.
4.1 Internalism
The competition here is between three main versions of time-slice internalism. They differ on the internal factors they say matter for justification. Let’s look at them from the least restrictive view to the most.
Least restrictive here is just the pairing of the time-slice view and internalism, a pairing familiar from Section 3. Call it mentalism. It says that a subject’s mental life at a time directly settles her justification then (Feldman & Conee Reference Feldman and Conee2001). A change in the external world that does not affect her mentally does not affect her justification. If two subjects differ in their justification at a time, they differ mentally then.
To see how the competitors differ, it’s helpful to see what justificatory roles each of them allows memory to play. It’s compatible with mentalism that a subject’s memory experiences matter for her justification. On mentalism, for example, it could be that a subject who seems to recall that p has justification for believing that p. Memory may matter in other ways also. A subject may not now seem to recall that p, but thanks to memory, she may at the moment still be disposed to seem to recall that p. This disposition is mental. On mentalism, having this disposition could explain why the subject has justification for believing that p even when she’s not recalling that p.
The next competitor is more restrictive than mentalism. Call it access internalism. As a time-slice theory, it, like mentalism, says the mental life of a subject at a time determines her justification at that time. But access internalism waggles a finger at some of the mental. It says only a certain portion of the subject’s mental life ultimately matters at a time. Specifically, what matters is what is in a certain way mentally accessible to the subject then (Feldman Reference Feldman and Austin1988; Feldman & Conee Reference Feldman and Conee2001). The exact nature of this accessibility is worth exploring fully. But I’ll save that adventure for another time. Examples may clarify enough. A memory experience, such as seeming to recall that p, is appropriately accessible to the subject. The seeming recollection is part of the subject’s current conscious life. It is making a difference to her conscious experience. She can attend to it and its features. So, on access internalism, this seeming recollection matters for a subject’s justification. Other aspects of her mind, however, are sidelined. A subject’s almost forgotten memory – one that would require an obscure and specific cue to bring to mind – might not be suitably accessible to her. The almost forgotten memory might have to become more or differently accessible before it could contribute to what she is justified in believing. It’s more controversial whether other aspects of the subject’s mind are accessible to her in the right way. For example, what are we to make of a subject’s disposition to seem to recall that p – is the disposition itself fittingly accessible to her? Fortunately we can table that. A readily available memory, something familiar and easily recalled, satisfies the accessibility condition, while a nearly forgotten memory does not. That is guidance enough. Mentalism, by contrast, can count the readily available memory, the recollective disposition, and the nearly forgotten memory as all mattering for justification.
The final competitor is the most restrictive. It counts even less of the mental as mattering for justification. Call the view strong internalism. It says only a certain portion of what is mentally accessible to the subject at a time matters for a subject’s justification then. What matters is what the subject is mentally accessing or aware of at the time. An occurrent memory experience, such as seeming to recall that p, likely satisfies the access condition. Mentalism, access internalism, and strong internalism can agree on the power of memory experience. But on strong internalism, a memory, however familiar and readily available, makes no difference to justification while the subject is not accessing it. Likewise, only the experiences, dispositions, and other mental phenomena that the subject is suitably aware of at the time matter for her justification then.
These three time-slice internalist theories differ. Still, as stated, they are in fact compatible. They are not quite rivals yet. Indeed, access internalism implies mentalism. What’s mentally accessible is, of course, mental. If what’s mentally accessible to one settles one’s justification, then one’s mental life settles one’s justification. Also, strong internalism implies both access internalism and mentalism. What one mentally accesses is both mentally accessible and mental. If mental access is central to justification, then mental accessibility is central. The accessed is a subset of the accessible.
But rivalry is just down the road. Some mentalist or access internalist theories imply this or that unaccessed mental entity does affect one’s justification. Strong internalism implies otherwise. Strong internalism competes with these internalist views. Let weak internalism be the claim that one of those internalist views is true. Weak internalism denies anything as restrictive as strong internalism. I will be comparing strong internalism to this internalist competitor (or gang of competitors). Because strong internalism is the more exclusive and less popular of the two, it would be surprising if it beat out weak internalism. And it would be surprising if a phenomenon so familiar as the tip of the tongue state supported it.
These are just three categories of time-slice internalist views. There are others. And then there are views that look beyond the time-slice or beyond the mental to explain justification. It might be tempting see whether the tip of the tongue phenomenon supports one of these theories. But Sections 2 and 3 cast doubt on them. So this section evaluates just strong internalism and its most popular time-slice internalist contenders here.
We saw that the contenders over the time and space dimensions of justification could concern either propositional or doxastic justification. The same goes for the internalist contenders that are the focus of this section. And, as before, the focus will be on propositional justification – having an attitude toward a proposition being justified for a subject. On internalism, this is usually understood to be an entirely synchronic matter, settled by the time-slice. Doxastic justification – the justification of a subject’s doxastic attitude – is often regarded on internalism as involving a belief’s being based on what propositionally justifies it. And this basing may tug us somewhat out of the time-slice, factoring in the past. Doxastic justification, then, may involve the past. If it does, the time-slice internalist views on the table here are poorly understood as views of doxastic justification. Regard them as views of propositional justification.
4.2 The Tip of the Tongue
I’ll argue that the tip of the tongue phenomenon supports strong internalism over weak internalism. Before that, though, a word about this phenomenon is in order. Philosophers and psychologists, since William James (Reference James1890), have discussed it. Here’s a paradigmatic example. You’re asked who starred in a movie you saw many times in childhood. You picture the actress, recall how many syllables are in her first and last names, and recall the letter that begins her first name. Her name eludes you, yet you feel like you have it. It’s on the tip of your tongue.
This is an example of the tip of the tongue phenomenon. The subject is in the tip of the tongue state, having a tip of the tongue experience. It involves retrieval failure, the failure to produce target information despite trying. The target information is a word or phrase, commonly a name. The retrieval failure may immediately precede the tip of the tongue experience. But it may persist alongside the experience. The experience involves the subject feeling confident that she has the information in memory (Schwartz & Metcalfe Reference Schwartz and Metcalfe2011). She may even feel she can produce it. This feeling may encourage her to keep looking for it in memory rather than to give up or to Google it.
Details on the nature of the tip of the tongue experience are up for debate. One view is that a subject having this experience is conscious of a certain mental state of hers that “carries” the target information, but she is not conscious of its content (Rosenthal Reference Rosenthal2000: 204). According to another view, the state is a feeling. It is a feeling unique to a gap that the target information completes, and in this gap is a “wraith” of the target that makes us at times “tingle with the sense of our closeness” to the target (James Reference James1890: 251). Another view is pluralist (Calabi Reference Calabi2016). The tip of the tongue state is not uniform. Different states, such as beliefs, perceptions, or metacognitive feelings, can make it up. What unites these states is a distinctive phenomenology.
Maybe the tip of the tongue phenomenon is not purely mnemonic. Maybe it involves brain systems, or mental processes or faculties or states, other than memory. It may seem that this phenomenon is therefore useless for deciding what version of internalism it is that the epistemology of memory favors. That seeming is misleading, however. Consider: A case of temporally extended reasoning is memorial. It is not purely memorial. After all, it involves reasoning! Still, the epistemology of memory can advance by reflecting on such a case. It may become clearer whether, say, the past is relevant to justification from memory in the present. That then makes clearer whether and how the past matters for justification at all. If a phenomenon is memorial at all, even if not purely, it can help reveal what the epistemology of memory upholds. And the tip of the tongue phenomenon is memorial. Perhaps focusing on the clearly non-memorial details of the phenomenon would be useless in our endeavor. But I will not be examining those.
I will remain neutral on as much of the nature of the tip of the experience as I can. I won’t defend necessary or sufficient conditions for having this experience. In the next section, however, I will state one common sense assessment of the tip of the tongue phenomenon, one that puts strong internalism on the backfoot.
4.3 An Apparent Weakness of Strong Internalism
The tip of the tongue state draws our attention to a part of a subject’s mental life. It’s a part that she does not, in some important sense, have before her mind. It’s plausible that whatever the tip of the tongue state ultimately is, some aspect of the subject’s mental life is outside her awareness. Usually the subject in the tip of the tongue state has some information in memory, is aware she has it, tries to access it, but does not access it.
Stare at this for long, and an obstacle for strong internalism takes shape. A subject is not aware of the whole of her mental life at a time. And we are wondering which internal conditions in the present matter for justification from memory. Initially, weak internalism will seem to have better answers. They can account for more of the justification we appear to have.
For example, here is one critical asset of weak internalism. It has the resources to account for the justification of dispositional (or stored or nonoccurrent) beliefs. A dispositional belief is perhaps best understood in contrast to an occurrent belief. An occurrent belief is, in an important sense, before the subject’s mind. It, with its specific content, makes a difference to her consciousness. I ask you: Who is your favorite composer? You reply: Tchaikovsky. You’re now thinking about Tchaikovsky being your favorite composer, judging this to be so, consciously endorsing it. You occurrently believe that he is your favorite composer.
This is not the first time you’ve believed that your favorite composer is Tchaikovsky. You realized he’s your favorite some while ago. It’s plausible that you’ve believed it ever since. You weren’t thinking about it right before I asked you. Still, right before I asked you, you believed that Tchaikovsky is your favorite composer. Your belief that Tchaikovsky is your favorite composer was, just prior to my asking, dispositional.
Whether there are dispositional beliefs at all is somewhat controversial (Frise Reference Frise2018a). But it is common sense to suppose they exist. What’s more, it’s common sense to suppose we have hosts of dispositional beliefs. After all, we have hosts of beliefs, but only so many are occurrent at a time. The rest must be dispositional. Finally, and critically, it’s common sense that many of our beliefs are justified even while dispositional.
If there are dispositional beliefs, any good theory of justification must account for their epistemic status. And if there indeed are many justified dispositional beliefs, weak internalism appears in better shape than strong internalism. Strong internalism may account for the justification of these beliefs by citing only that which the subject is currently accessing. This well dries too quickly. Before I asked about your favorite composer, your accessed mental life may have explained the justification of some of your beliefs, such as that I’m spending time with a philosopher, that it’s sunny, that I wish I wore sandals today, that warm weather is overrated. But your accessed mental life, before I posed my question, did not concern Tchaikovsky. It also didn’t concern familiar mathematical truths, world history, much of your autobiography, or most other topics. Yet on the common sense view, you had justified beliefs on these topics then. Strong internalism struggles to explain how.
Additionally, even if strong internalism could account for this justification, the account is suspect. You acquired most of your justified dispositional beliefs in considerably different circumstances, using considerably different evidence. It’s plausible that these beliefs differ considerably in what justifies them in the present. But your currently accessed mental life leaves you little to choose from. The beliefs would have an incorrectly narrow and uniform range of justifiers.
Weak internalism is free of these struggles. It is free to count diverse parts of your currently unaccessed mental life as justifiers for these beliefs at the moment. For example, a broad range of relevant but unaccessed memories or dispositions to recall could justify beliefs, on weak internalism. These phenomena are mental and, arguably, suitably accessible. Of course, it’s possible that they are not just accessible. It could be that a subject also accesses some of these phenomena at a time. So some of them can help strong internalism explain why certain dispositional beliefs are justified. But it’s implausible that a subject is at one time accessing all the mental phenomena that weak internalism might appeal to. Strong internalism, it seems, still fails to account for the justification of all justified dispositional beliefs. Weak internalism appears to have an advantage over strong internalism.
We’ve been thinking about how well different internalist theories explain some common sense views on dispositional beliefs. Let’s turn to a common sense interpretation of the tip of the tongue phenomenon. Weak internalism appears to have an advantage over strong internalism here too.
A common sense, or at least sensible, take on a typical tip of the tongue case goes like this. Typically, the subject in a tip of the tongue case has a particular dispositional belief. The dispositional belief involves the information that’s on the tip of the subject’s tongue. The content of the dispositional belief is a proposition about that information. If a name is on the tip of your tongue, it’s not just any name that’s on the tip of your tongue. It’s a name that completes a proposition you have a doxastic attitude about. The proposition is relevant to the inquiry you’re making. It may be that the tip of the tongue experience just is your being aware of your belief in a certain proposition, without your being aware of the belief’s content. But we needn’t muse much about the nature of this experience. However we understand its nature, the following is sensible. In the typical case, the subject has a dispositional belief in the unaccessed information. Without the belief, the subject would not feel she has the information that the belief is about.
If there is indeed this sort of dispositional belief in many tip of the tongue cases, its epistemic status matters. It’s plausible that in a typical case, this belief is justified. In fact, a tip of the tongue case may be one of the best examples of justified dispositional belief. For other dispositional beliefs, the subject has various potential justifiers, such as relevant memories or dispositions to recall. For the dispositional belief in the tip of the tongue case, all else being equal, the subject has these same sorts of potential justifiers. The subject has other support too. The subject has some metacognitive support for the belief by way of epistemic feelings, such as the feeling of knowing. Her dispositional belief is in a proposition about the answer to her inquiry, and she feels she knows the answer. She just does not access the answer and recognize it as such at the moment. To the subject, it can also seem that the answer to her inquiry has a certain number of syllables, or starts or ends with a certain letter. This supports her dispositional belief, as parts of its content fit these seemings. In short, compared to other dispositional beliefs, there is additional support for the dispositional belief in the typical tip of the tongue case.
Also, in the typical tip of the tongue case, it is especially plausible that the subject even has a dispositional belief. The subject has a belief that is not quite occurrent – the subject is not judging the belief’s content to be true, she is not endorsing its content, and so on. Yet the belief makes its presence known to the subject in the tip of the tongue case. The subject has an occurrent mental state about the dispositional belief, or about the belief’s content while that content is still not before her mind. Alleged dispositional beliefs in other cases aren’t ordinarily like this.
Weak internalism makes sense of the justification of the dispositional belief in a typical tip of the tongue case. It has more resources. It can cite more mental entities as potential justifiers, just as they can for other justified dispositional beliefs. However, the subject is not accessing these entities in the tip of the tongue case. So strong internalism cannot cite them as justifiers. Strong internalism does not explain the justification of the dispositional belief in the tip of the tongue case.
Let’s be a bit more precise. The internalist theories we’re considering are time-slice views. And the doxastic justification of a belief may in many cases involve more than the time-slice. Doxastic justification requires basing, which may have a causal element that involves temporal succession. Talk of the justification of the dispositional belief in a tip of the tongue case may misdirect us. It may incline us to think of doxastic justification in particular. But the focus should be on propositional justification. A subject in a typical tip of the tongue case dispositionally believes that p, where p is a proposition answering a target inquiry. And believing that p, it seems, is propositionally justified for the subject in the typical case. What accounts for this justification? Strong internalism has a harder time responding well than weak internalism does. Tip of the tongue cases seem to count against strong internalism, as do cases of justified dispositional beliefs in general. But here, it is all the more plausible that the subject has a justified dispositional belief. Strong internalism’s apparent disadvantage is that much starker.
4.4 Strong Internalism Strengthened
This section argues that appearances are misleading here. The tip of the tongue phenomenon does not count in favor of weak internalism over strong internalism. In fact, it’s the reverse. The tip of the tongue phenomenon supports strong internalism over its internalist competitors. This provides reason to reevaluate what it is that alleged cases of justified dispositional beliefs in general support. Perhaps they do not support weak internalism over strong internalism after all.
To see how things really stand, let’s look more closely at the possible epistemic statuses of the dispositional belief in the typical tip of the tongue case. Some of these statuses turn out to be implausible. At least one status is plausible. And it points toward strong internalism, not its internalist rivals.
Section 4.3 covered a common sense take on the typical tip of the tongue case. The take is that the subject has propositional justification for believing the dispositionally believed proposition. I’ll show that, if this is wrong, strong internalism is the better view. But first, let’s clarify the common sense take. A subject with justification has either just prima facie justification or both prima facie and overall justification. Overall justification is undefeated. The subject has support for having an attitude toward a proposition, and other considerations don’t count enough against that. Nothing for her fully undermines the support or counts at least as strongly in favor of having a different attitude toward that proposition. If something prima facie justifies a subject in having an attitude toward a proposition, this support may for her still stand defeated by other considerations. The subject’s counterevidence or her reasons to doubt that the support relation holds can do this defeating. If defeated, the justification is only prima facie, not overall.
With this in mind, there are three main options to consider. Again, in the typical tip of the tongue case, the subject dispositionally believes that p, where p answers her target inquiry, and the subject fails to access this answer despite her inquiring and her feeling that she knows it. Option 1, the subject has both overall and prima facie justification for believing that p. Option 2, the subject has prima facie but not overall justification for believing that p. Option 3, the subject has neither overall nor prima facie justification for believing that p. The common sense take on the typical tip of the tongue case is the first of these options. Let’s consider them in order.
4.4.1 Tip of the Tongue Cases: Overall Justification?
Option 1 is that the subject in the typical tip of the tongue case has both prima facie and overall justification for believing the relevant proposition. She has some support for believing, and the support is not fully defeated, perhaps not defeated at all. I will show this option is untenable.
It’s sensible that the subject believes the relevant proposition in the typical tip of the tongue case. She does not believe it occurrently, though. She believes it just dispositionally. Something is impeding the belief from shifting modes from dispositional to occurrent. We are supposing, on option 1, that this belief is justified.
It’s also sensible that whatever the justifier is, the subject is not accessing it. (The access requirement on strong internalism, then, is not met.) The subject’s dispositionally believing, then, is justified by some unaccessed mental entity if justified at all.
Crucially, if this unaccessed mental entity explains why dispositionally believing is justified for the subject, it should also explain why occurrently believing is justified for her. After all, a subject’s dispositional belief that p and occurrent belief that p are the same token belief, simply in different modes. To suppose they are distinct beliefs is to suppose it is possible for a subject to have multiple beliefs that p at once, since a subject can simultaneously occurrently and dispositionally believe that p. But it is implausible a subject could have a plurality of beliefs with identical content (Frise Reference Frise2021).
Justifiers do not discriminate between the modes of the doxastic attitude they justify a subject in having. If, say, an experience justifies your believing that p, the experience does not justify only occurrently believing that p but not also dispositionally believing that p, or vice versa. The experience justifies having the attitude in both modes or in neither. So, if an unaccessed mental entity justifies dispositionally believing for the subject, it also justifies occurrently believing for her.Footnote 11
And yet, contrary to common sense, in the typical tip of the tongue case, occurrently believing the relevant proposition does not in fact look justified for the subject. It appears instead that if something about the subject’s mental life contributes to her having justification for occurrently believing the relevant proposition, the subject must be accessing whatever that is. She must be accessing something, such as a recognition of the answer or the seeming truth of the answer. One way that a tip of the tongue case resolves is as follows. The subject goes on to suddenly occurrently believe the relevant proposition without awareness of a supporting mental entity. And in such a case, occurrently believing is not justified for her.
For example, suppose a subject is struggling to recall the capital of Armenia. She learned the answer in the past, and in a way she remembers it still, in the sense that the answer remains in her memory and is mentally accessible to her. The answer – Yerevan – is on the tip of her tongue. It seems to her that the name of the answer has a few syllables and begins with a “Y”. But occurrently believing that Yerevan is the capital of Armenia is not yet justified for her. If occurrently believing is to become justified for her, something must happen in her mental life. She needs to have a fitting “Aha!” moment, having the answer before her mind and something like recognizing it as such. Or it must seem to her that Yerevan is the capital of Armenia. Or she must recall that Yerevan is the capital of Armenia. Or some such. The recognition, seeming, and recollection here are the sorts of things that can fulfill strong internalism’s awareness condition on justification. In the absence of something like this recognition or seeming or recollection, occurrently believing that Yerevan is the capital of Armenia is not justified for the subject, even when that answer is on the tip of her tongue. If she happens to even entertain the correct answer (“Is it Yerevan?”), but does not have the “Aha!” moment, she has no reason to occurrently believe it. Many other cities have polysyllabic names beginning with a “Y” – Yakima, Yucaipa, Ypsilanti. The subject in this case is not justified in occurrently believing any of these is the capital of Armenia. And she has no more reason to occurrently believe that Yerevan is the capital of Armenia, so long as that answer remains just on the tip of her tongue.
In short, occurrently believing the relevant proposition in the typical tip of the tongue case is not justified for the subject. And occurrently believing and dispositionally believing are justified or unjustified together or not at all. So, dispositionally believing the relevant proposition is not justified in the typical case either. On option 1, however, the subject has both prima facie and overall justification. So option 1 is untenable.
Put differently, supposing option 1 is correct leads to a contradiction. Option 1 says the subject has both overall and prima facie justification for believing that which she dispositionally believes and has on the tip of her tongue. If believing for her is justified, it is not merely dispositionally believing that is justified. Occurrently believing is too. But occurrently believing is not justified. So the supposition is incorrect. The subject does not have overall and prima facie justification for believing in the typical tip of the tongue case. The common sense assessment of this sort of case is mistaken.
Right? Maybe not. Here’s a way, you might think, that the subject in the typical tip of the tongue case has prima facie and overall justification after all – a way to hang on to option 1. The subject’s justification is merely propositional, not doxastic. Believing the relevant proposition is justified for her, in the typical tip of the tongue case. But, because the answer is stuck on the tip of her tongue, her belief in the relevant proposition is not doxastically justified.
Still, abandon option 1. This way of hanging on to it fails. The idea is that the subject’s belief is doxastically unjustified in the tip of the tongue state, but the subject is still justified in believing its content. As we’ve done before, let’s return to a case and rewind the tape. What is the status of the subject’s belief that Yerevan is the capital of Armenia, prior to her entering the tip of the tongue state – prior to her thinking or inquiring about this proposition at all? Presumably, this dispositional belief is just like any of her others. If dispositional beliefs in general are justified, we should suppose her belief that Yerevan is the capital of Armenia is too, prior to her entering the tip of the tongue state. But then, on the proposal to retain option 1, this belief becomes doxastically unjustified when the subject enters the tip of the tongue state. Believing that Yerevan is the capital of Armenia is still propositionally justified for her, though. So what has happened? Apparently, by entering the tip of the tongue state, she ceases to believe on the basis of her propositional justification. Either she believes on some other basis, or she no longer has the belief at all.
Each possibility is preposterous. If she has discarded the belief, it’s a mystery why she’s in even the tip of the tongue state – why she feels she knows the answer to her inquiry; why she can correctly identify features of the answer, such as the letters and number of syllables in its name; why she is likely to have the “Aha!” moment when stumbling upon the answer.
So she still believes that Yerevan is the capital of Armenia. And the remaining lifeline for option 1 is that this belief has been rebased. But there’s no reason to suppose that entering the tip of the tongue state does this. The subject is trying to make occurrent a dispositional belief. Ordinarily, this effort does nothing to a belief’s basis. A better, but still incomplete, diagnosis is that the subject’s belief is simply failing for some reason to change modes. In general, when we search memory for information, our attitude toward that information should not take on a new basis during that very search and prior to our finding the information, particularly on a model in which memory retains dispositional beliefs. Consulting memory is not a shaking of a snow globe.
Denying that the subject has doxastic justification in the typical tip of the tongue case is a good idea. But it fails to make plausible that she has propositional justification. Reject option 1.
4.4.2 Tip of the Tongue Cases: Prima Facie Justification?
Option 2 is that the subject in the typical tip of the tongue case has prima facie justification for believing the relevant proposition, but not overall justification. This option may seem more appealing than option 1. It shares an attractive feature. Like option 1, option 2 attributes some positive epistemic status. But option 2 does not attribute any of the best epistemic statuses. Perhaps this concession accommodates our intuitions about ordinary tip of the tongue cases. What the subject in the case has is better than nothing, but not the best.
Option 2 breaks down into a pair of claims, and neither is plausible. The first in the pair is that the subject has prima facie justification for believing the relevant proposition. However, Section 4.4.1 casts doubt on this. This first claim of option 2 would incorrectly imply that the subject has something in favor of her occurrently believing the relevant proposition when it remains on the tip of her tongue. I won’t criticize this claim further. While its falsity is enough to scuttle option 2, perhaps my criticism of it was not fully persuasive. Perhaps it still seems there is justification in a typical tip of the tongue case, but the justification is just not as good as it could be. What I say in this section dismantles this take. The second claim of option 2 is that any justification the subject has for believing the relevant proposition is defeated. I will focus on this claim. If both claims in the pair are true, the subject has a defeater – something that limits her to having prima facie justification and not also overall justification.
At what point does the subject gain the defeater? And what is the defeater?
The best answer to this first question is that something about entering or being in the tip of the tongue state introduces the defeater. The subject is not entering the tip of the tongue state with the defeater already. After all, prior to entering the tip of the tongue state, her dispositional belief in the relevant proposition is just like any other in her memory. The content of the belief is like any other in her memory too. She has some support for believing the proposition. This support is in as good of shape as the support she has for anything else she reasonably dispositionally believes. Nothing, prior to her entering the tip of the tongue state, looks like a good candidate for a defeater. It’s implausible that the defeater is, say, the fact that the subject will enter the tip of the tongue state and will fail to retrieve the relevant proposition upon trying. Future contingencies do not defeat. (No Futurism!) Evidence of future contingencies can defeat, but the subject has no such relevant evidence before entering the tip of the tongue state.
So it is more plausible that being in the tip of the tongue state introduces a defeater. But what is the defeater? We will find only poor answers.
One kind of defeater suggests that the relevant proposition is false. This is a rebutting defeater. The other kind suggests certain support for believing the proposition fails. This is an undercutting defeater. It’s doubtful that the typical tip of the tongue case involves a rebutting defeater. It’s not as if the subject, upon entering the state, gains reason to believe that Yerevan is not the capital of Armenia! If she has a defeater in that state, it’s undercutting.
One possibility is that, while in the tip of the tongue state, the subject has evidence that she altogether lacks something critical. Perhaps the evidence suggests she doesn’t have any mental entity that could prima facie justify her in believing the relevant proposition, whatever that proposition may be. Having a disposition to recall the relevant proposition could, for instance, be the sort of mental entity that prima facie justifies believing the relevant proposition. And being in the tip of the tongue state may give the subject evidence that she has no such recollective disposition. It may also give her evidence that she has no other justification for believing the relevant proposition. This evidence may be misleading. But even so, it could defeat her prima facie justification.
It’s implausible that being in the tip of the tongue state provides this evidence. Being in the tip of the tongue state does not, for example, give the subject evidence that she lacks a disposition to recall the relevant proposition. In fact, being in the tip of the tongue state seems to confirm that she has the recollective disposition. On the best explanation available to her of her being in the tip of the tongue state, she is disposed to recall that relevant proposition. But in this instance, the disposition has led to the tip of the tongue state. Something simply impedes the full activation of this disposition. Something keeps her dispositional belief in the relevant proposition from becoming occurrent.
Maybe the undercutting defeater is something else. It’s not that the subject in the tip of the tongue state has the evidence described above. The subject does not have evidence that she altogether lacks something critical. The defeater is her evidence, roughly, that even if she has a mental entity that could normally justify her in believing the proposition, something about her present circumstances prevents that entity from justifying believing now. For instance, perhaps having a memory of learning that Yerevan is the capital of Armenia could ordinarily justify believing that Yerevan is the capital of Armenia. But perhaps the subject in the tip of the tongue state has evidence that her memory of this sort is inaccessible to her. Maybe this evidence of inaccessibility prevents the memory from providing her with overall justification.
This evidence could very well be a defeater. But it is doubtful the subject in the typical tip of the tongue case has it. The tip of the tongue phenomenology does not indicate that the relevant information is inaccessible. In fact, the phenomenology indicates the opposite. It feels as though the information is accessible – it’s right there, on the tip of the tongue. A feeling of its accessibility in part explains why the subject may persist in trying to mentally access the information rather than giving up or consulting an external store for the information. Of course, the subject in the tip of the tongue state is, despite trying, failing to access the information on the tip of her tongue. But it doesn’t follow that this information is inaccessible to her.
Also, suppose that last paragraph was hogwash. Suppose the subject has evidence that the information is inaccessible to her. Suppose that a one-off episode of retrieval failure provides this evidence. Still, at best it is weak evidence of inaccessibility. A single, brief failure only whispers against any chance of success. Maybe this access failure helps explain some reduction in the subject’s justification for believing the relevant proposition. But it explains too little. If common sense is correct, in many cases the subject’s justification for believing the relevant proposition, prior to her entering the tip of the tongue state, had been fairly high. In some cases it had been knowledge-level, high enough to fulfill the justification requirement for knowing. And knowing requires especially strong justification. The subject might dispositionally know the relevant proposition, prior to entering the tip of the tongue state. So even if the subject’s access failure in the tip of the tongue state plays some defeating role, it does not completely defeat what might otherwise have been knowledge-level justification for believing the relevant proposition. The defeater is too weak. On option 2 is there is supposed to be total defeat. We have not found a decent account of what does this defeating.
According to option 2, the subject in the typical tip of the tongue case has prima facie justification for believing the relevant proposition. But she lacks overall justification. We have found no good account of why the subject lacks overall justification. It is also worth noting that option 2 likely had only false appeal. Initially, option 2 may seem attractive because it is moderate. It attributes justification. But the justification is defeated. So the subject does not know the relevant proposition. Still, option 2 is as hopeless as option 1. Whatever we suppose the subject’s defeater is, on option 2, we can imagine cases in which that defeater is itself defeated. Suppose, for instance, the defeater is this. While in the tip of the tongue state, the subject has evidence that her relevant memories that support believing the relevant proposition are inaccessible to her. Because she has this evidence of this inaccessibility, she does not have overall justification for believing the relevant proposition. Now suppose a known expert on the subject’s memory testifies to her that the relevant information is indeed accessible to her. The expert cheers the subject on in her attempt to remember. “Keep trying!” The relevant memory’s apparent inaccessibility is misleading, the expert explains. The subject’s defeater, then, is defeated.
In cases with this testimony, option 2 reaps what option 1 sows: The subject has prima facie and overall justification, even for occurrently believing the relevant proposition.
If option 2 had seemed better than option 1, it might have been because option 2 predicts no overall justification for occurrent believing in typical tip of the tongue cases. It made sense of the intuition that occurrent believing is not overall justified without an appropriate “Aha!” moment, seeming recollection, or recognition of the answer. But in some cases, option 2 indeed allows for overall justification without these. Option 2 allows for this when the subject in the tip of the tongue state has a defeated defeater. These may not be typical tip of the tongue cases. But they are similar enough. Attributing overall justification for occurrent believing here is bad enough.
Option 2’s first claim shares the problems of option 1. Option 2’s second claim is difficult to support. And option 2’s initial appeal is misleading. Option 2 allows the subject in some tip of the tongue cases to have overall justification for believing the relevant proposition, even when the target information remains on the tip of her tongue. Reject option 2.
4.4.3 Tip of the Tongue Cases: No Justification
On option 3 the subject in the typical tip of the tongue case does not have prima facie justification. It follows that the subject lacks overall justification. On this option the common sense assessment is incorrect. And on this option, strong internalism looks better supported than weak internalism. Here is why.
Some versions of weak internalism predict that the subject in the tip of the tongue state has prima facie justification for believing the relevant proposition. The subject has this justification because she has some unaccessed mental entity with alleged justifying power. But the option we’re exploring, option 3, stipulates that believing the relevant proposition is not prima facie justified. Versions of weak internalism that imply otherwise imply something false here. So these versions are incorrect. Any defended version of strong internalism, however, predicts in accordance with the stipulation. It says the typical subject in the tip of the tongue state lacks justification, and that’s just what’s so on the present option. Strong internalism is better supported than these versions of weak internalism.
Of course, not all versions of weak internalism predict incorrectly. Not all versions attribute the subject in the tip of the tongue state with prima facie justification for believing the relevant proposition. On these versions of weak internalism, certain unaccessed mental entities have justifying power. The typical subject in the tip of the tongue state, for whatever reason, happens to lack these mental phenomena. So these internalist theories correctly imply that the subject lacks prima facie justification for believing the relevant proposition.
These internalist theories look worse than strong internalism. They are needlessly complex. And they are implausible.
Strong internalism, as noted, implies an absence of prima facie justification in the typical tip of the tongue case. The subject in the tip of the tongue state is not accessing a relevant mental justifier. So she has no justification for believing the relevant proposition.
The weak internalist theories that share the implication are bloated. They say some unaccessed mental entities still justify. But saying this does no clear good. It accounts for no justification, at least not here. The unaccessed mental entities do not confer even prima facie justification in typical tip of the tongue cases, cases where it is most plausible the subject has justified dispositional belief. And, other than in cases of dispositional belief, it’s unclear that citing unaccessed mental phenomena as justifiers is helpful. So attributing justificatory power to any unaccessed mental entity looks unnecessary. Strong internalism accounts for enough justification without this attribution.
Additionally, as noted, the weak internalist theories that do not attribute justification in the typical tip of the tongue case look implausible. Why, on these theories, does the subject in this case lack the right sort of mental justifier, while subjects in other sorts of cases have it? It’s hard to see a good answer. Tip of the tongue cases are among the most credible instances of justified dispositional belief. These internalist theories say there are unaccessed mental justifiers, but not where it would be most plausible. These theories offer no good reason for their claims that render them incompatible with strong internalism. They have little appeal.
Here’s a perhaps tempting way to plug the hole of these weak internalist theories: Deny that there is a dispositional belief in the typical tip of the tongue case. If there is no dispositional belief, perhaps it is less clear that the subject has prima facie justification for believing the relevant proposition. We might have thought there is justified dispositional belief in these cases only because we thought that there was belief, and also that the subject is not holding her attitude irrationally. But if there is no belief, then perhaps there is no justification for believing to account for. On this response, weak internalism looks fine when attributing no justification in the typical tip of the tongue case. There’s no belief in this case anyway. It is not one of those most plausible instances of justified dispositional belief after all.
The plug doesn’t fit. In many typical tip of the tongue cases, the subject believed the relevant proposition in the past, and is moments away from occurrently believing the relevant proposition. What explains this? It could be that belief in the relevant proposition disappeared at some point and is about to be generated again. Or it could be that this belief did not disappear and is currently dispositional. In moments it will simply switch modes, becoming occurrent. If there are such things as dispositional beliefs, and if they abound as much as we tend to think they do, it’s simpler to suppose that belief in the relevant proposition has endured since its most recent formation, that it did not for some reason disappear only to be generated again soon. It’s more plausible that the belief is just switching modes. The tip of the tongue experience may simply involve awareness of some difficulty in this switch. Still, there’s likely dispositional belief in the relevant proposition, if there are dispositional beliefs at all.
The attempt to plug the hole expresses a worry: Perhaps weak internalism is better supported on option 3 than I say. A different worry is this: Perhaps strong internalism is worse supported on option 3 than I say. Perhaps, contrary to what I say, strong internalism also implies that there is prima facie justification in a typical tip of the tongue case. I noted that in the typical tip of the tongue case, the subject feels she knows the answer to her inquiry. Maybe, on strong internalism, this feeling or something like it justifies her in believing the target of her inquiry. If so, option 3 does not favor strong internalism. Option 3 denies there is this justification.
It’s plausible that the feeling or seeming to know the answer matters for justification. It could help justify the subject in believing, specifically: That she knows the answer. And this fits with strong internalism. The subject is accessing or aware of a seeming, and so she could have justification for believing the content of the seeming, namely: That she knows the answer. But this seeming does not itself specify the answer. The seeming leaves the answer wide open. The seeming indicates just that the subject knows it, whatever it is. So, even if the subject knows the answer, it’s implausible that this seeming itself justifies the subject in believing it. But here’s the rub. The subject is not clearly accessing anything else that justifies her in believing the answer. She’s accessing nothing, then, that provides this justification. So it looks like strong internalism still fits option 3. It does not imply the subject in the typical tip of the tongue case has prima facie justification. (What’s more, on strong internalism, if the subject’s seeming is indeed that she knows the answer, the seeming is likely misleading. She does not know the answer, not at that very moment. Likely, the answer is accessible, and if she accesses it, she can easily come to know it. However, her seeming may be slightly different from how we’ve characterized it so far. With some of these differences, it is likely accurate. If, for example, it seems to her that the answer is accessible to her, or that she nearly knows the answer, or that she is on the verge of knowing it, then likely the seeming is correct.)
Someone could develop a version of strong internalism that conflicts with option 3. For example, a version could imply that, in the typical tip of the tongue case, by seeming to know the answer, the subject is accessing something that justifies her in believing the answer. But, as I just explained, that’s an implausible implication. That version of strong internalism is implausible as a result. So we’d simply end up in the following situation: Some implausible versions of strong internalism conflict with option 3. But the best versions fit option 3. So option 3 confirms the best versions. That’s a case for those versions of strong internalism. And it’s a notable outcome.
We’ve looked at option 3: The typical subject in the tip of the tongue state lacks both overall and prima facie justification for believing the relevant proposition. Strong internalism makes sense of this. Some versions of weak internalism fail to make sense of this. The rest are less plausible than strong internalism. Option 3 favors strong internalism. And option 3 is not simply a last resort. It is also tenable. We may be reluctant to embrace it because it upsets expectations. It denies justification in tip of the tongue cases. But now that we see what those expectations lead to, they look unreasonable. There is no reason to reject option 3.
We’ve considered the epistemic status of believing the relevant proposition for the subject in the typical tip of the tongue case. Believing for her could be both prima facie and overall justified. It could be just prima facie justified. Or it could be neither prima facie nor overall justified. Each of these options either has implausible consequences or supports strong internalism. This counts in favor of strong internalism.
4.5 Leaving the Tip of the Tongue Cases
The only plausible interpretation of the standard tip of the tongue case is one that confirms strong internalism. This interpretation is that the subject does not have even prima facie justification for believing the relevant proposition while the target of her inquiry remains on the tip of her tongue. This outcome is exactly what strong internalism predicts. And it is the best internalist option that predicts it. Some versions of weak internalism are incompatible with this outcome, and so they appear false. And the versions of weak internalism that are compatible with this outcome appear unnecessarily complicated and unmotivated. The tip of the tongue phenomenon supports strong internalism over the more permissive forms of internalism.
As mentioned, it seemed that if there are such things as dispositional beliefs, the relevant proposition in the typical tip of the tongue case was among the best candidates for what we might have justified dispositional belief in. However, there is no good story of this justification on the weak internalism. And accounting for the alleged justification of other dispositional beliefs is only more challenging. The subject does not have the tip of the tongue phenomenology that might justify them. (Imagine something like what we considered in the tip of the tongue case, but with an ordinary dispositional belief: The belief becomes occurrent without the subject recalling it, recognizing its correctness, and so on. The belief looks random, arbitrary, and unjustified. It has nothing more going for it than the belief of a clairvoyant [BonJour Reference Bonjour1980], who believes by some unwittingly reliable process.) If there are dispositional beliefs, they are unjustified.
This is not necessarily a skeptical outcome. Once we accept that there is no justification for dispositional beliefs, all it takes to avoid skepticism is to stop supposing we have dispositional beliefs in the first place. Our doxastic attitudes remain generally rational. It’s just that we have fewer beliefs than we tend to think.
As the pool of allegedly justified dispositional beliefs shrinks, the support for strong internalism over weak internalism swells. Strong internalism gets the cases right most elegantly.
5 Conclusion
We began large and ended small. We’ve looked everywhere and at each moment for what might matter for how memory provides justification. The longer we looked, less and less turned out to matter. The future and past do not matter. Only the present does. The way things happen to be beyond the mind doesn’t matter. Only the mind does. Most of the mind, however, does not matter. Only the accessed elements do. Memory justifies just by way of what’s accessed mentally now. The epistemology of memory provides a case for a time-slice and strong version of internalism. Strong internalism deserves more credit than it gets. If the epistemology of memory supports it, we should take it more seriously elsewhere. It may be the best theory of justification overall.
Acknowledgments
For Rowan and Freddie, memorable for all the right reasons. I thank Kevin McCain and Andrew Moon for helpful comments on drafts of parts of this Element. For helpful conversations, I thank audiences at the Midwest Epistemology Workshop at Washington University in Saint Louis, the Issues in Philosophy of Memory 3 Conference at Duke University, the Graduate Epistemology Conference at the University of Rochester, and the Southeastern Epistemology Conference at the University of Florida.
Stephen Hetherington
University of New South Wales, Sydney
Stephen Hetherington is Professor Emeritus of Philosophy at the University of New South Wales, Sydney. He is the author of numerous books, including Knowledge and the Gettier Problem (Cambridge University Press, 2016), and What Is Epistemology? (Polity, 2019), and is the editor of several others, including Knowledge in Contemporary Epistemology(with Markos Valaris: Bloomsbury, 2019), and What the Ancients Offer to Contemporary Epistemology (with Nicholas D. Smith: Routledge, 2020). He was the Editor-in-Chief of the Australasian Journal of Philosophy from 2013 until 2022.
About the Series
This Elements series seeks to cover all aspects of a rapidly evolving field, including emerging and evolving topics such as: fallibilism; knowing how; self-knowledge; knowledge of morality; knowledge and injustice; formal epistemology; knowledge and religion; scientific knowledge; collective epistemology; applied epistemology; virtue epistemology; wisdom. The series demonstrates the liveliness and diversity of the field, while also pointing to new areas of investigation.
