What should we believe about the future?

This paper discusses the ability of explanationist theories of epistemic justification to account for the justification we have for holding beliefs about the future. McCain’s explanationist account of the relation of evidential support is supposedly in a better position than other theories of this type to correctly handle cases involving beliefs about the future. However, the results delivered by this account have been questioned by Byerly and Martin. This paper argues that McCain’s account is, in fact, able to deliver plausible results in cases involving such beliefs and that explanationism, if properly articulated, is illuminating with respect to the justification we have for holding such beliefs, as it manages to correctly distinguish evidence that only supports believing probabilistic claims about the future from evidence that is sufficient to believe that a particular event will happen.


Introduction
Niels Bohr is supposed to have said that "prediction is very difficult, especially about the future". Besides the humoristic touch of this statement, there is indeed something peculiar concerning beliefs about the future. This is, it seems, because of the particular type of evidence we have for holding such beliefs. In contrast to beliefs about the past, we cannot rely directly on our memories to form an opinion about what will happen. No perception of future events can be remembered. In contrast to beliefs about the present, B Miloud Belkoniene miloud.belkoniene@unifr.ch 1 Université de Fribourg, Avenue de l'Europe 20, 1700 Fribourg, Switzerland we cannot rely directly on our own current experiences to form an opinion about what will happen. Future events cannot be perceived. But surely, if we can rationally hold beliefs about the future, it has to be because our present and past evidence supports, in some cases, believing propositions about future events.
Several philosophers have recently argued that explanationist theories of epistemic justification, in particular, fail to account for the support that beliefs about the future can receive from past and present evidence. While such theories have sparked the interest of some philosophers because of the results they deliver in other kinds of cases, their supposed inability to account for the justification we have for holding beliefs about the future casts serious doubts on their plausibility. This paper discusses a problem for explanationism raised by beliefs about the future by examining whether McCain's (2013McCain's ( , 2014b definition of the evidential support relation can correctly account for a case involving such beliefs offered by Byerly (2013). Its aim is to show that the results delivered by McCain's account in cases involving beliefs about the future are, in fact, plausible and that explanationism, if properly articulated, is illuminating with respect to the justification we have for holding such beliefs, as it manages to correctly distinguish evidence that only supports believing probabilistic claims about the future from evidence that is sufficient to believe that a particular event will happen.
To that end, I review, in Sect. 2, McCain's account of the evidential support relation and the problem that Byerly's case raises for it. In Sect. 3, I introduce the near neighborhood strategy considered by McCain to show that his account is, in the final analysis, able to meet the challenge constituted by Byerly's case, and I present Byerly and Martin's response to it. I then show that, if properly articulated, this strategy can succeed despite Byerly and Martin's concerns regarding it. In Sect. 4, I argue that the result delivered by McCain's account in the case offered by Byerly, given the near neighborhood strategy, has independent plausibility in light of the fact that statistical evidence, on its own, cannot support believing non-probabilistic claims. Finally, in Sect. 5, I show that McCain's account of the evidential support relation is able to correctly track the distinction that can be made between statistical and non-statistical evidence and that this account is, therefore, able to correctly distinguish cases in which one is justified in believing that a particular event will occur from cases in which one is only justified in believing that there is a certain chance that a particular event will occur.

Explanationism and beliefs about the future
The renewed interest in explanationist conceptions of epistemic justification gave rise to a debate between, inter alia, McCain on the one side and Byerly and Martin on the other side, about the ability of explanationism to account for the justification we have for holding beliefs about the future. Explanationism takes epistemic justification to be fundamentally a matter of explanatory considerations. This is because the relation of evidential support which is both necessary and sufficient for some evidence to support believing that something is the case is taken by explanationists to be definable in terms of relations of best explanation. Thus, in its simplest form, explanationism about epistemic justification states that some evidence e supports believing that p if and only if p is part of the best explanation available to a subject with e for why she has e. 1 Beliefs about the future are particularly difficult for explanationism to handle, for the following reason: if one's overall evidence is constituted by events that precede in time the state of affairs represented by p, then p cannot contribute to an explanation of that evidence. As beliefs about the future, if supported by one's evidence, are always supported by some evidence constituted by past and present events, it is doubtful that one's evidence can support believing that something will happen only because of the explanatory relation it bears to the content of that belief. 2 To illustrate this problem for explanationism, Byerly offers the following case: Golfer Case: Suppose I'm on the golf course on a sunny, calm day. My putting stroke has been working for me most of the day, and I'm now on the sixteenth green. It's not a long putt-just six feet. I'm fairly confident. I rotate my shoulders, pulling the putter back, and then accelerate through the ball. It rolls toward the cup. The speed looks good. The line looks on. Yes, I believe it's going in! (Byerly 2013, p. 235) According to Byerly, the golfer is justified in believing that the ball will go into the cup, given the evidence that is available to him in this case. This evidence consists, presumably, of his present observation of the golf ball rolling toward the cup and of his past observations of golf balls rolling toward a cup in similar circumstances. But, as the fact that the ball will go into the cup cannot contribute to explain these past and present observations, explanationism appears unable to account for the justification the gofer has for believing that the ball will go into the cup. While Byerly's Golfer Case undeniably constitutes a serious challenge for the simplest form of explanationism, more sophisticated versions of this view might be in a better position with respect to such cases. Other cases offered in the epistemological literature 3 led McCain (2013McCain ( , 2014b to argue that the relation of evidential support cannot be defined only in terms of relations of best explanation. According to him, a subject's evidence does not only support believing the propositions that are part of the best explanation available to her for that evidence. A subject's evidence also supports believing the propositions that are entailed by this explanation. McCain's (2015, p. 334) account, which he calls 'Explanationist Evidentialism' (EE), thus defines the evidential support relation in terms of both relations of best explanation and relations of logical consequence: EE: A person, S, with evidence e at t is justified in believing p at t iff at t S has considered p and either: 1 Conee and Feldman (2008) endorse such a view. 2 The problem for explanationism is, therefore, more general than cases involving beliefs about the future. But as this problem is particularly salient in such cases, the debate has tended to focus on the issue of beliefs concerning future events. Note, however, that any solution that can show how explanationism can correctly handle beliefs about the future will generalize to other cases as well (thanks to an anonymous referee for pointing this out). 3 McCain (2013, pp. 300-301) reacts specifically to a case offered by Lehrer (1974, p. 166) and a case offered by Goldman (2011, pp. 277-278). (i) p is part of the best explanation available to S at t for why S has e, or (ii) p is available to S as a logical consequence of the best explanation available to S at t for why S has e.
According to McCain (2014a), because EE defines the evidential support relation in terms of relations of logical consequence as well as in terms of relations of best explanation, it is able to handle correctly Byerly's case and, more generally, any case involving beliefs about the future. To show this, McCain distinguishes two possible version of the Golfer Case.
In the first version, the golfer observed that all the golf balls that have rolled toward the cup in circumstances C have gone into the cup and is observing that the golf ball is rolling toward the cup in circumstances C. McCain suggests that, in this configuration of the case, the best explanation available to the golfer for his evidence plausibly includes the generalisation 'all golf balls rolling toward the cup in circumstances C go into the cup' and the proposition 'the golf ball is currently rolling toward the cup in circumstances C'. It is, indeed, reasonable that the best explanation available to the golfer includes these two propositions as the first, if true, would explain why all the observed golf balls rolling toward the cup in circumstances C have gone into the cup and the second, if true, would explain the golfer's current visual experience. Now, if the best explanation available to the golfer for her evidence includes these two propositions, EE straightforwardly delivers the result that the golfer is justified in believing that the ball will go into the cup, as the proposition 'the ball will go into the cup' is entailed by these propositions. Thus, in the first possible version of the case, condition (ii) of EE is easily met.
In the second version of the Golfer Case, the golfer observed that most but not all of the golf balls that have rolled toward the cup in circumstances C have gone into the cup. This version is more difficult for EE to handle because the generalisation 'all golf balls rolling toward the cup in circumstances C go into the cup' can obviously not be part of the best explanation available to the golfer for his evidence. McCain (2014a, pp. 106-108) considers three possible strategies that are all deemed ultimately unsatisfying by Byerly and Martin (2015, pp. 778-781).) In reaction to Byerly and Martin's concerns, McCain (2015, p. 339) amends EE with respect to its second condition and offers a new definition of the evidential support relation which he takes to be in a better position regarding the second version of the Golfer Case: EE*: A person, S, with evidence e at t is justified in believing p at t iff at t S has considered p and either: (i) p is part of the best explanation available to S at t for why S has e, or (ii) p is available to S as an explanatory consequence of the best explanation available to S at t for why S has e. 4 Further objections have been raised by Byerly and Martin (2016) against EE* to which McCain (2017) recently responded. In the present paper, however, I will focus on one of the strategies considered by McCain for EE to handle the second version of the Golfer Case, which has been labelled 'the near neighborhood strategy' by Byerly and Martin. I will argue that, contrary to what Byerly and Martin claim, this strategy succeeds and actually shows that EE is illuminating regarding the justification we have for holding beliefs about the future.

The near neighborhood strategy
The near neighborhood strategy aims to show that, although EE does not deliver the result that the golfer is justified in believing the proposition 'the ball will go into the cup' in the second version of the Golfer Case, it delivers the result that he is justified in believing a proposition that is in the near neighborhood of this proposition-that is, that 'the ball will probably go into the cup'. For this strategy to succeed, it has to be the case that the proposition 'the ball will probably go into the cup' is a logical consequence of the best explanation available to the golfer for his evidence. In McCain's (2014a, p. 107) view, in the second version of the Golfer Case, it is reasonable to think that the best explanation available to the golfer for his evidence includes the propositions 'most golf balls rolling toward the cup in circumstances C go into the cup' and 'the golf ball is currently rolling toward the cup in circumstances C'. As he takes the proposition 'the ball will probably go into the cup' to be a logical consequence of these two propositions, McCain argues that EE delivers the result that the golfer is justified in believing that the ball will probably go into the cup and that, consequently, this account can handle the Golfer Case. Byerly and Martin (2015, pp. 778-779) are ready to grant McCain that the propositions 'most golf balls rolling toward the cup in circumstances C go into the cup' and 'the golf ball is currently rolling toward the cup in circumstances C' are part of the best explanation available to the golfer for his evidence. However, they deny that these propositions entail that 'the ball will probably go into the cup'. The reason is that a conjunction of the form 'most Fs are Gs and x is a F' does not entail a proposition of the form 'probably x is a G'. In support of this, they consider the following case: Sally is a woman over 35. Suppose most women over 35 are unable to run a 6-min mile. Do these claims entail that it is probable that Sally is unable to run a 6-min mile? They do not. Entailment is supposed to be monotonic. If p entails q, then for any r, p ∧ r entails q. But, suppose that in addition to being a woman over 35, Sally is a world-class Olympic runner, and that almost all world-class Olympic runners are able to run 6-min miles. If anything, then, it is likely that she can run a 6-min mile. (Byerly and Martin 2015, pp. 778) While Byerly and Martin are right in pointing out that the propositions 'most golf balls rolling toward the cup in circumstances C go into the cup' and 'the golf ball is rolling toward the cup in circumstances C' do not entail the proposition 'the ball will probably roll into the cup', I do not think that this undermines the viability of the near neighborhood strategy. This is because the propositions which are part of the best explanation available to the golfer for his evidence, and which are relevant for that strategy to succeed, are not the ones considered by McCain. To see why, let me first underline that the term "probably" in the proposition 'the ball will probably go into the cup' represents the objective chance of some event to occur. In fact, McCain considers the near neighborhood strategy as an alternative to another one that relies on an epistemic notion of probability. Therefore, in the context of the near neighborhood strategy, the proposition 'the ball will probably go into the cup' is equivalent to the proposition 'the chance that the ball will go into the cup is higher than 50%'.
Thus, for the near neighborhood strategy to succeed, what needs to be shown is that the best explanation available to the golfer for his evidence includes some propositions that jointly entail the proposition 'the chance that the ball will go into the cup is higher than 50%'. Recall that, in the second version of the Golfer Case, the golfer's evidence consists of her observations that most golf balls that have rolled toward the cup in circumstances C have gone into the cup and that the ball is rolling toward the cup in circumstances C. As outlined by McCain, it is reasonable to think that the proposition 'the ball is rolling toward the cup in circumstances C' is part of what could best explain these observations. It is also reasonable to think that the proposition 'most golf balls that have rolled toward the cup in circumstances C have gone into the cup' is part of that explanation. But, most importantly, it is plausible that the proposition 'all golf balls rolling toward the cup in circumstances C have a chance to go into the cup higher than 50%' is part of that explanation. This proposition, if true, would surely contribute to explaining why the golfer observed that the proportion of golf balls that have rolled toward the cup in circumstances C and gone into the cup is higher than the proportion of golf balls that have rolled toward the cup in the same circumstances and not gone into the cup. 5 Propositions that concern the chance of an event that belongs to a certain class occurring are, in fact, often cited to explain the observed proportion of events that belong to that class. Now, if the propositions 'all golf balls rolling toward the cup in circumstances C have a chance to go into the cup higher than 50%' and 'the golf ball is currently rolling toward the cup in circumstances C' are part of the best explanation available to the golfer for his evidence, it is clear that the proposition 'the chance that the ball will go into the cup is higher than 50%' is a logical consequence of that explanation. In addition, as, in the context of the near neighborhood strategy, the proposition 'the chance that the ball will go into the cup is higher than 50%' is equivalent to the 5 One might object that an explanation containing the proposition 'most golf balls rolling toward the cup in circumstances C have a chance to go into the cup (significantly) higher than 50%' is an equally good explanation of the golfer's observations. After all, this proposition could also contribute to explaining why the golfer observed that most golf balls that have rolled toward the cup in circumstances C have gone into the cup. Note that this constitutes an objection only if 'most golf balls rolling toward the cup in circumstances C have a chance to go into the cup (significantly) higher than 50%' entails that a minority of the golf balls rolling toward the cup in circumstances C have a chance to go into the cup lower than 50%. Otherwise, this proposition is equivalent to the proposition 'all golf balls rolling toward the cup in circumstances C have a chance to go into the cup higher than 50%'. It is, however, doubtful that an explanation which entails that a minority of the golf balls rolling toward the cup in circumstances C have a chance to go into the cup lower than 50% is an equally good explanation. If the golfer simply observed that most golf balls that have rolled toward the cup in circumstances C have gone into the cup, there is no reason to suppose that a minority of golf balls rolling toward the cup in these circumstances have a chance to go into the cup lower than 50%. An explanation stating that all golf balls rolling toward the cup in these circumstances have a chance to go into the cup higher than 50% is by far the simplest explanation of the golfer's observations. proposition 'the ball will probably go into the cup', this shows that the proposition 'the ball will probably go into the cup' is a logical consequence of the best explanation available to the golfer for his evidence. Given this articulation of the near neighborhood strategy, the proposition 'the ball will probably go into the cup' is not said to be a logical consequence of that explanation because, supposedly, a conjunction of the form 'most Fs are Gs and x is a F' entails a proposition of the form 'x is probably a G'. Instead, it is said to be a logical consequence of that explanation because a conjunction of the form 'all Fs are Gs and x is a F' entails a proposition of the form 'x is a G'. Thus, contrary to what McCain claims, the near neighborhood strategy does not succeed because the best explanation available to the golfer for his evidence in the second version of the Golfer Case contains the proposition 'most golf balls rolling toward the cup in circumstances C go into the cup' but, instead, because it contains the proposition 'all golf balls rolling toward the cup in circumstances C have a chance to go into the cup higher than 50%'.

The problem of statistical evidence
Although the near neighborhood strategy succeeds in showing that, in the second version of the Golfer Case, EE delivers the result that the golfer is justified in believing that the ball will probably go into the cup, Byerly and Martin could still claim that this result is ultimately unsatisfying. After all, they offered the Golfer Case because they take the golfer to be justified in believing that the ball will go into the cup in both versions of the case and not merely in believing that it will probably go into the cup. They could, therefore, argue that the near neighborhood strategy, even if successful, is not satisfying as it does not show how EE can account for the fact that the golfer is justified in believing that the ball will go into the cup in the second version of the Golfer Case.
To counter this possible objection, one has to provide independent reasons for thinking that, in the second version of the Golfer Case, the golfer is only justified in believing that the ball will probably go into the cup. In support of that conclusion McCain (2014a, p. 107) considers the view according to which knowledge requires probability 1. He claims that if such a view turns out to be correct, then, as epistemic justification is, presumably, what makes a belief a good candidate for knowledge, it is reasonable to think that justification also requires probability 1. As, in the second version of the Golfer Case, the probability that the ball will go into the cup given the available evidence is lower than 1, this could support the conclusion that the golfer is only justified in believing that the ball will probably go into the cup. However, I do not think that one has to endorse such a demanding view of knowledge and epistemic justification to have reasons to think that the golfer, in the second version of the Golfer Case, is only justified in believing that the golf ball will probably go into the cup. Instead, one needs to pay attention to the statistical nature of the evidence available to the golfer in that version of the case and to the fact that such evidence can only enable someone to know that there is a certain chance that a particular event will occur. Philosophers distinguish statistical evidence from non-statistical evidence because statistical evidence is particularly problematic when it comes to knowledge and epistemic justification. 6 Consider, for instance, a fair lottery involving a high number of tickets. Suppose that a participant in this lottery knows that n tickets were sold and that only one of them is a winning ticket. Is this participant in a position to know, solely on the basis of her evidence-i.e. what she knows about the lottery-that her ticket or any of the n lottery tickets will turn out to be a losing ticket? It seems not. But what prevents her from knowing that a particular lottery ticket will turn out to be a losing ticket? After all, given her evidence, the chance that any of the n lottery tickets will turn out to be a losing ticket is quite high. According to several philosophers, what prevents the lottery participant from knowing that a particular lottery ticket will turn out to be a losing ticket is the purely statistical nature of the evidence available to her. 7 As this evidence only conveys information about the proportion of lottery tickets that will turn out to be losing tickets relative to the proportion of lottery tickets that will turn out to be winning tickets and, in this sense, is purely statistical, it can only enable the lottery participant to know that there is a high chance that a particular lottery ticket will turn out to be a losing ticket.
In support of this, let me consider another case introduced in the literature concerning legal standards of proof to motivate the distinction between statistical and non-statistical evidence: Blue Bus Case: Suppose it is late at night and an individual's car is hit by a bus. This individual cannot identify the bus, but she can establish that it is a blue bus, and she can prove as well that 80 percent of the blue buses in the city are operated by the Blue Bus Company, that 20 percent are operated by the Red Bus Company, and that there are no buses in the vicinity except those operated by one of these two companies. Moreover, each of the other elements of the casenegligence, causation, and, especially, the fact and the extent of the injury -is either stipulated or established to a virtual certainty. (Schauer 2003, pp. 81-82) The evidence available in this case is also statistical in the sense that it only conveys information about the proportion of buses operated in the city by the Blue Bus Company relative to the proportion of buses operated in the city by the Red Bus Company. In addition, it is quite clear that a subject could not know that the bus responsible for the accident belongs to the Blue Bus Company solely on the basis of the evidence available in this case. Knowing that a higher proportion of buses operated in the city belong to the Blue Bus Company cannot enable someone to know that the bus responsible for the accident belongs to the Blue Bus Company. It can only enable one to know that there is a higher chance that the bus responsible for the accident belongs to that company. The reason need not be that knowledge requires probability 1. Nor is it, as Smith (2017, p. 17) emphasizes, that the nature of the circumstances described in the Blue Bus Case would make one's true belief that the bus responsible for the accident belongs to the Blue Bus Company lucky. Irrespective of the circumstances, the evidence available in the Blue Bus Case is not sufficient, on its own, to enable someone to know that the bus responsible for the accident belongs to the Blue Bus Company. This is because statistical evidence alone is not sufficient to support believing non-probabilistic claims.
In the second version of the Golfer Case, the evidence available to the golfer is no less statistical than the evidence available in the Blue Bus Case. It only conveys information about the proportion of golf balls that have rolled toward the cup in circumstances C and gone into the cup relative to the proportion of golf balls that have rolled toward the cup in the same circumstances and not gone into the cup. As there are reasons to deny that the evidence available in the Blue Bus Case can enable someone to know that the bus responsible for the accident is operated by the Blue Bus Company because of its purely statistical nature, there are also reasons to deny that the evidence available to the golfer in the second version of the Golfer Case can enable him to know that the ball will go into the cup. This is so, not because knowledge requires probability 1 but, instead, because statistical evidence alone can only enable someone to know that there is a certain chance that a particular event will occur. Thus, to have reasons to think that the golfer, in the second version of the Golfer Case, is only justified in believing that the golf ball will probably go into the cup, one does not need to endorse the view that knowledge requires probability 1. One only needs to pay attention to the fact that statistical evidence cannot, on its own, enable someone to have non-probabilistic knowledge.

Statistical evidence and beliefs about the future
The statistical nature of the evidence available in the second version of the Golfer Case is a reason to deny that EE should deliver the result that the golfer is justified in believing that the ball will go into the cup. Yet, one might worry that this actually leads to a generalized skepticism concerning beliefs about the future as the evidence we have for holding such beliefs often seems to be statistical. In other words, one might worry that the response developed in the previous section actually amounts to arguing that, in fact, the only things that can be known and rationally believed about the future are probabilistic claims.
To dispel this worry, let me first show that EE correctly tracks the distinction between statistical and non-statistical evidence in cases that do not involve beliefs about the future. Consider Buchak's variation of the Blue Bus Case: Green Bus Case: Suppose it is late at night, and an individual's car is hit by a green bus. The two bus companies in the area, the Green Bus Company and the Yellow Bus Company, each operate 50 percent of the green busses. There is an eyewitness, who identifies the bus as belonging to the Green Bus Company (the two bus companies operate busses with distinctive shapes). It is night-time, and so her vision is not ideal: let us say she makes mistakes 25% of the time. All of the other elements of the case remain the same. (Buchak 2014, p. 291) Given the eyewitness testimony, the overall evidence available in this case is not statistical and appears sufficient to support believing that the bus responsible for the accident belongs to the Green Bus Company. EE conforms with this intuition. In the absence of other evidence that could be best explained by the fact that the eyewitness lied or misperceived, the best explanation for the evidence available in this case plausibly contains the proposition 'the eyewitness saw that the bus responsible for the accident belongs to the Green Bus Company'. As this proposition entails the proposition 'the bus responsible for the accident belongs to the Green Bus Company', EE delivers the result that the evidence available in this case is sufficient to support believing that the bus responsible for the accident belongs to the Green Bus Company. In contrast, the proposition 'the bus responsible for the accident belongs to the Blue Bus Company' is neither part of, nor entailed by, what could best explain the evidence available in the Blue Bus Case. Thus, in cases like the Blue Bus Case and the Green Bus Case, EE correctly tracks the distinction between statistical and non-statistical evidence. 8 Now, recall that, in the first version of the Golfer Case, EE does not deliver the result that the golfer is only justified in believing a probabilistic claim about the future but, instead, delivers the result that he is justified in believing that the ball will go into the cup. In light of EE's ability to track the distinction between statistical and non-statistical evidence in other cases, this supports the conclusion that not all the evidence we have for holding beliefs about the future is statistical. If the best explanation available to a subject for her evidence entails a non-probabilistic claim about a future event, then that subject's evidence is not purely statistical with respect to that claim and supports believing that an event will occur. Let me highlight the fact that while in the first version of the Golfer Case, the best explanation available to the golfer for his evidence entails a proposition about the future by virtue of containing a universally quantified proposition, EE does not entail that a subject is justified in believing something about the future only when she is also justified in believing a universally quantified proposition. Suppose, for instance, that I am waiting for a shop to open and someone passing by tells me that the shop will open at 9 a.m. If I have no particular reason to suspect that this person is lying to me or that she is not in a position to know that the shop will open at 9 a.m., it seems that I am justified in believing that the shop will open at 9 a.m. on the basis of this person's testimony. Now, the best explanation I have in this case for the testimonial evidence I possess contains the proposition 'the person who told me that the shop will open at 9 a.m. knows that it will open at 9 a.m.'. This proposition, in the absence of reasons to suspect that that person lied to me could surely contribute to explaining why she told me that the shop will open at 9 a.m. As the proposition 'the person who told me that the shop will open at 9 a.m. knows that it will open at 9 a.m.' entails the proposition 'the shop will open at 9 a.m.', EE delivers the result that the testimonial evidence I possess in this case supports believing that the shop will open at 9 a.m. 9 But the reason is not, in this case, that my evidence supports believing a universally quantified proposition that entails a proposition about the future. Hence, it does not follow from EE that our justification for believing non-probabilistic claims about the future always depends on the justification we have for believing universally quantified claims.
The articulation of the near neighborhood strategy offered in the present paper and the defense of the result delivered by EE (given this strategy) in cases involving beliefs about the future thus does not lead to a generalized skepticism regarding the justification we have for holding such beliefs. In fact, EE provides an illuminating response to this skepticism. While it can be tempting to adopt a skeptical attitude concerning what we can know and rationally believe about the future because our evidence for holding such beliefs often seems to be merely statistical, EE shows to what extent some of the evidence we have for holding beliefs about the future is not merely statistical and actually supports believing non-probabilistic claims concerning future events. According to this account, we are epistemically entitled to hold beliefs concerning what will happen in the future-and not only about what will probably happen-when what we believe about the future follows logically from the best explanation we have for what already happened and what is currently happening.

Conclusion
In this paper, I showed why the near neighborhood strategy considered by McCain to handle cases of justified beliefs about the future succeeds, contrary to what Byerly and Martin claim. In addition, I argued that this strategy does not entail the view that knowledge requires probability 1 or lead to the skeptical conclusion that one should only believe probabilistic claims concerning the future. Instead, the appeal of this strategy becomes apparent when we pay attention to the fact that statistical evidence cannot, on its own, support believing non-probabilistic claims. I showed that McCain's initial account of the evidential support relation allows us to correctly distinguish situations in which one has only statistical evidence from situations in which one has evidence that is not merely statistical and, therefore, that the account allows us to see the extent to which our evidence for believing claims about the future is not always statistical. According to this account, if propositions about what will occur follow logically from what best explains our present and past evidence, we can be justified in believing that something will occur and not merely that there is a certain chance that it will occur. The ability of McCain's initial account of the evidential support relation to correctly track the distinction that can be made between statistical and non-statistical evidence as well as the plausibility of the results it delivers in cases involving beliefs about the future ultimately supports the conclusion that explanationism about epistemic justification, if properly articulated, has no problem accounting for the justification we have for holding beliefs about the future. In fact, explanationism is illuminating regarding the problem raised by the justificatory status of our beliefs about the future.