2. Conee and Feldman’s mainstream evidentialism

In a series of publications, Conee and Feldman (Reference Conee and Feldman2004, Reference Conee, Feldman and Smith2008, Reference Conee, Feldman and Dougherty2011) have defended a very influential evidentialist theory of epistemic justification. Although these authors have formulated their view by distilling several principles, the one that is most fit to convey their position is possibly this:

EJDoxastic attitude D towards proposition p is epistemically justified for S at t if and only if having D towards p fits the evidence S has at t.

(Reference Conee and Feldman2004, 83)

As D can stay for a number of different propositional attitudes, EJ is a very general principle. Not only does it say that believing a proposition is propositionally justified for a subject S if believing the proposition fits S’s evidence; it also says that whether or not any other propositional attitude – agnosticism, disbelief – towards a proposition is epistemically appropriate for S depends in the same way only on whether having that attitude fits the evidence available to S. In this paper, I shall primarily focus on the most central instance of EJ saying that the belief that P, in particular, is propositionally justified for S if and only if believing P fits S’s evidence.

As seen in the introductory section of this paper, a principle like EJ elicits two immediate questions that must be answered before it can give us a complete and testable theory of epistemic justification. The first question asks what entities constitute S’s evidence. The second question asks when believing P fits S’s evidence, or, in the different terminology that shall be adopted throughout this paper, when S’s evidence supports P.

In response to the first question, Conee and Feldman endorse a psychologistic ontology of evidence, which says that a subject S’s evidence consists of non-factive mental states or events; mental states by instantiating which, in other words, one can misrepresent the world.Footnote ¹ Here is a sample of the mental states that, according to Conee and Feldman, are included in one’s evidence: S’s beliefs (Reference Conee and Feldman2004, 84; Reference Conee, Feldman and Smith2008, 87), S’s visual experiences (Reference Conee and Feldman2004, 84; Reference Conee, Feldman and Smith2008, 91), S’s perceptual experiences and S’s memory experiences (Reference Conee and Feldman2004, 112; Reference Conee, Feldman and Smith2008, 96), S’s gustatory experiences (Reference Conee and Feldman2004, 84), S’s feeling of warmth (Reference Conee and Feldman2004, 2 and 84; Reference Conee, Feldman and Smith2008, 87), S’s feeling of confidence (Reference Conee and Feldman2004, 76 and 112), S’s sense of frustration (Reference Conee, Feldman and Smith2008, 87).Footnote ²

In response to the second question asking when a given mental state M of S supports a given proposition P – in Conee and Feldman’s wording, when believing P fits M – here is what they have said in a recent and already much quoted passage:

In the above quote, Conee and Feldman seem to defend the theory of epistemic support according to which, roughly, whether a subject’s evidence supports a proposition depends on how well the truth of this proposition explains the existence of the evidence. More specifically, as Byerly (Reference Byerly2013) has argued, they can reasonably be taken to defend the following necessary and sufficient condition:

(EXP)e is the evidence for S that P is true if and only if P is part of the best explanation available to S for e.Footnote ³

Since, as we have already seen, Conee and Feldman contend that S’s evidence is constituted by mental states of S, (EXP) can more perspicuously be rewritten in the following way:

(EXP1)Mental state M is the evidence for S that P is true if and only if the truth of P is part of the best explanation available to S of why S has M.

3. The mainstream evidentialist’s theoretical commitments

In the last section, I have described the main tenets of Conee and Feldman’s ME. In the penultimate section, I will say something more about what it is possibly meant by saying that P is ‘available to S as part of the best explanation of why they have M’. For the time being my principal aim is to make explicit the commitments that descend from ME as this view as been described so far. Consider the following intuitive principle linking propositional justification and doxastic justification:

(J _PD)If S’s belief that P based on S’s mental state M is prima facie doxastically justified, M gives S prima facie propositional justification for believing P.

(J _PD) just gives voice to the hardly deniable suggestion that the basis of a doxastically justified belief in a proposition can only be a mental state that gives to the subject propositional justification for believing that proposition. To exemplify: if Marguerite’s belief that she won’t find meat in the menu, inferred from the belief that she’s sitting in a vegetarian restaurant, is doxastically justified, then Marguerite’s second belief must give her some degree of propositional justification for entertaining the first belief. If the belief that she won’t find meat in the menu were based on a different belief that does not give her any degree of propositional justification – like for instance the belief that the waitress is wearing a pink uniform – the first belief would not be doxastically justified to begin with. The mainstream evidentialist is hence committed to accepting (J _PD). As seen in Section 2, however, the mainstream evidentialist is a psychologist about evidence who is committed by EJ to denying that S’s propositional justification for believing a proposition could originate otherwise than from evidence possessed by S. On ME the claim that mental state M gives S prima facie propositional justification for believing P, hence, naturally translates into the claim that mental state M is evidence for S that P is true. The mainstream evidentialist is thus committed to accepting the following evidentialist rewriting of (J _PD):

(J _DE)If S’s belief that P based on S’s mental state M is prima facie doxastically justified, M is evidence for S that P is true.Footnote ⁴

(J _DE) can be used to test the adequacy of ME. For suppose it is intuitive, for some proposition P and mental state M of S, that S is doxastically justified in believing P on the basis of M. Given (J _DE), the mainstream evidentialist can accommodate the relevant intuition only if on the explanationist theory of the support relation they advocate M is evidence for S that P is true. Conversely, if on the explanationist theory of the support relation they advocate M is not evidence for S that P is true, ME is not able to accommodate the relevant intuition.

4. Mainstream evidentialism at work

ME naturally accommodates intuitive cases of perceptual justification. Suppose Sam looks into the garden and has a visual experience as of a bird in the snow. Call EX_bird S’s visual experience having the proposition (BIRD) that there is a bird in the snow as its content. Suppose now that Sam forms the belief that there is a bird in the snow on the basis of EX_bird. It is strongly intuitive that:

(#)S’s belief that there is a bird in the snow based on EX_bird is prima facie doxastically justified.

Given (J _DE), the mainstream evidentialist can accommodate the intuition voiced by (#) only if on (EXP1) EX_bird is evidence that there is a bird in the snow. And, it is easy to see that the mainstream evidentialist can fulfil this commitment. If S has just ordinary background information concerning the normal functioning of her sense organs, the truth of BIRD – namely, the truth of the proposition that there is a bird in the snow – is part of the best explanation available to S of why S has EX_bird – of why she has a visual experience as of a bird in the snow. Hence, on (EXP1) EX_bird is evidence that BIRD is true.

So far, so good. Problems for ME begin to emerge as soon as we turn to cases in which it is intuitive that S is prima facie doxastically justified in believing a proposition P not on the basis of a perceptual experience as of P but on the basis of further justified beliefs of S. To see why, we just have to focus on the following instance of EXP1, which specifically pertains to the conditions on which a belief of S is evidence for S that a given proposition is true:

EXP1_B S’s belief B is evidence for S that P is true if and only if P is part of the best explanation available to S of why S has B.

Bearing in mind what it takes for a belief of S to be evidence for S that a give proposition is true, consider the following case.

Weather forecasts: Tim just heard the weather forecast saying that it will rain in Almeria the day to come. Not having special reasons to distrust the weather forecast, he forms the justified belief that it is going to rain in Almeria the day to come. Tim knows Almeria very well. In particular, he knows that owing to the extraordinary dry climate just a few people possess an umbrella there. So he draws the conclusion that the day to come Almeria’s local umbrella retailers will boost their sales.

The intuitive claim that ME must be able to accommodate, in this case, is this:

(*)Tim’s belief that Almeria’s local umbrella retailers will boost their sales based on Tim’s current justified belief that it will rain in Almeria the day to come is prima facie doxastically justified.

And again, given (J _DE) the mainstream evidentialist can accommodate the intuition voiced by (*) only if Tim’s (justified) belief that (RAIN) it will rain in Almeria the day to come is on (EXP1_B) evidence that (BOOST) Almeria’s local umbrella retailers will boost their sales. On (EXP1_B), however, the belief that it will rain in Almeria the day to come is not evidence that BOOST is true. For the truth of BOOST is not even part of the best explanation available to Tim of why he believes RAIN. If anything explains this episode of belief formation, this is the fact that Tim heard the weather forecasts reporting rain in Almeria the day to come, plus, perhaps, the fact that Tim has no special reason to distrust them. So, ME seems unable to accommodate the intuition voiced by (*). This would be a major shortcoming; for a sound theory of epistemic justification aspiring at generality should be able to account for the intuitive fact that Tim’s belief that Almeria’s local umbrella retailers will boost their sales is doxastically justified on the basis of Tim’s belief that it will rain in Almeria the day to come. So unless there is a way in which ME can try to evade this consequence, ME must be rejected as lacking the needed generality. The following section is devoted to explore a possible way out of the above objection based on recent proposals by McCain (Reference McCain2014a, Reference McCain2014b and Reference Mccain2015).

5. McCain to the rescue of mainstream evidentialism?

The claim that evidential relations are explanatory in nature has been challenged by Lehrer (Reference Lehrer1974), by Goldman (Reference Goldman and Dougherty2011) and, more recently, by Ryan Byerly (Reference Byerly2013) and by Byerly and Martin (Reference Byerly and Martin2015). The first two epistemologists, in particular, have contended that there are cases in which a given proposition Q is not part of the best explanation available to one of why a second proposition P is true, yet P epistemically supports Q because P entails Q. Here, in particular, is Lehrer’s case.Footnote ⁵

According to Lehrer, the proposition that the mouse is three feet from the flagpole and the proposition that the flagpole is four–foot-high support – because modulo Pythagoras’ theorem they entail – the proposition that the mouse is five feet from an owl that is sitting on the top of the flagpole. However, the proposition that the mouse is five feet from the owl does not explain why the mouse is three feet from the flagpole and why the flagpole is four feet high. Whence the conclusion, drawn by Lehrer, that epistemic support is not always explanatory.

Byerly, on the other hand, has argued against the explanationist theory of the support relation by presenting a case in which a subject S’s experiences clearly support a proposition whose truth is not available to S as part of the best explanation of why S has those experiences. Here is the relevant quote:

Beyerly contends that his experiences support the conclusion that the ball will go in. However, he contends, that the ball will go in does not explain why he has the experiences he does. Whence, again, the conclusion that epistemic support is not always explanatory.

According to McCain (Reference McCain2014a, 63), these counterexamples do show that (EXP1) and (EXP1_B) are too restrictive, as they just appeal to best explanations. The correct explanationist principle, according to McCain, must also appeal to logical consequences of best explanations, in the following way:

(EXP2)Mental state M is the evidence for S that P is true if and only if the truth of P is part of the best explanation available to S for why S has M or P is available to S as a logical consequence of the best explanation available to S for why S has M.

(EXP2) may cause some eyebrow raises. For (EXP2) would seem to presuppose the claim – which most epistemologists would regard with scepticism – that evidential relations are closed under known entailment. However, I don’t want to press this point any further. Rather, I want to address the different question of whether (EXP2) can be used to shield ME from the charge of not being able to accommodate (among other things) the intuition, voiced by (*), that Tim’s belief in BOOST based on Tim’s justified belief in RAIN is prima facie doxastically justified. To see a possible way in which one could try to articulate this defence of ME, it is instructive to consider the way in which, according to McCain, (EXP2) can be used to handle, in particular, Lehrer’s and Byerly’s counterexamples.

McCain grants the point that in Lehrer’s case one is completely justified in believing that the mouse is five feet from the top of the flagpole even if one’s justification does not depend on explanatory relations. The claim that one is justified in believing that the mouse is five feet from the top of the flagpole – McCain correctly observes – actually presupposes that in the situation described by Lehrer one has ‘evidence that justifies [one] in believing that the mouse is three feet from the flagpole, the owl is on top of the four-foot tall flagpole, and the Pythagorean Theorem is true’ (Reference McCain2014a, 74). This means that the proposition that the mouse is three feet from the flagpole, the proposition that the flagpole is four-foot tall and Pythagoras’ theorem are all part of the best explanation available to one of the existence of one’s evidence. But then, since the proposition that the mouse is five feet from the top of the flagpole is a logical consequence of these propositions, by the lights of (EXP2) one has evidence that the mouse is five feet from the top of the flagpole. For the latter proposition is available as a logical consequence of the best explanation of the evidence available to one. According to McCain (Reference McCain2014a and Reference McCain2014b) Byerly’s counterexample can be handled in similar ways. One first possibility, considered by McCain, is to suggest that the best explanation of Byerly’s evidence includes the proposition that ‘all golf balls rolling toward a cup in circumstances C have gone into the cup’, and the proposition that ‘the golf ball is rolling toward a cup in circumstances C’ (Reference McCain2014b, 143). Since these propositions entail that the golf ball will roll in the cup, on (EXP2) Byerly’s evidence supports the proposition that the ball will go in the cup, as this latter proposition is available to Byerly as a logical consequence of the best explanation available to him of his evidence.Footnote ⁶

Whether McCain’s strategy to handle Lehrer’s and, in particular, Byerly’s counterexamples is successful has turned out to be a matter of controversy (see Byerly and Martin Reference Byerly and Martin2015 and McCain Reference Mccain2015). Whether it is, however, is a question that for the limited purpose of this paper can be set aside. McCain’s strategy to handle Lehrer’s and Byerly’s counterexamples has been presented just for the purpose to introduce the following parallel strategy, based on (EXP2), that the mainstream evidentialist could invoke to vindicate the intuition voiced by (*). Here, it is (for ease of exposition, let us dub it ‘McCain’s reply’).

Suppose that in weather forecasts, Tim has evidence supporting RAIN – namely, the proposition that it will rain in Almeria the day to come. This, on (EXP2), means that RAIN is part of the best explanation available to Tim for the existence of Tim’s evidence; say, in particular, for why Tim instantiates a given mental state M*. As I shall argue in a moment, it is not entirely clear which mental state of Tim could be such a M*. For the time being, however, we can let this question hang in the air, and simply assume, for the purpose to see how far this assumption can take the advocate of ME, that Tim instantiates a mental state with the features ascribed to M*. As we have seen, Tim’s background information includes the propositions that just a few inhabitants in Almeria do have an umbrella, and the consequent proposition that if the day to come it’ll rain in Almeria, then local umbrella retailers will boost their sales. Against this background information, BOOST – the proposition that local umbrella retailers will boost their sales – is hence available to Tim as a logical consequence of a proposition – RAIN – that is part of the best explanation available to him of why he has M*. Thus, on (EXP2), M* is evidence for Tim that BOOST is true. If there is a mental state of Tim with the features ascribed to M* (EXP2), hence, provides the mainstream evidentialist with the resources needed to vindicate the claim that Tim has evidence for believing that BOOST is true. Is however this result enough to rescue the mainstream evidentialist from the allegation of not being able to accommodate the intuition voiced by (*)? I believe that it is not, and this for several independent reasons.

The first major worry is that, even if successful, McCain’s reply would entirely miss its target. Until now we have been assuming that Tim instantiates a given mental state M* and that part of the best explanation available to Tim of why he has M* is the truth of RAIN (a second worry I shall shortly explain is that this assumption is not at all innocent). When all this is granted, as we have just seen, (EXP2) allows one to say that M* is evidence for Tim that Almeria’s local umbrella retailers will boost their sales the day to come. Thus, on the assumption above the mainstream evidentialist who endorses (EXP2) can accommodate the intuitive claim that, in the situation described in weather forecasts, Tim has prima facie propositional justification for believing that Almeria’s local umbrella retailers will boost their sales the day to come. This, however, is not the intuitive claim that, along with the objection above, the mainstream evidentialist should be in position to accommodate in order to vindicate (*). The intuitive claim to which (*) gives voice – and which the mainstream evidentialist is allegedly not in position to accommodate – says that Tim is doxastically justified in believing BOOST. This, given (J _DE), can only be true if Tim’s belief in BOOST is based on evidence, namely only if Tim’s belief in RAIN – from which Tim has inferred the belief in BOOST – is evidence that BOOST is true. But the response above, inspired by McCain, falls short of showing that RAIN is evidence for Tim that BOOST is true. For on EXP2 it is M*, and not Tim’s belief in RAIN, that counts as evidence that BOOST is true. So, even if it incorporates McCain’s improved explanationist principle, ME remains incapable of accommodating the intuition that Tim is doxastically justified in believing BOOST on the basis of his justified belief in RAIN.

In response the mainstream evidentialist could protest that, when S is doxastically justified in believing P on the basis of some mental state M, ME need not necessarily be able to say that M is evidence for S that P is true in order to get the epistemic phenomena right. For instance, they might contend that when – as in the case above – M is a belief in a second proposition Q, and S’s belief that Q is in turn doxastically justified on the basis of a different mental state M*, all that ME has to be able to say, in order to save the phenomena, is that M* is evidence for S that Q is true – and so indirectly evidence that P is true –, not also that S’s belief that Q is evidence for S that P is true. The Mainstream Evidentialist could try to motivate this request to be (locally) exempted from the requirement issued by J _DE by suggesting that since S’s belief that Q is doxastically justified on the basis of M*, what gives S propositional justification for believing P is ultimately M*, not S’s belief that Q, the role of this latter belief just being that of relaying – to resort to Silins Reference Silins2005’s terminology – the justification it receives from M* to P.

One unwelcome consequence of this line of resistance is that it conflicts with a widely accepted thesis about the relation between doxastic and propositional justification. This thesis says that there are (at least) three conditions that have to be satisfied in order for a subject S to possess doxastic justification for a proposition P: (1) S must believe that P; (2) S must have propositional justification for P; and (3) S must believe that P on the basis of the reason in virtue of which S has propositional justification for P (Pollock and Cruz Reference Pollock and Cruz1999, 35–6; Alston Reference Alston1989, 108; Korcz Reference Korcz2000, 46). The above reply, by allowing Tim’s belief in BOOST based on TIM’s belief in RAIN to be doxastically justified while letting Tim’s propositional justification for believing BOOST depend on M* in fact infringes condition (3). For in the case at issue S does not believe P on the basis of the reason – namely M* – in virtue of which S has propositional justification for P.

The mainstream evidentialist, at this juncture, could perhaps try to evade this difficulty by distinguishing between the ultimate basis of S’s belief that P and the direct or proximate basis of S’s belief that P.Footnote ⁷ Let again, in order to illustrate this possible distinction, M be S’s belief that Q, and suppose M is justified on the basis of the different mental state M*; suppose further that S infers P from M. In a sense, in a similar situation the ultimate basis of S’s belief that P is not M, which is the mental state S’s belief that P is most directly based upon, but M*. After all, S believes P because she has inferred it from M, but M is just an intermediate step, because S would not have formed this belief, and so would not have inferred P, if she had not had M* to begin with. By invoking the distinction between an ultimate and a direct basis of a belief, the mainstream evidentialist could reformulate condition (3) thus:

(3*)S must believe that P on the ultimate basis of the reason in virtue of which S has propositional justification for P.

Condition (3*) is not infringed in the case above: the ultimate basis of Tim’s belief in BOOST is M*, and on EXP2 M* is evidence that BOOST is true. So Tim’s belief in BOOST is ultimately based on the reason in virtue of which he has propositional justification for BOOST.

One major difficulty of the proposal at issue is that it is not clear in which sense, if any, an ultimate basis of a belief, when it is distinct from its direct or proximate basis, can be regarded as an epistemic basis. So, it is not clear that adverting to the distinction between the ultimate basis of a belief and its direct basis can be of any assistance in reconciling J _DE – which requires that the epistemic basis of a justified belief be evidence that this belief is true – with examples like Tim’s, in which the direct basis of a subject’s belief is not evidence that the belief is true, and the evidence with which the subject is creditable is at best the ultimate basis of the belief in question. To conclusively assess the question about whether the ultimate basis of a belief, when distinct from its direct basis, could at all be characterized as an epistemic basis we’d need an adequate theory of the basing relation, but this, as everyone knows, is a desideratum that is far from having been achieved. However, I believe that a brief inspection of the current debate on this topic is bound to reveal that the prospects of a theory of the basing relation on which an ultimate basis – when distinct from a direct basis – non-problematically counts as an epistemic basis are indeed dim.

Begin to note that while most theorists agree that a reason R cannot be the epistemic basis of S’s belief that P unless R contributes causally to the formation of S’s belief that P (Harman Reference Harman1970; Goldman Reference Goldman and Pappas1979; Swain Reference Swain and Pappas1979; Alston Reference Alston1988; Moser Reference Moser1989; Huemer Reference Huemer2001), everybody agrees that a deviant causal connection between R and S’s belief that P isn’t sufficient to establish a basing relation between R and S’s belief that P. One important task for those who share the initial intuition – that to be a causal determinant is at least partly constitutive of what it takes to be the epistemic basis of a belief – is then to distinguish in principled terms between a deviant and a non-deviant causal connection. One promising strategy recently emerged within the literature is K McCain’s interventionist account of the basing relation. For this reason, and for the reason that this theory is an integral part of McCain Reference McCain2014a’s defence of ME, I think that McCain’s interventionist theory is a fair example of a causal theory of the basing relation against which to test whether an ultimate basis, when it does not coincide with a direct or proximate basis, is an epistemic basis.Footnote ⁸ Simplifying a little, McCain suggests that S’s belief that P is based on R if and only R is a direct and occurrent cause of this belief (Reference McCain2012, 364; also Reference McCain2014a, 89). On the interventionist account of causation that McCain takes from Woodward (Reference Woodward2003, 55), X is in particular a direct cause of Y with respect to a variable set V if and only if there is some intervention on X that changes Y, while all the other variables within V apart from X and Y are held fixed. McCain’s interventionist theory, as anticipated, helps to handle problematic cases of deviant causal connections. On such a theory, for example, when I see Sylvia and believe that I see her, and as a consequence of this I become rattled, inadvertently drop the cup of tea on my leg and then come to believe that my leg hurts, the latter belief doesn’t count as being based on the belief that I see Sylvia (see Plantinga Reference Plantinga1993, 69). Although such a belief is a causal ancestor of my belief that my leg hurts, it is no direct cause of this belief. Hence, on McCain’s theory it is not an epistemic basis of it. What is important, for present purposes, is that on McCain’s theory of the basing relation an ultimate basis of a belief, when it is distinct from its direct basis, is no epistemic basis of this belief. This is so because, in such a case, the ultimate basis of a belief is not a direct cause of this belief. To appreciate in particular the point that M* is not an epistemic basis of Tim’s belief in BOOST we just need to restrict the relevant variable set V to M* itself, Tim’s belief in RAIN and Tim’s Belief in BOOST. As is apparent, holding fixed Tim’s belief in RAIN and setting to zero M* does not affect a change in Tim’s belief in BOOST. Hence, M* is no direct cause of Tim’s belief in BOOST and is not, on McCain’s theory, an epistemic basis of it.Footnote ⁹

Of course, the causal account of the basing relation is not the only theory of the basing relation. Its main competitor, the doxastic theory of the basing relation, does not seem to offer to the Mainstream Evidentialist a particularly promising alternative though. Proponents of the doxastic theory maintain, roughly, that in order for a reason R to be the basis of S’s belief that P S need hold a meta-belief to the effect that R is a good reason to believe P. Naturally enough, that Tim believes that M* is a good reason to believe BOOST constitutes a possibility. So the doxastic theory is more hospitable to the suggestion that M* is an epistemic basis of Tim’s belief in BOOST. A first difficulty is however that Tim’s case can be re-described so as to make explicit that Tim does not hold the meta-belief that M* is a good reason to believe BOOST. If anything else in the example is held fixed, it remains exceedingly intuitive that Tim’s belief in BOOST is doxastically justified, but since Tim lacks the relevant meta-belief, M* does not count as an epistemic basis of Tim’s belief in BOOST on the doxastic theory. Hence, in this case, whether or not the doxastic theory is in itself viable, it proves unserviceable for the purpose to show that M* is an epistemic basis of Tim’s belief. Moreover, as it has largely been emphasized in the literature, the viability of the doxastic theory of the basing relation is highly debatable. One common line of objection against doxastic theories is that they under-generate in different ways. A first range of cases in which these theories give the wrong result it when it is intuitive that a subject S entertains a belief for a reason R but S lacks the conceptual repertoire necessary to hold the relevant meta-belief. Young children or unsophisticated adults, for instance, are arguably unable to believe that their perceptual states provide them with good reasons to believe that reality is how they represent it as being. However, it seems undeniable that young children and unsophisticated adults believe all sorts of things on the basis of their perceptual states. A second range of cases is when a subject S intuitively bases a belief on reasons of which she is unaware. One example described by Alston (Reference Alston1988) is when one justifiably believes that another person is upset on the basis of features of the way she looks that one is unable to discriminate. A different objection to doxastic theories of the basing relation is that they also threaten to over-generate. To illustrate this objection Korcz (Reference Korcz1997 and Reference Korcz and Zalta2015) has proposed an example in which a cult leader makes one of his followers believe that his belief in God is a good reason for believing whatever he believes. Korcz finds it incredible – and it seems independently absurd to believe – that every belief entertained by the follower should count as being based on his belief in God. Finally, McCain himself (Reference McCain2014a, 85) has suggested that doxastic theories of the basic relation are faced with the dilemma between requiring the relevant meta-belief to be justified – in which case an infinite regress is started – and allowing the relevant meta-belief not to be justified – in which case it is unclear how it could be relevant to the basing relation.

Let’s recap. The concession that Tim instantiates a mental state like M* has not proven, especially helpful to the mainstream evidentialist. A first difficulty with the suggestion that M* is what gives Tim propositional justification for BOOST is that Tim’s belief in BOOST is based not on M* but on Tim’s belief in RAIN. This would conflict with a widely accepted thesis about the relation between doxastic and propositional justification, as it would entail that Tim’s belief in BOOST is doxastically justified even if it is not based on the mental state that gives him propositional justification for BOOST. To distinguish between an ultimate and a direct basis of a belief, and to insist that M*, if not the direct basis, is at least the ultimate basis of Tim’s belief in BOOST has proven equally unhelpful. For the claim that the ultimate basis of a belief, when it does not coincide with its direct basis, can be taken to be a genuine epistemic basis has appeared debatable when assessed in the light of extant theories of the basing relation.

This, it must be stressed, is not the sole problem afflicting McCain’s reply. A second major problem, as widely anticipated, is that it is doubtful, to begin with, that Tim instantiates a mental state with the features ascribed to M*. To see this, consider the likeliest candidate for being such a M* – namely, Tim’s experience (EX_report) of hearing a journalist say that it will rain in Almeria the day to come. EX_report does not seem to fit the bill. Remember that M*, if it exists, must possess the following feature: the truth or RAIN must be available to Tim as part of the best explanation of why Tim instantiates it. What explains the existence of EX_report, however, is arguably not the truth of RAIN, but the truth of the different proposition (REPORT) that the journalist has said that it will rain in Almeria the day to come. So, on (EXP2) EX_report is evidence for Tim that REPORT is true, not evidence that RAIN is true. The mainstream evidentialist could perhaps try to circumvent this initial difficulty by observing that even if RAIN is not directly available to Tim as part of the best explanation of why he has EX_report, it is still available to Tim as part of the best explanation of why REPORT is true, which in turn is available to Tim as part of the best explanation of why he has EX_report. If the relation of being available to one as part of the best explanation of something is transitive, it would follow from this that RAIN is (perhaps indirectly) available to Tim as part of the best explanation of why he has EX_report, as required by (EXP2) for EX_report to be evidence for Tim that RAIN is true. Despite its initial plausibility, however, also this possible way out is fraught with problems. The main difficulty is that the relation of being available to one as part of the best explanation of something does not seem to be transitive. To see this, it is helpful to unpack a little more what it means that a given B is available to S as part of the best explanation of a given A. To prevent any charge of question-begging, here I resort to the useful analysis given by McCain (Reference McCain2014a). According to McCain, if B is available to S as part of the best explanation of A, then S has the disposition to have the seeming that B is part of the best answer to the question ‘why A?’ on the basis of reflection alone. Now consider that S can be disposed to have the seeming that B is part of the best answer to the question of why A is the case, and be disposed to have the seeming that C is part of the best answer to the question of why B is the case, while failing to be disposed to have the seeming that C is part of the best answer to the question of why A is the case. In a context in which McCain’s characterization is presupposed, this means that B may be available to S as part of the best explanation of why A is the case, C be available to S as part of the best explanation of why B is the case and C fail to be available to S as part of the best explanation of why A is the case. In other words: it means that the relation of being available to one as part of the best explanation of something is not transitive. To see how this could happen consider the following example. Suppose the question arises of why my neighbour, whom I know to be a reformed doctor, is washing his car on a Sunday morning. Witnessing the scene, and considering the question, I have no disposition to think that part of the explanation why (SUNDAY) my neighbour is washing his car on a Sunday morning is that (HELP) he likes helping out. This seems entirely compatible with my disposition to have the seeming that that my neighbour’s willingness to help out is part of the best answer to the question of why (HOSPITAL) my neighbour spends every day of the week but Sunday volunteering as a doctor at the local hospital; and this is entirely compatible with my disposition to have the seeming that the fact that my neighbour spends every day of the week but Sunday volunteering as a doctor at the local hospital is part of the best answer to the question of why he is washing his car on a Sunday morning. We have here a case in which I am disposed to have the seeming that a first proposition – HELP – is part of the best answer to the question of why a second proposition – HOSPITAL – is true, I am disposed to have the seeming that the second proposition – HOSPITAL – is part of the best answer to the question of why a third proposition – SUNDAY – is true, yet I have no corresponding disposition to have the seeming that the first proposition – HELP – is part of the best answer to the question of why the third proposition – SUNDAY – is true. Hence, we have a case in which the transitivity of the relevant relation is falsified.

A different way out of the second difficulty with McCain’s reply – to recap: the difficulty that on (EXP2) Ex_report is not evidence that RAIN is true – might advert to following explanationist principle stated by McCain (Reference Mccain2015).

(EXP3)Mental state M is the evidence for S that P is true if and only if the truth of P is part of the best explanation available to S for why S has M or P is available to S as an explanatory consequence of the best explanation available to S for why S has M.

A proposition P is available to one as an explanatory consequence of the best explanation available to one of why one has M when the best explanation available to one of why one has M would explain P significantly better than it would explain ~P (McCain Reference Mccain2015, 339). On (EXP3) it is hence sufficient for EX_report to count as evidence that RAIN is true that RAIN is available to S, if not as part of the best explanation of why she has EX_report, at least as an explanatory consequence of the truth of REPORT. So the relevant question turns out to be: that a journalist has said that it will rain in Almeria the day to come would explain the truth of RAIN significantly better that the truth of ~RAIN? Unfortunately, the truth of REPORT seems to be explanatory irrelevant with respect to the truth of RAIN (and, for that matter, with respect to the truth of ~RAIN). The question why it will tomorrow rain in Almeria is certainly potentially illuminated by the truth of many possible propositions reporting the values of the relevant meteorological parameters. But the proposition reporting the correct prediction of this meteorological phenomenon seems to contribute to no extent to its intelligibility.

In order to see a last difficulty with McCain’s reply, we just have to reformulate weather forecast by explicitly stipulating that RAIN is not available to Tim as part of the best explanation of why REPORT is true. In this slightly re-described scenario the intuition voiced by (*) does not seem to fade away: the detail I have added does not alter the intuitive fact that Tim is doxastically justified in believing BOOST on the basis of his justified belief in RAIN. However, this detail makes the strategy above unserviceable for the purpose to vindicate the claim that Tim’s belief in BOOST is doxastically justified. That strategy, it should be reminded, presupposes that EX_report is the ultimate epistemic basis of Tim’s belief that BOOST, and that if the truth of RAIN is available to Tim as part of the best explanation available to Tim of why REPORT is true then EX_report is evidence for Tim that RAIN (and so that BOOST) is true. In the pages that precede I have argued that none of this should be conceded. To begin with, I have argued that even if EX_report were evidence that BOOST is true, it wouldn’t constitute an epistemic basis for Tim’s belief in BOOST; moreover, I have argued that even if the truth of RAIN were available to Tim as part of the best explanation of why REPORT is true, EX_report wouldn’t count on EXP2 (or on EXP3) as evidence for Tim that RAIN, and so BOOST is true. If one is unconvinced, the scenario presently under consideration supplies an additional difficulty. For even if my criticisms are wrong, the strategy above to explain (*) still requires that the truth of RAIN be available to Tim as part of the best explanation of why REPORT is true. And the new scenario, although it verifies (*), has now been explicitly re-described as one in which RAIN is not available to Tim as part of the best explanation available to him of why REPORT is true.