§1 Nature and Sources of Belief

In the Treatise Hume claims that belief ‘most accurately defin'd’ is ‘[a] lively idea related to or associated with a present impression’. (T 1.3.7.5, SBN 96) but his exploration of different kinds of belief, their origins and their interactions with other cognitive and perceptual phenomena reveal that definition to be quite inadequate. At T 1.3.7.7 Hume concedes that he is ‘at a loss’ to express clearly what he means by belief and it is only with ‘considerable difficulty’ that belief can be identified as the subject matter of philosophical reflection (T 1.3.7.7, SBN 628 (App.)). According to how Hume describes this difficulty, it is a problem with capturing belief in satisfactory terminology, not with belief as such. Whilst we may ‘express something near it’, Hume insists that the nature of belief, in the end, is identifiable through ‘recourse to every one's feelings’, the most accurate understanding of belief thus being achieved through an exercise of clarifying interpersonal phenomenology. Belief ‘feels different’ from other mental acts and, Hume continues, we can distinguish belief from an idea about, say, a fiction because of its distinctive phenomenological signature. Despite the professed difficulty in providing a satisfactory definition of belief it is, Hume thinks, clear enough what belief is in contrast to other states, something that ‘every one sufficiently understands in common life’ (T 1.3.7.7, SBN, 629).

At a general level Hume identifies three cardinal belief-forming processes: Reason, Association and Education, all of which are natural causes of belief.Footnote ⁸ In the context of demonstrating the relations of ideas, Hume explains that beliefs can be generated by reasoning in a way that embodies equal force and vivacity to those beliefs arising from the force of association. Association involves belief formation that relates an idea with occurrent content provided by perception or memory.Footnote ⁹ Principally, such associations are resemblance, contiguity and constant conjunction. Belief formed by education is, in contrast, non-rational in the sense of non-demonstrative and also non-associative in the present sense; through education a person can come to believe that p entirely on the basis of repeated impressions, not by associating an idea with another object or transferring the vivacity of a present impression into a relevant idea.Footnote ¹⁰ The extent to which we might want to call what is produced in this case a belief is unclear, but at least it could be a firm, steady, and vivacious commitment to things being thus and so.

Hume's appeal to constant conjunction expresses how the strength of belief in many cases will reflect the depth of the habituated imaginative capacity: limited experience of constant conjunction will give rise to weak belief, whereas an extensive diet of instances produces firmer, stronger belief. Thus belief arises ‘by degrees’ and acquires a ‘new force from each instance, that falls under our observation. The first instance has little or no force: The second makes some addition to it: The third becomes still more sensible; and ’tis by these slow steps, that our judgment arrives at a full assurance’ (T 1.3.12.2, SNB 130). Hume writes:

Here Hume is developing a contrast between the belief forming process of a person with long-standing experience of events and a person with limited such experience. The difference is not just with the quantity of experience, so to speak: with ‘advanc'd knowledge’ arising from extensive experience a person develops a higher-order observational capacity that operates at a level unavailable to a person with limited experience. The content of the relevant kind of ‘observation’ is partly shaped by a form of generalized understanding of ‘the connexion of causes and effects’, an understanding that augments the inference, as Hume says, arising from ‘imperfect’ experience of particular phenomena and produces belief in causal connection even if its ground is but ‘one single experiment’.

Of course there are different senses of observation here. For a person of limited experience, they would need to observe in some quite literal and direct way both conjuncts in the relation, and then ‘sufficient instances’ thereof (something I will come back to) in order to form what would be recognizable as a belief concerning their relation. Now that is a quite different sense of observation to the one that, according to Hume, has as its object the fact that objects of experience are, ceteris paribus, generally related thus and so. Hume describes this augmentation in terms of how background general knowledge can shape the way that we experience particular events. So I am suggesting that the difference is not just in the kind of reasoning that occurs independently of and as a response to experience; that would leave open the possibility that the experience is identical both to a person of ‘advanc'd knowledge’ and to those without such understanding. It is hard to take seriously the idea that the background knowledge to which Hume appeals in this context is relevant only in distinguishing cognitive activity subsequent to experience, as if the empirical input is identical in the two cases yet the output, the inference, is different depending on whether or not a person has this standing knowledge. It is, I think, much more plausible to supppose that the meaning of the relevant experience must be different in each case such that the input itself is changed by the possession and influence of background knowledge.

Hume describes a case of a person who, after ‘sufficient instances’, is able to form a robust belief in the occurrence of an anticipated event where this belief is grounded in a form of understanding (A 12, SBN 651). The point here is not just about how belief arises from the observation of one event, which is immediately associated with another, because of the ‘frequent conjunction of objects’. It also describes the case of a person who is capable of higher-order observation in the sense just referred to: understanding has a role in anticipating experience in the context of a familiar series of previously conjoined events, and also in cases of unfamiliar or partial (‘imperfect’) experience perhaps constituting a single case. Once the relevant form of understanding is in place it structures perceptual and cognitive possibilities such that one is able to form beliefs on the basis of ‘one experiment’. Such an experiment is isolated in one sense but, for a mature agent, it occurs against the background of a pre-established understanding and is thus interconnected with a person's wider experience that grounds belief formation.

In cases of belief arising from the operation of a higher-order observational capacity, a person needs background understanding that cooperates with a relevant experience. Causal belief, for example, is not the product of Association or Education since the relevant experiential input is slender, perhaps singular. Neither could it be the product of Reason, understood in one of Hume's senses, since the belief is not generated by demonstration. But rationality more broadly construed is at least partly responsible for the relevant belief since it arises in virtue of pre-formed understanding. So parsing the salient roles of reason and rationality is not straightforward, particularly because Hume uses reason in a number of contrasting ways, yet such a broader construal is consistent with Hume's use.Footnote ¹¹

At Treatise Book 1 Sec. V in which Hume is discussing the origin and composition of ideas he makes use of a distinction in an attempt to improve on the cognitive taxonomy established by Descartes, Locke and Berkeley.Footnote ¹² The distinction is there to provide a more developed account of how, on the one hand, our ideas are related and, on the other, how we relate our ideas. These are crucially distinct for Hume and reflect the operation of principles of association as distinct from those of comparison. Hume tells us that ideas can be related by force of habit: a natural relation between ideas is exhibited when one idea ‘naturally introduces’ another, our minds having been entrained to associate ideas in this way (T 1.1.5.1, SBN 13). A ‘philosophical relation’ (T 1.1.5.2, SBN,14) in contrast, is established because of the (relatively) free operation of the imagination. In such cases a comparison of ideas is effected for the purposes of reflection independently of any natural associations that may or may not have been established between mental contents in the course of ordinary habituation. In other words, we are able to consider and compare ideas in a way largely (although not completely) unconstrained by association. The content and significance of the distinction between causation as natural and philosophical is of course the subject matter of a huge amount of critical literature. For present purposes what matters is clarifying the epistemic standing of natural causation.

Hume explains that transitions in thought that occur ‘without any reason’ are driven by natural relations and related principles of association (T 1.3.12; SBN 92). Invoking principles of association to explain the occurrence of mental phenomena signals that the relevant process is non-rational. But from other ways in which Hume describes natural causation it seems appropriate to construe it in more rational terms. Hume writes that we are able to reason from causation in so far as it is a natural relation, as opposed to a philosophical relation: ‘’tis only so far as [causation] is a natural relation, and produces an union among our ideas, that we are able to reason upon it, or draw any inference from it’ (T 1.3.6.16, SBN 94). But, as Hume makes clear, so too are we able to reason and make inferences on the basis of ideas related philosophically in his particular sense. In this case we compare ideas, as opposed to the ideas being associated and inferences are grounded on the basis of such comparison. Specifying the connections between these two kinds of relation is controversial. For present purposes the most salient interpretation is one provided by Henry Allison who draws on the Sellarsian distinction between ways that we can make mental phenomena intelligible: either, on the one hand, via a non-normative casual-explanatory approach and, on the other, by placing the subject matter in a normative ‘space of reasons’ such that its constituents are interrelated in an irreducibly justificatory network.Footnote ¹³ Sellars originally drew this distinction in the context of characterizing knowledge (more precisely: ‘knowing’) but the implications, exploited by Allison, are wide-ranging and particularly relevant for understanding the difference between natural and philosophical relations.

As a natural relation causal inference lies outside the ‘space of reasons’ and thus is not subject to the normative constraints that govern judgments and other forms of thought within it. Construed as a philosophical relation, causal belief can be characterized as lying within the space of reasons since the activity of ‘comparison’ distinctive of philosophical relations is an achievement of reflection.Footnote ¹⁴ Allison suggests that, in light of Sellars's distinction, causal belief thus ‘lies partly within and partly without’ the space of reasons.Footnote ¹⁵ Although Hume regards reason as subordinate to custom, that subordination cannot be total, according to Allison, since that would imply that our beliefs come into existence and have a role in our cognitive lives in a way which is entirely external to the space of reasons.Footnote ¹⁶ But that cannot account for the role that Hume assigns to reason and the understanding in the context of many of our beliefs ranging from the moral domain to inductive inference. Passages in the Treatise and the second Enquiry are emphasized by Allison in order to argue for an important ‘affinity’ between the account of reason and the understanding and that of morality and the passions which is presented in the Treatise.Footnote ¹⁷ The context of Allison's argument is a discussion of scepticism about induction, a topic which is not my direct concern here. But the general form of Allison's approach is relevant since it recognizes how belief formation can be viewed as an occurrence or achievement in a rational context and in a way that is not antagonistically related to an explanatory approach. This is important for what follows later. I now turn to the way in which Hume's explanatory account has been characterized.

§2 Interpreting Hume's explanatory account of belief

According to a fairly standard and broad interpretation Hume provides a naturalistic account of belief formation, referring to how natural, non-epistemic psychological properties underpin entitlement to believe that p. Here I take Norman Kemp-Smith and Louis Loeb to share core features of a standard interpretation in this sense. The former attributed to Hume the view that a basic class of beliefs, our ‘natural beliefs’ as he termed them, are simply impressed into our minds by nature.Footnote ¹⁸ Included is the belief and associated inferential practice that events are causally related, the belief that bodies continue to exist when unperceived and that natural processes continue uniformly.Footnote ¹⁹ For Kemp-Smith such beliefs are ‘irresistible’ and it is ‘inevitable’ that we come to hold them. This natural inescapability of believing implies that it is not coherent to ask whether or not we ought to hold the belief.Footnote ²⁰ Despite being unable to offer any positive rationalizations we are nevertheless entitled to these beliefs. So Kemp-Smith characterized Hume as providing an explanation of the existence of belief by adverting to the natural and inexorable process of belief formation. This process supplies, albeit negatively, a form of justification.

The spirit of Kemp-Smith's interpretation remains influential although it has undergone considerable refinement in recent years. For example, Loeb has proposed that, like Kemp-Smith, there is a basic non-epistemic psychological property that drives the explanation of belief formation which also has epistemic implications.Footnote ²¹ According to Kemp-Smith the basic psychological property is ‘irresistibility’, in Loeb's revision it is ‘stability’: a natural-psychological corrective to a feeling of ‘uneasiness’ in the face of unstable beliefs which potentially threaten the doxastic environment quite generally.Footnote ²² Stability in the relevant sense is ‘a kind of staying power in an idea’ and a way to understand Hume's description of belief throughout the Treatise as fast, firm, settled, solid and steady.Footnote ²³ Stable beliefs have a reliable connection to the truth; the epistemic implication, now characterized more positively, is that we are entitled to relevant beliefs just in case they maximize stability. According to Loeb this generates a naturalized constraint on belief formation and Hume is interpreted as providing an externalist and naturalistic theory of justification. Thus, according to Loeb:

Hume tells us that belief, in contrast to other mental states such as ‘fictions of the fancy’, has ‘superior force, or vivacity, or solidity, or firmness, or steadiness.’ (T 1.3.13.9, SBN 148, T 1.3.7.7, SBN 629). Thus Hume's explanation of the existence of beliefs appeals to the property of ‘steadiness’ which, on Loeb's view at least, also grounds how an agent can be justified in holding such beliefs. Stability is a property of doxastic states that plays a dual role: a key role in Hume's theory of belief formation and also in Hume's analysis of the epistemic status of belief.Footnote ²⁵

Overall Kemp-Smith and Loeb claim to show how, by considering the role of psychological properties, a belief's existence can be naturalistically explained. The relevant belief is also characterized as holding a rational status, that of being warranted or justified. There are different conceptions here of precisely how belief is justified. But at least for the purpose of this paper, they share the common idea that avoiding scepticism involves showing that, as an object of naturalistic explanation, our doxastic states can be granted a rational status other than that of being incurably unjustified. So the justificatory status of belief depends upon the right kind of underpinning naturalistic explanation, one that appeals to essentially non-epistemic properties. An implication here is that the causal-explanatory dimension of Hume's account connects to the epistemological aspect only, as it were, by accident. That is to say, it seems quite possible to formulate the kind of explanatory story under consideration without any mention of justification or the legitimacy of belief at all: a purely descriptive account, then, of how it is that we come by our beliefs. Such was Quine's vision for what had passed as ‘epistemology’ before the 1960's – a vision directly inspired by a particular reading of Hume. I think that there is a different kind of explanation, vindicatory explanation, that is relevant for understanding Hume's account of belief, one that I think is indicated in the Treatise and one in which persons and their experiential history are irreducible components.

A vindicatory explanation shows why a person thinks that p by showing that there is nothing other than that p which this person could have come to think. It is an explanation of the existence of belief or an idea that shows it to be inevitable and unavoidable, thus on the face of it similar to a ‘natural belief’ in Kemp-Smith's sense. But the kind of inevitability distinctive of vindicatory explanation is not characterized by some subpersonal process that implants states or propositional attitudes into a mind by force of nature. The relevant sense of inevitability does not refer to some psychic motivational force that makes it impossible for an agent to resist holding the relevant belief. As I will explain below vindicatory explanation shows how the relevant belief is a function, partly, of the understanding and thus construes the characteristic inevitability as falling partly within the space of reasons, as playing an irreducibly normative role. Overall what this kind of explanatory approach provides is a way to reassess one aspect in the complicated relation between the explanatory and the justificatory elements in Hume's account of belief.

§3 Vindicatory explanations of belief, Experience and Understanding

The idea that explanations of belief can be in some cases vindicatory is not new although its potential in the present context has been under-exploited if not ignored.Footnote ²⁶ The general idea is that formulating the relevant kind of explanation of S's belief that p will thereby provide justificatory grounds for S believing that p: a belief's existence can be explained in such a way that vindicates a person who holds it. Common to all such explanations of the existence of belief is that S's belief that p comes about because p, and comes about in such a way that S is thereby furnished with grounds for believing that p, as Wiggins put it.Footnote ²⁷ So, to the question ‘Why does S believe that p?’ a vindicatory explanation would show that S believes that p, precisely because p. The ‘because’ here is understood in the following way, as described by Wiggins:

A vindicatory explanation implies that S has (in a sense to be elucidated) “no choice” but to believe that p.Footnote ²⁹ Consider the following examples both of which are responses to the question ‘Why does S believe that p?’:

In the present context distinguishing the fact that p, on the one hand, from there being no room to deny that p on the other, is not a distinction between two explanans: one that invokes a worldly state of affairs independent of S and one a psychological state such that S is inescapably driven to possess a particular belief. When a vindicative explanation of S's belief that p invokes the fact that p, the referent is not just some independent fact or property which furnishes S with grounds for the belief that p via some form of non-normative causal relation. That, for Wiggins at least, is not the right species of causation. Vindicatory explanations, for instance, are a species of causal explanation but not in the sense of characterizing our beliefs as arising in virtue of some brute contact with an independent reality.Footnote ³¹

In the mathematical case the causal explanation of belief does not just invoke the fact that 7 + 5 = 12 as explanans. What informs the explanation includes the proper understanding and application of calculating rules, a subject's cognitive capacities, her opportunities and access to relevant information for example.Footnote ³² A vindicatory explanation involves the process of elucidating what leads to the state of mind such that there is nothing else to think but that p. The kind of explanation being proposed here is also a form of justification in so far as the person is shown, through the explanation, to be vindicated in holding the belief.

According to Hume reason (understood narrowly in the sense outlined in §1) cannot be the source of an everyday belief that a moving ball will cause a stationary ball to move when they collide. Rather such a belief can only arise on the basis of experience, although ‘experience’ is used by Hume in different ways. Experience sometimes means a discrete occurent episode of, say, looking at a ball. In such cases Hume is interested in the nature of the relevant perceptual content and its implications as the source of ideas and beliefs. But very often Hume employs ‘experience’ to refer to a diachronic process that reveals or shows us something about the origin of ideas and beliefs as they emerge in the course of ‘common life’: their part in our developing lives as human beings.Footnote ³³ Experience does not just refer to the alleged sensory origin of ideas. Hume refers to how experience teaches us about those belief-forming processes that constitute part of our epistemic agency.Footnote ³⁴

Part of what is included under Hume's key idea that a belief arises on the basis of experience is how this basis cultivates a felt expectation that a stationary ball, for instance, will be caused to move when struck by another ball. Hume explains in the first Enquiry that ‘no man, having seen only one body move after being impelled by another, could infer, that every other body will move after a like impulse’ (EHU 5.1, SBN 43). This is because the relevant inference is an achievement requiring a diet of relevant experience, ‘a sufficient number of instances’ (T 1.3.13.19, SBN 154; A 12, SBN 651). The process contrasts with the capacity of reason which is ‘invariant’: the number of instances over which reason is employed is irrelevant to what we can discover through its application. For example, the conclusions which we might reach through the application of reason in this narrow sense when considering the geometrical properties of one circle, are the same which we would reach ‘upon surveying all the circles in the universe’ (EHU 5.1, SBN 43). What distinguishes empirical inferences and beliefs regarding ‘matters of fact’ is that they do vary depending on the number of instances that constitute the developmental process underpinning the inferential capacity. There is a difference between an inference that we draw from experience of one instance and what we are able to infer from considering numerous instances despite the fact that, taken in a state of artificial isolation, each instance as such may be indistinguishable (EHU 5.1, SBN 43). The ‘higher order observational capacity’ referred to in §2 describes this difference. Hume writes that, after a sufficient number of instances, a person who then ‘saw the one ball moving towards the other … would always conclude without hesitation, that the second would acquire motion. His understanding would anticipate his sight, and form a conclusion suitable to his past experience’ (A 12, SBN 651). As Hume explains the anticipative phenomenology of what is formed by habit incorporates not just that an event is expected without hesitation, but that the expectation is, itself, belief:

Whether or not Hume's analysis of causal inference is defensible is, of course, the subject of an enormous amount of interpretation and commentary. I will not try to rehearse Hume's account or attempt any assessment of whether Hume's overall view here is the right one, or the right one for Hume to hold on his own terms. What I want to highlight is the difference that distinguishes the habituated experience from naïve experience for the purposes of clarifying what explanatory possibilities that difference implicates. We are now in a position to appreciate more fully the force of Hume's remarks about the background role of ‘advanc'd knowledge’ discussed in §1. For some beliefs at least, the explanation of their coming into existence cannot be adduced other than by invoking a constitutive background of person-level development or formation. A belief of that kind cannot be fully explained in a way that omits the previous experience of an agent where that does not just refer to the passive reception of serial sensory input. This past experience is the formative background without which some of our core beliefs would be not only inexplicable, but nonexistent. Without the anticipative structure of belief cultivated through experience of sufficient instances, we would not have Humean beliefs at all and so the explanatory ambition would become incoherent.

Suppose S is watching a marble rolling toward another stationary one on the table. If Hume is right S cannot help believing that the stationary marble will move because of the background experiential context in light of which S now perceives the first one moving. This context, one partially shaped by understanding, provides S with a different experience and way of thinking than the one S would have had if perception of the present rolling marble was not augmented by previous experience. It is true enough that, with effort, S can speculate about the occurrence of events other than that of the stationary marble moving when the rolling marble collides with it. But in the condition that S is now in as a developed agent, it is no longer possible for S to believe anything other than that the marble will move.

The phrase ‘it is no longer possible’ needs some qualification. Hume does allow that we can conceive that a stationary ball will not move, a conception that picks out a possible state of affairs albeit in a merely ‘metaphysical sense’ as he puts it (A 11, SBN 650–651). Hume insists that we can ‘conceive any effect to follow from any cause, and indeed any event to follow upon another’ (A 11, SBN 650–651). But we cannot believe any effect to follow from any cause, we cannot believe that the stationary ball will not move when struck. Causal beliefs of this sort involve an immediate felt expectation sewn into occurrent perception once habit and custom are operating; on seeing the ball move toward to the stationary ball we immediately come to believe that the latter will move. The fact that we can conceive of it not moving does nothing to alter the inevitability of our believing that it will move. On reflection, or ‘abstractly considered, and independent of experience’ (A 15, SBN 652), we might entertain the possibility that the ball will remain still. But what we abstractly consider cannot undermine what we in fact believe. We are free, through the use of the fancy, to conceive of whatever we like (other than what is contradictory) but we are not free to believe whatever we like.Footnote ³⁵ Belief is to be understood not as some affirmative attitude toward a proposition but as an achievement of persons in the midst of practical and theoretical lives shaped and guided by experience in the broad Humean sense. So this account of belief is quite different from standard contemporary accounts (many of which cite Hume as inspiration). For instance, consider a familiar characterization of belief:

This coincides with Hume's view of belief only marginally. The idea that S exhibits an attitude of acceptance to the proposition: ‘The stationary ball will move when struck by the moving ball’ almost entirely fails to capture what Hume means by belief. Hume emphasizes the feeling of believing that distinguishes it from ‘mere reverie of the imagination’; although ‘’tis impossible by words to describe this feeling’, belief is distinguishable from ‘mere conception’, Hume thinks, because of its distinctive phenomenology: belief is stronger, more lively, firmer, more vivid and more intense than mere conception (A 21, SBN 653–654).

Consider again Wiggins's ‘nothing else to think’ formula to explain belief now applied to the billiard ball example:

We should allow that S can conceive, in Hume's technical sense, of something other than that the stationary ball will move in the way anticipated. As he puts it in the Enquiry ‘may I not conceive, that a hundred different events might as well follow…? May not both these balls remain at absolute rest? May not the first ball return in a straight line, or leap off from the second in any line or direction?’ (EHU 4.1 SNB 29–30). But, if we follow Hume, we cannot allow that there is something for S to believe other than that the ball will move in the way anticipated. Since both conception and belief are species of thinking, a ‘nothing else to think’ formulation is too broad in scope; when considering an episode or state of affairs there are (almost always) a number of things that one can think or conceive. So we ought to restrict the formulation to: ‘There is nothing else to believe’. And put like this the formulation amounts to saying that S's belief is inevitable thus, on the face of it at least, characterizing it as a ‘natural belief’ in the sense employed by Kemp Smith. And so too are there implications for how we might think about the justificatory status of such a belief. But the implications are quite unlike those identified by Kemp Smith or by more contemporary versions of the idea that what underpins the normative status of belief is the role of independent non-epistemic psychological properties.

As explained previously, according to Hume part of what distinguishes conception from belief involves a person's exposure to ‘sufficient instances’ and that the anticipative structure of belief is grounded in a form of understanding, a rational capacity different from ‘imagination’ and ‘the fancy’. A plain contrast between these is made at Treatise 1.3.6.4, (SBN 88–89) where Hume is considering whether the idea of causation arises because of the understanding or imagination; or, as he puts it, whether the idea is ‘determin'd by reason’ or whether it arises because of ‘a certain association and relation of perceptions’ produced by the imagination. A full account of the nature of this understanding is beyond the scope of this paper but its presence as part of what explains belief blocks any proposed reduction to the role of sub-personal psychic motivational forces. It is arguable that belief could be an intensified or stabilized conception, one that is like a conception only with additional force and vivacity. As such belief could potentially share its source in whatever it is that explains conception only with a supplementary feeling of compulsion. But this is not consistent with what Hume says about the role of the understanding as the ground of belief. And since Hume at least on some occasions explicitly aligns understanding with reason, in contrast to the imagination and acts of association (e.g., T 1.3.6.4; SBN 88–89), it seems right to say that belief is, at least in part, a form of rational achievement. If this is right, the notion of ‘rational’ will not be one that refers narrowly to ratiocination or cogitation as it does in, say, the section ‘Scepticism with regard to Reason’ (T 1.4.1.1; SBN 180). Like ‘experience’, and as indicated above, Hume uses ‘reason’ in different ways and for different purposes, sometimes to refer to any transition in thought, thus making use of a wide notion of ‘reason’ and ‘reasoning’ that enables Hume to speak freely of the ‘reason of animals’ (T 1.3.16.1; SBN 176) and how reason is articulated ‘sensitively’ as well as ‘cogitatively’.Footnote ³⁷ In claiming that belief is in part a rational achievement I only want to emphasize the normative grounds of belief and to highlight how the recognition of such normativity is unavailable to a purely mechanistic or what is usually understood as a naturalistic conception of belief formation. The role of the understanding in this context suggests that the inevitability of belief formation incorporates reason, another way of parsing what it is for our thoughts to be ‘determin'd by reason’ and arguably an instance of how we ‘employ our reason only because we cannot help it’ (A 27, SBN 657).

I have been suggesting a way that the inevitability of believing that p can be characterized such that the relevant grounds are, in a suitable sense, rational. If this is along the right lines then our interpretive options are not exhausted by a disjunction between either an explanatory or an epistemological approach to belief. I have proposed that the scheme of vindicatory explanation can help avoid the disjunction but it needs to be distinguished from a related form of explanation, one that shows a person to be unjustified in holding a relevant belief. I consider this in the next section.

§4 Explanation and Epistemic Valence

Explaining a belief can not only be consistent with it being false but can demonstrate its falsehood. Such an explanation would not then vindicate but would rather remove S's epistemic entitlement – a destabilizing explanation as Kail puts it.Footnote ³⁸ The form of this kind of explanation is familiar from discussions elsewhere. For example, Harman famously attempted to deflate realism about moral properties via a destabilizing explanation.Footnote ³⁹ Suppose that the wrongness of an observed action is part of the explanation of the belief that the action is wrong. Such a belief, Harman says, can be explained in another way, one which does not invoke a controversial normative property (wrongness) as part of the explanans. The alternative explanation refers only to non-moral properties and a set of enculturated response dispositions that a subject learns during the process of ordinary upbringing. This explanation would destabilize the initial characterization of the origin of belief which, ex hypothesi, is that the causal history of the belief at some point mentions the presence of the relevant moral property. The destabilizing explanation shows the property to be explanatorily redundant and that the relevant belief is produced by mechanisms quite different from a sensitivity to such a property as supposed truth-maker.

Consider another case. Suppose that early humans, ignorant of the truth, formed beliefs about some kind of supernatural agency as explanation of natural events. Later, once cognizant of the truth, humans came to understand that the previous belief forming processes were not constitutively sensitive to the truth. Rather, what explained the belief's existence was, say, the pervasive anxiety to which the belief in supernatural agency was a reaction.Footnote ⁴⁰ Whatever it is that fixes or guides the belief in this case it is non-epistemic since the aim of belief here is not truth but feelings of (say) reassurance. The causal history of belief reveals that the relevant origin is truth-independent and so non-epistemic, thus at least jeopardizing (destabilizing) the entitlement that S would have for believing that p. So an immediate difficulty is now this: on what grounds do we say of one kind of explanation that it vindicates and of another that it destabilizes?

Some think that according to Hume's account our ‘natural state’ is for our beliefs to be guided by an intrinsic truth-conduciveness, to be guided, that is, epistemically. Garrett, for example, claims that this natural state obtains ‘because thinking one's beliefs to be true [or probable] is naturally more comfortable than thinking them to be false’.Footnote ⁴¹ There is arguably a difference between the discomfort experienced as part of anxiety and the discomfort one might feel in thinking one's belief to be false. But it is not clear why one psychological state considered as such is evidence of truth-independence whilst the other preserves a connection to truth. Since both explanations invoke psychological facts, why is one taken to reveal the origin of belief as non-epistemic, and the other taken to exemplify our natural state of being guided epistemically? ‘[W]hy aren't the superstitious … just psychologically different, rather than epistemically deficient?’, as Williams puts it.Footnote ⁴² Presumably, that S, as a matter of fact, finds it more comfortable to regard her beliefs as true rather than as false does not thereby make it the case that her beliefs are guided epistemically as opposed to not.

Our question is about the rationale for claiming that in one case a psychological state is part of what severs the link with truth and yet, in the other, that it exemplifies being guided epistemically. In the latter case we would say of the belief that it has a sound epistemic origin and, therefore, that it is at least potentially a justified belief. But being in a position to say this requires the right perspective. A person's consultation of the internally accessible qualities of their own psychological states may not (perhaps cannot) reveal that some of them have a distinctively truth-seeking character. Accordingly this person would not be in a position to say which of their beliefs if any has a sound epistemic origin. Denying that there is anything intrinsic to psychological states that distinguishes truth-guided belief from other motives leads naturally to the idea that an external relation between states of mind and the world is what fixes a belief's epistemic status. And an externalism about justification is one prominent way to portray the non-sceptical aspects of Hume's empiricism although others have emphasized the importance of internal aspects too.Footnote ⁴³

Externalism is usually characterized as a widening of what can count as sufficient for justification; being justified need not be a product of discovering internally accessible warrant but, at least in some cases, by having one's beliefs produced in a reliably truth-guided way. Externalism is not a rejection of internalism as such since if a person is aware that the history of a relevant belief includes the fact that it is based on epistemically dubious grounds, then that is an internally accessible reason for this person to at least withhold endorsement of if not reject that belief.Footnote ⁴⁴ Likewise it is perfectly consistent with a weak form of externalism that a person can locate and endorse internally accessible content that justifies their belief. Nevertheless, externalism allows for S to be justified in believing that p in a way that does not require S to be aware of and to endorse the justificans. On the face of it the relevant externalism must be of a fairly strong form excluding the role of content given internally, content that in the best case simply fails to provide justification and, in the worst, is immediately destabilizing. But the connection between the apparent failure of internalism and the role of reason needs careful consideration. In particular we should avoid the assumption that embracing externalism requires privileging non-rational, non-epistemic elements in belief formation, perhaps psychological properties of the sort identified by Kemp Smith and Loeb.

The way I have portrayed vindicatory explanation suggests that the failure of reason to illuminate the grounds of belief from the inside, as it were, does not imply that scepticism can only be avoided by exploiting the non-rational, so-called natural elements of belief formation. According to my proposal sufficient instances of experience provide a form of ‘cultivated rationality’, a term used in a different but related context by McDowell.Footnote ⁴⁵ Exercising cultivated rationality can result in a person being justified in a way that does not require explicit awareness and endorsement of the justificans yet is not the product of purely non-rational processes. Recall (§1) the emphasis that Hume puts on a form of generalized understanding that informs the inferences available to a person of ‘advanc'd knowledge’. Characterizing the belief that a stationary ball will move as an achievement of cultivated rationality is a way of acknowledging the role of understanding in this context. Warrant for a belief need not issue, then, either solely from the reflective intellect or from mindless mechanisms connecting an agent to the world. A vindicatory explanation of belief lays out the context of belief formation in such a way that gives the inevitability of belief a rational articulation in the sense elucidated above.