Internalism and Externalism in Early Modern Epistemology

1. Internalism and Externalism in the History of Epistemology

It is often said that Descartes is the archetype of epistemic internalism and that he inspired the epistemic turn in early modern philosophy that emphasizes the need for first-person awareness of evidence. Descartes has an inside-out program that starts with the awareness of my own thoughts and moves outward from there. Others then followed Descartes in pursuing this inside-out strategy. This oft-repeated story is true so far as it goes, but it leaves out the extent to which Descartes, Locke, and Hume, among others, are externalist about epistemic justification. For each of these figures, I will argue that the reliability of the belief-forming process contributes to the justification of beliefs.

Contemporary epistemologists are divided between internalism and externalism about the nature of justification. Internalism is, roughly, the view that a belief that p is justified by a mental state (something “internal” to the mind), such as another belief or a sensation. An especially common form of internalism is evidentialism, the view that it is the awareness of evidence that justifies beliefs. For example, as Lawrence BonJour (1992: 364) explains, “the main intuitive motivation for internalism” is “the idea that epistemic justification requires that the believer actually have in his cognitive possession a reason for thinking that the belief is true”.¹ By contrast, externalism is, roughly, the view that a belief that p is justified by facts about the belief-forming process (facts that a person may not be aware of, and hence “external” to the mind). Paradigmatically, the relevant facts about the belief-forming process are about (broadly speaking) its reliability. For example, according to reliabilism (Goldman 1979), a belief is justified by being formed in a reliable way; according to proper functionalism (Plantinga 1993), a belief is justified by being formed by a reliable and properly functioning belief-forming process.² In general, then, the paradigmatic externalist theories claim that a belief is justified by facts about (broadly speaking) the reliability of the belief-forming process; for ease of discussion, I will refer to these relevant facts as “reliability.”

There are two “master arguments” for internalism (Madison 2010), and both arguments will be relevant to the discussion in this paper. First, BonJour’s well-known case of Norman the clairvoyant is intended to show that justification requires the awareness of evidence. Suppose Norman has a reliable power of clairvoyance and, on this basis, forms a true belief that the president is in New York City. However, imagine Norman has no reason to believe he has a reliable power of clairvoyance or that the president is in New York City. BonJour (1980: 62–63) persuasively argues that, though formed in a reliable way, Norman’s belief is unjustified; for, from Norman’s own point of view (i.e., one without any evidence for the belief), his belief appears to be a lucky guess. This example seems to show that justification requires evidence. Most interpreters, myself included, recognize that Descartes, Locke, and Hume accept internalism of this kind. However, there is a second argument for internalism that, I will argue, Descartes, Locke, and Hume do not accept.

A second well-known argument for internalism attempts to show that evidence, even in the absence of reliability, is sufficient for justification. This is the so-called new evil demon problem (see Cohen 1984). Descartes famously raises the possibility that my sensations are merely the deception of an evil demon. Descartes takes the possibility of a deceiving evil demon to show that I cannot be certain, and hence I would be unjustified in believing, that my sensation of the fire corresponds to an actual fire. Thus the “old” evil demon problem takes the possibility of error to undermine justification. By contrast, the “new” evil demon problem assumes that sensations ordinarily do justify belief in external objects and, since the sensation would appear the same to me in an evil demon world, the sensation would also justify my belief in the fire even if I were being deceived. This conclusion, if correct, implies that evidence can justify beliefs even when the beliefs were formed in an unreliable way. Yet, I will argue, this is a point that Descartes and the other early modern philosophers discussed here refuse to concede, and hence they are not committed to this form of internalism.

The two master arguments support internalism in different ways. First, BonJour’s Norman example shows that evidence is necessary for justification. If E stands for evidence, and J for justification, then BonJour’s argument shows: ~E → ~J. Second, Cohen’s new evil demon example supports the claim that the evidence is sufficient for justification: E → J. If we accept both these arguments, then this commits us to the view that evidence alone is necessary and sufficient for justification: E ↔ J.

Similarly, reliability (or similar facts about the belief-forming process) could be related to justification in different ways. First, reliability might be necessary for justification; so, if R stands for the reliability of the causal process that produces the belief (or similar), then on this view: ~R → ~J. Second, reliability might be by itself taken to be sufficient for justification: R → J. Goldman (1979) and Plantinga (1993) both deny that justification (or warrant) requires being aware of evidence, and instead that hold that reliability (or similar) is both necessary and sufficient for justification, and thus are committed to: R ↔ J.

Although contemporary epistemologists tend to accept either a fully internalist or fully externalist view of justification, a partially internalist, partially externalist view is possible. A fully internalist position takes evidence alone to be necessary and sufficient for justification (so, E ↔ J), and hence is committed to:

Fully Internalist View: J ↔ (E & (R v ~R))

This is exactly what Cohen’s new evil demon problem purports to show: the awareness of evidence alone justifies belief, regardless of whether or not the belief was formed in a reliable way (hence the disjunction: R v ~R). Alternatively, a fully externalist position takes reliability to be necessary and sufficient for justification (so, R ↔ J), and hence is committed to:

Fully Externalist View: J ↔ (R & (E v ~E))

The fully externalist position is insensitive to the presence or absence of evidence (hence, E v ~E). Finally, a partially internalist, partially externalist view of justification is possible:

Partly Internalist, Partly Externalist View: J ↔ (E & R)

On this view, both evidence and reliability are required for justification, and they together explain why the belief is justified.

The question to be answered in this paper, then, is which view of epistemic justification do Descartes, Locke, and Hume hold? I argue that they all think that the awareness of evidence is required for justification, but none of them think that the awareness of evidence alone is sufficient for justification. Similarly, I argue that they all think that reliability of the belief-forming process is required for justification, but none of them think reliability alone is sufficient for justification. So, neither a fully internalist position nor a fully externalist position adequately captures the views of Descartes, Locke, and Hume; rather, both the supporting evidence and the reliability of the belief-forming process explain why we should hold those beliefs, and hence explain why those beliefs are justified. Thus, they each have a partly internalist, partly externalist view of justification.

The history of epistemology presented in this paper differs significantly from the standard narrative about early modern epistemology according to which Descartes and Locke are historically important proponents of internalism. Some have pushed back against the standard narrative, arguing for externalist interpretations of Descartes or Locke or Hume.³ By contrast, this paper offers an overarching narrative about the prevalence of externalism in early modern epistemology.⁴ Although I think the interpretation offered here is persuasive, I do not pretend it will be decisive; rather, the plausibility of the story told here calls for, at the very least, a reevaluation of the oft-repeated claim that epistemic externalism is a recent development in epistemology.

2. Descartes on Justification

Descartes is supposedly the paradigmatic internalist. Many see Descartes as an internalist because he demands that we prove the veracity of sensation and reason, and any such proof would be internalist justification. However, requiring evidence as a necessary condition for justification does not make evidence a sufficient condition (~E → ~J does not entail E → J). In my view, Descartes also thinks that reliability helps explain why a belief is justified, and hence he has a partly internalist, partly externalist account of justification (so, J ↔ (E & R)).

In the Meditations, Descartes tries to establish a foundation for knowledge. But not just any foundation will do. If the foundation is flawed, having “a large number of falsehoods”, then this makes the beliefs based on that foundation “highly doubtful”. He concludes that “if I wanted to establish anything at all in the sciences that was stable and likely to last”, he would have to “demolish everything completely and start again” (Meditations 1: 17/12, my emphasis).⁵ So, famously, Descartes questions the truth of all of his beliefs by raising skeptical doubts about the reliability of sensation and reason (First Meditation). He then tries to escape this skeptical doubt by identifying some things that cannot be doubted: my own existence (Second Meditation), God’s existence (Third Meditation), and that God is no deceiver (Fourth Meditation). He then uses these beliefs that cannot be doubted as a basis or foundation for the reliability of reason and sensation, and hence for the truth of the beliefs based on these mental faculties (Meditations 4–6).

One reason to interpret Descartes as an internalist is that he thinks the veracity of sensation is not sufficient to justify beliefs based on sensation. Suppose, for example, I have the sensation of sitting by the fire; my natural inclination is to believe that I really am sitting by the fire (Mediations 3: 38–39/26–27). Sense perception, then, is a natural belief-forming process. Furthermore, Descartes concedes that sensation is probably reliable (Meditations 1: 22/15, 6: 73/51), but in the First Meditation its reliability is not certain: “a man who sleeps at night . . . has all the same experiences” (Meditations 1: 19/13, emphasis added); or perhaps an omnipotent being makes “all these things appear to me to exist just as they do now” (Meditations 1: 22–23/15, emphasis added). Thus, if I were to form a belief that I am sitting by the fire, basing that belief solely on my experience, then I might be wrong. Here the possibility of error prevents me from knowing that there is a fire (Meditations 1: 18/12). Thus, Descartes denies that reliability of sensation alone justifies my belief that there is a fire. I can know the fire is there only after proving that God exists and would not allow my sensations to deceive me in this way (Meditations 6: 89–90/61–62).

A similar story goes for reasoning. When thinking carefully about some things, the truth seems so obvious, so transparent, so “clear and distinct” (as Descartes famously puts it) that I am psychological compelled to believe it is true. Yet, he suggests, perhaps this feeling of certainty is merely the deception of an evil demon; given this possibility, “may I not . . . go wrong every time I add two and three or count the sides of a square, or in some even simpler matter . . . ?” (Meditations 1: 21/14). Such a possibility of error undermines certainty. Now, according to Descartes, there is in fact a divine guarantee that my clear and distinct perception is reliable. But prior to being aware of this divine guarantee, there remains an apparent possibility of error and, for this reason, at the end of the First Meditation, Descartes concludes that I cannot know that 2 + 3 = 5. Therefore, the reliability of clear and distinct perception alone does not justify beliefs formed in this way, and instead I must be aware of evidence that clear and distinct perception is reliable.

These cases support an internalist interpretation of Descartes. Even if a belief that p is formed in a reliable way (either by sense perception or by clear and distinct perception), the reliability of this belief-forming process is not sufficient justification for knowledge. For the belief to be justified, and thus be knowledge, I must be aware of evidence that guarantees that p is true. This is similar to BonJour’s argument that Norman’s belief that the president is in New York City is unjustified because, even though the belief was formed by a reliable power of clairvoyance, Norman lacks evidence that his power of clairvoyance is reliable. Like BonJour, then, Descartes is committed to some form of internalism.

It is not yet clear, though, whether Descartes thinks the awareness of evidence alone provides the justification needed for knowledge. Descartes may have a fully internalist position that takes evidence alone to justify belief (so, J ↔ (E & (R v ~R)). If there is a belief that Descartes acknowledges is both fully justified and was formed in an unreliable way (so, ~R & J), this would be decisive evidence that reliability is not necessary for justification, and such a passage would therefore provide strong textual evidence for a fully internalist interpretation. The New Evil Demon (Cohen 1984) is supposed to provide this kind of example because beliefs are supposed to be justified by my evidence even though the belief is formed in an unreliable way. However, this is a point, I contend, that Descartes refuses to concede.

Harry Frankfurt interprets Descartes as having a fully internalist position. According to Frankfurt (2008), Descartes aims to establish a rational basis of belief rather than to establish the objective truth of belief. In support of this interpretation, Frankfurt cites Descartes’ response to the objection that clear and distinct perception might be false. Descartes seems to concede the point and insist that all that really matters is that we do not have any reason for thinking that what is clearly and distinctly perceived is false:

First of all, as soon as we think we are correctly perceiving something, we are spontaneously convinced that it is true. Now if this conviction is so firm that it is impossible for us ever to have any reason for doubting what we are convinced of, then there are no further questions for us to ask: we have everything we could reasonably want. What is it to us that someone may feign [fingat] that the perception whose truth we are so firmly convinced of may appear false to God or an angel, so that it is, absolutely speaking, false? Why should this alleged “absolute falsity” bother us, since we neither believe in it nor have even the smallest suspicion of it? For the supposition that we are making here is of a conviction so firm that it is quite incapable of being destroyed; and such a conviction is clearly the same as the most perfect certainty. (Meditations, Replies: 144–45/103, emphasis added)⁶

In this imagined scenario, I have sufficient justification for “the most perfect certainty” and yet my belief “is, absolutely speaking, false”. Frankfurt (2008: 248) says, “Descartes evidently recognizes that his position entails that from our knowing something with perfect certitude it does not follow that it is, ‘speaking absolutely’, true”. Therefore, on this interpretation, the awareness of evidence (here, the apparent impossibility of error) is alone sufficient justification for knowledge.⁷

Against Frankfurt’s interpretation, though, Lex Newman (2013: 27) persuasively argues that “Descartes does not contemplate this [imagined scenario] as an actual possibility.” Descartes says “someone may feign” clear and distinct perception is false, but “feigning” this scenario hardly commits Descartes to this being a genuine possibility. Furthermore, it is not even Descartes who feigns the alleged possibility! Descartes argues in the Meditations that clear and distinct perception is guaranteed by God to be true, and in the noted passage he is responding to the objection that he does not successfully “rule out the possibility that anyone of sound mind may be deceived on matters which he thinks he knows clearly and distinctly” (Meditations, Replies: 126/90; 143–44/102–3). So, someone else imagines that clear and distinct perception is false, but this is not a scenario that Descartes asserts in his own voice as an actual possibility. To the contrary, Descartes’ initial reply to the objection is that “this kind of explanation is impossible” and “it is impossible for us to be deceived” by clear and distinct perception (Meditations, Replies: 143–44/102–3). Thus, Descartes never clearly endorses the possibility that clear and distinct perception could be false and, to the contrary, he insists, including in the very passage at issue, that clear and distinct perceptions must be true.

Moreover, elsewhere Descartes implies that justification entails truth, and if justification entails truth, then a justified belief must be formed in a perfectly reliable way. In the Second Meditation, Descartes argues that I think and therefore “‘I exist’ is necessarily true [necessario est verum]” (Meditations 2: 25/17). As John Carriero observers, the necessario indicates that “I exist” must be the case, and “the verum makes explicit” that “I exist” is actually true (Carriero 2008: 304). If other beliefs can satisfy the same standard (i.e., like the cogito, I recognize that the beliefs cannot possibly be false), then I could likewise know those things. So, Descartes tentatively “lay[s] down as a general rule that whatever I perceive clearly and distinctly is true” (Meditation 3: 35/24, emphasis added), a rule that is vindicated in the Fourth Meditation so that, when this rule is followed, “it is quite impossible for me to go wrong” (Meditations 4: 62/43). Carriero rightly concludes, “Descartes is plainly assuming here that I cannot be certain of something false” (Carriero 2008: 307). For Descartes, the justification needed for knowledge entails truth.

The textual evidence for a fully internalist interpretation of Descartes therefore falls short. If Descartes said or implied that a belief can be fully justified when it is false or formed in an unreliable way (so, ~R & J), then this, together with his demand for evidence of reliability, would support a fully internalist interpretation. But when pressed on this very issue, Descartes denies that clear and distinct perception could ever be false. Instead, his point seems to be that a belief is justified when it is supported by such strong evidence that we could not possibly be wrong (so, E → R). It appears, then, that according to Descartes a belief is justified if and only if the belief is both supported by evidence and formed in a reliable way (so, J ↔ (E & R)). Since we have not been able to find a justified belief that is formed in an unreliable way, a fully internalist interpretation (so, J ↔ (E & (R v ~R)) is not well-supported. At the very least, then, there is reason to doubt the traditional interpretation of Descartes as a paradigmatic internalist.

Although I have argued that evidence and reliability together are necessary and sufficient for justification (so, J ↔ (E & R)), this does not yet get me everything I want in a partially externalist interpretation. Like Goldman (1979: 1), I want to identify what it is that explains why a belief is justified. For Goldman, a belief is justified by reliability; even if a belief is also supported by evidence, Goldman thinks it is the reliability that, fundamentally, explains why the belief is justified. So likewise, an internalist interpreter of Descartes may grant that justified beliefs are formed in a reliable way, and yet insist that it is the evidence that explains why the belief is justified. By analogy, think of the Euthyphro dilemma: even if being loved by the gods is necessary and sufficient for piety, being loved by the gods does not explain why it is pious. The current suggestion, then, is that even if E entails R, and so J ↔ (E & R), it is really only E that explains why a belief is justified. A persuasive case for a partially externalist interpretation, then, would show not only that J ↔ (E & R) but also that R helps explain why the belief is justified.

Consider the justification for a belief in my existence. In the First Meditation, Descartes raises the (epistemic) possibility of error, and this possibility undermines the justification for my belief there is a fire there and my belief that 2 + 3 = 5. In the Second Meditation, though, Descartes discovers a belief about which I cannot be mistaken: given that I am thinking, “I exist is necessarily true” (Meditations 2: 25/17). Descartes says he “recognizes it as self-evident” that he exists because “it is impossible that he should think without existing” (Meditations, Replies: 140/100). It is epistemically impossible because I cannot think of myself as not existing while thinking. It is also metaphysically impossible because it is true that I cannot think without existing. Further, the cogito is so persuasive because it is metaphysically impossible for me to be wrong about my existence. Descartes appeals to the necessity of the belief being true and the impossibility of error as part of the explanation for the justification of this belief. It appears, then, that the infallibility (reliability) of the belief helps justify my belief in my existence.

Descartes then takes the cogito as model for knowing other things:

I am certain that I am a thinking thing. Do I not therefore also know what is required for my being certain about anything? In this first item of knowledge there is simply a clear and distinct perception of what I am asserting. . . . So, I now seem to be able to lay it down as a general rule that whatever I perceive very clearly and distinctly is true. (Meditations 3: 35/24, my emphasis)

Descartes identifies the cogito as clear and distinct perception, and then concludes that other things that are clearly and distinctly perceived can also be known to be true. If clear and distinct perception in the cogito reveals that it is metaphysically impossible for my belief to be false, and this helps justify the belief in my existence, then we can expect that clear and distinct perception will do the same for other beliefs. This is confirmed later in the Meditations for my belief in the existence of God and material objects.

In the Third Meditation, Descartes argues that my belief that God exists is justified because, given my innate idea of God, it is impossible for God not to exist:

The whole force of the argument lies in this: I recognize that it would be impossible for me to exist with the kind of nature I have—that is, having within me the innate idea of God—were it not the case that God really existed. (Meditations 3: 51–52/35)

. . . a most important consideration—indeed, one on which the entire luminous power of the argument depends—namely, that this ability to have within us the idea of God could not belong to our intellect if . . . [it] did not have God as its cause. (Meditations, Replies: 105–6/77)

Descartes appeals to the necessity of God’s existence as justification for my belief in God’s existence. It is clear that God’s existence is epistemically necessary (“I recognize that it would be impossible” for God not to exist), but perhaps it is less clear whether Descartes regards God’s existence as metaphysically necessary. On the one hand, if Descartes does not think the argument shows that God’s existence is metaphysically necessary, then the argument would be exceedingly disappointing: if belief in God might turn out to be false, that belief is a rather shaky epistemic foundation, contrary to Descartes’ intention (cf. Meditations 1: 17–18/12). On the other hand, if Descartes thinks that he has shown the metaphysical necessity of God’s existence, then this would provide a satisfying explanation for why I can know that God exists and would secure the kind of unshakable foundation Descartes seeks. I argued above that clear and distinct perception guarantees a belief is true (so, E → R), and now I am suggesting that its reliability helps explain why I should believe that God exists.

Also, in the Sixth Meditation, Descartes argues that I can know that material objects exist because God is not a deceiver. He considers possible causes of sensation, and then argues that God would be a deceiver if material objects were not the cause of sensation:

I do not see how God could be understood to be anything but a deceiver if the ideas were transmitted from a source other than corporeal things. It follows that corporeal things exist. . . . the very fact that God is not a deceiver, and the consequent impossibility of their being any falsity in my opinions which cannot be corrected by some other faculty supplied by God, offers me a sure hope that I can attain the truth even in these matters. (Meditations 6: 80/55–56, my emphasis)

Because God is no deceiver, my sensations must be caused by material objects, and hence it is necessarily that material objects exist. Again, it would be very disappointing if Descartes were to admit that, despite all his effort, material objects might not exist. But if Descartes instead takes himself to prove that it is metaphysically necessary that material objects exist, then that would provide a satisfying explanation for why he thinks I can know that material objects exist.

I have presented a cumulative case that, for Descartes, the reliability of the belief-forming process justifies beliefs. In the cogito, in the proof for God, and in the proof of external objects, I recognize that the relevant belief is metaphysically necessary. Further, the recognition of this metaphysical necessity provides a plausible explanation for why I should believe it (in each case, the arguments is much less plausible otherwise). If reliability is both required for justification and helps explain why the belief is justified, then reliability is part of the justification for the belief. So, I suggest, Descartes has a partly externalist account of justification.

Michael Della Rocca also interprets Descartes as a kind of externalist, but for different reasons than I have given here. It will be worthwhile, then, to consider his interpretation. According to Della Rocca, Descartes justifies foundational beliefs (such as the belief that God exists) in the same kind of way as contemporary externalists. Contemporary externalists often argue the fact that sensation is reliable can justify perceptual beliefs without first proving that sensation is reliable (Goldman 1979; Plantinga 1993; etc.). Della Rocca argues that Descartes adopts the same kind of position for foundational beliefs. Knowledge that I exist, God exists, and God is no deceiver are foundational in that (1) this knowledge serves as the basis for my knowledge of other things and (2) this knowledge is based on clear and distinct perception even before verifying the truth rule:

This certainty [of the foundational beliefs] does not require Descartes to have at t₁ a proof of the claim that clear and distinct ideas in general are true. . . . Certainty at t₁ is simply a matter of clearly and distinctly perceiving at t₁ that p is true. In this way, we can see that Descartes is a kind of externalist with regard to the justification of current clear and distinct perceptions. A current clear and distinct perception gives us knowledge or certainty even without our “checking up” on that perception and realizing that clear and distinct ideas in general must be true. (Della Rocca 2005: 19, emphasis added)

Clear and distinct perception can justify a belief that p even before it is proven that clear and distinct perception is reliable, and so it appears to be the fact of its reliability justifies the belief. Hence, Della Rocca concludes that Descartes is an externalist.

While I agree with Della Rocca that Descartes accepts a form of externalism, my interpretation assigns a more important role to the awareness of evidence. Arguably, Descartes also thinks that God’s existence is self-evident (Newman 2013). Della Rocca could grant that, in these cases, I have both the awareness of evidence and a reliable belief-forming process, and yet insist that it is only the reliability of clear and distinct perception that justifies the foundational beliefs. It is more plausible, though, to grant that the clear and distinct perception that p is evidence for its truth and this evidence helps explains why I should believe it. In that case, the awareness of evidence helps explain why the belief is justified, supporting an internalist view of justification. For this reason, I prefer a partly internalist, partly externalist interpretation of Descartes.

The same line of reasoning applies to Louis Loeb’s externalist interpretation of Descartes. Loeb (2010) argues that Descartes’ goal in the Meditations is to establish an unshakable set of beliefs. The result of clear and distinct perception is a “psychological irresistible” belief (2010: 12) and, on this interpretation, this psychological property of irresistibility explains why Descartes takes the belief to be justified. This is an externalist interpretation because it is a fact about the belief-forming process (its irresistibility) that justifies the belief (see Loeb 2010: 27–29). I agree with Loeb that the irresistibility of clear and distinct perception helps explain why Descartes regards such beliefs as justified. But, like Della Rocca’s interpretation, this understates the role of evidence. Descartes thinks that clear and distinct perception makes the truth self-evident, and presumably my recognition of this evidence also helps explain why the belief is justified. Again, then, I prefer a partly externalist, partly internalist interpretation over a fully externalist one.

I conclude, then, that internalism and externalism can both be attributed to Descartes. He is an internalist because without the evidence a belief cannot be justified (so, ~E → ~J). He is an externalist because the belief would not be justified unless it was formed in a reliable way (so, ~R → ~J). Further, for Descartes, satisfying one of these conditions partly depends on satisfying the other. The degree of evidence required for justification entails that the belief cannot be wrong, and hence that the belief has been formed in a reliable way (so, E → R). The reliability of the belief also helps explains why the belief is justified. For, it is my recognition of the belief as reliable that provides me with the evidence I need to justify the belief. So, both evidence and reliability are necessary for justification and together they are sufficient to explain why the belief is justified (so, J ↔ (E & R)).

The oft-repeated narrative that externalist accounts of justification emerged in the middle of the twentieth century thus appears to be mistaken. Consider, for example, this commentary by Lawrence BonJour:

When viewed from the general standpoint of the western epistemological tradition, externalism represents a very radical departure. It seems safe to say that until very recent times, no serious philosopher of knowledge would have dreamed of suggesting that a person’s beliefs might be epistemically justified simply in virtue of facts or relations that were external to his subjective conception. Descartes, for example, would surely have been quite unimpressed by the suggestion that his problematic beliefs about the external world were justified if only they were in fact reliably related to the world—whether or not he had any reason for thinking this to be so. Clearly his [Descartes’s] conception, and that of generations of philosophers that followed, was that such a relation could play a justificatory role only if the believer possessed adequate reason for thinking that it obtained. Thus the suggestion embodied in externalism would have been regarded by most epistemologists as simply irrelevant to the main epistemological issue, so much so that the philosopher who suggested it would have been taken either to be hopelessly confused or to be simply changing the subject. (BonJour 1980: 56, emphasis added)

But, I have argued, Descartes is far from being the paradigmatic internalist, and so this narrative about the history of epistemology falls apart. We ought to wonder, then, whether other notable epistemologists of the time period, such as Locke and Hume, really embraced a fully internalist position. In the coming sections, I will argue that they did not.

3. Locke on Justification

Locke tells us that one of his primary goals in the Essay is to demarcate knowledge and judgment (Essay 1.1.2: 43).⁸ He repeatedly reminds us of the futility of seeking knowledge where it is not possible, and in those cases we must rely instead on probable judgment. In either case, though, rational beliefs require evidence. Thus, Locke is committed to some kind of internalism. However, as with Descartes, a fully internalist interpretation greatly understates the extent to which Locke also takes justification to depend on the cause of the belief. In this section, I argue that, according to Locke, knowledge requires both evidence and reliability (so, J ↔ (E & R)).

Locke takes all knowledge to require evidence, yet he distinguishes “three kinds of Knowledge” with “different degrees and ways of Evidence” (Essay 4.2.14: 538). The highest degree of knowledge comes from intuition, an a priori perception of an agreement (or disagreement) between ideas (Essay 4.2.1: 531). For example, I know, just by reflecting on my ideas, that “white is white” and “white is not black”. A Lockean “intuition” is the immediate grasp of a necessary truth of this kind. If I grasp that p is necessarily true, that provides me with evidence that it is true, and Locke repeatedly describes it as such (e.g., Essay 4.2.7, sect. title; 4.7.2; 4.18.5, etc.). The second degree of knowledge comes from demonstration, which is a series of intuitions (Essay 4.2.2: 531–32), and so a demonstration likewise provides evidence (Essay 4.2.4; 4.14.3). Finally, the third and lowest degree of knowledge is “sensitive” knowledge, or knowledge based on sensation (Essay 4.2.14: 537–38). Again, Locke repeatedly refers to the “evidence of our senses” or similar (e.g., Essay 4.11.6; 4.16.5, 4.20.8). Thus, all three degrees of knowledge require an awareness of evidence.

Rational belief that falls short of knowledge likewise requires evidence. Perhaps Locke’s most influential demand for evidence is in his account of religious belief. Revelation is guaranteed to be true, yet “our Assurance” that a revealed proposition is true “can be rationally no higher than the Evidence of its being a Revelation” (Essay 4.16.14: 667). Locke criticizes religious “enthusiasts” who claim to receive direct revelation from God. Suppose, just for the sake of argument, that God really does reveal that p to the enthusiasts. Locke says, “The question then here is, How do I know that God is the Revealer of this to me . . . ? If I know not this, . . . it is groundless” (Essay 4.19.10: 701, my emphasis). The enthusiasts here are in a similar position as Norman the clairvoyant. Even though the enthusiasts believe p because it was revealed to them (and thus is guaranteed to be true), the belief is still not justified because they do not have evidence that the belief is based on genuine revelation. Like BonJour, then, Locke rejects the view that objective reliability is sufficient for justification; instead, Locke holds that justification requires that Norman be aware of evidence that God revealed that p.

Yet, perhaps inferring internalism from Locke’s demand for evidence is too quick; Locke may not be using the term “evidence” in the same way as it is used in contemporary epistemology. For example, David Owen (1999: 51) argues that Locke’s use of “evidence” is a “mainly causal” relationship between the content a proposition and the belief in the proposition. Owen points out that, for Locke, “intuitive Evidence . . . infallibly determines the Understanding” (Essay 4.15.5: 656), and thus the intuition causes the belief. Similarly, “That which causes the Assent” to testimony is the “Veracity of the Speaker” (Essay 4.15.1: 654). So, Owen takes “evidence” as the cause of the belief rather than some logical support for the belief (Owen 1999: 51ff. and 185ff.).

I agree with Owen that, on Locke’s view, evidence is a cause of belief, but I also think “evidence” for Locke indicates logical support for the belief. Consider the proposition “white is not black”. According to Locke, the intuition (or the immediate a priori perception) that this proposition is true causes me to believe it. But intuition is also the perception that such a proposition is “self-evident” (Essay 4.7.2: 591). Presumably, the fact that I am aware that “white is not black” is self-evident helps explain why I should believe it. (It would be absurd to say otherwise.) In that case, the awareness of evidence (understood as logical support) helps justify the belief. Similarly, Locke describes probability in terms of observed frequency (see Essay 4.16.6–9: 661–63), and this observed frequency can provide logical support that a proposition is likely to be true. Further, as Frederick Schmitt points out, for Locke “evidence” is normative: there is something we should believe given the evidence (Schmitt 2014: 60–61; see also Owen 1999: 51). So, even if Locke does not use the term “evidence” in exactly the same way that contemporary evidentialists do, it is at least “akin to this contemporary vein in epistemology” (Schmitt 2014: 61).

When Locke demands evidence supporting knowledge and rational belief, he endorses internalism. The awareness of evidence does cause belief, as Owen suggests, but this does not rule out a notion of evidence as logical support. To the contrary, Locke thinks of a priori perception and probabilistic reasoning as logical support that helps explain why I should have some beliefs but not others. Evidence, taken as logical support, helps justify the belief.

It is not yet clear, though, whether Locke thinks an awareness of evidence alone is sufficient for justification. If he says or implies that a belief can be justified even though it was formed in an unreliable way (so, ~R & J), then that would provide strong evidence for a fully internalist position (so, J ↔ (E & (R v ~R)). But, in fact, he makes no such admission.

Locke grants that a belief can be justified even if it is false, and here he comes close to endorsing the view that a belief can be justified even if it is formed in an unreliable way. According to Locke, God wants us to “follow the clearer Evidence, and greater Probability.” Locke says:

He that does not this to the best of his Power, however he sometimes lights on Truth, is in the right but by chance; and I know not whether the luckiness of the Accident will excuse the irregularity of his proceeding. This at least is certain, that he must be accountable for whatever Mistakes he runs into: whereas he that makes use of the Light and Faculties GOD has given him, and seeks sincerely to discover Truth, by those Helps and Abilities he has, may have this satisfaction in doing his Duty as a rational Creature, that though he should miss Truth, he will not miss the Reward of it. (Essay 4.17.24: 688, my emphasis)

Suppose I believe, on the basis of the best available evidence, that p. Even if p is false, Locke seems to think the belief is justified. Given this possibility, it appears that Locke is admitting that a belief can be justified even it is not formed in a reliable way (Wolterstorff 1996: 11, 62, 64, 122). If so, then it appears that, for Locke, it is evidence alone that justifies belief.

However, although Locke grants that beliefs based on evidence can be justified but false, it is not clear that he concedes that a belief can be justified even if the belief-forming process is unreliable. When the evidence falls short of certainty, “the Mind if it will proceed rationally, ought to examine all the grounds of Probability, and . . . reject, or receive it” depending on “the greater grounds of Probability on the one side or the other” (Essay 4.15.5: 656). That is, I ought to believe whatever is most likely to be true given the evidence. The epistemic probability of a proposition depends on its observed frequency (Essay 4.16.6–9: 661–63), yet the observed frequency depends on the actual frequency. Although the observed frequency can differ in some cases from the actual frequency (e.g., Essay 4.15.5: 656–57), in the long run observed frequency and the actual frequency will converge. So, perhaps Locke thinks probable judgments are justified because, in the long run, they are objectively likely to be true. If so, then we have not identified an example of a justified belief formed in an unreliable way; for, on this view, believing on the basis of evidence would be a reliable way to form beliefs (so, E → R). Moreover, there is some reason to think this is Locke’s view.

Locke’s soliloquy on the “love of truth” (Essay 4.19.1) suggests that he would reject a fully internalist account of justification. He argues that we should be “lovers of Truth for the Truths sake,” and basing beliefs on the evidence is the best means to the end of forming true beliefs: “whatsoever degrees of Assent he affords it beyond the degrees of that Evidence, ‘tis plain all that surplusage of assurance is owing to some other Affection, and not to the love of Truth” (Essay 4.19.1: 697). On Locke’s view, beliefs should be based on evidence because this is a reliable way to form true beliefs.⁹

Contrast Locke’s view of with Cohen’s new evil demon argument for internalism. Cohen argues that even if I am being deceived by an evil demon, and hence my beliefs are formed in an unreliable way, my perception of the fire nonetheless justifies my belief that there is a fire there. On this view, what I should believe is determined by the evidence even if basing beliefs on evidence is unreliable. But, Locke might argue, Cohen mistakes the means for the end: evidence is not an end in and of itself to be pursued in the absence of reliability. Locke argues that we should base beliefs on the evidence because this is reliable, and so he is committed to J ↔ (E & R) rather than the fully internalist position J ↔ (E & (R v ~R)).

The reliability requirement is even clearer in Locke’s account of knowledge. Following Descartes, Locke takes knowledge to be certainty and certainty to entail truth. Locke says, “to know and be certain, is the same thing” (Works 4: 145) and “all along in my Essay I use certainty for knowledge” (Works 4: 273). He repeatedly describes intuition as infallible (Essay 4.1.4: 526; 4.7.4: 592; 4.7.10: 598–97, etc.). Since a demonstration consists in a series of intuitions, demonstration is likewise reliable. Finally, sensitive knowledge also requires sensation to be reliable. Locke contrasts the certainty of sensation with the uncertainty and mere probability of what is unobserved (Essay 4.11.9: 635–36 and 4.16.5: 656). This contrast shows how even a highly probable belief cannot count as knowledge (Essay 4.3.14: 546), and hence that certainty entails truth (Rockwood 2018). As with Descartes, then, a belief with justification sufficient for knowledge cannot be false, and hence a fully justified belief will always be formed in a reliable way. Thus, reliability is a necessary condition for justification.

Reliability also helps explain why a belief is justified. For example, when I have an intuition that p, “that intuitive Evidence, . . . infallibly determines the Understanding, and produces certain Knowledge” (Essay 4.15.5: 656, my emphasis). Above, I argued that the intuition as a kind of evidence or logical support helps justify the belief. Yet, as Owen (1999: 51) points out, the intuition is also a cause of the belief. Moreover, this cause of belief “produces certain Knowledge.” If the infallibility of the belief-forming process helps explain why the belief is justified, it is externalism. Locke hints at this view elsewhere, saying that when the mind “infallibly perceives” p is true it “cannot but certainly know” that p is true (Essay 4.7.10: 597). The infallibility of intuition seems to explain why I am certain, and hence why the belief is justified. Alternatively, Schmitt (2014: 54) argues that it is indubitability, rather than infallibility, that explains why intuition gives us knowledge. When Locke talks about sensitive knowledge, he frequently appeals to its indubitability, and so Schmitt argues that indubitability (not infallibility) is what makes something knowledge. But Schmitt (2014: 54) grants that the infallibility of intuition explains why intuition is indubitable. If the infallibility helps explain why intuition is indubitable and hence justified, though, then its infallibility should count as part of the justification of the belief.

Locke also appeals to the external cause of sensations to explain why the belief in external objects is justified (Wilson 2014; Rockwood 2016; 2018). Locke argues that I can know “the Existence of any other Being, but only when by actual operating upon him, it makes it self perceived by him” (Essay 4.11.1: 630, emphasis added). I can know that x exists only if x causes me to have the sensation of x. This makes the cause of the sensation a necessary condition for knowing that the object exists. More importantly, though, Locke thinks I know that an object x exists because the object causes my sensation:

’Tis therefore the actual receiving of Ideas from without that, gives us notice of the Existence of other Things, and makes us know, that something doth exist at that time without us, which causes the Idea in us. (Essay 4.11.2: 630, emphasis added)

The fact that the sensation of x is caused by x “makes us know” that x exists.

Aaron Wilson persuasively argues that the causal role in sensitive knowledge “seems not only causal, but also epistemic” (Wilson 2014: 440, my emphasis). He points out that it is not just having the ideas in our minds that justifies the belief, but it is “the receiving of ideas from without—i.e., from external things—that makes us know.” The external cause of the sensation helps to explain why I have knowledge that x exists. That is externalism: facts about the cause (or, more broadly, the reliability) of the belief-forming process justify the belief that external object exists.

Although in some places Locke appeals to the external cause of sensation to justify sensitive knowledge, it is not obvious how this account of sensitive knowledge is supposed to fit into his general account of knowledge. He defines knowledge as the “the perception of the connexion and agreement, or disagreement and repugnancy of any of our Ideas” (Essay 4.1.2: 525). Yet, the perception of an agreement between ideas seems to leave out any causal connection between those ideas and the external objects those ideas represent. Jennifer Marusic, for example, thinks that it follows from Locke’s definition that “sensitive knowledge, for Locke, consists in perceiving agreements among ideas” (2016: 236, my emphasis). So, she argues, Locke has a “strongly internalist” view of justification (2016: 235) that excludes a “broadly reliabilist” version of “epistemic externalism” (2016: 236).

However, two considerations support a partly externalist interpretation of Locke. First, even though Locke’s definition of knowledge does not include an externalist condition, it does not exclude it either. Newman’s (2007) interpretation of Locke distinguishes between knowledge per se (which consists solely in the perception of an agreement between ideas) and knowledge of a specific kind (which has some additional requirement). For example, demonstrative knowledge requires the perception of an agreement between ideas, yet it also requires that this agreement is perceived by means of an intermediate idea (see Essay 4.2.7: 533). So likewise, sensitive knowledge can require the perception of an agreement between ideas and, in addition, also require a causal connection between one of those ideas (namely, the sensation of the object) and an external object (namely, the object causing the sensation). Indeed, Locke argues that the sensation of simple ideas counts as “real ideas” that give us “real knowledge” because those ideas have an external cause and so conform to reality (Essay 2.30.2: 372 and 4.4.3: 563). So, if sensitive knowledge counts as a kind of “real knowledge” (see Newman 2007 and Rockwood 2016), then sensitive knowledge would require, in addition to the perception of an agreement between two ideas, this kind of causal connection (Rockwood 2016).

Second, I have argued that for Locke the reliability of sensation helps explain why a belief in external objects is justified, and so it ought to be included as part of the justification for that belief. Imagine, for example, Euthyphro succeeds in identifying a condition that is not only necessary and sufficient for an action to be pious, but the condition also explains why an action is pious. At that point, Socrates ought to concede he has found what he is looking for: he has found the property of an action that makes the action pious. Similarly, if we find what is not only necessary and sufficient for justification, but also what explains why the belief is justified, then we ought to take it (whatever “it” is) as the justification for the belief. Now, again, Locke says that x causing the sensation of x “makes us know” that x exists (Essay 4.11.2: 630). I suspect Marusic will concede that the reliability of sensation is supposed to help explain why the belief is justified. At that point, I say, the reliability condition should then count as part of the justification for the belief. If it is the perception of an agreement between ideas and the reliability of sensation together explain why the belief that x exists is justified, then they together justify the belief.

In addition to sensation having an external cause, Locke identifies the two ideas that are perceived to agree in sensitive knowledge (as required by his definition of knowledge in 4.1.2):

Now the two ideas, that in this case are perceived to agree, and thereby do produce knowledge, are the idea of actual sensation (which is an action whereof I have a clear and distinct idea) and the idea of actual existence of something without me that causes that sensation. (Works 4: 360, emphasis added)

How to interpret this passage is controversial, and I will not attempt to settle the issue here, but (following Nagel 2016) my view is that “the idea of actual existence of something without me” refers to the sensation of the object and “the idea of actual sensation” refers to a recognition of the sensation as a sensation.¹⁰ “When we see, hear, smell, [or] taste . . . , we know that we do so” (Essay 2.27.9: 335). According to Locke, then, we can identify a sensation as a sensation. An upshot of this interpretation is that the perceived agreement provides evidence for the existence of an external object. For, once I identify the sensation as a sensation (i.e., as an idea with an external cause), this identification gives me a reason to believe my sensation corresponds to an external object.

According to the standard story of early modern epistemology, Locke is an evidentialist who demands that rational beliefs be justified by evidence (so, ~E → ~J). This much of the standard story is true. But this version of the story leaves out the extent to which Locke is also committed to externalism. I have given two reasons for interpreting Locke as having a partly externalist position according to which reliability helps justify a belief. First, reliability is a necessary condition for justification (so, ~R → ~J). Further, I have argued, the reliability of the belief-forming-process also helps explain why the belief is justified. So, I conclude that Locke has a partly internalist, partly externalist account of justification (so, J ↔ (E & R)).

The standard history of epistemology, which says that the leading early modern epistemologists all held fully internalist views of justification and that externalism arose only in the mid-twentieth century, is beginning to unravel. Both Descartes and Locke can plausibly be interpreted as externalists who think that the reliability of the belief-forming process helps justify that belief. Next, I turn to Hume, who also has a partially externalist view of justification.

4. Hume on Justification

Like Descartes and Locke, Hume seeks to explain what justifies beliefs. With respect to certain “knowledge”, Hume largely follows the framework of Locke’s epistemology. For example, Hume famously distinguishes between (necessary) “relations of ideas” and (contingent) “matters of fact” (Enquiry 4.1: 25).¹¹ Hume’s “relations of ideas” seems to be a restatement of Locke’s definition of knowledge as the “perception of an agreement, or disagreement, between ideas” (Essay 4.1.2: 525) and knowledge of contingent matters of fact depends on experience (cf. Essay 4.11.9: 635 and Enquiry 4.16: 33). Like with Locke, both a priori perception of a necessary relation between ideas and experience of a matter of fact provide certainty. The most influential and innovative themes in Hume’s epistemology, though, concern the justification of beliefs that fall short of certainty: most notably, inductive causal inferences.

Hume has two goals in his analysis of causal inferences (see Smith 1941; Loeb 2002). First, most famously, Hume’s skeptical project is to show that causal inferences are not based on reason. Traditionally, interpreters assumed that if causal inferences are not based on reason, then they are unjustified (see Beebee 2006: ch. 5). The traditional interpretation, then, is that Hume is an internalist (so, ~E → ~J). Since Norman Kemp Smith (1941), though, commentators have come to see that Hume regards some causal inferences as justified. Second, then, there is a growing consensus that Hume has a constructive project that aims to explain why some beliefs are justified. Yet the nature of this justification remains controversial, with some arguing Hume is an internalist (e.g., Meeker 2006; Qu 2014; 2018) while others argue he is an externalist (e.g., Schmitt 1992; 2014; Loeb 2002; 2010; and Beebee 2006). Below, I argue that Hume accepts both internalism and externalism.

Hume famously argues that causal inferences are not based on reason, but instead such beliefs are produced by custom. Hume explains, “in all reasoning from experience, there is a step taken by the mind, which is not supported by any argument or process of the understanding”. Instead of some rational argument, Hume says, the inference “must be induced by some other principle” and, he adds later, “This principle is CUSTOM or HABIT” (Enquiry 5.2: 41, emphasis added, and 5.5: 43). Hume describes this inference as an “instinct” (Enquiry 5.8: 47; 5.22: 55), a “mechanical tendency” (5.22: 55), and a “propensity” (5.5: 43). Causal inferences, then, are produced by a natural but non-rational belief-forming process.

Hume has described the psychological process by which causal inferences are made. A discussion of this descriptive account inevitably leads to interpretive questions about normativity. Causal inferences are made by a non-rational belief-forming process: custom. But should I accept beliefs formed in this way? The traditional interpretation is that if causal inferences are not justified by reason, then they are not justified at all, and thus Hume holds that causal inferences are unjustified. However, there are good reasons to resist this skeptical interpretation.

Hume denies a causal inference is justified by an argument, yet he says the conclusion “may justly be inferred”, suggesting that the inference is epistemically justified. He says:

These two propositions are far from being the same, I have found that such an object has always been attended with such an effect, and I foresee, that other objects, which are, in appearance, similar, will be attended with similar effects. I shall allow, if you please, that the one proposition may justly be inferred from the other: I know, in fact, that it always is inferred. But if you insist, that the inference is made by a chain of reasoning, I desire you to produce that reasoning. (Enquiry 4.16: 34, bold added; cf. Treatise 1.3.6: 89)

If “just inference” is here interpreted as a justified inference, then Hume is saying that causal inferences are justified but not by a rational argument (Wolterstorff 1996: 166; Loeb 2002: 38–47; Schmitt 2014: 82–87).

Owen grants that Hume’s talk of ‘just inferences’ and ‘just conclusions’ “appears to be normative” (1999: 141, my emphasis), but, he claims, “Hume does not really address anything like our modern concerns with justification” (1999: 140, n. 38). Like his interpretation of Locke’s use of “evidence” (discussed above), Owen interprets Hume’s use of “just inferences” (and similar) as a causal explanation of the belief rather than an assertion about its normative justification.

However, Hume contrasts “just” causal inferences that we should accept and causal inferences we should not accept. In the Treatise, Hume develops, “Rules by which to judge of causes and effects” (1.3.15, section title: 173), and he says these “general rules . . . ought to regulate our judgment concerning causes and effects” (1.3.13: 149, my emphasis). In the Enquiry, he repeatedly refers to the “rules of just reasoning” (10: 109; 11: 145). In both cases, the rules of reasoning have normative force: there are causal inferences that should be made, and others that should not be made. As Loeb (2002: 43, n. 13) points out, Owen’s interpretation does not allow for this contrast. So, contrary to Owen’s interpretation, Hume seems to think that “just inferences” are normatively justified.

Causal inferences are justified and they are justified by custom, a non-rational belief-forming process. After granting “the authority of experience,” Hume proposes “to examine the principle of human nature, which gives this mighty authority to experience” (Enquiry 4.20: 36, emphasis added; see also Enquiry 5.2: 41). When Hume claims custom has “authority” to underwrite causal inferences, Hsueh Qu (2018: 600) rightly observes that this “is quite naturally read in a normative way, as claiming that a reliance on custom is as capable of justifying inductive inferences as reason is.” So, custom justifies causal inferences.

Custom justifies causal inference because it is a reliable belief-forming process. Hume argues:

Here, then, is a kind of pre-established harmony between the course of nature and the succession of our ideas; and though the powers and forces, by which the former is governed, be wholly unknown to us; yet our thoughts and conceptions have still, we find, gone on in the same train with the other works of nature. Custom is that principle, by which this correspondence has been effected . . .

[Nature has] implanted in us an instinct, which carries forward the thought in a correspondent course to that which she has established among external objects; though we are ignorant of those powers and forces, on which this regular course and succession of objects totally depends. (Enquiry 5.21–22: 54–55, emphasis added)

Helen Beebee (2006: 72) explains, “What Hume is clearly saying here is that there is a correspondence between causal reasoning on the one hand and the course of nature on the other. . . . This is significant because here Hume is expressing utter confidence in the reliability of causal reasoning.” She goes on to say, “I claim, then, that Hume offers a reliabilist justification of causal reasoning. Causal reasoning is justified because it works” (Beebee 2006: 73, emphasis added; cf. Wolterstorff 1996: 166; Qu 2014: 603 and 2018: 529).

Reliability also explains why Hume regards some beliefs as justified. As already noted, Hume’s use of “just” reasoning or inferences refers to (what we call) epistemic justification. Schmitt further documents that the use of “just” or justified beliefs are highly correlated with true or reliably formed beliefs. For example, demonstrations are necessarily reliable and thus maximally justified (2014: 87). Also, a causal inference or “proof” is a reliable belief-forming process and is justified. By contrast, “fallacy” or “fallacious” beliefs appear “not to be justifying because it is unreliable” (2014: 126). In short, reliably formed beliefs are often described as justified, whereas unreliably formed beliefs are unjustified, and the degree of justification corresponds with the degree of reliability. Schmitt argues that the simplest and most plausible explanation of the correlation in Hume between justification and reliability is that reliability explains why a belief is justified. I agree with Schmitt that, for Hume, the reliability of the belief-forming process helps explain why a belief is justified, and hence reliability is part of the justification for that belief.

Loeb (2002) offers an externalist interpretation of Hume, but for Loeb what justifies belief is stability rather than reliability. These two notions are related. According to Loeb (2002: 68), it is the repetition (and so reliability) of cause and effect that produces a stable disposition to make causal inferences. He further argues that the stability of this belief-forming process justifies the belief (Loeb 2002: 77). Loeb observes that Hume sets out to explain the psychological mechanism of causal inferences and, once Hume has found that, he concludes that such beliefs are justified. Apparently, then, Hume thinks that a feature of the belief forming process also explains why it is justified. According to Loeb, this feature is the stability of the disposition. The observed reliability of the causal connection creates a stable disposition to believe these causal inferences, but it is the stable disposition that explains why it is justified. By contrast, on the reliability view being defended here, the reliability explains why the belief is justified. Loeb and I agree that causal inferences are reliable and stable dispositions to believe, but Loeb takes stability to explain why the belief is justified whereas I take reliability to explain why the belief is justified. In either case, it is beliefs about the belief-forming process that justify the belief, and so to this extent Hume is externalist about justification (see Loeb 2010: 27).

We have seen that there are good reasons to interpret Hume as an externalist, but others have argued that Hume is an internalist. Kevin Meeker (2006) and Hsueh Qu (2014; 2018) argue that, on Hume’s view, reliability alone cannot justify causal inferences; I must be aware of evidence from past experience in order for causal inferences to be justified. They then conclude that Hume is an internalist rather than an externalist. In the best-case scenario, however, they show only that being aware of evidence is necessary for justification (~E → ~J), not that being aware of evidence alone is sufficient for justification (E → J). So, although they provide good reasons to interpret Hume as accepting an internalist requirement for justification, their arguments fail to show that Hume has a fully internalist position (so, J ↔ (E & (R v ~R)).

Meeker (2006: 129) argues that externalism implies that it is possible to have a justified belief without past experience and that, since Hume insists causal inferences depend on past experience, this implication creates a “devastating problem” for an externalist interpretation.¹² According to Goldman’s reliabilism, for example, if the belief is formed in a reliable way then it is justified (so, R → J). But Hume denies that Adam, had he been created as a fully-functioning adult, could make causal inferences without past experience:

Were a man, such as Adam, created in the full vigour of understanding, without experience, he would never be able to infer motion in the second ball from the motion and impulse of the first. It is not anything that reason sees in the cause which makes us infer the effect. . . .

It would have been necessary, therefore, for Adam (if he was not inspired) to have had experience of the effect which followed upon the impulse of these two balls. He must have seen, in several instances, that when the one ball struck upon the other, the second always acquired motion. If he had seen a sufficient number of instances of this kind, whenever he saw the one ball moving towards the other, he 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. (Treatise, Abstract 11–12: 650–51)

Hume says, “It would have been necessary . . . to have had experience” to make the causal inference and Adam “would never be able to infer” the effect from the cause (emphasis added). Thus, Hume considers past experience a necessary condition for causal inferences.

The problem here is that, as BonJour persuasively argues, a reliable belief-forming process without evidence of its reliability is not sufficient for justification. Meeker (2006: 134) argues that if Adam had the ability to reliably make causal inferences without past experience, Adam would be like Norman the clairvoyant: Adam “has no evidence (or ‘foundation’) to suppose that the mechanism that gave rise to the belief is at all reliable–even if the belief happens to be true.” Hume elsewhere implies that without past experience Adam’s inference would be “entirely arbitrary” (see Enquiry 4.9: 29), suggesting that even if he correctly guessed the effect that would not count as a justified belief. So, Meeker rightly concludes, Hume thinks that the evidence from past experience is necessary for the justification of causal inferences.

Qu (2014; 2018) also argues that evidence from past experience is necessary for a causal inference to be fully justified, but he uses other evidence to support his interpretation. Hume distinguishes between “antecedent” and “consequent” skepticism. According to antecedent skepticism, I ought to presume belief-forming processes are unreliable until proven otherwise, but Hume rejects this position as being excessively skeptical (Enquiry 12: 149–50). In order to avoid this excessively skeptical position, I ought to take belief-forming processes to be unreliable only after having evidence of their unreliability. This latter approach is consequent skepticism (Enquiry 12: 150). Some forms of consequent skepticism (Pyrrhonism) take the evidence to show that all belief-forming processes are unreliable (Enquiry 12, part 2), whereas Hume’s “mitigated” skepticism rejects the reliability of belief-forming processes only in the context which they are shown to be unreliable and only to the extent to which they are shown to be unreliable (Qu 2018: 520; see Enquiry 12.24–25: 161–62).

Qu argues that Hume’s mitigated form of consequent skepticism shows he is an internalist. Consequent skepticism leaves open the possibility that some beliefs are justified and others are unjustified. According to Qu (2018: 536, emphasis added), “consequent justification is a function of evidence of reliability.” If in my experience I find that causal inferences are unreliable, then I would be unjustified in making such inferences. But what I find instead is that custom (as the basis of causal inferences) is “infallible in its operations” and that there is a “correspondence” between “the course of nature” and “the succession of our ideas” (Enquiry 5.21–22: 54–55). Since I have evidence of the reliability of causal inferences, these inferences are justified. Qu (2018: 536) calls this “internalist reliabilism” because I need evidence of reliability “to fully justify” my causal inferences.¹³

In addition to the evidence presented by Meeker and Qu, I will point out that Hume often talks about past experience as a kind of evidence for causal inferences. Hume says, “A wise man . . . proportions his belief to the evidence”, with greater observed frequency of the events justifying greater confidence in a causal inference (Enquiry 10.4: 110). Elsewhere, he speaks of past experience as “evidence” (Enquiry 10.4: 110–11, 12.22: 159) and inferences from past experience as “reasoning” (Enquiry 10.16: 117) and an “argument” (Enquiry 12.29: 164) and a “proof” (Enquiry 10.12: 115). These descriptions suggest some kind of rational foundation, based on the evidence from past experience, supports causal inferences.

Thus, there are two forces at work in the dispute about how to interpret Hume’s account of causal inferences. On the one hand, Hume implies that custom justifies causal inferences. He argues that custom is “the principle of human nature, which gives this mighty authority to experience” (Enquiry 4.20: 36, emphasis added). Evidence of reliability can hardly qualify as a “principle of human nature”. So, it appears that it is a reliable belief-forming faculty, and not the evidence of the reliability of such a faculty, that Hume here identifies as the justification of causal inferences. This supports externalist interpretations of Hume. On the other hand, Hume seems to take past experience, specifically past experience that establishes a track record of reliability, as a necessary condition for causal inferences to be fully justified. This supports internalist interpretations of Hume.

The good news is that there is no conflict between these two interpretive forces; both can be accepted. Hume can consistently and plausibly hold that causal inferences are justified by the non-rational but reliable belief-forming process that produces them (externalist justification) along with my awareness of the evidence that causal inferences are reliable (internalist justification). If either the reliability or evidence was absent, then the belief in the conclusion of the causal inference would not be justified. Both are required and together they are sufficient for justification of causal inferences (so, J ↔ (E & R)). Thus, Hume accepts that both internalist and externalist conditions justify causal inferences.

The interpretation offered here again raises questions about the view that the major figures in early modern philosophy were all fully internalists in their accounts of justification, for we have seen reasons to attribute externalist justification to Hume. Still, notwithstanding the evidence that Hume is an externalist, some hesitate. Qu, for example, tells us that part of his motivation for his internalist interpretation of Hume is that “I am apprehensive about the anachronism of attributing externalism to Hume, given that early modern epistemology by and large tended to be wholly internalist (Descartes’ foundationalism being a prime example)” (Qu 2014: 619, n. 16). But if the preceding sections of this paper were successful, Qu need not worry about anachronism. The traditional view (i.e., a fully internalist view justification), it turns out, is not all that traditional.

4. Epistemic Externalism in Early Modern Philosophy

I hope to have shown that Descartes, Locke, and Hume can each plausibly be interpreted as having a partly internalist, partly externalist view of epistemic justification. All of them hold that a belief cannot be justified without evidence (so, ~E → ~J). Yet, they think, the awareness of evidence alone is not sufficient for justification. For, suppose a belief appears to be the result of an accurate belief-forming process, but the belief-forming process is actually unreliable; in that case, would my belief be fully justified? I have argued that for Descartes, Locke, and Hume the answer is no (so, ~R → ~J). The actual reliability of the belief-forming process is necessary for the belief’s justification. Further, the reliability helps explain both why it is justified and why there is evidence that the belief is formed in a reliable way. Hence, they each have a partly internalist, partly externalist account of justification (so, J ↔ (E & R)).

That Descartes, Locke, and Hume would agree in this way is rather remarkable because their respective epistemologies otherwise differ from each other in significant ways. Yet perhaps this conclusion should not be so surprising. Only recently have internalism and externalism about justification been clearly and sharply distinguished. Goldman (1967) developed the causal theory in response to the Gettier problem. Goldman, understandably, presents himself as doing something new: he rejects the view that justification requires being aware of evidence in favor of an avowedly externalist account. Later, Goldman (1979) contrasts his externalist account with the “Cartesian” view that justification is a matter of being aware of evidence. Others pushed back. BonJour (1980) defends internalism by arguing that a reliably formed belief is unjustified if there is no evidence for it. Shortly afterward, Cohen (1984) argues evidence is by itself sufficient for justification, even if believing on such evidence is systematically unreliable. And thus contemporary philosophers began to clearly distinguish internalism and externalism and to debate whether something “internal” or “external” to the mind justifies belief, either . . . or . . . , the one, but not the other.

Since Descartes and Locke so clearly demand evidence for justification, they have been widely interpreted as internalists. Descartes insists on first-person awareness of evidence that verifies the reliability of sensation and clear and distinct perception. Locke emphasizes the need for empirical evidence and insists that there is a duty to believe on the basis of such evidence. For these reasons, Plantinga (1993: vi) takes the “twin towers of Western epistemology, Descartes and Locke”, to be the “the roots of contemporary internalism.” Others likewise trace contemporary internalism back to Descartes or Locke (BonJour 1980: 365; Wolterstorff 1996: xi, xvi).

But the sharp distinction between internalist and externalist justification is a modern invention, and the forced choice between them is a false dilemma. It is possible to accept that evidence is required for justification while also maintaining that the reliability of the belief-forming process is required for justification. So, we are not forced to choose between either a fully internalist interpretation or a fully externalist interpretation of these early modern philosophers. Further, I contend that any such choice would fail to capture an important aspect of how Descartes, Locke, and Hume think about justification. Both evidence and reliability help explain why the belief is justified. So instead of choosing one or the other, we can simply attribute to them both internalism and externalism. Moreover, it is not clear to me that this view of justification is mistaken. Perhaps in this respect we have taken a step backward rather than forward; by forcing this false dilemma upon ourselves, we have both failed to appreciate the past and failed to make progress in developing a comprehensive account of epistemic justification.