1. Introduction
Let reliabilism be the view that a belief is justified if and only if it is formed by a reliable process. Keith Lehrer and Stewart Cohen (Reference Lehrer and Cohen1983: 192–3) have produced the following well-known and powerful counterexample to reliabilism:Footnote 1
Imagine that, unknown to us, our cognitive processes, those involved in perception, memory and inference, are rendered unreliable by the actions of a powerful demon … It would follow on reliabilist views that under such conditions the beliefs generated by those processes would not be justified. This result is unacceptable. The truth of the demon hypothesis also entails that our experiences and our reasonings are just what they would be if our cognitive processes were reliable, and therefore, that we would be just as well justified in believing what we do if the demon hypothesis were true as if it were false. Contrary to reliabilism, we aver that under the conditions of the demon hypothesis our beliefs would be justified in an epistemic sense.
Ralph Wedgwood Reference Wedgwood(2000: 349) uses the demon hypothesis to support epistemic internalism:
Consider two possible worlds, w 1 and w 2. In both worlds, you have exactly the same experiences, apparent memories, and intuitions, and in both worlds you go through exactly the same processes of reasoning, and form exactly the same beliefs. In this case, it seems, exactly the same beliefs are rational in both worlds, and exactly the same beliefs are irrational in both worlds. Now suppose that in w 1 you are bedeviled by an evil demon who ensures that many of your experiences are misleading, with the result that many of the beliefs that you hold in w 1 are false. In w 2, on the other hand, almost all your experiences are veridical, with the result that almost all the beliefs that you hold in w 2 are true. Intuitively, this makes no difference at all. Exactly the same beliefs are rational and irrational in both worlds … This intuition seems to support an ‘internalist’ conception of rational belief. According to this conception, the rationality of a belief supervenes purely on ‘internal facts’ about the thinker's mental states – in this example, on facts that hold in both these two possible worlds w 1 and w 2, not on facts about the external world that vary between w 1 and w 2.
The old evil demon problem used the demon hypothesis in an argument for skepticism. The new evil demon problem uses it both as a counterexample to reliabilism and also as an argument for internalism and against externalism. It has been very influential. Ernest Sosa Reference Sosa and Sosa(1991: 131) identifies it as one of the ‘three main problems’ for generic reliabilism, Laurence BonJour Reference BonJour and Moser(2002: 246) identifies it as one of the ‘three main objections’ to reliabilism, and Alvin Goldman Reference Goldman, Dancy and Sosa(1992: 435) lists it as one of the two ‘most influential counterexamples to reliabilism’.
In this paper, I aim to show that the new evil demon problem (from now on, NEDP) also afflicts the two most prominent forms of internalism: moderate internalism and historical internalism. Since virtually all internalists accept at least one of these two forms, it follows that virtually all internalists face the NEDP.Footnote 2 My secondary thesis is that many epistemologists – including both internalists and externalists – face a dilemma. The only form of internalism that is immune to the NEDP, strong internalism, is a very radical and revisionary view. This is not to say that strong internalism is implausible but only that a large number of epistemologists would have to significantly revise their views about justification if they accepted it.Footnote 3 Hence, either epistemologists face the NEDP or they must accept a very radical and revisionary view.
I will end §1 by making explicit both whom I am trying to persuade and also what I am taking myself to accomplish. My main audience is internalists and any epistemologists, whether internalist or externalist, who share internalist intuitions – those who feel the temptation to say that the demon victim has justified beliefs just as we do. Externalists who do not feel this temptation will likely not feel moved by my arguments and are not my primary audience. Still, these externalists should be interested to know that an argument, which a significant number of their epistemologist colleagues regard to be a significant threat to externalism, also applies to the two most prominent forms of internalism. Second, I do not take myself to be presenting knockdown arguments against these forms of internalism. Just as some externalists might think that other theoretical considerations in favor of externalism outweigh the NEDP against it, some internalists might think the same. I am only attempting to show that the problem does apply to these forms of internalism, whether it is knockdown or not.
2. Internalism, Externalism and the NEDP
In this section, I will define some terms. Internalism is the view that justificational properties supervene on internal properties of the believer.Footnote 4 In other words, only internal properties are directly relevant to whether a belief is justified. I say directly relevant, because any internalist will grant that external properties can be indirectly relevant to a belief's justificational status by affecting the internal properties. But only internal properties directly make a justificational difference.Footnote 5 Externalism, as I will understand it, is the denial of internalism. It entails the possibility that there are two individuals with identical internal properties but different justificational properties. The difference in justification would be a direct result of a difference in some external property, such as whether the belief was formed reliably.
What is it for a property to be internal? Perhaps T is internal if and only if T is mental. But Timothy Williamson (Reference Williamson2000) has argued that knowledge is a mental state, and most internalists would not regard knowledge as internal. It would be preferable to have a definition of ‘internal’ that did not depend on the failure of Williamson's argument. Here are two more plausible candidates. Perhaps T is internal if and only if T is a nonfactive mental state. Or perhaps T is internal to S if and only if T is introspectively accessible to S. For this paper, I will go with the former definition, simply because it is the more inclusive view.Footnote 6 Fortunately, an exact analysis of internality is unnecessary for the purposes of this paper, and an intuitive understanding should be enough.
What, exactly, is the NEDP? In its statement by Lehrer and Cohen, the NEDP took the form of a counterexample against reliabilism; this is how it is normally used. But we can also take it, following Wedgwood, as showing that no external property is directly relevant to justification. The NEDP would thereby show not only that being reliably formed is unnecessary for justification but also that it is irrelevant to it; reliable belief formation is a fifth wheel that does no justificatory work. Following Wedgwood, I understand a set of cases to exemplify the NEDP for a view V if they meet the following conditions: (1) V entails that some property P is directly relevant to justification; (2) in one case, some person S has a belief with a certain justificational status, and S (or S's belief) exemplifies P; (3) in another case, some person S* is exactly like S, except for any differences entailed by the fact that, due to the work of a demon, P is not exemplified by S (or S's belief); (4) there is the intuition that the justificational status of the beliefs of S and S* is the same; and (5) these cases can be generalized. P is normally the property being reliably formed, but it could in principle also be other external properties (e.g. being sensitive or being safe), as well as internal properties. This opens up at least the possibility that the NEDP could also apply to internalism.
Some might object to my liberal use of ‘NEDP’ and say that it should only be used as a counterexample to reliabilism or as an argument against externalist theories. I disagree; I think that cases that meet my above conditions are sufficiently similar to the original new evil demon cases to merit being called ‘instances of the NEDP’. How shall we resolve this dispute? Fortunately, nothing important stands on this issue. My objectors can take my first thesis to say that a problem that is similar to (but not the same as) the NEDP afflicts the two most prominent forms of internalism. This still leaves a serious challenge for virtually all internalists, which is what I care about most.
3. First Form of Internalism: Moderate Internalism
3.1 Defining moderate internalism
Let moderate internalism be the view that internalism is true and some unaccessed, internal properties can be directly relevant to the justificational status of beliefs. Stated as a supervenience thesis, it is the view that justificational properties supervene on internal properties, and it is not the case that justificational properties supervene on accessed properties. On moderate internalism, both accessed internal states (such as my being appeared to redly) and unaccessed internal states (such as a dispositional belief) can make the difference as to whether a belief is justified.Footnote 7
I understand a state T to be accessed by a person S if T is present to S's consciousness.Footnote 8 My being appeared to redly and my belief that I am typing a sentence (as I currently type this sentence) are accessed internal states. For T to be accessed by S, S does not need to have beliefs about T. For example, my being appeared to redly can be an accessed internal state even if I have not formed any belief about my being appeared to redly. Unaccessed internal states might include beliefs and memories that I am not currently thinking about. If asked why I believe Fred ate the last cookie, I can quickly bring to mind my memory of Fred eating the last cookie. This memory was an unaccessed internal state prior to my being asked about Fred and the cookie.
How do these terms relate to the common occurrent/dispositional distinction in epistemology? I take it that S's belief is occurrent if and only if the belief is present to S's consciousness; a belief is dispositional if and only if it is not occurrent.Footnote 9 Hence, a belief (or desire or memory) is occurrent if and only if it is accessed, and it is dispositional if and only if it is unaccessed. For example, my belief that 1 + 2 = 3 is currently occurrent (and accessed); but I continue to believe this even when I am asleep. During those times, the belief is dispositional (and unaccessed). (I also take ‘background belief’ and ‘stored belief’ to both be synonymous with ‘dispositional belief’; I will use these expressions interchangeably.)
Having explained my terminology, I think it is safe to say that virtually all internalists are moderate internalists. Otherwise, the set of potentially justifying (or defeating) beliefs would be severely restricted since only a relatively small number of our beliefs are occurrent at any given moment. So, most internalists will allow for nonoccurrent beliefs to be directly relevant to the justification of a belief. Hence, most internalists are moderate internalists.Footnote 10
3.2 The global NEDP for moderate internalism
Consider Augustine, who lived for seventy-six years. During this time, he held a great many beliefs; some were justified and some were not. Let us only consider the beliefs when they were occurrent. Now imagine Augustine*, a creature who also lives for seventy-six years and who experiences all of the same accessed internal states that Augustine had throughout his lifetime. The only difference is that the demon is manipulating Augustine* in such a way that he never has any unaccessed internal states. Augustine* might think that he has dispositional beliefs at various times in his life, but, in fact, he has none. Whenever he thinks that he is recalling one of his memory beliefs, the demon is actually creating a newly formed occurrent belief. This scenario seems metaphysically possible. Furthermore, it seems that the justificational status of Augustine's occurrent beliefs at any age X of his life will be identical to the justificational status of Augustine*'s parallel beliefs when he is also age X, even though Augustine* has no unaccessed internal states.Footnote 11
Consider that you, at this moment, might be fooled by a demon so that you are in a position similar to that of Augustine*. You might think that you have dispositional beliefs and memories, but you don't. You might try to prove that this isn't so by saying, ‘Look, I can bring to mind my dispositional belief that my childhood home was yellow.’ Unfortunately, you could still be duped by a demon. Whenever you try to bring this belief to mind, the demon could quickly create a brand-new occurrent belief. Trying to prove that the demon is not fooling you about your dispositional beliefs by recalling them is as futile as trying to prove that the demon is not fooling you about external objects by touching them. It would be to no avail. Now I do not think that you are actually fooled by a demon, but there is a possible world in which you have a twin, You*, who is in such a situation. And it seems that the justificational status of your occurrent beliefs and those of You* is the same.
Consider yet another example from the movie the Matrix. Many think that Neo, before he was freed from the Matrix, had justified beliefs. Suppose that Neo had a counterpart, Neo*, who was phenomenologically identical to Neo. It would seem to Neo* that he was typing into a computer, it would seem to him that he was recalling memories, and so on; they would share all the same sensory experiences. However, suppose Neo* had no unaccessed, internal states because the Matrix determined it to be so. It seems that the beliefs held by both Neo and Neo* would have the same justificational status. These three cases give us evidence that unaccessed internal states are not directly relevant to the justificational status of our occurrent beliefs.
Someone might object that my cases do not show that unaccessed internal states are not directly relevant to the justificatory status of dispositional beliefs. Suppose Augustine has two dispositional beliefs B1 and B2. Perhaps B2 is directly relevant to B1's justificatory status. In response, someone who thinks this and also agrees with my argument that unaccessed, internal states are not directly relevant to the justificational status of occurrent beliefs must think that if B1 were to become occurrent, then B2 would suddenly become irrelevant to B1's justificatory status. But this seems implausible. It seems more reasonable to think that B2 is not directly relevant to the justificatory status of B1 in the first place. So, we have reason to think that unaccessed, internal states are not directly relevant to the justificatory status of any of our beliefs.
Consider the following objection. Some might think that these cases are impossible for at least one of two reasons: wide-scale deletion of all of a person's unaccessed internal states will either alter the person's accessed internal states or result in the loss of the person; hence, my cases are impossible. Yet, it does still seem to me to be metaphysically possible that such deletions could happen because it seems possible that there is someone who is exactly like me but has only my accessed states. So, I take the claim that my cases are metaphysically impossible to be insufficiently grounded.
My objectors might not budge and insist that they are impossible. Instead of pursuing this discussion, I will turn to another way of pressing the NEDP for moderate internalism. The current section employed what I call the global NEDP, cases in which S and S* are internally identical except that all of S*'s unaccessed internal states are deleted by a demon. The following section employs the local NEDP, which only makes use of cases in which a limited number of specific, unaccessed internal states are deleted; these cases are immune to the above objections.
3.3 The local NEDP for moderate internalism
Moderate internalists normally have in mind certain cases in which unaccessed, internal states matter. Consider the following:
(Case 1) Fred has been outside working all day. He looks out over the horizon and has a rich sensory experience as of a field, of trees, of a drifting cloud, and of the sun. Fred forms the belief that that's a beautiful sunset. He also has the dispositional belief that it is evening.
(Case 2) Sally wakes up and she has no idea how long she's been sleeping. She considers that she may have been sleeping anywhere from three to twelve hours. Sally looks out into the horizon and has a rich sensory experience as of a field, of trees, of a drifting cloud, and of the sun. Sally forms the belief that that's a beautiful sunset. Sally has no dispositional belief that it is evening.Footnote 12
Neither Fred nor Sally has any background beliefs about which directions are east and west, and they cannot distinguish sunrises from sunsets by way of mere visual appearance. Furthermore, although Fred's belief that it is evening is dispositional, it plays a causal role in the formation of his belief that that's a beautiful sunset. It seems that Fred's belief about the sunset is justified and Sally's is not. If we asked Fred how he knows it is a sunset and not a sunrise, Fred would bring to mind some memory beliefs and say that sunrises do not occur in the evening, and it is evening. If we asked Sally the same question, she would say, ‘Oh, I didn't consider that it might be a sunrise. It looks like I judged too quickly!’
Furthermore, as I have explained the cases, the accessed internal states of Fred and Sally are the same when they form their beliefs about the sunset. They both look out over the horizon, they are appeared to in a certain way, and they form the belief. Although Fred's belief that it is evening plays a causal role in the formation of his belief that that's a beautiful sunset, he does not bring it to mind. The only relevant differences between Fred and Sally's internal states are their dispositional beliefs and memories. Moderate internalists will likely point to these states as the best explanation for the justificational difference between them.Footnote 13
However, other considerations might show that unaccessed, internal states are not actually doing the justificatory work. Consider:
(Case 3) Melissa has been outside working all day. She looks out into the horizon and has a rich sensory experience as of a field, of trees, of a drifting cloud, and of the sun. Melissa forms the belief that that's a beautiful sunset. She also has the dispositional belief that it is evening. An evil demon intervenes. First, he destroys Melissa's background belief that it is evening, and he himself causally sustains Melissa's occurrent belief that that's a beautiful sunset. After three seconds, the demon manipulates Melissa's mind so that (1) she once again has the background belief that it is evening and (2) the belief that that's a beautiful sunset is once again causally dependent on the background belief and not the demon. The demon is so skilled that throughout this process, Melissa's accessed internal experience remains uninterrupted as she enjoys the beautiful sunset.
Melissa's accessed states are exactly like Fred's. Furthermore, like Fred, Melissa's dispositional belief that it is evening plays a causal role in the formation and sustaining of her belief that that's a beautiful sunset. The only differences between them are the disappearance and reappearance of the background belief and the demon's causally sustaining her belief that that's a beautiful sunset when her background belief is absent. Plausibly, Melissa's belief that that's a beautiful sunset remains justified throughout the whole process.Footnote 14 Furthermore, Case 3 seems possible; a sufficiently talented and powerful demon could perform the task of deleting and creating a person's background belief while keeping her accessed mental states the same.Footnote 15 This possibility is easier to see than the possibility of Augustine* because it is easier to imagine the deletion of a limited number of unaccessed states rather than all of them.
Some may object that when Melissa's background belief disappears, it seems to them that her perceptual belief is not justified. I have two responses. First, note that if Melissa's belief is not justified because of the absence of the dispositional belief, then she epistemically ought to suspend judgment during those three seconds. This is counterintuitive.Footnote 16 Second, consider that you might be subject to a demon's tinkering in this way. The dispositional beliefs that you are inclined to think are relevant to the justification of your ordinary perceptual beliefs might be regularly deleted and recreated by a demon. It might be that in the last five minutes, you should have withheld some of your perceptual beliefs in increments of three seconds. Now, I do not think that you are actually in this situation, but your twin You** is. And it seems that the justificational status of your beliefs and You**'s beliefs are the same, regardless of those dispositional beliefs.
We can generalize from these examples. For any one of some subject S's occurrent beliefs that we are inclined to think is justified (or unjustified) from t1 to t2 at least partly in virtue of some dispositional mental state M, we can imagine a counterpart S* who is identical to S except for any differences entailed by the fact that M is deleted by a demon at t1 and recreated at t2. Like in the Melissa and You** cases, my intuition about any such case will likely be that the justificational status of S and S*'s occurrent beliefs from t1 to t2 is the same. It follows that unaccessed, internal properties are not directly relevant to justificational properties, and so moderate internalism is false.
In §3, I do not consider myself to have refuted moderate internalism. As I said in §1, other theoretical considerations could still tip the balance of evidence in favor of moderate internalism. But I believe I have shown that the NEDP (or something similar to it) is a problem for moderate internalism; it is a problem worth taking seriously. And since virtually all internalists are moderate internalists, I believe I have shown that there is a problem for the views of virtually all internalists.
4. Second Form of Internalism: Historical Internalism
4.1 Historical internalism
Let historical internalism be the view that internalism is true and some nonpresent internal properties can be directly relevant to justificational properties. Stated as a supervenience thesis, it is the view that justificational properties supervene on internal properties, and it is not the case that justificational properties supervene on present internal properties. One might object that historical internalism is not genuinely internalist because past internal states, though they are internal, are inaccessible now.Footnote 17 I would rather not engage in a dispute of how ‘internalism’ should be used. Instead, I will simply say that some internalists do think of historical internalism as a form of internalism, and I will argue that the NEDP applies to it, whether it is genuinely internalist or not.Footnote 18
Why might someone accept historical internalism? Gilbert Harman Reference Harman(1986: ch. 4) points out that forgotten evidence might play a role in the current justification of our beliefs. Following Harman, Thomas Senor gives the following example:
Harman notes that people often forget the original grounds of their beliefs even though, intuitively, those beliefs remain justified (for instance, I take myself to be justified in believing that my first-grade teacher's name was Mrs. McDonald, but I don't have any idea how I came by this belief, nor do I take myself to now have any other evidence for it). (Senor Reference Senor1993: 454)
This might move some philosophers to think the following:
Past Justification Principle: Justification S had for p at past time t 1 can be directly relevant to S's justification for believing that p at later time t 2.Footnote 19
On the other hand, John Greco Reference Greco, Steup and Sosa(2005, Reference Greco2009) and Alvin Goldman (Reference Goldman2009) both argue that past defeaters can be directly relevant to the justificational status of a present belief. Here is a recent example from Goldman Reference Goldman(2009: 324).
Ursula acquires by testimony an initially justified belief in a certain generalization, G. This occurs at t 1. She subsequently encounters a mountain of counterexamples to G, each of which is very obvious. The joint effect of these counterexamples, at t 5, is the total defeat of Ursula's original justification for G … hence she is unjustified in believing it. Nonetheless, Ursula continues to believe G … Much later, at t 20, Ursula continues to maintain the same high confidence in G as she has held right along … This is a clear case in which Ursula's belief in G at t 20 is unjustified. This assessment holds even if we also stipulate that at t 20 Ursula has long since forgotten the defeating evidence she encountered at t 5.
Goldman probably stipulates at the end that Ursula forgets the defeating evidence because he wants to show that the past encounters of the counterexamples at t 5 do justificational (or defeating) work, not the present memories or dispositional beliefs about the counterexamples. These cases support the following:
Past Defeat Principle: Defeaters S had for p at past time t 1 can be directly relevant to S's justification for believing that p at later time t 2.
Internalists who accept these principles should believe historical internalism.
A historical internalist can also explain the justificational difference between Fred and Sally in Cases 1 and 2 without appealing to unaccessed, internal states. Fred will likely have formed the occurrent belief that it is evening sometime in the recent past for it to have become a dispositional belief; Sally never formed this belief.Footnote 20 Even if moderate internalism is false, Fred's past accessed, internal state might make the justificational difference between him and Sally. And the Past Justification Principle explains why this is so. So, by embracing historical internalism, one can reject moderate internalism, retain the intuitions about Cases 1 and 2, and remain internalist.
4.2 The global NEDP for historical internalism
Unlike the hitherto unknown global NEDP for moderate internalism, the global NEDP for historical internalism has already been used by Michael Huemer.Footnote 21 Inspired by Bertrand Russell, he writes the following:
But now recall Russell's five-minute hypothesis. Suppose God created someone five minutes ago in exactly the state that I was in five minutes ago, surrounded by exactly the same kinds of things. Call this person Mike2. Mike2 was created complete with false memories of his past life, identical to my memories of my past life. He thinks his name is ‘Mike’ and is presently writing a paper about the problem of memory knowledge. His situation would be (to him) indistinguishable from my actual situation … What sorts of things would it be rational for Mike2 to believe? Pretty clearly, just the same things that it is rational for me, now, to believe (modulo appropriate changes in indexical references). Most of Mike2's beliefs about his own past are false, but he has no way of knowing that, and no more reason for suspecting it than I have for suspecting that my beliefs about my past are false. So if I am justified in believing that I ate a bagel this morning, Mike2 is justified (though mistaken) in believing that he ate a bagel this morning. (Huemer Reference Huemer1999: 350)
Here is another example. Consider that it is possible that you have a twin, You***, who is a molecule-by-molecule replica of you. You*** has all of the same memories and beliefs that you do, but You*** was created five minutes ago. (Note that you have no way of confirming that you are not You***!) The justificational status of your beliefs and You***'s beliefs seem to be the same.
We can also consider an alternative version of Case 1 (call it Case 4) in which a demon created the world a few minutes before Fred's internally identical counterpart, Fred*, looks out into the horizon and forms the belief that that's a beautiful sunset. Fred* shares Fred's stored memories and beliefs, but Fred*'s are mostly false; he also has the background belief that it is evening. Given the presence of this background belief, it seems that Fred*'s belief that's a beautiful sunset is justified. We can similarly formulate an alternative version of Case 2 (call it Case 5) in which a demon created the world a few minutes before Sally* formed her belief. Sally* shares Sally's stored memories and beliefs, but Sally*'s are mostly false; she also does not have the background belief that it is evening. Given the absence of this background belief, it seems that Sally*'s belief that's a beautiful sunset is not justified. It seems that the past does not make a direct justificational difference to Fred and Sally's beliefs. One might point out that what motivates these intuitions is that, when we imagine Cases 4 and 5, we imagine Fred* with unaccessed internal states (the background belief that it is evening) that Sally* does not have. Doesn't this bring us back to moderate internalism? I will discuss this question in §5. For now, I will say that it is clear that these sorts of cases can be multiplied. In reflecting on these three examples, it seems plausible that past internal properties are not directly relevant to justificational properties.
What, then, are we to say about the Past Justification Principle and the Past Defeat Principle? Or what should we say about the cases of forgotten evidence that motivated them? Recall that in wielding the NEDP against historical internalism, I am not trying to give a knockdown argument against it. I mainly just want to show that it is a problem worth taking seriously, and I acknowledge that there might be other theoretical considerations that make historical internalism an overall plausible view to hold.Footnote 22
Some might object that these cases are impossible. For example, they may say that Mike2 has the same beliefs and intentional states as the real Michael Huemer only if Mike2 has a certain sort of history, one precluded by Mike2's being created five minutes ago. For example, perhaps a being has intentional states only if it has had the right sort of evolutionary history.Footnote 23 Or perhaps a being can have a dispositional belief only if it was occurrent at some point in the past.Footnote 24 Despite these objections, it does still seem to me to be metaphysically possible that a molecule-by-molecule replica of me (created five minutes ago) could share the same beliefs and memories I have, even dispositional ones.
My objectors might not budge and insist that they are impossible. Instead of pursuing this discussion, I will turn to another way of pressing the NEDP for historical internalism. The current section employed what I call the global NEDP, cases in which S and S* are internally identical except that almost all of S*'s past internal states are absent due to a demon. The following section employs the local NEDP, which only makes use of cases in which a limited number of specific, past internal states are absent due to a demon; these cases are immune to the above objections.
4.3 The local NEDP for historical internalism
Historical internalists normally have in mind certain cases in which past internal states are supposed to matter, cases like Goldman's and Senor's cases. Since Goldman's case is the most detailed and described, I will focus on it.Footnote 25 In that case, Ursula justifiedly believed G at t 1 but encountered counterexamples to G at t 5. She continued to believe G until t 20, by which she had ‘long since forgotten the defeating evidence she encountered at t 5’. Now, the historical internalist might suggest a case with Ursula* who justifiedly believed G at t 1, never encountered the counterexamples at t 5, and has the same internal states as Ursula at t 20. To make this follow the structure of the local NEDP for historical internalism, we can even say that a demon was the reason why Ursula* never encountered the counterexamples at t 5. If the local NEDP is successful, then we should expect the justificatory states of Ursula and Ursula* to be the same, thus showing that the difference in past internal states didn't matter. However, we instead find that it seems that Ursula*'s belief is justified at this later time, and Ursula's belief is not. Furthermore, the only relevant differences between them are their past internal states. Historical internalists will likely point to these states as the best explanation for the justificational difference between them. They may also point out that the local NEDP for historical internalism fails.
I will show both that Goldman's case, once clarified, fails to provide support for historical internalism and also that the local NEDP is vindicated. I will clarify three points about Goldman's case. First, we sometimes use the word ‘forget’ in such a way that a person might forget that p but still dispositionally believe or know that p. For example, I might say, ‘I knew that we were meeting today at noon, but I forgot that when I left the house.’ After being told his new acquaintance's name for the ninth time, Fred exclaims, ‘No, I knew that your name's Norma! I just forgot for a second, and I needed a moment to remember it.’ Goldman's use of ‘forgotten’ may not properly leave in the mind of the reader the impression that Ursula's relevant beliefs and memories are completely wiped out.
Second, notice just how many memories and dispositional beliefs must be erased. It cannot just be Ursula's belief that in the example I just read, an instance of the antecedent of G is true and an instance of the consequent seems false. Other beliefs such as I read about a counterexample to G in that book and Smith and I talked about a counterexample to G the next day and even beliefs such as that I have at some time read about some putative counterexamples to G might also be evidence against G. These many beliefs might be deeply stored in the mind, and they too must disappear.
Third, if Ursula is an ordinary human being, memories of the encounter with the counterexamples will linger for a very long time, and the many relevant memories of the events surrounding the encounter will last even longer. So, Goldman's use of the word ‘forgotten’, the sheer number of beliefs and memories that must be erased, and the natural assumption that Ursula is an ordinary human being may detrimentally affect the reliability of our intuitions about his example.
We must stipulate, therefore, that all of the potentially relevant internal states have been completely erased from Ursula's mind so that she is internally identical to Ursula* at t 20. (Since ordinary humans do not lose memories this easily, we can even stipulate that a demon wipes out all the relevant memories.) From her internal point of view at t 20, it is as if the event never happened. At that moment, she has her initial reasons for the belief, and all of the potentially defeating dispositional memories and beliefs are gone. But when the case is specified this way, it seems to me that, at t 20, she is epistemically justified in believing G.Footnote 26 Hence, it seems that the justificational status of Ursula and Ursula*'s beliefs are the same at t 20.Footnote 27 They both justifiedly believe G, and Ursula's past encounter with the counterexamples seems justificationally irrelevant.
We can generalize. For any of some subject S's present beliefs that we are inclined to think is justified (or unjustified) at t at least partly in virtue of some specific past internal state, P, we can imagine a counterpart S* who is identical to S at t except for any differences entailed by the fact that S* was never in P because of a demon. Given that S and S* are completely internally identical at t, my intuition about any such case will likely be that the justificational status of S and S*'s beliefs is the same. We can now see that the local NEDP for historical internalism is also worth taking seriously.
5. Third Form of Internalism: Strong Internalism
Let us return to Cases 1 and 2. When I consider the global NEDP for historical internalism, I am inclined to think that unaccessed, internal properties make the justificational difference between Fred and Sally. When I consider the local NEDP for moderate internalism, I am inclined to think that past, accessed properties make the justificational difference between them. But both NEDPs combined seem to rule out any options for what could make the difference. If neither past accessed properties nor unaccessed, present properties are directly relevant, then how do we explain the justificational difference between Fred and Sally's beliefs?
Here is one answer: we don't. We accept that the justificational status of their beliefs is the same, and we take the lesson of the NEDP to be that only presently accessed, internal properties are directly relevant to justification. Richard Feldman (Reference Feldman, Conee and Feldman2004) has defended such a view. Briefly, Feldman first defended with his colleague Earl Conee (in Feldman and Conee Reference Feldman and Conee1985) the view that justification is solely a matter of the evidence one has. And in Feldman Reference Feldman, Conee and Feldman(2004: 219), he defended ‘a restrictive account that limits the evidence a person has at a time to the things the person is thinking of or aware of at that time’. In other words, he defended strong internalism, the view that justificational properties supervene on presently accessed internal properties.Footnote 28 I will not explore Feldman's argument for strong internalism here; I only point out that the view has been independently defended, and it is an option that an internalist could take.
The strong internalist will probably deny that Sally's belief is unjustified. Fred and Sally both undergo the same perceptual experience; it also seems or appears to both of them that that's a beautiful sunset. Feldman thinks that experiences such as ‘feelings of certainty’ count as evidence for a belief; Fred and Sally would both share these experiences (2004: 239). Given that Sally has no accessed reason to doubt her belief, the strong internalist will say that it seems that she is justified in believing that that's a beautiful sunset.Footnote 29 Furthermore, if someone were to ask her how she knows it is a sunset and not a sunrise, she might then form the occurrent belief that I don't know whether what I am seeing is a sunset or a sunrise. When she has this new accessed state, her belief that that is a beautiful sunset might then become unjustified, but not until then.Footnote 30 I do not find this explanation to be implausible. However, many will still consider this result to be a deal breaker. Some might think that if anything is true about justification, it is that Fred and Sally's beliefs are justificationally different. They will have a difficult time accepting strong internalism. I am not sure how to adjudicate this conflict.
I will end this section by defending my claim that strong internalism is radical and revisionary. First, consider that most epistemologists are foundationalists, coherentists or some mixture of the two. Coherentists will require that a belief be a member of a coherent set of beliefs in order to be justified; it is a very rare coherentist who restricts this set to presently occurrent beliefs. Foundationalists will require that a belief be based on other justified beliefs in order to be inferentially justified; it is a very rare foundationalist who restricts these justified beliefs to only presently occurrent ones. Someone who accepts a mixture of foundationalism and coherentism is also very unlikely to restrict the set of potentially justifying or defeating beliefs to only presently occurrent ones. We see that acceptance of strong internalism would require major revision in how many think about these traditional epistemological views, and the example with Fred and Sally only reveals one part of such a revision in our ways of thinking.Footnote 31
Second, nobody in the literature explicitly endorses strong internalism, including Feldman, the one person who has defended it. In their latest work, Conee and Feldman Reference Conee, Feldman and Dougherty(2011: 466–9) appeal to dispositional states in order to respond to one of Goldman's (Reference Goldman and Dougherty2011) arguments. This commits Feldman to a rejection of strong internalism. Some may think that the phenomenal conservativist or dogmatist views of Michael Huemer (Reference Huemer2001) and James Pryor (Reference Pryor2000) are strong internalist views. Both views attribute prima facie justification to anybody with the appropriate seeming or appearance, which are accessed internal states. Doesn't that make them endorsers of strong internalism? To answer this, note that these views are only stating a sufficient condition for prima facie justification, i.e. justification absent defeaters. Neither Huemer nor Pryor have ever said that nonoccurrent beliefs cannot serve as defeaters. They might or they might not hold to this. Hence, their views are only compatible with strong internalism, but nothing they say commits them to it. So, as of yet, there is nobody on record who endorses strong internalism.
I am not attempting to argue here that strong internalism is false because it is very radical and revisionary. The main point of this section is that, although strong internalism is immune to the NEDP, many epistemologists would have to significantly revise their views in order to accept it. Perhaps, however, the arguments of this paper will move more philosophers toward accepting it. I myself am undecided on what I think is the truth regarding the matter.
6. Conclusion
I have shown that the sorts of properties that many internalists would like to appeal to – unaccessed, internal properties and past internal properties – are sufficiently external that a demon could delete (or could have deleted) them without the believer's conscious awareness. These internalists thereby face the same (or a very similar) NEDP that they have been leveling against externalists. As a result, we might be tempted to deny all of these views: externalism, moderate internalism and historical internalism. But this leaves us only with strong internalism, which will be hard for many to swallow. The result is a dilemma: either accept strong internalism or face the force of the NEDP.Footnote 32
