A DILEMMA FOR INTERNALISM*

Michael Bergmann

[forthcoming in Knowledge and Reality: Essays in Honor of Alvin Plantinga eds. Thomas

Crisp, Matthew Davis, and David Vanderlaan (Kluwer Academic Publishers)]

 

Some objections to internalism in epistemology target only certain versions of it.  For example, some objections focus on versions of internalism that are wedded to a deontological conception of justification; others focus on versions of internalism according to which a necessary condition of justified belief is a further justified belief (or the potential for a further justified belief).[1]  But there are internalists who are very careful to sidestep these objections by defending positions to which they don’t apply.[2]  In this paper I develop an objection that applies to all species of internalism.

     As the title of this paper suggests, the objection is in the form of a dilemma:

(I). An essential feature of internalism is that it makes a subject’s actual or potential awareness of something a necessary condition for the justification of any belief held by that subject.

(II). This required awareness is either conceptual awareness (of a particular kind to be described later) or it is not. 

(III). If it is conceptual awareness (of the relevant kind), then internalism falls victim to regress problems.

(IV). If it is not, then internalism is subject to a prominent objection to externalism.

(V). If internalism is subject either to the regress problems mentioned in (III) or to the prominent objection to externalism mentioned in (IV), then we should not endorse internalism.

(VI). Therefore, we should not endorse internalism.

 

A familiar feature of this argument is its division of all awarenesses into two types and its conclusion that, for either type, some problem arises for an important epistemological thesis.  Wilfrid Sellars’ objection to what he called ‘the myth of the given’ employs a dilemma like this.[3] So does Laurence BonJour’s objection to internalist foundationalism (1985, ch. 4) and Paul Moser’s objection to internalist coherentism (1989, 173-76).[4]  But what hasn’t been widely observed is that an argument employing this sort of dilemma can be marshaled against a broader target, namely, internalism itself.[5]  The fact that it can be is striking.  It suggests that for many years, lurking in the writings of internalists (such as BonJour and Moser) have been the seeds of internalism’s own demise.

     The paper will proceed as follows.  I will defend premise (I) in section 1, premises (III) and (IV) in section 2, and premise (V) in section 3.  (Premise (II) needs no defense.)  I believe that this argument, along with the defense I’ll give of its premises, provides a formidable objection to internalism.  However, it’s been my experience that, upon hearing this argument, internalists tend to think that problems with it will become evident if we focus on specific beliefs (to see how the dilemma I pose is supposed to arise in a fleshed out example) or if we focus on specific attempts by internalist epistemologists to avoid this sort of dilemma.  For this reason, I will, in section 4, focus on two concrete examples of beliefs that can be used as test cases for my argument.  In addition to showing how these examples lend support to my conclusions, I will also consider whether the internalist positions defended by Paul Moser, Richard Fumerton, and Laurence BonJour can handle these examples in a way that enables them to resist my argument.  The reason I focus on these three philosophers is that they are among the most able defenders of internalism and they are particularly sensitive to the troublesome issues that lead to the sort of dilemma I propose in this paper.  Their failure to avoid being impaled on the horns of this dilemma provides further testimony to the strength of this objection to internalism.

 

1. Requiring Awareness

 

According to premise (I), an essential ingredient of internalism is the requirement (for justification) that there be some sort of actual or potential awareness of something on the part of the subject.  Two claims implicit in this premise are: (i) that, according to internalism, actual or potential awareness of something is required and (ii) that, according to internalism, it is such awareness on the part of the subject that is required.  In this section, I’ll defend these claims.  In order to do so, I’ll need to explain an alternative position they contradict.

     The alternative position I have in mind is one that is suggested by statements that are fairly common in the epistemological literature.  These common statements are probably not intended to stand up to rigorous criticism; they are more like rough and ready ways of characterizing certain aspects of the internalism/externalism controversy.  However, if we carelessly take them to be the truth strictly speaking, they lead to trouble.  In what follows, I’ll identify these common statements as well as the alternative position (on what internalism is) that they suggest.  But I want to emphasize that I am not so much criticizing anyone for endorsing them as highlighting a confusion we need to be wary of.

     The first common statement is that internalism has something to do with the conditions of justification being internal.  Another is that a condition is internal only if its satisfaction is in some fairly direct way (e.g., on reflection alone) cognitively accessible.  But accessible to whom? Clearly it is possible for a condition’s satisfaction to be accessible to one cognizer and inaccessible to another. It doesn’t make sense to say of a condition that it is internal simpliciter.  So how should we understand the fairly common practice of saying, without any explicit relativization, that certain conditions of justification are internal?  My recommendation is that we take such sayings as involving an implicit relativization.  Claims in the epistemological literature of the form ‘C is an internal condition’ should be understood as saying that C is internal to normal adult humans.  This seems to be the sort of thing that those who speak of internal conditions (without mentioning any relativization) have in mind.

     Now, returning to the first statement, what exactly does internalism have to do with the conditions of justification being internal?   Well, some externalists say that a reliability condition (e.g., one satisfied by a belief just in case it is produced by a reliable belief-forming process) is sufficient for justification.  Internalists complain that this condition is not an internal condition – its satisfaction isn’t cognitively accessible on reflection alone to normal humans  – and so it isn’t sufficient for the justification of our beliefs.  This sort of interchange suggests that the disagreement between internalists and externalists has to do with whether the conditions necessary for justification are internal conditions: internalists say the conditions necessary for justification are internal; externalists say they aren’t.  This way of understanding the internalism/externalism controversy is the alternative position that I said was suggested by some common statements in the epistemological literature.

     One way of expressing this position (or at least a part of it) is to say that a sufficient condition for being an internalist is endorsement of:

I1: At least one of the necessary conditions of justification is internal (i.e., internal to normal adult humans).

 

A little reflection shows that this position (i.e., that endorsement of I1 is sufficient for being an internalist) is contradicted by claims (i) and (ii) implicit in premise (I) of my dilemma.  Thus, by explaining what is wrong with this position I will be able to defend premise (I).

      Suppose that endorsement of I1 were sufficient for being an internalist.  Then an internalist could consistently endorse the following: 

Julie believes that p.  That belief satisfies each of the conditions that are together necessary and sufficient for justification.  Furthermore, all of these conditions are internal conditions insofar as they are internal to normal adult humans.  But Julie isn’t normal.  As a result of her abnormality, none of the necessary conditions of justification is internal to her.  In fact, Julie’s belief that p is justified even though she isn’t aware (or potentially aware) of anything that contributes to the justification of her belief that p.  What’s important for justification is that her belief satisfies each of the necessary conditions of justification – not that any of these conditions or anything contributing to the justification of her belief is internal to her (though it’s true that each of these conditions happens to be internal to normal adult humans).  

 

But no internalist would say this.  No internalist would allow that Julie’s belief is justified despite the fact that Julie isn’t aware of anything that contributes to the justification of her belief. 

     In response to the Julie example, one repair of I1 that naturally comes to mind is:

I2: At least one of the necessary conditions of justification is internal to the subject (i.e., to the person holding the belief whose justification is at issue).

 

Unfortunately, endorsement of I2 is also insufficient for being an internalist.  For a supporter of I2 could accept the following:

Tanner believes that p.  That belief satisfies each of the necessary conditions of justification.  Furthermore, each of these conditions is (contingently) internal to Tanner.  However, there is a possible world in which Tanner believes p and his belief that p satisfies each of the necessary conditions of justification although none of those conditions is internal to him.  In that world, Tanner’s belief that p is justified even though he isn’t aware (or potentially aware) of anything that contributes to the justification of his belief that p.  What matters for justification is that the belief in question satisfies each of the conditions necessary for justification – not that any of those conditions (or anything contributing to the belief’s justification) is internal to the person holding the belief.  It just happens to be the case that, in the actual world, each of the necessary conditions of justification is internal to Tanner.

 

But no internalist would say this either.  An internalist would insist that it couldn’t be the case that Tanner’s belief is justified when he isn’t aware (or potentially aware) of anything contributing to the justification of his belief.

      The Julie case supports claim (ii) from premise (I) (i.e., that it is the subject’s awareness that is at issue) and the Tanner case supports claim (i) (i.e., that this awareness on the part of the subject is required).  Together, these two cases make it clear that it isn’t sufficient for being an internalist to say that some necessary condition of justification is internal to normal humans. What’s required for being an internalist is to say that some necessary condition of justification or something contributing to the belief’s justification must be internal to the subject.[6]  Must be for what?  The obvious answer is that the condition or justification contributor must be internal if the belief in question is to be justified.  So if we construe cognitive accessibility in terms of being actually or potentially aware we can conclude that a necessary (though perhaps not a sufficient) condition for being an internalist is endorsement of:

I3: S’s belief B is justified only if (i) there is something, X, that contributes to the justification of B – e.g., evidence for B or a truth-indicator for B or the satisfaction of some necessary condition of B’s justification – and (ii) S is aware (or potentially aware) of X.

 

And if a necessary condition of being an internalist is endorsement of I3, then premise (I) of my objection to internalism is true.[7]

     Before moving on, I’d like to note briefly one implication of premise (I), namely, that some defenses of internalism use the term ‘internalism’ in an inappropriate way.  For example, Feldman and Conee claim (2001, 2) that it is sufficient for being an internalist that one endorses what they call ‘mentalism’, the view that justification is determined by the subject’s mental states whether or not the subject is aware (or potentially aware) of those mental states.  John Pollock (1986, 133-34) makes a similar claim when he says that a belief’s justification is a function of those states of the believer that are accessible to her automatic processors, whether or not those states are epistemically accessible (or potentially epistemically accessible) to the believer.[8]  If this were the right way to think about internalism, then someone endorsing the following would correctly be identified as an internalist:

The justification of our beliefs is determined by those of our mental states that are of kind K.  It is highly uncommon for a person to be aware (or even potentially aware) of mental states of kind K.  But it isn’t at all uncommon for a person to have mental states of kind K that justify her beliefs.  As a result, most of our justified beliefs are justified in virtue of our being in mental states we aren’t aware of (or even potentially aware of).  Thus, most of our justified beliefs are justified despite the fact that we aren’t aware of anything at all contributing to their justification.[9]

 

But no one who held such a view is plausibly construed as an internalist.[10]  So the fact that premise (I) conflicts with these accounts of internalism counts, if anything, in favor of premise (I) rather than against it.

 

2. Three Kinds of Awareness

 

Having defended premise (I), I now turn to premises (III) and (IV) – the two horns of my dilemma – since, as I noted earlier, premise (II) needs no defense.  In laying out these three premises, I alluded to a particular kind of conceptual awareness without saying what kind it is.  In order to say what kind it is, I’ll need to first say something about the distinction between conceptual and nonconceptual awareness. 

     Conceptual awareness of X is awareness of X that involves the application of a concept to X.  Nonconceptual awareness, by contrast, doesn’t involve the application of any concepts.  Cows and dogs presumably experience pain of some sort.  And presumably these animals are aware of such experiences.  Yet although they are aware of these experiences, it seems likely that they do not apply any concepts to them.  Humans too can be nonconceptually aware of experiences they undergo.  The difference is that we are also able to be conceptually aware of those experiences (by applying concepts to them) whereas dogs and cows presumably aren’t able to be conceptually aware of their experiences.[11]

     In light of this understanding of conceptual awareness, we can distinguish two species of it corresponding to two kinds of conceptual awareness requirements for a belief’s justification.  One kind of conceptual awareness requirement is satisfied by S’s belief B only if S conceives of the relevant object of awareness as contributing to B’s justification (or as indicating B’s truth or as being relevant in some way to the appropriateness of holding B).  I will call this sort of requirement a ‘conceptual1 awareness requirement’ and the awareness it involves ‘conceptual1 awareness’.  This is the sort of conceptual awareness that is the focus of premise (III).  All other conceptual awareness is conceptual2 awareness.  It too involves the application of a concept but it doesn’t involve application of the sort of concept associated with conceptual1 awareness.  For the purposes of my argument, therefore, there are three kinds of awareness: conceptual1 awareness, conceptual2 awareness and nonconceptual awareness.  I will discuss each in turn and, in doing so, defend premises (III) and (IV).

 

2.1 Conceptual1 Awareness

 

Suppose the awareness mentioned in I3 is actual conceptual1 awareness.  One way to guarantee this is to interpret I3 as follows:

I4: S’s belief B is justified only if (i) there is something, X, that contributes to the justification of B and (ii) S is aware of X in such a way that S justifiedly believes that X is in some way relevant to the appropriateness of holding B.[12]

 

Now consider the following familiar problem that arises in connection with I4.[13]  In order for S’s belief B to be justified, I4 says that S must have the further justified belief (with respect to something, X1, that contributes to the justification of S’s belief B) that:

P1: X1 is in some way relevant to the appropriateness of holding B.

And according to I4, in order for her belief that P1 to be justified S must have the further justified belief (with respect to something, X2, that contributes to the justification of S’s belief that P1) that:

P2: X2 is in some way relevant to the appropriateness of believing that éX1 is in some way relevant to the appropriateness of holding Bù.

 

And in order for her belief that P2 to be justified, S must have the further justified belief (with respect to something, X3, that contributes to the justification of S’s belief that P2) that:

P3: X3 is in some way relevant to the appropriateness of believing that éX2 is in some way relevant to the appropriateness of believing that éX1 is in some way relevant to the appropriateness of holding Bù ù.

 

And so on. On this actual conceptual1 awareness construal of I3, therefore, one has a justified belief only if one actually has an infinite number of justified beliefs of ever-increasing complexity.  But most of us find it exceedingly difficult even to grasp a proposition like P5 or P6 in such a series, much less believe it with justification.  Consequently, it’s very difficult to see how a supporter of I4 could resist the conclusion that none of our beliefs is justified.  The very ease with which this skeptical conclusion follows from I4 gives us a reason to reject it.[14]

     Now, as I said, the above sort of regress problem – like I4, the version of conceptual1 awareness internalism it afflicts – is familiar.  But perhaps other, less familiar forms of conceptual1 awareness internalism can escape this sort of difficulty.  Consider the suggestion that what internalists require for justification is merely the potential for conceptual1 awareness of the relevant fact (rather than actual awareness).  A person has potential conceptual1 awareness of a fact if she could have a justified belief that the fact obtained.  What exactly is involved in this ‘could have’?  The suggestion certainly isn’t that it is merely logically or metaphysically possible that the subject has a justified belief that the fact obtains.  Rather, the modality in question has to do with the subject’s abilities.  Thus, ‘S could do A’ should be understood as something like ‘on reflection alone S is able to do A’.  This gives us:

I5: S’s belief B is justified only if (i) there is something, X, that contributes to the justification of B and (ii) S is able on reflection alone to be aware of X in such a way that S justifiedly believes that X is in some way relevant to the appropriateness of holding B.

 

     Although I5 manages to avoid requiring for justification the actual possession of an infinite number of increasingly complicated beliefs, it still leads to trouble.  For in order to have the justified belief B, S must be able on reflection alone to justifiedly believe that P1. And to justifiedly believe that P1, S must be able on reflection alone to justifiedly believe that P2.  Thus, to justifiedly hold B, S must be able on reflection alone to be able on reflection alone to justifiedly believe that P2.  Given the plausible assumption that being able on reflection alone to be able on reflection alone reduces to being able on reflection alone, we may conclude that for every Pn in the series, S is justified in her belief B only if she is able on reflection alone to justifiedly believe that Pn.  But, as was noted above, one needn’t go very far in the series before one comes to a proposition that no human is able to grasp let alone justifiedly believe.  So although the potential awareness option avoids requiring the actual possession of an infinite number of justified beliefs, it is stuck with requiring the ability to justifiedly hold beliefs of ever-increasing complexity.  Like I4, therefore, I5 too has skeptical implications that give us a reason to reject it.

     Notice that there are two sorts of regresses associated with the above doxastic versions of the conceptual1 awareness requirement.  First, there is what we might call ‘the mental state regress’.  In the case of both actual belief and potential belief understandings of the conceptual1 awareness requirement, what is needed is either an infinite number of beliefs or the potential for that many beliefs.  But in addition to the mental state regress (either actual or potential), there is also a complexity regress.  And it is the latter that I’ve relied on in my objections to the above doxastic versions of the conceptual1 awareness requirement.  It is because the doxastic versions of the conceptual1 awareness requirement imply that a belief is justified only if one is able to hold justified beliefs of ever-increasing complexity that they are so implausible.

     One could try to save the internalist from these problems by dropping one or more of the assumptions I’ve made along the way. For example, one might suggest that mere belief, not justified belief, is required for conceptual1 awareness.  But if the internalist has the intuition that merely having a justification contributor isn’t enough – that the subject must also have some sort of conceptual1 awareness of that contributor (which includes believing that it is a justification contributor) – it seems highly doubtful that the internalist will be impressed by the mere belief (no matter how unjustified or insane) that the thing of which she is aware is a justification contributor. 

     One might also suggest that conceptual1 awareness could involve conceptualization without involving belief.  Paul Moser (1989, 186-87) distinguishes doxastic or propositional awareness of X (which involves predicating something of X) from conceptual awareness of X (which involves categorizing X according to some classificatory scheme). Perhaps I3 should be understood so that it requires merely that the subject be in some sort of cognitive contact with the justification contributor and that she categorize it in the appropriate way.  Then there is no requirement that she actually form any judgment.  In other words, perhaps the awareness required is conceptual1 but not doxastic.[15] 

     But concept application can be correct or incorrect; and it can be justified or unjustified in much the same way that believing can be justified or unjustified.  So even at the level of concepts, making conceptual1 awareness a requirement for justification gives rise to regress problems (involving ever-increasing complexity and, therefore, skepticism).[16]  To see this, consider the following nondoxastic conceptual1 awareness version of I3, which applies to justification for both concept application and belief:

I6: S’s belief or concept application, Y, is justified only if (i) there is something, X, that contributes to the justification of Y and (ii) S is aware of X in such a way that S justifiedly applies to X the concept of being in some way relevant to the appropriateness of Y.[17]

 

According to I6, S’s belief B is justified only if

A1: S’s application to X1 (a contributor to the justification of B) of the concept being in some way relevant to the appropriateness of B

 

occurs and is justified.  And according to I6, A1 is justified only if

A2: S’s application to X2 (a contributor to the justification of A1) of the concept being in some way relevant to the appropriateness of A1

 

occurs and is justified.  Likewise, I6 says that A2 is justified only if

A3: S’s application to X3 (a contributor to the justification of A2) of the concept being in some way relevant to the appropriateness of A2

 

occurs and is justified.  And so on.

     Now consider the concept that is applied in A3.   Spelled out more fully it is:

being in some way relevant to the appropriateness of S’s application to X2 of the concept ébeing in some way relevant to the appropriateness of A1ù

 

which, spelled out even more fully, is:

being in some way relevant to the appropriateness of S’s application to X2 of the concept ébeing in some way relevant to the appropriateness of S’s application to X1 of the concept ébeing in some way relevant to the appropriateness of Bù ù.

 

Not an easily digestible concept.  And of course it is only the tip of the iceberg when it comes to the complexity we are facing with this regress.  Thus, according to I6, justification for a belief B doesn’t depend on the subject having an infinite number of beliefs in propositions of ever-increasing complexity.  But it does depend on the application of an infinite number of concepts of ever-increasing complexity.[18]  And, for reasons similar to those discussed above in connection with the potential belief version of the conceptual1 awareness requirement, it won’t help to require merely the potential for applying an infinite number of concepts of ever-increasing complexity.  That way too lies skepticism.

     Thus, the familiar sort of regress problem that is associated with requiring actual awareness in the form of a further justified belief can’t be avoided if one insists that the awareness required for justified belief is conceptual1 awareness.  It won’t help to require only potential conceptual1 awareness.  Nor will it help to require only nondoxastic conceptual1 awareness. The problem is that the conceptual1 awareness that is being required for the justification of a concept application or a belief involves the application A of a concept whose content is at least slightly more complex than the content of the concept application or belief for whose justification A is required.[19] Thus, since the internalist is likely to deny that unjustified concept application can play a role in the justification of our beliefs (or concept applications) and since she requires conceptual1 awareness, she is forced into a skepticism-inducing complexity regress.  We may conclude, therefore, that requiring conceptual1 awareness – whether potential or actual, whether doxastic or nondoxastic – leads to regress problems.  This establishes premise (III) of my argument (since conceptual1 awareness is the sort of conceptual awareness that (III) is speaking of).

 

2.2 Nonconceptual awareness

 

Some internalists are familiar with the regress problems associated with requiring conceptual1 awareness.  Moser and Fumerton, for example, are internalists who recognize these problems and, because of this recognition, opt for the second horn of my dilemma by saying that the awareness required for justification is nonconceptual.[20]  In taking this route, such internalists face an objection from anti-foundationalists who say that what is nonconceptual – and therefore not capable of being justified – cannot confer justification on a belief based on it (cf. BonJour 1985, ch. 4).  The nonconceptual awareness internalist’s response to this objection is just to deny an assumption on which the anti-foundationalist objection relies – the assumption that only what is conceptual can confer justification (cf. Moser 1989, 193-94; Fumerton 1995, 74-75).

     Externalist foundationalists will have no complaint with this response.  For they agree that justification can be conferred by what is nonconceptual.  But an externalist foundationalist can point out that in endorsing that response, the internalist opens herself up to a prominent objection to externalism.  The objection I have in mind goes something like this:

The “Subject’s Perspective” Objection to Externalism: The externalist proposes an analysis according to which S1’s belief that p is justified so long as there exists something X contributing to that belief’s justification (where X is something like a reliable belief-forming process leading to the formation of the belief in question).  The idea is that S1’s belief is justified in virtue of X’s existence even if S1 isn’t aware of X – even if S1 doesn’t conceive of X as something relevant to the appropriateness of her belief that p.  Now I’ll grant [says the proponent of this objection] that if someone else, say S2, were aware of X and conceived of it as a contributor to the justification of S1’s belief that p, then S2 would have a reason for thinking that p is true.  But this doesn’t at all suggest that S1’s belief that p is justified.  True, X is relevant in some way to the appropriateness of S1’s belief. But this could be so even if S1 didn’t conceive of X as being in any way relevant to the appropriateness of her belief that p.  In such a case, it would – from S1’s subjective perspective – be an accident that her belief is true.

 

This objection, or something very much like it, is widely endorsed by internalists (cf. BonJour 1985, 43; Fumerton 1995, 116; Lehrer 1990, 162; Moser 1985, 129).  Indeed, I would go so far as to say that a putative internalist’s commitment to internalism would be in doubt if she said that there was nothing to the objection.  My claim is that the internalist who rejects the first horn of the dilemma I’ve proposed (and imposes only a nonconceptual awareness requirement on justification) is subject to this very same objection to externalism.  This should make her very unsympathetic to that objection.  But that lack of sympathy puts her allegiance to internalism in question.  And it leaves her without much of a complaint against externalism.

     Obviously, I need to defend my claim that the internalist who opts for a nonconceptual awareness interpretation of I3 is subject to the above objection to externalism. Consider an externalist view according to which it is necessary for the justification of S’s belief B that B is produced by a reliable belief-forming process.  And suppose that S’s belief B is produced by a belief-forming process token of a relevant process type that is, in fact, reliable (call this process token ‘RP’). [21]  Now imagine that a proponent of this reliabilist view is presented with the Subject’s Perspective objection to externalism and is impressed by it.  She decides to add to her account of justification the requirement that S is nonconceptually aware of the reliable process token in question – in this case RP.  Will that pacify those who endorse the Subject’s Perspective objection to externalism?

     It shouldn’t.  For since the awareness required is nonconceptual, a person can have the required awareness of RP without conceiving of RP in any way – without categorizing it according to any classificatory scheme.  But then a person can be nonconceptually aware of RP without conceiving of RP as relevant at all to the appropriateness of her belief.  According to the Subject’s Perspective objection to externalism given above, if S does not conceive of RP as something relevant to the appropriateness of her belief then, even if RP is relevant to the appropriateness of her belief, it is an accident from S’s perspective that her belief is true.  Clearly this supposed problem is not solved by adding the requirement that S is nonconceptually aware of RP.

 

2.3 Conceptual2 Awareness

 

Would it help if we added instead the requirement that S has a conceptual2 awareness of RP?  No.  For S could satisfy this sort of requirement simply by being aware of RP and applying some concept or other to it.  And that means that S can have a conceptual2 awareness of RP without conceiving of RP as relevant in any way at all to the appropriateness of her belief B.  But then, according to the Subject’s Perspective objection to externalism, even if this added requirement were satisfied, it would still be an accident from S’s subjective perspective that B is true.  For although S applies a concept to RP, she doesn’t apply the right sort of concept to it.  She doesn’t apply a concept that involves her conceiving of RP as contributing in some way to B’s justification (or as indicating that B is likely to be true or some such thing).  The only way to guarantee that she does apply such a concept to RP is to have B satisfy a conceptual1 awareness requirement.  Thus, we are forced to concede that by imposing only a conceptual2 awareness requirement, the internalist is vulnerable to the Subject’s Perspective objection.

     I argued above (in section 2.2) that if the awareness required is nonconceptual, then internalism falls prey to the Subject’s Perspective objection to externalism.  As we can now see, however, we get the same result if the awareness required is conceptual2 awareness.  Since, by definition, nonconceptual awareness and conceptual2 awareness are the only kinds of awareness other than conceptual1 awareness, this establishes premise (IV) of the main argument of this paper – namely, that if the awareness required for justification is not conceptual1 awareness, then internalism is subject to a prominent objection to externalism.

     To sum up, we have seen that internalism is motivated largely by the intuitions captured in the Subject’s Perspective objection to externalism.  But those intuitions, if taken seriously, imply that no belief is justified unless the person holding it is able to justifiedly believe infinitely many other propositions of ever-increasing complexity (or justifiedly apply infinitely many concepts of ever-increasing complexity).  The only way to avoid this implication is to make the required awareness something other than conceptual1 awareness.  But in doing so, one violates the very intuitions that motivated internalism in the first place.

 

3. Against Internalism

 

I have now defended premises (I), (III) and (IV) of my objection to internalism.  Those three premises (along with the uncontroversial second premise) entail that internalism is subject either to regress problems or to a prominent objection to externalism.  In this section I will complete my argument by defending premise (V) according to which this disjunction of problems implies that we shouldn’t endorse internalism.

 

3.1 Regress Problems

 

Why not simply admit that justification requires conceptual1 awareness and insist that internalism is true despite the fact that it leads to the sort of regress problems I mentioned in section 2.1?  My claim was that, if justification requires conceptual1 awareness, internalism leads to regress problems that suggest that we don’t have any justified beliefs.  I took these implications to be a problem for internalism.  But some internalists are known for looking askance at the suggestion that skeptical implications are sufficient to discredit a view (cf. BonJour 1985, 12-13; Fumerton 1995, 42-43).  So why can’t the internalist just confess that her internalism leads to the noted regress problems and then use modus ponens where I use modus tollens, accepting the skeptical consequences rather than rejecting her internalism?

     It’s worth noting that internalists such as Fumerton (1995, 80-81) resist this sort of response even though they think skeptical implications shouldn’t automatically discredit a position. And I think there is a good reason for this resistance.  It is one thing to think that justification clearly requires something such as a good reason and then to acknowledge that most of our beliefs lack such a reason.  An internalist (open to skepticism) who did this would insist that if we have no good reason for any of our beliefs, so much the worse for our beliefs; she wouldn’t be inclined to think we don’t need a good reason for them after all.  But it’s another thing to think that, in order for any belief to be justified, the person holding it must have (or have the ability to form) an infinite number of beliefs of ever-increasing complexity.  For in that sort of case, the problem isn’t merely that the requirement has skeptical implications.  It’s also that the requirement just seems excessive, especially when stated so bluntly.  I think this is why the internalist, in response to my discussion of conceptual1 awareness, is not (and shouldn’t be) inclined to use modus ponens where I use modus tollens.[22]

     It may be helpful to think of these matters in terms of Chisholm’s distinction between particularism and methodism (cf. Chisholm 1982, ch. 5).  Particularism is the view that we can use, as a starting point in our attempt to identify the criteria for justification, our knowledge of which beliefs are justified and which aren’t.  Methodism is the view that we should rely on our knowledge of what the criteria for justification are to determine which beliefs are justified and which aren’t.  The objection to premise (V) that I’m considering in this subsection is a methodist objection.  It says that we can just tell that the internalist’s proposed criterion for justification (namely, that the subject must have a conceptual1 awareness of something contributing to the justification of her belief) is correct and that if that criterion leads to the conclusion that none of our beliefs are justified, we should just accept that skeptical conclusion.

     Now perhaps the methodist approach is, on occasion, a good one.  But there are times when the extreme absurdity of the consequences implied by the criterion we started with demands that we use modus tollens with the particularist rather than use modus ponens with the methodist.  And this is just such a time.  For the skepticism implied by the conceptual1 awareness requirement is global.  Every belief of every believer is unjustified.  This is because every believer (with a finite mind) lacks the ability to grasp an infinite number of propositions or concepts of ever-increasing complexity.  It turns out, therefore, that if the conceptual1 awareness requirement on justification is correct it is literally impossible for a finite mind to have justified beliefs.  Surely this implication is a good reason to reject the proposed criterion.  It is doubtful in excelsis that the sort of justification epistemologists are trying to analyze is necessarily unexemplifiable by the beliefs of finite cognitive subjects such as ourselves.

     Furthermore, if the internalist were to accept the consequence that none of our beliefs is justified, her position would be self-defeating in an important way.  Her extreme confidence in the correctness of the conceptual1 awareness requirement on justification has led her to the conclusion that that very confidence was misplaced – misplaced in the sense that it is attached to a belief she is not justified in holding.  But once she concedes to her particularist opponent that that confidence is misplaced, it is difficult to see why anyone should approve of her persistence in endorsing the conceptual1 awareness requirement.  We can make sense of this sort of persistence in the face of absurd consequences only if the persister has extreme confidence in the correctness of the requirement she endorses.  But it isn’t easy to be understanding when we know that the person with this extreme confidence realizes that that confidence is misplaced.

 

3.2 The Subject’s Perspective Objection to Externalism

 

Given the conclusions reached so far, the internalist who wants to resist my conclusion is forced to say something like this: “I agree that my position is vulnerable to the Subject’s Perspective objection to externalism.  But that doesn’t mean I shouldn’t endorse internalism.”   Is this an acceptable response?   No.  For, as I mentioned briefly in section 2.2, this sort of response leads to two kinds of trouble: it puts one’s internalism in question and, more importantly, it leaves one without a good reason to prefer internalism to externalism.  Let’s look at both of these problems in a little more detail.

     In response to the suggestion that one’s internalism is in doubt if one’s view is subject to the Subject’s Perspective objection to externalism, an internalist might say the following: “My view is that noninferential justification supervenes on some sort of direct nonconceptual awareness.  Clearly, that isn’t an externalist view.  So my internalism can’t be in doubt.”[23]  The problem with this response is its assumption that, in denying that a position is an internalist one, I’m committed to saying that it is an externalist one.  But that assumption is false.  Let me explain. 

     In section 1, I identified one necessary condition for being an internalist, namely, endorsement of I3.  Now consider:

I7: Each of the necessary conditions of justification is internal (i.e., internal to normal adult humans).

 

Endorsement of I7 is insufficient for being an internalist for the same reason that endorsement of I1 is insufficient for being an internalist – namely, endorsement of either or both is compatible with rejecting I3.[24]  But endorsement of I7 seems to count very strongly against one’s being an externalist.  So it looks like it is possible for one to be neither an externalist nor an internalist: one need only endorse I7 and reject I3. And this shows that the inference from “position X is not an externalist position” to “it is a mistake to think that position X isn’t an internalist one” is illegitimate. 

     One might try to resist my suggestion that a view can be neither internalist nor externalist.  After all, the view that all nonexternalist epistemologists are internalists has been a fairly standard one in the epistemological literature.[25]  But it is also a fairly standard assumption in the philosophical literature (or at least it follows from entrenched views on what counts as an internalist and what counts as an externalist) that endorsing I7 is sufficient for not being an externalist and that rejecting I3 is sufficient for not being an internalist.  So something has to give.  And, although we are speaking of a technical term of art, I still think that, over the past twenty-five years, epistemologists have come to have some sense of what is intended by the terms ‘internalism’ and ‘externalism’.  This sense lends stronger support to the two views I just mentioned – i.e., that endorsing I7 is sufficient for not being an externalist and that rejecting I3­ is sufficient for not being an internalist – than it does to the view that one is automatically an internalist if one is not an externalist.  I therefore stand by my claim that one’s commitment to internalism is at least questionable if one’s view is vulnerable to the Subject’s Perspective objection to externalism. 

     The other, more significant, problem I mentioned is that one is left without a motivation for preferring internalism to noninternalist views.  Conceptual1 awareness internalism is out of the question.  That leaves us with either externalism or some form of what I will call ‘nonconceptual1 awareness internalism’ where by ‘nonconceptual1 awareness’ I mean awareness that isn’t conceptual1 awareness – i.e., awareness that is either nonconceptual awareness or conceptual2 awareness. But why prefer nonconceptual1 awareness internalism to externalism? As I read the literature, there are three main reasons for preferring internalism (of some kind or other) to externalism.  The first is that externalism is vulnerable to the Subject’s Perspective objection.  But of course that reason won’t help the nonconceptual1 awareness internalist we are now considering.  For she quite sensibly rejects the Subject’s Perspective objection since it applies to her own position and since endorsing it commits one to the regress problems discussed in section 2.1. 

     The second main reason for preferring internalism to externalism is the view that epistemic justification should be understood deontologically. There is much to be said in connection with this supposed reason for internalism.  But for now, the only thing we need to mention is that if this view lends any support at all to internalism, it is only to a conceptual1 awareness version of internalism.  For it is typically subjective deontological justification – i.e., deontological justification thought of as epistemic blamelessness – that is used in support of internalism.[26]  But I can have nonconceptual1 awareness of whatever you please – good reasons for my belief, strong evidence for it, the reliable process by which it is formed or its satisfying some other necessary condition of justification – without conceiving of the object of such awareness as being even remotely relevant to the appropriateness of my belief.  This makes it hard to see how having some sort of nonconceptual1 awareness necessarily moves one in the direction of being more epistemically blameless.  However, we can see how having conceptual1 awareness seems to help in this regard.  Conceptual1 awareness affects our subjective perspective on the appropriateness of our beliefs and, consequently, appears to be relevant to determining the degree to which we are blameless.  So we can see how a deontological conception of justification might motivate a conceptual1 awareness requirement even though it doesn’t seem to provide any motivation for a nonconceptual1 awareness requirement.  I don't have the space to develop this point more fully here (it probably deserves a paper in itself), but the basic idea should be clear.

     The third of the three main reasons for preferring internalism to externalism is that externalism is, allegedly, subject to the following objection:

The “Philosophical Seriousness” Objection to Externalism: Externalists have not correctly analyzed philosophically interesting epistemic properties.  They have either changed the subject to focus on properties that haven’t traditionally been the focus of epistemological inquiry or they have failed to fully appreciate and understand the philosophical depth and implications of the properties that have been the focus of traditional epistemology.  That externalists have missed the boat in this way is most evident when one looks at their flippant attitudes toward skepticism.

 

I have devoted an entire paper (Bergmann 2000b) to explaining why this objection fails. The basic idea of that paper is that the Philosophical Seriousness objection fails because the sort of reasoning on which it relies is, if successful at all, too powerful.  For if the internalist is right in saying that the epistemic properties on which externalists focus are philosophically uninteresting, then no epistemic properties at all are philosophically interesting – in which case we’ve been given no reason to prefer internalism to externalism.

     Without these three reasons for preferring internalism to externalism, it is very difficult to see why someone would do so.[27]  One wants to ask, “What good does having nonconceptual1 awareness do?  How does it do more to contribute to justification than the satisfaction of external conditions does?”  Again, the answer can’t be that it prevents it from being an accident from the subject’s perspective that her belief is true (or likely to be true or justified). But then why think awareness of something is so important if it doesn’t have the result that the subject conceives of the thing of which she is aware as relevant in some way to the appropriateness of her belief?

     One answer is suggested by Paul Moser.  Requiring that the subject have nonconceptual awareness of a justification contributor enables one to avoid an objection to externalism that happens to sound very much like the Subject’s Perspective objection.  That similar sounding objection is this: externalism is problematic insofar as it says that a belief can be justified even though, relative to all the subject is nonconceptually aware of, the belief isn’t likely to be true. (Moser 1989, 76)  The reason Moser’s view avoids that objection is that another requirement he imposes on justification (in addition to his nonconceptual awareness requirement) is, roughly, that P, the content of the subject’s belief, must explain E (the object of the subject’s required nonconceptual awareness) better than any other competing proposition S is aware of.  (1989, 136-37)  If P is the best explanation of E and if S is nonconceptually aware of E, then it is not the case that relative to all the subject is nonconceptually aware of her belief that P isn’t likely to be true.

     But the fact that Moser’s view avoids this other objection to externalism shouldn’t appease those with internalist sympathies.  For, as Moser himself acknowledges (1989, 164), S’s belief can satisfy Moser’s requirements for justification even if S has no idea that P is the best explanation of E.  In fact, given that S needs only nonconceptual awareness of E, S needn’t even conceive of E as something that might be explained by P or as something that is relevant in some way to P.  So, in satisfying Moser’s requirements for justification, the most that S’s belief that P is guaranteed to have going for it is that some object of awareness to which S may have applied no concepts at all is, unbeknownst to S, best explained by P.  That certainly needn’t make any relevant epistemic difference from S’s own perspective to the credentials of her belief that P – no more than the fact (supposing it were a fact) that, unbeknownst to S, the process by which S’s belief that P is formed is a reliable one.  So we are left without any motivation to prefer nonconceptual1 awareness internalism to externalism.

     This completes my defense of premise (V), the claim that a person shouldn’t endorse internalism if her position is vulnerable to the Subject’s Perspective objection to externalism.  She shouldn’t because it puts her internalism in doubt and because it leaves her without a reason to prefer internalism to externalism.  “But,” one might ask, “what advantages does externalism have over a nonconceptual1 awareness version of internalism?”  The main advantage is this: the nonconceptual1 awareness version of internalism is committed to imposing an unmotivated requirement – i.e., that the subject have some sort of nonconceptual1 awareness – that externalism isn’t committed to imposing.

 

4. Two Examples and Three Internalists

 

I have now defended the premises of the anti-internalist argument given in the introduction.  As I noted earlier, a common reaction from the internalists to whom I’ve shown this argument is the suspicion that its weaknesses will become manifest either when we apply the dilemma it proposes to particular beliefs or when we consider versions of internalism designed specifically to avoid a Sellarsian type of dilemma.  Thus, in order to make the force of my argument more evident to those inclined to resist its conclusion, it will be helpful to focus on some specific examples (I’ll focus on a physical object belief and a first-person mental state belief) and to consider the implications of the dilemma I’ve proposed in connection with them.  While doing so it will also be useful to evaluate the responses of three well-known internalist philosophers – Paul Moser, Richard Fumerton and Laurence BonJour – to the challenge of avoiding regress problems while imposing a plausible awareness requirement on justification, since seeing how their responses fail will lend further support to my argument.

 

4.1 A Physical Object Belief and Moser

 

Suppose that Jack forms the belief that there is before him a large spherically shaped object (call this belief ‘B1’).  And suppose that he forms this belief as follows.  He walks into a well-lit room that is empty except for a large white ball.  As a result of fixing his eyes on the ball in the room, he has a visual experience that is of the same type (color-experience-wise) as the experience that a normal human would have in such circumstances (call this visual experience of a large white ball ‘WB’).  And Jack’s having WB causes the formation of his belief B1. To alleviate any concerns an externalist might have, let’s add that B1 is a reliably formed belief.  Now, is Jack’s belief justified?  In order to answer that question, there is another important question that will naturally come to the mind of an internalist: does Jack conceive of WB as relevant in any way to the appropriateness of B1?

     The internalist faces a dilemma when she considers this last question in connection with a case like Jack’s.  For, since she is an internalist, she wants to insist that it does Jack no epistemic good that B1 is reliably formed unless B1 satisfies some sort of awareness requirement.  But the awareness requirement must be either a conceptual1 awareness requirement or not.  If it is a conceptual1 awareness requirement, then Jack must be aware of some contributor to the justification of B1 and justifiedly apply to it a concept such as being relevant in some way to the appropriateness of B1.  But that way lies the complexity regress and skepticism.  So the internalist is forced instead to require that Jack’s belief satisfy some nonconceptual1 awareness requirement (requiring either nonconceptual awareness or conceptual2 awareness).

     Suppose the internalist decides to insist on merely nonconceptual awareness of a contributor to the justification of B1.  This seems to be the approach adopted by Paul Moser.[28]  According to Moser, S’s belief that P is justified only if (a) S is nonconceptually aware of (or presented with) evidence E[29] and (b) S has a de re nonconceptual awareness of E’s supporting P.[30] In the case of Jack, we can think of this as requiring that he is nonconceptually aware of WB (the visual experience of a large white ball) and that he has a de re nonconceptual awareness of WB’s supporting B1 (the belief that there is a large spherical object in front of him).  But clearly B1 could satisfy those two requirements even if, according to the Subject’s Perspective objection, it’s the case that from Jack’s subjective perspective it is an accident that B1 is true.  For since Jack is required only to be nonconceptually aware of WB, he needn’t conceive of WB as evidence for anything at all.  The same comment applies to Jack’s nonconceptual awareness of WB’s supporting the proposition that there is a large spherical object in front of him. It is a little difficult to make sense of such a nonconceptual awareness.  I have some idea of what a nonconceptual awareness of a pain would be like.  But I’m not sure I have much of an idea of what a nonconceptual awareness of WB’s supporting a proposition would be like.  However, we know this much: one can have such an awareness without conceiving of WB’s supporting a proposition as WB’s supporting a proposition.  But if Jack doesn’t conceive of WB as evidence for anything and doesn’t conceive of WB’s supporting the content of B1 as WB’s supporting that content, it is hard to see how Jack’s subjective perspective on B1 has improved in any way.

     Moser is sensitive to this sort of problem and tries to solve it by requiring something more than nonconceptual awareness without requiring conceptual awareness of any sort. The something more he requires is the satisfaction of the following requirement for the justification of S’s belief that P:

as a nondeviant result of this awareness [i.e., the de re nonconceptual awareness of E’s supporting P], S is in a dispositional state whereby if he were to focus his attention only on his evidence for P (while all else remained the same), he would focus his attention on E.  (1989, 141)

 

How shall we understand this requirement?  Because he wants to avoid regress problems, Moser makes it very clear (1989, 142) that he wants it to be interpreted in such a way that no conceptual awareness is required.  But what exactly is it to be disposed to focus one’s attention on E if one were to focus one’s attention only on one’s evidence for P?  On one reading, it is just to be disposed to focus on E if one were to focus on E.  But that can’t be what Moser is saying.  The most plausible interpretation I can think of is this: to be in the state described is to be disposed to focus on E if one were to focus on what one takes to be one’s evidence for P.  Or, to put it another way, it is to be disposed to focus on E if one were to focus on what one conceives of as being one’s evidence for P.

     But that is just to be disposed to apply the concept of being evidence for P to E (if one were to apply the concept to anything at all).  And surely, it wouldn’t be sufficient to be disposed to apply that concept in an unjustified way.  So it must be a disposition to apply that concept in a justified way.  But that will involve a disposition to satisfy further awareness requirements and to have further dispositions to justifiedly apply concepts.  There is here a regress of dispositional concept application of the sort Moser is eager to avoid.  Thus, requiring merely nonconceptual awareness, even of WB’s supporting the content of B1, doesn’t enable B1 to escape the disapprobation of the supporters of the Subject’s Perspective objection.  And there doesn’t seem to be any way of requiring more (enough to satisfy these objectors) without imposing some sort of conceptual1 awareness requirement.

      Conceptual2 awareness won’t help Jack either.  For Jack’s conceptual awareness of WB will do him no good (according to the supporters of the Subject’s Perspective objection) unless he applies the right sort of concept to it – the sort of concept that will make his conceptual awareness a conceptual1 awareness.  So long as Jack applies only concepts other than ones like being indicative of B1’s truth or being a contributor to B1’s justification or being in some way relevant to the appropriateness of holding B1, B1’s truth will, according to the Subject’s Perspective objection, be an accident from Jack’s perspective.

     In short, because it is possible for Jack to have a nonconceptual1 awareness of an experience like WB without applying (or being able to apply) concepts like being indicative of B1’s truth to it, it is clear that such awareness – even if it is conceptual – doesn’t enable B1 to avoid the censure of the supporters of the Subject’s Perspective objection.  Therefore, with respect to physical object beliefs like Jack’s, the internalist’s only choice is to impose a conceptual1 awareness requirement leading to a complexity regress (and skepticism) or to impose a nonconceptual1 awareness requirement (of either the nonconceptual or conceptual2 variety) which leaves one vulnerable to the Subject’s Perspective objection to externalism.

 

4.2 A First-Person Mental State Belief

 

As we have just seen, the problem with requiring merely nonconceptual1 awareness for the justification of physical object beliefs is this: one could be aware (even conceptually aware) of an experience like WB without conceiving of it as being relevant in some way to the appropriateness of holding B1.  The only way to avoid this problem is to impose a conceptual1 awareness requirement, which leads to a complexity regress and skepticism.  Does this same sort of problem arise in connection with beliefs about first-person mental states?  Yes.  But it isn’t easy to see that it does because it is difficult to think clearly about the type of example needed to demonstrate that it does.  The remainder of this paper will be devoted to a careful consideration of this very important (and difficult to think about) type of example and its implications for internalists.

     Suppose that Lucy is being appeared to redly and that she believes that she is being appeared to redly (call this belief ‘B2’).  And suppose that there is nothing of which Lucy is aware to which she applies a concept like being in some way relevant to the appropriateness of holding B2.  It’s true that she is aware of her being appeared to redly.  And perhaps she even conceives of that experience in some way (e.g., as a rather uninteresting experience).  But, due to her severe cognitive malfunction, she doesn’t (and isn’t able to) conceive of that experience as something relevant to the appropriateness of holding B2. 

     This may seem a little hard to swallow.  We can see how, in the case of the perceptual belief discussed earlier, Jack might be conceptually aware of WB, the visual experience of a large white ball, without applying to it the concept of being in some way relevant to the appropriateness of holding B1 (where B1 is his belief that there is a large spherically shaped object in front of him).  But is it even possible for Lucy to be conceptually aware of her being appeared to redly without applying to it the concept of being in some way relevant to the appropriateness of holding B2 (where B2 is her belief that she is being appeared to redly)?  I don’t see why not.  It is one thing to have a nonconceptual1 awareness of one’s being appeared to redly.  It’s another to apply to one’s being appeared to redly a concept like being in some way relevant to the appropriateness of holding B2 (or being indicative of the truth of B2 or contributing to the justification of B2).  Granted, the malfunction involved in having such an awareness without applying (or being able to apply) such a concept would be extreme.  But that doesn’t make it impossible.  Perhaps, in addition to suffering from some sort of cognitive defect, Lucy is also in the grip of a philosophical theory which strongly supports her disinclination to apply to her experience of being appeared to redly a concept like being in some way relevant to the appropriateness of holding B2.  The philosophical theory (which she accepts on the authority of a philosophically astute but malicious prankster) gets Lucy interested in denying what seems to us to be utterly obvious.  And the cognitive malfunction makes it possible for her to do what we are incapable of doing, namely, genuinely refraining from a concept application that is so obviously correct.[31]

     This example provides us with a case of someone who is appeared to redly, who believes that she is appeared to redly, and who is conceptually aware of her being appeared to redly even though, according to proponents of the Subject’s Perspective Objection, it is an accident from her perspective that her belief that she is being appeared to redly is true.  For although Lucy is aware of her being appeared to redly and although she applies a concept to the object of that awareness, she doesn’t (and isn’t able to) apply to it a concept like being in some way relevant to the appropriateness of holding B2.  It’s true that the malfunction involved in this example is extreme. But the point is that it is possible and the proponent of the Subject’s Perspective Objection is committed to requiring (for the justification of B2) that such malfunction not be actual.

     So even in the case of first-person mental state beliefs, the internalist must choose between the extreme skepticism induced by conceptual1 awareness requirements and the vulnerability to the Subject’s Perspective objection that results from imposing only nonconceptual1 awareness requirements.  Let’s briefly consider what Fumerton and BonJour would say about this example of Lucy’s first-person mental state belief.

 

4.2.1 Fumerton

 

Because of his concern to avoid regress problems, Fumerton tries to steer clear of conceptual awareness requirements of any kind.  He says (1995, 75) that a person’s belief that p is (noninferentially) justified only if she is directly acquainted with the fact that p, her thought that p, and the relation of correspondence holding between her thought that p and the fact that p. Concerning this direct acquaintance he says the following:

Acquaintance is not another intentional state to be construed as a nonrelational property of the mind.  Acquaintance is a sui generis relation that holds between a self and a thing, property or fact.  To be acquainted with a fact is not by itself to have any kind of propositional knowledge or justified belief ...  One can be acquainted with a property or fact without even possessing the conceptual resources to represent that fact in thought ... Acquaintance is a relation that other animals probably bear to properties and even facts ...  (1995, 74-75).

 

Clearly, he has in mind some sort of nonconceptual awareness.

     In giving the Lucy example, I’ve already stipulated that she is nonconceptually aware of the fact that p (i.e., the fact that she is being appeared to redly).  We can add that she is also nonconceptually aware of her thought that she is being appeared to redly.  Finally, we can add that she is nonconceptually aware of the relation of correspondence holding between her thought that she is being appeared to redly and the fact that she is being appeared to redly.  As I noted earlier, it is a little difficult to make sense of nonconceptual awareness of things like this relation holding (vs. things like being in pain).  But for our purposes, what’s important is that we know that this awareness or direct acquaintance is nonconceptual (i.e., it needn’t involve the application of any concepts). 

     Now what happens to Lucy’s subjective perspective on B2 when we add to our example that Fumerton’s conditions on justification are satisfied by B2?  Because the direct acquaintance required is nonconceptual, Lucy can be directly acquainted with the fact that she is being appeared to redly without conceiving of the object of this awareness as being in any way relevant to the justification or truth of B2 (this is because nonconceptual awareness is the sort of thing that can occur without the application of any concepts to the object of awareness).  Furthermore, she can be directly acquainted with the relation of correspondence holding between her thought that she is being appeared to redly and the fact that she is being appeared to redly even if she has no idea that the relation of correspondence holds between these two items (again, this is because nonconceptual awareness is the sort of thing that can occur without the application of any concepts). Thus, Lucy’s belief that B2 can satisfy Fumerton’s requirements even if she conceives of her being appeared to redly as no more relevant to B2 than is the mild pain in her left knee.  It is, therefore, exceedingly difficult to see how these direct acquaintances improve things from Lucy’s subjective perspective.  If things were epistemically bleak from her perspective before she had these nonconceptual acquaintances Fumerton requires, there is no reason to think they will be less bleak afterwards.  The only way for things to improve from Lucy’s perspective is for her to apply to the objects of these direct acquaintances a concept such as being relevant in some way to the appropriateness of B2.  But to require that is to impose a conceptual1 awareness requirement on justification, something Fumerton refuses to do, knowing full-well that it leads to regress problems.   I conclude, therefore, that Fumerton’s account of the justification of first-person mental state beliefs is vulnerable to the Subject’s Perspective objection.

 

4.2.2 BonJour

 

According to BonJour (2001, 30-31), S’s belief that she is being appeared to redly is justified if S is being appeared to redly.  For to have an experience of being appeared to redly is to have a conscious experience of being appeared to redly.  And to have a conscious experience of being appeared to redly is to have the experience and to be aware of its nonconceptual sensory content.[32]  Thus, if Lucy is being appeared to redly, then she is aware of her being appeared to redly.  Furthermore, according to BonJour, an awareness of one’s being appeared to redly is an awareness of an excellent reason for believing that one is being appeared to redly.  The nonconceptual sensory content of that experience is an excellent reason for such a belief because it matches the description that is included in the propositional content of that belief.

      Let’s grant to BonJour that to be appeared to redly entails that one is aware of being appeared to redly and that to be aware of one’s being appeared to redly is to be aware of an excellent reason for believing that one is being appeared to redly.  That isn’t enough for BonJour to avoid the Subject’s Perspective objection to externalism. The fact that Lucy’s awareness of the sensory content of her experience is an awareness of what in fact counts as an excellent reason for holding B2, doesn’t imply that Lucy conceives of the sensory content of her experience as providing any indication of B2’s truth.  The supporter of the Subject’s Perspective objection will insist that it is only if Lucy applies to her experience of being appeared to redly a concept like being indicative of the truth of B2 that B2’s truth isn’t an accident from Lucy’s perspective.  And in the example given above, Lucy doesn’t apply that sort of concept to her experience of being appeared to redly. 

     According to the proponent of the Subject’s Perspective objection to externalism, even if Lucy is in possession of a good reason for B2 she is, at best, in the position of Dr. Watson after he has received all the evidence required for proving that the butler did it.  Watson is in possession of a good reason for thinking the butler did it but, unlike Sherlock Holmes, he can’t see that it is a good reason.  Now suppose Watson forms the belief that the butler did it.  From Watson’s subjective perspective, the truth of his belief that the butler did it is, according to the Subject’s Perspective objection, an accident.[33]  In the same way, although Lucy is aware of something that BonJour thinks is an excellent reason for B2, in my example she doesn’t conceive of her being appeared to redly as being any reason at all for B2.  So from Lucy’s perspective it is (according to the Subject’s Perspective objection) an accident that B2 is true; it is an accident even though that belief satisfies the conditions BonJour says are sufficient for its justification.  The only way for BonJour to avoid this result is to require that Lucy’s awareness of her being appeared to redly be accompanied by an application to that experience of a concept like being indicative of B2’s truth.  But this is for BonJour to insist that justification requires conceptual1 awareness.  And that leads to the complexity regress and to skepticism.  I conclude, therefore, that because BonJour’s awareness requirement isn’t a conceptual1 awareness requirement his view too falls victim to the Subject’s Perspective objection to externalism.

 

5. Conclusion

 

Why don’t internalists recognize that by avoiding regress problems they make their positions vulnerable to the Subject’s Perspective objection?  I think it’s because they aren’t recognizing the possibility of the kind of severe malfunction that occurs in my Lucy example.  To see the connection between their failure to recognize this possibility and their failure to foresee the problems I identify, let’s consider again what Fumerton does.  He wants to avoid imposing conceptual1 awareness requirements on justification.  But he doesn’t want (I assume) to allow for the possibility that it is an accident from the subject’s perspective that her beliefs are true (or at least not if those beliefs are to count as justified).  In order to guarantee that his view isn’t thus vulnerable to the Subject’s Perspective objection (and to do so while avoiding regress problems), he says that noninferential justification depends on nonconceptual awareness (or direct acquaintance).  The reason this guarantee seems to work is that it is tempting to think that it is impossible for it to be an accident from a person’s perspective that her belief that, say, she is being appeared to redly is true if she is (nonconceptually) aware of her being appeared to redly.  But as the Lucy example, with the severe malfunction it involves, shows us, such a scenario isn’t impossible.  So the supposed guarantee doesn’t work.  To avoid the Subject’s Perspective objection, something more (i.e., a conceptual1 awareness requirement) is needed.  And the reason the failure of that supposed guarantee isn’t noticed is that, unless one thinks about the matter carefully, it can seem almost unimaginable for one’s introspective belief-forming practices to be as severely damaged as Lucy’s are.  In short, although internalists such as BonJour, Fumerton and Moser refrain from actually imposing conceptual1 awareness requirements, they seem to have relied on the (false) assumption that our direct acquaintance beliefs cannot fail to satisfy such requirements.

     Internalists have two options.  They can say that conceptual1 awareness is required for justification; or they can say that nonconceptual1 awareness is required (i.e., awareness that isn’t conceptual1 awareness).  If they require the former, they are forced into a complexity regress according to which no belief is justified unless the person holding it is able to grasp concepts that are more complex than the human mind can grasp.  But if they require nonconceptual1 awareness instead, their position is subject to one of the most prominent and influential internalist objections to externalism – in which case they should reject that objection and give up their resistance to externalism.  Either way, they should not endorse internalism.  And the fact that even such able defenders of internalism as Moser, Fumerton and BonJour (who are particularly sensitive to the issues on which I’ve focused) cannot escape my dilemma provides further confirmation of the strength of the objection I’ve proposed.

 

Appendix: Hetherington’s Dilemma

 

I’m grateful to Stephen Hetherington for bringing to my attention two papers of his (see his 1990 and his 1991) that I hadn’t noticed when this paper was first subjected to public scrutiny.[34]  In them he proposes a dilemma for internalism that in certain ways resembles the one I’ve presented here. I think it’s fair to say that his argument and mine are two different versions of the same kind of objection to internalism.  However, as I shall argue in this appendix, internalists have good reasons not to be troubled by the dilemma Hetherington proposes (though none of those reasons count against the dilemma I’ve proposed above).

 

Hetherington’s Dilemma Explained

 

The basic structure of the dilemma Hetherington proposes in the two papers mentioned above can be captured (in a way that is parallel to my own dilemma) as follows:

(I*) An essential feature of internalism is its endorsement of the Transparency Thesis (TT):

TT: If some [reason or justification-contributor] W is epistemically internal to S, then its being epistemically internal to S is also epistemically internal to S.[35]

(II*) The internalist either gives up her commitment to TT or she doesn’t.

(III*) If she doesn’t give up her commitment to TT, then the internalist falls prey to regress problems that imply that “epistemic internalism is an empty concept”.[36]

(IV*) If the internalist does give up her commitment to TT, then, given (I*), she gives up her internalism and becomes an externalist.

(V*) If an internalist must either admit that epistemic internalism is an empty concept or give up her internalism, then internalism should be rejected.

(VI*) Therefore, internalism should be rejected.[37]

 

That Hetherington is proposing this dilemma in his 1991 is obvious.[38]  It is also obvious that in his 1990, he is proposing a dilemma of this same form.  For in that paper, not only does he explicitly formulate his objection as a dilemma, he also clearly says that: (i) internalists are committed to saying things that lead to regress problems, (ii) these regress problems imply that epistemic internalism is an empty concept and (iii) the only way to avoid the regress in question is to give up on internalism.[39]  But what isn’t so obvious is whether his 1990 says that the principle to which internalists are committed and from which a damaging regress can be generated is TT.  In order to see that both papers are focused on TT, we will need to take a careful look at what each paper says about the regress that Hetherington thinks is generated by internalism.  Once I’ve established that the above dilemma is the one Hetherington is in fact advancing in both papers, we can then turn to an evaluation of it.

     In his 1990, Hetherington describes the regress as follows:

The regress … is this:

According to epistemic internalism about a condition A of your having a justified belief,

(1) A contributes to your justification

only if

(2) You appreciate that A contributes to your justification

only if

(3) You appreciate that your-appreciating-that-A-contributes-to-your-justification (i.e., the appreciating which is (2)) contributes to your justification

only if

(4) You appreciate that the appreciating which is (3) contributes to your justification

only if

(5) You appreciate that the appreciating which is (4) contributes to your justification

only if ... And so on.  (1990, 246)

 

It is difficult to make sense of the sentence that appears between the colon and the second ellipsis.  However, I take it that what Hetherington had in mind was something like the following:

According to epistemic internalism about a (necessary) condition A of your having a justified belief B,

(1) A contributes to your justification

only if

(2) You appreciate that A contributes to your justification.

This means that (2), like A, is a necessary condition of B’s justification.  But (2)’s truth contributes to your justification for B only if

(3) You appreciate that your-appreciating-that-A-contributes-to-your-justification­ (i.e., the appreciating which is (2)) contributes to your justification.

This means that (3), like (2), is a necessary condition of B’s justification.  But (3)’s truth contributes to your justification for B only if

(4) You appreciate that the appreciating which is (3) contributes to your justification.

This means that (4), like (3), is a necessary condition of B’s justification.  But (4)’s truth contributes to your justification for B only if ...

 

Now suppose Hetherington is right in saying that the internalist – at least one who endorses “epistemic internalism about a condition A of your having a justified belief” B – is committed to thinking that A contributes to your belief B’s justification only if you appreciate that it does.  Then it seems right to say that, according to internalists, (2) is also a necessary condition of B’s justification.  But why conclude from that that internalists must say that (2)’s truth contributes to your justification for B only if you appreciate that it does?  Why can’t internalists say instead that A contributes to B’s justification only if you appreciate that it does but (2)’s truth contributes even if you don’t appreciate that it does?[40] 

     In the following passage, Hetherington gives his answer to that question:

Your appreciating-that-A-is-contributing-to-your-being-justified [i.e., (2)] is contributing to your being justified; need you in turn appreciate that it is doing so?  What the epistemic internalist should say is that you must.  Otherwise, …your appreciating-that-A-is-contributing-to-your-being-justified [i.e., (2)] is itself epistemically external to your being justified.  Yet a necessary condition of A’s being epistemically internal to you was your appreciating that it is contributing to your being justified.  And if the latter aspect of your situation [i.e., your appreciating-that-A-is-contributing-to-your-being-justified] is epistemically external to your being justified, then no doubt the former [i.e., A] is too. (1990, 246)

 

As I understand this passage, he is saying that if you don’t appreciate that (2)’s truth contributes to B’s justification (i.e., if (2) is epistemically external to you), then you don’t appreciate that A is contributing to B’s justification (i.e., then A is epistemically external to you).[41]  But this entails that:

P: If A is appreciated by you as contributing to B’s justification, then your appreciating that A is contributing to B’s justification is also appreciated by you as contributing to B’s justification.

 

And P is basically identical to TT from premise (I*).  This becomes clear once we recognize that, according to Hetherington, for something to be epistemically internal to a subject S “is for it to be appreciated by S as being his or her reason” (1991, 858).  Given this account of what it is to be epistemically internal to S, we can restate TT as follows:

TT*: If W is appreciated by S as being her reason, then S’s appreciating that W is her reason is also appreciated by S as being her reason.

 

Now clearly P and TT* (which is equivalent to TT) are saying pretty much the same thing.  And as we just noted above, Hetherington appeals to P in order to generate the regress he tries to identify in his 1990.  So even if it isn’t obvious from a casual reading, we may conclude that, according to his 1990, the principle to which internalists are committed and from which the regress is generated is TT. 

     Let’s turn now to Hetherington’s 1991 to see how the regress identified there (which is explicitly derived from TT) compares with the regress identified in his 1990.  If the regresses are the same, this will lend further support to my claim that both papers are proposing the same dilemma. 

     In his 1991 Hetherington argues (860-61) that according to TT, a potential justification-contributor (or J-contributor), W1, won’t be epistemically internal to S unless each of the following is too:

R:            the appreciating-of-W1-as-epistemically-internal-to-S,

the-appreciating-of-that-first-appreciating-as-epistemically-internal-to-S

the-appreciating-of-that-second-appreciating-as-epistemically-internal-to-S, ...  (860)

 

But that isn’t quite right.  To see why, it’s important to remember that, as I noted above, TT is equivalent to TT*.   By keeping TT* in mind, we can see more clearly which regress Hetherington seems to have had in mind in his 1991.  For according to TT*, a potential J-contributor, W1, won’t be epistemically internal to S unless each of the following is also epistemically internal to S (i.e., appreciated by S as being her reason):

R*:          (2*) W1’s-being-appreciated-by-S-as-being-her-reason,

(3*) (2*)’s-being-appreciated-by-S-as-being-her-reason,

(4*) (3*)’s-being-appreciated-by-S-as-being-her-reason,

(5*) (4*)’s-being-appreciated-by-S-as-being-her-reason, ...[42]

 

Furthermore, Hetherington thinks that internalists are committed not only to TT (and, therefore, to TT*) but also to:

Q: At least one of a belief’s potential J-contributors must be epistemically internal if the belief is to be justified.[43]

 

And if the internalist is committed to both TT and Q she must say that each of the appreciations in a regress like R* must be epistemically internal to S if some belief of hers is to be justified. 

     Notice that if (2*) must be epistemically internal to S in order for some belief B of hers to be justified, it follows that (3*) is necessary for B’s justification (since (3*) just is (2*)’s being epistemically internal to S).  And if (3*) must be epistemically internal to S in order for B to be justified, it follows that (4*) is necessary for B’s justification (since (4*) just is (3*)’s being epistemically internal to S).  And so on.  But to say that (3*), (4*), (5*), etc. are necessary for B’s justification is very much like saying that (3), (4), (5), etc. from Hetherington’s 1990 regress are necessary for B’s justification.  In fact, the two series of propositions are so similar that I think it’s fair to say they constitute the very same regress.  Thus, it looks like the regress Hetherington was trying to identify in his 1991 is the same regress he was trying to identify in his 1990.  This confirms my suggestion that both papers are proposing the dilemma identified above.

 

Hetherington’s Dilemma Criticized

 

Now that we have before us the dilemma Hetherington proposes we can consider why it is that internalists needn’t be troubled by it.   One weakness of this dilemma is that Hetherington’s standard for who counts as an internalist is implausibly strict.  Consequently, many philosophers typically viewed as internalists (such as Fumerton, Moser, and BonJour in his more recent work) don’t count as internalists according to his standard.  Thus, anyone who thinks that such philosophers do count as internalists will think that Hetherington’s objections pose no threat to internalism generally but only to versions of it that are extreme enough to endorse TT.

     The problem isn’t just that some who claim to be internalists aren’t internalists by Hetherington’s standards. Indeed, when I argue, at the end of section 1, that endorsing mentalism is not sufficient for being an internalist, I too say that not all who claim to be internalists are internalists.  But my standard for being an internalist (i.e., that one require, for justification, some sort of awareness of some J-contributor) is much less strict than Hetherington’s TT-requirement and much more in accord with accepted views about who counts as an internalist.  This makes my standard easier to defend and it prevents internalists from sidestepping my dilemma in the way they can easily sidestep Hetherington’s dilemma.

     Hetherington’s characterization of internalism in premise (I*) is not only too narrow, it is also inadequately supported.  In his 1990, he appeals to P – which, as I noted earlier, is basically the same thing as TT – when he explains how the regress he identifies there is generated by internalism.  So it is clear that, in his 1990, he is assuming that internalists are committed to TT.  However, he doesn’t really defend that claim until his 1991 where he is more explicit about the role of TT in generating the regress that he thinks follows from internalism. 

     Hetherington begins his 1991 defense of (I*) – the claim that internalists are committed to TT – like this:

For the Transparency Thesis [TT] to be false is for W1, say, to be epistemically within S without S appreciating it as epistemically within. (1991, 859)[44]

 

But this is a mistake. For, contrary to what Hetherington says, the falsity of TT does not imply that some potential J-contributor, such as W1, is epistemically within S without:

(a) S appreciating W1 as epistemically within.

Instead, the falsity of TT implies that W1 is epistemically within S without:

(b) S appreciating W1's being epistemically internal to S as her reason.

Now if the nonoccurrence of (b) entailed the nonoccurrence of (a), then the Hetherington passage just quoted would be correct.  But for (b) not to occur is for it to be the case that:

(c) S doesn't appreciate as her reason W1's being appreciated by her as being her reason.

And for (a) to occur is for it to be the case that:

(d) S does appreciate W1 as being appreciated by her as her reason.

Thus, if (c) and (d) can both be true at once, then the nonoccurrence of (b) doesn’t entail the nonoccurrence of (a), in which case the passage quoted above from Hetherington is mistaken.  And it seems that (c) and (d) can both be true at once.  A person can appreciate X as being F without appreciating X’s being F as her reason.  So a person can appreciate W1 as being appreciated-by-her-as-her-reason without appreciating W1’s being appreciated-by-her-as-her-reason as her reason.[45]  Hetherington seems to be conflating talk of “S's failure to appreciate an appreciation as a reason” with talk of “S's failure to appreciate something (which needn't be an appreciation) as being appreciated as a reason”.  

     In the end, this sort of conflation results in the failure of his 1991 defense of (I*).   To see this, consider the following claim Hetherington offers in support of (I*):

If the Transparency Thesis is [taken by the internalist to be] false then, for the epistemic internalist, justification can depend solely on facts about S which are in no way epistemically internalised by S.  That is, if the Transparency Thesis is [taken by the internalist to be] false, then the epistemic internalist becomes an epistemic externalist. (1991, 861)

 

This claim is false.  A person can insist that justification depends on facts which are epistemically internal to S (i.e., facts which are appreciated by S as her reasons) while at the same time denying that those appreciated facts are facts about something’s being appreciated by S as her reason.  Consider, for example, Hetherington’s case of a child’s belief that there is a cat in front of her.  (Call that belief ‘B’ and call whatever it is that B is based on ‘W’.)  The internalist can insist that B’s justification depends not just on W but also on the child’s appreciation of W as her reason (for B).  And the internalist can insist on this while denying that B’s justification depends on the child’s appreciating as her reason W’s being appreciated by her as her reason. In this way, the internalist can do what Hetherington, in the quotation above, says she must do (i.e., deny that “justification can depend solely on facts about S which are in no way epistemically internalised”) and yet still reject the Transparency Thesis.[46]

     I’ve argued that, in both his 1990 and his 1991, Hetherington proposes the dilemma mentioned in the second paragraph of this appendix – a dilemma in which the Transparency Thesis plays a key role.  And I’ve argued that internalists needn’t be troubled by that dilemma, first, because they won’t accept his characterization of what is essential to internalism and, second, because he gives them no good reason to think they should accept that characterization.  None of these remarks applies to my dilemma.  Thus, the fact that internalists can escape Hetherington’s objection shouldn’t make them optimistic about being able to respond adequately to mine.

 

REFERENCES

 

Alston, William. 1986.  “Internalism and Externalism in Epistemology.”  Philosophical Topics 14: 179-221.  Reprinted in Alston 1989, 185-226.  Page references are to reprint.

________. 1989.  Epistemic Justification:  Essays in the Theory of Knowledge.  Ithaca: Cornell University Press.

Bergmann, Michael. 1997.  “Internalism, Externalism and the No-Defeater Condition.”  Synthese 110:399-417.

________.  2000a.  “Deontology and Defeat.”  Philosophy and Phenomenological Research 60: 87-102.

________. 2000b  “Externalism and Skepticism.”  The Philosophical Review.  109: 159-94.

________.  2004.  “Externalist Justification Without Reliability.”  Philosophical Issues 14.

BonJour, Laurence. 1985.  The Structure of Empirical Knowledge.  Cambridge: Harvard University Press.

________.  2001.  “Towards a Defense of Empirical Foundationalism.” In Michael DePaul (ed.) Resurrecting Old-Fashioned Foundationalism pp. 21-38.  Lanham, Maryland: Rowman and Littlefield.

Chisholm, Roderick.  1982. The Foundations of Knowing.  Minneapolis: University of Minnesota Press.

Cohen, Stewart.  1984.  “Justification and Truth.”  Philosophical Studies 46: 279-95.

Conee, Earl and Richard Feldman.  1998.  “The Generality Problem for Reliabilism.”  Philosophical Studies 89: 1-29.

________.  2001. “Internalism Defended.”  American Philosophical Quarterly 38:  1-18.

Feldman, Richard. 1988a.  “Epistemic Obligations.” In James Tomberlin (ed.) Philosophical Perspectives, 2, Epistemology pp. 235-56.  Atascadero, CA: Ridgeview.

________.  1988b.  “Subjective and Objective Justification in Ethics and Epistemology.”  The Monist 71: 405-19.

________.  2000.  “The Ethics of Belief.”  Philosophy and Phenomenological Research 60: 667-95.

Foley, Richard.  1985.  “What’s Wrong With Reliabilism?”  The Monist 68: 188-202.

Fumerton, Richard. 1988.  “The Internalism/Externalism Controversy.” In James Tomberlin (ed.) Philosophical Perspectives, 2, Epistemology pp. 443-59.  Atascadero, CA: Ridgeview.

________.  1995.  Metaepistemology and Skepticism.  Lanham, Maryland: Rowman and Littlefield.

Goldman, Alvin.  1999. “Internalism Exposed.”  The Journal of Philosophy.  96: 271-93.

Hetherington, Stephen.  1990.  “Epistemic Internalism’s Dilemma.”  American Philosophical Quarterly 27: 245-51.

________.  1991. “On Being Epistemically Internal.”  Philosophy and Phenomenological Research 51: 855-71.

Lehrer, Keith. 1990. Theory of Knowledge.  Boulder: Westview Press.

Moser, Paul.  1985.  Empirical Justification.  Boston: D. Reidel.

________.  1989.  Knowledge and Evidence.  New York: Cambridge University Press.

Plantinga, Alvin.  1993a.  Warrant: The Current Debate.  New York: Oxford University Press.

________. 1993b.  Warrant and Proper Function.    New York: Oxford University Press.

Pollock, John. 1986.  Contemporary Theories of Knowledge.  Savage, MD: Rowman and Littlefield.

Pollock, John and Joseph Cruz. 1999.  Contemporary Theories of Knowledge, 2nd Edition.  Lanham, MD: Rowman and Littlefield.

Sellars, Wilfrid. 1963.  “Empiricism and the Philosophy of Mind.”  In Science, Perception and Reality.  New York: The Humanities Press.

________.  1975.  “The Structure of Knowledge.” In Hector-Neri Castaneda (ed.) Action, Knowledge, and Reality: Critical Studies in Honor of Wilfrid Sellars.  Indianapolis: Bobbs-Merrill.



* I’m delighted to include this paper in a volume honoring Alvin Plantinga, largely because he’s the person to whom I consider myself most indebted philosophically.  My thanks to Jeffrey Brower, Jan Cover, Richard Fumerton, Stephen Hetherington, Timothy McGrew, Kevin Meeker, Trenton Merricks, Alvin Plantinga, Joel Pust, Michael Rea and Matthias Steup for comments on earlier drafts.  Thanks also to my commentator, Paul Moser, and to participating audience members for their helpful discussion when I presented this paper at the Central Division Meeting of the APA in Chicago in April 2000.  Finally, I would like to thank the Pew Evangelical Scholars Program for providing support while I worked on this paper.

[1] For an example of the first sort of objection, see Alston 1986 and Plantinga 1993a, chapter 1.  For an example of the second, see Alston 1986 and Goldman 1999 (actually, Goldman attacks versions of internalism according to which a necessary condition for justified belief is a further bit of knowledge or, at least, the potential for this further bit of knowledge).

[2] See for example Moser 1989 and Fumerton 1995.

[3] It is, admittedly, difficult to find a very clear statement of this sort of dilemma in Sellars’ work.  But in his 1963 (131-32) he considers two ways to think of episodes of sensing what he calls ‘sense contents’: as cognitive episodes or as noncognitive episodes.  And one of the main burdens of his paper is to show that either way, the proponent of what he calls ‘the myth of the given’ runs into trouble.  See also his 1975.

[4] Let me emphasize that, in presenting this argument, I am not objecting to foundationalism per se or to the doctrine of the given.  It’s also worth highlighting the fact that my argument differs from both BonJour’s and Moser’s in that my dilemma has to do with being either conceptual or not whereas their dilemmas have to do with being either cognitive (where that means doxastic or propositional) or noncognitive.

[5] I say it hasn’t been widely observed.  But something like it has been noticed by Stephen Hetherington (see his 1990 and his 1991).  However, as I argue in the appendix to this paper, Hetherington’s papers needn’t worry internalists (even though my argument should).

[6] Thus, BonJour says (2001, 21) that internalism imposes the “requirement ... that the justification for a belief must be cognitively accessible to the believer”.  Notice that he here makes both of the two claims that I emphasized from premise (I).

[7] This corrects what I said in Bergmann 1997.  There I failed to distinguish between:

(A) At least one necessary condition of justification other than the no-defeater condition (NDC) is internal to normal humans

and

(B) At least one necessary condition of justification other than NDC must be internal to the subject (i.e., must be if her belief is to be justified).

(NDC is satisfied by S’s belief that p just in case S doesn’t take her belief that p to be defeated.)  Suppose, as internalists commonly do, that justification is necessary for warrant (that which makes the difference between knowledge and mere true belief).  Then my 1997 account entails that endorsement of (A) is sufficient for being an internalist (contrary to what I just asserted in the text) and that rejection of (A) is necessary for being an externalist.  I would now say the following three things:

(1) endorsement of (B) is necessary for being an internalist

(2) endorsement of (A) is necessary for being a nonskeptical internalist

(3) rejection of both (A) and (B) – or at least of their analogues with respect to warrant – is necessary for being an externalist. 

(Notice, by the way, that requiring NDC for justification is not to endorse I3 since NDC doesn’t require awareness of anything; at most it requires the absence of awareness of a certain thing.)

[8] See Pollock and Cruz 1999, 133 and 136 for elaboration that makes it clear that Pollock thinks awareness on the part of the subject is not required for a state to be accessible to one’s automatic processors – i.e., to the “cognitive mechanisms that direct our epistemic cognition”.

[9] This counterexample is designed so that the view being described satisfies the conditions Feldman and Conee say are sufficient for being an internalist position.  To make it apply to Pollock’s account, replace the term ‘mental states’ with ‘internal states’ where those are defined as states accessible (in a nonepistemic sense) to the subject’s automatic processors, not to the subject.

[10] It is irrelevant to point out that the view just described is an implausible one.  What matters is that it is clearly not an internalist one.  As for whether that entails that it is an externalist view, see my discussion below near the beginning of section 3.2.

[11] The point here isn’t to make dogmatic pronouncements on the inner life of dogs and cows.  I’m just working with common assumptions in order to give the reader some idea of what nonconceptual awareness is.

[12] Or that X is evidence for B or that X is a truth-indicator for B or that X contributes in some way to B’s justification.  I will often suppress these sorts of alternatives but they should be assumed.

[13] See Moser (1989, 173-76), Fumerton (1995, 64), and Alston (1986, 211) for discussions of this sort of regress problem.

[14] Even internalists shy away from imposing such exceedingly high standards for justified belief.  See Fumerton 1995, 64 for this sort of reaction.

[15] One response to this sort of move is to deny that there is any nondoxastic conceptual awareness – to insist that all conceptual awareness is doxastic involving some sort of belief.  For the purposes of this paper, I’ll simply waive this concern since taking it seriously only makes things worse for the internalist.

[16] Moser (1989, 80-81) makes a similar claim though the regress problem he is speaking of is not a complexity regress problem.

[17] In stating I6 so that it applies to the justification of concept applications as well as beliefs, I’m assuming that the internalist we have in mind here – the proponent of nondoxastic conceptual1 awareness requirement – thinks that the justification of concept application, like the justification of belief, requires conceptual1 awareness.  There seems to be no good reason to think justification of belief requires conceptual1 awareness if one thinks that justification of concept application does not. (Why would an externalist account of the justification of concept application be satisfactory if an externalist account of belief justification isn’t?)  I’m also assuming that the internalist we have in mind thinks that the conceptual1 awareness required for the justification of concept application, like the conceptual1 awareness required for belief justification, must itself be justified (i.e., that it must involve justified concept application).  Again, there seems to be no good reason to demand this in the case of belief but not in the case of concept application.  (If you think insane or irrational concept application is sufficient, why think concept application is even necessary?)

[18] Notice that even if for every justification contributor Xn, Xn = X1, the series still includes an ever-increasing complexity of concepts applied to X1.

[19] For example, if the belief for whose justification the concept application is required is the belief that p, then the content of the required concept application will be something like being in some way relevant to the appropriateness of that belief that p.

[20] For detailed discussion of their views, see section 4 below.

[21] Every process token is an instance of some reliable process type (it is due, in part, to this fact that reliabilism faces the generality problem – see Feldman and Conee 1998).  But reliabilism is committed to the view that for each process token, there is a relevant type of which it is an instance – a type whose reliability (or unreliability) determines the justification (or lack thereof) of the belief produced by the belief-forming process token in question.

[22] The internalist might protest that she didn’t originally start with the intuition that justification requires an infinite number of beliefs of ever-increasing complexity – she started with something that sounded more plausible.  But my claim is that once she sees that her view leads to that bizarre consequence, she should reject her view.  The consequence reveals the excessiveness implicit in the original intuition.

[23] This is exactly what Fumerton does say (1988, 449) in response to a point that is similar to the one I’m pressing.  To be honest, I’m not at all sure that a view like Fumerton’s – where justification supervenes on direct nonconceptual awareness of a certain kind (more details of his view are given in section 4) – is a clear case of a nonexternalist view (see Moser 1985, 147 for similar remarks).  But I won’t press the point here.

[24] My discussion of the Julie case establishes the compatibility of endorsing I7 and rejecting I3 in the same way that it establishes the compatibility of endorsing I1 and rejecting I3.

[25] One I myself endorsed at one point.  See my 1997.

[26] Objective deontological justification of the sort on which Feldman focuses (see his 1988a, 1988b and 2000) is typically not employed in support of internalism (certainly not by Feldman who, although he is an internalist and a deontologist, denies that deontology supports internalism).   In section 4.2 of my 2000a I discuss some strategies for trying to show that objective deontological justification supports internalism and explain why they fail.

[27] The three objections to externalism mentioned above say that external conditions aren’t sufficient for a philosophically interesting sort of justification.  Assuming, as internalists typically do, that justification is necessary for knowledge, these objections also say that external conditions aren’t sufficient for warrant (that which makes the difference between knowledge and mere true belief).  Two other objections that deserve mention are the generality problem for reliabilism (see Conee and Feldman 1998) and the charge that, since the beliefs of a Cartesian demon victim can be justified, a reliability condition isn’t necessary for justification (see Cohen 1984, 280-82; Foley 1985, section I; Moser 1985, 240-241; and Lehrer 1990, 166).  However, unlike the objections considered in the text, these latter two objections are not directed at externalism generally.  Instead, each draws attention to problems with reliabilism in particular.  Thus, even if successful, these objections don’t threaten all forms of externalism in the way in which my dilemma for internalism threatens all forms of internalism.  (See Plantinga 1993b, 29 for a defense of the claim that an externalist position – i.e., his own – avoids the generality problem.  And see my 2004 where I argue for an externalist account of justification that avoids the evil demon objection to reliabilist accounts.)

      Furthermore, the reliabilist can respond to the complaint about a reliability condition not being necessary for justification by remaining silent about justification and saying only that the reliability condition she proposes is necessary for warrant.  If the externalist does this, then she isn’t requiring too much for the internalist’s liking: she isn’t imposing any conditions on justification and, in imposing external constraints on warrant, she isn’t thereby contradicting the internalist’s own views (since everyone allows for external constraints on warrant).  This suggests that the principal point of disagreement between internalists and externalists does not have to do with whether externalists are requiring too much for justification; rather, it has to do with whether externalists are requiring enough for warrant.  This latter issue is the one on which the three objections to externalism mentioned in the text are focused.  It’s true that internalists tend to speak about justification, not warrant (just as externalists tend to speak about warrant and not justification).  But given that internalists think justification is necessary for warrant, their claim that awareness is required for justification has implications that contradict the externalist’s views on warrant.

[28] Moser (1989, 80) writes:

The awareness relevant to [internalism] must be essentially nonconceptual to enable those views to avoid the defects of the views criticized [earlier in the book].  Nonconceptual awareness is just awareness that does not essentially involve the application or the consideration of a concept.

[29] On p. 136 of his 1989 he says that being presented with E is necessary for E’s being a maximal unconditional probability-maker for P for S.  And on p. 141 he says that E’s being a maximal unconditional probability-maker for P for S is necessary for the justification of S’s belief that P.

      On p. 82 Moser says that the notion of presentation is conceptually basic in his system.  However, he likens it to what Bertrand Russell calls ‘acquaintance’ (Fumerton too points us to this Russellian notion for help in understanding his notion of direct acquaintance – see Fumerton 1995, 75).  And he (Moser) says that this presentation is a direct and nonconceptual sort of noticing.  On pp. 81-82 of his 1989 he defines awareness in terms of direct attention attraction and direct attention attraction in terms of presentation.  So in his system these three notions – i.e., awareness, direct attention attraction and presentation – are intimately linked.

[30] He mentions this requirement on p. 141 of his 1989.  I say the required awareness is nonconceptual because on p. 142 he says “the de re awareness of such a relation [i.e., of E’s supporting P] can be understood via the notion of direct attention attraction introduced [earlier] to clarify the notion of presentation”.  See the previous note where I point out that Moser thinks of presentation (and direct attention attraction) as a nonconceptual sort of direct awareness.

[31] What’s important here is to distinguish an awareness of being appeared to redly from both (a) the application of a concept to the object of that awareness and from (b) the belief—call it ‘B2’—that one is being so appeared to.  Once these three are distinguished, we need only consider some malfunction (due to brain damage or an evil demon) that prevents the subject from applying (and from being able to apply) to the object of the awareness in question a concept such as being in some way relevant to the appropriateness of holding B2.  Given that the awareness, the concept application and the belief are three distinct things, I don’t see why such malfunction is impossible (at the very least, the suggestion that it is impossible would require some defense).  And given the possibility of such malfunction, there is no reason to think the following aren’t compossible: (i) S’s holding the true belief B2 (that she is being appeared to redly), (ii) S’s being aware of her being appeared to redly, and (iii) S’s being unable to apply to the object of that awareness a concept like being in some way relevant to the appropriateness of holding B2.

[32] See his 2001, 29 and note 14.  He recognizes (note 14) that the term ‘content’ seems a little strange when applied to something nonconceptual and nonpropositional but he says that he uses it because he knows of “no better term for what one is conscious of in having sensory or phenomenal states of consciousness”.  But he warns us that “the two sorts of content [propositional and sensory] are importantly different and should not be conflated”.

[33]  We have to suppose that Watson doesn’t have as evidence for the conclusion that the butler did it the belief that Holmes endorses that conclusion (for Watson can see that that is a good reason for the conclusion that the butler did it).

[34]  A version of this paper was first presented at a meeting of the Central Division of the American Philosophical Association in Chicago in April 2000.

[35] This statement of TT is quoted from Hetherington’s 1991, 858.  As I mention below in the text, Hetherington explains that for something to be epistemically internal to S “is for it to be appreciated by S as being his or her reason” (1991, 858). 

[36] The claim that the regress problems in question imply that epistemic internalism is an empty concept is made by Hetherington in both his 1990, 247 and his 1991, 862.

[37] One strange thing about this sort of objection to internalism is that there seems to be no need to formulate it as a dilemma.  Why not simply drop premises (II*) and (IV*) and change (V*) so that it says “If internalists must admit that epistemic internalism is an empty concept, then internalism should be rejected”? 

[38] There he says (p. 859) that “epistemic internalism must accept the Transparency Thesis” and (p. 861) that:

Epistemic internalism, therefore, faces a dilemma.  Either it ceases to require that S epistemically internalise facts about his or her epistemic internalisings, or it does not.  (That is, either the Transparency Thesis is [taken by the internalist to be] false or it is not.)  ...  If the Transparency Thesis is [taken by the internalist to be] false then the epistemic internalist becomes an epistemic externalist. 

     So the Transparency Thesis will be retained ... But then there is the regress.

[39] Hetherington 1990, 246-47.

[40] You might think that Hetherington would answer this question by pointing out that, according to internalists, nothing can contribute to a belief’s justification unless the subject appreciates it as a justification-contributor.  But in his 1990, 245 he says that:

At least part of what [the internalist] thinks is that some aspect A of your circumstances is epistemically internal to you and is at least part of what makes you justified.

And he adds the following in a note attached to that sentence:

I say “at least part” because, for generality’s sake, I am happy to allow that an epistemically internalist condition could be necessary, sufficient, or even something looser still.

This makes it sound like an internalist could allow that there are other parts of what makes you justified that are not epistemically internal to you.  But then, according to internalists, something can contribute to the justification of your belief even if you don’t appreciate it as a justification-contributor.

[41] I don’t quite know what to make of the locution “epistemically external to your being justified” which appears in the passage just quoted and at other places in his 1990 (along with “epistemically internal to your being justified”).  I assume he intended to say “epistemically external to you” (and “epistemically internal to you”).

[42] By comparing R* to R, we can see why R isn’t quite right.  Consider, for example, the first appreciation mentioned in R. Given Hetherington’s account of what it is to be epistemically internal, that first appreciation can be restated as:

the appreciating-of-W1-as-being-appreciated-by-S-as-being-her-reason.

Notice that this appreciation differs from each appreciation mentioned in R*.  For in this appreciation something is being appreciated, not as being S’s reason, but as being appreciated by S as being her reason.

[43] See the first passage quoted in note 40.

[44] Hetherington uses ‘epistemically within S’ and ‘epistemically internal to S’ synonymously.

[45] Hetherington might want to argue that if TT is true, then this isn’t possible.  But even if that were true, how would it save his defense of  (I*)?  Does he want his argument that the internalist is committed to TT (a principle Hetherington thinks leads to regress problems), to rely on the truth of TT?

[46] For another example of how this sort of conflation spoils Hetherington’s defense of (I*), see his 1991, 859.  There he defends the conclusion that internalists are committed to thinking that a potential J-contributor, W1, contributes to the justification of S’s belief B only if S appreciates W1 as a J-contributor.  And he takes that to support the claim that internalists are committed to TT.  But TT is a thesis about appreciations entailing higher-level appreciations of appreciations.  Once again, Hetherington seems to be treating a discussion of appreciating facts about W1 as if it’s a discussion of appreciating facts about appreciating facts about W1.