Michael Bergmann


The new book, Epistemic Justification, by Laurence Bonjour and Ernest Sosa presents us with a terrific exchange between two of the best representatives of internalism (BonJour) and externalism (Sosa).[1]  Their discussion cuts to the heart of the disagreement between internalists and externalists and significantly advances the discussion of this important issue.  It is, therefore, a pleasure to have the opportunity to participate in this symposium on their book. Since I am squarely within the externalist camp and philosophy is prone to be critical, my comments will pertain almost entirely to BonJour’s illuminating contribution to their co-authored volume.

     For many years now, much of BonJour’s work has focused on ways of developing a dilemma he finds in the work of Wilfred Sellars.  In his earlier work, BonJour argued against internalist foundationalism using this Sellarsian dilemma.  But he has since switched his allegiance and now wants to offer a solution to this dilemma on behalf of internalist foundationalism.  Indeed, one of BonJour’s main tasks in the book that is the focus of this symposium is to propose and defend his solution to this dilemma.  He believes that if his solution fails, internalist foundationalism is in serious trouble.  I agree with that conditional and my aim in sections 1-3 will be to argue that his solution does fail.  In the final section, I will briefly defend externalism against one of BonJour’s objections to it.


1. The Sellarsian Dilemma for Internalist Foundationalism


So what is this dilemma for internalist foundationalism?  To answer this question, we must begin with an account of the difference between internalist and externalist foundationalism.  Internalist and externalist foundationalists agree that foundationalism is true: they agree that there are (or at least could be) justified beliefs that aren’t justified in virtue of being based on or inferred from other beliefs.  Furthermore, they also agree (says BonJour) that, in order for these beliefs to be justified, there must exist a reason for them.  Where they part company is that the internalist foundationalist thinks that this required reason contributes to the justification of such noninferentially justified beliefs only if the person holding the belief is aware of it.  It “must somehow be cognitively available to the believer himself, within his cognitive grasp or ken”. (24)  Externalist foundationalists deny that there is any such awareness requirement on justification.

     W can now state the dilemma for internalist foundationalists, the dilemma to which BonJour thinks he has a solution.  Either (i) the awareness (or grasp or apprehension) required by the internalist involves judging the object of awareness to be a certain way (or conceiving of it as being a certain way) or (ii) it doesn’t.  If we take option (i), then the required awareness involves something – i.e., a judgment or an act of conceiving of something as being a certain way – that admits of being correct or incorrect and, hence, of being justified or unjustified.  But if the awareness required for justification involves something else also in need of justification, we have a vicious regress of items in need of justification. (18-21)  If we take option (ii), we avoid the regress.  But it is difficult to see (says BonJour) how that “non-judging” sort of awareness of the supposed reason can confer justification on the belief in question.  To see what he’s getting at here, consider Jack’s introspective belief that he is having a sensory experience of a certain sort.  According to (ii), that belief can be justified if Jack has the sensory experience and is aware of it, even if he doesn’t make any judgments about it or conceive of it any particular way.  But this gives rise to the following questions:

If the direct apprehension [or awareness] of the experience involves no claim or assertion regarding its character, so that who thus has such an apprehension is apparently not thereby aware that it has such-and-such features, then in what way is his belief that he has an experience with those features justified by that apprehension?  The basic belief, after all, is judgmental: it has the assertive content that something, in this case a sensory experience, has one set of features rather than one of the various others that it might have had.  How can a state whose content does not in any way say or indicate that things are one way rather than another nonetheless provide a reason or any sort of basis for thinking that the propositional content of a belief that they are one specific way is true? (19)[2]


In short, the supposed problem with option (ii) is that if Jack’s awareness of his sensory experience doesn’t involve his conceiving of it in a certain way, then neither that experience nor that awareness of it justifies his belief that his experience is a certain way.


2. BonJour’s Solution


So what is BonJour’s solution?  Does he accept (i) or (ii)?  It seems to me that he changes his mind.  In fact, it seems that when he’s explaining how he avoids the problems associated with option (i), he endorses option (ii) and that when he’s explaining how he avoids the problems associated with option (ii), he endorses option (i).  That’s a serious charge. In what follows I’ll defend it.

     BonJour’s presentation of his solution comes in three stages.  First, there is the initial statement of his view in which he tries to avoid the vicious regress associated with option (i).  Here he makes a special effort to avoid any suggestion that judging is required for the justification of noninferential introspective beliefs.  In the second stage of his presentation, there is the attempt to show that, despite the fact that he rejects option (i), he doesn’t fall prey to the problems associated with option (ii).  But his way of doing this gives rise to the concern (mentioned above) that his account of how he avoids the problems associated with option (ii) forces him to go back on his initial claim that he is rejecting option (i).  And that brings us to the third stage of his presentation, in which he argues that the concern just noted is unfounded.


2.1 Stage One


Let’s begin with an example that can be employed in presenting BonJour’s view.  Suppose that Jack is having a visual experience E of the sort one normally has when looking at a large green ball in an empty but otherwise ordinary neutral-colored room. And suppose Jack also has the introspective belief B that he is having a visual experience that makes it look to him as if there is a green ball in front of him.  BonJour’s view is that what makes an experience like E a conscious experience is that it is partly constituted by an awareness of its sensory content. (70)[3]  This awareness is nonconceptual insofar as it doesn’t involve thinking in conceptual terms of the sensory content of the experience. (71-72) 

     Let’s turn now to BonJour’s initial account (72-73) of what makes a belief such as Jack’s justified.  At the very least, he thinks the following three things are required:

(a)     Jack has experience E.

(b)     Jack has belief B.

(c)     The propositional content of B is an accurate description of the sensory content of E.


Since E’s sensory content is accurately described by B’s propositional content, BonJour thinks that E’s content constitutes a good reason for B.  And since Jack has E and is, therefore, aware of E’s sensory content, he is aware of a good reason for B.  Because Jack is thus aware of this good reason for B, BonJour concludes that he is:

in a good, indeed an ideal, position to judge directly whether the conceptual description is accurate as far as it goes, and if so, to be thereby justified in accepting the belief (73).


BonJour then admits that it is possible for a belief about one’s own experiential states to be mistaken but he insists that this isn’t a serious worry unless there is some special reason to think such a mistake is likely.  Thus, to the above three conditions, he seems to want to add one more:

(d) There is no special reason to think it likely that B is mistaken.[4]

In his initial presentation of his view, he seems to be saying that these four conditions are together sufficient for the justification of Jack’s introspective belief B.[5]  The virtue of requiring only (a)-(d) is that they contain no hint of a requirement that the subject does any judging.  The negative consequences of option (i) are thereby avoided.


2.2 Stage Two


In the second stage, BonJour attempts to explain how his solution avoids the negative consequence of option (ii) – namely, the failure of the awareness in question to provide justification.  He develops this point in two ways.

     First, he says that so long as condition (d) is satisfied:

the fact that such a belief seems via direct comparison to accurately characterize the conscious experience that it purports to describe apparently provides an entirely adequate basis for thinking that the description is correct and hence an adequate basis for justification. (73-74)


Here he seems to be saying that justification for Jack’s belief B results from a “direct comparison” of B’s propositional content and E’s sensory content, which comparison has the result that it “seems” to Jack that B’s content accurately characterizes E’s content.  This way of understanding BonJour is confirmed a little later when he says:

if the foregoing account is correct, we seem to have found a case of exactly the sort of direct comparison or “confrontation” between a conceptual description and the non-conceptual element or chunk of reality that it purports to describe which seems intuitively to be essential if our conceptual descriptions are ever to capture reality in an ascertainable way. (74)


The point in these two passages seems to be that in addition to (a)-(d), the justification of Jack’s noninferential introspective beliefs about E also requires that:

(e) Jack directly compares B’s propositional content with E’s sensory content.

Now it’s true that adding (e) helps to avoid the problems associated with option (ii).  For if Jack compares the contents of B and E and the former seems to him to accurately characterize the latter, then the worries about B’s justification mentioned above in discussing option (ii) don’t arise or at least not with the same force.  But the problem is that (e) seems to require judging or conceiving of the experience in a certain way, thereby forcing those who require (e) to take option (i).  Below we’ll consider what BonJour has to say about this objection.  However, let’s first have a look at BonJour’s second way of explaining how his solution avoids the negative consequence associated with option (ii).

     The second approach appears in the section of the book in which BonJour replies to Sosa.  Sosa has pointed out that something like (a)-(d) above isn’t sufficient for the justification of introspective beliefs.  BonJour concedes the point (192) and responds as follows:

So what then needs to be added … in order to yield internalist justification for the belief?  The answer, I suggest, is quite simple: the person must apprehend or recognize the agreement or fit between the aspect of experience being attended to and the conceptual description given by the belief.  (193)


In fact, he says this has been his position all along:

On my view, as set forth in my main essay, a foundational belief results when one directly sees or apprehends that one’s experience satisfies the description of it offered by the content of the belief.  (191, emphasis added)


It seems, therefore, that he thinks that, in addition to (a)-(d), the following condition is necessary for the justification of Jack’s belief B:

(f) Jack sees (or recognizes or apprehends) that E satisfies (or fits or agrees with) the description of it given in B’s content.


Clearly this too alleviates the concerns associated with option (ii).  However, for Jack’s belief to satisfy (f), it seems that he must judge that E fits the description given in B’s content or conceive of E as satisfying that description.  Thus, once again, BonJour’s attempt to avoid the problems associated with option (ii) seems to force him to take option (i).


2.3 Stage Three


BonJour, with his usual adeptness in philosophical debate, has foreseen both of the objections I raised in the previous subsection.  However, his replies to them are unsuccessful.  Let’s consider them in turn.

     Here is what he says in response to the charge that direct comparison requires an act of judging:

I am inclined to regard the suggestion that a direct comparison of two conscious states need involve an independent judgment that must in turn be justified by something other than the conscious contents of the states themselves as a clear case of objectionable overintellectualization. If any intellectual comparison or assessment can ever be direct and unmediated by a further judgment, surely this one can.  And to deny that this is ever possible is to guarantee vicious regresses in all directions, rendering the operation of the intellect inherently futile. (65)


The question at issue is this: is there, in addition to the belief B and the experience E, another act of judging required in order for B to be justified?  BonJour’s answer is that a direct comparison of B’s content and E’s content is required.  But can there be a direct comparison of those two contents without there being, in addition to B and E, another act of judging (or of conceiving of E in a certain way)?  It seems clear that the answer is ‘No’.  For comparison (whether direct or not) involves noting similarities and differences.  But that can’t be done if one isn’t aware that the items in question have certain features.  And that requires one to judge that they have the features in question or to conceive of them as having those features.  But again, such acts of judging or conceiving are precisely the sorts of things that can be correct or not and, hence, justified or not.  And this is what leads immediately to the regress problems associated with option (i).  BonJour may be correct in saying that requiring these acts of judging is objectionable overintellectualization.  But if that’s right, then requiring a direct comparison of the contents of B and E is objectionable overintellectualization since direct comparison requires such acts of judging.  BonJour says that those who say that direct comparison requires judging are “rendering the operation of the intellect inherently futile”.  But that’s not true.  If anything, it is the internalist who says justified belief requires direct comparison who is rendering the operation of the intellect inherently futile.

     Consider next what BonJour says in response to the second objection, according to which requiring condition (f) is requiring an act of judging:

When I am e-aware[6] of an experiential feature, it is genuinely present in my consciousness.  Because many other things are present there as well, I may have to selectively focus in order to pick out that feature and come to see that its character fits a particular description, but the awareness of it that allows me to do these things is just the original e-awareness itself.  To insist that a distinct cognitive act, requiring an independent justification, was involved would be in effect to say that e-awareness in itself has no cognitive significance at all, that I can be genuinely conscious of the feature in question but be unable to do anything further with that consciousness. (193, emphasis added)


The claim here is that the sort of awareness of an experience that one has simply in virtue of having a conscious experience involves, in itself, “coming to see that its character fits a particular description”.  But that is false.  BonJour concedes, indeed emphasizes, that what is here called ‘e-awareness’ involves no judging or conceiving of the experience in a certain way. (71-72)  But clearly, coming to see that the experience’s character fits a particular description does involve an act of judging or conceiving of the experience in a certain way.  Hence, contrary to the main contention of the passage quoted above, e-awareness of an experience does not involve coming to see that its character fits a particular description. 

     BonJour goes on to say that by denying that e-awareness of an experience involves coming to see that its character fits a particular description, one is saying that (1) such awareness has no cognitive significance at all and that (2) one can have it and yet be unable to do anything further with it.  Not so.  Those who deny that e-awareness of an experience involves coming to see that its character fits a particular description are simply clarifying the point that, as BonJour himself insists, e-awareness is nonconceptual.  It is, instead, a certain kind of internalist who is committed to saying that (1) and (2) are true—the kind who insists that, unless the awareness in question involves the subject coming to see that the experience’s character fits a particular description, (1) and (2) follow.  Externalists, on the other hand, think sensory experience can have cognitive significance even if the subject doesn’t come to see that the experience’s character fits a particular description.

     I have, in this subsection, defended my charge that BonJour’s proposed solution to his dilemma for internalist foundationalism fails.  Indeed, I’ve showed that in order to avoid the problems associated with first horn, he grabs the second and that when the time comes to show how he’s thereby avoided the problems connected with the second horn, he reaches again for the first.  His attempts to defend himself against these charges fail.


3. A Dilemma for All Internalists


Looked at from certain perspective, it is no surprise that BonJour’s attempted solution to the Sellarsian dilemma fails.  For the failure of any such solution is entailed by a more general objection—one employing a similar dilemma—that applies to all versions of internalism.  I will develop that objection briefly here in order to make it clear that it is not open to BonJour (or, for that matter, any other internalist) simply to regroup and offer another solution in a similar vein.[7] 

     Internalists are those who require for justification some sort of awareness of a potential justifier.  But all such awareness either involves conceiving of the potential justifier as in some way relevant to the truth or justification of the belief in question (call this sort of awareness ‘strong awareness’) or it doesn’t (call this ‘weak awareness’).[8]  If internalists require strong awareness, vicious regress problems result for exactly the reasons BonJour lays out. (18-21)  If they require instead only weak awareness, then, for reasons I shall give below, the internalist’s awareness requirement is unmotivated.  This gives externalism an advantage because, unlike internalism, it refrains from imposing a requirement on justification that is either impossible to satisfy (due to regress problems) or unmotivated.

     Why think that a weak awareness requirement is unmotivated?  Because the very same objections BonJour lodges against externalism (raising worries that are supposedly solved by imposing an awareness requirement) apply equally well against an internalism that imposes only a weak awareness requirement on justification.[9]  Thus, the only way to avoid the results BonJour says we need to avoid is to adopt a strong awareness requirement leading to vicious regress problems.  This suggests that the results BonJour says we need to avoid, don’t need to be avoided after all.  For it is implausible to think that the only possible way a belief can be justified leads to vicious regress problems.

     I’ve claimed that what BonJour thinks of as the difficulties of externalism afflict any form of internalism that imposes only a weak awareness requirement on justification.  What are these difficulties?  It seems that BonJour thinks that the most fundamental difficulty with externalism is that it allows that a belief, like Jack’s belief B above, can be justified even if one or both of the following are true of it:

(g) The provenance of Jack’s belief B is a total mystery to him.


(h) The status of Jack’s belief B is, as far as he can tell, no different from that of a stray hunch or arbitrary conviction.[10]


But the very same difficulty arises for those internalists who impose only a weak requirement on justification.  Consider first (g).  Does imposing a weak awareness requirement on justification guarantee that Jack’s belief B lacks justification if (g) is true of it?  Clearly not.  For a weak awareness requirement requires only that one be aware of the potential justifier (which in Jack’s case is E); there is no requirement that Jack conceive of E as part of the provenance of his belief B.  In fact, there is no requirement that Jack have any views at all about the provenance of his belief B.

     What about (h)?  Does imposing a weak awareness requirement on justification guarantee that Jack’s belief B lacks justification if (h) is true of it?  Again, clearly not.  Jack can have a weak awareness of E without conceiving of E (or anything else) as being the least bit relevant to the justification or truth of B.  As a result, imposing a weak awareness requirement on B doesn’t force Jack to think any differently of B than he would of a stray hunch or an arbitrary conviction.[11]

     I take it that these supposedly negative results of externalism constitute the main motivation for rejecting externalism and imposing an awareness requirement on justification.  Given that assumption, we may conclude that the main motivation for rejecting externalism and imposing an awareness requirement on justification fails to motivate a weak awareness requirement (since the latter is plagued with the same negative results).  Internalism, therefore, is either faced with vicious regress problems (by imposing a strong awareness requirement) or unmotivated (because it imposes only a weak awareness requirement).  Either way, externalism is to be preferred because it doesn’t impose unmotivated or impossible to satisfy awareness requirements.


4. Qualified Externalism


In closing, I wish to offer a brief defense of what BonJour takes to be the most plausible form of externalism—what he calls ‘qualified externalism’—since certain of his criticisms of it might seem to provide one final motivation for internalist foundationalism.

     Internalism, says BonJour, requires both of the following for justification: (i) that the subject be aware of positive reasons for her belief and (ii) that the subject not be aware of any negative reasons against her belief (or for the unreliability of its source).  The qualified externalist is one who accepts the latter “no-defeater” requirement but not the former awareness requirement on justification.  BonJour thinks such an externalist is in effect accepting “one part of the internalist view” (31); and he suggests that this results in “an untenable halfway house”. (32)  To demonstrate its untenability, he argues that the only reason for thinking there is a no-defeater requirement on justification is also a reason for thinking there is an awareness requirement on justification.  From this he concludes that it is “hard to see how the qualified externalist can defend one part of his position without undermining the other”. (33)

     By way of response, notice first that from the fact that internalism imposes a requirement R on some epistemic property, we can’t conclude that every other view which imposes requirement R is in some way conceding ground to internalism.  For example, the fact that internalists think true belief is a necessary condition of knowledge doesn’t show that anyone who thinks true belief is necessary for knowledge is making a damaging concession to internalists.  What is distinctive about internalism is its insistence on an awareness requirement on justification.   But a no-defeater requirement isn’t an awareness requirement.  A no-defeater requirement doesn’t require awareness of anything.  Instead, it requires the absence of awareness of something (i.e., defeaters).  The no-defeater requirement is an independent requirement on justification which, like the true belief requirement on knowledge, is accepted by many on both sides of the internalist-externalist divide.  The correctness of the no-defeater requirement isn’t what is at issue in the disagreement between internalists and externalists.  So by allowing that there is a no-defeater requirement on justification, the externalist isn’t conceding any ground to internalists or moving any closer to an internalist position.

     What then of BonJour’s argument that the only reason for thinking there is a no-defeater requirement is also a reason for thinking there is an awareness requirement?  Consider what he says is the only reason for thinking there is a no-defeater requirement:

the only clear reason for the negative [no-defeater] requirement is that accepting beliefs that are, as far as one can tell from one’s own cognitive perspective, unlikely to be true is plainly irrational and irresponsible from an epistemic standpoint that aims at truth – even if those beliefs happen to be, unbeknownst to the person in question, reliably caused. (32, emphasis added)


How does this reason for thinking there is a no-defeater requirement differ from saying that it just seems (perhaps after considering many examples) that a belief isn’t justified if one has either a reason for thinking it false or a reason for doubting the reliability of its source – and that this is so whether or not the belief is in fact reliably formed?  That’s what I would say in support of a no-defeater requirement on justification.  And I can’t see how it differs from the reason BonJour gives (though if it does differ, then he’s mistaken to say he’s identified the only reason). 

     Thuso, assuming I’ve understood him correctly, BonJour’s complaint goes like this: “The reason given for imposing a no-defeater requirement on justification is its intuitive obviousness.  But the awareness requirement is also supported by intuitive obviousness.  Since both requirements are supported by the same reason (i.e., their intuitive obviousness), it is untenable for the externalist to endorse the no-defeater requirement without also endorsing the awareness requirement.”

     But that way of arguing isn’t very convincing.  One perfectly sensible externalist reply is to say that although the no-defeater requirement seems intuitively obvious, the awareness requirement does not.   Another sensible response is to allow that, when one first considers the matter, there is intuitive support for the awareness requirement but to insist that this intuitive support is completely outweighed by the argument given above in section 3 for the conclusion that there is no such requirement on justification.  And since there is no parallel argument that undermines the intuitive obviousness of the no-defeater requirement, qualified externalism is not an untenable halfway house.  Indeed, such externalism isn’t really even properly called ‘qualified’ since endorsement of a no-defeater condition doesn’t in any way force one to back off from the externalist rejection of awareness requirements or to any make concessions to internalism.[12]

[1] Laurence BonJour and Ernest Sosa, Epistemic Justification: Internalism vs. Externalism, Foundations vs. Virtues (Malden, MA: Blackwell Publishing, 2003).  All parenthetical references in the paper are to this volume.

[2] It is interesting to note how similar this objection (to requiring only the “non-judging” sort awareness) is to BonJour’s objection to externalism: in each case, the objector thinks that even if there exists a reason for a belief B, the believer lacks the sort of awareness of it that is required in order for it to contribute to B’s justification.

[3] BonJour emphasizes that sensory content isn’t propositional or conceptual in nature. It is what one is aware of when one is aware of a sensory experience.  See BonJour and Sosa, p. 70, n. 6.

[4] There are at least two ways to read (d): so that the nonexistence of such a reason is sufficient to satisfy (d) in this case, even if Jack thinks there is such a reason; and so that Jack’s lack of awareness of any such reason is sufficient to satisfy (d) in this case, even if such a reason actually exists.  I’m not sure which, if either, of these readings BonJour has in mind.

[5] I should mention here – in support of my claim that, in his initial presentation, BonJour seems to be saying that (a)-(d) are sufficient for justification – that when Sosa summarizes (218) what he takes BonJour’s official position to be, he doesn’t mention any conditions besides these four.  My characterization of BonJour is not, therefore, idiosyncratic. 

[6] E-awareness is awareness of an experience; it is the sort of awareness that is constitutive of having that conscious experience.

[7] For a detailed presentation of this more general objection, see my paper “A Dilemma for Internalism” in Knowledge and Reality: Essays in Honor of Alvin Plantinga, eds. Thomas Crisp, Matthew Davidson, and David Vander Laan (Kluwer Academic Publishers, forthcoming).

[8] Notice that according to these stipulative definitions, every awareness is of one of these two sorts.

[9] Cf. note 2 above.

[10] I derive (g) and (h) from BonJour’s discussion (32) of the reliable clairvoyant who lacks both reasons for and reasons against her clairvoyant beliefs.  Two other propositions, very similar to (h), which BonJour thinks are in conflict with B’s justification are:

(h*) It is a matter of cognitive luck from Jack’s standpoint that B is true.

(h**) It is an accident from Jack’s subjective perspective that B is true.

I take (h*) from p. 27 of BonJour and Sosa and (h**) from p. 43 of BonJour’s Structure of Empirical Knowledge (Cambridge: Harvard University Press, 1985). My discussion below of (h) is easily adapted to apply to (h*) and (h**).

[11] Of course it is important that Jack doesn’t think there are good reasons for rejecting B or for doubting the reliability of B’s source.  But in the discussion taking place in this part of the text (30-33) BonJour notes that it is consistent with externalism to require for justification that the subject does not have a defeater for her belief.  So this concern is easily handled without adding an awareness requirement.  A “no-defeater” requirement, consistent with externalism, will do the trick.

[12] Thanks to Jeffrey Brower, Trenton Merricks, and Michael Rea for comments on earlier drafts.