Spring 2016, NYU Abu Dhabi

BonJour's Critique of Goldman

The Regress Problem and Externalism

BonJour begins his article "Externalist Theories of Empirical Knowledge" by talking about the Regress Problem. We will be talking about this in more detail in a week or two. The basic idea is this. Take some belief or yours which purports to be justified. We can ask: what makes it justified?

As emerged in our discussion of Goldman, some beliefs are justified by being based on or inferred from further supporting beliefs. (They're the beliefs that are formed by processes that take other beliefs as input.) This kind of justification is called inferential or mediate justification. The inferred beliefs get their justification from the lower-level beliefs that they're based on or inferred from.

We can raise our question again, about the lower-level beliefs: what makes those beliefs justified? If they in turn are based on or inferred from other beliefs, we ask the question again. And again.

Will this process ever end? There are four possibilities:

The foundationalist is someone who thinks that only the last way is legitimate. He thinks that if some belief of yours is justified, then its justification has to trace back to some basic, immediately justified belief. Infinite chains and circles and so on won't cut it. Those can't be ways for your beliefs to be justified.

Now suppose the foundationalist is right. Then your structure of beliefs will look something like this:

What sort of story should we tell about the basic beliefs B'', at the bottom of the picture? What makes them justified?

One sort of story you could tell would be a reliabilist story, like Goldman does. You'd say that those basic beliefs are justified because they were formed in a reliable way.

Another sort of story would be a more internalist story about what makes those beliefs justified. Their being justified doesn't depend on factors "external" to you, like reliability and so on. It depends on factors "internal" to you. It's just that these factors don't include other supporting beliefs.

How might such a story go? Well, suppose you have a toothache, which feels quite painful. You would be justified in believing you have a toothache. But what justifies you in believing it? In this case, it doesn't seem to be any other belief which does the work. It's not like you're justified in believing something else, which gives you evidence for thinking you have a toothache. What justifies you is just the toothache itself, the way it feels. This is a factor "internal" to you; and it's not a belief.

Many philosophers think that your basic beliefs B'' are justified in something like that way. For those basic beliefs, it's the sensations and experiences you have which do the justificatory work, not other beliefs.

When we talk about perceptual experiences, these concern how the world looks and feels to you. These experiences are different from beliefs. It could look to you like the following two lines are different lengths, even though you've measured them and so believe that they have the same length:

You experience these lines as having different lengths even if you believe them to have the same length. This shows that experiences are different from beliefs. (Of course, you can have beliefs about your experiences. But the experience itself is not a belief.)

Traditionally, only the second kind of story has been called "foundationalist." But some people, like BonJour, call the first, externalist, story a kind of "foundationalism," too. You'll see the word used in both ways. Myself I prefer to use the word in the traditional way: so by my lights, you have to be an internalist in order to be a foundationalist.

Here's a little organizational chart which might help:

Externalism about Justification Internalism about Justification
for example, Goldman's Reliabilism

(BonJour thinks of these views as
one species of "foundationalism")
(Coherentism--we'll talk about this view later) Foundationalism
(as traditionally construed)

These guys give some internalist story about why the basic beliefs are justified


BonJour has a quick argument (on pp. 54-5) that it's not possible for us to have any basic beliefs. We'll be coming back to this argument in later weeks. (It plays a bigger role in the second BonJour article we'll be reading.) BonJour thinks that the foundationalist has only two strategies for avoiding that argument. One is to be an externalist kind of "foundationalist."The other is to adopt some form of the Given Theory. He doesn't think any of those strategies ultimately succeed. (So in the end, he thinks there are no basic beliefs. We have to give up foundationalism and become coherentists. We'll be talking about that later, too.) But in this paper, he focuses on the arguments against the externalist.


Earlier we mentioned a number of problem cases for the reliabilist. One sort of case involves subjects who have reliable "hunches" at the race track, or a reliable faculty of clairvoyance or telepathy or something like that. Many people have the intuition that it would be irresponsible and unreasonable to rely on faculties like that, unless you had independent evidence that they are reliable. Their merely being reliable is not enough to justify you in accepting them.

BonJour takes up these sorts of cases.

Three Stances

In order to asses BonJour's argument here, let's step back a little and think again about how to understand the disagreement between the internalist and the externalist.

Sometimes this disagreement is formulated in the following way. We ask: To be justified in believing P, does a subject have to be able to "justify" his belief that P, that is, to defend it by argument?

Here is a second way of understanding the disagreement between the internalist and the externalist. We ask: Are epistemic properties (like justification) shared between all "internal duplicates"? If two subjects are the same "on the inside," does that entail that they're equally reasonable in believing as they do?

If you say yes to that question, then you're some sort of internalist. But internalism in this second sense is much weaker than internalism in the first sense. Internalism in this second sense does not entail that whenever your beliefs are reasonable or justified, you will always be able to "justify" them, or defend them in argument.

Let's distinguish three different stances you can have in epistemology. Suppose you believe that P, and D is some potential defect in your belief that P. For example, D might be "Your senses are unreliable,"or it might be "You are a BIV."

In between the conservative stance and the liberal stance we can put a third position:

The intermediate view differs from the liberal view in that the intermediate view says that it has to be true that your beliefs lack the defect, for those beliefs to be justified. So brains in vats, and subjects who formed their beliefs unreliably, do not have justified beliefs, on the intermediate view. The liberal view, on the other hand, says that so long as there's no evidence that your beliefs have the defect, that's good enough.

If you tend to think of justification in terms of giving justifying arguments, then you will feel strong pressure towards the conservative view here. If you're giving an argument for your belief, and someone asks you, "Hey how do you know you can trust your senses here?" you have to respond to that challenge. This is why BonJour thinks that for Norman to be justified in believing the President is in NYC, Norman would have to have independent reasons for thinking that his clairvoyant faculty is reliable.

If, on the other hand, you think of justification more in terms of a good epistemological status, which some beliefs have and others lack, then it will be an independent question whether you'd be able to demonstrate by argument that your beliefs have this status. Remember the analogy of the little child: He can have a good character even if he can't defend his character against attack. Similarly, your beliefs might be justified even if you're not able to defend them against attack.

So now let's go back to the question of how to understand the disagreement between the internalist and the externalist. If the question is:"For your belief to be justified, do you have to have evidence that would enable you to demonstrate that your belief lacks all these defects?" the conservative says yes, but the intermediate view and the liberal view say no. If the question is: "Will epistemic properties be shared between all internal duplicates," then the intermediate view says no, because whether or not a subject has the defect is an "external" fact. There could be two internal duplicates, one of whom had the defect and the other not. The conservative and the liberal, on the other hand, can say yes here.

I like to think of the internalism/externalism debate in terms of the question about internal duplicates. So by my lights, the conservative and the liberal are both internalists. It's just that the one is a much more conservative internalist than the other. He thinks the requirements for having a justified belief are much more demanding. The intermediate view, on the other hand, is clearly an externalist view.

Now, BonJour holds the conservative view. He's arguing against the intermediate, externalist view. He doesn't really consider the most liberal view. I guess he thinks that the liberal view is too liberal. He thinks that many of the beliefs it counts as being "justified" would not really be justified. Its standards are too loose.

I want you to appreciate the role that BonJour's conservativism is playing in his argument against Goldman. We will look at this more closely in a moment.

I also want you to recognize that BonJour's conservativism and Goldman's externalism (the intermediate view) aren't the only options here. There's also the possibility of going for the liberal view. Of course, it might turn out, as BonJour believes, that the liberal view's standards are too loose, and so we should reject it. But that's a matter we'll have to investigate and argue about.

BonJour's Conservativism

Recall when we were asking whether Norman was "subjectively irrational," whether from his own standpoint, it was unreasonable for him to be accepting his clairvoyant belief. BonJour said that if Norman had no beliefs about whether or not he had a reliable clairvoyant faculty, then from Norman's standpoint, there is no way he could know the President's whereabouts. But this is tricky. We need to distinguish:

i. Norman has evidence that there is no reliable way for him to know the President's whereabouts

ii. Norman lacks evidence that there is a reliable way for him to know the President's whereabouts

These are importantly different. Suppose you're on Temptation Island. That's this TV show where you go on one side of the island, and your boyfriend or girlfriend goes on the other side of the island, and people try to tempt both of you to cheat.

There's a big difference between:

iii. getting evidence that your partner was unfaithful

iv. failing to get evidence that your partner was faithful

To illustrate, suppose one of the other contestants on your side of the island gets evidence that their partner has been especially faithful. You don't get any such evidence. This bums you out. However, by itself that wouldn't be enough to justify you in cheating, would it? Perhaps your partner has also been faithful, too. You just don't know yet. (iv) isn't the same as (iii).

Similarly, just because (ii) Norman lacks evidence that his belief was formed reliably, that's not the same as saying (i) he has evidence that his belief was formed unreliably. BonJour treats these on a par. In both cases, BonJour wants to say that "From Norman's own standpoint, there is no way he could reliably know the President's whereabouts." But BonJour seems to be glossing over an important difference here.

For a conservative like BonJour, if either (i) or (ii) is true, that's enough to show that Norman's belief is unjustified. The conservative thinks that Norman can have a justified belief only if he does has evidence that the way he formed his belief is reliable. If you held the liberal view about justification, though, then it would only be (i) which made Norman's belief unjustified. If all that's true is (ii), then Norman can still be justified in his belief. From the liberal perspective, BonJour is trying to trick us into confusing (ii) with (i).

Of course, the liberal view is not a reliabilist view. The liberal agrees with BonJour that external factors like reliability are not what make our beliefs justified. But the liberal has to reject BonJour's criticisms of the reliabilist. Because if those criticisms are right, then the liberal view would also fall. BonJour's criticism assume that, if your belief is going to be justified, then you need to have independent reasons to believe it was formed in a reliable way. Otherwise, BonJour argues, the belief will be "subjectively irrational." From your own standpoint, it will be unreasonable for you to hold that belief. This is something that both the externalist and the liberal internalist have to deny.

BonJour has another argument against the externalist, which works in a similar way. I call this second argument BonJour's Accident Argument. It occurs on pp. 63-4 of this paper (and also in some other papers). The argument goes roughly as follows:

Take some belief for which you have no supporting evidence, but which was as a matter of fact reliably formed. If the belief was reliably formed, then you're unlikely to go wrong in accepting it, and in a sense it's not an accident that this is so. From your subjective perspective, however, it would seem an accident if your belief turned out to be true, since you have no evidence in support of the belief. Perhaps it wouldn't seem an accident from the standpoint of an external observer, who knew that the belief was reliably formed. But the rationality or justifiability of your belief should be judged from your own perspective, rather than from a perspective unavailable to you. And from your own perspective, your belief is one you're not be justified in holding.

What exactly is BonJour's argument here? There seem to be three steps.

1. Your belief was reliably formed, but you have no evidence that it was.

2. Hence, from your subjective perspective, it would seem an accident if your belief turned out to be true.

3. Hence, your belief is unjustified.

A lot turns here on what exactly (2) means.

There are clearly some interpretations of (2) where it would support (3). We can interpret "from your perspective, it would seem an accident if your belief turned out to be true" to mean: you have good reason to believe that your belief was formed in an unreliable way. If that's true, then your belief is unjustified. For example, if I form a belief about your mother's maiden name by picking names at random from the phone book, then I have good reason to believe that my belief about your mother's maiden name was formed unreliably, and my belief about your mother's maiden name is clearly unjustified. So this interpretation of (2) makes (2) support (3). But on this interpretation of (2), (2) doesn't follow from (1);and (2) isn't anything to which the reliabilist is otherwise committed. As I've been emphasizing, just because you lack evidence that your belief was reliably formed, it doesn't follow that you have evidence that your belief was formed unreliably. (The reliabilist can agree that when one does have evidence that a belief was formed unreliably, then the belief is unjustified. For example, he can agree that my belief about your mother's maiden name is unjustified. This is like Goldman's Jones case, and the second and third of BonJour's clairvoyance examples.)

On the other hand, there are other interpretations of (2) where it clearly does follows from (1). We can interpret "from your perspective, it would seem an accident if your belief turned out to be true" to mean: you lack any good reason to believe that your belief was formed reliably. On this interpretation, (2) does follows from (1). However, we haven't yet seen any argument that beliefs which are "accidental" in this sense are unjustified. The rhetorical talk of "luck" and "accidents"may tempt us to conclude that beliefs of this sort are unjustified. But BonJour hasn't given us any argument to back up the rhetoric.

The problem with BonJour's argument, then, is that there's no clear interpretation of (2) where it both follows from (1) and obviously supports (3). It may be true that in the cases BonJour is considering, the subjects' beliefs aren't justified. But BonJour hasn't here given us any convincing argument for this. The only sense of (2) where it obviously supports (3) is not a sense in which (2) follows from (1).

This "Accident Argument" is another place where BonJour is glossing together the following very different situations:

i. you have evidence that your belief was formed unreliably

ii. you lack evidence that your belief was formed reliably

From a conservative perspective, these are equally bad. In neither case would your belief be justified. But from an intermediate (externalist) perspective, like Goldman's, or from a liberal internalist perspective, the two situations are importantly different. In the case of (i), BonJour is right, your belief isn't justified; but in the case of (ii) your belief might be justified, after all.

Clairvoyance and Perception

One thing we haven't spoken much about is how clairvoyance is supposed to work, exactly. How does it feel from the inside? Does it have its own peculiar sensory quality, like vision and touch do? Or is it more like having a hunch pop into your head at the race track?

You should think this issue over. If you're not inclined to count Norman as justified in his beliefs, then ask yourself what is the epistemic difference between Norman's spontaneous clairvoyant beliefs, and your spontaneous visual beliefs. Why should the one count as justified and the other not? Why does BonJour think there is a difference? Why do you?