The regress argument in epistemology

All of the foregoing has prepared us to come to grips with a very important argument in epistemology, called the regress argument or the infinite regress argument. It goes like this:

(1) Suppose that the belief that Q is justified by the belief that P; so P justifies Q.

(2) But if P justifies Q, then P is justified.

(3) So P is justified.

(4) But if P is justified, then it must be justified some other belief; and that belief must be justified by some other belief; and so on. There is a chain of justifying beliefs. And then there are three possibilities: (a) the chain goes on forever; (b) the chain loops around on itself, forming a circle; or (c) the chain begins with a belief that is justified, but which is not justified by another belief.

(5) Possibility (a) (called regressism) is obviously incorrect.

(6) Possibility (b) (called coherentism) is incorrect (for various reasons, explained below).

(7) Possibility (c) (called foundationalism) is the only possibility left and must be correct.

(8) Therefore, there must be some beliefs that are justified, but which are not justified by other beliefs: these are called basic beliefs. All other beliefs are justified by basic beliefs.

So the regress argument, as I've presented it here (and as it is usually presented) is an argument for a theory of justification called foundationalism. But I don't want to talk about foundationalism yet. Let's go back over this argument carefully, first.

Now, we've already explained steps (1) through (3). It advances that principle we talked about before, that if a belief is justified by another belief, then the justifying belief itself must be justified. Now let's look at premise (4). This is the really important step of the argument: "If P is justified, then it must be justified some other belief; and that belief must be justified by some other belief; and so on. There is a chain of justifying beliefs. And then there are three possibilities: (a) the chain goes on forever; (b) the chain loops around on itself, forming a circle; or (c) the chain begins with a belief that is justified, but which is not justified by another belief."

Let me try to put this premise in entirely different words. Pick a belief of yours -- any old belief; call it the belief that Q. Suppose you say that Q is justified. OK, then there's something that justifies it. Suppose you want to say it is justified by another belief, P. But in that case, you'd have to know what justifies P. And what justifies that. And so on. You can't have an infinite regress of justifying beliefs -- you don't have an infinite number of beliefs supporting any given belief of yours! So there's another choice; you could say that the belief justifies itself somehow; either directly, so that Q justifies Q; or indirectly, so that Q justifies R, S, and so forth, down to Z say, and Z then justifies Q. The whole set of beliefs then forms a circular chain of justification. That's at least a possibility. And there is one other possibility, namely, that Q is justified not by another belief, but by something else, something that isn't a belief. We'll have to see what that something else might be later. Anyway, all this is in explanation of what premise (4) says.

Now to premises (5) and (6). These two premises eliminate options (a) and (b), respectively, leaving us with option (c). To begin with premise (5), which eliminates option (a). Premise (5) says: "Possibility (a) (called regressism)" -- which is the possibility that the chain goes on forever -- "is obviously incorrect." In other words, it is obviously wrong to say that a chain of justification goes on forever. I could but will not elaborate; virtually no philosophers take it seriously. I think you should be able to see generally why this is so obviously wrong, though. I mean, for one thing, we don't have an infinite number of beliefs.

Now for premise (6), which eliminates option (b). Premise (6) says: "Possibility (b) (called coherentism)" -- which is the possibility that the chain loops around on itself, forming a circle -- "is incorrect." Now, coherentism is taken much more seriously than regressism. So I'm going to discuss coherentism at greater length.

Remember that the word "coherentism" can mean either a theory of truth or of justification. Here is a definition of the coherence theory of justification, or coherentism for short:

A belief is justified iff it is part of a coherent system of mutually supporting beliefs (i.e., beliefs that support each other).

A common analogy used to illustrate coherentism is an arched doorway. Each of your beliefs is like a brick in an arched doorway; all the bricks support each other, and if you chip away at one brick, the arch becomes much weaker or even comes crashing down. But when the bricks are all intact and support each other, the doorway stands up; in the same way, when your beliefs all support each other, they are all justified, except of course for those of your beliefs that don't cohere with your other beliefs.

Now look again possibility (b) from premise (4): that the chain of justification loops around on itself, forming a circle. That is essentially what happens in a coherent system of mutually supporting beliefs. So let me try to give you an example of part of a coherent system of beliefs. I believe that we are at Ohio State; I believe that University Hall is on the Ohio State campus; I believe that I have seen many papers, booklets, and other materials indicating that I am a philosophy instructor at Ohio State; I believe I am seeing you all right now in front of me, and that you are students; moreover, I believe you are Philosophy 101 students, and that I am your teacher; and so on. There is a whole bunch of information I have about you, about myself, about this place, about what we're doing, and so forth. These beliefs are all mutually supporting. What does that mean? For example, the belief that I am teaching you Philosophy 101 supports the belief that I have seen many materials indicating that I am a philosophy instructor; and vice-versa, the belief that I am a philosophy instructor supports the belief that I am teaching you Philosophy 101.

And of course this is only one small portion of what I believe. I believe things not just about Ohio State, my role as a teacher, and so forth. I also have many, many beliefs about my own personal history, about the political system of the United States, about the natural world, about any number of academic subjects, and of course about philosophy itself. And it's not like these beliefs are in some nice little boxes, separated from each other. It's not like I have a set of beliefs about economics, and those beliefs do not support, and are not supported by, my beliefs about our political system, and our legal system. No, my beliefs about virtually everything are interwoven into a truly massive web of belief. It's incredible that all those beliefs could somehow be stuffed into my brain; but of course they are.

Now let's ask the coherentist a question: What justifies an individual belief in my web of belief? Here is the coherentist's answer: Virtually all the other beliefs in the web of belief. All your beliefs are interconnected; if your belief system is coherent, then each of your beliefs makes each other belief, to some extent, more plausible, more rational to believe; that's what it means to say the system is coherent. If you wanted to, you could trace justification relations from my belief that Timbuktu is in Africa, to my belief that relativism about truth is false -- and you could trace the justification relations back again. And what that means is that, ultimately, every belief in my web of belief supports itself. Not directly. Our beliefs don't support themselves directly -- we don't say, "P justifies P," for example. But every belief in my web of belief supports itself indirectly. So P justifies Q, and Q justifies R, and R returns to justify P.

Quite a few philosophers, especially in the past forty or fifty years, have found the coherence theory of justification very plausible. But there have been a lot of powerful objections to coherentism. I want to review three of the objections.

The first objection is that coherentism seems to imply that circular justification is just fine. There's nothing wrong with saying that P supports P, ultimately. But that seems wrong. Think of it like this. If you say, for example, that P justifies Q, and Q justifies R, and R justifies P, then it seems like you're saying that you could argue: P, therefore Q; Q, therefore R; R, therefore P. If all that is true, then we could argue: P, therefore P. Well, that's obviously a fallacious sort of argument; it's called "begging the question" or "arguing in a circle." But coherentism seems to imply that such an argument is just fine -- that there's nothing wrong with it. But there is something wrong with it; therefore, coherentism must be rejected.

That is one objection to coherentism. The way that coherentists typically reply to it is by saying that if the circle is big enough, then circularity isn't a problem. If you've got a thousand different beliefs in a circle, then the fact that #1 supports #1 via the other 999, well, that's not really a problem. My reply (which others make) is that I just don't see why the size of the circle should make any difference. A circular argument is a circular argument. It doesn't matter if there is one intermediate belief or a thousand. It's still a fallacy!

Well, whatever you think of that, here is a second objection. According to coherentism, a belief is justified if it is part of a coherent system of beliefs. But isn't it possible for there to be a coherent system of beliefs that is totally divorced from reality? Let me explain. P can cohere with Q and R, of course, without P, or Q, or R, being true. The coherence of a few beliefs does not imply that any of the beliefs in question are true.

Think of the conspiracy theorists -- you know, people who think that there is an enormous interlocking directorate of semi-secret organizations, corporations, government agencies, foundations, universities, and powerful individuals who have together conspired to take over the world and make it into a totalitarian one-world system. Now I tend to think that people who advocate this theory are probably wrong and at least a little bit paranoid. But if you listen to them, they definitely have a coherent body of beliefs. One part of their system of beliefs, for example their beliefs about semi-secret organizations, is quite consistent with, and strongly supports, their beliefs about other parts of their system of beliefs, for example their beliefs about corporations. So I think we might grant that the conspiracy theory is coherent. But does that mean that belief in the conspiracy theory is justified? Does mere coherence make a body of beliefs justified?

Well, no, I don't think so anyway: it has to be grounded in, based on, derived from experience. If your coherent system of beliefs ignores experience it's bound to be wrong and unjustified. So here's the point: It's perfectly possible to have a coherent body of beliefs that is not grounded in experience, and hence not justified.

This leads to the third objection. My third objection is this: There are some beliefs, namely beliefs that arise immediately from experience, which appear to be justified, but not all of their justification comes from other beliefs. In other words, some beliefs we have from seeing and hearing and other kinds of experience; and those experience-based beliefs are very well justified; but clearly, not all of their justification comes from other beliefs.

Let me give you an example. Say you are looking through a keyhole into a room; you have absolutely no idea what's in the room. Unfortunately, the lights are off in the room so you can't see anything. But then, suddenly, the lights turn on for a second; and you can very clearly see a white canopy bed in the room. You don't notice anything else. So then you form a belief. You believe: There is a white canopy bed in the room. Now I think we'd all want to say that this belief is justified. There's no reason to think there's anything wrong with that belief. It might turn out to be false; but if you looked through a keyhole into a room and the lights were turned on for a second, and it definitely seemed that you were seeing a white canopy bed, then by golly you are very well justified in your belief that there is a white canopy bed in the room.

But you don't know anything about the room, what kind of room it is, whether it has windows, and so on. So you don't have any beliefs that would be enough to justify your new belief, that there's a white canopy bed in the room. But that new belief is justified. So justifies it? Obviously, it's a piece of sense-experience. Namely, you saw the bed! Here's the justifier: you saw the bed through the keyhole. Here's the belief it justifies: There is a white canopy bed in the room. So there's a belief that is justified, but not justified by another belief.

Now remember what coherentism says: it says that if a belief is justified, then what justifies it is a bunch of other beliefs, in a belief system. But only other beliefs. I think you should be able to see that that is rather implausible now: it certainly looks as though there are some beliefs which are not justified by any other beliefs. Rather, they are justified directly by experience. If that's true, then coherentism must be wrong. Coherentism says that there are no beliefs that are justified by anything other than beliefs.

So let's go back to the regress argument. It says we have options (a), (b), and (c). We have eliminated options (a) and (b). Only (c) is left. As you will recall, option (c) reads as follows: "the chain" -- the chain of justification -- "begins with a belief that is justified, but which is not justified by another belief." This, we said, is called foundationalism. So here is what the foundationalist theory of justification says:

A belief is justified iff (1) it is a basic/foundational belief (i.e., it is justified by a nonbelief), or (2) it is justified by a basic belief or beliefs, or (3) it is justified by a chain of beliefs that is ultimately justified by a basic belief or beliefs.

To understand what foundationalism says, we might as well use another analogy. And the analogy explains why it's called "foundationalism." You remember that enormous belief system I have, that we were just talking about -- about myself, about the natural world, about our socio-political system, about philosophy, and so forth? Compare that belief system to a building. Ordinary individual beliefs occupy the upper stories of the building; basic, or foundational beliefs are down in the basement, in the foundation of the building, holding everything else up. In a similar way, individual beliefs, say about economics or ethics, rest on more basic beliefs, say about the nature of human beings; and those rest on still more basic beliefs, say about the mind; and in the end the entire system rests on a set of beliefs, basic beliefs, which are not justified by other beliefs, but instead by something else.

So what is essential to foundationalism? It's the view that there are basic beliefs, foundational beliefs. And what exactly are basic beliefs? Here is a definition:

A belief is basic iff it is justified, but it is not justified by other beliefs.

So if a belief isn't justified by other beliefs, then what could it be justified by? I already introduced one source of justification: sense-perception. It seems to me I see a white canopy bed through the keyhole, and that's not a belief, it's an event of perception. Or, if you believe representational theory of perception, then you could say that it is a sense-datum! And that's what a lot of foundationalists have said, classically: that sense-data justify our basic beliefs. At the foundation of knowledge, they say, is a whole body of sense-data, or what you and I call "experience." And all beliefs are ultimately justified by sense-data.

Well, we've already looked at the representational theory of perception and I, at least, found it doubtful. And in any case you don't have to believe in sense-data in order to be a foundationalist. Really all you have to believe is that it's something that isn't a belief that justifies a basic belief. Doesn't have to be a sense-datum. It could be, as I say, an event of perception.

Now, you might object: What if your supposed event of perception wasn't an event of perception, but instead a hallucination? A hallucination couldn't justify anything. Well, that's a problem we are going to have to face later. Believe it or not it's a very deep and difficult problem: the whole problem of skepticism.

Now how about some arguments in favor of foundationalism? Well, there is, first of all, the regress argument. But all the regress argument says is: well, it looks like regressism and coherentism are out; so we're stuck with foundationalism, apparently. So are there any positive arguments for foundationalism? Some argument for foundationalism that wouldn't involve simply saying that the other theories are wrong so this is what we're left with?

What we want then is a positive argument that there are basic beliefs. And remember what that means: beliefs which are justified, but not justified by other beliefs. Well, we have given an argument for foundationalism, then, haven't we? I mean the third argument against coherentism. It stated that there is at least one belief that is justified, but which can be justified only by reference to sense experience, and not to other beliefs. Sense experience is, after all, a different mental process from believing -- even if we do typically believe almost everything we perceive.

I'm sure you can think of other examples. Here's an example of how memories can be justifiers. I'll bet you can remember some fairly minor childhood event in your life, which no one has any records for at all, and no one but you can remember. Say you fell of your bike and scraped your knee. Say you remember this vividly. So you believe you did fall off your bike and scrape your knee when you were a child; you have a belief that that occurred. Now, granted, you might be misremembering. But surely you're justified at least a little bit in believing that that childhood event occurred. And what justifies you in believing it? The fact that you remember it.

You might object: well no, it's my belief that I remember it that justifies my belief that the event occurred. In other words, make a distinction. On the one hand there's your memory of the event; on the other hand, there is your belief that you remember the event. And you might tell me: what justifies my belief that the childhood event happened is that I think I remember the event. Not that I do remember the event.

Well, suppose that that is true. So we'll suppose that your belief that Q, the belief that you fell off your bike as a child, is justified by another belief, namely the belief that you remember falling off your bike as a child. But then I'll just raise a question about the latter belief: what makes you justified in believing that you remember falling off your bike as a child? I mean, if you really are justified in believing you remember that? Well, what else could it be, besides your having a mental state that was a memory. And the remembering is itself not a belief; even if it results in a belief, there's a difference between remembering and believing. So in any case, what ultimately justifies you in believing that you fell off your bike as a child is that you remember falling off your bike as a child. So it is a nonbelief, a memory, that ultimately justifies you in holding that belief.

Now again, you might have a skeptical reaction: you might wonder, "What if I was misremembering? Then I'm not justified in my belief." Well, that might be the case. That's the sort of problem we're going to have to deal with when we talk about skepticism next week. But in any case, it does seem that if you're not misremembering, if your memory is reliable, then it can justify your beliefs in past events.

It's at this point that I should raise a bunch of objections to foundationalism I suppose. One objection, as I have already hinted, is skepticism. But there have been a lot of other objections to foundationalism; almost all of them, however, presuppose that the ultimate justifiers are sense-data, rather than events of perceiving, remembering, and so forth, themselves. In other words, it has been very fashionable to bash foundationalism, in the past thirty or forty years. But these criticisms have been of representationalist versions of foundationalism. They also presuppose that foundationalism says that the foundations are absolutely certain and not open to any question. But nothing I've said should lead you to believe that you have to be a representationalist, in order to be a foundationalist. Nor should anything I've said lead you to believe that foundational beliefs have to be absolutely certain and unrevisable. For all I've said, those foundational beliefs might be merely probable. And I think that's definitely the case: there are lots of basic beliefs that are not absolutely certain. But they still do lend weight, at least some weight, to other of our beliefs. That's all foundationalists have to claim.

So actually, in order to criticize foundationalism, in the bare-bones version that I've presented it here, we'd have to get pretty technical. And we don't have the time to get so technical. So let's just take skepticism as our main objection to foundationalism; and that, as I said, is something we'll deal with next time.

I said that there are three different possible kinds of justifiers, right? They were: beliefs; non-belief mental states; and other sorts of things. The other sorts of things I haven't discussed. We don't have time to go into this in much depth, but I just want you to be aware of something. There are some kinds of foundationalists, called externalists, who say that basic beliefs are justified by processes or states in the world that aren't even conscious mental states. In fact, we may not even be aware of some of the things that make us justified in our beliefs! Or so say the externalists. Externalism is opposed to internalism. So let me define these two terms:

Externalism is the view that we need not have access to the justifiers of our beliefs.

Internalism is the view that we do always have access to the justifiers of our beliefs.

The debate between externalists and internalists has been raging in recent years -- the last few decades. On the one hand you have the externalists saying, for example, that I can be justified in believing that Stockholm is the capital of Sweden even if I have totally forgotten any reason I originally had for believing this; I'm not aware of the justifiers of this belief, but it can still be justified. On the other hand you have the internalists who counter that if you're going to be justified in any belief, you've definitely got to be at least able to become aware of what justifies your belief. If you don't have access to your justifiers, say the internalists, then you aren't really justified.