Lecture Notes, UC Davis Philosophy 102, Theory of Knowledge

Foundationalism

G. J. Mattey, Senior Lecturer

William Alston

Text: William Alston, "Two Types of Foundationalism" (1976)

We have seen that Descartes proposed to place knowledge on a new foundation. Most commentators have taken that foundation to be the indubitable truth of one's own existence, when one is contemplating it. Descartes stated that "I shall have the right to conceive high hopes if I am happy enough to discover one thing only which is certain and indubitable." What made him so sanguine was the fact that Archimedes had said that in order to move the earth, he needed only one point that is fixed and immovable.

With an indubitable truth in hand, Descartes moved on to declare that whatever he perceived very clearly and very distinctly is true. We can say that the class of propositions which are known through clear and distinct perception forms the foundation for his other knowledge. This knowledge is generated through the use of deductive inference from what is foundational. To make sure the derived truths are as secure as the foundational ones, Descartes had to show that our memory of previous clear and distinct perceptions is trustworthy. To do this, he thought he had to prove that God exists and that God is no deceiver.

Modern Foundationalism

Alston presents us with an updated account of foundationalism. There are several differences between his account and that of Descartes. First, the modern account is stated in terms of justified belief, rather than knowledge. What lies at the foundation is beliefs, and what is supported by foundational beliefs is other beliefs. The supported beliefs are said to be justified by the foundational beliefs.

Second, the relation of support is understood in terms of justification rather than deductive proof. So one belief supports another insofar as the former justifies or provides justification for, the latter. Justification can include other modes of support besides logical deduction. It may be non-deductive, so that the truth of the supporting beliefs is compatible with the falsehood of the beliefs that are supported by them.

Third, neither the supporting beliefs nor the supported beliefs need be indubitable. As for foundational beliefs, they do not have to be indubitable or certain. If they are uncertain, than what they support will inherit some uncertainty from them. And the fact that the support relation can be non-deductive introduces a further element of uncertainty. Beliefs supported by certain beliefs may nonetheless themselves be uncertain. And beliefs supported non-deductively by uncertain beliefs become doubly uncertain.

Alston's Formulation of Foundationalism

For Alston, foundationalism is a doctrine about the structure of justification. In this structure, some beliefs do not depend for their justification on other justified beliefs. These will constitute the foundations. Note that in phrasing it this way, Alston leaves open whether foundational beliefs are justified by their relation to other beliefs which are unjustified, and whether they may be justified by their relation to something other than beliefs (perhaps sense-experience). Alston calls foundational beliefs "directly (or immediately) justified." The account of being directly justified, then, is not very restrictive.

Indirectly (mediately) justified beliefs comprise the rest of the justified beliefs. Every indirectly justfied belief depends for its justification on at least one justified belief. What does the justifying is called "reasons or grounds." Correspondingly, if a directly justified belief is a case of knowledge, then the knowledge is direct, while it an indirectly justified belief is a case of knowledge, then the knowledge is indirect.

The justification of indirectly justified beliefs will have a tree-structure. We begin with the belief at issue, and trace its grounds back to other justified beliefs. If any of those beliefs is directly justified, then the tree terminates. If not, then the intermediate indirectly justified belief is then traced back to its grounds. The process ends when every ground is a directly justified belief.

The precise formulation of this account of foundationalism is as follows:

(II) Every indirectly justified belief stands at the origin of a (more or less) multiply branching tree structure at the tip of each branch of which is a directly justified belief.

Aune's Argument Against Directly Justified Belief

Having sketched out an account of foundationalism, Alston turns to an anti-foundationalist argument presented by Bruce Aune, a retired professor at the University of Massachusetts, Amherst. Aune was a student of Wilfrid Sellars at the University of Minnesota. Sellars was one of the most celebrated opponents of foundationalism in the middle of the 20th century.

Aune's argument is directed at the view held by the "empiricist" that there must be "indispensable basic premises" if there is to be empirical knowledge. The reason is that "empirical truth" is established by arguments, and the conclusion of an argument is only as good as its premises. "If the premises you start with are false, you will have no guarantee that the conclusions you reach are not false either." For there to be such a guarantee, "you must ultimately have premises whose truth is acceptable independently of any inference and whose status is accordingly indubitable." These ultimate premises are called "intrinsically acceptable."

As Alston notes, intrinsically acceptable premises are different from what he himself calls directly justified beliefs. There are two chief differences. The first is that intrinsically justified beliefs must in some sense justify themselves. But as we have noted, Alston's account of foundationalism requires only that directly justified beliefs not be justified on the basis of other justified beliefs. The second difference is that intrinsically acceptable premises are supposed to be indubitable. This is said to follow from the fact that they need no argument to support them. Alston has no such requirement for directly justified beliefs.

With these qualifications in mind, we can look at Aune's basic argument. He thinks that the standpoint of common sense shows that there are no intrinsically acceptable beliefs. For empiricists, here are two leading candidates for such beliefs. One is introspective beliefs: beliefs about one's present state of mind, such as that I am in agonizing pain right now. The second is "the observation claims of ordinary life," such as that I am sitting in my dressing gown by the fire (Descartes).

What makes introspective and observational claims acceptable, according to Aune, is more than the mere fact that they are made. What is important is the context in which they are made. If we assume that the person making them is normal and reliable, then the claims are acceptable. But in that case, they are not intrinsically acceptable. Rather, they are acceptable due to "our confidence that a complex body of background assumptions . . . and, often, a complex mass of further observations all point to the conclusion that it is true."

Alston responds by noting that this argument against intrinsically acceptable premises does not apply to directly justified beliefs. The reason is that for him, beliefs are justified for the very person having the belief. But for Aune, if he is talking about justification, it would be justification for another person trying to evaluate someone else's claims. This involves "a higher level of belief to the effect that someone is justified in believing that p." But one need not rise to this higher level in order to have beliefs of one's own that are justified without being justified by any other justified belief.

Iterative Foundationalism

So Aune's argument is not an obvious threat to foundationalism. It would become one, though, if a further assumption is made: that direct justification is "iterative." That is, when one is directly justified in believing that p, one is thereby directly justified in believing that one is directly justified in believing that p. A foundationalism which mades this assumption is called by Alston "iterative foundationalism," as opposed to the "simple foundationalism" which he initially introduced.

It is pretty obvious how iterative foundationalism is threatened by a "second level argument" which has its kernel in Aune's argument. "Justified" is an evaluative term, and so the proposition that one is justified in believing that p is a proposition which evaluates the belief. What makes a belief justified is that it has what Roderick Firth terms "warrant-increasing properties." For example, a belief may be justified if it is produced by a reliable process of belief-formation. Determining whether it has such a property Q requires further information: both that the belief has Q and that Q is is a warrant-increasing property. So, the belief that one is justified in believing p is not directly justified.

Some philosophers have held that if S is justified in believing that p, then one must also be justified in believing that one is justified in believing that p. (We might call this the "JJ thesis," on an analogy with the "KK thesis" regarding knowledge: if one knows that p, then one must know that one knows that p.

Alston finds the JJ thesis is not plausible in itself. But even if it were true, it is consistent with the conclusion of the second-level argument. Everything depends on what it means to for one to have the higher-level justification. Perhaps all that the JJ thesis says is that being justified in believing that p only gives one grounds for being justified at the higher level. In that case the higher-level justification could be mediate.

A second point Alston makes about the conclusion of the second-level argument is that it does not require that all our justified beliefs about what is justified ("epistemic beliefs") be the consequence of conscious inference. Nor must one be conscious that the lower-level belief has the warrant-increasing property Q. The reason this is important for Alston is that most of the time we do not make such inferences or have the information about our lower-level beliefs. If we want to avoid the conclusion that we "have precious little mediate knowledge," we have to allow that "one's grounds can be possessed more or less implicitly."

The Regress Argument

The argument presented by Alston in favor of foundationalism, which he calls the "regress argument," dates back to Aristotle. The thesis which the argument is supposed to establish is that:

The original belief is mediately justified only if every branch in the tree structure terminates in an immediately justified belief. Hence every mediately justified belief stands at the origin of a tree structure at the tip of each branch of which is an immediately justified belief.

Basically, there are three alternatives with respect to the nodes downward from the original node of a tree. The first (b) is that one of its downward nodes is unjustified. The second (c) is that somewhere down the line, the original belief is encountered, making the justification circular. Finally, (c) has it that some branch does not terminate at all, but proceeds through infinitely many nodes.

So we are left with four cases. The positive case (1) is made that if all the branches do terminate at beliefs which are directly justified, then the original belief is justified. The justification is transferred back upward from the directly justified beliefs to the original, and the regress is ended.

The first negative case (2) is that justified belief cannot be based on belief that is unjustified. So if there is a branch that ends with an unjustified belief, the entire tree is corrupted, and there is no justification of the original belief. (There may, however, be justification of intermediate beliefs all of whose trees terminate in a directly justified belief. This point holds for the other two negative cases.)

The second negative case (3) is that a belief cannot be justified on the basis of itself. There is a chain of justification from the original belief back to itself, so that all we have is the true but trivial fact that if p is justified, then p is justified. If we are entitled to assert the antecedent, we must have already been in a position to tell that p is justified.

The final negative case (4) is that a belief cannot be justified if there is no end to the tree of justification on one of its branches. In the absence of a stopping point, there is no mediate justification.

The regress argument supports simple foundationalism. The regress itself is stopped simply by there being the requisite directly justified beliefs. It says nothing about any justified belief in there being such things. So the argument as it stands does not support iterative foundationalism.

Iterative Regress Arguments

One might try to generate regress arguments to establish the truth of iterative foundationalism. There are two strategies in the literature to this end. The first would take as original the higher-level belief that one is justified in believing that one is justified in believing that p. The second is to change the focus from being justified in a belief to the activity of showing that a belief that p is true.

Let us recall what iterative foundationalism is. It is the thesis that necessarily, along with directly justified beliefs of simple foundationalism, we also have directly justified beliefs about those directly justified beliefs. So any regress argument for this claim would have to hold that there is some epistemically undesirable consequence if we do not have directly justified higher-level beliefs about our directly justified beliefs.

Chisholm makes an argument that in order to justify indirectly claims to know, by appeal to other things we know, we need to be able to justify directly our claims to know. There must reach a an end-point in the tree of second-level justification where for some proposition n, "What justifies me in thinking I know that n is true is simply the fact that n is true." If there were no such stopping point, then the second-level claim would either be unjustified, circular, or infinitely unattainable. We can call this end-point an immediately justified "epistemic" belief, because it is a belief about what we know to be true. Note that the fact that n is true is not itself the end-point, because it is not a belief.

Alston asks why the end-point of the justification of an epistemic belief must itself be an epistemic belief. Take for example, the claim that I know that it is dark outside now. This claim might be justified by appeal to my knowledge that it is now 10:27 P.M. in the autumn in California, and that it looks dark outside. Then it will be asked why I know that it looks dark outside. According to Chisholm, I would have to appeal to the brute fact that it looks dark outside to me.

But why not say that I know that it looks dark outside because I believe it looks dark outside and that my belief is justified. Then the claim that I believe it looks dark outside could be taken as foundational. It is justified, but not justified by any other belief. Rather, it is justified by a psychological state I am in. This belief about how it looks outside to me is not epistemic, yet it seems to be the sort of belief that is a natural-stopping point on the downward path of the tree of justification.

Alston does not put it this way, though. He points out that, for example, knowledge of one's perceptual experience justifies claims to knowledge of the present perceptual situation. The two kinds of knowledge are different, but one is the foundation of the other. This example may be misleading, though, because it looks as if there are epistemic beliefs involved at each level. Alston does seem to be right in claiming that no reason has been given why the foundational beliefs that support epistemic beliefs must themselves be epistemic. The general conclusion is that Chisholm's version of the regress argument does not make the case for iterative foundationalism.

The second sort of regress argument is generated to establish that foundationalism is required in order to show that one is justified in believing that p. The assumption is that in order to be justified in believing some p, one must show that he is justified in believing it. And this requires a foundational epistemic belief. So then the question becomes why showing that one is justified requires directly justified epistemic beliefs.

In order to show that p, one must find some proposition q that is true and that constitutes adequate grounds for believing that p. There is no regress generated by the first requirement. "Even if no proposition can be true without some other proposition's being true, there is nothing repugnant about the notion of an infinity of true propositions." Alston strangely does not mention the second requirement. Instead, he brings up a third, that one is justified in believing that q. This does generate a regress, but there is no reason to think that it must terminate with directly justified epistemic beliefs.

So a fourth requirement is brought forth. According to this one, to be justified in believing that p, one must be justified in believing that one is justified in believing that q. Alston has already stated that he does not think this is a legitimate requirement. Moreover, even if a regress is generated, it would be of the type we just discussed in connection with Chisholm, which was found not to support iterative foundationalism.

A final proposed requirement is that not only must a true proposition q be adequate grounds for p, but (5) one must also be able to show that q. This gives rise to a regress in the activity of showing. Either you arrive at something you do not show to be true, or your showing involves a circle, or there is an infinite regress of showings. So the only way I can show that q is that I "arrive at a proposition that I can show to be correct without appealing to some other proposition." Whether this supports iterative foundationalism depends on whether the "showing" requirement is a real requirement, and whether the regress is stopped by a directly justified epistemic belief.

Alston argues against (5) as a requirement for the justification of p. To show that p, I would invoke q, where q is an adequate ground for p. Suppose I want to show you that it is raining. Then I might point to the window and claim that there is water falling from above. You might protest that I must show you that there is water falling from above. I would respond that this is needed only to show you that I have shown you that it is raining. To put the matter slightly differently, if you challenge what I use to prove something to you, you are challenging my claim to have proved it, and so you are asking me to prove that I have proved it to you. "Can't I prove a theorem in logic without being able to prove that I have proved it? . . . Similarly, it would seem that I can show that p, by adducing true adequate grounds I am justified in accepting, without being able to show that those grounds are true."

A deeper problem lies in the end-point itself. Suppose that a regress is ended by a directly justified belief about what I know or am justified in believing. The fact that I have such a belief does not constitute a showing that the epistemic belief Z is true. To show that belief Z is true, I would have to adduce further grounds. The activity of showing generates a regress that cannot be stopped, because to show something, you always have to bring in something else. So either I will stop at something that is not shown, or I will use what I originally tried to show as evidence for itself, or an infinite regress will be generated. Alston concludes that a regress based on showing cannot possibly generate an argument in favor of iterative foundationalism.

Answering Skepticism

The preceding argument is taken by Alston to show that simple foundationalism is the only viable form of foundationalism. But is simple foundationalism of any value beyond being a way to stop the regress argument? The first use to which Alston puts it is in giving an answer to skepticism. The kind of skepticism it might answer is not extreme skepticism, in which the skeptic questions every proposition put forward for acceptance. There is no way to defeat extreme skepticism, and Alston thinks it is futile to try.

A second kind of skepticism is limited. For example, a skeptic might ask how I know that minds other than my own exist. I appeal in response to things that I observe: sounds that I hear and interpret as linguisitic communication, actions that I observe and interpret as intentional, etc. Whether or not knowledge has foundations is not relevant here. What is important is the way in which my observational knowledge (which may itself be indirect) supports claims to knowledge of other minds.

The kind of skepticism that might be answered by appealing to the foundational structure of knowledge "is that in which the skeptic doubts that we have any knowledge, a successful answer being a demonstration that there is some." So how do we demonstrate that we have knowledge that p? There are two main conditions. (1) p has to be true if we are to demonstrate that it is known. (2) the demonstration proceeds by citing another proposition q in support of p. This proposition must be (A) true, (B) an adequate ground for p and (C) itself justified.

The skeptic rightfully demands that the credentials for q be given. If this means that q must be demonstrated or shown, then there is no answer to the skeptic, as was seen above. So Alston looks for what might satisfy a "reasonable skeptic" but which falls short of showing that q is true for every q. Rather q must be shown to be true only when there is some real doubt about q in the mind of a "reasonable person who has reflected on the matter." If we can find some ultimate grounds about which there is no real doubt by a reasonable person, then there is a prospect for showing that p.

The issue is whether iterative foundationalism is needed to show that S knows that a. It is not needed to satisfy the first condition, i.e., that it be true that S knows that a. It was argued earlier that knowledge in general does not require any directly justified epistemic beliefs. If might be held that when S knows that a, S thereby knows that S knows that a. This does not require any directly justified epistemic beliefs either, since, as was claimed before, it may require only directly justified non-epistemic beliefs.

The issue of whether the grounds q are true does not depend on my knowledge of those grounds, so it could not require directly justified epistemic beliefs. The adequacy of the grounds q is likewise independent of any of my knowledge, as it is a relation between propositions. As a result it does not call for there being directly justified epistemic beliefs. Being justified in believing that q does not, as argued earlier, require directly justified epistemic beliefs, either.

The last thing that might require directly justified epistemic beliefs is the requirement that S is able to show the truth of anything about which there is real doubt. There is nothing in this that is different from showing that S knows that a. So none of the conditions for showing that S knows that a requires directly justified epistemic beliefs, and as a result, iterative foundationalism plays no essential role in showing that we have knowledge. If we cannot show that S knows that a, it is not for lack of directly justified epistemic belief.

Reconstructing Knowledge

A second answer to the skeptic that might require iterative foundationalism is the attempt to place all our knowledge on the ultra-secure foundation of certainty. We need to be justified, directly, that some of our beliefs are known with certainty in order to place it at the foundations of other beliefs. The reason the epistemic belief must be directly rather than indirectly justified is that we are building the foundations of our knowledge, and there is nothing to justify the belief indirectly at this point of the investigation.

What is described here is a transcription of the program of Descartes. As has already been mentioned, Descartes worked with a concept of knowledge, rather than justification. A second difference is that Descartes tried to prove that his foundational beliefs are certain. This led him into one of the dead-ends of the regress argument, in that (it appears) his justifications of his epistemic beliefs are ultimately circular. (This is a very controversial point that is by no means settled).

The problem with the process of building an edifice of knowledge from scratch, even with the aid of post-Cartesian insights, is that it depends on an errant notion of iterative foundations. Alston has argued that there is are "crushing objections" to it. It generates a regress that cannot be stopped in the way that simple foundationalism stops its regress. If we can't attempt a reconstruction of knowledge from nothing, what is the loss? To be sure, we want to know what our foundations are, but we can draw on our knowledge and justified beliefs to determine this. There is no need for ultimate epistemic beliefs.

Dogmatism

Alston describes simple foundationalism as a way to stop the regress problem without resorting to a stopping-point that is not justified. Instead, it stops with the directly justified. If one were to base one's indirect justification on what is unjustified, one would be believing dogmatically. So simple foundationalism promises an antidote to dogmatism. Alston raises a final challenge regarding his solution: that simple foundationalism "must allow that some beliefs may be accepted in the absence of any reasons for supposing them to be true." If it does, then it is open to the charge of dogmatism.

The idea behind this criticism is that the notion of direct justification is very skimpy. A directly justified belief has some justification, but the justification is not based on any other justified belief. If it is not based on any other justified belief, then when it is believed, it would seem that it is believed for no reason, since presumably a reason would be a justified belief.

Alston responds that even if S does not have a reason for accepting that p, S may have reasons to accept that he is immediately justified in believing that p. This second-level epistemic belief is not subject to the negative constraint that stops the regress on the first level. If one has this second-level epistemic belief, then we should not say that the acceptance of p is arbitrary or dogmatic. "The curse (of dogmatism) is taken off immediate justification at the lower level, just by virtue of the fact that propositions at the higher level are acceptable only on the basis of reasons." What makes the belief justified directly is that it satisifies some epistemic principle or principles of justification. Recognizing that the belief satisfies the epistemic principle(s) requires reasons to think that it does.

We may contrast the reasons one has for accepting a directly justified belief with those which one has for accepting an indirectly justified belief. The first kind of reasons can only be found at the higher level, as we have just seen. The second kind of reasons are found at the lower level: they are the justified beliefs which support the indirectly justified beliefs. There may also be higher-level reasons to think that one has an indirectly justified belief.

It may be that, for any given epistemic proposition, one will fail to find reasons to accept it. Indeed, this may be the cae for any proposition. But the anti-dogmatism argument is supposed to show that there is nothing in principle to stop one from finding a reason for the belief that a belief is directly justified.

In demanding that every belief have a reason for its being accepted, the critic is appealing to an ideal. In ordinary life, we should not demanded to produce a reason for every belief we have. All the critic is asking is that a theory of justification make having a reason possible. And simple foundationalism does make it possible.

The critic could finally maintain that the regress ends in dogmatism when one attempts to show that p. To do this, one must give grounds for believing that p, and that this process must end with some foundation f which cannot be shown to be true. All that can be done is to assert its truth. But Alston has argued that more can be done than mere assertion, namely, to establish the higher-level belief that one is directly justified in believing that f. This is all we can ask, unless we are going to surrender to the skeptic by requiring that no matter what grounds are given at what level, grounds must still be given for them. A final point is that it is not up to the foundations theorist to say exactly where the end-points of justification may be. He just has to allow that there may be some.

[ Course Home Page | Menu of Lectures ]