1. Some philosophers formulate the Epistemic Regress Argument in terms of knowledge, rather than in terms of justified belief. This is true, for example, in the case of Lawrence BonJour, in his article "A Critique of Foundationalism", American Philosophical Quarterly 15 (1978), pp 1-13, and reprinted in Louis Pojman (ed.), The Theory of Knowledge, Third edition, (Belmont, CA: Wadsworth, 2003), even though the only aspect of the concept of knowledge that enters into BonJour's formulation of the argument is that the belief be justified.
2. An exposition of the Epistemic Regress Argument:
First, there seem to be only four possibilities with regard to inferentially justified beliefs:
(1) The regress of justification is a finite one, with the justification of a given belief terminating in one or more beliefs that are justified, but without being justified on the basis of any other beliefs. So justification terminates in noninferentially justified beliefs.
(2) The regress of justification is infinite: every inferentially justified belief is justified on the basis of one or more beliefs that are also inferentially justified – and only inferentially justified – and no belief ever occurs twice in the resulting sequences of justified beliefs.
(3) Every inferentially justified belief is justified on the basis of one or more beliefs that are also inferentially justified, but there is no infinite regress, since any such sequence of inferentially justified beliefs contains only a finite number of members, and one is thus always led back either to the first belief, or else to some later belief, in the chain of justification. Inferential justification, in short, involves one or more circles or loops.
(4) The regress of justification is finite, and one is never led back to the starting point. Nor does the regress terminate in noninferentially justified beliefs. It terminates, instead, in one or more beliefs that are not justified – either inferentially or noninferentially.
Secondly, the argument then runs as follows:
(1) If possibility (4) obtained, then the belief in question would not be justified. One cannot arrive at justified beliefs on the basis of unjustified beliefs.
(2) The infinite regress required by possibility (2) seems most unlikely – at least for humans – but, even if there were such an infinite regress of beliefs justified relative to other beliefs, it would not suffice to justify any of the beliefs in the regress.
(3) Justification involving a circle is unacceptable for essentially the same reason.
(4) Hence the only way to avoid a skeptical conclusion concerning the possibility of inferentially justified beliefs is possibility (1): justification must terminate in beliefs that are noninferentially justified, beliefs that are justified without being justified on the basis of evidence.
Comments
1. If a belief is not justified unless its probability is above a certain number k – say, ½ – then the claim just made would appear to be false.
Reason
There can, it would seem, be cases where there are a number of beliefs, none of which is more likely to be true than false, but which nonetheless render some other belief likely to be true. For consider the following:
B1: Mary told me that p;
B2: John told me that p;
B3: Bruce told me that p;
.....
Bn: Suzanne told me that p.
Suppose now that I am in a cognitively impaired state – perhaps I am slightly drunk – such that I am somewhat inclined to believe each of these – it seems to me that it may be that I remember the corresponding event – but, given that I am aware of my present, cognitively dubious state, I form the belief, concerning each potential memory, that it probably is not reliable. Nevertheless, though each individual potential memory is probably mistaken, what about the following, disjunctive belief:
B: Either B1 or B2 or B3 or ... Bn ?
For this belief to be mistaken, all of B1, B2, B3, ... Bn would have to be mistaken, and while it is likely that each one is mistaken, the probability that they are all mistaken might very well be much less than one half. So it may be likely that B is true, even though none of the Bi upon which it is based is likely to be true. (Then, as a consequence, if one is justified in believing that the individuals in question – Mary, John, Bruce, ... Suzanne – are sufficiently trustworthy, one may be justified in going on to conclude that p is true.)
In short: It seems that beliefs that are unlikely to be true can justify beliefs that are likely to be true.
2. What support can one offer for the claim that if either possibility (2) or possibility (3), above, obtained, the original belief would not be justified? One natural idea is to say that for one belief to be justified relative to another is just for a statement like the following to be true:
If E is justified, then B is justified.
If so, then the point can be put as follows. All the relations of relative justification – of one belief's being justified in terms of some other belief – are a matter of a certain conditional statement's being true. But to show that B is justified, one has to establish an unconditional statement. No unconditional statement, however, of the relevant form, is entailed by the conditional statements in question, regardless of whether the latter are finite in number or infinite in number. So neither relative justification in a circle, nor an infinite regress of relative justification, will suffice to justify a given belief.
2. Skepticism, Foundationalism, and Coherentism
1. As regards epistemic justification, there are three main options: skepticism, foundationalism, and coherentism.
2. It is important to note that these are options not only with regard to knowledge, but also with regard to justified belief. (Many philosophers – such as Keith Lehrer in his article “Why Not Skepticism?” – define the term “skepticism” too narrowly, in that they take skepticism as being concerned only with knowledge. Mike Huemer, by contrast, defines skepticism correctly – namely, in terms of theses about the possibility of having certain sorts of justified beliefs.)
3. The present Epistemic Regress Argument, though it has been put forward as a justification of foundationalism, can be turned into an argument in support of skepticism. For if it is successful against coherentism, and if one can also argue – as BonJour attempts to do – that the idea of noninferentially justified beliefs is, in the end, untenable, both foundationalism and coherentism will be ruled out, and only skepticism will remain.
4. Such an argument for skepticism is broader than the types of skeptical argument previously considered, since the conclusion would be, not merely that there are no inferentially justified beliefs, or no inferentially justified beliefs of certain sorts, but that there are no justified beliefs at all – either inferentially justified or noninferentially justified.
3. Foundationalism
1. Foundationalism comes in three varieties: classic, moderate, and weak.
2. Classic and moderate foundationalism share the following features:
(1) They affirm the existence of noninferential knowledge, or at least of noninferentially justified beliefs;
(2) All other knowledge is justified on the basis of noninferential knowledge, and all other justified beliefs are justified on the basis of noninferentially justified beliefs.
3. Classic foundationalism involves the following two additional elements:
(1) The noninferential starting points are (a) indubitable and (b) infallible.
(2) Inferential knowledge and inferentially justified beliefs are arrived at via deductive reasoning.
4. The main problem with classic foundationalism is that the link between what we are noninferentially justified in believing, and what we are inferentially justified in believing, cannot, it would seem, be deductive in nature. (To see that the reasoning cannot be deductive, consider the following possibilities: (1) Brain in a vat type possibilities; (2) Bishop Berkeley's view of the world; (3) Puppets controlled by some sort of deity, or powerful extra-terrestrial being, in the case of beliefs about other minds; (4) Gosse's hypothesis in the case of knowledge of the past.)
5. Moderate foundationalism. This may differ from classic foundationalism in either or both of the following ways:
(1) It may be held that noninferential knowledge is not indubitable, or that noninferentially justified beliefs are not infallible.
(2) It may be held that the relation between what is noninferentially known (or justified) and what is inferentially known (or justified) is not deductive.
6. Weak foundationalism. This differs from both classic and moderate foundationalism in holding that the starting points for the justification of beliefs are (a) beliefs that have some epistemic warrant independently of their relations to other justified beliefs, but (b) where the warrant is not sufficient to make the beliefs in question such as are more likely to be true than false, and thus such as are prima facie credible.
The idea is that beliefs that have some degree of justification, beyond their a priori probability, can serve as a starting point for justification even if the extent to which they themselves are justified is not such as to render those beliefs themselves prima facie credible,. The earlier discussion, at the end of Section 1 above, indicates how that might be possible.
Compare, also, Laurence BonJour's characterization and discussion of weak foundationalism in his article "A Critique of Foundationalism”. (See Section II, pages 185-6, in Pojman’s anthology.)
4. Two Arguments in Support of Foundationalism
Foundationalism is typically supported by two arguments:
(1) The Epistemic Regress Argument
Perhaps the most commonly used argument in support of foundationalism is the epistemic regress argument, which we have already considered.
(2) The Evidentially Isolated, Justified Belief Argument
The epistemic regress argument, if sound, shows only that if one is to be justified in believing anything, there must be some beliefs that are justified even if one cannot offer any evidence in support of them. That argument provides one, however, with no indication of what sorts of beliefs might be noninferentially justified. This second argument, by contrast, focuses upon certain states of affairs, and appeals to intuitions to the effect that beliefs about such states of affairs are justified, even in the absence of other beliefs that would provide evidence for them.
Suppose, for example, that you are experiencing a tingling sensation. It would seem that it might very well be the case that there is nothing else that one believes at the time that would provide evidence for the claim that one is experiencing a tingling sensation. But even if that were so, wouldn't it still be true that one was justified in believing that one was experiencing a tingling sensation? So aren't there some beliefs that are noninferentially justified?
A slightly different way of putting this argument is to imagine, for example, that one is suffering from total amnesia, that one is in a perceptually deprived state, and that the only experience one is having is a tingling sensation. Then, if one believes that one is experiencing a tingling sensation, that belief may be evidentially isolated because it is the only belief that one has. So if the belief that one is experiencing a tingling sensation would be justified in such circumstances, it would have to be noninferentially justified, since it is the only belief that one has.
In contrast to the epistemic regress argument, the present argument does not show that all of one's justified beliefs are ultimately justified on the basis of beliefs that are noninferentially justified. For the present argument does not itself rule out the possibility that while there are noninferentially justified beliefs, some inferentially justified beliefs, rather than being justified on the basis of noninferentially justified beliefs, are justified in the manner proposed by coherentism. However, given that the main way of supporting coherentism, as we shall see later, involves setting aside the epistemic regress argument on the ground that the notion of noninferentially justified belief is, in the end, problematic, it would seem that, once the possibility of some noninferentially justified beliefs is admitted, there is no good reason not to adopt a foundationalist approach, and thus to hold that all justified beliefs are either noninferentially justified, or else inferentially justified on the basis of noninferentially justified beliefs.
5. Some Possible Characteristics of Noninferential Knowledge, or of Noninferentially Justified Belief
1. Infallibility: Noninferentially justified beliefs cannot possibly be mistaken.
2. Objective certainty: One is justified in betting at any odds on the truth of a noninferentially justified belief.
3. Subjective certainty: One is willing to bet at any odds on the truth of a noninferentially justified belief.
4. Indubitability. When a belief is spoken of as indubitable, this can be interpreted in either a psychological sense or a logical sense. On the former interpretation, it would be a matter of its being psychologically impossible to entertain any doubts about the truth of the belief in question, and that sounds very similar to subjective certainty. On the logical interpretation, on the other hand, the idea is that if one attempts to entertain a doubt about the belief, it turns out that such a doubt is self-defeating, in the sense that one is led back to the very belief that one is attempting to doubt. (The familiar Cartesian illustration is this: if you doubt whether you are thinking, then that very act of doubting whether you are thinking is itself an act of thinking.)
Comment: Indubitability in this logical sense of doubt being self-defeating would seem to be very limited in scope. Thus, while it may be psychologically impossible for Mary to doubt whether she is in pain, if she were to be able to entertain such a doubt, it doesn't look as if the doubt would be self-defeating: doubting whether one is thinking may itself be a case of thinking, but doubting whether one is in pain is not a case of being in pain.
5. Indefeasibility. A belief B that is noninferentially justified at time t is indefeasible if and only if there is no possible evidence such that if the person in question had that evidence at time t, he or she would no longer be justified in accepting B. An indefeasible belief, in short, is one that is not susceptible to being overthrown by additional evidence at the time in question.
Illustration
Suppose that devices have been developed – call them cerebroscopes – that when placed on a person’s head indicate what mental state the person is in. Suppose further that they have been tested for many years, and have been found to be completely reliable. A cerebroscope is placed on your head, and it indicates that you are having no visual experiences at all. It seems to you, however, that you are enjoying a complex visual field. When you indicate that the cerebroscope must be mistaken, it is checked out, but it is found to be suffering from no physical defect. Moreover, when other cerebroscopes are placed on your head, they yield the same results – namely, that you are not having any visual experiences at all.
What would you conclude in such a case? If you would hold, in spite of the cerebroscopic evidence, that you are enjoying visual experiences, then your view is presumably that the belief that one is oneself having visual experiences at a given time is indefeasible: if the belief is noninferentially justified for a person at a time, then there is no evidence that that person could have at that time that would undercut the justification.
6. Coherentism
What is coherentism? This non-skeptical alternative to foundationalism is a little more difficult to characterize precisely than foundationalism is. But one way of approaching the task of characterizing coherentism is via a bit of a historical digression.
6.1 Coherence Theories of Truth
1. Coherence theories of truth are an alternative to correspondence theories of truth.
2. What is a correspondence theory of truth? The basic idea involved in a correspondence theory of truth is that the bearers of truth-values – that is, such things as sentences, statements, and propositions – are entities with structure, as are states of affairs in the world, and truth is a matter of a certain relation between constituents of the former and constituents of the latter.
Illustration
Consider the sentence, "Fido is a dog". According to a correspondence theory, that sentence is true provided the following three conditions are satisfied:
(1) The world contains an object that is the referent of the term “Fido”.
(2) There is some property (or concept) associated with the predicate expression “is a dog”.
(3) The object referred to the term “Fido” has the property, or falls under the concept, associated with the predicate expression “is a dog”.
What if one focuses upon propositions, rather than upon sentences? Then the idea is that propositions are structured entities composed of concepts (and perhaps other things). So, for example, the view then is that the proposition that Fido is a dog involves an individual concept – the concept of Fido – and also a general concept – the concept of being a dog – and the proposition that Fido is a dog is true if and only if there is some entity that falls under the concept of Fido, and that entity also falls under the concept of a dog.
3. A coherence theory of truth, on the other hand, claims that truth is not a matter of a relation between propositions and something else, but simply a matter of a certain relation among propositions – namely, the relation of coherence.
4. But what is coherence? The usual answer here starts out from the idea that if one considers any two propositions, p and q, then there are three possibilities with respect to the bearing of proposition p upon proposition q: (1) proposition p might make q more likely, might raise the probability that q is true; (2) proposition p might make q less likely, might lower the probability that q is true; (3) proposition p might be neutral with respect to proposition q – neither making it more likely that q is true, nor making it less likely that q is true. One can then say that propositions p and q provide mutual support for each other if each makes it more likely that the other is true. Finally, one can describe a large system or set of propositions as coherent to the extent that subsets of that set of propositions mutually support one another.
5. One standard objection that was advanced to a coherence theory of truth is that given any set of statements or propositions that it might be claimed were true, given a coherentist account, in view of the relations of mutual support that they exhibited, one can find another set of statements or propositions that are incompatible with the first, but that exhibit equally impressive relations of mutual support among the propositions in the set.
How might one support that claim? One idea is to make systematic substitutions in the original set of statements. Here’s an illustration, in which two sets are related via an interchange of the terms “green’ and “red”:
First set: Second set:
(1) Fido is green. (1*) Fido is red.
(2) Socksie is red. (2*) Socksie is green.
(3) Fido looks green to Socksie. (3*) Fido looks red to Socksie
(4) Socksie looks red to Fido. (4*) Socksie looks green to Fido
6.2 Coherence Theories of
Justification
So much for the coherence theory of truth, and one important objection to it. As a result of such criticisms, some philosophers concluded that perhaps what coherence was really relevant to was not truth, but either (i) knowledge, or (ii) justified belief.
1. But won't the same problem arise? That is to say, if what beliefs are justified is a matter of relations of mutual support among propositions, won't one run, once again, into the Fido/Socksie sort of objection? The answer is that once one shifts from a coherence theory of truth to a coherence theory of knowledge, or of justified belief, one can shift from considering relations of coherence among statements or propositions to considering relations of coherence, or mutual support, among beliefs.
2. The coherence view of knowledge, or of justified belief, is therefore roughly as follows. Consider the total set of everything that you presently believe. It can happen that not all of your beliefs fit together in a very satisfactory fashion, in the sense that there may be some belief, B, that is really rather improbable, given the other things that you believe. (Consider the belief in a young earth by a person who generally accepts scientific beliefs.) So it may be that if you were to abandon some of your present beliefs, the result would be a set of beliefs that would be more coherent, in that each belief in the resulting set would be supported, to a reasonable extent, by the remaining beliefs. Tension, or conflict between different beliefs would thereby have been eliminated, or at least reduced to a minimum.
3. So how does shifting from a coherence theory of truth to a coherence theory of justified belief improve the situation? In the case of truth, one can generate alternative sets of propositions, and so the question of why some such alternative set isn't true is a pressing one. If one generates, in the same fashion, an alternative set of beliefs, then the point can be made that the alternative set may not be a set of beliefs that anyone actually accepts, and so there is no immediate problem.
One might try to press the point by saying that they are at least possible beliefs – beliefs that someone could accept. In response, the advocate of a coherentist approach to justified belief may say that beliefs that are merely logically possible are not relevant to the theory that he or she is proposing. For beliefs that are merely logically possible may not be such as anyone could actually come to have in the world as it is, and it is only beliefs that are possible in that latter, much more robust sense, that one needs to be concerned about.
6.3 Two Arguments Against Foundationalism, and in Support of Coherentism
Let us now turn to consider the defense of coherentism that Larry BonJour once offered, and which is as follows. First, BonJour grants that the epistemic regress argument does show that a foundationalist approach is prima facie plausible. He argues, however, that when foundationalism is carefully examined, it turns out to suffer from fatal flaws. In support of the latter claim, BonJour offers two, related arguments, each of which concerns the issue of whether it is reasonable, in the final analysis, to believe that some beliefs can be justified without being supported by evidence.
6.3.1 Argument 1: Doxastic Ascent
Formulations of this type of argument are found in Ernie Sosa's article, "The Raft and the Pyramid: Coherence Versus Foundations in the Theory of Knowledge", (Midwest Studies in Philosophy, Volume 5, Minneapolis: University of Minnesota Press, 1980, pages 3-25, and reprinted in Louis Pojman (ed.), The Theory of Knowledge, Third edition, Belmont, CA: Wadsworth, 2003, pages 222-37) and in Larry BonJour's previously mentioned article, "A Critique of Foundationalism". Sosa, for example, offers the following summary:
"A belief B is foundationally justified for S in virtue of having a property F only if S is justified in believing (1) that most at least of his beliefs with property F are true, and (2) that B has property F. But this means that belief B is not foundational after all, and indeed that the very notion of (empirical) foundational belief is incoherent." (Pojman's anthology, 231)
A similar, but fuller statement of the argument is offered by BonJour:
"If we let ‘F’ represent this feature, then for a belief B to qualify as basic in an acceptable foundationalist account, the premises of the following justificatory argument must themselves be at least justified:
(i) Belief B has feature F.
(ii) Beliefs having feature F are highly likely to be true.
Therefore, B is highly likely to be true.
Notice further that while either premise taken separately might turn out to be justifiable on an a priori basis (depending on the particular choice of F), it seems clear that they could not both be thus justifiable. For B is ex hypothesi an empirical belief, and it is hard to see how a particular empirical belief could be justified on a purely a priori basis. And if we now assume, reasonably enough, that for B to be justified for a particular person (at a particular time) it is necessary, not merely that a justification for B exist in the abstract, but that the person in question be in cognitive possession of that justification, we get the result that B is not basic after all since its justification depends on that of at least one other empirical belief. If this is correct, strong foundationalism is untenable as a solution to the regress problem (and an analogous argument will show weak foundationalism to be similarly untenable)." (Pojman's anthology, 187)
Comments: The Structure of BonJour's Argument
1. Let us consider a concrete version of BonJour's argument. In particular, let B be the belief that I am now experiencing a tingling sensation, and let feature F be the property of being a belief about one's own present experience. The above schema then becomes:
A. My belief that I am now experiencing a tingling sensation is a belief by me about my own present experience.
B. Beliefs by me about my own present experience are highly likely to be true.
C. Therefore, my belief that I am now experiencing a tingling sensation is highly likely to be true.
2. BonJour’s argument then appears to be as follows. First, why does the foundationalist need to be able to appeal to the above sort of schema? BonJour's answer is that, first, any satisfactory account of the notion of epistemic justification must connect up with the idea of being epistemically responsible, and secondly, a particular standard of epistemic justification can be defensible only if beliefs that are in accordance with that standard are likely to be true. So if the belief that I am now experiencing a tingling sensation is to be noninferentially justified, then it must be the case that it is likely to be true. In addition, the reason why this is so must in some sense be accessible to me, for otherwise my acceptance of that belief would not be epistemically responsible.
3. Given this starting point, the second step in BonJour's argument involves asking what sort of justification might be offered for premises A and B. One possibility is that both A and B might both be justified in an a priori fashion – that is, in a way that did not depend upon the way the world is, as a matter of fact. But it would then follow, since A and B together entail C, that C would have to be a necessary truth. But this is impossible, since C expresses a contingent proposition that might very well be false.
4. Thirdly, if it is not possible to justify both A and B in a priori fashion, then either A or B is an empirical belief, and so given that an argument of the above sort is needed if my belief that I am now experiencing a tingling sensation is to be justified, it follows that the belief that I am now experiencing a tingling sensation stands in need of empirical support, and so cannot be a noninferentially justified belief.
Critical Evaluation of BonJour's Argument
1. The above argument from A and B to C is capable of being interpreted in two different ways. The first interpretation can be put as follows:
Interpretation 1
A. I do in fact believe that I am now experiencing a tingling sensation and that belief is a belief by me about my own present experience.
B. Beliefs by me about my own present experience are highly likely to be true.
C. Therefore, I do in fact believe that I am now experiencing a tingling sensation, and that belief is highly likely to be true.
The second interpretation can be put in either of two slightly different ways. One involves making use of subjunctive conditional statements to formulate both premise A and the conclusion:
Interpretation 2a
A. If I were to believe now that I am now experiencing a tingling sensation, that would be a belief by me about my own present experience.
B. Beliefs by me about my own present experience are highly likely to be true.
C. Therefore, if I were to believe now that I am now experiencing a tingling sensation, that would be a belief that was highly likely to be true.
Alternatively, one can refer, in premise A, not to a belief, but to the corresponding proposition:
Interpretation 2b
A. The proposition that I am now experiencing a tingling sensation is a proposition about my own present experience.
B. Whenever I believe a proposition about my own present experience, that belief is highly likely to be true.
C. Therefore, the proposition that I am now experiencing a tingling sensation is such that if I were to believe it, that belief would be highly likely to be true.
On interpretation 1, statement A clearly cannot be justified in purely a priori fashion, since it involves the contingent claim that I do in fact have a certain belief at the present moment. The same is true of the conclusion – statement C – for the same reason. But on interpretation 2, in contrast – both in the form 2a and in the form 2b – neither statement A nor statement C has any such implication, and the question then becomes what is wrong with its being the case that both premises A and B are justified in a priori fashion if either version of interpretation 2 is adopted. It will still follow, of course – just as in the case of interpretation 1 – that conclusion C must be a necessary truth, but that is not problematic in the case of interpretation 2 in the way it is in the case of interpretation 1, since now the conclusion no longer says that I actually have a belief that is very likely to be true. Moreover, it is not at all obvious that there is anything wrong with the view that the conclusion expresses a necessary truth.
2. The upshot is that, if interpretation 2 is the correct interpretation of BonJour's argument, then there doesn't seem to be any reason, as far as the preceding argument goes, why both A and B, and therefore also C, mightn't be justified in an a priori fashion.
3. Although BonJour does not say so in the present argument, he does think, I believe, that premise B does not express an a priori truth. For in his article "A Holistic Coherentism" (Philosophical Studies 30 (1976), pages 281-312), he considers the following claim
(*) Introspective beliefs (of certain sorts) are very likely to be true
and he explicitly says that this is an empirical premise. But, as I shall be arguing later, this latter view seems very problematic. If I am right about this, then no problem arises if one adopts interpretation 2.
4. Perhaps, however, BonJour has interpretation 1 in mind? But then the argument seems very suspicious, for two reasons. First of all, if, in order to be justified in believing that
E: I am now experiencing a tingling sensation
it is being asserted that I have to be justified in accepting both A and B, under interpretation 1, then it is being asserted that I have to be justified in believing:
B(E): I now believe that I am now experiencing a tingling sensation.
But what reason is there to think that my being justified in believing that I am now in some mental state M is based upon, or rests upon, my being justified in believing that I now believe that I am in mental state M? For even if having a belief about one's own present mental state involves being aware of that mental state, and even if awareness of a mental state entails (at least) potential awareness of the awareness, this would show only that being justified in believing that B(E) is a logically necessary condition of being justified in believing that E. It would not show that one's being justified in believing that E is in any way based upon one's being justified in believing that B(E). Moreover, the idea that one's being justified in believing that E is based upon one's being justified in believing that B(E) is surely very implausible, since it would seem that any awareness that one is aware that one is in a certain mental state itself derives from the awareness that one is in the mental state in question. So how can any justification of the former possibly underpin one's justification of the latter?
Secondly, suppose that, given interpretation 2, premises A and B are susceptible of an a priori justification. They then entail C, which on the second interpretation asserts that if I were now to have a belief that I am now having a tingling sensation, that belief would very likely be true. But, then, must not the truth of this conclusion suffice to ensure that I am acting in an epistemically responsible way in having the belief in question, and, if so, why isn't the foundationalist correct in holding that the belief in question is noninferentially justified?
6.3.2 Argument 2: Is the Idea of the Given Ultimately Coherent?
One way in which one might try to support the idea that some beliefs can be noninferentially justified – that is, that one can be justified in believing some things even if one cannot offer any evidence in support of the beliefs in question – is by appealing to the idea that some states of affairs are immediately given, that the mind is directly presented with some things. Thus it is often claimed, for example, that one is directly acquainted with one's own present experiences, and possibly with other mental states, such as thoughts. Talk about something’s being "immediately given", or "directly presented" to the mind, or about one's being "directly acquainted" with some state of affairs is, admittedly, not easy to cash out. Is it, for example, a matter of causal immediacy? That may be part of it – though even that is not clear. But even if causal immediacy is involved, it seems clear that that is not the whole story. For the idea that one is trying to capture here is intimately connected with consciousness, and it seems clear that consciousness is not to be explained in terms of – or, at least, not simply in terms of – causal immediacy. For any causal process whatsoever can be divided into two parts such that one is causally immediately before the other.
So the idea of the immediately given is not unproblematic. BonJour, however, is advancing a stronger claim – namely, that it is very unlikely that one can, in the end, make sense of the notion of the given. In support of this claim, he argues that the idea of "the given" involves the idea of states that seem to oscillate between being cognitive states and being non-cognitive states, and neither way of conceiving them is satisfactory. For, BonJour argues, on the one hand, it is hard to see how the given could be a non-cognitive state – and so a state with no propositional content – since the given is supposed to justify one in having a certain belief, and it is hard, he contends, to see how anything that is not a cognitive state can provide support for a belief, can make it likely that a belief is true. But, on the other hand, if what is given is itself a cognitive state, then it involves propositional content, and the question arises as to what reason there is for thinking that the proposition in question is true, and one is once again off on a regress.
Could there be some sort of intermediate state – some sort of quasi-cognitive state – that involves both cognitive elements and non-cognitive elements? This is a natural way of thinking about a state such as an awareness that one is having, say, a tickling sensation. For on the one hand, if one is aware that one is having a tickling sensation, doesn't that imply that one believes that one is having a tickling sensation? And, on the other hand, mustn't there be more to the awareness that one is having a tickling sensation – namely, the having of the tickling sensation – and mustn't the latter involve something more than – or something different from – the mere presence of beliefs, or other “propositional attitude” sorts of states.
BonJour argues, however, that the idea of a quasi-cognitive state will not help here. For the postulation of any state that involves propositional content means that one is confronted with the question of how the propositional content that is present in the quasi-cognitive state is itself justified.
What the foundationalist needs is a state that has both of the following two properties: (1) it is a state that can serve to justify beliefs; (2) it is a state that does not itself stand in need of justification. BonJour contends that no state can possibly possess both of those properties, since, he claims, to possess the first property it must have propositional content, but having propositional content means that it is a state that itself stands in need of justification.
Comments
1. The idea of the given is closely connected with the idea of consciousness, and I think it is reasonable to believe that one cannot expect to have a satisfactory account of givenness until one has a satisfactory account of consciousness. But to provide the latter is, arguably, the most difficult problem in philosophy of mind – and perhaps in all of philosophy.
2. Although one may not be able, at present, to offer a satisfactory account of consciousness, or of the idea of the mind's being immediately presented with something, or being directly acquainted with something, if one considers any experience that one has, the idea that the mere fact that one is having that experience serves to justify certain beliefs about that experience has great intuitive appeal.
(3) BonJour explicitly contends that the response that I have just made to his attack upon the idea of givenness is not satisfactory:
"Third. Notice carefully that the problem raised here with respect to givenism is a logical problem (in a broad sense of 'logical'). Thus it would be a mistake to think that it can be solved simply by indicating some sort of state which seems intuitively to have the appropriate sorts of characteristics; the problem is to understand how it is possible for any state to have those characteristics." (A Critique of Foundationalism, page 193.)
But is BonJour right here? The foundationalist points to the state of being aware that one's own present experience has, say, the property of being of the tingling sort, and claims that the belief that one is now experiencing a tingling sensation – which is part of that state of awareness – is justified simply by being part of such a state of awareness. BonJour contends that there is an argument of a broadly 'logical' sort that shows that this is not possible. But BonJour's argument involves the crucial contention that a propositional claim can only be justified by appealing to another propositional claim, and, in the present context, that contention appears to be simply question begging. Awareness of qualities of one's present experiences certainly seems to be a state that justifies a given belief, and that does so without appealing to other justified beliefs as evidence, and given that this is so, it seems reasonable to conclude that what BonJour claims is impossible is in fact possible.
4. This is not to say that there is nothing more that needs to be done here. The notion of consciousness, or of “immediate”, or “direct” awareness of a property, is a notion that one would really like to be able to analyze. But, as I noted in the first comment above, the problem of consciousness has proven quite intractable. The difficulty posed by the problem of consciousness gives one no reason for concluding, however, that the notion of direct awareness does not make any sense.
7. Possible Objections to Coherentism
Objection 1: The Existence of Alternative, Equally Coherent Systems of Beliefs
This parallels the objection that was set out earlier in connection with coherence approaches to truth. There, the objection seems decisive. But coherentists maintain that when one switches to the case of coherence approaches to justified belief, the objection loses its force. One can, it is suggested, insist that consideration be restricted to something like beliefs that people actually have, and ignore systems of beliefs that are merely potential.
Is this move convincing? I can't see that it is. For surely we can imagine technological advances that would enable a person to have his or her beliefs altered – including present memory beliefs – and, then, if it is really coherence that makes for justification, why wouldn't it be reasonable to have one's memories changed in quite dramatic ways if doing so resulted in a more coherent system of beliefs? Yet surely the idea of a potentially wholesale alteration of one's present beliefs in order to have a more coherent system of beliefs is very dubious, to say the least. For many of one's beliefs, surely, reflect the actual impact of the world upon one, and thus play a crucial role in ensuring a measure of correspondence between what one believes and the way the world really is.
Objection 2: How Does Coherence Connect Up with Truth?
Mightn't the set of one's beliefs be just a grand fairy tale? That is to say, if truth is a matter of correspondence, not of coherence, then what reason is there for believing that because someone's beliefs constitute a harmonious, mutually supporting set, that they are likely to correspond to the way that the world really is? But if the fact that certain beliefs are coherent does not make it more likely that they will be true, then how can a coherence account of justification possibly be satisfactory?
Larry BonJour sketches an answer to this objection in his article, "Holistic Coherentism", the thrust of which seems to be that the world gives rise to one's observational beliefs, and it is this causal connection that makes it likely that coherent beliefs will, at least over the long haul, be true.
It is unclear whether BonJour is appealing here to (1) the existence of an actual causal connection, or instead to (2) a belief, within a coherent system of beliefs, that there is such a causal connection. If it is the former, so that BonJour is appealing to the existence of an actual causal connection, then it would seem that he is lapsing into an externalist account of justification – according to which there can be factors that are crucial to the justification of a given belief, but of which one has no awareness. On the other hand, if he is appealing instead to a belief that there is such a causal connection, and to the coherence of that belief with one's other beliefs, then it seems that one is back with the grand fairy-tale problem, and there is no reason for thinking that a coherent set of beliefs is likely to be true.
Objection 3: Shouldn't Observational Beliefs Be Assigned a Special Place?
A foundationalist account of the justification of belief is appealing, I would suggest, for two main reasons. First, there is the epistemic regress argument discussed earlier, the thrust of which was that there does not seem to be any satisfactory alternative to the idea that all inferentially justified beliefs must be justified, ultimately, on the basis of beliefs that are justified (or, at least, have a prima facie credibility that is greater than their a priori probability), but that are not justified (or do not have that prima facie credibility) on the basis of evidence. In the present context, the thing to notice about the epistemic regress argument is that it does not provide any indication of what the content of noninferentially justified beliefs might be. In this respect, it contrasts sharply with a second way in which one might support a foundationalist approach, namely, by claiming that there are some beliefs that it is reasonable to accept even if one cannot offer any evidence in support of them. Among the most plausible candidates are certain beliefs about one's own present experience – such as the belief that one's own present experiences involve different properties, or that parts of one's visual field resemble one another quite closely, of that several parts of one's visual field are red, etc.
This second line of support for a foundationalist approach gives rise, accordingly, to the following question. Don't beliefs about one's own present experiences have to be assigned a special epistemic position? And if this is so, then isn't a coherentist approach unsatisfactory, for doesn't it fail to do so?
Comment
Larry BonJour offers an answer to this objection, but his answer involves the same idea as that mentioned in connection with the previous objection.
Objection 4: Direct Acquaintance and Semantically Basic Terms
An objection that is closely related to the preceding one focuses upon the importance of observational beliefs in fixing the meaning of semantically basic terms – that is, those terms with which analysis starts. Here the idea is that analysis should start with terms that pick out properties or relations with which one can be directly acquainted. But if there are properties or relations with which one can be directly acquainted, is it not plausible that such direct acquaintance makes it the case that one is justified in believing that the property or relation in question exists, and is instantiated in a given particular, or ordered set of particulars?
In short, the objection is that rejection of the idea that there are epistemically basic beliefs appears to commit one to the rejection, as well, of the idea that there are semantically basic terms with which analysis can, and must, start.
Objection 5: Coherence Is Too Easily Achieved
This objection, which is set out by Richard Fumerton in his article, "A Critique of Coherentism" (Louis Pojman (ed.), The Theory of Knowledge, Third edition, Belmont, CA: Wadsworth, 2003, pages 215-21) runs as follows. First, the coherentist generally sets out his or her position in terms of relations of support. The reason for doing so, Fumerton suggests, may be that the coherentist thinks that the stronger relations of entailment would be too hard to come by. But, secondly, is it true that relations of entailment are hard to come by? Fumerton argues that they are not. His argument is as follows:
(1) Let p and q be any two unrelated propositions. For example:
p = There are cars in the world
q = There is a golden teacup in orbit around Venus
(2) Suppose that the only beliefs that Bruce accepts are the belief that p and the belief that q. Not a very exciting set of beliefs, but perhaps Bruce is a cautious believer. We ask Bruce, however, to consider accepting two other beliefs:
r = Either there are cars in the world, or else it is not the case that there is a golden teacup in orbit around Venus
s = Either there is a golden teacup in orbit around Venus, or else it is not the case that there are cars in the world.
Since r = (p or not q), r cannot be false if p is true. Similarly, since s = (q or not p), s cannot be false if q is true. So, since Bruce already believes p and q, he has good reason also to accept both r and s
(3) But once Bruce believes the following four propositions – p, q, r, and s – he has a system of beliefs that, though small, displays striking coherence. For we have the following relationships:
(a) p entails r
(b) q entails s
(c) p together with s entails q
(d) q together with r entails p.
So every belief in the system of beliefs is either entailed by one of the other beliefs, or by a combination of two of the other beliefs.
The thrust of Fumerton's argument, accordingly, is that given that any set of two or more beliefs can be extended into a highly coherent system simply by adding beliefs that are logically entailed by beliefs that are already in the set, coherence is surely not a good reason for thinking that a set of beliefs is justified. For then anyone's beliefs would qualify as justified provided that that person makes some elementary logical inferences.
Comment
This is an interesting and important argument. It does show at the very least, I believe, that the coherentist needs a very different account of coherence if coherentism is to be acceptable. But if relations of mutual support won't do, what will be the key idea in such a revised account? Something along the lines of the idea of explanatory relevance would seem to be the most plausible answer.
Objection 6: Another Regress
This objection, which is also set out by Richard Fumerton in his article, "A Critique of Coherentism", may be put as follows:
(1) Suppose that my beliefs are beautifully coherent. Am I then justified in accepting the propositions in question?
(2) Answer: No, because though my beliefs are beautifully coherent, I may not believe that they are, or I may believe that they are, but not be justified in doing so.
(3) So in addition to my beliefs being beautifully coherent, I have to have certain "meta-beliefs", to the effect that I have certain beliefs, and that they are coherent. Moreover, those meta-beliefs must be justified. But on the coherentist approach, those meta-beliefs will be justified only if they are coherent.
(4) But it is not enough that those meta-beliefs be coherent. For I might not believe that I have any such beliefs, or I might not believe that they are coherent. Or I might believe that, but not be justified in doing so.
(5) So now I have to have meta-meta-beliefs, and they have to be coherent.
(6) This process never stops. So a coherentist approach to justification generates an infinite regress of its own – and one that surely entails that no belief is ever justified if the coherentist approach is correct.
Schematically:
1. According to the coherentist approach, B is justified only if it belongs to a coherent set of beliefs, S, that I accept.
2. But if one adopts an internalist view of justification – according to which the only factors that can justify a belief for a person are factors that the person is aware of – then my belief that B is justified only if I believe that S is a coherent set of beliefs to which B belongs, and all of which I accept. Call this meta-belief B1.
3. So my belief B would not be justified if the meta-belief B1 were not justified.
4. Applying the coherentist approach to B1, however, will, in precisely parallel fashion, generate a meta-meta-belief B2 that is such that I am not justified in accepting B1 unless I am justified in accepting B2.
5. Because of the nature of the new beliefs that are generated – namely, the fact that there is, at every step, a deeper embedding within a "belief operator" – one will never come back to earlier beliefs in the sequence. So the sequence generated cannot be any sort of circle: it must be an infinite regress that involves, at every step, some new belief.
6. Coherentism leads, according, to a vicious infinite regress. So if coherentism were correct, there would be no way of avoiding skepticism.
Comment
Might one be able to get around this objection by the following strategy?
(1) Let S be the set of all of A's beliefs.
(2) Suppose that S is a very coherent set, and that it contains the following beliefs:
C0: S is a coherent set of beliefs.
C1: A believes that S is a coherent set of beliefs.
C2: A believes that A believes that S is a coherent set of beliefs.
Etc., for C3, C4, ... Cn, ... !
(3) Then S will be such that A can show, along coherentist lines, that any belief in S is justified.
The cost of this maneuver is that, first, A has an infinite number of beliefs of a certain sort – C0, C1, C2, ... etc. – and that seems problematic in the case of humans. Secondly, self-reference is involved in the beliefs in question, since S is defined as the set of all of A’s beliefs, but some of the beliefs in set S are themselves beliefs about set S – namely, all of the Ci. But self-reference is known to generate various paradoxes, and many philosophers think that self-reference should not be allowed. (Compare the following sentence, which refers to itself: “This very sentence is false”. Is that sentence true or false? If it is true, then it follows that it is false, since it says that it is. If, on the other hand, it is false, then since it says that it is false, ,it must be true. Either way, there is a contradiction.)
Objection 7: Consistency is Not Even A Necessary Condition for Rational Beliefs.
This objection is also set out by Richard Fumerton in his article, "A Critique of Coherentism". It turns upon a consideration of the lottery case – sometimes referred to as "the lottery paradox". Thus, suppose that there is a lottery in which there are 100 tickets, and the draw is a random one. Then it would seem that one is justified in believing that ticket 1 will not be the winning ticket, since the chances against it are 99 to 1. (If odds of 99 to 1 against do not seem sufficient to justify believing that the proposition in question is false, the number of entries in the lottery can be made as large as one wants.) But similarly, one is also justified in believing that ticket 2 will not be the winning ticket, and so on for every ticket. So, for every ticket n, one is justified in believing that ticket n will not be the winning ticket. But the resulting 100 rational beliefs, taken together, entail that no ticket will be the winning ticket. Yet one is surely justified in believing that one of the tickets will win. Conclusion: There is a set, containing 101 beliefs, such that it is logically impossible for all 101 beliefs to be true, and yet one is nevertheless justified in accepting all 101 beliefs. Therefore one is sometimes justified in accepting inconsistent beliefs.
Comments
1. One response to the lottery case is to say, first of all, that one has to shift from talking about beliefs simpliciter to talking about degrees of assent to different propositions, and then, secondly, that one then has to introduce an idea that deals with the consistency of degrees of assent to different propositions. When this is correctly done, there will no longer be any inconsistency.
2. Alternatively, one can take the view that it is a mistake to believe that ticket 1 will not be the winning ticket: what one should believe, rather, is that the probability that ticket 1 will win is only 1 in 100. If this is done, then the 101 beliefs in question are not mutually inconsistent.
3. Either way, the argument for the thesis that consistency is not even a necessary condition for rational beliefs has been shown to be unsound.
Objection 8: The Impossibility of "Pooling Evidence" to Resolve Disagreements
Mightn't different people accept different beliefs that formed equally coherent sets, but which were incompatible with one another? (Differing sets of religious beliefs are a common, proposed illustration here.) On a foundationalist view, there would, presumably, be a way of resolving such disagreements, at least in principle, since people could, so to speak, pool their evidence, and determine what beliefs were justified in the light of all of the evidence. But is there any way of resolving the disagreement, even in principle, on a coherentist approach? Perhaps one could pool all of the beliefs? But, in the first place, the result will be an inconsistent set of beliefs. (Contrast this with the situation in the case where one pools evidence. In that case, there will also be inconsistency, given certain views concerning what sorts of things one can have noninferential knowledge of. In particular, this will be the case on a direct realist approach. But on other views concerning noninferential knowledge, the pooling together will not lead to any inconsistency. In particular, if one holds that the only beliefs that can be noninferentially justified for a given person are beliefs about that person’s mental states at the present time, no inconsistency will arise.)
Secondly, couldn't there be cases where one could cut back to a set of mutually supporting beliefs in two or more very different ways? But if so, which would be the correct way?
Objection 9: The Possibility of Evidentially Isolated, Justified Beliefs
Finally, there is an objection that consists simply of one of the two arguments offered earlier in support of foundationalism – namely, the argument to the effect that it seems that I might very well be justified in believing that I am now experiencing a tingling sensation even if, as a matter of fact, I have no other beliefs that would provide evidence for that belief. So it seems that one can have beliefs that are noninferentially justified.
A slightly more dramatic version of that argument – as was also noted earlier – is one where a person is suffering from complete amnesia, and is in a situation where the only experience that he or she is having is a tingling sensation. Then the belief in question might very well be necessarily isolated, evidentially, since the person might not have, at that time, any other beliefs at all. Yet even if that were so, wouldn't the belief still be a justified one? If so, then noninferentially justified beliefs must be possible.