Special foundations theories of justification can be contrasted with no-foundations theories. The two sorts of theories differ in the answers they give to the question, “Which of a person’s beliefs and inferential methods require justification in terms of other beliefs and/or methods?” No-foundations theories answer that all beliefs and methods require such justification. Special foundation’s theories answer that some beliefs and methods--the foundational beliefs and methods--do not require such justification and all other beliefs and methods require justification that ultimately terminates in foundational beliefs and methods.
A third more conservative kind of theory, general foundationalism, answers that none of one’s beliefs and methods require such justification, at least in the absence of special reasons to distrust such beliefs or methods. My plan is to try to explain the issue that divides these views and to argue that the third conservative approach is the only one that is viable. In particular, intractable skeptical problems arise for the first two approaches that do not arise for the third.
I assume that justification has a normative significance. If S is not justified in believing B or using inferential process P, then S ought not to believe B or use P and, if S continues to believe B or use inferential process P, there is something wrong with S for doing so; S is at fault, criticizable for believing B or using P.
Furthermore, I am concerned with an “internalist” understanding of justification in contrast with an “externalist” understanding (Bonjour, 1992).
An externalist account might say that S is justified in using inferential procedure P just in case P is a reliable process, whether S is aware of its reliability or not, and that S is justified in believing B if S came to believe B by virtue of a reliable process, where any beliefs relied on in that process are justified, where it may be irrelevant whether or not S knows that the process is reliable, and where S may at this point not even remember how he or she came to believe B (Goldman, 1986).
Imagine two cases that are exactly alike as far as S can tell. In the first case, the process by virtue of which S comes to believe B is reliable. In the second case, the process by virtue of which S comes to believe B is not reliable. I am here concerned with a notion of justification such that, if there is no way for S to tell whether he or she is in the first case or the second, then S must be equally justified in believing B in the two cases. If S is not at fault for believing B in the first case, S is not at fault for believing B in the second case. If S is criticizable for believing B in the first case, S is criticizable for believing B in the second case.
It may be that S believes something true in the one case and not in the other, but that does not affect whether S’s belief is justified in the sense of “justified” I am concerned with.
Issues about internalist justification are methodological issues--issues about what to do--although the methods need not be deliberately or consciously followed. Externalist justification may be more relevant to whether one knows, as opposed to being relevant to what one should now do.
Internalist justification need not be conceived as entirely a matter of what is available to consciousness or deliberate thought. But such justification must at least be internal to the (mostly unconscious) procedures by which one forms and modifies beliefs. Justification has to do with the internal workings of that system. Relevant states and processes will include beliefs and inferences and may also include perceptual and other experiential states along with desires, intentions, and other affective states (Pollock, J., 1986, Chapter 5; Harman, 1986).
One issue for an internalist account of justification is whether the account requires or allows for foundational beliefs and foundational inferential methods--beliefs and methods that are at least prima facie justified without appeal to other beliefs and methods--in contrast with non-foundational beliefs and methods that have to be justified by appeal to other justified beliefs and methods.
There are at least three possible positions on this issue. (1) No beliefs and inferential methods are foundational. (2) Some beliefs and methods are foundational and all other beliefs and methods have to be justified ultimately in terms of foundational beliefs and methods. (3) All beliefs and inferential methods are foundational. The term “foundationalism” is often used for the second, intermediate position, which I will call a “special foundations” theory, in contrast with a “general foundations theory” (the third position) and a “no foundations theory” (the first position).1
The intermediate, special foundations, approach to epistemology supposes that most beliefs and inferential methods are justified in terms of other beliefs and methods, which in turn are justified in terms of still other beliefs and methods, until ultimately one reaches foundational beliefs and foundational methods that are justified either directly or perhaps by perceptual or other experience, without having to be justified in terms of other beliefs and methods.
In some versions, foundational beliefs and methods have to be absolutely certain--justified no matter what. In other versions, foundational beliefs and methods are provisionally or prima facie justified, possessing an initial justification that can be overcome or defeated by further considerations.
Different special foundations theories may provide different characterizations of the foundations, but foundational beliefs are normally taken to be limited to two sorts, namely, beliefs about experience and beliefs in principles that one takes to be basic a priori truths. Foundational methods are normally taken to have a justificatory force that is apparent directly a priori.
Many traditional skeptical problems about justification arise within this sort of foundational framework. I should emphasize that I am concerned here only with skeptical problems concerning justification and leave aside additional skeptical problems about knowledge. From this perspective, the problem of other minds is the problem of justifying beliefs in other minds. Similarly, the problem of the external world is the problem of justifying beliefs about objects in the external world. The traditional problem of induction is whether there is a justification for conclusions arrived at via induction, where induction is not itself treated as a foundational method.
An important question for any theory of justification is, “What conditions have to be satisfied for S to be justified or not justified in believing B or in inferring using P?” According to a foundationalist theory, a nonfoundational belief or inferential process is justified only if S has a justification for believing B or for using P, where a justification is an argument of some sort. (What is it for S to “have” a justification J? I put this extremely important issue to one side for the time being.)
Foundational beliefs and methods are supposed to be at least prima facie justified for S without the need for S to have justifications for them that appeal to other of S’s beliefs or methods. Exceptions are possible, when for example there is sufficient evidence against a foundational belief or method to override its prima facie justification. In such a case, S needs to have some sort of justification for continuing to have a belief or use a method.
As already mentioned, the foundations framework is intermediate between the other two with respect to the question, “How many of one’s beliefs and inferential methods require justification in terms of others?” To this question, a special foundations approach answers that some of one’s beliefs and methods require such justification while others, the foundations beliefs and methods, do not.
The competing no-foundations view holds that that none of one’s beliefs and methods are foundational; all require justification in terms of other beliefs and methods. In one version the no-foundations view allows for circular justifications, “as long as the circle is big enough.” So developed, the view is sometimes called the “coherence” theory of justification on the grounds that the justification of any belief or method ultimately rests on the way that belief or method coheres with other beliefs or methods. Another variant allows potentially infinite justifications.
On the other side, the general foundationalist approach holds that all of one’s beliefs are foundational. They are all prima facie justified. One is justified in continuing to believe as one believes and to continue to use whatever epistemic methods one uses, in the absence of any special reason not to do so. This approach might also be called “general conservatism” for obvious reasons.
Any special foundations theory faces the familiar skeptical problems of providing foundational justifications for beliefs in other minds, the past, and the external world, and for inductive inferential practices.
The seriousness of such skeptical problems depends on what the foundations are taken to be. In the strictest theories, foundational beliefs have to be either self-evidently and directly a priori or must reflect direct awareness of the subjective character of immediate sensory or other experience. Furthermore, in the strictest foundational theories, the only acceptable foundational methods appeal to obvious self-evident deductive implications.
Now, at any given time, S can be aware of only a few aspects of his or her current subjective experience and perhaps only one or two self-evident truths. If S is reflecting on memory experience, S cannot be reflecting on very much of it. So, the strictest foundations must be very limited indeed. Little of what S believes could have a justification that is based on the strictest foundation.
Defenders of special foundations approaches therefore tend to settle for somewhat less strict foundations, perhaps allowing S to appeal to what S could now remember if S tried to do so, or perhaps by allowing S’s past experiential beliefs and self-evident insights into the foundations.
A further step might be to include in the foundation certain perceptual beliefs about perceived external objects, like “There is a tree in front of me,” even though such beliefs are not merely about subjective experience. To be sure, such perceptual beliefs are to be abandoned in the face of certain additional evidence about viewing conditions, but still they might be directly justified in the absence of specific challenge to them.
Beliefs about the past arising from certain sorts of episodic memory might also be treated as foundational. An even more liberal version of this approach would count all beliefs that are based on memory as foundational. Since all beliefs that are retained for a while are in a sense “based on memory,” that version allows quite general foundations and is similar in its treatment of beliefs to general conservatism, the third of the views I am discussing.
Similarly, one might extend the foundational methods of justification to include more than strict deduction. One might allow induction or inference to the best explanation. Some theorists argue that the way in which we normally rely on what others tell us has to be treated as a foundational method that needs no further justification.
Consider this last case in which a belief is based on testimony. S reads that the Battle of Hastings occurred in 1066. Later S is able to answer correctly when asked on a test when the Battle of Hastings occurred. What justifies S in this belief B?
Some foundationalists might say that S infers B from the fact that there is testimony to that effect and testimony is generally reliable. But what justifies S in believing that testimony is generally reliable? Coady (1992) argues that there is no non-circular way to argue for the reliability of testimony. Burge (1993) and Foley (1994) accept what Goldman (1999, p. 128) calls “testimonial foundationalism,” which counts any belief B derived from testimony as a foundational belief, one that S is directly prima facie justified in accepting.
One might object that much inference must go into understanding what is said and so the acceptance of testimony, given the ambiguities of ordinary linguistic communication. S effortless chooses among various interpretations of what has been said given the surrounding context, where one important principle is the likelihood that the interpretation makes what is said true. Consider an example of Tom Patton’s. “I watched her duck when they were throwing rotten eggs; it swam out to the middle of the lake” (cited in Ziff, 1972, p. 29.) The first clause is ambiguous but is initially heard as having a particular interpretation. Given the second clause, hearers immediately reinterpret the first, where some sort of inference is involved. Some such inference to reduce ambiguity is involved in all linguistic understanding. The objection, then, is that beliefs based on testimony cannot be foundational because they are (and have to be) based on inference.
A variation on testimonial foundationalism might work here, however. A theorist might simply take the processes of inference involved in the acceptance of testimony to be foundational processes, processes that S is directly prima facie justified in using.
This and the previously mentioned extensions of foundationalism might be defended on grounds of necessity: without such extensions, foundationalism would imply that we are hardly ever justified in our beliefs or inferences. However, such extensions move special foundationalism toward general foundationalism, which treats as prima facie justified all current beliefs and methods S uses. A view properly called special foundationalism must restrict the foundations.
An important problem facing a special foundations approach lies in motivating an adequate account of sufficiently limited special foundations for justification such that, on the one hand, it makes sense that justification should be able to rest on this special foundation and, on the other hand, the foundation is adequate to prevent an immediate fall into justificational skepticism.
Difficulties that arise in distinguishing foundational from other beliefs may lead one to try to give up any distinction between foundational beliefs and methods and other beliefs and methods. One way to abandon the distinction is to treat all of a person’s current beliefs and methods as foundational, that is, as prima facie justified in the absence of any specific objection to them. Such is the response of general conservatism or general foundationalism. The other response is that of a no-foundations theory: treat no beliefs and methods as foundational. In a no-foundations theory, all beliefs and methods require justification in terms of other beliefs and methods, where such justification is ultimately potentially circular or infinite.
Like general foundationalism, a no-foundations theory must assign S’s current beliefs some sort of special justifying status as compared with other propositions that S does not believe. Otherwise there would be no way to make out that any of S’s beliefs are any more justified than their opposites. A large complex web of beliefs and procedures might admit of circular justification relations or a potentially infinite chain of justification relations without any part of the web or chain being justified. Consider any conceivable web or potentially infinite chain of propositions that are related by possible methods of inference in a coherent fashion, where these are not propositions believed by S or methods of inference used by S. If no special status were given to S’s current beliefs, S would be equally justified in any accepting any such web of propositions and methods. That would mean that S was equally justified in believing anything and using any inferential methods whatsoever.
To avoid this result, it must be said that S’s justifications are to appeal to other actual beliefs of S and methods that S actually uses, so that S’s current beliefs and methods are being privileged in comparison with other possible beliefs and methods. S’s current beliefs and methods are therefore being treated as something like a foundation. We might call this a “foundational-no-foundational” approach. It is foundational in the special role it gives to existing beliefs. It is no-foundational in that it requires that non-trivial justifications be associated with all beliefs and methods.
No-foundations and special foundations theories require that all or most beliefs be associated with non-trivial justificatory arguments, if they are to be justified. Consider the implication of such a requirement for things one came to believe for reasons one has now forgotten. One no longer associates reasons or justificatory arguments with such beliefs. The requirement that such beliefs be associated with justificatory arguments if they are to be justified implies that such beliefs are not justified. Given the normative aspect of justification, one ought to give up any such belief and one is at fault for continuing to believe it.
People tend to have many beliefs of this sort, beliefs about historical dates (the Battle of Hastings) and various other factual beliefs (about state or national capitals). The skeptical implications of the justificatory argument requirement can be quite substantial. On the other hand, there are no such skeptical implications to general foundationalism, since it does not impose any general requirement that beliefs be associated with justificatory arguments.
Notice that an internalist approach cannot be satisfied with the thought that a belief is now justified as long as one had a good argument for it when one came to believe it, if one cannot now tell whether one had such a justificatory argument.
A further worry emerges when we ask what it is for S to “have” a justification (justificatory argument) J. Must S be consciously aware of J? Given the relatively limited capacity of S’s awareness at any given moment, this would mean that at any given moment S was justified in very little.
One might want to distinguish occurrent conscious judgments from dispositional beliefs and then say that S is justified in consciously judging B if and only if S consciously associates with B a conscious justificatory argument J. And S is justified in dispositionally believing B if and only if S’s disposition to consciously judge that B involves also a disposition to be consciously aware of a justificatory argument J for B.
There are complications in that judgments might well be occurrent but unconscious (Harman, 1986, p. 13-14). Furthermore, beliefs might affect how one reasons without being reflected in occurrent judgments, either conscious or unconscious. Such beliefs might be different from merely implicit beliefs that can be attributed to S because they are in some relevant sense implied by S’s implicit beliefs.
A related idea would be that S has a justification J for believing B or for using P if and only if S is disposed to produce J if asked to justify B or P. However, two worries arise. First, and perhaps less seriously, S might currently be unjustified in believing B even though, if asked to justify B, S might be able to come up with a justification that would then justify S in believing B (and similarly for inferential process P). This is really a worry only if the truth of that conditional implies that S has the relevant disposition. However, such an implication does not hold in general (Lewis, 1997, and references therein.).
Second, and more seriously, people in general seem to be inept at producing good justifications for their beliefs. Do we really want to say that they are therefore not justified in those beliefs?
It might be enough if S’s beliefs were structured in such a way as to associate justifications with certain beliefs and with certain inferential procedures, whether or not S consciously realized that this was so or could consciously formulate the relevant justifications. S would have to have procedures that are sensitive to these associated justifications so as to abandon beliefs without adequate justifications and to adopt new beliefs for which there are adequate justifications. Alas, it is doubtful that people do associate justifications with beliefs as required by this idea.
As soon as one tries to formulate any sort of requirement that nonfoundational beliefs are to count as justified only if associated with a good justification, a serious issue arises about how to attribute the needed justifications to someone in a way which avoids the immediate conclusion that most beliefs are unjustified.
In a general foundations framework all of S’s current beliefs are foundational. Similarly for S’s current inferential methods. These beliefs and methods form the basis from which S must begin any reasoning. (One must start where one is.) These beliefs and methods are prima facie justified, so S is justified in continuing to accept them in the absence of specific challenges to them, where challenges can only arise out of S’s other beliefs and methods.
If conflicts arise within S’s beliefs and methods, they are not to be resolved by producing a justificatory argument from some special foundation. S is to start with his or her current beliefs and methods and making minimal adjustments to them so that the resulting beliefs and methods become more coherent with each other, aiming at something like what Rawls (1971, citing Goodman, 1955) calls a “reflective equilibrium.”
Given general foundationalism there is no general epistemological problem about the justification of beliefs about objects in the external world. S believes there are various things external to himself or herself; having such beliefs, S is prima facie justified in having them.
S faces no special “problem of induction.” S uses certain procedures for making inductive inferences. Whatever procedures they are, they are procedures that S starts out using, so they are part of S’s initial foundation and S is prima facie justified in using those inductive methods in the absence of specific reason not to.
Similarly, S faces no special “problem of other minds.” S has various beliefs about other people and their mental states. So, S is prima facie justified in these beliefs. S makes inferences about the mental lives of others on the basis of their behavior. Whatever methods S uses to make these inferences are methods S uses and so methods that S is prima facie justified in using.
With respect to testimony, S’s initial acceptance of B based on testimony relies on the sorts of complex inferences mentioned earlier. Later, when S simply remembers that the battle of Hastings was in 1066 without remembering what led to that belief, this belief has become foundational for S.
It might be objected that general foundationalism counts too many beliefs as justified. People sometimes are unjustified in their beliefs. But how can general foundationalism account for that?
The answer is that general foundationalism counts a belief B of S’s as unjustified only when S’s believing B conflicts with other things that S believes or ought to believe given methods of reasoning that S uses. Given such conflict, which (if any) beliefs or methods are justified all things considered for S depends on how S ought to resolve the conflict in trying to achieve reflective equilibrium.
Suppose that S unjustifiably decides to believe B. Once S believes B mustn’t the general foundations framework wrongly count S justified in believing B?
The answer is that, if S comes to think that his or her reasons for coming to believe B were insufficient, then that is a reason for S to doubt the belief. In other words, there is an internal conflict in S’s beliefs between continuing to believe B and believing that the belief was accepted for insufficient reasons. In this case, S is justified in giving up B and so not justified in continuing to believe B (assuming that S has not come up with new good reasons to believe B).2
Suppose however that S continues to believe B and forgets how it was arrived at. Suppose also that nothing S believes at the later time counts against continuing to believe B. Then general foundationalism does count S justified in continuing to believe B, for B is in the same category as other beliefs of S for which S has forgotten the original reasons.
We normally suppose that perception plays a special role in justification, as illustrated in such remarks as “Seeing is believing” and “I’m from Missouri; show me!” Perception plays a special role in special foundationalism because perceptual beliefs are part of the foundations of justification and because, at least in certain versions of special foundationalism, perceptual and other experiences have special relevance to what one is justified in believing. And it may seem that there is no way to capture the special role of perception in a general foundations framework.
But, if we have a practice of treating observation as special, then according to general foundationalism we are prima facie justified in doing so. Furthermore, suppose that perceptual experience has content and represents the environment in a certain way and suppose that we normally believe things are as represented by our perceptual experience, perhaps by using our perceptual representations to get around in the environment, etc. (Harman, 1990). Then we are prima facie justified in doing this and so prima facie justified in believing that the environment is as represented by perceptual experience.
In a foundations approach, S’s religious beliefs are justified only if S has a justification for them. According to general foundationalism, all of S’s current beliefs are prima facie justified, including S’s religious beliefs, and the mere fact that S cannot find a foundational justification for those beliefs does not prevent S from being justified in believing them. Some philosophers have thought this was a serious objection to general foundationalism. I myself do not agree.
General foundationalism must indeed suppose that S’s religious beliefs are prima facie justified in the absence of some specific challenge to them deriving from S’s other beliefs and inferential practices. It may be that S is aware that certain other people have religious beliefs that conflict with S’s religious beliefs in various ways, and S may come to think that he or she is no better placed to know which of these beliefs are true than those others do, which might be enough to give S reason to become somewhat agnostic. But maybe not.
Someone with paranoia may have a system of beliefs and methods of reasoning that are in excellent reflective equilibrium. According to general foundationalism, such a person is completely justified in his or her beliefs and inferential methods, which may seem to be a bizarre result.
I am not sure what to say about this objection. There is certainly much that is wrong with paranoid beliefs and thinking, but is the trouble that a paranoid in reflective equilibrium is not justified in his or her beliefs in the relevant internalist sense? I am inclined to think that the situation is more like that of someone who is color blind without having any reason to think that he or she is color blind.
Suppose that S starts out believing that beliefs about the external world are unjustified. Perhaps S starts out accepting special foundationalism and sees that this entails that beliefs about the external world are unjustified. So general foundationalism implies that S is prima facie justified in being a skeptic about the external world. Doesn’t that mean S should abandon all beliefs about things in the external world?
No. S’s skeptical opinions conflict with other things that S believes. The best resolution of this conflict is for S to abandon skepticism and the special foundationalism that leads to it. Not because there is a foundational justification for S’s beliefs about things in the external world but because the way to resolve such conflicts is by finding the minimal adjustment in S’s beliefs and methods that provides the most coherent upshot.
Ruth Chang (personal communication) suggests that at least in some cases, believing B may involve believing that there exists a justification of some sort for B.
If so, when S is immediately justified in believing B, S may also be immediately justified in believing that there exists a justification for B. It would be no objection to S’s believing B that S cannot come up with such a justification, but if S were to come to think that there is no justification for B, that would be an objection to S’s continuing to believe B.
If someone thought that all beliefs were like this, so that believing anything involved believing that there was a justification for that belief, and also thought that for most of our beliefs there is no justification of the relevant sort, then there would be the sort of internal conflict that is best resolved by abandoning the idea that believing anything involves believing that there is a justification for that belief.
Chang makes her proposal only about certain normative beliefs, especially moral beliefs. Her idea is that if one believes that certain conduct, for example, incest, is morally wrong, then one is committed to supposing that there is a justification of a certain sort for thinking that conduct wrong. S may not be able to formulate that justification, but if S were to become justified in thinking that there was no such justification, S would no be longer justified in believing that incest is wrong.
I don’t know what to say about Chang’s idea. For certain moral beliefs and perhaps even more so for certain legal beliefs, it seems to me to have a certain plausibility. Even here, however, if one were to decide that the needed justifications are not forthcoming, then again I suggest that it might sometimes be more reasonable to abandon Chang’s principle than to give up the moral and legal beliefs.
I have contrasted the special foundations framework of justification with the no-foundations framework, on the one side, and the general foundations framework on the other side. The frameworks differ in the answers they give to the question, “Which of a person’s beliefs and inferential methods require justification in terms of other beliefs and/or methods?” The no-foundations framework answers that all beliefs and methods require such justification. The special foundations framework answers that, while some beliefs and methods may require such justification, others (the foundational beliefs and methods) do not. The general foundations framework answers that none of one’s beliefs and methods require such justification, at least in the absence of special reasons to distrust such beliefs or methods. I have argued in favor of the general foundations framework on the grounds that intractable skeptical problems arise for the first two approaches that do not arise for the third.3