EPISTEMIC CIRCULARITY: MALIGNANT AND BENIGN
Consider the following dialogue:
Juror #1: You know that witness named Hank? I have doubts about his trustworthiness.
Juror #2: Well perhaps this will help you. Yesterday I overheard Hank claiming to be a trustworthy witness.
Juror #1: So Hank claimed to be trustworthy did he? Well, that settles it then. I’m now convinced that Hank is trustworthy.
Is the belief of Juror #1 that Hank is a trustworthy witness justified? Most of us would be inclined to say it isn’t. Juror #1 begins by having some doubts about Hank’s trustworthiness, and then he comes to believe that Hank is trustworthy. The problem is that he arrives at this belief on the basis of Hank’s own testimony. That isn’t reasonable. You can’t sensibly come to trust a doubted witness on the basis of that very witness’s testimony on his own behalf.
Now consider the following soliloquy:
Doubter: I have some doubts about the trustworthiness of my senses. After all, for all I know, they are deceiving me. Let’s see ... Hey, wait a minute. They are trustworthy! I recall many occasions in the past when I was inclined to hold certain perceptual beliefs. On each of those occasions, the beliefs I formed were true. I know that because the people I was with confirmed to me that they were true. By inductive reasoning, I can safely conclude from those past cases that my senses are trustworthy. There we go. It feels good to have those doubts about my senses behind me.
Doubter seems to be making the same mistake Juror #1 was making. He is relying on a doubted source to confirm its own trustworthiness. For he depends on the testimony of the people around him to confirm the trustworthiness of his senses. But it was only by relying on his senses that he believed there were people with him and that they confirmed what his senses told him.
Epistemic circularity, as I shall be thinking of it, arises in connection with the formation of beliefs about the trustworthiness or reliability of one’s own belief sources. Epistemic circularity can afflict both arguments and beliefs. Suppose I form a belief – either inferentially or noninferentially – in the trustworthiness of one of my belief sources, X. If, in the formation of that belief, I depend upon X, then that belief is infected with epistemic circularity. (I’ll call beliefs infected with epistemic circularity ‘EC-beliefs’.) Likewise, if I propose an argument for the conclusion that X is trustworthy and my belief in one or more of the premises depends upon X, then that argument is epistemically circular. And if I base my belief in the argument’s conclusion on such an argument, then my belief in that conclusion is an EC-belief. Although the most common examples of EC-beliefs are like Doubter’s in that they are based on epistemically circular arguments, it is possible (as I shall show later in the paper) for there to be noninferential EC-beliefs that aren’t based on arguments of any sort.
Doubter’s belief that his senses are trustworthy – like the belief of Juror #1 that Hank is to be trusted – seems to be unjustified. And the reason appears to be that it is an EC-belief. That suggests that epistemic circularity is a bad thing. (To say that epistemic circularity is a bad thing is just to say that it prevents beliefs infected by it from being justified.) In this paper, I will be arguing that – appearances to the contrary notwithstanding – epistemic circularity is not, in itself, a bad thing.
In section I, I will argue that, given the plausible assumption that justified belief isn’t impossible, it follows that justified EC-belief is possible – which entails that epistemic circularity is not, in itself, a bad thing. In section II, I will give an account of the difference between malignant and benign epistemic circularity and then use that account to explain why it is so tempting to mistakenly think epistemic circularity is, in itself, a bad thing. In the final section, I will explain why a good way to complete my defense of epistemic circularity is to add to it a Reidian account of how we know that our belief sources are to be trusted.
I. In Defense of Epistemic Circularity
My defense of epistemic circularity runs as follows: It is exceedingly plausible to think that justification isn’t impossible. But this commits you to a certain foundationalist thesis. And that thesis commits its sensible adherents to thinking justified EC-beliefs are possible.
A. An Objection to Reliabilism
An important component of my defense of epistemic circularity is modeled after a key premise in a recent objection to reliabilism, a premise according to which reliabilism’s adherents are committed to approving of track record arguments for the trustworthiness of our faculties (which tend to be epistemically circular arguments). A track record argument is an argument of the sort employed by Doubter in the soliloquy presented above. In such an argument, one infers from the past success of a belief source in producing true beliefs – i.e., its track record – that it is a trustworthy source. Such arguments tend to be epistemically circular because, in order to confirm that the source in question did produce true beliefs on those past occasions, the subject seems forced to depend on that very source. This objection to reliabilism concludes that, because reliabilists are committed to approving of epistemically circular track record arguments, reliabilism is to be rejected. Richard Fumerton proposes this sort of objection to reliabilism in his 1995 and Jonathan Vogel has also objected to reliabilism in this way in his 2000.
Why think reliabilists are so committed? Take as an example a simplistic version of reliabilism according to which a belief is justified if and only if it is reliably formed. From this it follows that a perceptual belief is justified if it is reliably formed, even if the person holding it doesn’t hold the belief that her sense perception is reliable. Consider the implications of this position for an epistemically circular track record argument that your sense perception is reliable. Suppose that you believe each of the premises of such an argument and that each of those beliefs is reliably formed, including the one(s) formed via sense perception. (Since the argument is epistemically circular, there will be at least one such premise.) It follows that (a) you’ve got a rather simple and inductively valid argument for the conclusion that sense perception is reliable and that (b) your beliefs in its premises are justified. What could keep you from forming, on the basis of this argument, a justified belief in the conclusion? It isn’t reasonable to deny that justification can be transferred from premises to conclusion via this simple inductively valid argument. This is why reliabilists are viewed, by critics such as Fumerton and Vogel, as being committed to approving of epistemically circular track record arguments, that is, as committed to thinking that beliefs based on such arguments can be justified beliefs.
In what follows, I will argue for a similar conclusion, namely, that anyone who accepts the foundationalist thesis:
F: There can be noninferentially justified beliefs
should think that epistemically circular track record arguments needn’t be a bad thing. On the basis of this conclusion, I will defend epistemic circularity by arguing that since F is exceedingly plausible, the fact that it commits its sensible adherents to approving of (at least some) epistemically circular arguments suggests that epistemic circularity isn’t, in itself, a bad thing.
B. A Reason for Thinking Epistemic Circularity isn’t a Bad Thing
My reason for thinking that epistemic circularity isn’t a bad thing depends, therefore, on two points: first, that (sensible) proponents of F are committed to thinking that epistemic circularity is acceptable under certain circumstances and, second, that F is an exceedingly plausible view. Let’s consider the first point first.
Whether you are an internalist or an externalist, if you endorse F, you think a belief can be justified noninferentially. Now suppose you also thought the following were compossible:
(a) A subject S has several belief sources, X1-Xn, each of which directly produces noninferentially justified beliefs.
(b) On the basis of the noninferentially justified beliefs produced by those belief sources (including beliefs produced by X1), S relies on valid inductive or deductive reasoning to infer (and, thereby, to come to believe for the first time) that source X1 is reliable.
(c) Justification transfers from an argument’s premises to its conclusion when S relies on valid inductive or deductive reasoning to infer the conclusion (which is then believed for the first time) from the already justified premises (unless S is disposed to resist believing that conclusion once she considers it).
(d) S is not disposed to resist believing, once she considers it, the proposition that X1 is reliable.
(a)-(c) entail that, so long as the “unless” clause in (c) isn’t satisfied, there exists a justification-producing epistemically circular track record argument. And (d) tells us that the “unless” clause isn’t satisfied in the case in question. Thus, if you allow for the compossibility of (a)-(d), you will be committed to thinking that it is possible for there to be justified EC-beliefs. From this we may conclude that any proponent of F who thinks (a)-(d) are compossible is committed to thinking that epistemic circularity needn’t be a bad thing.
It’s difficult to see why any proponent of F would think either (a) or (d) is impossible. And once we consider the example of a track record argument of the sort employed by Doubter, it’s also difficult to see why a proponent of F would think (b) is impossible. (c) is a little more complicated. But consider a person with some justified beliefs who sees that those beliefs genuinely imply p – a proposition she has never before considered yet to which she feels no resistance when she does consider it. If she comes to believe p on the basis of her recognition that it is implied by her previously held justified beliefs, how could her belief that p fail to be justified on the basis of that inference?
Furthermore, it’s difficult to see why a proponent of F who thinks each of (a)-(d) is possible would think they aren’t compossible. It seems, therefore, that sensible proponents of F will think that (a)-(d) are compossible. But then it follows that sensible proponents of F are committed to allowing that there are possible circumstances in which epistemic circularity is acceptable.
I now want to argue that the combination of F with the view that (a)-(d) are compossible is far more plausible than its competitors. I’ve already explained why it is much more plausible than the combination of F with the view that (a)-(d) aren’t compossible. But it’s also far more plausible than the denial of F. To deny F is to say that all justification must be inferential. Unfortunately, the denial of F together with a few widely accepted premises leads directly to the radical skeptical position that justified belief is not merely unavailable but impossible:
1. A belief can be justified only if it is inferentially justified. [i.e., ~F]
2. A belief can be inferentially justified only if the belief from which it is inferred is a justified belief.
3. Therefore, either a belief can be justified via circular reasoning or it can be justified via an infinite chain of reasoning or no belief can be justified. [from 1 and 2]
4. Our beliefs can’t be justified via circular reasoning.
5. Our beliefs can’t be justified via infinite chains of reasoning.
6. Therefore, no belief of ours can be justified. [from 3, 4 and 5]
Since this argument is valid and premises 2, 4 and 5 are so widely accepted, the conclusion that denying F implies radical skepticism will also seem very plausible. And since the radical skeptical conclusion seems so implausible, F has been a very popular view among contemporary epistemologists. Even coherentists accept it when they propose holistic coherentism in place of linear coherentism.
It seems, therefore, that the options for those who want to insist that epistemic circularity is, in itself, a bad thing are the following:
I. Accept F but insist that (a)-(d) are not compossible.
II. Reject F and endorse one or more of the following:
· our beliefs can be justified by being inferred from unjustified beliefs
· our beliefs can be justified via circular reasoning
· our beliefs can be justified by an infinite chain of reasoning
· none of our beliefs can be justified.
The fact that each of those options is so unpalatable gives us a good reason for thinking that epistemically circularity needn’t be a bad thing.
II. Why Epistemic Circularity Seems Like a Bad Thing
I’ve argued in the previous section that epistemic circularity needn’t be a bad thing. But if that conclusion is right, what should we say about the cases of Juror #1 and Doubter? Surely their beliefs were unjustified. And surely that had to do with the fact that they were EC-beliefs. So how can I defend epistemic circularity when, in light of those examples, it seems like such a bad thing?
In this section of the paper I will explain why it (misleadingly) seems that epistemic circularity is, in itself, a bad thing. The explanation comes down to this: there are some contexts (e.g., Doubter’s context) in which EC-infection is malignant; and, for a variety of reasons, when we think about epistemic circularity, we are inclined to think about its occurrence in such contexts.
A. Questioned Source Contexts and Unquestioned Source Contexts
The sort of belief that can get infected with epistemic circularity is a belief that one’s belief source, X, is trustworthy. A context in which epistemic circularity is a bad thing is one in which the subject begins by doubting or being unsure of X’s trustworthiness. For example, someone who has recently been persuaded, by some skeptical argument, that her sense perception is unreliable might be considering an argument that has been proposed to convince her that her perception is trustworthy after all. If she can see that the argument is epistemically circular, she will (if she’s sensible) consider it to be useless as a means to help her regain lost confidence in her perception. The reason is simple. She doesn’t trust perception. She thinks it is unreliable, or, at the very least, she is uncertain about whether it is reliable. So long as that is the case, it wouldn’t be reasonable for her to depend on the testimony of perception to learn things. What she is (or should be) looking for is some other testimony – some testimony that is independent of perception – that supports the reliability of her sense perception. This is just the sort of context that Doubter is in. That is why we conclude that epistemic circularity is a bad thing in his case.
But not all contexts are like that. Suppose that a person who has no doubts at all about the trustworthiness of her sense perception – in fact, she has never before considered the proposition that her sense perception is reliable – comes to believe that her sense perception is reliable. We can add, if we like, that although she has no doubt that this newly acquired belief is justified, she wonders how she came to hold that belief and, also, how it came to be what she thinks it obviously is, namely, a justified belief. And we can note that, even if she discovers that she formed the belief in a way that involved epistemic circularity, there is no reason for her to be troubled by this. For she wasn’t looking for some independent verification of the reliability of her senses. She was merely curious about how it was that she came to hold, with justification, the obviously justified belief that her sense perception is reliable.
What we have, in the previous two paragraphs, are two different contexts in which a belief in the trustworthiness of a belief source is formed. The first is a context in which the subject has doubt or is uncertain about the source’s trustworthiness. As such, it is what I’ll call a ‘questioned source context’. In the second context, the subject has no such doubt or uncertainty so it isn’t a questioned source context. In virtue of its not being a questioned source context it is what I’ll call an ‘unquestioned source context’. My proposal is that epistemic circularity in a questioned source context is malignant and that epistemic circularity in an unquestioned source context is benign.
What explains this difference? Well, in a questioned source context, the EC-belief in a source X’s trustworthiness is produced, at least in part, by a source (i.e., X) whose trustworthiness the subject questions. Because she questions X’s trustworthiness, the subject has an undercutting defeater for all her beliefs produced (even in part) by source X, including the belief that X is a trustworthy source. And this defeater keeps the belief in X’s trustworthiness from being justified. Since epistemic circularity in a questioned source context prevents (in the way just described) the belief it infects from being justified, it is, in that context, a bad thing. But there is no reason for thinking EC-beliefs in unquestioned source contexts are defeated in this way. Furthermore, we have the reason given in section I for thinking epistemic circularity needn’t be a bad thing.
In most discussions in the literature concerning beliefs in a source’s trustworthiness, the participants have in mind questioned source contexts, such as the contexts of Juror #1 and Doubter. Even if the participants themselves have no doubts about the trustworthiness of sense perception, when they are discussing how one can justifiedly come to believe that sense perception is reliable they are interested in ways a person in a questioned source context can come to hold such a belief. Perhaps this is because they are seeking to identify a way that anyone – whether in a questioned source context or not – can sensibly come to believe in the trustworthiness of perception. At any rate, because such discussions almost always focus on questioned source contexts, there is a strong inclination to think epistemic circularity is a bad thing. And since epistemic circularity is a bad thing in such contexts, that inclination is understandable. The problem is that such discussions typically fail to consider unquestioned source contexts and to give some reason for thinking that epistemic circularity is a bad thing in them.
B. Track Record Arguments
Another reason it is so tempting to think epistemic circularity is, in itself, a bad thing is that it is exceedingly difficult to imagine a context in which a belief in a source’s trustworthiness formed via an epistemically circular track record argument is justified. A defender of epistemic circularity might say, as I have, that a belief formed via such a track record argument would be unproblematic if formed in an unquestioned source context. But that claim seems unconvincing for at least two reasons.
First, an argument is commonly evaluated in terms of how useful it would be in convincing someone who initially doubts its conclusion. An epistemically circular track record argument fails abysmally by that standard. It is of no use whatsoever to anyone who begins by questioning its conclusion. For this reason, epistemically circular arguments seem to be pathetic arguments. And it is difficult to imagine a context in which it would be a good thing to depend upon a pathetic argument.
Second, it seems that most people who believe in the trustworthiness of their belief sources do not arrive at such beliefs via a track record argument. For one thing, many people will be tempted (in light of the considerations just mentioned) to think very poorly of such arguments. That will incline them not to employ them. For another, many people just don’t take the time to form the required beliefs in the premises of track record arguments and to base a belief in the trustworthiness of some belief source on such an argument. Instead, they seem to form noninferential beliefs in the trustworthiness of their belief sources without basing such beliefs on track record arguments (or any arguments for that matter).
So, the suggestion that epistemically circular track record arguments would be acceptable in an unquestioned source context seems implausible, in part because no one seems actually to rely on such arguments in any context and in part because, as arguments go, they seem so useless that they shouldn’t be relied upon in any context.
But doesn’t this spell trouble for all those who accept F and that (a)-(d) are compossible? Aren’t they committed to approving of these dismal arguments if they’re used in an unquestioned source context? Well, it all depends on what it means to say they approve of epistemically circular track record arguments. What they are committed to saying is the following:
One can justifiedly come to believe the premises of a track record argument before believing – or even considering – its conclusion. And one can then come to see that the conclusion can be validly inferred from its premises. At this point, one can make the relevant inference and, thereby, come to justifiedly believe the conclusion for the first time (on the basis of such an inference).
But they aren’t committed to the suggestion that such arguments are useful for convincing doubters (they can even agree that, in that sense, such arguments are pathetic). Nor are they committed to the suggestion that track record arguments are commonly (or likely to be) relied upon by people in coming to believe that their faculties are reliable.
III. Epistemic Circularity and Common Sense
A nagging question remains for the defender of epistemic circularity: Given the implausibility of the suggestion that justified EC-beliefs in a source’s trustworthiness are in fact formed on the basis of track record arguments, how are such beliefs formed in unquestioned source contexts? Without a satisfactory answer to this question, my defense of epistemic circularity leaves us without an account of how people actually know (as they sometimes seem to) that their belief sources are trustworthy. I propose that we handle this difficulty by looking to Thomas Reid for inspiration.
A. Reidian Common Sense
According to Reid, it is a first principle that our faculties are reliable. And first principles, says Reid, are properly believed noninferentially. Just as we have noninferential knowledge about our immediate physical environment by means of sense perception and about our past by means of memory and about our own minds by means of introspection, so also we have a faculty by means of which we have noninferential knowledge of first principles. Reid thinks of first principles as self-evident truths. He thinks some are contingent and some are necessary. The one mentioned above (concerning the reliability of our natural faculties) is contingent. And the faculty by which we know these first principles (whether necessary or contingent) he calls ‘common sense’.
How exactly does this faculty of common sense work? What is the process by which it leads us to beliefs in first principles? Sense perception seems to work as follows: we have sensory experiences and, on the basis of such experiential evidence, we form noninferential perceptual beliefs. Something similar can be said about the faculty of common sense:
We may observe, that opinions which contradict first principles are distinguished from other errors by this; that they are not only false, but absurd: and, to discountenance absurdity, nature has given us a particular emotion, to wit, that of ridicule, which seems intended for this very purpose of putting out of countenance what is absurd, either in opinion or practice.
The idea is that when we entertain the contrary of a first principle, we experience what Reid calls ‘the emotion of ridicule’. On the basis of this experience we do two things: we dismiss as absurd the contrary of the first principle and we believe the first principle itself. Thus, noninferential common sense beliefs, like noninferential perceptual beliefs, are based on experiential evidence.
According to Reid, how is it that we come to know that our faculty of common sense is reliable? Recall that the first principle Reid mentioned earlier was that all our natural faculties are reliable; this includes the faculty of common sense itself. Thus, as I understand him, Reid thinks that one noninferential output of the faculty of common sense is the experience-based belief that the faculty of common sense itself is reliable. This is where the epistemic circularity enters in on Reid’s account.
Reid’s account is an improvement over “track record argument” accounts of how we come to know that our belief sources are trustworthy. For one thing, the suggestion that our EC-beliefs in the trustworthiness of our belief sources are formed noninferentially is much more in accord with what actually seems to take place than is the suggestion that such beliefs are based on track record arguments. We don’t seem to formulate to ourselves arguments that our faculties are reliable. Rather, we seem just to take it as obvious (without inference) that our faculties are reliable. Another virtue of Reid’s account of how we know our belief sources are reliable is that, because he doesn’t say these beliefs are argument-based, he is able to avoid the implausible suggestion that EC-beliefs are acceptable even in questioned source contexts. For these reasons, a Reidian common sense account nicely completes my defense of epistemic circularity: it answers the nagging question mentioned at the opening of section III without the troublesome problems accompanying the suggestion that we rely on track record arguments.
There are, of course, many objections to the Reidian account just outlined. I haven’t the space to deal with them all here. So I’ll focus on the anti-Moorean complaint according to which commonsensism is not even in the ballpark of philosophically respectable responses. In proposing this objection, certain nonskeptics join with skeptics in accusing Moore (and Reid) of question-begging – of making the philosophically irresponsible move of merely asserting things for which one needs to argue.
By way of response, notice first that it isn’t considered philosophically irresponsible (at least not by anti-Moorean nonskeptics) for a philosopher, in a context in which she has no doubts about the reliability of her faculty of sense perception, to believe one of its outputs (e.g., that there is a police car with flashing lights following her) even though she hasn’t independently confirmed its reliability. So why think it is philosophically irresponsible for her, in a context in which she has no doubts about the reliability of her faculty of common sense, to believe one of its noninferential outputs (e.g., that all of her faculties are trustworthy) when she hasn’t independently confirmed its reliability?
Perhaps the anti-Moorean’s complaint is that, by responding to the skeptic’s charges, the defender of common sense is suggesting that her views could be used in an argument aimed at persuading someone in a questioned source context to give up her skepticism. But Reid is clear that philosophy, arguments and logic are of no use in delivering someone in a questioned source context from her skepticism. So Reid would concede that if by a “philosophically responsible response to the skeptic” one means a response that will be useful in a questioned source context, then he doesn’t have such a response. Nevertheless, the commonsensist can sensibly point out that there is no good reason to think that a philosophically responsible response to the skeptic’s charges must be one that is useful in a questioned source context.
In reply, the anti-Moorean might point out that when epistemologists are considering the problem of skepticism, the answers they are looking for are ones that will be acceptable in questioned source contexts. But if that’s true, then, unless at least some belief sources are trusted before their reliability is verified, the game is over. We simply can’t check on the reliability of our belief sources without relying on our belief sources. But if it is acceptable (as many worried about skepticism concerning the external world seem to think) to trust rational intuition or introspection or direct acquaintance without first verifying their reliability, why isn’t it acceptable to trust sense perception and common sense without first verifying their reliability? Just as a direct acquaintance theorist (like Fumerton) would consider himself artificially handicapped if asked to explain how one can know, without relying on direct acquaintance, that direct acquaintance is reliable, so also the commonsensist considers herself artificially handicapped when asked to explain how one can know, without relying on common sense, that one’s faculty of common sense is reliable.
A central disagreement between nonskeptics and skeptics (of various kinds) is over which, if any, of one’s belief sources can sensibly be trusted before they have been verified. The disagreement between commonsensists and nonskeptical anti-Mooreans is over what to do when skeptics and nonskeptics disagree in that way. The anti-Moorean nonskeptics think we need to rely on philosophy (i.e., ways of knowing employed in doing philosophy) in responding to such skeptics. The commonsensists happily rely on philosophy when thinking about what makes their beliefs in the reliability of their (trusted) sources justified. But when it comes to helping skeptics trust their faculties, commonsensists think philosophy is useless. Nature needs to take over and cause the skeptics to trust their cognitive faculties. At best, philosophy can help them see that they have no good reasons to mistrust their faculties.
Commonsensism, of the Reidian sort advocated here, gives an account of what makes our EC-beliefs in the trustworthiness of our faculties justified. But it makes no pretense of providing an account that will convince a skeptic in a questioned source context. This isn’t because the Reidian is being philosophically irresponsible. It’s because she is being consistent with the implications of her position. It’s because she recognizes the dialectical situation for what it really is.
Alston. 1986. “Epistemic Circularity.” Philosophy and Phenomenological Research 47:1-30. Reprinted in Alston 1989.
________. 1989. Epistemic Justification: Essays in the Theory of Knowledge. Ithaca: Cornell University Press.
________. 1993. The Reliability of Sense Perception. Ithaca: Cornell University Press.
Bergmann, Michael. 2002. “Commonsense Naturalism.” In James Beilby (ed.), Naturalism Defeated? Essays On Plantinga's Evolutionary Argument Against Naturalism, pp.61-90. Cornell University Press.
________. Unpublished. “Defeaters and Higher-Level Requirements.”
Cohen, Stewart. 2002. “Basic Knowledge and the Problem of Easy Knowledge.” Philosophy and Phenomenological Research 65:309-29.
Fumerton, Richard. 1995. Metaepistemology and Skepticism. Lanham, Maryland: Rowman and Littlefield.
________. Forthcoming. “Epistemic Internalism, Philosophical Assurance and the Skeptical Predicament.” In Thomas Crisp, Matthew Davidson, and David Vanderlaan (eds.), Knowledge and Reality: Essays in Honor of Alvin Plantinga. Kluwer Academic Press.
Goodman, Nelson. 1983. Fact, Fiction and Forecast (4th edition). Cambridge, MA: Harvard University Press.
Klein, Peter. 1999. “Human Knowledge and the Infinite Regress of Reasons.” In James Tomberlin (ed.), Philosophical Perspectives, 13, Epistemology, pp. 297-325. Cambridge, Massachusetts: Blackwell Publishers.
Lemos, Noah. Forthcoming. The Commonsense Tradition: A Contemporary Perspective.
Plantinga, Alvin. 1993a. Warrant: The Current Debate. New York: Oxford University Press.
________. 1993b. Warrant and Proper Function. New York: Oxford University Press.
________. 2000. Warranted Christian Belief. New York: Oxford University Press.
Pollock, John. 1990. Nomic Probability and the Foundations of Induction. New York: Oxford University Press.
Reid, Thomas. 1969 . Essays on the Intellectual Powers, introduction by Baruch Brody. Cambridge: MIT Press.
________. 1970 . An Inquiry into the Human Mind, edited with an introduction by Timothy Duggan. Chicago: University of Chicago Press.
Sorenson, Roy. 1991. “‘P, Therefore P’ Without Circularity.” The Journal of Philosophy 88: 245-66.
Vogel, Jonathan. 2000. “Reliablism Leveled.” The Journal of Philosophy 97:602-23.
 You might think that by using a lie detector, you can come to trust a doubted witness on the basis of that witness’s testimony on his own behalf. But that is a case where you have testimony from another source confirming that person’s testimony. The other source is that person’s heart rate, etc. which is monitored by the lie detector and which can either confirm or disconfirm claims made by that person. Likewise, some people think that they can look a person in the eye and tell whether that person is lying. What’s happening in such cases is similar to what happens when using a lie detector. The listener is using another source (body language, eye appearance) to confirm or disconfirm the witness’s claim that he is trustworthy.
 To depend upon a belief source X in forming a belief B is for B either to be an output of X or to be held on the basis of an actually employed inference chain leading back to an output of X.
 Our understanding of epistemically circular arguments can benefit by contrasting them with logically circular arguments. An argument is logically circular only if it includes its conclusion as a premise. If we assume that “a belief formed via source X is warranted (i.e., it has what it takes to turn true belief into knowledge) only if X is a trustworthy source,” then we may distinguish logically circular arguments from epistemically circular arguments as follows: a logically circular argument’s conclusion is included as one of its premises whereas an epistemically circular argument’s conclusion must be true if belief in its premises is to be warranted. (Notice that, for epistemically circular arguments, warranted belief in the premises requires only that the conclusion be true, not that it be believed.) See Alston 1986 and 1993, ch. 2 for a characterization of epistemic circularity that is very similar to the one given here.
 A belief B is based on an argument A iff the content of B is A’s conclusion and B is inferred from the content of A’s premises (via the reasoning A lays out). Thus, a belief in A’s conclusion is not based on A (in my sense) when it is based solely on an awareness of some other feature of A (or a particular tokening of A) and is not inferred from the content of A’s premises. Sorenson (1991) identifies several arguments that are logically circular (in my sense – see the previous note), each of which can rationally persuade a person of its conclusion but only if that person’s belief in the conclusion is not due to its being based on the argument in my sense. Here are a few examples he gives (see his 1991, 248-50):
Some arguments are written in black ink.
Some arguments are written in black ink.
There are at least two tokens of an eleven-word sentence.
There are at least two tokens of an eleven-word sentence.
Some arguments are composed solely of existential generalizations.
Some arguments are composed solely of existential generalizations.
 Remember that the examples are to be understood so that just as Juror #1 is relying on Hank’s testimony in arriving at his belief in Hank’s trustworthiness, so also, Doubter is relying on the reasoning noted in arriving at his belief in the trustworthiness of his senses. Prior to that reliance, Juror #1 had doubts about Hank’s trustworthiness and Doubter had doubts about the trustworthiness of his senses. Furthermore, throughout this paper I’ll be assuming (for simplicity’s sake) that beliefs aren’t causally overdetermined by two or more different belief sources (or combinations of sources). Thus, if Doubter’s belief that his senses are trustworthy is based in part on his senses, we may conclude that that belief isn’t also produced (in an overdetermining way) by a distinct set of belief sources that doesn’t include his senses.
 In his 1993, Alston defends this point at length in connection with sense perception and argues for it briefly in connection with other beliefs sources as well.
 Obviously, this objection takes for granted that epistemic circularity is a bad thing. And of course this is where I part company with such objectors. However, as will become clear shortly, my view resembles the view held by this sort of objector to reliabilism insofar as we both select a philosophical thesis and argue that its proponents are committed to approving of some epistemically circular track record arguments.
 Fumerton (in his Forthcoming) has since modified his presentation of this sort of objection to reliabilism.
 A belief is noninferentially justified if it is justified despite its not being based on another belief. Notice, by the way, that although linear coherentists who endorse circular reasoning reject F, holistic coherentists who reject circular reasoning (and infinite chains of reasoning) accept F. For they acknowledge that a belief can be justified noninferentially in virtue of the fact that it coheres with one’s other beliefs. See Plantinga 1993a, ch. 4 and Klein 1999, 298 for a defense of the view that what I’m calling ‘holistic coherentism’ is a version of foundationalism insofar as it disapproves of circular reasoning and infinite chains of reasoning while endorsing F.
 Insofar as I point out that the objection to reliabilism described above applies to a wider target, what I do here is similar to what Stewart Cohen does in his 2002. There he speaks of views that allow for what he calls ‘basic knowledge’. He defines basic knowledge (2002, 309) as knowledge produced by a belief source prior to one’s knowing that the source in question is reliable. Thus, views that allow for basic knowledge are views that reject this principle:
KR: A potential knowledge source K can yield knowledge for S, only if S knows K is reliable (2002, 309).
What Cohen argues (2002, 317-18) is that any view which allows for basic knowledge – i.e., any view which rejects KR – is committed to approving of epistemically circular track record arguments. This conclusion is very similar to my position that any view that accepts F is committed to approving of epistemically track record arguments. For a brief discussion of how my position differs from Cohen’s, see the next footnote.
 Cohen’s strategy (in his 2002) is different from mine. He argues that all those who reject KR (see the previous note) are committed to approving of epistemically circular arguments. But, unlike me, he doesn’t conclude that, since not-KR is extremely plausible, epistemic circularity needn’t be a bad thing. Instead, he recommends endorsing KR and he attempts to explain how one can do so while avoiding both epistemic circularity and skepticism.
The problem is that, as Cohen understands KR, it is consistent with the following scenario:
Scenario A: A potential knowledge source K has, as one of its noninferential outputs, the belief that K is reliable. Despite the fact that that belief’s only source is K, it counts as noninferential knowledge.
This scenario is consistent with Cohen’s reading of KR because Cohen does not read KR as saying “… only if S first knows K is reliable”. Instead, he reads it as saying “… only if S doesn’t wait until later to know K is reliable” (see his 2002, 323-25). In Scenario A, S doesn’t wait until after K produces knowledge before knowing that K is reliable. Instead, knowing that K is reliable is simultaneous with K first producing knowledge, since the fact that K is reliable is (or is a component of) the first knowledge K produces. By understanding KR in this way, Cohen is able to avoid the skeptical implications typically associated with KR (although I would say that those skeptical implications are typically associated with KR only when it is understood as saying “… only if S first knows K is reliable”). But Cohen is mistaken in thinking that this way of understanding KR enables him to endorse KR without permitting epistemic circularity. It’s true that, in Scenario A, S doesn’t depend on source K in order to know something before finding out that K is reliable. But S still depends on K in order to find out (for the first time) that K is reliable. Clearly, that is as much a case of epistemic circularity (or bootstrapping or whatever you want to call it) as is the juror case given in the opening paragraph of my paper.
 Both here and in (c), I’m understanding valid inductive reasoning in such a way that, in addition to satisfying the obvious formal constraints on valid inductive reasoning, it must also satisfy the following two conditions: (i) it doesn’t involve the projection of unprojectable properties (such as Goodman’s grueness and bleenness) and (ii) it isn’t impossible for counterexamples (to the inductively inferred conclusion) that are in fact possible to show up in the sample class on which the induction is based (so we can’t conclude from the fact that all ravens in our sample class have been observed that all ravens have been observed, since, even if there are unobserved ravens, it is impossible for unobserved ravens to show up in our sample class). See Goodman 1983 chs. 3-4, Plantinga 1993b ch. 7, and Pollock 1990 chs. 3-5 for a discussion of these conditions on appropriate inductive reasoning.
 Notice that this parenthetical clause rules out the possibility of justification transferring from premises to conclusion in logically circular arguments. For in logically circular arguments, the conclusion is just a repeat of one of the premises. So coming to believe that conclusion on the basis of the premises of the logically circular argument won’t be a case of coming to believe the proposition in question for the first time.
 The reason for the “unless” clause is this. If you come to see that some beliefs you are justified in holding entail a proposition you have never before considered but which you find utterly absurd (i.e., now that you consider it, you think that its falsity is much more plausible than the premises you were justified in believing), you aren’t thereby justified in believing that absurd conclusion. Instead, you lose your justification for believing one or more of the premises.
 For a belief to be justified via circular reasoning is for it to be justified in virtue of its being based on a logically circular argument (or in virtue of its being based on itself). See note 3 for an explanation of what a logically circular argument is and how it is different from an epistemically circular argument; and see note 4 for an explanation of the relevant sense of “basing a belief on an argument”. See also Plantinga 1993a, ch. 4 for a discussion of what is objectionable about circular reasoning.
 See Plantinga 1993a, ch. 4 and Klein 1999, 298 for a defense of this claim.
 Would taking this option be enough to enable the opponent of epistemic circularity to avoid implausible commitments? Perhaps in order to avoid being committed to approving of epistemic circularity, proponents of F will need to deny (implausibly) the possibility of things besides the joint truth of (a)-(d). If so, then things are even more difficult for the opponent of epistemic circularity than I represent them as being.
 Or at least this is so in Doubter’s case. Strictly speaking, the belief of Juror #1 isn’t an EC-belief since it isn’t a belief in the trustworthiness of one of the subject’s own belief sources.
 Just how much doubt or uncertainty about the source’s trustworthiness is required for the context to count as a questioned source context? If the subject believes that the belief source in question is trustworthy, then there isn’t enough doubt or uncertainty. But if, as a result of her doubt or uncertainty, the subject disbelieves or withholds that proposition about the trustworthiness of her belief source, then there is enough doubt or uncertainty for the context to count as a questioned source context.
 In order to qualify as an unquestioned source context, it is enough that the context isn’t a questioned source context. There is no need for the subject to believe that her source is reliable. It is the absence of sufficient doubt (and sufficient uncertainty) about a source, not the presence of a belief in its trustworthiness, that makes a context an unquestioned source context (see the previous note for a discussion of what counts as sufficient doubt or uncertainty). Thus, disbelieving or withholding (because of doubt or uncertainty) the proposition that one’s belief source is trustworthy results in a questioned source context. An unquestioned source context is one without such disbelieving or withholding – one in which the subject either believes the proposition that her belief source is trustworthy or has no attitude at all toward that proposition, not even the attitude of withholding. (See my Unpublished for a discussion of withholding and how it differs from taking no attitude at all toward a proposition.)
 I don’t have the space here to discuss defeaters and defeater types at length (but see my Unpublished as well as Plantinga 2000, pp. 359-66). For our purposes here it will be sufficient to note that, roughly speaking, a defeater for a belief B is a propositional attitude or experience a person comes to have that makes B cease to be justified. A person has an undercutting defeater for B when her defeater consists in her disbelieving or withholding (due to doubt or uncertainty) the proposition that B’s source is trustworthy. (A person has a rebutting defeater for a belief B when her defeater consists in her having reasons for thinking B is false.)
 Consider, for example, the cases Cohen gives (in his 2002) involving conversations with his son who expresses a lack of confidence in a belief source.
 I suspect that it just doesn’t seem very interesting to most epistemologists to explain how a person with no doubts at all about the trustworthiness of sense perception can come to justifiedly believe that her perception is reliable.
 Notice that if you thought that you had to add – to the original premises of an epistemically circular argument – further premises stating that the sources of your beliefs in the original premises are trustworthy, then, because the original argument is epistemically circular, you’d have to add the conclusion as one of the premises. This would transform the epistemically circular argument under discussion into a logically circular argument (see note 3 above for a definition of logically circular arguments). This provides another possible explanation for why people are inclined to think epistemic circularity is a bad thing. For they may be inclined to think there is some pressure to add (or to be able to add) to the premises of an epistemically circular argument further premises about the trustworthiness of your belief sources. Since doing so turns an epistemically circular argument into a logically circular argument (and since it seems that a belief cannot be justified in virtue of its being based on a logically circular argument), this would incline such people to think epistemic circularity is a bad thing. (See note 4 for an explanation of the relevant sense of “basing a belief on an argument”.)
 I am more interested in developing a line of thought that seems to me to be suggested by Reid than in defending my belief that Reid intended to propose it. I expect that I’d be attracted to the ideas that I (think I) see in Reid even if I became convinced that I haven’t understood him correctly.
 “Another first principle is, that the natural faculties [e.g. sense perception, memory, introspection, etc.], by which we distinguish truth from error are not fallacious.” (Reid 1969, 630)
 Reid 1969, 593.
 “We ascribe to reason two offices, or two degrees. The first is to judge of things self-evident; the second to draw conclusions that are not self-evident from those that are. The first of these is the province, and the sole province, of common sense ... and is only another name for one branch or degree of reason.” (Reid 1969, 567)
Thus, Reid, as I understand him, disagrees with Alston’s conclusion (in his 1993) that one can’t know that sense perception is reliable without relying on sense perception. According to Reid, one can know directly and noninferentially, via the cognitive faculty we have for knowing first principles, that sense perception is reliable. However, as we shall see below, Reid agrees with Alston’s other conclusion, namely, that in knowing our faculties are reliable, we eventually get involved in epistemic circularity of some sort.
 It is important not to be put off by Reid’s name – i.e., ‘common sense’ – for the faculty by which we know first principles. We tend to classify as ‘common sense beliefs’ beliefs that are peculiar to our own culture or upbringing. Reid doesn’t (or at least he doesn’t want to). His intention is to include only propositions that most everyone believes (and knows) noninferentially – things that are immediately accepted by sane persons once considered and understood. That 2 + 2 = 4, that modus ponens is a valid form of inference, that the thoughts of which I am conscious are my thoughts, that I have some degree of control over my actions – these are examples of what Reid considers the dictates of common sense. The first two are examples of necessary truths known by common sense; the latter two are contingent truths. A more familiar name for the faculty by which we have noninferential knowledge of necessary truths is a priori intuition. So the branch of reason Reid calls ‘the faculty of common sense’ encompasses both what we call ‘a priori intuition’ and something akin to it that produces beliefs in contingent rather than necessary truths.
 Reid 1969, 606.
 See his 1969, 632 where he says:
How then come we to be assured of this fundamental truth on which all others rest [i.e., the truth that our natural faculties are not fallacious]? Perhaps evidence, as in many other respects it resembles light, so in this also, that as light, which is the discoverer of all visible objects, discovers itself at the same time: so evidence, which is the voucher for all truth, vouches for itself at the same time.
 Especially not epistemically circular track record arguments which don’t seem to be very impressive as arguments given that they are useless at convincing doubters.
 Recall that arguments are typically viewed as the sorts of things that are useful when the conclusion is in doubt as is the case in a questioned source context.
 It’s worth emphasizing that we don’t need to adopt all aspects of Reid’s account in order to be Reidians. For example, we don’t need to agree with his entire list of first principles. It is enough that we think we have a belief source (which we can call ‘common sense’) by which we know noninferentially that our belief sources, including common sense, are trustworthy. Those philosophers (and this includes the majority of analytic epistemologists) who proceed as Chisholmian particularists do – proposing epistemic principles with the aim of having them fit our commonsense judgments about which beliefs count as justified and which don’t – seem, at the very least, to be friendly to this minimalist sort of Reidian commonsensism.
 But see my 2002 (section VIII) where I outline a defense of a minimalist Reidian account, one that says merely that belief in the trustworthiness of our belief sources can be noninferentially justified. There I also argue (section VII) that we have no reason to think a Reidian account is incompatible with metaphysical naturalism. See also Lemos’s Forthcoming for a qualified defense of a Reidian position.
 See Reid 1970, 10-13 and 81-83 and Reid 1969, 30-31.
 See his 1995, 73-79, 162 and 185.
 Notice, by the way, that although the commonsensist has an account of what makes such beliefs justified, it is no part of her position that one needs justified beliefs in the trustworthiness of one’s faculties in order for beliefs produced by those faculties to be justified.
 Thanks to Jeffrey Brower, Tully Borland, Trenton Merricks, Kevin Meeker, Myron Penner, Alvin Plantinga, Michael Rea, and Matthias Steup for helpful comments on earlier drafts.