Recent Work On The Basing Relation

by

Keith Allen Korcz

(published inThe American Philosophical Quarterly, Vol. 34, No. 2, April,1997. All rights reserved.)
 
 
 

The basing relation is the epistemic relation which holds between a reason and a belief when the reason is the reason for which the belief is held (Harman 1973, 26). The basing relation is important because it marks the distinction between justified and justifiable belief (Pollock 1986, 81). A belief is merely justifiable for a person S when S possesses reasons  sufficient to justify the belief, but has not made any appropriate connection between the reasons and the belief, and consequently remains unjustified in holding the belief. The appropriate connection would be the belief's being based on the reason. A belief is justified for S when S possesses reasons sufficient to justify the belief and has made an appropriate connection between her reasons and her belief (i.e., the belief is based on the reasons).
Presuppositions regarding the basing relation have been essential to a number of important epistemological theories and arguments presented in the literature. For example, important discussions of the principle of closure (e.g., Nozick 1981, 204-207), objections to reliabilism (e.g., Lehrer 1971 and 1990, 168-172; Jacobson 1993), and critiques of holistic coherentism (Bender and Davis 1989, 57-58) have all involved explicit appeals to a causal account of the basing relation (or psychological inference). Theories of the basing relation have also been prominent in some notable theories of epistemic justification (e.g., Alston 1988; Swain 1979). Furthermore, the correct account of the basing relation will help us to understand the relation between basic and non-basic belief, and may shed light on the internalism/externalism controversy.  Despite the fact that a belief's being based on a reason is a necessary condition of the belief's being justified, and despite its usefulness in understanding and developing theories of justification, epistemologists have just recently begun to attend to the intricacies the basing relation (Audi's 1986 paper, discussed below, was the most detailed discussion of the basing relation in the literature prior to the book length treatment given in Korcz 1996). Currently, there is no generally accepted analysis of the basing relation thought to solve what has come to be called the problem of the basing relation (Pollock 1986, 37).
The standard view is that the correct analysis of the basing relation will be some sort of causal analysis. It is important to note that a causal analysis of the basing relation in no way commits one to a causal theory of justification or knowledge, for the basing relation concerns the connection between reasons and belief, not what makes reasons sufficient to justify the belief. The aim of this paper is to discuss recent work on the basing relation published between 1980 and 1995.

1. Causal Theories Of The Basing Relation

1.1 Moser's Theory Of The Basing Relation

Paul Moser briefly sketches a causal theory of the basing relation in his book Knowledge and Evidence (Moser 1989, 156-158). Moser distinguishes between nonpropositional reasons (e.g., sensation or perceptual states) and propositional reasons (e.g., beliefs), and offers a distinct definition of the basing relation for each. I shall focus on his account of propositional reasons, with the understanding that my comments apply mutatis mutandis to his account of non-propositional reasons.  For propositional reasons, Moser's account is as follows (Moser 1989, 157):

S's believing or assenting to P is based on his justifying propositional reason Q = df. S's believing or assenting to P is causally sustained in a nondeviant manner by his believing or assenting to Q, and by his associating P and Q.

Moser limits his account of the basing relation to justifying reasons, unlike the more general characterization of the basing relation I gave above. Moser's definition contains two conditions for a belief's being based on a reason, which I will refer to, respectively, as the causal sustaining condition and the association condition. The idea behind the association condition is that the individual must be aware (and thereafter be disposed to become aware) of the support the reason provides for the belief. Moser defines the association condition as follows:

S occurrently satisfies an association relation between evidence E and belief P = df. (i) S has a de re awareness of E's supporting P, and (ii) as a nondeviant result of this awareness, S is in a dispositional state whereby if he were to focus attention only on his evidence for P (while all else remained the same), he would focus his attention on E. (Moser 1989, 141).

The de re awareness is presumably a non-propositional, direct awareness of S's own mental state (Moser 1989, 142). In addition to occurrent (conscious) association relations, there may be non-occurrent association relations, where condition (i) obtained at some time in the past and condition (ii) currently obtains. Both occurrent and non-occurrent relations may satisfy the requirements of the association condition.
One difficulty for the association requirement is its internalist requirement that in order to base a belief on evidence, one must at some time have been occurrently aware of the evidence as supporting the belief. It seems that we often base beliefs on evidence when we are not aware of the evidence as supporting the belief. For example, I could justifiably come to believe that the sun has spots because I read it in an astronomy textbook, and only later come to consciously realize that, all along, my belief was also based on other evidence I possessed which was necessary for me to be justified in this particular instance, e.g., that the textbook is reliable, that the sun exists, etc. Similarly, one might not be aware of one's reasons if they involve subliminal perceptions, yet we can imagine cases in which a belief is based on such perceptions. For example, suppose patrons of a bar are fed subliminal messages to the effect that the bar sells hot pretzels. Many such patrons may come to believe that the bar sells hot pretzels, even though they are unaware of the subliminal messages that cause their belief. To a patron that reliably responds to such messages we might plausibly attribute knowledge that the bar sells pretzels, and knowledge in light of a reason requires that the known belief be based on the reason.

1.2 Two Standard Objections To Causal Theories Of The Basing Relation

The two standard objections to causal theories of the basing relation such as Moser's are the problem of deviant causal chains and the problem posed by counter-examples such as the case of the gypsy lawyer.
The problem of deviant causal chains plagues causal analyses in general. Alvin Plantinga, John L. Pollock and others argue against causal theories of the basing relation that a reason's causing a belief is not sufficient to make it the case that the belief is based on the reason, even if such causal relations are necessary for a belief to be based on a reason.  Not every causal chain from reason to belief establishes a basing relation, and those that fail to do so are deviant causal chains. Plantinga gives the following illustration of one of the kinds of problems in question:

Suddenly seeing Sylvia, I form the belief that I see her; as a result, I become rattled and drop my cup of tea, scalding my leg. I then form the belief that my leg hurts; but though the former belief is a (part) cause of the latter, it is not the case that I accept the latter on the evidential basis of the former. (Plantinga 1993a, 69n8)

Such counterexamples show that causal theories of the basing relation cannot plausibly assert that basing relations get established by just any causal chain of events from reason to belief.
A second difficulty facing causal theories of the basing relation involves gypsy lawyer style counter-examples, which show that a reason need not cause a belief for the belief to be based on the reason. The classic illustration of the point is that given by Keith Lehrer (Lehrer 1971; 1974, 124-125; 1990, 169).  His original case of the gypsy lawyer goes like this: suppose a series of eight grisly crimes has been committed, all the available evidence indicates that the lawyer's client committed the first seven of those crimes, and everyone believes that he committed all eight crimes. However, the lawyer, being a practicing member of a gypsy religion, has absolute faith in the cards. The cards indicate that his client is innocent of the eighth crime, and the lawyer comes to believe this on the basis of his faith in the cards. The lawyer then searches through the evidence to find a plausible argument presentable in court showing that his client is innocent. The lawyer finds such an argument and it involves a complicated line of reasoning. The lawyer recognizes that the complicated line of reasoning shows that his client is innocent, but the complicated line of reasoning cannot cause the lawyer to believe that his client is innocent due to the emotional factors surrounding the case. Nonetheless, we can plausibly imagine that the lawyer is justified on the basis of the complicated line of reasoning in believing that his client is innocent, given that the lawyer takes the complicated line of reasoning seriously.
The case of the gypsy lawyer is very complicated and often criticized. Alvin Goldman, for example, has claimed that " ... I find this example unconvincing. To the extent that I clearly imagine that the lawyer fixes his belief solely as a result of the cards, it seems intuitively wrong to say that he knows - or has a justified belief - that his client is innocent" (Goldman 1979, 22n8). Similarly, Robert Audi presents several arguments attempting to show that the lawyer is not justified in believing that his client is innocent (Audi 1983, 223-225).
It might initially be thought that concern over whether the lawyer is justified in believing his client is innocent misses the point of the counter-example, since the lawyer's belief that his client is innocent could be based on the complicated line of reasoning even if he is not justified in holding that belief. However, the criticism misses the point of Audi's (and perhaps Goldman's) concern. In the case of the gypsy lawyer, there is no doubt that if the lawyer bases his belief that his client is innocent on the complicated line of reasoning, then he is justified in believing that his client is innocent. As Audi notes: "Granted, what (impersonally) justifies the belief that p [here, the gypsy lawyer's belief that his client is innocent] would, other things being equal, justify S [here, the gypsy lawyer] in his belief that p should S believe p on that basis" (Audi 1983, 222). Audi's notion of impersonal justification is also sometimes called propositional justification, and involves only a logical relation among propositions. S being justified in his belief involves a psychological relation.
We can understand Audi's strategy as involving an argument something like this:

P1 If the lawyer bases his belief that his client is innocent on the complicated line of reasoning, then he justifiably believes that his client is innocent.
P2 The lawyer does not justifiably believe that his client is innocent.
C The lawyer does not base his belief that his client is innocent on the complicated line of reasoning.

Note that the complicated line of reasoning is the only reason the lawyer has that can justify his belief that his client is innocent. The circumstances of the gypsy lawyer are such that P1 is true. So Audi's discussion focuses on the question of whether P2 is true.
Audi presents three reasons to believe that the gypsy lawyer does not justifiably believe that his client is innocent:

...even though S (here the gypsy) has (objectively) good evidence for p [that the client is innocent of the eighth crime], given a contrary verdict from the cards he would (other things equal) have had the false belief that not-p. Second, given his faith in the cards, he would have believed p even if it had been false, indeed, even if, on the basis of the cards, it had not been rendered so much as objectively likely (to any degree) to be true, i.e., very roughly, likely to some degree given the actual facts relevant to p (I leave open how relevance is to be specified). Third, S would have believed other falsehoods about the crime, had the cards pointed to them, e.g. that the client's spouse committed them. These points, especially the second, strongly suggest that S does not justifiably believe p. (...) Surely if one's belief that p is justified by good evidence, it cannot simply be good fortune that one did not believe something false instead. (Audi 1983, 223)

One limitation of these objections is that they stem from the unreliability of the cards. But the point of Lehrer's example is not undermined if we suppose that the cards are completely reliable, the lawyer knows this, and his faith in the cards is rational. We can then understand the complicated line of reasoning to increase the degree to which the lawyer is justified in believing his client to be innocent. Alternatively, we could plausibly hold that the lawyer bases his belief on the complicated line of reasoning even if the reasoning does not significantly effect the degree to which he justifiably believes that his client is innocent. The fact that the lawyer sincerely recognizes that the reason supports the belief indicates that the belief is based on the reason (given that the lawyer is not rationalizing his belief). Either way, the lawyer's belief that his client is innocent is based on the complicated line of reasoning.
Audi has also objected to the case of the gypsy lawyer that the lawyer is merely rationalizing his belief (Audi 1983, 221-227). But while the lawyer could be rationalizing his belief that his client is innocent, he need not. Perhaps he does not doubt the quality of the complicated line of reasoning, nor is he dissatisfied with his original reasons for believing that his client is innocent, and so on, in which case the lawyer could not plausibly be said to be offering a rationalization. The case of the gypsy lawyer, or some plausible variation of it, poses a genuine problem for causal theories of the basing relation. I now turn to a causal analysis of the basing relation designed to avoid gypsy lawyer style counter-examples.

2. Swain's Pseudo-Overdetermination Theory

One of the most widely discussed causal analyses of the basing relation is Marshall Swain's (Swain 1979, 1981, 1985). On Swain's theory, a belief is based on a reason if the reason either causes the belief, or (in a certain sense) would have caused it. The appropriate counterfactual causal relation requires that there be some possible world W in which the belief occurs, the reason occurs, nothing other than the reason causes the belief, and the person's epistemic situation (and everything else) remains the same insofar as this is logically possible.  If the reason causes the belief in W, and W is closer to the actual world than any possible world in which the other conditions are satisfied but the reason does not cause the belief, then the belief is based on the reason in the actual world. Swain calls such counterfactual causal relations pseudo-overdeterminants. His formal definition of pseudo-overdetermination is as follows:

(DPO) Where c and e are occurrent events, c is a pseudo-overdeterminant of e iff:
(1) c is not a cause of e; and
(2) there is some set of occurrent events, D = {d1, d2, ..., dn} (possibly having only one member), such that
(a) each di in D is a cause of e; and
(b) if no member of D had occurred, but c and e had occurred anyway, then there would have been a causal chain from c to e, and c would have been causally prior to e. (Swain 1981, 70).

Swain's account of pseudo-overdetermination must be distinguished from similar but importantly different counterfactual relations with which it is easily confused. In particular, (DPO) does not imply the counterfactual relation (CR): a person S's reason would have caused the belief had the actual causes of the belief not done so. For example, it could be the case that all the conditions of (DPO) are met and, nonetheless, had the actual cause of person S's belief that p not occurred, S's actions would have been effected such that S would have been run over by a bus and so would have never formed the belief that p.
Swain emphasizes that one's epistemic situation must not be altered any more than is absolutely necessary in the possible world picked out by the counterfactual. Swain does not provide a detailed account of what one's epistemic situation includes, but it presumably consists of everything relevant to whether anything one believes is known or justified, including all of one's reasons and beliefs, as well as certain facts about (or states of affairs regarding) one's own mind and mental operations, and certain facts about (or states of affairs in) the world.
Pseudo-overdetermination fits into Swain's overall account of the basing relation as follows:

(DB) S's belief that h is based upon the set of causal reasons r at t =df.
(1) S believes that h at t; and
(2) For every member, ri of R, there is some time tn (simultaneous or prior to t) such that
(a) S has (or had) rj at tn; and
(b) Either
(i) S's having rj at tn is a cause or genuine overdeterminant of S's believing h at t or S's having rj at tn is a pseudo-overdeterminant of S's believing that h at t;
or
(i + 1) for some ri and ti that satisfy condition (i), S's having rj at tn is either a cause or a pseudo-overdeterminant of S's having ri at ti. (Swain 1981, 74, 86-87).

One of Swain's motives for introducing pseudo-overdetermination is to account for counter-examples such as the case of the gypsy lawyer discussed by Lehrer (Lehrer 1971). Swain attempts to solve the problem posed by the case of the gypsy lawyer by arguing that the lawyer's belief that his client is innocent is pseudo-overdetermined by the complicated line of reasoning, hence based on it. We can further clarify the notion of pseudo-overdetermination by examining Kvanvig's objections to Swain's pseudo-overdetermination condition.

2.1. Kvanvig's Objections To Swain's Theory Of The Basing Relation

In his paper "Swain On The Basing Relation," Jonathan Kvanvig presents several objections to Swain's theory (Kvanvig 1985). Kvanvig states that event c counterfactually causes event e if c does not in fact cause e, but would have caused e under certain conditions. Pseudo-overdeterminants are thus a kind of counterfactual cause. Kvanvig observes that counterfactual causes can be buried by each other (Kvanvig 1985, 154-155). One counterfactual cause is buried by another "if the second must fail to obtain in order for the first counterfactual cause to be the actual cause" (Kvanvig 1985, 154).  For example, suppose a lamp is connected to a timer switch which will turn the lamp on at 7:00 PM, it is 7:00 PM, and the timer switch turns the lamp on. Suppose I would have switched on the lamp only if the timer had failed to switch on the lamp. Then, the actual cause of the lamp's being switched on (namely, the operation of the timer switch) buries a counterfactual cause of the lamp's being switched on (namely, my switching the lamp on had the timer switch failed to do so).
On Swain's theory, the counterfactual cause (the pseudo-overdeterminant) must be buried only by the actual cause or causes (Swain 1981, 70).  But, Kvanvig notes, counterfactual causes can be buried by other counterfactual causes established by occurrent events. For instance, in the case of the gypsy lawyer, the actual cause of the lawyer's belief that his client is innocent is the reading of the cards. But suppose that the lawyer also believes in fortune tellers and has gone to one who told him that his client is innocent. We may suppose that, for whatever reason, the fortune teller's reading is not a genuine overdeterminant but a pseudo-overdeterminant of the lawyer's belief that his client is innocent, and that the complicated line of reasoning is also buried by this counterfactual cause (the fortune teller's reading). In this situation, the actual cause (the card reading) of the lawyer's belief that his client is innocent buries a counterfactual cause (the fortune teller's reading) of the lawyer's belief that his client is innocent, and that counterfactual cause (the fortune teller's reading) itself buries yet another counterfactual cause (the complicated line of reasoning) of the lawyer's belief that his client is innocent. Thus, if the lawyer had not gone to the card reading, the statements of the fortune teller, not the complicated line of reasoning, would have caused the lawyer to believe that his client is innocent. Intuitively, the lawyer's belief is still based on the complicated line of reasoning even though the requirements for pseudo-overdetermination have not been met.
As Kvanvig notes, the difficulty just described is avoidable if Swain's theory is modified to allow that buried pseudo-overdeterminants (established by occurrent events) of a given belief may count as the basis of that belief (Kvanvig 1985, 155-156). This would require that an additional condition be added to clause (2) of (DPO), as follows:

and, (c) if there is a set F (possibly consisting of only one member) of occurrent events (f1, ...., fn) which establish one or more counterfactual causes each of which buries occurrent event c, then if none of the members of F had occurred, but c and (occurrent event) e had occurred anyway, then there would have been a causal chain from c to e and c would have been causally prior to e.

Clause (c) makes it clear that Swain can allow pseudo-overdeterminants buried by any number of other pseudo-overdeterminants to count as the basis of a belief. All such buried pseudo-overdeterminants which meet the conditions of (DB) get counted by (c) as a reason for the belief. Such a modification of Swain's theory is necessary, and throughout the remaining discussion I assume that such a modification has been made.
Kvanvig argues that even with such a modification, Swain's theory faces a further difficulty in that counterfactual causes established by occurrent events may be buried by other counterfactual causes not established by occurrent events. For example, suppose, in the original case of the gypsy lawyer, that if the lawyer had not read the cards, he would still have believed that his client is innocent, not because of the complicated line of reasoning but because instead of reading the cards he would have read and believed a story in the National Enquirer claiming that his client is innocent. In this case, the reading of the Enquirer (a counterfactual cause not established by occurrent events) buries the complicated line of reasoning as a cause of the lawyer's belief. It seems obvious that the lawyer is still justified in believing that his client is innocent on the basis of the complicated line of reasoning, yet, Kvanvig claims, the complicated line of reasoning is not a pseudo-overdeterminant of the lawyer's belief.
Kvanvig's objection appears to misconstrue Swain's theory. What matters is not merely what would have caused the person's belief had the actual cause been absent (as Kvanvig assumes), but what would have caused his belief had the actual cause been absent given the person's current epistemic situation (ignoring trivial changes in the epistemic situation, such as the actual cause of the pseudo-overdetermined belief being absent). Thus, counterfactual causes not established by occurrent events are irrelevant on Swain's theory.
Although Kvanvig's objections to Swain's theory of the basing relation do not ultimately succeed, they are helpful in clarifying how Swain's theory works.

2.2 Tolliver's Objection To Swain's Theory

Joseph Tolliver's objection to Swain's theory stems from basing relations between beliefs which imply each other (Tolliver 1982, 151-155). He illustrates his objection with The Pendulum Case. Suppose a physics student has learned that from the period of a pendulum (i.e., the time it takes to complete a swing) one can calculate its length and vice versa.  The student observes that a particular pendulum has a length l, and calculates that it must have period p. The student also has two general beliefs about pendulums, namely (1) that if x is a pendulum of period P, then x is a pendulum of length L, and (2) that if x is a pendulum of length L, then x is a pendulum of period P. We may suppose that it is clear in this case that the student's belief about the period is based (at least in part) on her belief about its length, but her belief about its length is not based on her belief about the period. But, Tolliver claims, the student's belief about the period pseudo-overdetermines her belief about the length of the pendulum, hence gets counted, on Swain's theory, as the basis of his belief about the length of the pendulum. This is so because, to paraphrase Tolliver, if no other states were causes of the student's belief about the length, and if the student still had her belief about the period, then her belief about the period would be a cause of her belief about the length (Tolliver 1982, 155).  This objection appears to pose a difficult obstacle for Swain's theory.

2.3 Another Counter-Example To The Pseudo-Overdetermination Condition

An additional counter-example to the pseudo-overdetermination condition is the case of the doubtful gypsy lawyer. The case of the doubtful gypsy lawyer is much like the case of the gypsy lawyer except that the lawyer does not recognize that the complicated line of reasoning is a good reason to believe that her client is innocent. Suppose that the complicated line of reasoning indicates that the client is innocent because he was unable to acquire the eighth murder weapon. However, the particular manner in which the cards are read causes the lawyer to believe that the reason her client is innocent is due solely to the clients character. Since the complicated line of reasoning has nothing to with his clients character, the lawyer dismisses the apparent evidence that her client could not have acquired the eighth murder weapon, thinking it is unlikely, but nonetheless true, that her client did manage to acquire the eighth murder weapon. Thus, the doubtful lawyer is not justified in believing that her client is innocent on the basis of the complicated line of reasoning. Yet the complicated line of reasoning still pseudo-overdetermines the doubtful lawyer's belief that her client is innocent. Further suppose that the lawyer knows that she has no uncaused doubts. Then the features of the actual world (AW) relevant to the counter-example are these:

AW:
(1) the card reading occurs
(2) the lawyer doubts the complicated line of reasoning showing her client's  innocence
(3) the lawyer's doubt described in (2) is based on the card reading
(4) the lawyer has no uncaused doubts, and knows that she has no uncaused doubts

Intuitively, the doubtful gypsy lawyer does not base her belief that her client is innocent on the complicated line of reasoning in AW. The possible world Swain picks (call it W1) to show that his theory matches our beliefs about the case of the doubtful gypsy lawyer is as follows:

W1:
(1') the card reading does not occur
(2) the lawyer's doubt about the complicated line of reasoning remains
(3') the lawyer's doubt described in (2) is not based on the card reading
(4') the lawyer no longer has no uncaused doubts, hence no longer knows that she has no uncaused doubts
(5) everything else remains as in the actual world AW

W1 involves three notable changes in the lawyer's epistemic situation from the actual world AW, reflected in (1'), (3'), and (4'). However, there is a closer possible world in which the pseudo-overdetermination relation does hold, namely W2:

W2:
(1') the card reading does not occur
(2'') the lawyer's doubts about the complicated line of reasoning cease to  exist once their actual cause (the card reading) is removed
(3'') there is no basing relation from the card reading to the  lawyer's doubt (obviously, since there is no card reading and no doubt in  W2)
(4) the lawyer has no uncaused doubts, and knows that she has no uncaused doubts
(5) everything else remains as in the actual world AW

W2 also involves only three notable changes in the lawyer's epistemic situation from the actual world AW, reflected in (1'), (2'') and (3"). Yet W1 is further away from AW than is W2 because W1 contains both uncaused causes (the lawyer's doubts) and a significant change in the lawyer's epistemic situation, whereas W2 (along with AW) does not. (If one feels otherwise about uncaused causes, note that which feature(s) we select to distinguish W1 from W2 are quite arbitrary: other factors could be stipulated which secure the same point). In W2 the complicated line of reasoning causes the lawyer to believe that her client is innocent. Thus, the reasoning pseudo-overdetermines the lawyer's belief in the actual world (AW). The result is that Swain's theory gives us the incorrect result in the case of the doubtful gypsy lawyer, for the lawyer does not base her belief on the complicated line of reasoning in the actual world (AW).
Of course, a defender of Swain's theory is not limited to appealing to W1. If there is any possible world closer to the actual world than W2, and meeting the conditions of Swain's theory, in which the belief in the client's innocence is not based on the complicated line of reasoning, then Swain's theory handles the case of the doubtful gypsy lawyer. A likely alternative to W1 would be W3:

W3:
(1') the card reading does not occur
(2) the lawyer's doubt about the complicated line of reasoning remains
(3') the lawyer's doubt described in (2) is not based on the card reading
(4') the lawyer possesses some reason state r she does not possess in the actual world
(5') the lawyer's doubt is based on reason state r
(6) everything else remains as in the actual world AW

W3 avoids any mention of uncaused causes, and the lawyer's belief in her client's innocence is not based on the complicated line of reasoning because of her doubts. Nonetheless, it seems that W2 is closer to AW that is W3, given the number and quality of changes to the lawyer's epistemic situation in W3.
It seems that, in general, one could rig up a possible world closer to the actual world than the possible world that accommodates the pseudo-causal theorist, but that nonetheless fails to accommodate the pseudo-causal theorist. Perhaps the most plausible strategy for replying to the foregoing objection would be to give an alternative account of which features of one's epistemic situation must be kept constant in the relevant possible worlds. However, I suspect that all such proposals will be vulnerable to further modified versions of the case of the doubtful gypsy lawyer. For example, suppose we modify Swain's theory so that only those elements of the lawyer's epistemic situation that seem to the lawyer to be appropriately related to the truth of her belief that her client is innocent must remain unchanged. This would avoid the problem posed by the case of the doubtful gypsy lawyer, given that doubts about the complicated line of reasoning purporting to show that the lawyer's belief is true would remain but the lawyer's knowledge that she has no uncaused doubts would be irrelevant. But this version of Swain's theory would fail in the face of another counter-example, the modified case of the doubtful gypsy lawyer, identical to the case of the doubtful gypsy lawyer except that the complicated line of reasoning involves the claim that the lawyer knows she has no uncaused doubts. Then in W2, the lawyer does base her belief in her client's innocence on the complicated line of reasoning and in W1 she does not, and since W2 is closer to AW than W1, we have a counter-example to the proposed revision of Swain's theory.
Swain's theory is in some sense correct in that any reason for a belief will at least be a counter-factual cause of the belief. However, I do not see how the counter-factual can be elucidated in such a way as to provide a satisfactory analysis of the basing relation.

3. Doxastic Theories Of The Basing Relation

Doxastic theories of the basing relation hold (at least) that a meta-belief to the effect that a reason is a good reason to hold a belief is a necessary condition of the belief being based on the reason. Audi's theory is the most detailed doxastic theory I am aware of, and so I will concentrate on his account.

3.1 Audi's Doxastic Theory

In his paper "Belief, Reason And Inference", Audi presents a version of the doxastic theory of the basing relation (Audi 1986).  Audi equates reason states with (causally efficacious) beliefs, contrary to Swain's use of 'reasons' to include non-belief states (Audi, 1986, 234). Reason states are not to be equated with reasons in Audi's terminology: reasons for Audi are propositions. Furthermore, since propositions may be obviously false, reasons need not be good reasons (Audi 1986, 234). In discussing Audi's theory, I will follow his usage.
Audi suggests that while the appropriate meta-belief is necessary for a belief to be based on another belief, such a meta-belief is not necessary for a belief to be based on a perceptual state (Audi 1986, 250). Thus, Audi would propose two distinct theories of the basing relation: a doxastic theory for beliefs based on other beliefs and (presumably) a causal theory for beliefs based on non-belief states. However, Audi discusses only the doxastic theory in detail, and so I shall focus on that.
Audi begins by developing an account of "a reason for which S believes p." Audi does not equate a reason for which S believes p with S's belief that p being based on a reason for S. Audi stipulates that having a reason for which one believes p is sufficient, but not necessary, for S's belief that p to be based on the reason. Audi's strategy in elucidating the basing relation is to provide an account of what it is for a reason to be a reason for which S believes p, and then to extend that account in order to provide a theory of the basing relation.
The analysis that Audi presents of S's believing p for a reason is this:

S believes that p, for a reason, r, at time t, if and only if, at t, there is a support relation, C, such that (1) S believes that p and that r; (2) S believes C to hold between r and p (or believes something to the effect that r bears C to p); (3) S's basis belief, that r, and at least one connecting belief, i.e. his believing C to hold between r and p, or his believing something to the effect that r bears C to p, are part of what explains why S believes p, and, together, fully explain this; (4) S is non-inferentially disposed, independently of seeking reasons he has, had, or might have had, at or before t, for believing p, to attribute his belief that p to the explaining beliefs specified in (3); (5) those explaining beliefs do not (singly or together) accidentally sustain or produce S's belief that p; and (6) those beliefs do not (singly or together) sustain or produce it via an alien intermediary. (Audi 1986, 262)

Audi adds to this already formidable theory some additional conditions to account for believing something for several reasons as well as believing something for partially supporting (contributing) reasons. He allows that some elements of this account will remain vague, and certainly there is much to be discussed (Audi 1986, 262). Audi's theory amounts to an amalgam of almost all of the other theories of the basing relation presented in the literature.
Condition (3) of Audi's account stipulates (among other things) that there must be a causal connection from reason to belief. Complementing (3) are conditions (5) and (6), both intended to rule out deviant causal chains. Although Audi does provide several examples of deviant causal chains, he does not offer an analysis designed to overcome the problem of deviant causal chains, so I shall not pursue what he has to say about that issue here. As condition (1) of Audi's theory merely stipulates that S must possess both reason and belief, this leaves conditions (2), (3) and (4) as the primary focus of this discussion.
Although Audi's theory of the basing relation initially appears to be a hodge-podge, it is the distinction between a reason's guiding a belief and the belief's merely being an effect of the reason that motivates both conditions (2) and (4). For ease of reference, I'll call this the guide/effect distinction. Iíll discuss each of these conditions, observing how the guide/effect distinction comes into play, and then turn to a discussion of the distinction.
I'll begin with condition (4), what Audi calls the subjective grounding requirement, which requires that one be non-inferentially disposed, independently of seeking reasons, to attribute one's belief to both one's reasons and a meta-belief to the effect that a support relation holds from the reason to the proposition believed. One objection to condition (4) is that an individual may be mistaken about the basis of his or her belief, and so would not attribute the belief to the reason state even though, intuitively, the belief is based on the reason state.  For example, suppose that Wiley comes to know many propositions concerning the nature and appearance of road runners: he reads books about roadrunners, he looks at pictures of road runners, he watches movies of road runners, he goes to the zoo and watches the road runners, to the point that his head is swimming with images of road runners. Wiley then goes on a drive in the desert in search of road runners and sees a blur across the road. Wiley's propositional evidence tells him that any small blur across the road in this desert is probably a road runner, and on the basis of this he comes to know that he saw a road runner. But since his head is swimming with images of road runners, Wiley mistakenly believes that he actually perceived a clear image of the road runner (when all he actually saw was a blur). When asked to justify his belief that he saw a road runner, he appeals not to his propositional evidence, but rather to the clear image he mistakenly believes is of the road runner he saw. Yet Wiley's belief that he saw the roadrunner is in fact only based on (and justified by) his propositional evidence about roadrunners and the blur he saw, not any clear visual impression of the roadrunner. This is so even though he would never appeal to his propositional evidence, having unknowingly deceived himself regarding the reasons for his belief.  Wiley's situation is somewhat like that of a person who is unsure of their reasons, but nonetheless consistently responds correctly to questions regarding the issues they are unsure about. We are inclined to attribute knowledge to such persons in order to explain their consistently getting it right.
Audi might reply to objections such as the example of Wiley that if a person, given their current beliefs, cannot in some sense make the connection between reason and belief, then "we do not conceive the putative reason as capable of the appropriate guiding role, and the case becomes one of believing because of, but not for, a reason" (Audi 1986, 254). And Audi thinks that one cannot make the connection without having a disposition to attribute one's belief to both one's reasons and the appropriate meta-belief. The belief might be an effect of one's reasons, but it would not be guided by them in the appropriate sense. The guide/effect distinction, then, serves as the motivation for including condition (4).  I believe this reply fails because, as I shall argue, the guide/effect distinction does not require that one have a disposition to attribute one's belief to both one's reasons and the appropriate meta-belief.
Let's turn to condition (2) of Audi's theory - the doxastic element of the theory - which Audi calls the connecting belief requirement. According to the connecting belief requirement, where r is a reason for which S believes p, there is a support relation C such that either S believes C to hold between r and p or S believes something to the effect that r bears C to p. The support relation C may include such things as the reason implying, confirming, justifying, entailing, explaining, indicating, evidencing, or etc., proposition p (Audi 1986, 241).
The connecting belief requirement is disjunctive, and the difference between the disjuncts is important. Audi states that the first disjunct (S believes C to hold between reason r and proposition p) is intended to account for de dicto beliefs about C holding between r and p while the other disjunct (S believes something to the effect that r bears C to p) is intended to include de re beliefs to the effect that r bears C to p (Audi 1986, 241). De dicto beliefs are beliefs that something is the case whereas de re beliefs are about things out in the world. An example of such a de re meta-belief would be taking a reason "r to be such that if r, then p" (Audi 1986, 241). The provision for de re beliefs is meant to allow persons (such as children) with no epistemic concepts whatsoever to have some beliefs based on others, even though they are unable to conceptualize epistemic relations by means of de dicto beliefs.
Audi's motivation for making the connecting belief requirement part of the analysis of a reason for which one believes is again the guide/effect distinction. Audi states:

... believing for a reason is not just having a belief that is an effect of a reason (strictly, a reason state), though it is believing on account of a reason. It is, in a way, belief that takes some account of, in the sense that it is guided by, a reason. (Again, it is strictly the reason state that guides, on the assumption I am making here that abstract entities are not causal elements.) [...] The account I shall develop, then, will construe beliefs for reasons as responses to reasons, and not mere effects; as held in the light of reasons, and not merely explainable in terms of them; and as, in a special way, under the control of reasons (Audi 1986, 237).

Audi appends the causal theory of the basing relation so that reasons for which one believes are analyzed not simply in terms of causal states and causal chains of events, but rather in terms of those causal states and causal chains of events that occur in and are influenced by a certain cognitive context. The cognitive context consists of the relevant meta-belief and a disposition to appeal to one's reasons. Audi holds that it is only if we take this context into account that we can capture the subtleties of belief formation in light of our (propositional, epistemic) reasons.
Audi argues that the connecting belief requirement is necessary if the analysis of a reason for which one believes is to explain how reasons guide belief formation rather than merely assert that beliefs are effects of reasons (Audi 1986, 242). And the guide/effect distinction is motivated by an important property of believing for a reason, namely that it is discriminative. Thus Audi's argument for including both the connecting belief requirement (condition (2)) and the subjective grounding requirement (condition (4)) is ultimately that they are needed to account for this property of being discriminative.
According to Audi, there are three aspects of the property of being discriminative (Audi 1986, 242-243). I will discuss each of these three aspects individually arguing that none alone are sufficient to justify the connecting belief requirement, and then argue that a cumulative case argument involving all three also fails.
The first aspect of being discriminative is that "the belief that r does not, qua reason belief ... tend to give rise to or sustain just any belief, but only those S takes r to support...." Thus, the argument goes, if the analysis of a reason for which one believes does not include the connecting belief requirement, this property of being discriminative cannot be accounted for.
In reply, a reason state will typically give rise to only those beliefs S would take it to support insofar as one has accurate introspective access to his reasons, hence knows how he or she goes about reasoning. One difficulty is that one might lack accurate introspective access to one's reasons, and yet believe on the basis of those reasons, as in the case of Wiley or in cases in which the belief is formed subliminally. In such situations, one may lack the access to one's reasons needed to form a meta-belief about them. Another concern is that reason states such as perceptual states are able to give rise to beliefs without a corresponding meta-belief, as Audi recognizes. And perceptual states do not give rise to or sustain just any belief. But then the causal powers of states of persons are sufficient to give rise to beliefs in a non-arbitrary way in the absence of meta-beliefs. Why, then, couldn't beliefs give rise to other beliefs in a similar way?
The second aspect of the property of being discriminative, according to Audi, is that "... where r is the only reason for which S believes p, so far as S is disposed to explain or justify his belief that p (e.g., on being asked why he believes it), he (i) spontaneously tends to appeal to r, and (ii) does not spontaneously tend to appeal to other beliefs, in the explanatory or justificatory attempt" (Audi 1986, 242). For instance, S 's belief that the car's brakes squeak naturally generates or sustains the belief that the brakes are worn, but does not generate the belief that S is being a nuisance unless S believes that having squeaky brakes is sufficient for being a nuisance, in which case the appropriate meta-belief is present (Audi 1986, 242). However, it seems plausible to suppose that spontaneous appeals to reasons for beliefs may occur in the absence of connecting beliefs. For example, I might, while caught up in the excitement of a TV commercial for Pepsi, come to believe that Pepsi is good as a result of believing they have such clever ads. Yet I might be such that were I to form the relevant connecting belief (if the ads are clever, the soda is good), I would do so consciously, would immediately recognize it as ridiculous, and reject my belief about the ads as a reason for my belief about Pepsi. But in the absence of the connecting belief, I would still excitedly talk about the ads when asked why I think Pepsi is good. It seems clear that such associations do not require connecting beliefs. Furthermore, people might base beliefs on reasons and yet have no tendency to appeal to those reasons, as in the Wiley example discussed above. As Audi seems to recognize, this aspect of the property of being discriminative need not be present when a reason is a reason for which one believes.
The third aspect of being discriminative is that connecting beliefs explain how people are able to distinguish what they believe from what they don't (Audi 1986, 242-243). For instance, if you believe that the squeaking indicates that your brakes are worn, and you mention this to someone and they ask you "Why should their squeaking indicate that they need to be replaced?," you might reply that you don't think that they need to be replaced. Audi states that we explain, in part, your making this reply by observing that you do not have a meta-belief to the effect that a brake's squeaking implies that it needs to be replaced. It is the fact that you have a meta-belief to the effect that its squeaking implies that it is worn that allows you to distinguish what you believe from what has been attributed to you.
In reply, one difficulty concerns the question of how we are to distinguish connecting beliefs from things we don't believe. Does this require an additional connecting belief? If so, we face a regress. It seems clear that we can have beliefs distinct from our other beliefs such that each plays a full and unique role in our mental life, even if we lack meta-beliefs regarding those beliefs. But then we must be distinguishing among our beliefs, or they would not each play a full and unique roll in out mental life. And if some beliefs can be distinct from others in the absence of meta-beliefs about them, then one can distinguish among them in the absence of meta-beliefs about them. But then we should be able to distinguish, by means of direct introspective awareness, what we believe from mistaken attributions without meta-beliefs about what we believe.
I don't believe that any of the three aspects of being discriminative alone will justify the connecting belief requirement, and we have seen reasons for thinking the connecting belief requirement need not be met for a reason to be a reason for which one believes.
I now turn to condition (3) of Audi's analysis, which states that "S's basis belief, that r, and at least one connecting belief, i.e. his believing C to hold between r and p, or his believing something to the effect that r bears C to p, are part of what explains why S believes p, and, together, fully explain this" (Audi 1986, 262). Gypsy lawyer style counter-examples would appear to be effective against this causal requirement. Audi argues that the case of the gypsy lawyer is no counter-example to causal theories of the basing relation, but I have argued above that his objections are ineffective. I believe, therefore, that the case of the gypsy lawyer still shows that condition (3) of Audi's analysis is not a necessary condition of a belief's being based on a reason.

3.2 Causal Theories Of The Basing Relation And The Guide-Effect Distinction

Perhaps the central issue raised by Audi's discussion of the connecting belief requirement is not whether the connecting belief requirement is a necessary condition of having a reason for which one believes, but whether causal theories of the basing relation can account for the guide/effect distinction and the intuitions supporting the claim that reasons for which one believes have the property of being discriminative which motivate that distinction.
Causal theories of the basing relation can account for the property of being discriminative, at least on many occasions. To see this, one must recognize that never will a single belief be the sole reason for another belief. The actual causing of a belief will presumably involve a large set of beliefs corresponding to the presuppositions we make when forming beliefs. We can account for the fact that a reason state is likely to give rise to not just any belief but only certain beliefs (the first aspect of being discriminative) in terms of the causal properties it has by virtue of being a causally efficacious state appropriately related to a particular (propositional, epistemic) reason state. These causal properties will make it the case that the reason state is likely to give rise to only certain other belief states. What makes it the case, for example, that my belief that my mug is blue will not give rise to just any belief is that it is a belief that my mug is blue and as such, it will play only a certain sort of role in my cognitive system. This role is constrained by numerous other elements of my cognitive system, my other reasons and beliefs about the nature of mugs, being blue, etc., and also the processes by which I reason. Were it to cause me to believe just anything, then it would cease to be a belief that my mug is blue.
The second aspect of being discriminative is that where a reason is the only reason for which a belief is held, so far as S is disposed to explain or justify his belief, he spontaneously tends to appeal to his reason and not to other beliefs. Again, I think we can account for this in terms of reason states causing beliefs. Insofar as I have a disposition to explain or justify my belief, and I have access to my reasons for that belief, I will tend to associate my belief with my reasons for it. It seems natural to suppose that a reason state's causing a belief might often (but not always) result in one's associating the belief with the reason in such a way that one tends to spontaneously appeal to one's reasons in support of the belief.
The third aspect of being discriminative is that we are able to distinguish false attributions of reasons from correct ones. If causal relations among reason states and beliefs are sufficient to account for the association of one's beliefs with one's reasons, as I have just argued, then there is no reason to suppose that they cannot also account for the dissociation of one's reasons from false attributions. In addition, where the sort of meta-beliefs Audi has in mind occur, they may well be the result of the attribution rather than something that exists prior to it.
I have argued that the guide/effect distinction, understood in terms of the property of being discriminative, is a distinction that causal theories of the basing relation can account for. Belief formation, while not holistic, is at least be pluralistic, involving a confluence of numerous reason states, both those directly supporting the belief and those involved in the process of reasoning. So conceived, it is easier to imagine how numerous reasons causing a belief can guide the formation of the belief in the way Audi has in mind. Thus, perhaps guidance can be reduced to a complicated confluence of causes which may, but need not, involve meta-beliefs or dispositions to appeal to reasons.
If I am correct in this, then Audi's cumulative case argument for the connecting belief and subjective grounding requirements fails. The cumulative case argument would be that the three aspects of being discriminative taken together suggest that an analysis of the basing relation merely in terms of reasons causing beliefs cannot account for the manner in which beliefs get based on reasons. However, since causal theories of the basing relation can account for each of the three aspects of the property of being discriminative, and since there is no apparent reason to believe that causal theories of the basing relation cannot account for all three aspects when they coincide, the cumulative case argument fails.
Foley presents an objection to causal accounts of the basing relation which seems particularly appropriate as an objection to my argument that causal relations from reasons to belief may be sufficient to account for the guide/effect distinction (Foley 1987, 178). Foley suggests that "it is presumably too stringent" to hold that all the evidence needed to make a belief propositionally justified (e.g., needed to entail the belief) will be appropriately related to (causally efficacious) reasons which, in turn, causally contribute to the belief. It follows that beliefs, if they are to be epistemically justified, will often be based on reasons which do not cause them (and, by the way, which are not the object of appropriate meta-beliefs capable of establishing basing relations). This conflicts with my claim that numerous reasons are usually involved in the causing of a belief, and in this way are able to guide the formation of the belief.
Whether such an objection is effective ultimately depends on the empirical question of what causes what. But one reason for thinking that there may be multiple reasons contributing to the formation of a belief is the relations that often seem to occur among beliefs. It seems that having one belief may be nomologically or logically sufficient for having other beliefs. The point can be illustrated by making some controversial, but ultimately inessential, suppositions about mind/body relations. Suppose that being in brain state A is sufficient for being in brain states B through K, and a different belief supervenes on each of A through K. Thus, my belief (A) that my Liberace mug is in the sink may also be sufficient for holding belief (B) that my Liberace mug exists, (C) that at least one mug exists, (D) that my sink exists, (E) that more than one thing exists, etc. It is not that (A) causes additional states (B) through (E), but that (B) through (E) are contained in, and in some sense essential to (A)s being (A). Then, the fact that I am caused to hold some belief (Q) by belief (A) would be sufficient for (Q) also being caused by beliefs (B) through (E). If so, then what might initially be thought of as one reason's causing a belief may actually be sufficient for a multiplicity of reasons causing the belief. Then it may be that, at some point in the causal history of the belief and the immediate reasons for it, all of the reasons required for the belief to be justified contribute causally to its formation.

3.3 The Prospects For Doxastic Theories

Several other doxastic theories of the basing relation (or related concepts) have been discussed in the literature (Foley 1987, 180; Pollock 1986, 82; Fumerton 1985, 48-52; Winters 1980 and 1983; Tolliver 1982; Stroud 1979; Pappas, 1979b; Longino 1978, 22). Some of them (e.g., Tolliver and Longino) have suggested that having the appropriate meta-belief is sufficient for the belief to be based on the reason. However, such meta-beliefs could be mistaken, or the result of pure coincidence, and such meta-beliefs will not establish basing relations. For example, suppose Ezekiel belongs to a religious cult, and slavishly believes everything his cult leader, Exidor, tells him. Exidor tells Ezekiel that all his beliefs are based on his belief in God, and Ezekiel obediently comes to believe this. But it seems clear that Ezekiel could simply be mistaken about that.
Also, as we have seen, there are numerous kinds of counterexamples which show that meta-beliefs are not necessary for a belief to be based on reasons. The most obvious sorts of cases are those in which the reason is a perceptual state and the belief is about what has been perceived. Other cases include individuals, such as children, who may lack any notion of the epistemic concepts needed to form the appropriate meta-belief. It seems clear that such persons may nonetheless be able to base one belief on another.
There are also cases in which persons are mistaken about the basis of their beliefs, as in the example of Wiley given above, or in the case of subliminal reasons for belief. Such persons would not form the requisite meta-belief because they are mistaken about the reasons for their belief, but it seems that, nonetheless, their belief may be based on those reasons. Also, there may be cases in which forming the requisite meta-belief results in one's rejecting a reason as the reason for which one believes, when prior to the formation of the meta-belief, the reason does seem to be a reason for which one believes, as in the Pepsi example discussed above.
Finally, as I have argued above, there is no apparent need to appeal to meta-beliefs. It seems clear that a reason's bearing an appropriate causal relation to a belief may be sufficient for the belief to be based on the reason in the absence of any meta-belief to the effect that the reason is a good reason to hold the belief.

4.0 Conclusion

It seems clear that the basing relation often involves a reason's causing a belief. Yet causal theories of the basing relation face two major obstacles: the problem of deviant causal chains and the problem posed by gypsy lawyer style counterexamples. I am not aware of an adequate solution to the former problem. However, it is not a problem unique to theories of the basing relation. The problem of gypsy lawyer style counter-examples has been the focus of most of the literature on the basing relation. Such counterexamples seem to show that a meta-belief to the effect that a reason is a good reason to believe p is sometimes sufficient to make it the case that the belief that p is based on the reason, even where the reason does not cause the belief. Yet, as the objections to doxastic theories show, it is a mistake to suppose that such meta-beliefs are necessary for a belief to be based on a reason, or that any such meta-belief will establish a basing relation. This leads to the conclusion that either an appropriate causal relation or an appropriate meta-belief or some combination of the two may be sufficient to establish a basing relation. The remaining challenge for those who would propose a causal analysis of the basing relation, then, is to develop a theoretical understanding of the basing relation sufficient to show that such a disjunctive account of the basing relation is not utterly ad hoc. ,
 
References:

Alston, W. 1986. "Internalism and Externalism in Epistemology", in Alston 1989.
_____. 1988. "An Internalistic Externalism", in Alston 1989.
_____. 1989. Epistemic Justification: Essays In The Theory Of Knowledge (Ithaca, NY: Cornell University Press).
Audi, R. 1993. The Structure Of Justification (Cambridge, UK: Cambridge University Press).
_____. 1986. "Belief, Reason And Inference", in Audi 1993.
Bender, J. W. 1989. The Current State Of The Coherence Theory (Dordrecht, The Netherlands: Kluwer).
Bender, J. W. and Davis, W. "Fundamental Troubles With The Coherence Theory", in Bender, J. W. 1989.
Davidson, D. 1973, "Freedom To Act", in Davidson 1980.
_____ 1980. Essays On Actions And Events (Oxford, UK: Oxford University Press).
Foley, R. 1987. The Theory of Epistemic Rationality (Cambridge, MA: Harvard University Press).
Fumerton. R. 1985. Metaphysical And Epistemological Problems Of Perception (Lincoln, NE: University Of Nebraska Press).
_____ 1995. Metaepistemology and Skepticism (Lanham, MD: Rowman and Littlefield Publishers, Inc.)
Goldman, A. 1979. "What Is Justified Belief", in Pappas 1979a.
Harman, G. 1973. Thought. Princeton: Princeton University Press.
Jacobson, A. J. 1993. "A Problem For Causal Theories Of Reasons And Rationalizations", The Southern Journal Of Philosophy, Vol. XXXI, No. 3, 307-321.
Kim, K. 1993. "Internalism And Externalism In Epistemology", American Philosophical Quarterly, Vol. 30, No. 4, October 1993, 303-316.
Korcz, K. (1996). The Epistemic Basing Relation. Ph.D. Dissertation, The Ohio State University. UMI # 9639273.
Kvanvig, J. 1985. "Swain On The Basing Relation", Analysis, Vol. 45, 153-158.
_____. 1987. "On Lemke's Defence of a Causal Basing Requirement", Analysis, Vol. 47, No. 3, June 1987, 162-167.
_____. 1992. The Intellectual Virtues and the Life of the Mind. Savage, Maryland: Rowman & Littlefield.
Lehrer, K. 1965. "Knowledge, Truth And Evidence", in Roth and Gallis.
_____. 1971. "How Reasons Give Us Knowledge, Or The Case Of The Gypsy Lawyer". The Journal of Philosophy, Vol. 68, no. 10, May 20, 1971, 311-313.
_____. 1974. Knowledge. New York, NY: Oxford University Press.
_____. 1990. Theory Of Knowledge. Boulder, CO: Westview.
Lemke, L. 1986. "Kvanvig And Swain On The Basing Relation", Analysis, Vol. 46, No. 3, June 1986.
Longino, H. 1978. "Inferring", Philosophy Research Archives, Vol. IV, 19-26.
McGrew, T. 1995. The Foundations Of Knowledge (Lanham, MD: Littlefield Adams Books).
Moore, G. E. 1962. Commonplace Book 1919-1953, edited by Casimir Lewy, (London, UK: George Allen & Unwin, Ltd.).
Moser, P. 1989. Knowledge and Evidence. (New York, NY: Cambridge University Press).
Nozick, R. 1981. Philosophical Explanations (Cambridge, MA: Harvard University Press).
Pappas, G. 1979a. "Basing Relations", in Pappas 1979b.
_____. 1979b. Justification and Knowledge, (Dordrecht, Holland: D. Reidel).
Plantinga, A. 1993. Warrant: The Current Debate. Oxford: Oxford University Press.
Pollock, J. 1986. Contemporary Theories of Knowledge. Savage, MD: Rowman and Littlefield.
Roth, M. and Galis, L. 1970. Knowing: Essays In The Analysis Of Knowledge. (New York, NY: Random House).
Shope, R. K. 1983. The Analysis Of Knowing. Princeton: Princeton University Press.
Steele, J. D. 1872. Fourteen Weeks In Natural Philosophy. (New York, NY: A. S. Barnes and Co.).
Stroud, B. 1979. "Inference, Belief And Understanding", Mind, Vol. 88, 179-196.
Swain, M. 1979. "Justification and the Basis of Belief", in Pappas 1979.
_____ 1981. Reasons and Knowledge. Ithaca, NY: Cornell University Press.
_____ 1985. "Justification, Reasons and Reliability", Synthese, Vol. 64, No. 1, p. 69-92.
Tolliver, J. 1982. "Basing Beliefs on Reasons", Grazer Philosophische Studien, Vol. 15, p.149-161.
Winters, B. 1980. "Reasonable Believing", Dialectica, Vol. 34, No. 1, 3-15.
_____. 1983. "Inferring", Philosophical Studies, Vol. 44, 201-220.