THE BASIC NOTION OF JUSTIFICATION

 

I. Introduction

 

Epistemologists often offer theories of justification without paying much attention to the variety and diversity of locutions in which the notion of justification appears. For example, consider the following claims which contain some notion of justification: B is a justified belief, S's belief that p is justified, p is justified for S, S is justified in believing that p, S justifiably believes that p, S's believing p is justified, there is justification for S to believe that p, there is justification for S's believing p, and S has a justification for believing that p. In addition to these passive uses of the notion of justification, there are active uses as well: S justified his belief in p, believing e justifies believing p, etc. The syntactic variety involves semantic difference as well. For example, the proposition S has a justification for believing that p does not entail that S believes p, whereas the proposition S justifiably believes that p does entail that S believes p.

Our ultimate goal is to show that this diversity is only superficial by arguing that there is a basic kind of justification. On the way, however, we shall argue that there are three central uses of a notion of justification in the above list: propositional justification (as in p is justified for S), personal justification (as in S is justified in believing that p) and doxastic justification (as in S's believing p is justified). Our preliminary argument will be that the multiplicity above can be explained in terms of these three locutions, and the substance of our argument will be to show that one of these three is the basic kind of justification. Success in this task will thereby justify, at least in part, the practice of contemporary epistemologists. Our conclusions, however, shall not be of much comfort to contemporary epistemology, for the way in which the apparent diversity in the uses of the notion of justification is eliminated undermines much of recent epistemology. In recent years, a new approach to the theory of justification has become popular; an approach on which one first evaluates the intellectual powers and faculties common to human beings, and holds that a belief is justified just in case it is produced or sustained by an appropriate faculty or power or by following a method which is acquired through the use of appropriate faculties or powers. According to Sosa, this brand of epistemology is appropriately termed 'Aristotelian', for it claims that appropriately virtuous basic cognitive equipment is fundamental to justification.¹ Readers may recognize the view under the more familiar name of reliabilism.

This approach to epistemology differs from traditional epistemology which holds that what is fundamental to justification is some abstract relation between propositions² which we can call the evidence relation. Justification obtains for a person's belief, according to such theories, when these abstract relations are instanced in some specified type of psychological reality. For example, one such theory might claim that a belief is justified for a person S just in case S is aware of the evidence for that belief and bases the belief on that evidence.

Both kinds of theories assume that there is a basic notion of justification by which we can explain the variety of ways in which the notion of justification is used. If our conclusions are correct, no theory which explains justification in the way distinctive of Aristotelian epistemology can be adequate, and hence a very plausible version of reliabilism must be rejected.

In order to understand the structure of our argument, it is important to distinguish it from a recent defense of reliabilism. According to several recent theorists, the notion of justification is ambiguous, so that the notion reliabilists attempt to clarify is a different notion than that which some other epistemologists attempt to clarify.⁸ Our argument will be compatible with the ambiguity of the notion of justification, for it is important to distinguish between an irreducibil-ity claim about the kind of justification involved in certain syntactic structures and an ambiguity claim about justification. Most importantly, a theorist may claim to find an ambiguity may be found in any of the particular locutions cited to this point, while maintaining at the same time that the variety of locutions are interdefinable, with one such locution constituting the primitive locution from which all the others can be defined. In sum, a reducibility theorist is not in the least committed to hold that there is one and only one univocal sense of the term 'justified'. We shall say more about the ambiguity in this claim later, and we shall also have more to say about the relationship between ambiguity and irreducibility; for now, though, the important point to recognize is that a claim regarding the ambiguity of the notion of justification is not necessarily of any relevance to the issue of the reducibility claim.

 

II. Preliminaries: Setting the Task

 

Our task, then, is to see if the above multiplicity can be simplified by taking one locution as basic and defining the others in terms of it. Our purpose in this section is to organize the multiplicities introduced in the last section in such a way that the issues before us and the options available can be clearly seen. We can begin by noting that the active uses of the notion of justification are easily understood in terms of the passive uses. Believing e justifies believing p just in case p is justified by e; a person justifies her believing p just in case she uses some other claim e in defense of believing p, and p is justified by e for her. So, the active uses of the notion of justification do not prove difficult for unifying the variety of locutions in which the notion of justification appears. We are left then with the passive uses of the notion of justification. An initial glance at the locutions above suggests, as we noted in the previous section, three different uses of the term 'justified'. In the propositional case, justification seems predicated of a proposition; for example, when we say that a proposition p is justified for S, we do not commit ourselves to S's believing p and so can at most be committing ourselves to a justification for the proposition which would be the content of S's belief were he to hold it. In the doxastic case, justification attaches to the believing itself; for example, when we claim that S's believing that it is raining is justified. In the personal case, justification appears to reside in the person in question. One example of this appears in the claim that S is justified in believing B.

The multiplicity involved in the original list of locutions involving the notion of justification is no broader than the three above, for the other locutions can be explained in terms of one of these three. Two are easily explained in terms of propositional justification:

Jl: There is justification for S to believe p =_df p is justified for S.⁴

J2: S has a justification for believing p =_df p is justified for S, and S believes or is aware

of that which justifies p for S.⁵

Other cases are easily understood in terms of doxastic justification:

J3: B is a justified belief (of S's) =_df where p is the content of S's belief B, S's believing p is

justified.

J4: S's belief that p is justified =_df S's believing p is justified and S believes p.

J5: S justifiably believes p =_df S's believing p is justified and S believes p.

The only remaining proposition from our original list is the proposition there is justification for S's believing p. This claim is type/token ambiguous in its reference to S's believing p. If the reference to S's believing p is read as a reference to a token believing of p by S, then this claim implies that S believes p. However, if the reference to S's believing p is read as a reference to a type of belief, then no implication is present regarding whether S believes p or not. Given the ambiguity of the sentence, our final example can be explicated in the following two ways:

J6a: There is justification for S's believing p =_df p is justified for S, and S believes p.

 

J6b: There is justification for S's believing p =_df p is justified for S.

We have not defended that the list we have acccounted for is exhaustive, but we believe that list is complete in the following sense. Any examples from ordinary language which would suggest multiple and irreducible kinds of justification are represented on that list. Thus, we claim that any examples ignored by the discussion to this point can be easily construed in terms of the three kinds of justification which we have isolated as the most basic kinds to this point.

So, the issue before us is the relationship between the three kinds of justification noted above, and we shall argue that propositional justification is the basic kind of justification. We maintain that personal justification can be defined in terms of doxastic justification, and that doxastic justification can be defined in terms of propositional justification as follows:

J7: A token of S's believing of p is justified =_df (i) p is justified for S, and (ii) either (a)

that token of S's believing of p is based on that which justifies p for S, or (b) p is justified

for S, but not by anything other than itself and S believes p.

This definition claims that doxastic justification amounts to propositional justification plus proper basing, if the belief in question is not a basic belief. If the belief is a basic belief (i.e., not based on any evidence or anything else which might be taken to be epistemically supportive of belief), then doxastic justification is just propositional justification plus belief.

We can argue for one aspect of J7 at this point, and that aspect is that the left side of J7 is logically both necessary and sufficient for the right side. First, in order to deny that the right side is necessary for the left side, one would claim either that proper basing is not necessary for justifiably believing a proposition or one would have to claim that a believing could be justified even though that believing has a content which fails to be justified. Failure of proper basing, however, is just the classic example of a case in which propositional justification obtains and doxastic justification fails to obtain; hence there is ample reason to rule out the first possibility. Further, unless one is employing the notion of justification in two different senses, it is unimaginable what it would be like for a believing to be justified while its content failed to be justified. Hence, there is good reason for thinking that the right side of J7 is necessary for the left side. Further, if a particular believing occurs when the content of the believing is justified and the believing is based on that which justifies its content, there could be no good reason for claiming that the believing itself was not justified. For there is no plausible further condition a believing must meet in order to be justified. Hence, the right side of J7 is sufficient for the left side.

The only other objection there could be to the claim that the two sides of J7 are logically equivalent concerns the clause which might be taken to imply the possibility of justified beliefs not based on anything other than themselves. This objection, however, is not an objection to J7, but rather an objection to certain types of foundationalism. So, we maintain, J7 is an exceptionless equivalence claim about the relationship between doxastic and propositional justification.

However, to defend our thesis that propositional justification is the basic kind of justification, much more is required than mere logical equivalence. First, we must address the possibility that personal justification is irreducible to doxastic or propositional justification. We take up that question in the following section, where we argue for the logical equivalence of personal and doxastic justification. A second, more subtle issue concerns the direction of reduction J7 implies between propositional and doxastic justification. Even if the right side of J7 provides necessary and sufficient conditions for doxastic justification as we claim, it does not follow that the direction of reduction should be from doxastic justification to propositional justification, as J7 claims. It might well be that the two notions are interdefinable, so that if we take doxastic justification as primitive, propositional justification can be defined in terms of it as well. We shall argue that propositional justification cannot be defined in terms of doxastic justification in section four. Our remaining strategy, then, is as follows: we will show that (i) personal justification is no more recalcitrant to reduction than is doxastic justification and (ii) propositional justification cannot be defined in terms of doxastic justification. If we can demonstrate these two claims, given that the logical equivalence claim implied by J7 is correct, we will have shown that propositional justification is the basic kind of justification, i.e., that the direction of reduction implied by J7 is correct.

 

III. On Reducing Personal to Doxastic Justification

 

Our concern in this section is to consider whether personal justification is distinct from doxastic justification. Recently, Kent Bach has offered an original defense of reliabilism against certain counterexamples to it, a defense which is best taken as affirming a logical distinction between personal and doxastic justification. Should his arguments prove sound, we would find our hopes for demonstrating that there is a basic kind of justification dashed, so a useful starting point for our discussion is with his arguments. He says,

I propose that we distinguish between a person being justified in holding a belief and the belief itself being justified. What makes a person justified in holding a belief resides in the quality of his epistemic action. There is much that this can involve, including asking fruitful questions, considering plausible alternatives, and properly evaluating evidence. Without trying to spell out precisely what good epistemic action involves, let's just say that a person is justified in believing something to the extent that he holds the belief rationally and responsibly. However, a belief can be justified even in the absence of any action on the part of the believer, as in the case of beliefs formed automatically or routinely, without any deliberate consideration.⁶

Bach claims that there is a distinction between personal justification and the sort of justification which attaches to a belief. Personal justification is a kind of justification which attaches to, or is predicated of, persons and not of propositions or beliefs; doxastic justification is not predicated of beliefs and not persons. Of course, it does not follow from the fact that Bach draws this distinction that he also holds, or should hold, that the kinds in question are irreducibly distinct. However, if Bach thought the two kinds of justification were not distinct, it would be crucial for the force of his argument that he noted the way the two kinds were related in order to defend his claim that the distinction between the two is as he sees it. In particular, he would need to argue that the way the two kinds were related definitionally implies that only doxastic justification can obtain in the purported counterexamples to reliabilism. Since Bach engages in no such discussion, it is reasonable to conclude that he intends to be drawing a distinction between two distinct kinds of justification. And, if Bach is right, we have good reason for thinking that personal justification is not equivalent to doxastic justification and perhaps not to propositional justification either.

            Bach claims that S is justified in holding a belief (i.e., S is the locus of personal justification in the situation of holding the belief in question) roughly if and only if that person's epistemic action, which resulted in the belief, is of sufficiently high quality. Since basic beliefs are not products of action at all, it follows (and Bach agrees that it follows) that no person can be justified in holding a basic belief. This claim is belied by ordinary parlance. For instance, Descartes can be justified in believing that he thinks even though he performs no epistemic action in arriving at that belief. Further, it is worth noting that the appeal to ordinary parlance is appropriate for the task before us, for the question of the reducibility of personal justification concerns the relationship between the syntactic formulations of ordinary English which employ the word 'justified' and variants on it. Thus, even if there is a sense of 'justification' on which basic beliefs all fail to be justified, this fact would not show the irreducibility of personal justification to other kinds of justification. One way to put this point is this: the reducibility claim is a different claim than an ambiguity claim about justification, and at best, Bach's arguments give us a reason for claiming ambiguity. The reason this ambiguity claim seems important in our context is that Bach couples the ambiguity claim with a claim at least apparently relating that ambiguity to the locutions we have identified as distinctive of personal and doxastic justification. This latter claim is false: if there is an ambiguity in the notion of justification, it is not reflected in ordinary locutions distinctive of personal and doxastic justification. Hence, even if Bach should have found a defense of reliabilism against certain counterexamples to it, his defense fails to provide any definitive argument for the irreducibility of personal justification to propositional and doxastic justification.

We turn then to a direct examination and evaluation of the relation between personal and doxastic justification. Some formal machinery will aid us in this task. We will assume a first-order language with variables ‘s’ ranging over persons, ‘p’ and ‘q’ ranging over propositions, and ‘x’ ranging over all objects. This language also contains complex expressions of the form ⌈[λx₁...x_n ψ]⌉ ,⁷ where ‘λ’ here is an abstraction operator and ψ any formula.⁸ λ-expressions will be considered to be both terms and predicates so they can occur both in subject and predicate position in atomic formulas. Intuitively ⌈[λx₁...x_n ψ]⌉ denotes the n-place relation that holds between objects a₁..., a_n just in case ψ(x_i/a_i).⁹ Where n = 0, ⌈[λ ψ]⌉ denotes the proposition that ψ. Thus, where L is the loving relation, '[λ Ljm]' can be read as "(the proposition) that John loves Mary," and '[λx Lxm]' as "(the property of) being an x such that x loves Mary."¹⁰

Doxastic justification has an agent's belief as the subject of predication, to which a property is attributed. What is not clear, however, is whether the belief in question is a belief-type or a belief-token. To get at the distinction, suppose Sid stops every day at the same stop sign, and does so justifiably. When we assert on one such occasion that he stops justifiably, are we attributing justification to that particular token of his stopping, or to that type which is common to all such stoppings?

Decisions regarding cases of this sort may not seem that pressing, for it may seem that not much hinges on one answer as opposed to another in this case. However, there are other cases in which the type/token distinction is critical because there is no token of the type in question of which to predicate anything. For example, suppose Joe is driving from Dallas to Houston at a rate of 45 miles per hour. Given that the speed limit is 55, we might say that Joe's driving 55, or his driving 10 miles per hour faster than he is driving, is justified, even though he is not in fact driving that fast. Analogously, when we say that a person's believing a proposition is justified, we may mean to imply that the person has the belief in question, but we may not mean to imply this either. Hence, doxastic justification locutions are ambiguous.

To capture the ambiguity, we need a notation which will allow us to distinguish between types and tokens. To this end, terms for token states of affairs will have the same form as our A-expressions above, only the scope of the 'A' will be indicated by braces rather than square brackets. We then get two readings the nature of doxastic justification:

(1)J₂[λ Bsp] and

(la)J₃{λ Bsp}.

(la) implies that an actual belief of S's is justified; (1) does not imply that there is any actual belief, but rather claims that an abstract state of affairs of S's believing p is justified.¹¹ We have chosen to use different subscripts on the justification predicate in each of (1) and (la). The reason for this difference is not because it is obvious that the relations these predicates express are different, but rather because it would be hasty at this point to assume their identity. It may be that these relations are the same, but that should be left an open question at this point.

Personal justification occurs in claims of the form S is justified in believing p. This claim is ambiguous as well, and its ambiguity is the same sort of ambiguity that infects attributions of doxastic justification. Consider again the case of Joe's driving 45 when the speed limit is 55. We might say, in such a case, that Joe is justified in driving 10 miles per hour faster than he is; or we might just say that Joe is justified in driving 55. In either case, we have an attribution of a property to Joe which involves only an act-type and not an act-token. In the belief case, the matter is similar. When we say that S is justified in believing p, we may or may not mean to imply that S believes p.

It may be objected here that the purported implicature, or lack of one, is only conversational in nature, and not a semantic implication or lack of one. This objection may be correct, bat need not be pursued here. For if it is correct, then the task of finding a basic sense of justification which can be employed to define all others is much easier, for there will be fewer readings about which to worry. Since one of our complaints about standard treatments of justification is that they obscure complexities without proper investigation, it is best to allow the

ambiguity to remain and deal with the resulting complexities as they stand rather than covering them up from the start.

The issue, then, is how to represent the two readings of 'S is justified in believing p\ If we are to give the fairest hearing to those who believe that personal justification is irreducible to propositional or doxastic justification, we should perhaps heed their intuitions regarding personal justification. It will be recalled that, for Bach, personal justification involves attributing a property to a person. If that is so, then personal justification involves the attribution of a property, being justified in believing p, to a person S. This property, however, is complex and should be represented as such. One way to understand its complexity is as follows: it involves the attribution of some form of justification to S's (token or type) believing of p. Since we are here assuming the identity of abstract states of affairs and propositions,¹² the abstract state of affairs S's believing p is just the proposition that S believes p. Further, the representation of the same claim read as an attribution of justification to a token state of affairs should have the same logical form as that of an attribution to the correlative abstract state of affairs, with the exception that the abstract state is replaced by a corresponding token state. So, in the abstract case, the complex property in question can be understood as involving the attribution of some form of justification to the proposition that S believes p. We then get the following for the abstract case:

(2) [λx J₄[λ Bxp]]s.

And for the token case we have by the above reasoning:

(2a) [λx J₅{λ Bxp}]s.

(2) reads as follows: S has the property of being an x such that x's believing p (a type, rather than token, state of affairs) is justified. (2a) reads as follows: S has the property of being an x such that a token believing of x's that p is true is justified. In both cases, these representations capture the intuitive points above, for they attribute a complex property--the property of being an x such that x believes that p is true is justified--of S.

Given the representations in (2) and (2a), we are in a position to see that personal justification is easily explainable in terms of the two readings of doxastic justification. For by λ-conversion,¹³ (2) and (2a) are equivalent to:

(2') J₄[λ Bsp] and

(2a’) J₅{λ Bsp}.

Consider, then, the following claims of logical equivalence between personal and doxastic justification:

J8: [λx J₄[λ Bxp]]s ⇔ J₂[λ Bsp] and

39: [λx J₅{λ Bxp}]s. ⇔ J₃{λ Bsp}

Note that the only difference between the right sides of J8 and J9, and (2') and (2a'), respectively, is the different subscripts on the justification predicates in each. Note further, however, that the reason for employing different subscripts on the justification predicates was in order to avoid prejudging the issue whether it is the same justification property in the diverse kinds of propositional, personal, and doxastic justification. What we have shown is that the properties distinctive of personal justification attach to the very same object in the very same way as those properties distinctive of doxastic justification. Hence, there remains no reason to deny that the predicates which ascribe justification to a person are the same predicates which ascribe justification to a belief. Hence, J8 and J9 follow from (2) and (2a).

Given the adequacy of J8 and J9, personal justification obtains if and only if doxastic justification obtains as well. The question which would remain, if this conclusion is correct, is the relation between doxastic and propositional justification. However, before moving to an investigation of the relationship between doxastic and propositional justification, we must face an objection to the course of the argument for the logical equivalence of personal and doxastic justification. It might be objected here that the reducibility claim we defend is too easily achieved by misrepresenting the nature of personal justification. For if we pay closer attention to the syntactic structure involved, there are two importantly different ways to parse those sentences other than the one underlying the representations on which we have focused. First we have

SI: S is (justified in believing) p, and on the other hand we have

S2: S is (justified in) believing p.

In either case, it might seem that personal justification should be understood as a relation. If read as SI, the objection would be that we should analyze personal justification as involving the relation being justified in believing which holds between a person and a proposition. If read as S2, the objection would be that it should be analyzed as involving the relation being justified in which holds between a person and a believing of p. As before, both readings would still be subject to the type/token ambiguity; the heart of the objection, however, is that by ignoring the relational character of personal justification, our conclusion about the logical equivalence of personal and doxastic justification is too easily achieved.

A quick response focuses on a point we have already made: those who believe that personal justification is irreducible are more likely to think of personal justification in terms of a property residing in a person. If personal justification is treated as a relation between a person and some other thing, perhaps a belief state, this point is lost.

But there is a more substantive response. Nothing we have said implies that there are no relations which obtain in virtue of the obtaining of personal justification as represented above. The relation being justified in believing obtains between S and p; the relation being justified in obtains between S and the belief-state of believing p. In each case, these relations will be at the very least logically equivalent to, and perhaps definable in terms of, those representations we have used in (2) and (2a) to represent the nature of personal justification. For, if they are not, there must be some more direct objection to these representations, an objection showing that our representations of personal justification somehow get it wrong. So the objection we are considering has no force against our conclusion unless it is bolstered by an as yet unformulated objection. Until we see how that objection goes, there is nothing in the present objection to which we can or must respond.

Hence this objection does not undermine the basic thrust of our discussion: namely, that personal justification locutions are logically equivalent to attributions of doxastic justification. What remains, then, is the examination of the relationship between doxastic and propositional justification. To this question we now turn.

 

IV. On Reducing Propositional to Doxastic Justification

 

In order to address the relationship between doxastic and propositional justification, we must first understand propositional justification. This kind of justification involves a justification relation between a proposition and a person. We can represent the claim that p is justified for S as follows:

(5) J₁ps.

In introducing the distinction between propositional and other kinds of justification, we claimed that the distinctive feature of propositional justification involves the predication of justification of a proposition. (1) does not do that, for according to (5), justification is a relation between a person and a proposition. However, by λ-conversion, (5) is equivalent to

(5a) [λx J₁xs]p.

where a relational property of justification is predicated of p. Hence, representing propositional justification as we have in (5) is equivalent to the claim that propositional justification involves a property predicated of a proposition.

Our question about the relationship between doxastic and propositional justification is really two questions, since our representation of doxastic justification comes in two forms. One form occurs when justification is predicated of a type of belief, and the other form occurs when justification is predicated of a token belief. The meaty issues, however, are those involved in token doxastic justification, as we will now demonstrate.

When we say that S's believing is justified, but do not mean to imply that S has the belief in question, what we are saying is that the content of the belief in question is justified for S. For, when we say that S's believing of p is justified, and mean not to imply that S believes p, what we are saying is that a certain belief type is justified for S, and not a particular token of that type (since sometimes there will be no token). The important question, then, is the identity conditions for types of beliefs, and the answer is obvious and straightforward: the identity condition for types of beliefs is identity of content. For example, to say that two token beliefs are of the same type is just to say that they have the same content. Thus, to say that a certain type of belief is justified is just to say that the propositional content distinctive of that type is justified. So, even

though attributions of doxastic justification are ambiguous, the weaker reading is easily explicable in terms of propositional justification. Therefore, we can accept:

J10: J₃ [λ Bsp] =_df. J₁ps.

Just as clearly, however, an attribution of doxastic justification as represented by (1) cannot be defined in terms of (1) alone. For, when we say that an actual belief is justified, that says more than that the propositional content in question is justified for the person in question. So, as claimed, the important issue for this section is the relationship between propositional justification and that kind of doxastic justification which characterizes some belief tokens.

Let us take a moment to remind ourselves of the task before us. In section two of this paper, we presented a definition of doxastic justification in terms of propositional justification. In order to accept that definition, two problems had to be handled. The first was to show that personal justification could be explicated in terms of doxastic justification, and we showed that such a reduction is possible in the last section. The remaining problem concerns direction of reduction, reduction between propositional and doxastic justification—where, since J10 shows how to understand doxastic justification regarding types of beliefs, 'doxastic justification' here and henceforth refers only to doxastic justification which attaches to belief tokens. Even if there are necessary and sufficient relations between propositional and doxastic justification as claimed in section two, it does not follow that doxastic justification should be reduced to propositional justification rather than the other way around.

We will attempt to show that the reduction must go in the direction posited by J7, for, we shall argue, it is not possible to define the notion of propositional justification in terms of doxastic justification. If our arguments are sound here, it will follow as well that propositional justification cannot be defined in terms of personal justification either, for doxastic and personal justification are logically equivalent.¹⁴ Thus, if our arguments are sound, we will have shown that propositional justification is the basic kind of justification.

Not surprisingly, Goldman believes that propositional justification is to be defined in terms of doxastic justification. He says,

The account I have offered is a theory of when an actual belief is justified. It is a theory of what I call . . . ex post justification. But [there is also a sense of 'justification' which] deals with being justified in believing a proposition which one does not actually believe. The idea is that it is a proposition one would (or could) be justified in believing, given one's present cognitive state, although one does not in fact believe it. This is what I call ex ante justification.¹⁵

Goldman holds that doxastic justification is primary and that propositional justification is to be defined in terms of it. According to Goldman, the definition is in the form of a counterfactual. If we let '□→' be the symbol for counterfactual implication, the most straightforward counterfactual approach to reducing propositional justification to doxastic is:

J11: J₁ps =_df. Bsp □→ J₂{λ Bsp}.¹⁶

J11 claims that propositional justification can be defined in terms of the claim that if S were to believe p, S's believing of p would be justified. This definition is, however, inadequate. Suppose superstitious Sam has sufficient evidence for the claim that he will have a bad day today, but believes it instead because a black cat crossed his path. We can then say both that the proposition in question is justified for Sam, but also that his believing is not justified. If this is so, however, the counterfactual in J11 will be false, for its antecedent is true in virtue of the fact that Sam believes he will have a bad day today, but his believing is not justified. Hence, the definition above does not give us a sufficient condition for propositional justification.

             It does not provide a necessary condition either. Suppose Joe does not believe that it is raining and would not believe that it is raining unless he had good evidence for that claim. Further, if he had good evidence for the claim, he would accept that claim because of his evidence for it. At present, though, he lacks any evidence that it is raining and hence that claim has no justification for Joe at present. Yet, if Joe were to believe that claim at present, he would have doxastic justification for it, for he would not believe it unless he had good evidence for it and if he had good evidence for it, he would accept the claim because of the good evidence for it. These counterexamples to J11 show that two different restrictions are needed on the right side of J11 if the doxasticist is to succeed in offering an account of propositional justification. The first point is that the epistemic support for the content of the belief which is held in the counterfactual situation must be the same as in the actual situation; in brief, when moving to the counterfactual situation, sameness of epistemic situation for p must be preserved. Second, it must be specified that in adding the belief p, S come to believe p on the basis of that which epistemically supports it.

The problem for doxasticism is that both of these restrictions are put in a form acceptable only to the propositionalist, for both restrictions appeal as stated to the notion of propositional justification. We shall first say why there is an implicit appeal to that notion, and then argue that the prospects are not good for finding a doxasticist way of satisfying these requirements without employing the notion of propositional justification.

The proposal we are considering attempts to explicate propositional justification in terms of the following counterfactual: were S's total epistemic situation to remain the same and were S's belief p based on that which epistemically supports it, S's believing of p would be justified. The appeal to the notion of propositional justification in this definition is, however, thinly veiled at best. First, the notion of a total epistemic situation regarding p must preserve identity of epistemic support relations regarding p, where the notion of an epistemic support relation regarding p is quite obviously to be explicated in terms of propositional justification for p.¹⁷ Second, the appeal to basing a belief on that which epistemically supports it is equally an appeal to the notion of propositional justification. For consider the case of superstitious Sam again. Sam has a belief which is propositionally, though not doxastically, justified. The explanation of why the belief is propositionally though not doxastically justified involves two parts. The first part denies that the possession of the evidence justifies the belief state; instead, it only justifies the content of the belief. Once the possibility is opened up in this way for a belief state to obtain and yet fail to be justified even though the person who has the belief has sufficient evidence for it, the basing relation is brought into the picture in order to bridge the gap between merely having evidence for a belief and that belief being doxastically justified. The first part of this explanation is the crucial part. For it simply cannot be granted that the evidence possessed by Sam justifies his believing, for otherwise there would be no room to maintain that he fails to believe justifiably. Hence it must be maintained that the epistemic support relation between the evidence and his belief is properly thought of only as a relation between the evidence and the propositional content of his belief. Without this view of the epistemic support relation, there could be no room for thinking that the basing relation makes up the gap between justified content and justified belief state. Hence any reference to an epistemic support relation will vitiate doxasticists' attempt to explain propositional justification.

The task facing the doxasticist is, then, the task of finding a way to restrict the antecedent of the counterfactual on the right side of Jll to meet the requirements above without appealing to the notion of propositional justification. We can begin to see how a doxasticist might try to do this by returning to the Goldman quote earlier. For in formulating and objecting to J11, we have been unfair to Goldman, in that in the quote above Goldman does not signal any interest in defending J11. Instead, his claim was that propositional justification should be understood in terms of what one would justifiably believe, given one's present cognitive state. So Goldman shows no interest in defending Jll, for no such qualifier is to be found on the right side of J11. Let us, then, consider whether Goldman's approach to restricting the counterfactual in J11 is adequate.

The most obvious point to note is that Goldman's restriction is at best a way of dealing with the first problem, the problem put above in terms of requiring sameness of epistemic situation regarding p. As to the problem of proper basing, Goldman has nothing to say. However, his restriction concerning sameness of epistemic situation regarding p is instructive, for it is an attempt to include that restriction without appeal to any notion of propositional justification. The problem for his proposal is not that it appeals to the very notion he is attempting to define, but rather that his proposal is just woefully inadequate. In particular, it is both too strong and too weak. It is too strong in that the addition of any belief engenders difference of cognitive state, so there is no possibility of a counterfactual situation in which sameness of cognitive state is found and a belief is added. And, as already noted, it is too weak in that it fails to address the problem of improper basing. This failure is due to Goldman's starting point, for in the passage above, he explicitly claims that the kind of justification he is going to explain is that kind which "deals with

being justified in believing a proposition which one does not actually believe." As we have seen, however, the problems facing the doxasticist include cases of both kinds: cases where the proposition is not believed, but also cases where it is believed but for the wrong reasons.

So Goldman's attempt is far from successful in relieving doxasticism of its problems. Goldman, however, does give us insight into how the doxasticist might attempt to solve the first problem, the problem of securing sameness of epistemic situation regarding p in the counterfac-tual situation in question. Goldman's suggestion is to attempt to avoid employing the notion of sameness of epistemic situation by employing the notion of sameness of cognitive state instead. As we have seen, this approach will not work as it stands, but perhaps we can substitute the notion of similarity of cognitive state and solve the first problem. The trick is to specify the kind of similarity negatively. One might claim, for example, that two total cognitive states are sufficiently similar with respect to p just in case the second cognitive state involves no changes other than the addition of belief p and those changes entailed by the proper basing of that belief.

            This proposal puts the weight of the solution to the first problem on the possibility of a solution to the second problem, the problem of specifying what proper basing comes to without employing the notion of an epistemic support relation. Contemporary approaches to the basing relation fall into two categories: psychological accounts and causal accounts.¹⁸ On a psychological account, to base one's belief p on e is to believe p, to believe or be aware of e, and also believe or be aware of the force of evidence e, i.e., to believe or be aware that e is sufficient evidence for p, that e shows that p is true, that e confirms p, or some such formulation in this general ballpark. Clearly, no such psychological account of the basing relation can be employed by the doxasticist, for to be aware of the force of one's evidence is just to be aware of what that evidence proposi-tionally justifies. Hence, employing a psychological construal of the basing relation would result in a proposed account of propositional justification which appealed to that very notion. Thus the most convenient move here is for the doxasticist to employ a causal account of the basing relation. We can conjoin these responses to the two problems facing J11 to obtain:

J12: p is justified for S =_df there is some proposition q which S believes or of which S is aware which is such that were (i) S to believe p, (ii) S's belief p caused by S's belief or awareness q, and (iii) S's cognitive state to remain the same except for the addition of the belief p, entailments from the addition of the belief that p, and changes required in S's cognitive state by #'s being a cause of the belief p; S would justifiably believe p.

 J12 avoids the problems facing J11, and does so without obviously employing the notion of propositional justification. Thus, at the very least, it is an instructive attempt at reformulating Jl 1 in the face of the counterexamples to it.

J12, however, is not adequate. In propositionalist language, the problem is that it fails to claim that q is evidence for p. Because of this lack, it is subject to the following alteration of the case of Joe. We already know that Joe has no evidence for believing that it is raining. But suppose he does believe that it is sunny outside, and further suppose that this belief that it is sunny outside could cause the belief that it is raining, and that it would so cause it in the following way. If Joe were to have his present belief that it is sunny outside cause a belief that it is raining, it would be because sunny times and rainy times, though rarely conjoined in this world, are so consistently conjoined that at least one rational view to hold is that there is an explanatory connection between sun and rain (this explanation fits in with the fact that Joe is assumed to have an unusually high degree of cognitive admirability). In J12, then, let q = it is sunny outside, and let p = it is raining outside. The right side of J12 is satisfied in the case as described, and yet, by hypothesis, p is not justified for Joe.

There are two ways to avoid this problem. The first is to insert the claim that q must be evidence for p in J12, evidence adequate to justify p. Clearly, however, this move will not be attractive to doxasticists for it obviously appeals to the notion of propositional justification. The other way is to make the connection between q and p a more intimate one: instead of saying that believing p would be justified if caused by q, the doxasticist can say that believing p would necessarily be justified if caused by q. More exactly, the doxasticist might replace J12 with:

J13: p is justified for S =_df there is some proposition q which S believes or of which S is aware which is necessarily such that were (i) S to believe that p, (ii) S's belief that p caused by S's belief or awareness that q, and (iii) S's cognitive state to remain the same except for the addition of the belief that p, entailments from the addition of the belief that p, and changes required in S's cognitive state by q's being a cause of the belief that p; S would justifiably believe that p.

J13 avoids the above problem with J12, for there is nothing in the case as described which suggests that Joe is essentially such that he would only come to believe that it is raining if he had adequate evidence for the claim. Instead, all that was claimed above was that he would only come to believe that it is raining if he had adequate evidence for the claim.

We must ask, however, if the case would suffer in terms of possibility if we added the claim that Joe is so cognitively admirable and so inordinately meticulous about rain-beliefs that he is essentially such that he would only come to believe that it is raining if he had adequate evidence for the claim. Perhaps Joe would be no ordinary human being if he could have that property, but that is no objection to the case. Further, note that it is just this kind of property which theists ascribe, consistently it would seem, to God regarding all of his beliefs.¹⁹ Further, even if there is some problem with the universal inability to believe without evidence which theists ascribe to God,²⁰ it would still seem possible that some persons be essentially such that they would not believe some particular claim without adequate evidence. If so, however, J13 must be rejected. For if such cases are possible, we can let q on the right side of J13 be any actual belief which might have caused such an essentially-evidence-sensitive belief that p, and all the conditions on the right side of J13 will have been satisfied. Hence, J13 is still too weak without the additional restriction that q is adequate evidence for p, a restriction the doxasticist cannot employ.

This objection shows that J13 is too weak in that the right side of J13 lets in some cases which the left side does not. There is reason to think that it is too strong in two other, quite different ways. The first way affects only those doxasticist theories which, like many versions of reliabilism, include a claim to the effect that justification involves the propensity for a belief to be true (perhaps by being produced by a mechanism which is likely, or which has the propensity, to produce true belief). It is well-known that the propensity towards truth is a world-bound property, i.e., it is not a property which applies to the same beliefs or belief-producing mechanisms in all possible worlds.²¹ If so, however, no propensity theorist can be happy with J13, for it requires that the justification producing power of q obtain in all possible worlds. So, if J13 were adequate on other grounds and were the best the doxasticist could do, such propensity theorists would still be in deep trouble. For if J13 were adequate, it would follow quite straightforwardly that justification does not involve any propensity for a belief to be true.

The second way in which J13 is too strong is relevant to all forms of doxasticism and not just to propensity versions of it. The difficulty here is a version of the problem of deviant causal chains. Causal theories of the basing relation must specify when a causal relation between beliefs or awarenesses is appropriate for the production of justified belief, for not every case in which an awareness of evidence, even sufficient evidence, causes belief is a case in which the belief state itself is justified. The burden on J13 is even greater than the usual burden on causal theories, for J13 requires that it be a necessary truth that a justified believing of p result when a belief or awareness that q causes the belief that p in the context of S's being in a particular total cognitive state. This modal enhancement of the causal theory makes it suspect, to say the least; on the face of it, it is wholly implausible to think that any and all possible causal connections between believing q and believing p would result in justified believing of p, even given the kind of cognitive state specified. The general point is that causal routes from one belief to another can be either relatively normal or wildly exotic, and until some way of ruling out the wildly exotic routes is forthcoming, there should be a great deal of suspicion about the prospects for a successful doxasticist account of propositional justification.

If this problem weren't deep enough already, there is a way to show that it is even more severe than what one might think given only the content of the last paragraph. For the problems facing causal theories of the basing relation regarding deviant causal chains ordinarily count as evidence in favor of a psychological account of the basing relation. The general strategy in inferential cases of belief, for example, is to describe a causal route from awareness of evidence to belief which obtains even though the person in question has no idea what the force of the evidence is. It might seem that such cases are the downfall of the causal theory, for it would seem to be a paradigm example of a case where the content of belief may be justified and yet the belief state itself, or the believing itself, is not justified when there is a causal route from evidence to belief and yet the person has no idea how the evidence is connected to the belief. However, we need not draw that conclusion here for there is a response available to the causal theorist to solve this difficulty. In particular, at least some causal theorists can exploit psychological connections in describing which causal routes are the appropriate ones. Such theorists can claim that a causal route in inferential cases from evidence to belief is appropriate only when it proceeds through a psychological awareness of the force of the evidence.

Note however that the doxasticist cannot resurrect the causal theory in this way, for as we have seen, appealing to notions such as an awareness of the force of the evidence clearly reintro-duces the notion of propositional justification into the discussion. So the most attractive strategy for salvaging the causal theory from one kind of objection concerning deviant causal chains cannot be employed by the doxasticist.

We can put our second objection concerning how J13 is too strong, then, as follows. The problem is a problem of deviant causal chains, but it is not just any ordinary deviant causal chain problem. For causal theories of the basing relation have an inviting path to take in avoiding objections intended to support psychological construals of the basing relation, but it is a path on which the doxasticist cannot brook. Some causal theories can exploit the view of their opponents by insisting that the appropriate causal route is through the very psychological connection the opposition is intent on emphasizing. The plight of the doxasticist in this regard is severe. For doxasticists must delineate the appropriate kinds of causal routes from awareness of evidence to belief without employing any notions of evidence or of psychological awareness of the force of evidence. And, even worse, they must solve this problem not just for this world, but for all possible worlds. We suggest that the reasonable conclusion to draw is that the task is hopeless.

So there are several reasons for thinking that J13 is inadequate. First, it is not implausible to think that a person could be essentially such that they would not believe a proposition without adequate evidence for it. Second, propensity theorists, a major group within the camp of the doxasticists, will find J13 wholly contrary to their epistemological tastes. And finally, the formulation used in J13 in combination with the demands placed on a satisfactory doxasticist account of propositional justification makes solving the problem problem of deviant causal chains

fairly well hopeless. To put it simply, as we see it, the best the doxasticist can offer is J13, and if so, doxasticism fails.

In summary, then, one cannot satisfy the two restrictions necessary in order to avoid the problems with 111 and remain a doxasticist. Further, if one cannot satisfy the two restrictions needed to avoid the problems with Jl 1 and remain a doxasticist, one will not be able to offer an account of propositional justification in terms of personal justification either. For since personal and doxastic justification are logically equivalent, no advantage could be gained by substituting the notion of personal justification throughout the preceding discussion wherever the notion of doxastic justification appeared. In particular, the analogue of 111 which results when the concept of doxastic justification is replaced with that of personal justification would require similar restrictions to those necessary for Jll. And any attempt to satisfy those restrictions using only the notions of personal or doxastic justification will suffer from an inability to account for sameness of epistemic situation and proper basing. Hence, the only consistent viewpoint regarding the basic notion of justification is the propositionalist viewpoint: the basic kind of justification is propositional justification.

 

V. Conclusion

 

Our tasks are then complete. We offered, in J7, a defensible claim that there is a relation of necessity and sufficiency between doxastic and propositional justification. We claimed as well that propositional justification is the basic kind of justification. In order to defend this claim, several tasks were undertaken. First, a preliminary investigation showed that there are at most three irreducible kinds of justification. In order to show that there is only one kind, we first showed how personal justification is equivalent to doxastic justification. We then showed that, whereas doxastic justification can be defined in terms of propositional justification, propositional justification cannot be defined in terms of that kind of doxastic justification predicable of tokens of belief. We conclude that propositional justification is the basic kind of justification.

These results have implications both for the substance of epistemology and for the methodology employed in constructing an epistemological theory. First, no version of reliabilism can be defended by claiming that there are fundamentally distinct and irreducible notions of

justification. Further, the procedure adopted by those versions of reliabilism which count as instances of Aristotelian epistemology cannot be correct. Such theories begin by discussing our intellectual powers, faculties, virtues, or cognitive processes and then define a primitive notion of justification in terms of the doxastic products of such powers, faculties, virtues, or processes. This approach cannot succeed, for, as we have seen, no such theory can give an adequate explanation of the connection between doxastic and propositional justification. For these theories to succeed, one of three alternatives must be available. The first is that propositional justification is explicable in terms of doxastic justification; as we have seen, that claim is false. The second and third alternatives are that Aristotelians either might claim irreducibility between propositional and doxastic justification, or the might hold that, for some reason or other, there is no such thing as propositional justification. Both of these alternatives are unattractive because of the plausibility of 37. If it is adequate, then doxastic justification is explicable in terms of propositional justification. Further, if J7 is adequate, claiming that there is no such thing as propositional justification puts the notion of doxastic justification in jeopardy as well, for doxastic justification is definable in terms of propositional justification. In sum: present reliabilist theories which hold that justification is the result of the proper employment of our cognitive equipment, and which are thereby committed to constructing a theory of justification with doxastic justification as the basic notion of justification, are incompatible with the fact that propositional justification is the basic kind of justification.²²

ENDNOTES

 

           ¹See, , e.g., Ernest Sosa, "Knowledge and Intellectual Virtue," The Monist, Vol. 68, No. 2,

 

April 1985, pp. 226-245.

²Some traditional epistemologists claim that the relations hold between ideas, but the most charitable interpretation of this claim is that the relations hold between the content of the ideas, and these contents are either propositional in nature or proposition-like by being abstract entities of one sort or another which are shared in common among many concrete ideas.

³See, for example, Alvin Goldman, "Strong and Weak Justification," pp. 51-71, and Alvin Plantinga, "Positive Epistemic Status and Proper Function," pp. 1-51, both in James E. Tomberlin, ed., Philosophical Perspectives, Vol. II, (Atascadero, CA: Ridgeview Publishing Co., 1988). Richard Foley, in The Theory of Epistemic Rationality, (Harvard, 1986), defends an ambiguity view about the nature of epistemic rationality, and also suggests that the ambiguity of the notion allows for the adequacy of quite a broad range of views about rationality, including reliabilist conceptions.

⁴Until we introduce a more formal approach below, we take 'p' to be a schematic letter whose acceptable substituends are declarative sentences, and ‘p’ to abbreviate the expression 'the proposition that p'.

⁵The last conjunct of J2 is formulated to be compatible with the view that the mental states involved in sensation are not themselves belief states but can only be counted as awarenesses.

⁶Kent Bach, "A Rationale for Reliabilism," The Monist Vol. 68, No. 2, April 1985, p. 251.

⁷In the formal details of this language, not every formula can be legitimately substituted for ψ in '[λ x₁...x_n ψ]’. Nonetheless, all formulas relevant to our project can function in such expressions, so the additional complexities required by the language and logic we are assuming will not be introduced here. For more on the assumed logic and language, see Christopher Menzel, "A Complete, Type-free "Second-Order" Logic and Its Philosophical Foundations," Report No. CSLI-86-40, Center for the Study of Language and Information, Stanford University, 19S6.

⁸There are certain restrictions which must be placed on ψ in a rigorous presentation of the logic underlying our discussion. We will not state these restrictions overtly, but we note that our use of the logic here follows them.

⁹Where ψ(x_i/a_i) is the result of evaluating ψ with a_i as the value of x_i.

¹⁰Here as elsewhere throughout the paper we are assuming that formulas are implicitly time-indexed, so to say that John loves Mary is to say that John loves Mary at a particular time. We do not insert the index in our discussion since it does not affect any of our discussion and thus would only complicate the presentation unnecessarily.

^J1For our purposes, we can identify states of affairs and propositions. Should the reader find this identification objectionable, we could introduce a further propositional operator which takes one from a proposition to the correlative abstract state of affairs. We see no reason for such an operator, and introducing one in the present context will only make for unnecessary complexities.

¹²See footnote 7.

¹³The general form of A-conversion is

[(λ x₁...x_n) ψ]y₁...y_n ↔ ψ(x₁...x_n/y₁...y_n)

 

where ψ(x₁...x_n/y₁...y_n) is the result of simultaneously replacing each x_i with y_i in ψ. This general form

does not hold in the logic we are assuming, but the inference from (5) to (5a) licensed by A-conversion does hold in that logic nonetheless. The same is true of all other such inferences in this paper.

¹⁴Some may be puzzled by this form of argument, for it may seem that if one can define one notion in terms of another, one should also be able to define the first notion in terms of the second. This viewpoint is mistaken, for when one notion can be defined in terms of another with the aid of auxiliary concepts, it is often the case that the second concept cannot be defined in terms of the first. A simple example out of number theory is this. The relation less than or equal to can easily be defined in terms of addition: x < y just in case ∃z(x + z = y). There is, however, no way to define addition in terms of the relation less than or equal to.

¹⁵Alvin Goldman, Epistemology and Cognition, (Cambridge: Harvard University Press, 1986), p. 112.

¹⁶To make this definition work properly here, we need a principle which correlates true propositions about beliefs with appropriate token states of affairs. The following seems to capture what we need: □(Bsp ↔ ∃x(x={λBsp})), i.e., a person S believes p (at t) necessarily iff the token state which is s's believing p (at /) exists.

¹⁷It may be that the notion of an epistemic support relation is to be clarified in terms of the notion of evidence, and it may also be that a proper account of propositional justification is to be

clarified in terms of the notion of evidence. In claiming that the restrictions needed on 111 appeal to the notion of propositional justification, we do not claim that the notions of epistemic support and evidence are to be defined in terms of propositional justification. Instead, all we mean to claim is that all these notions are in the same region of conceptual space, a region distinct from that of doxastic justification. We make no assumptions about the structuring of this region; in particular, it would be amenable to our view if the notion of propositional justification were to be in need of a clarification in terms of the notion of evidence.

¹⁸For a defense of a causal basing requirement, see Robert Audi, "The Causal Structure of Indirect Justification," Journal of Philosophy, 1983, pp. 398-415. For a defense of a psychological awareness requirement, and objections to a causal requirement, see Keith Lehrer, Knowledge, (Oxford: Clarendon Press, 1974), esp. pp. 124-125.

¹⁹For an argument that God could have this essential property, see Jonathan L. Kvanvig, The Possibility of an All-Knowing God, (London: Macmillan Press Ltd., 1986).

²⁰Patrick Grim has argued in several recent papers that the concept of omniscience is subject to paradox. See "Some Neglected Problems of Omniscience," American Philosophical Quarterly, Vol. 20., No. 3, pp. 265-277; and "Logic and Limits of Knowledge and Truth," Nous, Vol. XXII, No. 3, (September 1988), pp. 341-368.

²¹Just consider Cartesian evil demon worlds to see that the mechanisms we take to involve a propensity toward truth need not have that characteristic.

²²We extend our thanks to Robert Burch, Susan Hale, Peter Markie, and Scott Sturgeon for helpful comments and suggestions.

 

Jonathan L. Kvanvig & Christopher P. Menzel Texas A&M University