Section 3: A Dilemma About Side Defeat We’ve just identified a distinctive kind of justification defeat — the defeat of one’s justification that does not involve opposing defeat, but also does not involve any defeat of one’s justification for the propositions that constitute one’s reasons. Whenever we believe something for a reason, our justification for this belief can be defeated or augmented in this distinctive kind of way, viz., by defeating or augmenting one’s justification for a proposition that is neither one’s belief, nor one’s reason to believe, but that is related (in some yet to be specified way) to one’s reason to believe. What we want to know is: how does this phenomenon of “side defeat” work? What is the basing relation such that it admits of such defeat? In this section, I will articulate a dilemma that we must confront in answering this question.7 Our question now is: how can an agent’s C’ing (based on R) have its RDC justification defeated, but not by gaining reasons to doubt R, nor by gaining reasons against C’ing? We cannot plausibly avoid this question by claiming that side defeat is a brute normative phenomenon, insusceptible of deeper explanation. There may be some brute normative phenomena, but side defeat is not among them. One way to answer our question about how side defeat works is by claiming that, for an agent to C, based on reason R, involves the agent’s believing that R supports, or is a good normative reason to, C. Side defeat is possible because, when this latter belief is defeated, then so is her C. Let’s call this the “representationalist” explanation of side defeat. Part of what makes this representationalist explanation plausible is that, when an agent C’s for the reason R, the agent is, in some sense, committed to R’s supporting, C; and defeating this commitment will at least typically defeat the agent’s justification for C’ing. Another way to answer this question is by claiming that, for an agent to C, based on a reason R, involves the agent’s exercising a disposition to C when she accepts R. Side defeat is possible because, when she has reason to distrust this disposition, then she has reason to doubt C. Let’s call this the “dispositionalist” explanation of side defeat. Part of what makes this dispositionalist explanation plausible is that, when an agent C’s for the reason R, the agent is exercising a disposition to C when R8; and acquiring reasons to distrust this disposition will at least typically defeat the agent’s justification for C’ing. In this section, I will argue that neither of these two explanations – neither the representationalist nor the dispositionalist explanation – can work. Suppose that, with the representationalist, we try to explain side defeat by claiming that basing C on R involves believing that R supports, or is a normative reason to, C, and side defeat involves defeat of this belief. On that view, whenever one C’s for the reason R, one also believes that R is a (normative) reason to C, and one’s justification for C can be defeated by defeating one’s justification for believing that R is a reason to C. The problem with this view, as I will now argue, is that the explanation that it gives for side defeat specifies a condition that is not necessary for side defeat. Suppose that we gain extremely compelling empirical evidence for the following psychological hypothesis: someone who has an accurate belief about the support relations between a particular reason R and a particular rationally determinable condition C is thereby very likely to suffer from a very localized cognitive disorder in which their being in C is not responsive to, or controlled by, their having reason R. Indeed, the better justified a person’s beliefs about the support relation between R and C, the more likely the person is to suffer from this localized cognitive disorder concerning their being in C. People who do not at all suffer from this disorder – and so whose rationally determinable conditions are fully responsive to, and controlled by, their normative reasons for those conditions – tend to have almost no accurate beliefs about support relations (either because they don’t have beliefs about such support relations at all, or because the beliefs that they have are inaccurate). If we gain enough evidence to become justified in believing this psychological hypothesis, then we will find ourselves in the following bizarre but nonetheless metaphysically possible type of scenario: we can recognize what normative reasons we have for our RDC’s, and we can recognize the support that those reasons provide to those RDC’s, but, by virtue of our justification for believing the psychological hypothesis above, we will also be justified in believing that the very RDC’s that we take to be supported by the reasons that we have are also not responsive to our having those reasons. And if we are justified in believing that an RDC is not responsive to our reasons for it, then, even if we have justification for being in that RDC, that is not a justification that the RDC itself, once we’re in it, can enjoy. (Smithies 2012 provides cases of propositional justification that cannot be used to create doxastic justification: e.g., evidential justification for believing both p and that I do not believe that p. This sort of “finkish” propositional justification is the kind that I’m describing in my hypothetical sort of case here.) In such a case, our RDC’s themselves will be side defeated, even though we will have compelling reason to be in those RDC’s.9 And so we cannot successfully explain side defeat by saying that it is defeat of the belief that R supports my C’ing, since defeating such a belief is not necessary for side defeat. Side defeat could instead result from my justifiably believing that my RDC is not responsive to whatever reasons I have for it. The representationalist could try to fix her view as follows: for an agent to C, based on reason R, involves the agent’s believing that R supports, or is a good normative reason to, C, and the agent’s C’ing is properly based on this latter belief. But this modified representationalist view takes the notion of proper basing for granted, and so cannot be used to explain what it is for the basing relation to be proper, and so what goes wrong in cases of side defeat. Next, let’s suppose that, with the dispositionalist, we try to explain side defeat by claiming that basing C on R involves not a belief but rather a disposition to C when one accepts R. In that case, whenever one C’s for the reason R, one also has the disposition to C when R, and one’s justification for C can be defeated by giving one reason to distrust this disposition, i.e., to think that this disposition leads one to C erroneously or incorrectly. The problem with this view, as I will now argue, is that the explanation that it gives for side defeat specifies a condition that is not sufficient for side defeat. Suppose that, although we have the disposition to C when R, and we do not have the disposition to C when R* (≠R), we also have justified, false beliefs about what dispositions we have, and we lack some true beliefs about what dispositions we have. In particular, we falsely, though justifiably, believe that we do not have the disposition to C when R; we also falsely, though again justifiably, believe that we do have the dispositions to C when R*; and finally, we do not believe (what is in fact the case) that we have the disposition to C when R, and we do not believe (what is in fact the case) that we do not have the disposition to C when R*. The dispositionalist cannot plausibly deny that this is possible: People generally have plenty of justified, false beliefs about what dispositions they have, so why should not they also have plenty of justified, false beliefs about what RDC-forming dispositions they have? Now, if we have all of these justified, false beliefs, and lack all of these true beliefs, then we could find ourselves in the following bizarre but nonetheless metaphysically possible type of scenario: we have R, we are disposed to C when R, we exercise that disposition and consequently C, but we do not believe that we are exercising that disposition, and we also falsely but justifiably believe that we C not on account of R but rather on account of R*. Furthermore, let’s suppose, we now acquire evidence for the hypothesis that R does not support C. In such a case, since we are fully justified in taking there to be no connection whatsoever between the evidence that R does not support C, on the one hand, and our C’ing, on the other hand, our acquisition of evidence for the hypothesis that R does not support C gives us no reason whatsoever not to C, and so our C’ing is not defeated. A fortiori, C’ing is not side defeated. And so we cannot successfully explain side defeat by saying that it is constituted by our getting reasons to distrust the reliability of our disposition to C when R, since our getting such reasons is not sufficient for side defeat. Again, the dispositionalist could try to fix her view as follows: for an agent to C, based on a reason R, involves the agent’s exercising a disposition to properly base her C’ing on her acceptance of R. But again, this modified dispositionalist view takes the notion of proper basing for granted, and so cannot be used to explain what it is for the basing relation to be proper, and so what goes wrong in cases of side defeat. We are trying to explain side defeat by specifying what it is about the basing relation that makes it liable to suffer from side defeat. What we have shown so far is that this feature of the basing relation is neither the belief that the reason supports C, nor is it a disposition to C when one accepts the reason: defeat of the former is not necessary for side defeat, and having a reason to distrust the latter is not sufficient for side defeat. But what, then, is the feature of the basing relation that explains side defeat? Could it be some simple combination of the two features considered: both the belief that R supports C, and the disposition to C when accepting R? Such a view is put forward in Schroeder 2007: “For R to be the (motivating) reason for which X did A is for the fact that R was a subjective normative reason for X to do A to constitute an explanatory reason why X did A.” But this attempt to combine dispositionalism and representationalism is bound to fail, for it doesn’t fix either of the problems affecting the preceding proposals. Suppose that an agent has undefeated justification for accepting R, and undefeated justification for accepting that R supports C, and she in fact accepts both R, and that R supports C. The agent then has a subjective normative reason to C. Now suppose that the agent has, and exercises, the disposition to C when she has the subjective normative reason just described (viz., she accepts R, and accepts that R justifies C, and has undefeated justification for accepting both of these things). Even in such a case, the agent could still suffer side defeat by having compelling evidence for the psychological hypothesis (which would, in this case, be false, given the stipulations of the case) that her C’ing is unresponsive to her reasons for C’ing. So a view like Schroeder’s cannot give us an adequate account of side defeat. Recall that at least part of what makes the representationalist explanation of side defeat plausible is that, when an agent C’s for the reason R, the agent is, in some sense, committed to R’s supporting, C; and defeating this commitment will at least typically defeat the agent’s justification for C’ing. But the problem with the representationalist account of side defeat is that it locates side defeat as defeat of a belief concerning a particular relation of normative support between R and C, but side defeat can be constituted not by defeating that belief, but rather by calling into question the explanation for C’ing. This is a problem that can arise so long as the representationalist takes the belief that is defeated in side defeat to be something that does not itself fix the reasons why one C’s. Recall that at least part of what makes this dispositionalist explanation of side defeat plausible is that, when an agent C’s for the reason R, the agent is exercising a disposition to C when R10; and acquiring reasons to distrust this disposition will at least typically defeat the agent’s justification for C’ing. But the problem with the dispositionalist account of side defeat is that it locates side defeat as involving reasons for calling into question the reliability of one’s disposition to C when R, but side defeat can fail to be constituted by such reasons, since one can have justified, false beliefs about what dispositions one has. This is a problem that can arise so long as the dispositionalist takes the dispositions to C when R to be dispositions about which one can have fully justified, false beliefs. In order to avoid the objection that we’ve posed, the representationalist must take the belief that is defeated in side defeat to be a belief that fixes the explanation of one’s C’ing. In order to avoid the objection that we’ve posed, the dispositionalist must take the dispositions that explain one’s C’ing to be dispositions about which the agent cannot have fully justified but false beliefs.In the next two sections, I will show how we can perform both of these two fixes at once, and finally arrive at a view of the basing relation that enjoys the plausibility of both representationalism and dispositionalism, while suffering from neither of the objections posed above.
Section 4: Basing as the Use of a Demonstrative Concept As I’ve argued in the preceding section, while basing might involve the belief that R supports C, and it might involve the disposition to C when one accepts R, neither of these conditions on basing can explain side defeat.
So what is it about the basing relation that explains side defeat? On the account that I develop in this section and the next, the basing relation involves the use of a demonstrative concept to refer to something in one’s own psychology. In order to spell out this account, I must say specifically what it is in one’s psychology to which one is thus referring, and I must also say more about the distinctive kind of demonstrative concept by means of which one refers to it. But before turning to either of these two main tasks of this section, first a few background observations.
Whether or not there are non-conceptual demonstratives, one feature of every demonstrative concept is that it involves some general, non-demonstrative concept. To conceptually ostend an object is always to ostend it as of some general kind or other. You can ostend that color, that shape, that occurrence, that thing, that sound, and so on, but all of these demonstrative concepts involve some non-demonstrative concept (i.e., color, shape, occurrence, thing, sound, etc.) Note that this does not imply that demonstrative reference requires that the referent actually fall into the extension of the general concept that partly constitutes one’s demonstrative concept, nor does it imply that one be justified in believing that it fall into that extension. Demonstrative reference can, for all I say here, presuppose lots of false and unjustified belief about the object to which one refers. When an agent uses a demonstrative concept, that agent can be justified or not, and correct or not, in applying the concept that is partly constitutive of that demonstrative concept to the thing ostended by that demonstrative concept. If the correctness of the application manifests the concept-applier’s skill in applying it, then the application is not merely correct and justified, but also knowledgeable. The basing relation, on the view that I propose here, is the use of a demonstrative concept to refer to a particular thing in one’s own psychology. Given what I’ve said just now, this form of deixis will have to employ a general concept, and it will be more or less accurate, more or less justified, and more or less knowledgable, depending upon whether the thinker is correct, justified, or knowledgable, in applying that general concept to the particular referent of the deixis. But what general concept will this deixis employ? And to what will it refer? Answering those questions will require just a bit more background. We’ve so far been using the term “RDC” to refer to anything that can be based upon a reason, and we’ve used the phrase “RDC justification” to refer to the kind of justification that such rationally determinable conditions can enjoy by virtue of being based in the right way on adequate reasons. When a particular RDC (i.e., belief, intention, emotion, etc.) is RDC justified, there is something that makes it so. This is, typically at least, its being based upon the reason upon which it is based. This is what we’ll call the “RDC justifier” of the RDC. The RDC justifier of a particular condition will include not merely the normative reason that one has for being in the condition, but also everything that makes the condition be based in the right way on that normative reason. It has required some work to isolate the concept of a RDC justifier. But, while it has required work to isolate this concept, and there is no easy way of expressing this concept in ordinary English, this does not imply that the concept itself is not an ordinary one. In fact, the concept is possessed by anyone who is capable of asking a particular kind of ordinary “why?” question, or understanding a particular kind of ordinary “because” statement: those questions and statements that concern reasons for which. (Children who are old enough to ask “why did the chicken cross the road?” have the relevant concept, though I don’t know whether pre-linguistic infants have it.) Someone can understand such questions well enough to know when they arise and when they do not, and to be able to assess potential answers to them as more or less relevant, even if she does not have any term in her vocabulary corresponding to our technical term “RDC justifier”. With these remarks in the background, I can now start to spell out my account of the basing relation. For an agent A to C for the reason that R involves A’s ostending an explanatory relation between R, on the one hand, and her own C’ing, on the other, and to ostend it under the concept RDC justifying. While performing this act of ostension might require an agent to have various beliefs and dispositions, the act of ostension itself is neither a belief nor a disposition. Notice that it is possible for you to treat something as a reason to C even when you cannot say what your reason to C is, and even when you believe that you have no good reason to C: especially irrational agents do this sort of thing often, and most of us do it sometimes. There might be reasons for which I am angry at my neighbor, but I might think that, whatever those reasons are, they are almost certainly not good reasons. Still, if they are reasons for which I feel that way, and not merely reasons why I feel that way, then there must be some part of me that is treating those reasons, whatever they are, as RDC justifying my anger. So basing involves using a demonstrative concept that contains the general concept RDC justifier to ostend an explanatory relation between one’s C’ing and one’s reason R. Just as it is possible to ostend the visible distance between two objects even when one is ignorant or mistaken about what those two objects, so too is it possible to ostend the explanatory relation between R and one’s C’ing, even when one is ignorant or mistaken about what R and C are. The basing relation can therefore obtain even between relata that are unknown to, or misidentified by, the agent. It may be worried that what I’ve said about the basing relation so far involves some kind of circularity. The basing relation is that explanatory relation that enables its explanandum to enjoy RDC justification, but the basing relation also involves an agent’s using the concept of RDC justification. Isn’t this a kind of problematic circularity? It is no more problematic than the kind of circularity plausibly involved, say, in accounting for colors (something is red if it appears red to normal observers under normal circumstances), artifacts (something is a hammock if it is used as a hammock), artworks (something is an installation if it is created and received as an art installation), relationships (people are in a relationship of a certain kind if they conceive of themselves as in such a relationship), and more generally, for those things about which some form of projectivist account is plausible. I’ve so far given one necessary condition on the basing relation: it involves using a demonstrative concept containing the general concept RDC justifier to ostend an explanatory relation between one’s C’ing and one’s reason R. But does basing involve any further conditions on the explanatory relation that is thereby ostended? What do we need to add to the claim above in order to get a full account of basing? The next section answers that question.
Section 5: Completing my Account of Basing The basing relation obtains between some reason R, and some RDC C, whenever R is the reason for which an agent C’s. Now it is time to say precisely what this relation amounts to. Recall that one thing we wanted from an account of basing is that it explain side defeat, and we introduced the use of a certain kind of demonstrative concept as the component of basing that gets defeated in cases of side defeat. Another condition of adequacy on our account is that basing involve an explanatory relation: for R to be the reason for which A C’s, it must at the very least be a reason why A C’s. So how shall we build an account of basing that satisfies both of these two conditions? I will consider three proposals. Proposal 1: Basing is simply the conjunction of our two conditions, viz., R is the reason for which A C’s = R is a reason why A C’s, and A uses the concept RDC justifier to ostend an explanatory relation between R and her C’ing. This first proposal is subject to clear counterexample in cases in which R is a deviant reason why A C’s, but A nonetheless incorrectly ostends a distinct explanatory relation between R and C under the concept RDC justifier. Consider, for instance, Davidson’s example of the climber who wants to let go of the rope and let his companion fall; this desire makes the climber so nervous that he trembles, and this trembling causes him to let go of the rope. The climber might mistakenly ostend some explanatory relation between his desire to let go of the rope and his letting go of the rope under the concept RDC justifier, but this would not suffice to make it the case that the reason for which he let go of the rope was that he wanted to do so. The problem with this proposed account of basing — what seems to render it subject to counterexample — is that the explanatory condition and the ostending condition are treated as independent.11 An adequate account of basing should connect these conditions to each other more closely. This suggests a second possible account of basing. Proposal 2: Basing is the obtaining of the ostending condition, caused by the obtaining of the explanation condition, viz., R is the reason for which A C’s = R is a reason why A C’s, and A uses the concept RDC justifier to ostend that very explanatory relation between R and her C’ing. This second proposal has the virtue of satisfying our two constraints on an account of basing. And it also has the virtue of relating the two conditions so as to avoid the sorts of counterexample just described. But it has the vice of being subject to still other counterexamples. In particular, it fails to handle cases in which R is a deviant reason why A C’s, but A treats R as RDC justifying her C’ing because of R’s being a reason why A C’s. We can construct an example of this kind by modifying Davidson’s case of the climber slightly. Suppose that the climber’s desire to let go of the rope not only causes him to be so nervous as to let go of the rope, but furthermore, in causing him to do this, it also causes him to use the concept RDC justifier to ostend that very same explanatory relation between his desire and his letting go of the rope. This still would not make his desire the reason for which he lets go of the rope. So, though the two conditions on basing need to be connected in order to secure a proper account of basing, the connection needs to be of the right kind. Let’s try once more. Proposal 3: Basing is the obtaining of the explanation condition in virtue of the obtaining of the ostending condition, viz., R is the reason for which A C’s = A uses the concept RDC justifier to ostend an explanatory relation between R and her C’ing, and in virtue of that fact, R is a reason why A C’s.12 Here, at last, we’ve reached an account that satisfies our two conditions on an account of basing, and can also handle all the cases correctly. The basing relation is an explanatory relation (a “reason why” relation) that obtains in virtue of our demonstratively referring to that very relation under the concept RDC justifier. Side defeat happens when, and because, the agent’s justification for applying the concept of RDC justification to the ostended explanatory relation is defeated. Recall that defeating the agent’s reason to believe that R supports C was not necessary for side defeat (since side defeat could occur even while the agent was still justified in believing that R supports C, if she was also justified in believing that her C’ing is not RDC justified), whereas defeating the agent’s reason for trusting her disposition to C when R was not sufficient for side defeat (since side defeat could fail to occur, so long as the agent was justified in trusting what she justifiably took to be the causally relevant disposition). The present account of side defeat avoids the problems of each of these other proposals. The ostended explanatory relation is identical to the real explanatory relation, since the latter is real only by virtue of being ostended. And the application of the concept of RDC justification to that explanatory relation is defeated by any justification for suspending the resulting RDC. This proposal may strike some philosophers as metaphysically odd: how can an explanatory relation obtain in virtue of our referring to it in thought? It can help to mitigate the sense of oddity to think of other cases in which an explanatory relation obtains in virtue of our referring to it. I say “I hereby pronounce you husband and wife”, and in virtue of saying these words, I bring into being the very same relation that the words describe. I think “I hereby think a self-referential thought”, and in virtue of thinking this, I bring into being the very same thought to which my thought refers. These are all cases of ostension, but they have their direction of fit is the opposite of the direction of fit of most of the examples of ostension that philosophers typically think about. These cases are all examples of “conjuring”, in the literal sense of that term. And so I say that, on my account, the basing relation is an act of conjuring.13 I conclude that proposal 3 is correct. Basing is when an explanatory relation between R, on the one hand, and A’s C’ing, on the other, obtains by virtue of A’s ostending that explanatory relation under the concept of RDC justification. You can ostend a thing only if that thing exists. And so you can ostend an explanatory relation only if that explanatory relation exists. Thus, an agent’s ostending an explanatory relation under the concept RDC justifying is possible only if that explanatory relation obtains. But, according to this proposal, the explanatory relation is itself grounded in the ostensive act. It follows that the ostensive act that we’ve described is both necessary and sufficient for the obtaining of the ostended explanatory relation. Since the explanatory relation is the basing relation, it follows that the basing relation obtains when and only when an agent ostends that relation under the concept RDC justifying. How does basing, so understood, explain side defeat? Side defeat is the defeat that occurs when the agent’s act of conceiving of this explanatory relation – the explanatory relation between R and C – as one of RDC justification is itself defeated. Side defeat directly attacks an agent’s justification for subsuming this explanatory relation under the concept RDC justification. Consider again the representationalist’s proposal that side defeat is constituted by having reason to doubt one’s belief that R supports C. This proposal failed because one could have no such reason, and still suffer from side defeat by virtue of having evidence that one’s C’ing was not responsive to R. The only way to fix this problem with representationalism is to say that one’s belief about the relevance of R to C fixes the reason why one C’s. And this is what we’ve done here: the explanatory relation between R and one’s C’ing is itself grounded in one’s ostending this explanatory relation under the demonstrative concept of RDC justifier, and to perform this ostensive act is to be committed (perhaps correctly, perhaps incorrectly) to a relation of normative support between R and C.14 Consider again the dispositionalist’s proposal that side defeat is constituted by having reason to doubt the reliability of one’s disposition to C when R. This proposal failed because one could have such reason, and yet be fully justified in taking it to be irrelevant to the RDC justification of one’s C’ing. The only way to fix this problem with dispositionalism is to say that one’s disposition to C when R is not a disposition about which one could have fully justified but false beliefs. And this is what we’ve done here: the explanatory relation between R and C is the exercise of a disposition that occurs in virtue of one’s ostension of that very exercise as justificatory, and so in virtue of a fact to which one has privileged access. One might, of course, have false beliefs about various matters of fact to which one has privileged access – but these false beliefs cannot be fully justified, so long as one has privileged access to the facts that belie these beliefs. Finally, let’s return to the issue of how we can distinguish proper basing – the kind of basing the enables RDC justification – from improper basing – the kind that doesn’t. Our answer is simple: proper basing is not side defeated, while improper basing is.15