Skepticism, Logical Independence, and Epistemic Priority
Radical skepticism about the external world is founded on two assumptions: one is that the mind and the external world are logically independent; the other is that all our evidence for the nature of that world consists of facts about our minds. In this paper, I explore the option of denying the epistemic, rather than the logical assumption. I argue that one can do so only by embracing externalism about justification, or, after all, by rejecting the logical independence assumption. Since (I argue) externalism is not a solution to the problem of skepticism, this means that skepticism is false only if the mind and the world are not logically independent.
Skepticism, Logical Independence, and Epistemic Priority
Department of Philosophy
University of Florida
Gainesville, FL 32611
Skepticism, Logical Independence, and Epistemic Priority
This paper is concerned with the logical space of solutions to the traditional problem of skepticism about the external world, which I will argue is much smaller than one might have supposed. Skepticism about the external world I take to be the problem, not of showing how knowledge or justified belief is possible, but of providing a reflective grounding for our beliefs about the external world. The salient fact about this project for our purposes is that success in carrying it out requires that we be able to show or give reasons to believe that a belief is justified, if it is. This requires minimally that we be able to show that we have more reason than not to believe what we do about the world around us.
Radical skepticism about the external world, in this sense, is founded on two assumptions about the mind-world relation:
(L) Propositions solely about the contents of the mind are logically independent of propositions solely about the nature of the external world.
(E) Facts solely about the contents of our minds are all our evidence for facts about the external world.
I will call these the assumptions of logical independence and epistemic priority.
I believe that (L), appropriately understood, is correct, but I do not argue for that in the following. Instead, I explore the possibility of rejecting (E), the assumption of epistemic priority. I argue that it is not possible to reject (E) without either embracing epistemic externalism or denying (L). Since (as I will argue) epistemic externalism is not a solution to skepticism as I have characterized it, the logical space of solutions to skepticism is very small. Skepticism is false only if (L) is false.
My procedure is the following. First, I clarify the premises (E) and (L) of the skeptical argument. Then I show how these lead to radical skepticism about the external world. Next, I argue that epistemic externalism does not offer a solution to this traditional skeptical problem. Finally, I examine the considerations in favor of (E), and explore the options for rejecting it, given (L).
I begin with a clarification of the notions of logical independence and epistemic priority at work in the skeptical argument. The relations of logical independence and epistemic priority are in the first instance relations among propositions.
The proposition that p is logically independent of the proposition that q if and only if neither entails the other or its negation or anything about its probability. If (L) is true, then no proposition solely about the contents of our minds entails any proposition about external world, and no proposition solely about the external world entails any proposition about the mind. We return in a moment to the question what it is for a proposition to be solely about the mind or world.
I use ‘epistemic priority’ to indicate an asymmetric relation of support between propositions for an epistemic agent. I will always treat this relation as relative to an agent, not a relation that holds between propositions independently of relativization to an agent. If one proposition is epistemically prior to another for an agent, that implies something about which could serve as evidence for the other for the agent. I will say that the proposition that p is epistemically prior to the proposition that q for x if and only if the proposition that p would, certain other conditions being met, be evidence for the proposition that q for x. The other conditions include that x be justified in believing that p and that it is a fact that p. If x has evidence for the proposition that q, then there is a proposition, let us say the proposition that p, such that the proposition that p is evidence for x for the proposition that q. I will not undertake any further analysis of either of these notions here. This will be enough for our purposes.
What is it for a proposition to be solely about the mind, or solely about the external world? Each of these categories of propositions has to be relativized to a thinker. When we are interested in the problem of the external world, we are each of us interested in the relation between facts about our own minds and facts about the rest of the world, if any. Thus, for example, for one thinker, a proposition about another’s mind would be a proposition about the world external to him, while for that other, the same proposition would be about his own mind and not about the external world.
In speaking of propositions about someone’s mind, we do not take ‘mind’ seriously as a count noun. We mean rather a proposition that entails that he has a pure psychological state. A state of an individual at a time t consists in his instantiating some property at t. An individual has a psychological state at a time t then just in case he instantiates a psychological property at t. A state is a pure psychological state just in case it involves being in a psychological state and having it does not entail instantiating an epistemic property, such as being justified, having warrant or knowing that p. Henceforth I drop the qualifier ‘pure’, but it should be understood throughout. Psychological states are either conscious states or intentional states. For present purposes, we will want to focus on those psychological states of a person to which he has non-inferential access. I will leave open how to analyze this sort of access we have to some of our psychological states. Examples of states to which we have this sort of access are certain occurrent beliefs and desires, states of feeling pain, anxiety, joy, many perceptual experiences, and so on. I will call these basic psychological states. Among the psychological states that are not in this sense basic are propositional attitudes whose contents are characterized by direct reference to individuals distinct from the thinker, so that the individuals themselves are involved in the individuation of the content. For example, my belief that Mars is the fourth closest planet to the Sun is not a basic psychological belief because to know that I have it I must know or be in a position to know that Mars and the Sun exist, and these are not things I could know non-inferentially. Basic psychological states are those that are supposed to be preserved invariant under traditional skeptical scenarios such as the brain in a vat hypothesis, and Descartes’s Evil Demon hypothesis. To say this much is not to presuppose, however, that we could have the states that they suppose invariant in those worlds very different from how we believe the world to be. I will say then, that a proposition is about the mind for x (or about x’s mind) iff it entails that x has a basic psychological state.
A proposition is about the external world for x just in case it entails the existence or probable existence (or non-existence) of a contingent existent distinct from x. Henceforth, I will omit ‘for x’, but the relativization to individuals should be understood throughout.
A proposition is solely about the mind (or external world) just in case it is about the mind (or external world) and does not have as a logical consequence a proposition that is about the external world (or mind). This leaves it open whether any proposition solely about the mind entails a proposition about the external world, or vice versa. I will drop the modifier ‘solely’ from ‘solely about’ below, but I should be understood to mean ‘solely about’ throughout the following discussion.
If (E) is true, then the only propositions we can appeal to in justifying our beliefs about the external world are propositions about our own minds and necessary truths—that is, in any argument we give to show that our beliefs about the external world are justified all of the premises must be either necessary truths or about the contents of our own minds.
If the mind-world relation is correctly characterized by (L) and (E), then we can give neither an a priori nor an a posteriori grounding for our beliefs about the external world. (E) requires that we give an account of our knowledge of the external world in terms of the contents of our minds and necessary truths. If (L) is true, then no proposition about any mind entails any proposition about the external world, and, since (L) allows that there could be a world in which only mental propositions were true, no propositions about the external world are necessarily true. Since (E) constrains our starting point to propositions about our minds and necessary truths, there is no possibility of knowing or arguing a priori from the contents of the mind to the nature of the external world. To provide an a posteriori argument we would have to establish a projectible correlation between events and states of our minds and the external world. To establish such a correlation we would have to have independent reflective access to both the contents of our minds and the external world. Independent access to the external world is ruled out by (E). Therefore, if (L) and (E) are true, it is impossible to show that we have even minimally justified belief about the world around us. If the possibility of a reflective grounding of our beliefs about the external world is a necessary condition on their being justified, then it is impossible to be justified in believing anything about the world around us.
If this account of the assumptions underlying skepticism about the external world is correct, then skepticism is false only if (L) or (E) is false. Traditional responses to skepticism about the external world have denied (L) by making the nature of the world depend logically on our minds. More recently, externalist theories of thought content have held out the hope that we could deny (L) without the traditional commitment to idealism. Though I will not argue for it here, in my opinion, neither of these responses is successful. If this is right, then skepticism about the external world is false only if (E) is.
There is one other response to skepticism that might be thought to succeed without denying either (L) or (E), namely, epistemic externalism. My goal in the rest of this paper will be, first, to argue that epistemic externalism does not offer a way of refuting skepticism which satisfies the constraints on the project given above, and, second, to show that we cannot accept (L) and deny (E) compatibly with carrying out the project of providing a reflective grounding of our beliefs about the external world. The attempt to do so requires us to embrace epistemic externalism after all. This shows that the skeptical challenge can be met only if (L) is false.
The epistemic externalist argues that the skeptic has misconceived the conditions necessary for justified belief, and, in particular, argues typically that what is required is not that we be able to construct an argument from what we know or are justified in believing about our minds to facts about the external world, but simply that certain non-epistemic relations hold between our beliefs and what they are about (e.g., for our beliefs to be justified, that they be formed by a process that reliably produces true beliefs). This maneuver strongly suggests identifying externalist theories of justification with theories that seek to reduce those epistemic relations the skeptic denies we can bear to the world to non-epistemic relations. One of the benefits of our investigation will be that it shows clearly that though this is a sufficient condition for a theory to be externalist, it is not necessary. On such a view whether or not we have justified beliefs about the external world is up to the nature of the external world itself, and not something that we need to be able to cite legitimately in an argument to show that our beliefs are justified. Externalist theories of justified belief aim to solve the problem of skepticism by showing not that our beliefs are justified but that it is possible that they are.
I will not argue that externalist theories are false, although I do not think any satisfactory reductive externalist theory has been proposed. We can show that externalism does not respond to our problem without showing that it is false. Skepticism of the sort we are concerned with is directed against the possibility of providing a reflective grounding for our beliefs about the external world. If epistemic externalism is true, this would show that the skeptic would be wrong to put his conclusion by saying that justified belief about the external world is impossible. But this would not show that the project that gave rise to the skeptical problem could be carried out. The strategy that externalists employ to show how justified belief can be possible in the face of the skeptical challenge deliberately puts the conditions for justification on the external world side of the traditional appearance-reality gap. Thus, from the point of view of the traditional project, the problem of showing that our beliefs are justified remains untouched.
The problem is not that externalist theories do not have the resources to allow that we be justified in believing that our beliefs about the external world are justified. If we interpret ‘justified’ throughout as the externalist urges, then there is no more difficulty in being justified in believing a belief about the world is justified than there is in being justified in believing something about the world. In each case, the requirements are that the belief be appropriately non-epistemically related to its object, perhaps that it be produced by a reliable mechanism. Only an inept formulation of the externalist condition would prevent this from being logically possible. However, we want to show not just that we could be justified in believing that our beliefs about the world around us are justified, but that they are justified, and in particular in the light of (L) and (E). (E) constrains the reflective project to show this by constructing an argument whose premises are all about our own minds or are necessary truths. The properties that on typical externalist accounts make our beliefs justified are not psychological properties. To show that our beliefs about the external world are justified, given (E), we would have to show that our beliefs had those properties by appealing only to facts about our minds and necessary truths. But since these properties are not psychological, and our beliefs do not have them necessarily, epistemic externalism provides no solution to a skepticism founded on (L) and (E).
If my account of the reasoning underlying radical skepticism is correct, then if we retain the logical assumption and epistemic internalism, but deny skepticism, then we must deny the epistemic priority of the mind to the world. If we want to avoid raising a skeptical problem about our own minds, this requires that we hold that our epistemic access to the world and to our own minds is independent, that is, that we can know things about our own minds and about the external world independently, just as we can know things independently, for example, about the heavens and the earth. This position I call direct realism. One way of putting my conclusion is that the only way to be a direct realist and retain (L) is to be an epistemic externalist, which, as we have seen, is not a solution to the skeptical problem.
The source of the difficulty for direct realism lies in the possibility of making a certain kind of mistake about the external world, in which our error is referred not to some misleading feature of the world, but to a misleading feature of our experience. If I mistake a decoy duck for a real duck, then what I knew that misled me, what I went on in forming my beliefs, so far as explaining this error goes, was a fact about the external world, that a certain object I was in perceptual contact with looked like a duck. In explaining my error we are not forced to cite any proposition about experience as epistemically prior to a proposition about the external world. We do not, that is, have to cite any experience as evidence for the mistaken belief. In contrast, if I hallucinate a duck, or dream that I see a duck, I cannot cite perceptual contact with an object in my environment which appeared to be something other than it was, since I am not in perceptual contact with anything. To explain my error in terms of misleading evidence I must cite my experience as what misled me. In doing so, I treat it as part of my evidence, and, consequently, treat propositions about my experience as epistemically prior to propositions about the world.
These familiar facts do not lead to skepticism. It is compatible with propositions about my experiences being sometimes epistemically prior to propositions about the external world, or my experiences always being part of my evidence for the nature of the external world, that they are not my sole evidence. It would be a mistake to infer from the fact that my experiences are sometimes or even always part of my evidence for the nature of the world around me that they are my sole evidence.
The difficulty arises when we conjoin this model for explaining error with the apparent coherence of globally misleading hallucinations or dreams, the logical possibility that all of our experiences and other psychological states are the same as they are, although all of our beliefs about the world around us are false. The coherence of traditional thought experiments such as the dream doubt and the evil demon hypothesis illustrates both the logical independence of our minds and the external world and provides the strongest case for the epistemic priority of the one to the other. Let us call a world in which we have all of the experiences, beliefs, and other psychological states that we in fact have, although our beliefs about the external world are false, and in which we have no way to discover that they are from how the world appears to us, a d-world. In a d-world we would neither know nor be justified in believing anything about the external world. In a d-world all of our beliefs about the external world are false, so we cannot know anything about the external world in a d-world. Moreover, there is no connection, logical, nomological, or probabilistic, between our having the beliefs and experiences we do in a d-world and the truth of any proposition about the external world. We could say that in a d-world we were internally justified in believing what we do, in the sense that given our most firmly held beliefs, our practices in forming and rejecting propositions about the external world would be to varying degrees epistemically responsible. But such justification would promote the epistemic goal of believing truths and avoiding falsehoods only relative to beliefs whose truth or probability is assumed. In a d-world those assumptions are false. If justification does not guarantee truth, it must still retain some connection with it on pain of divorcing justification from the epistemic goals it must promote. If a belief is justified for an agent, there must be some true beliefs of his relative to which it is more likely than not to be true. In a d-world, this condition is not met, for there would be no connections between our beliefs and what they were about. Consequently, in a d-world, we would not know or be justified in believing anything about the external world.
Why would we have incorrect beliefs about the external world in a d-world? It is natural to take this as a request for a source of misleading evidence. But then the only possible answer seems to be that, in a d-world, we would be led by our experiences to believe that the world was very different from the way it was. For in a d-world, if we have evidence, this cannot consist of any facts we know or are justified in believing about the external world. There are no such facts. If we have evidence, it must consist of facts about our minds. None of our beliefs about the external world are justified a priori. So our beliefs about the external world are not our ultimate evidence for other beliefs about the external world. If anything is our evidence, it must then be our perceptual experiences, how the world perceptually appears to us to be. Thus we must cite (if anything) our experiences as our evidence for the nature of the world around us. That it is appropriate to cite our experiences in a d-world as our evidence is suggested by the legitimacy of citing it in cases of more limited error. And if in a d-world there is no possibility of finding out that we are misled, it must be that our experiences as a whole in d-worlds are epistemically prior to any knowledge of the external world. If we had any other evidence, then it would be possible to correct our beliefs, but by hypothesis this is not possible in a d-world. So if we have evidence for our beliefs about the external world in a d-world, (E) is true of us in that world.
But if (E) would be true of us in a d-world, how can we resist the conclusion that (E) is true of us in the actual world, and that consequently there is no escaping the skeptical argument if we accept (L)? To resist this, we must hold that in the actual world, in contrast to a d-world, some of our evidence for the nature of the world around us consists in facts about the external world that we are justified in believing, but not on the basis of any facts about our minds. We must hold that we have independent access to the external world. But if in a d-world we would not be justified in believing anything about the external world, we would not there have independent access to the external world. So our having independent epistemic access to the external world would have to be a contingent fact about us. It follows that epistemic facts do not logically supervene on facts about our psychological states, since these are by hypothesis the same in d-worlds and in worlds in which we are supposing we would have independent epistemic access to the world around us. Such knowledge or justified belief would be synthetic a priori, for it would be about contingent facts but not based on perceptual experience. So far as anything we have said goes, it can be taken to involve a primitive epistemic relation to the external world.
This is an extremely unsatisfying philosophical position. To hold it is to accept that there can be no philosophical explanation for our knowledge of the external world; it is to hold that our knowledge of the external world is a brute fact. There can be no philosophical explanation of our knowledge because if we are forced to this position, the only explanation we could give would have to rely on the very knowledge whose possession we were trying to understand.
But this in itself would not show that this response to the skeptic is inadequate. Such a response must serve us in some areas, as in the case of knowledge of our own minds. The greatest obstacle to adopting this position as a response to skepticism about the external world is not that it is philosophically unsatisfying. It is more surprising than this, and reveals something important about what it is to take up the reflective standpoint. The problem is this. If we accept (L) and reject (E), then, as we have said, epistemic facts do not logically supervene on facts about experience. However, if epistemic facts do not logically supervene on facts about experience, then the justification of our beliefs about the external world, if any, can make no difference to us, in the sense that it cannot be anything that we can, after all, show that we have in the sense required to carry out our project.
The reason is twofold. (1) First, what is available to us from the standpoint of providing a reflective grounding of our beliefs about the external world is the evidence we have available to us without any further investigation, that is, how things initially seem to us. (2) But, secondly, how things seem to us is not a fact about the world around us but rather a fact about our mental lives, and in particular our conscious mental lives. If our conscious mental lives are the same in two possible worlds, then how things seem to us is the same in each world. In saying that all our mental states could be the same in a d-world and in the actual world, we allow that how things seem to us is the same in both worlds. But if we maintain that in the actual world, in contrast to a d-world, we have epistemic access to the external world, then we must hold that the fact that, e.g., the belief that I have a body is justified makes no difference to how things seem to me. But if this makes no difference to how things seem to me, then whether I am justified in believing I have a body is inaccessible to me from the standpoint of reflection. Seemings are on the appearance side of the traditional appearance-reality distinction, and if the epistemic status of my beliefs about the external world could make no difference to how things seem to me, then their epistemic status is no more present or available to us than is the world of which we sought knowledge. Such knowledge or justification must fail the test of a reflective endorsement of our beliefs about the external world.
We rejected externalism above as a solution to the skeptical challenge when we took it to be a way of trying to avoid denying either of the two assumptions (L) or (E). Now we have tried to reject (E). Yet, paradoxically, this does not seem to offer us a solution to the traditional problem in its own terms. We have no sense of having escaped being trapped behind or within the endless stream of our own experiences. Our problem is that, on this view, whatever epistemic status our beliefs about the external world have is not, if it cannot impinge on our conscious mental lives, something that is available to us from the point of view of reflection. This shows us that to have evidence about the world in a way that will meet the skeptical challenge what we have must be accessible to consciousness. It must be something that, as we review our beliefs and experiences, makes a difference to what we find.
Why should the reflective standpoint require that having evidence for something make a difference to consciousness? The difficulty is that the standpoint of reflection is how things seem to me. How things seem to me is how things appear to me; how things appear to me is a matter of the states of my mind, not of the world, and specifically conscious mental states, since things do not seem any way to me at all when I am knocked unconscious. Thus accepting (L) while trying to reject (E) drives us to embrace a form of externalism. Externalism frustrates the traditional project by putting beyond the standpoint of reflection the properties we seek to show our beliefs possess. Consequently, unlike (L), (E) is dictated by the nature of the project itself.
We can see now why it would be a mistake to characterize externalist theories as reducing epistemic relations to non-epistemic relations, although in practice this is the externalist’s procedure. Reducing epistemic properties of beliefs about the external world to non-epistemic facts about the external world is sufficient to place them beyond the reach of appearances. But the position we have reached, which postulates primitive access to the external world, shows that this is not necessary; for by allowing that our primitive access makes no difference to how things seem to us, it gives this primitive epistemic property, in the context of our project, the same status as any other fact about the external world.
Let us say that things seem the same to us in possible worlds w and wN provided that the sequence of conscious mental states we have in w is the same as that in wN; and that if it is possible for things to seem the same to us whether or not it is the case that F, then we cannot tell whether or not it is the case that F. This provides us with the resources for a simple characterization of externalist theories of justification: An epistemic theory of justified belief about a domain of facts W is externalist if and only if we cannot tell whether or not the conditions hold that distinguish a justified belief from an unjustified belief about facts in W. Thus, what makes a theory externalist is that it entails that we cannot tell (not: cannot know—for this difference the externalist can accommodate) the difference between a world in which we are justified in believing W-facts and one in which we are not.
Can we reject (E) but avoid embracing epistemic externalism? To do so we would have to hold that the epistemic status of my beliefs must make a difference to how things seem to me. But if the epistemic status of my beliefs about the external world must make a difference to how things seem to me, it cannot be that there is a possible world in which all of my psychological states are the same, and, hence, a possible world in which everything seems the same to me, and yet in which I am not justified. So to insist that the epistemic status of my beliefs about the external world must make a difference to how things seem to me is to deny the possibility of a d-world, and therefore to rebuild the logical link between experience and the world we had rejected earlier. We are faced with a dilemma. Either we once again find ourselves forced to deny the logical assumption or we open as wide a gap between our experiences and the epistemic status of our beliefs as that between our experiences and the external world, and so embrace a form of externalism.
Is there another option? Can we deny that in a d-world all our evidence for the external world would be our experiences? We could do this only by denying that we had any evidence at all. This would amount to charging that the question we ask about agents in a d-world, “Why do they believe these things about the external world?”, when interpreted as a request for their grounds, has a false presupposition, that there is some evidence that they go on. But it might be said that they go on no evidence at all, any more than someone who acquires a belief by being knocked on the head, and that consequently we are not committed to the claim that all their evidence for the nature of the world around them is their experiences. We would have no difficulty then in accounting for why our experiences are not our sole evidence.
However, this is not a way out of the dilemma. For if we are to provide a reflective account of the epistemic status of our beliefs, we must be able to show that we do have evidence for the nature of the external world. In a d-world, it remains true that we would not. So if we do, we must contingently have independent access to the external world. Whether or not we have access to the external world either makes a difference to how things seem to us, or not. If not, then our knowledge, if we had it, would be “as nothing to us.” If so, then a genuine d-world is not possible, and we have given up the logical independence of the mind and the world.
There is no solution to the skeptical problem in this domain, or in any other domain in which it has the same structure, which respects both our desire to give a reflective account of our knowledge and to maintain logical independence of our evidence and what it is evidence for. Another way to put the conclusion is that the connection we require between justification and truth for genuine justification can be secured only by denying the logical independence of thought and the world or by accepting externalism and giving up the reflective project.
 I have benefited greatly from discussion about the topics of this paper with Tony Brueckner, Steve Jacobson, Peter Klein, and Russell Wahl. I owe a more general debt to Thompson Clarke and Barry Stroud for their influence on my thinking about skepticism.
 For stylistic variation and sometimes for brevity I use ‘external world’, ‘world around us’, and ‘world’ interchangeably
 For convenience I drop the second disjunct below.
 My discussion throughout focuses on justified belief. Justified belief is a necessary condition on knowledge, and so to that extent the discussion also applies to skepticism about knowledge of the external world. However, I believe that the central problem about our knowledge of the external world has always been how to show that our beliefs are justified, rather than that they constitute knowledge, and so this will be the focus of my discussion.
 I do not consider in this paper responses to skepticism which question the intelligibility of the concepts or terms in which the skeptical puzzle is developed, and in particular I will not consider responses which deny the intelligibility of the modal concepts of conceptual or broadly logical necessity and possibility, which are central to this formulation of the traditional problem.
 For convenience I sometimes talk about the mind and the world being logically independent or the mind being epistemically prior to the world, but in each case these expressions are to be interpreted as about propositions about the mind and propositions about the external world.
 To say that the proposition that p entails the proposition that q is not to say merely that necessarily, if it is the case that p, then it is the case that q, interpreting the conditional as the material conditional. The latter may be true even if there is no ‘internal’ connection between the proposition that p and the proposition that q, if, for example, the proposition that p is necessarily false, or the proposition that q is necessarily true. I take entailment to be basic for present purposes, and I understand it as a priori entailment in the sense that if a proposition that p entails a proposition that q, then anyone who grasps both propositions is in a position to know that the latter follows from the former on a priori grounds. In the special case that the propositions that p and that q are both contingent, then that necessary, that p is true only if that q is true is sufficient for the proposition that p to entail the proposition that q. I will mean by ‘necessity’ throughout conceptual necessity, or what has been called ‘broadly logical necessity’. I take conceptually necessary truths to be a priori knowable, though I do not take all a priori truths to be necessary (e.g., it is a priori that all actual philosophers are philosophers but it is not necessary that all actual philosophers are philosophers). I do not think there are any a posteriori necessities, but I cannot go into the reasons for that in this paper; in any case, I mean to exclude these, if there are any, from the extension of my use of ‘necessary truths’.
 See my “Singular Thought and the Cartesian Theory of Mind” Nous, 30(4), 1996: 434-460.
 Here I invoke the technical notion of logical consequence extended from sentences to propositions through the ‘expressing’ relation. That p is a logical consequence of that q just in case in some language a sentence that expresses the proposition that q has as a logical consequence a sentence that expresses the proposition that p. A sentence C in L is a logical consequence of a sentence or set of sentences S in L iff on all interpretations of non-logical terms in L on which S or sentences in S are true in L, C is true in L. I am here assuming that propositions are as finely individuated as sentence meanings so that two sentences express the same propositions iff they are the same in meaning (relativized to their respective languages). Differences in compositional structure will be counted as differences in meaning, so that two sentences of different logical forms will not count as synonymous.
 It might be thought that this argument fails because it does not block the possibility of arriving at justified belief about the external world by way of inference to the best explanation. For inference to the best explanation to be a mode of acquiring justified belief about the external world that meets the skeptical challenge, we must be able to identify features of the best explanation of our mental lives without independently establishing that the explanation is true or likely to be true. So by ‘best’ in ‘best explanation’ we cannot mean simply ‘correct explanation’ or ‘probably correct explanation’. The best explanation then would be identified on the basis of formal features of it such as its scoring highest on some measure of simplicity. Then in addition one would have to have reason to think that scoring highest with respect to such features ensured that or made it likely that the explanation was correct. Let us say that the relevant features are F. Then we need to establish the principle [BE] that the explanation with the highest F-score is likely to be correct. We cannot appeal to an a posteriori justification of [BE] because that would presuppose we had contrived independently to meet the skeptical challenge, making the appeal to inference to the best explanation otiose. Thus, we would have to establish [BE] a priori. I think the prospects for this are dim. But independently of this assessment, if [BE] could be established a priori, and it could be established that the best explanation of our mental lives required most of our beliefs about the external world to be true, then it would in effect show that certain facts about the actual character of our mental lives entailed that it was likely that our beliefs about the external world were true. For what determines which explanation of a domain of facts has the highest F-score is the structure of the facts in that domain itself. Thus, if (L) is true, [BE] is not a priori. To put this another way, as I noted above, if that p and that q are contingent propositions, neither necessarily true nor necessarily false, then that p entails that q provided that necessarily, if the proposition that p is true, then the proposition that q is true.
 This amounts to a weak transparency requirement on justified belief. This does not require that if we are justified in believing that p then we are justified in believing that we are justified in believing that p, nor is it necessarily met if being justified in believing that p entails being justified in believing that one is justified in believing that p; it does require that to be justified in believing that p we be in a position to show that we are justified in believing that p. To show that we are justified in believing that p would be sufficient for us to be justified in believing we are justified in believing it, but not necessary.
 The conclusion does not require the claim that none of our beliefs about the external world in a d-world would be justified. What is needed is the observation that instances of (2) follow from instances of (1).
(1) that p is evidence that x has for its being the case that q
Since none of our beliefs about the external world would be true in a d-world, we could not have any evidence other than facts about our own minds in a d-world.
 Barry Stroud makes this point persuasively in “Understanding Human Knowledge in General,” in Understanding Human Knowledge: Philosophical Essays (2000), New York: Oxford University Press.