Erkenn (2013) 78:1185–1196 DOI 10.1007/s10670-012-9393-4 ORIGINAL ARTICLE

The Argument from Abomination Michael Veber

Received: 24 April 2012 / Accepted: 18 August 2012 / Published online: 12 September 2012  Springer Science+Business Media B.V. 2012

Abstract The conclusive reasons view of knowledge entails the ‘‘abominable conjunction’’ that I know that I have hands but I do not know that I am not a brain in a vat. The argument from abomination takes this as a reason to reject the view. This paper aims to buttress the argument from abomination by adding a new sort to this list: the logical abominations. These include: ‘‘I know that argument is sound and that sound arguments have true conclusions but I don’t know whether the conclusion of that argument is true’’. Two standard replies to the argument from abomination are raised. It is argued that the logical abominations open new holes in both.

1 Introduction According to Dretske (1971, 2005), S knows that P if and only if P is true, S believes that P and S’s belief that P is grounded on a conclusive reason. Conclusive reasons are defined in the following way: R is a conclusive reason for P = df R would not be true unless P were true (alternatively, if P were false, R would be false). The conclusive reasons requirement famously leads Dretske to claim that I can know that P, know that P entails Q but be in no position to know that Q. Consider this old example: My basis for believing that I have hands is my perceptual experience.1 Here is one hand; here is another. This provides a conclusive reason for thinking that I have hands because were it not true that I have hands then I would 1

Dretske’s view allows that propositions about experiences as well as the experiences themselves can function as reasons to believe. That sort of distinction will not be important in this discussion.

M. Veber (&) Department of Philosophy, East Carolina University, Greenville, NC 27858, USA e-mail: [email protected]

123

1186

M. Veber

not have these perceptual experiences. In the language of possible worlds, the nearest world where I do not have hands—a world, say, where I lost them in a horrible shark fishing accident—is also a world where my perceptual experiences are noticeably different. Or, if you prefer, a world where I do not have hands and I do not have these kinds of experiences is nearer to the actual world than any world where I do not have hands but I have these experiences anyway. Since it is an obvious and necessary truth, I also know that if I have hands then I am not a brain in a vat. But, and here is where the surprise comes, my perceptual experience does not provide conclusive reason to believe that I am not a brain in a vat. This is because, if I were a brain in a vat, things would look just as they do now. I know that I have hands and that having hands entails not being a brain in a vat but this puts me in no position to know that I am not a brain in a vat.2 One nice consequence of the conclusive reasons account of knowledge is that it seems to preserve both our sense that radical skepticism is mistaken and the widely shared suspicion that I cannot know I am not a brain in a vat just by taking a gander at my various appendages and reasoning from there. A common objection to this position, however, is that it entails that certain ‘‘abominable conjunctions’’ (DeRose 1995) are true. For example, ‘‘I know that I have hands but I don’t know that I am not a brain in a vat’’. According to some, the abominable consequences of the conclusive reasons view of knowledge show that it is unacceptable. I shall call this ‘‘the argument from abomination’’. My aim here is to buttress the argument from abomination by adding a new variety to the list: the logical abominations. After those are presented, I shall consider two replies to the argument from abomination. I will argue that the logical abominations open new holes in both. Dretske is not the only epistemologist who has been put on the receiving end of the argument from abomination. Nozick’s (1981) famous sensitivity requirement on knowledge also has the result that I can know that I have hands, know that having hands entails not being a brain in a vat and yet be unable to know that I am not a brain in a vat. Although Dretske’s epistemology will be the target of this discussion, the criticisms offered here should apply equally well mutatis mutandis to Nozick’s epistemic theory and any other relevantly similar theory of knowledge.

2 Abominations Abound Consider the following anti-skeptical argument. 1. 2. 3.

I have hands. If I have hands, then I am not a brain in a vat. Therefore, I am not a brain in a vat.

2

Some may complain that what was just said assumes that the actual world is a world where I have hands and that a world where I am a brain in a vat is a far out possibility. And this assumption is not one we are entitled to help ourselves to in this sort of discussion. While I have some sympathy for this complaint, I will not press it here. But I will follow the conclusive reasons theorist and adopt that assumption in the discussion that follows.

123

The Argument from Abomination

1187

On Dretske’s view, I know that the premises of this argument are true. And, although he denies various epistemic ‘‘closure’’ principles, Dretske is careful to distinguish those principles from modus ponens which he does not deny (Dretske 2005: 18). I do not either. So I also know that this argument is valid. But, according to Dretske, I do not know that the conclusion is true. This is logical abomination number one: A1: I know that argument is valid and that it has true premises but I don’t know that its conclusion is true In calling this a ‘‘logical abomination’’, I do not mean to imply that it is a formal contradiction.3 I only mean to convey that this proposition is similar to other abominable consequences of the conclusive reasons view and it concerns an elementary logical concept, namely, validity. Of course, we could imagine someone who knows that an argument is valid and knows that its premises are true but fails to know that its conclusion is true because of some interfering factor. Perhaps he never bothers to make the connection or forgets what validity is or he attempts to make the connection but acquires a defeater in the process. But nothing like that is happening here. I know what validity is and thus I know that if an argument is valid and its premises are true, then its conclusion must also be true. And, according to Dretske’s epistemology, I know that this argument is valid and that its premises are true. But I am in no position to know that its conclusion is true. In other words: A2: I know that if an argument is valid and its premises are true then its conclusion must be true as well. And I know that argument is valid and that its premises are true. But I don’t know that its conclusion is true. The abomination here can be highlighted if we introduce the ‘‘for I all I know’’ operator where for all I know, P = df I do not know that P and I do not know that not-P. This gets us: A3: I know that argument is valid and that it has true premises. But, for all I know, its conclusion is false Do I know whether the anti-skeptical argument is sound? To answer this question, we must apply the conclusive reasons test and ask how things would be if it were unsound. If the argument were unsound, then it would either have a false premise or be invalid. Since there is no way for a modus ponens argument to be invalid, if the argument were unsound, it would have to contain a false premise. And the second premise is a necessary truth. So, if the anti-skeptical argument were unsound then I would not have hands. And in that case, my perceptual experiences would not be as they are. In other words, if the anti-skeptical arguments were unsound, I would not possess the reasons I now have for believing that I have hands. This means that my perceptual experience provides conclusive reason to believe that 3

Dretske (2005: 18) seems to be using the phrase ‘‘logical abomination’’ in that way when he distinguishes ‘‘logical’’ from ‘‘conversational’’ abominations. The latter notion will be discussed later on in this paper.

123

1188

M. Veber

the argument is sound and, since I still do not possess a conclusive reason to believe the argument’s conclusion, two more logical abominations result. A4: I know that argument is sound but I don’t know whether the conclusion is true. A5: I know that argument is sound and that arguments, by definition, have true conclusions. But for all I know, that argument has a false conclusion. George Edward is a philosopher who rejects Dretske’s epistemological theory. He is perfectly happy to believe things like 3 on the basis of things like 1 and 2. And that is what he does. Dretske’s epistemology says that although George Edward knows that he reasoned validly from true premises, his belief that he is not a brain in a vat is not an item of knowledge. Now consider what this epistemological theory entails that I do and do not know about George Edward’s inference: A6: I know that George Edward reasoned validly from premises that are true. But, for all I know, he ended up with a false conclusion. A7: I know that George Edward employed a sound argument. And I know that anytime someone employs a sound argument, he ends up with a true conclusion. But I don’t know that he ended up with a true conclusion. Now consider another sort of argument: 1. 2.

I have hands. Therefore, I have hands or I am not a brain in a vat.

For the kinds of reasons already given, Dretske’s view has it that I know that the premise of this argument is true and I know that it is valid. Now, for a disjunction to be false, both disjuncts must be false.4 Therefore, if the conclusion of this argument were false, then both disjuncts would be false. And in that case, I would be a brain in a vat without hands. Thus, I have a conclusive reason to believe the premise, I know that the argument is valid but I do not have a conclusive reason to believe its conclusion.5 This argument generates all of the same logical abominations as the one that preceded it. To appreciate just how abominable this consequence is, it helps to consider the conclusion of this argument in isolation. Here it is: I have hands or I am not a brain in a vat. Dretske’s view entails that I am not in any position to know that this disjunction is true even though I do know that its first disjunct is true and I know that all that it takes for a disjunction to be true is for one of its disjuncts to be true. Thus we have our eighth abomination: 4

An anonymous reviewer points out that some non-classical logics do not require that a false disjunction have all false disjuncts but only no disjuncts that are true. This does not affect the overall point. It is also worth noting that non-classical logic is not taken to be a component of the conclusive reasons view. (As noted above, Dretske emphasizes the distinction between rejecting epistemological closure principles and rejecting principles of classical logic.) For that reason, I shall assume the correctness of classical logic throughout this discussion.

5

Hawthorne (2004: 41) uses a different example to make a similar point.

123

The Argument from Abomination

1189

A8: I know that the first disjunct is true. And I know that’s all it takes for a disjunction to be true. But for all I know, the disjunction is false. As a young logic student, I can remember doing exercises where we determined the truth value of a compound sentence by writing little T’s and F’s over its component sentences and then applying the definitions of the connectives. We would signify truth or falsity for the whole sentence by writing a great big T or F over the main connective. Once you learned the way the connectives worked, there were some easy short cuts. If you wrote a little T over any component sentence in a disjunction you could go ahead and write a big T over the whole thing. Under Dretske’s epistemic theory, this elementary exercise gets a bit more difficult. When I confront the disjunction above, I know I can put a little T over the first disjunct but then I do not know what to do. And notice, it is not because I have forgotten the meaning of the operators or my evidence has significantly changed or anything like that. It is because I know that the first disjunct is true but I am not in any position to know that the disjunction is true. This gives us abomination number nine. A9: I know that a little T goes over the first disjunct and that means a big T goes over the whole thing. But, for all I know a big F goes over the whole thing. Here’s another little logical exercise. All of us remember doing truth tables. Let us interpret the second premise of the anti-skeptical argument as a material conditional. We can then construct the following truth-table.

I have hands

I am not a brain in a vat

I have hands . I’m not a brain in a vat

T

T

T

T

F

F

F

T

T

F

F

T

This table provides a complete account of all the ways the world can be with respect to the truth values of the conditional and each of its component statements. And thus one of the horizontal rows on this table provides the actual truth values for the conditional and its components. Which one is it? According to Dretske, I know that the conditional is true and I know that its antecedent is true. This rules out all rows but the first. Do I know that the first row provides the actual truth values? It is not crystal clear what the conclusive reasons theorist should say. Since we see that ‘‘I am not a brain in a vat’’ gets assigned a T on the first row, it might look like Dretske’s view entails that I do not know that the first row provides the actual truth values. If that is right, then we get the following abomination:

123

1190

M. Veber

A10: I know that the conditional and its antecedent are both true. And I know that if the conditional and its antecedent are both true, then the first row provides the actual truth values for the conditional and each of its component sentences. But, for all I know, the first row doesn’t provide the actual truth values. To see why it is not obvious that the view has this consequence, let us ask whether I have conclusive reason to believe the following: (FRA)

The first row of the truth table provides the actual truth values for the conditional and each of its components.

There is good reason to think that, contra A10, Dretske’s epistemology allows that the same perceptual experiences that enable me to know that I have hands also provide me with a conclusive reason to believe FRA. And thus I can know which row provides the actual truth values. To see why, consider this. If FRA were false, then some row of our truth table other than the first would have to describe the actual world. If the first row did not provide the actual truth values, which one would? There are of course only three possible candidates: the second, the third and the fourth. Since brains in vats lack hands by definition, the second row describes an impossible situation. Thus if the first row did not provide the actual truth values, the second one would not. The third row describes worlds where I do not have hands but I am not a brain in a vat either. The fourth row describes worlds where I do not have hands and I am a brain in a vat. Which sort of world is nearer to the actual one? Given the assumption that brains in vats are far-fetched possibilities,6 a world where the third row provides the correct truth values is much nearer to the actual world than any world where the fourth row does. In other words, if FRA were false and the first row did not provide the actual truth values for the conditional and each of its components, then the third row would. Now, how would things look to me if the third row were actual? Or, equivalently, how would things look to me if it were false that I have hands but still true that I am not a brain in a vat? The conclusive reasons theorist has already provided an answer to this question. If it were false that I have hands (but still true that I am not a brain in a vat), then I would not have the perceptual experiences that I now enjoy. My perceptual experiences provide a conclusive reason for believing FRA because if FRA were false, I would not have hands and thus I would not have these perceptual experiences. Therefore, provided I believe FRA on the basis of my perceptual experiences, I know that the first row on the truth table provides the actual truth values for the conditional and each of its components. But I still lack a conclusive reason to believe that I am not a brain in a vat. This gets us our eleventh logical abomination. A11: I know that the first row provides the actual truth values and I know that the first row says that it’s true that I’m not a brain in a vat. But for all I know, ‘‘I am not a brain in a vat’’ is actually false. 6

See footnote 2.

123

The Argument from Abomination

1191

Now let us move to set theory. Dretske’s epistemology allows that I know that I have hands. In other words, I know that I am not a member of the set of all handless beings. After all, if I were a member of the set of all handless beings, it would not be because some alien mad scientist has made me a brain in a vat. It would be because of my often careless approach to shark handling. Therefore, if I were a member of the set of all handless beings, my perceptual experiences would be much different. And, of course, the set all handless beings includes any and all brains in vats. But, according to this view, I do not know that I am not a brain in a vat. In other words, A12: I know that I am not a member of the set of all handless beings (a set which, by the way, includes all brains in vats). But I don’t know that I am not a brain in a vat. Or, to be a bit more sophisticated, A13: I know that if something is not a member of a given set, then it is not a member of any subsets of that set. And I know that I am not a member of the set of all handless beings. And I know that the set of all brains in vats is a subset of that set. But for all I know, I am a member of the set of all brains in vats. There will likely be many more abominations entailed by this theory of knowledge (more perhaps than are dreamt of in your epistemology). But it seems appropriate to stop at thirteen. At this point, some might wonder what is really new here. Everybody (including Dretske) already knows that, according to this view, I know I have hands but I do not know I am not a brain in a vat. These might be new examples but do they not just make the same point over (and over and over) again? As painful as the more familiar abominations are, the logical ones hit even harder. Epistemology is tough and every view will have likely have some unwelcome consequences especially when it comes time to deal with brains in vats. That said, it is much harder to accept the conclusive reasons view of knowledge once we realize that the howlers do not stop there. This view also has shocking epistemological consequences for propositions involving elementary logical concepts. To allow that you can know that you have hands but are unable to know that you are not a brain in a vat is pretty bad. But to allow that you can know that an argument is sound but you are unable to know that its conclusion is true is downright abominable. Other ways in which these examples strengthen the argument from abomination are revealed when we consider some common replies.

3 Abomination Amelioration: Truth versus Assertibility One way to try and ameliorate the abominable consequences of the conclusive reasons view of knowledge is to emphasize the distinction between truth and assertibility. (Dretske 2005: 18; Adams and Clark 2005) The aim here is to acknowledge the ‘‘absurdity’’ in these sorts of consequences but to explain it away as merely conversational. Sure, these things sound awful but no one ever advised

123

1192

M. Veber

you to go around saying them. If you want to show that the conclusive reasons view is mistaken, you must do more than show that it has consequences that are conversationally absurd, you must show that it has consequences that are false. Since he regards them as true, it is incumbent upon the conclusive reasons theorist to explain why it is absurd to assert the abominable consequences of the view. A corollary of the distinction between truth and assertibility is that between knowing something and claiming to know it. My asserting any of the abominations above involves my claiming to know something. Perhaps any situation in which it is appropriate to assert that I know that I have hands is also one where, even though it is true, it is inappropriate to assert that I do not know that I am not a brain in a vat. The knowledge account of assertion might provide some help here.7 According to this view, it is appropriate to assert only what you know. If that is correct, then one can properly assert that he knows that P only if he knows that he knows that P. And perhaps knowing that one knows that he has hands is possible only if one knows that he is not a brain in a vat. Whatever the details, there are two problems for this way of responding to the argument from abomination. First, it is odd that Dretske or anyone out to defend his epistemic theory would contend that the various abominations it entails are conversational absurdities. When we consider such old favorites as ‘‘I know I have hands but I don’t know I’m not a brain in a vat’’ or ‘‘I know that’s a zebra but I don’t know it’s not a cleverly painted mule’’ we must not forget that Dretske says that kind of thing all the time. He says it at philosophy conferences, dinner parties and in our best philosophical journals and books. And, after forty odd years, he does not seem to be showing any signs of taking it back. If it is absurd to say those things, why does he keep doing it? A second problem is that there are abominable consequences of the conclusive reasons view that do not involve knowledge claims. Consider the following analogue of our fifth abomination: A5’:

Sound arguments, by definition, have true conclusions. And the antiskeptical argument is sound. But, for all I know, it has a false conclusion.

A5’ does not involve a claim to know anything. But it is just as awful or worse. All of our other abominations can be similarly modified. To emphasize this point, we can also run the abominations in dialogue form. A is on the bus and sees that his fellow passenger B has just finished filling out our little truth table. A: Excuse me sir, but I couldn’t help but notice you filling out that truth table. Could you please tell me which row provides the actual truth values? B: Why, the first one of course. A: So that material conditional (which I assume is about you) is true? B: Yeah that’s right. It’s about me and the conditional is true just like the table says. In fact, that conditional is true by definition. A: Great. Now is the antecedent true too? 7

See Williamson (1996).

123

The Argument from Abomination

1193

B: Well duh, that’s what the first row on the truth table says. A: Fantastic. Now I notice that the first row also says that ‘‘I am not brain in a vat’’ is true. Now, is the truth table right about that? B: I don’t know. A: But didn’t you say that the first row provides the actual truth values? And doesn’t the first row say ‘‘I am not a brain in a vat’’ is true? B: You are correct sir. A: But you don’t know whether ‘‘I am not a brain in a vat’’ is true? B: Yeah that’s right. The first row provides all the actual truth values for those sentences. And it says that ‘‘I am not a brain in a vat’’ is true. But, for all I know, ‘‘I am not a brain in a vat’’ is false. Lemme know if I can help you with anything else. Notice that B never makes a claim to know. Perhaps abominations that do not involve a claim to know can be handled by offering even tighter restrictions on assertion. Even though B never claims to know, perhaps his claiming not to know violates some sort of conversational norm. Or perhaps we might contend that B cannot claim anything unless he not only knows it but understands how he knows it.8 It is difficult to tell what exactly such restrictions would look like9 or whether they would cohere with the overall epistemic theory. But I do not wish to complain too much about that because I think these sorts of considerations are beside the point. In fact, I think it is unfortunate that so much of the discussion surrounding the argument from abomination has focused on assertion. The problem raised by the argument, as I see it, has very little to do with that. Let us grant that no one should ever assert any of the abominable consequences of the conclusive reasons view and ask a different question. What should one think or believe in these sorts of situations? Consider for example our troublesome disjunction. Since I know that the first disjunct is true I assume I am well within my rights to think and believe that it is true. And I do. But then what am I to think of the disjunction? Since the view entails that I do not know whether the disjunction is true, I presume that I should not think or believe that it is. I must, I suppose, withhold judgment and think that maybe the disjunction is true and maybe it is false. But how am I supposed to pull this off? The conclusive reasons view seems to demand that I suppress my knowledge of elementary logic and pretend that, even though one of the disjuncts is true, maybe the disjunction isn’t. But I cannot do that. And of course, the logical suppression does not stop there. For instance, I must also think that even though the argument is valid and all of its premises are true, maybe the conclusion is false. I cannot do that either. Can you?

4 Abomination Amelioration, Second Attempt: Information Although it has not taken center stage here, an important part of Dretske’s account of empirical knowledge has always been the idea that our perceptual mechanisms 8

See Kvanvig (2006: 264 ff) for a suggestion along these lines.

9

Hawthorne (2005: 30) also makes this sort of complaint.

123

1194

M. Veber

are signals that carry information. One might appeal to this idea in an effort at abomination amelioration. According to Dretske, a signal can inform me that P but fail to inform me that Q even though I know that P entails Q. He offers the following illustrative example: Tree rings carry information about the age of the tree. Suppose the tree stump has twenty rings. Assuming everything normal, this carries the information that the tree is 20 years old. That the tree is 20 years old implies, logically, that the past (at least 20 years of it) is real. So Russell’s skeptical hypothesis about the world being only a few minutes old must (logically) be false. Do the tree rings carry the information that Russell’s hypothesis is false? No! Why not? Because information … isn’t closed under logical entailment. A can carry information B without carrying information C despite the fact that B implies C. (2006: 411) Dretske goes on to call the reality of the past a ‘‘condition’’ in which things likes trees and newspapers can carry information about the past. Since it is a condition of their carrying information, it is not something about which those things can carry information. Applied to our example, the idea is that my not being a brain in a vat is a condition in which my perceptual experiences carry information about the world I inhabit. My perceptual experiences cannot inform me that this condition is met. Even those unwedded to information-theoretic epistemology can probably see where Dretske is coming from here. And to some, this theoretical grounding might help make the more well-known abominable conjunctions seem a good bit less abominable. But when we consider the abominations offered here, a serious problem arises. Reflection on the logical abominations reveals how the account just sketched both presupposes and conflicts with a ‘‘Humean’’ view of deduction. According to this view, deductively valid arguments contain no information in their conclusions that is not already contained in their premises. Because of this, valid deduction never supplies the reasoner with any new information but only unpacks what is already there. On the one hand, this assumption is crucial to Dretske’s explanation for why I cannot know that the past is real by deducing it from what I learn when I count the tree rings. Even if it is granted that the tree rings do not inform me that the past is real, we do not get the result that I cannot come to know that the past is real by deducing it unless we also assume that the process of deduction never supplies new information. If I can gain new information by deduction then, even though a given signal does not supply me with the information that P, I might still come to know that P by deducing it from the information the signal sends. Most will regard that assumption as a fair one to make. It sits well with widely held empiricist scruples: we have no information about the world save what perception supplies and mere ratiocination contributes nothing really new. It also jibes with commonplace views on why deductive arguments can be valid but inductive ones cannot. Inductive arguments ‘‘go beyond’’ what is stated in their premises (they don’t call it ‘‘ampliative’’ for nothing) and deductively valid arguments do not. But reflection on the logical abominations reveals how the conclusive reasons view of knowledge also conflicts with this common conception of deduction.

123

The Argument from Abomination

1195

The conflict arises in the following way. I know that the anti-skeptical argument is deductively valid. With the aid of perception, I am informed that the premises of this argument are true. But, according to this epistemic theory, I am not informed of the truth of its conclusion and I cannot become so informed by making the inference. But if the conclusion of a deductively valid argument contains no information not already in the premises, this should be impossible. I can be informed that the premises of the anti-skeptical argument are true but fail to be informed of the conclusion only if there is information in the conclusion not already in the premises. But if the Humean view is correct, to be informed of the premises of a deductively valid argument is already to be informed of its conclusion. In response to this problem, one might claim that the signal contains the information in question but we are unable to unpack or extract it. (Adams et al. 2011) Applied here, the idea would be that perception of my hands carries the information that I am not a brain in a vat but this is not information that I am in a position to retrieve. This is a way to preserve commitment to the Humean view of deduction and at the same time maintain that I can know the premises of the antiskeptical argument but not its conclusion. Two problems arise for this attempt to resolve the conflict. First, this idea directly contradicts what appears to be an essential component of Dretkse’s account of information, namely, that signals cannot carry the information that the various conditions required for them to be information carriers are met. Second, the response leaves us wondering what it is to ‘‘unpack’’ or ‘‘extract’’ information. Although any of our examples would serve to make the point, let us consider again the troublesome disjunction. I know that its first disjunct is true and I know that having a true disjunct is all it takes to for a disjunction to be true. What is there to unpack? Or consider again our old friend George Edward who, just like me, knows that the antiskeptical argument is valid and that its premises are true. He also makes the inference and believes the conclusion. On the Human view of deduction, when George Edward makes a deductively valid inference like this, he thereby unpacks the information contained in the premises. But then to say that we are unable to unpack the relevant information would be to say that we are unable to make a deductively valid inference. And this, of course, is false. Something else must be meant by ‘‘unpacking information’’. But if that’s not unpacking, then what is? The metaphor of extraction is not helpful here either because, once I know that a disjunct is true, why must I go in and remove anything from anywhere? ‘‘Extracting’’ the information that the disjunction is true would appear to be simply a matter of applying what I know is the correct definition of a simple logical constant. And in the case of George Edward, ‘‘extracting’’ the information that the conclusion is true appears to be simply a matter of making an obviously valid inference from known premises. That’s not like pulling teeth.

5 Conclusion The logical abominations strengthen the argument from abomination in two ways. They provide further examples of the troubling consequences of the conclusive

123

1196

M. Veber

reasons view and they also reveal weaknesses in some common replies to that argument. Acknowledgments This paper benefited from conversation with John Collins, Nick Georgalis, Jay Newhard and Michael Pendlebury. I also thank two anonymous referees for their remarks on earlier versions.

References Adams, F., Barker, J.A., & Figurelli, J. (2011) Towards closure on closure. Synthese, on-line first, http://www.springerlink.com/content/px8611348u3w90t0/. Accessed February 2, 2012. Adams, F., & Clark, M. (2005). Resurrecting the tracking theories. Australasian Journal of Philosophy, 83(2), 207–221. DeRose, K. (1995). Solving the skeptical the problem. Philosophical Review, 104(1), 1–52. Dretske, F. (1971). Conclusive reasons. Australasian Journal of Philosophy, 49(1), 1–22. Dretske, F. (2005). The case against closure. In M. Steup & E. Sosa (Eds.), Contemporary debates in epistemology (pp. 13–25). New York: Blackwell. Dretske, F. (2006). Information and closure. Erkenntnis, 64(3), 409–413. Hawthorne, J. (2004). Knowledge and lotteries. New York: OUP. Hawthorne, J. (2005). The case for closure. In M. Steup & E. Sosa (Eds.), Contemporary debates in epistemology (pp. 26–42). New York: Blackwell. Kvanvig, J. (2006). Closure principles. Philosophy Compass, 1(3), 256–267. Nozick, R. (1981). Philosophical explanations. Cambridge, MA: Harvard University Press. Williamson, T. (1996). Knowing and asserting. The Philosophical Review, 105(4), 489–523.

123