Philosophy Compass 2/6 (2007): 781–791, 10.1111/j.1747-9991.2007.00076.
Knowledge and Subjunctive Conditionals
University of Wisconsin – Madison
What relation must hold between a fact p and the corresponding belief that p for the belief to amount to knowledge? Many authors have recently proposed that the relation can be captured by subjunctive conditionals. In this paper I critically evaluate the main proposals along those lines.
Introduction A recent development in epistemology is the claim that there is a constitutive connection between knowledge and the truth of some subjunctive conditional. More precisely, the claim is that for a subject to know a certain proposition, some subjunctive conditional related to that subject and that proposition has to be true. In this paper I present and evaluate the main proposals of this type. Subjunctive Conditionals Subjunctive conditionals are usually marked in English by the presence of verbs in the subjunctive mood, as in the constructions ‘. . . were (not) . . . would (not)’, (‘If it were the case that kangaroos have no tails, then it would be the case that they topple over’) and ‘. . . had (not) . . . would (not)’ (‘If it had not rained, then the yard would have been dry’), and we will represent them as follows: A ! B.1 The groundwork for giving a semantics for that kind of conditionals was laid down by work on possible world semantics for modal logic (see Kripke). The main idea of such a semantics is that a necessity operator ‘ ’ can be defined as a restricted universal quantifier over a domain of possible worlds, as follows:
" is true at a world w if and only if " is true at every world accessible from w,
where different restrictions on the accessibility relation give rise to different modal logics. For instance, if every world in the domain is accessible to every other world, then the resulting logic is the familiar S5. In what follows I will not explicitly mention the restriction to accessible worlds, because it does not matter for our purposes.
© 2007 The Author Journal Compilation © 2007 Blackwell Publishing Ltd
and both of them are subjunctive conditionals. If p were true. but it raises an important issue that we will address. p is true. Condition 3 (which has come to be called the ‘sensitivity’ condition) has been widely discussed in the literature. S would believe that p via M. besides belief and truth Nozick imposes two further conditions on knowledge. 3. with the sensitivity condition. then S wouldn’t believe that p via M. Problems for Sensitivity: Counterexamples and Closure Let us begin. Some
© 2007 The Author Philosophy Compass 2/6 (2007): 781–791. provided that those worlds are different enough from the actual world (for instance ‘If Mike Tyson were to fight David Letterman.
A subjunctive conditional does not have the same truth-conditions as a strict conditional. (178)3
As can be seen. S believes that p via M. 4. if and only if every world where A is true is a world where B is true).x Journal Compilation © 2007 Blackwell Publishing Ltd
.2007. then if p were false then I wouldn’t believe that p via method M? Or are there cases where I know that p via some method M and yet it is not the case that if p were false then I wouldn’t believe that p via that method? Most philosophers (besides Nozick) that have written on the topic believe that condition 3. 2. Is it true that. if I know that p via method M. then we can capture that difference between subjunctive and strict conditionals in this definition:
A ! B is true at a world w if and only if all the worlds that are most similar to w where A is true are worlds where B is true as well. The definition is the following:
A subject S knows that p via method M if and only if: 1. and S were to use M to arrive at a belief whether (or not) p.00076. then.1747-9991. But if we assume that there is an overall similarity ordering of the possible worlds. for a subjunctive conditional can be true even if there are some worlds where the antecedent is true and the consequent false. If p were false. 10.1111/j. is false. even though there surely are possible worlds where Letterman wins). then Mike Tyson would win’ is certainly true. we can get the following definition of a strict conditional:
(A # B) is true at a world w if and only if (A # B) is true at every world (in other words.782 Knowledge and Subjunctive Conditionals
From the definition of ‘ ’ and the usual understanding of the material conditional.2
Nozick’s Account I: Sensitivity The most influential part of Nozick’s book Philosophical Explanations is the one where he gives and defends an interesting definition of propositional knowledge. as it stands.4 Condition 4 has been less widely discussed.
and yet it does amount to knowledge. Skeptical arguments often appeal to ‘skeptical hypotheses’: propositions that claim that most of my beliefs are undetectably false. it might happen that the proposition that is sensitively believed doesn’t satisfy some other condition that the account posits as necessary for knowledge.00076. 10. (Sosa. and know. on Nozick’s account knowledge does fail to be closed under competent deduction. and I would still believe that it is in the basement. but rather to refine it. Another reason for dissatisfaction with the sensitivity requirement is that it doesn’t respect a very plausible closure principle. then I wouldn’t believe that there are). Now. I sensitively believe that there are cookies in the jar (if there were none. This failure of the sensitivity condition of being closed under competent deduction was welcomed by Nozick. I can competently deduce from that proposition that I don’t falsely believe that there are cookies in the jar (without ceasing to sensitively believe that there are cookies in the jar). and this failure can be traced back to the sensitivity condition. Nevertheless. and that the correct reaction to the counterexamples shouldn’t be to abandon the sensitivity condition altogether. Some moments later I believe.5 Notice first that it is not true that if one sensitively believes that p and competently deduces that q from p without ceasing to sensitively believe that p. take the propositions that there are cookies in the jar and that I do not falsely believe that there are cookies in the jar. that the trash bag is in the basement.2007. One such hypothesis is the brain-in-a-vat hypothesis: the proposition that the world is such that I am a brain in a vat being fed experiences as if I were a
© 2007 The Author Philosophy Compass 2/6 (2007): 781–791. One main reason for thinking that the sensitivity condition is false as it stands stems from a family of counterexamples that can be traced back to Vogel. then I would still believe that there are cookies in the jar. however. If the trash bag were not in the basement. and I would still deduce from that proposition that I don’t falsely believe that there are cookies in the jar). To take one extreme example (also due to Vogel). then one sensitively believes that q.1111/j.1747-9991. believe that it is on the right track.6 the fact that the sensitivity condition is not closed under competent deduction doesn’t mean that any account of knowledge that incorporates the sensitivity condition is similarly not closed – in any case where a proposition is sensitively believed but a consequence of it is not.Knowledge and Subjunctive Conditionals 783
philosophers. One of the counterexamples runs as follows:
GARBAGE CHUTE: I throw a trash bag down the garbage chute of my condo. because it afforded him an explanation of both the allure and the ultimate failure of skeptical arguments. as Vogel noticed. And yet I do not sensitively believe that I don’t falsely believe that there are cookies in the jar (if I did falsely believe that there are cookies in the jar. though. ‘Skepticism and Contextualism’ 13)
My belief that the trash bag is in the basement is not sensitive.x Journal Compilation © 2007 Blackwell Publishing Ltd
. that would be because it is stuck somewhere in the chute.
I do not know that I have hands. Therefore. I do not know that biv is false. that belief of mine is sensitive: if I didn’t have hands.x Journal Compilation © 2007 Blackwell Publishing Ltd
. 3. but I would have no way of detecting that they are.2007. If I do not know that biv is false. because if it were true then (given that I would be a brain in a vat being fed experiences as if I were a normal human being in a normal environment) I would still believe that it is false. notice that that conjunction satisfies sensitivity. Say that if S satisfies all the conditions
© 2007 The Author Philosophy Compass 2/6 (2007): 781–791. according to Nozick.7 But although Nozick (as well as Dretske) welcomed the fact that their accounts of knowledge don’t respect a closure principle. premise 2 is a conditional with a true antecedent and a false consequent. Nozick’s account of knowledge licenses particularly egregious failures of closure. For instance. With respect to the problem that sensitivity is not closed under competent deduction.1747-9991. 2. in which case I wouldn’t believe it). the argument’s second premise is false. I do know that I have hands according to Nozick. then most of my beliefs would be false. Even setting aside the question of whether I can know that I have hands even if I don’t know that I am not a biv. Therefore. I do not know that biv is false. because if it were false that would be because I am not writing. some philosophers think that the condition of sensitivity is sufficiently on the right track as to be worthy of refinement – as opposed to rejection.784 Knowledge and Subjunctive Conditionals
normal human being in a normal environment. then I would not believe that I do. as explained above. If biv were true. because my belief that I have hands satisfies all the clauses of his definition of knowledge. knowledge doesn’t distribute over conjunction.00076.1111/j. according to Nozick’s account. and so I fail the sensitivity condition with respect to that proposition. I can never know that I am not a brain in a vat (see Nozick 228). and so it is false. 10.
Armed with his account of knowledge. and that not much more goes on in the world (let’s call this proposition ‘biv’). most philosophers think that this constitutes a very serious problem for those accounts. it is interesting to note that Nozick himself considered (in a few sentences) the possibility of incorporating the sensitivity condition in a recursive account that has the consequence that knowledge is closed under competent deduction. I can know that I am writing and I am not a brain in a vat (in particular. Nozick can explain the argument’s appeal by noting that it is a valid argument (it is an instance of Modus Ponens) and its first premise is true. In particular. Defending Sensitivity Despite the counterexamples and the fact that it violates a very plausible closure principle. One powerful skeptical argument uses that (or similar) skeptical hypothesis in the following way:
1. but. then I do not know that I have hands. On the other hand.
then S wouldn’t believe that p via M.10
This version of the sensitivity condition has the consequence that my belief that I don’t falsely believe that I have hands is sensitive. the garbage chute case: I know that the trash bag is in the basement.1111/j. despite the fact that if it were not in the basement I would still believe that it is. A plausible candidate is the proposition that the trash bag is in the basement itself. and (iii) that the trash bag is not in the basement doesn’t explain why I would falsely believe that q. ¬ p entails some q. DeRose says that some beliefs that are insensitive nevertheless strike us as known when their negations entail something that we take ourselves to know to be false. then S tracks that p. Recall.2007.
Under this recursive account.x Journal Compilation © 2007 Blackwell Publishing Ltd
.9 With respect to the counterexamples to sensitivity. without explaining how we came to falsely believe it (see DeRose 23). the new condition seems to be of no help in dealing with other kinds of counterexamples. or 2. S tracks that p. 10. a proposition is non-inferentially known if and only if it is tracked (and is not inferred from another proposition that is tracked). and I take myself to know that I have hands. That proposition certainly satisfies conditions (i) and (ii). but propositions can also be inferentially known by being competently deduced from propositions that are tracked. and ¬ p doesn’t explain why S would falsely believe that ¬ q if p were false. and q the proposition that I have hands. or 2. For the hypothesis that I falsely believe that I have hands doesn’t explain why I falsely believe that I have hands.11 However. we let p be the proposition that I don’t falsely believe that I have hands. for instance. If we follow DeRose’s suggestion and amend the sensitivity condition accordingly.8.00076. the result is the following:
DeRose-style sensitivity condition: S knows that p via M only if either: 1.1747-9991. In this example.Knowledge and Subjunctive Conditionals 785
of Nozick’s original account with respect to a proposition p. What we need to find if the DeRose-style sensitivity condition is going to fare better with respect to this counterexample is a proposition q such that: (i) the trash bag is not in the basement entails that ¬ q. (ii) I take myself to know that q (and I would continue to believe that q if p were false). S takes himself to know that ¬ q (and S would continue to believe that ¬ q if p were false). Keith DeRose has made (in passing. The new account can then be formulated as follows:
S knows that p if and only if either: 1. and without committing himself to its ultimate adequacy) a proposal that has the potential to handle some of them. we judge that my belief that the trash bag is in the basement is insensitive (as the case is described) because if it were not in the basement
© 2007 The Author Philosophy Compass 2/6 (2007): 781–791. S knows that q and S competently deduces that p from q. But is it the case that if the trash bag were not in the basement that would not explain why I still believe that the trash bag is in the basement? Well. if p were false.
Perhaps the most interesting issue raised by the adherence condition is whether or not it is a trivial condition. Subjunctive conditionals with true antecedent and true consequent (‘true-true conditionals’ in what follows) are hard to evaluate. but it does raise interesting issues.1111/j. Now.2007. So. and then defines subjunctives in terms of neighborhoods. One suggestion starts by defining the p-neighborhood of a world. w¬p is closer to w1 than w2 is to w1. wp is at least as close to w1 as w¬p is to w1. it is natural to conclude that the DeRose-style sensitivity condition fails to account for this kind of cases.1747-9991. Therefore. “If Bush were President?” He is the President!’. Consider the true-true conditional: If Bush were President.13 So there are two related tasks that a friend of the adherence condition must undertake: first.12 Nozick’s Account II: Adherence Remember that Nozick’s account of knowledge requires not only that if the proposition believed were false then the subject wouldn’t believe it (via the method by which he actually believes it). he must provide a semantics that doesn’t make true-true conditionals trivially true. as follows:
For any worlds w1 and w2. if the proposition believed were true (and the subject were to use the method that he actually uses in order to arrive at a belief ) the subject would still believe it. he must explain why it is legitimate to impose a condition that strike us as unassertible. (see Nozick 680–1. and any world is closer to itself than any other world. be stuck in the chute on its way to the basement. it is a subjunctive conditional whose antecedent and consequent are guaranteed to be true by the account of which the condition is a part. Nozick did offer a sketch of a semantics that doesn’t have the consequence that all true-true conditionals are trivially true. 10. p is true in wp and false in w¬p. because in the actual world both the antecedent and the consequent are true. n8)14
© 2007 The Author Philosophy Compass 2/6 (2007): 781–791. any true-true conditional is trivially true. With respect to the second task. and we might well be at a loss to judge its truth-value. unbeknownst to me. under the semantics for subjunctives introduced in section 2. w2 is in the p neighborhood of w1 if and only if p is true in w2 and there are no worlds wp and w¬p such that: 1. he would have invaded Iraq.00076. After all. known as the ‘adherence condition’ hasn’t been as widely discussed as the sensitivity condition. A ! B is true at w if and only if B is true in the A neighborhood of w. second.x Journal Compilation © 2007 Blackwell Publishing Ltd
. 3. This condition. unless more is said about what it takes for a proposition to explain another. but also that. 2. Our first reaction upon hearing such a conditional might well be: ‘Whaddaya mean.786 Knowledge and Subjunctive Conditionals
it would. the closest situation were the trash bag is not in the basement is one that does explain why I would still falsely believe that it is in the basement (because it is a situation where the trash bag miseladingly appears to be in the basement).
you strike the match and it lights.x Journal Compilation © 2007 Blackwell Publishing Ltd
. of course. but in order to know which true-true conditionals are true and which are false we would now need to know. If p is true in w.00076. then w’s p neighborhood will not include w itself. that it obliterates distinctions that we do find). Those definitions do have the consequence that not all true-true conditionals are trivially true. but if p is false in w (and possibly true). But there are worlds in the bird-in-the-yard neighborhood of the actual world where I do not believe that there is a bird in the yard (namely. no amount of formal semantics is going to rescue the adherence condition from this problem. Besides this problem with true-true conditionals.1747-9991.15 But the difficulties with true-true conditionals are not over. suppose that we flip a coin to decide whether you or I will struck this match. the following:
BIRD IN THE YARD: There is a pelican in the yard. for instance. those conditionals are not equivalent to their contrapositives. But this construes the relationship between the formal semantics and our pre-theoretic judgments about the truth-values of conditionals exactly backwards: the closeness relation among worlds in the model should be construed so that it delivers the truth-values that we pre-theoretically judge different conditionals to have. and we are equally at a loss to judge their truthvalues. and we are at a loss to judge their truth-value. and it shouldn’t be the case that the model introduces distinctions where we don’t find any (or.17 We find all conditionals that we know to be true-true equally unassertible.
© 2007 The Author Philosophy Compass 2/6 (2007): 781–791. the adherence condition is also subject to potential counterexamples.2007. the p neighborhood of a world w is the largest stretch of worlds where p is true that is uninterrupted by any world where p is false that is closest to w.Knowledge and Subjunctive Conditionals 787
Roughly speaking. (Sosa’s ‘Replies’ in Greco 280)
It seems clear that in the situation described I know that there is a bird in the yard. those worlds where the pelican is gone but the canary is still there).1111/j. then w will be in the p neighborhood of w.18 Sosa’s Account: Safety Under the standard semantics for subjunctives. Therefore.16 The Nozickian semantics under consideration assures us that not all true-true conditionals will be trivially true. Consider. The coin comes up heads. 10. which are the worlds in the A neighborhood of the actual world. I take a glance at the yard and form the belief that there is a bird in the yard. heads you strike it. tails I strike it. There is also a canary behind the pelican. for any conditional A ! B. Thus. Unless this pre-theoretic reaction of ours is massaged to lead us to accept differences between conditionals that we know to be true-true. I do not satisfy the adherence condition. which is too small for me to see from where I am standing. Remember our first problem with such conditionals: they strike us as unassertible.
perhaps the one sketched in the previous section. on pain of making the safety condition trivial. (Sosa. Because of this. Throughout the worlds that are in the I-believe-that-I-am-not-a-biv neighborhood of the actual world. the closest possible world where I struck the match is a world where it lights. it seems clear that Sosa (like Nozick) cannot accept the standard semantics for subjunctives. that is required for knowledge. I seriously consider disguising myself as Michael. S believes that p on basis e ! p).x Journal Compilation © 2007 Blackwell Publishing Ltd
. But it need not be true that if the match hadn’t lit then I wouldn’t have struck it. If the match hadn’t lit. the safety condition is a crucial ingredient in Sosa’s ‘Moorean’ response to skepticism: it is true that we know that we are not brains in a vat.00076. then that could have been because it was wet (although it actually wasn’t). but there are possible worlds where the match doesn’t light and I strike it that are as close to actuality as are worlds where the match doesn’t light and you strike it. which Sosa calls a ‘safety’ condition. in general. then.788 Knowledge and Subjunctive Conditionals
In this situation. and given that belief and truth are also necessary for knowledge. Despite these advantages of the safety condition over its contrapositive. several counterexamples to safety have appeared in the literature. and whereas it is safety. so he hires Judy to stand at a crossroads and direct people towards the house ( Judy’s job is to tell people that the party is at the house down the left road). so he also tells Judy that if she sees Michael she should tell him the same thing she tells everybody else (that the party is at the house down the left road). that in assessing the safety condition we are to assume some other semantics for subjunctives.1747-9991. then it would have lit.1111/j. it is true that if I had struck the match. and either of us could have struck it. Unbeknownst to me. but we might be fooled into thinking that we don’t know it because that belief of ours is not sensitive. ‘Tracking’)19
Given that the safety condition is offered as a necessary condition on knowledge. Ernest Sosa had proposed that we should replace Nozick’s sensitivity condition with its contrapositive. the two conditions are easily confused (because one is the contrapositive of the other). I am not a brain in a vat. Sosa’s safety condition will always be a true-true conditional.20 Thus. After noticing the failure of subjunctives to contrapose. Let us suppose. and I am invited.2007. but she should immediately phone Andy so that the party can be moved to Adam’s house. 10. The following formulation seems to capture Sosa’s intent:
Safety: S’s belief that p based on e is safe if and only if S would not easily believe that p based on e without it being so that p (in symbols. and not sensitivity. Andy’s house is very difficult to find. In the possible worlds terminology. but at the last
© 2007 The Author Philosophy Compass 2/6 (2007): 781–791. which is down the right road. One of them runs as follows:
Halloween Party: There is a Halloween party at Andy’s house. One advantage of the safety condition over sensitivity is that my belief that I am not a brain in a vat (and. Andy doesn’t want Michael to go to the party. any belief that I am not the victim of undetectable deception) is safe.
Comesaña is an Assistant Professor at the University of Wisconsin – Madison. I have knowledge but my belief is not safe. The idea that Dretske. the standard semantics for subjunctives will have to be replaced. Tim Black. in which case Judy would have lied to me).x Journal Compilation © 2007 Blackwell Publishing Ltd
. and he holds a Ph. a fortiori. Philosophy and Phenomenological Research. Andy Egan. however. no verb in the subjunctive mood). I know that the party is down the left road. Sosa. in Philosophy from Brown University. no A-bomb’. if it is true that S knows that p. he has authored papers in these areas for Philosophical Studies. Nozick. then p would be true. Conclusion Unconstrained. WI 53706. I ask Judy where the party is. 600 North Park Street. Madison. and she tells me that it is down the left road. 5163 Helen C. and others have put forward. as well as Avram Hiller. For instance. is that interesting and illuminating conditions on knowledge can be captured in terms of subjunctive conditionals. and Peter Murphy. Short Biography Juan Comesaña’s research interests are centered in questions about justification and knowledge. and the Encyclopedia of Philosophy. especially my commentators John Greco. such as ‘No Hitler.Knowledge and Subjunctive Conditionals 789
moment I don’t. In particular. and there are powerful counterexamples to every proposed condition. White Hall. and the participants in the Second Online Philosophy Conference <http://experimentalphilosophy. then the following subjunctive is also true: if S knew that p. however. Philosophical Perspectives.2007.
1 As David Lewis noted. Email: jmcomesana@wisc. It is fair to say that these accounts must face important obstacles. University of Wisconsin – Madison. Synthese. Notes
* Correspondence address: Department of Philosophy. When I get to the crossroads. the thesis that there is a connection between knowledge and the truth of some subjunctive conditional about the proposition known is trivially true. (Comesaña 397)21
In the case as described.D. for very helpful comments. USA. and yet I could have easily believed the same thing without its being true (because I could have easily have come disguised as Michael. 10.1111/j.com/2nd_annual_online_philoso/>.1747-9991. there are shortened conditionals that have no verb (and. In other words. and which should
© 2007 The Author Philosophy Compass 2/6 (2007): 781–791. Acknowledgements Many thanks to Carolina Sartorio.00076. typepad.edu.
4 The sensitivity condition. This is true. 9 Inductive knowledge presents a problem for Nozick’s own account.00076. who claims (implausibly) that you can know that the sun will come out tomorrow only if. 17 The traditional semantics also finds truth-values where we don’t (clearly) see any. which behave rather like indicative conditionals (see Lewis 4). if it weren’t the case that the sun will come out tomorrow.790 Knowledge and Subjunctive Conditionals
still receive the same semantic treatment as subjunctive conditionals. ‘Epistemic Operators’. but at least it doesn’t introduce any distinctions that we don’t see.1111/j. but inferred from tracked propositions according to sound inductive cannons. Also. were anticipated by Fred Dretske – see Dretske. then there would have been signs of that (a ‘back-tracking’ counterfactual. as well as the claim (also made by Nozick) that knowledge is not closed under known logical implication. Iatridou argues that what distinguishes a subjunctive conditional is the presence of past morphology in both the antecedent and the consequent clause. 10 The parenthetical qualification is my addition. ‘Conclusive Reasons’. 2 The definition is similar to the one put forward by Stalnaker (and differs from the one put forward by Lewis) in that it incorporates the assumption that. In the text I ignore this complication by assuming that S takes himself to know only propositions that we would take him to know. there is a set of worlds whose members resemble w more than any world not in the set. 5 For a discussion of closure principles in epistemology. The modified version of Nozick’s account under consideration should also deal with inductive knowledge: presumably. 7 The fact that my belief that I have hands is sensitive whereas my belief that biv is false is not is underscored by the fact that subjunctive conditionals are variable strict conditionals and not any kind of fixed strict conditionals. but it just highlights the need to say much more about method-individuation than Nozick ever said. it resembles the one in Lewis) in allowing that set to contain more than one member (that is. there would be trouble’. 11 For a development of the DeRose-style sensitivity condition. conditionals that we know to be true-true conditionals strike us as unassertible and we are at a loss to judge their truth-values.1747-9991. 3 The definition in the text is a slightly revised version of Nozick’s. That is the kind of conditional in play in Nozick’s adherence condition. . see Adams and Clarke. 14 Nozick goes on to consider a further refinement to allow for not-p worlds that are just as close to the world in which the conditional is evaluated as are some of the worlds in the p neighborhood. it means that there is no easy translation from DeRose’s proposal to a condition on knowledge (as opposed to knowledge attribution). where the consequent refers to a time that comes before the time referred to by the antecedent) – see Nozick 222–3. 16 Or at least. 6 See also Warfield. 18 The adherence condition could be defended by claiming that my method of belief formation is different if I see a pelican than if I don’t see anything. 15 A different theory that allows for non-trivially true (and false) true-true conditionals is the one put forward in von Fintel – although only in contexts where other counterfactuals have been previously asserted. I ignore that complication. DeRose’s proposal is that we tend to judge that S knows that p only if either S believes that p sensitively or we take it that there is some q such that . but it seems clear that is needed.2007. it allows for ties).x Journal Compilation © 2007 Blackwell Publishing Ltd
. This avoids the difficulty that S may know anything as long as he takes himself to know the right proposition.
© 2007 The Author Philosophy Compass 2/6 (2007): 781–791. . in that respect. but it differs from it (and. 10. see Black and Murphy. 13 This is the assumption that Lewis calls ‘centering’. for any given world w. At the same time. one would want to allow for knowledge of propositions that are neither tracked nor deduced from propositions tracked. and there are also subjunctive conditionals about the future. see Kvanvig. 8 For a development of this idea in the framework of a probabilistic interpretation of Nozick’s subjunctive conditions. such as ‘If our troops invade Iran next year. 12 For a different defense of the tracking theory from these (and other) counterexamples. however. When we evaluate the truth-conditions of the proposition that if biv were true then I would not believe that it is false we have per force to consider worlds that are much more dissimilar to the actual world than when we evaluate the truthconditions of the proposition that if I didn’t have hands then I wouldn’t believe that I did. see Roush.
Kripke. New York: Oxford UP. Neta. Ed. ‘Resurrecting the Tracking Theories’. which goes against Sosa’s project. T. D. Stalnaker. Linguistic Inquiry 31. Philosophy Compass 1. F. ——. 2002. Competence and Knowledge’. 197–215. Analysis 64 (2004): 35– 41. 1968. L. 1987. Williamson. 264–86. 1973. Clarke. Synthese 146 (2005): 395– 404.
Adams. ‘Closure Principles’. Sosa. The Oxford Handbook of Epistemology. S. Ed. Murphy. Oxford: Oxford UP. R. under the assumption that p is truly believed. ‘Tracking. and M. The Philosophical Review 104. However. ——. S. S. Kvanvig. ‘Tracking. ‘Conclusive Reasons’.1111/j. Ernest Sosa and His Critics. J. Synthese 154 (2007): 53– 71. Luper-Foy. and he cites Sosa approvingly. Australasian Journal of Philosophy 49 (1971): 1–22. Tracking Truth: Knowledge.2007. Iatridou. M.
© 2007 The Author Philosophy Compass 2/6 (2007): 781–791. T. J. Williamson might not be adverse to understanding safety in terms of knowledge. Ed. F. K. counterexamples to safety as a necessary condition on knowledge are also (further) counterexamples to sensitivity. Rohrbaugh. Kenstowicz. Nozick. NJ: Rowman and Littlefield. It is interesting to note that. Roush. ‘A Theory of Conditionals’. Greco. Dretske. Australasian Journal of Philosophy 83. Rescher. Philosophical Issues 10 (2000): 1–18. 20 Another advantage may be that safety respects plausible principles of closure. ‘Epistemic Operators’.x Journal Compilation © 2007 Blackwell Publishing Ltd
. Therefore. 2006. von Fintel. 21 See also Neta and Rohrbaugh.1747-9991.3 (2006): 256– 67. Oxford: Blackwell. ‘Unsafe Knowledge’. T. 2004. 10. Totowa. ‘In Defense of Sensitivity’. Ed. Oxford: Oxford UP. Closure.2 (2000): 231–70. Oxford: Blackwell. Counterfactuals. Philosophical Explanations. ‘Solving the Skeptical Problem’.00076. R. Acta Philosophica Fennica 16 (1963): 83–94. E. R. Studies in Logical Theory.Knowledge and Subjunctive Conditionals 791
19 Williamson also proposes what he calls a ‘safety’ condition on knowledge. Ken Hale: A Life in Language. ‘When Epistemic Closure Does and Does not Fail: A Lesson from the History of Epistemology’. Cambridge. 2001. 1981. Moser. ‘Semantical Considerations on Modal Logic’. J. ‘Counterfactuals in a Dynamic Context’. MA: MIT Press. ed. 98–112. if p is sensitive then it is safe (under the revised semantics offered in the previous section). The Possibility of Knowledge. Journal of Philosophy 67 (1970): 1007–23. P.1 (1995): 1–52. but whether that is so or not depends on details of the individuation of bases of belief formation. 123– 54. Comesaña. Vogel. Knowledge and Its Limits. Warfield.. K. S. ‘Luminosity and the Safety of Knowledge’. and G. and P. Pacific Philosophical Quarterly 85 (2004): 396– 406.2 (2005): 207–21. Evidence. ‘The Grammatical Ingredients of Counterfactuality’. and Science. 2000. J. N. and Inductive Knowledge’. DeRose. Oxford: Blackwell. Lewis. Cambridge. Black. MA: Harvard UP. ‘Skepticism and Contextualism’.