1. Hyperbolic vs. Realistic Skepticisms

In the first two chapters of his aforementioned book, Sosa makes a couple of attempts to “resolve the problem of dream skepticism” (2007, p. 16). He begins by distinguishing two kinds of skepticism: hyperbolic and realistic. The first type involves conjuring up possible scenarios that are too remote and thus irrelevant. “Possibilities like that of the evil demon or the brain in a vat are said to pose no real threat, being so remote” -and a possibility would be too remote when it might not really or easily happen ^{(Sosa 2007, p. 2)}. On the other hand, some dreaming type scenarios are not too remote and thus not irrelevant to our claims to knowledge. The dreaming possibility is not too remote -it is “too close for comfort”- since dreaming is a daily part of our lives, Sosa says.

However, it is not clear what Sosa means when he suggests that because “dreaming is a daily part of our lives”, while brains in vats are not, the possibility that I am dreaming all the time and having the sorts of beliefs I actually have on a sensory basis but where most of those beliefs are false, is less remote than the possibility that I am a brain in a vat and have the beliefs I actually have on a sensory basis but where most of those beliefs are false. Why would one think that the former is closer than the latter? Could it be that Sosa is thinking that a possibility is too remote when it involves situations with which we are not familiar? Perhaps Sosa thinks that the brain-in-a-vat scenario is not a familiar one because in it my brain is out of my body and in a vat, which is a situation of a type with which I have no familiarity. But this does not seem to be a good reason since the brain-in-a-vat scenario can easily be modified in such a way that my brain is in my body while someone is tampering with it in suitable ways. My brain’s being without a body and in an vat is not an essential part of the type of possibility that is being contemplated. Besides, familiarity differs from person to person. For example, brain surgeons may have familiarity with some of the brain-in-a-vat type scenarios or at least with some aspects of them -which does not mean that the brain-in-a-vat skeptical scenarios are less remote- less threatening to claims to knowledge -for brain surgeons than for the non-medical part of humanity.

At other times Sosa says that the brain-in-a-vat possibility -in contrast with the dreaming possibility- is too remote because “it might not easily happen”. One way in which one could understand the idea of a possible scenario being remote or close involves using the idea of some possible worlds being closer to the actual world than other possible worlds -a closeness understood in terms of some notion of similarity among possible worlds (e.g., ^{Lewis 1983}). In this way, we could say that a possibility is close to the extent to which it resembles the actual world (in interesting or significant respects). However, under this manner of understanding the closeness or remoteness of possible scenarios, it is not clear that brain-in-a-vat type scenarios are all uniformly more remote to the actual world than dreaming type scenarios. Some brain-in-a-vat type scenarios may be closer to the actual world than some dreaming type scenarios. For example, a brain-in-a-vat scenario where my brain is being tampered with by a neuroscientist but where the rest of the world is very much like we believe it to be, including most of its particular facts and all of its natural laws, is closer to the actual world than a dreaming type scenario where I have been dreaming all along but most of the particular facts and the laws in that scenario are different from those in the actual world.

There is another way to understand the idea of a possible scenario being less remote to us than other scenarios; to wit:

A possibility P1 is more remote than possibility P2 when, given all we believe, P1 is less likely than P2.

What we are talking about here is the conditional subjective probability of certain scenarios. In relation to this second criterion of remoteness one could ask: what makes the dreaming scenario more likely (given all I believe on a sensory basis) than the brain-in-a-vat scenario? Sosa does not say much that could be used for an answer to this question -although something can be said; e.g., the brain-in-a-vat scenario is less likely than the dreaming scenario since, given all I believe on an empirical basis, we still do not have either the knowledge or the technical capability to manipulate the human brain in such sophisticated ways as are required for the brain tampering scenario to come true, while we do dream all the time, so that is a possibility within our actual reach, not as unlikely. Maybe this is what Sosa means when he says that the dreaming scenario is less remote since “dreaming is a daily part of our lives” (^{Sosa 2007, p. 2}).

But there are a number of different dreaming skeptical scenarios; there is the one where I have been dreaming all along and where all of my empirical beliefs are false so that the world where I live is radically distinct (even in its laws) from what I believe the world to be like; there is another dreaming scenario where I have been dreaming all along but not all of my empirical beliefs are false, only those that refer to my proximate environment, but where all of the physical laws are such as I believe them to actually be, etc. Indeed, there is a considerable number of dreaming skeptical scenarios, which have different probability assignments relative to the appropriate set of background beliefs. Thus it appears that there is no acceptable criterion of remoteness that Sosa could be using in saying that “the” dreaming scenario -which dreaming scenario?- is less remote than the brain-in-a-vat scenario.

Let us consider again any of the rankings of skeptical possibilities generated by the nearness/remoteness of a possible scenario understood in terms of either (a) the degree of resemblance of that scenario to the actual world, or (b) the subjective conditional probability of that scenario relative to most of our beliefs about the actual world. Any of these rankings are epistemologically irrelevant for the type of skeptic who poses an argument such as the Cartesian dreaming argument: such a skeptic challenges us to give good reasons for holding any and all the sensory-based beliefs we have about the empirical world -and thus leaves us in the meager epistemological position of not being able to appeal to any of those beliefs. In this position, we cannot judge any possible scenario as being closer to what we believe the world to be like, or to the world we happen to live in, than any other scenario. To begin with, we cannot assume we are in a position to tell which one in fact is the actual world.

Furthermore, Sosa would need to use either one of these rankings of possibilities to discard some of the skeptical possible scenarios as epistemologically non-threatening (e.g., the evil demon scenario or the brain-in-a-vat scenario); he says we should not consider them as a threat to knowledge because they are too remote, relative to (most of our sensory based beliefs about) the actual world. But, to this skeptic, all possibilities concerning the properties and characteristics of the empirically accessible part of the “actual” world are on the same epistemological footing. To use some of (our beliefs about) the facts in the world to rank possibilities as more or less remote would be to assume that some of them have a positive epistemological status -something that, once we accept this skeptic’s challenge, we cannot assume. Why should we reject some skeptical possibility simply because it is very unlikely relative to a bunch of beliefs which, as far as the skeptic can tell, may all be false, and for which we may not presume to have any good reasons? After all, we still have not answered the skeptical challenge. And in order to answer this challenge, it is not permissible simply to say that the possibilities to which the skeptic appeals in the formulation of his skeptical argument are “too remote”. We may try to answer him by questioning some of his premises, but that is all.

Notice, then, that the skeptic who offers a rational argument whose conclusion is that we have no good reasons to hold any and all of our sensory-based beliefs -a skeptic for whom the relevant possible scenario would be one where all of my sensory-based beliefs are false (which is, I think, the Cartesian type argument as formulated in the first Meditation)- would get a ranking of a “very remote” scenario. Notice then that some dreaming skepticisms -and not only the brain-in-a-vat or the evil demon types of skepticisms- can be very hyperbolic, if we accept that more or less hyperbolic be understood in terms of the remoteness/closeness of the possible scenarios to which the skepticism in question appeals.

But perhaps Sosa could say that there are certain possibilities that can be properly ignored, and use David Lewis’s rules concerning which possibilities can, and which ones cannot, be properly ignored (^{Lewis 1996}). Some of the rules as to the possibilities that cannot be properly ignored, are:

Additionally, concerning the possibilities that can be properly ignored (that are “too remote”), Lewis proposes some rules, like:

Lewis proposes other rules, but these are enough to appreciate why I think this proposal will not do, if what Sosa is trying to do is to answer the dreaming skeptic who gives us reasons to doubt any and all our sensory-based beliefs. The Rule of Actuality, for example, presupposes that we somehow know (or have at least a good indication as to) which possibility is actual -something that we cannot simply assume before having given a satisfactory answer to the hyperbolic dreaming skeptic- and much the same can be said concerning the Rule of Reliability¹. According to this radical skeptic, we cannot assume, on pain of circularity, that we know which of all possibilities is actual before answering his challenge; at this point and as far as we are concerned, any possibility could be actual for us.

Finally, in his book Putting Skeptics in Their Place: The Nature of Skeptical Arguments and Their Role in Philosophical Inquiry (2000), John Greco argues that a reason can be formulated for why only “close” possible scenarios should be considered as really threatening to our knowledge claims. He says:

In the language of possible worlds, someone has an ability to achieve some result under relevant conditions only if the person is very likely to achieve that result across close possible worlds. But if knowledge essentially involves having cognitive abilities, and if abilities are dispositions to achieve results across close possible worlds, then this explains why possibilities are relevant only when they are true in some close possible world. Specifically, only such possibilities as these can undermine one’s cognitive abilities. (^{Greco 2000, p. 207}; my emphasis)

I personally find this explanation convincing since I share the naturalistic bent that it evinces -that of considering knowledge as crucially involving the use of some of our cognitive abilities, the nature and extent of which is to be discovered by empirical investigation. However, I do not think that the dreaming skeptic will share this -somewhat contentious- starting point. For one, as we already argued, knowing what possible worlds are close to ours -and what our world is like- is a matter of empirical investigation; and the skeptic in question has challenged us to show him that any and all empirical investigations can result in knowledge. What the shape of our cognitive abilities is -and thus which possible scenarios are close to our actual scenario- is a question to which we cannot presume to know the answer, unless either we have decided not to accept the skeptical challenge to begin with, or we have already answered the challenge satisfactorily. If, for good reason, one does not accept the challenge, then the skeptic does not even get to formulate his argument. But it is not clear to me what Sosa’s position is concerning whether or not we should accept the dreaming skeptic’s challenge, and why. If he does not accept it, why not? And the reasons given by Greco appear to me to beg the question against this radical dreaming skeptic. Again, simply saying with Sosa that certain possibilities can properly be ignored (because, in his view, they are “too remote” from our actual situation) begs the question. We are not in a position to say what “our” actual situation is like.

But maybe this radical skepticism is not the type of skepticism Sosa is trying to answer. Perhaps the type of dreaming skeptic that he is attempting to rebut is one that is not trying to make me doubt all of my sensory-based beliefs, but only some, i.e., only those that I am contemplating at this moment, the ones to which I am consciously assenting right now on the basis of (or suitably related to) my present sensory experiences. To this skeptic, Sosa’s position is acceptable; the only skeptical possibilities I need consider in this connection are those that are closest or more probable in view of those sensory-based beliefs of mine that I am not putting in doubt right now. I need only consider the more likely possibilities, like some limited dreaming scenario, which differs from what I take to be the actual world in that I am not awake now but dreaming.

I think this may be a more acceptable way of understanding what Sosa calls a “realistic skepticism” in contradistinction to “hyperbolic skepticism” (^{Sosa 2007, p. 2}). Henceforth, I will use “realistic” vs. “hyperbolic” as follows:

In what follows, we will ask ourselves whether Sosa’s attempts to “resolve the problem of dream skepticism” (^{Sosa 2007, p. 16}) are really successful against either the hyperbolic or the realistic dreaming skeptic. To answer this question, let us consider Sosa’s first of two strategies against the dreaming skeptic.

2. Believe It or Not

In the first chapter of his aforementioned book, Sosa offers an interesting strategy to answer the dreaming skeptic; he argues that dreaming resembles imagining in that, when one imagines, one does not thereby acquire beliefs, but only certain propositionally contentful states that are known as “make-beliefs” (M-beliefs). Similarly, adds Sosa, when dreaming we do have certain mental states that do not constitute beliefs, but something like make-beliefs: states that are propositionally contentful but are not beliefs. Let us call these propositionally contentful states that we have while dreaming, but are not beliefs, “D-beliefs”. These pseudobeliefs are not beliefs, Sosa argues, because they lack some of the requisite properties that real beliefs (R-beliefs) have, in a way similar to that in which M-beliefs lack some of the crucial properties of R-beliefs. What crucial properties do R-beliefs have and D-beliefs lack?

To begin with, one is never praised nor blamed for things that one “does” in one’s dream as a result of one’s D-beliefs and D-intentions -for, e.g., having stolen a thing in one’s dream- just as one is not blamed for the fact that the character one imagined oneself to be in a story stole something in the story. And this is because D-beliefs do not have the requisite causal connections to (actual, real) actions, but only (if at all) to actions one appears to oneself to undertake while dreaming -D-actions, so to speak- that is, to some things that are not actions at all.

Note, however, that Sosa accepts that D-beliefs are conscious episodes of assent to a proposition (^{Sosa 2007, p. 5}) -which does not make them identical to beliefs, since having a belief does not amount to consciously assenting to a proposition; indeed, one can consciously assent to propositions one does not actually believe, and believe a proposition to which one never consciously assents. Thus, it is possible for me to appear in a dream to have real beliefs, when all I am having is D-beliefs. What I do in the dream is to consciously assent to a proposition, a conscious assent that does not amount to a belief, but is something that appears to be a belief. But if so, then I may be misled in my dream into D-believing that I have some actual beliefs -things that very much appear to me to be beliefs.

Sosa argues that there is at least one compelling reason to reject that D-beliefs are R-beliefs: while dreaming we typically (albeit not necessarily) have D-beliefs that directly contradict our R-beliefs -the long standing beliefs we have- so that, accepting that D-beliefs are R-beliefs would lead us to view ourselves as incurring in massive doxastic contradictions every night, that is, as being radically, albeit momentarily, irrational (^{Sosa 2007, p. 5}). But this unsavory conclusion could be avoided, adds Sosa, if we reject the view that D-beliefs are R-beliefs. To go back to the imagining analogy: nobody would think I am being massively irrational if I temporarily make believe that I live in a magical world populated by fairies and ogres, given that I also R-believe that no such world exists -and that’s because M-beliefs are not R-beliefs.

Suppose then that we accept the view that D-beliefs are not real beliefs. How would this constitute an answer to the dreaming skeptic? In his article “Dreams and Philosophy” (2005), Sosa argues that his denial that we have beliefs in dreams is meant to dispose “of the threat posed by dreams for the safety of our beliefs” (2005, p. 12). If S’s belief B that p is safe, according to Sosa, then if B were (really) believed by S in any nearby possible world, then p would likely be true. How would the skeptic be threatening the safety of our sensory beliefs? The skeptic is saying that there is a (nearby) possible world where I really believe I am awake but I am not awake -further, that I cannot know that that world is not the actual world. Sosa counters by saying that no such possible world exists, since in all (nearby) possible worlds, if I really believe anything, then I am awake.

Has Sosa shown now that we cannot be deceived by our dreams? If being deceived meant R-believing something that is false, then since I cannot R-believe anything in my dreams, then in particular I cannot R-believe anything false in my dreams. But why should we agree to understand “being deceived” in this narrow manner? Truth and falsity can be properly applied both to D-beliefs and to R-beliefs; both are mental items having propositional contents. Indeed, Sosa grants that my dreams may lead me to consciously assent to false propositions -they are not R-beliefs, yes, but they are as false as can be! Furthermore, how can I tell whether the act of assent I undertake now to a proposition p is (or is not) a real belief? To me, it would now taste like a real belief, look like a real belief, smell like a real belief. But I cannot on these bases conclude that it is a real belief. We can express the same point using Sosa’s distinction between thin and thick thoughts: “On the thicker notion of thinking, if I imagine that p, hypothesize that p, or dream that p, I do not thereby think that p… On the thinner notion of thinking, by contrast, in imagining that p one does thereby think that p. And the same is now true of dreaming” (^{Sosa 2007, p. 10}).

The thin sense of “thought” (in the widest, almost Cartesian, sense of “thought”) is wide enough to refer to, e.g., D-beliefs as well as R-beliefs. So if I believe (thin sense) that one of my thoughts (thin sense) is a belief (thick sense), I may be wrong. In Sosa’s account, there is a nearby possible world where I think I am awake but I am not awake -because the thought I have is only a thin thought. And it appears I have no reason to think (thin) that this possible world is not the actual world -I have no sensory beliefs to which I can appeal for this purpose. How can I tell whether the thought I am now having is not merely a thin thought but a thick one? I cannot, it appears, unless I can appeal to some of my sensory-based beliefs. And that, ex hypothesi, I cannot do. Can I be deceived by my dreams -i.e., can I be led by my dreams to falsely believe (thin) I am awake? Certainly. Can I be led by my dreams to falsely believe (thin) I am having some beliefs (thick sense)? Certainly. So, why does Sosa say I cannot be deceived by my dreams? I do not see the logic of this.

We can now begin to appreciate that there is another, more radical way in which we can be deceived and which Sosa opens up in distinguishing between having beliefs and its appearing to us (in dreams) that we have beliefs; between R-beliefs and D-beliefs. Like R-beliefs, D-beliefs are truth-evaluable and thus epistemically evaluable, since they are propositionally contentful. They are conscious episodes of assent to propositions. How do I know that all of the mental states, M₁,…, M_n, that up to now I thought were R-beliefs have actually been R-beliefs and not mere D-beliefs? I would have to have good reasons to assent to the following content “M₁,…, M_n have the requisite properties that, in my conceptual book, R-beliefs have”; in other words, I would have to have good reasons to affirm that M₁,…, M_n have the appropriate causal connections with intentions, actions, etc. But this is hopeless: requiring me to have good reasons to hold that M₁,…, M_n have the requisite connections with actual actions and intentions -i.e., with R-actions and R-intentions- would require me to know which of the things I appear to do I actually do, which in turn requires me to appeal to empirical considerations I cannot appeal to in this context. Furthermore, some of my D-beliefs appear to have the appropriate causal connections with some of my D-actions, D-intentions, etc. The skeptic is back but now with a vengeance, with a skepticism that has somehow wormed itself inside our heads.

If all I have been arguing so far is true, then the lesson to learn is that once we accept the stringent conditions of the hyperbolic skeptic’s challenge, we are in deep trouble -and denying we have either beliefs or experiences in our dreams in the way Sosa does is not the answer. Furthermore, we can begin to appreciate why it does not seem to help Sosa to say that the dreaming skeptic he is facing is not the hyperbolic but only the realistic one. It does not help to make his argument successful or even plausible because, in order to appeal to certain beliefs as reasons to hold anything, I need to be able to distinguish those of my mental states that are R-beliefs from those that are not -otherwise, I may be appealing to certain pseudobeliefs (or D-beliefs) as reasons to hold a belief (whether real or not)- and that, Sosa believes, would not do.

But consider the following alternative interpretation of what Sosa is doing in denying that we have real beliefs in dreams: first, as we saw, showing that our perceptual beliefs are safe, and then arguing that all that is left for him to show after this is that it is not irrational to (1) believe ourselves to be awake and at the same time (2) acknowledge that “nothing in the content of our conscious states would seem to reveal the difference” between being awake and asleep, that is, that “I cannot distinguish my state internally from that of a realistic dream” (^{Sosa 2005, p. 13}).

To show that it is not irrational for us to accept those two beliefs, Sosa begins by explaining one sense in which, he claims, we can say that two scenarios are indistinguishable: “Two scenarios are indistinguishable if, and only if, one can tell neither that one is in the first and not the second when that is so, nor that one is in the second and not the first when that is so” (^{Sosa 2005, p. 14}). In other words, scenarios s1 and s2 are indistinguishable for A at t if and only if

But, claims Sosa, in this sense, being awake and dreaming are distinguishable by me since although when I dream I cannot tell I am dreaming and not awake, when I am awake I can tell that I am awake and not dreaming. But what is this telling that, according to Sosa, I can do when awake? He says: “In the dream state we do not affirm anything… whereas in waking life we do knowingly perceive our surroundings” (^{Sosa 2005, p. 15}). Again, this has to do with his claim that we cannot believe, affirm, or tell anything in our dreams, but only when awake. Let us see.

When I am awake, Sosa says, I can tell I am awake and not dreaming -and what he means (what he should mean) is that I can have a real belief that I am awake and not dreaming.² Let us not quibble with him on this point; I can have real beliefs only while awake. But can I conceivably have any reasons for the belief that I am awake? If I could not have any reasons for this belief, then the belief would be irremediably irrational for me. What could these reasons be for me? Since Sosa grants that “nothing in the content of our conscious states would seem to reveal the difference” between being awake and dreaming, what could Sosa’s answer to this question be? Sosa may say that the belief that he is awake now is not irrational for him, since he has many other sensory-based beliefs he is not questioning right now that can be used as reasons for that belief. But even if this strategy works -and it may work- it works only against the realistic skeptic, not against hyperbolic skepticism, as we have been arguing all along. Thus, why is it not irrational for Sosa to assert that he believes he is awake and at the same time admit he does not have reasons of any sort for that belief? Does he have an answer to hyperbolic skepticism?

3. Dreaming Skepticism and Animal Knowledge

Against the dreaming skeptic, Sosa’s second strategy is to construct a virtue epistemology according to which we can be said to know we are awake, in a suitable sense of the term “knowledge” -i.e., animal knowledge- while in a different sense of the same term -reflective knowledge- we do not know any such thing. Sosa’s distinction between animal and reflective knowledge goes as follows:

-where a belief’s being apt is defined using the notion of epistemic virtue:

Furthermore, the intellectual virtues of a subject are those cognitive competences or abilities of his -that may be either innate or learned- that would produce a high ratio of true beliefs (^{BonJour and Sosa 2003, p. 156}).

Sosa’s answer to the “dreaming skeptic” then consists in arguing that our perceptual beliefs can constitute animal knowledge since they are apt beliefs, although they do not constitute reflective knowledge since they are not defensibly apt -i.e., they are not “apt belief aptly noted” (^{Sosa 2007, p. 32}). Thus, any of my true perceptual beliefs constitutes animal knowledge only if the belief’s being true is due to the fact that the belief itself is the result of the exercise of some of my cognitive competences or abilities that are intellectually virtuous -i.e., that the belief is true is importantly explained by the fact that the belief was caused by the exercise of some of my intellectual virtues, e.g., my perceptual capacities.

Does this strategy of Sosa’s work against the hyperbolic dreaming skeptic -i.e., the one that puts in doubt any and all of our sensorybased beliefs? To answer this question, I shall ask which of the considerations that the hyperbolic dreaming skeptic brings into his argument (his “premises”) has to be rejected if we are to accept that sensory-based beliefs constitute animal knowledge. Let us now consider the hyperbolic dreaming argument:

Which of these premises will not work if we accept with Sosa that sensory-based beliefs constitute animal knowledge? For any of my true sensory-based beliefs to constitute animal knowledge, it has to fulfill the condition that its being true is saliently explained by the fact that the belief itself was produced by those of my abilities or capacities (for example, my perceptual capacities) that would tend to produce true beliefs in normal circumstances. More importantly, for the belief that p to constitute animal knowledge, it is not necessary that I (either animally or reflectively) know that it is unlikely that p is false -nor is it necessary that I know that I am awake and not dreaming. Thus, it appears that the premise to be rejected is (2) (and thus (3)). I can (animally) know that p is true even though I do not know that the corresponding belief is not too close to falsehood. Thus, assuming that animal knowledge is knowledge, then it appears that premises (2) and (3) of the hyperbolic skeptic’s dreaming argument are unacceptable -and Sosa’s answer to the hyperbolic dreaming skeptic will ring true.

Something very similar can be said of the argument that a realistic dreaming skeptic can construct in order to give us reasons to doubt the perceptual beliefs we are contemplating at present. Such an argument shares at least premises (2) and (3) of the hyperbolic skeptic’s dreaming argument. Thus, if Sosa’s strategy works against the hyperbolic dreaming skeptic, then it should also work against the realistic dreaming skeptic.

But if all of the above is right, then the dispute between Sosa and the hyperbolic dreaming skeptics will boil down to the issue as to whether (2) has, or does not have, to be a part of any adequate concept of knowledge (and/or of having good reasons). Put in general terms, (2) asserts the following:

(2) S knows that p only if S knows that it is unlikely that p is false.

For example, this principle would say that I do not know that there is a cup in front of me unless I know I am not dreaming, since if I were dreaming then it would be very unlikely that there is a cup in front of me -and my risk of failing in my belief that there is a cup would be too great.

I think we cannot deny that there is at least one concept of knowledge in ordinary use that has this connotation. And according to this concept (granting the other premises of the skeptical argument) we can have no perceptual knowledge. But this goes against common sense; so, if we are to respect common sense -Sosa seems to be suggesting- we need to adopt a different concept of knowledge; one which allows us to say we have perceptual knowledge. Of course, the hyperbolic dreaming skeptic rejects common sense; he is not too impressed by its epistemological credentials -and thus he is in no hurry to adopt a concept of knowledge that does not incorporate (2).

But the skeptic -together with all those who claim that (2) ought to be a part of any concept of knowledge- could argue as follows: if a subject knows that p is true, then she has to know that the probability of p is high, and thus that the probability of not-p is low -and this is just what (2) asserts. To deny (2), would amount to saying that one can know a proposition and yet not know that its falsehood is unlikely; to saying that, for example, it could be that, for all I know, a cup is in front of me even though I do not know it is unlikely that it is not there. Notice that (2) is not equivalent to the principle of closure that says:

(C) If S knows that p and S knows that p implies q, then S knows that q.

Indeed, (2) is weaker than (C): for one to be said to know something, (2) requires only that one knows that its falsehood is unlikely; while (C) requires that one knows all of those things that that thing implies. And the dreaming skeptic only needs (2) (and not (C)), together with a conceptual truth about dreaming, namely:

(D) necessarily, if S is dreaming that p, then it is unlikely that p.

(D) is simply another way of saying that, since dreams do not have a reliable connection with reality (if any), any content that is part of a dream is likely to be false. Are we then to reject (D)? That does not seem like a promising strategy. How about (2)? To deny it, is to accept the following scenarios as possible:

All of these scenarios sound very paradoxical. Their paradoxical air is due to the fact that both the notion of likelihood and that of knowledge are strongly associated to the notion of having good reasons -thus saying that you know something but have no good reasons to believe it is not false sounds almost contradictory.

But perhaps, in adopting the notion of animal knowledge and in claiming that perceptual knowledge is animal knowledge, Sosa is not trying to deny (2) but premise (4), that is, denying that

(4) My knowing that that mental state is not part of a dream cannot appeal to empirical considerations, since in appealing to any empirical consideration I will be assuming that I know that some sensory-based beliefs of mine are true -something which one cannot assume (premise, part of the skeptical challenge).

Under this manner of interpreting Sosa’s strategy, what he is doing is fashioning a concept of knowledge in accordance with which one can say that we have some empirical (animal) knowledge and appeal to other bits of empirical knowledge (animal or not) as grounds. But this strategy -if it were Sosa’s- would amount to rejecting the hyperbolic skeptical challenge; rejecting the challenge of showing the hyperbolic skeptic we have sensory knowledge (and/or good reasons to have sensory-based beliefs) without appealing to any sensory-based beliefs. But as far as I can see, Sosa does not appear to be engaging in any (either good or bad) argument to the effect that the hyperbolic skeptic’s challenge is either unacceptable, or irrational, or unintelligible, or anything of the sort.

Therefore, the dispute between Sosa and the hyperbolic skeptic appears to be at a standoff: Sosa would insist that there is a suitable concept of knowledge that does not incorporate (2) and rescues common sense; the skeptic would insist that our long-standing, traditional concept of knowledge incorporates (2) and opens the door to his argument, and would reject any “artificially fashioned” concept as a mere desperate attempt to avoid his argument.

Are we then to accept that we are reduced to being in this uncomfortable position with regard to hyperbolic dreaming skepticism -having the only consolation that the skeptic is reduced to a similar position?

4. Hyperbolic and Realistic Dreaming Skepticisms

We have tried to argue that there are two types of dreaming skepticism that are epistemologically relevant; that one of them -the realistic one- appeals to a possible scenario that is less remote than the scenario to which the other one appeals, in a plausible sense of remoteness that is characterized in terms of a notion of conditional subjective probability that is relative to the rest of our empirical beliefs about the world. But what we have called “realistic” dreaming skepticism is not a very threatening form of skepticism -it is easily answerable. The reason for this is that the realistic skeptic would have an argument similar to the hyperbolic skeptic’s argument -which we presented above- except that the realistic argument would not appeal to the following premise:

(4) My knowing (my having good reasons to believe) that that mental state is not part of a dream cannot appeal to empirical considerations, since in appealing to any empirical consideration I will be assuming that I know (have good reasons to hold) that some sensory based beliefs are true -something which one cannot assume (premise, part of the skeptical challenge).

In other words, the realistic skeptic -in contradistinction with the hyperbolic skeptic- does not pose a sweeping challenge on the basis of making us doubt any and all of our sensory-based beliefs. But if so, then the conclusion of the hyperbolic skeptic’s argument simply does not follow for the realistic skeptic -it does not follow that we have no reasons of any sort to believe that we are not now dreaming; indeed, we have most of our sensory-based beliefs to support our belief that we are not dreaming at this moment.

So, the really threatening form of dreaming skepticism is the hyperbolic one. The hyperbolic skeptic tries to convince us to accept his challenge of showing him that all our sensory-based beliefs constitute knowledge. But, should we accept his challenge? If we accept it in his terms, then discarding some possible scenarios as epistemologically “non threatening” due to being “too remote” is a forbidden move. In accepting his challenge, we accept putting in doubt all our sensory-based beliefs and thus accept putting ourselves in the position of not being able to use (if only temporarily) any empirical reason to do the discarding. If, on the contrary, we do some discarding of possible scenarios, then we reject (2). But is there any good reason to do so?

Perhaps there is some other way of rationally rejecting the hyperbolic skeptic’s challenge; the idea is that, in order to accept the skeptic’s challenge, he has to give us first a good argument whose conclusion is that we have no good reasons for all those beliefs. Before he has formulated the argument, we as yet have not good reasons to doubt all our sensory-based beliefs; the skeptic has yet to convince us to give up -or at least to not make epistemological use of- all those beliefs. But if so, before the skeptic has finished formulating his argument, we have no good reason to accept (4), i.e., to accept that we cannot appeal to any empirical consideration as a reason to hold any sensory-based belief. And so, there is a premise in the argument we have no good reason to accept at that point, and the skeptic cannot finish formulating his argument.

Of course, this is a dialectical move against the hyperbolic skeptic. But the point is that the skeptic cannot forbid us to use some of our sensory based beliefs as reasons for other such beliefs before he has given us reasons to doubt all of them -and we concede him too much by conceding him (4); that is, by accepting to put all our sensory-based beliefs in an epistemological parenthesis (so to speak) even before he has given us good reasons to do so. I think that this is what is wrong with the hyperbolic skeptic’s argument. The argument is compelling only if we accept in advance to do something that can be done rationally only if the conclusion of the argument is itself acceptable.

But could it not be argued that it is rational to ask ourselves whether we have any reasons to hold all of our beliefs of a certain type -in our case, all our sensory-based beliefs? But what is the rationale for doing that? What is the rationale for asking a question as to our reasons for holding a belief of type S that in advance rules out that other beliefs of type S can function as reasons for that belief? Until the hyperbolic skeptic can answer for us this question, his skeptical argument remains in doubt.

Indeed, I find Michael Williams’ classification of kinds of responses to skepticism illuminating: “[There are those responses that are] direct. They take skeptical arguments at face value, choosing as a point of attack some acknowledged premise or premises. By contrast, diagnostic responses treat skeptical arguments as deeply misleading” (Williams 1999, p. 49).

Furthermore, I agree with a number of philosophers -e.g., ^{Barry Stroud (1984)} and ^{Michael Williams himself (1999)}- that hyperbolic dreaming skepticism appears to be very compelling because it presumably appeals only to the most trivial and mundane of considerations. But this means that any minimally acceptable direct response to skepticism has to rely on assumptions that are at least as trivial and as mundane as the skeptic’s. Sosa’s two responses -particularly the second one- appear to me to score low in this regard.

I also agree with Michael Williams that the best type of response to the skeptic is theoretically diagnostic: “The theoretical diagnostician suspects the skeptic of trading on theoretical commitments that he either does not acknowledge or tries to pass off as commonsense platitudes” (^{Williams 1999, p. 50}). Indeed, in my view, the hyperbolic skeptic assumes the following:

(α) From the point of view of knowledge and/or justification, all our doxastic cognitive states³ are grouped into a number of large mutually exclusive classes, C₁,..,C_n, in such a way that a doxastic cognitive state D of a certain class C_i (i = 1,…, n), if known/justified, will have to be ultimately known/justified in a way such that none of the knowledge/justificatory chains starting with D will end in doxastic cognitive states belonging to C_i

-where we say that a doxastic cognitive state D₁ of S is part of the ultimate justification⁴ of another doxastic cognitive state D₂ of S at t if and only if D₁ and D₂ are justified, and there is a justificatory chain starting with D₂ leading to D₁ at t such that D₁ is not justified by reference to any other doxastic cognitive state of S. This also assumes that if D₁ is an ultimate justifier of D₂, then D₂ cannot be a justifier (ultimate or not) for D₁.⁵

Notice that (α) is not the strongest assertion we can say the skeptic is assuming; indeed, (α) is compatible with a view according to which the ultimate justifiers of any sensory-based belief are a variety of beliefs belonging to widely different epistemological classes. More importantly, (α) is not sufficient by itself to rule out the possibility that there is a system of justified beliefs in which each belief of a certain epistemological class is ultimately justified by beliefs of distinct classes in such a way that the system of beliefs looks, from the point of view of their epistemological connections, more like a web of beliefs than like a “pyramid” -to use one of Sosa’s visual metaphors- and this possibility seems to cancel out the very idea of there being distinct, mutually exclusive epistemological classes; an idea that the hyperbolic dreaming skeptic needs. The hyperbolic dreaming skeptic thus needs a stronger assumption:

(β) From the point of view of knowledge and/or justification, all our doxastic cognitive states are grouped into a number of large mutually exclusive classes, C₁,…,C_n, in such a way that a doxastic cognitive state D of a certain class C_i (i = 1,…, n), if known/justified, will have to be ultimately known/justified in a way such that the knowledge/justificatory chains starting with D will all end in doxastic cognitive states belonging to some other class C_j (j = 1,…, n and j ≠ i) and to no other class.

Now, if you grant me that the hyperbolic skeptic is assuming something like (β), then the shape of the skeptical considerations to which he appeals in his argument will start to make perfect sense, e.g., premise (4) of his argument above, that is, the premise:

(4) But my knowing that that mental state is not part of a dream cannot appeal to empirical considerations, since in appealing to any empirical consideration I will be assuming that I know that some sensory-based beliefs of mine are knowledge -something which one cannot assume (premise, part of the skeptical challenge).

Applied to the dreaming skeptic, the point is that he presupposes that empirical beliefs are of a class, C₁, whose members are ultimately justified by the class of sensory experiences S (or perhaps by the class of immediate perceptual beliefs C₂ understood as distinct from the other empirical beliefs of C₁) in such a way that the ultimate justification of the empirical beliefs of C₁ will have to appeal only to that other class of justifiers -whether it is S or C₂- and not the other way around.

Assumption (β) is, I contend, neither mundane nor uncontroversial: first, it presupposes the idea that all our doxastic cognitive states are cut up into a number of mutually exclusive classes -where the cutting up is motivated already by vague epistemological considerations. For example, the skeptic starts out asking what reasons we can possibly give for all our sensory-based beliefs -i.e., as if somehow they formed an interesting epistemological class. Failure to answer this question by pointing to beliefs of some other unique class, the skeptic contends, would mean that they do not amount to knowledge or are not justified. This is a substantial assumption.

Secondly, there is a strong foundationalist assumption according to which there are ultimate justifiers; i.e., that the justification of any doxastic cognitive state has to end with a number of other such states whose justification does not appeal to any doxastic cognitive state.

Thirdly, (β) also presupposes what Laurence BonJour calls “a linear conception of justification” according to which, a belief can be justified only by other individual beliefs, and not by the holistic properties of the system of beliefs to which the former belief belongs -e.g., coherence (^{BonJour 1976}). This is implicitly assumed by the idea that an ultimate justifier cannot -on pain of vicious circularity- be justified by the doxastic states it ultimately justifies.

All these presuppositions are, I contend, far from trivial or uncontroversial and are thus in urgent need of argument. At this point, we can take two distinct routes in our dialogue with the skeptic: (a) either we stop here and consider ourselves satisfied since we have already shown that the skeptic is -to use Williams’ phrase- treading on assumptions that are far from mundane and trivial; or (b) we do not stop here since we are still not satisfied insofar as we have not yet articulated an interesting view concerning what knowledge and/or justification are that does not presuppose (β).

Concerning alternative (a): what we have done to skepticism is to make it lose its pre-philosophical appeal, the irresistible charm it appears to have to most people, philosophers and laypeople alike. However, I suspect that the philosopher would not be satisfied were we to stop at this point. As for (b), more needs to be said concerning which views are compatible with the denial of (β) above. There are a number of ways of denying (β) since, as we saw, (β) itself embodies a number of different assumptions. Adopting a type of epistemological contextualism -of the sort ^{Williams (1999)} defends- is one such view -but not the only view. The discussion concerning which of all the epistemological views that reject (β) is the best cannot be undertaken here for obvious limitations of space.

5. Conclusions

I have argued that Sosa’s first attempt to respond to the dreaming skeptic by arguing that we have no beliefs in dreams, does not appear to be successful against (what we have called) either the hyperbolic or the realistic dreaming skeptic -that, furthermore, it opens the door to more radical forms of skepticism. I also argued that Sosa’s second attempt is successful only if one is willing to make some changes to our concept of knowledge that appear to be somewhat radical. However, I did not argue that the changes in question are not theoretically useful or that they are unacceptable -indeed, there may be other good reasons to accept these changes. All I claim is that they may be unnecessary if all that one aims to do with them is to avoid hyperbolic dreaming skepticism. I argued as well that, contrary to what many a philosopher thinks, the hyperbolic skeptic’s argument does not require us to assume the principle of closure, but a weaker principle, i.e., that one knows (has reasons to believe) that p only if one knows (has reasons to believe) that the falsehood of p is unlikely.

Finally, I sketched some considerations why I think that the hyperbolic skeptic’s dreaming argument is not acceptable and why, as far as I know, there is no reason to accept his challenge.