David and Jean Blumenfeld, "Can I Know That I Am Not Dreaming?"

Here is David and Jean Blumenfeld's version of the Dreaming Argument:

1. I have had dreams which were experientially indistinguishable from waking experiences. (This loosely corresponds to step 2 in our earlier formulation of the argument, when we were discussing the First Meditation.)

2. So the qualitative character of my experience does not guarantee that I'm now not dreaming.

3. If the qualitative character of my experience does not guarantee that I'm now not dreaming, then I can't know that I'm now not dreaming.

4. So, from 2 and 3, I can't know that I'm now not dreaming. (This is step 3a in our earlier formulation of the argument.)

5. If I can't know that I'm now not dreaming, then I can't know that I'm not always dreaming.

6. So, from 4 and 5, I can't know that I'm not always dreaming. (This is step 3b in our earlier formulation.)

7. If I can't know that I'm not always dreaming, then I can't know to be true any belief about the external world which is based on my experience. (This loosely corresponds to step 4 in our earlier formulation of the argument.)

8. So, from 6 and 7, I can't know to be true any belief about the external world which is based on my experience.

Austin rejects premise 1.

Austin denies that dreams are experientially indistinguishable from waking experiences. He argues for this as follows: we have the expression "a dream-like quality." This expression would have no use unless there was some special quality that dream experiences had and waking experiences lacked. But clearly the expression does have a use. Hence, the qualitative character of dreams must be quite different from that of waking experiences.

The Blumenfelds reply: perhaps some dreams are vague and unclear, and in virtue of this are said to have "a dream-like quality,"whereas other dreams are clear and vivid. All the skeptic needs for his argument is that waking experiences are indistinguishable from dreams of the second sort. He can allow that there are also some dreams of the first sort, and that we use the expression "a dream-like quality" to describe them.

Moore argues that the skeptic is not entitled to premise 1.

Moore says that the skeptic, in offering his skeptical argument, is at least implicitly committing himself to knowing that its premises are true. (Perhaps the skeptic doesn't say that he knows the premises to be true; but he has no business offering the argument if he doesn't know the premises to be true.) And Moore points out that, if the argument were sound, then by its own conclusion, the skeptic would not be in a position to know anything about how dreams compared to waking experiences. In particular, the skeptic would not be in a position to know premise 1:

1. I have had dreams which were experientially indistinguishable from waking experiences.

So the skeptic's argument is self-defeating. If the conclusion were true, then the skeptic wouldn't be in a position to know the argument's premises. (See also p. 7 in the Rosenberg Dialogue.)

The Blumenfelds reply: That's fine. The skeptic can concede that he doesn't know premise 1 to be true. He doesn't really need any premise as strong as premise 1 for his argument to work. He can replace premise 1 with the weaker premise:

1*. It's possible for there to be dreams which are experientially indistinguishable from waking experiences.

and the skeptical argument goes through as before. The skeptic can claim to know that premise 1* is true, because his knowledge of 1* is not based upon perception. It's based upon imagination and reasoning critically about the concept of a dream.

A few classes ago, I mentioned a view called fallibilism. That's the view that you can sometimes know that P on the basis of defeasible evidence, evidence that fails to make you infallible. The fallibilist will reject premise 3 in the skeptic's argument, the premise which says:

3. If the qualitative character of my experience does not guarantee that I'm now not dreaming, then I can't know that I'm now not dreaming.

That premise seems to rely on the following sort of principle:

And if you're a fallibilist about knowledge, then of course you're going to reject this Certainty Principle.

The Blumenfelds concede that the argument as they've presented it does rest on a principle like the Certainty Principle. However, they think that the skeptic doesn't have to appeal to so strong a principle. They think that the skeptic's argument can still go through even with a weaker principle, which even fallibilists ought to accept. So the skeptic doesn't have to rest his case on a strong infallibilist construal of "knowledge." The Blumenfelds think that skepticism poses a problem for you even if you are a fallibilist.

How does the skeptic's argument work if we give up the appeal to the Certainty Principle?

The fallibilist says you can know things on the basis of evidence which merely makes those things very likely, without guaranteeing their truth. According to the Blumenfelds, the skeptic will ask: Well, how do you know it's even very likely that you're not dreaming? You would be having exactly the same experiences if you were dreaming. Those experiences don't merely fail to guarantee that you're awake. The skeptic thinks they don't give you any reason whatsoever to believe you're awake.

If you're a fallibilist, and you want to say that the skeptic is wrong, we do know we're not dreaming, we do know what the external world is like, then you'll have to find some way to resist the move that the Blumenfelds are making here. You have to argue that our experiences do give us reason to believe we're not dreaming, even though a dreaming person might have those very same experiences.

Perhaps the best explanation of your having the experiences that you do is that you're not dreaming but awake.

Sometimes you're justified in believing P because P is part of the best explanation of some observed evidence. For instance, suppose you remember you left the headlights on your car on. Now the headlights are off, and the radio doesn't work. Furthermore, when you turn the key nothing happens. What is the best explanation of all this evidence? The best explanation is that the car's battery has died. There are other possible explanations: for instance, maybe a Good Samaritan broke into the car and turned off the headlights. But then some mischievous kids also broke into your car and disconnected the wires leading to the radio and the starter engine. That is all possible. But the best explanation of what happens seems to be: the lights were on all night, and now the battery is dead. So that's what you seem to be justified in believing.

Perhaps you're justified in believing that there's an external world in something like the same way. Perhaps the hypothesis that there's an external world which you sometimes perceive is part of the best explanation of why you have the experiences that you do.

This is an attractive line of response to the skeptic, and it's one that many contemporary philosophers accept. We'll look again at this line of response later in the course.

The important question a fan of this response has to address is: Why is the hypothesis that we sometimes perceive an external world a better explanation of our experiences than any of the skeptical alternatives?

Slote offers an answer to this question. His answer is that the skeptical hypotheses are "inquiry-limiting" hypotheses, and that insofar as you're an inquirer after truth, it's reasonable to avoid accepting inquiry-limiting hypotheses.

The Blumenfelds reply: Perhaps it is reasonable in some sense to avoid hypotheses that prevent you from engaging in any further inquiry. But how do we know that it's epistemically reasonable? We need to attend to an important distinction.

(i) One kind of reason for having a belief is a practical or instrumental reason. You have this kind of reason when there would be a good "pay-off" to having the belief. For instance, if Tom believes that his wife is being faithful to him, that has a good pay-off, because it makes him happy. Pascal argued that it's rational to believe in God because the expected pay-off of having that belief is so high. William James discusses a case where a climber gets stuck in the Alps and he has to jump across a crevasse. Suppose it's more likely that the climber will make it if he believes that he'll make it; and suppose the climber knows this. Then, James argues, what it's rational for the climber to do is to form the belief that he will succeed. Having this belief will make his life better off.

All of these examples have to do with subject's practical reasons for believing things.

(ii) The other kind of reason for believing something is what we'll call an epistemic reason. This consists of things like good evidence, evidence that makes your belief likely to be true. Pascal's and James' arguments do not show that we have good epistemic reasons to believe in God, or that the climber has good epistemic reasons to believe he can successfully jump the crevasse. Their arguments don't do anything to show that those beliefs are likely to be true. (Although the climber is more likely to jump the crevasse if he believes he can, that does not mean that he is very likely to do so. Perhaps believing that he will succeed only raises his chances from 5% to 10%.)

This distinction is relevant in the present context because it's not clear that Slote has given us any epistemic reason for rejecting the skeptic's hypotheses. It may very well be practically reasonable to avoid the skeptic's hypotheses, given that we have an interest in continued inquiry. But that doesn't show that the skeptic's hypotheses are any worse off epistemically.

On p. 6 of the Rosenberg Dialogue, Gemma says she can know she's not dreaming. She knows this because she remembers waking up, having a bagel for breakfast, and so on.

Many people when they hear this think it's just begging the question. But Gemma can defend herself as follows. Sometimes philosophers have very clever arguments, where the premises all seem to be true, but it's just utterly clear to us that the conclusion is false. In a case like that, we know that it's much more likely that the philosopher has made some mistake than it is that the conclusion is true. For instance, I may argue:

1. If a person promises to do something, then she is obligated to do it.
2. No one is obligated to do things which it's impossible for them to do.

These premises both seem plausible. Now my argument continues:

3. People sometimes promise to do things which it's impossible for them to do.

That also seems to be true. But from iii and i it follows that:

4. Those people are obligated to keep their promises.

And from iii and ii it follows that:

5. Those people are not obligated to keep their promises.

So I triumphantly conclude:

6. Some contradictions are true. There are people who both are and are not obligated to keep their promises.

Now, I don't know about you, but I'm much more suspicious of the premises of this argument than I am ready to accept the conclusion that some contradictions are true. I think it's much more likely that there's a flaw in this reasoning than it is that the conclusion is correct.

So too Gemma can argue that it's much more likely that there's a flaw in the skeptic's reasoning somewhere than it is that his conclusion is correct, that we can't know anything about the external world. (See pp. 16-17 of the Rosenberg dialogue.)

Well, Gemma may be right about that. Perhaps it is likely that the skeptic has made some sort of mistake. But we can't just say that and then go home. For the skeptic's argument looks pretty compelling. It starts from premises that most of us are willing to accept. So we can't just ignore that argument. If there's a mistake in it, it would be good to figure out what the mistake is. It's likely to be a mistake in some pretty fundamental belief of ours, which the skeptic relies on to make his argument seem plausible. When we discover that mistake, it will probably mean revising a bunch of things we've so far just taken for granted.

Williams objects to premise 5.

Williams points out that there's a difference between the following two claims:

i. For each experience, for all I know it is possible that that experience is a dream.

ii. For all I know it is possible that: All of my experiences are dreams.

Compare:

iii. For each child, for all I know it is possible that he or she lives longer than average.

iv. For all I know it is possible that: all the children live longer than average.

Claim (iv) is false, but claim (iii) might nonetheless be true. Claims of form (iii) don't in general entail claims of form (iv). Similarly, Williams points out, all the skeptic has shown at step 4 is something like (i). That doesn't entail anything like (ii).

If Williams is right that the skeptic is not entitled to infer (ii) from (i), then we should reject step 5 in the Blumenfelds' argument:

5. If I can't know that I'm now not dreaming, then I can't know that I'm not always dreaming.

The Blumenfelds reply: Williams is right to say that claims like (i) and (iii) do not in general entail claims like (ii) and (iv). But the skeptic can happily accept this. The skeptic doesn't think that 5 is true because claims of the one form generally follow from claims of the other form. The skeptic thinks that 5 is true because of the specific content of what it says.

Compare the following example. Claims with the form:

a. X was a man.

do not in general entail claims with the form:

b. X must have been a cross-dresser.

But the claim that "If Marilyn Monroe was a man, then Marilyn Monroe must have been a cross-dresser" might for all that be true. It's not true because claims with the form of (a) generally entail claims with the form of (b), but rather because of the specific content of what it says.

Well, why then does the skeptic think that if you can't know you're not now dreaming, then you can't know that you haven't always been dreaming?

The skeptic asks, how could you acquire knowledge that you haven't always been dreaming? There are two possible ways. First, you might acquire this knowledge on the basis of empirical evidence. Second, you might acquire it by some a priori philosophical argument.

The skeptic says that the first route won't work, unless there's some time at which you can rely on your empirical evidence to give you knowledge of what things are like beyond what you're then experiencing. There has to be some time t such that you can know at t that the empirical evidence you have then is not all just the product of a dream. But the skeptic has argued that there can be no such time. If his argument up to step 4 is sound (and if that argument works at any time t), then at no time t can you know that you're not dreaming at t. So there's no time at which you can rely on your empirical evidence. Hence, the skeptic concludes, there's no time at which you can acquire empirical evidence for believing that you haven't always been dreaming.

Perhaps, though, there's some non-empirical, a priori argument that enables you to know that you haven't always been dreaming, that you must have been awake at some point in your life. We'll consider this possibility next.

Austin and Ryle argue that you can know a priori that you haven't always been dreaming, because the supposition that you have always been dreaming is somehow incoherent.

• Austin's argument goes as follows:

  i. It makes sense to talk of deception (or error) only if it's possible to recognize cases of deception (or error).

  ii. It's possible to recognize cases of deception only if there is a background of general non-deception.

  iii. So it makes sense to talk of deception only if some experiences are non-deceptive.

  The Blumenfelds reply: Why should we accept premise (i)? Austin seems to be assuming that the claim "Belief B is erroneous" makes sense only if it's possible to tell whether the claim is true. That's a claim the verificationist would like. The verificationist thinks that if it's not possible to tell whether a certain claim is true, then the claim is meaningless. But if we haven't already been persuaded to accept verificationism, then it's unclear why we should accept Austin's premise (i).

• Ryle argues that in order for counterfeit coins to exist, there must also exist some real coins. He apparently intends this as an analogy for our sense-experiences: in order for some experiences to be dreams, some experiences must be veridical.

  The Blumenfelds reply: First, even if we accepted Ryle's argument, it's unclear whether the argument would establish that any of your experiences are veridical. Perhaps, in order for some experiences to be dreams, somebody somewhere has to have veridical experiences. But that doesn't show that you do.

  Second, it's not at all clear that in order for counterfeit coins to exist, there must also exist some real coins. The Blumenfelds tell the following story:

  Just as the first money is about to be printed, a band of criminals seizes the press and issues its own currency, which is a facsimile of the original design. Later, when the shady origins of the currency are exposed, the community forever drops the institution of money. In such a situation, it would appear that all the money that has ever existed has been counterfeit. (247)