*From Barry Stroud, "The Problem of the External World" in Stroud, The Significance of Philosophical Skepticism.

I. Stroud's Version of Descartes' Argument

Where p = some physical object proposition, Stroud is thinking of Descartes' argument as being something like this:

(1) If I am dreaming p, then I do not know that p.

(2) It is possible that I am dreaming.

=>(3) Therefore, it is possible that I do not know that p. (from (1) and (2))

(4) If it is possible that I do not know that p, then I do not know p.

=>(5) Therefore, I do not know that p. (from (3) and (4)).

Stroud's analysis focuses on premises (1), (2) and (4). He wants to know to what extent Descartes is entitled to hold these. Since the argument (1)-(5) is valid, we will be forced to accept Descartes' skeptical conclusion unless we can find good reason to suppose that at least one of these premises is false.

PREMISE (2): Stroud argues that Descartes is surely correct about premise (2). Stroud's point is an important one here. It is easy to conclude from a reading of Descartes' First Meditation that Descartes' argument presupposes that Descartes knows that he has had dreams in the past and that he knows that these dreams were distinct from the reality he experienced in his waking state. However, if Descartes thought this, then his skeptical argument would be infected with self-defeat. Descartes' skeptical argument cannot rely on any "alleged" knowledge he has acquired from sense experience, such as that he has had dreams in the past. Stroud points out that, despite appearances to the contrary, Descartes' argument does not require any such assumptions. He can simply posit the possibility of dreaming as a conceptual truth. Stroud believes that Descartes' argument cannot be undermined at this juncture.

PREMISE (1): The more controversial, or at least less obvious, premise is (1). Why should dreaming p undermine knowing p? More specifically, why should dreaming p, where p is a physical object proposition, undermine knowing p.

We might dream what is false, and in that case our belief would not be knowledge. But this is not the primary problem. After all, we might dream what is true. Stroud provides several examples that involve a person dreaming p and p happens to be true. The Duke of Devonshire dreams that he is speaking in the House of Lords, but it turns out that he is speaking in the House of Lords. His belief in the dream happens to be true, but Stroud maintains that he would nonetheless not know that he is speaking in the House of Lords. Stroud doesn't say why such true beliefs would fail to constitute knowledge. Presumably knowledge requires something above and beyond true belief, something that is compromised when we dream what turns out to be true.

One possibility here is something that connects the true beliefs we form in our dreams with the external world. In the case of the Duke of Devonshire, there is no causal connection between the fact that he is speaking in the House of Lords and his coming to believe this in his dream. This might lead us to expect that knowledge is a true belief that is caused by some fact about the external world. However, sometimes there is a causal connection between the external world and true beliefs we form in our dreams. A shutter making noise or a light shining on our eyelids may affect the events in our dreams and hence the beliefs we form in our dreams. The banging shutter may cause us to dream about a banging shutter and thus form a true belief that the shutter is banging. So it looks like imposing a causal requirement on knowledge will not be sufficient to explain why dreaming undermines knowledge. Stroud doesn't tell us much about why (1) is still correct. One possibility is that knowledge requires reliability of belief formation. This is distinct from a causal requirement. The external world may cause things in our dreams, but it doesn't seem to do this regularly or in such a way that we consistently form true beliefs. So although we can dream what is true, and sometimes as the result of a causal connection between the external world and the dream world, it certainly looks like dreams are not a reliable source of true beliefs about the physical world. Most of the beliefs we form in our dreams are not true.

Hence, Stroud argues that premise (1) is not open to objection. If Descartes' skeptical argument is flawed there must be something mistaken about premise (4).

Premise (4) is crucial to allow Descartes to move from <it is possible that we do not know that p> to <we do not know that p>. As Stroud explains, according to Descartes knowledge requires that there be no ground for doubt. Hence, the mere possibility that we might be mistaken carries enough force to undermine knowledge. Put otherwise, according to Descartes knowledge requires that we eliminate certain possibilities that count against our beliefs.

Stroud identifies two sorts of possibilities relevant to a proposition that must be eliminated if we know that p. There are some possibilities which if true would entail that our belief that p is false. These are possibilities that are incompatible with p, so if such possibilities were actual (true), then our belief p would have to be false. Call this possibility a rebutting defeater for the belief that p. Stroud sites the example of a person's belief that the object in front of him is a goldfinch. One possibility that is inconsistent with this is that it is a canary. Hence, knowing that the object in front of one is a goldfinch requires eliminating the possibility that it is a canary.

However, the dream argument proposes a possibility, namely that we are dreaming, that is not incompatible with our holding a true belief. (See above). So it represents a different sort of possibility. Compare the example of the Duke of Devonshire. He is dreaming, but this does not entail that all his beliefs in his dreams are false. The possibility that one is dreaming is a possibility that counts against one's belief in a way other than entailing that it is false. Such a possibility removes our reasons for supposing that our belief is true by postulating a possibility, which though consistent with our belief being true, removes the likelihood that the belief is true. Call this an undercutting defeater.

Stroud recognizes that the demand that we rule out all possibilities (rebutting and undercutting defeaters) is somewhat unreasonable. However, he thinks that this intuition is a strong one. He thinks that we normally think of knowledge as something that implies that we have eliminated various possibilities that would count against our belief in a significant way.

It seems, though, that Stroud makes a crucial error at this juncture. He began the discussion by proposing that eliminating certain possibilities was a necessary condition for us to know p. However when he appeals to common intuitions or the way "we" normally think of knowledge, he says, "we know what would be a valid challenge to a claim to know something, and we can recognize the relevance and force of objections made to our claims to know." (p. 17). Later he says that we cannot escape premise (4) "if it is nothing more than an instance of a general procedure we recognize and insist on in making and assessing knowledge-claims" (p. 19). Notice that in both of these quotes Stroud argues that the requirement that we eliminate certain possibilities (such as the possibility that we dreaming) is a condition for "assessing" or "evaluating" knowledge-claims. It seems that Stroud may have made at least one of two possible mistakes here. First, to insist that eliminating certain possibilities is necessary to defend our beliefs or knowledge-claims is distinct from saying that it is necessary for us to know a proposition. What it takes to defend a belief or knowledge-claim may be more than what is required to know the relevant proposition. Secondly, Stroud shifts the focus in his paper from conditions needed to know to conditions needed to assess or make a determination as to whether we know. But the two are not the same. "Knowing p" and "claiming to know p" are conceptually distinct. I can know p even if I don't claim to know p, and I can claim to know p without actually knowing p. "Claiming to know p' also implies that one believes that one knows p, whereas one can know p without believing that one knows p. Hence, it may be that Stroud has confused "knowledge" with "knowledge-claims" because of a more fundamental mistake, a failure to see the distinction between a belief that p and belief about whether one knows p or not. To know p may not require what is necessary to know that I know p.

At the most Stroud has shown that eliminating various possibilities is necessary to defend our beliefs or claims to know, or that this is necessary for us to determine or know that we know. But it does not follow from this that the requirement in question is necessary for us to know. Hence, there appears to be at least one way to avoid Stroud's skeptical conclusion.