Consider the following case.

bq. S has generally reliable vision, but she is subject to a small but serious deception. When she is on a boat over salt water, she is prone to hallucinate objects in the distance. The hallucinations are quite convincing, and S has often formed false beliefs on this account. She does not know the cause of the hallucinations. In fact, she hasn’t even considered that it may be the salt in the salt water that is responsible for them. That’s too bad, because the salt is the cause of it; her vision, even at large distances, is well above average when she’s over fresh water.

bq. Today she is sailing on Lake Huron (a fresh water lake). She forms a visual representation of land in the distance, about 20 miles ahead. She checks on her map and sees (correctly) that the map does not record any land there. And she knows that this is the kind of thing she’s disposed to hallucinate when over sea water. But she decides to trust her eyes, and forms a firm belief that there is some land ahead of her, and that her map must be mistaken. Both of these beliefs are of course true, since her eyes are reliable in these circumstances, and her eyes are telling her that there is land there.

Let p be the proposition that there is land about 20 miles ahead of S. Consider the following four questions.

(1) Does S see that p?
(2) Does S know that p?
(3) Is p part of S’s evidence?
(4) Can S take p for granted in practical and theoretical deliberation, if this is a question of some importance to her?

My initial reaction is to say “Yes” to (1) and “No” to (4). Both of these seem like fairly secure judgments actually.

Vision, like most senses, is fairly strongly informationally encapsulated. Even if S has reasons to doubt that p, those don’t affect what she sees. Since she’s formed a visual representation that p, and that representation was caused, in a non-deviant way, by p being true, she sees that p. (Is this the way these cases are standardly classified in the perception literature?)

On the other hand, if anything at all turns on the question of whether p is true, she should get more information before proceeding. She knows that her eyes are unreliable in circumstances like these, and she has direct evidence that her eyes are faulty here, namely the conflict with the map. The situation is one that calls out for further investigation, not simply trusting her eyes.

I don’t have immediate judgments about (2) or (3). But I do sort of think that if the answer to (3) were “Yes”, the answer to (4) would be “Yes”. So the answer to (3) must be “No” as well.

Whatever we say about (2), there’s a problem here for the views on knowledge and evidence that Williamson has put forward in recent work. He says that knowledge is the most general factive mental state. Seeing is a factive mental state. So if the answer to (1) is “Yes”, the answer to (2) is “Yes”. He also says that all knowledge is evidence. So if the answer to (2) is “Yes”, the answer to (3) is “Yes”. But that doesn’t seem to be the correct answer.

There’s a further challenge here for a broader Williamsonian view of evidence. Consider a straightforward case where we learn something by visual perception. So I just looked out the window and saw clouds. I now know that its cloudy outside. Is my evidence (a) that there are clouds, or (b) that I see there are clouds? Or perhaps both?

It’s not too hard to be motivated, on ordinary language grounds if nothing else, to think that the answer is (a). But if we agree that S’s evidence does not include p, there is a hard question that needs to be answered. Under what circumstances does seeing that p make it the case that p is part of your evidence? Williamson suggests the answer “All circumstances”, but I don’t think that can be right, because of S’s case. And I’m not sure there’s another answer around.

There’s a related question about philosophical methodology. T considers a case, and judges that q. That’s the right judgment about the case, and T makes it for the right reason. Is her philosophical evidence that q, or that she’s judged that q. Williamson again wants to argue that it is q, not merely the judgment that q. But again we have to ask, under just what circumstances does a judgment that q get to be part of your evidence? I suspect that thinking about cases like S’s will make us think that the answer is not completely obvious. More on this to follow.