I've seen red. But as I am writing this sentence, I am not seeing red (my eyes are closed). So do I know what it's like to see red?
Let's try "no". Then the knowledge of what it's like to see red is really evanescent: it's only present when actually perceiving red. Moreover, it seems that what's relevant is not just the perceiving of red, but my attending to apparent perception of red. So it seems that I only know what it's like to see red when I attend to an apparent perception of red. But when I do attend to an apparent perception of red, then I surely know what it's like to see red. So I know what it's like to see red when and only when I attend to an apparent perception of red. This now makes me worry that "I know what it's like to see red" is just a more colloquial way of saying "I attend to an apparent perception of red". And if that's so, then the Mary argument for the nonphysicality of qualia is undermined. Furthermore, I think a lot of the intuitive plausibility of the argument comes from imagining oneself in Mary's pre-seeing-red stage, and imaging the kind of curiosity we'd have about what it's like to see red. But if this curiosity is a desire for knowledge that one doesn't have, and if I don't know what it's like to see red, then it's surprising that in Mary's position we'd have curiosity, but when my eyes are closed and I am not seeing red I have little curiosity about what it's like to see red, even though I don't have that knowledge. This suggests, in turn, that the curiosity that we would have in Mary's position isn't a desire for knowledge, but a desire for perception. So, all in all, the "no" answer seems harmful to the Mary argument.
What about "yes"? Intuitively that's the right answer. Surely people know what it's like to have perceptions that they aren't occurrently having. But now it's not clear what my knowledge of what it's like to see red consists in. Does it consist in the fact that even when I'm not looking at anything red, I can bring to mind a memory of seeing red? I'm not very good at it. I search my memory and find a memory of looking at a red object. For a split second, a flash of a red v-shaped piece of tape on a climbing wall shows up in my mind, before disappearing. Is my knowing what it's like to see red constituted by my possession of the skill of producing such evanescent memory images? Then it sounds like know-how rather than the kind of knowledge that's relevant to the Mary argument. And my skills in this direction are quite limited. I've seen very good approximations to circles: for instance, the clocks in the classrooms I teach. But when I bring such seeing-a-circle experiences back to memory, the images are far from being good approximations to circles--instead, I get foggy images of arcs that don't even meet up.
All in all, I am puzzled. I just can't put my finger on what it is that I have when I know what it's like to see red...