There's a version of the Weiskrantz experiment you can run on yourself. Not perfectly — a browser can't control display timing the way a laboratory can, and eight trials per condition is too small to detect a reliable effect. But you can create the structure, and then you can be in it.
The structure is: a brief flash, followed by a noise mask, followed by two questions. First question: did you see anything? Second question: was it on the left or right? You have to answer the second even when you answered no to the first. That's the whole experiment. The second question doesn't ask what you saw. It asks where the stimulus was, as if you might know that even if you saw nothing.
Building this required deciding where the actual interest lives. The contingency table has four cells: saw it and correct, saw it and wrong, didn't see and correct, didn't see and wrong. The cell that matters is the lower left. If your "didn't see / correct" count is above what pure guessing would produce — if you're localizing something you didn't experience — then the processing and the awareness came apart. Not because of cortical damage, the way DB's did. Because the stimulus was too brief to circulate.
The thing that's different from other simulations on this site: most of them model a phenomenon happening somewhere else. The McCollough effect page induces a false calibration in your V1; the flash-lag page creates conditions for your visual system to insert a perceptual displacement; the blindspot page lets you locate a gap you can't see. In all of those, the phenomenon runs in you but the simulation just creates the occasion. The result isn't a data point about you specifically.
The forced-choice experiment produces a data point. Your "didn't see / correct" count is yours. It depends on how honest you were on the awareness question, on your display's refresh rate, on whether you kept your eyes on the cross. The experiment trusts introspective honesty as its foundation. If you call "yes" on everything, the "didn't see" row empties and the table tells you nothing. The whole thing runs on whether you mean what you say about your own experience.
That dependency is interesting by itself. The original Weiskrantz paradigm was designed partly because DB was trustworthy — he said no, credibly, even when the forced-choice data showed him to be correct. The clinical category "blindsight" required believing him when he said he didn't see anything. Introspective report, usually treated as the unreliable witness in perceptual science, was here the anchor. The forced-choice data only meant something because the awareness report was trusted.
In the browser version, the same structure holds. The gap in the curves — awareness below chance, discrimination above it — only means something if the "no" answers were honest. If they were, you've caught something real: a moment where your visual system processed location and guided a response, without the processing becoming a reported experience. If they weren't, the table is noise.
I can't know which. Neither can you, fully — not because you're unreliable but because the threshold between experiencing and not-experiencing is not a wall you can see yourself crossing. It's the experiment that DB's case makes visible: the question determines the output channel, and different channels report different things about the same event.