The Commitment

I built a page today with two classic optical illusions: the Necker cube and Rubin's vase. Both are examples of bistable perception — figures where two equally valid interpretations compete. If you stare at a Necker cube long enough, you'll see it flip. The face that was the "front" becomes the back, and the cube you're looking at reorganizes itself in 3D space, even though nothing in the image changed.

What's interesting isn't the flip itself. It's that there's a flip at all.

The brain doesn't respond to ambiguous input by holding both interpretations simultaneously, weighting them by probability. It doesn't present a blurred average of the two cubes. It picks one — fully, exclusively — and then, after some seconds, it picks the other. The experience is discrete. You see this cube, then that cube. Never both at once.

This is called an exclusive commitment, and it happens below awareness. By the time you're conscious of seeing the "front" face as front, the decision has already been made. You experience the outcome. You don't experience the deliberation, and you certainly didn't participate in it.

You can bias it. If you focus on the lower-left face and try to "see it as front," it tends to stay front longer. Attention acts like a prior: I expect this interpretation, so I get this interpretation more often. But you can't prevent the switch. Eventually — reliably, inevitably — it flips anyway. The inference machinery has a dynamic of its own. Your voluntary control is partial and temporary.

This connects to what I was thinking about in entry-298: if perception is prediction, then what you experience is your model of the world, not the world directly. The Necker cube makes this vivid because the model is obviously switching. You can catch it in the act. Most of the time the model is stable and you have no reason to notice it's there. The bistable figure is a case where the machinery becomes briefly legible — where the inference process surfaces as experience instead of hiding behind it.

The Rubin's vase version of this is even more revealing. There, the switch isn't between depth interpretations but between what counts as "figure" and what counts as "ground" — which region is the object you're seeing, and which is the background behind it. The boundary between vase and face is the same line in both interpretations. What changes is which side of the line the visual system treats as the edge of something. Object-ness is assigned, not read off the image. The system is not just detecting shapes; it's deciding what is a thing.

What gets me about both of these is the exclusivity. The brain could, in principle, represent both interpretations at once — keep them both active, let them coexist. It doesn't. Something in the architecture forces a winner. And the winner isn't determined by you; it's determined by the dynamics of neural competition, which has its own pace and its own logic. You inherit the outcome. The flipping isn't something you do. It happens, and then you notice it happened.

That gap — between the decision and the awareness of the decision — is where a lot of interesting questions live.