Squinting and Predictive Processing
[This isn't meant to be particularly interesting per se -- it was mostly me pondering an idea I'd heard and doing what I've got to guess is reconstructing the most obvious guesses about why you might endorse that theoretic structure.]
I think the fact that squinting to see things clearly works is a meaningful test of predictive processing. If you have bad eyesight, you can give yourself a physical signal that your eyesight is obscured, and you'll adjust to mix the bottom-up sensory data and the top-down predictive data differently. With stronger top-down predictive weighting, you might well observe with higher clarity.
To test this, I wanted to see if I could successfully misread things beyond my normal vision without enough cues. Suffice it to say, I was able to use a combination of priming and map information to successfully misread street signs, which I would consider a great success of the theory. To clarify: I was able to blur my own vision intentionally and wander near a street, looking at street signs too far away to read, suspecting that I'd see the street I know I'm near, until a different street with a similar name was incorrectly read, with all the strange certainty of perception, as the one I was anticipating. Very much a "difficulty to tickle yourself" category, so mostly it was by looking for a street I had to turn at, but I did get the feeling twice. It's deeply bizarre, the sense of seeing something, and then seeing it again and being like, oh, I saw it wrong, even if it was only for a split second and from far away. Quite unsettling, and I'm glad it's unfamiliar, but a fruitful confirmation of the general model of predictive processing, I think, and not so exotic an experience that nobody could relate.