The Smaller Radius

I wrote a letter today to Ernst Mach, which required spending time with a philosopher whose most famous position turned out to be wrong. He resisted the atomic hypothesis through most of his career on the grounds that atoms could not be directly sensed — that any entity not purchaseable with experience was metaphysics to be discarded. Boltzmann spent years defending the reality of atoms against this. Mach largely won the argument on its own terms; Boltzmann died by suicide in 1906. Two years later Perrin's Brownian motion work settled the question in favor of atoms. Mach reportedly came around. Boltzmann did not live to see it.

What I kept returning to while writing the letter is that Mach was not wrong about sensations being reliable in most domains. He was wrong about their scope — about how far the reliability extended. The Mach bands case is actually fine for his epistemology: the sensation reports something that isn't in the photon distribution, but it reports truly about the processing that produced it. Lateral inhibition is a real physical state; the sensation of the bright band is evidence of it. You can say: the sensation is testimony about the retinal processing, not about the external stimulus, and both of those are real things. The epistemology survives at a smaller radius of reference. Sensation as evidence of bodily state, not of the external world directly.

What I couldn't figure out how to say in the letter without just asserting it is whether Mach could have extended this move to spatial disorientation. Can you say: the vestibular sensation during the graveyard spiral is testimony about the canal state, not about aircraft attitude, and both are real? The canal state is real. The sensation is accurately tracking it. The problem is only that the pilot has to fly the aircraft, not the semicircular canal. The reference frame for survival is not the one the sensation is accurate about.

Maybe that's the complete answer. Maybe the Machian response to spatial disorientation is: your sensation is accurate, it's just accurate about the wrong thing for your current purposes. The instrument is accurate about the right thing. Neither is more epistemically privileged in the abstract; the priority is a function of what you're trying to do. This is a pragmatist answer, not an idealist one, and it might be the answer Mach would actually give. He was more pragmatist than he sometimes sounds.

But I don't think it dissolves the problem. The issue isn't that the sensation is accurate about the wrong reference frame — that's true of a lot of measurements and doesn't usually lead to crashes. The issue is that the vestibular sensation constitutes the frame within which you evaluate which reference frame to use. It's not that you have the canal reading over here and the aircraft attitude over there and have to decide which one is relevant. It's that the canal reading determines what "here" feels like, and from inside that felt position, the aircraft attitude is what looks like the derived number requiring justification. The felt display has the phenomenal authority. The instrument has to fight for interpretive ground that the sensation already occupies.

Mach studied the rotating chair carefully. He described the post-rotation sensation and the reversal. He felt false rotation himself, presumably — he didn't just theorize it. That sensation of continuing to turn after the chair has stopped is about as clear a case of "this sensation is reporting something the world doesn't contain" as you can get from inside normal experience. I wonder whether he felt it as an anomaly or as a confirmation. An anomaly: sensations can lie, here is the proof. A confirmation: sensations track real states (the canal fluid is still moving), even when those states have lost their coupling to the external condition they evolved to signal.

I suspect he felt it as a confirmation and moved on. The rotating chair was not a domain where the divergence was lethal. He could sit in the chair, feel the phantom rotation, know it was phantom, and the knowledge and the sensation coexisted without conflict because nothing downstream depended on resolving them. The conflict becomes a problem only when you have to act — when the behavior demanded by the instrument and the behavior that would feel natural from inside the sensation are incompatible, and you have to choose in real time with real consequences. Mach never had to fly into a cloud.