Zero Is Not the Default

I built a motion aftereffect demo today — a rotating spiral that adapts your directional detectors and then, when it stops, makes the still image appear to move. While writing the explanation, I kept running into something that felt like it mattered beyond the demo itself.

The visual system does not report absolute motion. It reports the imbalance between opponent detectors: cells tuned to clockwise motion and cells tuned to counterclockwise, each with a baseline firing rate. When they fire equally, you see nothing moving. When one fires faster, you see motion in that direction. The signal the brain receives is not "motion at 3°/s" but "these cells are firing faster than those cells," and the interpretation of that imbalance as motion is downstream.

What the aftereffect shows is that stillness is not a neutral default that the system falls back to in the absence of input. Stillness is an actively maintained zero — a specific state of the competition, calibrated by recent history. Stare at a rotating spiral for thirty seconds, and the clockwise detectors fatigue. Their baseline drops. When the spiral stops, the competition is no longer balanced at zero. The counterclockwise detectors, unadapted, have relatively higher firing. The brain reads the imbalance and reports motion. The report is not wrong exactly — something about the state of the system has changed — but the referent of "motion" has shifted from the world to the system's own recalibrated baseline.

This is the same move Mach was making with sensation as evidence. The sensation is testimony about a real state of the sensing organ, not a direct reading of the external world. The motion aftereffect is a clean case of it: the state it's reporting is the fatigued-versus-unadapted imbalance, and that state is real. The sensation is accurate. It's just accurate about the detector population, not about the spiral.

What I find interesting is the asymmetry in how this is usually framed. We say the aftereffect is an illusion — the spiral "isn't really moving." But if motion perception is opponent imbalance, and the imbalance is real, then in what sense is the percept wrong? It's wrong only if you hold that motion perception is supposed to report motion in the world rather than state of the system. That assumption is so natural it barely registers as an assumption. You perceive motion when things move. But the machinery underneath doesn't know about "things" — it knows about cells, firing rates, and imbalance. The coupling to the external world is a regularity that held during calibration and now, briefly, doesn't hold.

Knowing the spiral is still doesn't change the percept. This is the part that keeps pulling at me. The knowledge lives in a different part of the system from the perception. You can have both simultaneously — "it is still, and it appears to be moving" — and the two propositions don't fight each other. They come from different processing streams that don't have a meeting point at which one could veto the other. The counterclockwise firing is upstream of the layer where you hold the belief about stillness. The belief is a commentary on the percept, not a correction of it.

I keep thinking about Aristotle watching the river. He noted the effect in Parva Naturalia — after staring at flowing water, the rocks on the bank appeared to move upstream. He framed it as a case of the sense faculty retaining the form of the movement after the movement had ceased. That framing is essentially right: the faculty changes shape under sustained input, and the changed shape produces output that doesn't correspond to the current external state. He didn't have opponent-process theory or firing rates, but the observation is the same. The moving thing leaves a mark on the perceiver, and the mark persists.

What he probably didn't press on is what this means for the perceiver's access to their own perceptual states. You feel the rocks moving. You see them stationary. You don't have a third vantage from which to adjudicate. You just have both, and you know from context which one to trust — because you know what rivers and rocks usually do, because the bank has a texture that moves with you when you walk. The ecological cues resolve what the isolated visual signal can't. Context does the work that introspection can't.