The Motion Aftereffect: Why Stopped Motion Looks Like It Reverses

The waterfall illusion

Stare at a waterfall for thirty seconds, look at the static rocks beside it, and the rocks appear to creep upward. This is the motion aftereffect (MAE), first written up by Aristotle and formally described by Robert Addams in 1834 after watching the Falls of Foyers in Scotland. The simulation on this page reproduces it directly: a spiral rotates for an adaptation phase, then stops, and the now-static spiral appears to spin the other way for several seconds before settling.

Direction-selective neurons and opponent coding

Area MT/V5 of the visual cortex contains neurons individually tuned to a preferred direction of motion. Motion is not decoded from one neuron's firing rate alone; the brain reads a population code, comparing the output of pools of neurons tuned to opposite directions, roughly "how much more do the rightward cells fire than the leftward cells." At rest, with no motion present, these opponent pools are balanced and the net signal reads zero, which the brain interprets as stationary.

Sustained exposure to motion in one direction fatigues the neurons tuned to that direction — their firing rate for a given stimulus strength drops, a process called neural adaptation or repetition suppression. The opposite-tuned pool, having seen no matching stimulus, keeps its normal sensitivity. When the display then stops moving, the fatigued pool responds more weakly than the un-fatigued opposite pool even though neither is seeing real motion, and the imbalance is read out as movement in the opposite direction. The illusion is not a trick of the eye; it is a direct read-out of adaptation in a population code, which is why it can be measured electrophysiologically in single MT neurons of behaving monkeys.

Why it decays, and why it depends on adaptation time

The strength and duration of the aftereffect both grow with adaptation time, but with strongly diminishing returns — roughly logarithmically, so 60 seconds of adaptation buys much more aftereffect than 5 seconds does, but 600 seconds buys only a little more than 60. The aftereffect itself decays exponentially once the real stimulus stops, typically over several seconds, as the fatigued neurons recover their baseline sensitivity and the opponent balance is restored.

perceived_speed(t) ≈ k · log(t_adapt) · exp(−t / τ)

t_adapt   how long you stared at the moving stimulus (seconds)
t         time since the motion stopped
τ         recovery time constant, typically 5–15 s
k         a gain set by contrast, eccentricity and attention

What it reveals about vision in general

The MAE is one instance of a much broader family of adaptation aftereffects tied to almost every visual dimension that has a dedicated neural population: colour (afterimages), orientation (the tilt aftereffect), size, and even face identity and gender. All share the same signature — sustained exposure biases perception in the opposite direction once the adapting stimulus is removed — and all are used by vision scientists as a non-invasive probe for which neural populations exist and how they are organised, since you can only get an aftereffect out of a dimension the brain actually encodes with a dedicated, adaptable population.

Frequently asked questions

Try it live

Everything above runs in your browser — open Motion Aftereffect and change the parameters while it is running. Nothing is installed, nothing is uploaded, the whole model lives in one tab.