Still There

In 1965, Celeste McCollough published a short paper in Science. The setup is simple enough to do yourself. You alternate between two images: a red horizontal grating (black-and-white stripes with a red tint) and a green vertical grating. You do this for about fifteen minutes. Then you look at an ordinary black-and-white grating — no color at all. The horizontal bars look faintly greenish. The vertical bars look faintly pinkish. The opposite of what you looked at.

That's unusual but not shocking. Stare at red long enough and you see green — opponent-channel rebound, the exhausted detectors swinging back to their resting state. Color aftereffects are common. They're also brief. A minute, maybe a few minutes. The system recovers.

Here's what's strange. Jones and Holding ran this experiment in 1975 and tested how long the McCollough effect lasted. Some subjects were tested repeatedly over days and lost the effect within about a week. But they also had a group that was inducted — same fifteen-minute procedure — and then not tested until eighty-five days later. Two and a half months. That group still had it.

Eighty-five days from fifteen minutes of looking at colored stripes. That ratio has no parallel in ordinary sensory adaptation. Photoreceptor fatigue doesn't store anything for weeks. It doesn't store anything for more than a few minutes. Whatever is holding the McCollough effect, it is not what holds a standard afterimage.

MacKay and MacKay added a detail in 1977: eight hours of sleep doesn't reduce the effect. Neither does wearing an eyepatch for twenty-five hours. Whatever it is, it doesn't decay from disuse the way fatigue would. The storage isn't running down. It's just sitting there.

The effect is also specific in a way that makes the ordinary explanation harder to apply. When you look at a standard red afterimage, the green tint follows your gaze — it floats, it appears on whatever surface you look at. The McCollough effect doesn't do that. The tint appears only when horizontal or vertical stripes are present in the scene. Rotate the test grating ninety degrees and the color switches. The effect is locked to orientation, not to a retinal location or a gaze direction. This is why it's called a contingent aftereffect: the color isn't free-floating. It's conditional on a structural feature of what you're seeing.

No photoreceptor mechanism produces orientation-selective effects. That property belongs to cortical neurons — cells in primary visual cortex that fire selectively when bars of a particular angle fall on the retina. So the McCollough effect is happening at a different level than ordinary afterimages. But that still doesn't explain the duration. Orientation-selective neurons in visual cortex also shouldn't stay adapted for months.

One account treats the effect as perceptual learning rather than adaptation — something closer to associative conditioning than sensory fatigue. The visual system pairs a color with an orientation during induction and stores that pairing the way a conditioned response is stored. On this account the McCollough effect is not an afterimage at all. It's learned. The months-long duration, the sleep-insensitivity, the specificity — these would all follow if what's being stored is a trained association rather than a depleted detector.

But this account has a problem. If the visual system is capable of learning a color-orientation pairing from fifteen minutes of exposure, and the learning lasts for months, then we should expect such pairings to accumulate constantly — every colored object against every textured background, throughout every day, for an entire lifetime. Perception would be saturated with them. The world would look wrong all the time. That's not what happens. So either the McCollough effect triggers some special learning mechanism that ordinary experience doesn't reach, or the perceptual learning account is missing something about why this particular stimulus is retained when normal experience isn't.

There is a third possibility, sometimes called the chromatic aberration account. Natural scenes very rarely have color perfectly correlated with orientation — when a red thing is always horizontal and a green thing is always vertical, that's unusual enough that it might be a symptom of a defect in the optical system. The visual system could be treating the induction as evidence of a chromatic aberration in its own lens and compensating accordingly. On this account the McCollough effect is a correction, not a storage — the system is trying to undo what it thinks is a systematic error in its input. That would explain why the effect goes in the opposite direction from what was viewed: not an afterimage but a correction that overshoots.

But this account also doesn't explain the months-long persistence. Why would the visual system stay in a correction mode for three months? If the aberration it was correcting is no longer present, shouldn't it recalibrate?

Fifty-plus years in, the mechanism isn't settled. Vul and colleagues found evidence for two timescales of learning in early visual cortex contributing to the effect — a fast component and a slow one — which suggests something more structured than simple fatigue. But the full picture isn't there.

What I keep returning to is the ratio. Fifteen minutes. Eighty-five days. The visual system encodes something about that brief, artificial, laboratory stimulus and holds it longer than it holds most explicit memories of comparable events. There's no reason you'd remember, months later, that you looked at colored gratings in a lab. But some part of your visual processing does. It remembers it without knowing it. And it remembers it in such a specific way — not "I saw red" but "red was horizontal" — that it tints horizontal lines slightly green when it encounters them.

The question I don't know how to answer: what does the visual system think it's doing?