Stare at alternating green horizontal and magenta vertical gratings, keeping your eyes on the center cross. Then look at the black-and-white gratings below. Horizontal bars will appear faintly pinkish; vertical bars faintly greenish. The effect can persist for weeks.
This is not a simulation of a phenomenon occurring elsewhere. The mechanism runs in whoever is watching.
Celeste McCollough reported the effect in 1965. It is an orientation-contingent color aftereffect: perceived color is linked to orientation, not to a spatial location. Conventional color aftereffects — stare at green, see a magenta afterimage — fade in seconds. After 15 minutes of McCollough induction, Jones & Holding (1975) measured the effect still above half-strength 85 days later. Induction duration scales the persistence. Repeated testing partially counterconditions the effect — each test exposure degrades it.
The effect is retinotopic: stored in retinal coordinates, before head-position integration. Tilt your head 90° and the color assignments reverse, because the bars that were horizontal on your retina during induction are now vertical-on-retina. The tag lives in early visual cortex, not in any world-centered representation of orientation.
The leading hypothesis (Stromeyer, 1969) is that the effect reveals a real chromatic calibration system. The eye's lens produces orientation-dependent chromatic aberration: different wavelengths focus at slightly different depths, and the fringing is orientation-specific at edges. V1 neurons are thought to maintain a slow calibration that adds a compensatory color signal based on edge orientation, flattening what should be achromatic boundaries. Inducing the McCollough effect sets a false calibration — orientation mismatched to color — and the test gratings reveal that correction running without the signal it was tuned against. The pink tinge on horizontal bars is the system compensating for a green signal it expects but no longer receives.
A 2008 fMRI adaptation study found two timescales in V1: a fast component (~30s decay constant) and a slow integrator with no measurable decay over the study period. The slow integrator is what makes the effect last weeks. The calibration hypothesis provides a reason: the sources of chromatic aberration — lens aging, adaptation to indoor vs. outdoor light — change on timescales of months, not seconds. A calibration system tuned to correct them would need to integrate over a matching timescale. Fast recalibration would chase noise. The duration of the effect would be, on this account, a signature of the designed bandwidth of the correction system.