One image should show one orientation of grating (here horizontal) with a colored background (red) and the other should show the other orientation of grating (here vertical) with a different, preferably oppositely colored background (green).
The subject should stare approximately at the center of each image, allowing the eyes to move around a little.
It depends on retinal orientation (tilting the head to the side by 45 degrees makes the colors in the above example disappear; tilting the head by 90 degrees makes the colors reappear such that the gravitationally vertical grating now looks green).
A typical test stimulus might show adjacent patches of black-and-white vertical and horizontal gratings (as above).
The effect is strongest, however, when the colors are complementary, such as red and green, or blue and orange.
A related version of the McCollough effect also occurs with a single color and orientation.
This property led to non-redundant effects being reported by people who had used computer monitors with uniformly colored phosphors to do word processing.
People noticed later when reading text of the same spatial frequency, such as in a book, that it looked pink.
A variety of similar aftereffects have been discovered not only between pattern and color contingencies, but between movement/color, spatial frequency/color and other relationships.
[8] Neurophysiological explanations of the effect have variously pointed to the adaptation of cells in the lateral geniculate nucleus designed to correct for chromatic aberration of the eye, to adaptation of cells in the visual cortex jointly responsive to color and orientation (this was McCollough's explanation) such as monocular areas of cortical hypercolumns, to processing within higher centers of the brain (including the frontal lobes[9]), and to learning and memory.
Given that AMEs do transfer interocularly,[8] it is reasonable to suppose that they must occur in higher, binocular regions of the brain.