Main

McKay5 reported that viewing a stationary pattern of radial lines causes the illusion of fine grains or wavy lines moving in circles. This has been explained as being due to antagonism between pattern and movement channels within an orientation column6. We have discovered a new visual after-effect by extending this model to explore antagonism between different orientation columns.

With the subject fixating at the centre, a windmill pattern is presented that slowly rotates at 0.2 Hz in one direction for 5 s, and then in the other direction (Fig. 1a). If this is followed by a pattern orthogonal to the windmill, namely concentric rings, diverging and converging for 5 s each at 2 Hz, viewing a blank screen after the concentric rings causes a vivid after-effect of a stationary windmill for a few seconds. This after-effect is quite different from that seen after viewing the concentric rings alone at 2 Hz, if indeed any after-effect is visible. With a cyclical presentation of the sequence shown in Fig. 1a, a striking windmill-like after-effect is always seen after the concentric rings (bottom left blank in Fig. 1a), but no comparable after-effect is visible immediately after the windmill (top right blank in Fig. 1a). We found that the after-effect could be produced if the rings were delayed by up to 30 s, but not after 60 s.

Figure 1: The stimuli used and the after-effect.
figure 1

a, b, Two different sequences of images that were used to test for the after-effect. The stimuli were produced with a VSG series 3 stimulus generator (Cambridge Research System, Rochester, UK). The radial patterns rotate clockwise or anticlockwise and the concentric rings move either inwards or outwards. The grey image is the blank screen of the same average luminance (31.3 cd m−2) as the patterns. The screen was placed at a distance of 85 cm. c, A simulated enlarged image of the windmill after-effect that most observers see (lower left blank in a and b). The contrast approximates to the strength of the percept. The Supplementary Information includes a demonstration of the illusions as in a and b, as well as the script for viewing them on the VSG system.

Full size image

The after-effect in Fig. 1a resembles the windmill pattern and is not related to the spatial frequency of the rings. If the angular frequency of the windmill changes, so does that of the after-effect. An essential requirement for the after-effect we describe is that the windmill pattern should be moving slowly, at 1 Hz or less, indicating that the stimulation of sustained pattern channels7 is crucial to the illusion.

The version shown in Fig. 1b demonstrates the effect more vividly. If the windmill and concentric patterns are both presented at a slow and then at a faster rate (for example, 0.2 Hz and then 2 Hz) with intervening blank screens, and the routine is cycled, after-effects can be seen in each blank period. In the period after the windmill (Fig. 1b, top right), a concentric ring after-effect can be seen; in the period after the concentric rings (bottom left), a windmill after-effect is seen. We presented these patterns to 13 naive subjects, who all reported seeing the illusions. Figure 1c approximates the windmill after-effect that most subjects perceive. It is unlikely that a retinal after-image could produce the illusion because the inducing patterns were not stationary and the after-effects seen were not of the immediately preceding patterns.

Even though the effects could be obtained with simple sine-wave gratings, we recommend radial patterns to avoid retinal after-images from pursuit eye movements. The after-effect is also stronger with radial patterns, which might be related to the meridional bias reported at many levels of the visual system4,8,9.

If, after the windmill pattern, a moving grating is presented instead of concentric rings, the after-effect is seen in an hourglass fashion, with the radial lines only at orientations orthogonal to the grating. This orientation selectivity, together with our finding that the effects are seen better binocularly than monocularly, suggests that a cortical locus is involved.

We propose the occurrence of two neuronal processes to explain the illusion. First, the stimulation of pattern channels not only leads to a decrease in sensitivity for that pattern1, but also causes the deposition of a trace lasting for about half a minute, possibly at the synaptic level, akin to some form of short-term plasticity10. Second, this trace is brought to perceptual threshold by a rebound from inhibition caused by orthogonal orientations.

Georgeson6 explained McKay's rays by proposing that there is mutual antagonism between pattern and movement channels within the same orientation column. We suggest that this antagonism between pattern and movement channels also exists between orthogonal orientation columns.

The phenomenon that we have observed might serve as a strategy of the visual system for preventing the spurious excitation of contour detectors. Visual cortical cells, although best stimulated by bars moving perpendicular to the long axis of their receptive fields, often respond to small stimuli moving along this axis11. During eye movements, texture elements in the scene could therefore inappropriately stimulate cells whose optimum orientation is the same as the direction of movement. We propose that this excitation is prevented by inhibition from cells tuned to orthogonal orientations when the latter are activated by moving stimuli.

The transient existence of a stimulus trace is hard to explain. It might be an inevitable consequence of synaptic processes underlying short-term or long-term potentiation. It might also be a way of rapidly facilitating the perception of pre-existing contours after an eye movement.