Clues to the site of origin of the complementary image

Abstract

PATTERNS of near-parallel black-and-white lines induce an anomalous state of the visual system, in which a superimposed test field of dynamic noise (such as the ‘snowstorm’ on a detuned TV receiver) seems to stream in directions roughly perpendicular to the inducing contours1. The illusory streamers form a ‘complementary image’ (CI) which for the stimulus of Fig. 1a comprises a family of wavy circles, as sketched by J. P. Wilson2 in Fig. 1b. When two orthogonal gratings are superimposed, the streamers lie at 45° to both, as if the resulting bias obeys the rule of vector summation3. J. P. Wilson4 has observed rudimentary signs of directional bias and streaming even at a single black/white contour; and M. E. Wilson5, using only simple pairs of test spots illuminated in sequence, found that the illusory ‘phi’ motion seen between the spots was enhanced for directions orthogonal to the contours of a superimposed grating. When the inducing pattern is presented to one eye and the noise field to the other, however, little or no CI is observed1,6. It has seemed logical1 to attribute these phenomena to some form of adaptive cooperative activity among orientation-sensitive cells of the visual system, leading to a kind of ‘simultaneous contrast’ in the orientation domain, that is, a bias of the signalling system for motion and contour orientation in favour of the orientation(s) not present in the inducing stimulus.