Abstract
There is a dissociation between the orientation of contours and their contrast: suddenly changing the orientation of a grating during brief presentations has no effect on its apparent contrast at suprathreshold levels even though its threshold contrast increases (Fiser et al, Vis. Res. 2003). We call this supra-threshold phenomenon contrast conservation. Here we ask whether there is any limit on the differences between stimuli over which contrast is conserved. Two naïve subjects matched the apparent contrast of a succession of 2 images that were spatially coextensive but different in Fourier spectra (a plaid, a natural scene, or a random noise pattern) to a composite formed by superimposing the two images; or they matched a succession of spectrally similar stimuli that differed orientation or spatial position (a ring of 8 evenly spaced Gabor patches whose elements were either rotated by 90 deg in place or moved 22.5 deg along the ring). All stimuli were presented for 53 msec at 25% RMS contrast, 2 deg from fixation. The spectral change had no effect: the curve describing the growth of apparent contrast was the same whether the stimulus was a simple plaid, a natural scene, a random noise pattern, or one of these followed by a different one. The same was true when the orientation but not the position of the ring of Gabor elements was switched. However, when the positions of the Gabor elements changed, but not their orientation, contrast conservation was abolished. As the spectra of a plaid and random noise are as different as possible (maximally concentrated and evenly distributed), we find that the shape of the Fourier spectra of images places no limit on contrast conservation. However, conservation occurs only over a limited area, even for stimuli that are spectrally similar. Natural scenes are no different from either of these artificial stimuli in this respect.
Supported by NIH EY-1319 and the Wellcome Trust