Abstract
When a transparent filter is moved over a background, observers report a scission into layers, and attribute a color to the overlay which is separate from the colors of the background. Similar percepts occur when filters and materials are simulated on a CRT screen. Is the color attributed to the transparent layer independent of background colors, or is there a systematic relationship? To simulate background materials we used reflectances of 40 everyday objects whose MacLeod-Boynton chromaticities under equal-energy light were distributed uniformly around equal-energy white. On half of the screen, overlapping ellipses simulated either the complete set of materials, or materials restricted to one of the four quadrants around white. The other half consisted of luminance matched achromatic ellipses of uniform reflectance. One of six Kodak CC30 filters (R, G, B, Y, M, C) or a Neutral Density filter was simulated on the colored background as a moving transparent layer. Observers adjusted the spectral transmittance of another filter moving on the achromatic side, so that the two filters looked identical. The spectral transmittance of the matching filter was varied between linear combinations of the six CC30 filters and the Neutral Density filter. Matching results were consistent across 3 observers. Filters overlaid on the complete set of colored objects were matched by filters of broader transmittance spectra, despite the two backgrounds being equated for mean chromaticity and luminance. Background color variation thus reduced the perceived saturation of the transparent layers, akin to chromatic contrast-contrast. Backgrounds consisting of materials from chromaticity quadrants had a biasing effect. Matched transmittances were broadest for filters that shifted the mean chromaticity to the opposite quadrant. The results showed that the inferred color of a transparent layer depends systematically on the background materials, and is not a constant function of transmittance.
NIH EY07556.