Initial distinctness and subsequent fading of minimally distinct borders*

Non-cardinal color mechanism elicitation by stimulus shape: Bringing the S versus L+M color plane to the table

Neurons in the cortex typically respond best to elongated stimuli, or gratings, whereas neurons in the lateral geniculate nucleus (LGN) typically prefer circular stimuli, or spots. Further, neural mechanisms specifically tuned for non-cardinal colors largely do not emerge until the cortex; therefore, the use of gratings should better reveal non-cardinal color mechanisms. This hypothesis has been tested in the isoluminant color plane in macaque monkeys (Stoughton, Lafer-Sousa, Gagin, & Conway, 2012) and in the L–M versus L+M color plane in human subjects (Gegenfurtner & Kiper, 1992). Here, this hypothesis was tested in the third color plane, S versus L+M, in human subjects in two experiments. Experiment 1 tested 10 subjects across four directions in this color plane; Experiment 2 tested three subjects in eight to twelve color directions. Consistent with data from the other two color planes, in both experiments in the S versus L+M color plane, gratings revealed the presence of non-cardinal mechanisms more strongly than did spots.

A role for cortical crosstalk in the binding problem: stimulus-driven correlations that link color, form, and motion

The putative independence of cortical mechanisms for color, form, and motion raises the binding problem-how is neural activity coordinated to create unified and correctly segmented percepts? Binding could be guided by stimulus-driven correlations between mechanisms, but the nature of these correlations is largely unexplored and no one has (intentionally) studied effects on binding if this joint information is compromised. Here, we develop a theoretical framework which: (1) describes crosstalk-generated correlations between cortical mechanisms for color, achromatic form, and motion, which arise from retinogeniculate encoding; (2) shows how these correlations can facilitate synchronization, segmentation, and binding; (3) provides a basis for understanding perceptual oddities and binding failures that occur for equiluminant and stabilized images. These ideas can be tested by measuring both perceptual events and neural activity while achromatic border contrast or stabilized image velocity is manipulated.

What do catastrophic visual binding failures look like?

Ordinary vision is considered a binding success: all the pieces and aspects of an image are bound together, despite being processed by many different neurons in several different cortical areas. How this is accomplished is a key problem in visual neuroscience. The study of visual binding might be facilitated if we had ways to induce binding failures. A particularly interesting failure would involve a loss of the physical integrity of the image. Here, we identify conditions that induce such perceptual failures (e.g. the melting together of equiluminant colored images and the fragmentation of retinally stabilized images) and we suggest that these should studied using electrophysiological measures of binding.

Perception of forbidden colors in retinally stabilized equiluminant images: an indication of softwired cortical color opponency?

In color theory and perceptual practice, two color naming combinations are forbidden-reddish greens and bluish yellows-however, when multicolored images are stabilized on the retina, their borders fade and filling-in mechanisms can create forbidden colors. The sole report of such events found that only some observers saw forbidden colors, while others saw illusory multicolored patterns. We found that when colors were equiluminant, subjects saw reddish greens, bluish yellows, or a multistable spatial color exchange (an entirely novel perceptual phenomena); when the colors were nonequiluminant, subjects saw spurious pattern formation. To make sense of color opponency violations, we created a soft-wired model of cortical color opponency (based on winner-take-all competition) whose opponency can be disabled.

Visibility of borders: separate and combined effects of color differences, luminance contrast, and luminance level

The strength of the border between two regions was assessed by measuring the tendency of the border to disappear during a 5-min fixation period. We measured the total time the border was judged visible when subjects viewed two hemifields with different chromaticities (six wavelength pairs), five luminance contrasts, or with differences in both, at five levels of luminance. For a homochromatic field, border visibility increased linearly with the logarithm of the luminance contrast, regardless of the particular wavelength of the field. For fields with only chromatic differences, border visibility increased linearly with the logarithm of the tritanopic-purity difference (the relative activity of only the long- and middle-wavelength-sensitive cones). For all borders, visibility increased with an increase in luminance level. Thus borders formed by the combination of chromatic differences and luminance contrasts are more visible than borders that have the same chromatic difference or the same luminance contrast alone. For any chromatic difference, calculations of a luminance contrast that would yield the same border visibility were made. These equivalent luminance-contrast functions for border visibility provide a metric for chromatic differences across a range of luminances that allows chromatic and luminance-contrast sensitivities to be compared in a nonarbitrary way. Relative to chromatic differences, luminance contrast is less effective in maintaining border visibility at lower luminance levels. When the borders combine both chromatic differences and luminance contrasts, a root-mean-square model can be used to account for the data. This is consistent with the idea that chromatic and luminance systems make independent contributions to the visibility of borders.