Abstract
Individual differences in color appearance are large and reliable among color-normal observers, but it remains unclear what processes underlie these differences. In a recent factor analysis of hue-scaling functions, we found that variability depended on multiple, narrowly-tuned processes, consistent with a multichannel or population code in which different "directions" in color space are represented separately. To gain insights into this representation, we examined individual differences in the perception of motion, which like color is defined by three cardinal, and in this case, Cartesian dimensions. We designed a motion scaling task similar to hue scaling, where observers judged the strength of up/down and left/right percepts in a moving stimulus (analogous to decomposing a hue into red/green and blue/yellow percepts). The stimulus was a 2° circular aperture of dots moving coherently in one of 36 possible directions at 10° intervals along the 2D plane. The directions were displayed in random order and were pulsed repeatedly along the same trajectory until observers recorded their response. Settings for twenty-three observers were factor-analyzed using PCA and Varimax rotation. The analysis revealed approximately four systematically-tuned factors (i.e. with significant loadings on two or more adjacent stimuli). Together these accounted for 60% of the total variance. The first factor's loadings varied roughly sinusoidally with a period of 180°, consistent with differences in the relative weighting of horizontal vs. vertical. The second factor corresponded to variations relative to both the cardinal axes and diagonals (~sinusoidally varying with a period of 90°). Unlike the narrow factors for color, the factors for motion had a global pattern of loadings consistent with a metrical representation of space in terms of the underlying Cartesian axes. These results suggest that while color and motion share a common dimensional structure, the perceptual representation of color- and motion-defined space may be fundamentally different.
Meeting abstract presented at VSS 2018