Abstract
The distinction between warm vs. cool colors has been widely invoked as a fundamental dimension of both color appearance and color emotion. We examined how this dimension varied among observers and the relationship to both the cone-opponent dimensions of early color coding (LvsM and SvsLM cardinal axes) and the perceptual opponency of color appearance (red-green and blue-yellow). Observers viewed a sample of 36 stimuli with a fixed nominal (~multiple of threshold) contrast of 80 and spanning the cone-opponent plane in 10-deg steps. Each was shown individually on an equiluminant (20 c/m2) gray background and pulsed 500-ms on, 150-ms off while observers rated the warm-cool attribute on a 7-point scale (ranging from very cool to neutral to very warm). Four measurements were made for each stimulus in random order. In a second set of measurements, the full set of 36 hues were displayed as a circle on the screen and the same observers varied a pointer to select their best example of the four unique (e.g., red or yellow) and four balanced binary hues (e.g., a 50%-red, 50%-yellow orange), again with four settings for each. The two boundaries for warm vs. cool were estimated from polynomial fits to each observer’s ratings. These loci varied substantially across observers but in almost all cases fell within the first and third quadrants of the cone-opponent space (corresponding to purplish reds and green to yellowish-greens). Inter-observer variability was larger for the warm-cool boundaries than the hue loci and showed weak to no correlations with each other or the hue loci. These results suggest that warm-cool judgments correspond to a dimension of color appearance independent of both the cone-opponent and perceptual-opponent axes; and to a dimension that tends to vary along bluish to yellow/orange axes, which are prominent axes of variation in natural illumination and environments.