Abstract
The distinction between warm versus cool colors is widely considered a robust and fundamental property of color experience, yet whether this dimension reflects a perceptual or conceptual property remains unclear. Previously we measured warm-cool colors by scaling for this dimension for a range of hues evenly sampled in the LvsM, SvsLM cone-opponent space (Fite et al. VSS 2021). The ratings peaked at orangish (warm) and bluish (cool) hues but were uncorrelated with observers’ individual hue loci, suggesting an independent dimension. Here we compare these ratings to the asymmetries in color perception predicted by uniform color space (CIELAB). Contours of constant chroma in LAB project to ellipsoids in the cone-opponent space, consistent with weaker sensitivity to bluish-yellowish chromatic contrasts where LvsM and SvsLM signals are negatively correlated. Both the maxima and minima of these ellipsoids closely align with the boundaries and peaks of the warm-cool ratings (estimated from polynomial fits to the mean ratings). This close correspondence points to an actual perceptual analog for warm-cool colors. Notably, the peak warm and cool colors lie at hue angles that require the largest cone-opponent contrasts for constant chroma, meaning that they are chromatic directions of lowest perceived contrast. Conversely, the warm-cool boundaries fall at hue angles with the minimum required contrast, or chromatic directions of highest sensitivity. This pattern could arise if the visual system is more strongly adapted to chromatic variations along the warm-cool axis. Natural illuminants and scenes vary more in bluish-yellowish contrast, predicting reduced contrast gain for this dimension, but these variations are typically too rotated toward the SvsLM axis to account for the orange axis of warm-cool colors and uniform color metrics. This discrepancy could arise if typical natural scenes or their global color statistics are not typical of the environments or selective color sampling we are adapted to.