Abstract
A number of recent studies have pointed to asymmetries in color constancy for blue vs. yellow illuminants, including weaker sensitivity for detecting changes in blue illuminants and a tendency for blue tints to appear more achromatic. We compared how blue and yellow percepts are parsed between lighting and surfaces in scenes. Stimuli were uncalibrated images sampled from the internet of outdoor or indoor scenes or objects with a dominant yellow to orange or blue to blue-green hue. For each, chromatic contrast was rescaled or inverted to create a range of images varying along the complementary bluish to yellowish axis. Observers were shown each image and asked to identify the dominant color of the lighting and surface, and also the “known” color of the object. In scenes with strong blue tints (e.g. because of artificial lighting, shadows, or water) there is a strong bias to attribute the hue to the lighting while the objects appear achromatic. Conversely, in the same scenes with yellow tints, the hue is instead more likely to be attributed to the surface color, with the lighting perceived as more achromatic. For example, a room might appear white in blue light while yellow in white light, and translucent water might change to chocolate. These differences may partly reflect specific knowledge of object color (e.g. skin tones), but were also evident for objects which could be associated with any hue (e.g. clothing), or which are more likely to have neutral reflectances (e.g. interiors). Thus, they may reflect more general priors for a blue bias for lighting and shadows, while a yellow bias for surfaces, and may be consistent with recent evidence that objects are defined by predominantly warmer colors (e.g. Rosenthal et al., JOV 2018).
Acknowledgement: EY010834