What is white? We find that when subjects set white in a dark surround, settings are well constrained to be neither red nor green, but the amount of yellow or blue varies for each subject as well as between subjects. This scatter of settings according to unique hue, rather than along physiological cone-opponent axes, is at odds with discrimination experiments that result in threshold ellipses with an orientation nearer to the cone-opponent axes.

Discrimination experiments are commonly done with two bordering half-fields, while our white settings were done without a reference. Is the abutting reference field critical? In this second experiment, subjects discriminated abutting and non-abutting near-white stimuli. Resulting threshold ellipses are nearer to the cone-opponent axes than to the unique hue directions.

The difference between absolute judgments of white and discrimination of near white stimuli has two possible explanations. 1) White settings use unique-hue mechanisms while discrimination uses the physiological cone-opponent mechanisms. 2) Both white settings and discrimination use cone-opponent mechanisms. The trial-by-trial scatter in a subject's white settings may be generated by the subject choosing, on each trial, a slightly different assumed illuminant from the set of likely natural daylights. One would then expect subjects' settings to be scattered in the blue/yellow direction, as we found, since that is the direction of greatest variance of natural illuminants.