Abstract
Purpose: To measure how well humans can isolate a target color among heterogeneous distracter colors.
Method: We used a centroid-estimation task to quantify this ability. For each subject, we first used a minimum motion paradigm to obtain 8 equiluminant lights evenly spaced around an ellipse in the space spanned by the (Stockman-Sharpe) cone fundamentals. This ellipse had an achromatic center, and the elliptically arrayed lights had the maximum saturation available on our monitor. Each subject was thoroughly trained with trial-by-trial feedback (showing the true centroid, the stimulus dots and the participant’s response) to estimate the centroid of a cloud of dark dots on a gray background. In the color-selection task, on each trial, 3 dots of each color (24 in all) with locations drawn from a bivariate Gaussian distribution with a roved mean were shown for 300ms after an initial fixation period. The participant used a mouse-click to locate the centroid of the three target-color dots. Feedback was displayed following the response. Linear regression was used to derive the weights exerted by dots of different colors in the centroid computation used by participants.
Results: Subjects weighted the target color substantially more than the distracter colors with ratio between the single-target color and the SUM of all distracter-color weights ranging from 2.5 to 4 on average. In control conditions, subjects performed the dot centroid task but the 8 colors were replaced with 8 achromatic contrasts that spanned the maximum range of the monitor. For all the target contrasts except the darkest, the ratio ranged from 0.3 to 1. Only the darkest contrast yielded performance comparable in effectiveness to that observed with each individual color. We conclude that subjects can effectively attend to single hues but not to single gray levels, with the exception of black.
Meeting abstract presented at VSS 2013