Color channels, not color appearance or color categories, guide visual search for desaturated color targets

Abstract

In this article, we report that in visual search, desaturated reddish targets are much easier to find than other desaturated targets, even when perceptual differences between targets and distractors are carefully equated. Observers searched for desaturated targets among mixtures of white and saturated distractors. Reaction times were hundreds of milliseconds faster for the most effective (reddish) targets than for the least effective (purplish) targets. The advantage for desaturated reds did not reflect an advantage for the lexical category "pink," because reaction times did not follow named color categories. Many pink stimuli were not found quickly, and many quickly found stimuli were not labeled "pink." Other possible explanations (e.g., linear-separability effects) also failed. Instead, we propose that guidance of visual search for desaturated colors is based on a combination of low-level color-opponent signals that is different from the combinations that produce perceived color. We speculate that this guidance might reflect a specialization for human skin.

Figure 1

Search for a desaturated target among white and saturated distractors. (a) Examples of desaturated red (left) and purple targets (right), along with their distractors; these colors are only illustrative, and will not be correctly displayed in print or on an uncalibrated video monitor. (b) The colors used in these experiments. Target colors and distractor colors were at constant separation from white in CIELAB color space (horizontal lines). The colors of the symbols show the approximate colors of the stimuli. Disks and squares represent data from two different stimulus sets and different observers; black symbols: constant-saturation settings of stimuli for observer DTL. (c) Reaction times (left ordinate, ± 1 s.e.m.) for correct responses, target-present searches at set size = 40. Small dots on the x-axis show the chromaticities of lips (L) and skin (S) from Gozalo-Diaz et al., 2007. Observers named each target color as blue, purple, pink, peach, orange, brown, yellow or green. Pastel areas plot the frequency with which observers used each color term (percentages on right ordinate).

Figure 2

Replication of the main result. Symbol colors are the RGB values of the desaturated targets (circles, triangles, and small squares) and distractors (big squares) These colors will not be chromatically correct on the reader's monitor or printout. Circles: reaction time data, +/- 1 s.e.m. Bold line, model described in text and Supplement. Inset, stimulus chromaticities in xyY coordinates. Squares, y (left ordinate) vs. x (abscissa); Triangles, luminance (Y, right ordinate) vs. x (abscissa).

Figure 3

Fits of a quantitative model described by Equ.1 in the text to the results of the main experiment. Symbols, the reaction time data from Figure 1c. (a) The reciprocals of the responses of the LM- (red line) or S-cone (blue line) components of the model. (b) Fit of the model (solid black line) to the reaction time data, based on selection of “fastest” channel response. Dotted lines in each panel indicate fits based on Equ. 1, but without g(s).