Abstract
Sun and colleagues (PNAS, 2016) reported that attentional filters are considerably narrower for hue than for saturation. They used the ‘centroid paradigm’: Participants judge the center of gravity of an array of dots of a target color, embedded in distractor dots of other colors. The weights applied to dots of each color can then be modeled. One problem with Sun et al's application of this valuable method to color is that their hue targets were all of high saturation, whereas saturation targets included ones that would be difficult to detect against the equiluminant background.
We implemented a centroid procedure under conditions where the two types of stimulus could be expressed in the common metric of the MacLeod-Boynton diagram. We scaled the diagram so that thresholds for discrimination along the two cardinal axes were the same at the white point. Target chromaticities fell on either a 45° or a −45° line passing through the white point. Saturation distractors fell on the same line, while hue distractors fell on an orthogonal line through the target, at the same distances from the target as the saturation distractors. Thus, saturation and hue distractors differed from the target by equal distances in S/(L+M) and L/(L+M) channels, but the underlying signals were in different phases. (Danilova, Mollon, ProcRoySocB, 2016).
We could not replicate the dramatic differences between hue and saturation reported by Sun et al: Selectivities were similar for the two types of filter. There were reliable secondary differences between quadrants of the MacLeod-Boynton diagram.