The purpose of this study was to test the hypothesis that chromaticity or luminance cues can be used to segregate a subset of relevant stimuli from randomly intermixed irrelevant stimuli. We used a 2AFC procedure to estimate thresholds for detecting a target differing from distractors in chromaticity in all experiments. Stimuli were small disks displayed briefly (0.2 sec) within an annular region centered on a fixation point on a uniform gray background. Subjects indicated whether a single target appeared to the right or the left of fixation. In control conditions 2, 4, 8, or 24, green stimuli were presented. The target was slightly bluer (+S chromaticity) than the rest of the stimuli. Thresholds for detecting the bluish target in these conditions were compared with thresholds for detecting the target when 2, 4, or 8, green stimuli were presented randomly intermixed with enough red stimuli to make a total of 24 stimuli in each display. A cue indicated that one of the green stimuli contained the blue increment and the red stimuli were irrelevant. In a third condition, variation in the color of the irrelevant stimuli was introduced. In all three conditions, thresholds increased with the number of green stimuli presented. The presence of the irrelevant stimuli had little or no effect on the thresholds. A similar experiment was performed with luminance contrast as the dimension defining relevant (dim) and irrelevant (bright) stimuli. The stimuli were all bluish in appearance with a single target defined by an increment in redness (+L chromaticity). Thresholds again increased with the number of stimuli presented in the control condition. Thresholds were higher when the relevant stimuli were presented among irrelevant stimuli and did not vary much with the number of relevant stimuli. We conclude that chromaticity can be used effectively to segregate a subset of stimuli to be attended, but luminance contrast cannot.