Using a modified Humphrey perimeter, we evaluated 16 eyes with primary open-angle glaucoma and visual field loss (defects 0.5-3.0 log units in depth), and 14 normal eyes. Each eye was tested twice in random order with conventional luminance-increment static perimetry and with the perimeter modified to produce a high-luminance yellow adapting background and a blue test stimulus. The background was a broad-spectrum light of 500 nm and above (yellow), while the stimulus was a broad-spectrum light of 500 nm and below (blue). Paired comparisons were made between conventional and blue/yellow sensitivities for every point examined (1184 points in 16 diseased eyes and 1036 points in 14 normal eyes). Defect depths were determined by using the age-corrected norms distributed in the Humphrey Statpac software. In glaucomatous eyes, blue/yellow sensitivity showed greater impairment than did conventional perimetric sensitivity, in which defect depths were less than 1.0 log unit. However, for defects greater than 1.0 log unit in depth, conventional perimetric sensitivity and blue/yellow sensitivity showed equivalent degrees of damage. Receiver operating characteristic (ROC) analysis was used to compare the ability of blue/yellow and of conventional perimetry in distinguishing between glaucomatous and normal eyes. Results indicated that although blue/yellow color-contrast perimetry may be more sensitive for the detection of incipient glaucomatous damage, in the manifest stages of visual field damage blue/yellow color-contrast perimetry is no more sensitive than is conventional (luminance-increment) perimetry for defining the extent of glaucomatous visual field defects.