PURPOSE. To determine the light sensitivity of poorly reflective cones observed in retinas of normal subjects, and to establish a relationship between cone reflectivity and perceptual threshold. METHODS. Five subjects (four male, one female) with normal vision were imaged longitudinally (7–26 imaging sessions, representing 82–896 days) using adaptive optics scanning laser ophthalmoscopy (AOSLO) to monitor cone reflectance. Ten cones with unusually low reflectivity, as well as 10 normally reflective cones serving as controls, were targeted for perceptual testing. Cone-sized stimuli were delivered to the targeted cones and luminance increment thresholds were quantified. Thresholds were measured three to five times per session for each cone in the 10 pairs, all located 2.2 to 3.38 from the center of gaze. RESULTS. Compared with other cones in the same retinal area, three of 10 monitored dark cones were persistently poorly reflective, while seven occasionally manifested normal reflectance. Tested psychophysically, all 10 dark cones had thresholds comparable with those from normally reflecting cones measured concurrently (P =0.49). The variation observed in dark cone thresholds also matched the wide variation seen in a large population (n = 56 cone pairs, six subjects) of normal cones; in the latter, no correlation was found between cone reflectivity and threshold (P = 0.0502). CONCLUSIONS. Low cone reflectance cannot be used as a reliable indicator of cone sensitivity to light in normal retinas. To improve assessment of early retinal pathology, other diagnostic criteria should be employed along with imaging and cone-based microperimetry.