Several studies document rudimentary color vision under dim illumination. Here, hue perceptions of paper color samples were determined for a wide range of light levels, including very low light levels where rods alone mediate vision. The appearances of 24 paper color samples from the OSA Uniform Color Scales were gauged under successively dimmer illuminations from 10-0.0003 Lux. Triads of samples were chosen representing each of eight basic color categories; red, pink, orange, yellow, green, blue, purple, and gray. Samples within each triad varied in lightness. Observers sorted samples into groups that they could categorize with specific color names. Above 0.32 Lux, observers sorted the samples into the originally chosen color groups with few exceptions. For 0.1-0.01 Lux, the red and orange samples were usually correctly identified as either red or orange. The remaining samples tended to be grouped into two categories, associated with the scotopic sample reflectance. The lowest reflectance samples were below threshold and were named black. The higher reflectance group was named predominately as green or blue-green (three observers; the fourth observer used blue or achromatic). At the three dimmest levels (< or = 0.0032 Lux) there continued to be conspicuous color percepts. Color categories were reliably assigned based on relative sample scotopic lightness. Of the samples above threshold, those with lower reflectance were classified as red or orange (all observers) and the higher reflectance samples as green or blue-green (three observers) or achromatic or blue (the fourth observer). Rods and L-cones presumably mediated color percepts at the intermediate light levels used in the study. At the three lowest light levels there were distinct color appearances mediated exclusively by rods. We speculate that at these light levels the visual system estimates probable colors based on prior natural experience.