The human retina contains three classes of cone photoreceptors each sensitive to different portions of the visual spectrum: long (L), medium (M) and short (S) wavelengths. Color information is computed by downstream neurons that compare relative activity across the three cone types. How cone signals are combined at a cellular scale has been more difficult to resolve. This is especially true near the fovea, where spectrally-opponent neurons in the parvocellular pathway draw excitatory input from a single cone and thus even the smallest stimulus projected through natural optics will engage multiple color-signaling neurons. We used an adaptive optics microstimulator to target individual and pairs of cones with light. Consistent with prior work, we found that color percepts elicited from individual cones were predicted by their spectral sensitivity, although there was considerable variability even between cones within the same spectral class. The appearance of spots targeted at two cones were predicted by an average of their individual activations. However, two cones of the same subclass elicited percepts that were systematically more saturated than predicted by an average. Together, these observations suggest both spectral opponency and prior experience influence the appearance of small spots.