We quantified the spatial distribution of presumed ganglion cells and displaced amacrine cells in unstained whole mounts of six young normal human retinas whose photoreceptor distributions had previously been characterized. Cells with large somata compared to their nuclei were considered ganglion cells; cells with small somata relative to their nuclei were considered displaced amacrine cells. Within the central area, ganglion cell densities reach 32,000--38,000 cells/mm2 in a horizontally oriented elliptical ring 0.4--2.0 mm from the foveal center. In peripheral retina, densities in nasal retina exceed those at corresponding eccentricities in temporal retina by more than 300%; superior exceeds inferior by 60%. Displaced amacrine cells represented 3% of the total cells in central retina and nearly 80% in the far periphery. A twofold range in the total number of ganglion cells (0.7 to 1.5 million) was largely explained by a similar range in ganglion cell density in different eyes. Cone and ganglion cell number were not correlated, and the overall cone: ganglion cell ratio ranged from 2.9 to 7.5 in different eyes. Peripheral cones and ganglion cells have different topographies, thus suggesting meridianal differences in convergence onto individual ganglion cells.
Low convergence of foveal cones onto individual ganglion cells is an important mechanism for preserving high resolution at later stages of neural processing. Our improved estimates for the density of central ganglion cells allowed us to ask whether there are enough ganglion cells for each cone at the foveal center to have a direct line to the brain. Our calculations indicate that (1) there are so many ganglion cells relative to cones that a ratio of only one ganglion cell per foveal cone would require fibers of Henle radiating toward rather than away from the foveal center; and (2) like the macaque, the human retina may have enough ganglion cells to transmit the information afforded by closely spaced foveal cones to both ON- and OFF-channels. Comparison of ganglion cell topography with the visual field representation in V1 reveals similarities consistent with the idea that cortical magnification is proportional to ganglion cell density throughout the visual field.