We studied the anatomical substrates of spatial vision in a New World monkey, the marmoset Callithrix jacchus. This species has good visual acuity and a foveal specialization which is qualitatively similar to that of humans and other Old World primates.We measured the spatial density of retinal ganglion cells and photoreceptors, and calculated the relative numbers of these cell populations. We find that ganglion cells outnumber photoreceptors by between 2.4:1 and 4.2:1 in the fovea. The peak sampling density of ganglion cells is close to 550,000 cells/mm2. This value falls by almost 1000-fold between the fovea and peripheral retina; a value which approaches recent estimates of the centroperipheral ganglion cell gradient for human and macaque monkey retina and primary visual cortex. The marmoset shows a sex-linked polymorphism of color vision: all male and some female marmosets are dichromats. Six of the retinas used in the present study came from animals whose chromatic phenotype was identified in electrophysiological experiments and confirmed by polymerase chain reaction (PCR) amplification of cone opsin encoding genes. One animal was a trichromat and the others were dichromats. Antibodies against short wavelength-sensitive (SWS) cones labeled close to 8% of all cones near the fovea of onedichromat animal, consistent with electrophysiological evidence that the SWS system is present inall marmosets. The topography and spatial density of cone photoreceptors and ganglion cells was similar to that reported for macaque retina, and we found no obvious difference between dichromatic and trichromatic marmoset retinas. These results reinforce the view that the main determinate of primate foveal topography is the requirement for maximal spatial resolution.

The distributions of rod and cone photoreceptors have been determined in the retina of the California ground squirrel, Spermophilus beecheyi. Retinas were fixed by perfusion and the rods and cones were detected with indirect immunofluorescence using opsin antibodies. Local densities were determined at 2-mm intervals across the entire retina, from which total numbers of each receptor type were estimated and isodensity distributions were constructed. The ground squirrel retina contains 7.5 million cones and 1.27 million rods. The peak density for the cones (49,550/mm2) is found in a horizontal strip of central retina 2 mm ventral to the elongated optic nerve head, falling gradually to half this value in the dorsal and ventral retinal periphery. Of the cones, there are 14 M cones for every S cone. S cone density is relatively flat across most of the retina, reaching a peak (4500/mm2) at the temporal end of the visual streak. There is one exception to this, however: S cone density climbs dramatically at the extreme dorso-nasal retinal margin (20,000/mm2), where the local ratio of S to M cones equals 1. Rod density is lowest in the visual streak, where the rods comprise less than 5% of the local photoreceptor population, increasing conspicuously in the ventral retina, where the rods achieve 30% of the local photoreceptor population (13,000/mm2). The functional importance of the change in S to M cone ratio at the dorsal circumference of the retina is compromised by the extremely limited portion of the visual field subserved by this retinal region. The significance for vision, if any, remains to be determined. By contrast, the change in rod/cone ratio between the dorsal and ventral halves of the retina indicates a conspicuous asymmetry in the ground squirrel's visual system, suggesting a specialization for maximizing visual sensitivity under dim levels of illumination in the superior visual field.