Müller cells from 22 mammalian species were subjected to morphological and electrophysiological studies. In the 'midperiphery' of retinae immunocytochemically labeled for vimentin, estimates of Müller cell densities per unit retinal surface area, and of neuron-to-(Müller) glia indices were performed. Müller cell densities were strikingly similar among the species studied (around 8000-11,000 mm-2) with the extremes of the horse (< or = 5000 mm-2) and the tree shrew (> or = 20,000 mm-2). By contrast, the number of neurons per Müller cell varied widely, being clustered at 6-8 (in retinae with many cones), at about 16, and at up to more than 30 (in strongly rod-dominated retinae). Isolated Müller cell volumes were estimated morphometrically, and cell surface areas were calculated from membrane capacities. Müller cells isolated from thick vascularized retinae (carnivores, rats, mice, ungulates) were longer and thinner, and had smaller volumes but higher surface-to-volume ratios than cells from thin paurangiotic (i.e. with blood vessels only near the optic disc) or avascular retinae (rabbits, guinea pigs, horses, zebras). In whole-cell voltage-clamp studies, Müller cells from all mammals studied displayed two dominant K+ conductances, inwardly rectifying currents and delayed rectifier currents. TTX-sensitive Na+ currents were recorded only in some species. Based on these data, the following hypotheses are presented, (a) neuron-to-(Müller) glia indices are determined by precursor cell proliferation rather than by metabolic demands; (b) Müller cell volumes depend on available space rather than on the number of supported neurons; and (c) it follows that, the specific metabolic activities of Müller cells must differ greatly between species, a difference that may contribute to distinct patterns of retinal vascularization.