The retinas of teleost fish grow continuously, in part, by neuronal hyperplasia and when lesioned will regenerate. Within the differentiated retina, the growth-associated hyperplasia results in the generation of new rod photoreceptors only, whereas injury-induced neurogenesis results in the regeneration of all retinal cell types. It is believed, however, that both new rod photoreceptors and regenerated neurons originate from the same populations of intrinsic progenitors. Experiments are described here that attempt to identify in the normal retina of goldfish neuronal progenitors intrinsic to the retina, particularly those which have remained cryptic because they divide infrequently. Long-term, systemic exposure to bromodeoxyuridine (BrdU) was used to label these cells. Five populations of proliferative cells were labeled: microglia, which are briefly described but not studied further; retinal progenitors in the circumferential germinal zone (CGZ); and rod precursors in the outer nuclear layer (ONL), both of which have been well characterized previously; and two populations of slowly-dividing cells in the inner nuclear layer (INL). The majority of these cells have a fusiform morphology, whereas the remaining ones are spherical. Longitudinal BrdU labeling suggests that the fusiform cells migrate to the ONL to replenish the pool of rod precursors. A subset of the spherical cells express pax6, although none are stained with markers of differentiated amacrine or bipolar cells. It is hypothesized that these rare, pax6-expressing cells are retinal stem cells, which give rise to the pax6-negative fusiform cells. Based on these data, two models are proposed: the first describes the lineage of rod photoreceptors in goldfish; the second is a consensus model of neurogenesis in the retinas of all teleosts.