In all mammalian species the projections of the two eyes to the dorsal lateral geniculate nucleus are initially overlapping before gradually forming the eye-specific domains evident at maturity. It is widely thought that retinal waves of neuronal activity play an instructional role in this developmental process. Here, I discuss the myriad reasons why retinal waves are unlikely to have such a role, and suggest that eye-specific molecular cues in combination with neuronal activity are most probably involved in the formation of eye-specific retinogeniculate projections.