We propose that there exists within the avian, and perhaps more generally in the vertebrate retina, a two-state nonadapting flip-flop circuit, based on reciprocal inhibitory interactions between the photoreceptors, releasing melatonin, the dopaminergic amacrine cells, and amacrine cells which contain enkephalin-, neurotensin-, and somatostatin-like immunoreactivity (the ENSL1 amacrine cells). This circuit consists of two loops, one based on the photoreceptors and dopaminergic amacrine cells, and the other on the dopaminergic and ENSLI amacrine cells. In the dark, the photoreceptors and ENSL1 amacrine cells are active, with the dopaminergic amacrine cells inactive. In the light, the dopaminergic amacrine cells are active, with the photoreceptors and ENSLI amacrine cells inactive. The transition from dark to light state occurs over a narrow (<1 log unit) range of low light intensities, and we postulate that this transition is driven by a graded, adapting pathway from photoreceptors, releasing glutamate, to ON-bipolar cells to dopaminergic amacrine cells. The properties of this pathway suggest that, once released from the reciprocal inhibitory controls of the dark state, dopamine release will show graded, adapting characteristics. Thus, we postulate that retinal function will be divided into two phases: a dopamine-independent phase at low light intensities, and a dopamine-dependent phase at higher light intensities. Dopamine-dependent functions may show two-state properties, or two-state properties on which are superimposed graded, adapting characteristics. Functions dependent upon melatonin, the enkephalins, neurotensin, and somatostatin may tend to show simpler two-state properties. We propose that the dark-light switch may have a role in a range of light-adaptive phenomena, in signalling night-day transitions to the suprachiasmatic nucleus and the pineal, and in the control of eye growth during development.