A quantitative description of the electrical properties of the photoreceptor layer in the salamander retina was obtained from earlier data on the characteristics of isolated rods and cones and on rod-rod coupling, and from new data on rod-cone and cone-cone coupling and on the rod photocurrent. Injecting -1 nA current into a rod elicits hyperpolarizations of about 20 mV in an adjacent rod and 4 mV in an adjacent cone. Responses of more distant receptors are smaller. Injecting -1 nA into a cone elicits hyperpolarizations of about 4 mV in an adjacent rod and 0.4 mV in a nearby cone. Depolarizing current evokes smaller responses. Assuming, in agreement with anatomical evidence, that each rod is electrically coupled to four rods and to four cones around it, and that there is no direct electrical coupling between cones, we found these results could be predicted from the properties of isolated rods and cones if adjacent rods are coupled by a resistance of 300 M omega and adjacent rods and cones are coupled by a resistance of 5000 M omega. The small cone-cone coupling seen is due to coupling via intervening rods. The two halves of double cones are not electrically coupled. The spectral sensitivity of both halves is a maximum around 620 nm wave-length. The rod photocurrent has been characterized by voltage-clamping rods isolated from the retina. In agreement with Bader, MacLeish & Schwartz (1979) we found the time course of the photocurrent to be approximately independent of voltage between -35 and -85 mV. The voltage responses of rods, single cones and double cones isolated from the retina obey the principle of univariance. Responses of receptors in the retina do not obey univariance. The main deviations from univariance observed can be explained if adjacent rods and cones are coupled by a resistance of 5000 M omega. Our data demonstrate that rod-cone coupling is relatively weak. We simplified our description of the photoreceptor network, by omitting cones, to investigate the spatiotemporal processing that the rod network is capable of. Computer simulations predict, as is found experimentally, that the rod voltage response to a large spot of bright light should show a much more pronounced initial transient hyperpolarization than the response to a small spot of light of the same intensity. This difference is produced by the combination of electrical coupling of the rods with the existence of a voltage-gated current, IA, in the rod membrane.(ABSTRACT TRUNCATED AT 400 WORDS)