Photoreceptors not only transduce and relay light signals, they also receive a dazzling array of synaptic input from the retinal network including electrical and chemical synaptic signals from other photoreceptors, lateral feedback from horizontal cells, and long range feedback from interplexiform cells. In this issue of The Journal of PhysiologyGao et al. (2013) quantitatively tackle the interplay between some of these circuits. Gap junction coupling in the retina is pervasive and perplexing. It starts at the first cell layer, the photoreceptors, and is observed in every cell type as visual signals travel through this network. In a tissue designed for acuity, photoreceptor coupling could blur images and dampen signals. Instead, evidence indicates that coupling reduces noise and leads to an increase in both gain and dynamic range: an engineering marvel (Attwell et al. 1987). Coupling can be either homotypic (rod–rod, cone–cone) or heterotypic (rod–cone). Rods are high sensitivity, low resolution light detectors where strong homotypic coupling has clear advantages in improving dim light communication. Furthermore, heterotypic coupling allows rods to pass signals through the cone circuitry with a shift in sensitivity that expands the dynamic range of rod vision (Bloomfield & Dacheux, 2001). Gao et al. (2013) have now presented detailed measurements of this heterotypic coupling in salamander retina. Studies over several decades, often emanating from this laboratory, have made this animal a model system. In addition to the previously described strong heterotypic coupling of a subset (∼20%) of salamander rods, there is also coupling of variable strength with all cone subtypes. The coupling is linear and symmetrical. The rod–cone coupling is less than rod homotypic coupling, ranging from approximately half in the strongly coupled rod subset to about 10% in principal double cones. Surprisingly, the coupling is not under circadian control, unlike fish or rodents, although it is altered by background illumination. But in addition to sign-conserved coupling, Gao et al. (2013) explore the unidirectional ‘sign-inverting’ synapse from rods to cones (Attwell et al. 1983). They found that this feedback, generated by B-type horizontal cells, affects all but accessory cones (these receive A-type horizontal cell feedback). Surprisingly, the horizontal input to cones induces a chloride current, mediated by a glutamate transporter. A similar glutamate-gated channel has been described in bipolar cells, where it is an autoreceptor that suppresses transmitter release (Veruki et al. 2006). At the photoreceptor feedback synapse the mechanism may be more complex because it has to compete with glutamate release from photoreceptors and the potential stimulation of metabotropic receptors. The synaptic calcium channel that induces photoreceptor glutamate release is also under the direct control of horizontal cells. A further complication is that the chloride reversal potential sits slightly below the cone's dark membrane potential, instead of the more negative potential in bipolar cell terminals. Thus, while horizontal cell depolarization linked to surround responses hyperpolarizes and/or reduces transmitter release from cones, this B-type horizontal cell pathway may depolarize cones. The horizontal cell input would occur at light offset and oppose the effect of rod–cone coupling. As the authors suggest, horizontal cell lateral pathways are diverse and involve activation of multiple mechanisms. There is no shortage of candidates, including feedback by pH, GABA and ephaptic synapses (Thoreson & Mangel, 2012). And if unexpectedly glutamate is the transmitter, what role does this leave for GABA found in B-type horizontal cells? The report of Gao et al. (2013) fills in many of the gaps, but (Gao et al. 2013) the saga of horizontal cell feedback continues and you know that with the introduction of a new set of characters the story will not end soon.