Although all bipolar cells in the retina probably use the excitatory transmitter glutamate, approximately half of the cone bipolar cells also contain elevated levels of the inhibitory transmitter glycine. Some types of cone bipolar cells make heterologous gap junctions with rod amacrine cells, which contain elevated levels of glycine, leading to the hypothesis that the bipolar cells obtain their glycine from amacrine cells. Experimental support for this hypothesis is now provided by three independent lines of evidence. First, the glycine transporter GLYT1 is expressed by the glycine-containing amacrine cells but not by the glycine-containing bipolar cells, suggesting that only the amacrine cells are functionally glycinergic. Second, the gap-junction blocker carbenoxolone greatly reduces exogenous 3H-glycine accumulation into the bipolar cells but not the amacrine cells. Moreover, when the endogenous glycine stores in both cell classes are depleted by incubating the retina with a glycine-uptake inhibitor, carbenoxolone blocks the subsequent glycine replenishment of the bipolar cells but not the amacrine cells. Third, intracellular injection of rod amacrine cells with the gap-junction permeant tracer Neurobiotin secondarily labels a heterogenous population of cone bipolar cells, all of which show glycine immunoreactivity. Taken together, these findings indicate that the elevated glycine in cone bipolar cells is not derived by high-affinity uptake or de novo synthesis but is obtained by neurotransmitter coupling through gap junctions with glycinergic amacrine cells. Thus transmitter content may be an unreliable indicator of transmitter function for neurons that make heterologous gap junctions.