H1 horizontal cells in goldfish retina are probably GABAergic and receive excitatory synaptic input from red cones. This input should affect the synaptic release of GABA from H1 cells. We studied the uptake and release of [3H]GABA from the isolated goldfish retina by use of autoradiography. When retinas were incubated in the light for 15 min in 0.72 microM [3H]GABA, heavy label was found over the somata (HS) and axon terminals (HAT) of H1 horizontal cells, and over pyriform amacrine cell bodies and their processes in sublamina b of the IPL. Postincubation of retinas, preloaded with [3H]GABA, in 0.5-10 mM L-glutamate or 0.1-10 mM L-aspartate, resulted in a dose-dependent and selective loss of [3H]GABA from HS and very little loss from HAT. This loss was not due to an efflux of metabolites of [3H]GABA or to any calcium-dependent vesicular release of [3H]GABA from HS. The glutamate-evoked release of [3H]GABA by H1 cells was sodium dependent, sensitive to substitution of lithium for sodium, and inhibited by nipecotic acid. In addition, [3H]GABA was released from HS by 0.1 mM ouabain but not by 50 mM potassium chloride. Our results suggest that the chemically evoked release of [3H]GABA from HS is mediated by a sodium-dependent transport carrier which may be responsible for the high affinity uptake of [3H]GABA by H1 cells as well. Since synaptic vesicles are not found at presumed synaptic release sites in H1 cells, we suggest that the GABA which is released synaptically from H1 cells may derive from a cytoplasmic pool of GABA and is released by means of a transport carrier. This carrier appears to depend primarily on the sodium concentration gradient across the H1 cell membrane rather than on the membrane potential of the H1 cell for its action. The relevance of the carrier-mediated release of GABA from HS in regard to the synaptic function of H1 cells is discussed.