1. When toad retinae were incubated first with veratrine, then with antibodies that reacted with the outer segments of photoreceptors, and finally with complement, horizontal cells survived and most other neurones died. This preparation of 'isolated' horizontal cells accumulated radioactive GABA from the incubation medium. The subsequent release of radioactive GABA could then be measured. 2. The efflux of GABA was increased by exposure to an elevated potassium concentration or added glutamate. Both procedures are known to depolarize horizontal cells. 3. GABA in the external medium also increased the efflux of GABA. 4. The increase in GABA efflux produced by an elevated potassium concentration was unaffected with calcium in the external medium was replaced with cobalt and when sodium was replaced with choline or lithium. 5. The increase in GABA efflux produced by glutamate was unaffected when calcium was replaced with cobalt and when sodium was replaced with lithium, but was inhibited when sodium was replaced with choline. 6. The increase in GABA efflux produced by external GABA was unaffected when calcium was replaced with cobalt but required sodium. Neither choline nor lithium would substitute for sodium. 7. An increase in GABA efflux was accompanied by an increase in sodium efflux. 8. After a high concentration of GABA (2-20 mM) had produced a maximal increase in GABA efflux, the addition of glutamate produced no further effect. Conversely, after a high concentration of glutamate (2-20 mM) had produced a maximal increase in efflux, the addition of external GABA produced only a small further increase. These and the preceding results could occur if GABA release were mediated by a carrier system which could be activated by either depolarization or homoexchange.