The cholinergic amacrine cells of the rabbit retina may be labeled with 3H-choline, and the activity of the cholinergic population may be monitored by following the release of 3H-ACh. In magnesium-free medium, the glutamate analog NMDA caused massive ACh release, up to 50x the basal efflux. Magnesium blocked the NMDA-evoked release of ACh with an IC50 of 151 microM. The NMDA-evoked release of ACh was unchanged in glycine-free medium or in the presence of 500 microM glycine. However, the block of NMDA-evoked release by 7-chlorokynurenic acid (7-Cl-Kyn) was reversed by exogenous glycine. This suggests that the NMDA receptors mediating ACh release possess an allosteric glycine binding site, but under normal conditions, it is saturated by endogenous glycine. Submaximal doses of NMDA were used to determine the potency of NMDA antagonists and their specificity was established with submaximal doses of other glutamate agonists. DL-2-amino-7-phosphonoheptanoate (DL- AP-7) was a competitive NMDA antagonist, with an IC50 of 33 microM and (+)5-methyl-10,11-dihydro-5H-dibenzo[a,d]cyclohepten-5,10-imine hydrogen maleate (MK-801) was a noncompetitive NMDA antagonist, with an IC50 of 10.6 nM. Neither antagonist blocked the light-evoked release of ACh from the retina. Furthermore, light stimulation did not activate the use-dependent block characteristic of MK-801, indicating that the endogenous transmitter did not open the NMDA channel. These results suggest that NMDA receptors do not mediate the physiological input to cholinergic amacrine cells in the rabbit retina.