The time course of local end-plate hyperpolarization after d-tubocurarine application measured by an intracellular microelectrode was followed in vitro in anticholinesterase-treated mouse diaphragm pinned to the bottom of the perfusion chamber. The d-tubocurarine-induced hyperpolarization, which served as an indicator of non-quantal acetylcholine release, started to decline from 6 mV after 1 h and was negligible after 3 h in continuously perfused preparations. This decline was slowed down by 10 mumol l-1 choline and almost completely prevented by long-term nerve stimulation with a frequency of 3 Hz. The rapid decrease of the d-tubocurarine-induced hyperpolarization was observed within 10-15 min after the application of 1 mumol l-1 hemicholinium-3 and substitution of lithium for sodium. Both these procedures inhibit the fast choline uptake into nerve terminals. Our data suggest that the amount of available acetylcholine for non-quantal release is proportional to the rate of its synthesis and to the number of available carriers in the nerve terminals. Some of our observations might also be explained by postulating that the choline-uptake system as such is responsible for the non-quantal release.