1. Rats were injected once every 48 h with alpha-bungarotoxin (alpha BTX) for periods up to 6 weeks. Injections caused weakness of facial muscles which lasted about 8 h. Hemidiaphragms were dissected for biochemical and electrophysiological measurements. 2. In muscles from animals treated for 2-3 weeks with toxin, the binding of 125I-alpha BTX was reduced to 58%, and the ACh content to 81% of control values. Choline acetyltransferase activity was unchanged. ACh release evoked by 3 Hz nerve stimulation was increased to 175% of control values. 3. The use of mu-conotoxin, which specifically blocks muscle action potentials, enabled the recording of full-sized endplate potentials (EPPs) and miniature endplate potentials (MEPPs) at normal muscle membrane potentials (-70 to -80 mV). The amplitude of MEPPs was decreased to 57% in muscles from animals treated for 3 weeks with alpha BTX. The mean of the quantal contents, calculated from the ratio of the corrected EPPs and the MEPPs, was increased to 154%. 4. Within individual muscles of both alpha BTX-treated and control rats, there was an inverse relationship between the quantal content of an endplate and its MEPP amplitude. 5. The MEPP frequency of endplates from control muscles was positively correlated with the quantal content. However, this correlation was not found in alpha BTX-affected muscles. 6. Three hours after a single injection of alpha BTX the amplitude of the MEPPs was reduced to about 60% of control values but no increase of the quantal content was found. During the first few days of alpha BTX treatment the quantal content gradually increased; it reached a plateau between 20 and 30 days. 7. The results suggest the existence of an adaptive mechanism, operating at individual endplates, in which retrograde signals at the motor nerve terminals modulate ACh release when neuromuscular transmission is endangered by block of acetylcholine receptors.