1. Following the addition of crotoxin (250 nM) at the frog neuromuscular junction, there was an initial fall in frequency of miniature endplate potentials (m.e.p.ps), followed by a secondary rise which was characterized by the appearance of large spontaneous potentials (giants, g.m.e.p.ps) and an occasional large potential of the burst type. 2. In the presence of 2-(4-phenylpiperidino)cyclohexanol (AH5183, vesicamol), an inhibitor of vesicular acetylcholine uptake, the frequency of g.m.e.p.ps induced by crotoxin was reduced. 3. The characteristic changes in m.e.p.p. frequency and amplitude distribution were absent with crotoxin in Sr-EGTA Ringer. In the presence of high concentrations of Mn (3.6 or 5.4 mM with 0.9 mM Ca), the crotoxin-induced initial fall and the onset of the secondary rise in m.e.p.p. and g.m.e.p.p. frequencies were slower. The timing of these phases was unaffected by Ca concentrations ranging from 6.3 to 0.9 mM. 4. High concentrations of Mn ions partially inhibited the phospholipase A2 activity of crotoxin on artificial phospholipid membranes. This also supports the involvement of the Ca-dependent phospholipase A2 subunit in both phases of the physiological action of the toxin. 5. G.m.e.p.ps were associated with a moderate increase in m.e.p.p. frequency (2-3 s-1) and were of a time-course similar to that of m.e.p.ps. They persisted after washing with medium lacking Ca ions and in the presence of Ca-Mn Ringer that blocked evoked responses. 6. It is concluded that crotoxin, acting through its phospholipase A2 subunit, produces specific disturbances of synaptic exocytosis and vesicle formation in the axolemma of the motor nerve terminal which lead to biphasic changes in m.e.p.p. frequency and the onset of large spontaneous potentials.