1. Frog cutaneous pectoris motor nerve terminals were loaded with the fluorescent dye FM1-43, which produced a series of discrete spots along the length of terminals, each spot evidently marking a cluster of synaptic vesicles. Terminals were imaged for 2-10 min as they destained during repetitive nerve stimulation. Endplate potentials (EPPs) were recorded simultaneously from the muscle fibres innervated by these terminals; their summed amplitudes provided a measure of cumulative transmitter release. 2. Individual fluorescent spots in any one terminal varied in initial brightness but destained at similar fractional rates. 3. The rates of cumulative transmitter release and destaining increased with stimulus frequency in the range 2-30 Hz. At 40 Hz, however, both transmitter release and destaining were slower than at 30 Hz. 4. In twenty-six experiments, rates of dye loss and transmitter release were compared quantitatively. When the time course of summed EPPs was scaled to fit the time course of dye loss during the first 30-60 s of destaining, the two curves usually diverged at later times, the dye loss curve falling below the summed EPP curve. Thus, assuming that dye loss and transmitter release are proportional at early times, at later times the rate of dye loss decreases relative to the rate of transmitter release. 5. At stimulus frequencies from 2 to 30 Hz, the results could be fitted by a simple model in which vesicles lose their dye during exocytosis and, after a fixed recycle 'dead time', they re-enter the vesicle pool, mixing randomly with other vesicles. 6. Unlike stimulation at lower frequencies, at 40 Hz dye loss and summed EPP amplitude curves did not significantly diverge. Stimulation periods lasted up to about 2 min. Interpreted according to the model of vesicle recycling, this suggests that vesicle recycling is inhibited at 40 Hz. 7. The model led to predictions about the relative number, N, of vesicles (labelled and unlabelled) in the terminal at any time during stimulation. The calculated value of N decreased at times less than the recycle 'dead time', and then increased, reflecting the appearance of recycled vesicles in the vesicle pool. 8. From estimates of N and recorded EPP amplitudes, the fraction of vesicles released per shock, F, could be calculated during the entire stimulation period. At low stimulus frequencies (2-5 Hz), after an initial rapid fall, F decreased slowly and monotonically by about 50% in 6 min. At higher stimulus frequencies, a different process was observed.(ABSTRACT TRUNCATED AT 400 WORDS)