SUMMARY1. a-Latrotoxin (a-LTx) was applied to frog cutaneus pectoris muscles bathed at 1-3°C in either Ringer solution, Ca2"-free Ringer solution with 1 mM-EGTA and 4 mM-Mg2+ or Ringer solution plus 4 mM-Mg2+, and its effects on miniature end-plate potential (MEPP) frequency, nerve terminal ultrastructure and uptake of horseradish peroxidase (HRP) were studied.2. Large concentrations (2 ,ug/ml) of a-LTx increased MEPP rates to levels above 100/s at all junctions, but the time course of the increases depended upon the divalent cation content of the bathing solution. However, similar numbers of MEPPs (0-3-407 x 106) were recorded at all junctions during 2 h of secretion.3. Nerve terminals exposed to a-LTx for 2 h lost 60-75 % of their synaptic vesicles and were swollen; their presynaptic membranes were deeply infolded and they often contained many large vesicular structures. Terminals in Ringer solution retained the largest number of synaptic vesicles; terminals in Ringer solution plus Mg2+ swelled the least and contained the largest number of coated vesicles. The average number of synaptic vesicles lost was approximately equal to the average number of MEPPs recorded.4. Few vesicles became loaded with HRP when this extracellular tracer was present in the bathing solution and the muscles were fixed near the peak of secretion.5. When the terminals were warmed to 20°C, those in the Caa2+-free solution with Mg2+ secreted additional quanta and lost almost all their residual vesicles; those in Ringer solution without Mg2+ secreted few additional quanta and retained most of their residual vesicles.6. These results suggest that recycling was blocked at these terminals and that for each quantum secreted a vesicle became permanently incorporated into the axolemma.