The number of quanta secreted from selected sites along terminal branches at toad (Bufo marinus) neuromuscular junctions was determined. Terminal branches were visualized by prior staining with the fluorescent dye, 3-3 Diethyloxadicarbocyanine iodide (DiOC2(5)); neither impulse conduction nor quantal release were affected by DiOC2(5) at concentrations less than 10 microM. The evoked quantal release recorded with an extracellular micro-electrode (me) at different sites along the length of terminal branches was determined in an external calcium concentration, [Ca]o, of 0.35-0.45 mM. For short branches (40-80 microns), me remained approximately constant for over 60% of the branches; for the rest, me declined approximately exponentially with an average length constant of 17 +/- 2 microns (mean +/- S.E. of mean). For both medium (81-120 microns) and long branches (121-160 microns), me declined in nearly all cases approximately exponentially with length constants of 39 +/- 5 and 54 +/- 8 microns respectively. These changes in me were observed at synapses having a wide range of terminal branching patterns. Some DiOC2(5)-stained branches possessed discontinuous cholinesterase staining. In general, me declined along these branches in the same way as along DiOC2(5)-stained branches with continuous cholinesterase staining. It is suggested that because of the decline in me along most medium and long terminal branches, many release sites have a very low probability for secretion in low [Ca]o. Release sites near the point of nerve entry, which have a relatively high probability, therefore make the main contribution to secretion recorded with an intracellular micro-electrode. As a consequence, transmitter secretion from the whole terminal does not fluctuate from impulse to impulse as much as expected if there were a large number of release sites, each with a low probability of secretion. Transmitter secretion then follows binomial rather than Poisson statistics.