1. The number of quanta secreted from visualized release sites along terminal branches at toad (Bufo marinus) neuromuscular junctions in different extracellular concentrations of calcium ions. [Ca2+]o, and during facilitation was determined. Terminal branches were visualized by prior staining with the fluorescent dye, 3-3 diethyloxardicarbocyanine iodide (DiOC2(5)). 2. Increasing [Ca2+]o between 0.25 and 0.4 mM gave a similar proportional increase in the mean quantal content of the end-plate potential recorded with an extracellular electrode (me) at all sites along terminal branches. Thus the length constant for the experimental decline in me along terminal branches (the quantal length constant) remained constant with an increase in [Ca2+]o. The increase in m with [Ca2+]o followed a fourth power relation at both proximal and distal release sites. 3. The increase in me with [Ca2+]o was almost entirely due to an increase in the binomial probability for secretion, pe, which increased as the third to fourth power of [Ca2+]o. However, at higher [Ca2+]o there was an increasing tendency for the binomial parameter ne to increase. It is shown that when ne increases by about 1 there is very little change in pe, suggesting that the new release site recruited at high [Ca2+]o has a relatively low probability for secretion. 4. Test impulses gave a similar proportional increase in me following a conditioning impulse at all sites along terminal branches. The quantal length constant remained constant for both conditioning and test values of me. The increase in me for the test impulse increased linearly with an increase in me for the conditioning impulse at all release sites. 5. Facilitation of me declined exponentially with an increase in the test-conditioning interval. The time constant for this decline (30-40 ms) was similar at both proximal and distal release sites. Changes in facilitation of me were almost entirely due to changes in pe except at very short test-conditioning intervals of about 10 ms. At these intervals ne frequently increased by about 1 and there was very little change in pe. Again, this suggests that additional release sites recruited at short intervals have relatively low probability for secretion. 6. The results indicate that relatively low probability release sites exist in close juxtaposition to relatively high probability release sites which themselves decline in probability along the length of terminal branches.