In order to determine the relative probability of evoked transmitter release from different parts of frog motor nerve terminals, a technique has been developed in which single quantum end-plate potentials (e.p.p.s) are recorded by two intracellular electrodes, located at opposite ends of identified junctions. The log of the ratio of the amplitudes recorded simultaneously at the two electrodes is a linear function of the distance of the site of origin of the event from each of the two electrodes. Using online computer data acquisition and analysis, and current pulses at known locations for spatial calibration, it is possible to localize the site of single quantum e.p.p.s to within +/- 10-20 micron. Using the frog cutaneous pectoris neuromuscular preparation and a low calcium, high magnesium Ringer solution to ensure mostly single quantum events and failures, several thousand responses were recorded from each junction, allowing construction of a profile of the numbers of single quantum events arising from each portion of the junction. By comparison of junctional morphology and release profiles, it is possible to construct a probability of release per unit length profile for the entire junction. This technique has several advantages over localization of release events by measurements of extracellular synaptic currents. It was found that, for most junctions, the central 60-90% of the terminal exhibited relatively uniform probability of release, with highest levels typically near the point where the axon first contacted the muscle fibre, or in regions with many short terminal branches. However, no instances have been found in which a small region of terminal (10% or less) showed extraordinarily high release levels (30-50% of the total release from the junction). Characteristically, but not invariably, there is reduced release near the ends of terminal branches, especially the longer branches, where release per unit length could be as little as 5-10% of that in proximal portions. Some junctions had large regions of terminal that released very little transmitter. These also showed multiple myelineated axonal inputs, and may have been polyneuronally innervated junctions in which one of the inputs was much weaker than the other.