1. Transmitter release was analysed at frog neuromuscular junctions pre-treated with 400 mM glycerol Ringer. In the absence of added drugs, end-plate potentials (e.p.p.s) and miniature e.p.p.s (m.e.p.p.s) could be recorded at selected junctions. 2. E.p.p.s were unusually large and calculations of quantal content indicated a high level of release. Also recorded were anomalous action potentials resembling e.p.p.s but these could be distinguished using a summation test. 3. Plots of coefficient of variation of e.p.p.s (0-5 Hz stimulation) versus direct quantal content (M1) showed a marked deviation from Poisson expectations with high M1. Analysis of these results with amplitude-frequency histograms showed a progressively better fit to binomial predictions with increasing M1. 4. The use of binomial statistics allowed direct calculations of the mean probability of release (p) and the readily available store (n). Increasing Ca/Mg caused increases in both n and p. 5. Plots of M1 vs. Ca/Mg showed a power relationship of 3.58. Maximum m.e.p.p. amplitude occurred at control Ca/Mg. Both results were consistent with studies in muscles not treated with glycerol and indicated that glycerol treatment caused no major alterations pre- or post-junctionally. 6. Estimates of quantum size using Poisson assumptions showed an over-all increase when stimulus frequency was raised, indicating a shift from binomial to a Poisson distribution. 7. The combined findings demonstrate that the glycerol treated preparation can be used to examine the release process during high output. Such release conforms to binomial statistics and allows direct determinations of the parameters n and p.