1. A study has been made of the effect of blocking endogenous adenosine on the statistics of quantal secretion at nerve terminals in toad (Bufo marinus) muscle during summer and winter. 2. Exogenous adenosine (10-50 microM) reduces the mean quantal content of the endplate potential (EPP) recorded with an intracellular microelectrode (m) by 36 +/- 6% (mean +/- S.E.M.), independent of the control value of m in both summer and winter. The variance of the EPP (S2) was reduced by adenosine in proportion to m, so that the probability of quantal secretion (p) remained relatively constant. 3. Exogenous adenosine reduces the mean quantal content of secretion recorded with an extracellular electrode (m(e)) to a similar extent (47 +/- 6%) at different relatively high secreting sites along nerve terminal branches in both summer and winter. 4. Both theophylline (20-100 microM) and adenosine deaminase (2.5 i.u./ml) increase the amplitude of the EPP in summer in a [Ca2+]o of 0.5 mM or greater; a maximum increase of about 40% is reached at a [Ca2+]o of 1.2 mM. The amplitude of the EPP in winter is usually reduced by theophylline in a [Ca2+]o of 0.5 mM, but is always increased in a [Ca2+]o greater than 0.9 mM to reach a maximum increase of about 40% at high [Ca2+]o. 5. The variance of the EPP (S2) was always increased by theophylline to a greater extent than m in summer, so that p decreased and Poisson rather than binomial statistics could be used to describe the distribution of EPP amplitudes. In winter, theophylline generally increased m and S2 to about the same extent, so that p did not change much. 6. An autocorrelation analysis of the amplitude of successive EPPs in a long train at 0.5 Hz in high [Ca2+]o showed that these are likely to be independently distributed. Adenosine secreted by one impulse is then unlikely to affect secretion by a subsequent impulse in the train. 7. These observations are discussed in terms of the hypothesis that endogenously secreted adenosine at a release site inhibits secretion by nearby release sites in summer; this has the effect of reducing m and to a greater extent S2 so that binomial rather than Poisson statistics can describe the frequency distribution of EPP amplitudes.