1. The statistical nature of transmitter release during facilitation was studied at single synaptic sites by recording extracellular excitatory junctional potentials from the claw opener muscle in crayfish.2. At low temperatures, single quanta could be counted in the responses to nerve impulses. The distribution of the number of quanta observed (x) was most accurately described by assuming that x is a binomial random variable.3. A quantitative estimate was made of the effects of errors in counting quanta due to the simultaneous release of quanta and the release of quanta which were not individually detectable above the noise of the recording system. Such errors of observation cannot account for the deviation of quantal release from a Poisson distribution.4. Facilitated release occurred in the responses to the second of two closely following nerve impulses and in the responses to successive impulses in a tetanus. In both cases, the increase in the average number of quanta released (m) could be attributed entirely to an increase in the probability (p) that available quanta were released.5. The results can be interpreted most easily in terms of a model in which the maximum number of releasable quanta is limited by a finite number of discrete release sites within recording distance of the microelectrode. In this model, the binomial parameter n is an estimate of the number of these sites, and the statistical parameter p is a compound probability depending on the rate of re-occupying sites after a nerve discharge and the probability that an impulse activates an occupied site.