Recent studies of the mechanism of quantal neurotransmitter release have assumed that the number of quanta released at each stimulation is binomially distributed and have sought to estimate the binomial parameters n and p. Mathematical analysis and computer simulations show that temporal variation in the number of eligible or filled release sites and either spatial or temporal variation in the probability of release at a site can drastically bias such estimates, while the experimental histograms remain statistically indistinguishable from those predicted by the binomial law. Interpretation of the estimates n and p in terms of ultrastructural or physiological characteristics of the nerve terminal is liable to significant error if departures from the binomial assumptions are not suitably assessed.