Quantal content of transmission was estimated for three synaptic systems (crayfish and Drosophila neuromuscular junctions, and rat dentate gyrus neurons) with three different methods of measurement: direct counts of released quanta, amplitude measurements of evoked and spontaneous events, and charge measurements of evoked and spontaneous events. At the crayfish neuromuscular junction, comparison of the three methods showed that estimates from charge measurements were closer to estimates from direct counts, since amplitude measurements were more seriously affected by variable latency in evoked release of quantal units. Thus, charge measurements are better for estimating quantal content when direct counts cannot be made, as in crayfish at high frequency of stimulation or in the dentate gyrus neurons. At the Drosophila neuromuscular junction, there is almost no latency variation of quantal release in realistic physiological solutions, and the methods based upon amplitudes and charge give similar results. Distributions of evoked synaptic quantal events obtained by direct counts at the crayfish neuromuscular junction were compared to statistical distributions obtained by best fits. Binomial distributions with uniform or non-uniform probabilities of release generally provided good fits to the observations. From best fit distributions, the quantal parameters n (number of release sites) and p (their probability of release) can be calculated. We used two algorithms to estimate n and p: one allows for non-uniform probability of release and uses a modified chi-square (chi 2) criterion, and the second assumes uniform probability of release and derives parameters from maximum likelihood estimation (MLE). The bootstrap estimate of standard errors is used to determine the accuracy of n and p estimates.