In many types of central neurons, the coefficient of variation (CV) of stimulus-evoked uniquantal events inferred from quantal analysis is small, frequently less than 20%. In contrast, spontaneous putative uniquantal events (minis) from the same neurons are much more variable in amplitude, having a CV of roughly 50% or more. One explanation for this discrepancy is that, if the variance in mini amplitude were generated by differences between release sites, the small number of sites activated during stimulation would sometimes fortuitously have similar quantal amplitudes. Only in these fortuitous cases where uniquantal variance is small could quantal peaks be resolved, and therefore the uniquantal CV seen in the subset of cells where quantal analysis can be performed would systematically be much smaller than predicted by the mini distribution. We have explored this possibility by Monte Carlo simulation assuming all variance in mini amplitude to be due to intersite differences in uniquantal amplitude. We find that when a small number of release sites are activated under these conditions, there is a reduction in the expected uniquantal variance. However, the expected uniquantal CV is highly variable from one experiment to the next, and low uniquantal CVs are not expected to be seen often enough to account for the high frequency with which quantal peaks with a uniquantal CV < 20% are observed experimentally. We conclude that variance in mini amplitude between release sites cannot account for the small uniquantal CV seen in quantal analysis of many central synapses.