Puffs are localized, transient elevations in cytosolic Ca(2+) that serve both as the building blocks of global cellular Ca(2+) signals and as local signals in their own right. They arise from clustered inositol 1,4,5-trisphosphate receptor/channels (IP3Rs), whose openings are coordinated by Ca(2+)-induced Ca(2+) release (CICR). We utilized total internal reflection fluorescence imaging of Ca(2+) signals in neuroblastoma cells with single-channel resolution to elucidate the mechanisms determining the triggering, amplitudes, kinetics, and spatial spread of puffs. We find that any given channel in a cluster has a mean probability of ∼66% of opening following opening of an initial "trigger" channel, and the probability of puff triggering thus increases steeply with increasing number of channels in a cluster (cluster size). Mean puff amplitudes scale with cluster size, but individual amplitudes vary widely, even at sites of similar cluster size, displaying similar proportions of events involving any given number of the channels in the cluster. Stochastic variation in numbers of Ca(2+)-inhibited IP3Rs likely contributes to the variability of amplitudes of repeated puffs at a site but the amplitudes of successive puffs were uncorrelated, even though we observed statistical correlations between interpuff intervals and puff amplitudes. Initial puffs evoked following photorelease of IP3-which would not be subject to earlier Ca(2+)-inhibition-also showed wide variability, indicating that mechanisms such as stochastic variation in IP3 binding and channel recruitment by CICR further determine puff amplitudes. The mean termination time of puffs lengthened with increasing puff amplitude size, consistent with independent closings of channels after a given mean open time, but we found no correlation of termination time with cluster size independent of puff amplitude. The spatial extent of puffs increased with their amplitude, and puffs of similar size were of similar width, independent of cluster size.