1. The elementary release events underlying inositol 1,4, 5-trisphosphate (InsP3)-mediated calcium signalling were investigated in Xenopus oocytes by means of high-resolution confocal linescan imaging together with flash photolysis of caged InsP3. 2. Weak photolysis flashes evoked localized, transient calcium signals that arose at specific sites following random latencies of up to several seconds. The duration, spatial spread and amplitude of these elementary events varied widely. Event durations (at half-maximal amplitude) were distributed exponentially between about 100 and 600 ms. Fluorescence magnitudes (F/F0 of Oregon Green 488 BAPTA-1) showed a skewed distribution with a peak at about 1.5 and a tail extending as high as 3.5. 3. Individual release sites exhibited both small events (blips) and large events (puffs). The spatiotemporal distribution of calcium signals during puffs was consistent with calcium diffusion from a point source (< a few hundred nanometres), rather than with propagation of a microscopic calcium wave. 4. Estimates of the calcium flux associated with individual events were made by integrating fluorescence profiles along the scan line in three dimensions to derive the 'signal mass' at each time point. The smallest resolved events corresponded to liberation of < 2 x 10-20 mol Ca2+, and large events to about 2 x 10-18 mol Ca2+. The rise of signal mass was more prolonged than that of the fluorescence intensity, suggesting that calcium liberation persists even while the fluorescence begins to decline. Rates of rise of signal mass corresponded to Ca2+ currents of 0.4-2.5 pA. 5. Measurements of signal mass from different events showed a continuous, exponential distribution, arising through variability in magnitude and duration of calcium flux. 6. We conclude that localized calcium transients in the oocyte represent a continuum of events involving widely varying amounts of calcium liberation, rather than falling into separate populations of 'fundamental' and 'elementary' events (blips and puffs) involving, respectively, single and multiple InsP3 receptor channels. This variability probably arises through stochastic variation in both the number of channels recruited and the duration of channel opening.