1. The activation of elementary calcium release events ('puffs') and their co-ordination to generate calcium waves was studied in Xenopus oocytes by confocal linescan imaging together with photorelease of inositol 1,4,5-trisphosphate (InsP3) from a caged precursor. 2. Weak photolysis flashes evoked no responses or isolated calcium puffs, whereas flashes of increasing strength evoked more frequent puffs, often occurring in flurries as abortive waves, and then a near-simultaneous calcium liberation originating at multiple sites. The numbers of sites activated increased initially as about the fourth power of photoreleased [InsP3]. 3. Following repeated, identical photolysis flashes, puffs arose after stochastically varying latencies of a few hundred milliseconds to several seconds. The cumulative number of events initially increased as about the third power of time. No rise in free [Ca2+] was detected preceding the puffs, suggesting that this co-operativity arises through binding of multiple InsP3 molecules, rather than through calcium feedback. 4. The mean latency to onset of calcium liberation shortened as about the square of the flash strength, and the dispersion in latencies between events reduced correspondingly. 5. Weak stimuli often evoked coupled puffs involving adjacent sites, and stronger flashes evoked saltatory calcium waves, propagating with non-constant velocity. During waves, [Ca2+] rose slowly between puff sites, but more abruptly at active sites following an initial diffusive rise in calcium. 6. Initial rates of rise of local [Ca2+] at release sites were similar during puffs and release induced by much (> 10-fold) greater [InsP3]. In contrast, macroscopic calcium measurements averaged over the scan line showed a graded dependence of rate of calcium liberation upon [InsP3], due to recruitment of additional sites and decreasing dispersion in activation latencies. 7. We conclude that the initiation of calcium liberation depends co-operatively upon [InsP3] whereas the subsequent regenerative increase in calcium flux depends upon local calcium feedback and is largely independent of [InsP3]. Wave propagation is consistent with the diffusive spread of calcium evoking regenerative liberation at heterogeneous discrete sites, the sensitivity of which is primed by InsP3.