Ca2+ liberation by inositol 1,4,5-trisphosphate (IP3) is 'quantal', in that low [IP3] causes only partial Ca2+ release, but further increasing [IP3] evokes more release. This characteristic allows cells to generate graded Ca2+ signals, but is unexpected, given the regenerative nature of Ca2+-induced Ca2+ release through IP3 receptors. Two models have been proposed to resolve this paradox: (i) all-or-none Ca2+ release from heterogeneous stores that empty at varying [IP3]; and (ii) phasic liberation from homogeneously sensitive stores. To discriminate between these hypotheses, we imaged subcellular Ca2+ puffs evoked by IP3 in Xenopus oocytes where release sites were functionally uncoupled using EGTA. Puffs were little changed by 300 mu M intracellular EGTA, but sites operated autonomously and did not propagate waves. Photoreleased IP3 generated flurries of puffs-different to the prolonged Ca2+ elevation following waves in control cells-and individual sites responded repeatedly to successive increments of [IP3]. These data support the second hypothesis while refuting the first, and suggest that local Ca2+ signals exhibit rapid adaptation, different to the slower inhibition following global Ca2+ waves.