Secretion of catecholamines from neuroendocrine cells is relatively slow and it is likely that redistribution and buffering of Ca2+ is a major factor for delaying the response after a stimulus. In fact, in a recent study (Chow, R. H., J. Klingauf, and E. Neher. 1994. Time course of Ca2+ concentration triggering exocytosis in neuroendocrine cells. Proc. Natl. Acad. Sci. U.S.A. 91:12765-12769) Chow et al. concluded that the concentration of free calcium ([Ca2+]i) at a release site peaks at < 10 microM during short-step depolarizations, and then decays to baseline over tens of milliseconds. To check whether such a time course is consistent with diffusion theory, we modeled buffered diffusion in the vicinity of a Ca2+ channel pore. Peak [Ca2+]i and the slow decay were well simulated when release-ready granules were randomly distributed within a regular grid of Ca2+ channels with mean interchannel distances of 300-600 nm. For such large spacings, however, the initial rise in [Ca2+]i was underestimated, suggesting that a small fraction of the release-ready pool (approximately 10%) experiences much higher [Ca2+]i, and thus might be colocalized with Ca2+ channels. A model that accommodates these findings then correctly predicts many recent observations, including the result that single action potentials evoke near-synchronous transmitter release with low quantal yield, whereas trains of action potentials lead to desynchronized release, but with severalfold increased quantal yield. The simulations emphasize the role of Ca2+ not only in triggering, but also in modulating the secretory response: buffers are locally depleted by residual Ca2+ of a preceding stimulus, so that a second pulse leads to a larger peak [Ca2+]i at the fusion sites.