1. Measurements of membrane capacitance, as an indicator of exocytosis, and intracellular Ca2+ concentration ([Ca2+]i) were used to determine the Ca2+ dependence of secretion in single pancreatic B‐cells. 2. Exocytosis was dependent on a rise in [Ca2+]i and could be evoked by activation of voltage‐dependent Ca2+ currents. The threshold for depolarization‐induced release was 0.5 microM [Ca2+]i. Once the [Ca2+]i threshold was exceeded, exocytosis was rapidly (< 50 ms) initiated. When individual pulses were applied, exocytosis stopped immediately upon repolarization and the Ca2+ channels closed, although [Ca2+]i remained elevated for several seconds. 3. During repetitive stimulation (1 Hz), when [Ca2+]i attained micromolar levels, exocytosis also took place during the interpulse intervals albeit at a slower rate than during the depolarizations. 4. Exocytosis could be initiated by simulated action potentials. Whereas a single action potential only produced a small capacitance increase, and in some cells even failed to stimulate release, larger and more consistent responses were obtained with > or = four action potentials. 5. Comparison of the rates of exocytosis measured in response to depolarization, mobilization of Ca2+ from intracellular stores or infusion of Ca2+ through the patch pipette suggests that [Ca2+]i at the secretory sites attains a concentration of several micromolar. This is much higher than the average [Ca2+]i detected by microfluorimetry suggesting the existence of steep spatial gradients of [Ca2+]i within the B‐cell. 6. Inclusion of inhibitors of Ca2+/calmodulin‐dependent protein kinase II in the intracellular solution reduced the depolarization‐induced exocytotic responses suggesting this enzyme may be involved in the coupling between elevation of [Ca2+]i to stimulation of the secretory machinery. 7. The size of the unitary exocytotic event was 2 fF, corresponding to a secretory granule diameter of 250 nm. 8. Over short periods, exocytosis may be extremely fast (1 pF/s or 500 granules/s), which is much higher than the rate of endocytosis (18 fF/s or 9 granules/s). Since the latter is in better agreement with the maximum rate of insulin secretion from islets (approximately 2 granules/s), we suggest that membrane retrieval may set an upper limit on the rate of exocytosis during extended periods of secretion.