When the nerves of isolated frog sartorius muscles were stimulated at 10 Hz, synaptic vesicles in the motor nerve terminals became transiently depleted. This depletion apparently resulted from a redistribution rather than disappearance of synaptic vesicle membrane, since the total amount of membrane comprising these nerve terminals remained constant during stimulation. At 1 min of stimulation, the 30% depletion in synaptic vesicle membrane was nearly balanced by an increase in plasma membrane, suggesting that vesicle membrane rapidly moved to the surface as it might if vesicles released their content of transmitter by exocytosis. After 15 min of stimulation, the 60% depletion of synaptic vesicle membrane was largely balanced by the appearance of numerous irregular membrane-walled cisternae inside the terminals, suggesting that vesicle membrane was retrieved from the surface as cisternae. When muscles were rested after 15 min of stimulation, cisternae disappeared and synaptic vesicles reappeared, suggesting that cisternae divided to form new synaptic vesicles so that the original vesicle membrane was now recycled into new synaptic vesicles. When muscles were soaked in horseradish peroxidase (HRP), this tracerfirst entered the cisternae which formed during stimulation and then entered a large proportion of the synaptic vesicles which reappeared during rest, strengthening the idea that synaptic vesicle membrane added to the surface was retrieved as cisternae which subsequently divided to form new vesicles. When muscles containing HRP in synaptic vesicles were washed to remove extracellular HRP and restimulated, HRP disappeared from vesicles without appearing in the new cisternae formed during the second stimulation, confirming that a one-way recycling of synaptic membrane, from the surface through cisternae to new vesicles, was occurring. Coated vesicles apparently represented the actual mechanism for retrieval of synaptic vesicle membrane from the plasma membrane, because during nerve stimulation they proliferated at regions of the nerve terminals covered by Schwann processes, took up peroxidase, and appeared in various stages of coalescence with cisternae. In contrast, synaptic vesicles did not appear to return directly from the surface to form cisternae, and cisternae themselves never appeared directly connected to the surface. Thus, during stimulation the intracellular compartments of this synapse change shape and take up extracellular protein in a manner which indicates that synaptic vesicle membrane added to the surface during exocytosis is retrieved by coated vesicles and recycled into new synaptic vesicles by way of intermediate cisternae.