The SNARE hypothesis holds that a transport vesicle chooses its target for fusion when a soluble NSF attachment protein (SNAP) receptor on the vesicle (v-SNARE) pairs with its cognate t-SNARE at the target membrane. Three synaptosomal membrane proteins have previously been identified: syntaxin, SNAP-25 (t-SNAREs), and vesicle-associated membrane protein (VAMP) (v-SNARE); all assemble with SNAPs and NSF into 20S fusion particles. We now report that in the absence of SNAP and NSF, these three SNAREs form a stable complex that can also bind synaptotagmin. Synaptotagmin is displaced by alpha-SNAP, suggesting that these two proteins share binding sites on the SNARE complex and implying that synaptotagmin operates as a "clamp" to prevent fusion from proceeding in the absence of a signal. The alpha-SNAP-SNARE complex can bind NSF, and NSF-dependent hydrolysis of ATP dissociates the complex, separating syntaxin, SNAP-25, and VAMP. ATP hydrolysis by NSF may provide motion to initiate bilayer fusion.