Membrane fusion is broadly envisioned as a two- or three-step process proceeding from contacting bilayers through one or two semistable, nonlamellar lipidic intermediate structures to a fusion pore. A true fusion event requires mixing of contents between compartments and is monitored by the movement of soluble molecules between trapped compartments. We have used poly(ethylene glycol) (PEG) to rapidly generate an ensemble aggregated state A that proceeds sequentially through intermediates (I₁ and/or I₂) to a final fusion pore state (FP) with rate constants k₁, k₂, and k₃. Movement of moderately sized solutes (e.g., Tb³⁺/dipicolinic acid) has been used to detect pores assigned to intermediate states as well as to the final state (FP). Analysis of ensemble kinetic data has required that mixing of contents occurs with defined probabilities (αi) in each ensemble state, although it is unclear whether pores that form in different states are different. We introduce here a simple new assay that employs fluorescence resonance energy transfer (FRET) between a 6-carboxyfluorescein (donor) and tetramethylrhodamine (acceptor), which are covalently attached to complementary sequences of 10 bp oligonucleotides. Complementary sequences of fluorophore-labeled oligonucleotides were incorporated in vesicles separately, and the level of FRET increased in a simple exponential fashion during PEG-mediated fusion. The resulting rate constant corresponded closely to the slow rate constant of FP formation (k₃) derived from small molecule assays. Additionally, the total extent of oligonucleotide mixing corresponded to the fraction of content mixing that occurred in state FP in the small molecule assay. The results show that both large "final pores" and small (presumably transient) pores can form between vesicles throughout the fusion process. The implications of this result for the mechanism of membrane fusion are discussed.