The dynamics of intramolecular energy transfer in vinyl bromide are investigated using projection methods and results obtained from classical trajectories. The method is based on the calculation of the temporal variation of a diagonal kinetic energy matrix. Energy transfer rates and pathways are extracted from the envelope functions of this temporal variation. Average mode energies are also obtained from these data using the virial theorem. Total energy decay rates and the pathways of energy flow for initial excitation of each of the 12 vibrational modes in the equilibrium configuration and in initial configurations corresponding to points in the near vicinity of the minumum-energy structure on the optimum dividing surfaces for three-center H-2 and HBr elimination are reported. The results for the equilibrium structure show that the total relaxation rate for each mode can be characterized with fair accuracy by a first-order rate law. The minimum decay rate among the 12 modes is at least 3.1 times larger than the decomposition rate of vinyl bromide with 6.44 eV of excitation energy present. However, it is also found that energy transfer is not globally rapid. In configurations near the minumum-energy structure on the optimum dividing surface for three-center H-2 elimination, the intramolecular energy transfer rate for some mode-to-mode processes is slower than the unimolecular dissociation rate. In contrast, energy transfer in configurations near the minimum-energy structure on the optimum dividing surface for three-center HBr elimination is globally rapid relative to the HBr elimination rate for all modes except the C-C-Br bend. The energy transfer dynamics for the system therefore suggest that three-center HBr elimination may be accurately described by statistical theories, but the corresponding three-center H-2 elimination reaction will probably behave nonstatistically. These results again demonstrate that the existence of a total intramolecular vibrational relaxation rate that is fast relative to the unimolecular reaction rate may not be a sufficient condition to guarantee statistical behavior.