A new algorithm has been developed for determining the cavity size distribution in liquids. The computational method is based on energetic rather than geometric considerations. It is applicable to any liquid structure, including polymers, that can be simulated by molecular dynamic or Monte Carlo methods. To illustrate its utility, it has been applied to hard sphere (HS) and Lennard-Jones (LJ) fluids, SPC/E water, as well as to the two isomeric polyimides. To verify its ability to locate cavities and to determine cavity size frequency, it has also been applied with success to crystalline systems. Significant differences in liquid structure can be detected among HS, LJ, and SPC/E liquids, particularly at low densities where LJ and SPC/E liquids show evidence of "clustering." The supercooled LJ liquid clearly reveals the development of a solid phase by the preferential appearance of tetrahedral and octahedral cavities. Cavitation in water under negative pressures call be detected. Differences in cavity size distributions are seen in two isomeric polyimides of identical chemical composition. Molecular dynamic simulations of carbon dioxide diffusion in these polyimides confirms that diffusion is faster in the isomer with the larger average cavity size.