An extensive series of simulations was performed of underground explosions in salt, using both chemical and nuclear explosives. In both cases, the radius of the initial emplacement cavity was varied from the fully tamped configuration to as large as 80 m/kt1/3; when not fully tamped, the cavity was assumed initially to contain air at ambient temperature and pressure. In the nuclear source case, the simulations are shown to be in good agreement with the Salmon/Sterling events conducted by the United States and with recently released Russian data on a similar pair of explosions in an Azgir salt dome. Simulation of the U.S. Cowboy series of chemical explosions in a Louisiana salt mine are also shown to be in very good agreement with the experimental data; however, the constitutive model for the salt that best explains these data is different from that derived for Salmon; both salt models are amply supported by laboratory and field data. The main result of these simulations is that cavity decoupling with chemical explosives is much less efficient than with nuclear explosives. Although maximum decoupling factors, fmax, near 200 may be attainable with either of the two sources, the cavity size required to achieve this value appears to be ≥40 m/kt1/3. For cavity radii half as large, fmax is roughly 4 times lower with nuclear explosives, and lower by another factor of 4 with chemical sources. Moreover, if the initial cavity radius is a more modest 10 m/kt1/3, fmax≤3 even with a nuclear source. ¿ American Geophysical Union 1994