There are similarities and differences between chemical and nuclear explosions underground. Host of the differences are in the early stages of the explosions. The later stages are similar with respect to seismic wave generation. Three sources of seismic waves from explosions are coincident in space and time, or nearly so: the explosion itself, explosion-induced tectonic strain release, and (probably) spall-closure following explosion-produced spall. Cavity collapse and explosion-induced aftershocks are two sources of delayed seismic signals. Theories, computer calcula tions, and measurements of spherical stress waves from explosions are described and compared, with emphasis on the transition from inelastic to almost-elastic relations between stress and strain. Two aspects of nonspherical explosion geometry are considered: tectonic strain release and surface spall. Tectonic strain release affects the generation of surface waves; spall closure may also. The forward problem in seismology can, in principle, be solved by calculations beginning with explosive detonation and ending with the synthetic seismogram. The inverse problem can also, in principle, be solved by inverting observed seismic data to obtain an "equivalent elastic source," but the solution cannot extend backward in space and time into the nonlinear inelastic processes of the explosion. The reduced-displacement potential is a common solution (the "equivalent elastic source") of the forward and inverse problems, assuming a spherical source. Measured reduceddisplacement potentials are compared with potentials calculated as solutions of the direct and inverse problems; there are signifi cant differences between the results of the two types of calcula tions and between calculations and measurements. The simple