Concerns about nuclear proliferation and negotiations toward a Comprehensive Test Ban Treaty (CTBT) emphasize the need for robust seismic monitoring and regional discrimination capabilities, especially at low magnitudes. Standard regional discrimination techniques have many shortcomings. For example, they cannot always discriminate shallow, low-magnitude earthquakes or mining-related events. I show that slopes of P- to S-wave spectral ratios can be used to discriminate shallow- and normal-depth earthquakes from explosions and possibly to discriminate explosions from mining-related rockburst or collapse events. P- to S-wave slopes and conventional, high-frequency P- to S-wave spectral amplitude ratios discriminate even better in combination than alone. A comparison of Pn and Lg spectra from Nevada Test Site (NTS) explosions and earthquakes shows that this method discriminates concentrated explosions from earthquakes and other shallow seismic events such as mine collapses because S waves from explosions decay more rapidly at high frequencies than do S waves from earthquakes. Spectra of one-dimensional reflectivity synthetic seismograms show that frequency-dependent attenuation of explosion-generated P- to S-wave converted phases, in the near-surface layers, may be responsible for the differences in Lg (S-wave) spectral falloff. These simulations also show that absolute amplitudes of earthquake P- to S-wave spectral ratios are very sensitive to mechanism and depth, which may help explain the variable performance of conventional methods. ¿ American Geophysical Union 1995