We have examined broadband regional waveforms of recent (since 1988) Nevada Test Site (NTS) underground explosions and earthquakes throughout the southwestern United States and Baja, Mexico, recorded by TERRAscope and other IRIS stations in order to characterize seismic sources for the purposes of event identification. As expected, earthquakes tended to be richer in long-period surface-wave and short-period shear-wave energy relative to explosions of comparable P-wave strength. Also, explosions, in general, were found to be richer in 1- to 6-sec surface-wave (Rg) energy and other late-arriving coda energy than were earthquakes. Most earthquakes show relatively little long-period (T > 6 sec) Rg and surface-wave coda energy, which we attribute to their deeper source depths, whereas known shallow earthquakes do exhibit these phases. We have developed several seismic discriminants based on our observations. The most promising discriminant is the ratio of short-period (f ≧ 1.0 Hz), vertical component, Pnl wave-train energy (EspPz) to long-period (0.05 to 0.167 Hz), three-component, surface-wave energy (Elp−3). For this ratio, explosions tend to have a higher value than do earthquakes. This discriminant works on the same premise as the teleseismic mb:MS ratio, for which earthquakes are richer in long-period surface-wave energy relative to explosions with the same body-wave magnitude. The long-period passband was chosen to limit the effect of longer-period noise and to remove the effect of the coda surface waves. Another potential discriminant examined is the ratio of short-period (f ≧ 1.0 Hz), vertical-component, P-wave to S-wave energy (EspPz:EspSz). We find that this criterion only yields marginal separation of the source populations but becomes more effective at higher frequency bands (f ≧ 4.0 Hz) or when looking at single-station observations. It does, however, help to quantify significant short-period waveform differences between the three test subsites, with Pahute Mesa shots generating relatively little S-wave energy compared to those of Yucca Flat for which the S wave (or Lg) is often the largest phase, while Rainier's shots are intermediate in character with distinct but less prominent S waves. This S-wave generation is thought to be caused by near-source scattering to converted phases and appears to be highly dependent on the near-source geology. These two discriminants are useful in that they are simple and fast to calculate. Using regional stations for sources 200 to 1300 km away, the magnitude threshold for the EspPz:Elp−3 discriminant is roughly ML ≧ 4.0, the limiting factor being the signal level of the Airy phase, while that for the EspPz:EspSz discriminant is roughly ML ≧ 3.0 for the same distance ranges.