A basic statistical method is presented for the classification of observations into one of two multivariate, normal populations. This theory is applied to the identification of seismic events by considering one of the “populations” to consist of measurements on seismic recordings of earthquakes, the other of explosions. The measurements, or parameters, consist of ratios of the “energies” contained within predetermined “velocity windows” on the seismograms. The choice of velocity windows is guided by the assumption that earthquake source mechanism is extended both in time and space and generates a larger fraction of energy in shear waves as compared to explosion source mechanism. The best separation of twenty earthquakes and twenty-seven explosions is achieved when only seven of the nine ratios calculated are used in the statistical discriminator. Based on this data we have about 85% probability of correctly classifying a given event either as an explosion or as an earthquake.