Explosions can be differentiated from natural earthquakes easily and efficiently using S/P amplitude ratios, owing to the fact that shear‐dislocation source radiates S‐wave energy and explosions do not. However, explosion seismograms often contain S waves generated by either a secondary source signal after the explosion or scattering during propagation. This sometimes leads to failure in seismic discrimination using the average S/P ratio discriminant. In this study, I develop a new seismic discrimination method based on the azimuthal distribution of S/P amplitude. The Fourier spectra of the azimuthal distribution of S/P amplitude ratios show strong 90° periodicity for strike‐slip natural earthquakes, unlike those of explosion events. I generate a natural earthquake model by averaging the spectra of azimuthal S/P ratios measured from natural earthquake seismograms. The relative errors of individual event spectra are estimated, and the error is used as a discriminant. The criterion for the discriminant is empirically determined. The results show that over 90% of natural earthquakes and 80% of explosions are properly discriminated. The method is also verified for small earthquakes. Sensitivity tests for the number of S/P measurements per event and the azimuthal coverage are performed, indicating that 17 S/P measurements and an azimuthal coverage of 250° are sufficient for successful discrimination. I consider this method applicable for most earthquakes occurring in inland areas of the southern Korean Peninsula. By comparing with traditional average S/P ratio discriminant, it is also found that azimuthal ratio discriminant is less sensitive to event magnitude than that of average ratio.