This paper describes a method to determine the usefulness of seismological event identification criteria for underground test ban control with or without on-site inspection. The method rests upon the notion that such control should be deterrent, by confronting a prospective violator with a certain disclosure probability at a politically determined level. Simple decision theory is then applied to the statistical properties of identification measures, to find the required compromise between a sufficient probability to disclose explosions and a not too high incidence of false alarms about earthquakes. A clean separation between the political requirements and the seismological capabilities is obtained. The latter are expressed by identification curves similar to the receiver operating characteristics curves employed in telecommunications analysis. The political requirements appear as geometrical conditions on the identification curves, expressing the required deterrence, the number of yearly explosions and earthquakes to be considered and the permitted number and quality of inspections, in the case of control with inspection. In the case of control without inspection, the acceptable rate of false alarms is included. The method also shows how identification criteria are most efficiently exploited.
Application to some published observations shows identification by short-period body-wave magnitudes and long-period surface-wave magnitudes to be most efficient. It remains, however, to extend the analysis to weak events and low signal-to-noise ratios. For applications the acquisition of proper statistics is essential.