In making pulsed-neutron source measurements, counting rates of such magnitude are often encountered that the characteristics of the data-acquisition system must be properly identified before corrections for coincidence losses can be accurately made. To account properly for coincidence losses at very high counting rates, it is necessary to determine how closely a perfectly paralyzable or completely nonparalyzable system represents the real detection system used in the measurements. A maximum observed counting-rate technique is presented which, in conjunction with a double-pulse method, permits the system to be characterized relative to these two theoretically limiting models.