To estimate the frequency of firearm retrieval because of a known or presumed intruder, the authors analyzed data from a 1994 national random digit dialing telephone survey (n = 5,238 interviews). Three mutually exclusive definitions of firearm retrieval were constructed: (1) retrieved a firearm because there might be an intruder, (2) retrieved a firearm and saw an intruder, and (3) retrieved a firearm, saw an intruder, and believed the intruder was frightened away by the gun. Of 1,678 (34%) households with firearms, 105 (6%) retrieved a firearm in the previous 12 months because of an intruder. National projections based on these self-reports reveal an estimated 1,896,842 (95% CI [confidence interval] = 1,480,647-2,313,035) incidents in which a firearm was retrieved, but no intruder was seen; 503,481 (95% CI = 305,093-701,870) incidents occurred in which an intruder was seen, and 497,646 (95% CI = 266,060-729,231) incidents occurred in which the intruder was seen and reportedly scared away by the firearm. Estimates of the protective use of firearms are sensitive to the definitions used. Researchers should carefully consider both how these events are defined and the study methods used.