criminals in the course of their crimesrecall that in Wright and Rossi's sample of incarcerated felons, 48 percent of those who fired guns cited self-defense as a motive. Police report anecdotally hearing "It was him or me" as an increasingly common excuse offered by alleged youthful killers with guns.
Because self-defense is such an ambiguous term, it is not surprising that there is great dispute over how often firearms are actually used for self-defense. Using National Crime Survey data, Cook (1991) estimates incidents of self-defense firearm use at 78,000 per year, just below the number of people killed and wounded by firearms. Drawing on a number of surveys that ask about self-defense uses of guns without tying them to specific incidents, Kleck puts the annual number of self-defense uses about 10 times higherbetween 700,000 (G. Kleck, personal communication to Jeffrey A. Roth, National Research Council, 1990; Kleck, 1991) and 1 million (Kleck, 1988) per year, about equal to the number of violent crimes involving guns.
Part of the discrepancy is no doubt accounted for by methodology. The National Crime Survey excludes commercial robberies, and it undercounts attempted and completed rapes and sexual assaults, particularly by family members; self-defense gun uses in those incidents would be missed as well. The surveys reviewed by Kleck (1988, 1991) have smaller sample sizes and longer reference periods, and they leave definitions of self-defense and use to respondents. Besides the ambiguity in defining self-defense, the counts may be inflated by respondents' inadvertent telescoping of incidents of self-defense into the reference period from previous periods. Also, some reported instances of self-protection are not comparable to specific criminal victimizations: respondents in some surveys could well have been concerned about animal rather than human attackers; others may have simply brought the gun nearby in anticipation of an encounter that never occurred.5 To correct for these and various other methodological artifacts, Kleck uses a number of adjustment factors, some of which rely on untested assumptions. He presents no sensitivity analyses of alternative adjustment procedures, and some of them are incompletely documented. Our calculations indicate that making alternative plausible assumptions would substantially decrease Kleck's estimates of annual self-defense uses.6
Because of the likely undercounts in the National Crime Survey