We know not only that a large percentage of criminal homicides are committed with a prior felony conviction, but that a sizable percentage of the victims have criminal records as well. Organized criminals kill competitors, and the most common target for "home invasions" are dwellings which the invaders believe to contain drugs and/or the proceeds their sale. If the target of an attempted hit or home invasion defends himself (and possibly also his loved ones), that's a legitimate act of self-defense. What makes a defensive gun use (DGU) illegal in such a context is if the defending arty possesses the firearm illegally, which is very likely.
Studies like Kleck & Gertz's 1993 "National Self-Defense Survey" (
http://www.saf.org/LawReviews/KleckAndGertz1.htm) have undoubtedly counted "illegal DGUs," in the sense that the respondent was illegally in possession of a firearm at the time of the incident. Note that in 1993, far fewer states had adopted "shall issue" laws, making CCW permits difficult to impossible to acquire even for citizens with perfectly clean backgrounds; thus, many of such DGUs would be legal now, seventeen years later. Some of the respondents were very possibly prohibited from possessing a firearm
at all. Nevertheless, those studies were designed to exclude responses that did not meet certain criteria for legitimate self-defense, one of which being that the the respondent had to be able to articulate (in some detail) the offense that he thought was being attempted against him (e.g. assault, robbery, confrontational burglary, homicide).
By making reference to studies like the Kleck-Gertz one immediately after his talk about self-defense using firearms being a "nebulous concept," Hemenway falsely suggests that the estimates produced by such studies include cases of criminals claiming that their use of a firearm in the commission of a crime was self-defense.
Interestingly, Hemenway himself (in collaboration with Deborah Azreal) published a study in 1999 which included the results of two telephone surveys of their own, conducted in 1994, which yielded an estimate of 900,000 DGUs annually--close to the number of incidents of violent crimes committed with firearms that year (~1,076,000). Somewhat predictably, Hemenway and Azreal spent the rest of the article trying to hand-wave away their own findings, but it does not speak well of their ability as researchers if they could not even compile a survey questionnaire that would exclude such instances.
So either Hemenway is an incompetent researcher (in which case one must take any critique on his part of others' research as being of dubious value), or, when he claims that "gun use in self-defense is rare," he uses a
very unusual and selective definition of "rare," by which 900,000 DGUs in a year is "rare" but 18,765 firearm suicides in that same year is not "rare."
Hemenway shows remarkable persistence in arguing why research data yielding estimates of 900,000-2.5 million DGUs a year must be flawed, but in over 15 years of doing so, he has produced only speculation; what he has
not produced is better research providing hard empirical evidence proving material like the Kleck-Gertz study to be wrong.