the firearms safety programs we discuss has actually utilized this method of evaluation, however, usually because of policy regulations at schools prohibiting even disabled firearms on campus. Nonetheless, direct observation may be the most accurate method of discerning what a child or adolescent would do when confronted with a firearm. Researchers who have directly observed children’s behaviors around firearms following an intervention have found high rates of gun play (see Hardy et al., 1996; Hardy, 2002b).
The best evaluation of a firearm violence prevention program should assess its impact on knowledge, attitudes, and behavior from a variety of sources, particularly since these variables are not highly correlated. Inconsistencies between children’s knowledge and behavior following participation in more general violence prevention programs is well documented (Arcus, 1995). Moreover, Wilson-Brewer and colleagues (1991) found in a survey of 51 programs that fewer than half claimed to reduce actual violence levels. Those that did claim to do so had limited empirical data to support their claims.
The correlation between children’s knowledge about guns and the likelihood that they will handle a gun is less well studied. However, a recent study by Hardy (2002b) suggests that the two outcomes following a firearm violence prevention program are unrelated. In this study, 70 children ages 4 to 7 were observed in a structured play setting in which they had access to a semiautomatic pistol. Observers coded several behaviors, including gun safety statements (“Don’t touch that!”) and gun touching. Assuming that children who say “Don’t touch that gun!” to another child have some knowledge that guns are dangerous (or for some other reason should not be touched), one might expect that these children would themselves not touch the guns. Nonetheless, 15 of the 24 children who made such comments in the study subsequently touched the gun themselves during the 10-minute interval.
Another way Hardy (2002b) assessed the correlation between firearms safety knowledge and behavior was to examine the relationship between a child’s belief that a gun is real and his or her behavior around that gun. Again, however, the evidence suggests no significant relationship. Specifically, the children who correctly identified the real gun as such were no less likely to play with the gun (n = 19) than were children who believed the gun was a toy (n = 16). These findings were later replicated in a study with children ages 9 to 15 (Hardy, 2002a).
Once the appropriate outcome measures are identified and operationally defined, program developers must decide on the design of the evalua-