The problem in discovering the relationship between behavior and victimization is one of research design, as cross-sectional surveys of a single point in time can only examine the effect of past victimization on current behavior. Research and data could encompass the immediate context within which incidents occur and describe the routine activities of victims and nonvictims. While new opportunity or routine activity theories of victimization emphasize the importance of processes which bring together potential offenders with potential victims under facilitating circumstances, only sketchy data can be bound on the location of offenses, and no data will support how often victims and nonvictims are in those locations and under what facilitating circumstances. Also, little is known about the direct, individual benefits of crime prevention activities. The question of to what extent people's fates are in their own hands or how much of the variation in victimization can be explained by victims' routine activities or purposive actions remains unanswered. One standard by which behavior measures could be judged, however, refers to specific recall periods brief enough to promise accurate recall, where quantitative estimates of the incidence of the activity of interest should be gathered to maximize their utility to evaluators. Wherever possible, individual items measuring behavior should be validated against observations or other independent records of behavior to establish the margins of error with which they truly reflect the activity of interest. Multiple-component measures reduce the relative size of the error component, allow for the substitutability of related behaviors, and increase the generality of the analysis. Moreover, examination of the routine activity or conscious anticrime efforts on victimization requires long-term panel data on a sample of respondents. Footnotes are given. (Author abstract modified)