Particular emphasis is placed on the accuracy of determining whether the respondent had been a victim or not. Information is presented about the amount, direction, and correlates of error in regard to victim recall of the details of the crime. This analysis of the accuracy of victim recall is accomplished through an examination of an LEAA-funded program that compared victimization survey data with police records of the same victimization. The characteristics of the victim in relation to the absolute amount and direction of error are analyzed, as well as the implications of data distortion in relation to the length of the period of victim recall. The article concludes that victimization surveys using short recall periods (i.e., 6 months), no boundary for the elimination of telescoping, and no special memory-aids other than usual screening questions may contain considerable error as to which respondents have been victims during the reference period. Also, theories of assaultive violence may be more difficult to support from the data than are theories of property offenses. However, information about victimization details, such as those used to classify the offense and calculate its seriousness, tend to be accurate. Moreover, less serious offenses tend to be subject to more extreme patterns of memory decay; information about offenders may contain considerable error, thereby requiring a large sample size for significant results; and the amount of time lapse between the victimological incident and the interview of the victim is not related to the amount or direction of error in the data. Numerous notes, 20 tables, 34 references and an appendix presenting the seriousness scale used in the analysis are provided.