This research note addresses the accuracy of arrestees' self-reports of contacts with the criminal justice system as a means of exploring the relative importance of various sources of inaccurate responding. Erroneous self-report of sensitive behaviors has been linked to deception, memory problems, and faulty criterion measures, among other things. However, the existent literature provides limited guidance for investigating the relative importance of these factors in a given study. Further, variations in the amount and types of inaccuracy cannot be distinguished by commonly used summary agreement statistics, such as kappa. These issues were examined using data from the Policing Project, a National Institute of Justice-funded research study designed to explore new means of evaluating police behavior. The project interviewed 892 New York City arrestees during the second half of 1999. Subjects were asked about several forms of criminal justice system contact, and gave informed consent for researchers to obtain their official criminal histories, which were acquired from the state agency as an anonymous data set. A key finding was that the accuracy of arrestee self-reports compared to official criminal histories varied according to specific measures. Agreement regarding arrest in the prior six months was substantial, but other measures were less accurate. Overreporting was about equal to underreporting of criminal justice contacts across several measures. We conclude that arrestee self-reports continue to be valuable for criminological research. While arrestee self-reports may lack the precision and accuracy that criminal justice practitioners might prefer, the limitations of official records contribute substantially to inaccuracies between self-reports and criminal histories.