The survey used structured interviews to determine the attitudes and practices of approximately 60 judges and other officials involved in the pretrial release process in Tuscon, Ariz., San Jose, Calif., Baltimore, and Miami. Judges and commissioners in three of the jurisdictions also provided recommendations on simulated defendants. The case summaries of the 102 defendants were developed from official records of actual cases. Data analysis showed that two of the jurisdictions had low rates of decisions to rerelease bail offenders as before and low rates of multiple pretrial arrests. The other two jurisdictions had high rates of both phenomena. The decisions to rerelease offenders as before resulted mainly from gaps in information regarding the offender's status. The criminal justice system needs to develop Automated Law Enforcement Release/Rearrest Tracking (ALERT) systems that will provide pending release and case information in time for its use in decisionmaking. High-risk defendants need additional supervision. The system should respond effectively to the first rearrest rather than increasing detention following the initial arrest. Twenty-two additional recommendations, data tables, footnotes, appended study instruments and detailed studies of each site. (Author abstract modified)