The program targets prosecutorial attention on serious, repeat offenders. The main objectives are early identification, vertical prosecution and reduced caseloads, reduced plea barganing, increased convictions, and appropriate prison sentences. The MVU in San Diego focuses only on habitual robbery offenders. A standard form is used for rating all robbery defendants. Robbery and robbery-related homicide cases are prosecuted by the MVU if (1) the suspect is under arrest for committing three or more separate robbery offenses or (2) the suspect is under arrest for robbery and within the last 10 years has been convicted of one or more of a series of specified offenses. In addition, prosecutors may accept cases involving great bodily harm or a defendant with several prior felony convictions. Other responsibilities of MVU prosecutors include attending bail hearings to ensure that an appropriate bail is set, maintaining contact with witnesses, and handling pretrial motions. MVU adheres to a policy called descriptive pleading designed to tighten the bargaining process. It requires defendants to plead to the top felony count or more than one count. The report briefly recounts traditional felony processing in San Diego County. In addition, it discusses data collected routinely by the MVU research analyst and the formal evaluation performed. The MVU's outcomes are contrasted with those of other career criminal programs. Finally, some of the issues that must be considered by potential replicators of a career criminal program are raised: unit composition, selection criteria, and variations in operating procedures. The report concludes with a summary of the feasibility of MVU replication and of career criminal prosecution programs, generally. Tables, footnotes, and charts are provided. Appendixes present information on PROMIS (Prosecutor Management Information System) and career criminal programs, information available to the MVU via the Justice Records Information System, general characteristics of MVU defendants, evaluation methodology, and other related data. (Author abstract modified)