In March 1989 the Maricopa County Attorney's Office (Phoenix, Ariz.) began a program of pre-file deferred prosecution to treatment for felony drug possession cases. Qualifications for deferred prosecution varied somewhat by the drug possessed, but their general purpose was to offer diversion to only casual users with no prior criminal history. Eligible offenders could reject the diversion program, but those who entered the program had to submit to random drug tests, attend all required meetings or counseling sessions, and pay both a Drug Enforcement Fund Fee and the costs for their treatment. A sliding schedule of fees was used for low-income offenders. During the first 2 years of the diversion option, 7,012 arrests that alleged felony drug possession were reviewed for prosecution by the County Attorney's Office. Of these, 1,275 were deemed to be "not prosecutable," typically due to a low likelihood of conviction or questions concerning search-and-seizure violations. This analysis focuses on 5,554 prosecutable cases that involved Anglo, African- American, and Hispanic defendants. Of these cases, deferred prosecution (diversion to treatment) was offered in 4,026 cases. For the remaining 1,528 cases, the prosecuting attorney filed felony charges. The dependent variable in the analysis was the prosecuting attorney's decision to defer prosecution. The analysis includes defendant characteristics, offense characteristics, and process decisions that may influence prosecutorial discretion to defer analysis. The findings show that partial support for hypotheses derived from the theoretical perspectives pursued. Male and older defendants, as well as defendants with a prior record, were less likely to be diverted from prosecution and into treatment. Also, consistent with the uncertainty avoidance perspective, the study found that persons with multiple charges had a decreased likelihood of diversion; whereas, persons charged with possession of marijuana had an increased likelihood of diversion from prosecution; however, the data do not support the hypothesis that defendant's minority status directly influences the diversion decision. In addition, these findings point to a more complex model of the subjective nature of the exercise of prosecutorial discretion, a model that benefits from understanding the salience of minimizing uncertainty in the decision to criminals. 6 tables and 42 references