REFORM WITHIN THE BAIL SYSTEM HAS RECEIVED MUCH ATTENTION IN RECENT YEARS. MUCH OF THIS ATTENTION HAS BEEN FOCUSED ON THE SUBSTITUTION OF INTERVIEWS DESIGNED TO DETERMINE AN OFFENDER'S LIKELIHOOD OF RETURNING FOR TRIAL FOR THE TRADITIONAL SETTING OF BAIL. AFTER A BRIEF OVERVIEW OF THE EXISTING BAIL SYSTEM IN THE UNITED STATES AND AN EXAMINATION OF SOME OF THE PROBLEMS WITH CURRENT BAIL PRACTICES, THE AUTHOR DESCRIBES THE PROCEDURES AND PHILOSOPHY OF THE MANHATTAN BAIL PROJECT. THE VERA HYPOTHESIS USED BY THE PROJECT STATES THAT RETURN FOR TRIAL IS DEPENDENT ON THE OFFENDER'S COMMUNITY STABILITY. THUS, THE PROJECT USED SOCIOECONOMIC DATA GATHERED FROM INTERVIEWS TO DETERMINE AND RATE THE COMMUNITY STABILITY OF OFFENDERS, AND USED THIS DATA TO MAKE ITS RELEASE DECISIONS. TO TEST THIS, THE AUTHOR EXAMINED A BASE OF DATA COLLECTED ON 519 PRISONERS SUBJECTED TO A PRETRIAL RELEASE PROGRAM, STATISTICALLY SEPARATING THE RELEASED AND THE NON-RELEASED. THE CLUSTER OF VARIABLES WHICH BEST DISCRIMINATES BETWEEN THE RELEASED AND THE NON-RELEASED ARE IDENTIFIED, AND VARIABLES WITHIN THE CLUSTER ARE RANKED FROM MOST TO LEAST IN TERMS OF INDICATING STABILITY. AN OPTIMAL FORMULA FOR CLASSIFYING INDIVIDUALS IS DETERMINED AND APPLIED TO EACH PRISONER IN THE DATA BASE. THE PERCENTAGE OF PRISONERS CORRECTLY CLASSIFIED IS USED TO INDICATE THE POWER OF THE VARIABLES TO DISCRIMINATE BETWEEN THE RELEASED AND THE NON-RELEASED. THE RESULTS SHOWED THAT ACTUAL USE OF THE VARIABLES TO CLASSIFY PRISONERS WITHIN THE DATA BASE WAS 76.11 PERCENT EFFECTIVE. (AUTHOR ABSTRACT MODIFIED)...DMC