THE FIRST ANALYTICAL MODELS FOR CRIMINAL JUSTICE SYSTEMS APPEARED IN THE PRESIDENTIAL COMMISSION'S TASK FORCE REPORT ON SCIENCE AND TECHNOLOGY, ISSUED IN 1967. THESE MODELS AND THE ONES WHICH FOLLOWED WERE BASICALLY SIMPLIFIED EMPIRICAL MODELS AND, AS SUCH, DID NOT CAPTURE THE COMPLEXITIES OF THE CRIMINAL JUSTICE PROCESS. THE AVI-ITZHAK AND SHINNAR MODEL WAS THE FIRST TO POSSESS RECOGNIZABLE POLICY VARIABLES, BUT IT DID NOT DIFFERENTIATE BETWEEN THE TREATMENT OF CLASSES OF OFFENDERS. THE MODEL FORMULATED BY BLUMSTEIN AND NAGIN, BASED ON AVI-ITZHAK AND SHINNAR'S WORK, WAS UNABLE TO DETERMINE AN OPTIMAL POLICY OVER A PLANNING HORIZON. SIMULATION MODELS FOCUSING ON THE ENTIRE CRIMINAL JUSTICE SYSTEM--THE COURTSIM MODELS--, JUSSIM AND JUSSIM II, AND DOTSIM, ATTEMPT TO TRACK AN OFFENDER THROUGH THE SYSTEM AND DETERMINE THE EFFECTS OF VARIOUS ALTERNATIVES. THESE MODELS ARE LIMITED BY THE AMOUNT OF DATA REQUIRED AND THE COST OF RUNNING THEM. HOWEVER, SIMULATIONS ARE MORE USEFUL THAN ANALYTICAL MODELS FOR POLICY STUDIES. THE GENERALIZED NETWORK SIMULATOR IS RECOMMENDED AS AN EASILY MODIFIED SIMULATION MODEL WHICH COMBINES THE RESOURCE ALLOCATION, QUEUEING, AND COSTING FEATURES NECESSARY FOR DISCRETE EVENT SIMULATION. REFERENCES ARE APPENDED. (GLR)