SENTENCING DATA FOR BANK ROBBERY, BANK EMBEZZLEMENT, COUNTERFEITING, AUTO THEFT, LARCENY, NARCOTICS VIOLATIONS, MARIHUANA VIOLATIONS, AND SELECTIVE SERVICE OFFENSES ARE EXAMINED IN RELATION TO OFFENDER, OFFENSE, PROCESS, AND COURT-RELATED VARIABLES. THE SENTENCING DECISION ITSELF HAS TWO DISTINCT STEPS: THE JUDGE DECIDES WHETHER TO IMPRISON THE OFFENDER; IF SO, THE JUDGE THEN DETERMINES THE LENGTH OF THE PRISON TERM. THIS DISTINCTION IS IMPORTANT BECAUSE THE VARIABLES THAT BEST PREDICT THE INCARCERATION DECISION DIFFER FROM THE VARIABLES THAT BEST PREDICT LENGTH OF INCARCERATION. THE LENGTH OF THE PRISON TERM WAS FOUND TO BE MORE PREDICTABLE THAN WAS THE DECISION TO (OR NOT TO) INCARCERATE. IN DETERMINING THE PRISON TERM, THE TYPE OF OFFENSE WAS THE BEST PREDICTOR, WHILE THE METHOD OF PREDICTION (JURY TRIAL OR OTHER) WAS THE SECOND BEST PREDICTOR. HOWEVER, THE BEST PREDICTOR OF WHETHER OR NOT AN OFFENDER WOULD BE INCARCERATED WAS PRIOR CRIMINAL RECORD. METHOD OF CONVICTION AND TYPE OF OFFENSE WERE SECOND AND THIRD BEST PREDICTORS OF INCARCERATION, BUT THEY APPEAR TO ONLY MARGINALLY INFLUENCE THIS SENTENCING DECISION. SENTENCING DECISIONS WERE FOUND TO VARY SUBSTANTIALLY ACCORDING TO THE TYPE OF OFFENSE: DRUG OFFENSE SENTENCES WERE MOST PREDICTABLE, WHILE BANK ROBBERY, COUNTERFEITING, AUTO THEFT AND LARCENY SENTENCES FELL INTO A MIDDLE RANGE OF PREDICTABILITY; BANK EMBEZZLEMENT AND SELECTIVE SERVICE VIOLATIONS WERE THE LEAST PREDICTABLE. OVERALL, TYPE OF OFFENSE, PRIOR CRIMINAL RECORD, AND METHOD OF CONVICTION WERE THE BEST PREDICTORS OF SENTENCE. ALTHOUGH RACE WAS NOT SIGNIFICANT, SEX WAS WITH RESPECT TO THE IMPRISONMENT OF BANK ROBBERS. RESEARCH PERSPECTIVES AND NATIONAL LEVEL AGGREGATE ANALYSIS ARE DISCUSSED, ALONG WITH OFFENSE-SPECIFIC VARIABILITY IN SENTENCING FOR THE EIGHT FOCAL OFFENSES. DEFINITIONS OF INDEPENDENT VARIABLES ARE APPENDED. A USER EVALUATION QUESTIONNAIRE AND TABULAR AND GRAPHIC DATA ARE PROVIDED. SEE ALSO NCJ-33683, 33685, AND 33686. (AUTHOR ABSTRACT MODIFIED--KBL),