THREE SAMPLES OF 100 PROTOCOLS EACH WERE DRAWN FROM A MMPI DATA POOL CONSISTING OF PROFILES FOR 2,500 OFFENDERS, 18 TO 27 YEARS OLD, AT THE TALLAHASSEE (FLA.) FEDERAL CORRECTIONAL INSTITUTION. IN PHASE 1, EACH SAMPLE WAS SUBJECTED TO A HIERACHICAL PROFILE ANALYSIS USING VELDMAN'S (1967) PROGRAM H-GROUP. THIS RESULTED IN OPTIMAL SUBGROUPINGS RANGING FROM 10 TO 14 IN NUMBER. ELIMINATION OF SUBGROUPS CONTAINING FIVE OR LESS SUBJECTS RESULTED IN EIGHT TO NINE SUBGROUPS IN EACH SAMPLE. IN PHASE 2, THE SUBGROUPS IN THE FIRST SAMPLE WERE MATCHED AND RULES WERE FORMULATED TO CHARACTERIZE EACH PROFILE TYPE. THE SUBGROUPS IN THE SECOND AND THIRD SAMPLES WERE THEN MATCHED WITH THOSE IN THE FIRST SAMPLE. TO EXAMINE IF INDIVIDUAL, AS OPPOSED TO GROUP PROFILES, COULD ALSO BE RELIABLY CLASSIFIED USING THE PREVIOUSLY ESTABLISHED RULES, AN ADDITIONAL 85 PROFILES WERE SORTED INDEPENDENTLY BY 2 CLINICIANS. THERE WAS AN 87 PERCENT AGREEMENT BETWEEN THE TWO SORTERS AS TO THE CORRECT CLASSIFICATION, SUGGESTING THAT THE RULES COULD RELIABLY BE USED TO DISCRIMINATE BETWEEN INDIVIDUALS. RATHER THAN PROVIDE OPERATIONAL DEFINITIONS, SUBGROUPS WERE ALPHABETICALLY LABELED AT THIS POINT. THIS WAS DONE TO AVOID PREMATURE CHARACTERIZATION WITH THE HOPE THAT THE ATTRIBUTES OF EACH SUBGROUP WOULD BECOME APPARENT DURING LATER RESEARCH. THESE PRELIMINARY RESULTS INDICATED THE POTENTIAL FOR AN MMPI-BASED TAXONOMY. REFERENCES ARE PROVIDED. FURTHER RESEARCH IS REPORTED IN NCJ 47695-47696 AND NCJ 47698-47700. (JAP)