A LARGE SERIES OF BLOOD GROUP PROTEIN SYSTEMS WAS ANALYZED AS PART OF A PROJECT FUNDED BY THE CALIFORNIA OFFICE OF CRIMINAL JUSTICE PLANNING. ONE PHASE OF THIS RESEARCH ATTEMPTED TO USE PHENOTYPE DATA TO DISCRIMINATE AMONG FOUR ETHNIC GROUPS (WHITE, BLACK, CHICANO AND AMERINDIAN, AND ASIAN). TWO PROBABILITY CALCULATIONS WERE CONSIDERED: (1) THE PROBABILITY THAT TWO RANDOMLY SELECTED INDIVIDUALS WILL MATCH WITH RESPECT TO A SET OF PHENOTYPES, USING A LINEAR DISCRIMINANT FUNCTION, AND (2) THE PROBABILITY THAT A SINGLE INDIVIDUAL RANDOMLY SELECTED FROM SOME POPULATION WILL MATCH A SET OF PREDETERMINED PHENOTYPES (BAYES' THEOREM). THE STATISTICAL RATIONALE FOR EACH METHOD IS DISCUSSED IN DETAIL. BOTH METHODS WERE THEN TESTED BY ASSIGNING 5,043 SAMPLES OF MALE BLOOD FROM VARIOUS BLOOD BANKS IN CALIFORNIA AND HAWAII TO ONE OF THE FOUR RACIAL TYPES. THE MEN HAD DECLARED THEIR RACIAL CLASSIFICATION ON THE DONOR RECORDS.) BOTH BAYES' THEOREM AND THE LINEAR DISCRIMINANT FUNCTION GAVE PRACTICALLY THE SAME RESULTS, ALTHOUGH THE ROUTES WERE ENTIRELY DIFFERENT. CHARTS GIVE THE PREDICTED AND THE OBSERVED CLASSIFICATIONS FOR BOTH METHODS. EACH SYSTEM HAD AN ERROR PROBABILITY OF ABOUT 22 PERCENT. HOWEVER, THE CALIFORNIA POPULATION IS 75.9 PERCENT WHITE: DECLARING ALL INDIVIDUALS AS WHITE WOULD YIELD A MISCLASSIFICATION PROBABILITY OF 0.241. THUS, IT IS CLEAR THAT, FOR THE 11 BLOOD PROTEIN SYSTEMS EXAMINED, THERE IS NOT SUFFICIENT RACIAL HETEROGENEITY FOR EITHER METHOD TO TYPE INDIVIDUALS HIGHLY ACCURATELY. SUCH CLASSIFICATIONS ARE MORE USEFUL WHEN A PHENOTYPE OCCURS MAINLY IN ONE POPULATION, SUCH AS SICKLE CELL TRAITS IN BLACKS. THE DISCUSSION CONTAINS BOTH TABLES AND REFERENCES. (GLR)