STATES AND LOCALITIES ARE RESPONSIBLE FOR AVOIDING INCORRECT PAYMENTS AND PAYMENTS TO INELIGIBLE PERSONS. AVERAGE ERROR RATES HAVE FALLEN FROM 33.5 PERCENT OF CASES IN THE FIRST HALF OF 1974 TO 22.4 PERCENT IN THE FIRST HALF OF 1976. ALTHOUGH ERROR PROBLEMS IN LARGE URBAN CENTERS ARE ABOUT 20 PERCENT HIGHER THAN IN NONURBAN AREAS, URBAN AREAS HAVE REDUCED THEIR ERROR RATES AS RAPIDLY AS HAVE NONURBAN AREAS. THE MOST PROMISING MEANS FOR REDUCING ERRORS ARE (1) ELIMINATING OVERDUE REDETERMINATIONS OF ELIGIBILITY AND PAYMENT LEVELS; (2) REWRITING DOCUMENTS AND FORMS TO REDUCE CLIENT ERRORS DUE TO THE WRITTEN MATERIALS' COMPLEXITY; (3) RAISING THE SKILL LEVEL OF ELIGIBILITY WORKERS THROUGH FORMAL TRAINING, IMPROVED SALARIES, AND PROMOTIONAL OPPORTUNITIES; (4) SIMPLIFYING RULES TO STREAMLINE ADMINISTRATION; AND (5) DEVELOPING SYSTEMS FOR FLAGGING CASES MOST SUBJECT TO ERRORS. DATA FROM STATES WHICH HAVE IMPLEMENTED ONE OR MORE OF THESE APPROACHES INDICATE THAT THEIR USE WOULD REDUCE THE NATIONAL ERROR RATE BY 50 PERCENT. CONVERSION OF STATE SUPERVISED PROGRAMS TO STATE ADMINISTERED PROGRAMS WOULD REDUCE THE ERROR RATE BY ANOTHER 3 PERCENT. CONTINUED PARTIAL STATE FUNDING ALONG WITH IMPROVED QUALITY CONTROL AUDITS WOULD ALSO REDUCE ERROR RATES. FINALLY, THE FEDERAL GOVENRMENT SHOULD UNDERTAKE A MORE COMPREHENSIVE ROLE IN REDUCING ERROR RATES BY GIVING SPECIFIC GUIDANCE TO STATES ON QUALITY CONTROL AND CORRECTIVE ACTIONS AND BY DEVELOPING TOOLS FOR ACHIEVING THESE IMPROVEMENTS. TWENTY SEVEN RECOMMENDED ACTIONS FOR REDUCING AFDC ERRORS ARE NOTED. FOOTNOTES, GRAPHS, TABULAR DATA, AND A BIBLIOGRAPHY ARE INCLUDED. APPENDIXES CONTAIN RESEARCH METHODS AND SUPPLEMENTARY DATA. (CFW)