EMPHASIS IS ON RANDOMIZATION IN A 'TRICKLE-PROCESSED' SETTING WHERE UNITS DEEMED ELIGIBLE FOR TREATMENT TRICKLE IN, I.E., APPEAR AT ANY TIME. THE U.S. COURT OF APPEALS IS EXPERIMENTING WITH A PROCEDURE DESIGNED TO SETTLE OR OTHERWISE IMPROVE THE QUALITY OF SELECTED CIVIL CASES AT THE APPELLATE LEVEL. THE PROCEDURE REPRESENTS A COMBINATION OF TREATMENTS. ONE TREATMENT IS A HEARING IN WHICH ADVERSARIES IN SELECTED CASES MEET TO DISCUSS THEIR APPEALS. ANOTHER TREATMENT IS THE COURT ORDER ISSUED IN THESE SELECTED CASES IN WHICH THE COURT OFFICER SETS DEADLINES FOR CERTAIN STAGES IN THE APPELLATE PROCESS. VARIOUS EVALUATION DESIGNS WERE CONSIDERED BY THE COURT TO DETERMINE THE EFFECTIVENESS OF THE PROGRAM. IN THE END, THE COURT AGREED TO ASSIGN A TOTAL OF 300 CASES MEETING PROGRAM REQUIREMENTS TO A TREATMENT GROUP AND A CONTROL GROUP IN A RANDOM FASHION. A BLANK LIST WAS DRAWN AND NUMBERED IN ORDER FROM 1 TO 300. EACH BLANK LINE WAS DESIGNATED AS TREATMENT OR CONTROL. WHEN THE COURT OFFICER DEEMED A CASE ELIGIBLE FOR THE PROGRAM, IT WAS PLACED ON THE NUMBERED LIST. THE EVALUATION OFFICE WAS THEN CALLED AND RELEVANT CASE INFORMATION AND ENTRY DATA WERE ENTERED ON A DUPLICATE LIST IN THAT OFFICE. THE DUPLICATE LIST PROCEDURE FOR RANDOMLY ASSIGNING TREATMENTS IN FIELD EXPERIMENTS IS SUFFICIENTLY FLEXIBLE TO MEET ANY EXIGENCY WITHOUT SACRIFICING HIGH STANDARDS OF RESEARCH DESIGN. REFERENCES ARE CITED. (DEP)