OF APPROXIMATELY 18,000 STUDENTS CATEGORIZED AT TIME 1 OF THE STUDY (1971) AS ELIGIBLE FOR FOLLOWUP, SOME 10,000 WERE LOST TO THE STUDY BY TIME 2 (1973). THIS CONSTITUTED A FAIRLY HIGH ATTRITION RATE OF 55 PERCENT. THE PERSONAL CHARACTERISTICS AND REPORTED DRUG BEHAVIOR, PARTICULARLY MARIHUANA USE, OF STUDENTS WHO WERE LOST TO THE STUDY AFTER TIME 1 ARE COMPARED WIH THOSE OF STUDENTS WHO REMAINED IN THE STUDY. REPORTED DRUG USE BY STUDENTS REMAINING AT TIME 2 IS COMPARED WITH REPORTED DRUG USE BY STUDENTS NEW TO THE STUDY AT TIME 2. ATTRITION IS TRACED TO THREE PRINCIPAL SOURCES: (1) SCHOOL TRANSFERS OR DROPOUTS AFTER TIME 1; (2) TIME 2 ABSENTEES WHO HAD PARTICIPATED AT TIME 1; AND (3) FAILURE OF SOME STUDENTS TO PRODUCE MATCHING CODE NUMBERS. BECAUSE OF THE RESEARCHERS' GUARANTEE OF ANONYMITY, IT WAS IMPOSSIBLE TO TRACE STUDENTS WHO LEFT SCHOOLS OR TO IDENTIFY ABSENTEES. IT WAS IMPOSSIBLE ALSO TO IDENTIFY THE DEMOGRAPHIC CHARACTERISTICS OF ABSENTEES IN ORDER TO CORRECT FOR THEIR LOSS. REMAINING STUDENTS AND LOST CASES DIFFERED SIGNIFICANTLY ON A NUMBER OF CHARACTERISTICS: AGE, ATTITUDES TOWARD SCHOOL; BEHAVIOR IN SCHOOL; ATTITUDES TOWARD DRUGS; AND, ESPECIALLY, REPORTED USE OF DRUGS. LOST CASES WERE MUCH MORE LIKELY THAN REMAINING SUBJECTS TO HAVE REPORTED DRUG USE. THE FINDINGS, TOGETHER WITH A REVIEW OF OTHER RESEARCHERS' EXPERIENCES WITH SAMPLE ATTRITION, SUGGEST THAT THE LENGTH OF THE INTERVAL BETWEEN DATA COLLECTIONS IS NOT IN ITSELF SUFFICIENT TO EXPLAIN SAMPLE LOSS. THE POSSIBILITY EXISTS THAT PANEL LOSSES WILL BE REDUCED IF AND WHEN ILLICIT DRUG USE COMES TO BE SEEN AS LESS DEVIANT. INVESTIGATORS ARE URGED TO REPORT MORE FULLY ON THE CAUSES AND MAGNITUDE OF SAMPLE LOSSES, PARTICULARLY ON THE CHARACTERISTICS OF LOST CASES AND THE DRUG USERS AMONG THEM. SUPPORTING DATA AND A LIST OF REFERENCES ARE INCLUDED. (LKM)