THIS EXPERIMENT WAS BASED ON 2 HYPOTHESES--(1) INTERNS WHO LEARN AND APPLY FLANDERS' INTERACTION ANALYSIS WILL BECOME MORE INDIRECT IN THEIR VERBAL CLASSROOM BEHAVIOR THAN INTERNS NOT LEARNING THIS TECHNIQUE. (2) THERE IS A RELATION BETWEEN INTERNS' PERCEPTIONS OF THEIR TEACHING PROBLEMS AND THEIR CLASSROOM VERBAL BEHAVIOR. TWO EXPERIMENTAL GROUPS OF 12 INTERN TEACHERS EACH WERE TRAINED IN INTERACTION ANALYSIS, WHILE A CONTROL GROUP OF 12 WAS NOT. ALL INTERNS TAUGHT AT THE ELEMENTARY LEVEL, HALF OF EACH GROUP IN "MIDDLE-CLASS" SCHOOLS AND HALF IN "UNDERPRIVILEGED" SCHOOLS. DURING THE FALL SEMESTER, INTERNS WORKED 2 TO A BUILDING UNDER A COORDINATING TEACHER. AN OBSERVATION TEAM (PROGRAM DIRECTOR, COLLEGE-CITY SUPERVISOR, GRADUATE ASSISTANT) VISITED 24 EXPERIMENTAL GROUP CLASSROOMS FOR 4 COMPLETE DAYS (12 IN THE FALL, 12 IN THE SPRING) AND 12 CONTROL GROUP CLASSROOMS FOR 2 COMPLETE DAYS (IN THE SPRING) TO RECORD VERBAL BEHAVIOR IN TERMS OF INTERACTION OBSERVED. RESULTS WERE REPORTED IN 4 AREAS--(A) READING, (B) SOCIAL STUDIES, (C) LANGUAGE ARTS, AND (D) TOTAL TALK. ANALYSIS OF VARIANCE WAS USED TO TEST THE FIRST HYPOTHESIS AND A Q-SORT TO TEST THE SECOND HYPOTHESIS. EXPERIMENTAL DATA DID NOT SUPPORT THE HYPOTHESES. A DISCUSSION OF RESULTS AND SUGGESTIONS FOR INTERN TRAINING ARE INCLUDED. (AF)