THE RELATIVE EFFECTIVENESS OF SIX MODES OF TRAINING TEACHERS TO USE PROBING QUESTIONS WAS INVESTIGATED. THE MODES INVOLVED SYMBOLIC MODELING, PERCEPTUAL MODELING, OR BOTH, COUPLED WITH FEEDBACK. AFTER RATINGS OF PERTINENT BEHAVIOR IN A 5-MINUTE LESSON WERE COLLECTED AS PRETRAINING MEASURES, STANFORD TEACHER INTERNS WERE RANDOMLY DISTRIBUTED AMONG 6 TRAINING GROUPS, WHOSE EXPECTED EFFECTIVENESS, IN INCREASING ORDER, WAS--(1) MINIMUM SYMBOLIC MODELING (SAW PRETEST VIDEOTAPE PLAYBACK ALONE, STUDIED WRITTEN INSTRUCTIONS, PLANNED, AND RETAUGHT), (2) MAXIMUM SYMBOLIC MODELING (SAW PLAYBACK WITH EXPERIMENTER WHO GAVE CUES AND REINFORCEMENT, STUDIED WRITTEN INSTRUCTIONS, PLANNED, AND RETAUGHT), (3) MINIMUM PERCEPTUAL MODELING (SAW PLAYBACK AND PERCEPTUAL MODEL ALONE, PLANNED, AND RETAUGHT), (4) STRONG SYMBOLIC AND MAXIMUM PERCEPTUAL MODELING (SAME AS 2 BUT ALSO VIEWED PERCEPTUAL MODEL ALONE), (5) MAXIMUM PERCEPTUAL MODELING (VIEWED PLAYBACK ALONE BUT SAW PERCEPTUAL MODEL WITH EXPERIMENTER), (6) STRONG SYMBOLIC AND MAXIMUM PERCEPTUAL MODELING (SAW PLAYBACK AND PERCEPTUAL MODEL WITH EXPERIMENTER). EACH TAPE WAS RATED FOR RELEVANT BEHAVIOR BY 2 TRAINED RATERS. THESE EXPECTATIONS WERE LARGELY CONFIRMED, (EXCEPT MODE 4 DID BETTER THAN MODE 3) WITH PERCEPTUAL MODELING APPEARING SUPERIOR TO SYMBOLIC. (AF)