In this study a training program for new teaching assistants in freshman chemistry was developed and assessed. All new assistants in the Freshman Division were assigned by a stratified random technique to either the control or treatment group, with the latter receiving inservice training in the skills of teaching recitation classes. This training consisted of seminars, videotaped microteaching sessions, and inclass observations by and individual conferences with the author. The participants in both groups were observed before significant instruction of the treatment group and twice after the training program was concluded. An audio recording was made of each classroom observation and the verbal behaviors of teacher and students were encoded by the Flanders Interaction Analysis Category System and the Question Category System for Science. The data thus obtained were submitted to a two-way analysis of variance and a 90 X 90 Pearson product moment correlation matrix. A conservative interpretation allowed the following statements: (1) the training program appeared to provide assistants who were more successful in drawing students into the verbal activity of the classroom; (2) the program provided assistants who lectured less and used more techniques of praise or encouragement; and (3) the program provided assistants who asked a greater number of questions. (Author/MLH)