One of the main advantages of the naturalistic approach to classroom research is that the observations are of natural phenomena, occuring in circumstances which the researcher ultimately hopes to generalize. Some problem areas are associated with this approach, however. Observers must obtain complete and accurate records of classroom activities and teacher and student behavior. They must also maintain objectivity in their narrative records. The difficulty also exists that the presence of the observer may intrude or interfere with normal classroom functioning. This paper presents some of the considerations made and techniques used to select and train observers for a major naturalistic classroom observation study, the Junior High Classroom Organization Study. Priorities used for selecting observers are described, including classroom experience, either as a teacher or observer, and writing skills. A description is given of the activities which were engaged in during the training week. It is reported that each of the three problem areas that are associated with naturalistic studies was dealt with and some success was achieved in ameliorating them. It is pointed out that it is necessary to choose observers carefully, train them extensively, and supervise them closely. (JD)