This paper reports upon the evaluation and implementation of a 30-item Likert-type rating scale for teachers to use in assessing children's behaviors within preschool classrooms. The Preschool Observation Scale (POS) was developed to evaluate programs of the Mt. Druitt Early Childhood Project, North Ryde, Australia. Items were constructed on the program-related, socio-emotional variables of creativity, dependency, language, prosocial behaviors, need achievement and anxiety. Reliability and construct validity tests of the scale are described. Inter-rater reliability coefficient estimates, stability coefficient estimates, and factor analyses of the pilot test indicate that the instrument has construct validity and is reliable. Use of the instrument in the large scale 1977 assessment of the classroom behavior of 259 children attending 4 project preschools is described and results are reported and discussed. Language, need-achievement, anxiety and prosocial factors were conceptually consistent in ratings across all 4 sites. Dependence and creativity did not have the same consistency. Need achievement correlated positively with the language dimension and negatively with anxiety type behaviors. Inter-rater reliability decreased as time between ratings increased. The paper suggests that the POS is a useful device for monitoring selected child behaviors. Directions for further research are indicated. (RH)