The construction and validation of a theoretically based sign system for the analysis of teaching roles in childhood education is described. A theoretical and empirical approach to validation were developed. In the first, the general concept of teacher role was identified as a viable construct for investigating characteristic patterns of classroom teaching behavior. Four recognizable roles were derived: acquisition role mother-surrogate role, inquiry role, and authenticity role. After extensive field testing, and revision of the measurement instrument, 84 items were assumed to be congruent with the theory of teaching described by the role paradigm. The instrument was designated the acronym TRACE - Teaching Role Analysis in Childhood Education. Empirical validation was attempted by making observational records of the classroom behavior of teachers in selected Follow Through models which were tentatively identified as prototypes for three of the hypothetical roles, i.e., all except the mother-surrogate role. A seven-model sample was selected to provide empirical validation of both the theoretical framework and the discriminative power of the classroom observation system TRACE. For three of the specified teaching roles, the predicted mode of teaching emphasis emerged as predominant within the patterns of behavior characteristic for the groups of teachers in the first sample. For the mother-surrogate role group, the predominant mode of teaching behavior was maneuvers to induce Conformity and Compliance. Results of the data analysis indicate that the observation system TRACE has great potential for discriminating among teachers' classroom behaviors. Both theoretically and empirically, the observation system seemed to be a valid measure of the construct teaching role. (CK)