Covert checks of observational methodology reveal declines in reliability of observations. This appears to be particularly true when complex codes are used to track social interaction. The present study was undertaken to see whether reliability could be maintained through a combination of technological advancements and the development of improved observer training and maintenance procedures. Eight people were enrolled in a highly refined, 150-hour training program developed at the University of Oregon's Child Research Laboratory. Using automated data collectors, the trainees learned to code the interactions of young children. Following completion of the training program, observers remained in close contact with the project director through weekly meetings, allowing for discussion of unusual circumstances, performance feedback, and refinement of procedures. Subsequent covert checks failed to indicate any decline in agreement when scores were compared with data from overt checks. While agreement was high under all circumstances, this does not mean that complex coding systems should be applied in all observational studies. However, findings do suggest that when an appreciation of the subtleties inherent in social process are sought, complex systems are essential. (Author/RH)