Seven mature women were recruited to be trained in a family interaction code and collect data for a research project evaluating a program of child behavior therapy. Observations were made in the homes of normal families and families participating in the therapy program. Two of the observers were kept uninformed of the normal vs. deviant andbaseline vs. treatment status of the families to be observed. The remaining five observers were fully informed regarding the status of the families. Comparisons of the reports of the informed and uninformed observers were examined under two conditions: overt and covert monitoring of observer reliability. No significant interactions were obtained under either condition, suggesting that observer bias of large magnitude was not present. However, the informed observers reported significantly higher rates of deviant behavior than uninformed observers in the overt monitoring condition independent of the selection or treatment status of the families. The possibility that information about a study sensitizes observers to the variables involved was discussed. Both comparisons were limited by the small N and a possible selection confound.