A replication in three parts of a clinical experiment testing the effect of nursing interaction on pain is reported. Three treatment groups were designed according to categories of an interaction analysis scheme, the Nurse Orientation System. Patients who complained of pain were randomly assigned to one of the three approaches, and measurements of pulse, respiration, verbal and nonverbal behavior were taken at the beginning of the interaction, at the end, and about one hour later. Findings tended to be differentiated according to nursing approach with some variation across studies. Secondary analysis of the sample characteristics and of the conduct of the three studies provided information on the differences among studies. The use of interaction analysis categories seems to contribute methodological rigor to definitions of the independent variable, nursing approach. Suggestions for future studies are made, as well as suggestions for consideration of some issues in replication.