Establishing the reliability of interaction analysis

This study developed a method for the determination of the reliability of Flanders' Interaction Analysis which preserves the sequential nature of the data. The reliability is determined by use of a computer program which produces a point by point matching of the original observations and identifies unmatched points as to type of error. The reliability coefficient is reported as a decimal which represents the ratio of matched points to total points. This method of reliability determination, the Data Analysis Method, as well as those of Bales, Scott, and Darwin, was then used for the determination of intra- and inter-observer reliability. Three experienced observers and three trainees independently recorded interaction data from the same audio tape on three different days. Each observer was then tested for intra-observer reliability by each of the four coefficients. All four methods correlated well and indicated acceptable intra-observer reliability for all trained observers and for two of the three trainees. ..