The data of two experiments of dyadic group discussion have been reanalysed. An extended proportional change model was designed to explain the actual process of attitude change. The model is defined by two parameters. The first represents the impact of single pro-arguments and single con-arguments on the attitude or decision preference. The second describes the resistance to further change that increases with the distance from the initial position. It was hypothesized that the first parameter should be higher and the second lower, with a similar partner than with a dissimilar one. The prediction was confirmed for the first parameter only. A comparison of the extended proportional change model to related models concludes the report.