Introduction: In the Achievement Goal Theory, different reasons for learning are contrasted. Mastery and Performance are most commonly used while less attention is paid to Social Responsibility goal orientations, despite that this is an integral part of many curricula. Aims: The purpose of this paper is to examine the psychometric properties of a scale of Social Responsibility goal orientation by means of the polytomous Rasch model. Mehods: This study is based on data from the longitudinal ETF project (Evaluation Through Follow up) among Swedish students. Two cohorts (born in 1992 and 1998) of students in school year 9 were subjected to analysis. Data was collected by a self-administered postal questionnaires to randomly selected students. In total, 6,010 students responded in 1992 and 4,573 in 1998. A scale consisting of 6 polytomous items is analysed. General fit statistics as well as their graphical representations (ICC) are used to evaluate the fit to the Rasch model. Strategies to handle large samples in statistical test of fit are discussed.   Results: The social responsibility scale seems to fit the Rasch model fairly well, with good separation of individuals, and showing no reversed item thresholds, i.e. the response categories work as intended. The estimates are also relatively stable over cohorts. However, there are indications of Differential Item Functioning (DIF) by sex.    Conclusions: Although, the Social Responsibility scale fits the Rasch model fairly well, there are room for improvements. In particular, targeting may be improved by the inclusion of additional items of appropriate difficulty.