The purpose of this study was to analyse the predictive power of responsibility on the index of psychological mediators, this on the autonomous motivation, and this on the school climate and the perception of school violence through the Self-Determination Theory. A second objective was to verify the possible differences that may exist according to gender and educational stage. To this end, data were collected by means of self-reported questionnaires in different Secondary and Primary Education centres in Physical Education classes, with a sample of 1007 students from Primary (n = 294), Secondary (n = 601) and Baccalaureate and training cycles (n = 112), with a mean age of M = 13.30 (SD = 2.86). Of the participants, 53.6% were girls (n = 540) and 46.4% boys (n = 467). The data were analysed with IBM SPSS 25.0 and AMOS 23.0 software, using structural equation analysis. The results verified the predictive ability of accountability on the index of psychological mediators, autonomous motivation and on school climate and perception of school violence. The analysis of factorial invariance of the model confirmed its robustness to gender, but not to educational stage. According to educational stage, the analysis of variance showed significant differences between Primary School and Secondary School and Baccalaureate/Training, with younger students showing higher values for all variables except violence. It is concluded that responsibility has a relevant role in the educational environment, due to its predictive power on the school climate and school violence, as well as on the satisfaction of basic psychological needs and autonomous motivation, following the Self-Determination Theory. At the same time, students in higher educational stages presented lower values of responsibility, making it necessary to work on it in the classroom. It is recommended that future research should investigate, on the one hand, interventions where the central core is the promotion of responsibility, in addition to conducting longitudinal analyses with samples of students in different contexts, checking whether the model of this study is corroborated and allows creating a line of work for the improvement of school climate and the reduction of the perception of violence.