Rubrics are popular tools to assess students' social-emotional skills. However, little research has been devoted to investigating whether rubrics' performance level descriptions validly represent the intended skills. In this study, we addressed the validity of Self-management rubrics by examining the presence of construct-irrelevant variance in their scores. To this end, we decomposed rubrics into single items, and those indicating low versus high levels of a construct were used to compute an acquiescence index to control for response sets. The sample consisted of 947 middle-school students (54% female; M[subscript age] = 13.8; SD = 0.89). This innovative methodological approach allowed us to correct rubrics' scores for acquiescence and demonstrate its biasing effects on rubrics' structural psychometric properties. After correction, reliabilities were preserved, and the underlying factor structure was more unidimensional. The opportunities and implications of the use of decomposed rubrics for the process of rubric development are broadly discussed.