Little is known about how teachers construct tests. For that reason, this study addresses the use of teacher-constructed tests for assessing educational goals, expressed in terms of student mathematical competencies. The focus is on meanings that upper secondary school mathematics teachers assign to their own test construction practices for assessing educational goals, expressed in terms of mathematical competencies in the curriculum. The methodological approach of grounded theory, underlined by symbolic interactionism, is applied to semi-structured interviews with teachers. The core category, the emerging taxonomy, is derived by revealing distinctions in degree of paying attention to competencies: no attention, superficial attention, and qualitative attention, as well as two different phases of the assessment: constructional and marking. Finally, a couple of possible implications for developing and improving test construction are offered. This includes collaborative work, inside and outside of schools, with both prospective and in-service teachers, for improvement of competence implementation in regular teaching and learning in alignment with mathematical content.