In the Nordic countries (like in most of Central and Eastern Europe) Didaktik has been the key to define effective instruction, not at least in the case of planning and evaluating teaching units (lessons). Besides German (Klafki) and US-American influences (Goodlad), the most used Scandinavian model is probably the ”relational Didaktik” as developed by Bjørndal and Lieberg in the 1970’s. This predominant position of Didaktik is facing two challenges which might alter the place of Didaktik fundamentally. On the one hand, the emerging emphasis on ”student achievement” as measured by tests moves the focus of quality assessment away from the actual conditions of teaching in a given environment. This de-coupling supports at the same time a move towards generalized models of evidence and / or research based teaching, whereby ”evidence” or ”research” means a proven record of enhancing student achievement as measured. Both moves may lead to a devaluation of situated or ”restrained” lesson-planning, i.e. the key asset of Didaktik as we know it, and thereby to a reductionist concept of teaching and learning. The paper will discuss this issue based on recent research on teaching and accountability in the Nordic countries and elsewhere.