You propose a shift from traditional university lectures to a system that teaches the methods of scientific enquiry to students of science, technology, engineering and mathematics (STEM; see Nature 523, 272–274 and 282–284; 2015). This move has clear merits, but systematic transfer of the requisite knowledge should not be abandoned entirely.

A pioneer in active-learning practices, Roskilde University in Denmark has been using problem-based teaching and successfully involving students in research since 1972. Our experience shows, however, that there are potential pitfalls.

Unless critical thinking is allied with a strong fundamental knowledge base, there is a risk that students will develop a 'Swiss cheese' understanding of science — with a good grasp of their chosen subject areas but major gaps in others. This can produce niche researchers who lack a proper understanding of their wider field.

We therefore advise retaining aspects of traditional education in an appropriate balance, which is then adjusted on the basis of student and course evaluations.