In the ideal world, teaching materials should be evaluated with reference to general principles of learning. When such principles are not available, teaching materials should be evaluated using controlled empirical studies. However, the design of the studies should be such as to render the principles more certain. This said, two questions follow for any domain of knowledge: where in the domain uncertainties lie, and by what empirical methods they should be resolved. The author asks these questions in relation to the explanatory concepts of physics. She starts by considering what might, given recent research, be taken as a principle, namely that 'Socratic dialogue' between pupils facilitates learning. Subjecting the supportive research to methodological review, the author not only debates the legitimacy of its claims. She also clarifies the methods by which some specific uncertainties might be further explored. In doing this, it draws attention to the long term nature of the evaluative strategy that its working assumption has led to, and asks whether the assumption should be suspended. Arguing that apparent shortcuts are illusory, the author concludes that it should not. (31 references)