Part of the problem with using 'scientific evidence' is that it mistakes efficiency for learning. Nick Shackleton-Jones makes this point quite nicely: "If someone wants to pass a test, the most efficient process is to give them the answers so that they can copy them into the paper." This is essentially what direct instruction amounts to, and so it's not surprising that the evidence people like Paul Kirschner cite reach this conclusion. The other side of the coin, though: "So what is it then – if it’s not storing facts in your head? (This question will not be on the test)." That's what connectivism seeks to answer.