Academic papers report language-learning experimental results using methods like those described here. Often they take a class of students (say, 30) and randomly assign some task for one half, and the other half remains unchanged. I worry such experiments give highly correlated experimental resu...