I don't really want to play into Dana's fantasies, but the
graphs in the second of his papers are feckin' weird.
Look at the last graph (bottom right). The line for control data goes straight across, the experimental data drop rapidly then rise. Two points are marked as p<0.01. The nadir of the drop and the last line in the sequence, where each of the four intervening points are higher than the former and lower than the latter. Indications of data spread (SD or SEM) are given for
none of the data. Even so, I struggle to imagine what data structure can yield these artefacts. There must be wild differences in variance among the points.