“I have always held that it is only when one sees one's own mistakes with a convex lens, and does just the reverse in the case of others, that one is able to arrive at a just relative estimate of the two.”