Tests whether the differences between two measurements of the same subjects lean one way, using ranks. The absolute differences are ranked, and the ranks of the positive and negative differences are compared. It is the paired t-test without the assumption of normality.