Suppose you have a sample of 50 measurements of matched-pairs data, and the histogram of differences is fairly, but not perfectly, normal. The t-test of µD leads to a rejection of the null hypothesis, indicating that the average difference between the populations is not zero. You decide to conduct a Wilcoxon Signed Rank Sum Test, just to be sure, but you find that you fail to reject the null hypothesis of no difference. How could this happen? What does it mean?