I recently decided to do a Friedman test in a pretest-posttest-follow up examination. The overall test was significant (p=.045). Then I looked for significant pairwise differences (which I expected, at least for the largest difference) using a Wilcoxon. I always thought that the significant overall test indicated at least one significant pairwise difference, but none showed up, although the largest difference (pre-post) was coming close (without Bonferroni correction, p=.065). If this is technically possible, and it seems that way, what then should be the interpretation of the Friedman outcome? Grateful for any suggestion.