of I have carried out a Kruskal-Wallis analysis and the resulting probability p=0.027 would indicate a significant difference between the 7 groups that I am investigating. To try and identify which groups were significantly different I used the Behrens-Fisher test, as implemented in R, and there were no significantly different groups. I do not understand why this should be the case. Am I missing something in the philosophy of multiple comparisons, or could somebody suggest how I could identify where the differences are? Thank you. Paul _________________________________________________________________ Are you using the latest version of MSN Messenger? Download MSN Messenger 7.5 today! http://messenger.msn.co.uk