Hi Steve, No, at the 2nd level i only did a mean of the 2 runs for each session for each subject. So after the 2nd level analysis i have 21 session1 and 21 session2 Cope images.I want to find out if session1 results are reproducible in session2.so i did a paired group difference between all session1 vs session2 at the 3rd level using FLAME. I didnt get any significant difference in the roi's i specified( using FEATQUERY). at first glance this indicates that the results of session1 are reproducible in session2. However, i wanted to check if the COPE and VARCOPE were similar between the 2 sessions or not. So i did 2 separate 3rd level analyses to find group mean of session1 and group mean of session2. I ran featquery using an ROI on session1 group mean gfeat and session2 group mean gfeat and looked at the 90th percentile values for COPE and Varcope. the values were Session1 Session2 COPE 38 25 VARCOPE 19 20 As you can see, the varcope is the same but COPE value reduces in session2. Therefore, shouldnt i expect to see significant differences in the paired group difference 3rd level analysis knowing that the mean COPE between the two sessions are different while varcopes are same? -vish