First, it seems to me that you would be better off running “3dANOVA2 -type 3 ...” instead of “3dANOVA3 -type 5 ...”. The reason is that you basically have 2 factors here: task (4 levels) and group (number of subjects in each group -- assuming equal number of subjects in each group). Of course you can run “3dANOVA3 -type 5 ...”, but the problem is that, with repeat = 1 for each observation, you would lose some statistical power and testing availability as well.

With 3dANOVA2, Other than testing the significance of main and interaction effects, you can also obtain the task differences between the two groups by adding an option “-bdiff i j prefixname ...” into your running script (assuming group is your 2nd factor). This option will generate the difference between two level means of factor B and its corresponding T-statistic values.