Dear FSL experts, We have functional connectivity data on 3 different groups (1=depressed, 2=manic, 3=controls) and want to examine differences in connectivity between them using FEAT. Following the advice on the FSL GLM website we have made 3 contrasts and one F-test to see whether the group averages account some significant effect (see below the design). The output of the FEAT higher-level analysis are 4 connectivity maps: zstat1 (depressed group), zstat2 (manic group), zstat3 (controls), and zfstat1 which should identify areas in which the groups may differ. However, the zfstat1 seems more like a combination of zstats 1, 2, 3, rather than their difference, as practically all the voxels of zstat 1, 2, and 3 are depicted in zfstat 1. Have we done something wrong? Also, is this most optimum way of testing between group differences in case one has more than 2 groups or are there more efficient/robust alternatives? Any assistance would be more than welcome! Regards, Mauricio Design: Groups EV1 EV2 EV3 F1 1 1 * 2 1 * 3 1 *