Hi there,

I'm working with 3 column EV files for my experiment and have 5 EVs (say A,B,C,D and E). My problem is that the design efficiency (and consequently the effects I see) look very different depending on what EVs I  include in each run.

1. For instance, if I input just an EV for A, I get a design efficiency of ~.7 and see lots of activity in the ROIs I expect to be active.

2. However, if I include EVs for events A-E (and look for the effect of A by running a contrast of 1 0 0 0 0), I get a design efficiency of 2 or 3, and of course, see no activation.

The same happens for my A-B contrast: when I include just A and B (1 -1) I see the activation I expect, but if I add C-E (i.e. 1 -1 0 0 0), the activation disappears and design efficiency becomes prohibitively high again...

How is FSL treating the empty zeroed out EVs, and in the case of both the above approaches, what is the "baseline" to which activation in A is relative to?

I don't know if this confounds things further, but I am also interested in looking at events B and C, and the same thing happens when I try the above 2 approaches for these events. D and E are essentially just coding for missed responses, so aren't of any interest, but I have been adding them in to completely account for any specific events in the data.

I've looked through the archive but can't find an article that addresses this specifically, so would really appreciate your help. Thanks very much,

am