> I'd like some advice in setting up the design for a blocked
> fmri-experiment with 9 conditions in total. Any comments are welcome.
> One condition is 'rest', all others are 'active conditions'. All these
> active conditions are closely related to each other, so we are
> interested in all contrasts between these active conditions and also in
> the contrasts active-rest. Presenting all 9 conditions in 1 session is
> not preferable, since we cannot repeat the conditions within 1 session
> and defining a HPF would become very difficult. This means that we
> have to spread the conditions over sessions. We have a TR=3.5 s and 8
> One appoach to overcome the problem of contrasting conditions from
> different sessions (fixed effect) and perform conjunctions between
> these contrasts, is to do a second level analysis with one image per
> condition per session. Is this correct?? But what if we want a fixed
> effect analysis?
> What is the best approach, remembering the contrasts we 're interested
> The conditions are:
> R : rest
> A1 A2 : single cognitive component task
> B1 B2 : single cognitive component task
> A : task A1 and A2
> B : task B1 and B2
> AB : two different cognitive component task
> BA : two different cognitive component task
> Furthermore we 'd like to specify the contrasts "A-(A1+A2)" ,
> "B-(B1+B2)", "AB-(A1+B2)" and "BA-(A2+B1)"
> and we are also interested in the conjunction of the contrasts:
> A-(A1+A2) & B-(B1+B2)
> A-(A1+A2) & AB-(A1+B2)
> One possibility could be to set up different sessions in such way that
> all direct contrasts (of interest) can be defined within sessions, e.g.
> R-A1-A2-A-B2-AB % sessions of type 1
> R-B1-B2-B-A2-BA % sessions of type 2
> However, since these two types of sessions are analysed seperately, we
> cannot define the desired conjunctions. For that reason we rename
> the conditions in the second set of sessions, e.g.
> r -b1 -b2 -b -a2 -ba
> If we now define 12 conditions (for each session), it becomes possible
> to specify direct contrasts within similar sessions (e.g. A-(A1+A2)
> or b-(b1-b2)) and to define a conjunction between these two.
> So, I have 2 questions:
> 1) Is this latter approach correct? Is there another solution to
> this problem?
> 2) What about the second level analysis with one image per condition
> and per session?
I would recommend you include all 9 conditions in each session. Do not
worry about high-pass filtering. The recurring R conditions (between
the 8 active e.g. R-A1-R-B2-R ...) define the fundamental frequency.
This determines a good cutoff (not the number of conditions). You
would then be free to analyse your results at the first or second
level. For example your 8 active conditions could be finger movement
at 8 different frequencies. This would be a perfectly viable
single-session parametric design, even if each frequency was used
It is quite possible to do conjunctions at the second level if you
qualify your results by noting you have assumed sphericity (i.e. the
difference between session-specific contrasts is the same as the
difference among contrasts within session.
I hope this helps - Karl