My problems with SPM are two-fold. 
Firstly, I don't understand F-contrasts well. I thought I did but now I am not 
sure I do. 
Secondly, I don't know how to do a GLM analysis on multiple subjects. Let me 
elaborate further.
 
So far, what I'm doing is simple: subjects answered some questions on a 
computer screen; questions were either "counting" questions (they must press 
a button to show how many objects they see on the screen) or "reasoning" 
questions (they must choose one of several answer choices that best fit a 
given pattern); a block design was used, a question was displayed for 9 
seconds (during which the subject had to press a button to answer) and there 
was a 3-second rest period after each question; 36 questions in all.
 
From my understanding of what the SPM5 manual says, T-contrasts are 
suitable when I use only the canonical HRF, and F-contrasts are suitable when 
I use the canonical HRF and its partial derivatives. F-contrasts are also the 
way to go for any analysis of multiple subjects.
I am also of the understanding that GLM of multiple subjects is a 2-level 
process: I have to do a 1st level analysis on each subject (presumably, using 
only F-contrasts), generate contrast images from each subject, and then use 
those contrast images in a 2nd-level analysis.
There is one example in the manual (Chapter 30 of the SPM5 manual) where 
they show how to do a 2nd-level multi-subject analysis. In that example, got 
three con*.img images from each subject; one for the canonical HRF, one for 
its time derivative and one for its dispersion derivative. I really wonder how 
they got those images because it appears SPM doesn't let me choose either of 
the derivatives on its own. [In my case, I used an F-contrast for all canonical 
HRF and all its derivatives, and I got ess*.img images and not con*.img 
images.]
Well, I proceeded with my ess*.img images and did the 2nd-level analysis. I 
got some results but I don't know how to make sense out of them.
 
The reason I don't know how to make sense out of them is my understanding 
of T- and F-contrasts. Let's assume I have 3 conditions: counting, reasoning 
and rest. (Another question I have is whether or not to model the "rest" as a 
separate condition. So far, I've been doing that. What do you think?) Well, if I 
want to, say, know what brain areas are activated during counting, my T-
contrast vector is [1 0 0]. What would my F-contrast vector be? Since I have 
the canonical HRF and its two derivatives, I'm assuming my F-contrast vector 
would now be [1 0 0 0 0 0 0 0 0; 0 1 0 0 0 0 0 0 0; 0 0 1 0 0 0 0 0 0]. But 
when I did the 2nd-level analysis, I couldn't at all enter such an F-contrast!
 
So, in view of all this background, my questions are:
 
1. How do I conduct a GLM analysis of multiple subjects? In fact, in answering 
this question, you'd automatically be answering my other questions....
 
2. Is my understanding of F-contrasts correct?
 
3. In the study I have above, would you model "rest" as a 3rd condition?
 
 
Thank you very much.