Dear Samuel,

> We 've recently conducted  a neuroimaging study with one control task (C)
> and three experimental task (F1 F2 F3). Each task was repeated three times.
> 
> We are interested by the following question: which brain regions, commonly
> activated in the three experimental tasks (F1 F2 F3) when compared to the
> control one (C),  are more activated in F2 than in F1. I have two question
> regarding this procedure.
> 
> a) Can we only delineate the common areas of activations to the three F tasks
> by comparing each of them to the control task and looking for common
> activations in the following contrasts F1 - C, F2 - C, F3 - C or can we use
> the contrast (F1+F2+F3)-3C ?

I would use a conjunction analysis of the following sort.  Take the main
effect (F1+F2+F3)-3C and mask with F1-C, F2-C and F3-C.  Note that this may
be interpreted as deactivations due to C.

> b) We then want to use the masking procedure to look for regions commonly
> activated for the three experimental tasks which are also differentialy
> activated between experimental tasks. It seems to us that there are two
> possibilities:

Use the above three masks and apply them to the difference of interest: namely
F2-F1.  The SPM{Z} and masks are not orthogonal but this would identify the
voxels of interest.


>        first: we take the contrast of interest (F2-F1) and then we mask it
> with the main task effect (F1+F2+F3)-3C with a thresholds
> at p<0.001 for both contrasts, or even a lower treshold for the masked
> (interest) contrast.
> I suppose the question here is: which brain regions, activated in the
> common network at p<0.001 are also activated in the contrast F2-F1 at p<0.001
> (or less).
>         second, the opposite: we take the main contrast (F1+F2+F3)-3C and
> then, we mask with the contrast of interest (F2-F1). I suppose the asked
> question  here is
> also : which brain region, activated in the contrast of interest (F2-F1),
> are also activated in the main contrast (F1+F2+F3)-3C.
> 
> We think the first way should be used in our study but I'm
> not sure to understand the exact nature of the difference between this two
> ways. Could you tell me the what are the differences (statistical and
> conceptual)  between these two masking procedures ?


The difference is in the 't' {Z} values displayed.  In the first
instance these pertain to F2-F1 so these voxels represent differential
activation in F2 relative to F1, in regions that are jointly
activationed with respect to C.  Similarly for the second approach.

I hope this helps - Karl


%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%