Thank you very much for your answer; however it is really ambiguous.
Did you mean that:
(1) It is fine to use the peak voxels from the GLM contrast
"condition_A vs condition_B" in the PPI with psychological context
"condition_A vs condition_B" ?
(2) It is fine to use the peak voxels from the GLM contrast
"conditions_A&B vs control_condition" in the PPI with psychological
context "condition_A vs condition_B" ?
I know that option (1) is fine but still have some doubts about option (2).
The way I understand the paragraph from Friston et al (2007) (pasted
in my previous message, below), the time-course of the seed region
should be uncorrelated with the stimulus which is used in the PPI.
Therefore, if we use the "condition_A vs condition_B" GLM contrast to
determine the seed voxels there might be a high linear dependency
between the physiological variable and the psychological context used
in the PPI. Wouldn't that be bad?
My apologies if something is not clear or if I have missed something
from the literature but I need a definitive answer.
Are there any other strong opinions on this matter?
On Wed, Jul 20, 2011 at 12:28 AM, MCLAREN, Donald
<[log in to unmask]> wrote:
> This is fine. You are simply asking whether or not an active region
> has differential connectivity with other regions.
> Best Regards, Donald McLaren
> On Tue, Jul 19, 2011 at 6:13 PM, <Firstname> <Lastname>
> <[log in to unmask]> wrote:
>> Dear PPI methods experts,
>> Suppose, we have performed a pretty standard GLM mixed effects
>> analysis and have identified peak voxels in a contrast of "condition_A
>> vs condition_B".
>> Is it theoretically sound to then proceed by using those voxels in a
>> PPI analysis, where the psychological context is again "condition_A vs
>> Even though conceptually this seems intuitive, after studying
>> Friston's et al (1997) "Psychophysiological and Modulatory
>> Interactions in Neuroimaging" paper in detail, it seems like we should
>> not be using as source area a seed voxel whose activity is known to
>> have the most systematic variation between the levels of the factor
>> used in the PPI (i.e. condition_A vs condition_B).
>> This issue is best summarized in a paragraph from page 227 of the paper:
>> Psychophysiological Interactions and Factorial Designs
>> Psychophysiological interactions generally depend on factorial
>> experimental designs, wherein one can introduce neurophysiological
>> changes in one brain system that are uncorrelated with the stimulus or
>> cognitive context one hopes to see an interaction with. We make this
>> point explicit, suggesting that this is another example of the
>> usefulness of factorial experiments: Although it is possible to test
>> for psychophysiological interactions in almost any experimental
>> design, the use of factorial designs ensures that any
>> psychophysiological interactions will be detected with a fair degree
>> of sensitivity. This is because the activities in the source area, the
>> psychological context, and the interaction between them, will be
>> roughly orthogonal and therefore one can use the first two as
>> confounds with impunity. The converse situation, in which only one
>> stimulus or cognitive factor has been changed, may render the activity
>> in the source area and changes in the factor correlated. If this is
>> the case, there is no guarantee that the interaction will be
>> independent of either and its effect may be difficult to detect in the
>> presence of the ‘‘main effects.’’
>> To me, it would seem that this paragraph precludes the approach I
>> briefly described, however the phrase "although it is possible to
>> test for psychophysiological interactions in almost any experimental
>> design" does leave some room for doubt.
>> Please forgive me if I am misinterpreting something, but to me it
>> would seem like it would be more appropriate to take peak voxels from
>> a GLM contrast such as "conditions_A&B vs control_condition" and then
>> subject those to the PPI with "condition_A vs condition_B" as the
>> I am looking forward to reading your opinion on this matter.
>> kind regards,