Dear Steven,
> Suppose we thought mood had an effect on working memory and had the
> N-back and a control task that used the same stimuli without a memory
> component in each order-counterbalanced mood/session. Would
> interaction contrasts of Mood/Session by Task (Nback vs control task)
> be valid in fMRI?
Absolutely. I think most people would make inferences using a
random-effects analysis in this instance (by assuming sphericity).
> If so, suppose orbitofrontal cortex is significantly
> activated in the Happy/Sad by Task interaction. Is there a way in SPM
> to choose between interpreting the effect of Sad Mood (relative to
> Happy Mood) as increasing activation of OFC during N-back, or
> deactivating it during the control task? I'm trying to understand what
> information can be gleaned across sessions in fMRI experiments.
The SPM{T} testing for the interaction will identify both effects
(increased activation with increases in the other factor or decreased
activation by decreases in the other factor). As usual these changes
are all relative. I cannot think of a principled way of distinguishing
between the effects of increasing happiness and decreasing sadness.
You can however mask (or indeed do a conjunction) between the main
effects and the interaction. For example you could look at regions
that activate AND are negatively modulated by happiness (note
these are not deactivations but decreased activations).
I hope this helps - Karl
|