Hi Lesley,
It might be good to be a little more explicit...
You seem to be using the time series from a single voxel as the seed
(as opposed to a small ROI). And because of that, as you said in your
original post, the correlation at the seed voxel with the seed time
series is 1.0, as expected.
The point to make is that since the seed time series exactly matches
the data at the seed voxel, _all_ other betas and R values will be 0
(at that voxel), including the interaction. Since the seed explains
everything there, there is nothing left for the interaction. The
interaction must be 0, by your design.
It isn't that there doesn't happen to be an interaction at that voxel,
it is that there cannot be one.
- rick