Hi Colm,

That looks like a result of Fourier interpolation on data that appears to
be masked. Those bands outside are presumably still very close to zero
(though they can still show statistical significance, if the edges that
the bands come from do).

Did you mask the data before this step? If so, why?

Note that Fourier interpolation, while considered the most accurate, has
the side effect of such banding from sharp edges or spikes in the data
(things that are far from smooth), which applies to interpolation over
space as well as over time (3dTshift). For this reason, we do not use
Fourier interpolation as the default with afni_proc.py.

I should also mention that with afni_proc.py, the default at the single
subject level is not to mask at all (well, except for the EPI extents).

- rick