The ring effects certainly look like they are from motion,
but it is hard to be specific without the time series and
seed locations in hand. You should pick a few subjects
and play with InstaCorr a bit to get a feel for what is
driving the individual results.
1) Extract the seed time series from the errts dataset,
i.e. after all other signals have been regressed out,
and generate the correlations of that with the rest of
the errts dataset. The signal of interest should not be
part of the analysis, you already know what looks like
that.
2) The censor threshold is somewhat group specific. It
might be a guide to try setting it as low as possible
while still leaving plenty of TRs for the correlation.
Some subjects will be dropped when they have too many
censored TRs.
The default in afni_proc.py for healthy adults is 0.3
for task analysis and 0.2 for rest analysis. But for
children, that might have to go closer to 1.0, say.
- rick