Hi Hyejeen,

This is a common problem, especially for a main effect. Note
that even if you were able to reasonably increase the number
of clusters, there will still probably be some clusters that
span multiple regions of interest.

If you want to define ROIs functionally, then maybe you could
consider using 3dExtrema or 3dmaxima to drop down spheres at
extrema with some minimum separation distance.

Another option is to run the clustering for each condition,
instead of on just the main effect. From what I know, this
is most common (though it does not lend to the tabular result
you seem to be looking for).

Otherwise, maybe anatomical ROIs might be worth consideration.

Since the main effect is so large, it might not be appropriate
to do an analysis of functional ROIs.

- rick