> Do I need any additional bonferroni-like correction because I am dealing with
> 3 groups, and if so how would I adjust the clustering options?
The current practice in brain imaging about this is that people simply don't pay attention to or care for this kind of multiple comparisons issue. I guess the rationale is that, it is the number of tests in terms of number of voxels that is the severest, so, compared to that severity, the number of tests in terms of group comparisons is pretty much negligible. Of course you can show some self-discipline and correct it yourself by lowering the p value (dividing the number of group comparisons) when clustering.
Gang