Hi Alfonso,
Thanks for your helpful response. I checked our denoising histogram and global MPFC connectivity results and they look fine (see attached screenshot). It could be that the analyses we're trying to run are too stringent. We have six scans (rest, and 5 emotion conditions (neutral, anger, anxiety, etc.) in which subjects are listening to emotionally evocative music while undergoing a task-free scan). We ran the standard preprocessing and denoising pipelines and set up ROI-to-ROI, seed-to-voxel, and voxel-to-voxel first level analyses for all conditions and for many different seeds/source ROIs. When we only consider one condition at a time (for example, on the second-level results page, just clicking each condition to look at their main effects one by one), we see some relatively robust activations, but when we try to put in a differential contrast (even something like anger > rest) we aren't seeing much of anything. 

Again, thanks for clearing up my initial question but I figured I'd include this information in case you have any thoughts about using conn for this design.

Thanks,
Adam