afni.nimh.nih.gov

> Would it make sense to use across-subjects means for each condition separately?

If the average RT varies a lot across conditions, it would make more sense to just use the condition-specific average, not the global mean across all conditions.

Gang