Is the following an accurate description of your experiment?
1) Are nice and mean two separate big categories in your experimental design?
2) Aggression is a behavioral measure at the trial level that is applicable to both mean and nice categories?
As for modulation analysis, the basic model at the subject level is the following formulation
y = a + b * x + e
where x is the modulation variable (aggression in your case). What is your effect of interest at the population level between the following two?
1) a: BOLD response after controlling the aggression, or
2) b: aggression effect
Gang