Very interesting.

Experiments with good controls don’t require many subjects, at least if the effect being measured is of a decent magnitude. This design isn’t purely experimental (because you can’t randomly assign participants to the ‘bully’ group), but it is much closer than what you see in most social science research designs. I’m guessing that most of the human subjects research you’re used to is either completely non-experimental (where the researcher does not have control over either the composition of the groups or the application of the independent variable), involves variables that only explain a small amount of variability, or is attempting to examine several variables simultaneously.

The problems with this study (at least in the summary of the summary here), are that any given region of the brain does multiple “things,” and even if you’ve got the right cognitive/affective processes, it can be tricky to make reliable linkages between neural activation and behavior. For instance, emotion regulation is not entirely automatic. It’s quite possible that bullies could regulate their emotions if they found them to be aversive, they just don’t find their emotional reaction to the stimuli to be aversive. If that were the case, it wouldn’t really be a neurological difference, at least not in the areas that were “lighting up.”

Bottom line… many very smart full-time researchers are quick to dismiss small sample sizes. You’re certainly not alone there. But that bias reflects experience operating under certain conditions and assumptions that aren’t applicable to experimental and quasi-experimental designs. The research may be questionable, but the number of participants is not a big factor in that.