Beautiful faces are average faces...but...

How did they control for whether the hyper-attractive group was "average" for a real world sample, but not their sample?