But, I find his analysis a little suspect. It appears to confuse corellation and causation. It notices a difference in behavior between the "great" and the "average" and assumes those differences to cause the difference in ability.
It completely ignores the possibility that the behavioral difference might be caused by a difference in ability, rather than the behavior changing the ability.
Also, another problem is that student classification is based on faculty opinion. The fact that students with a class of behavior tend to be liked more by their instructors implies something about the instructors, not the students. It does not measure music ability. It measures the ability to be perceived positively by music professors.
I would be interested to see a study that actually measures music ability, using set learning time frames with unfamiliar pieces and double blind assessment of music samples.
If that was correlated with behavior, then it might mean something.
The best the study can say is "if you want music teachers in Berlin to rate you highly, then behave this way".