If I read his argument correctly, he was essentially saying that:
i) If intelligences capable of developing apocalyptic technologies continue to be colonistic they will eventually suicide themselves. His assumption is that colonistic behaviour is both aggressive and irrevocably conflict inducing.
ii) If intelligences capable of developing apocalyptic technologies cease to be colonistic they may survive, but at the cost of remaining local to their solar system of origin.
iii) Thus, a selectional effect is in place here, where the only true *advanced* intelligences that can exist in the Universe are a) non-colonistic and b) local to their solar system of origin.

He admits, however, that his argument is potentially non-exclusive; all it would take is just one successful civ to overcome this risk and the Galaxy is colonized in short-order. Moreover, a successful colonization attempt via Von Neumann probes would solve this problem, also.

And his argument is largely a sociological one. Using terms like "colonization" and "aggression" is a risky proposition at best when trying to infer the motives of extremely advanced intelligences living in socio-economic conditions radically different from our own (Kardashev civ types come to mind, as does Post-Singularity intelligences).

That being said, his mechanism for a Darwinian selectional effect against interstellar colonization is a provocative one and it's provided me with much food for thought.