This GDC panel dicussion raised a very interesting point. A review score can dissuade people from actually reading the review. I would think a low score is the most likely to make a reader skip the actual article, with a medium-high score being the most inviting.
What this means is that game reviewers have an incentive to score games a particular way, aside from developer/publisher pressures. If they score a game too low (or too high?), their articles will be read by fewer people and thereby make less advertising money.
I'm not being accusative... just laying it out there for everyone to chew on.