Can't comment on other areas but in EECS my experience was 2 - 3 was the norm. (Never ever heard of 5 - 6 although I presume you'd get that many if you submitted a paper with "cold fusion" in the title.)

Stuff is becoming so technical these days that from those 2-3 reviewers you are sometimes lucky to get a single person with a real understanding of the subject matter and competent to vote on it. The whole peer review system is almost as broken as the patent system imho.