You know, every time I point out a paper from PNAS, there are always a few comments to the effect of "Why do you bother reading that garbage heap, anyway?" Since I keep citing papers from the journal, it's obvious that I disagree, but I suppose I should take a minute to explain why.
The reason people are down on PNAS is the way that members of the National Academy can, if they choose, sort of jam things into the journal through a side entrance. Here are all the details. The unusual thing about the journal is the existence of "Track I". Basically, a member of the NAS can publish up to four of their own papers per year. Each of these have to be submitted with the comments of two qualified referees, but the author gets to pick them. So a reasonable member should be able to get any sort of interesting or at least non-insane paper in there, by judicious choice of colleagues for review. Members can also pass along up to two papers a year by others in their field, with a similar review process (Track III). Some NAS members take full advantage of these privileges, and some hardly ever do, even (so I'm told, in some cases) for their own papers.
It's a lot less rigorous than the open (Track II) submissions, that's for sure. For those, you're supposed to name three editorial board members, three NAS members, and five external referees, and the editorial board can still do whatever it wants with your paper or with the lists you've sent. (To be sure, they can also reject those direct-submission papers from members, although no figures are available on how often that happens). Two thirds of the Track II submissions are rejected before being sent out for review at all.
But hold on: according to the journal, 80% of the submissions are via Track II, but those make up only 40% of the published contents. Doing the math, that means that the most of Track I and Track III submissions have to get in. Assume 80 Track II manuscripts and 20 of the others. Rejecting two-thirds of the first group will give you about 27 papers to send out for review. If you've accepted all 20 of the others, that means that about half of those 27 will have to get canned during the later review process, to make that 40/60 proportion come out right. So the overall acceptance rate for open submissions has to be, at most, 16%.
But if you ditch some of the 20 member-track papers, you have to come down even harder on the open submissions, of course. If you only (only!) take 75% of the member submissions, that gives you 15 manuscripts. Now you have to reject not half of the open submission papers that made the first cut, but 63% of them, to knock it down to ten published Track II submissions. So with an acceptance rate of 75% for member submissions, it has to be about 12% for everyone else. And so on.
So much for the numbers - it's clear that NAS members must put a lot of things of their own (or from their friends) into PNAS. The real question is: what does this do to the quality of the journal? As far as I can see, it's still a very interesting read, and definitely cannot be safely ignored. And the publication routes are out on the table: if you want to keep score and adjust your perceptions accordingly, the Track I papers are identified as "Contributed by" the member, and the Track IIIs are "Communicated by". I think, myself, that the advantage of letting members publish unusual or possibly controversial work outweighs the temptation to fill the journal with junk.