PLoS Biology has an article with data that supports the hypothesis that open access articles receive more citations than articles hidden behind a toll (summary available here). The author compared open access and non-open access articles in PNAS, controlling for any confounding variable he could think of. The article is open access, of course, so you should check it out.
My own exercise in open access -- publishing some original research on this blog -- has been put on the back burner as I take care of some other research. You can read the background here. The next step involves a bit of data mining, and I'll get around to it when I've got some spare time.
I'm still a little skeptical about this study. They did mention my major concern, but I'm not convinced that they actually addressed it very well. The issue is that PNAS charges authors extra -- and quite a bit extra: A thousand dollars (US) -- for open access articles. That's got to impose a major selective force on which papers get moved into open access, and by who. I just went through this with PNAS, and did not select open access, even thoguh in principle I'm all for the concept; but I couldn't justify spending a thousand dollars on it.
If I was a large and well-funded principle investigator (the kind most likely to get highly-cited work published in PNAS) the bar would have been much lower. More, if I felt my paper was really earth-shaking, high impact, I'd have thought a lot longer about spending the $1000.
The authors of the study do note that there's a cost to going open access, and they do try to normalize for some of the factors that might confound the choice; but I'm a long way from being convinced that they've ruled out this self-selection as a major force on the impact rate of open-access papers.
A thousand dollars is just too much.
The first link is broken. I believe it should be http://dx.doi.org/10.1371/journal.pbio.0040157, which has an extra slash.
The link has been fixed. Thanks for pointing that out.