In a study published in Science magazine (
no link available yet, PDF (subscription required)), Evans and Reimer find that:
when a research article is offered online after being in print for one year, the use of an open-source format increases citations to the article by 8 percent. But when a paid-subscription format is used to distribute a year-old print article, the citations increase by 12 percent. (from The Chronicle of Higher Education)
Some of the Chronicle comments note that Evans and Reimer elected to publish in a fee-based journal themselves.
There is a complicated self-fulfilling prophecy mechanism at work in the pattern they find. For a scholar seeking tenure, citation count matters more than simple exposure. If the fee-based journals get cited more, they will remain more attractive publication vehicles for scholars. As the best scholars and rising scholars continue to publish in fee-based journals, then those journals will continue to get cited more.
Update (Feb. 25, 2009): Evans describes his research in this video interview from National Science Foundation.