This is a topic which is attracting a fair amount of interest in academia
"The recent decision by Elsevier to start sending take down notices to sites like Academia.edu, and to individual universities, demanding that they remove self-archived papers from their web sites has sparked a debate about the copyright status of different versions of a scholarly paper.
Last week, the Scholarly Communications Officer at Duke University in the US, Kevin Smith,published a blog post challenging a widely held assumption amongst OA advocates that when scholars transfer copyright in their papers they transfer only the final version of the article. This is not true, Smith argued.
If correct, this would seem to have important implications for Green OA, not least because it would mean that publishers have greater control over self-archiving than OA advocates assume.
However Charles Oppenheim, a UK-based copyright specialist, believes that OA advocates are correct in thinking that when an author signs a copyright assignment only the rights in the final version of the paper are transferred, and so authors retain the rights to all earlier versions of their work, certainly under UK and EU law. As such, they are free to post earlier versions of their papers on the Web.