Check out this paper, which inspired Jeff Atwood's "two humps" model of programming education: https://web.archive.org/web/20090401003425/http://www.cs.mdx...

It's been retracted, but it's the only case I've ever seen where I found the retraction unconvincing. The retraction effectively says "After this paper came out, I made a lot of wild claims about it that the data didn't support. Therefore ignore the paper, even though the results replicated successfully."