Dehnadi & Bornat (2006) was the seminal two-hump paper, which was widely circulated but never peer reviewed. In 2014, Bornat (who was Dehnadi's advisor) retracted it.

However, the situation is an odd one. None of the data was ever called into question. The basic methodology has seen criticisms raised, but has not been clearly invalidated. Several subsequent experiments (Dehnadi, 2009; Dehnadi et al., 2009; Bornat et al., 2012) confirmed, at least, that the aptitude test consistently produces two humps.

Rather, Bornat issued his retraction in the face of his commentary on the studies. He described having a bad reaction to SSRIs in 2006, and subsequently making a large number of wild and grandiose claims about the work. These included calling the aptitude test "100% accurate", flatly stating that some students couldn't learn to program, and publicizing third-party claims about gender in the data set.

This is all very weird. The study stands uncorrected. Dehnadi seems to have gotten caught in the crossfire, since his paper was retracted for extra-textual reasons. No one has really settled the question of how predictive the aptitude test is. The only major follow-up I've seen, out of Toronto, is simply wretched. (They used CS course grades after curving to argue about distributions!)

So: double bell curve. Not proven, not retracted, still widely supported by anecdote.

http://wiki.t-o-f.info/uploads/EDM4600/The%20camel%20has%20t...

http://retractionwatch.com/2014/07/18/the-camel-doesnt-have-...

http://www.eis.mdx.ac.uk/staffpages/r_bornat/papers/camel_hu...

http://www.cs.toronto.edu/~sme/papers/2016/icer_2016_bimodal...