Results from a recent research paper claim that fluid intelligence could be improved by training working memory.
hback is a dual n-back memory game based on the protocol described in the paper.
It is written in Haskell with gtk2hs.
Currently the project is a little slim on documentation. You can find a short write-up here: http://pithyless.com/blog/2008/05/18/hback-haskell-n-back-memory-game/
Also, if you're interested in the psychology and implementation details, do read the comments and responses.