Nature published a study on brain training via computerized tests, such as those found in Brain Age or Brain Challenge, this week. Unfortunately, their conclusion was not what I would have hoped for; these games do not increase cognitive function. It's a shame, because I actually find those types of games entertaining.
The study shows that any skills obtained from training on one particular task does not transfer to other tasks even when the tasks are cognitively closely related! That's surprising to me.
Anyway, read the actual paper here: Putting brain training to the test.