From Scientific American:
In the market for more brain power? In what’s being touted as “a landmark” result, University of Michigan at Ann Arbor (U.M.) researchers report that a specific memory exercise may improve so-called fluid intelligence—the capacity to succeed at new cognitive tasks and in new situations. The finding flies in the face of conventional wisdom in psychology that training for one brain task cannot be transferred to improvement in other mental abilities. If proved, the finding could lead to new therapies and prevention of learning disorders and age related memory loss.
The study contradicts decades of research showing that attempts at crossover training effects, known as far transfer, do not work well. Previous research has shown that improving on one kind of cognitive task does not improve performance on other kinds—for example, memorizing long strings of numbers does not help people learn strings of letters.
Researchers gave 35 volunteers a standardized intelligence test and gave them another such test after training them on a complex memory task for a variable number of days (eight, 12, 17 or 19). Thirty-five other study participants simply took the tests. Both improved on the second one, but those who did the exercise showed far more improvement—and the more they trained, the better they got.
More here.