(3 of 5)
They asked right-handed test subjects to type a sequence of numbers (for example, 4-1-3-2-4) with their left hand over and over again as fast as they could. No matter what time of day they learned the task, their accuracy improved 60% to 70% after six minutes of practice.
When subjects who learned the sequence in the morning were retested 12 hours later, they hadn't significantly improved. But when those who learned the sequence in the evening were retested following a night's sleep, they were an extra 15% to 20% faster and 30% to 40% more accurate.
Much to the researchers' surprise, the greatest improvements appeared in those who spent the most time in the second stage of non-REM sleep. Other procedural tasks that depended more heavily on visual or perceptual ability required periods of deeper sleep or both slow-wave and REM sleep. Sometimes even just an hour of shut-eye made a big difference. Other times a full night's rest was needed. "It's probably going to turn out that different types of memory tasks need different kinds of sleep," says Stickgold.
The search continues for other cognitive skills that might be linked to sleep. In January, Jan Born and his colleagues at the University of Lubeck in Germany published a clever study that shows why sleeping on a problem often brings such good results. They asked 106 test subjects to transform a string of numbers into a different string of numbers, using a simple but tedious mathematical equation.
Unbeknownst to the study volunteers, there was a hidden trick to the calculations that could cut their response time dramatically. A good night's sleep between practice sessions more than doubledfrom 23% to 59%the probability that participants caught on to the trick. In other words, sleep isn't absolutely necessary to gain insight into a problem, but it can be a big help.
So can new technology, which is allowing researchers to study sleep at a microscopic level for the first time. Neuroscientists have long been able to record the firing of a single nerve cell, using a tiny electrode implanted in a laboratory animal's brain. But it's only recently that they have had electrodes small enough and computers powerful enough to record scores of individual neurons at once. The goal is to identify the changing patterns of neuronal firing during sleep. "There are days when we can record up to 500 neurons, but that's not typical," says Bruce McNaughton, a psychologist and physiologist at the University of Arizona in Tucson, who studies rats. More typically, he is able to tap between 50 and 100 neurons.
That's not a lot when you consider that even a rodent's brain has 125 million neurons. But it was enough to get him started.