A diversion? Yes. Addictive? Maybe. But a research tool for delving into the purpose of dreaming? In fact, the game Tetris has proved to be just that. Robert Stickgold and his colleagues at Harvard Medical School recently conducted a clever set of experiments in which they used the game to guide the content of people's dreams: among 17 subjects they trained to play Tetris, more than 60 percent reported dreaming of images associated with the game. And the researchers found that when and how the study's sleeping participants saw these images helps confirm the idea that the brain uses dreaming to reinforce learning. They reported their findings in the October 13 issue of Science.

The idea that sleep, and in particular dreaming, serves to cement new information and skills in the brain first gained a lot of attention when Stickgold and his colleagues described another set of findings in the March 1999 issue of the Journal of Cognitive Neuroscience. In this set of experiments, they showed that subjects who had slept for six hours or more after learning a new task—in this case, spotting a visual target on a screen as quickly as possible—improved, whereas those who didn't sleep on it didn't. Moreover, they found that those who improved the most slept for eight hours, with ample time for both slow-wave and rapid eye movement (REM) periods of sleep.

Stickgold hypothesized why these sequential phases of sleep were so crucial, describing a two-step process by which memories important to learning were consolidated and integrated in the brain. During slow-wave sleep, the hippocampus—a region of the brain that stores recent, episodic memories about discrete events—replays its files for the neocortex, home to more permanent memories. The communication between the two brain areas at this time is one way, from the hippocampus to the neocortex. During the REM dreaming that follows, though, the flow of information flips, from the neocortex back to the hippocampus. Stickgold suggested that once the neocortex connects the new memories to others in storage, it sends a message back to the hippocampus to erase them.

In this latest round of experimentation, Stickgold and his team probed yet a third phase of dreaming—the hypnagogic period that occurs within the first hour of sleep. They studied three different sets of subjects who all played Tetris over the course of three days—playing for two hours in the morning and in the evening on the first day, and for an hour each morning and evening on the following days of the study. Twelve of their subjects had never before played the game. Another 10 were experts who had logged between 50 and 500 hours of Tetris prior to the experiment. And the remaining five were amnesiacs, having no short-term memory due to lesions in the hippocampus.

Seventeen of these 27 subjects reported seeing the same images during hypnagogic sleep—namely falling geometric pieces that, if placed properly, rack up points in Tetris. And, interestingly, most of these reports occurred after the second night of the study. This delay suggests to the researchers that the need to learn may in part prod the brain into dreaming. "It's as if the brain needs more time or more play before it decides, 'Okay, this is something that I really need to deal with at sleep onset,'" Stickgold says.

In fact, the learning curve for the game—measured by total points earned—was quite different for the three groups. Whereas both the experts and the novices showed considerable improvement, the amnesiacs did not. And this progress was somewhat reflected in the dream reports. The nine novices who were initially worst at the game were the very same who reported seeing falling pieces during sleep onset—suggesting again that the more a subject needed to learn, the more his or her brain reviewed the material. Only five experts saw the imagery. Two of them, however, described Tetris images associated not with the version they played in the lab but with the version they had played on Nintendo machines—a twist that Stickgold attributes to the integrative process.

Perhaps most surprising, three of the five amnesiacs described having the same kinds of hypnagogic dreams as the normal subjects. The researchers had assumed that the amnesiacs' dreams—especially those during the hypnagogic phase—would have nothing to do with recent events, if they occurred at all, due to the damage to their short-term memory centers. "We thought that if there's one part of sleep that depends on episodic memories, which amnesiacs lack, it's sleep onset," Stickgold says.

But even for these individuals, most of whom did not remember the game from one day to the next and had to be taught all over again, the Tetris dreams seemed to affect their waking behavior. Co-author David Roddenberry, an undergraduate at Harvard, noticed that one of the amnesiacs who didn't remember the game nevertheless placed her fingers on the computer keys used in playing at the start of a session. "She did not quite know what she was doing and yet she did know what she was doing," Stickgold comments. "In a way, this is Freud's unconscious—things activated in our brain that are in fact memories that guide our behavior but are not conscious."

To try to understand this barrier between waking and sleep, the researchers also compared the differences in reports of images or thoughts of Tetris both before sleep onset and right after. Curiously, thoughts about Tetris not associated with seeing falling pieces were more prevalent before sleep, whereas reports of images were more common during sleep. "What was most striking about the data," the researchers write in the Science paper, "was the strong similarity in reports from different individuals." All the subjects dreamed of pieces falling and sometimes rotating or fitting into empty spaces—and none reported seeing the picture surrounding the pieces, the scoreboard or the keyboard.

"What we're really looking at here is the age-old mind-body problem: the mind-brain connection," Stickgold notes. "We think of our mind as being ours. But there are real ways in which the brain has a set of rules of its own. We're getting an idea of what the brain uses as its rules for picking out cortical memory traces to reactivate and bring into our conscious mind, and we're trying to see across wake-sleep cycles how that process happens." That game is far from over.