Japanese Researchers Develop Computer Program To Read Your Dreams


"It's the stuff that dreams are made of."

In 1941, Humphrey Bogart delivered those famous lines in the classic film The Maltese Falcon. More than 70 years later, neuroscientists are still trying to figure out just what that stuff is. Dreams, however, largely remain a mystery. They're as elusive to us as they are to the people who study them. But the other week a team of Japanese researchers announced a potential breakthrough: they've developed a program that can detect dreams.

In the 1950s–60s, several promising sleep studies were published. Yet, after many unsuccessful attempts at sustained progress, hope dwindled and scientists moved on to other areas of research. It wasn't until 2000 that Robert Stickgold's groundbreaking study revived this field.

Stickgold is an associate professor at Harvard Medical School, and a pioneer in the field of dreams, as it were. He found that 63% of participants who played Tetris during the day saw images from the game in their dreams at night. This discovery established a new way of thinking about dreams, and what role they may play in learning and memory.

More recently, Matt Wilson, a professor at the Massachusetts Institute of Technology, followed up on this work by going right to the source: the brain. Wilson studies the brains of rats by connecting electrodes to collections of their neurons. These allow him to record electrical signals and uncover patterns of brain activity. Wilson has become so fluent in rat neuron chatter that he can accurately predict what the rat is doing just by listening to the recording.

In one study, Wilson and his team recorded rats' brain activity as they ran through a maze. Later, as the rats slept, Wilson found similar, almost identical, electrical patterns of activity as if they were running through the maze again. Similar to Stickgold's findings, this study suggested that rats, like humans, replay activities from their day while they sleep.

Neuroscientists now believe this is how the brain consolidates memory. By repeating the day’s activities while we sleep and dream, the brain reinforces the important skills learned.

One of the main problems with dream studies is that most rely on self-reporting, (e.g. you're monitored by researchers as you sleep and they wake you up throughout different sleep cycles and ask you to report what you were dreaming about). It's not the most pleasant experience and not the most reliable data since many of us find it hard to remember our dreams in the first place.

Last week's study bypassed this clunky method with math. They used both fMRI and EEG machines to record brain activity in volunteers while they slept. Then, they woke volunteers up several times and asked them to report what they were seeing. Each object reported was assigned to the specific pattern of brain activity recorded. The team created an algorithm and trained a computer program to make these associations automatically. When they scanned the participants again, the program was able to detect what they were dreaming with 60% accuracy.

Keep in mind this study included only three participants and examined sleep stage one. Thus, these findings may not represent REM sleep, where the most complex dreams and coordination of many neural networks are thought to occur. However, the findings give us some real insight to what make up our dreams. This may one day help to develop treatments for people who suffer from parasomnias, or sleep terrors. More importantly, this gives scientists more understanding into sleep’s role in learning and memory.

For more insight on dreams and sleep in general, check out one of my favorite RadioLab podcasts, Sleep: