A team at Northwestern University played sounds into the ears of sleeping volunteers during REM sleep, watched those volunteers respond from inside their dreams using pre-arranged physical signals, and then measured what happened when they woke up. The people whose dreams were successfully steered toward specific unsolved puzzles were twice as likely to solve those puzzles the next morning. The ones who forgot their dreams entirely but still responded to cues while unconscious solved at the highest rate of all.
The study, published in February 2026 in Neuroscience of Consciousness by Oxford University Press, is led by Karen R. Konkoly at Northwestern’s Psychology Department, with collaborators at the University of Notre Dame. It is one of the most direct experimental demonstrations ever produced that REM dreams actively contribute to creative problem-solving, not as folklore or anecdote, but as a laboratory finding backed by statistical analysis and physiological recordings.
Twenty participants took part, most of them frequent lucid dreamers capable of becoming aware they are dreaming while still asleep. Each came to the lab for two overnight sessions about a week apart. On each visit, they arrived two hours before their usual bedtime and were given a series of creative puzzles designed to resist analytical, step-by-step solving. Matchstick problems, rebus challenges, spatial and verbal brain-teasers, the sort of thing where you either get the flash of insight or you stay stuck. Each puzzle came with its own unique soundtrack, a 15-second clip of music or environmental sound. Participants had three minutes per puzzle, listening to the soundtrack on repeat while they worked. When they failed, they listened to the soundtrack one more time and memorized the pairing. The evening continued until each participant had banked four unsolved puzzles.
Then came the wiring. Electrodes were applied to monitor brain waves, eye movements, muscle activity, heart rate, and nasal airflow. Participants watched either Inception or Waking Life while the setup was completed, then went to bed with specific instructions. At 4am they would be woken for 20 minutes of lucidity training, a technique that pairs a specific triple-beep sound cue with a lucid state of mind as they drift back toward sleep. After that, sounds might be played during their REM periods. If they became lucid, they should signal by performing rapid left-right eye movements. If they heard a puzzle soundtrack, they should start working on that puzzle and signal by sniffing rapidly in and out through the nose.
Of the four unsolved puzzles each participant carried into sleep, researchers randomly selected two for cueing. The other two served as controls. When REM sleep arrived, the team first played about ten lucidity cues, then began feeding in the puzzle soundtracks and title recordings for the two selected problems. Cues were delivered every 15 to 30 seconds, starting barely above the threshold of hearing and increasing only while the sleeper showed no signs of arousal. Any twitch of muscle tone, any flicker of alpha waves, and the volume dropped back down.
After each REM period, participants were woken and asked to recount everything they could remember.
Fifteen of the twenty participants had at least one dream involving an unsolved puzzle. Cued puzzles were incorporated into dreams at a significantly higher rate than uncued ones, with the effect reaching p less than .001. The cueing worked exactly as designed: playing a puzzle’s soundtrack during REM sleep made people dream about that puzzle.
The dream reports read like transmissions from somewhere profoundly strange. One participant signaled lucidity with her eye movements, heard a puzzle name played into her ears, then confirmed with sniffing signals that she was working on it. When she woke, she described recruiting a friend’s daughter in her dream. The girl, who in the dream had autism and did not speak, climbed onto a seesaw. Watching this, the dreamer realized something about the puzzle: the scales were not flat. The actual puzzle required moving five matchsticks to balance a hanging scale. Whether the seesaw vision helped remains unclear. She never solved that one.
Another participant’s case was more striking from a data standpoint. She responded to two puzzle cues in real-time while asleep, confirming through sniffing signals that she was processing them. When she woke, she remembered almost nothing. She had been looking through a dresser. Papers with puzzles were inside. She solved one and got a little celebration. She could not say which puzzle. The next morning, she solved both cued problems. Something happened in her sleeping brain that her waking memory could not access, but the cognitive benefit was real.
This connects to the study’s most unsettling finding. When solving rates were broken down by dream type, the numbers ran against every intuitive expectation. Puzzles worked on during lucid dreams, where participants were consciously aware and deliberately trying to solve the problem, were solved at just 11% the next morning. Puzzles appearing in ordinary non-lucid dreams, with no conscious awareness at all, were solved at 46%. And puzzles where participants responded to cues in real-time but remembered nothing afterward hit 67%.
The people who tried hardest did worst. The people who processed the problems without any awareness of doing so did best. The researchers suggest that non-lucid dreams may enhance unconscious incubation by allowing incorrect mental pathways to fade while the brain ranges across distant associations. Deliberate effort during lucid dreams may do the opposite, dragging the dreamer back into the same wrong approaches that created the impasse while awake.
The headline numbers: puzzles that were dreamed about, by any measure, were solved the next morning at 42%. Puzzles not dreamed about were solved at 17%. That difference was statistically significant.
When the team divided participants into targeted dreamers, the twelve whose dreams responded to the cues, and nonresponders, the eight whose dreams did not, the picture sharpened further. Targeted dreamers solved cued puzzles at roughly 40%, compared to 20% for uncued puzzles. That interaction was significant at p equals .04. For nonresponders, cueing made no difference at all. This is a critical detail because it rules out the explanation that participants simply expected to do better on cued puzzles and tried harder. If expectation were driving the results, nonresponders would show a similar pattern. They did not.
The experiment also included home dream diaries. After each lab session, participants recorded their dreams every morning for about a week. Puzzle incorporation continued outside the laboratory. Sixteen dreams from twelve participants referenced unsolved puzzles during this period, suggesting that the cueing effect had a lingering influence on dream content even days later.
A prior study by some of the same researchers had found that playing puzzle sounds during slow-wave sleep, a different stage from REM, boosted next-day solving from 21% to 32%. The current study did not find an overall benefit from REM cueing alone. But when dreams actually incorporated the cued puzzles, solving rates jumped in a way the earlier slow-wave study never measured. The two findings may be connected. Previous research has shown that memory reactivation during non-REM sleep sometimes depends on a subsequent period of REM sleep to produce its full effect. The slow-wave cueing may have been setting up content for REM dream processing all along.
The neuroscience of why REM sleep lends itself to this kind of creative processing connects to how the brain handles associations during different sleep stages. Earlier research found that people woken from REM showed stronger priming from weak, unexpected word associations than from obvious ones, a pattern absent in waking and non-REM states. REM sleep appears to be a state where the brain naturally ranges further from the conventional, linking concepts that would never meet during conscious thought. That is the exact cognitive shift that creative problem-solving demands: the ability to abandon the obvious path and find the hidden one.
What Konkoly and her team have demonstrated is a method. Play associated sounds during REM sleep. Steer dream content toward specific problems. Measure the results. The dreams that were successfully steered produced a measurable doubling of problem-solving success. The technique worked, and it worked most powerfully in the participants whose sleeping brains were most responsive to external input.
The fact that forgotten dreams produced the best outcomes may be the strangest part of all. Something is happening in the sleeping brain during REM that the waking mind cannot retrieve but that leaves a clear mark on next-day cognitive performance. Whatever process is running during those lost dreams, it appears to be doing more useful work than conscious effort ever could.
Source:
Konkoly, K.R., Morris, D.J., Hurka, K., Martinez, A.M., Sanders, K.E.G., & Paller, K.A. (2026). Creative problem-solving after experimentally provoking dreams of unsolved puzzles during REM sleep. Neuroscience of Consciousness, 2026(1), niaf067. https://doi.org/10.1093/nc/niaf067
