Controlling your dreams to solve problems: what the Northwestern study really shows

What did the Northwestern team actually test?

Karen Konkoly, Ken Paller and colleagues at Northwestern University published their study in Neuroscience of Consciousness in February 2026. Twenty volunteers, most of them frequent lucid dreamers, spent nights in a sleep lab.

In the evening, each person tried creative puzzles (rebuses, matchstick problems, word and spatial puzzles), each paired with its own short soundtrack. They went to sleep with four puzzles still unsolved. During REM sleep, the researchers played the sounds of two of them; the other two served as a comparison. Participants were woken to report their dreams, then tried the puzzles again in the morning.

The clearest result concerns the dreams themselves. Cued puzzles showed up in dreams more often than the others, and 15 of the 20 participants dreamed at least once about an unsolved puzzle. A sound played during sleep can therefore steer what a dream is about, at least in part.

And did they solve more puzzles?

This is where the headlines ran ahead. Across the whole group, about 30% of cued puzzles were solved the next morning, against 22% of the others, a gap the authors did not consider reliable. Among the 12 participants whose dreams followed the cues most clearly, cued puzzles were solved more often.

That is encouraging, but it comes from an after-the-fact analysis of a small group. It is a lead worth replicating, not proof that dreams double your problem-solving.

How can a sound reach a dream?

The principle: reactivating a memory during sleep

The method is called targeted memory reactivation (TMR). It rests on one observation: during sleep, the brain replays recent memories. If a sound or a smell was present while you learned something, presenting it again during sleep can favor that memory.

In a well-known 2007 study, Björn Rasch and colleagues had people learn with a scent in the air, then released the same scent during deep sleep. Their recall of what they had learned improved compared with a control night.

What the 2026 study changed

Most TMR research targets deep sleep and memory. Here the cues were played during REM sleep, when vivid dreams are most common, with another aim: to bring an unsolved problem into the dream. The study does not show that REM dreams beat other states for creative work. A 2021 Paris study (Lacaux et al.) points to another window: people who spent a short time in the very first stage of sleep were more likely to spot a hidden rule in a math task.

How it differs from dream incubation

Dream incubation is much older: you hold a question in mind as you fall asleep and hope a dream takes it up. In 1993, Deirdre Barrett asked 76 students to incubate a personal problem every night for a week. About half recalled a dream they judged related to their problem, and most of those felt it held a solution. It was their own judgment, with no control group. TMR adds an outside cue during the night; incubation relies on intention alone.

Can you talk to someone who is dreaming?

The 2026 study builds on an experiment published in 2021 in Current Biology by four teams: Northwestern in the US, and groups in Paris (France), Osnabrück (Germany) and Nijmegen (Radboud University, the Netherlands). Their question: can a lucid dreamer receive a question and answer it without waking up?

People trained in lucid dreaming heard spoken questions, or saw flashes of light, during REM sleep. They answered from inside the dream with agreed eye movements or facial muscle twitches picked up by sensors. Across 36 participants, the teams recorded 29 correct answers from 6 people, to simple questions such as subtractions.

That is not many, and most attempts drew no clear reply. But it shows that a dreaming brain can take in a question from outside and send back an answer. It also explains why the 2026 study recruited lucid dreamers: they can signal what is happening in the dream while it happens.

Kekulé, McCartney: what the famous stories are worth

Dream solutions make good stories. The chemist August Kekulé said a vision of a snake biting its own tail led him to the ring structure of benzene. Paul McCartney has said the melody of Yesterday came to him in his sleep. Elias Howe's sewing machine needle is often credited to a nightmare. These accounts were told years later, sometimes embellished, and none can be tested. Our article on dreams and creativity looks at them more closely.

What the lab adds is not a miracle but a method: compare cued and uncued problems, count the results, publish the limits. The findings are more modest than the legends, and more useful.

How to try it at home

Without a lab, you can't detect REM sleep or know when a sound reaches you. These steps borrow the principle from the research without promising its results.

• Pick a real, specific problem. "How do I start the first chapter?" works better than "my career." The 2026 study used puzzles with one answer; open-ended problems have been tested far less.
• Write the question down before bed. One sentence. Read it again, picture the situation for a minute, then let it go. That is classic incubation.
• Add a sound if you like. Play a short, unusual sound (a chime, a few notes) while you work on the problem in the evening. If you replay it at night, keep it barely audible and drop it if it wakes you: broken sleep costs more than it gives.
• Note everything when you wake, before your phone. The scene, the feeling, any link with the problem, even a loose one. Our article on why we forget dreams explains why the first minutes count.
• Test the idea in daylight. A dream can offer an angle, not a guarantee. Check it while awake.

If a door, a key, a staircase or a mirror appears, don't read it as a coded answer. Note what you did with it instead: did the door open, did you climb or go down? Those details say more about how you are approaching the problem than any symbol list.

A journal example

Fictional example: how to record an incubation night without forcing the link.

• Question the night before: "How do I tell my team the project is running late?"
• Dream: "I was carrying boxes up a narrow staircase. Someone below kept asking what was inside, and I didn't dare open them."
• Feeling: "Unease, then relief when I put one box down and opened it."
• Possible link: "Maybe I'm less afraid of the delay than of showing unfinished work."
• Question to keep: "What if I showed the work in progress instead of waiting until it's done?"

One night proves nothing. Over several nights, log the question, any cue you used, the dream and what you tried the next day. If the problem touches on health, work or relationships, check how your journal stores those entries: our article on AI dream journal privacy explains what to look for.

What this research does not show yet

A small, unusual sample. Twenty people, mostly frequent lucid dreamers. Most of us don't sleep or dream like them.

Lab puzzles, not life problems. Nobody knows yet whether the approach helps with a conflict, a business decision or writer's block.

A fragile effect on solving. The overall effect was not reliable, and the positive result in one subgroup comes from an after-the-fact analysis. Participants also knew what the study was about, which can color their reports.

Rough timing at home. Sleep trackers estimate sleep stages; they don't measure them the way a lab recording does.

Sleep comes first. Don't turn your nights into work sessions. If a problem keeps you awake, if you sleep badly for weeks or wake up distressed, talk to a doctor before trying any dream technique.

Dream science is moving from "what do we dream?" to "can we steer what we dream?" The first answers are interesting, and still modest.