Adam Haar Horowitz is the first to admit that whispering to strangers as they fall asleep “seems a little creepy.” He’d been mulling over the idea with fellow MIT master’s student Ishaan Grover a few years ago while thinking about ways to influence the dreamlike visions people see at sleep onset, a state known as hypnagogia. The pair wondered if quietly saying words or phrases to people in hypnagogia might influence the content of their thoughts and visions, thereby serving both as a tool to investigate human cognition and, ultimately, as a means to help people wield control over their dreaming brains.

Haar Horowitz didn’t end up whispering into strangers’ ears, but he, Grover, and other collaborators did find a way to execute the basic concept, using a more practical solution: a device that fits into a person’s hand to monitor changes in heart rate,...

In experiments detailed in Haar Horowitz’s master’s thesis and a scientific paper published earlier this year, people interacted easily with the setup, known as Dormio. They chatted with it, albeit somewhat nonsensically, about what they could see and feel as they slipped in and out of wakefulness. One volunteer, prompted by Dormio to think about a fork, described dreaming about a family that was “happy to see the fork. And they’re putting it in a pumpkin,” according to Haar Horowitz’s thesis. Another participant, told to think about a tree, described “a tree from my childhood, from my backyard. It never asked for anything.”

We treat studying dreams as another way to understand what the mind and brain are doing during sleep.

—Erin Wamsley, Furman University

Dream researchers who spoke with The Scientist say Dormio marks an exciting step for a field traditionally limited by scientists’ inability to interact with their study participants. Many experts define dreams broadly as any subjective experiences people have while asleep, although most projects rely on dream reports collected specifically from people woken up from rapid-eye movement (REM) sleep, the stage of sleep at which people are most likely to experience emotional, narrative dreams (though not the only sleep stage in which dreams can occur; see “The Stages of Dreaming” below). This time-consuming approach has been something of an obstacle to researchers interested in manipulating dreams—an important aspect of dream research, says Haar Horowitz, now a research assistant in the Fluid Interfaces research group at MIT’s Media Lab. After all, he says, “you can’t do controlled experimentation on dreams without an ability to control dreams.”

Adam Haar Horowitz (left) works on Dormio alongside (from left to right) Ishaan Grover, Oscar Rosello and Pedro Reynolds-Cuéllar. Other team members include Abhinandan Jain, Tomas Vega Galvez, Eyal Perry, Matthew Ha, Kathleen Esfahany, Christina Chen.

Haar Horowitz is one of a small but growing group of researchers who call themselves dream engineers and are exploring various methods to influence people’s thoughts at sleep onset and during sleep itself. Some tools, like Dormio, use aural stimuli, while others harness sights or smells, or employ more-complex technologies such as noninvasive brain stimulation. With a recent wave of studies demonstrating the promise of such approaches, neuroscientists and psychologists may be able to learn more not only about how and why dreams are generated, but about possible health and cognitive applications of dream control. 

Such techniques could help give dream researchers the control they’re after, says Harvard Medical School dream researcher Robert Stickgold, Haar Horowitz’s mentor and collaborator. With these approaches, “we can use the scientific method.” 

Lucidity and other dream experiences

Attempts to influence dreaming are by no means new. People dating as far back as the ancient Egyptians have been known to fast to induce vivid dreams, while scientists, philosophers, and artists have been experimenting for centuries with hashish, opium, and other drugs to conjure dreamy visions in and out of sleep. Many cultures continue to hold beliefs about the dream-altering effects of certain foods—folklore in many Western societies holds that cheese can induce vivid dreams, although there’s little scientific research on the topic. And demand for lucid dreaming classes designed to help people take control of their in-dream environments has taken off in the last few years, helped along by Christopher Nolan’s dream-twisting 2010 blockbuster, Inception.

Dream engineers are interested in finding reliable, researcher-controlled ways to induce lucidity—where a dreamer becomes aware of being in a dream and may be able to exert control over their actions and their environment, as well as other sensations such as flying, to investigate how those sensations are generated and whether they’re associated with any benefits for the person experiencing them. One method that’s received significant interest as a way to manipulate dream sensations is noninvasive brain stimulation, which uses a magnetic coil or scalp electrodes to influence electrical activity in the dreamer’s brain. 

In 2013, a small study applied 10 minutes of transcranial direct current stimulation (tDCS) to people in REM sleep, and concluded, based on reports people made after being woken up from REM sleep, that the procedure increased lucidity in dreams when compared to a sham procedure. A similar study the following year that applied either a sham procedure or bursts of transcranial alternating current stimulation (tACS)—which is thought to be better than tDCS at influencing brain oscillations—concluded that 40 Hz currents during REM could also promote dreamers to become more self-aware. However, in both studies, the effect was weak, and in the 2013 study it was only observed among people who said they already frequently experienced lucid dreams. That group is unlikely to be representative of the general population, for which researchers estimate that up to 50 percent may never have experienced dream lucidity. 

The University of Montreal’s Tore Nielsen, who directs the Dream and Nightmare Laboratory at the Center for Advanced Research in Sleep Medicine, is unconvinced that noninvasive brain stimulation works as a lucidity inducer. Like many dream researchers, Nielsen says he has experienced his fair share of lucid and otherwise extraordinary dreams. He and his colleagues recently carried out their own study of tACS using stringent study conditions: for researchers to confirm a participant’s report of lucid dreaming, that person had to give a signal on becoming lucid—flicking their eyes from left to right under their closed eyelids three times—and that had to happen during REM sleep, as determined by electroencephalography (EEG) analyses of their brain activity. “Much to our chagrin, we failed to replicate” the earlier findings, Nielsen says. Although some participants did the eye-flick signal during REM sleep and subsequently described vivid dreams, people were no more likely to have lucid dreams after receiving tACS than they were if they’d had the sham procedure.


Neuroscientists used to think that dreaming took place almost exclusively during rapid-eye movement (REM) sleep, a stage of slumber that is often accompanied by complex emotional, narrative-heavy dreams that can involve sensations such as flying or other movements. But in the last few decades, research has shown that people can also have subjective dream-like experiences in non-REM sleep, albeit less frequently and of a different nature. For example, a person thinking about a cat as they doze off into the first stage of sleep—a hallucinatory state known as hypnagogia—may see strange cat visions and experience sensations such as falling. Dreams experienced later in non-REM sleep tend to be more mundane and may involve people or objects that are familiar to the dreamer. Once in very deep sleep, people are more likely to have conceptual thoughts than to experience emotional narratives, if they have any memorable dreams at all.


Noninvasive brain stimulation may have other uses in dream manipulation, particularly for studying the relative roles of different brain regions in generating common dream experiences. Queen Mary University of London’s Valdas Noreika and colleagues, for example, recently used 10-minute sessions of tDCS to disrupt activity in the sensorimotor cortex of 10 volunteers while they were in REM sleep. The researchers woke people from REM sleep shortly after each session, and asked them to fill out questionnaires on what they’d been dreaming about—and specifically, whether they’d been engaged in movements such as lifting objects or walking. The results showed that people who had received tDCS reported experiencing less movement in their dreams than people receiving a sham procedure, suggesting that normal sensorimotor cortex activity is required for those dream sensations, Noreika says. Specifically, “we found that this sensorimotor cortex is responsible for repetitive actions of the dream self . . . such as walking, running, swimming.” 

Simpler technologies likely also have their place in the manipulation of dream experiences. Michelle Carr, who did her PhD in Nielsen’s lab and is now a postdoc at the University of Rochester, has been experimenting for the last couple of years with techniques to induce self-awareness in dreamers in a lab setting. She’s found that using behavioral training to get people to associate sensory stimuli such as lights or sounds with a sense of heightened awareness seems to be an effective way to trigger lucid dreaming.

Some people who had never before had one had their first lucid dream in the lab.

—Michelle Carr, University of Rochester

In a recent study, for example, Carr and her colleagues trained awake volunteers to try to become particularly aware of their surroundings whenever researchers presented them with alternating cues of flashing red LEDs and a beeping noise. Participants subsequently were allowed 90 minutes to doze off for a nap in the lab, while researchers monitored their sleep stages using several techniques including EEG and measurements of electrical activity in the muscles. When a participant entered REM sleep, experimenters triggered the LED and the audio cues in the same alternating pattern they’d played during training. By monitoring eye movements for the agreed-upon eye-flicking signal, collecting dream reports from people woken up by researchers for brief periods mid-nap, and administering questionnaires after the 90 minutes was up, the team found that around 50 percent of the treatment group experienced lucid dreams, compared with just 17 percent in a control group of participants who’d completed the training but hadn’t had the lights and sounds played to them during their naps. “It was really cool—some people did [the cue-signal response] up to eight times,” Carr says. “Some people who had never before had one had their first lucid dream in the lab.” 

While the research is still in early stages, Carr says she hopes the findings will encourage further studies intended to trigger certain sensations in dreams, with an eye toward the possible benefits. She and her colleagues recently analyzed dream and mood diaries kept by 20 people over the course of a week and found that higher lucidity correlated with elevated waking mood the following day. The researchers plan to use their lucidity-inducing techniques to investigate whether the relationship is to some degree causal, Carr says, and whether inducing lucidity has other applications, such as helping people suffering from recurrent nightmares—a common symptom of many mental health conditions including anxiety disorder and post-traumatic stress disorder. 

“If we can get [these dreams] to be induced reliably,” Carr says, “then we can use them for beneficial purposes.”

Incubation of specific dream content

For some dream researchers, it’s not just the overall dream experience that’s worth manipulating, but also a dream’s specific content. Many ancient human civilizations experimented with this idea too, and documented attempts to promote in-dream encounters with various deities, for example. But for Harvard’s Stickgold, it was a family trip to Vermont in the 1990s that made him start thinking about the idea. 

Falling asleep one evening after a hike up Camel’s Hump in the Green Mountains, Stickgold was surprised to feel as though he were scrambling up the side of the mountain, just as he had earlier that day, with the distinct sensation of rocky ground under his hands. Waking up and then dozing off again, he found that he was able to regain this sensation several times before falling into a deeper sleep. Intrigued by the experience, Stickgold says, he wondered about how to try to capture it in an experiment.

There was a stumbling block, however: he’d be unlikely to obtain ethical and administrative approval to lead a gaggle of undergrads on a rock-climbing expedition just to see if they’d go on to dream about the experience. It was only a few years later that an alternative presented itself. “I was in a meeting with a bunch of students one day, pissing and moaning about what a great experiment this would be, but how I would never be able to do it,” Stickgold says. “One of the students sitting there just said, ‘What about Tetris?’ They proceeded to tell me that this happens when you start playing Tetris: you see [the pieces] floating down before your eyes.”

That conversation was the seed for what would become a famous study in dream research. Stickgold and colleagues recruited 27 people—10 Tetris experts, 12 Tetris novices, and five patients with memory loss from brain damage—to play seven hours of the computer game over the course of three days. For an hour at the beginning of each night, participants were prompted by an experimenter or by a digitized voice recording to say what they were thinking about into a microcassette recorder or to an experimenter as they fell in and out of sleep. Almost two-thirds of the participants reported dream-like visions of Tetris during sleep onset, and three of the five amnesiacs also reported seeing Tetris-inspired imagery, despite having no conscious memory of the game. One described “thinking about little squares coming down on a screen and trying to put them in place,” while another said they’d seen “images that are turned on their side. I don’t know what they are from, I wish I could remember, but they are like blocks.” Several Tetris experts reported thinking not only of the Tetris they’d been playing during the experiment, but also of older versions of the game they’d played previously. 

Nudging the brain to incorporate specific content—a trick known as dream incubation—has proven to be surprisingly practical using computer and virtual reality games. Erin Wamsley, previously a postdoc with Stickgold’s group who now runs a lab at Furman University in South Carolina, says that many researchers previously assumed dreams would be most influenced by more-intense experiences. “You can show someone horrible graphic images or very disturbing films with very high emotional content that participants would agree is disturbing or emotional,” says Wamsley. “But [that’s] not something that triggers people to dream directly about that experience, necessarily. On the other hand, we’ve had a lot of success causing participants to incorporate new learning experiences into their dreams.” Wamsley’s now looking into what determines whether a particular experience will be incorporated into a dream. 

The Dream Engineer’s Toolbox

Researchers use a variety of technologies to monitor (teal) and attempt to modulate (purple) people’s dream experiences. While many protocols include pre-sleep training—to encourage people to become more aware of their dreaming selves, for example, or to incubate specific ideas using virtual reality or computer games—a number of dream-influencing approaches can be applied during sleep. Scientists also monitor participants during sleep and collect dream reports as soon as they awake.

See full infographic: WEB | PDF

In 2010, Stickgold, Wamsley, and colleagues got 43 volunteers to play an arcade skiing game called Alpine Racer. Around a third of the dream reports collected from subjects woken up from non-REM sleep over the following nights were related to the game. The nature of the dream content changed as people fell into deeper sleep, however, going from typical comments such as “I get like flashes of that . . . game in my head, virtual reality skiing game,” to oblique skiing references such as, “I was picturing stacking wood this time. . . . I felt like I was doing it at . . . a ski resort that I had been to before, like five years ago maybe.”

In the last couple of years, the same researchers have also used a simple maze navigation task that participants carry out on a computer to explore how content is incorporated into dreams that occur during different sleep stages. A dream report from someone just falling asleep contained thoughts of swimming above the maze, for example, while one participant woken from REM sleep reported dreaming about walking through it. A typical report from later stages of non-REM sleep involved the dreamer just standing in the middle of a maze waiting for a friend to find them. 

This and many other studies have also reported an association between the incorporation of task content into dreams and task performance post-sleep—a finding that adds weight to the prevalent view among sleep researchers that sleep, and perhaps dreaming specifically, plays an important role in memory consolidation. On the basis of current evidence, it’s not clear whether dreaming helps drive that consolidation, or is perhaps instead a reflection or byproduct of the process. Stickgold, who explores theories of dreaming with coauthor Antonio Zadra in a new book, When Brains Dream, slated for publication in January, hypothesizes that REM sleep plays an active role in consolidating emotional memories and extracting patterns from recent experiences, and that perhaps the dreamlike visions of hypnagogia are the brain’s way of tagging relevant content for processing later on in the sleep cycle. Other researchers posit that dreams serve different functions—Noreika is one of several scientists who think they offer simulation of potential threats and social interactions that the dreamer might encounter in waking life—or perhaps no function at all.

See “Dreaming of Possibilities”

Exploring potential functions of dreams and dream content is a key purpose of dream-influencing technologies such as Dormio, notes Haar Horowitz, who says that the device’s ability to interact with dreamers in real time offers the possibility of collecting data more easily compared to traditional dream research, even if hypnagogia and REM sleep aren’t exactly equivalent. He’s recently launched a number of collaborations, not only with sleep scientists curious about how changing dream content could alter memory or learning, but also with artists and philosophers interested in how dream incubation might boost their creativity. 

Making dream engineering mainstream

Carr, Haar Horowitz, and others organized a workshop at MIT last year for engineers and dream researchers to discuss technologies available to the field, and the group put together a special issue of scientific papers on dream engineering for the journal Consciousness and Cognition this summer. “A lot of collaborations developed from that workshop,” says Carr, who was managing guest editor for the issue. “I think it’s the start of something new.” 

With this momentum, dream researchers are hopeful that their field will overcome a lingering image problem in sleep science. Even now, “a lot of people view dreaming as a fringe topic, kind of like studying ESP [extrasensory perception] or out-of-body experiences,” says Wamsley. “Of course, in my opinion, it’s nothing like that at all. In our research on dreaming, we treat studying dreams as another way to understand what the mind and brain are doing during sleep.”

Nevertheless, the subjectivity of self-reported dreams remains an issue, she acknowledges. While neuroscientists’ attempts to objectively predict what people are dreaming about on the basis of brain imaging techniques such as functional MRI have made strides in the last few years, they’re a long way from matching the detail in dreamers’ own descriptions, she says. Aware of this obstacle, several groups working on dream engineering seek to demonstrate the value and feasibility of collecting dream reports as part of regular sleep studies. 

You can’t do controlled experimentation on dreams without an ability to control dreams.

—Adam Haar Horowitz, MIT

In a recent study from Björn Rasch’s lab at the University of Fribourg in Switzerland, for example, researchers trained people on a word-picture association task, and then subsequently woke them up for dream reports during the night. The team found that people’s memory of the task the following morning didn’t seem to be affected by the awakenings themselves. The researchers also reported that there was a positive relationship between dreaming of the task during non-REM sleep and memory performance the following morning, but they found no such association when it came to dreams of the task during REM sleep—a clue about sleep’s role in memory that would have been overlooked had dream reports not been gathered.

Dream researchers are also looking toward some of the extraordinary implications of manipulating the minds of sleeping people. With the prospect of devices such as Dormio allowing people to interface with their own or other people’s dreams, ethical considerations “are paramount here,” notes computer scientist Pattie Maes, the head of the Fluid Interfaces group at MIT’s Media Lab and a coauthor of a review of the field in Consciousness and Cognition.

Stickgold agrees, noting that even after having done it for decades, there’s something unique, and even unsettling, about interacting with the minds of people in the not-quite-conscious, not-quite-unconscious world of dreams. “It has an edge of scariness,” he says. “We’re tapping into an aspect of people’s minds that we don’t have much control over and they don’t have much control over when they’re sleeping. We’re almost voyeurs, watching their minds do what they decide to do.”  

Interested in reading more?

Magaizne Cover

Become a Member of

Receive full access to digital editions of The Scientist, as well as TS Digest, feature stories, more than 35 years of archives, and much more!
Already a member?