Stanford University scientists triggered visual hallucinations in mice by activating a few brain cells with a light signal. The researchers only had to stimulate neurons to make a mouse behave as if it perceived things that weren’t there—implying that the brain may be more malleable than previously thought.

So few cells were involved in implanting the visual image in the mouse’s brain that “it raised the question: Why are we not walking around hallucinating all the time?” says Karl Deisseroth, the study’s senior author and a professor of bioengineering, psychiatry and behavioral sciences at Stanford.

It’s not yet clear how the healthy brain avoids such hallucinations, but “as a psychiatrist, this makes me think about all the disorders where we have spontaneous, unwanted, inappropriate activity,” Deisseroth says, regarding schizophrenia and other conditions. “We’re very interested in diving deeper,” he adds.

In the study, published on July 18 in Science, Deisseroth and his colleagues controlled a mouse’s perception while simultaneously capturing its brain activity—and then used that recording to stimulate the same neurons to elicit, in effect, a “hallucination.” The new work will help scientists begin to “understand how the circuits in the cortex help us to perceive the world,” says Michael Häusser, a professor of neuroscience at University College London, who was not involved in the research. (A similar study was published this month in the journal Cell.)

In the experiment, the mice looked at simple visual stimuli—either vertical or horizontal black bars on a white background. The researchers trained the mice to lick a tube to get water when they saw the vertical bar but not the horizontal one or an absence of the two. The neural activity in the visual cortex was recorded, and then the same areas were stimulated with a technique called optogenetics without displaying the actual images, which generated the perception of something that was not there. The resulting activity—even in neurons that had not been directly stimulated—was similar to when an actual image was displayed on the screen.

At one point, the researchers gradually reduced the contrast of the image until it was difficult for a mouse to distinguish the vertical or horizontal bars. But when, for instance, the neurons that had fired in response to a vertical bar were stimulated with optogenetics, the mice were better able to correctly make out the hazy image of the vertical bar.

Deisseroth’s paper created new tools that will be useful for trying to understand how information is represented and stored in brain circuits, Häusser says. “If we can use these tools to understand the neural code for perceptions, then we can actually simulate sensation,”says Häusser, who recently conducted similar research. “If we can do that, we can build prostheses. We can basically help the brain to perceive the world.”

For the past dozen years or so, the optogenetics technology that Deisseroth helped pioneer has allowed researchers to control different types of neurons by genetically engineering them to switch on or off when stimulated with light. Back in 2012 Deisseroth and his colleagues showed that they could activate a single mouse brain cell. Then they gained control of a group of cells, stimulating them sequentially.

In the latest paper, Deisseroth and his team demonstrated a new way of controlling brain cell activity. They did so by testing a variety of natural proteins and identifying one that complemented, rather than interfered with, the typical optogenetic protein, so the two could act at the same time. This achievement allowed the researchers to stimulate and image the neurons simultaneously.

The group also demonstrated it could use this new approach to learn more about how the brain perceives sensory information. “To understand—and, certainly, to influence or restore—perception, affecting the activity of a few neurons may be sufficient. And that’s a profound implication,” says Mriganka Sur, a neuroscience professor at the Massachusetts Institute of Technology, who was not involved in the research.

It is unclear, Häusser notes, whether the sensory capacities of the human brain, including smell hearing and touch, react in the same way. “Maybe it’ll turn out that the auditory cortex is very different. We don’t know yet,” he says.

But Deisseroth says he’s seen clues that the human brain functions similarly to a mouse’s. He once had a patient with a brain disease known as Charles Bonnet syndrome, which can occur when someone becomes blind in adulthood, leading to spontaneous, well-formed visual hallucinations. Antipsychotic drugs don’t usually work to treat the disease, Deisseroth says, but an epilepsy drug that inhibited the activity of certain brain cells made a difference for his patient. This observation fits with what he saw in the mouse brain, where turning up the activity of some cells created hallucinations.

Next, Deisseroth plans to look at other aspects of the mouse cells that he stimulated, including the proteins that they produce and their wiring, in an attempt to track the signals they transmit to their neighbors and discern how that message travels through the brain—gaining an even more detailed picture of the mammalian brain.