Hallucinations implanted in mouse brains using light

Behavioural evidence suggests that targeting just 20 neurons prompted animals to ‘see’ an image.

Search for this author in:

Fluorescence light micrograph of a section through brain tissue, showing numerous glial cells and neurons.

Neurons (shown here in a fluorescence micrograph) can be manipulated with pulses of light.Credit: Dr. Chris Henstridge/Science Photo Library

Scientists have induced visual hallucinations in mice by using light to stimulate a handful of cells in the animals’ brains. The feat could improve researchers’ understanding of how the brain interprets and acts on what the eyes see — and perhaps even lead to the development of devices that would help visually-impaired people to see.

The authors of the study, published in Science on 18 July1, used a technology known as optogenetics that controls individual brain cells with pulses of light. The technique works with mice that have been modified so that their neurons produce a protein that causes them to fire when exposed to light.

In this case, neuroscientist Karl Deisseroth of Stanford University in California and his colleagues attempted to implant images into the brain’s visual cortex. This region normally knits pictures together from information produced by the retinas.

Deisseroth’s team showed mice images of either horizontal or vertical bars, and trained the animals to lick from a tube of water whenever they saw the vertical bars. The scientists monitored the animals’ brains and recorded which neurons fired when the mice saw the vertical bars. They eventually identified about 20 cells per animal that seemed to be consistently associated with the vertical image.

To create the hallucinations, the researchers shone light on only these neurons — stimulating them to fire. This caused the mice to lick the tube of water as if they were seeing vertical bars, even though the animals were sitting in darkness. The mice didn’t lick the tube when the scientists stimulated the neurons linked to the image of horizontal bars.

Christof Koch, president of the Allen Institute for Brain Science in Seattle, Washington, says that the paper is a technical tour de force and an advance in optogenetics. “It’s playing the piano of the mind,” he says.

Taking the long view

Anil Seth, a neuroscientist at the University of Sussex in Brighton, UK, says it is not clear whether the mice in the latest study ‘saw’ vertical bars consciously or subconsciously, and finding this out might require a different behavioural test.

But he is enthusiastic about the potential applications of the approach. “These optogenetic techniques really are game-changing,” he says, because they allow scientists to manipulate the brain rather than just observing it. That could lead to prostheses that input sensory information directly into the brain.

For his part, Deisseroth was surprised that stimulating only 20 neurons seemed to make the mice hallucinate. Given the chance that this number of neurons could randomly fire, he wonders why mice are not constantly hallucinating.

But Koch says that cells in the visual cortex are only part of what the brain uses to perceive and interpret an image — the first master switch in a cascade of neurons.

A mouse next to one of the social custom hologram generators used in the study

A mouse sits next to a sensor that generated the images used in the study. Credit: Sean Quirin, James Marshel, Cephra Raja, and Karl Deisseroth, Stanford University.

Other regions of the brain connected to the visual cortex assess the meaning of an image by putting it into context. In some cases, such as in dreams, the brain can generate images without any input from the eyes.

And master-switch neurons in the visual cortex can be very specific. In 2005, Koch’s group published a study showing that a single neuron fired whenever a person saw an image of actress Jennifer Aniston2. It’s unclear whether mice can recognize faces in this way, he says, but vision is less important to mice than it is to primates.

Future visions

The next challenge for the Stanford team will be to determine how neurons that sense specific images connect to regions of the brain that interpret the meaning of visual information. “We’re just scratching the surface here,” Deisseroth says.

The technique that the researchers devised relies on a set of proteins that are sensitive to dim, red pulses of light, to reduce the risk of overheating the brain. The scientists hope that the proteins will enable them and others to explore the function of neurons associated with the perception of other visual factors such as colour and shape, and other types of sensory input — including sound and touch.

For now, optogenetics is far from ready for use in people. But research is under way into other methods to supplement the senses by stimulating the human brain.

In June, a company called Second Sight in Los Angeles, California, revealed early clinical results from a device that uses electrodes implanted in the visual cortex to restore some vision to people who are blind. The electrodes stimulate the brain in response to information gleaned from a camera worn near a person’s eye.

The system improved the vision of six people to the point that they could see a white square on a black screen. The company hopes that the device will one day restore sight by sending more complex visual information directly into the brain.

Nature 571, 459-460 (2019)


  1. 1.

    Marshel, J. H. et al. Science (2019).

  2. 2.

    Quian Quiroga, R., Reddy, L., Kreiman, G., Koch, C. & Fried, I. Nature 435, 1102–1107 (2005).

Download references

Nature Briefing

An essential round-up of science news, opinion and analysis, delivered to your inbox every weekday.