Imagine That: Brain Uses Neurons from Vision System When Forming Mental Imagery

Creative endeavors, like making art, writing music, or penning a poem, require the recall of memories to fuel imagination. Many other human behaviors, including problem solving, also rely on mental imagery to complete tasks, but little was known about how imagery works at the level of single neurons in the brain—until now.

Varun Wadia (PhD '23), a former graduate student in neurobiology at Caltech and now a postdoctoral scholar at Cedars-Sinai, and a team of scientists and physicians have found that many of the same neurons that are active when looking at an object are also active when imagining that object from memory. The findings could help develop defenses against diseases that cause memory loss, like Alzheimer's, and assist in building more efficient artificial intelligence platforms. A paper published April 9, 2026, in the journal Science outlines the researchers' process and findings.

"We were very interested in trying to understand the mechanisms of mental imagery because they permeate many interesting human behaviors," says Wadia, who is first author of the study and whose thesis work forms the core of the paper. "What we saw is that when you imagine something you've seen before, your visual system is being put into the state that it was in when you first looked at it."

The new study builds on work by Doris Tsao (BS '96), a professor in neurobiology at UC Berkeley who was a Caltech faculty member from 2009 to 2021 and is a senior author of the paper. Tsao, who served as Wadia's PhD advisor, studies the representations of visual objects in nonhuman primates. Her research has found the mechanism that the brain uses to represent facial identity and the mathematical system used by the brain to organize visual objects, among other discoveries.

The first step in the team's work was to compare how humans process objects in the brain with the framework that Tsao had found in nonhuman primates—called a distributed axis code—in which individual neurons encode a specific dimension, or axis, of object space. Wadia and Tsao collaborated with electrophysiologist and neuroscientist Ueli Rutishauser (PhD '08), who is a faculty associate in biology and biological engineering at Caltech and a faculty member at Cedars-Sinai where he directs the Center for Neural Science and Medicine. The researchers, along with Rutishauser's clinical colleagues at Cedars-Sinai, recorded neuron activity in patients with epilepsy who have electrodes temporarily implanted in their brains to monitor seizures.

The electrodes allowed the researchers to employ single-neuron recordings to document the activity of many individual neurons at the same time in a region called the ventral temporal cortex (VTC) that is critical to visual recognition and memory. This tool let the team examine what VTC neurons were doing when study participants looked at an object, such as a bird or a saxophone.

"It was very surprising how well the model for nonhuman primates mapped to and worked for humans," says Rutishauser, a senior author on the study. "It means this entire body of knowledge that has been developed over many years applies to the human brain, which is far from trivial."

Wadia and the team used neural activity to figure out how the neurons were representing viewed objects and then applied that knowledge to reconstruct the objects patients were viewing. They then had patients imagine a subset of the objects shown to them while recording activity from the same neurons. Roughly 40 percent of the neurons reactivated during this imagining phase and had similar responses as during vision, implying that both processes used the same distributed axis code. Furthermore, the responses were so strong that the researchers were then able to reconstruct the objects that people were imagining, a first for these types of studies.

"We did analysis to demonstrate that this reactivation is visual in nature and involves the same neurons, which suggests that we have a generative model in our heads," Wadia says. "That is an intriguing conclusion because it means we have a way to conceptualize how the nervous system implements creative tasks, whether it's making a song or a painting or imagining how to solve a problem in your head. This insight can serve as a hook into understanding all of those super interesting behaviors."

He says that having a mechanistic understanding of how creative and intelligent behaviors happen could help inform a more efficient way of developing artificial intelligence in computing systems. On the clinical side, Wadia and Rutishauser believe their findings represent the first step in decoding memory in the brain.

"If you really understand how memory works, then you can start to think about how we might consolidate memory or prevent it from being eroded by Alzheimer's and other diseases," Wadia says. "Tomorrow's clinical care is today's science project, and we are immensely grateful for the patients who were willing to participate in our work and understand that their participation could lead to benefits for others in the future."

The findings could also help reveal new solutions for mental illnesses such as schizophrenia. "There are many devastating conditions where people imagine things that don't exist and that has a negative effect on their well-being," Rutishauser explains. "Our work could have significant relevance in the field of psychiatry."

The team is also planning follow-up studies with additional data collected from their study participants to try and find where the trigger signal for reactivation is coming from and how different areas of the brain might be working together to implement imagination.

"This paper is really a good example of the kind of discoveries we can achieve when scientists and engineers at Caltech work closely together with clinicians and patients," Rutishauser says. "It's something that neither Caltech nor Cedars-Sinai could have done on its own."

The Science paper is titled "A shared code for perceiving and imagining objects in human ventral temporal cortex." Additional authors from Cedars-Sinai are neurologists Chrystal Reed, Jeffrey Chung, Lisa Bateman, and neurosurgeon Adam Mamelak. The authors also acknowledge Ralph Adolphs, Bren Professor of Psychology, Neuroscience, and Biology at Caltech, for his support of this study. Funding was provided by the National Institutes of Health's BRAIN Initiative, the National Institute of Mental Health, the Howard Hughes Medical Institute, the Simons Foundation Collaboration on the Global Brain, and the T&C Chen Center for Systems Neuroscience at Caltech.