According to recent research, when a person looks at an object and later visualizes it, many of the same brain cells are activated. This overlap between seeing and imagining gives mental images a stronger biological foundation and helps explain why memories, works of art, and intrusive visions can feel so real. The study’s findings were published in the journal Science.
In hospital epilepsy units, 16 adults first viewed hundreds of images and then recalled some of them from memory. Based on these recordings, Ueli Rutishauser from Cedars-Sinai Medical Center discovered that many cells that responded to visual perception were reactivated during mental imagery.
This reactivation did not occur throughout the entire visual system, but it was strong enough to show that remembered objects partially reuse the same neural pattern. The exact mechanism triggering this pattern remains unclear, raising a new question about the results.
Deep within the ventral temporal cortex, a visual area responsible for sorting complex objects, 456 out of 714 recorded neurons selectively responded to the presented objects. Many of these neurons were located in the fusiform gyrus, a brain fold that plays a key role in face recognition.
Earlier studies using brain scans suggested some overlap between visual and imagined perception, but these methods could not prove that the same cells were involved. Recording individual neuron activity changed this understanding, revealing not only the regions but also the specific cells where perception and imagination meet.
About 80 percent of neurons that responded to visual information followed a pattern already observed in the primate brain cortex. In many images, this pattern formed a neural code—a recurring activation pattern that carries information.
“Our study identified a code that we use to recreate images,” said Ueli Rutishauser, a neuroscientist at Cedars-Sinai Medical Center.
Once this pattern is established, the research can test whether memory restores the same design or creates a different one.
By using the responses of many cells simultaneously, researchers reconstructed visible objects with remarkable accuracy based solely on brain activity. In 487 out of 500 cases, the decoded pattern was closer to the correct image than a random match among multiple alternatives.
Artificial intelligence assisted by converting images into numerical descriptions and then generating new images that more strongly activated the selected neurons. These tests were significant because correctly guessing previously unseen images indicated that the code was genuine, not a feature of any single task.
Advertisement
aliexpress.ru
Later, six participants imagined a small set of images after viewing them, allowing the team to compare the results of viewing with those of recall. About 40 percent, or 43 out of 107 neurons tuned to specific images, were reactivated according to the original pattern.
Rutishauser stated that the result showed visualization works by reactivating cells first used during actual vision. Some cells responded only to visual perception or only to images, indicating a partial overlap without complete similarity.
Using a single code did not make the two states identical, as the overall population still had enough features to distinguish them. In the study of 231 neurons, the average firing rate during perception and imagination appeared similar, yet computer tests could still differentiate these states.
Even a single randomly selected neuron provided responses with above-chance accuracy, and after about 100 neurons, accuracy stopped improving significantly. Thus, the brain can reuse a sensory code while simultaneously indicating whether the image came from the eyes or from memory.
The reused code appeared predominantly visual rather than verbal or semantic, narrowing down what exactly the imagination revived. Models based on image features explained neuron activity better than models built on word meanings or category labels.
This code also aligns with the experience of mental images, which often arise as shapes, textures, and arrangements before language can catch up with the process. Since this study measured individual cells, it linked this everyday experience to a specific biological process rather than a metaphor.
These findings provide a better understanding of the biology of visual imagination, which could help explain both the process of creating creative images and unwanted intrusive scenes. Researchers describe this reproduction system as a generative model, a neural process capable of creating detailed sensory content based on memory.
Doctors already know that post-traumatic stress disorder can involve vivid, involuntary mental images, making this mechanism medically relevant. Developing treatments will require much more evidence, but a cellular target offers a clearer benchmark than symptoms alone.
One major puzzle remains unsolved, as the study could not pinpoint the signal that commands these visual cells to restart. Possible sources include the hippocampus, a deep memory center responsible for storing experiences, and frontal areas involved in retrieving information.
Another limitation of the study was the volunteers themselves, as all had epilepsy, and most reported strong visual imagery. These limitations leave room for further work with people who suffer from weak visual memory, hallucinations, or disorders that blur the boundary between imagination and reality.
By demonstrating that imagination engages many of the same cellular mechanisms used in vision, this work makes mental images more measurable. A clearer map may help develop future treatments and tools inspired by brain function, but it also shows how thin the line is between reality and imagination.
About the Author
