Neuroscientists find: the brain encodes vision and imagination with the same neurons

Neuroscientists obtained one of the most direct confirmations of how imagination works: when a person imagines an object in their mind, the brain can engage the same neural codes as when actually looking at it. This brings together the mechanisms of perception and mental visualization not only at the level of brain regions, but also at the level of individual neurons.

How It Was Tested

For a long time, researchers saw only the general picture: fMRI and other methods showed that imagination and vision engage similar areas of the brain. But such tools provide too coarse a resolution. They help understand which area is active, but they do not answer the main question: do literally the same neurons work in both modes, and do they encode the image in the same way?

This was precisely the missing link in the overall picture. To verify this, intracranial recording of single neuron activity in humans was needed. This approach allows observation of the work of specific cells, rather than an averaged signal from a large area of tissue.

Researchers focused on the ventral temporal cortex—a region that plays a key role in recognizing visual images. They then compared the response of neurons in two states: when participants actually saw stimuli and when they only mentally imagined them.

A Common Code for Images

The key finding turned out to be stronger than simply "imagination also activates the visual system." In the ventral temporal cortex, researchers discovered a common sensory code: some neurons that responded to real visual stimuli became active again during mental visualization. Moreover, this did not appear to be a random coincidence.

The amplitude of the response during imagination correlated with how strongly the same cells responded during normal perception. That is, not only the fact of activation matched, but also its profile. This is important because it is not about vague general activation, but about the reuse of the same way of representing information.

In other words, the brain does not simply "think about a picture" in an abstract sense, but partially reproduces the sensory format in which the picture is encoded when actually looking at it. This makes the mental image a more physiologically concrete phenomenon than previously thought. This is the main shift in interpretation.

• When looking at an object, specific neurons in the ventral temporal cortex became active
• During mental visualization, some of these same cells reactivated
• The strength of the response during imagination was related to the response to real stimuli
• Therefore, the brain uses a common encoding format for external and internal images

The Brain's Generative Model

The results fit well with the hypothesis of a generative model of the brain. In this logic, perception works not only bottom-up, from eyes to interpretation, but also top-down: more abstract representations can restart sensory patterns. Imagination then ceases to be a side effect of memory and looks like a separate working mode in which the brain itself assembles an internal version of the image.

Such a mechanism well explains why images can be so vivid. This finding has broad practical significance. It helps better understand how memory retrieval, spatial planning, and creative synthesis are organized, where a person needs not just to remember a fact, but literally to "see" the scene in their head.

For AI, this is also a useful reference point: if effective imagination in the brain is built on reusing sensory codes, then future architectures could more strongly link perception, memory, and generation, rather than keeping them as almost isolated modules.

What This Means

If the finding is confirmed in further work, imagination could be described not as a vague superstructure above vision, but as a mode of restarting an already existing sensory system. For neuroscience, this is a more precise map of the mechanism of mental images, and for AI—a clue for building models that not only recognize the world, but also internally reproduce its representations. This is no longer a metaphor, but a testable working scheme.