visual experience physically shapes the brain’s feedback loops

Brains constantly predict what the eyes will see next, relying on internal feedback networks that physically rewire themselves to match the patterns they encounter in the world. An experiment manipulating the sight of mice reveals that our visual system behaves like an active learning machine, providing a biological basis for how the mind generates contextual expectations. The animal findings were published in Current Biology.

When light hits the retina, the information travels to the back of the brain directly into the primary visual cortex. This region identifies basic visual elements like primitive edges and simple geometric shapes. From there, the data moves to higher-order brain regions that piece together the simple inputs to process more abstract concepts like entire moving objects or complex scenes.

Information does not just flow in one ascending direction. The higher-order areas constantly send dense bundles of nerve signals back down to the primary visual cortex. These descending connections are known as feedback pathways.

This feedback architecture helps the brain provide necessary context to what the eyes currently perceive. If a portion of a shape is hidden in shadow, the feedback network uses statistical probabilities learned from the natural world to guess what is missing. The system uses past visual scenes to predict what belongs in the blind spots.

Biologists recognized that early visual experience shapes how these feedback networks mature in young animals. It remained largely unknown if this visual input simply acted as a generic switch to start a rigid biological sequence. The alternative hypothesis was that the feedback networks took direct instruction from the visual world, molding their physical architecture to mimic specific statistical properties of their native environment.

Radhika Rajan and Rodrigo F. Dias, at the Champalimaud Neuroscience Programme in Portugal, led a research team to resolve this uncertainty. They designed an experiment to see what happens within the physical wiring of the brain when the visual world is artificially restricted during the early weeks of life.

The researchers fitted juvenile mice, at about 45 days old, with custom-designed steel goggles. These goggles featured specialized cylindrical lenses that warped incoming light before it reached the eyes of the animals. The lenses restricted the animals’ vision to a narrow range of specific orientations, somewhat similar to seeing the world through a stationary sequence of parallel bars.

One group of mice wore goggles that only allowed them to see elements angled entirely at 45 degrees. Another group of mice wore goggles restricting their sight completely to 135-degree angles. A third control group wore goggles with flat, unaltering plastic lenses that provided normal vision.

The mice lived within specially enriched environments for over a month while wearing these goggles. The cages included patterned walls and enough space to explore, ensuring the animals continuously engaged with their altered visual surroundings. The goal was to fully immerse the animals in an environment entirely dominated by a single geometric angle.

To observe how this restricted diet of images affected the neural circuitry, the scientists used a technique called two-photon imaging. This method involves introducing fluorescent proteins into targeted groups of brain cells. These special proteins briefly emit light whenever a cell experiences a spike in electrical activity, allowing researchers to watch living brain cells communicate in real time.

The scientists focused their microscopes on the exact connection points where nerves from a higher-order region, called the lateromedial area, terminated upon cells in the primary visual cortex. They recorded the microscopic flashes of light while the lightly anesthetized mice viewed moving geometric patterns presented on a laboratory monitor.

The researchers measured the responses of individual nerve endings to see which visual patterns triggered the most electrical activity. Brain cells generally display a preferred orientation, meaning they fire most rapidly when presented with a line drawn at a specific angle. They also possess a receptive field, which is the specific tiny patch of visual space they are responsible for monitoring.

In the goggle-wearing mice, the cells in the primary visual cortex entirely shifted their orientation preferences. The cells became highly responsive to the specific angles forced upon them by their optical lenses. This meant a mouse raised seeing only 45-degree angles developed a massive surplus of brain cells devoted exclusively to 45-degree angles.

The feedback nerves descending from the higher-order visual areas perfectly mirrored this sensory shift. The entire feedback population adjusted its tuning toward the angle of orientation the mouse experienced in its cage. The actual physical parameters of the receptive fields for these feedback nerves changed their geometry as well.

Normally, these receptive fields are somewhat rounded and symmetrical. In the experimental mice, the boundaries of the receptive field shapes stretched and elongated. They became tilted ovals aligning in the exact direction of their forced optical experience.

The experiment also revealed structural changes in the spatial mapping of the descending feedback network. Nerves sending top-down signals usually overlap precisely with the primary visual cells located directly beneath them in the cortical tissue. The scientists expected to see a generic blanket of feedback covering the primary areas.

Instead, they found that the visual experience altered how differing sets of nerves distributed information about the surrounding visual background. The feedback nerves that actively responded to the monitor patterns became less likely to transmit visual data from the spatial axis corresponding to the goggle’s specific angle. The feedback system reorganized itself around the overabundant geometric signals.

To understand how these physical changes might take place on a biological level, the researchers turned to computer modeling. They built a mathematical simulation of the visual cortex to test different governing rules for how brain synapses might adapt over time. Synapses are the tiny chemical junctions between individual brain cells where electrical signals are passed.

The computer models showed that the changes likely stem from two competing learning behaviors happening simultaneously in the same brain region. The ascending pathways, traveling from the primary visual cortex upward, appeared to follow a rule known as Hebbian plasticity.

Hebbian plasticity is a principle dictating that neurons firing together wire together. The constant visual exposure to a single specific angle strengthened the chemical connections between ascending cells repeatedly activated by that angle. This mechanism explains why the lower brain developed an exaggerated sensory preference for the single geometric orientation.

Conversely, the descending feedback pathways coming from the higher areas likely utilize a mechanism called anti-Hebbian plasticity. In this scenario, the feedback synapses actively weaken and retreat if they fire at the exact same time as the primary visual cells they target. The system essentially penalizes redundancy.

This anti-Hebbian rule functions as an active decorrelation mechanism. It stops the higher-order brain regions from simply echoing the exact same visual information back into the lower regions. By reducing redundant overlapping signals, the feedback network selectively amplifies novel visual information that deviates from what the animal normally expects to see.

The experimental results give a mechanical explanation for how visual expectations are sculpted, though there are limitations to the approach. The mathematical modeling demonstrates what is computationally possible given known biological rules, but it does not directly measure the microscopic molecular changes happening at individual synapses. Proving the existence of anti-Hebbian processes in live animals is notoriously difficult.

The methodology also focused exclusively on vertical connections between differing layers and regions of the visual brain. The researchers acknowledge that horizontal connections between neighboring cells within the exact same brain layer likely contribute to shaping these shifting responses. The model currently treats the neural layers as isolated vertical columns.

Future studies will need to untangle how horizontal mapping and vertical feedback networks interact during normal development. Moving beyond screen-based geometric patterns will help researchers understand how these wiring rules apply to active visual perception in the wild.

The study, “Visual experience exerts an instructive role on cortical feedback inputs to the primary visual cortex,” was authored by Radhika Rajan, Rodrigo F. Dias, Nikos Malakasis, Margarida Baeta, Xinyun Zhang, Julijana Gjorgjieva, and Leopoldo Petreanu.