Abstract
Organisms process sensory information in the context of their own moving bodies, an idea referred to as embodiment. This concept is increasingly central to developmental neuroscience, robotics, and systems neuroscience. A key manifestation of embodiment is the observation in mice of brain-wide neuromodulation, including in the primary visual cortex, driven by task-irrelevant spontaneous body movements. Here, we tested this hypothesis in macaque monkeys—a primary model for human vision—by simultaneously recording visual cortex activity and facial and body movements. We found that while activity in the visual cortex (V1, V2, V3/V3A) was associated with the animals’ own movements, this modulation was largely explained by the impact of those movements on the retinal image. These results suggest that embodiment in primate vision may be realized by input provided by the eyes themselves, rather than direct motor-to-sensory modulation.
Introduction
Recent discoveries in rodent neuroscience have sparked a minor revolution in our understanding of sensory processing. Researchers found that when a mouse engages in spontaneous behaviors—such as whisking, running, or grooming—neurons across its entire brain, including the "purely" sensory primary visual cortex (V1), fire in synchronization with those movements. This supports the theory of embodiment: the notion that sensory processing is inextricably linked to the body’s motor state.
However, these findings raised significant questions for primate researchers. For decades, vision science in monkeys and humans has operated under the assumption that the visual cortex primarily processes retinal images. If primates were like mice, decades of research might require a fundamental re-evaluation. This study asks: Is the primate visual cortex actually driven by spontaneous movements, or is it more modular and "visual" than the rodent brain?
Methods
The study utilized a newly collected dataset from two adult male rhesus macaques (Macaca mulatta), recording from 900 high-quality neural units across areas V1, V2, and V3/V3A. To ensure a direct comparison with previous mouse studies, the authors employed a similar analytical pipeline:
- Movement Decomposition: High-resolution videography of the face and body was analyzed using Singular Value Decomposition (SVD) to extract "movement components" (e.g., mouth twitches, hand movements).
- Linear Encoding Models: Ridge regression models were built to predict neuronal firing rates based on three categories: task/stimulus, slow activity drifts, and movement components.
- Unique Variance: The key metric was "unique variance"—the amount of neural activity explained only by movement after accounting for the stimulus and other factors.
- Retinal Control: The researchers compared epochs where the monkey was fixating (controlled retinal input) versus epochs where the monkey was looking around freely (uncontrolled retinal input).
Results
The analysis revealed that the primate visual cortex remains remarkably "insulated" from spontaneous body movements.
The "Retinal Image" Effect
In mice, movement appears to drive V1 activity directly. In macaques, however, the authors found that "movement-related" activity almost entirely disappeared when the retinal image was stable. Most of what appeared to be movement-driven activity was actually the visual system responding to the room moving across the retina because the monkey moved its eyes. When the monkeys' eyes moved freely, 67% of neurons appeared movement-sensitive; when they fixated on a dot, that number plummeted to just 5%.
Quantitative Divergence from Rodents
In macaque V1, the "unique variance" explained by movement was near zero (~0.1% or less) once retinal input was controlled. This stands in sharp contrast to mouse V1, where movement can explain more variance than the actual visual stimulus. In primates, the variance explained by the task and stimulus was roughly eight times larger than the variance explained by movement and slow drifts combined.
Attention vs. Fidgeting
The researchers also compared the effect of spontaneous movements to spatial attention. They found that attention modulates V1 activity about ten times more strongly than body movements do. Furthermore, these two effects are independent—a neuron sensitive to attention is not necessarily sensitive to movement, suggesting distinct neural pathways for cognitive vs. motor modulation.
Discussion
These results suggest that the primate visual system is much more modular than the rodent system. It appears evolved to "filter out" the noise of bodily movements to maintain a stable representation of the visual world—a requirement that may be more stringent for animals relying on high-resolution foveal vision.
While this study focused on early and mid-level visual areas, embodiment signals might emerge further up the visual hierarchy in the parietal or temporal lobes. Additionally, while head-fixation was used to match mouse study protocols, future work could investigate if "natural" movement (where the head and body move together) involves different neural circuits.
Data Availability: The dataset is available on Figshare and the analysis code is hosted on GitHub.
Summary updated