Summary: A new SISSA study shows that rats possess a cluster of visual neurons that function like primate “pattern cells,” enabling precise perception of motion direction.
By recording neurons in the rat visual cortex and testing their responses with state-of-the-art artificial neural network models, researchers identified cells that integrate local motion signals into coherent, global motion representations. This supports the idea that rodents have more advanced motion-processing capabilities than previously appreciated.
The finding strengthens the case for using rats as models in vision research and may inform the design of artificial vision systems as well as studies of visual disorders.
Key Facts:
- Scientists discovered neurons in the rat visual cortex that behave like primate “pattern cells,” which are critical for accurate motion perception.
- Deep-learning models were used to validate that these neurons perform nonlinear integration of motion signals, not just simple linear filtering.
- The results reinforce the utility of rodents as model organisms for studying visual processing and for developing computational and artificial-vision models.
Source: SISSA
Rats show specialized neural mechanisms for perceiving motion
Researchers from SISSA, led by Prof. Davide Zoccolan, investigated how rats estimate the direction of moving objects—a task that presents a known challenge in vision science called the “aperture problem.” By recording from neurons in primary visual cortex (V1) and the lateromedial area (LM), the team tested whether some rat neurons integrate motion signals across local features to produce a unified sense of global direction.
The team found a small but distinct population of neurons whose responses could not be explained by simple linear receptive-field properties. Instead, these neurons displayed tuning consistent with true nonlinear integration of motion signals—an essential feature of pattern cells in the primate dorsal stream. The investigators used deep neural network models that mimic dorsal-stream processing to predict neuronal responses, and the models matched the behavior of these rat neurons.
These results indicate that, despite notable anatomical and organizational differences between rodent and primate visual cortex, rodents can implement similar computational strategies for motion representation. In rats the pattern-like neurons are fewer and more widely distributed across visual areas than in primates, where they form more concentrated populations in higher visual regions.
Understanding the aperture problem and its solution
The aperture problem arises because neurons early in the visual pathway have small receptive fields and respond to oriented image segments. Each such neuron can detect only the component of motion perpendicular to the contour it encodes, making it ambiguous to infer the true global motion of an object from any single local signal.
Solving this ambiguity requires integration across many local signals. In primates, this integration occurs along the dorsal visual stream, where higher-order areas combine component signals into a pattern-invariant representation of global motion. The SISSA study shows that rats also possess neurons capable of performing this computation, offering a realistic representation of object motion rather than only local motion components.
Artificial intelligence helped confirm true integration
To distinguish genuine pattern-like integration from artifacts of receptive-field geometry, the researchers used deep learning models trained to emulate dorsal-stream processing. Predictive models based on these artificial neural networks successfully accounted for the responses of the candidate pattern cells, allowing the team to reject alternative linear explanations. This computational approach was key to confirming that these rat neurons perform nonlinear integration of motion signals.
Why the study matters
The findings have several important implications. First, they show that a relatively simple cortical architecture can still support high-level motion-processing units, expanding our comparative understanding of visual systems across species. Second, the results motivate further computational and experimental work to reveal the circuit mechanisms that produce pattern-like responses. Finally, because rodents are widely used in neuroscience with powerful experimental tools available, demonstrating comparable motion-processing capabilities strengthens their role as model systems for visual neuroscience and for studying related pathologies.
Future research will need to map the specific circuits that give rise to these pattern-like neurons in rats and to determine how they contribute to behavior in naturalistic tasks. For now, the study provides strong evidence that rodents use specialized global-motion detectors, and that AI models can be indispensable tools to uncover and validate such neural computations.
About this visual neuroscience research news
Author: Donato Ramani
Source: SISSA
Contact: Donato Ramani – SISSA
Image: The image is credited to Neuroscience News
Original Research: Open access.
“Truly pattern: Nonlinear integration of motion signals is required to account for the responses of pattern cells in rat visual cortex” by Davide Zoccolan et al., Science Advances
Abstract
Truly pattern: Nonlinear integration of motion signals is required to account for the responses of pattern cells in rat visual cortex
Pattern cells are a defining feature of advanced motion processing in the primate dorsal stream: they integrate local motion signals into representations that preserve global direction. Pattern-like neurons have been reported in rodent visual cortex, but until now it was unclear whether their tuning arose from genuine nonlinear integration or from particular linear receptive-field geometries. This study demonstrates that pattern cells in rat V1 and LM exhibit tuning that cannot be explained by linear spatiotemporal receptive fields. Instead, their behavior matches units in a contemporary neural network model of dorsal-stream processing, suggesting shared cortical mechanisms for motion representation across mammals and supporting the use of rodents as model systems to study the circuit-level basis of motion perception.