Summary: New research shows the brain begins summarizing complex visual scenes far earlier than previously believed. The study demonstrates that the primary visual cortex (V1)—long considered responsible mainly for basic features like edges and local motion—actually computes ensemble statistics. V1 compresses many individual motion signals into a concise “gist,” encoding both the average direction and the amount of variability (noise). These summary statistics are then forwarded to the posterior parietal cortex (PPC), where they are transformed into abstract category representations that support decision-making.
When animals navigate visually busy environments, the nervous system cannot inspect every element in detail. Instead, it forms rapid statistical summaries—such as the mean direction of motion across many elements—to represent the overall structure of the scene. This ability, known as ensemble perception, enables fast, efficient perception, but where and how these summaries are computed has been uncertain until now.
Key Facts
- Early compression in V1: The primary visual cortex does more than relay raw inputs; it encodes statistical summaries (mean and variance) of complex motion patterns at an early cortical stage.
- Value of untuned neurons: Neurons that appear untuned at the single-cell level still contribute meaningfully to the population code, underscoring the importance of distributed representations.
- Behavioral influence: Task demands bias V1 activity—during active categorization, V1 representations shift toward learned category centers, showing that early sensory processing is shaped by context and learning.
- Hierarchical processing: V1 computes the numerical summaries of sensory input, while PPC converts those summaries into category-level signals that guide behavior.
A research team led by Doyun Lee and Yee-Joon Kim at the Center for Memory and Glioscience, Institute for Basic Science (IBS), investigated how and where these visual summaries arise. Their results indicate that the transformation from detailed sensory input to compact summary statistics begins in V1, the first cortical area to receive visual information from the eyes.
Traditionally, V1 was thought to represent simple, local features such as edges or the direction of a single moving element, while higher areas like PPC integrated those features into decision-relevant signals. The new data reveal that V1 already carries population-level information about the global mean direction of motion and the variability of that motion. PPC then builds on these summaries, reorganizing them into more abstract, task-relevant categories.
To study this process, the researchers trained head-fixed mice to classify random-dot motion displays according to overall direction. Unlike typical coherent-motion stimuli, each dot in these displays moved in a direction drawn from a controlled distribution. By varying both the distribution mean and its spread, the team could separate the brain’s representation of average direction from its representation of uncertainty.
Mice learned to group eight possible mean motion directions into two categories. They successfully categorized stimuli even when individual dot directions were highly variable, demonstrating that they relied on a global statistical summary of the scene rather than a few prominent local cues.
Using miniscope calcium imaging, the team recorded neural activity in both V1 and PPC while animals performed the task and while they passively viewed the same stimuli. Although only a modest fraction of individual neurons showed clear selectivity for the global mean direction, population-level activity in both areas robustly encoded the mean direction. Importantly, V1 population responses also carried reliable information about stimulus variance.
During active categorization, V1 representations were systematically biased toward the center of the learned category, indicating that even early visual signals are modulated by behavioral context. The study also highlights the role of neurons that lack strong single-cell tuning: when combined across the population, these neurons contributed substantially to accurate representations of global motion.
The researchers conclude that the visual system progressively reorganizes information: initial cortical stages compute compact summary statistics, and higher areas bind those statistics into abstract category signals that can drive decisions. This hierarchical encoding of summary statistics provides an efficient strategy for extracting meaningful structure from noisy, complex scenes and may offer inspiration for improvements in artificial vision systems.
Key Questions Answered:
A: No. The brain summarizes the swarm by computing the average direction of motion and the overall level of variability, treating the group as a single statistical object rather than tracking each individual.
A: An “untuned” neuron is one that does not show strong preference for a particular stimulus when measured alone. This study shows that such neurons still contribute to the population code: collectively, they improve the fidelity of the brain’s summary representations.
A: Yes. Emulating early statistical compression—summarizing noisy inputs into concise statistics—could help artificial vision systems become faster and more robust when processing complex, noisy environments.
Editorial Notes:
- This article was edited by a Neuroscience News editor.
- The journal paper was reviewed in full by staff.
- Additional context was added to help readers interpret the findings.
About this visual neuroscience research news
Author: William Suh
Source: Institute for Basic Science
Contact: William Suh – Institute for Basic Science
Image: The image is credited to Neuroscience News
Original Research: Open access. “Hierarchical summary statistics encoding across primary visual and posterior parietal cortices” by Young-Beom Lee, Oliver James, Gaeun Jung, Doyun Lee, Yee-Joon Kim. Advanced Science. DOI: 10.1002/advs.202512369
Abstract
Hierarchical summary statistics encoding across primary visual and posterior parietal cortices
Although mounting evidence suggests the visual system pools sensory inputs into compact summary statistics, the neural mechanisms supporting this process were not fully understood. The authors characterized how summary statistics are represented at single-cell and population levels by imaging calcium signals in primary visual cortex (V1) and posterior parietal cortex (PPC) while head-fixed mice either passively viewed or actively categorized eight mean motion directions of randomly moving dots into two groups.
A minority of neurons in both regions showed selectivity for the global mean direction that exceeded what would be expected from linear summation of local motion responses. Despite variability in single-cell selectivity across stimulus conditions and trials, population activity in both areas robustly encoded the global mean. V1 population representations depended on stimulus variance and shifted toward category centers during active categorization. These results, along with population-level coding of stimulus variance, indicate that multivariate activity in V1 is well suited for encoding summary statistics. Redundant encodings across V1 and PPC suggest accumulation of summary information along the visual hierarchy, allowing PPC to bind multiple levels of statistical summaries into task-oriented category signals.