🧠 neuroscience

Neuronal avalanches in awake macaque V1 and V4

This study demonstrates that neuronal avalanches in the awake macaque early visual cortex (V1 and V4) exhibit robust power-law scaling and critical dynamics similar to those previously observed in motor areas, despite the presence of strong structured afferent drive.

Original authors: Solovyeva, K., Popov, P., Makarova, A., Lebedev, M.

Published 2026-10-16

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Solovyeva, K., Popov, P., Makarova, A., Lebedev, M.

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ⚕️ This is an AI-generated explanation of a preprint that has not been peer-reviewed. It is not medical advice. Do not make health decisions based on this content. Read full disclaimer

The brain is not a static machine; it is a living network that never truly rests. Even when we are not thinking about anything in particular, billions of neurons fire in complex, shifting patterns. For decades, scientists have wondered if this background noise is random chaos or if it follows a hidden order. One compelling idea suggests that the brain operates at a "critical point," a delicate balance between order and chaos where information flows most efficiently. In this state, small sparks of activity can sometimes die out quickly, but other times they can cascade into massive waves of synchronized firing, much like a small snowflake triggering a large avalanche. These cascades, known as neuronal avalanches, appear to follow a specific mathematical pattern where the frequency of small events is high, but large events, while rare, still happen often enough to fit a predictable curve. If the brain truly lives at this critical point, it would mean the network is perfectly tuned to process information, neither too rigid to change nor too chaotic to hold a thought.

Until recently, this phenomenon had been clearly observed in the motor areas of awake monkeys, which control movement, and in brain slices from animals under anesthesia. However, a crucial question remained unanswered: does this same critical organization exist in the visual cortex of an awake animal? The visual system is unique because it is constantly bombarded by strong, structured signals from the eyes. Scientists needed to know if this heavy external input would disrupt the delicate balance of the avalanches or if the brain's internal rules for organizing activity were strong enough to persist even while the animal was looking at the world.

To answer this, researchers turned to a massive dataset recorded from two adult macaque monkeys. These animals had been implanted with arrays containing over a thousand tiny electrodes, spread across two key areas of the visual brain: the primary visual cortex, which receives the first signals from the eyes, and a higher area that processes more complex visual details. The monkeys were sitting quietly in a dark room, not performing any specific task, allowing the scientists to listen to the brain's spontaneous activity. The researchers focused on moments when the animals were awake, carefully filtering out periods of sleep or drowsiness, and they paid close attention to whether the monkeys' eyes were open or closed. By analyzing the electrical signals from these electrodes, they looked for the tell-tale signs of avalanches: clusters of activity that started small and grew, following the specific statistical pattern that suggests a system operating at a critical point.

The results were striking. In both monkeys, the spontaneous electrical activity formed these cascades exactly as predicted. The sizes of the avalanches followed the expected pattern over a range spanning more than ten times the smallest event, confirming that the visual cortex, even while receiving strong signals from the eyes, is organized in the same way as the motor cortex. The researchers tested the robustness of this finding by changing the rules of their analysis. They looked at the data with different time windows, raised the threshold for what counted as an event, and even removed large numbers of electrodes to see if the pattern held up. In every case, the power-law relationship remained intact. When they shuffled the timing of the signals so that events on different electrodes no longer happened together, the pattern vanished, proving that the organization was real and not just a statistical fluke. They also found that the cutoff point, where the largest avalanches stopped, scaled perfectly with the number of electrodes they were watching, suggesting that the size of the biggest events is limited only by how much of the brain they can see, not by the brain's internal dynamics.

One of the most intriguing parts of the study involved the monkeys' eye state. The researchers compared the brain activity when the animals' eyes were open versus when they were closed. Intuitively, one might expect the brain to behave very differently when the flood of visual input is cut off. However, the study found that the fundamental statistics of the avalanches remained virtually unchanged. Whether the eyes were open or closed, the size of the cascades, the speed at which they grew, and the balance of activity within the network stayed the same. The differences observed were so small that they were far smaller than the natural variations seen between different recording sites in the same brain. This suggests that the brain's internal organization is remarkably stable and robust, capable of maintaining its critical state even when the external world changes dramatically.

The researchers also noted that while the overall pattern was consistent, there were subtle differences between the two monkeys and between different recording locations within the brain. These variations were larger than the changes caused by opening or closing the eyes, hinting that the specific architecture of the tissue or the technical details of the recording might shape the exact numbers. The study did not find evidence that the brain was operating in a state of perfect, unchanging criticality, but rather that it maintained a stable, critical-like organization that is resilient to changes in sensory input. This work extends our understanding of how the brain works, showing that the elegant, scale-free organization of neuronal avalanches is not limited to motor control or quiet, anesthetized states, but is a fundamental feature of the awake, sensory-processing brain as well.

Problem and Motivation
Neuronal avalanches—cascades of synchronized activity following power-law size distributions—are a hallmark of cortical networks operating near a critical point. While these dynamics and their associated scaling invariances (temporal, spatial, and threshold) have been robustly demonstrated in the motor and premotor cortices of awake primates, it remains untested whether the early visual cortex of awake primates shares these properties. The visual cortex presents a unique challenge: it receives strong, structured afferent drive from the retina, which could theoretically modulate avalanche statistics or prevent the system from operating near criticality. Furthermore, differences in receptive field size and intrinsic timescales between primary visual cortex (V1) and higher visual area V4 might yield distinct dynamical regimes. This study investigates whether spontaneous local field potential (LFP) activity in the awake macaque visual cortex exhibits the same avalanche organization and invariances previously observed in motor areas.

Methodology
The authors analyzed a public, high-density dataset (Chen et al.) comprising resting-state recordings from two adult male macaques (monkeys L and A). Each animal was implanted with sixteen 8×8 Utah arrays (1,024 channels total) spanning V1 and V4. The analysis focused on three sessions per animal, utilizing pupil-based segmentation to distinguish between eyes-open and eyes-closed epochs.

Key methodological steps included:

Preprocessing: Channels with a signal-to-noise ratio (SNR) < 2 were discarded. Hardware cross-talk (synchrofacts) was mitigated by removing the top 50 electrodes from a ranked removal list provided by the dataset authors. Slow-wave sleep/drowsiness periods were excluded based on spectral criteria (power in 1–10 Hz).
Event Detection: Negative LFP deflections (nLFPs) were detected as contiguous excursions below a threshold (z = 3 standard deviations, tested from 1.5 to 4.5 SD).
Avalanche Definition: Avalanches were defined as maximal runs of consecutive time bins (widths $\Delta t$ = 2, 4, 8, 16, 32 ms) containing at least one nLFP, bounded by empty bins.
Statistical Analysis: Size and lifetime distributions were fitted using maximum likelihood estimation. The study tested for three specific invariances characteristic of criticality:
1. Temporal Invariance: The power-law exponent $\alpha$ scales with bin width as $\alpha \propto \Delta t^\beta$ .
2. Spatial Invariance: The distribution cutoff scales proportionally with the number of electrodes analyzed.
3. Threshold Invariance: The power-law form is preserved when the detection threshold is raised, whereas random thinning destroys it.
Controls: Per-electrode time-shuffled surrogates (preserving rates but destroying cross-electrode timing) and random event removal controls were employed to verify that observed heavy tails were not artifacts of single-channel statistics or random sampling.
Comparisons: Parameters were compared across eye states (using equivalence testing), cortical areas (V1 vs. V4), and animals, while accounting for array-to-array variability via linear mixed models.

Key Results

Power-Law Distributions: In both animals, avalanche sizes followed power laws over more than a 10-fold range. The maximum-likelihood exponents ( $\alpha$ ) ranged from approximately -1.82 to -1.49 depending on the bin width and animal. Shuffling electrode events abolished these heavy tails, converting distributions to rapidly decaying exponentials.
Temporal Scaling: The exponent $\alpha$ decreased monotonically with increasing bin width. The scaling exponent $\beta$ was estimated at -0.045 and -0.046 for monkey L (eyes open/closed) and -0.031 for monkey A. While these values are generally consistent with motor cortex findings, $\beta$ in monkey A fell slightly outside the previously reported range, and the amplitude rescaling factor from the original motor cortex study did not collapse the visual cortex data.
Spatial Scaling: The cutoff of the avalanche size distribution scaled linearly with the number of electrodes analyzed ( $D \approx 0.99$ ), confirming that the largest avalanches are bounded by the observation window rather than intrinsic dynamics.
Threshold Invariance: Raising the detection threshold from 1.5 to 4.5 SD reduced the event rate by over 100-fold but preserved the power-law form. The slope of $\alpha$ versus threshold was small (~~0.03 per SD), significantly distinct from the steep slope (~~1.0 per SD) observed when events were removed randomly. This indicates that large avalanches are not random aggregates of small events.
Branching Ratio: Three estimators of the branching ratio ( $\sigma$ ) were computed. The "Beggs-Plenz" estimator yielded values near unity ( $\sigma \approx 0.95–1.04$ ) at small bin widths, consistent with criticality, though values varied significantly with bin width and estimator choice.
Eye State and Area: Equivalence testing revealed that eye closure (eyes open vs. closed) did not produce statistically significant changes in $\alpha$ , $\beta$ , or $\sigma$ within predefined bounds. Similarly, no significant differences were found between V1 and V4 in monkey L. The largest source of variance in the parameters was the specific recording array (site), which accounted for 83–95% of the variance in $\alpha$ , far exceeding the variance explained by eye state or cortical area.

Significance and Claims
The paper claims that the organization of neuronal avalanches extends to the awake early visual cortex of primates, despite the presence of strong structured afferent drive. The study demonstrates that V1 and V4 exhibit power-law size distributions, heavy-tailed lifetimes, and the three key invariances (temporal, spatial, and threshold) that characterize critical branching processes in motor cortex.

However, the authors maintain a modest stance regarding the exactness of these findings:

The invariances are described as approximate rather than exact. Quantitative departures were noted, including a shallower $\beta$ in one animal, a statistically detectable but small drift in $\alpha$ with threshold, and the failure of the original amplitude rescaling to collapse the data.
The authors suggest that these parameters are shaped partly by technical factors, such as electrode count and array position, rather than purely by functional architecture.
The lack of effect from eye closure is interpreted as either evidence of robustness to changes in afferent drive or an indication that spontaneous eye closure is too weak a perturbation (and too confounded with vigilance) to displace the network from its operating point.
The study concludes that while sensory cortex shares the general avalanche motif with motor cortex, the specific scaling parameters may be approximate and sensitive to recording conditions, suggesting that the invariances are robust features of cortical dynamics but not necessarily exact mathematical constants across all contexts.

More like this