Response function as a quantitative measure of consciousness in brain dynamics

Wenkang Du 1, Haiping Huang 1,2 · 2025

Read Response function as a quantitative measure of consciousness in brain dynamics by Wenkang Du 1, Haiping Huang 1,2 (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2509-00730.txt.

Introduction

Wenkang Du and Haiping Huang propose a measure of consciousness based on neural response. They analyze recordings from inside the brains of non-human primates. The recordings cover wakefulness, anesthesia, and recovery states. A recurrent neural network model is fit to the cortical dynamics. The fit extracts how strongly the cortex responds to disturbance.

Response strength tracks the conscious state closely in their data. It is strong and widely spread during wakefulness. It is suppressed under anesthesia, while dynamics grow more chaotic. It rises again as the animal recovers. The pattern matches theory on response near a critical border. That border separates ordered regimes from chaotic ones. The pair suggests such a response may be a needed condition for consciousness.

Catalog record

CREATORWenkang Du 1, Haiping Huang 1,2
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE25 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2509-00730

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Response function as a quantitative measure of consciousness in brain dynamics Wenkang Du 1, Haiping Huang 1,2 Abstract Understanding the neural correlates of consciousness remains a central challenge in neuroscience. In this study, we investigate the relationship between consciousness and neural responsiveness by analyzing intracranial ECoG recordings from non-human primates across three distinct states: wakefulness, anesthesia, and recovery. Using a nonequilibrium recurrent neural network (RNN) model, we fit state-dependent cortical dynamics to extract the neural response function as a dynamics complexity indicator. Our findings demonstrate that the amplitude of the neural response function serves as a robust dynamical indicator of conscious state, consistent with the role of a linear response function in statistical physics. Notably, this aligns with our previous theoretical results showing that the response function in RNNs peaks near the transition between ordered and chaotic regimes—highlighting criticality as a potential principle for sustaining flexible and responsive cortical dynamics. Empirically, we find that during wakefulness, neural responsiveness is strong, widely distributed, and consistent with rich nonequilibrium fluctuations. Under anesthesia, response amplitudes are significantly suppressed, and the network dynamics become more chaotic, indicating a loss of dynamical sensitivity. During recovery, the neural response function is elevated, supporting the gradual re-establishment of flexible and responsive activity that parallels the restoration of conscious processing. Our work suggests that a robust, brain-state-dependent neural response function may be a necessary dynamical condition for consciousness, providing a principled framework for quantifying levels of consciousness in terms of nonequilibrium responsiveness in the brain. I Introduction Understanding the neural basis of consciousness remains one of the most challenging questions in neuroscience [ 1 , 2 , 3 ] . Despite remarkable advances in neuroimaging and electrophysiology, reliably distinguishing conscious from unconscious brain states purely on the basis of measurable neural activity remains elusive [ 4 , 5 , 6 ] . Clinically, this gap limits our ability to monitor depth of anesthesia, diagnose disorders of consciousness, and understand the nature of brain-state transitions among wakefulness, sleep, anesthesia, and coma. Brain, as a hierarchical system of interacting neurons, lacks a clear theoretical foundation to determine what should be defined and measured in mathematical terms [ 7 ] . A growing number of works suggest that consciousness is not defined by static structural features alone, but by the brain dynamics far from equilibrium [ 8 , 9 , 10 , 11 , 12 ] . Empirical evidences now indicate that conscious states are underpinned by rich, flexible, and metastable patterns of neural activity, and thus, dynamics at the edge of instability are essential for maintaining consciousness [ 5 ] . For example, dynamical system approaches have revealed that cortical activity during wakefulness is poised near the edge of instability—a critical regime that maximizes responsiveness to external stimuli [ 13 , 14 ] . Complexity analysis showed that attractors in conscious and anesthesia-induced unconscious states exhibit significantly different shapes, which affects the information processing capability [ 5 , 15 ] . These empirical observations can be studied within the perspective of nonequilibrium statistical physics. Recent studies show that the human brain, like all living systems, fundamentally operates out of thermodynamic equilibrium [ 8 , 12 ] . Measures such as entropy production, broken detailed balance, and probability flux irreversibility have emerged as principled signatures of cognitive complexity and conscious awareness [ 9 , 11 ] . For instance, healthy conscious brain dynamics break time-reversal symmetry [ 16 ] , generating a preferred direction in time (i.e., arrow of time), whereas proximity to equilibrium—as in anesthesia, non-rapid-eye-movement sleep, or neurodegenerative disorders like Alzheimer’s disease—is accompanied by a loss of temporal irreversibility [ 17 ] . Our recent theoretical works further support this dynamic complexity by showing that in recurrent neural networks (despite random connections among neurons), the neural response function itself peaks exactly at the edge of chaos, where the system exhibits maximal sensitivity to perturbations [ 18 , 19 ] . Within an optimization-based framework for non-equilibrium dynamics, this critical regime naturally emerges as the point where the responsive property of the network is maximized, indicating a deep link between dynamical instability and the network’s capacity for flexible information processing. This is also consistently supported by an analysis of cortical electrodynamics [ 14 ] showing that the information richness occurs at the edge of chaos, and the conscious brain states are located near the edge of chaos. Another study focusing on a clinical index of conscious level (but on recurrent neural networks) also identifies this picture [ 20 ] . In this work, we verify this theoretical concept in real brain dynamics by building a data-driven recurrent neural network, where the couplings between neurons (or brain regions) are learned from a predictive processing of the time series collected from experiments [ 21 ] . The recurrent neural network (RNN) model fits high-density electrocorticography (ECoG) recordings from the cortex of non-human primates during reversible loss and recovery of consciousness. The main methodology is based on a predictive learning principle [ 22 ] . After the network is reconstructed, we obtain the functional connectivity among all involved brain regions. As derived by our previous theory [ 18 ] , the response function is measured with a tiny current perturbation to the neural dynamics equation. Within the statistical physics framework, we provide a principled and mechanistic approach to quantify how cortical dynamics sustain responsiveness far from equilibrium, and how this capacity deteriorates under anesthesia and re-emerges during recovery. While our modeling currently focuses on primate data, the methodology and analysis may lay the foundation for future extensions to human datasets and for understanding how the responsiveness of nonequilibrium brain dynamics is correlated with conscious awareness. II Model and methodology In this section, we describe the network model used to fit the observed time series data. We first give an overall introduction to the training dataset, and then to the RNN model and training method. Finally, we introduce the response function as a brain complexity index and show how to measure this function from the reconstructed networks. Figure 1: Sketch of experimental data: 2D ECoG electrode array and neural activity signals for Monkey George [ 23 ] . Left: Spatial layout of the 128‑channel ECoG grid covering frontal, parietal, and temporal cortices (image adapted from [ 23 ] ). This map serves as the anatomical reference for interpreting the fitted RNN dynamics across cortical regions. Right: Representative neural activity traces from channels 1 and 128 during three behavioral states—wakefulness, anesthesia, and recovery. The purple shaded region marks the time window used for model fitting in this study. II.1 Data description In this study, we analyzed electrocorticography (ECoG) recordings from a publicly available dataset of non-human primates provided by the Neurotycho project [ 21 , 23 ] . The data were collected at the Laboratory for Adaptive Intelligence, Brain Science Institute, RIKEN, from four male monkeys under different states of consciousness, including wakefulness, deep sleep, and anesthesia induced by anaesthetic drugs (propofol, ketamine, and ketamine combined with medetomidine). The ECoG array consists of 128 electrodes covering the prefrontal cortice, primary motor and somatosensory cortices, parietal and temporal cortices, and visual cortices, providing high-density cortical coverage. ECoG signals were recorded at a sampling rate of 1\mathrm{kHz}

Read the complete text