Neuron Signals Break Established Rules of Random Movement

Partha Ghose, at the Indian Statistical Institute, Kolkata, and colleagues have found that Leggett-Garg inequalities, commonly employed in quantum mechanics, can be applied to analyse temporal correlations in neuronal dynamics. Diffusive dynamics conform to these inequalities, but persistent stochastic dynamics may breach them, indicating memory and contextual temporal structure. The work offers a new experimental method for understanding non-Markovian behaviour in neural systems, focusing on the possibility of finite-velocity persistent stochastic transport as a mechanism for these observed correlations.

Detecting neuronal memory using Leggett-Garg temporal correlations

Leggett-Garg-type temporal correlations served as the key analytical framework for this investigation, providing a mathematical rule to detect persistent behaviour or memory within neuronal signals. Originally developed to explore the foundations of quantum mechanics and the measurement problem, these inequalities provide a stringent test for the degree to which a system’s future behaviour is predetermined by its past. In the context of neuronal dynamics, the technique examines the relationship between a neuron’s state at one moment and its state at later moments, revealing whether it behaves randomly or retains a ‘memory’ of its past. Specifically, the analysis focuses on correlations in the neuron’s electrical signals over time, determining if its activity can be predicted by simple diffusion, or if more complex, time-dependent processes are at play. The Leggett-Garg inequality, in its simplest form, posits a limit on the correlation between measurements made at different times; violations of this inequality suggest the presence of memory or non-classical correlations. This approach allows researchers to quantify the degree to which a neuron’s present state is influenced by its previous states, providing insights into the mechanisms underlying neuronal information processing.

Adapting this concept from quantum mechanics, scientists circumvent the need to prove quantum effects within the brain, instead focusing on identifying non-diffusive dynamics through statistical analysis. The Telegrapher’s equation, a model describing finite-velocity particle movement, explains the observed non-diffusive dynamics. This equation, derived from the diffusion equation under the constraint of a finite particle velocity, provides a more accurate description of systems where particles do not instantaneously propagate information. The investigation examined whether neuron activity followed random diffusion, or exhibited ‘memory’ of past states. The choice of the Telegrapher’s equation is crucial, as it allows for the modelling of persistent stochastic processes where a particle’s velocity remains constant for a certain duration before randomly changing direction, contrasting with the continuous, instantaneous velocity changes assumed in standard diffusion models. This persistence is key to generating the observed temporal correlations.

Neuronal signalling exceeds diffusive limits via persistent stochastic dynamics

Evidence now challenges the key number of -1, representing the maximum bound for purely diffusive dynamics, in single neutron behaviour. Prior research established this limit as definitive for systems following simple, trajectory-based diffusion; however, persistent stochastic dynamics can generate oscillatory temporal correlations capable of exceeding this threshold. This offers a new analytical framework to probe neural dynamics without invoking quantum coherence, focusing instead on the possibility of finite-velocity persistent stochastic transport. The value of -1 arises from the assumption that correlations decay monotonically with time in a purely diffusive system. Any deviation from this value suggests the presence of non-diffusive processes that maintain or even enhance temporal correlations.

Tests reveal contextual and non-Markovian structure, mathematically analogous to quantum systems, within neuronal signalling. Single neuron behaviour exhibits temporal correlations exceeding -1, a previously established limit for purely diffusive processes. This violation of the Leggett-Garg inequality is a significant finding, as it suggests that neuronal dynamics are not solely governed by random diffusion but are influenced by persistent stochastic processes. Neuronal signalling analysis revealed oscillatory patterns, indicating a degree of ‘memory’ in the system’s evolution; specifically, the observed correlations persisted for durations beyond those expected from simple, trajectory-based diffusion. These oscillations suggest that the neuron’s past activity continues to influence its present state, creating a form of temporal context.

A dichotomic observable, based on spike occurrence, was employed to probe temporal structure in neural activity without invoking quantum effects. This simplification allows for a clear and quantifiable measure of neuronal activity, focusing solely on the presence or absence of action potentials. Further analysis demonstrated that a Kac-type stochastic process, modelling persistent random movement, naturally generates these non-diffusive correlations, unlike standard diffusive models which predict monotonic decay. Kac-type processes are characterised by a ‘waiting time’ distribution, where a particle moves at a constant velocity for a random duration before changing direction. This persistence is crucial for generating the observed temporal correlations. Currently, these results focus on single neurons and do not yet demonstrate how such dynamics scale to larger networks, or whether they offer a pathway to improved computational models of brain function. Understanding how these dynamics interact within complex neural circuits is a crucial next step in this research.

Detecting signal persistence and non-Markovian dynamics in neuronal activity

Researchers are developing new ways to understand how neurons process information, moving beyond established models that treat signal transmission as simple diffusion. This research proposes a method to detect whether neuronal activity exhibits ‘memory’ of past signals, using mathematical tools borrowed from quantum mechanics; however, the current focus remains firmly on classical stochastic processes, not quantum effects within the brain itself. The emphasis on classical stochasticity is important, as it avoids the complexities and uncertainties associated with invoking quantum phenomena in biological systems. Despite this conservative approach, a key tension arises from the mathematical analogy between these neuronal dynamics and quantum systems, specifically the potential for complex, non-Markovian behaviour.

Acknowledging the mathematical analogy does not imply quantum effects are actually occurring within neurons. Concerns about equating neuronal behaviour with quantum systems are valid, given the vastly different scales and environments involved. The brain operates in a warm, wet, and noisy environment, which is fundamentally different from the carefully controlled conditions required for observing quantum effects. Identifying these ‘memory’ effects, persistence in signals beyond simple diffusion, is important for building more accurate brain models. Demonstrating non-diffusive processes clarifies how neurons might integrate past inputs, influencing present responses and offering a richer understanding of neural computation. This integration of past information could be crucial for tasks such as pattern recognition, decision-making, and learning.

This research establishes a new method for examining how signals travel within single neurons. Adapting the Leggett-Garg inequality, originally used in quantum physics, scientists can now investigate whether neuronal activity retains a ‘memory’ of past signals without claiming quantum processes are at play within the brain. Demonstrating that persistent stochastic dynamics can violate this inequality suggests neurons may utilise more complex, time-dependent processes than previously thought, opening the possibility of contextual and non-Markovian structure within neural signalling. Non-Markovian dynamics imply that the future state of the neuron depends not only on its present state but also on its entire past history, making it a more complex and potentially more powerful information processing unit.

The research demonstrated that single-neuron dynamics can exhibit temporal correlations which violate Leggett-Garg inequalities, suggesting activity is not simply diffusive. This finding indicates neurons may retain a ‘memory’ of past signals and utilise more complex, time-dependent processes than previously understood. By adapting a tool from quantum physics, scientists were able to probe for these ‘memory’ effects without invoking quantum coherence in the brain. The authors suggest these tests offer a way to examine contextual and non-Markovian structure in neural dynamics, potentially improving the accuracy of brain models.