The study, authored by Cristian Mendico of the Institut de Mathématique de Bourgogne at Université Bourgogne Europe, is the computational counterpart to a companion theoretical paper on branched optimal transport for brain mapping. Its central claim is provocative: what is often treated as a problem of steering dynamics on a prescribed connectome may actually be a problem of architecture inference. Rather than asking “how does signal travel along this network?”, the framework asks “which network best explains how stimulation becomes reaction?”

The technical foundation rests on an old idea from applied mathematics that is finding surprising new life in neuroscience. In classical network control models, researchers fix a structural substrate — usually derived from diffusion MRI tractography — and then compute the cost of driving the brain from one state to another. Those models, influential since landmark work on structural brain controllability, leave the substrate untouched. Mendico’s formulation makes the substrate the unknown variable. The optimization variable is not a trajectory or a control signal, but an oriented one-dimensional transport current whose support defines what the author calls a “brain reaction map.”

Mathematically, the problem is posed as a balance-constrained optimization. External stimulation is represented by a nonnegative source measure, and the reaction-producing neural configuration by a nonnegative target measure, both estimated directly from neuroimaging data and assumed to be balanced so the problem is conservative. The task is to find a transport structure that moves the source mass into the target configuration at minimal cost, subject to a conservation law. The crucial ingredient is a concave dependence of the cost on transported flux: because of this concavity, carrying two signals together along a shared segment is cheaper than carrying them separately. Minimizers therefore spontaneously form ramified, tree-like structures — shared “neural highways” that aggregate signal before redistributing it at branching points. This branching economy, rooted in the mathematics of branched transport developed by analysts such as Qinglan Xia, is precisely what the framework borrows to model neural propagation.

What distinguishes the new work is that the abstract measures are replaced by multimodal, data-driven estimates. The pipeline begins with task-related blood-oxygen-level-dependent (BOLD) responses analyzed through a block-design general linear model, which separates a stimulation regressor from a reaction regressor and produces region-wise contrast statistics with strong spatial selectivity. Early task epochs concentrate in visually and auditorily driven regions, while later epochs emphasize sensorimotor and default-mode-related areas. But BOLD alone localizes activity without resolving its timing, so the study fuses it with source-reconstructed EEG/MEG data — electrophysiological signals inverted onto the cortical surface using a lead-field forward model and a Tikhonov-regularized minimum-norm inverse. This second modality resolves the temporal separation cleanly: stimulus-locked components peak around 100 milliseconds in sensory-entry regions, while reaction-locked components dominate around 350 milliseconds in motor and higher-order regions.

The fusion step is geometric rather than additive. Modality-specific regional scores are combined through weighted geometric averaging — an exponent of roughly 0.55 on fMRI scores and its complement on EEG/MEG scores — and then normalized to produce balanced probability measures on a common regional support. Sensitivity analyses across the full range of fusion weights show that the dominant support of the resulting measures is stable, meaning the transport problem is not driven by a fragile, modality-specific artifact. The fusion of fMRI’s spatial specificity with EEG/MEG’s temporal precision yields supply and demand profiles that are structured enough to generate a non-trivial transport problem.

The anatomical prior enters through diffusion-informed anisotropy. Each region is assigned a synthetic diffusion tensor, and the transport cost of moving signal at a given position in a given direction is computed from the inverse of the tensor field, quadratic in form. This is more than a rescaling of distances: in the isotropic baseline, short edges are generically favored, but under anisotropic costs a short edge cutting across an implausible direction becomes expensive, while a longer edge aligned with dominant tensor axes becomes favorable. The anatomical prior changes the directional logic of admissible propagation. Midpoint quadrature along each candidate edge converts the continuous, direction-dependent cost density into a discrete edge cost on a candidate graph built by a k-nearest-neighbour rule.

The central result of the paper is the comparison between isotropic and anisotropic branched transport solutions. Both produce branched architectures, but they differ qualitatively, not merely quantitatively. The isotropic solution stays relatively close to direct geometric routing, while the anisotropic solution reorganizes the entire backbone around relay regions aligned with tractography-derived geometry. Branches that are weakly expressed under isotropic costs become dominant, and some direct alternatives vanish entirely. Comparison of edgewise fluxes reveals substantial redistribution of transported mass — exactly the kind of architectural reorganization that fixed-substrate control models are structurally incapable of revealing, since once the substrate is prescribed, anisotropy can modulate dynamics on it but cannot alter which routing map is selected in the first place.

Perhaps the most biologically striking finding concerns where the bottlenecks emerge. In the synthetic setting, the strongest branching interfaces preferentially appear in dorsal attention, salience/ventral attention, frontoparietal and thalamic nodes — systems commonly interpreted as integrative bridges between sensory input and distributed action or cognitive output. The model, in other words, does not simply recover short routes between sources and sinks; it selects a mesoscale backbone in which anatomically and functionally plausible association systems act as shared highways for signal aggregation before redistribution. That the optimization, armed only with multimodal activity measures and a diffusion-informed cost, gravitates toward these known integrative systems lends the synthetic proof-of-concept a degree of biological credibility that pure mathematical demonstrations often lack.

The study goes one step further by asking whether the geometrically optimal map is also dynamically plausible. Once a graph is inferred, it serves as the substrate for a graph-induced stochastic dynamics — a linear system with a graph-Laplacian propagation term, stimulus forcing, a control input, and graph-dependent noise. The dynamic cost is quantified as the minimum path-space control effort required to steer the stochastic process between prescribed endpoint distributions, in a spirit related to recent Schrödinger bridge formulations for brain state transitions. The resulting hybrid functional balances geometric transport efficiency against dynamical controllability, and its Pareto frontier — the set of non-dominated candidate maps in the plane of geometric cost versus control cost — exhibits a non-trivial geometry with local trade-offs across branching regimes.

The most consequential outcome is the discovery of rank reversals. As the weight given to the dynamical term increases, the ordering of candidate graphs changes: a graph that is geometrically optimal can lose to a dynamically cheaper competitor, and vice versa. Controlled stochastic trajectories do reach terminal states substantially closer to the prescribed reaction profile than uncontrolled ones, confirming the inferred map is dynamically usable. But geometric efficiency and dynamical controllability turn out to be related yet non-equivalent criteria. The author argues this is a conceptual message rather than a technical curiosity: geometry and dynamics should be treated as coupled principles of large-scale propagation, not interchangeable surrogates, and the natural selection object is the coupled geometric–dynamical landscape.

The study is explicitly a proof of principle, and the author is candid about its limits. The demonstrations use synthetic multimodal data on a low-dimensional cortical support of 18 regions of interest; the candidate graph is finite; the stochastic dynamics is linear; and the tractography prior is a coarse tensor field rather than subject-specific whole- brain tract reconstruction. Notably, although the transport optimization runs on directed arcs, the weighted adjacency matrix used in the dynamic layer is symmetrized before Laplacian construction — a choice the author frames as reflecting the fact that diffusion MRI does not resolve axonal polarity in vivo, not as a claim of biological symmetry. Scaling to human connectomes with 200 or more regions is described as conceptually straightforward but computationally demanding, pointing toward anatomically informed graph sparsification, warm-start strategies across branching exponents, and parallel evaluation of dynamic costs as practical extensions.

Even with these caveats, the reframing is significant. If the propagation substrate is allowed to vary, the correct inverse problem for stimulus-to-reaction transformation is no longer only to steer dynamics on a graph — it is to determine which graph best explains the transformation itself. That question, long hidden inside assumptions about the connectome, has now been given a variational, data-driven, and explicitly testable formulation.

Traditionally, neuroscientific studies have employed a variety of empirical methods to quantify timescales of neural activity, ranging from autocorrelation analyses of spike trains to spectral decompositions of local field potentials. Yet, despite this rich trove of data, a consensus on the precise definitions and measurement protocols for neural timescales remains elusive. Different brain regions exhibit remarkably distinct temporal signatures, which themselves shift with behavioral states such as attention, arousal, or task engagement. This complexity challenges the interpretation of timescales as simple readouts of brain function and calls for a rigorous computational approach to decipher their underlying mechanisms.

Computational models, especially those grounded in biophysical realism, provide a powerful lens for dissecting the origins of temporal diversity in the brain. Ion channel kinetics, synaptic dynamics, and network connectivity all impose constraints and affordances that shape the temporal profile of neuronal responses. For example, the interplay between excitatory and inhibitory neurons can give rise to slow fluctuations in population activity, which may underlie longer timescales observed during resting states. In contrast, faster timescales often emerge in circuits specialized for rapid sensory processing or motor execution. By fine-tuning model parameters to replicate empirical timescales, researchers can identify mechanistic principles that govern temporal diversity in neural circuits.

Beyond mechanistic insights, understanding neural timescales holds profound implications for cognition and behavior. Recent breakthroughs in machine learning and task-optimized neural networks demonstrate that varying timescales within an artificial system enable flexible information integration across different temporal contexts. In natural brains, this flexibility likely manifests in the capacity to maintain working memory, predict future inputs, or adjust behavior based on evolving contingencies. By explicitly incorporating timescale diversity into computational models, scientists can probe how temporal hierarchies support memory formation, attention modulation, and decision making in dynamic environments.

Importantly, the relationship between brain structure and timescales is not merely correlative but may reflect causal influences. Anatomical features such as dendritic arborization, synapse density, and the modular organization of cortical layers impose characteristic delays and integration windows that manifest as distinct timescales. Long-range connections across brain areas further enrich the temporal landscape by enabling feedback loops and recurrent processing over extended periods. Computational frameworks integrating structural connectivity data and dynamical simulations stand poised to reveal how anatomical architecture sculpts the neural timescale repertoire.

Equally critical is the role of behavioral state and environmental context in modulating neural timescales. Empirical studies show that timescales adapt dynamically as animals shift between rest, exploration, or task performance. Neuromodulatory systems, including cholinergic and noradrenergic pathways, sculpt neuronal excitability and synaptic efficacy, thereby transforming the temporal dynamics of neural ensembles. Simulations incorporating state-dependent parameter changes reproduce these transitions, highlighting how internal and external signals jointly regulate the temporal dimension of brain activity.

The synthesis of data analysis methods, biophysical modeling, and task-driven network modeling marks a turning point in the study of neural timescales. Such integrative computational approaches go beyond descriptive statistics to generate testable hypotheses about causal mechanisms. For instance, manipulations in simulated networks can demonstrate how specific ionic currents or connectivity motifs extend or compress timescales, providing experimentally verifiable predictions. Similarly, task-optimized artificial networks trained to perform cognitive tasks reveal how timescale diversity facilitates flexible information routing and robust performance under uncertainty.

At a broader level, these computational perspectives challenge conventional reductionist views of neural dynamics by emphasizing temporal complexity as a core organizational principle. Neural timescales are not mere epiphenomena but critical features that enable the brain to reconcile immediacy with history, sensory input with memory, and action with anticipation. Capturing this complexity requires tools that marry theory with data, and models with behavior—a challenge that the new computational paradigm tackles head-on.

Moreover, understanding neural timescales potentially illuminates pathophysiological processes underlying neurological disorders. Aberrant temporal dynamics have been implicated in conditions such as schizophrenia, autism, and neurodegenerative diseases, where disruptions in the balance of excitation, inhibition, or connectivity alter normal temporal processing. Computational models provide a sandbox for exploring how perturbations in timescale-generating mechanisms might lead to cognitive deficits or maladaptive behaviors, offering avenues for targeted interventions.

The promise of these computational frameworks extends into emerging technologies as well. Brain-machine interfaces, neuroprosthetics, and adaptive neurostimulation devices can benefit from insights into the temporal structure of neural signals. Designing algorithms that respect and leverage natural timescale hierarchies could enhance the fidelity, responsiveness, and interpretability of these systems, fostering seamless integration with biological neural networks.

Finally, the integrative view of neural timescales propels neuroscience toward a more holistic understanding of brain function—one that recognizes the intertwining of structure, dynamics, behavior, and cognition across time. It urges researchers to transcend static snapshots of activity and embrace models that capture brain function as a living, evolving process, inherently shaped by temporal complexity. As computational techniques evolve and enrich empirical inquiry, the secrets encoded within neural timescales will undoubtedly unlock new vistas into the brain’s inner workings and its remarkable capacity for adaptive intelligence.

Subject of Research: Neural timescales and their computational underpinnings in brain dynamics and cognition.

Article Title: Neural timescales from a computational perspective.

Article References:
Zeraati, R., Levina, A., Macke, J.H. et al. Neural timescales from a computational perspective. Nat Neurosci (2026). https://doi.org/10.1038/s41593-026-02343-8

Image Credits: AI Generated

DOI: https://doi.org/10.1038/s41593-026-02343-8

The study introduces a sophisticated meta-reward prediction error model that integrates a meta-learning parameter into traditional reinforcement learning frameworks. This innovation enables a nuanced understanding of how the brain dynamically adjusts its expectations and response strategies when previously reliable cues no longer predict reward outcomes as strongly. By applying this model to mouse behavioral data, the researchers demonstrated unprecedented accuracy in predicting cue-evoked licking behavior as the contingency between cues and rewards degraded or enhanced. This advancement provides a computational foundation for dissecting the neuronal substrates of flexible behavior.

Utilizing longitudinal two-photon calcium imaging, the team tracked neural activity in the mPFC with exquisite temporal and spatial resolution. This method allowed for the identification of a subpopulation of neurons that exhibited selective encoding of contingency degradation signals. These neurons displayed robust activity changes precisely when the relationship between cues and subsequent rewards shifted, highlighting their role in signaling the need to adapt behavior. This selective encoding underpins the brain’s capacity to halt established actions when they no longer yield expected outcomes.

The study further leveraged single-cell holographic optogenetics to causally test the involvement of these identified neuron ensembles. By selectively activating or inhibiting mPFC neurons encoding contingency degradation signals, researchers demonstrated a direct and significant impact on the animals’ behavioral flexibility. These manipulations accelerated or impeded the updating of learned behaviors, firmly establishing the causal role of these neurons in mediating cognitive flexibility.

Recognizing that adaptive behavior is not solely the province of cortical areas, the researchers investigated the interactions between the mPFC and the ventral tegmental area (VTA), a pivotal structure in reward processing. The VTA is known for its dopaminergic projections, which modulate learning and motivation. Imaging data revealed that mPFC neurons projecting to the VTA prominently carry contingency degradation signals, suggesting a functional communication pathway critical for updating reward expectations.

Optogenetic stimulation experiments further corroborated this functional link. When subsets of mPFC→VTA neurons encoding contingency degradation were selectively activated, animals exhibited an accelerated adjustment to degraded cue–reward contingencies. This finding underscores a direct top-down influence from the prefrontal cortex on subcortical reward circuits, orchestrating the adaptive suppression of no longer beneficial behaviors.

These results offer a paradigm-shifting perspective on the neural architecture of cognitive flexibility, demonstrating how frontocortical circuits interface with dopaminergic midbrain centers to implement behavioral adaptation. The work bridges gaps in our understanding of how executive control regions communicate with reward networks to flexibly modulate behavior in dynamic environments.

Moreover, the findings have broad implications for neuropsychiatric disorders characterized by impaired cognitive flexibility, such as obsessive-compulsive disorder, addiction, and schizophrenia. Understanding the precise circuitry and signaling mechanisms may ultimately inform targeted therapeutic interventions to restore adaptive behavioral control in these populations.

Methodologically, the combination of advanced computational modeling, state-of-the-art in vivo imaging, and cutting-edge optogenetic manipulation provides a powerful toolkit for dissecting complex brain functions. This integrated approach sets a new standard for exploring how distributed neural networks coordinate to produce flexible, goal-directed behavior.

Importantly, this research exemplifies a translational bridge from computational theory to neural implementation and behavior, illuminating the brain’s capacity to deploy meta-learning processes at the cellular circuit level. It opens avenues for future work to explore how these mechanisms operate across species and contribute to higher-order cognitive functions.

In sum, this study uncovers a critical neural pathway through which the prefrontal cortex exerts executive control over subcortical reward circuitry, driving the suppression of obsolete learned behaviors in favor of more adaptive responses. The demonstration that mPFC→VTA dynamics underlie contingency degradation enriches our neurobiological understanding of flexibility and sets the stage for innovative approaches in neuroscience and psychiatry.

As adaptive decision-making continues to be a central challenge in both basic and clinical neuroscience, these findings represent a major advance, providing mechanistic insight into how the brain reconfigures its expectations and behaviors in the face of changing environments. This work not only elucidates fundamental brain functions but also holds promise for addressing cognitive rigidity in disease.

Subject of Research: Neural circuit mechanisms underlying cognitive flexibility and contingency degradation.

Article Title: Prefrontal to ventral tegmental area dynamics drive contingency degradation.

Article References: Hjort, M.M., Garrett, Z.Q., Gordon, A.G. et al. Prefrontal to ventral tegmental area dynamics drive contingency degradation. Nature (2026). https://doi.org/10.1038/s41586-026-10443-5

Image Credits: AI Generated

DOI: https://doi.org/10.1038/s41586-026-10443-5