Brain-wide organization of intrinsic timescales at single-neuron resolution

Yan-Liang Shi, Roxana Zeraati, The International Brain Laboratory, Anna LevinabioRxiv 2025 · DOI

Abstract

Variations in intrinsic neural timescales across the mammalian forebrain reflect the anatomical structure and functional specialization of brain areas and individual neurons. Yet, the organization of timescales beyond the forebrain remains unexplored. We analyzed intrinsic timescales of single neurons across the entire mouse brain. Median timescales were up to fivefold longer in the midbrain and hindbrain than in the forebrain. Spatial patterns of gene expression predicted timescale variation at a resolution finer than brain-area boundaries. Across neurons, the diversity of timescales revealed a multiscale architecture, in which fast timescales determined regional differences in medians, while slow timescales universally followed a power-law distribution with an exponent near 2, indicating a shared dynamical regime across the brain consistent with the edge of instability or chaos. These organizing principles for the dynamics of single neurons across the brain provide a foundation for linking cellular activity with regional specialization and brain-wide computation.

Introduction

How long does a thought last? In neuroscience, "intrinsic timescales" refer to the duration over which a neuron’s activity remains correlated with its own past. These timescales are critical for brain function: a neuron processing rapid sensory input must reset quickly, while one involved in maintaining a motor plan or a mood may need to integrate information over seconds or minutes. While previous research established a temporal hierarchy in the cortex—moving from fast sensory areas to slow associative regions—the temporal organization of the rest of the brain has remained largely uncharted.

A new study by Yan-Liang Shi and colleagues provides the first brain-wide map of these timescales at single-neuron resolution. By analyzing massive datasets from the International Brain Laboratory (IBL), the researchers uncovered a "multiscale architecture" that governs neural dynamics from the isocortex to the hindbrain.

Methods and Data

The authors utilized Neuropixels recordings of spontaneous activity from 27,960 well-isolated single neurons across 223 brain regions in 115 mice. To quantify the temporal fingerprints of these neurons, they employed a sophisticated autocorrelation fitting technique. Rather than forcing a single exponential decay, the team fitted a mixture of up to four exponential functions to the spike-count autocorrelation of each neuron. This allowed them to define an "effective timescale" ($\tau_{eff}$) that captures the overall persistence of a neuron's activity.

Diversity and quantification of intrinsic timescales across the mouse brain.

To investigate the biological drivers of these timescales, the team integrated electrophysiological data with spatial densities from the Allen Gene Expression Atlas (4,353 genes) and the Allen Brain Cell Atlas (338 cell types). They used ridge regression to determine how well anatomical and molecular features could predict a neuron's temporal properties.

Key Results

Regional Specialization: The Slow Subcortex

A primary finding of the study is the surprising slowness of subcortical regions. While the isocortex and thalamus are often viewed as the centers of complex computation, the midbrain and hindbrain exhibited the longest intrinsic timescales. Median timescales in the midbrain (0.76s) and hindbrain (0.89s) were up to five times longer than those in the forebrain (0.16s). This suggests that these phylogenetically older structures are specialized for the long-term integration of internal states and homeostatic signals.

Molecular Predictors of Temporal Dynamics

The study found that timescales are not merely a property of broad anatomical regions. Gene expression profiles were able to predict a neuron's timescale at a resolution finer than the boundaries of brain areas. Ridge regression models using gene expression explained significantly more variance ($R^2 = 0.14$) than models using only regional labels ($R^2 = 0.025$), indicating that the molecular machinery of a cell is a primary determinant of its temporal integration window.

Universal Scaling and the Edge of Chaos

Perhaps the most striking discovery is the universal distribution of slow timescales. While the median timescale varies significantly by region, the distribution of slow timescales follows a power law with an exponent near 2.0 across all 12 major brain structures. In dynamical systems theory, such power laws often emerge at critical points. The authors’ modeling suggests this consistency indicates a shared dynamical regime across the entire brain, likely operating near the "edge of chaos" or instability, which may allow for maximum computational flexibility.

Discussion and Future Directions

This research provides a foundational map of the brain's temporal architecture, linking cellular-level dynamics with brain-wide specialization. The findings demonstrate that neurons selective for "choice" and "reward" possess longer intrinsic timescales than non-selective neurons, further cementing the link between temporal persistence and functional role.

While the study identifies gene expression as a powerful predictor, the authors note that future work combining single-cell electrophysiology with single-cell RNA sequencing will be necessary to identify the specific synaptic motifs or membrane properties that drive these dynamics. By revealing a universal scaling law alongside regional diversity, this work offers a new lens through which to view brain-wide computation and the biological basis of thought duration.

Summary updated

Abstract

Variations in intrinsic neural timescales across the mammalian forebrain reflect the anatomical structure and functional specialization of brain areas and individual neurons. Yet, the organization of timescales beyond the forebrain remains unexplored. We analyzed intrinsic timescales of single neurons across the entire mouse brain. Median timescales were up to fivefold longer in the midbrain and hindbrain than in the forebrain. Spatial patterns of gene expression predicted timescale variation at a resolution finer than brain-area boundaries. Across neurons, the diversity of timescales revealed a multiscale architecture, in which fast timescales determined regional differences in medians, while slow timescales universally followed a power-law distribution with an exponent near 2, indicating a shared dynamical regime across the brain consistent with the edge of instability or chaos. These organizing principles for the dynamics of single neurons across the brain provide a foundation for linking cellular activity with regional specialization and brain-wide computation.
BibTeX
@misc{shi2025brain,
  author = {Yan-Liang Shi and Roxana Zeraati and The International Brain Laboratory and Anna Levina},
  title = {{Brain-wide organization of intrinsic timescales at single-neuron resolution}},
  year = {2025},
  doi = {10.1101/2025.08.30.673281},
  howpublished = {bioRxiv},
}
Figure