Intrinsic timescales of spiking activity in humans during wakefulness and sleep

Annika Hagemann, Marcel Stephan Kehl, Jonas Dehning, F. Paul Spitzner, Johannes Niediek, Michael Wibral, Florian Mormann, Viola PriesemannarXiv 2022 · DOI · arXiv

Abstract

Information processing in the brain requires integration of information over time. Such an integration can be achieved if signals are maintained in the network activity for the required period, as quantified by the intrinsic timescale. While short timescales are considered beneficial for fast responses to stimuli, long timescales facilitate information storage and integration. We quantified intrinsic timescales from spiking activity in the medial temporal lobe of humans. We found extended and highly diverse timescales ranging from tens to hundreds of milliseconds, though with no evidence for differences between subareas. Notably, however, timescales differed between sleep stages and were longest during slow wave sleep. This supports the hypothesis that intrinsic timescales are a central mechanism to tune networks to the requirements of different tasks and cognitive states.

Introduction

To navigate and interpret the world, the brain must integrate information across various temporal windows. For instance, linguistic comprehension requires the brain to maintain the beginning of a sentence in "active" memory until the conclusion is reached. This temporal persistence of neural signals is formally quantified as the intrinsic timescale ($\tau$).

There exists a fundamental computational trade-off: short timescales enable rapid responses to environmental changes (low latency), whereas longer timescales support complex information integration and storage. While these dynamics have been extensively studied in animal models using direct spike recordings, human research has largely relied on indirect, low-resolution measures like fMRI or EEG. This study provides the first direct quantification of intrinsic timescales from individual neuronal spikes in the human brain, specifically investigating how these scales vary across brain regions and states of consciousness.

Methods

The researchers analyzed intracranial recordings from 24 patients with focal epilepsy, who were implanted with depth electrodes in the medial temporal lobe (MTL)—including the hippocampus, amygdala, entorhinal cortex, and parahippocampal cortex—for clinical monitoring.

The methodology for estimating $\tau$ involved several key steps:

  1. Population Activity ($A_t$): Spikes from all recorded single units were summed into 4ms time bins.
  2. Autocorrelation Function (ACF): The correlation of network activity with itself was calculated across various time lags ($\Delta t$).
  3. Exponential Fitting: The intrinsic timescale $\tau$ was defined as the decay constant of this autocorrelation, determined by fitting an exponential function: $f(\Delta t) = A \exp(-\Delta t/\tau) + B$.
Figure 1: Quantification of intrinsic timescales from spiking activity. (a-b) Spike rasters and population activity. (c-f) Autocorrelation functions (ACF) and exponential fits. (g) Distribution of timescales across recordings.

Results

Diversity of Neural "Clocks"

The study revealed that intrinsic timescales in the human MTL are remarkably diverse, ranging from approximately 10ms to 700ms, with a median value of 54 ms during wakefulness. This wide spectrum suggests that even within localized brain regions, different neural ensembles are specialized for processing information at different speeds. Interestingly, no significant differences in $\tau$ were found between subregions like the amygdala and hippocampus, though high internal variability may have obscured subtle distinctions.

State-Dependent Tuning during Sleep

A primary finding of the research is that intrinsic timescales are dynamic rather than static. They shift significantly depending on the brain's state of consciousness. Specifically, timescales were found to lengthen during sleep, reaching their maximum during Slow Wave Sleep (N3).

Figure 5: Intrinsic timescales across sleep stages. (b) Average ACFs for REM, N1, N2, and N3 sleep stages. (c) Swarm plot showing significant increases in timescale during deep sleep (N3) compared to other stages.

The median $\tau$ in N3 was significantly longer than in REM, N1, or N2 ($p < 10^{-4}$). This "stretching" of the neural clock during deep sleep suggests the network is tuned toward a state of high integration, potentially to facilitate critical functions such as memory consolidation.

Discussion

While this study provides a groundbreaking direct look at human neural timescales, several limitations remain. The data was collected from a clinical population on anti-epileptic medications, which might influence neural excitability and shorten observed timescales. Additionally, the human data was recorded during rest, whereas comparative animal data often comes from active task states.

Future research should focus on measuring human intrinsic timescales during active cognitive tasks. This would determine if $\tau$ dynamically adjusts during active memory or decision-making processes, similar to the shifts observed during the transition from wakefulness to deep sleep.

Summary updated

Abstract

Information processing in the brain requires integration of information over time. Such an integration can be achieved if signals are maintained in the network activity for the required period, as quantified by the intrinsic timescale. While short timescales are considered beneficial for fast responses to stimuli, long timescales facilitate information storage and integration. We quantified intrinsic timescales from spiking activity in the medial temporal lobe of humans. We found extended and highly diverse timescales ranging from tens to hundreds of milliseconds, though with no evidence for differences between subareas. Notably, however, timescales differed between sleep stages and were longest during slow wave sleep. This supports the hypothesis that intrinsic timescales are a central mechanism to tune networks to the requirements of different tasks and cognitive states.
BibTeX
@misc{hagemann2022intrinsic,
  author = {Annika Hagemann and Marcel Stephan Kehl and Jonas Dehning and F. Paul Spitzner and Johannes Niediek and Michael Wibral and Florian Mormann and Viola Priesemann},
  title = {{Intrinsic timescales of spiking activity in humans during wakefulness and sleep}},
  year = {2022},
  eprint = {2205.10308},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2205.10308},
}

Loading related content...

Figure