Gamma and beta bursts during working memory read-out suggest roles in its volitional control

Mikael Lundqvist, Pawel Herman, Melissa R. Warden, Scott L. Brincat, Earl K. MillerbioRxiv 2017 · DOI

TL;DR

This study reveals that working memory isn't a constant hum of neural activity, but rather a dynamic interplay of brief gamma and beta oscillations. These oscillations, particularly their balance, appear to act as a volitional control mechanism, regulating the flow of information in and out of working memory, and even predicting behavioral errors.

What Problem Does This Paper Address?

For decades, the dominant model of working memory (WM) has been sustained neuronal firing – the idea that neurons keep "on" to represent information. However, recent research has hinted at a more complex picture involving transient oscillatory bursts. This paper delves into the specific roles of gamma (∼55–120 Hz) and beta (∼20–35 Hz) oscillations during the "read-out" phase of WM, a process that's been difficult to study due to its overlap with motor responses. The authors aim to disentangle these oscillatory dynamics and their relationship to information processing and volitional control within WM.

Methods

The researchers used electrophysiological recordings (local field potentials and spiking activity) from monkeys performing a challenging working memory task. The task involved remembering a sequence of two objects and then determining if a subsequent test sequence matched.

Key methodological aspects include:

  • Oscillatory Burst Analysis: Identifying discrete bursts of gamma and beta oscillations.
  • Information Encoding (PEV): Quantifying how much information about the remembered objects was carried by neuronal spiking activity using Percentage of Explained Variance (PEV).
  • Correlation Analysis: Examining the relationships between burst rates, spiking information, and behavioral outcomes.
  • Task Design: A clever task design that allowed for the examination of WM read-out and decision-making independent of immediate motor responses.

Datasets

The study utilized electrophysiological data collected from monkeys performing a working memory task. No new datasets were introduced or benchmarked.

Key Insights and Ideas

  • WM is Dynamic, Not Static: The findings strongly suggest that working memory is not a sustained, constant state but rather a series of brief, dynamic events orchestrated by gamma and beta oscillations.
  • Beta as a "Gatekeeper": Beta oscillations appear to act as a regulatory mechanism, potentially suppressing or "clearing out" information from WM when it's no longer needed, or when preparing for new information.
  • Gamma as the Information Carrier: Gamma bursts are closely linked to spiking activity that carries specific information about the remembered items.
  • Push-Pull Relationship: Gamma and beta oscillations exhibit an anti-correlated, "push-pull" relationship, especially at sites where information is being actively processed.
  • Predicting Errors: Deviations from the typical gamma/beta dynamics during WM read-out accurately predicted behavioral errors, suggesting these oscillations are crucial for accurate decision-making.
  • Volitional Control: The authors propose that the balance between beta and gamma oscillations is under top-down control, allowing for the volitional regulation of WM content.

Key Results

  • Gamma and Beta Anti-correlation: Gamma and beta bursts were consistently anti-correlated, with high gamma coinciding with low beta, and vice-versa, particularly at "informative" sites.
  • Information Encoding: Neurons at "gamma-modulated" sites showed significantly higher information content about the remembered objects compared to "non-gamma-modulated" sites.
  • WM Read-out Dynamics: In anticipation of needing to use an object for a match decision, there was an increase in spiking information, gamma, and a decrease in beta. Conversely, when an object was no longer needed, beta increased while gamma and information decreased.
  • Error Prediction: Incorrect behavioral responses were associated with deviations from these expected gamma/beta dynamics. For instance, in trials where a non-match was mistakenly identified as a match, the neural activity initially followed the "non-match" trajectory but then shifted towards a "match-like" profile in the critical delay period.
  • Match/Non-match Differentiation: Gamma burst rates differed based on whether the test sequence was a match, an "order non-match," or an "identity non-match," with higher rates for non-matches. Beta, on the other hand, distinguished between matches and non-matches in the timing of its bursts.

Limitations and Future Work

  • Causality: While the study shows strong correlations between gamma/beta dynamics and WM function/errors, establishing direct causal links would require further investigation (e.g., using stimulation techniques).
  • Generalizability: The study was conducted in monkeys. Further research is needed to confirm these findings in humans and across different cognitive tasks.
  • Mechanisms of Control: While the authors propose top-down control of the beta/gamma balance, the precise neural circuits and mechanisms underlying this control remain to be fully elucidated.
  • Role of Other Oscillations: The study focused on gamma and beta. Future work could explore the interplay of these with other frequency bands (e.g., theta) in WM.
  • Individual Neuron Dynamics: While the study highlights site-level dynamics, understanding how individual neurons contribute to these population-level oscillations and their role in WM is an ongoing area of research.

Summary updated

Abstract

Abstract Working memory (WM) activity is not as stationary or sustained as previously thought. There are brief bursts of gamma (∼55–120 Hz) and beta (∼20–35 Hz) oscillations, the former linked to stimulus information in spiking. We examine these dynamics in relation to read-out from WM, which is still not well understood. Monkeys held a sequence of two objects and had to decide if they matched a subsequent sequence. Changes in the balance of beta/gamma suggested their role in WM control. In anticipation of having to use an object for the match decision, there was an increase in spiking information about that object along with an increase in gamma and a decrease in beta. When an object was no longer needed, beta increased and gamma as well as spiking information about that object decreased. Deviations from these dynamics predicted behavioral errors. Thus, turning up or down beta could regulate gamma and the information in working memory.
BibTeX
@misc{lundqvist2017gamma,
  author = {Mikael Lundqvist and Pawel Herman and Melissa R. Warden and Scott L. Brincat and Earl K. Miller},
  title = {{Gamma and beta bursts during working memory read-out suggest roles in its volitional control}},
  year = {2017},
  doi = {10.1101/122598},
  howpublished = {bioRxiv},
}
Figure