Theta-Gamma Coding Meets Communication-through-Coherence: Neuronal Oscillatory Multiplexing Theories Reconciled

Douglas McLelland, Rufin VanRullenPLOS Computational Biology 2016 · DOI · OpenReview

Introduction

The human brain is tasked with processing complex visual scenes containing multiple overlapping objects. A fundamental challenge in neuroscience is the "binding problem": how does the brain correctly associate features (e.g., color and shape) belonging to the same object without creating "illusory conjunctions"? Two prominent theories involving cross-frequency coupling between theta (4–14 Hz) and gamma (30–80 Hz) oscillations have emerged to explain this.

The Lisman & Jensen (LJ) multiplexing theory suggests that multiple items are processed sequentially, each occupying a distinct gamma cycle within a single theta cycle. Conversely, the Fries (F) multiplexing theory, rooted in Communication-through-Coherence, proposes that the brain selectively synchronizes with only the most salient item at a time, using the slow oscillation as a periodic reset. Historically viewed as competing models, this research demonstrates that both are physiologically plausible operating modes of the same neural architecture.

Methods and Modeling Approach

The researchers developed a hierarchical computational model approximating the pathway from V1 to higher-order areas like V4. The model utilizes leaky integrate-and-fire spiking neurons and a Pyramidal-Interneuron Gamma (PING) mechanism to generate intrinsic gamma oscillations.

To simulate theta/alpha rhythms, an external 9 Hz inhibitory conductance was applied to the lower area. The "binding problem" was specifically tested using a feature-binding task where a higher-level "conjunction" area required simultaneous input from specific orientation and color neurons to fire. The study utilized synthetic visual stimuli consisting of discrete objects with varying contrasts and Gabor filters to simulate V1 receptive fields.

Reconciling Multiplexing Theories

The core insight of this work is that the transition between "exploratory" (LJ-multiplexing) and "selective" (F-multiplexing) processing is controlled by a single parameter: global inhibition.

Schematic of F-multiplexing where selective synchronization occurs within a slow oscillatory cycle.

When global inhibition is increased in the higher-level area, the network enters the F-multiplexing mode. In this state, the most salient signal arrives first and suppresses competing inputs, ensuring unambiguous processing of a single item. Alternatively, increasing global inhibition in the lower area facilitates LJ-multiplexing, where different neural ensembles compete to fire in a sequence across successive gamma cycles. Both modes effectively solve the binding problem by ensuring that only one object's features are active in the higher area at any specific millisecond, preventing the formation of illusory conjunctions.

Key Results and Physiological Implications

The model demonstrates that both multiplexing styles can emerge from biologically realistic parameters, using synaptic and membrane time constants derived from experimental literature. Furthermore, the model naturally reproduces the observation that gamma frequency increases with stimulus contrast. This contrast sensitivity provides the necessary phase offset for the sequential ordering of items in the LJ-mode.

The ability of a single circuit to switch between these modes suggests a dynamic flexibility in brain function. The authors hypothesize that neuromodulatory systems, such as cholinergic pathways, may act as the biological "knob" that adjusts global inhibition levels to toggle between exploratory and selective attention.

Limitations and Future Directions

While the model provides a robust framework for reconciliation, it exhibits higher levels of oscillatory synchronization than typically seen in vivo, where neural activity is often noisier. The "higher area" represented is also relatively simple; future research should investigate how these dynamics scale in larger, recurrent networks. Additionally, exploring the system's robustness to noise and the specific role of neuromodulators in switching between these modes remains a critical area for experimental validation.

Summary updated

Abstract

Author Summary There is a growing consensus that neuronal oscillations constitute a fundamental computational mechanism in the brain. Beyond this, recent experimental evidence has highlighted interactions between oscillations at high and low frequencies (e.g. gamma oscillations, 40–80 Hz, are modulated by theta oscillations, 4–10 Hz), and two major theories have developed regarding the functional role of this kind of cross-frequency coupling. Here, we present a computational modelling study of these theories with strong implications for biological studies. Firstly, we demonstrate for the first time that each of these theories is physiologically plausible, in that they can be implemented in a spiking network model with parameters guided by experimental data. Secondly, we show that they are each computationally useful, able to overcome a feature-binding ambiguity in a presented stimulus. Finally, we implement both theories within a single network model, and find that only a single parameter change is required to switch between the two processing states. This leads to the exciting new proposal that both theories may be correct, both implemented in the brain, with dynamic switching between modes according to processing and attentional requirements.
BibTeX
@article{mclelland2016theta,
  author = {Douglas McLelland and Rufin VanRullen},
  title = {{Theta-Gamma Coding Meets Communication-through-Coherence: Neuronal Oscillatory Multiplexing Theories Reconciled}},
  journal = {PLOS Computational Biology},
  year = {2016},
  doi = {10.1371/journal.pcbi.1005162},
}
Figure