Theta-Phase Dependent Neuronal Coding During Sequence Learning in Human Single Neurons

Leila Reddy, Matthew W. Self, Benedikt Zoefel, Marlène Poncet, Jessy K. Possel, Judith Peters, Johannes C. Baayen, Sander Idema, Rufin VanRullen, Pieter R. RoelfsemaSSRN 2019 · DOI

Introduction

How does the human brain maintain the order of events in short-term memory? While neuroscientists have long understood that specific neurons fire in response to specific objects, the mechanism for encoding the temporal relationship between those objects—the "what came first"—has remained elusive. A growing body of evidence suggests that the brain does not rely solely on which neurons fire, but also on when they fire relative to underlying brain oscillations. This study investigates the "spike-phase code" in the human medial temporal lobe (MTL) to determine how individual neurons organize sequences of information.

The Rodent Precedent: Phase Precession

To understand the human mechanism, researchers often look to rodent models of navigation. In rats, a phenomenon known as phase precession occurs as an animal moves through space. When a rat approaches a specific location, the neurons representing that spot (place cells) fire at progressively earlier phases of the 6–10 Hz theta brain wave. This timing-based code allows the brain to represent a spatial sequence within a single oscillatory cycle. For years, it was hypothesized that humans might use a similar mechanism to organize non-spatial sequences, such as a list of items or images.

Experimental Design and Methodology

This study utilized a rare clinical opportunity to record single-neuron activity directly from the human brain. The researchers worked with nine patients with drug-resistant epilepsy who were already implanted with depth electrodes for clinical monitoring. This allowed for the recording of 249 neurons across the hippocampus, parahippocampal gyrus, and temporal cortex.

The Sequence Task

To simulate navigation through a conceptual space, participants viewed a sequence of five to seven images displayed on a virtual rotating wheel. Their task was to learn the fixed order of these images. The researchers then analyzed the Local Field Potentials (LFP) and individual neuronal spikes, focusing on 63 "selective" neurons that responded strongly to specific images in the sequence.

Using Pairwise Phase Consistency (PPC), the team measured whether neurons were phase-locked to the 6–10 Hz theta oscillation and how that phase changed depending on the image's position in the learned sequence.

Results: Temporal Multiplexing and Phase Succession

The study revealed that the human MTL indeed uses a spike-phase code to represent sequences, but with a surprising twist that differentiates it from the rodent model.

1. Time-Division Multiplexing

The researchers found that the brain uses "time-division multiplexing" to represent different items. In a single 125ms theta cycle (8 Hz), different images in the sequence are represented by spikes occurring at distinct phases. This allows the brain to keep multiple items active in short-term memory simultaneously without them interfering with one another.

2. The "Succession" Discovery

While the rodent model shows phase precession (spikes moving earlier as the goal is approached), the human sequence task demonstrated phase succession. The researchers found that the neuron's response to the preceding item in the sequence occurred at an earlier phase (~ -55°) than the response to the current preferred item (~ -3°). This ~52° difference corresponds to a timing gap of approximately 20 milliseconds.

3. Learning-Dependent Shifts

Interestingly, this phase separation was not a default state. As subjects learned the sequence, the phase of firing for the preceding stimulus shifted to this distinct, earlier position. This suggests that the phase shift is a physical signature of a learned temporal relationship.

4. Frequency Specificity

This coding mechanism was found to be highly specific to the theta band (6–10 Hz). No similar phase-coding effects were observed in the slower delta (1.6–3 Hz) or faster beta (15–23 Hz) ranges, highlighting the unique role of theta oscillations in organizing human memory.

Discussion and Limitations

The discovery of phase succession suggests that the human brain may have a unique way of encoding conceptual or episodic sequences compared to how rodents encode spatial navigation. While the rodent model suggests an entire sequence might be visible in one cycle, this study primarily observed the relationship between two consecutive items (the current and the preceding). This "sparse representation" might be due to the limitations of recording a finite number of neurons, or it may reflect a fundamental difference in human cognitive processing.

Furthermore, because the data was gathered from a clinical population (patients with epilepsy), future research is required to confirm these findings in the broader population. Additionally, it remains to be seen whether humans navigating a physical environment would show the rodent-like precession or the succession found in this conceptual task.

Conclusion

By demonstrating that human single neurons use the precise timing of spikes relative to theta waves to encode order, this study provides a vital link between low-level neuronal activity and high-level memory organization. The shift from precession to succession suggests that the human brain adapts oscillatory timing to meet the specific demands of sequence learning, providing a robust framework for how we remember the order of our experiences.

Summary updated

BibTeX
@misc{reddy2019theta,
  author = {Leila Reddy and Matthew W. Self and Benedikt Zoefel and Marlène Poncet and Jessy K. Possel and Judith Peters and Johannes C. Baayen and Sander Idema and Rufin VanRullen and Pieter R. Roelfsema},
  title = {{Theta-Phase Dependent Neuronal Coding During Sequence Learning in Human Single Neurons}},
  year = {2019},
  doi = {10.2139/ssrn.3383799},
  howpublished = {SSRN},
}
Figure