Authors :
Presenting Author: Jiyun Shin, PhD – New York University Langone Health
Forouzan Farahani, Ph.D. – New York University Langone Health
Ritika Rohatgi, Undergraduate student – Barnard College, Columbia University
Simon Henin, Ph.D. – New York University Langone Health
Charan Ranganath, PhD – Center for Neuroscience, UC Davis
Anli Liu, M.D. – New York University Langone Health
Rationale:
The brain organizes continuous experience into discrete events to form memories, a process known as event segmentation. Functional MRI studies using naturalistic learning paradigms, such as narrative listening or movie watching, have shown that the hippocampus is activated at event boundaries. However, the exact neural mechanisms underlying event segmentation and its relevance to episodic memory in humans remain unclear. Here, we leveraged intracranial EEG (iEEG) and single-unit recordings in epilepsy patients to investigate hippocampal dynamics during viewing of short films.
Methods:
We recruited 13 surgical epilepsy patients (6 patients with Behnke-Fried electrodes) who were (1) between 18-65 years old (median 30 y, 6F), (2) fluent in English, and (3) capable of providing informed consent. Subjects watched either one block of 8 short films (3 patients) or 3 blocks of 5 short films (10 patients). Following a brief distraction task, they were asked to verbally recall as much as they could remember from the films. Films were 2-5 minute narratives, presented a variety of content, and included dialogue and silent films. Film event boundaries were separately defined by 100 independent raters (recruited via Prolific). Subjects’ spoken recall was transcribed using Whisper.ai with human validation and compared against the annotated event log. High-gamma broadband activity (HGA; 70-150 Hz) around event boundaries (±1 s window) in the medial temporal lobe was extracted after removing line noise. A non-parametric permutation test was used to compare hippocampal HGA change at event boundaries compared to a randomly selected window within the prior event. Single-unit activity was recorded via microwires and processed using Kilosort for spike sorting.
Results:
We found that 36% of hippocampal (HPC) electrodes (44 out of 122) and 46% parahippocampal gyrus (PHG) electrodes (17 out of 37) showed significant HGA responses around event boundaries compared to a null distribution. These findings align with previous fMRI studies demonstrating activation in both regions at event boundaries. None of the 90 single-units recorded from six patients implanted with Behnke-Fried electrodes (n=19 microwire bundles; mean 4.7 units/bundle) showed significant firing modulation around event boundaries.
Conclusions:
Our preliminary findings suggest that the HPC and PHG exhibit robust high-gamma activity at event boundaries, consistent with previous fMRI studies and iEEG studies implicating these regions in event segmentation. Unlike a recent report of 10% of hippocampal neurons responding to event boundaries (Zheng et al., 2022), we did not find any hippocampal single unit modulation around event boundaries. This dissociation suggests that event segmentation may not be driven by strong rate coding in individual neurons, but rather by alternative mechanisms such as sparse coding, subthreshold activity, or population-level dynamics like phase locking. These preliminary results will guide detection of medial temporal lobe electrodes involved in boundary processing for future mechanistic work.
Funding:
R01 NS1007806 (AL)
1K23NS104252 (AL)
FACES (AL)
Department of Neurology (AL)