Repeated Viewing of a Film Clip Changes Event Timescales in The Brain

Narjes Al-Zahli, Mariam Aly, Christopher BaldassanobioRxiv 2025 · DOI

TL;DR

Our brains dynamically adjust how they segment continuous experiences, like watching a movie, based on familiarity. Repeated viewing can lead to either finer (more detailed) or coarser (more generalized) event representations in different brain regions, and this flexibility is linked to better memory recall.

What Problem Does This Paper Address?

We constantly encounter repeated experiences, from listening to music to rewatching films. A fundamental question in neuroscience is how our brains adapt their internal representations of these experiences as they become more familiar. Specifically, how do the brain's "event timescales" – the way it breaks down continuous streams of information into discrete moments – change with repeated exposure? Previous research suggests that different brain regions have preferred timescales, but it's unclear how flexible these are. This paper investigates whether repeated viewing of a narrative movie alters these event timescales and, if so, how these changes relate to memory.

Methods

The researchers employed a sophisticated approach combining fMRI data with advanced computational modeling.

  • Hidden Markov Models (HMMs): They used HMMs to segment continuous neural activity (fMRI signals) into discrete "event states" at multiple timescales. This allowed them to identify event boundaries and quantify the strength of event structure at both slow (fewer, longer events) and fast (more, shorter events) timescales.
  • Within- vs. Between-Event Similarity: This metric was used to assess the quality of the event segmentation. It measures how similar brain activity patterns are within an identified event compared to patterns between different events. A higher value indicates stronger, more distinct event representations.
  • Searchlight Analysis: This technique allowed them to analyze event structure at a fine-grained, voxel-wise level across the entire brain, rather than just in predefined regions of interest.
  • Conjunction Analysis: To identify brain regions that showed consistent changes across different movie clips, regardless of their specific content, a conjunction analysis was performed.

Datasets

The study utilized data from a previously published experiment (Aly et al., 2018) involving 30 participants who underwent fMRI while watching three 90-second clips from "The Grand Budapest Hotel" six times each. The clips were presented in three conditions: "Intact" (original narrative order), "Scrambled-Fixed" (segments shuffled but in a consistent order across viewings), and "Scrambled-Random" (segments shuffled differently each viewing).

Key Insights and Ideas

  • Brain's Event Segmentation is Flexible: Contrary to the idea of fixed temporal receptive windows, the brain actively adapts how it segments continuous experiences based on familiarity.
  • Finer vs. Coarser Representations: Repeated viewing can lead to either more detailed (finer) or more generalized (coarser) event representations, depending on the brain region and the nature of the stimulus.
  • Narrative Coherence Matters: The degree of narrative coherence in the movie clip significantly influenced how event timescales changed. Coherent narratives led to more spatially organized changes, while scrambled narratives resulted in more widespread coarsening.
  • Finer Event Structure Aids Memory: A loss of slow-timescale event structure (leading to finer segmentation) in specific brain regions was associated with better recall of narrative details. This suggests that the brain's ability to break down experiences into more granular units can enhance memory encoding.

Key Results

  • Widespread Stability, but Notable Flexibility: While many brain regions showed stable event structures across repeated viewings, a significant number exhibited changes.
  • Differential Changes: In the "Intact" clip, visual and auditory regions (like the lateral occipital cortex) showed finer event segmentation (loss of slow-timescale structure), while higher-order regions (like the inferior frontal cortex) showed coarser segmentation.
  • Scrambled Clips Lead to Coarsening: The "Scrambled-Random" condition, in particular, showed a strong tendency towards coarser event representations with repeated viewing, suggesting the brain attempts to impose structure on incoherent information.
  • Memory Correlation: In the lateral occipital cortex and middle temporal gyrus, a greater loss of slow-timescale event structure with repeated viewing was significantly correlated with better recall of narrative details.

Limitations and Future Work

  • Memory Metric: The study used the number of recalled details as a measure of memory. Future work could incorporate other metrics like narrative coherence or recognition memory accuracy to provide a more comprehensive understanding of the relationship between neural event timescales and memory.
  • Participant Engagement: The authors acknowledge the possibility of participant disengagement (boredom) during later viewings but argue against it explaining their findings due to the selective nature of the observed changes.
  • Future Research Directions: The authors propose leveraging their new methodology to study learning, schema formation, and narrative processing. They also suggest exploring how event timescale changes relate to the perception of elapsed time and whether these changes persist long-term, predicting memory consolidation or forgetting.

Summary updated

Abstract

Many experiences occur repeatedly throughout our lives: we might watch the same movie more than once and listen to the same song on repeat. How does the brain modify its representations of events when experiences are repeated? We hypothesized that, with repeated viewing of a narrative movie, brain regions would adapt their event representations by becoming either finer (more detailed) or coarser (more generalized). To test this hypothesis, we analyzed data from 30 human participants who underwent functional magnetic resonance imaging (fMRI) while watching three 90-second clips from "The Grand Budapest Hotel" six times each. We used hidden Markov models and pattern similarity analysis applied to searchlights across the brain to quantify the strength of event structure at different timescales for each clip presentation. We then tested how event structure strength changed at both slow and fast timescales with repeated viewings. Most brain regions exhibited stability in the strength of event structure at both slow and fast timescales. Other regions, however, showed flexible event representations that became more or less granular across repeated clip presentations. Notably, several brain regions exhibited consistent changes in the strength of event structure at a slow timescale across different movie clips. Furthermore, in lateral occipital cortex and middle temporal gyrus, greater loss of event structure at a slow timescale predicted more detailed memory recall. These results highlight that event dynamics in the brain are not fixed, but can change flexibly with experience.
BibTeX
@misc{alzahli2025repeated,
  author = {Narjes Al-Zahli and Mariam Aly and Christopher Baldassano},
  title = {{Repeated Viewing of a Film Clip Changes Event Timescales in The Brain}},
  year = {2025},
  doi = {10.1101/2025.08.27.672403},
  howpublished = {bioRxiv},
}

Loading related content...

Figure