Fig. 1: Study design and memory performance.
From: Neural and behavioral reinstatement jointly reflect retrieval of narrative events
A Task design. Participants first viewed 48 minutes of the BBC show "Sherlock" (Task 1) and then recalled it verbally (Task 2). The movie clip was segmented into 48 narrative events by an independent viewer36. B Dataset overview. Eye-tracking data and voice recordings were acquired for two datasets in independent participants. In addition, Dataset 1 included simultaneously acquired fMRI data. Table shows participant counts for the respective data types. C Recall quantification 1: Summary of the order and duration of the events in both tasks. Each box represents an event and is scaled according to the duration in the respective task. Participants were color-coded within each dataset. D Recall quantification 2: Language-model results of the spoken recall data expressed as event-by-event sentence-embedding similarity. Matrices show the ranks of Pearson correlations between sentence embeddings estimated for each event within each participant, averaged across them. Ranking was performed after computing the correlations to visually highlight similarities between matrices (blue to yellow colors show low to high ranks). Spoken recall was highly similar between the datasets.
