Auricular Transcutaneous Vagus Nerve Stimulation Stabilizes Event Segmentation Through Modulation of Working Memory Representations

Figures

• 

  Figure 1. Schematic overview of the creation of no-boundary intervals. Boundary intervals are depicted in black, and no-boundary intervals in gray. In step 1, the data were segmented into 2-second intervals. Each interval containing a boundary (key press) was selected and randomly paired with an interval that did not include a boundary. A virtual marker was then placed within the no-boundary interval at the same relative time point as the key press in its paired boundary interval (left panel). In step 2, data were re-segmented around these virtual markers, allowing for time-locked analysis from −1 to +1 second relative to each marker’s position (right panel).

• 

  Figure 2. Behavioral results of mixed-effects logistic regression. (A) Model predicted probability of segmentation (y-axis) with 2 predictors-situational change counts per 2 seconds and session-with separate curves for active atVNS (magenta) and sham (pink). (B) Violin plots display the distribution of within-subject differences in predicted segmentation probabilities (active minus sham). Each dot represents one subject, and the horizontal black bar marks the mean difference. Overall, active atVNS reduced the likelihood of segmentation across levels of situational change, consistent with increased stability of the working event model. atVNS, auricular transcutaneous vagus nerve stimulation; VNS, vagus nerve stimulation.

• 

  Figure 3. MVPA results comparing active and sham sessions for both boundary and no-boundary intervals. (A) Boundary interval. Left: time-resolved classification performance expressed as area under the curve (AUC); bold lines mark above change periods, with shading area reflecting the standard deviation across participants. Right: temporal generalization matrix; color encodes AUC, with brighter regions indicating greater between-session discriminability across train-test times. (B) No-boundary interval. Same layout; neither the time-resolved nor the temporal generalization analysis shows significant classification. These results indicate that atVNS selectively modulated neural representations during event boundaries, rather than producing a global session difference. atVNS, auricular transcutaneous vagus nerve stimulation; MVPA, multivariate pattern analysis.

• 

  Figure 4. Session-wise comparison of MVPA decoding accuracy with the corresponding sLORETA result. (A) Time-resolved BI–NBI decoding (AUC): magenta curve for active session and pink curve for sham session; gray shading highlights time periods with a significant AUC difference between sessions; sLORETA maps show brain regions exhibiting significant active-sham contrasts within the gray-marked window. (B) Temporal generalization BI–NBI decoding (AUC): Cohen’s d effect size map for active vs sham across train-test matrix; the color scale indicates effect magnitude within cluster-corrected significant regions. Active atVNS was associated with reduced BI–NBI decodability, consistent with increased stabilization of event representations prior to segmentation. atVNS, auricular transcutaneous vagus nerve stimulation; BI, boundary interval ; NBI, no-boundary interval; MVPA, multivariate pattern analysis; sLORETA, standardized low-resolution brain electromagnetic tomography.