Investigators recorded heart rate from 439 riders and retained 418 usable traces for analysis. The ride itself lasted roughly two minutes; during that interval mean heart rate rose by 15.2 beats per minute (95% CI 14.2 to 16.3). That peak was transient — the study team observed a return to baseline within nine minutes after ride end. Short, sharp arousal characterized the cohort's cardiac response, and the aggregate curve implies a rapid activation followed by recovery in the minutes that follow.
The authors reported analyses aligned to ride start so that temporal changes could be compared across individuals and across trains. Real-time alignment permitted the authors to distinguish ride-phase effects from other between-person variance. The source did not report additional continuous physiological measures beyond heart-rate traces in the abstract.
The study team used k-means clustering to probe whether riders separated into distinct heart-rate response types. Two clusters emerged, but the silhouette score was low (0.29), indicating weak separation. In practical terms, heart-rate curves did not form clearly distinct types; the clusters should be read as descriptive groupings rather than robust phenotypes.
One identifiable pattern was that riders who appeared calm while waiting in the queue tended to show larger increases during the ride and stayed elevated longer. Conversely, riders who were already more aroused in the queue showed smaller rises during the ride. The abstract does not report cluster sizes, within-cluster variance, or additional validation metrics beyond the silhouette score.
To separate coupling between co-riders from the common influence of the ride itself, the investigators compared real pairs of riders with surrogate pairs assembled from different trains and aligned on ride start. This surrogate benchmark controlled for shared, time-locked ride effects and tested for additional synchrony attributable to co-presence or social coupling.
Pairs of real co-riders gained more synchrony than the surrogate pairs, indicating a modest increment beyond what the ride alone produces. The synchrony boost was quantified with a Fisher z of 0.096 (95% CI 0.050 to 0.142). The peak of coupling occurred just after the ride and dissipated within five minutes, suggesting the interpersonal component is transient and temporally proximal to the shared event.
The cohort-level synchrony signal varied with social context. Pairs in which both riders had come with companions displayed a clearer synchrony boost compared with the surrogate benchmark. For pairs where one or both riders were solo, the synchrony increase was smaller and the effect estimate was described as uncertain in the abstract.
These results suggest that pre-existing social ties or shared social intent (arriving with companions) may amplify short-term physiological coupling during a shared, arousing event. The abstract does not provide detailed subgroup counts or tests of interaction beyond the summary statistics reported.
Participants rated happiness, stress and closeness on 0–10 scales immediately before and after the ride. Group means shifted after the ride: riders reported higher happiness (mean increase +1.3 points; Cohen's d = 0.87) and greater closeness to fellow riders on the train (+2.2 points; d = 0.70). The abstract notes these are substantial effect sizes on brief self-report scales.
The source collected stress ratings as well but the abstract focuses on happiness and closeness increases; details about stress change magnitudes or statistical estimates are not emphasized there. The reported psychological changes were measured in parallel with physiology, enabling correlation testing between subjective change and cardiac metrics.
Despite concurrent measurement, the authors reported no association between phase-specific heart-rate responses, response-profile membership, or synchrony and the observed increases in self-reported happiness or closeness. In other words, physiological arousal and momentary interpersonal coupling ran in parallel with reported social changes, but the study team did not find evidence that cardiac dynamics explained the psychological shifts in this dataset.
Analyses used linear mixed models to account for nested structure (individuals within trains and pairs). The abstract does not list model covariates, random-effect structure, or sensitivity analyses; those details may appear in the full manuscript.
This work is exploratory and was conducted in a naturalistic field setting. The authors explicitly caution that without a non-riding comparison group, psychological changes cannot be causally attributed to the ride itself. The study is also a preprint and has not been peer reviewed. The abstract does not report finer-grained demographic breakdowns, cluster counts, or additional physiological measures; those details were not in the abstract and should not be inferred.
Open questions remain: does synchrony persist beyond the minutes after the ride in other contexts? Would a matched non-riding control group—waiting together without riding—show similar increases in happiness or closeness? The authors flagged the absence of a non-riding control as the main limitation reported in the abstract, and that limitation frames the interpretive boundaries of the findings.