Sex differences in pediatric EEG inter-subject correlation during naturalistic movie watching: a large-scale characterization of the Healthy Brain Network EEG dataset

The recent investigation into EEG inter-subject correlation (ISC) during naturalistic movie watching in pediatric populations has shed new light on the sex differences observed in this phenomenon. A comprehensive analysis of the Healthy Brain Network EEG dataset, comprising 1143 children and adolescents aged 5 to 21, revealed three key findings.

Firstly, ISC proved to be a potent differentiator between narrative and abstract stimuli, surpassing a resting-state baseline by a factor of 42 to 99 times. This finding builds upon previous research that documented a developmental decline in ISC. Secondly, the study unveiled a pronounced sex-dependent effect, with males consistently outperforming females across all movie clips (t = 9.7 to 13.5; Cohen's d = 0.61 to 0.85).

This disparity was most evident in delta and theta frequency bands, peaking during adolescence, specifically between the ages of 11 and 14. The observed pattern corroborated an earlier, albeit marginal, report, and the current study successfully elucidated the underlying structure that had been previously undetectable.

Remarkably, this sex difference in ISC persisted despite three rigorous robustness analyses, encompassing demographic and clinical factors, same-sex and size-matched templates, ocular component removal, and the exclusion of overlapping recordings from the aforementioned report. While adjusting for interpolation count attenuated ISC by approximately 10 percent, and excluding recordings devoid of cluster interpolation reduced it by 23 percent, neither adjustment could explain the disparity.

Interestingly, evoked response magnitude, which had been previously hypothesized as an alternative explanation, accounted for only 8.6 to 18.1 percent of the observed effect, further weakening its plausibility.

Lastly, and perhaps most significantly, ISC failed to predict any of the four CBCL bifactor dimensions after false discovery rate correction at any feature resolution. This outcome, which excludes standardized associations beyond |β| = 0.106, underscores the fact that ISC, despite its robustness and sex-dependent nature, does not appear to be a suitable biomarker for the domains examined in this investigation.