Sleepiness but neither fluid nor crystallized intelligence can be predicted from resting-state electroencephalography - Evidence from the large scale CoScience EEG-Personality Project.

Abstract

Previous electroencephalogram (EEG) studies linked measures of spectral power under rest and fluid intelligence; however, subsequent high-powered studies challenged this relationship. The present study aimed to address previous limitations (low statistical power, lack of preregistration) and investigated the predictability of intelligence measures from resting-state EEG in the CoScience data set (N = 772). Support vector regressions were applied to analyze 8 min of resting-state EEG with eyes open and closed before and after unrelated tasks. The decoding performance between the spectral power of 59 EEG channels within 30 frequency bins and fluid and crystallized intelligence, was evaluated with a tenfold cross-validation. We could not identify any meaningful associations between resting-state EEG spectral power and either fluid or crystallized intelligence-a null finding that is unlikely to be entirely due to a relatively modest restriction of fluid intelligence variance in our student sample. Moreover, we did replicate the previously reported association between state sleepiness and theta power, attesting to the integrity of the CoScience data set. Furthermore, the decomposition of the EEG signal into its periodic and aperiodic components revealed that the aperiodic offset parameter is significantly correlated with state sleepiness, emphasizing the relevance of aperiodic signal components in understanding states of alertness versus sleepiness.