Can personality traits be predicted from resting-state EEG oscillations? A replication study

Abstract

Personality neuroscience seeks to uncover the neurobiological underpinnings of personality. Identifying links between measures of brain activity and personality traits is important in this respect. Using an entirely inductive approach, Jach et al. (2020) attempted to predict personality trait scores from resting-state spectral electroencephalography (EEG) using multivariate pattern analysis (MVPA) and found meaningful results for Agreeableness. The exploratory nature of this work and concerns about replicability in general require a rigorous replication, which was the aim of the current study. We applied the same analytic approach to a large data set (N = 772) to evaluate the robustness of the previous results. Similar to Jach et al. (2020), 8 min of resting-state EEG before and after unrelated tasks with both eyes open and closed were analyzed using support vector regressions (SVR). A 10-fold cross-validation was used to evaluate the prediction accuracy between the spectral power of 59 EEG electrodes within 30 frequency bins ranging from 1 to 30 Hz and Big Five personality trait scores. We were not able to replicate the findings for Agreeableness. We extended the analysis by parameterizing the total EEG signal into its periodic and aperiodic signal components. However, neither component was meaningfully associated with the Big Five personality traits. Our results do not support the initial results and indicate that personality traits may at least not be substantially predictable from resting-state spectral power. Future identification of robust and replicable brain-personality associations will likely require alternative analysis methods and rigorous preregistration of all analysis steps.