Multidimensional Analysis of Personality Prediction from Electroencephalographic Representations

Associations between personality and resting brain activity have proved difficult to reproduce, leaving unresolved whether electrophysiological information is broadly absent or selectively detectable across personality dimensions. Here, we investigated this question in 254 adults undergoing clinical assessment, examining 15 personality-pattern scales of the Millon Clinical Multiaxial Inventory–IV using resting-state electroencephalography (EEG). We evaluated eight complementary EEG feature families across four representations of personality scores—continuous, two binary thresholds and ordinal—using nested participant-level cross-validation. Most of the 60 scale–formulation combinations showed little predictive performance beyond chance. The strongest evidence concerned Masochistic scores dichotomized at a base-rate score of 60 (area under the curve, 0.630 and 0.594 across two model-aggregation widths), the only combination meeting the within-formulation false-discovery-rate criterion at both widths in fixed-candidate permutation tests. Weaker evidence emerged for continuous Turbulent scores and thresholded Negativistic scores, while an ordinal Histrionic association was identified in a post hoc three-class analysis. Although particular EEG feature families recurred among better-performing configurations, the exact configurations varied across folds. These findings do not support broadly accurate prediction of clinically oriented personality dimensions from resting EEG. Instead, they reveal modest, selective associations whose detection varies across personality dimensions and their operational definitions. Whether these associations generalize, reflect enduring individual differences or provide clinically useful information remains to be established.