Psychometric, pre-processing, and trial-type considerations in individual differences studies of EEG mid-frontal...

EEG mid-frontal theta-band activity (MFT; 4-8 Hz) is of considerable interest as a possible biomarker in translational research on cognitive control. However, because most of the MFT literature has focused on experimental within-subjects effects, the impact of particular data processing choices on individual difference analyses is not well understood. This study aimed to reduce that gap by examining the psychometric properties of different pipelines for measuring individual differences in MFT power and latency. Ninety-three adults aged 60 or older completed a flanker task during EEG recording. Stimulus-locked MFT was extracted in three primary pipelines via the fast Fourier transform (FFT), linear- and log-spaced wavelet, and filter-Hilbert analyses. The effects of frequency resolution, electrode choice, overall versus peak power, and trial type (overall, congruent, incongruent, and subtraction- and regression-based residual scores contrasting congruent and incongruent activity) were examined, as was the degree of overlap among related variables. Internal consistency reliabilities and associations with reaction times (RT) during the Flanker were assessed for select measures. Results indicated no benefit of higher frequency resolutions or region of interest over single-electrode measurements from FCz. Two-part coefficient alpha reliabilities ranged from 0.63 to 1.00 for MFT power variables, and from 0.02 to 0.83 for latency variables. Contrary to hypotheses and common criticisms of derived scores, correlations with RT were generally strongest for the difference and residual scores, with an additional benefit of time-frequency-based peak power relative to FFT-based overall power (r ~ = 0.30 vs. 0.45). These findings add to the growing literature on psychometric properties of EEG biomarkers, and help clarify measurement strategies that may enhance detection of behavioral and clinical correlates of MFT power and latency.