Sex-based Measurement Invariance in Reinforcement Sensitivity Theory: An Exploratory Structural Equation Model based on Biological Variables.

This study investigates sex-based differences in the neuroanatomical correlates of Reinforcement Sensitivity Theory (RST) using Exploratory Structural Equation Modeling (ESEM). By focusing on gray matter volumes as biological indicators, we aim to examine how RST-related brain structures differ between sexes, and whether these differences affect the theory's structural validity across males and females. Regions of interest were defined using the Neuromorphometrics Atlas, and latent constructs representing the Behavioral Inhibition System/Fight-Flight-Freeze System, the Behavioral Approach System, and a dual Constraint System with dorsal and ventral cortical streams were modeled in 172 males and 128 females. We conducted multigroup invariance analyses using a four-factor model, both with and without correction for intracranial volume (ICV) via the power-corrected proportion method. Results revealed lower model error in females and acceptable configural fit in both models (RMSEA = 0.07; CFI = 0.97 in the non-ICV model and 0.95 in the ICV-corrected model). Nonetheless, measurement invariance was largely supported across sexes, particularly at the metric level (ΔCFI ≤ 0.003 in both models). Observed differences were more pronounced when ICV was not accounted for, underscoring the importance of biological covariates in morphometric studies. Given the age differences between the male and female groups in the whole sample, robustness analysis of 75 age- and education-matched male-female pairs reinforced the results. These findings support the robustness of RST constructs across sexes, underscoring the value of incorporating neurobiological indicators and multigroup ESEM in personality neuroscience. This study advances current understanding of sex-based brain structure variability within the RST framework and highlights the relevance of anatomical correction when modeling psychological constructs from neuroimaging data.