Our aim was to investigate the understudied working mechanisms of schema therapy by empirically testing the models proposed by Arntz et al. (2021). We tested the six hypothesized mediation and moderated mediation models from the position paper, and also assessed the psychometric properties of the newly developed Internalization and Externalization Inventory. Data from 139 participants were analyzed using Structural Equation Modeling with a latent Vulnerable EMS factor. Results suffered a consistent and pervasive lack of acceptable model fit across all six structural models, suggesting that their specified relationships do not fully capture the observed data. Furthermore, the IEI's two-factor structure and the Externalizing subscale's reliability were not confirmed. Notably, strong direct effects of Vulnerable EMS on most child modes persisted, indicating unmodeled influences. We attribute the lack of overwhelming support to potential model misspecification and methodological limitations including a very limited sample size and consequent non-normal data. Future research requires larger samples, refined questionnaires, and longitudinal designs to shed more light on these complex mechanisms, with temperament and resignation remaining promising variables for the investigation of schema therapy’s working mechanisms.