A Unified Framework for Psychometrics in Experimental Psychology: The Standardized Generalized Hierarchical Factor Model

To address Cronbach’s longstanding call to unify experimental and correlational psychology, Hierarchical Factor Models (HFMs) have emerged as a promising approach to estimate individual differences while accounting for trial-level noise. As formalized by Rouder et al. (2025), this framework assumes Gaussian response times (RTs) and operates in raw units. Here, we generalize this framework by introducing the Standardized Generalized HFM (GenHFM), which incorporates two natural extensions: explicitly modeling asymmetric distributions and implementing a fully standardized latent structure. First, modeling asymmetry allows us to capture the true shape of RTs and, crucially, quantify the bias introduced by fitting Gaussian HFMs to skewed data. Second, the standardized parameterization ensures transparent prior specification and facilitates the analysis of factor loadings, enabling researchers to directly compare the degree of association between experimental effects and common factors, regardless of their original metrics. Consequently, this allows researchers to evaluate the discriminative capacity of each task regarding the latent process, mirroring the assessment of construct validity in classical psychometrics within experimental settings. We conducted a simulation study showing that Gaussian HFMs can underestimate true correlations by up to 50%, whereas GenHFM yields unbiased and more efficient estimates. Finally, we re-analyze data from two published studies using executive control tasks to illustrate the empirical application of these models. Furthermore, we show that GenHFM achieves better predictive accuracy than complex models typically used in the field, such as hierarchical models with covariates or Drift Diffusion Models.