Measuring Individual Differences with Bayesian Hierarchical Cognitive Models

Cognitive models have been widely used to understand human cognition and, in principle, offer tools for measuring cognitive abilities at the individual level, but they have rarely been used for that purpose. This study uses three simulations to examine the validity of applying cognitive measurement models to individual-differences research. Specifically, we assessed the extent to which parameter estimates reflect true individual differences in model parameters (parameter recovery) and the extent to which relations between parameters and other variables were recovered correctly (correlation recovery). We further considered the implications of correlation recovery for the accuracy of estimating structural equation models used in individual-differences research. Taken together, the findings provide practical guidance for designing individual-differences studies using cognitive models, with implications for model choice, task design, latent-variable modeling, model specification, and cross-model parameter relations.