Accessible Cognitive Measurement Models

Easing the application and use of formal models

Implementing commonly used cognitive measurement models in accessible R packages so more researchers can use them in their work.
Published

December 2, 2022

Cognitive measurement models provide a mathematical formalization of the cognitive processes underlying observed behavior.

Cognitive measurement models (the diffusion model for processing speed is a well-known example) translate verbal theories of cognitive processes into mathematical form. Rather than treating observed behavior as a direct readout of some construct, these models describe response times, accuracy rates, and choice patterns as the joint product of several interacting processes. The parameters of the model correspond to specific cognitive operations — and that correspondence is what makes the model theoretically useful, rather than just descriptive.

The problem

Researchers who develop cognitive measurement models usually provide code, but running it typically requires fluency in JAGS or Stan. That is a real barrier. Researchers who would benefit from these models often cannot use them without weeks of investment in a new programming environment — so they fall back on behavioral summary statistics that are theoretically ambiguous.

The bmm package

Together with Ven Popov, I co-develop the R package bmm (CRAN), which implements these models in a hierarchical Bayesian framework via brms. The package started with models for visual working memory, but it has since grown into a broader collection covering both memory and speeded decision-making.

Visual working memory (continuous reproduction tasks):

Working memory tasks with categorical responses:

  • Memory Measurement Model (M3) (Oberauer & Lewandowsky, 2019), with ready-made versions for simple and complex span tasks and a custom version for other categorical decision tasks

Speeded decision-making (new in the upcoming release, already available in the development version):

Each model ships with documentation, default priors, a worked-example vignette, and density and random generation functions for exploring what the model predicts before fitting it.

Teaching these models to PhD students

Making the models accessible in software is only half the job — researchers also need training in how to use them well. I currently pursue this through:

An overview of my other workshops is on the teaching page.

Publications & preprints

Published:

  • Frischkorn, G. T., & Popov, V. (2025). A tutorial for estimating Bayesian hierarchical mixture models for visual working memory tasks: Introducing the Bayesian Measurement Modeling (bmm) package for R. Behavior Research Methods, 57(5), 144. https://doi.org/10.3758/s13428-025-02643-0
  • Nunez, M. D., Schubert, A.-L., Frischkorn, G. T., & Oberauer, K. (2025). Cognitive models of decision-making with identifiable parameters: Diffusion decision models with within-trial noise. Journal of Mathematical Psychology, 125, 102917. https://doi.org/10.1016/j.jmp.2025.102917

Preprints:

  • Frischkorn, G. T., & Li, C. (invited for revision at Advances in Methods & Practices for Psychological Science). Applying the Memory Measurement Model (M3): A tutorial using the bmm R package. https://osf.io/tcuzx_v1
  • Li, C., Frischkorn, G. T., & Oberauer, K. (under review at Behavior Research Methods). Measuring individual differences with Bayesian hierarchical cognitive models. https://osf.io/24w3z_v1
  • Göttmann, J., Frischkorn, G. T., Oberauer, K., Schaefer, S. B., & Schubert, A.-L. (invited for revision at Journal of Mathematical Psychology). Modeling individual differences in working memory: Subject-level parameter recovery within the Memory Measurement Model framework (M³). https://doi.org/10.31234/osf.io/945d2_v2
  • Frischkorn, G. T., Rebmann, R., & Oberauer, K. (under review at Psychological Review). Validity is a theoretical problem: A computational psychometrics perspective on how to measure cognition. https://osf.io/pczua_v3