Applying the Memory Measurement Model (M3): A Tutorial Using the bmm R Package

The memory measurement model (M3) decomposes categorical response data from memory tasks into latent activation parameters, separating item memory from context-specific binding via competitive retrieval. This tutorial introduces the M3 framework and demonstrates how to apply it to empirical data using the bmm R package, which provides a standardized interface for fitting Bayesian hierarchical measurement models. Four tutorial examples build progressively: In Tutorials 1 and 2, we fit M3 to two common working memory paradigms, simple span and complex span, demonstrating how different choice rules can be applied and how distractor categories extend the basic model. In Tutorial 3, we use the complex span data to illustrate how to build a custom specification based on different theoretical assumptions. Tutorial 4 provides a simulation-based workflow for validating custom specifications through parameter recovery. Complete R scripts accompany each tutorial. After completing the tutorials, readers will be able to apply standard and custom M3 to their own data and validate new model specifications through simulation.