Bayesian Modeling of the Mnemonic Similarity Task Using Multinomial Processing Trees

Area of Science:

• Cognitive psychology
• Neuroscience
• Computational modeling

Background:

• The Mnemonic Similarity Task (MST) is a key tool for assessing pattern separation, crucial for distinguishing similar memories.
• Its sensitivity and reliability make it valuable for clinical applications, but deeper understanding of performance is needed.

Purpose of the Study:

• To develop novel cognitive models for two versions of the Mnemonic Similarity Task (MST).
• To apply these models using Bayesian graphical methods for enhanced inference on behavioral data.
• To explore individual differences in decision strategies and lure detection within the MST framework.

Main Methods:

• Development of cognitive models within the multinomial processing tree framework.
• Implementation of models as generative probabilistic models.
• Application of Bayesian graphical modeling to behavioral data from the MST.
• Inclusion of latent-mixture and hierarchical extensions for detailed analysis.

Main Results:

• The combination of cognitive modeling and Bayesian methods provides flexible and powerful inferences for MST performance.
• Latent-mixture extensions successfully identified individual differences in decision strategies.
• Hierarchical extensions enabled fine-grained measurement of lure detection abilities.
• The inclusion of a "similar" response option in the MST was found to reduce individual differences in decision strategies.

Conclusions:

• Cognitive modeling coupled with Bayesian inference offers a powerful approach to analyzing Mnemonic Similarity Task data.
• The MST, particularly with a "similar" response option, is a refined tool for measuring recognition memory and pattern separation.
• These models advance our understanding of cognitive processes underlying memory and have implications for clinical assessment.