A Multinomial Processing Tree Model of the 2-back Working Memory Task

Area of Science:

• Cognitive Psychology
• Cognitive Neuroscience
• Psychometrics

Background:

• The 2-back task is a standard measure of working memory and inhibitory control.
• Existing models may not fully dissociate underlying cognitive processes.

Purpose of the Study:

• To develop a novel computational model for the 2-back task using Multinomial Processing Trees.
• To provide interpretable parameters for memory, decision-making, and response bias.
• To apply and validate the model on diverse datasets.

Main Methods:

• Developed a three-parameter model (memory, decision, base-rate) within the Multinomial Processing Trees framework.
• Tested parameter recovery using existing 2-back experimental designs.
• Applied the model to datasets from social psychology (racial face perception) and cognitive neuroscience (Human Connectome Project).

Main Results:

• Demonstrated the model's ability to recover parameters and infer individual-level cognitive processes.
• Extended the model hierarchically to compare stimulus conditions (e.g., different races).
• Utilized a multivariate regression extension to link model parameters with performance on other cognitive tasks (List Sorting, Flanker).

Conclusions:

• The novel model effectively dissociates cognitive processes like encoding, inhibition, and binding failures in the 2-back task.
• The model provides a flexible framework for analyzing working memory and inhibition across different research domains.
• This approach facilitates a deeper understanding of individual differences in cognitive control.