A new toolbox to distinguish the sources of spatial memory error

Area of Science:

• Cognitive Psychology
• Neuroscience
• Computational Modeling

Background:

• Understanding working memory (WM) mechanisms relies on analyzing memory recall errors.
• Existing mixture modeling toolboxes can analyze one-dimensional memory features (e.g., color, orientation) to differentiate imprecision, guessing, and misbinding.
• However, these toolboxes are not available for two-dimensional (2D) spatial WM tasks.

Purpose of the Study:

• To present a novel method for isolating sources of spatial error in 2D spatial WM tasks.
• To utilize 2D mixture models to analyze spatial location recall accuracy.
• To provide a computational tool for advancing the study of spatial working memory.

Main Methods:

• Developed and applied 2D mixture models to analyze spatial error sources in WM tasks.
• Validated the method by recovering simulated parameters and testing its robustness against response distributions, biases, and trial parameters.
• Integrated the method into the freely available MemToolbox2D package for MATLAB.

Main Results:

• The 2D mixture model method successfully recovered simulated parameters with high accuracy.
• The model demonstrated robustness against various confounding factors like response biases and the number of non-target items.
• Analysis of a complex spatial WM task showed recovered parameters aligned with previous findings and a 1D analogue, indicating convergent validity.

Conclusions:

• The presented 2D mixture modeling approach effectively isolates spatial error sources in WM tasks.
• 2D spatial tasks offer superior separation of misbinding from imprecision and guessing compared to 1D tasks due to the added spatial dimension.
• The freely available MemToolbox2D facilitates the application of this method in cognitive psychology and neuroscience research.