Measurement models for visual working memory-A factorial model comparison.

Area of Science:

• Cognitive Psychology
• Computational Neuroscience
• Psychometrics

Background:

• Existing measurement models for visual working memory (WM) include the mixture model, interference measurement model (IMM), and target confusability competition (TCC) model.
• These models are used to analyze data from the continuous-reproduction paradigm, a common method for studying WM.
• A unified framework is needed to compare and understand the relationships between these diverse WM measurement models.

Purpose of the Study:

• To define a comprehensive space of possible measurement models for visual working memory data.
• To systematically evaluate existing and novel models within this space using empirical data.
• To identify the most empirically adequate measurement model for the continuous-reproduction paradigm.

Main Methods:

• A model space was constructed based on three key dimensions: activation function (von-Mises vs. Laplace), response-selection function (Luce's choice vs. signal-detection theory), and memory precision constancy.
• A factorial combination of these dimensions generated all possible models within the defined space.
• All generated models were fitted to eight independent datasets from visual working memory experiments.

Main Results:

• A novel model, combining a von-Mises activation function with a signal-detection response-selection rule, emerged as the most empirically adequate.
• Memory precision could be treated as a constant across many experimental manipulations, although individual differences in precision are likely.
• The proposed model space successfully encompasses previously proposed models, providing a unified framework for their comparison.

Conclusions:

• The newly identified model provides a superior fit to continuous-reproduction visual working memory data compared to existing models.
• The findings suggest that the choice of activation and response-selection functions significantly impacts the accuracy of WM measurement.
• This research offers a flexible framework for developing and testing future visual working memory measurement models.