Discrete-slots models of visual working-memory response times

Area of Science:

• Cognitive Psychology
• Neuroscience
• Computational Modeling

Background:

• Debate on visual working memory (WM) representation: discrete slots vs. shared resources.
• Limited consideration of response-time (RT) predictions in differentiating WM models.
• Need for models that explain both choice accuracy and RT distributions.

Purpose of the Study:

• Formalize and compare discrete-slots and shared-resources models for visual WM change detection.
• Investigate model predictions for choice and RTs.
• Determine which model class best accounts for empirical data.

Main Methods:

• Formalized mixed-state, discrete-slots models and continuous shared-resources models.
• Applied models to visual WM change detection tasks.
• Analyzed individual subject data using qualitative contrasts and quantitative fits to RT distributions.

Main Results:

• Discrete-slots models provided superior qualitative and quantitative fits to RT and choice data compared to shared-resources models.
• Observed RT distributions were better explained as mixtures of memory-based and guessing processes.
• Some evidence for hybrid 'slots plus resources' models emerged at very small memory set sizes.

Conclusions:

• Discrete-slots models offer a more compelling explanation for visual working memory behavior in change detection tasks.
• Response time distributions are crucial for distinguishing between WM theoretical frameworks.
• The findings support a slot-based architecture for visual working memory, with potential resource-sharing mechanisms under specific conditions.