A spatially continuous diffusion model of visual working memory.

Area of Science:

• Cognitive Psychology
• Computational Neuroscience
• Psychophysics

Background:

• Visual working memory (VWM) models traditionally focus on accuracy, often neglecting response time (RT) data.
• Existing discrete and resource models of VWM do not inherently account for the temporal dynamics of memory retrieval.

Purpose of the Study:

• To evaluate discrete and resource VWM models by integrating them with the spatially continuous diffusion model (SCDM).
• To investigate how RT data can refine and discriminate between different VWM models.

Main Methods:

• Conducted five VWM experiments presenting 2-6 colored squares, followed by cued recall on a continuous color wheel.
• Collected response proportions and RT data as a function of choice angle.
• Implemented and evaluated VWM models within the SCDM framework, which models evidence accumulation with continuous distributions and Gaussian noise.

Main Results:

• The best-fitting models from both discrete and resource classes converged on a common structure.
• This unified model included a "guessing component" (modeled as a zero-drift process) and allowed memory strength (distribution height) to vary with the number of items.
• Integrating choice and RT data enabled discrimination between models not identifiable from choice data alone.

Conclusions:

• A hybrid VWM model incorporating a guessing process and item-dependent memory strength provides a superior account of both accuracy and RT data.
• The SCDM framework is effective for evaluating and comparing VWM models using comprehensive behavioral data.
• Response time data are crucial for a complete understanding of visual working memory dynamics.