Adaptive allocation of human visual working memory capacity during statistical and categorical learning.

Area of Science:

• Cognitive Neuroscience
• Computational Neuroscience
• Information Theory

Background:

• Human brains possess finite capacity, necessitating efficient resource management.
• Visual working memory (VWM) is crucial for retaining visual information but is prone to errors due to capacity limits.

Purpose of the Study:

• To investigate how limited VWM capacity is dynamically reallocated based on statistical regularities of stimuli and task requirements.
• To test the hypothesis that VWM adapts flexibly and robustly to optimize performance.

Main Methods:

• Utilized a normative computational modeling framework based on rate-distortion theory.
• Conducted two experiments using controlled, naturalistic images of plants.
• Manipulated stimulus feature distributions and task relevance of different features.

Main Results:

• Experiment 1: Participants adapted VWM performance based on the statistical distribution of stimulus features (leaf width).
• Experiment 2: Participants showed increased sensitivity to task-relevant stimulus dimensions (leaf width) over irrelevant ones (leaf angle).

Conclusions:

• VWM capacity is not fixed but dynamically reallocated in a task-dependent manner.
• Memory performance demonstrates robustness, adapting to statistical regularities and task demands, highlighting the role of learning in VWM.