Contextual effects in visual working memory reveal hierarchically structured memory representations.

Area of Science:

• Cognitive Psychology
• Neuroscience
• Visual Perception

Background:

• Current models of visual working memory (VWM) assume items are stored independently.
• These models predict performance is determined solely by set size (number of items).
• This overlooks potential interactions and relationships between items.

Purpose of the Study:

• To investigate whether display difficulty varies within a constant set size in VWM.
• To determine if inter-item relationships, beyond individual item storage, influence VWM capacity.
• To challenge the assumption of independent item storage in VWM models.

Main Methods:

• Hundreds of participants were tested on remembering identical sets of visual displays.
• Analysis focused on identifying consistent patterns of display difficulty and item memorability.
• Computational modeling explored hierarchical representations and inter-item features (e.g., color statistics).

Main Results:

• Significant variability in display difficulty was observed even with the same set size.
• Participants showed high consistency in identifying easier and harder items/displays.
• A hierarchical model incorporating ensemble and texture representations better explained display variability than simple chunking.

Conclusions:

• Working memory storage is not solely based on independent item representations.
• Inter-item representations, such as perceptual grouping, ensemble, and texture, are crucial for VWM performance.
• Existing VWM models need to incorporate these relational representations to fully account for memory capacity.