Visual working memory for real-world objects is resistant to visual but vulnerable to semantic interference.

Area of Science:

• Cognitive Psychology
• Neuroscience
• Computer Science

Background:

• Visual working memory (VWM) capacity is higher for meaningful objects than simple stimuli.
• This is thought to be due to the high dimensionality of real-world object representations, which shields visual features from interference.
• The potential for object meaning to introduce semantic interference in VWM remains unexplored.

Purpose of the Study:

• To investigate whether object meaning creates semantic interference in visual working memory.
• To quantify low-level visual and high-level semantic similarity of real-world objects.
• To determine how these similarities impact memory performance.

Main Methods:

• Utilized a recurrent neural network to quantify object similarity.
• Assessed semantic similarity using human-provided object feature descriptions.
• Conducted four experiments where participants recalled object locations in VWM.

Main Results:

• High-level semantic similarity predicted memory errors when object meaning was intact.
• Low-level visual similarity did not predict errors under normal conditions.
• When object meaning was degraded, low-level visual similarity became a significant predictor of memory errors.

Conclusions:

• Meaningful object representations in VWM offer protection against low-level visual feature interference.
• Simultaneously, these representations increase susceptibility to high-level semantic interference.
• The interplay between visual and semantic similarity is crucial for understanding VWM capacity and errors.