Evaluating and excluding swap errors in analogue tests of working memory.

Area of Science:

• Cognitive Psychology
• Neuroscience
• Visual Perception

Background:

• Working memory recall involves two error types: feature variability and swap errors.
• Previous methods for distinguishing these errors relied on unverified assumptions about noise distribution.

Purpose of the Study:

• To develop and apply novel non-parametric methods for independently assessing feature variability and swap errors in visual working memory.
• To re-evaluate the prevalence and nature of swap errors using a less assumption-dependent approach.

Main Methods:

• Introduction of novel non-parametric (NP) methods to estimate swap error frequency and feature variability.
• These NP methods require fewer prior assumptions and no fitting procedures, allowing for more robust disentanglement of error sources.

Main Results:

• Swap errors are significantly more prevalent than previously estimated, comprising over a third of responses at a set size of 8.
• When location cues are used, items spatially closest to the target are most frequently mistaken for it, suggesting a link between spatial noise and swap errors.

Conclusions:

• Non-parametric methods offer a more accurate way to study the structure of internal representations and item binding in working memory.
• Swap errors are a substantial component of working memory recall errors, potentially driven by noise in the feature dimension, particularly spatial proximity.