Perceptual similarity judgments do not predict the distribution of errors in working memory

Area of Science:

• Cognitive Psychology
• Neuroscience
• Computational Vision

Background:

• Population coding models explain visual working memory (VWM) retrieval errors.
• Recent work links population coding to signal detection theory, proposing a new hypothesis for recall error distributions based on stimulus space similarity.

Purpose of the Study:

• To test the hypothesis that psychophysical similarity functions explain VWM recall errors.
• To compare a novel exponential-like responsiveness function with the canonical von Mises function for VWM models.

Main Methods:

• Measured psychophysical similarity and VWM errors in participants across four visual feature spaces.
• Analyzed group and individual level data to assess relationships between similarity and recall.
• Compared goodness-of-fit for exponential-like and von Mises functions.

Main Results:

• Found a common source of variability in similarity judgments and VWM recall estimates.
• Did not find a consistent relationship between psychophysical similarity functions and VWM errors.
• Demonstrated that the novel exponential-like function offered no advantage over the von Mises function.

Conclusions:

• The hypothesis linking psychophysical similarity directly to VWM recall error distributions was not supported.
• Existing population coding models, particularly those using the von Mises function, remain viable for explaining VWM errors.