Forgetting in visual working memory: Internal noise explains decay of feature representations

Area of Science:

• Cognitive Psychology
• Neuroscience
• Visual Perception

Background:

• Visual working memory (VWM) precision declines over time.
• Forgetting is often attributed to random noise or systematic feature loss.
• Previous research on VWM decay models yielded conflicting results.

Purpose of the Study:

• To rigorously investigate the mechanisms of VWM decay.
• To differentiate between systematic feature decay and random internal noise accumulation.
• To determine if VWM decay affects all features equally or targets specific ones.

Main Methods:

• Employed signal detection theory methods, including psychophysical classification images and ROC analysis.
• Utilized a modified same-different memory task with short (500 ms) and long (4000 ms) retention intervals.
• Tested VWM decay using both compound grating and radial frequency pattern stimuli.

Main Results:

• Memory performance decreased by approximately 15% with longer retention times.
• Classification images indicated consistent feature weighting across retention intervals, refuting feature-specific decay.
• A significant 77% increase in stimulus-independent internal noise was observed at longer retention times.
• Shallower ROC curve slopes confirmed the rise in internal noise.

Conclusions:

• VWM decay is primarily driven by an increase in random internal noise, not a systematic loss of specific stimulus features.
• The findings support a model where uniform internal noise accumulation underlies forgetting in VWM.
• This research clarifies the fundamental mechanisms governing the fidelity of visual working memory over time.