Drift in Neural Population Activity Causes Working Memory to Deteriorate Over Time

Area of Science:

• Neuroscience
• Cognitive Science
• Computational Neuroscience

Background:

• Short-term memories are theorized to be maintained by sustained neural spiking activity.
• Recall precision decreases with more memorized items, potentially due to limited spiking activity.
• The impact of retention duration on recall precision and the underlying neural mechanisms remain unclear.

Purpose of the Study:

• To investigate the mechanisms behind short-term memory degradation over time.
• To differentiate between signal decay and value drift as explanations for reduced recall precision with longer retention intervals.
• To test predictions derived from a spike integration model of decision-making.

Main Methods:

• Human participants performed a variable-delay cued recall task with saccadic eye movement responses.
• Response latency and recall precision were precisely measured.
• A computational model based on limited neural resources was used to reproduce experimental results.

Main Results:

• Recall precision decreased with longer retention durations, but response latency did not systematically increase.
• Increasing the number of memorized items increased response latency.
• Results were quantitatively reproduced by a model where working memories drift over time, rather than decay.

Conclusions:

• Short-term memory degradation is primarily driven by random drift of encoded values, not a simple decay of neural signal.
• This finding supports a dynamic, spiking-based model of working memory.
• The study contrasts with theories proposing activity-silent storage mechanisms for working memory.