Computational models can replicate the capacity of human recognition memory.

Area of Science:

• Cognitive Psychology
• Computational Neuroscience

Background:

• Human recognition memory capacity has been previously modeled using power-law relationships.
• Understanding the neural mechanisms underlying memory recall is crucial for cognitive science.

Purpose of the Study:

• To investigate if existing computational models of familiarity discrimination can replicate Standing's findings on human recognition memory.
• To explore the impact of different assumptions about neural representations on model performance.

Main Methods:

• Simulating three models: Familiarity discrimination based on Energy (FamE), Anti-Hebbian, and Info-max.
• Testing model replication of Standing's power-law relationship under varying assumptions of neural stimulus representation.

Main Results:

• Under simplified assumptions, FamE, Anti-Hebbian, and Info-max models reproduced the power-law relationship.
• With more realistic neural representations, only Anti-Hebbian and Info-max models maintained the power-law replication.
• Model-generated power-law slopes consistently differed from experimental observations.

Conclusions:

• Computational models can partially replicate the power-law nature of recognition memory.
• The ability of models to replicate findings depends on assumptions about neural representations.
• Further research is needed to account for discrepancies in power-law slopes, potentially involving familiarity and recollection processes.