Serial memory: Putting chains and position codes in context.

Area of Science:

• Cognitive Psychology
• Neuroscience of Memory
• Computational Neuroscience

Background:

• Serial memory research has long debated chaining vs. position coding theories.
• Chaining models struggled to explain specific error patterns in serial recall tasks.
• Position coding models became dominant due to their ability to account for these errors.

Purpose of the Study:

• To analyze a retrieved context model that integrates both chaining and position coding mechanisms.
• To demonstrate how this unified model can account for findings explained by both theoretical frameworks.
• To propose a reconciliation of competing theories in serial memory.

Main Methods:

• Analysis of a retrieved context model with varying parameter values.
• Mathematical derivation of position codes from context representations.
• Simulation of model predictions to match existing data.

Main Results:

• The retrieved context model can produce classic chaining under certain parameters.
• Under most parameters, the model generates context representations containing position information.
• Extracted position codes are mathematically equivalent to those in established position coding models.

Conclusions:

• Chains, position codes, and retrieved contexts may represent different information extraction strategies.
• Item-dependent context representations, built from fading memory traces, are crucial.
• The retrieved context model offers a unified framework for understanding serial memory.