Strategy execution in cognitive skill learning: an item-level test of candidate models

Area of Science:

• Cognitive Psychology
• Human Learning and Memory
• Computational Cognitive Science

Background:

• Many tasks initially rely on multistep algorithms before transitioning to memory-based performance with practice.
• Understanding the learning mechanisms behind this transition is crucial for cognitive science and educational psychology.

Purpose of the Study:

• To investigate the transition from algorithmic processing to memory-based performance in a practiced task.
• To test the predictive power of component power laws against parallel strategy execution models.
• To determine if the shift to retrieval is a task-general or item-specific learning phenomenon.

Main Methods:

• An alphabet arithmetic task was used to measure item response times.
• T. C. Rickard's (1997) component power laws model was applied to analyze performance changes.
• Strategy probes were used to validate inferred strategy shifts based on speed-up patterns.

Main Results:

• Pronounced step-function decreases in item response times were observed after practice.
• These speed-up patterns were uniquely predicted by the component power laws model.
• The shift to memory retrieval was found to be an item-specific learning phenomenon.

Conclusions:

• The findings challenge existing parallel strategy execution models and smooth speed-up functions as global learning laws.
• Component power laws provide a more accurate model for explaining practice-based performance transitions.
• The study validates the use of strategy probes in conjunction with speed-up patterns for inferring learning processes.