Simulating procedural discovery in early language acquisition: Domain-general cognition with contextual learning

Area of Science:

• Cognitive Science
• Computational Linguistics
• Developmental Psychology

Background:

• Early language acquisition involves complex learning processes.
• Existing models often focus on specific phenomena, lacking unification.
• Understanding developmental trajectories is crucial for typical and atypical language development.

Purpose of the Study:

• To present a unified cognitive architecture for early language acquisition.
• To simulate and explain phenomena observed in laboratory and naturalistic settings.
• To explore the influence of processing efficiency on language learning.

Main Methods:

• Developed a simulation study based on a cognitive architecture.
• Employed adaptive learning through trial-and-error with general-purpose operators.
• Utilized learned contextual associations to optimize performance.
• Simulated preferential focusing to explain specific learning behaviors.

Main Results:

• The model explains delayed statistical learning and age-related decreases in algebraic processing.
• Demonstrated continuous, implicit learning rather than explicit strategy acquisition.
• Showcased procedural evolution and emergence of multiple task strategies.
• Provided a proof-of-concept for word-level phonological learning from infant-directed speech.

Conclusions:

• The cognitive architecture successfully unifies diverse early language acquisition phenomena.
• Age-related processing efficiency significantly influences language learning trajectories.
• The model offers insights into both typical and atypical early language development.
• The framework has broader implications for modeling real-world language acquisition.