Dynamic self-organization and early lexical development in children.

Area of Science:

• Computational Neuroscience
• Developmental Psychology
• Linguistics

Background:

• Early lexical development in children exhibits complex patterns, including vocabulary spurts and variability.
• Existing models may not fully capture the dynamic and interactive nature of word acquisition.

Purpose of the Study:

• To present DevLex-II, a self-organizing connectionist model simulating early lexical development.
• To account for empirical patterns in children's word acquisition, such as vocabulary spurts and individual differences.

Main Methods:

• Development of DevLex-II, an enhanced connectionist model based on the earlier DevLex.
• Simulation of word acquisition patterns, including effects of word frequency, length, and memory capacity.
• Analysis of developmental plasticity using lesioned network models.

Main Results:

• DevLex-II successfully simulates vocabulary spurts and the influence of word frequency and length.
• Individual differences in acquisition are modeled as a function of phonological short-term memory and associative capacity.
• Lesioned models demonstrate non-monotonic developmental plasticity in recovery from damage.

Conclusions:

• Interactive learning dynamics are crucial for DevLex-II's ability to model early lexical development.
• Input characteristics, representation consolidation, and lesion-related factors modulate vocabulary growth.
• DevLex and DevLex-II offer a robust computational account of early word learning in children.