Lumping and splitting: Developmental changes in the structure of children's semantic networks

Area of Science:

• Developmental science
• Cognitive science
• Computational modeling

Background:

• Mature cognition involves organized semantic representations.
• Computational models explain semantic representation development.
• Models' predictions lack direct testing in young children.

Purpose of the Study:

• To directly test computational model predictions in young children.
• To investigate the developmental emergence of semantic representations.
• To examine the balance between across- and within-domain distinctions.

Main Methods:

• Two experiments were conducted with young children.
• Investigated the representation of broad across-domain vs. fine within-domain distinctions.
• Analyzed how children exploit covariation patterns.

Main Results:

• Results support the hypothesis that broad distinctions are more represented early in development.
• Findings indicate finer-grained distinctions are less represented in early development.
• Evidence aligns with the exploitation of covariation for semantic acquisition.

Conclusions:

• Early semantic development prioritizes broad, across-domain distinctions.
• Children's acquisition of structured semantic representations relies on identifying entity covariation.
• Findings provide direct evidence for computational model predictions in early childhood development.