Contextual dynamics in lexical encoding across the ageing spectrum: A simulation study.

Area of Science:

• Psycholinguistics
• Cognitive Science
• Computational Linguistics

Background:

• Traditional models emphasize word frequency (WF) in word recognition.
• Recent research highlights contextual diversity (CD) as a superior predictor.
• CD accounts for the variety of contexts a word appears in, unlike simple occurrence counts.

Purpose of the Study:

• To investigate the predictive power of a semantic distinctiveness model trained on social media data for word recognition.
• To compare the performance of different theoretical models in explaining word recognition latency and accuracy across age groups.
• To explore the developmental trajectory of attention mechanisms in lexical organization.

Main Methods:

• Trained a semantic distinctiveness model on over 55 billion word tokens from social media.
• Analyzed word recognition data from over 1 million participants (Mandera et al. Crowdsourcing Project).
• Examined data across six age bands (10-60 years) for approximately 59,000 words.

Main Results:

• Contextual diversity (CD) significantly outperformed word frequency (WF) in predicting word recognition.
• A developmental trend was observed: younger models relied on social environment attention, while older models showed dominance of discourse-based attention.
• This indicates a shift in cognitive mechanisms for lexical organization with age.

Conclusions:

• Contextual diversity is a more robust predictor of word recognition than word frequency.
• Lexical organization involves a dynamic interplay between cognitive mechanisms and environmental information.
• Attention mechanisms in word processing evolve across the lifespan, moving from social to discourse-based focus.