Superior Statistical Learning Relies on Rejecting Partwords.

Area of Science:

• Cognitive Science
• Psycholinguistics
• Speech Perception

Background:

• Speech segmentation is crucial for language acquisition.
• Individual differences in statistical learning (SL) performance are observed but not fully understood.
• Existing models suggest memory-based mechanisms influence learning outcomes.

Purpose of the Study:

• To investigate the learning mechanisms underlying individual differences in speech segmentation.
• To explore how memory-based models explain performance variations in SL tasks.
• To differentiate learning strategies between superior and regular learners.

Main Methods:

• Seventy-four participants were exposed to a novel language.
• A 2-alternative-forced-choice (2AFC) task assessed speech segmentation ability.
• A familiarity rating task (7-point Likert scale) was administered.
• Participants were categorized into superior and regular learning groups based on 2AFC performance.
• Linear mixed-effects models analyzed rating differences for target words, nonwords, and partwords.

Main Results:

• Both superior and regular learners demonstrated above-chance performance in 2AFC and familiarity tasks.
• Superior learners rated partwords as significantly less familiar compared to regular learners.
• No significant differences in familiarity ratings were found for target words or nonwords between the groups.
• Perceived familiarity of partwords emerged as the key differentiator in overall SL task performance.

Conclusions:

• Individual differences in speech segmentation are significantly influenced by the perceived familiarity of partwords.
• A memory-based model provides a viable framework for understanding these performance variations.
• The study introduces a refined approach to examining individual differences in statistical learning tasks.