Individual differences in artificial and natural language statistical learning

Area of Science:

• Cognitive Science
• Psycholinguistics
• Computational Neuroscience

Background:

• Statistical learning (SL) is crucial for cognition, but its connection to real-world phenomena like language remains unclear.
• Previous research often assumes SL is a general ability, potentially overlooking domain-specific processing.
• Mixed findings may stem from mismatched task structures in experimental designs.

Purpose of the Study:

• To investigate the relationship between statistical learning and language processing.
• To determine if sensitivity to artificial trigram patterns predicts sensitivity to natural language trigrams.
• To explore whether comparable statistical structures facilitate cross-domain learning.

Main Methods:

• Adult participants learned artificial syllable trigrams.
• Sensitivity to similar statistical structures in natural language was assessed via a multiword chunking task.
• The study focused on trigram pattern learning in both artificial and natural language contexts.

Main Results:

• A positive correlation was found between learning artificial syllable trigrams and sensitivity to high-frequency word trigrams in natural language.
• This suggests that similar computational processes underlie learning comparable statistical structures.
• Short-term statistical learning contributes to long-term language acquisition when structures align.

Conclusions:

• Aligning statistical structures across tasks is key to understanding the link between statistical learning and cognition.
• Specific computations for processing statistical patterns may span across different domains, including language.
• This approach offers a pathway to clarifying the broader role of statistical learning in cognitive functions.