More than words: Adults learn probabilities over categories and relationships between them

Area of Science:

• Cognitive Science
• Psycholinguistics
• Language Acquisition

Background:

• Human statistical learning is crucial for language acquisition.
• Previous research focused on learning individual word frequencies or simple sequential patterns.
• The ability to learn abstract grammatical category statistics remained less understood.

Purpose of the Study:

• To investigate if human learners can acquire statistical regularities over abstract grammatical categories.
• To determine if learners track probabilities at the category level or item level.
• To explore the role of abstract linguistic categories in statistical learning.

Main Methods:

• Adult participants were exposed to artificial miniature languages with varying word order frequencies (Subject, Object, Verb).
• Language structure necessitated tracking statistics over abstract categories, not just individual words.
• Testing involved assessing participants' reproduction of learned input probabilities.

Main Results:

• Participants accurately reproduced the statistical probabilities present in the artificial language input.
• Learners matched probabilities associated with individual verbs, not abstract categories broadly.
• Individual nouns did not influence the word orders produced by participants.

Conclusions:

• Statistical learning mechanisms can track relationships between abstract linguistic categories.
• Learners demonstrated an ability to associate specific word order probabilities with individual verbs.
• Findings suggest a nuanced statistical learning capability extending beyond individual items to abstract grammatical structures.