SPEECH SEGMENTATION IN A SIMULATED BILINGUAL ENVIRONMENT: A CHALLENGE FOR STATISTICAL LEARNING?

Area of Science:

• Cognitive Science
• Psycholinguistics
• Second Language Acquisition

Background:

• Infants and adults typically segment words using statistical learning from a single language.
• Bilingualism is common, yet how individuals segment words from multiple languages is less understood.
• Combining input across languages may alter unit statistics, potentially hindering word segmentation.

Purpose of the Study:

• To investigate how adult second language (L2) learners segment words from bilingual input.
• To determine if learners can track statistical information from multiple artificial languages simultaneously.
• To explore the formation of multiple representations during early stages of bilingual language acquisition.

Main Methods:

• Utilized artificial language streams to simulate early bilingual input exposure.
• Conducted four experiments with adult L2 learners tracking statistics from two languages.
• Analyzed participants' ability to maintain separate statistical tracking for each language.

Main Results:

• Adult learners successfully tracked two independent sets of statistical information concurrently.
• Evidence suggests the formation of distinct mental representations for each language.
• Performance indicates robust statistical learning capabilities in a bilingual context.

Conclusions:

• Adults can form and maintain multiple representations when exposed to simultaneous language input.
• This ability supports effective word segmentation in early bilingual language processing.
• Findings inform critical questions in bilingualism research regarding the development of multiple language representations.