Statistical language learning in neonates revealed by event-related brain potentials

Area of Science:

• Cognitive Neuroscience
• Developmental Psychology
• Linguistics

Background:

• Statistical learning is crucial for rapid spoken language acquisition.
• Infants aged 8 months can segment fluent speech into words using statistical learning.
• This study investigated statistical learning abilities in neonates at birth.

Purpose of the Study:

• To determine if neonates possess statistical learning capabilities for speech segmentation.
• To examine brain responses to statistical cues in auditory speech streams.
• To understand the role of statistical learning in early language development.

Main Methods:

• Recorded event-related brain responses (ERPs) in sleeping neonates.
• Exposed neonates to continuous streams of syllables with statistical word boundary cues.
• Analyzed syllable-specific ERPs to assess differential brain processing based on syllable position.

Main Results:

• Neonates demonstrated the ability to extract statistical properties from speech input.
• Detected word boundaries in continuous syllable streams lacking morphological cues.
• Syllable-specific brain responses indicated differential processing of syllables within pseudowords.

Conclusions:

• Neonates can learn transitional probabilities between syllables to identify word boundaries.
• This ability allows newborns to isolate words from fluent natural speech.
• Statistical learning of speech structures is a fundamental early prerequisite for language acquisition.