Infant Phonetic Learning as Perceptual Space Learning: A Crosslinguistic Evaluation of Computational Models

Area of Science:

• Cognitive Science
• Developmental Psychology
• Linguistics

Background:

• Infants' speech perception tunes to their native language within the first year of life.
• This attunement is traditionally explained by phonetic category acquisition.
• Alternative theories propose perceptual space learning without distinct categories.

Purpose of the Study:

• To investigate perceptual space learning as a model for early phonetic learning.
• To implement and test computational models of perceptual space learning.
• To compare model performance against established findings in infant speech perception.

Main Methods:

• Implemented five distinct perceptual space learning models.
• Tested models on three phonetic contrasts from infant speech perception literature.
• Evaluated models' ability to replicate crosslinguistic differences.

Main Results:

• A model using distributional acoustic information explained some crosslinguistic differences.
• A second model incorporating word-level guidance performed comparably.
• Both models support perceptual space learning but require further refinement.

Conclusions:

• Perceptual space learning offers a viable framework for early phonetic learning.
• Acoustic and word-level information are both important for this process.
• Future research needs finer data to differentiate between specific perceptual space learning models.