Input features mediate word segmentation efficacy, peaking at 14 months: A computational modeling study

This study investigates age-related variations in the properties of Child-Directed Speech (CDS) and their impact on the processing efficiency, particularly in simulated word segmentation. By analyzing a corpus of Korean CDS directed at 35 children aged 6 to 30 months, significant age-related shifts were identified. The results revealed an inverse U-shaped pattern in CDS segmentability by age, with a peak at around 14 months. Further analysis indicated that the effect of age was significantly attenuated when linguistic properties were taken into account. This study highlights the age-dependent effects of CDS properties on word segmentation, and underscores the importance of considering the interaction between age and linguistic input in language acquisition models.