Investigating Lexical Segmentation in 7.5-Month Old Infants

Abstract

The present study aims to better pinpoint the amount of exposure a 7.5-month-old infant requires to become familiar with an unfamiliar word such that it helps infants correctly identify, or segment, a subsequent novel word in running speech. Previous literature has shown that statistical learning and segmentation are instrumental in language acquisition in humans. Infants detect the predominant strong-weak stress pattern of their native language (Jusczyk, Cutler, and Redanze, 1993). At certain points in development, infants weight segmentation strategies and stress patterns differently leading them to incorrectly segment TARis after hearing “The guitar is…”. Infants have shown that they are able to override a metrical bias (that otherwise leads them to missegment words) when presented with a highly familiar phrase containing "Mommy" or "Mama" but not when presented with an unfamiliar word like "Lola" (Sandoval, LeClerc, & Gómez, 2016). We tested the hypothesis that infants could learn "Lola" with sufficient exposure and whether or not this previously unfamiliar word would similarly allow them to override the metrical bias to accurately segment a weak-strong word. Infants became familiar enough to "Lola" after two weeks of exposure that it assisted them in segmenting weak-strong patterned words following it in running speech.