Neural entrainment as a measure of speech segmentation in infants

Poster Session A - Saturday, April 13, 2024, 2:30 – 4:30 pm EDT, Sheraton Hall ABC

Sarah Breen¹ (sarah.breen@manchester.ac.uk), Szilvia Linnert¹, Anna Theakston¹, Alissa Ferry¹; ¹University of Manchester

A first step in language acquisition is breaking down the continuous stream of sounds of speech into individual words, a process called speech segmentation. While traditional end state measures of speech segmentation are informative at a group level (e.g. determining what an age group can do), they might not capture individual differences effectively (Aslin, 2007; Pérez-Edgar et al., 2020). For instance, standard looking preference measures have large within subject variability (DeBolt et al., 2020). There is also an increased risk that related cognitive abilities may confound segmentation based on end state measures (Kabdebon et al., 2022). As an alternative, neural entrainment (NE) can measure speech segmentation while infants discover word boundaries in real-time (Choi et al., 2020; Kabdebon et al., 2015, 2022). The current study investigates whether NE, captured with electroencephalogram, accurately measures speech segmentation in infants. Eight to 9-month-old infants complete two testing sessions 5-7 days apart. Each session includes two experiments that test different cues to word boundaries: transitional probability and stress pattern. Each experiment consists of a 3.6-minute stream of an artificial language followed by 32 test trials in which words and part-words are presented in isolation. NE is quantified from familiarisation phase data while ERP analysis is performed on the test phase data. This study will evaluate NE's reliability, compare it with ERPs and vocabulary to determine whether NE can be used to study individual differences in speech segmentation. Data collection and analysis is ongoing (40% of target sample collected); preliminary results will be presented.

Topic Area: LANGUAGE: Development & aging