Word Segmentation and Multilingualism

When we listen to someone speak, there are no clear pauses between words, yet we can easily figure out where one word ends and another begins. This skill, called word segmentation, is essential for learning any language. Our research explores how growing up with more than one language shapes the ability to find words in continuous speech. By examining the different cues that learners use to recognize word boundaries, we aim to better understand how multilingual experiences influence language learning and brain development.

To investigate this, we study how infants and adults process speech using a variety of methods. In some studies, we use eye-tracking technology to measure where participants look and how their pupils respond while they listen to speech. We focus on languages like English and Cantonese, which differ in the availability and reliability of segmentation cues such as stress patterns and statistical regularities. By comparing how listeners segment speech in these different languages, we can explore how experience with multiple language systems shapes sensitivity to different types of linguistic information.

In addition to stress and statistical patterns, we are also interested in how learners use other kinds of regularities, such as vowel harmony, to segment speech. Vowel harmony refers to a pattern where vowels within a word share similar properties, like tongue position or lip rounding. In these studies, infants are first familiarized with streams of artificial speech that follow vowel harmony patterns. During the test phase, they hear sequences that represent different possible ways of segmenting the familiarization stream: some sequences are harmonic (preserving vowel harmony within words), while others are disharmonic (breaking vowel harmony across word boundaries). All other properties, such as frequency and transitional probabilities, are carefully matched across conditions. By comparing younger and older infants, as well as monolingual and bilingual learners, we examine how sensitivity to different types of linguistic structure develops with age and experience.