Native Language Influence in the Segmentation of a Novel Language

Abstract

A major problem in second language acquisition (SLA) is the segmentation of fluent speech in the target language, i.e., detecting the boundaries of phonological constituents like words and phrases in the speech stream. To this end, among a variety of cues, people extensively use prosody and statistical regularities. We examined the role of pitch, duration, and transitional probabilities (TPs) between adjacent syllables in the segmentation of a novel language by native speakers of German and compared their responses with the segmentation by the listeners with a phonologically different native language: Italian. We used an artificial language with different prosodic cues marking the boundaries of statistically defined words. In artificial-language learning experiments, we compared how Germans and Italians use prosodic and statistical cues for segmenting continuous speech. We show that native phonology modulates the processing of prosodic cues in novel languages. While native speakers of Italian interpret prosodic cues at both the word and the phrasal level, native speakers of German interpret them exclusively at the phrasal level. Phrasal prosody can facilitate the segmentation of a novel language when prosodic and statistical cues lead to the same segmentation solution. Word-level prosody does not necessarily facilitate segmentation, but it can disrupt it when statistical and prosodic cues lead to different segmentation solutions.