Can within-person production cues facilitate infants’ and adults’ word segmentation of two artificial languages?

• In Event: 3-235 - Contending with talker variability across development

Integrative Statement

Infants and adults are sensitive to syllable co-occurrence, or transitional, probabilities when segmenting words from a single language (e.g., Saffran, Aslin & Newport, 1996). Words have higher syllable transitional probabilities (i.e., their constituent syllables co-occur often) than word boundaries. Yet, most studies have examined statistical learning using a single input language. Given the global prevalence of bilingualism (Grosjean, 2010), it is important to investigate how learners track transitional probabilities from two input languages. Adults can segment two languages when supported by strong between-speaker contextual cues: two languages distinguished by voices from only one gender respectively (Weiss, Gerfen & Mitchel, 2009) or by distinct faces (Mitchel & Weiss, 2010). However, by nature, a bilingual individual has to alternate between two languages, and often does so in one conversation. To address this research gap, we explored how phonetic cues within one individual’s speech could facilitate statistical word segmentation of two languages. Sets of phonetic variations can be unique to each language; these variations therefore effectively highlight the differences between two languages in a bilingual environment. For example, /ba/ is produced with shorter voice onset time and vowel duration in French than in English. Further, as a potential bilingual advantage in statistical learning may exist (Antovich & Graf Estes, 2017; Poepsel & Weiss, 2016), we also examined whether bilingual experience improves learners’ use of phonetic cues to segment words from two languages.

We presented a dual-language statistical word segmentation task to adults (n=99) and an adapted infant version to 9.5-month-olds (n=68). The word boundaries of each artificial language were only marked by transitional probabilities. As the two languages shared some common syllables, successful word segmentation required learners to separate the languages. Here, the two languages, produced by a single speaker, were separated by phonetic cues; one language used English cues and the other used French cues.

Adults successfully learned words from the two languages [t(98)=3.70, p<0.01] (Figure 1). Participants’ language backgrounds influenced their segmentation performance [F(2,95)=3.54, p<0.01]. Simultaneous French-English bilinguals who learned both languages since birth (n=32) outperformed English monolinguals (n=33; p=0.03); however, sequential French-English bilinguals who learned their second language after age 1 (n=34) did not (p=0.39).

Overall, infants differentiated between part-words (word boundaries) and words in the two languages [t(67)=2.38, p=0.02] (Figure 2). English monolinguals (n=34) and French-English bilinguals (n=34) did not differ in terms of their ability to segment words [t(66)=1.36, p=0.18].

While both adults and infants can use language-specific phonetic cues produced by one speaker to segment words from two languages, early experience affects this skill. Early bilingualism (under age 1) leads to maintenance of this skill into adulthood, whereas monolinguals only succeed in infancy. We believe that both findings relate to infants’ refinement of native-language phonetic cues by 12 months, where they decrease sensitivity to non-native cues and increase sensitivity to native (Werker & Tees, 1984). Simultaneous bilinguals would therefore optimally refine these cues, whereas sequential bilinguals would not; and monolinguals of 9.5 months would still possess sensitivity to non-native phonetic variation, allowing them to distinguish the languages.

Authors

• Angeline Sin Mei Tsui, University of Ottawa
  Presenting Author

• Lucy Erickson, Carnegie Mellon University
  Non-Presenting Author

• Erik Theissen, Carnegie Mellon University
  Non-Presenting Author

• Christopher Fennell, University of Ottawa
  Non-Presenting Author