School of Psychology, University of Ottawa, Ottawa, ON, Canada.
Department of Psychology, Stanford University, Stanford, CA, USA.
Dev Sci. 2021 May;24(3):e13050. doi: 10.1111/desc.13050. Epub 2020 Nov 23.
Infants are sensitive to syllable co-occurrence probabilities when segmenting words from fluent speech. However, segmenting two languages overlapping at the syllabic level is challenging because the statistical cues across the languages are incongruent. Successful segmentation, thus, relies on infants' ability to separate language inputs and track the statistics of each language. Here, we report three experiments investigating how infants statistically segment words from two overlapping languages in a simulated language-mixing bilingual environment. In the first two experiments, we investigated whether 9.5-month-olds can use French and English phonetic markers to segment words from two overlapping artificial languages produced by one individual. After showing that infants could segment the languages when the languages were presented in isolation (Experiment 1), we presented infants with two interleaved languages differing in phonetic cues (Experiment 2). Both monolingual and bilingual infants successfully segmented words from one of the two languages-the language heard last during familiarization. In Experiment 3, a conceptual replication, we replicated the findings of Experiment 2 with a different population and with different cues. As before, when 12-month-old monolingual infants heard two interleaved languages differing in English and Finnish phonetic cues, they learned only the last language heard during familiarization. Together, our findings suggest that segmenting words in a language-mixing environment is challenging, but infants possess a nascent ability to recruit phonetic cues to segment words from one of two overlapping languages in a bilingual-like environment. A video abstract of this article can be viewed at https://www.youtube.com/watch?v=92pNcpxZguw.
婴儿在从流畅的言语中分割单词时对音节共现概率敏感。然而，在音节层面上分割两种重叠的语言具有挑战性，因为两种语言的统计线索不一致。因此，成功的分割依赖于婴儿分离语言输入并跟踪每种语言的统计数据的能力。在这里，我们报告了三项实验，研究了婴儿在模拟的语言混合双语环境中如何从两种重叠语言中统计分割单词。在前两项实验中，我们调查了 9.5 个月大的婴儿是否可以使用法语和英语语音标记从由一个人产生的两种重叠的人工语言中分割单词。在表明婴儿可以在语言孤立呈现时分割语言（实验 1）之后，我们向婴儿呈现了两种具有不同语音线索的交错语言（实验 2）。单语和双语婴儿都成功地从两种语言中的一种中分割单词——在熟悉过程中听到的最后一种语言。在实验 3 中，一个概念复制，我们使用不同的人群和不同的线索复制了实验 2 的发现。与之前一样，当 12 个月大的单语婴儿听到两种具有英语和芬兰语语音线索的交错语言时，他们只学习在熟悉过程中听到的最后一种语言。总之，我们的发现表明，在语言混合环境中分割单词具有挑战性，但婴儿具有一种初生的能力，可以在双语样环境中利用语音线索从两种重叠语言中的一种分割单词。本文的视频摘要可以在 https://www.youtube.com/watch?v=92pNcpxZguw 观看。
