Decoding continuous speech in infancy

想象一下，一位母亲和她6个月大的孩子在看一只鸟啄人行道上的面包屑。妈妈说：看那只鸽子！那是一只鸽子。看到鸽子了吗？”后来，孩子和她爸爸在公园里吃午饭。一只鸟俯冲下来，落在他们旁边。爸爸开玩笑说：“我应该给鸽子一点我的泡菜吗？”在这个假设的例子中，孩子学到了什么？她能理解多少？她现在知道“鸽子”这个词了吗？这些问题看似简单，其实不然。正如我将在下面解释的那样，这个例子说明了我的研究计划旨在解决的一些关键理论问题。首先，语音中没有关于单词边界的可靠线索。成年人把语言理解为一串离散的单词，只是因为他们知道单词的典型发音。通过听一门不熟悉的语言，你可以体会到缺乏对单词边界的物理暗示，单词边界混在一起，模糊了一个单词的结束和下一个单词的开始。因此，在我们问这个例子中的婴儿是否理解“鸽子”之前，我们必须首先问他们是否从父母的话语中提取了这个词。过去的研究表明，6个月大的婴儿还没有做到这一点，但我假设，当单词反复与话语边缘对齐时，婴儿确实会将单词分割出来（即边缘假设）。我们假设的妈妈做了妈妈们在称呼婴儿时经常做的事情，鸽子是反复定位的话语。假设婴儿已经区分出了“pigeon”这个词，那么下一个问题就是当爸爸说“pigeon”这个词时，她是否能听懂。6个月大的婴儿可以理解一些单词，所以可以想象，当妈妈说话时，婴儿可能会理解鸽子（特别是如果妈妈经常给鸽子贴上标签）。但是，当爸爸说这个词时，孩子是否能认出并理解这个词完全是另一个问题。首先，父亲把话语这个词放在中间，使其难以识别。此外，根据分布模型，父亲也会发出音节pih‘ in pickle’和‘ gin’ in smidgen‘，这一事实可能会导致孩子假设音节pih’和‘ gin’都是一个单词。最后，过去的研究表明，婴儿很难识别跨性别单词（即，用女性声音学的单词在另一个女性声音中被识别出来，但很难从男性声音中识别出来）。如果妈妈和爸爸说不同的英语变体（例如，妈妈说加拿大英语，爸爸说英国英语），这可能会使事情进一步复杂化。然而，尽管有这些挑战，婴儿还是很容易学会语言。在我的研究中，我问：1)婴儿如何以及何时从言语中分词？2)婴儿是如何以及何时学会应对言语变异性的？3)接触多种英语变体的影响是什么？这些问题的答案将促进我们对婴儿发育的理论理解，并为语言病理学的更多应用进展奠定基础。

Imagine a mom and her 6-month-old watching a bird peck at crumbs on the sidewalk. The mom says Look at the pigeon! That is a pigeon. See the pigeon?' Later, the child is having lunch with her dad in the park. A bird swoops down and lands next to them. Dad jokes Should I give the pigeon a smidgen of my pickle?' In this hypothetical example, what has the child learned? How much does she understand? Does she now know the word pigeon? These questions seem simple, but they are not. As I will explain below, this example illustrates some key theoretical issues my research program is designed to address.First, speech contains no reliable cues to word boundaries. Adults perceive speech as a string of discrete words only because they know how words typically sound. The lack of physical cues to word boundaries can be appreciated by listening to an unfamiliar language word boundaries blend together, obscuring where one word ends and the next begins. Thus, before we can ask whether the infant in this example understands pigeon', we must first ask whether they have extracted the word from their parents' utterances. Past work suggests that 6-month-olds do not yet do this, but I hypothesize that infants do segment out words when they are repeatedly aligned with an utterance edge (i.e. the Edge Hypothesis). Our hypothetical mom has done what moms often do when addressing their infants pigeon' is repeatedly positioned utterance-finally. Assuming the infant has segmented out pigeon', the next question is whether she will understand pigeon' when spoken by dad. 6-month-olds understand some words, so it is conceivable that the infant may understand pigeon when spoken by mom (especially if mom labels pigeons regularly). But whether or not the child will recognize and understand this word when spoken by dad is another question altogether. First, the dad places the word utterance-medially, making it hard to identify. Moreover, according to distributional models, the fact that dad also produces the syllable pih' in pickle' and gin' in smidgen' may lead the child to hypothesize that the syllables pih' and gin' are each a word. And finally, past research has shown that infants struggle with cross-gender word identification (i.e., a word learned in a female voice is recognized in another female voice, but hard to recognize in a male voice). And if mom and dad speak different variants of English (e.g., mom speaks Canadian English, dad speaks British English), this could complicate matters further. And yet, infants readily learn language despite these challenges.In my research, I ask: 1) How and when do infants segment words from speech? 2) How and when do infants learn to cope with speech variability? 3) What is the impact of being exposed to more than one variant of English? Answers to these questions will advance our theoretical understanding of infant development, and lay the groundwork for more applied advances in speech-language pathology.