Doctoral Dissertation Research: The acquisition of phoneme categories

语言对于儿童来说是毫不费力的，只有学习第二语言的成年人才能一瞥学习语言所涉及的无数错综复杂的事情。范围从单词到意义的关联，到像语音分类这样看似简单的事情。为了更好地说明这一点，我们可以看看两个例子：说英语的人和说印地语的人在感知上的差异，以及说英语和日语的人在感知上的差异。说英语的人很难听出“pear”中的“p”和“spare”中的“p”之间的区别，尽管它们实际上发音完全不同。另一方面，说印地语的人会毫不费力地听出这种差异。再举一个例子，说英语的人可以很容易地告诉你“light”和“right”是两个不同的发音，而说日语的人却很难听清这两个词之间的区别。所有这些困难都是人类婴儿在第一次学习语言时非常有效的学习机制的副产品(S)。甚至在我们做出有意识的决定之前，人类的大脑就会无意识地、自动地将我们不断接收的大量信息分类成有序的数据。这项研究致力于确定印地语婴儿接收了哪些关键信息，从而(无意识地)确定这是两个不同的p音，而英国婴儿则确定它们是同一个音；或者是什么信息导致日本婴儿在习得日语时忽略了“L”和“r”之间的差异，但不会导致英国婴儿忽略相同的差异。这项研究通过解决人类如何学习类别来解决认知科学中广泛的理论问题，也导致了实际应用，例如更有效的第二语言学习技术和语音病理学的进步。以前的工作已经确定了至少两个人类在对声音进行分类时使用的主要线索：学习者不断增长的词汇量(“词汇线索”)和在语音流中发现的声学特性(“声学线索”)。词汇提示将出现在相似单词环境(“light”和“right”)中的发音组合在一起，组成一个发音类别。声学提示将声音映射到一些声学地图中，并在该地图上挑选出高频区域。这项研究利用了过去的观察结果，即成年人可以在实验室环境中无意识地模拟婴儿学习经历的各个方面，至少在某种程度上是这样。这项实验将通过向参与者呈现一种人工语言来测试成年人对这两种线索的依赖程度，在人工语言中，词汇线索和声音线索给参与者提供了相互矛盾的信息。通过这种方式，这项研究将有助于更好地理解人类大脑在对语音进行分类时做出的无意识决定。这项研究解决了语言学和认知科学中的两个基本问题：(1)人脑如何形成复杂的范畴；(2)有多少语言是通过专门的语言模块习得的，有多少语言可以归因于一般的认知机制。

Language comes effortlessly to children, and only adults learning a second language get a glimpse of the numerous intricacies involved in learning a language. These range from correlation of word to meaning, to something as deceptively simple as the categorization of speech sounds. To best illustrate this, we can look at two examples: differences in perception between English and Hindi speakers, and differences in perception between English and Japanese speakers. English speakers have trouble hearing the difference between the "p" in "pear" and the "p" in "spare," even though they are in reality pronounced quite differently. A speaker of Hindi on the other hand would have no trouble hearing this difference. To take another example, while English speakers could easily tell you that "light" and "right" begin with two different sounds, Japanese speakers find it very difficult to hear the difference between these two words. All of these difficulties arise as a byproduct of the human infant's very efficient learning mechanism when (s)he is first learning a language. The human brain unconsciously and automatically sorts the large amount of information we are constantly receiving into ordered data, even before we make conscious decisions. This study is concerned with determining what critical pieces of information were received by the Hindi infant to lead him or her to (unconsciously) determine that those were two different "p" sounds, while the English infant determined that they were the same sound; or what information led the Japanese infant to ignore the difference between "l" and "r" when acquiring Japanese, but did not lead the English infant to ignore the same difference. This study both contributes to broad theoretical issues within cognitive science by addressing how categories are learned by humans, and also leads to practical applications, such as more efficient second language learning techniques and advancements in speech pathology.Previous work has identified at least two main cues that humans use when categorizing sounds: the learner's growing vocabulary ("vocabulary cue") and acoustic properties found within a stream of speech ("acoustic cue"). The vocabulary cue groups together sounds occurring in similar word environments ("light" and "right") into a single sound category. The acoustic cue maps sounds into some acoustic map and picks out regions of high frequencies on this map. This study makes use of past observations that adults can unconsciously simulate, at least to some degree, aspects of the infant's learning experience within a laboratory setting. This experiment will test how much adults rely on each of these two cues by presenting participants with an artificial language where the vocabulary cue and the acoustic cue give the participant conflicting information. In this way, the research will contribute to a better understanding of what unconscious decisions the human brain makes when categorizing speech sounds. This study addresses two fundamental issues within linguistics and within cognitive science: (1) how the human brain forms complex categories, and (2) how much of language is acquired through specialized language-specific modules and how much of language can be attributed to general cognitive mechanisms.