How much input is required for infant language learning: Exploring the distribution of types and tokens

婴儿和儿童非常擅长学习语言，实际上比他们的父母熟练得多。直到最近，两种理论还在相互竞争地解释幼儿的语言学习技能。在一种观点中，婴儿和儿童就像海绵，吸收了他们环境的每一个细节，包括语言环境。在一种相反的观点中，婴儿出生时已经具备了人类语言(或语法)的范围，这些语言(或语法)已经“内置”了。他们的语言环境只允许他们选择他们母语的特定语法，例如将使用什么语音，动词是出现在句子的中间还是结尾等。最近，第三种观点开始得到支持。这一观点与第一种观点相同，即婴儿和儿童对环境的每一个细节都高度敏感。它赞同第二种观点的立场，即婴儿和儿童以决定其社区的特定语言系统或语法为目标。区分这三种理论的一种方法是最佳语言学习所需的输入数量和种类。第一种“海绵”观点(或联想理论)表明，需要大量的输入，并且无论该输入是由相同示例(类型；例如，多次将狗作为狗的复数来听)还是不同的例子(例如，将狗、猪和马克杯作为它们各自的单数的复数)的许多重复(记号)组成的，这并不重要。第二种“内建语法”观点(或先天领域理论)表明，只需要极少的输入，即使只重复一个例子，对婴儿或儿童来说也足以学习母语的特定组成部分。第三种“寻找语法的海绵”观点(或假设选择理论)认为，学习只需要极少量(例如三个)不同的例子，不同的例子比重复一个例子更有帮助。这个项目将通过询问语言学习所需的输入量来比较这三种理论。它还试图将第三个理论嵌入到一个更现实的人类学习者模型中，通过询问学习是否总是随着例子数量的增加而改善，或者在例子的数量和每个例子的重复次数之间是否存在权衡。这项研究将通过询问人类是否是他们的环境、他们的生物学的反映，或者大脑是否是以及如何是生物大脑的产物，其主要目标是对环境进行“理解”，从而为我们的科学知识做出贡献。此外，由于这项研究借鉴了已有的关于婴儿视觉和婴儿社交的研究，它将语言发展的研究置于更广泛的发展背景下。最后，研究探索了能最大限度地促进学习的特定类型的输入。因此，它有可能最终告诉我们如何向婴儿、学童甚至成年人提供语言和其他信息，以便在尽可能短的时间内实现最佳学习。

Infants and children are remarkably adept at learning language, much more adept in fact than their parents. Until recently, two theories competed to explain the language learning skill of young children. In one view, infants and children are like sponges, soaking up every detail of their environment, including the linguistic environment. In a competing view, infants come into the world with the range of human languages (or grammars) already "built-in." Their linguistic environment serves only to allow them to select the particular grammar of their native language, such as what speech sounds will be used, whether verbs come in the middle or the end of sentences, etc. Recently, a third view has begun gaining support. This view shares the position of the first view that infants and children are highly sensitive to every detail of their environment. It shares the position of the second view that infants and children have as their goal determining the particular linguistic system or grammar of their community. One way to discriminate among these three theories is the amount and kind of input required for optimal language learning. The first "sponge" view (or Associative theories) suggests that a great deal of input is needed and that it does not matter much whether that input consists of many repetitions (tokens) of the same example (type; e.g., hearing the word dogs as the plural of dog multiple times) or different examples (e.g., hearing dogs, pigs, and mugs as the plurals of their respective singulars). The second "built-in grammars" view (or Innate Domain theories) suggests that extremely little input is needed and that even a single repetition of a single example is sufficient for the infant or child to learn a particular component of the native language. The third "grammar-seeking sponge" view (or Hypothesis Selection theories) suggests that a very small number (e.g., three) of different examples are needed for learning and that different examples are much more helpful than repetitions of an example. This project will compare these three theories by asking about the amount of input needed for language learning. It also attempts to embed the third theory in a more realistic model of human learners by asking if learning always improves as the number of examples increases, or whether there is a trade-off between the number of examples and the number of repetitions of each example.This research will contribute to our scientific knowledge by asking whether humans are reflections of their environment, their biology, or whether and how the mind is a product of a biological brain with a primary goal of "making sense" of the environment. Moreover, because the research draws on existing studies of infant vision and infant social interaction, it places the study of language development in a broader developmental context. Finally, the research explores the specific kinds of input that maximize learning. Therefore, it has the potential to ultimately inform how we present linguistic and other information to infants, school children, and even adults in order to result in the best possible learning in the shortest possible time.