Statistical learning of multiple patterns in infants, adults, and monkeys

婴儿、成人和猴子多种模式的统计学习

• 批准号: 8116119
• 负责人: DANIEL J WEISS
• 金额: $ 22.16万
• 依托单位: PENNSYLVANIA STATE UNIVERSITY, THE
• 依托单位国家: 美国
• 财政年份: 2011
• 资助国家: 美国
• 起止时间: 2011-04-01 至 2016-03-31
• 项目状态: 已结题
• 来源: https://reporter.nih.gov/project-details/8116119
• 关键词: Adult; American; Applications Grants; Benchmarking; British; Cues; Data; Development; Disease; Economics; Elements; Environment; Experimental Designs; Exposure to; Eye; Goals; Head; Head Movements; Hour; Human; Human Development; Infant; Judgment; Language; Language Delays; Language Development; Language Disorders; Lead; Learning; Linguistics; Location; Machine Learning; Measures; Methods; Monkeys; Parents; Pattern; Performance; Phonetics; Procedures; Recovery; Research; Rivers; Role; Saguinus; Semantics; Signal Transduction; Speech; Stimulus; Stream; Structure; Sum; System; Testing; Time; Variant; Voice; Work; auditory stimulus; bilingualism; design; lexical; man; nonhuman primate; novel; phonology; preference; research study; response; theories

DESCRIPTION (provided by applicant): The overall goal of the present grant application is to understand how a naive learner collects distributional information from the environment and makes an implicit decision that the corpus of input contains either a single structure or multiple structures. Mature learners are incredibly facile at interpreting information in a context-specific manner, thereby partitioning the input into two or more sub-structures. We will investigate this question of context-specific statistical learning by studying two types of naove learners - human infants and tamarin monkeys - as well as mature adults. The specific objective of the proposed research is to determine whether and how infants learn that there are multiple patterns of information embedded in streams of speech, or that there are multiple words that refer to the same object, and to determine whether context-specific statistical learning has species-specific biases. Two types of experimental designs will be used to study context-specific statistical learning. The first uses a single change in the underlying structure. A variety of contextual cues will be introduced to signal that the underlying structure has undergone a change, and the dependent measure is whether the learner has acquired the first, the second, both the first and the second, or neither structures. The second design uses two alternating structures that are signaled by a variety of stimulus cues to partition the two underlying structures. It is important to note that in both of these designs, if the learner aggregates the structural information across the entire corpus, rather than partitioning the corpus into two subsets, no learning is possible. Thus, these designs test the ability of the learner to extract the contextual cues that partition the input into subsets. The implications of the proposed studies are fundamental to any theory of learning, but particularly to the kind of implicit (passive exposure) statistical learning that is thought to characterize much of early human development in many domains. Infants must learn - by a combination of sensitivity to distributional patterns and innate biases - that patterns of information are context-specific, as in the case of bilingualism. Our proposed experiments will extend our recent studies of human adults by determining (a) whether infants show the same pattern of learning biases (primacy effects) and context-sensitivity (to talker voice), (b) whether tamarin monkeys show these same biases and context effects, and (c) what the limits of context-specific statistical learning are in human adults and infants in both word segmentation and referential tasks. PUBLIC HEALTH RELEVANCE: Language development is one of the hallmarks of the human species, yet it is difficult to study because of the huge variation in early exposure to different amounts of linguistic input. The use of artificial languages that are acquired in the lab over a few hours provides a window on the mechanisms of language development. We will study language learning in the lab to gain a unique perspective on how the infants and adults learn the patterns of words in streams of speech and contrast this with performance in nonhuman primates. These studies will not only reveal a basic mechanism of language learning, but also establish benchmarks against which language delay can be compared. Moreover, understanding the mechanisms that lead to successful acquisition in normal infants and adults can help to identify loci of language disorders and design methods for remediating disorders.

描述（由申请人提供）：本资助申请的总体目标是了解天真的学习者如何从环境中收集分布信息，并做出输入语料库包含单个结构或多个结构的隐式决定。成熟的学习者非常容易以特定于上下文的方式解释信息，从而将输入划分为两个或多个子结构。我们将通过研究两种类型的幼稚学习者（人类婴儿和狨猴）以及成熟的成年人来研究特定情境统计学习的问题。该研究的具体目标是确定婴儿是否以及如何学习语音流中嵌入的多种信息模式，或者有多个单词指代同一对象，并确定特定情境的统计学习是否具有物种特定的偏差。将使用两种类型的实验设计来研究特定背景的统计学习。第一个使用底层结构的单一更改。将引入各种上下文线索来表明基础结构已经发生了变化，并且依赖的度量是学习者是否已经掌握了第一个、第二个、第一个和第二个结构，或者两者都没有。第二种设计使用两个交替的结构，这些结构由各种刺激线索发出信号来划分两个底层结构。值得注意的是，在这两种设计中，如果学习者聚合整个语料库的结构信息，而不是将语料库划分为两个子集，则不可能进行学习。因此，这些设计测试学习者提取将输入划分为子集的上下文线索的能力。所提出的研究的意义对于任何学习理论都是基础，特别是对于隐式（被动暴露）统计学习，这种学习被认为是许多领域早期人类发展的特征。婴儿必须通过对分布模式的敏感性和先天偏见的结合来学习信息模式是特定于上下文的，就像双语的情况一样。我们提出的实验将通过确定（a）婴儿是否表现出相同的学习偏差（首因效应）和情境敏感性（对说话者的声音）模式，（b）狨猴是否表现出相同的偏差和情境效应，以及（c）在分词和参考任务中，成人和婴儿的情境特定统计学习的局限性，来扩展我们最近对成人的研究。 公共健康相关性：语言发展是人类的标志之一，但由于早期接触不同数量的语言输入存在巨大差异，因此研究起来很困难。使用在实验室几个小时内获得的人工语言为了解语言发展机制提供了一个窗口。我们将在实验室中研究语言学习，以获得关于婴儿和成人如何学习语音流中的单词模式的独特视角，并将其与非人类灵长类动物的表现进行对比。这些研究不仅将揭示语言学习的基本机制，还将建立比较语言延迟的基准。此外，了解正常婴儿和成人成功习得的机制有助于识别语言障碍的位点并设计治疗障碍的方法。