Infant statistical learning: Resilience, longevity, and specificity

婴儿统计学习：弹性、寿命和特异性

• 批准号: 9313700
• 负责人: Jessica S Hay
• 金额: $ 26.77万
• 依托单位: UNIVERSITY OF TENNESSEE KNOXVILLE
• 依托单位国家: 美国
• 财政年份: 2015
• 资助国家: 美国
• 起止时间: 2015-08-01 至 2020-07-31
• 项目状态: 已结题
• 来源: https://reporter.nih.gov/project-details/9313700
• 关键词: Address; Affect; Chiroptera; Cognitive; Complex; Development; Environment; Felis catus; Future; Gender; Health; Hour; Individual; Infant; Knowledge; Label; Language; Language Delays; Language Development; Learning; Longevity; Machine Learning; Memory; Nature; Neurologic; Noise; Outcome; Output; Pattern; Play; Population; Probability; Public Health; Research; Risk; Role; Sensory; Signal Transduction; Specificity; Speech; Stream; Testing; United States National Institutes of Health; Variant; clinical application; clinical practice; critical developmental period; design; evidence base; experience; experimental study; hearing impairment; indexing; lexical; long term memory; natural language; novel; public health relevance; resilience; sound; statistics; theories; tool; word learning

 DESCRIPTION (provided by applicant): Typically developing infants acquire language at a remarkable rate despite numerous perceptual and cognitive challenges. Infants may begin to learn language by tracking regularities in their environment. Specifically, research suggests that infants possess powerful computational mechanisms that may support the segmentation of words from fluent speech and facilitate word learning. The problem is that there is little research on the extent to which statistical regularities support early language acquisition under the challenging learning conditions often faced by young infants. The objective of the proposed research is to advance integrative and comprehensive theories of infant language acquisition by assessing how statistical learning supports (1) speech segmentation and word learning in background noise, (2) infants' ability to encode lexical representations in long-term memory, and (3) infants' abilities to represent newly segmented words with the appropriate level and type of detail to facilitate subsequent language learning. Three Aims will be addressed across nine experiments designed to test how statistical regularities found in natural language input support resilience, longevity, and representational specificity within a developmental framework. Infants will be familiarized with a short natural Italian language corpus and then tested on their ability o either discriminate words that have strong versus weak internal co-occurrence patterns (8- and 11-month-olds), or associate those words with novel objects (17-month-olds). Experiments are designed to tests how infants cope with simultaneous learning challenges. We will test the predictions that strong syllable co-occurrence patterns will bolster (1) speech segmentation and word learning in noise and (2) long-term memory for newly extracted words, and (3) that infants' word form representations will become more robust and specific. Results from the proposed project will advance our understanding of the learning mechanisms underlying normative language development. Individuals who are, for a variety of sensory, neurological, or developmental reasons, less adept at tracking and representing statistical regularities when faced with real-world learning challenges may be at greater risk for atypical language development. Results from the proposed research will be used to help generate and test hypotheses about the causal mechanisms for specific language delays in atypical populations, such as for infants with hearing loss or infants who, for various reasons, receive sub-optimal language input during critical developmental periods.

 描述（由申请人提供）：尽管存在许多感知和认知挑战，但通常发育中的婴儿以显著的速度获得语言。婴儿可能通过跟踪环境中的声音开始学习语言。具体来说，研究表明，婴儿拥有强大的计算机制，可以支持从流利的语音中分割单词，并促进单词学习。问题是几乎没有研究 在何种程度上，统计数字支持早期语言习得的挑战性学习条件下，往往面临的幼儿。本研究的目的是通过评估统计学习如何支持（1）背景噪声中的语音分割和单词学习，（2）婴儿在长期记忆中编码词汇表征的能力，以及（3）婴儿以适当的细节水平和类型来表征新分割的单词的能力，以促进随后的语言学习。三个目标将在九个实验中得到解决，这些实验旨在测试自然语言输入中发现的统计数据如何支持发展框架内的弹性，寿命和表征特异性。婴儿将熟悉一个简短的自然意大利语语料库，然后测试他们区分具有强与弱内部共现模式的单词的能力（8个月和11个月大），或者将这些单词与新物体联系起来的能力（17个月大）。实验旨在测试婴儿如何科普同时学习的挑战。我们将测试以下预测：强音节共现模式将支持（1）噪音中的语音分割和单词学习，（2）新提取单词的长期记忆，以及（3）婴儿的单词形式表征将变得更加强大和具体。从拟议的项目的结果将推进我们的理解规范的语言发展的学习机制。由于各种感官、神经或发育原因，在面对现实世界的学习挑战时，不太擅长跟踪和表示统计学上的错误的人可能会面临更大的非典型语言发展风险。拟议研究的结果将用于帮助生成和测试关于非典型人群中特定语言延迟的因果机制的假设，例如患有听力损失的婴儿或由于各种原因在关键发育期接受次优语言输入的婴儿。