Statistical Learning in Infant Language Acquisition

项目概要/摘要 第一语言习得是人类发展的一个标志。大量研究表明 婴儿在语言输入中发现统计错误的能力有助于语言学习。但 这篇文献的影响力受到了限制，因为它未能将统计学习与新兴的 研究和理论集中在婴儿的实时语言处理。目前的申请是有动机的 假设统计学上的差异有助于婴儿有效地编码和处理语言 输入.为此，婴儿对可能的下游输入产生预测。这些预测往往 不正确的，渲染预测错误-一个潜在的有效数据来源，为后续的学习。的输入 概率性的和/或先前导致预测错误的信息提供了信息丰富的数据， 随后的关注和学习。我们假设，统计学上的错误是一个重要的来源， 影响这一过程的信息，沿着其他语境线索， 非语言环境。迄今为止，没有先前的研究操纵婴儿期的统计学数据， 语言处理任务;研究是必要的，以裁定之间的可能关系 开发过程中输入和实时语言行为的统计分析。拟议 实验中，我们将测量婴儿在预测性语言过程中使用顺序统计的能力， 处理任务（目标1），评估统计学习对基于预测误差的学习的影响（目标2）， 并研究了由于统计结构和预测误差引起的不确定性在激励婴儿主动学习中的作用。 语言学习过程中的探索行为（目标3）。拟议研究的结果将促进 积极的发展成果，扩大我们对统计数据之间关系的理解， 语言输入的结构和在语言发展过程中展开的实时过程。在我们 以前的统计学习研究，这些研究的结果与典型的发展婴儿将 推动未来的研究，包括婴儿和幼儿的非典型语言发展的风险 轨迹

Project Summary/Abstract First language acquisition is a hallmark of typical human development. A substantial body of research suggests that infants’ ability to detect statistical regularities in language input facilitates language learning. However, the impact of this literature has been limited by its failure to connect statistical learning with the burgeoning body of research and theories focused on infants’ real-time language processing. The current application is motivated by the premise that statistical regularities facilitate infants’ attempts to efficiently encode and process language input. To this end, infants generate predictions about likely downstream input. These predictions are often incorrect, rendering prediction errors – a potentially potent source of data for subsequent learning. Input that is probabilistic and/or that has previously led to prediction errors provides information-rich data, guiding infants’ subsequent attention and learning. We hypothesize that statistical regularities are an important source of information influencing this process, along with the other contextual cues available in both the linguistic and nonlinguistic environment. To date, no prior studies have manipulated statistical regularities during infant language processing tasks; research is necessary to adjudicate amongst the possible relationships between statistical regularities in the input and real-time language behaviors during development. In the proposed experiments, we will measure infants’ use of sequential statistical regularities during predictive language processing tasks (Aim 1), assess the impact of statistical regularities on prediction error-based learning (Aim 2), and examine the role of uncertainty due to statistical structure and prediction error in motivating infants’ active exploratory behavior during language learning (Aim 3). The results of the proposed research will promote positive developmental outcomes by expanding our understanding of the relationship between the statistical structure of language input and real-time processes that are unfolding during language development. As in our previous statistical learning research, the outcomes of these studies with typically-developing infants will motivate future investigations that include infants and young children at risk for atypical language development trajectories.