Learning mechanisms in language acquisition

描述(申请人提供)：近50年来对语言发展的研究揭示了年轻语言学习者的两种潜在的相互冲突的能力。一方面，婴儿对其输入的特定表面属性及其统计数据极其敏感。另一方面，婴儿能够基于高度抽象的形式属性快速地概括他们的输入。对表面和抽象语言属性的敏感性的相对重要性是关于人类语言本质的理论辩论的核心。 这里提出的研究反映了我们正在进行的尝试，即语言学习者如何将他们对语言表面属性的敏感性与他们对这些属性之外的抽象能力结合起来。根据本研究期间的研究，我们形成了以下三个假设：(1)记忆需求、任务需求和先前经验都会影响形式相同输入的抽象水平。(2)语言学习者只能形成最抽象的表征来解释他们所遇到的一组数据。(3)在某些情况下，轻度语言发育障碍的成年人比语言正常的成年人更不容易根据语言的抽象性质进行概括。 为了检验这些假设，参与者需要熟悉来自人工语言系统的刺激。然后，他们在新的刺激上进行测试，这些新的刺激与正在研究的熟悉刺激的特定性质一致或不一致。对一致和不一致测试项目的显著区分被认为是学习的证据。使用这一范式，研究比较了不同输入条件(两个形式上相同的语言系统，不同的表层实例化，符合或不符合自然语言特性的正式系统)和不同人群(6至12个月大的婴儿、语言正常的成年人、语言能力受损的成年人)之间的学习。这项研究的发现可能会揭示重要的信息，即婴儿和成年人在概括语言输入的过程中，如何受到语言输入细节的影响。这些发现也有可能为学习在语言障碍中的作用提供新的见解。

DESCRIPTION (provided by applicant): The last 50 years of research on language development have revealed in young language learners two, potential conflicting, abilities. On the one hand, infants are exquisitely sensitive to specific surface properties of their input and the statistics thereof. On the other hand, infants are able to rapidly generalize beyond their input based on highly abstract formal properties. The relative importance of sensitivity to surface versus abstract linguistic properties is at the heart of theoretical debates about the nature of human language. The research proposed here reflects our ongoing attempt to determine how language learners combine their sensitivity to surface properties of language and their ability to abstract beyond these properties. From the research performed during the current period of funding, we have formed the following three hypotheses: (1) Memory demands, task demands and previous experience all influence the level of abstraction achieved for formally identical input. (2) Language learners form only the most abstract representation necessary to account for the set of data they have encountered. (3) Under some circumstances, adults with subtle developmental language disabilities are less likely than adults with normal language to generalize based on abstract properties of language. To test these hypotheses, participants are familiarized with stimuli from an artificial language-like system. They are then tested on new stimuli that are either consistent or inconsistent with the particular properties of the familiarization stimuli under study. Significant discrimination of consistent versus inconsistent test items is taken as evidence of learning. Using this paradigm, the research compares learning across different input conditions (two formally identical language systems with different surface instantiations, formal systems that conform versus do not conform to properties of natural languages) and across populations (6- to -12-month-old infants, adults with normal language, adults with impaired language abilities). The findings from this research are likely to reveal important information about how infants and adults are influenced by the specifics of their language input as they generalize beyond it. The findings also have the potential to provide new insights into the role of learning in language disorders.