Insights into statistical learning from neural entrainment

学习一门新语言需要我们对许多不同类型的语言模式变得敏感。例如，说英语的人必须知道-s表示复数名词，并且某些声音的组合不能出现在单词的开头（例如，tl）。从环境输入中提取这些类型模式的能力被称为统计学习。虽然统计学习是一项核心的人类能力，有助于语言习得和认知的许多其他方面，但这种重要能力背后的神经机制很少被研究。大多数以前的研究都是在学习发生后才测试学习者对模式的知识，而没有检查学习过程中参与的机制。我的长期目标是通过描述支持从语言中提取模式的神经机制来发展和告知统计语言学习的理论。我的短期目标是了解语言模式的神经蕴涵如何反映和指导统计学习。神经夹带是指大脑节律与外界刺激之间的同步。我最近的研究表明，通过脑电图（EEG）测量，当学习者发现连续讲话的模式时，他们的大脑逐渐对潜在的模式表现出更大的兴趣。这一指标直接反映了学习者在接触语言过程中对语言规律的感知，因此是方法论上的一大进步。在这项工作的基础上，我和我的学员将开展四个相互关联的项目，以了解神经夹带如何有助于统计学习。在项目1中，我们将测试神经携动的个体间差异，如评估随机刺激序列，是否预测统计学习能力。在项目2中，我们将通过评估睡眠中的学习来测试语言模式的神经携动是否至少部分地自动出现。在项目3中，我们将评估先验知识对神经夹带的影响。在项目4中，我们将使用感觉刺激方法和非侵入性脑刺激，通过外部操纵不同频率的神经夹带，测试结构的神经夹带是否对统计学习有因果关系。提出的研究将显著推进我们对在线、动态神经机制驱动统计学习的理解。由于统计学习在语言以外的认知的许多方面起着重要作用，因此该研究项目将影响多个领域。这项研究对提高语言习得，特别是对成人第二语言学习者，也有令人兴奋的启示。这些发现将有助于确定统计学习的最佳条件，揭示以前被忽视的学习机会（例如，在睡眠期间），并产生促进统计学习的新方法。

Learning a new language requires us to become sensitive to many different types of linguistic patterns. For example, English speakers must learn that -s denotes a plural noun and that certain sound combinations cannot occur at the beginnings of words (e.g., tl). The ability to extract these types of patterns from environmental input is known as statistical learning. Although statistical learning is a core human ability that contributes to language acquisition and many other aspects of cognition, the neural mechanisms underlying this important ability have rarely been investigated. Most previous studies have tested learners' knowledge of patterns only after learning has occurred, and have not examined the mechanisms engaged during the learning process. My long-term goal is to develop and inform theories of statistical language learning by characterizing the neural mechanisms that support the extraction of patterns from language. My short-term goal is to understand how neural entrainment to linguistic patterns may both reflect and guide statistical learning. Neural entrainment refers to the synchrony between brain rhythms and external stimuli. My recent work shows that as learners discover patterns in continuous speech, their brains gradually show greater entrainment to the underlying patterns, as measured through electroencephalography (EEG). This measure directly indexes learners' perceptions of linguistic regularities during exposure to language, and thus represents a major methodological step forward. Building on this work, my trainees and I will carry out four interrelated projects to understand how neural entrainment may contribute to statistical learning. In Project 1, we will test whether interindividual differences in neural entrainment, as assessed to random stimulus sequences, predict statistical learning ability. In Project 2, we will test whether neural entrainment to linguistic patterns can emerge at least partially automatically by assessing learning during sleep. In Project 3, we will assess effects of prior knowledge on neural entrainment. In Project 4, we will test whether neural entrainment to structure causally contributes to statistical learning by externally manipulating neural entrainment at different frequencies, using both sensory stimulation methods and noninvasive brain stimulation. The proposed research will significantly advance our understanding of the online, dynamic neural mechanisms that drive statistical learning. Because statistical learning plays a fundamental role in many aspects of cognition beyond language, this research program will impact multiple fields. This research also has exciting implications for improving language acquisition, particularly for adult second language learners. These findings will help identify the optimal conditions for statistical learning, uncover previously overlooked learning opportunities (e.g., during sleep) and produce novel methods of boosting statistical learning.