Transfer of statistical learning from perception to production

无论是在咖啡馆与朋友聊天，在教室讨论一个话题，还是与亲戚辩论政治，我们都依赖于与许多不同的人交流的能力。在75亿英语使用者中，只有5%的人以英语为母语，而且不同地区的英语也存在巨大的方言差异。因此，我们经常会遇到一个说话的人，他的讲话和我们自己的不同。先前的研究表明，大脑如何协调听和说的联系可以使对话伙伴听起来更像对方。一个人听到的东西对他说话方式的影响通常被称为“转移”。迁移提供了一个窗口，通过它可以更好地理解语音感知和语音产生之间的机制联系。这一点很重要，因为它为在各个领域产生更广泛的影响奠定了基础，例如：1)在语言多样性的教室中开发有效教育方法的新技术；2)帮助有口吃等沟通障碍的人；3)为完善人工智能和机器语音识别系统提供约束；工程大脑假肢可以恢复因中风而丧失的语言能力。目前的项目重点是了解大脑是如何将我们听到的和我们说话的方式联系起来的，即使变化是微妙的，无法被意识到。研究人员采用一种新颖的方法，通过微妙但系统的方式操纵语言流的构成来研究迁移。这种方法允许对生产的预期变化进行清晰的预测，如果它隐含地受到感知的影响。初步数据表明，情况确实如此：通过控制声音的音高来制造微妙的口音，会导致说话者产生相同音高的强烈变化。这种影响不仅仅是立即模仿听到的话语（例如，重复刚刚听到的单词）。相反，这意味着说话人的生产系统发生了更根本的变化，这可能涉及到一种通过对大脑进行小而持久的神经可塑性改变来进行增量学习的形式。目前的项目将系统地调查何时以及如何发生转移，以揭示负责任的机制。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

Whether chatting with a friend at a café, discussing a topic in a classroom, or debating politics with a relative, we depend on our ability to communicate with many different people. Only 5 percent of the 7.5 billion English speakers are native talkers and there are also tremendous dialect differences in regional English. Thus, we quite often encounter a speaker whose speech differs from our own. Prior research demonstrates that links in how the brain coordinates listening and speaking can lead conversation partners to sound more like one another. The influence of what one hears on how one talks is often referred to as “transfer.” Transfer provides a window through which to better understand the mechanistic links between speech perception and speech production. This is important because it provides a foundation for broader impacts across a variety of domains, such as: 1) developing new technologies for effective education approaches in classrooms with language diversity; 2) assisting individuals with communication challenges like stuttering; 3) providing constraints to refine artificial intelligence and machine speech recognition systems; and 4) engineering brain prostheses that can restore speech lost to stroke.The current project focuses on understanding how the brain links what we hear with how we speak, even when changes are subtle and not consciously identifiable. The investigators take a novel approach to the study of transfer by manipulating the make-up of the speech stream in subtle but systematic ways. This method allows clear predictions about the expected changes to production if it is implicitly influenced by perception. Preliminary data suggest this is the case: creating a subtle accent by manipulating the pitch of a voice induces robust changes in how speakers produce the very same dimension. This influence goes beyond immediate imitation of heard speech (for example, repeating a just-heard word). Rather, it implies a more fundamental change to the speaker’s production system that likely involves a form of incremental learning through small, but persistent, neuroplastic changes to the brain. The current project will systematically investigate when and how transfer occurs in order to reveal the responsible mechanisms.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.