Predicting language under difficult conditions: Effects of cognitive load, noise, and hearing impairment

理解言语是非常复杂的，然而人们通常一听就能理解某人所说的话。他们之所以能做到这一点，部分原因是他们预测了演讲者接下来要说的话。例如，如果说话者说“我想出去放一架……”，那么听者可能会预测说话者的下一个单词可能是“风筝”。听众使用这些预测来“领先于游戏”，从而使理解更迅速、更有效。但我们不知道当听者无法做出这些预测时(因为他们的努力在其他地方)，例如当他们不得不努力集中注意力才能听到话语时，当他们专注于另一项任务时，或者当他们有听力障碍时。事实上，鉴于听力受损的人已经接受了退化的听觉信息，必须付出额外的努力才能理解他们所说的话，他们可能会特别受益于“保持警惕”的预测。在这个项目中，我们调查了在听语音时扰乱预测的因素，以更好地理解预测在有听力障碍和没有听力障碍的听者中是如何工作的。通过确定影响预测的困难类型，我们可以开始确定帮助解决这些困难和提高言语理解的方法，无论是对于听力正常的人来说，在挑战听力的情况下，还是对于听力受损的老年人来说都是如此。我们使用了一种被称为视觉世界范式的成熟技术，在这种技术中，参与者听句子，在屏幕上观看物体，同时记录他们的眼动。在以前的研究中，人们听到(例如)这样一句话：“我想穿……”一边看着领带、连衣裙、钻头和吹风机的照片。他们倾向于在说话者说出可穿戴物品(即领带或连衣裙)的名字之前先看照片，而且他们的反应非常迅速。我们称之为联想预测，因为它依赖于与“磨损”这个词的简单联系。然而，我们的工作表明，人们在预测时也会考虑他们对说话人的了解。例如，当他们听到男性说话者说“我想穿……”时，他们倾向于看领带的图片(一种刻板印象中的男性物体)，但当女性说话者说一件连衣裙时，他们倾向于看裙子的图片(一种刻板印象中的女性物体)。我们称之为战略预测，因为它要求听者另外从说话人的角度出发。这些策略预测特别有价值，因为它们更准确地估计了说话人可能会说什么，但它们速度慢且费力。我们的目标是调查阻碍简单联想预测和更复杂的策略预测的因素。在一系列实验中，参与者在听力困难(通过在背景噪音中听句子)和任务难度增加(通过在听的同时完成第二项任务)的条件下执行视觉世界范式。我们的一些实验使用了听力正常的年轻参与者(在爱丁堡大学进行)。其他实验使用了有听力障碍和没有听力障碍的老年参与者(在位于格拉斯哥的诺丁汉大学听力科学-苏格兰分部进行)。这些研究涉及以下问题：(1)不利的听力条件和增加的任务难度如何影响这两种类型的预测？(2)困难对有听力障碍的人和没有听力障碍的人的影响是相同的还是不同的？以及(3)听力损失的影响是与知觉困难、认知困难或两者兼而有之？回答这些问题将使我们找到克服这些困难的方法来提高言语理解能力，特别是对因听力损失而挣扎的人。

Understanding speech is extraordinarily complex, yet people can generally understand what somebody has said as soon as they hear it. They manage to do so, in part, because they predict what the speaker is going to say next. For example, if a speaker says "I would like to go outside to fly a...", then the listener may predict that the speaker's next word will likely be "kite". Listeners use these predictions to "get ahead of the game", making comprehension more rapid and efficient. But we do not know what happens when listeners are unable to make these predictions (because their efforts are elsewhere), such as when they have to concentrate hard to hear the utterance, when they are paying attention to another task, or when they have a hearing impairment. In fact, given that people with hearing impairment already receive degraded auditory information and have to apply extra effort to understand what is being said, they could particularly benefit from predictions to "stay on the front foot". In this project, we investigate the factors that disrupt prediction when listening to speech to better understand the how prediction works in listeners with and without hearing impairment. By identifying the types of difficulty that impair prediction, we can start to identify methods to help resolve these difficulties and improve speech understanding, both for people with normal hearing in challenging listening situations, and for older adults with hearing impairment. We use a well-established technique known as the visual-world paradigm, in which participants listen to sentences and view objects on a screen while their eye movements are recorded. In previous studies, people heard (for example) the sentence "I would like to wear..." while seeing pictures of a tie, a dress, a drill, and a hairdryer. They tended to look at pictures of wearable objects (i.e., the tie or dress) before the speaker actually named the object, and they did so very quickly. We call this associative prediction, because it is dependent on simple associations with the word "wear". However, our work has shown that people also take their knowledge of the speaker into account when predicting. For example, they tended to look at the picture of a tie (a stereotypically masculine object) when they heard a male speaker say "I would like to wear...", but the picture of a dress (a stereotypically feminine object) when it was said by a female speaker. We call this strategic prediction, because it requires the listener to additionally take the perspective of the speaker. These strategic predictions are particularly valuable because they more accurately estimate what the speaker is likely to say, but they are slow and effortful.Our goal is to investigate the factors that hinder both simple associative predictions, and more complex strategic predictions. In a series of experiments, participants perform the visual-world paradigm under conditions of hearing difficulty (by listening to sentences in background noise) and increased task difficulty (by completing a second task while listening). Some of our experiments use young participants with normal hearing (conducted at the University of Edinburgh). Other experiments use older participants with and without hearing impairment (conducted at Hearing Sciences - Scottish Section, part of the University of Nottingham located in Glasgow). The studies address the following questions: (1) How do adverse listening conditions and increased task difficulty affect the two types of prediction?; (2) Does difficulty affect people with and without hearing impairment in the same or different ways?; and (3) Are the effects of hearing loss comparable to perceptual difficulty, cognitive difficulty, or a combination of both?Answering these questions will allow us to identify methods for overcoming such difficulties to improve speech understanding, particularly for people who struggle because of hearing loss.