Split listening: A cognitive investigation of speech perception in adverse conditions

“认知听力”是指研究记忆和注意力过程如何在具有挑战性的条件下（例如，噪音、分心）。良好的认知能力是成功倾听的基础，这一说法在老年人和/或听力受损的听众中得到了广泛的证实。在这些群体中，高工作记忆（WM）容量与噪声中语音任务的更好表现相关。然而，对于听力正常的年轻人来说，认知和听力水平之间的联系尚不清楚。这有四个原因：（1）年轻，听力正常的成年人可能不会产生足够的认知测量的个体差异，以检测可靠的关联与听力表现。因此，大多数实验都动力不足。因此，认知听力研究需要更大的样本。(2)大多数研究都集中在涉及能量掩蔽（EM）的不利条件上，即信号降级的条件（例如，背景噪声）。有证据表明，认知对本质上“数据有限”的任务几乎没有帮助。因此，当挑战来自非EM源时，也应该研究认知倾听。(3)语义语境可以减轻不利听力条件的影响，从而使认知支持变得多余或难以检测。因此，在对认知听力进行建模时，应考虑语义上下文。(4)听力表现（例如，句子中被正确识别的单词数量）可能不是认知功能如何支持听力的最敏感的衡量标准，因为给定的表现水平可能隐藏个体认知努力的有意义的差异。因此，在对认知听力进行建模时，应包括对心理努力的测量。该项目的目的是系统地解决上述挑战，并提供听力正常成人听力和认知过程之间的相互作用的全面了解。该项目将包括大规模在线实验和有针对性的实验室实验。在线实验将检测个体工作记忆差异对听力表现的任何调节作用，并将建立工作记忆有助于言语理解和感知努力的条件。然后，这些条件将在实验室中使用听力努力的生理测量，任务诱发的瞳孔反应，一种公认的“此刻”努力的标记（即，听的时候）。在线和实验室实验都将使用我们新开发的“分裂听”范式。在这个范例中，参与者必须跟踪两个同时出现的声音。两个声音的相对空间位置以四个增量步骤来操纵，从并置，即，从听音者的正前方（双耳分音）到完全偏侧（双耳分音）。空间操纵提供了一种使EM对抗注意力驱动（即，认知）需求，双耳分听中具有高EM和低注意力需求，而双耳分听中具有低EM和高注意力需求。我们实验室的试点数据已经证实了这种范式量化能量和认知过程之间竞争动态的能力。试点数据还表明，这一模式适合大规模在线测试。其EM和注意力需求的参数操纵使其成为瞳孔测量评估的理想候选者。作为一个整体，该项目提供了一个程序化的方法来努力倾听的认知和生理构建块。在线实验中收集的关于个体差异的数据在阐明传统听力测试之外对健康听力的认知贡献方面将是有价值的，并且在这样做的过程中，为临床人群中的靶向干预铺平了道路（例如，患有听力损失和/或痴呆症的人）。

"Cognitive Listening" refers to the study of how memory and attention processes contribute to speech perception in challenging conditions (e.g., noise, distraction). The claim that good cognitive abilities underpin successful listening is broadly confirmed for older and/or hearing-impaired listeners. In those groups, high working memory (WM) capacity is associated with better performance in speech-in-noise tasks. For young adults with normal hearing, however, the link between cognition and listening proficiency is unclear. There are four reasons for this: (1) Young, normal-hearing adults might not generate sufficient individual variability on cognitive measures to detect reliable associations with listening performance. A majority of experiments are, thus, under-powered. Cognitive listening studies therefore need to be run with larger samples.(2) Most studies have focused on adverse conditions involving energetic masking (EM), that is, conditions in which the signal is degraded (e.g., background noise). Evidence suggests that cognition is of little help for tasks that are intrinsically "data-limited". Cognitive listening should therefore also be investigated when challenges arise from a non-EM source.(3) Semantic context can mitigate the effects of adverse listening conditions, hence making cognitive support redundant or difficult to detect. Semantic context should therefore be taken into account when modelling cognitive listening.(4) Listening performance (e.g., number of words in a sentence that are correctly recognised) may not be the most sensitive measure of how cognitive functions support listening because a given performance level can hide meaningful differences in individual cognitive effort. A measure of mental effort should therefore be included when modelling cognitive listening. The aim of this project is to systematically address the above challenges and provide a comprehensive understanding of the interaction between hearing and cognitive processes in adults with normal hearing. The project will include large-scale online experiments and targeted in-lab experiments. The online experiments will be powered to detect any modulating effects of individual WM differences on listening performance and will establish the conditions under which WM contributes to speech understanding and perceived effort. These conditions will then be inspected more closely in the lab using a physiological measure of listening effort, the task-evoked pupil response, a recognised marker of effort 'in the moment' (i.e., during listening). Both online and in-lab experiments will use our newly developed "split-listening" paradigm. In this paradigm, participants have to track two simultaneous voices. The relative spatial location of the two voices is manipulated in four incremental steps, from collocated, i.e., straight ahead of the listener (diotic), to fully lateralised (dichotic). The spatial manipulation provides a powerful way of pitting EM against attention-driven (i.e., cognitive) demands, with high EM and low attention demands in diotic listening vs low EM and high attention demands in dichotic listening. Pilot data from our lab have confirmed the ability of this paradigm to quantify the competing dynamics between energetic and cognitive processes. The pilot data have also shown that this paradigm is amenable to online testing on a large scale. Its parametric manipulation of EM and attention demands makes it an ideal candidate for pupillometric assessment. As a whole, the project offers a programmatic approach to the cognitive and physiological building blocks of effortful listening. The data on individual differences gathered in the online experiments will be valuable in elucidating the cognitive contributors to healthy hearing beyond the traditional hearing test and, in doing so, pave the way for targeted interventions in clinical populations (e.g., individuals with a hearing loss and/or dementia).