Doctoral dissertation research: Evoked Category Representations

人类通过将细粒度的声学细节映射到存储在长期记忆中的音素（用于区分单词的最小单位）来理解语音。关于这个过程的一个基本问题是音素的性质。在经典生成音系学中，音素是抽象和离散特征的结合，缺乏声学细节。从这个角度来看，语音感知是一个将梯度信息分类为非梯度类别的过程。然而，有证据表明，说话者在决定听什么时，对语音的细微和梯度声学特性很敏感，这表明音素本身可能包含了声学细节和它们的概率分布，可以通过说话者的经验来重塑。这种观点认为语音和音素之间有更直接的关系。目前的项目将进行测量大脑活动的实验，旨在测试这两种模型的预测，促进对语音感知这一核心机制的理解。实验方法利用事件相关脑电位，通过古怪范式测量听觉皮层的自动声音变化检测。一个实验将比较一个类别与来自同一类别的标记：在这里，变化检测反应将要求类别编码物理刺激属性。第二个实验将处理导致类别表示的标记的两个不同统计分布。如果古怪的反应是由产生“临时”类别的刺激中的统计信息调制的，那么大脑必须能够在类别表征中编码这些信息。如果没有观察到对声学或统计学的敏感性，那么一定是音素类别表示不包含梯度声学信息。这些发现可能会增加对人类认知中类别是如何形成的理解。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。

Humans understand speech by mapping fine-grained acoustic details to phonemes – the smallest units used to distinguish words – stored in long-term memory. A fundamental issue regarding this process is the nature of the phoneme. In classical generative phonology, a phoneme is a combination of abstract and discretized features, devoid of acoustic details. On this view, speech perception is a process of sorting gradient information into non-gradient categories. However, evidence has emerged that speakers are sensitive to minute and gradient acoustic properties of speech sounds when making decisions about what they hear, which suggests that the phoneme itself may encapsulate acoustic details and their probability distributions that can be reshaped by speakers’ experience. This view assumes a more direct relationship between the speech sound and the phoneme. The current project will conduct experiments measuring brain activity designed to test the predictions of both models, facilitating an understanding of this core mechanism in speech perception.The experimental methodology utilizes event-related brain potentials which measure automatic sound change detection in auditory cortex via oddball paradigms. One experiment will compare a category to a token from the same category: here, a change detection response will require that the category encodes physical stimulus properties. A second experiment will manipulate two different statistical distributions of the tokens that lead to a category representation. If the oddball response is modulated by the statistical information in the stimuli that gave rise to the "temporary" category, then the brain must be able to encode this information in category representations. If no sensitivity to acoustics or statistics is observed, then it must be the case that phoneme category representations do not contain gradient acoustic information. The findings can potentially increase the understanding of how categories are formed in human cognition.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.