Neural mechanisms underlying vocalization perception in realistic listening conditions

摘要 试听中的一个基本问题是，复杂的声音如何被准确识别，尽管存在着很大的差异 听音条件。准确和健壮的识别对于行为关键的声音尤其重要 例如语音或动物发声(叫声)。在说话或叫声的产生上有很大的个体差异， 以及由于混响和噪声造成的环境扭曲，导致声学上的相当大的可变性 当它们到达耳朵的时候。听觉系统的一个关键功能是将这些不同的信号组合在一起 分为功能类别，例如动物交流中的呼叫类型，或人类语音中的单词。在这 提案之后，我们将开始阐明实现呼叫不变分类的机制 在听觉皮质。在目标1中，我们将首先开发和验证一个神经回路如何实现呼叫的模型 分类，同时保持对生产变异性的不变。该模型对以下方面进行了具体预测 听觉皮层单个神经元和神经群的调谐特性。我们将检验这些预测 在清醒的豚鼠身上实验使用活体录音。初步数据表明，神经元在 初级听觉皮质(A1)的皮质浅层(L2/3)编码着构成其基础的“中层”特征 产生式不变调用分类。在目标2中，我们将检验这两个生产不变的假设 A1中同时出现了对特定呼叫类型的选择性以及对监听条件的高度不变性 L2/3。我们将分析丘脑A1层4中单个神经元以及神经元群体的反应 (L4)，和A1 L2/3呼叫刺激呈现在理想和系统退化的听条件下。初步 数据表明，A1 L2/3神经元，而不是丘脑神经元，能够在 广泛的收听条件。最后，在目标3中，我们将检验皮层抑制至关重要的假设 用于生成监听条件的不变性，而不是呼叫选择性。具体地说，我们将使用光遗传 用新的病毒载体表征抑制降低对选择性和 A1单个神经元和A1群体中的不变性。加在一起，理论模型和行为， 电生理和光遗传学数据将产生一个通用的计算原理，通过这个原理， 听觉皮质从噪声输入中提取有用的特征。从这些实验中获得的见解将提供一个 理解正常电路和电路功能障碍中的听觉处理的新框架 在沟通障碍方面，如自闭症、阅读障碍和特定语言障碍。

ABSTRACT A fundamental question in audition is how complex sounds are accurately recognized despite wide variations in listening conditions. Accurate and robust recognition is especially important for behaviorally critical sounds such as speech or animal vocalizations (calls). Large individual differences in the production of speech or calls, and environmental distortions due to reverberations and noise, result in considerable variability in the acoustics of calls when they arrive at the ear. A critical function of the auditory system is to group these diverse signals into functional categories, such as a call type in animal communication, or words in human speech. In this proposal, we will begin to elucidate the mechanisms by which invariant categorization of calls is accomplished in auditory cortex. In Aim 1, we will first develop and validate a model for how neural circuits might achieve call categorization while remaining invariant to production variability. This model makes specific predictions about the tuning properties of single neurons and neural populations in auditory cortex. We will test these predictions experimentally using in-vivo recordings in awake Guinea pigs. Preliminary data suggests that neurons in superficial cortical layers (L2/3) of primary auditory cortex (A1) encode ‘mid-level’ features that underlie production-invariant call categorization. In Aim 2, we will test the hypothesis that both production-invariant selectivity to specific call types, as well as a high degree of invariance to listening conditions co-emerge in A1 L2/3. We will analyze the responses of single neurons as well as neural populations in thalamus, A1 layer 4 (L4), and A1 L2/3 to call stimuli presented in ideal and systematically degraded listening conditions. Preliminary data suggest that A1 L2/3 neurons, but not thalamic neurons, are able to maintain their feature selectivity in a wide range of listening conditions. Finally, in Aim 3, we will test the hypothesis that cortical inhibition is crucial for generating invariance to listening conditions, but not call selectivity. Specifically, we will use optogenetic approaches with novel viral vectors to characterize the effects of decreased inhibition on selectivity and invariance in A1 single neurons and the A1 population. Together, the theoretical model, and behavioral, electrophysiological, and optogenetic data will give rise to a general computational principle by which the auditory cortex extracts useful features from noisy inputs. Insights gained from these experiments will provide a novel framework for understanding auditory processing in normal circuits, as well as during circuit dysfunction in communication disorders such as autism, dyslexia and specific language impairment.