Interaction of neuronal filters involved in perceptual decisions of grasshoppers

一个相对较小的神经系统如何处理和识别声音信号的快速变化，以及它如何在不可靠的神经元组件中获得高的整体可靠性？这些问题将使用蚱蜢作为模型系统的声学通信进行调查。在这些动物中，特定物种的通信信号（歌曲）用于启动性伴侣的会议和繁殖。对于雌性蚱蜢来说，它们是否应该对潜在伴侣的歌声信号做出反应的决定尤为重要，因为它们对大卵的投入很大。由于蚱蜢的听觉通路相对简单，因此它提供了了解这些高度相关的感觉信号是如何在小神经系统中提取和表示的机会。与最优编码理论雅阁，我们发现在蝗虫听觉通路的早期，不同神经元的反应变得越来越去相关和稀疏。这类似于脊椎动物感觉系统高级阶段的情况。听觉神经元的可管理数量使我们能够更好地理解引导信号表示中的这种转换的神经元操作。通信信号的时间和种群稀疏表示可能允许下游神经元更有效地读出信息，并且可以促进行为决策。我们开始调查的决定，发生在这些处理步骤的顶部通过线性非线性（LN）模型的方法。仅使用两个特征检测器，该LN模型产生了很高的预测能力，解释了被调查物种行为数据的~ 90%的方差。此外，该模型解释了迄今为止对行为和神经生理学实验的几个神秘观察。该提案的一个主要部分将是寻找从LN模型得出的特定预测的神经元实现。此外，我们将这种建模方法扩展到其他刺激功能，特别是其他蝗虫物种。密切相关的物种往往在它们的歌声模式和信号接收器的相应偏好方面存在很大差异。这些测试旨在揭示模型的普适性以及潜在的限制。通过分析不同物种的听觉处理，我们还希望获得洞察的进化约束的歌曲进化和物种形成事件。

How can a relatively small nervous system process and discern the fast changes which are characteristic for acoustic signals, and how can it attain a high overall reliability in spite of its unreliable components, the neurons? These questions will be investigated using the acoustic communication of grasshoppers as a model system. In these animals, species-specific communication signals (songs) serve to initiate the meeting of sexual partners and reproduction. The decision of whether or not they should respond to a song signal of a potential partner is particularly important for females, due to their high investment into large eggs.Since the auditory pathway of grasshoppers is relatively simply organized it offers the opportunity to understand how information about these highly relevant sensory signals is extracted and represented in a small nervous system. In accord with optimal coding theory we find that responses of different neurons become increasingly decorrelated and sparse early in the auditory pathway of grasshoppers. This resembles the situation at higher stages of sensory systems of vertebrates. The manageable number of auditory neurons allows us to better understand the neuronal operations that guide such transformations in signal representation. A temporally and population-sparse representation of the communication signals probably allows for a more efficient information read-out by downstream neurons and may facilitate behavioral decisions. We began to investigate the decisions that occur on the top of these processing steps by means of a linear-nonlinear (LN)-model approach. Using only two feature detectors, this LN-model yielded a high predictive power, explaining ~ 90% of the variance in behavioral data of the investigated species. In addition, the model explained several so far enigmatic observations of behavioral as well as of neurophysiological experiments. A main part of the proposal will be to search for the neuronal implementations of the specific predictions derived from the LN-model. Furthermore, we will extend this modeling approach to additional stimulus features, and in particular also to other grasshopper species. Closely related species often differ considerably in their song patterns and the corresponding preferences in the signal receivers. These tests aim at revealing the models universality as well as potential constraints. By analyzing the auditory processing in different species we also expect to obtain insights into evolutionary constraints of song evolution and speciation events.