Neuronal representation of sound source location in the auditory cortex during active navigation

Lopes Amaro, Diana Inês (2021): Neuronal representation of sound source location in the auditory cortex during active navigation. Dissertation, LMU München: Graduate School of Systemic Neurosciences (GSN)

Abstract

The ability to localize sounds is crucial for the survival of both predators as well as prey. The former rely on their senses to lead them to the latter, which in turn also benefit from locating a predator in the vicinity to escape accordingly. In such cases, the sound localization process typically takes place while the animals are in motion. Since the cues that the brain uses to localize sounds are head-centered (egocentric), they can change very rapidly when an animal moves and rotates. This constitutes an even bigger challenge than sound localization in a static environment. Up to now, however, both aspects have mostly been studied separately in neuroscience, thus limiting our understanding of active sound localization during navigation. This thesis reports on the development of a novel behavioral paradigm – the Sensory Island Task (SIT) – to promote sound localization during unrestricted motion. By attributing a different behavioral meaning (associated to different outcomes) to two spatially separated sound sources, Mongolian gerbils (Meriones unguiculatus) were trained to forage for an area (target island) in the arena that triggered a change in the active sound source to the target loudspeaker and to report its detection by remaining within the island for a duration of 6 s. Importantly, the two loudspeakers played identical sounds and the location of the target island in the arena was changed randomly every trial. When the probability of successfully identifying the target island exceeded the chance level, a tetrode bundle was implanted in the primary auditory cortex of the gerbils to record neuronal responses during task performance. Canonically, the auditory cortex (AC) is described as possessing neurons with a broad hemispheric tuning. Nonetheless, context and behavioral state have been shown to modulate the neuronal responses in the AC. The experiments described in this thesis demonstrate the existence of a large variety of additional, previously unreported (or underreported) spatial tuning types. In particular, neurons that were sensitive to the midline and, most intriguingly, neurons that were sensitive to the task identity of the active loudspeaker were observed. The latter comprise neurons that were spatially tuned to only one of the two loudspeakers, neurons that exhibited a large difference in the preferred egocentric sound-source location for the two loudspeakers as well as spatially untuned neurons whose firing rate changed depending on the active loudspeaker. Additionally, temporal complexity in the neuronal responses was observed, with neurons changing their preferred egocentric sound-source location throughout their response to a sound. Corroborating earlier studies, also here it was found that the task-specific choice of the animal was reflected in the neuronal responses. Specifically, the neuronal firing rate decreased before the animal successfully finished a trial in comparison to situations in which the gerbil incorrectly left the target island before trial completion. Furthermore, the differential behavioral meaning between the two loudspeakers was found to be represented in the neuronal tuning acuity, with neurons being more sharply tuned to sounds coming from the target than from the background loudspeaker. Lastly, by implementing an artificial neural network, all of the observed phenomena could be studied in a common framework, enabling a better and more comprehensive understanding of the computational relevance of the diversity of observed neuronal responses. Strikingly, the algorithm was capable of predicting not only the egocentric sound-source location but also which sound source was active – both with high accuracy. Taken together, the results presented in this thesis suggest the existence of an interlaced coding of egocentric and allocentric information in the neurons of the primary auditory cortex. These novel findings thus contribute towards a better understanding of how sound sources are perceptually stable during self-motion, an effect that could be advantageous for selective hearing.

• Dokumententyp: Dissertationen (Dissertation, LMU München)
• Keywords: auditory cortex, sound localization, selective listening, sensory-based decision making, active navigation, chronic recording, auditory scene analysis, active sensing, auditory object
• Fakultäten: Graduate School of Systemic Neurosciences (GSN)
• Sprache der Hochschulschrift: Englisch
• Datum der mündlichen Prüfung: 19. März 2021
• 1. Bericht­erstatter:in: Pecka, Michael