Exploring the Processing of Objects, Settings, and Global Properties in Natural Auditory Scenes

Our auditory system has a remarkable ability to make sense of complex environments, allowing us to identify what we are hearing and where sounds come from. This process of auditory scene analysis is essential for situational awareness, decision-making, and effective communication in daily life. While much is known about how listeners segregate and identify individual sound sources (e.g., voices), less is understood about how global properties of a scene (e.g., openness, naturalness) contribute to perception. This dissertation examined how object- and setting-level information in natural auditory scenes are processed, and whether they rely on distinct or overlapping mechanisms. In Experiment 1, participants listened to 200 scenes of varying durations (1, 2, or 4 sec) and listed the setting (e.g., park) and objects (e.g., dog bark, wind, birds chirping) they heard in each scene. Overall, object identification was more accurate and benefited more from longer scene durations than setting identification. Different low-level (pitch, frequency) and mid-level (spectrotemporal patterns) acoustic features predicted performance across the two tasks, suggesting that distinct but potentially interacting mechanisms support object and setting perception in natural scenes. In Experiment 2, participants completed separate forced-choice object and setting identification tasks during electroencephalography recording. Although no significant differences in neural activity were observed between tasks, the spectral complexity of scenes modulated the P2 event-related potential, indicating that processes relevant to both object and setting identification share a sensitivity to acoustic features at a mid-level stage of processing. Together, these findings suggest that object and setting identification may rely on partially overlapping mechanisms. Understanding how the auditory system integrates both object- and scene-level information offers insights into real-world listening and can inform the development of more effective artificial intelligence systems and hearing assistive devices for navigating complex environments.