Models optimized for real-world tasks reveal the task-dependent necessity of precise temporal coding in hearing

Area of Science:

• Neuroscience
• Computational Auditory Neuroscience
• Machine Learning in Auditory Perception

Background:

• Neurons encode information using both firing rate and precise spike timing.
• Auditory nerve action potentials exhibit sub-millisecond phase locking to sound, but its behavioral significance is unclear.
• Understanding the role of neural coding in auditory perception is a key challenge.

Purpose of the Study:

• To determine the behavioral relevance of auditory nerve spike timing precision in hearing.
• To assess the required temporal precision for reproducing human-like auditory behaviors.
• To investigate how different perceptual tasks utilize phase locking.

Main Methods:

• Optimized machine learning models using simulated cochlear input to perform real-world hearing tasks.
• Compared model performance with and without high-fidelity phase locking.
• Assessed the impact of varying temporal precision on task performance and human-likeness.

Main Results:

• Models with high-fidelity phase locking demonstrated more human-like sound localization and speech perception.
• The necessary temporal precision for human-like behavior varied across different auditory tasks.
• Task-specific benefits of precise spike timing were observed, suggesting differential incorporation of phase locking.

Conclusions:

• Auditory nerve spike timing precision plays an essential role in human auditory perception.
• The extent to which phase locking is utilized varies depending on the specific demands of auditory tasks.
• Optimizing computational models for realistic tasks is a valuable approach to elucidating the function of neural codes.