Level-Dependent Latency Shifts Quantified Through Binaural Processing

Ida Siveke, Christian Leibold, Katharina Kaiser, Benedikt Grothe, Lutz WiegrebeJournal of Neurophysiology 2010 · DOI

TL;DR

This paper investigates how the timing of neural responses to sound changes with loudness, a phenomenon known as level-dependent latency shifts. They found that these shifts are reflected in our perception of sound location and in the neural processing within the auditory brainstem, suggesting that a level-dependent firing threshold is crucial for accurate binaural processing across different sound levels.

What Problem Does This Paper Address?

The ability to pinpoint the location of a sound in space, particularly its left-right position (azimuth), relies heavily on the brain's ability to precisely compare the arrival times of sound at our two ears. This comparison is incredibly sensitive, working with microsecond precision. However, a significant challenge arises because the timing of neural responses in the auditory system itself changes with the loudness (sound pressure level) of the incoming sound. This means that a sound arriving at a certain time might trigger a neural response with a different latency depending on whether it's loud or soft. The core problem this paper addresses is understanding how these level-dependent latency shifts in individual ears are processed and compensated for by the binaural system to maintain accurate sound localization across a wide range of sound levels. Previous models struggled to explain this level invariance, particularly at higher sound levels.

Methods

The researchers employed a multi-pronged approach combining psychophysics, electrophysiology, and computational modeling.

  • Psychophysics: Human listeners were presented with specially designed binaural stimuli. These stimuli were trains of high-frequency tone pips with asymmetric envelopes (different rise and fall times). Crucially, in one ear, the envelope was temporally reversed relative to the other. This "incongruent" stimulus couldn't be perfectly matched by simply shifting the entire sound in time. Listeners were asked to adjust the interaural time difference (ITD) until the perceived sound image was centered. This "adjusted eITD" revealed how their brains compensated for the envelope timing differences.
  • Electrophysiology: Similar stimuli were used to record neural responses from two key auditory brainstem nuclei in gerbils: the Lateral Superior Olive (LSO) and the Dorsal Nucleus of the Lateral Lemniscus (DNLL). The LSO was used to study monaural responses (inputs to the binaural system), while the DNLL was used to investigate binaural processing. They measured how the timing of neural spikes (specifically, phase locking to the sound envelope) changed with sound level and envelope asymmetry.
  • Modeling: They adapted an existing model of temporal integration and neural latency. The key modification was introducing a "level-dependent firing threshold," meaning the loudness of the sound influenced the threshold a neuron needed to cross to fire. This was compared against a model with a fixed firing threshold.

Datasets

Not discussed in the paper. The study used custom-generated auditory stimuli for both psychophysical and electrophysiological experiments.

Key Insights and Ideas

  • Perceptual Compensation is Level-Dependent: Our perception of sound location isn't perfectly invariant to loudness. Listeners consistently adjusted the ITD to compensate for envelope timing differences, and this adjustment decreased as sound level increased, but never reached zero.
  • Neural Latency Mirrors Perception: The timing of neural responses in the auditory brainstem (LSO) also showed level-dependent shifts that mirrored the perceptual findings. This suggests that the neural processing of sound timing at the input stage to the binaural system is directly affected by loudness.
  • The "Level-Dependent Firing Threshold" is Key: The most significant insight is that existing models of neural latency fail to explain these findings unless a level-dependent firing threshold is incorporated. This suggests that the auditory system dynamically adjusts its firing threshold based on sound intensity to maintain accurate temporal coding.
  • Binaural System Compensates, But Not Perfectly: While the binaural system (DNLL) does process these envelope timing differences, the observed eITDs in the DNLL were smaller than predicted by the monaural LSO data and the perceptual adjustments. This hints at further processing or potential limitations in the DNLL's ability to fully compensate for all level-dependent latency shifts.

Key Results

  • Psychophysical: Adjusted eITDs decreased monotonically with increasing sound sensation level, from approximately 1,150 µs at 20 dB SPL to 580 µs at 60 dB SPL for 6-ms pips. Even at the highest levels, the eITD did not asymptote to zero.
  • Electrophysiological (LSO): Time shifts in neural responses decreased from ~810 µs at 26 dB above threshold to ~357 µs at 50 dB above threshold for 6-ms pips. These shifts also decreased with increasing sound level and were more pronounced for shorter pip durations.
  • Modeling: A model with a fixed firing threshold could not reproduce the observed level-dependent eITDs. However, a model incorporating a level-dependent firing threshold provided a good qualitative match to both psychophysical and LSO electrophysiological data. The fitting error for the level-dependent threshold model was significantly lower (149 µs) compared to the fixed threshold model (415 µs).
  • Electrophysiological (DNLL): Time differences derived from DNLL neurons also decreased with increasing sound level, but were generally smaller than those observed in the LSO or in psychophysics.

Limitations and Future Work

  • Stimulus Specificity: The study used specific "pip train" stimuli. While effective for probing envelope timing, their generalizability to more complex natural sounds needs further investigation.
  • Neural Basis of Threshold Adjustment: The precise physiological mechanisms underlying the proposed level-dependent firing threshold remain to be fully elucidated. The authors suggest the cochlear nucleus as a potential site for this adjustment.
  • DNLL vs. LSO Discrepancy: The smaller eITD changes observed in the DNLL compared to the LSO and psychophysics warrant further investigation. Potential reasons include differences in stimulation paradigms (static vs. dynamic eITDs) and the DNLL's role as a processing stage that might not be the sole determinant of binaural envelope processing.
  • Future Directions: The authors suggest exploring the role of dynamic compressive mechanisms and neural adaptation in threshold adjustment. They also propose investigating the interaction of eITDs with interaural level differences (ILDs) to better understand the dynamics of binaural processing. The inclusion of a model based on auditory nerve responses (Zilany et al., 2009) showed promising qualitative agreement, suggesting that refining such models with level-dependent adaptation mechanisms is a fruitful avenue.

Summary updated

Abstract

The mammalian binaural system compares the timing of monaural inputs with microsecond precision. This temporal precision is required for localizing sounds in azimuth. However, temporal features of the monaural inputs, in particular their latencies, highly depend on the overall sound level. In a combined psychophysical, electrophysiological, and modeling approach, we investigate how level-dependent latency shifts of the monaural responses are reflected in the perception and neural representation of interaural time differences. We exploit the sensitivity of the binaural system to the timing of high-frequency stimuli with binaurally incongruent envelopes. Using these novel stimuli, both the perceptually adjusted interaural time differences and the time differences extracted from electrophysiological recordings systematically depend on overall sound pressure level. The perceptual and electrophysiological time differences of the envelopes can be explained in an existing model of temporal integration only if a level-dependent firing threshold is added. Such an adjustment of firing threshold provides a temporally accurate neural code of the temporal structure of a stimulus and its binaural disparities independent of overall sound level.
BibTeX
@article{siveke2010level,
  author = {Ida Siveke and Christian Leibold and Katharina Kaiser and Benedikt Grothe and Lutz Wiegrebe},
  title = {{Level-Dependent Latency Shifts Quantified Through Binaural Processing}},
  journal = {Journal of Neurophysiology},
  year = {2010},
  doi = {10.1152/jn.00392.2010},
}
Figure