Across-species differences in pitch perception are consistent with differences in cochlear filtering

Kerry M. M. Walker, Ray González, Joe Kang, Josh H. McDermott, Andrew J. KingeLife 2019 · DOI

TL;DR

This paper investigates why different species perceive pitch differently, proposing that variations in cochlear filtering are the key. By simulating cochlear responses and conducting behavioral experiments with ferrets and humans, they demonstrate that species' auditory systems are tuned to different cues (temporal vs. harmonic) for pitch perception, directly linking this to their cochlear filter bandwidths.

What Problem Does This Paper Address?

The fundamental basis of pitch perception, especially how it's processed across different species, remains an open question. While pitch is crucial for understanding speech, music, and animal vocalizations, studies have yielded conflicting results when comparing pitch perception abilities and strategies between species. This makes it difficult to generalize findings from animal models to human hearing. The authors aim to reconcile these differences by exploring whether variations in the auditory periphery, specifically cochlear filter bandwidths, explain why different species might rely on different acoustic cues (resolved harmonics vs. temporal envelope) to determine a sound's fundamental frequency (F0), which we perceive as pitch.

Methods

The study employed a two-pronged approach:

  • Cochlear Modeling: They adapted a standard cochlear filter bank model (Patterson et al., 1992) to simulate the auditory nerve's representation of complex tones in both human and ferret cochleas. Crucially, they used species-specific parameters for cochlear filter bandwidths, derived from human psychophysics and ferret auditory nerve recordings. This allowed them to visualize and predict how resolved and unresolved harmonic cues would be represented in each species.
  • Behavioral Experiments: They trained ferrets and humans on an analogous two-alternative forced choice (2AFC) pitch classification task. This task involved discriminating between two fundamental frequencies (F0s) of harmonic complex tones. To pinpoint which acoustic cues were being used, they introduced "probe trials" with modified stimuli where either resolved harmonics or temporal envelope cues were manipulated (e.g., by removing low harmonics or randomizing phases). Ferrets received water rewards, while humans used keypresses.

Datasets

The study did not explicitly train models on large datasets. Instead, they used:

  • Existing Psychophysical Data: Parameters for the cochlear model were derived from published human psychophysical data (e.g., Glasberg and Moore, 1990) and ferret auditory nerve recordings (e.g., Sumner and Palmer, 2012).
  • Generated Stimuli: They created specific harmonic complex tones and probe stimuli for the behavioral experiments.
  • Publicly Available Data: The psychophysical data and stimuli generated for this study are available on the Dryad Digital Repository.

Key Insights and Ideas

  • Cochlear Bandwidth as the Driver: The core idea is that differences in cochlear filter bandwidths between species directly influence the availability and salience of different pitch cues. Broader filters in ferrets lead to fewer resolved harmonics but a stronger temporal envelope representation, while narrower human filters resolve harmonics better.
  • Species-Specific Pitch Strategies: This difference in peripheral processing leads to distinct strategies for pitch perception. Ferrets rely more on temporal envelope cues, while humans prioritize resolved harmonic cues.
  • Unambiguous Dissociation: The study provides the first clear behavioral dissociation of pitch mechanisms across species using the same task, directly linking it to cochlear filtering.
  • Implications for Animal Models: The findings highlight ferrets as a valuable model for studying temporal pitch extraction mechanisms, similar to how cochlear implant users might process sound.

Key Results

  • Cochlear Model: Simulations showed that ferret cochlear filters are wider than human ones, leading to fewer resolved harmonics and a stronger temporal representation of F0.
  • Ferret Behavior: Ferrets performed poorly when temporal envelope cues were disrupted (e.g., 'High Harmonics Random Phase' probes) but were unimpaired when resolved harmonics were removed ('Low Harmonics' probes). This indicates a strong reliance on temporal cues.
  • Human Behavior: Humans showed impaired performance when resolved harmonic cues were removed ('High Harmonics' probes) but performed well when temporal envelope cues were disrupted, as long as resolved harmonics were present. This confirms their reliance on harmonic cues.
  • Performance Index: A normalized performance index showed that ferrets' performance dropped significantly when temporal cues were degraded, while humans' performance was largely unaffected by the degradation of temporal cues but suffered when harmonic cues were removed.

Limitations and Future Work

  • Ferret Harmonic Estimation: The authors acknowledge that their estimation of resolved harmonics in ferrets might be conservative, as ferret auditory nerve tuning curves were used, which might not fully capture the effects of lateral suppression in complex tones.
  • Stimulus Confound in Humans: For human participants, it was not possible to perfectly match the spectral edge of all stimuli across conditions due to the smaller F0 difference required. However, they argue this confound was unlikely to have driven the key results.
  • Future Work: The authors suggest that further investigation into primate pitch perception would be valuable, as they might share more similarities with humans. They also propose that applying normative models of pitch perception could further illuminate the relative importance of different cues. The study also opens avenues for exploring how different training regimes might influence species' reliance on specific pitch cues.

Summary updated

Abstract

Pitch perception is critical for recognizing speech, music and animal vocalizations, but its neurobiological basis remains unsettled, in part because of divergent results across species. We investigated whether species-specific differences exist in the cues used to perceive pitch and whether these can be accounted for by differences in the auditory periphery. Ferrets accurately generalized pitch discriminations to untrained stimuli whenever temporal envelope cues were robust in the probe sounds, but not when resolved harmonics were the main available cue. By contrast, human listeners exhibited the opposite pattern of results on an analogous task, consistent with previous studies. Simulated cochlear responses in the two species suggest that differences in the relative salience of the two pitch cues can be attributed to differences in cochlear filter bandwidths. The results support the view that cross-species variation in pitch perception reflects the constraints of estimating a sound's fundamental frequency given species-specific cochlear tuning.
BibTeX
@inproceedings{walker2019across,
  author = {Kerry M. M. Walker and Ray González and Joe Kang and Josh H. McDermott and Andrew J. King},
  title = {{Across-species differences in pitch perception are consistent with differences in cochlear filtering}},
  booktitle = {eLife},
  year = {2019},
  doi = {10.7554/elife.41626},
}
Figure