The Role of Musical Experience in the Perceptual Weighting of Acoustic Cues for the Obstruent Coda Voicing Contrast in American English

Michelle Cohn, Georgia Zellou, Santiago BarredaInterspeech 2019

TL;DR

This study investigates whether musical training influences how people weigh different acoustic cues when distinguishing between voiced and voiceless consonants in American English. Researchers found that musicians, compared to non-musicians, show steeper categorization functions and a more pronounced shift in cue weighting when speech is presented in noisy conditions, suggesting that musical experience enhances auditory processing and attentional strategies relevant to speech perception.

What Problem Does This Paper Address?

The relationship between musical training and speech perception is a complex and debated topic. While some research suggests that musicians have enhanced auditory abilities that transfer to speech processing, especially in noisy environments, other studies show no significant differences between musicians and non-musicians. This paper aims to clarify this by focusing on how individuals weigh acoustic cues for distinguishing between voiced and voiceless obstruent codas (like the difference between "beat" and "bead") in American English. Specifically, it asks whether musical experience alters the relative importance listeners assign to cues such as vowel duration, fundamental frequency (f0), and spectral composition, particularly under challenging listening conditions like background noise. Understanding this could shed light on the nature of experience-driven neuroplasticity and individual differences in auditory perception.

Methods

The researchers recruited 44 adult native English speakers, divided into two groups: musicians (with 10+ years of instrumental practice) and non-musicians (with minimal musical experience). They presented participants with synthesized speech stimuli of the words "beat" and "bead," where key acoustic cues were systematically manipulated: vowel duration (ranging from 180 to 300 ms), fundamental frequency (low at 120 Hz, high at 140 Hz), and spectral composition (preserved cues from the original recordings).

These stimuli were presented in two listening conditions: silence and multitalker babble (MTB), a realistic background noise. Participants were asked to identify whether they heard "beat" or "bead." The core of the analysis involved using mixed-effects logistic regression to model the probability of a "bead" response. This allowed the researchers to examine the main effects of group (musician vs. non-musician), listening condition (silence vs. MTB), and the acoustic cues, as well as their interactions. Crucially, they analyzed how the "weight" or importance of each acoustic cue changed based on musical experience and the presence of noise.

Datasets

Not discussed in the paper. The stimuli were synthesized for the experiment.

Key Insights and Ideas

  • Musicians are "sharper" listeners: Musicians showed steeper categorization functions for all three acoustic cues (duration, f0, spectrum). This means they were more precise in using these cues to distinguish between voiced and voiceless sounds, suggesting enhanced perceptual acuity.
  • Noise changes the game for everyone, but musicians adapt more: Both musicians and non-musicians shifted their cue weighting in noise. Specifically, in multitalker babble, vowel duration and f0 became stronger cues, while spectral cues became less important. This re-ranking was more pronounced in musicians.
  • Music-to-speech transfer is multifaceted: The findings support theories suggesting that musical training can transfer to speech perception, enhancing both subcortical/cortical encoding of acoustic features (bottom-up) and attentional mechanisms (top-down). However, the results also challenge theories that predict transfer only for specific "music-like" features (like f0), as musicians showed enhanced weighting for duration and spectral cues as well.
  • Language experience is a strong driver: Despite differences, the overall pattern of cue re-ranking in noise was similar for both groups, suggesting that native language experience plays a significant role in shaping these fundamental perceptual strategies.

Key Results

  • Musicians displayed significantly fewer "bead" responses than non-musicians overall, indicating a general difference in their categorization.
  • Increased vowel duration and preserved spectral information were significant predictors of a "[+voice]" (bead) response for both groups. A lower f0 also showed a trend towards increasing voiced judgments.
  • In multitalker babble (MTB), listeners relied more heavily on vowel duration and lower f0 compared to silence. Conversely, they relied less on spectral cues in MTB than in silence.
  • A significant three-way interaction revealed that while musicians generally relied more on spectral cues, this reliance decreased in MTB compared to silence, relative to non-musicians.

Limitations and Future Work

The authors acknowledge several limitations and suggest future research directions:

  • Stimulus limitations: The study used synthesized stimuli and a limited set of words ("beat"/"bead"). Future work could explore a wider range of sounds and natural speech.
  • Task design: The identification task forced a binary choice. Future studies could use continuous rating scales to investigate finer gradations in perception and cue weighting.
  • Feedback: The experiment did not provide feedback to participants. Incorporating feedback could reveal how flexible cue weighting is and how quickly listeners can adapt.
  • Discrimination vs. Identification: Future research could use discrimination tasks to further tease apart the contributions of bottom-up sensory encoding versus top-down attentional processing differences between musicians and non-musicians.
  • Generalizability: The findings need to be tested with other obstruent coda contrasts and potentially across different languages to assess the generalizability of musical training's impact on speech perception.

Summary updated

BibTeX
@inproceedings{cohn2019role,
  author = {Michelle Cohn and Georgia Zellou and Santiago Barreda},
  title = {{The Role of Musical Experience in the Perceptual Weighting of Acoustic Cues for the Obstruent Coda Voicing Contrast in American English}},
  booktitle = {Interspeech},
  year = {2019},
}
Figure