Perception of the moraic obstruent /q/: a cross-linguistic study

Makiko Sadakata, Mizuki Shingai, Alex Brandmeyer, Kaoru SekiyamaInterspeech 2012

TL;DR

Japanese speakers perceive geminate consonants differently than Dutch speakers, relying on an abstract "silent duration" representation rather than just acoustic length. This suggests language-specific auditory processing for phonemes, even for seemingly straightforward acoustic features like consonant duration.

What Problem Does This Paper Address?

Languages like Japanese, Finnish, and Italian distinguish between single and geminate (doubled) consonants. While acoustic duration is a primary cue for this distinction in many languages, the paper questions whether this is the only or most important cue for native speakers. Specifically, for Japanese, there's a concept of a "moraic obstruent /Q/" which suggests geminates are perceived as a moraic unit plus a short consonant, implying an abstract representation that might include a silent duration. This study investigates whether Japanese listeners rely on this abstract "silent duration" when perceiving geminate consonants, particularly fricatives, and how this perception differs from that of Dutch speakers, whose language doesn't have this specific moraic concept.

Methods

The researchers conducted two experiments using synthesized pseudo-words.

  • Stimuli: Pseudo-words were created to represent three conditions:
  • /asu/: A singleton fricative consonant.
  • /assu/: A geminate fricative consonant (acoustically longer).
  • /a_su/: A representation of a "silent duration" followed by a singleton fricative. This was designed to mimic the hypothesized Japanese perception of a geminate fricative.
  • Tasks:
    • Discrimination Test: Participants listened to a standard sound (/assu/) and had to discriminate it from either a deviant sound (/a_su/ or /akku/ in Exp 2) or a distractor sound (/asu/). This tested their ability to detect differences in speech sounds.
  • Categorization Test: Participants were trained to associate sounds with labels and then had to categorize new sounds. This tested how they grouped sounds into categories.
  • Participants: Native Japanese and Dutch speakers were recruited. In Experiment 2, participants were further divided into musicians and non-musicians to explore the potential influence of musical training on speech perception.

Datasets

Not discussed in the paper. The stimuli were synthesized using the Praat application.

Key Insights and Ideas

  • Japanese listeners rely on an abstract "silent duration" for geminate fricatives: The core finding is that Japanese speakers perceive a "silent duration + singleton fricative" (/a_su/) as very similar to a "geminate fricative" (/assu/). This suggests they are not just listening to the acoustic length but are applying a language-specific rule that interprets geminates as having an inherent silent component.
  • Cross-linguistic differences in perception: Dutch listeners, unlike Japanese listeners, were able to discriminate and categorize /a_su/ and /assu/ more distinctly. This supports the idea that the moraic obstruent concept is specific to Japanese.
  • Discrimination is easier than categorization: Across both experiments and language groups, participants were generally better at discriminating between sounds than categorizing them. This is a common finding in speech perception research, suggesting that the perceptual system can detect fine acoustic differences but grouping them into meaningful categories is more complex.
  • Musical experience doesn't significantly alter the effect: While musicians are often expected to have enhanced auditory processing, the study found no significant difference in how musicians and non-musicians perceived the geminate contrasts, suggesting the moraic effect is deeply ingrained in native language processing.

Key Results

  • Discrimination: All participants, regardless of native language, could discriminate between the different consonant types (e.g., /a_su/ from /assu/). Error rates were generally low (below 10% in Experiment 2).
  • Categorization (Experiment 1): Japanese natives showed significantly higher error rates when categorizing /a_su/ and /assu/, often confusing them. Dutch natives performed much better, categorizing them more distinctly.
  • Categorization (Experiment 2): The results were consistent with Experiment 1. Japanese natives again showed higher error rates for the "Deviant" (/a_su/) condition compared to Dutch natives, indicating confusion between the abstract silent duration and the actual geminate fricative.

Limitations and Future Work

  • Stimulus Specificity: The authors acknowledge that the specific way /a_su/ was constructed might have influenced results, suggesting the need for testing with more varied stimuli.
  • Task Difficulty: The categorization task, particularly the 3AFC in Experiment 1, might have been challenging. Future work could explore easier categorization paradigms.
  • Dutch Training: Dutch participants in Experiment 1 underwent significant training to perceive Japanese geminate-singleton contrasts. This extensive exposure might have influenced their performance, making it harder to draw definitive conclusions about their native perception without such training. Experiment 2 aimed to mitigate this by using less training and more varied stimuli.
  • Future Directions: The authors suggest further research could explore the influence of musical experience more rigorously and investigate how this moraic representation interacts with other phonological features.

Summary updated

BibTeX
@inproceedings{sadakata2012perception,
  author = {Makiko Sadakata and Mizuki Shingai and Alex Brandmeyer and Kaoru Sekiyama},
  title = {{Perception of the moraic obstruent /q/: a cross-linguistic study}},
  booktitle = {Interspeech},
  year = {2012},
}
Figure