Acoustics and Articulation of Medial versus Final Coronal Stop Gemination Contrasts in Moroccan Arabic

Mohamed Yassine Frej, Christopher Carignan, Catherine T. BestInterspeech 2017

TL;DR

This study investigates how speakers of Moroccan Arabic distinguish between single and doubled consonants (geminates) at the end of words, a phenomenon that's often debated. Using acoustic and ultrasound data, the researchers found that speakers use different strategies for medial vs. final geminates, with one speaker relying on release duration for final geminates, a novel finding.

What Problem Does This Paper Address?

The paper tackles the question of whether the contrast between single and doubled consonants (gemination) at the end of words in Moroccan Arabic (MA) is still actively used by speakers. While gemination is common in Arabic, research has largely focused on consonants in the middle of words. There's a debate about whether final geminates are neutralized or if they still carry meaning. Previous studies have been impressionistic, lacking instrumental data. This paper aims to fill that gap by using acoustic and articulatory methods to see how MA speakers produce and distinguish these sounds, and how this differs from medial gemination.

Methods

The researchers employed a dual approach:

  • Acoustic Analysis: They analyzed recordings of speakers producing words with single and geminate consonants in medial and final positions. Key acoustic features measured included the duration of the preceding vowel, the closure period of the consonant, and the release burst.
  • Articulatory Analysis (Ultrasound): To get a more direct look at how the tongue moves, they used ultrasound. A novel aspect here is the application of TRACTUS (Temporally Resolved Articulatory Configuration Tracking of UltraSound). This method allows for time-varying analysis of articulatory movements directly from ultrasound video without the need for tedious manual tracing of tongue contours frame by frame. This is particularly useful for studying duration-based contrasts like gemination.

Datasets

The study used recordings from two native Moroccan Arabic speakers (one female, one male) producing minimal pairs of words contrasting single and geminate coronal stops (/t/, /d/ vs. /tt/, /dd/) in medial and final positions. These words were recorded in isolation and within a carrier phrase. No new benchmark datasets were introduced; the focus was on analyzing existing speech data.

Key Insights and Ideas

  • Medial Gemination Strategy is Consistent: Both speakers consistently lengthened the closure duration of geminate consonants compared to singletons in the middle of words. This is a common and expected finding for medial gemination.
  • Final Gemination Strategies Diverge: This is where things get interesting. While one speaker (S1) continued to use closure duration to distinguish final geminates, the other speaker (S2) shifted their strategy. S2 primarily lengthened the release duration of final geminates, a less common cue.
  • Release Duration as a Final Gemination Cue: The finding that S2 uses release duration for final geminates is a significant insight, suggesting a potential speaker-specific adaptation or a more nuanced phonetic realization of this contrast.
  • Ultrasound Reveals Articulatory Nuances: The TRACTUS analysis provided visual evidence of articulatory configurations, showing that while the overall gestures for single and doubled consonants are similar, there might be subtle differences in constriction (e.g., pharyngeal and tongue blade approximation for geminates).
  • Laminal Coronal Stops: The study observed that MA coronal stops appear to be produced with a "laminal" articulation (using the blade of the tongue) rather than the more commonly assumed "apical" (using the tip of the tongue) for singletons, though this needs further validation.

Key Results

  • Medial Position: Both speakers significantly lengthened closure duration for geminates compared to singletons. Preceding vowel and consonant closure durations were also longer for geminates.
  • Final Position:
    • Speaker S1 continued to lengthen closure duration for final geminates.
  • Speaker S2 lengthened release duration for final geminates, while closure duration did not significantly differ.
  • For S1, release duration was longer in the final position than medial. For S2, release duration was longer medially than finally, but S2 used it contrastively for final geminates.
  • Articulatory Data: LDA heatmaps showed comparable gestures for singletons and geminates, with potential subtle differences in constriction. LDA scores over time indicated a clearer distinction between geminate and singleton lines in medial positions for both speakers, with a less pronounced but present distinction for S1 in final positions.

Limitations and Future Work

  • Limited Speaker Sample: The study only included two speakers, making it difficult to generalize the findings about divergent strategies for final gemination.
  • Reliability of Release Duration Contrast: The authors express caution about the reliability of the release-duration contrast for final gemination, given its novelty and the limited data.
  • Perceptual Studies Needed: The paper highlights the need for perceptual studies to determine if MA speakers can actually distinguish between final geminates and singletons based on these acoustic and articulatory cues.
  • Replication and Broader Scope: Future work should involve replicating the experiment with more participants to see if other speakers adopt similar or different strategies. The authors also plan to extend this analysis to word-initial gemination contrasts.

Summary updated

BibTeX
@inproceedings{frej2017acoustics,
  author = {Mohamed Yassine Frej and Christopher Carignan and Catherine T. Best},
  title = {{Acoustics and Articulation of Medial versus Final Coronal Stop Gemination Contrasts in Moroccan Arabic}},
  booktitle = {Interspeech},
  year = {2017},
}
Figure