Physiological and acoustic study of word initial post-lexical gemination in Moroccan Arabic

Chakir Zeroual, Diamantis Gafos, Phil Hoole, John EslingInterspeech 2012

TL;DR

This paper investigates the physiological and acoustic differences between single and geminate consonants at the beginning of words in Moroccan Arabic. Using EMA (Electromagnetic Articulography), they found that while geminates have longer acoustic closure, their articulatory movements (like tongue tip height and velocity) are surprisingly similar to single consonants, challenging some existing theories.

What Problem Does This Paper Address?

The paper tackles the question of how languages distinguish between single and "doubled" (geminate) consonants, particularly when they appear at the start of a word. Geminates are known to have longer acoustic durations, but the underlying articulatory mechanisms and how they interact with surrounding sounds are not fully understood. Previous research has focused mainly on geminates between vowels, and findings have been mixed. This study specifically looks at word-initial geminates in Moroccan Arabic, a context where consonant length contrasts are rare cross-linguistically, to see if articulatory differences exist and how they might relate to theories of speech production and syllable structure.

Methods

The researchers employed Electromagnetic Articulography (EMA) to track the precise movements of articulators (tongue tip, mid, and back, lower lip, and jaw) at a sampling rate of 200 Hz. They focused on the vertical movement of the tongue tip (TTIPy). Using custom software, they identified key temporal landmarks in the articulatory gestures, such as the onset of movement, target positions, and release points. They then analyzed durations of different phases of the articulatory gesture (closing, plateau, opening), amplitude, and peak velocity. Acoustic measurements of closure duration, burst, and total duration were also taken.

Datasets

The study used data from four native Moroccan Arabic speakers, with results presented for three of them (S1, S2, S3). The specific linguistic material analyzed was the minimal pair [duda] ('an earthworm') versus [dduda] ('the earthworm'), pronounced within the carrier phrase [zibi hnaja] ('bring here'). This minimal pair was chosen because [dd] in [dduda] is a derived geminate resulting from the assimilation of the definite article's /l/ to the following /d/.

Key Insights and Ideas

  • Acoustic vs. Articulatory Discrepancy: While geminates are acoustically longer, their physical articulation doesn't always reflect this in a straightforward way. The tongue tip doesn't necessarily move higher or faster for a geminate.
  • Challenging the "Virtual Target" Hypothesis: The idea that geminates require a higher articulatory target to maintain contact for longer is not strongly supported for these Moroccan Arabic coronals. This might be due to the rigid surfaces of the tongue and teeth.
  • No Strong C-Center Effect: The study found no significant "C-center effect" (a measure of how consonant gestures overlap with surrounding vowels) between the initial geminate consonant and the following vowel. This aligns with previous findings suggesting complex onsets are not typical in Moroccan Arabic.
  • Geminate as a Hetero-syllabic Cluster: The lack of a strong C-center effect and the similar kinematic properties of single and geminate consonants suggest that word-initial geminates in this context might be best understood as a sequence of two distinct, overlapped consonants belonging to different syllables.

Key Results

  • Acoustic Durations: Geminate [dd] had substantially longer acoustic closure duration (over 100% longer for S1 and S3) and total duration compared to single [d]. The following vowel [u] was slightly shorter in the geminate context, but this difference was not consistently significant across speakers.
  • Articulatory Durations: The plateau phase of the tongue tip gesture was significantly longer for [dd] than for [d] across all speakers, contributing to the overall longer constriction duration. Closing and opening phases were generally unaffected.
  • Velocity and Amplitude: For most speakers, there were no significant differences in the peak velocity or amplitude of the tongue tip movements between single [d] and geminate [dd]. One speaker (S1) showed higher peak velocities for [dd], but this was not accompanied by significantly different amplitudes.
  • Tongue Tip Height: The height of the tongue tip at key points (target, midpoint, release) during the gesture was not significantly different between [d] and [dd] for any speaker.
  • C-Center Effect: Temporal intervals measuring the overlap between the initial consonant gesture and the following vowel were consistently longer when the initial consonant was a geminate [dd], but this difference was not statistically significant for all intervals, suggesting a lack of a strong C-center effect.

Limitations and Future Work

  • Limited Speakers and Data: The study analyzed data from only three speakers and focused on a single minimal pair. While this allowed for detailed analysis, a larger sample size and a broader range of phonetic contexts would strengthen the conclusions.
  • Focus on Coronal Consonants: The findings are specific to the coronal stop [d]/[dd]. Further research on other consonant types (labials, velars) and in different positions (intervocalic) would provide a more complete picture of gemination in Moroccan Arabic.
  • Future Work: The authors suggest that their results support the idea of geminates being sequences of hetero-syllabic consonants. Future work could explore this further by examining other phonetic contexts and languages, and by developing more sophisticated models to account for the observed articulatory patterns.

Summary updated

BibTeX
@inproceedings{zeroual2012physiological,
  author = {Chakir Zeroual and Diamantis Gafos and Phil Hoole and John Esling},
  title = {{Physiological and acoustic study of word initial post-lexical gemination in Moroccan Arabic}},
  booktitle = {Interspeech},
  year = {2012},
}
Figure