Acoustic Cues to the Singleton-Geminate Contrast: The Case of Libyan Arabic Sonorants

Amel IssaInterspeech 2017

TL;DR

This study investigates the acoustic differences between single and double (geminate) consonants in Libyan Arabic, focusing on sonorant sounds. The findings strongly suggest that duration is the primary acoustic cue distinguishing these consonants, with other acoustic features like amplitude and formant frequencies playing a minimal role. This research contributes to our understanding of phonological contrasts in Arabic dialects.

What Problem Does This Paper Address?

The distinction between single and geminate consonants is a crucial phonological feature in many languages, including Arabic. While it's generally accepted that duration is a key acoustic cue for this contrast, some research suggests that other non-durational cues, such as differences in articulation or resonance, might also contribute. However, these findings are not consistent across all languages, and the specific phonetic realization of gemination in Libyan Arabic (TLA), particularly for sonorant consonants, has been underexplored. This paper aims to fill this gap by systematically examining both durational and non-durational acoustic correlates of the singleton-geminate contrast in TLA sonorants.

Methods

The researchers employed a combination of acoustic measurements and statistical analysis.

  • Acoustic Measurements: They extracted several acoustic features from recorded speech:
  • Duration: Target consonant duration and the duration of the preceding vowel.
  • RMS Amplitude: The root-mean-square amplitude of the target consonants, normalized by the preceding vowel's amplitude.
  • Formant Frequencies: F1, F2, and F3, measured at the midpoint of the target consonant.
  • Segmentation: Consonant boundaries were identified using an "oral constriction criterion," focusing on the onset and release of the oral constriction.
  • Statistical Analysis: Independent analyses of variance (ANOVAs) and independent t-tests were used to determine the significance of differences between singleton and geminate consonants across various acoustic measures and sound categories.

Datasets

The study utilized a dataset of trisyllabic, word-medial, intervocalic minimal or near-minimal words containing the sonorant sounds /l, m, n, r/. These words were spoken by four native Libyan Arabic speakers. The stimuli were designed to include both singleton and geminate consonants preceded by short and long vowels. The extracted tokens were analyzed individually. No new benchmark datasets were introduced; rather, existing speech data was collected and analyzed.

Key Insights and Ideas

  • Duration Reigns Supreme: The most striking finding is the overwhelming importance of duration. Geminate consonants are consistently and significantly longer than their singleton counterparts, with a ratio of approximately 1:2.42. This holds true across different sonorant types.
  • Preceding Vowel Duration as a Secondary Cue: The duration of the vowel preceding the consonant also appears to be a significant factor. Short vowels are noticeably shorter before geminate consonants, suggesting a compensatory mechanism. Long vowels, however, do not show this significant durational difference.
  • Amplitude is Not a Differentiator: Contrary to some findings in other languages, RMS amplitude did not show significant differences between singleton and geminate sonorants in TLA. The authors hypothesize this might be because sonorants, unlike stops, don't involve significant air pressure build-up, which is often linked to higher amplitude in geminate stops.
  • Formants Show No Gestural Differences: Formant frequencies (F1, F2, F3) did not reveal any systematic differences between singleton and geminate sonorants. This suggests that gemination in TLA sonorants does not involve distinct gestural configurations or resonance patterns, unlike what has been observed for geminate laterals in some other languages.

Key Results

  • Consonant Duration: Geminate consonants were significantly longer than singletons (ratio C:CC of 1:2.42). The specific duration varied by consonant type, with /r/ being the shortest and /m/ the longest.
  • Vowel Duration: Short vowels preceding geminates were significantly shorter (t = 10.157, p<0.001) compared to those preceding singletons. Long vowels showed no significant difference.
  • RMS Amplitude: No significant differences in RMS amplitude were found between singleton and geminate consonants (F(1,3)=0.186, p=0.806).
  • Formant Frequencies: No significant differences were found for F1, F2, or F3 between singleton and geminate consonants, indicating that gemination does not affect the formant structure of TLA sonorants.

Limitations and Future Work

  • Limited to Sonorants: The study exclusively focused on sonorant consonants (/l, m, n, r/). The authors acknowledge that further research involving other consonant types (stops, fricatives) is necessary to provide a more comprehensive understanding of gemination in TLA.
  • Potential for Gestural Differences in Other Segments: While no gestural differences were found for sonorants, the possibility of such differences existing for other consonant types remains open.
  • Language-Specific Findings: The lack of non-durational cues for gemination in TLA sonorants highlights that these phonetic effects can be language-specific. Future research could explore how these findings compare to other Arabic dialects or languages with similar phonological contrasts.

Summary updated

BibTeX
@inproceedings{issa2017acoustic,
  author = {Amel Issa},
  title = {{Acoustic Cues to the Singleton-Geminate Contrast: The Case of Libyan Arabic Sonorants}},
  booktitle = {Interspeech},
  year = {2017},
}
Figure