Arabic Obstruents: Laryngeal Contrast and Representation

Abstract

This thesis examines the phonetics and phonology of Arabic obstruents in light of Laryngeal Realism. It focuses on how the voicing contrast is laryngeally modulated across several acoustic metrics. I draw from three acoustic experiments conducted to unpack the phonetic realisation and phonological representation of Arabic obstruents. The larger picture of this thesis is comparative in nature in that it seeks to investigate whether stops and fricatives exhibit similar laryngeal behaviours. The overarching goal involves examining how phonetics may inform formal representation, and how phonological patternings may explain phonetic differences for obstruents in Jazani Arabic. This thesis also disentangles the relative contribution and robustness of each acoustic correlate in predicting the domain of laryngeal contrast. Bayesian and Generalised Additive Mixed Models (GAMMs) were used as statistical approaches to answer the research questions. For visualisation, Bayesian estimation figures and smoothing spline figures were used to visualise the predicted results.

Experiment One examines how the voicing contrast is modulated in stops. The results show that 7 out of 16 voicing correlates exhibited a reliable difference between the effect of the voiced vs. voiceless stops on the respective measure with 0% of the 95% HDI of the posterior distribution falling within the ROPE range, indicating that these acoustic measures are robust and reliable for uncovering voicing distinction in the present dialect. In addition, the larger picture of the pattern of VOT and voicing%, inter alia, in all the contexts examined suggests that voiced stops in the Jazani dialect are specified as [voice] whilst the voiceless stops are specified as [tense].

Experiment Two examines how the voicing contrast is modulated in fricatives. The results show that 9 out of 14 voicing correlates exhibit a reliable difference between the effect of the voiced vs. voiceless fricatives on the respective measure with 0% of the 95% HDI of the posterior distribution falling within the ROPE range. In addition, the pattern of fricative duration, voicing%, zero-crossing rate, inter alia, suggests that voiced fricatives are specified as [voice] whilst voiceless fricatives are specified as [tense].

Experiment Three examines the effects of consonant voicing on f0 perturbation dynamically and whether it mirrors the VOT pattern and (dis)confirms the predictions of Laryngeal Realism. Laryngeal Realism predicts that acoustic correlates such as VOT, f0, etc. should exhibit a comparable pattern for either feature. The results show that voiced obstruents exhibit a pronounced lowering effect compared to the nasal baseline, indicating that voiced obstruents are specified as [voice]. In contrast, the results of the voiceless context show that voiceless obstruents exhibited an insignificant raising pattern at vowel onset. The results of f0 perturbation show that f0 perturbation exhibits a similar pattern as VOT. The results, furthermore, provide further evidence demonstrating that the feature driving these patterns in the voiceless obstruents is [tense]. These results are in line with the predictions of Laryngeal Realism.

While the combined results of these experiments show that stops and fricatives appear to exhibit relatively similar laryngeal behaviours and each category can be predicted from the other, each category, nonetheless, exhibits its own characteristics at the microscopic level.