PUB - Publikationen an der Universität Bielefeld

Obstruents’ closure duration is reported to vary depending on voicing, place of articulation, and position. Stathopoulus & Weismer (1983) showed that voiced stops in English have longer closure compared with their voiceless counterparts and initial stops are produced with longer closure duration than the medial and final ones. While most of the existing studies are based on word lists produced in European languages, we analyze closure duration in plosives produced in semi-spontaneous speech in Punjabi, an under-studied language. We investigate how closure duration is affected by plosives’ position (word initial, final), place of articulation (bilabial, dental, velar), and voicing (voiced, voiceless). We extracted normalized closure duration (closure/word duration) from 584 plosives produced by two female Punjabi speakers from Lahore, Pakistan. Linear Mixed-Effects Regression analysis showed no significant three-way interaction between position, voicing, and place of articulation (p = 0.2) for closure duration. However, there was significant main effect of voicing (p < 0.0001) as voiceless plosives were produced with longer closure compared with the voiced plosives, regardless of their position. Moreover, place of articulation interacted with position (p = 0.0001). Word initial bilabials were produced with shorter closure compared with their word final variants. However, initial dental plosives had longer closure than final dental plosives, whereas velars showed no effect of position. Significant interaction was found between voicing and place of articulation (p = 0.02). Voiceless dental plosives were produced with longer closure than bilabial and velar plosives. These results portray a complex interplay between position, voicing, and place of articulation in the closure of plosives in Pakistani Punjabi. While we could replicate the previous results regarding position, we found the reverse effect of voicing in our data compared to the findings of Stathopoulus & Weismer. This highlights the need for further research using semi-spontaneous speech data on under-researched languages.