Reversed sonority in Pashto initial clusters

Pashto possesses both resonant clusters of rising sonority such as lw‐ and clusters of reversed sonority, such as wl‐. The highly preferred status of the first clusters has been incorporated in the syllabic theories of Bybee Hooper, Vennemann, Fujimura and Lovins, and Kiparsky. The question arises whether the dispreferred clusters manifest phonetic adaptations to compensate for their reversed sonority. The clusters ml‐, lm‐, lw‐, wl‐, and wr‐ and the singlet initials m‐, l‐, r‐, and w‐ were analyzed. Waveforms and LPC spectral analyses were obtained for 120 tokens recorded by the second author, using the Haskins Laboratories speech analysis programs. No frication was observed for any onset. All onsets were partially devoiced. The devoicing duration depended only upon the initial segment, not the sonority type of the cluster it began. Comparing the segment durations of the preferred ml‐ and lw‐ clusters to the reversed lm‐ and wl‐ clusters showed that the more sonorous segments were relatively longer beginning reversed clusters than ending preferred ones; the less sonorous segments showed the opposite relationship. No adaptation to reversed sonority was found. [NICHD support acknowledged.]