Farsi vowel compensatory lengthening: an experimental approach

This paper addresses a question concerning phonological operations involved standard Farsi vowel Compensatory Lengthening (CL) due to glottal consonant deletion in syllabic coda. Specifically, I question two current views related to CL process. One view suggests that glottal allophonic weakening is compensated by vowel lengthening. Another view suggests that glottal consonant is deleted, but vacated mora is filled by preceding vowel. Based on an experimental approach, I claim that Farsi CL is a gradient process in which different magnitude of glottal gesture variation is accompanied by F0 decreasing of vowel onset and offset and lengthening of the vowel. Evidence for this claim comes from the results of two experiments. For experiment 1, ten sentences having ten CaC words were made. /a/ duration was doubled while each synthetic CaaC word supposed to be a phonetic counterpart of another CaCC lexical word. Ten other sentences were generated by such CaaC words. Ten subjects listened to these twenty sentences. Subjects’ lexical ambiguity to understand CaC and CaCC words was 37.8%. For experiment 2, subjects colloquially uttered those twenty sentences with CaC and CaCC words. I found that, in addition to significant vowel lengthening for colloquial CaCC , /a/ onset and offset F0 was significantly decreased, while different magnitude of glottal gesture was realized in speech signal from a weak through complete deletion of glottals. This finding can be used to improve Farsi ASR and TTS systems.