The Three Way Tonal System of Sylheti Priti Raychoudhury 1 , Shakuntala Mahanta 2 1 IIT Guwahati, India 2 IIT Guwahati, India raychoudhury19@iitg.ac.in, shakunmahanta@gmail.com Abstract This study reports data collected and analyzed from 7 native speakers from a corpus of 70 Sylheti noun words. Our work shows that the loss of [+spread glottis] feature from [-voice] and [+voice] onsets have resulted in independent tone association patterns. The [-voice, +spread glottis] onset associated to a Low tone, as opposed to the tone association pattern of the voiced onsets [1]. Our study looks into the tonal pattern of Sylheti which arose from the merger of [+spread glottis] and [-spread glottis] contrast in voiceless obstruents. We built linear mixed effect models of f0 and duration to examine the acoustic factors affecting tone. We found that disyllables trigger a three way tonal contrast depending on the historical features and positions of voiceless onsets in Sylheti. The tone of the syllable spreads throughout the word rather than the syllable of origin. Word is thus reclaimed to be the TBU in Sylheti. The Mean f0 for the intercept was about 274.345 Hz which represented the High tone. It differed from the Low tone by about 75 Hz, and the Mid tone by about 25 Hz. Tone affected pitch by (χ2 (1) = 927.07, p < 0.0001). Index Terms: voiceless onsets, tonogenesis, disyllables, tone 1. Introduction Sylheti is an Indo-Aryan language. It is the main language spoken in the Surma and Kushiara valleys of Sylhet Division in Bangladesh and the Barak valley region of Assam, India. The language has been regarded as a sub-division of Bangla for long but was also considered to be phonologically and grammatically unintelligible to other Bangla dialects [2]. The Sylheti language had made great advances in its literature where a distinct script was used in administrative and religious sectors during the 6th century AD [3], [4]. The ‘Sylheti Nagari’ script has recently been revived for educational purposes [5]. The ‘Sylheti Nagari’ script matches the phonology of spoken Sylheti and consists of 5 vowels and 28 consonants and was in use in printing and publishing, especially in the Sylhet division until the independence of Bangladesh. Sylheti is structurally related both to Assamese and to the rural dialects of eastern Bengal, but with a high proportion of words derived from Persian and Arabic, and has its own distinct grammar [5]. Apart from the robust differences in the lexicon and morpho-syntactic structure, Sylheti also has a noticeably reduced inventory of phonemes as compared to its cognate languages which has led to tonogenesis in the language [6]. Our present study shows that the historical [-voice, +spread glottis] onset associates to a Low tone as opposed to the tone association pattern of the voiced onsets as established in previous studies [1]. Voiceless consonant onsets as in *p h alɸàl ‘jump (noun)’ led to a Low lexical tone in Sylheti and the onsets with [-voice, -spread glottis] consonants thus resulted in a contrastive high tone as in *palɸāl ‘group of animals’. When we extended the research to disyllables, we found that this contrastive high tone associates to the lexical Mid tone in Sylheti. The study shows that in disyllables the onsets with [-voice, -spread glottis] consonants resulted in a lexically contrastive Mid tone (and not High tone) as in the word for crooked stick, xūʈā which was historically *kuʈa. The onsets with a historical [-voice, +spread glottis] onset of the leftmost syllable led to a lexical Low tone as in the word for taunt, xùʈà which was *k h uʈa. The words with underlying [- voice, +spread glottis] feature in the onset of the second syllable results in a contrastive High tone as in the word for room which is xúʈá was *kuʈ h a. The High tone followed this constraint only when the leftmost syllable had a [-voice, - spread glottis] onset for the same word. As against monosyllables which exhibit only two tones, our study of disyllables shows that the loss of aspiration contrast in different onset positions, led to three lexically contrastive tones in Sylheti. The study confirms that tone is spread throughout the word and thus word is the TBU in Sylheti as reported in earlier studies on Sylheti tonogenesis. The pitch contrast in terms of mean f0, maximum f0 and minimum f0 for the three tones was very significant (p < 0.00001). The significance of pitch contrast within syllables was found to be only about 3 Hz, p = 0.1254. 2. Background The earlier well received studies on Sylheti tonogenesis claim Sylheti has a two way lexical tone contrast as a consequence of the merger of aspirated and unaspirated obstruents. The studies [7] on the acoustics of Sylheti phonemes reported that Sylheti has a complete absence of [+spread glottis] feature from its phonemic inventory. Gope reports that the Sylheti phonemic inventory consists majorly of fricative consonants. Gope’s work shows that there is a complete absence of affricates in the language as they reduced to homorganic fricatives. For instance, he shows that *tʃ de-affricates to [s] as a result historical words like *tʃal changed to [sāl] ‘roof’; similarly the word *tʃ h al changed to [sàl] ‘skin’. The voiced counterpart of the affricate *dz h was reduced to [z] as in the tonal pair *dz h al[zál] ‘spicy’ and *dzal[z̀al] ‘net’. The voiceless velar stop was reduced from */k/ to [x] as in the word for ‘drain’, *k h alx̀al but the voiced velar [g] remained unchanged. Similarly the retroflex stops *ʈ and *ɖ have merged with the alveolar stops [t] and [d] respectively as in *paʈaɸā.tā ‘grinding-stone’ or in *ɖ h ald̀al ‘pour’. The dental stops [t̪] and [d̪] too remained unchanged. The studies show that onsets which did not undergo fricativization triggered tonogenesis in Sylheti as well. As fricativization did spread throughout all the obstruents in the phonemic inventory, this sound change was not considered as a factor for triggering tonogenesis in the language. Gope studies Sylheti tonogenesis with special focus on the merger of aspirated and unaspirated voiced consonants and provides an explanation for Speech Prosody 2020 25-28 May 2020, Tokyo, Japan 503 10.21437/SpeechProsody.2020-103