Phonology of Deori: an endangered language Prarthana Acharyya and Shakuntala Mahanta Department of Humanities and Social Sciences, Indian Institute of Technology Guwahati, India a.prarthana@iitg.ernet.in, smahanta@iitg.ernet.in Abstract Deori is a Tibeto-Burman (henceforth TB) language belonging to the Bodo-Garo sub family [5]. It is spoken in the eastern parts of Assam and some parts of Arunachal Pradesh. It has a population of 41,161 (2001 census) and is spoken by approximately 28,000 people [9]. Deori has been explicitly listed as a „definitely endangered‟ language [9] and as a „severely endangered language‟ [3]. It has been claimed that the linguistic features of Deori has given an individual identity to the language. This paper studies the current use of the lexical tone in Deori and describes the language change attested in the speech of the younger generation speakers. It also presents an overview of the segmental phonology of Deori. Deori has seventeen consonants /p, t, k, b, d, g, m, n, ŋ, tʃ, s, z, h, l, w, r, j/ and has seven oral vowels /a,e,ɛ,i,o,ɔ,u/ and five nasal vowels /ɑ ̃ ,ɛ ̃ ,,,/. Deori exhibits a process of [ATR] harmony in which vowels in a certain domain (such as „word‟) agree with a [±ATR] feature specification. Presence of contrastive nasal vowel in Deori is considered as a “rare linguistic feature” [2]. It is seen that Deori follows many of the observed processes of vowel nasalization similar to many other languages with nasal harmony. At the same time there are aspects of vowel nasalization in Deori which are not so frequently observed. While nasalization spreads across and to glides under all conditions /gr͂ ɔ ̃ / „throat‟, /ʧi᷉j͂ a᷉ / „wife of younger brother‟, fricatives are transparent segments in non-derived domains, as in /dsɔ ̃ / “pot”, ũzũ „bracelet‟ and they are opaque in derived domains, for instance /tadũ+si/ “spoon.SEL, /bibɔ ̃ +si/ „granary.SEL‟. It has been claimed that languages belonging to the TB family range from having many tonal contrasts to none, and from displaying emerging tonal contrasts to disappearing ones [4]. While tonogenesis is a widely acclaimed feature in the literature on tone, a large body of literature also records a gradual loss of the already existing tone, a process known as tonoexodus [4]. In the existing literature on Deori there have been many observations about tone. Different scholars have expressed doubts about the presence of two lexical tones in Deori [8] [10] and therefore tone has been considered a moribund feature in Deori by some others [2]. Most recently it has been shown that Deori manifest level tones with special reference to the older generation speakers, though minimally and the Tone Bearing Unit (TBU) is the entire phonological word. Tonal contours in Deori do not manifest a fixed tonal characteristic as compared to the other genetically related Tibeto-Burman languages like Bodo, Dimasa and Tiwa [6]. Following the minimal tonal contrast prevalent in Deori a production experiment was conducted to look into the tonal realization in the speech of the younger generation. The hypothesis formulated for the study is that due to constant contact with the surrounding non-tonal and dominant official state language of Assam, there is a gradual loss of tonal realization among the younger generation speakers. Deori is hypothesized to be in an intermediate stage where tonoexodus is present but the language is also beginning to emerge as a stress accent language without tonal distinction [6]. For the present study we used the same set of words (16 disyllabic and 10 monosyllabic words) that were examined in the previous study on tone [6] as the tonal contrasts for these words were already established in the previous study. The target word bearing the tonal contrast was embedded in a fixed sentence frame “I X said” [ã X nina itʃabɛm] where X is the target word. The use of the frame ensured that intonational interference in the target words was uniform and hence predictable. Tone was not marked in the orthography presented to the participants. A method of using pictures of the target word was integrated in the experimental design in order to have one to one correspondence with the actual meaning in an appropriate context. Each word with the carrier sentence along with the relevant picture was randomized and was presented to the participants on the computer screen and they were instructed to read each word in the predetermined sentence frame. Three of the four repetitions of each test word were used for measurement; the last repetition was discarded to avoid an intonational boundary effect. Recording was done using a Shure SM-10 head-mounted microphone connected to a Tascam DR 100 MK II solid state recorder. Five speakers (18-30 years) participated in the production experiment. The result of the production test shows a trend of underlying tone reversal [+H]