The Xumi 旭米 language (/EPʃʉ-hĩ ketɕɐ/ ‘the language of the Shu people’) is spoken by approximately 1,800 people who reside along the banks of the Shuiluo River (水洛河) in Shuiluo Township (水洛乡) of Muli Tibetan Autonomous County (木里藏族自治县; smi li rang skyong rdzong in Written Tibetan, hereafter, WT).Footnote 1 This county is located in the South-West of Sichuan Province (四川省) in the People's Republic of China (see Figure 1).
Figure 1 Location of Shuiluo Township (Xumi villages marked with solid black circles).
The language of the group was first described by Sun Hongkai (孙宏开) (Sun Reference Hongkai1983), who labeled it ‘Shixing 史兴’ based on the autonym of the group (/EPʃʉ-hĩ/ ‘Xumi people’). However, this label is generally unknown in the county where the group resides and where its official name in the national Mandarin Chinese language is Xumi (旭米) (Muli Gazetteers 2010: 560, 563–564, 568).
Xumi is currently classified as a member of the putative Qiangic subgroup of the Sino-Tibetan language family (Bradley Reference Bradley and Bradley1997: 36–37; Sun Reference Hongkai2001). However, the recent migration history of the group from the areas historically populated by the Naxi and Mosuo ethnic groups, in combination with salient typological similarities between Xumi and the Naxi and Mosuo languages, rather suggest that Xumi is more closely related to those two languages (Guo & He Reference Dalie and Zhiwu1994: 8–9; Chirkova Reference Chirkova2009, Reference Chirkova2012).
Of old, Shuiluo Township is a multi-ethnic and multi-lingual area. Hence, Xumi villages are interspersed with villages of other ethnic groups. The ethnic neighbors of the Xumi include (in the order of population size, see Figure 1): (i) Gami (嘎米) Tibetans in the upper reaches of the Shuiluo river (the village of Dulu),Footnote 2 (ii) Pumi (普米) in the upper and lower reaches of Shuiluo river (the villages of Siweng and Gangba), (iii) Naxi (纳西, the villages of Qiubao and Guni) in the lower reaches of the Shuiluo river, and (iv) Muli Mongolians (speakers of the Mosuo (摩些 or 摩梭) language, the village of Lianmu) in the middle reaches of the Shuiluo river (Weckerle et al. Reference Weckerle, Huber, Yongping and Weibang2006, Büeler Reference Büeler2010).
Possibly reflecting these respective contact influences, the Xumi language can be divided into two varieties with restricted mutual intelligibility: (i) the variety of the upper reaches of the Shuiluo River (hereafter, Upper Xumi, spoken in the village of Lanman); and (ii) the variety of the lower and middle reaches of the Shuiluo River (hereafter, Lower Xumi, spoken in the villages of Xinzang, Pingweng, Liangbao, and Mianbang). The two varieties differ both in their segmental inventories and phonotactic constraints. At the lexical level, the two varieties have loanwords from different donor languages. Both varieties have a large number of Tibetan loanwords in their lexicon, but the number of Tibetan loanwords is higher in Upper Xumi, due to its close contact with Gami Tibetan. For example, ‘head’ is /RPʁu-
o/ in Lower Xumi (a native Xumi word), but /EPthɑpɔ/ in Upper Xumi (which is a loanword from Tibetan, WT thod pa ‘forehead’). Conversely, Lower Xumi has many Pumi loanwords, e.g. ‘wolf’: Lower Xumi /Hlɐ/ (loan from the local dialect of Pumi, [lɐ⁵Footnote 3], see Lu Reference Shaozun2001: 373), Upper Xumi /Rphu/. As a result of its exposure to multiple linguistic influences, Xumi has large and complex consonant and vowel inventories, much subphonemic variation in the realization of individual phonemes and tonemes, and a number of marginal phonemes in both varieties. Given that most languages with which Xumi is in contact are non-written (with the exception of Tibetan) and little researched, it remains difficult at this stage to realiably assess their respective impact on the phonetics and phonology of Xumi. On the whole, of the two varieties, Lower Xumi has clearer distinctions between vowels, and therefore lends itself better for an introductory overview of the Xumi language. The more complex Upper Xumi is the topic of a follow-up Illustration (see Chirkova, Chen & Kocjančič Antolík Reference Chirkova, Chen and Antolík2013).
The Xumi use their native language as the primary language of oral communication for family and community events. A considerable percentage of Xumi women are monolingual, whereas most men are proficient in Mandarin Chinese (the local variety of South-West Mandarin). In addition, Lower Xumi speakers often have a good command of Pumi (and some even of Naxi and Mosuo), whereas Upper Xumi speakers often have a good command of Gami Tibetan. The language of the Xumi is traditionally considered by its speakers and ethnic neighbors alike as mixed and combining elements of various local languages. It does not have its own writing system.
Xumi is little researched, with only three outlines to date. Of these, one (Sun Reference Hongkai1983) focuses on Lower Xumi, whereas two other (Huang & Renzeng Reference Bufan, Wangmu, Qingxia, Bufan, Ailan, Wangmu and Juhuang1991, Chirkova Reference Chirkova2009) focus on Upper Xumi.
The present description is based on the first author's fieldwork. The word list and the text provided with this paper were recited by a sixty-two-year-old male native speaker of Lower Xumi, who was born and raised in Shuiluo Township (Liangbao village 两保村). Examples in the sections on syllable structure and prosodic organization were pronounced by a female speaker in her sixties, also born and raised in Shuiluo Township (Mianbang village 免邦村; sound files marked as ‘FEMALE_SPKR’). A chart of all C–V combinations is provided in the appendix.
Consonants
The consonant inventory comprises of 50 consonants, as listed in the table below. In the table, segments that are low in frequency (the voiced uvular stop) or restricted to loanwords (retroflex affricates, and the voiceless alveolopalatal and velar nasals) are put in parentheses.
There is a general thee-way contrast in stops and affricates: voiceless aspirated, voiceless unaspirated, and voiced. Velar and uvular stops contrast before /ɐ u ʉ o/, as in /Hkhɐ/ ‘foot’ vs. /Hqhɐ/ ‘feces’. Elsewhere, they are in complementary distribution. Velar stops are found before front vowels, whereas uvular stops are found before the low back vowel /ɑ/, as in /Hkhe/ ‘to pour’, /Rɡjɛ/ ‘eggplant’ vs. /Hqhɑ/ ‘to slaughter’. The voiced uvular stop occurs only in a few words. It shows varying degrees of frication at the beginning of the following vowel, after the release (e.g. /Rɢo/ [Rɢ
o] ‘to stew’).Footnote 4
Xumi affricates include three-way contrasts in place of articulation: alveolar, postalveolar, and alveolopalatal. Postalveolar and alveolopalatal affricates mostly occur before front and central vowels. No constrast is found before /i ɛ/ (in which cases we use symbols for alveolopalatal affricates in our transcriptions) (e.g. /Htɕi/ ‘to be afraid’, /Htɕɛ/ ‘earth’). Postalveolar and alveolopalatal affricates contrast before /e ʉ o/ (as well as marginally also before /ɐ/). Examples include:
Xumi has an additional set of retroflex affricates in Tibetan loanwords, which mostly occur before back vowels, but occasionally also with high front vowels (in loanwords). In such cases, retroflex affricates correspond to WT initial clusters with the medial -r-, as in /RPtʂiwɑ/ ‘business, affair’ (WT brel ba), /RPtʂupɑ/ ‘sixth lunar month’ (WT drug pa), /RPtʂɑpjɛ/ ‘monk’ (WT grwa ba). In a very few cases (e.g. before the syllabic consonant /
/), retroflex affricates contrast with postalveolar affricates, e.g. /Htʃh/ [Htʃh
] ‘to sell’ vs. /Htʂh/ [Htʂh
] ‘dollar’ (WT khrid).
Xumi fricatives are pronounced at six different places of articulation: alveolar, postalveolar, alveolopalatal, velar, uvular, and glottal. Most fricatives contrast in voicing (namely, /s/ vs. /z/, /ʃ/ vs. /ʒ/, /ɕ/ vs. /ʑ/, /x/ vs. /ɣ/, /h/ vs. /ɦ/), whereas the uvular voiced fricative /ʁ/ does not have a voiceless counterpart. Alveolar and alveolopalatal fricatives (/sz ɕ ʑ/) have the broadest distribution, co-occurring with most vowels, e.g. /Hse/ ‘breath’, /(Htɐ) Hɕe/ ‘to spend a night (there)’, /Hsjɛ/ ‘blood’ vs. /Hɕɛ/ ‘louse’, /EPsoje/ ‘saw’ (WT sog le) vs. /Hɕo/ ‘meat’. /ʃ/ vs. /ʒ/ have a more restricted distribution, as they are essentially attested with central vowels and the syllabic consonant //, e.g. /RPʃʉ-ʃʉ/ ‘paper’ (WT shog shog), /Hʒʉ/ ‘to stamp’, /Hʃ/ ‘tongue’, /RPʎɐ-ʒ/ ‘to sleep’. The voiced glottal fricative /ɦ/ has the most restricted distribution of all fricatives, co-occurring only with /
/ and /
/, e.g. /Hɦ/ ‘silver’, /Hɦ/ ‘pigeon’.
There is a complex relationship between velar and glottal voiceless fricatives and nasalization. Xumi essentially follows the areal tendency to nasalize glottal-initial words, as is the case in a number of Ngwi (Yi) and Na languages (e.g. Matisoff Reference Matisoff1973: 20–21, Reference Matisoff, Ferguson, Hyman and Ohala1975; Bradley Reference Bradley, Bradley, Henderson and Mazaudon1989), and in some neighboring Qiangic languages (e.g. Lizu, see Chirkova & Chen Reference Chirkova and Chen2013). Words beginning with /h-/, /ɦ-/ or a zero-initial (preceded by a non-phonemic glottal stop) mostly co-occur with nasal vowels and tend to be nasalized throughout. Examples include: /Hh/ ‘deep’, /Hɦ/ ‘pigeon’, /H/ [Hʔ] ‘self’. Conversely, words beginning with /x-/ and /ɣ-/ (as well as some words with a zero-initial) co-occur with oral vowels, e.g. /Hxu/ ‘rain’, /Hɣu/ ‘lake’, /RPphi-u/ [RPphi-ʔu] ‘patch’. Overall, velar fricatives are infrequent in the word-initial position, where they mostly co-occur with the vowel /u/ (see the examples above).
Xumi nasals are pronounced at four places of articulation: bilabial, alveolar, alveolopalatal, and velar. /m/ has the broadest distribution, and /ŋ/ has the most restricted distribution of all nasals. /ŋ/ only occurs in a few words and before low vowels (/ɐ ɑ/), e.g. /Hŋɐ/ ‘I’, /RPŋɑpɑ/ ‘fifth lunar month’ (WT lnga pa). /n/ and /ɲ/ contrast before /ɐ /, e.g. /Rɲɐ/ ‘fire’, /Rnɐ/ ‘brain’ (loanword from the local dialect of Pumi, [nɐ13] ‘brain’, see Lu Reference Shaozun2001: 382–383), /(RPdʑɐ) n/ ‘inside (water)’, /RPbu-ɲ/ [RPbə-ɲ] ‘to push (outward)’.Footnote 5 The contrast is not found before /i ɛ ʉ/ (in which cases, we use symbols for alveolopalatal nasals in our transcriptions), as in /Hɲi/ ‘you, thou’, /Hɲɛ/ ‘milk’, /LPɲʉ-ɲʉ/ ‘breast’. Finally, before /e/, only /n/ is found (e.g. /Hne/ ‘snivel, snot’).
All Xumi nasals show the contrast of voiced vs. voiceless nasals, as the (near) minimal pairs show: /Hm
/ ‘today’, /H
/ ‘medicine’; /(RPdʑɐ) n/ ‘inside (water)’, /R
/ ‘fur, hair’; /Hɲɛ/ ‘milk’, /EPʃ
ɛ/ ‘clean’; /RPŋɑpɑ/ ‘fifth lunar month’, /EP
ɑm/ ‘camel’ (WT rnga mo). For voiceless nasals, there is a sustainable amount of nasal airflow, due to velum lowering before the oral release, resulting in a sequential timing relationship between oral and velic articulations. We therefore transcribe them as voiceless nasals. Voiceless nasals are infrequent and mostly restricted to loanwords from Tibetan, where the original initial was a cluster beginning with s- or r-, e.g. /H/ (WT sman) ‘medicine’, /EP
ku/ ‘pen’ (WT smyug gu), /EP ɑm/ ‘camel’ (WT rnga mo). / / and // are also attested in the native vocabulary, e.g. /EPɹu ɑ/ ‘lower part of the body’, /RPjɛts/ ‘tail’, /R/ ‘fur, hair’.
Xumi has three central approximants and four lateral approximants: /wj ɹ/, and /l ʎ 
/. The approximant /j/ occurs before front and central oral vowels (/ie ɛ ɐ/) as well as before the nasal vowels /ĩ / (as in /Rji/ ‘wart’, /Rje/ [Rjɪ] ‘to lick’, /Rjɛ/ ‘vegetable oil’, /Hjĩ=ji/ ‘to plough’, /EPkj/ ‘oil lamp’ (WT kong?), /Hj/ ‘to bark’). Before /i/, /j/ contrasts with a zero initial, e.g. /Rji/ ‘wart’, /RPine/ ‘in the past’.
/w/ and /ɹ/ have a broad distribution, occurring before most oral vowels (/ie ɐ ʉ uo ɑ/). For example, /LPowe/ ‘downstairs shed’, /RPbʉ-wʉ/ ‘to go out’, /RPdiwu/ ‘bullet’ (WT mde'u), /EPthuwɑ/ ‘hammer’ (WT tho ba). Before /ɐ/, /w/ is realized as [v], as in /Hwɐ/ [Hvɐʔ] ‘tooth’. Examples with /ɹ/ include: /Hɹi/ ‘now’, /Hɹe/ ‘cloth’, /Hɹʉ/ ‘to walk’, /RPbɐɹo/ ‘snake’, /Rɹɐ/ ‘good’. We note that in word-initial position, /ɹ/ is realized as preceded by a schwa (e.g. /Hɹi/ [Həɹi] ‘now’). In the transcribed text, a similar addition of schwa is also observed in the words /Hthi/ [Həthi] ‘that’ and /Htɐ/ [Hətɐ] ‘there’. No such effect is observed in word-medial position, where /ɹ/ is realized as [ɹ]. In addition to being an initial, /ɹ/ is also found as a rhyme, i.e. // (see below).
Xumi laterals are pronounced at two places of articulation: alveolar and alveolopalatal. They contrast before /ɐ o/, as in /Hlɐ/ ‘wolf’ vs. /RloHʎɐ/ ‘come again’. No contrast is found before /i ɛ ʉ/ (in which cases we use symbols for alveolopalatal laterals in our transcriptions), e.g. /LPneɡuRPbe-ʎi/ ‘to dry clothes near fire’, /Rʎɛ/ ‘predestined affinity’, /RPdʑi-ʎʉ/ ‘one item of clothing’. Before /e/, we only observe /l/, as in /Rle/ ‘hand’.
All laterals show the correlation voiced–voiceless (/l/ vs. //, /ʎ/ vs. //), which is similar to that of Xumi nasals. Examples of (near) minimal pairs include: /Hlɑ/ ‘tiger’, /RPɡiɑ/ ‘mountain spirit’ (WT dge? lha); /RPlo-lo/ ‘to wrap’, /RPʁu-o/ ‘head’; /RPdʑi-lu/ ‘one time’, /RPuβi/ ‘timber’; /Rle/ ‘hand’, /He/ ‘wind’; /Rʎo/ ‘musk deer’, /Ho/ ‘spirit, soul’.
Syllabic consonant
As a syllabic consonant, // may occur with a zero-initial (e.g. the topic marker //, see the transcribed text). We take the position that this syllabic consonant is phonologically the same as the fricative vowel after alveolar sibilants and retroflexes (/sztstshdz ʃ ʒ tʃhtʂh/). In the latter case, // is realized as homorganic to the preceding consonant onset. Consider the following examples: /Hts/ [Hts
] ‘leopard, panther’, /Hdz/ [Hdz] ‘wheat’, /Htsh/ [Htsh] ‘goat’, /Hʃ/ [Hʃ] ‘tongue’, /RPʎɐ-ʒ/ [RPʎɐ-ʒ] ‘to sleep’, /Htʃh/ [Htʃh] ‘to sell’, /Htʂh/ [Htʂh ] ‘dollar’. Fricative vowels are found in many languages of Southwest China. For example, they are attested in Northern Ngwi (Nuosu) (Bradley Reference Bradley1979: 70; Li & Ma Reference Min and Ming1983: 36), Lisu (Bradley Reference Bradley, Thurgood and LaPolla2003: 224), Naxi and Mosuo (Michaud Reference Michaud2008). In these languages, syllabic fricatives are often analyzed as allophones of high vowels (of /i/ in Chinese, of /i/ or /y/ in Nuosu and Lisu, and of /ɯ/ in Naxi and Mosuo; see references above). This option, however, does not appear possible for Xumi, where the syllabic consonant // contrasts with the entire range of high vowels (/i ʉ u/), as in /Htsi/ ‘lock’, /Hts/ ‘leopard, panther’, /RPʎɐ-tsʉ/ ‘to filter tea’, /Rtsu/ ‘to be in debt’. We therefore analyze it as a separate phoneme in this language.
Consonant clusters
The approximants /w/ and /j/ occur in the second position in consonant clusters, where they may be realized as secondary labialization or palatalization of the first position consonant.
/j/ occurs after bilabials, alveolars, alveolopalatals, velars, nasals, and laterals. In most cases it is restricted in co-occurrence with the vowel /ɛ/ (see ‘Vowels’ section below). In addition, it marginally co-occurs in clusters with alveolars, followed by the vowels //, as in /RPmjɛ-tj/ ‘to add’ (compare /Rd/ ‘wild cat’), /Rtj/ ‘to add up’ (compare /EPt-t/ ‘to chat’).
/w/ occurs after alveolars, postalveolars, alveolopalatals, velars, uvulars, laterals, and /ɹ/; most frequently before /i ɐ/, as well as before /e ɛ ɑ /. If preceded by a palatal initial and/or followed by the front vowels /ie ɛ/, /w/ is palatalized and realized as [ɥ], e.g. /Rdʑwɛ/ [Rdʑɥɛ] ‘fang’ vs. /Rdʑwɐ/ [Rdʑɥɐ] ‘snow’, /LPbutɕwɛ/ [LPbutɕɥɛ] ‘whip’ vs. /Htɕwɐ/ [Htɕɥɐ] ‘sweat’, /EPɡwɛlɑ/ [EPɡɥɛlɑ] ‘to hunt’ vs. /RPɐ-ɡwɐ/ [RPɐ-ɡwɐ] ‘uncle’, /RPdʑi-tɕhwɐ/ [RPdʑi-tɕhɥɐ] ‘to hang’, /Rʎwe/ [Rʎɥe] ‘heavy’; /Htwi/ [Htɥi] ‘mule’ (compare /Htwɐ/ ‘to plant’), /Htshwi/ [Htshɥi] ‘paint, lacquer’ (WT tshos) (compare /Rtshwɐ/ ‘fat’).
Consonant lenition
Xumi has a set of productive lenition rules, which transform most voiced and voiceless aspirated stops and affricates into spirants. This happens when these initials appear intervocalically – in compounds (most commonly in disyllabic numeral–classifier and verb–object collocations) as well as when the verb root is preceded either by a prefix or by a negator. Lenition appears to mostly operate within disyllabic domains. The outputs of the above rules are by and large allophones of the corresponding stops and affricates (such as [β ɸ χ]), but also independent phonemes that can occur word-initially; the following list illustrates these cases and includes examples involving the remaining phonemes:
Lenition in Xumi is regular and affects most stops and affricates (with the only exception of /tsh/ and /tʃh/), although the degree of contrast varies slightly. (In addition, we have no examples of the lenition of the voiced uvular stop, possibly, due to its low frequency.) For example, while bilabial and alveolar stops and affricates show a clear case of alternation between a stop or an affricate in the word-initial position and its corresponding lenited allophone in the word-medial position, the lenition in velar and uvular stops is less categorical. For example, in the compound /RPdʑi-qhʉ/ ‘one year’, the initial /qh/ of the words /Hqhʉ/ ‘year’ does not undergo the otherwise pervasive change to [χ]. In a similar fashion, in the compound /LPskhʉ/ ‘root of a tree’, the initial /kh/ of the word /Rkhʉ/ ‘root’ does not undergo lenition to [x]. Overall, the stability of spirantization may be attributed to the frequency of the item in question (high stability in high frequency phrases and variable in low frequency phrases).
Vowels
The Xumi vowel system comprises eight oral vowels and six nasal vowels.
Oral vowels
Front mid vowels contrast two degrees of vowel height: /e/ vs. /ɛ/, as in the following examples (where these vowels are also contrasted with /i/): /Rji/ ‘wart’, /Rje/ ‘to lick’, /Rjɛ/ ‘vegetable oil’.
The front vowel /ɛ/ seems to be limited to co-occur only with [j] (with only one exception in our present corpus of c. 1,500 words, e.g. /RPlɑɹɛ/ ‘turban’); for example, /Hpjɛ/ ‘to climb’, /Htjɛ/ ‘(large old-style) cutting board (for cutting bones)’, /Rɡjɛ/ ‘eggplant’, /Hmjɛ/ ‘bamboo’. /ʉ/ co-occurs with bilabial, alveolar, velar and uvular stops, and with alveolar, postalveolar, and alveolopalatal affricates and fricatives. Consider some examples of the contrast between /u/ and /ʉ/ (also contrasted with /o/): /Rbu/ ‘crops’, /Rbʉ/ ‘fly’, /LPboxo/ ‘gourd’; /Hsu/ ‘to count’, /Hsʉ/ ‘to wipe’, /EPsoje/ ‘saw’ (WT sog le); /LPqhulɑ/ ‘crow’, /Hqhʉ/ ‘year’, /Hqho/ ‘bowl’.
Contrast between /ɐ/ and /ɑ/ can be illustrated with the following minimal pairs: /Hpɐ/ ‘to speak’ vs. /Hpɑ/ ‘to arrive’, /Hlɐ/ ‘wolf’ vs. /Hlɑ/ ‘tiger’, /LPʎɐ-zɐ/ ‘have washed’ vs. /LPʎɐ-zɑ/ ‘thin, skinny’, /Hqhwɐ/ ‘horn; to steal’ vs. /Hqhwɑ/ ‘far’.
Nasal vowels
The nasal vowels include /ĩ /, as in the following examples: /LPmĩdɑ/ ‘pitiful’, /Hb/ ‘measure cup for grain’, /Hb/ ‘millet’, /Hb/ ‘thick’, /Hb/ ‘yak’; /Hhĩ/ ‘man, person’, /Hh/ ‘to cut’, /Hh/ ‘to blow’, /Hh/ ‘deep; vegetable’. /ĩ/ can be realized as [ĩ] or [
], in free variation. Compare the two realizations of the word /Hhĩ/ ‘man, person’: (i) [Hhĩ] in isolation (as above) and (ii) [Hh] in the compound /EPʃʉ-hĩ/ [EPʃʉ-h] ‘Xumi people’. After postalveolar and alveolopalatal initials, // is realized as [] (as in /Hdʒ/ [Hdʒ] ‘to fly’). In addition, Xumi has one marginal nasal vowel, /
/, which occurs in only one word in our corpus (e.g. /LPmdɑ Rʁo/ ‘on the roof, upstairs’, compare with /LPmĩdɑ/ ‘pitiful’).
Syllable structure
The canonical Xumi syllable minimally consists of an initial consonant, a nucleus and a tone. It may also contain an optional element in the following linear structure: (C1)(C2)V, where C1 can be any consonant, and C2 can only be or -w- or -j-; V stands for vowel (or the syllable consonant /
/), and parentheses indicate optional constituents. A non-phonemic glottal stop can be inserted at the left edge of a vowel-initial stressed syllable (e.g. /H
/ [Hʔ
] ‘collar’, /H
/ [Hʔ
ʔ] ‘sheep’) and at the left edge of a vowel-initial root (e.g. /RPmjɛ-
/ [RPmjɛ-ʔ
] ‘to swallow’, /RPʎɐ-
/ [RPʎɐ-ʔ
ʔ] ‘to be drunk’).
Similar to its linguistic neighbors, Xumi is phonologically monosyllabic with a strong tendency towards disyllabicity in its lexicon. Disyllabic feet are domains for the phonological processes of lenition and vowel assimilation.
Polysyllabic words are mostly composite, e.g. /EPm-tʃwe/ ‘butter tea’ (from /Hm/ ‘butter’, /Htʃwe/ ‘tea’), /RPlɑwu-dʑwɐ/ ‘peach’ (from the bound root /dʑwɐ/ ‘fruit’), /LPɲɐmibu-bu/ ‘sunflower’ (from /LPɲɐmi/ ‘sun’, /RPbu-bu/ ‘flower’). Xumi also has a handful of di-, tri- and tetra-syllabic monomorphemic words (e.g. /EPɐβi/ ‘ear’, /RPɲiɡe/ ‘morning’, /LPɕtspɑpɑ/ ‘bat’, a loan from the local Pumi dialect, [ɕ3⁵ɡue⁵⁵pa⁵⁵pa⁵⁵]; see Lu Reference Shaozun2001: 375).
In a few cases in our corpus (mostly involving low vowels), we observe that some syllables end rather abruptly, as if followed by a glottal stop. The examples include /Hdʑɐ/ [Hdʑɐʔ] ‘water’, /Hwɐ/ [Hvɐʔ] ‘tooth’, /Hlɑ/ [Hlɑʔ] ‘tiger’, /H/ [Hʔʔ] ‘sheep’. However, no contrastive pairs involving a glottal stop coda have been found. Furthermore, one and the same word may be realized with or without a glottal stop. For example, compare the two realizations of the word /Hlɑ/ ‘tiger’: [Hlɑʔ] and [Hlɑ]. In our corpus, these syllables typically associate with the high tone.
Prosodic organization
Xumi is a tone language. Xumi shares the areal prosodic type of languages of Southwestern Sichuan in having a sparse tone system (see Evans Reference Evans2008, Reference Evans2009 for an overview and discussion). In such a system, no more than one pitch contour is pronounced per word (or longer phonological unit). This is similar to its neighboring languages, such as Lizu (Chirkova & Chen Reference Chirkova, Chen and Antolík2013). Most Xumi syllables (roots) have etymological tones, whereas affixes are toneless and may carry all surface tones, depending on the tone of the root. Xumi function words and discourse particles (e.g. the genitive particle /ji/, the topic marker // in the transcribed text) are never pronounced in isolation. Their surface pitch contour depends on the tone of the preceding (host) lexical word (similar to tonal realization in compounds).
Given the current state of our knowledge and the limited corpus that we have at our disposal, the present overview is not meant to be exhaustive in its description. Instead, it is limited to lexical tonal contrasts and tonal patterns in disyllabic compounds, leaving a broader range of morphosyntactic contexts and intonational patterns for more detailed follow-up work.
Tone and tonal melodies on lexical words
On monosyllabic roots, we observe a two-way tonal contrast: (i) rising (R), as in /Rje/ ‘to lick’, /Rwɐ/ ‘cow’, and (ii) high (H), as in /Hje/ ‘tobacco’, /Hwɐ/ [Hvɐʔ] ‘tooth’. Of the two contrastive tones, the rising tone is more consistent in its rising pitch contour across lexical items. The high tone, on the other hand, is subject to considerable variation. Its pitch contour may be (a) rising (as in /Hje/ [je2⁴] ‘tobacco’ and /Hʒʉ/ [ʒʉ2⁵] ‘to stamp’) or (b) falling (as in /Hdʑo/ [dʑo2⁵2] ‘swan’). Note that perceptually, the high contour in the word /Hje/ [je2⁴] ‘tobacco’ sounds like a high-level tone. This may be due to its early f0 peak alignment, especially in comparison to the rising contour in the word /Rje/ ‘to lick’. Here we label this tonal category as ‘high’ in order to highlight the difference from the rising tone category. The consistent difference between the rising tone and the high tone appears to be in the alignment of f0 peak with regard to the segmental anchor, where the rising tone shows a later peak than the high tone. Conversely, the pitch level of the start of rising as well as the rising peak may vary. This is illustrated in Figures 2 and 3, where we observe a higher rising peak in the rising tone in the word /Rje/ ‘to lick’ (in the minimal pair /Hje/ ‘tobacco’ vs. /Rje/ ‘to lick’), but a lower peak in the rising tone in the word /Rwɐ/ ‘cow’ (in the minimal pair /Hwɐ/ ‘tooth’ vs. /Rwɐ/ ‘cow’). For both pairs, the early alignment of the f0 peak seems to lead to the perception of a rather level tone, despite the fact that there is a clear rising pitch contour on both the high tone and the rising tone.
Figure 2 Pitch contours of the high (H) and rising (R) tones illustrated with the minimal pair /ʜje/ ‘tobacco’ and /ʀje/ ‘to lick’.
Figure 3 Pitch contours of the high (H) and rising (R) tones illustrated with the minimal pair /ʜwɐ/ ‘tooth’ and /ʀwɐ/ ‘cow’.
We note, furthermore, that the high tone may be realized with a continued falling after the alignment of the f0 peak in the early part of the syllable, giving the perception of a falling tone. This is illustrated in Figure 4 with two instantiations of the high rising tone in the word /Hdʑo/ ‘swan; goose’.
Figure 4 Pitch contours of the two instantiations of the word /ʜdʑo/ ‘swan; goose’: A rising f0 contour vs. a rising-falling f0 contour.
Similar to Lizu, the rich latitude of tonal variation in Xumi may be due to the fact that there are only two tonal contrasts to be made in this language over monosyllabic words. Given this amount of variation, we refrain from using the five-point-scale pitch system developed by Yuen Ren Chao (Chao Reference Chao1930), as it would overrepresent the actual amount of lexical tonal contrasts in Xumi.
In polysyllabic monomorphemic words, we observe a three-way contrast of tonal melodies. No minimal three-way contrast of the three tonal melodies has been attested in Xumi, whereas binary tonal contrasts are relatively common (see examples below). The pattern is broadly similar to that in the neighboring Lizu language (Chirkova & Chen Reference Chirkova, Chen and Antolík2013), but it also displays a number of differences, as detailed below. To give one example, while in Lizu, the three-way contrast of tonal melodies is correlated with relative durational differences between the syllables, Xumi appears to rely instead on the melodic difference between the three patterns, without the accompanying durational differences between the syllables.
In our analysis, such tonal melodic difference is reminiscent of the stress effect in polysyllabic words in stress languages. Similar, again, to Lizu, this suggests the existence of a hybrid prosodic system in Xumi, in which the pitch contour of the prosodic domain is correlated with the prominence pattern. In order to better reflect this correlation and to bring about, what appears to us, to be the crucial prosodic contrasts within polysyllabic domains in this language, we opt for the notation system that represents the prominent syllable(s) within the domain, which we originally developed for the Lizu language (Chirkova & Chen Reference Chirkova, Chen and Antolík2013). This hybrid analysis and the associated notation system are distinct from an alternative autosegmental analysis of the data (as in, for instance, Chirkova & Michaud Reference Chirkova and Michaud2009). In our opinion, the approach adopted here accounts better for the strong correlation between the prominence pattern and the pitch contour of the domain. Naturally, further investigation is in order to confirm these preliminary observations and to gain a better understanding of the prosodic organization of this language.
The three tonal melodies on polysyllabic words are as follows:
(i) Equally-Prominent Contour (EP). There is no salient rise or fall over any of the syllables. Rather, it seems to be high-level pitch contours throughout the two syllables. This is different from Lizu, where we usually observe a slightly lowered level pitch contour over the second syllable (Chirkova & Chen Reference Chirkova, Chen and Antolík2013). This pattern is mostly attested in monomorphemic words and in loanwords. For example, /EP
ɐβi/ ‘ear’, /EPpoqɑ/ ‘shoe’, /EP ɑm/ ‘camel’ (WT rnga mo), /EPbɐkɐɹ/ ‘spider’.(ii) Left-Prominent Contour (LP). The high f0 peak is realized before the end of the first syllable, where the pitch starts to fall already and it continues to fall in the second syllable. For example, /LPɲɐq
/ ‘nose’, /LPbemi/ ‘axe’, /LPqhɐ-χɐ/ ‘to itch’, /LPʎwɐʁumi/ ‘lizard’.(iii) Right-Prominent Contour (RP). The high f0 peak is realized over the last syllable of the word, which also sounds more prominent. The high peak can be realized earlier, which gives the perception of a high-level tone or later, which lends to the perception of a rising tone. No contrastive pairs for the two patterns have so far been found. We analyze them here as belonging to the same pattern. For example, /RPʁu-
o/ ‘head’, /RPdʑh/ ‘house’, /RPbe-mi/ ‘sow’, /RPɲiɡe/ ‘morning’. This is different from Lizu where the last prominent syllable typically shows a falling f0 contour.
Tone in compounds
Xumi generally conforms to the areal characteristics, in which the prosodic pattern of the leftmost root determines in many cases the tonal melody of the whole compound domain. The observed patterns are reminiscent of tonal spreading effect over post-lexical domains, as discussed in previous work on the neighboring languages (see Evans Reference Evans2009 for an overview). Nonetheless, we refrain from analyzing such spreading effect as the decomposition of the lexical tones and the reassociation of these individual tones to the following syllables sequentially. This is because our data suggest that the leftmost lexical tone is realized over the entire derived compound domain. This pattern is different from the prediction of the decomposition analysis.
If the tone of the leftmost monosyllabic root is high, the resulting compound has the left-prominent pattern. Conversely, if the tone of the leftmost monosyllabic root is rising, the resulting melody is right prominent (see examples below). Note that on the compound (i.e. postlexical) level, we adopt the same notation system that we use on the lexical level. However, at the current stage of research, it is unclear whether these two levels share the same level of prosodic organization. It is in fact plausible that the lexical-level prominence difference may be different from the post-lexical difference, an issue to be clarified in follow-up studies.
Transcription of the recorded passage
The original recording (made with a solid-state recorder Fostex FR-2LE and a Beyerdynamic M88 N microphone) has been made available to the JIPA along with this analysis. In the transcription, only lexical items are marked for tone, whereas function words are not.
Semi-narrow phonetic transcription
North Wind and the Sun
Interlinear morphemic glossing
Abbreviations used in the gloss below follow the Leipzig Glossing Rules (LGR, http://www.eva.mpg.de/lingua/resources/glossing-rules.php). Non-standard abbreviations (those not included in the LGR) are: agt = agentive, conj = conjunction.
Acknowledgements
We would like to thank our principal language consultants, Mr. Waxia Danyou 瓦下旦友 and Mrs. Dazhen Lachu 打珍拉初, for their patience and assistance. We are grateful to Jonathan Evans and the anonymous reviewers of this journal for helpful comments and suggestions. Thanks are also due to Franz Huber (ETH Zürich) for creating the map in this paper. We gratefully acknowledge the financial support of the Agence Nationale de la Recherche (France) as part of the research project ‘What defines Qiang-ness? Towards a phylogenetic assessment of the Southern Qiangic languages of Muli’ (ANR-07-JCJC-0063) to Katia Chirkova. Yiya Chen is currently supported by the Netherlands Organization for Scientific Research (NWO-VIDI 016084338) and the European Research Council (ERC-Starting Independent Researcher Grant, 206198).
Appendix. C–V and CC–V combinations in Lower Xumi
C–V combinations
The table is based on a list of basic vocabulary of c. 1,500 words.
CC–V combinations
